Mercurial > games > semicongine
changeset 1278:772bb32c4368
replcae lodepng with stb_image, some cleanup
| author | sam <sam@basx.dev> |
|---|---|
| date | Mon, 29 Jul 2024 00:00:28 +0700 |
| parents | 4cfa8db3fc9e |
| children | e0ba4aead324 |
| files | config.nims semicongine/contrib/lodepng.c semicongine/contrib/lodepng.h semicongine/gltf.nim semicongine/image.nim semicongine/resources.nim semicongine/text/font.nim semicongine/text/stb_truetype.h semicongine/thirdparty/stb/stb_image.h semicongine/thirdparty/stb/stb_truetype.h tests/test_gltf.nim |
| diffstat | 11 files changed, 13103 insertions(+), 14185 deletions(-) [+] |
line wrap: on
line diff
--- a/config.nims Sun Jul 28 23:18:05 2024 +0700 +++ b/config.nims Mon Jul 29 00:00:28 2024 +0700 @@ -2,7 +2,7 @@ import std/strutils import std/os -import semiconginev2/build +import semicongine/build task build_dev, "build dev": semicongine_build_switches(buildname = "dev")
--- a/semicongine/contrib/lodepng.c Sun Jul 28 23:18:05 2024 +0700 +++ /dev/null Thu Jan 01 00:00:00 1970 +0000 @@ -1,6991 +0,0 @@ -/* -LodePNG version 20230410 - -Copyright (c) 2005-2023 Lode Vandevenne - -This software is provided 'as-is', without any express or implied -warranty. In no event will the authors be held liable for any damages -arising from the use of this software. - -Permission is granted to anyone to use this software for any purpose, -including commercial applications, and to alter it and redistribute it -freely, subject to the following restrictions: - - 1. The origin of this software must not be misrepresented; you must not - claim that you wrote the original software. If you use this software - in a product, an acknowledgment in the product documentation would be - appreciated but is not required. - - 2. Altered source versions must be plainly marked as such, and must not be - misrepresented as being the original software. - - 3. This notice may not be removed or altered from any source - distribution. -*/ - -/* -The manual and changelog are in the header file "lodepng.h" -Rename this file to lodepng.cpp to use it for C++, or to lodepng.c to use it for C. -*/ - -#include "lodepng.h" - -#ifdef LODEPNG_COMPILE_DISK -#include <limits.h> /* LONG_MAX */ -#include <stdio.h> /* file handling */ -#endif /* LODEPNG_COMPILE_DISK */ - -#ifdef LODEPNG_COMPILE_ALLOCATORS -#include <stdlib.h> /* allocations */ -#endif /* LODEPNG_COMPILE_ALLOCATORS */ - -#if defined(_MSC_VER) && (_MSC_VER >= 1310) /*Visual Studio: A few warning types are not desired here.*/ -#pragma warning( disable : 4244 ) /*implicit conversions: not warned by gcc -Wall -Wextra and requires too much casts*/ -#pragma warning( disable : 4996 ) /*VS does not like fopen, but fopen_s is not standard C so unusable here*/ -#endif /*_MSC_VER */ - -const char* LODEPNG_VERSION_STRING = "20230410"; - -/* -This source file is divided into the following large parts. The code sections -with the "LODEPNG_COMPILE_" #defines divide this up further in an intermixed way. --Tools for C and common code for PNG and Zlib --C Code for Zlib (huffman, deflate, ...) --C Code for PNG (file format chunks, adam7, PNG filters, color conversions, ...) --The C++ wrapper around all of the above -*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* // Tools for C, and common code for PNG and Zlib. // */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ - -/*The malloc, realloc and free functions defined here with "lodepng_" in front -of the name, so that you can easily change them to others related to your -platform if needed. Everything else in the code calls these. Pass --DLODEPNG_NO_COMPILE_ALLOCATORS to the compiler, or comment out -#define LODEPNG_COMPILE_ALLOCATORS in the header, to disable the ones here and -define them in your own project's source files without needing to change -lodepng source code. Don't forget to remove "static" if you copypaste them -from here.*/ - -#ifdef LODEPNG_COMPILE_ALLOCATORS -static void* lodepng_malloc(size_t size) { -#ifdef LODEPNG_MAX_ALLOC - if(size > LODEPNG_MAX_ALLOC) return 0; -#endif - return malloc(size); -} - -/* NOTE: when realloc returns NULL, it leaves the original memory untouched */ -static void* lodepng_realloc(void* ptr, size_t new_size) { -#ifdef LODEPNG_MAX_ALLOC - if(new_size > LODEPNG_MAX_ALLOC) return 0; -#endif - return realloc(ptr, new_size); -} - -static void lodepng_free(void* ptr) { - free(ptr); -} -#else /*LODEPNG_COMPILE_ALLOCATORS*/ -/* TODO: support giving additional void* payload to the custom allocators */ -void* lodepng_malloc(size_t size); -void* lodepng_realloc(void* ptr, size_t new_size); -void lodepng_free(void* ptr); -#endif /*LODEPNG_COMPILE_ALLOCATORS*/ - -/* convince the compiler to inline a function, for use when this measurably improves performance */ -/* inline is not available in C90, but use it when supported by the compiler */ -#if (defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L)) || (defined(__cplusplus) && (__cplusplus >= 199711L)) -#define LODEPNG_INLINE inline -#else -#define LODEPNG_INLINE /* not available */ -#endif - -/* restrict is not available in C90, but use it when supported by the compiler */ -#if (defined(__GNUC__) && (__GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ >= 1))) ||\ - (defined(_MSC_VER) && (_MSC_VER >= 1400)) || \ - (defined(__WATCOMC__) && (__WATCOMC__ >= 1250) && !defined(__cplusplus)) -#define LODEPNG_RESTRICT __restrict -#else -#define LODEPNG_RESTRICT /* not available */ -#endif - -/* Replacements for C library functions such as memcpy and strlen, to support platforms -where a full C library is not available. The compiler can recognize them and compile -to something as fast. */ - -static void lodepng_memcpy(void* LODEPNG_RESTRICT dst, - const void* LODEPNG_RESTRICT src, size_t size) { - size_t i; - for(i = 0; i < size; i++) ((char*)dst)[i] = ((const char*)src)[i]; -} - -static void lodepng_memset(void* LODEPNG_RESTRICT dst, - int value, size_t num) { - size_t i; - for(i = 0; i < num; i++) ((char*)dst)[i] = (char)value; -} - -/* does not check memory out of bounds, do not use on untrusted data */ -static size_t lodepng_strlen(const char* a) { - const char* orig = a; - /* avoid warning about unused function in case of disabled COMPILE... macros */ - (void)(&lodepng_strlen); - while(*a) a++; - return (size_t)(a - orig); -} - -#define LODEPNG_MAX(a, b) (((a) > (b)) ? (a) : (b)) -#define LODEPNG_MIN(a, b) (((a) < (b)) ? (a) : (b)) - -#if defined(LODEPNG_COMPILE_PNG) || defined(LODEPNG_COMPILE_DECODER) -/* Safely check if adding two integers will overflow (no undefined -behavior, compiler removing the code, etc...) and output result. */ -static int lodepng_addofl(size_t a, size_t b, size_t* result) { - *result = a + b; /* Unsigned addition is well defined and safe in C90 */ - return *result < a; -} -#endif /*defined(LODEPNG_COMPILE_PNG) || defined(LODEPNG_COMPILE_DECODER)*/ - -#ifdef LODEPNG_COMPILE_DECODER -/* Safely check if multiplying two integers will overflow (no undefined -behavior, compiler removing the code, etc...) and output result. */ -static int lodepng_mulofl(size_t a, size_t b, size_t* result) { - *result = a * b; /* Unsigned multiplication is well defined and safe in C90 */ - return (a != 0 && *result / a != b); -} - -#ifdef LODEPNG_COMPILE_ZLIB -/* Safely check if a + b > c, even if overflow could happen. */ -static int lodepng_gtofl(size_t a, size_t b, size_t c) { - size_t d; - if(lodepng_addofl(a, b, &d)) return 1; - return d > c; -} -#endif /*LODEPNG_COMPILE_ZLIB*/ -#endif /*LODEPNG_COMPILE_DECODER*/ - - -/* -Often in case of an error a value is assigned to a variable and then it breaks -out of a loop (to go to the cleanup phase of a function). This macro does that. -It makes the error handling code shorter and more readable. - -Example: if(!uivector_resize(&lz77_encoded, datasize)) ERROR_BREAK(83); -*/ -#define CERROR_BREAK(errorvar, code){\ - errorvar = code;\ - break;\ -} - -/*version of CERROR_BREAK that assumes the common case where the error variable is named "error"*/ -#define ERROR_BREAK(code) CERROR_BREAK(error, code) - -/*Set error var to the error code, and return it.*/ -#define CERROR_RETURN_ERROR(errorvar, code){\ - errorvar = code;\ - return code;\ -} - -/*Try the code, if it returns error, also return the error.*/ -#define CERROR_TRY_RETURN(call){\ - unsigned error = call;\ - if(error) return error;\ -} - -/*Set error var to the error code, and return from the void function.*/ -#define CERROR_RETURN(errorvar, code){\ - errorvar = code;\ - return;\ -} - -/* -About uivector, ucvector and string: --All of them wrap dynamic arrays or text strings in a similar way. --LodePNG was originally written in C++. The vectors replace the std::vectors that were used in the C++ version. --The string tools are made to avoid problems with compilers that declare things like strncat as deprecated. --They're not used in the interface, only internally in this file as static functions. --As with many other structs in this file, the init and cleanup functions serve as ctor and dtor. -*/ - -#ifdef LODEPNG_COMPILE_ZLIB -#ifdef LODEPNG_COMPILE_ENCODER -/*dynamic vector of unsigned ints*/ -typedef struct uivector { - unsigned* data; - size_t size; /*size in number of unsigned longs*/ - size_t allocsize; /*allocated size in bytes*/ -} uivector; - -static void uivector_cleanup(void* p) { - ((uivector*)p)->size = ((uivector*)p)->allocsize = 0; - lodepng_free(((uivector*)p)->data); - ((uivector*)p)->data = NULL; -} - -/*returns 1 if success, 0 if failure ==> nothing done*/ -static unsigned uivector_resize(uivector* p, size_t size) { - size_t allocsize = size * sizeof(unsigned); - if(allocsize > p->allocsize) { - size_t newsize = allocsize + (p->allocsize >> 1u); - void* data = lodepng_realloc(p->data, newsize); - if(data) { - p->allocsize = newsize; - p->data = (unsigned*)data; - } - else return 0; /*error: not enough memory*/ - } - p->size = size; - return 1; /*success*/ -} - -static void uivector_init(uivector* p) { - p->data = NULL; - p->size = p->allocsize = 0; -} - -/*returns 1 if success, 0 if failure ==> nothing done*/ -static unsigned uivector_push_back(uivector* p, unsigned c) { - if(!uivector_resize(p, p->size + 1)) return 0; - p->data[p->size - 1] = c; - return 1; -} -#endif /*LODEPNG_COMPILE_ENCODER*/ -#endif /*LODEPNG_COMPILE_ZLIB*/ - -/* /////////////////////////////////////////////////////////////////////////// */ - -/*dynamic vector of unsigned chars*/ -typedef struct ucvector { - unsigned char* data; - size_t size; /*used size*/ - size_t allocsize; /*allocated size*/ -} ucvector; - -/*returns 1 if success, 0 if failure ==> nothing done*/ -static unsigned ucvector_reserve(ucvector* p, size_t size) { - if(size > p->allocsize) { - size_t newsize = size + (p->allocsize >> 1u); - void* data = lodepng_realloc(p->data, newsize); - if(data) { - p->allocsize = newsize; - p->data = (unsigned char*)data; - } - else return 0; /*error: not enough memory*/ - } - return 1; /*success*/ -} - -/*returns 1 if success, 0 if failure ==> nothing done*/ -static unsigned ucvector_resize(ucvector* p, size_t size) { - p->size = size; - return ucvector_reserve(p, size); -} - -static ucvector ucvector_init(unsigned char* buffer, size_t size) { - ucvector v; - v.data = buffer; - v.allocsize = v.size = size; - return v; -} - -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_PNG -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - -/*free string pointer and set it to NULL*/ -static void string_cleanup(char** out) { - lodepng_free(*out); - *out = NULL; -} - -/*also appends null termination character*/ -static char* alloc_string_sized(const char* in, size_t insize) { - char* out = (char*)lodepng_malloc(insize + 1); - if(out) { - lodepng_memcpy(out, in, insize); - out[insize] = 0; - } - return out; -} - -/* dynamically allocates a new string with a copy of the null terminated input text */ -static char* alloc_string(const char* in) { - return alloc_string_sized(in, lodepng_strlen(in)); -} -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -#endif /*LODEPNG_COMPILE_PNG*/ - -/* ////////////////////////////////////////////////////////////////////////// */ - -#if defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_PNG) -static unsigned lodepng_read32bitInt(const unsigned char* buffer) { - return (((unsigned)buffer[0] << 24u) | ((unsigned)buffer[1] << 16u) | - ((unsigned)buffer[2] << 8u) | (unsigned)buffer[3]); -} -#endif /*defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_PNG)*/ - -#if defined(LODEPNG_COMPILE_PNG) || defined(LODEPNG_COMPILE_ENCODER) -/*buffer must have at least 4 allocated bytes available*/ -static void lodepng_set32bitInt(unsigned char* buffer, unsigned value) { - buffer[0] = (unsigned char)((value >> 24) & 0xff); - buffer[1] = (unsigned char)((value >> 16) & 0xff); - buffer[2] = (unsigned char)((value >> 8) & 0xff); - buffer[3] = (unsigned char)((value ) & 0xff); -} -#endif /*defined(LODEPNG_COMPILE_PNG) || defined(LODEPNG_COMPILE_ENCODER)*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / File IO / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_DISK - -/* returns negative value on error. This should be pure C compatible, so no fstat. */ -static long lodepng_filesize(const char* filename) { - FILE* file; - long size; - file = fopen(filename, "rb"); - if(!file) return -1; - - if(fseek(file, 0, SEEK_END) != 0) { - fclose(file); - return -1; - } - - size = ftell(file); - /* It may give LONG_MAX as directory size, this is invalid for us. */ - if(size == LONG_MAX) size = -1; - - fclose(file); - return size; -} - -/* load file into buffer that already has the correct allocated size. Returns error code.*/ -static unsigned lodepng_buffer_file(unsigned char* out, size_t size, const char* filename) { - FILE* file; - size_t readsize; - file = fopen(filename, "rb"); - if(!file) return 78; - - readsize = fread(out, 1, size, file); - fclose(file); - - if(readsize != size) return 78; - return 0; -} - -unsigned lodepng_load_file(unsigned char** out, size_t* outsize, const char* filename) { - long size = lodepng_filesize(filename); - if(size < 0) return 78; - *outsize = (size_t)size; - - *out = (unsigned char*)lodepng_malloc((size_t)size); - if(!(*out) && size > 0) return 83; /*the above malloc failed*/ - - return lodepng_buffer_file(*out, (size_t)size, filename); -} - -/*write given buffer to the file, overwriting the file, it doesn't append to it.*/ -unsigned lodepng_save_file(const unsigned char* buffer, size_t buffersize, const char* filename) { - FILE* file; - file = fopen(filename, "wb" ); - if(!file) return 79; - fwrite(buffer, 1, buffersize, file); - fclose(file); - return 0; -} - -#endif /*LODEPNG_COMPILE_DISK*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* // End of common code and tools. Begin of Zlib related code. // */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_ZLIB -#ifdef LODEPNG_COMPILE_ENCODER - -typedef struct { - ucvector* data; - unsigned char bp; /*ok to overflow, indicates bit pos inside byte*/ -} LodePNGBitWriter; - -static void LodePNGBitWriter_init(LodePNGBitWriter* writer, ucvector* data) { - writer->data = data; - writer->bp = 0; -} - -/*TODO: this ignores potential out of memory errors*/ -#define WRITEBIT(writer, bit){\ - /* append new byte */\ - if(((writer->bp) & 7u) == 0) {\ - if(!ucvector_resize(writer->data, writer->data->size + 1)) return;\ - writer->data->data[writer->data->size - 1] = 0;\ - }\ - (writer->data->data[writer->data->size - 1]) |= (bit << ((writer->bp) & 7u));\ - ++writer->bp;\ -} - -/* LSB of value is written first, and LSB of bytes is used first */ -static void writeBits(LodePNGBitWriter* writer, unsigned value, size_t nbits) { - if(nbits == 1) { /* compiler should statically compile this case if nbits == 1 */ - WRITEBIT(writer, value); - } else { - /* TODO: increase output size only once here rather than in each WRITEBIT */ - size_t i; - for(i = 0; i != nbits; ++i) { - WRITEBIT(writer, (unsigned char)((value >> i) & 1)); - } - } -} - -/* This one is to use for adding huffman symbol, the value bits are written MSB first */ -static void writeBitsReversed(LodePNGBitWriter* writer, unsigned value, size_t nbits) { - size_t i; - for(i = 0; i != nbits; ++i) { - /* TODO: increase output size only once here rather than in each WRITEBIT */ - WRITEBIT(writer, (unsigned char)((value >> (nbits - 1u - i)) & 1u)); - } -} -#endif /*LODEPNG_COMPILE_ENCODER*/ - -#ifdef LODEPNG_COMPILE_DECODER - -typedef struct { - const unsigned char* data; - size_t size; /*size of data in bytes*/ - size_t bitsize; /*size of data in bits, end of valid bp values, should be 8*size*/ - size_t bp; - unsigned buffer; /*buffer for reading bits. NOTE: 'unsigned' must support at least 32 bits*/ -} LodePNGBitReader; - -/* data size argument is in bytes. Returns error if size too large causing overflow */ -static unsigned LodePNGBitReader_init(LodePNGBitReader* reader, const unsigned char* data, size_t size) { - size_t temp; - reader->data = data; - reader->size = size; - /* size in bits, return error if overflow (if size_t is 32 bit this supports up to 500MB) */ - if(lodepng_mulofl(size, 8u, &reader->bitsize)) return 105; - /*ensure incremented bp can be compared to bitsize without overflow even when it would be incremented 32 too much and - trying to ensure 32 more bits*/ - if(lodepng_addofl(reader->bitsize, 64u, &temp)) return 105; - reader->bp = 0; - reader->buffer = 0; - return 0; /*ok*/ -} - -/* -ensureBits functions: -Ensures the reader can at least read nbits bits in one or more readBits calls, -safely even if not enough bits are available. -The nbits parameter is unused but is given for documentation purposes, error -checking for amount of bits must be done beforehand. -*/ - -/*See ensureBits documentation above. This one ensures up to 9 bits */ -static LODEPNG_INLINE void ensureBits9(LodePNGBitReader* reader, size_t nbits) { - size_t start = reader->bp >> 3u; - size_t size = reader->size; - if(start + 1u < size) { - reader->buffer = (unsigned)reader->data[start + 0] | ((unsigned)reader->data[start + 1] << 8u); - reader->buffer >>= (reader->bp & 7u); - } else { - reader->buffer = 0; - if(start + 0u < size) reader->buffer = reader->data[start + 0]; - reader->buffer >>= (reader->bp & 7u); - } - (void)nbits; -} - -/*See ensureBits documentation above. This one ensures up to 17 bits */ -static LODEPNG_INLINE void ensureBits17(LodePNGBitReader* reader, size_t nbits) { - size_t start = reader->bp >> 3u; - size_t size = reader->size; - if(start + 2u < size) { - reader->buffer = (unsigned)reader->data[start + 0] | ((unsigned)reader->data[start + 1] << 8u) | - ((unsigned)reader->data[start + 2] << 16u); - reader->buffer >>= (reader->bp & 7u); - } else { - reader->buffer = 0; - if(start + 0u < size) reader->buffer |= reader->data[start + 0]; - if(start + 1u < size) reader->buffer |= ((unsigned)reader->data[start + 1] << 8u); - reader->buffer >>= (reader->bp & 7u); - } - (void)nbits; -} - -/*See ensureBits documentation above. This one ensures up to 25 bits */ -static LODEPNG_INLINE void ensureBits25(LodePNGBitReader* reader, size_t nbits) { - size_t start = reader->bp >> 3u; - size_t size = reader->size; - if(start + 3u < size) { - reader->buffer = (unsigned)reader->data[start + 0] | ((unsigned)reader->data[start + 1] << 8u) | - ((unsigned)reader->data[start + 2] << 16u) | ((unsigned)reader->data[start + 3] << 24u); - reader->buffer >>= (reader->bp & 7u); - } else { - reader->buffer = 0; - if(start + 0u < size) reader->buffer |= reader->data[start + 0]; - if(start + 1u < size) reader->buffer |= ((unsigned)reader->data[start + 1] << 8u); - if(start + 2u < size) reader->buffer |= ((unsigned)reader->data[start + 2] << 16u); - reader->buffer >>= (reader->bp & 7u); - } - (void)nbits; -} - -/*See ensureBits documentation above. This one ensures up to 32 bits */ -static LODEPNG_INLINE void ensureBits32(LodePNGBitReader* reader, size_t nbits) { - size_t start = reader->bp >> 3u; - size_t size = reader->size; - if(start + 4u < size) { - reader->buffer = (unsigned)reader->data[start + 0] | ((unsigned)reader->data[start + 1] << 8u) | - ((unsigned)reader->data[start + 2] << 16u) | ((unsigned)reader->data[start + 3] << 24u); - reader->buffer >>= (reader->bp & 7u); - reader->buffer |= (((unsigned)reader->data[start + 4] << 24u) << (8u - (reader->bp & 7u))); - } else { - reader->buffer = 0; - if(start + 0u < size) reader->buffer |= reader->data[start + 0]; - if(start + 1u < size) reader->buffer |= ((unsigned)reader->data[start + 1] << 8u); - if(start + 2u < size) reader->buffer |= ((unsigned)reader->data[start + 2] << 16u); - if(start + 3u < size) reader->buffer |= ((unsigned)reader->data[start + 3] << 24u); - reader->buffer >>= (reader->bp & 7u); - } - (void)nbits; -} - -/* Get bits without advancing the bit pointer. Must have enough bits available with ensureBits. Max nbits is 31. */ -static LODEPNG_INLINE unsigned peekBits(LodePNGBitReader* reader, size_t nbits) { - /* The shift allows nbits to be only up to 31. */ - return reader->buffer & ((1u << nbits) - 1u); -} - -/* Must have enough bits available with ensureBits */ -static LODEPNG_INLINE void advanceBits(LodePNGBitReader* reader, size_t nbits) { - reader->buffer >>= nbits; - reader->bp += nbits; -} - -/* Must have enough bits available with ensureBits */ -static LODEPNG_INLINE unsigned readBits(LodePNGBitReader* reader, size_t nbits) { - unsigned result = peekBits(reader, nbits); - advanceBits(reader, nbits); - return result; -} -#endif /*LODEPNG_COMPILE_DECODER*/ - -static unsigned reverseBits(unsigned bits, unsigned num) { - /*TODO: implement faster lookup table based version when needed*/ - unsigned i, result = 0; - for(i = 0; i < num; i++) result |= ((bits >> (num - i - 1u)) & 1u) << i; - return result; -} - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Deflate - Huffman / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -#define FIRST_LENGTH_CODE_INDEX 257 -#define LAST_LENGTH_CODE_INDEX 285 -/*256 literals, the end code, some length codes, and 2 unused codes*/ -#define NUM_DEFLATE_CODE_SYMBOLS 288 -/*the distance codes have their own symbols, 30 used, 2 unused*/ -#define NUM_DISTANCE_SYMBOLS 32 -/*the code length codes. 0-15: code lengths, 16: copy previous 3-6 times, 17: 3-10 zeros, 18: 11-138 zeros*/ -#define NUM_CODE_LENGTH_CODES 19 - -/*the base lengths represented by codes 257-285*/ -static const unsigned LENGTHBASE[29] - = {3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31, 35, 43, 51, 59, - 67, 83, 99, 115, 131, 163, 195, 227, 258}; - -/*the extra bits used by codes 257-285 (added to base length)*/ -static const unsigned LENGTHEXTRA[29] - = {0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, - 4, 4, 4, 4, 5, 5, 5, 5, 0}; - -/*the base backwards distances (the bits of distance codes appear after length codes and use their own huffman tree)*/ -static const unsigned DISTANCEBASE[30] - = {1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193, 257, 385, 513, - 769, 1025, 1537, 2049, 3073, 4097, 6145, 8193, 12289, 16385, 24577}; - -/*the extra bits of backwards distances (added to base)*/ -static const unsigned DISTANCEEXTRA[30] - = {0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, - 8, 9, 9, 10, 10, 11, 11, 12, 12, 13, 13}; - -/*the order in which "code length alphabet code lengths" are stored as specified by deflate, out of this the huffman -tree of the dynamic huffman tree lengths is generated*/ -static const unsigned CLCL_ORDER[NUM_CODE_LENGTH_CODES] - = {16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15}; - -/* ////////////////////////////////////////////////////////////////////////// */ - -/* -Huffman tree struct, containing multiple representations of the tree -*/ -typedef struct HuffmanTree { - unsigned* codes; /*the huffman codes (bit patterns representing the symbols)*/ - unsigned* lengths; /*the lengths of the huffman codes*/ - unsigned maxbitlen; /*maximum number of bits a single code can get*/ - unsigned numcodes; /*number of symbols in the alphabet = number of codes*/ - /* for reading only */ - unsigned char* table_len; /*length of symbol from lookup table, or max length if secondary lookup needed*/ - unsigned short* table_value; /*value of symbol from lookup table, or pointer to secondary table if needed*/ -} HuffmanTree; - -static void HuffmanTree_init(HuffmanTree* tree) { - tree->codes = 0; - tree->lengths = 0; - tree->table_len = 0; - tree->table_value = 0; -} - -static void HuffmanTree_cleanup(HuffmanTree* tree) { - lodepng_free(tree->codes); - lodepng_free(tree->lengths); - lodepng_free(tree->table_len); - lodepng_free(tree->table_value); -} - -/* amount of bits for first huffman table lookup (aka root bits), see HuffmanTree_makeTable and huffmanDecodeSymbol.*/ -/* values 8u and 9u work the fastest */ -#define FIRSTBITS 9u - -/* a symbol value too big to represent any valid symbol, to indicate reading disallowed huffman bits combination, -which is possible in case of only 0 or 1 present symbols. */ -#define INVALIDSYMBOL 65535u - -/* make table for huffman decoding */ -static unsigned HuffmanTree_makeTable(HuffmanTree* tree) { - static const unsigned headsize = 1u << FIRSTBITS; /*size of the first table*/ - static const unsigned mask = (1u << FIRSTBITS) /*headsize*/ - 1u; - size_t i, numpresent, pointer, size; /*total table size*/ - unsigned* maxlens = (unsigned*)lodepng_malloc(headsize * sizeof(unsigned)); - if(!maxlens) return 83; /*alloc fail*/ - - /* compute maxlens: max total bit length of symbols sharing prefix in the first table*/ - lodepng_memset(maxlens, 0, headsize * sizeof(*maxlens)); - for(i = 0; i < tree->numcodes; i++) { - unsigned symbol = tree->codes[i]; - unsigned l = tree->lengths[i]; - unsigned index; - if(l <= FIRSTBITS) continue; /*symbols that fit in first table don't increase secondary table size*/ - /*get the FIRSTBITS MSBs, the MSBs of the symbol are encoded first. See later comment about the reversing*/ - index = reverseBits(symbol >> (l - FIRSTBITS), FIRSTBITS); - maxlens[index] = LODEPNG_MAX(maxlens[index], l); - } - /* compute total table size: size of first table plus all secondary tables for symbols longer than FIRSTBITS */ - size = headsize; - for(i = 0; i < headsize; ++i) { - unsigned l = maxlens[i]; - if(l > FIRSTBITS) size += (1u << (l - FIRSTBITS)); - } - tree->table_len = (unsigned char*)lodepng_malloc(size * sizeof(*tree->table_len)); - tree->table_value = (unsigned short*)lodepng_malloc(size * sizeof(*tree->table_value)); - if(!tree->table_len || !tree->table_value) { - lodepng_free(maxlens); - /* freeing tree->table values is done at a higher scope */ - return 83; /*alloc fail*/ - } - /*initialize with an invalid length to indicate unused entries*/ - for(i = 0; i < size; ++i) tree->table_len[i] = 16; - - /*fill in the first table for long symbols: max prefix size and pointer to secondary tables*/ - pointer = headsize; - for(i = 0; i < headsize; ++i) { - unsigned l = maxlens[i]; - if(l <= FIRSTBITS) continue; - tree->table_len[i] = l; - tree->table_value[i] = pointer; - pointer += (1u << (l - FIRSTBITS)); - } - lodepng_free(maxlens); - - /*fill in the first table for short symbols, or secondary table for long symbols*/ - numpresent = 0; - for(i = 0; i < tree->numcodes; ++i) { - unsigned l = tree->lengths[i]; - unsigned symbol, reverse; - if(l == 0) continue; - symbol = tree->codes[i]; /*the huffman bit pattern. i itself is the value.*/ - /*reverse bits, because the huffman bits are given in MSB first order but the bit reader reads LSB first*/ - reverse = reverseBits(symbol, l); - numpresent++; - - if(l <= FIRSTBITS) { - /*short symbol, fully in first table, replicated num times if l < FIRSTBITS*/ - unsigned num = 1u << (FIRSTBITS - l); - unsigned j; - for(j = 0; j < num; ++j) { - /*bit reader will read the l bits of symbol first, the remaining FIRSTBITS - l bits go to the MSB's*/ - unsigned index = reverse | (j << l); - if(tree->table_len[index] != 16) return 55; /*invalid tree: long symbol shares prefix with short symbol*/ - tree->table_len[index] = l; - tree->table_value[index] = i; - } - } else { - /*long symbol, shares prefix with other long symbols in first lookup table, needs second lookup*/ - /*the FIRSTBITS MSBs of the symbol are the first table index*/ - unsigned index = reverse & mask; - unsigned maxlen = tree->table_len[index]; - /*log2 of secondary table length, should be >= l - FIRSTBITS*/ - unsigned tablelen = maxlen - FIRSTBITS; - unsigned start = tree->table_value[index]; /*starting index in secondary table*/ - unsigned num = 1u << (tablelen - (l - FIRSTBITS)); /*amount of entries of this symbol in secondary table*/ - unsigned j; - if(maxlen < l) return 55; /*invalid tree: long symbol shares prefix with short symbol*/ - for(j = 0; j < num; ++j) { - unsigned reverse2 = reverse >> FIRSTBITS; /* l - FIRSTBITS bits */ - unsigned index2 = start + (reverse2 | (j << (l - FIRSTBITS))); - tree->table_len[index2] = l; - tree->table_value[index2] = i; - } - } - } - - if(numpresent < 2) { - /* In case of exactly 1 symbol, in theory the huffman symbol needs 0 bits, - but deflate uses 1 bit instead. In case of 0 symbols, no symbols can - appear at all, but such huffman tree could still exist (e.g. if distance - codes are never used). In both cases, not all symbols of the table will be - filled in. Fill them in with an invalid symbol value so returning them from - huffmanDecodeSymbol will cause error. */ - for(i = 0; i < size; ++i) { - if(tree->table_len[i] == 16) { - /* As length, use a value smaller than FIRSTBITS for the head table, - and a value larger than FIRSTBITS for the secondary table, to ensure - valid behavior for advanceBits when reading this symbol. */ - tree->table_len[i] = (i < headsize) ? 1 : (FIRSTBITS + 1); - tree->table_value[i] = INVALIDSYMBOL; - } - } - } else { - /* A good huffman tree has N * 2 - 1 nodes, of which N - 1 are internal nodes. - If that is not the case (due to too long length codes), the table will not - have been fully used, and this is an error (not all bit combinations can be - decoded): an oversubscribed huffman tree, indicated by error 55. */ - for(i = 0; i < size; ++i) { - if(tree->table_len[i] == 16) return 55; - } - } - - return 0; -} - -/* -Second step for the ...makeFromLengths and ...makeFromFrequencies functions. -numcodes, lengths and maxbitlen must already be filled in correctly. return -value is error. -*/ -static unsigned HuffmanTree_makeFromLengths2(HuffmanTree* tree) { - unsigned* blcount; - unsigned* nextcode; - unsigned error = 0; - unsigned bits, n; - - tree->codes = (unsigned*)lodepng_malloc(tree->numcodes * sizeof(unsigned)); - blcount = (unsigned*)lodepng_malloc((tree->maxbitlen + 1) * sizeof(unsigned)); - nextcode = (unsigned*)lodepng_malloc((tree->maxbitlen + 1) * sizeof(unsigned)); - if(!tree->codes || !blcount || !nextcode) error = 83; /*alloc fail*/ - - if(!error) { - for(n = 0; n != tree->maxbitlen + 1; n++) blcount[n] = nextcode[n] = 0; - /*step 1: count number of instances of each code length*/ - for(bits = 0; bits != tree->numcodes; ++bits) ++blcount[tree->lengths[bits]]; - /*step 2: generate the nextcode values*/ - for(bits = 1; bits <= tree->maxbitlen; ++bits) { - nextcode[bits] = (nextcode[bits - 1] + blcount[bits - 1]) << 1u; - } - /*step 3: generate all the codes*/ - for(n = 0; n != tree->numcodes; ++n) { - if(tree->lengths[n] != 0) { - tree->codes[n] = nextcode[tree->lengths[n]]++; - /*remove superfluous bits from the code*/ - tree->codes[n] &= ((1u << tree->lengths[n]) - 1u); - } - } - } - - lodepng_free(blcount); - lodepng_free(nextcode); - - if(!error) error = HuffmanTree_makeTable(tree); - return error; -} - -/* -given the code lengths (as stored in the PNG file), generate the tree as defined -by Deflate. maxbitlen is the maximum bits that a code in the tree can have. -return value is error. -*/ -static unsigned HuffmanTree_makeFromLengths(HuffmanTree* tree, const unsigned* bitlen, - size_t numcodes, unsigned maxbitlen) { - unsigned i; - tree->lengths = (unsigned*)lodepng_malloc(numcodes * sizeof(unsigned)); - if(!tree->lengths) return 83; /*alloc fail*/ - for(i = 0; i != numcodes; ++i) tree->lengths[i] = bitlen[i]; - tree->numcodes = (unsigned)numcodes; /*number of symbols*/ - tree->maxbitlen = maxbitlen; - return HuffmanTree_makeFromLengths2(tree); -} - -#ifdef LODEPNG_COMPILE_ENCODER - -/*BPM: Boundary Package Merge, see "A Fast and Space-Economical Algorithm for Length-Limited Coding", -Jyrki Katajainen, Alistair Moffat, Andrew Turpin, 1995.*/ - -/*chain node for boundary package merge*/ -typedef struct BPMNode { - int weight; /*the sum of all weights in this chain*/ - unsigned index; /*index of this leaf node (called "count" in the paper)*/ - struct BPMNode* tail; /*the next nodes in this chain (null if last)*/ - int in_use; -} BPMNode; - -/*lists of chains*/ -typedef struct BPMLists { - /*memory pool*/ - unsigned memsize; - BPMNode* memory; - unsigned numfree; - unsigned nextfree; - BPMNode** freelist; - /*two heads of lookahead chains per list*/ - unsigned listsize; - BPMNode** chains0; - BPMNode** chains1; -} BPMLists; - -/*creates a new chain node with the given parameters, from the memory in the lists */ -static BPMNode* bpmnode_create(BPMLists* lists, int weight, unsigned index, BPMNode* tail) { - unsigned i; - BPMNode* result; - - /*memory full, so garbage collect*/ - if(lists->nextfree >= lists->numfree) { - /*mark only those that are in use*/ - for(i = 0; i != lists->memsize; ++i) lists->memory[i].in_use = 0; - for(i = 0; i != lists->listsize; ++i) { - BPMNode* node; - for(node = lists->chains0[i]; node != 0; node = node->tail) node->in_use = 1; - for(node = lists->chains1[i]; node != 0; node = node->tail) node->in_use = 1; - } - /*collect those that are free*/ - lists->numfree = 0; - for(i = 0; i != lists->memsize; ++i) { - if(!lists->memory[i].in_use) lists->freelist[lists->numfree++] = &lists->memory[i]; - } - lists->nextfree = 0; - } - - result = lists->freelist[lists->nextfree++]; - result->weight = weight; - result->index = index; - result->tail = tail; - return result; -} - -/*sort the leaves with stable mergesort*/ -static void bpmnode_sort(BPMNode* leaves, size_t num) { - BPMNode* mem = (BPMNode*)lodepng_malloc(sizeof(*leaves) * num); - size_t width, counter = 0; - for(width = 1; width < num; width *= 2) { - BPMNode* a = (counter & 1) ? mem : leaves; - BPMNode* b = (counter & 1) ? leaves : mem; - size_t p; - for(p = 0; p < num; p += 2 * width) { - size_t q = (p + width > num) ? num : (p + width); - size_t r = (p + 2 * width > num) ? num : (p + 2 * width); - size_t i = p, j = q, k; - for(k = p; k < r; k++) { - if(i < q && (j >= r || a[i].weight <= a[j].weight)) b[k] = a[i++]; - else b[k] = a[j++]; - } - } - counter++; - } - if(counter & 1) lodepng_memcpy(leaves, mem, sizeof(*leaves) * num); - lodepng_free(mem); -} - -/*Boundary Package Merge step, numpresent is the amount of leaves, and c is the current chain.*/ -static void boundaryPM(BPMLists* lists, BPMNode* leaves, size_t numpresent, int c, int num) { - unsigned lastindex = lists->chains1[c]->index; - - if(c == 0) { - if(lastindex >= numpresent) return; - lists->chains0[c] = lists->chains1[c]; - lists->chains1[c] = bpmnode_create(lists, leaves[lastindex].weight, lastindex + 1, 0); - } else { - /*sum of the weights of the head nodes of the previous lookahead chains.*/ - int sum = lists->chains0[c - 1]->weight + lists->chains1[c - 1]->weight; - lists->chains0[c] = lists->chains1[c]; - if(lastindex < numpresent && sum > leaves[lastindex].weight) { - lists->chains1[c] = bpmnode_create(lists, leaves[lastindex].weight, lastindex + 1, lists->chains1[c]->tail); - return; - } - lists->chains1[c] = bpmnode_create(lists, sum, lastindex, lists->chains1[c - 1]); - /*in the end we are only interested in the chain of the last list, so no - need to recurse if we're at the last one (this gives measurable speedup)*/ - if(num + 1 < (int)(2 * numpresent - 2)) { - boundaryPM(lists, leaves, numpresent, c - 1, num); - boundaryPM(lists, leaves, numpresent, c - 1, num); - } - } -} - -unsigned lodepng_huffman_code_lengths(unsigned* lengths, const unsigned* frequencies, - size_t numcodes, unsigned maxbitlen) { - unsigned error = 0; - unsigned i; - size_t numpresent = 0; /*number of symbols with non-zero frequency*/ - BPMNode* leaves; /*the symbols, only those with > 0 frequency*/ - - if(numcodes == 0) return 80; /*error: a tree of 0 symbols is not supposed to be made*/ - if((1u << maxbitlen) < (unsigned)numcodes) return 80; /*error: represent all symbols*/ - - leaves = (BPMNode*)lodepng_malloc(numcodes * sizeof(*leaves)); - if(!leaves) return 83; /*alloc fail*/ - - for(i = 0; i != numcodes; ++i) { - if(frequencies[i] > 0) { - leaves[numpresent].weight = (int)frequencies[i]; - leaves[numpresent].index = i; - ++numpresent; - } - } - - lodepng_memset(lengths, 0, numcodes * sizeof(*lengths)); - - /*ensure at least two present symbols. There should be at least one symbol - according to RFC 1951 section 3.2.7. Some decoders incorrectly require two. To - make these work as well ensure there are at least two symbols. The - Package-Merge code below also doesn't work correctly if there's only one - symbol, it'd give it the theoretical 0 bits but in practice zlib wants 1 bit*/ - if(numpresent == 0) { - lengths[0] = lengths[1] = 1; /*note that for RFC 1951 section 3.2.7, only lengths[0] = 1 is needed*/ - } else if(numpresent == 1) { - lengths[leaves[0].index] = 1; - lengths[leaves[0].index == 0 ? 1 : 0] = 1; - } else { - BPMLists lists; - BPMNode* node; - - bpmnode_sort(leaves, numpresent); - - lists.listsize = maxbitlen; - lists.memsize = 2 * maxbitlen * (maxbitlen + 1); - lists.nextfree = 0; - lists.numfree = lists.memsize; - lists.memory = (BPMNode*)lodepng_malloc(lists.memsize * sizeof(*lists.memory)); - lists.freelist = (BPMNode**)lodepng_malloc(lists.memsize * sizeof(BPMNode*)); - lists.chains0 = (BPMNode**)lodepng_malloc(lists.listsize * sizeof(BPMNode*)); - lists.chains1 = (BPMNode**)lodepng_malloc(lists.listsize * sizeof(BPMNode*)); - if(!lists.memory || !lists.freelist || !lists.chains0 || !lists.chains1) error = 83; /*alloc fail*/ - - if(!error) { - for(i = 0; i != lists.memsize; ++i) lists.freelist[i] = &lists.memory[i]; - - bpmnode_create(&lists, leaves[0].weight, 1, 0); - bpmnode_create(&lists, leaves[1].weight, 2, 0); - - for(i = 0; i != lists.listsize; ++i) { - lists.chains0[i] = &lists.memory[0]; - lists.chains1[i] = &lists.memory[1]; - } - - /*each boundaryPM call adds one chain to the last list, and we need 2 * numpresent - 2 chains.*/ - for(i = 2; i != 2 * numpresent - 2; ++i) boundaryPM(&lists, leaves, numpresent, (int)maxbitlen - 1, (int)i); - - for(node = lists.chains1[maxbitlen - 1]; node; node = node->tail) { - for(i = 0; i != node->index; ++i) ++lengths[leaves[i].index]; - } - } - - lodepng_free(lists.memory); - lodepng_free(lists.freelist); - lodepng_free(lists.chains0); - lodepng_free(lists.chains1); - } - - lodepng_free(leaves); - return error; -} - -/*Create the Huffman tree given the symbol frequencies*/ -static unsigned HuffmanTree_makeFromFrequencies(HuffmanTree* tree, const unsigned* frequencies, - size_t mincodes, size_t numcodes, unsigned maxbitlen) { - unsigned error = 0; - while(!frequencies[numcodes - 1] && numcodes > mincodes) --numcodes; /*trim zeroes*/ - tree->lengths = (unsigned*)lodepng_malloc(numcodes * sizeof(unsigned)); - if(!tree->lengths) return 83; /*alloc fail*/ - tree->maxbitlen = maxbitlen; - tree->numcodes = (unsigned)numcodes; /*number of symbols*/ - - error = lodepng_huffman_code_lengths(tree->lengths, frequencies, numcodes, maxbitlen); - if(!error) error = HuffmanTree_makeFromLengths2(tree); - return error; -} -#endif /*LODEPNG_COMPILE_ENCODER*/ - -/*get the literal and length code tree of a deflated block with fixed tree, as per the deflate specification*/ -static unsigned generateFixedLitLenTree(HuffmanTree* tree) { - unsigned i, error = 0; - unsigned* bitlen = (unsigned*)lodepng_malloc(NUM_DEFLATE_CODE_SYMBOLS * sizeof(unsigned)); - if(!bitlen) return 83; /*alloc fail*/ - - /*288 possible codes: 0-255=literals, 256=endcode, 257-285=lengthcodes, 286-287=unused*/ - for(i = 0; i <= 143; ++i) bitlen[i] = 8; - for(i = 144; i <= 255; ++i) bitlen[i] = 9; - for(i = 256; i <= 279; ++i) bitlen[i] = 7; - for(i = 280; i <= 287; ++i) bitlen[i] = 8; - - error = HuffmanTree_makeFromLengths(tree, bitlen, NUM_DEFLATE_CODE_SYMBOLS, 15); - - lodepng_free(bitlen); - return error; -} - -/*get the distance code tree of a deflated block with fixed tree, as specified in the deflate specification*/ -static unsigned generateFixedDistanceTree(HuffmanTree* tree) { - unsigned i, error = 0; - unsigned* bitlen = (unsigned*)lodepng_malloc(NUM_DISTANCE_SYMBOLS * sizeof(unsigned)); - if(!bitlen) return 83; /*alloc fail*/ - - /*there are 32 distance codes, but 30-31 are unused*/ - for(i = 0; i != NUM_DISTANCE_SYMBOLS; ++i) bitlen[i] = 5; - error = HuffmanTree_makeFromLengths(tree, bitlen, NUM_DISTANCE_SYMBOLS, 15); - - lodepng_free(bitlen); - return error; -} - -#ifdef LODEPNG_COMPILE_DECODER - -/* -returns the code. The bit reader must already have been ensured at least 15 bits -*/ -static unsigned huffmanDecodeSymbol(LodePNGBitReader* reader, const HuffmanTree* codetree) { - unsigned short code = peekBits(reader, FIRSTBITS); - unsigned short l = codetree->table_len[code]; - unsigned short value = codetree->table_value[code]; - if(l <= FIRSTBITS) { - advanceBits(reader, l); - return value; - } else { - advanceBits(reader, FIRSTBITS); - value += peekBits(reader, l - FIRSTBITS); - advanceBits(reader, codetree->table_len[value] - FIRSTBITS); - return codetree->table_value[value]; - } -} -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_DECODER - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Inflator (Decompressor) / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -/*get the tree of a deflated block with fixed tree, as specified in the deflate specification -Returns error code.*/ -static unsigned getTreeInflateFixed(HuffmanTree* tree_ll, HuffmanTree* tree_d) { - unsigned error = generateFixedLitLenTree(tree_ll); - if(error) return error; - return generateFixedDistanceTree(tree_d); -} - -/*get the tree of a deflated block with dynamic tree, the tree itself is also Huffman compressed with a known tree*/ -static unsigned getTreeInflateDynamic(HuffmanTree* tree_ll, HuffmanTree* tree_d, - LodePNGBitReader* reader) { - /*make sure that length values that aren't filled in will be 0, or a wrong tree will be generated*/ - unsigned error = 0; - unsigned n, HLIT, HDIST, HCLEN, i; - - /*see comments in deflateDynamic for explanation of the context and these variables, it is analogous*/ - unsigned* bitlen_ll = 0; /*lit,len code lengths*/ - unsigned* bitlen_d = 0; /*dist code lengths*/ - /*code length code lengths ("clcl"), the bit lengths of the huffman tree used to compress bitlen_ll and bitlen_d*/ - unsigned* bitlen_cl = 0; - HuffmanTree tree_cl; /*the code tree for code length codes (the huffman tree for compressed huffman trees)*/ - - if(reader->bitsize - reader->bp < 14) return 49; /*error: the bit pointer is or will go past the memory*/ - ensureBits17(reader, 14); - - /*number of literal/length codes + 257. Unlike the spec, the value 257 is added to it here already*/ - HLIT = readBits(reader, 5) + 257; - /*number of distance codes. Unlike the spec, the value 1 is added to it here already*/ - HDIST = readBits(reader, 5) + 1; - /*number of code length codes. Unlike the spec, the value 4 is added to it here already*/ - HCLEN = readBits(reader, 4) + 4; - - bitlen_cl = (unsigned*)lodepng_malloc(NUM_CODE_LENGTH_CODES * sizeof(unsigned)); - if(!bitlen_cl) return 83 /*alloc fail*/; - - HuffmanTree_init(&tree_cl); - - while(!error) { - /*read the code length codes out of 3 * (amount of code length codes) bits*/ - if(lodepng_gtofl(reader->bp, HCLEN * 3, reader->bitsize)) { - ERROR_BREAK(50); /*error: the bit pointer is or will go past the memory*/ - } - for(i = 0; i != HCLEN; ++i) { - ensureBits9(reader, 3); /*out of bounds already checked above */ - bitlen_cl[CLCL_ORDER[i]] = readBits(reader, 3); - } - for(i = HCLEN; i != NUM_CODE_LENGTH_CODES; ++i) { - bitlen_cl[CLCL_ORDER[i]] = 0; - } - - error = HuffmanTree_makeFromLengths(&tree_cl, bitlen_cl, NUM_CODE_LENGTH_CODES, 7); - if(error) break; - - /*now we can use this tree to read the lengths for the tree that this function will return*/ - bitlen_ll = (unsigned*)lodepng_malloc(NUM_DEFLATE_CODE_SYMBOLS * sizeof(unsigned)); - bitlen_d = (unsigned*)lodepng_malloc(NUM_DISTANCE_SYMBOLS * sizeof(unsigned)); - if(!bitlen_ll || !bitlen_d) ERROR_BREAK(83 /*alloc fail*/); - lodepng_memset(bitlen_ll, 0, NUM_DEFLATE_CODE_SYMBOLS * sizeof(*bitlen_ll)); - lodepng_memset(bitlen_d, 0, NUM_DISTANCE_SYMBOLS * sizeof(*bitlen_d)); - - /*i is the current symbol we're reading in the part that contains the code lengths of lit/len and dist codes*/ - i = 0; - while(i < HLIT + HDIST) { - unsigned code; - ensureBits25(reader, 22); /* up to 15 bits for huffman code, up to 7 extra bits below*/ - code = huffmanDecodeSymbol(reader, &tree_cl); - if(code <= 15) /*a length code*/ { - if(i < HLIT) bitlen_ll[i] = code; - else bitlen_d[i - HLIT] = code; - ++i; - } else if(code == 16) /*repeat previous*/ { - unsigned replength = 3; /*read in the 2 bits that indicate repeat length (3-6)*/ - unsigned value; /*set value to the previous code*/ - - if(i == 0) ERROR_BREAK(54); /*can't repeat previous if i is 0*/ - - replength += readBits(reader, 2); - - if(i < HLIT + 1) value = bitlen_ll[i - 1]; - else value = bitlen_d[i - HLIT - 1]; - /*repeat this value in the next lengths*/ - for(n = 0; n < replength; ++n) { - if(i >= HLIT + HDIST) ERROR_BREAK(13); /*error: i is larger than the amount of codes*/ - if(i < HLIT) bitlen_ll[i] = value; - else bitlen_d[i - HLIT] = value; - ++i; - } - } else if(code == 17) /*repeat "0" 3-10 times*/ { - unsigned replength = 3; /*read in the bits that indicate repeat length*/ - replength += readBits(reader, 3); - - /*repeat this value in the next lengths*/ - for(n = 0; n < replength; ++n) { - if(i >= HLIT + HDIST) ERROR_BREAK(14); /*error: i is larger than the amount of codes*/ - - if(i < HLIT) bitlen_ll[i] = 0; - else bitlen_d[i - HLIT] = 0; - ++i; - } - } else if(code == 18) /*repeat "0" 11-138 times*/ { - unsigned replength = 11; /*read in the bits that indicate repeat length*/ - replength += readBits(reader, 7); - - /*repeat this value in the next lengths*/ - for(n = 0; n < replength; ++n) { - if(i >= HLIT + HDIST) ERROR_BREAK(15); /*error: i is larger than the amount of codes*/ - - if(i < HLIT) bitlen_ll[i] = 0; - else bitlen_d[i - HLIT] = 0; - ++i; - } - } else /*if(code == INVALIDSYMBOL)*/ { - ERROR_BREAK(16); /*error: tried to read disallowed huffman symbol*/ - } - /*check if any of the ensureBits above went out of bounds*/ - if(reader->bp > reader->bitsize) { - /*return error code 10 or 11 depending on the situation that happened in huffmanDecodeSymbol - (10=no endcode, 11=wrong jump outside of tree)*/ - /* TODO: revise error codes 10,11,50: the above comment is no longer valid */ - ERROR_BREAK(50); /*error, bit pointer jumps past memory*/ - } - } - if(error) break; - - if(bitlen_ll[256] == 0) ERROR_BREAK(64); /*the length of the end code 256 must be larger than 0*/ - - /*now we've finally got HLIT and HDIST, so generate the code trees, and the function is done*/ - error = HuffmanTree_makeFromLengths(tree_ll, bitlen_ll, NUM_DEFLATE_CODE_SYMBOLS, 15); - if(error) break; - error = HuffmanTree_makeFromLengths(tree_d, bitlen_d, NUM_DISTANCE_SYMBOLS, 15); - - break; /*end of error-while*/ - } - - lodepng_free(bitlen_cl); - lodepng_free(bitlen_ll); - lodepng_free(bitlen_d); - HuffmanTree_cleanup(&tree_cl); - - return error; -} - -/*inflate a block with dynamic of fixed Huffman tree. btype must be 1 or 2.*/ -static unsigned inflateHuffmanBlock(ucvector* out, LodePNGBitReader* reader, - unsigned btype, size_t max_output_size) { - unsigned error = 0; - HuffmanTree tree_ll; /*the huffman tree for literal and length codes*/ - HuffmanTree tree_d; /*the huffman tree for distance codes*/ - const size_t reserved_size = 260; /* must be at least 258 for max length, and a few extra for adding a few extra literals */ - int done = 0; - - if(!ucvector_reserve(out, out->size + reserved_size)) return 83; /*alloc fail*/ - - HuffmanTree_init(&tree_ll); - HuffmanTree_init(&tree_d); - - if(btype == 1) error = getTreeInflateFixed(&tree_ll, &tree_d); - else /*if(btype == 2)*/ error = getTreeInflateDynamic(&tree_ll, &tree_d, reader); - - - while(!error && !done) /*decode all symbols until end reached, breaks at end code*/ { - /*code_ll is literal, length or end code*/ - unsigned code_ll; - /* ensure enough bits for 2 huffman code reads (15 bits each): if the first is a literal, a second literal is read at once. This - appears to be slightly faster, than ensuring 20 bits here for 1 huffman symbol and the potential 5 extra bits for the length symbol.*/ - ensureBits32(reader, 30); - code_ll = huffmanDecodeSymbol(reader, &tree_ll); - if(code_ll <= 255) { - /*slightly faster code path if multiple literals in a row*/ - out->data[out->size++] = (unsigned char)code_ll; - code_ll = huffmanDecodeSymbol(reader, &tree_ll); - } - if(code_ll <= 255) /*literal symbol*/ { - out->data[out->size++] = (unsigned char)code_ll; - } else if(code_ll >= FIRST_LENGTH_CODE_INDEX && code_ll <= LAST_LENGTH_CODE_INDEX) /*length code*/ { - unsigned code_d, distance; - unsigned numextrabits_l, numextrabits_d; /*extra bits for length and distance*/ - size_t start, backward, length; - - /*part 1: get length base*/ - length = LENGTHBASE[code_ll - FIRST_LENGTH_CODE_INDEX]; - - /*part 2: get extra bits and add the value of that to length*/ - numextrabits_l = LENGTHEXTRA[code_ll - FIRST_LENGTH_CODE_INDEX]; - if(numextrabits_l != 0) { - /* bits already ensured above */ - ensureBits25(reader, 5); - length += readBits(reader, numextrabits_l); - } - - /*part 3: get distance code*/ - ensureBits32(reader, 28); /* up to 15 for the huffman symbol, up to 13 for the extra bits */ - code_d = huffmanDecodeSymbol(reader, &tree_d); - if(code_d > 29) { - if(code_d <= 31) { - ERROR_BREAK(18); /*error: invalid distance code (30-31 are never used)*/ - } else /* if(code_d == INVALIDSYMBOL) */{ - ERROR_BREAK(16); /*error: tried to read disallowed huffman symbol*/ - } - } - distance = DISTANCEBASE[code_d]; - - /*part 4: get extra bits from distance*/ - numextrabits_d = DISTANCEEXTRA[code_d]; - if(numextrabits_d != 0) { - /* bits already ensured above */ - distance += readBits(reader, numextrabits_d); - } - - /*part 5: fill in all the out[n] values based on the length and dist*/ - start = out->size; - if(distance > start) ERROR_BREAK(52); /*too long backward distance*/ - backward = start - distance; - - out->size += length; - if(distance < length) { - size_t forward; - lodepng_memcpy(out->data + start, out->data + backward, distance); - start += distance; - for(forward = distance; forward < length; ++forward) { - out->data[start++] = out->data[backward++]; - } - } else { - lodepng_memcpy(out->data + start, out->data + backward, length); - } - } else if(code_ll == 256) { - done = 1; /*end code, finish the loop*/ - } else /*if(code_ll == INVALIDSYMBOL)*/ { - ERROR_BREAK(16); /*error: tried to read disallowed huffman symbol*/ - } - if(out->allocsize - out->size < reserved_size) { - if(!ucvector_reserve(out, out->size + reserved_size)) ERROR_BREAK(83); /*alloc fail*/ - } - /*check if any of the ensureBits above went out of bounds*/ - if(reader->bp > reader->bitsize) { - /*return error code 10 or 11 depending on the situation that happened in huffmanDecodeSymbol - (10=no endcode, 11=wrong jump outside of tree)*/ - /* TODO: revise error codes 10,11,50: the above comment is no longer valid */ - ERROR_BREAK(51); /*error, bit pointer jumps past memory*/ - } - if(max_output_size && out->size > max_output_size) { - ERROR_BREAK(109); /*error, larger than max size*/ - } - } - - HuffmanTree_cleanup(&tree_ll); - HuffmanTree_cleanup(&tree_d); - - return error; -} - -static unsigned inflateNoCompression(ucvector* out, LodePNGBitReader* reader, - const LodePNGDecompressSettings* settings) { - size_t bytepos; - size_t size = reader->size; - unsigned LEN, NLEN, error = 0; - - /*go to first boundary of byte*/ - bytepos = (reader->bp + 7u) >> 3u; - - /*read LEN (2 bytes) and NLEN (2 bytes)*/ - if(bytepos + 4 >= size) return 52; /*error, bit pointer will jump past memory*/ - LEN = (unsigned)reader->data[bytepos] + ((unsigned)reader->data[bytepos + 1] << 8u); bytepos += 2; - NLEN = (unsigned)reader->data[bytepos] + ((unsigned)reader->data[bytepos + 1] << 8u); bytepos += 2; - - /*check if 16-bit NLEN is really the one's complement of LEN*/ - if(!settings->ignore_nlen && LEN + NLEN != 65535) { - return 21; /*error: NLEN is not one's complement of LEN*/ - } - - if(!ucvector_resize(out, out->size + LEN)) return 83; /*alloc fail*/ - - /*read the literal data: LEN bytes are now stored in the out buffer*/ - if(bytepos + LEN > size) return 23; /*error: reading outside of in buffer*/ - - /*out->data can be NULL (when LEN is zero), and arithmetics on NULL ptr is undefined*/ - if (LEN) { - lodepng_memcpy(out->data + out->size - LEN, reader->data + bytepos, LEN); - bytepos += LEN; - } - - reader->bp = bytepos << 3u; - - return error; -} - -static unsigned lodepng_inflatev(ucvector* out, - const unsigned char* in, size_t insize, - const LodePNGDecompressSettings* settings) { - unsigned BFINAL = 0; - LodePNGBitReader reader; - unsigned error = LodePNGBitReader_init(&reader, in, insize); - - if(error) return error; - - while(!BFINAL) { - unsigned BTYPE; - if(reader.bitsize - reader.bp < 3) return 52; /*error, bit pointer will jump past memory*/ - ensureBits9(&reader, 3); - BFINAL = readBits(&reader, 1); - BTYPE = readBits(&reader, 2); - - if(BTYPE == 3) return 20; /*error: invalid BTYPE*/ - else if(BTYPE == 0) error = inflateNoCompression(out, &reader, settings); /*no compression*/ - else error = inflateHuffmanBlock(out, &reader, BTYPE, settings->max_output_size); /*compression, BTYPE 01 or 10*/ - if(!error && settings->max_output_size && out->size > settings->max_output_size) error = 109; - if(error) break; - } - - return error; -} - -unsigned lodepng_inflate(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGDecompressSettings* settings) { - ucvector v = ucvector_init(*out, *outsize); - unsigned error = lodepng_inflatev(&v, in, insize, settings); - *out = v.data; - *outsize = v.size; - return error; -} - -static unsigned inflatev(ucvector* out, const unsigned char* in, size_t insize, - const LodePNGDecompressSettings* settings) { - if(settings->custom_inflate) { - unsigned error = settings->custom_inflate(&out->data, &out->size, in, insize, settings); - out->allocsize = out->size; - if(error) { - /*the custom inflate is allowed to have its own error codes, however, we translate it to code 110*/ - error = 110; - /*if there's a max output size, and the custom zlib returned error, then indicate that error instead*/ - if(settings->max_output_size && out->size > settings->max_output_size) error = 109; - } - return error; - } else { - return lodepng_inflatev(out, in, insize, settings); - } -} - -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_ENCODER - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Deflator (Compressor) / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -static const size_t MAX_SUPPORTED_DEFLATE_LENGTH = 258; - -/*search the index in the array, that has the largest value smaller than or equal to the given value, -given array must be sorted (if no value is smaller, it returns the size of the given array)*/ -static size_t searchCodeIndex(const unsigned* array, size_t array_size, size_t value) { - /*binary search (only small gain over linear). TODO: use CPU log2 instruction for getting symbols instead*/ - size_t left = 1; - size_t right = array_size - 1; - - while(left <= right) { - size_t mid = (left + right) >> 1; - if(array[mid] >= value) right = mid - 1; - else left = mid + 1; - } - if(left >= array_size || array[left] > value) left--; - return left; -} - -static void addLengthDistance(uivector* values, size_t length, size_t distance) { - /*values in encoded vector are those used by deflate: - 0-255: literal bytes - 256: end - 257-285: length/distance pair (length code, followed by extra length bits, distance code, extra distance bits) - 286-287: invalid*/ - - unsigned length_code = (unsigned)searchCodeIndex(LENGTHBASE, 29, length); - unsigned extra_length = (unsigned)(length - LENGTHBASE[length_code]); - unsigned dist_code = (unsigned)searchCodeIndex(DISTANCEBASE, 30, distance); - unsigned extra_distance = (unsigned)(distance - DISTANCEBASE[dist_code]); - - size_t pos = values->size; - /*TODO: return error when this fails (out of memory)*/ - unsigned ok = uivector_resize(values, values->size + 4); - if(ok) { - values->data[pos + 0] = length_code + FIRST_LENGTH_CODE_INDEX; - values->data[pos + 1] = extra_length; - values->data[pos + 2] = dist_code; - values->data[pos + 3] = extra_distance; - } -} - -/*3 bytes of data get encoded into two bytes. The hash cannot use more than 3 -bytes as input because 3 is the minimum match length for deflate*/ -static const unsigned HASH_NUM_VALUES = 65536; -static const unsigned HASH_BIT_MASK = 65535; /*HASH_NUM_VALUES - 1, but C90 does not like that as initializer*/ - -typedef struct Hash { - int* head; /*hash value to head circular pos - can be outdated if went around window*/ - /*circular pos to prev circular pos*/ - unsigned short* chain; - int* val; /*circular pos to hash value*/ - - /*TODO: do this not only for zeros but for any repeated byte. However for PNG - it's always going to be the zeros that dominate, so not important for PNG*/ - int* headz; /*similar to head, but for chainz*/ - unsigned short* chainz; /*those with same amount of zeros*/ - unsigned short* zeros; /*length of zeros streak, used as a second hash chain*/ -} Hash; - -static unsigned hash_init(Hash* hash, unsigned windowsize) { - unsigned i; - hash->head = (int*)lodepng_malloc(sizeof(int) * HASH_NUM_VALUES); - hash->val = (int*)lodepng_malloc(sizeof(int) * windowsize); - hash->chain = (unsigned short*)lodepng_malloc(sizeof(unsigned short) * windowsize); - - hash->zeros = (unsigned short*)lodepng_malloc(sizeof(unsigned short) * windowsize); - hash->headz = (int*)lodepng_malloc(sizeof(int) * (MAX_SUPPORTED_DEFLATE_LENGTH + 1)); - hash->chainz = (unsigned short*)lodepng_malloc(sizeof(unsigned short) * windowsize); - - if(!hash->head || !hash->chain || !hash->val || !hash->headz|| !hash->chainz || !hash->zeros) { - return 83; /*alloc fail*/ - } - - /*initialize hash table*/ - for(i = 0; i != HASH_NUM_VALUES; ++i) hash->head[i] = -1; - for(i = 0; i != windowsize; ++i) hash->val[i] = -1; - for(i = 0; i != windowsize; ++i) hash->chain[i] = i; /*same value as index indicates uninitialized*/ - - for(i = 0; i <= MAX_SUPPORTED_DEFLATE_LENGTH; ++i) hash->headz[i] = -1; - for(i = 0; i != windowsize; ++i) hash->chainz[i] = i; /*same value as index indicates uninitialized*/ - - return 0; -} - -static void hash_cleanup(Hash* hash) { - lodepng_free(hash->head); - lodepng_free(hash->val); - lodepng_free(hash->chain); - - lodepng_free(hash->zeros); - lodepng_free(hash->headz); - lodepng_free(hash->chainz); -} - - - -static unsigned getHash(const unsigned char* data, size_t size, size_t pos) { - unsigned result = 0; - if(pos + 2 < size) { - /*A simple shift and xor hash is used. Since the data of PNGs is dominated - by zeroes due to the filters, a better hash does not have a significant - effect on speed in traversing the chain, and causes more time spend on - calculating the hash.*/ - result ^= ((unsigned)data[pos + 0] << 0u); - result ^= ((unsigned)data[pos + 1] << 4u); - result ^= ((unsigned)data[pos + 2] << 8u); - } else { - size_t amount, i; - if(pos >= size) return 0; - amount = size - pos; - for(i = 0; i != amount; ++i) result ^= ((unsigned)data[pos + i] << (i * 8u)); - } - return result & HASH_BIT_MASK; -} - -static unsigned countZeros(const unsigned char* data, size_t size, size_t pos) { - const unsigned char* start = data + pos; - const unsigned char* end = start + MAX_SUPPORTED_DEFLATE_LENGTH; - if(end > data + size) end = data + size; - data = start; - while(data != end && *data == 0) ++data; - /*subtracting two addresses returned as 32-bit number (max value is MAX_SUPPORTED_DEFLATE_LENGTH)*/ - return (unsigned)(data - start); -} - -/*wpos = pos & (windowsize - 1)*/ -static void updateHashChain(Hash* hash, size_t wpos, unsigned hashval, unsigned short numzeros) { - hash->val[wpos] = (int)hashval; - if(hash->head[hashval] != -1) hash->chain[wpos] = hash->head[hashval]; - hash->head[hashval] = (int)wpos; - - hash->zeros[wpos] = numzeros; - if(hash->headz[numzeros] != -1) hash->chainz[wpos] = hash->headz[numzeros]; - hash->headz[numzeros] = (int)wpos; -} - -/* -LZ77-encode the data. Return value is error code. The input are raw bytes, the output -is in the form of unsigned integers with codes representing for example literal bytes, or -length/distance pairs. -It uses a hash table technique to let it encode faster. When doing LZ77 encoding, a -sliding window (of windowsize) is used, and all past bytes in that window can be used as -the "dictionary". A brute force search through all possible distances would be slow, and -this hash technique is one out of several ways to speed this up. -*/ -static unsigned encodeLZ77(uivector* out, Hash* hash, - const unsigned char* in, size_t inpos, size_t insize, unsigned windowsize, - unsigned minmatch, unsigned nicematch, unsigned lazymatching) { - size_t pos; - unsigned i, error = 0; - /*for large window lengths, assume the user wants no compression loss. Otherwise, max hash chain length speedup.*/ - unsigned maxchainlength = windowsize >= 8192 ? windowsize : windowsize / 8u; - unsigned maxlazymatch = windowsize >= 8192 ? MAX_SUPPORTED_DEFLATE_LENGTH : 64; - - unsigned usezeros = 1; /*not sure if setting it to false for windowsize < 8192 is better or worse*/ - unsigned numzeros = 0; - - unsigned offset; /*the offset represents the distance in LZ77 terminology*/ - unsigned length; - unsigned lazy = 0; - unsigned lazylength = 0, lazyoffset = 0; - unsigned hashval; - unsigned current_offset, current_length; - unsigned prev_offset; - const unsigned char *lastptr, *foreptr, *backptr; - unsigned hashpos; - - if(windowsize == 0 || windowsize > 32768) return 60; /*error: windowsize smaller/larger than allowed*/ - if((windowsize & (windowsize - 1)) != 0) return 90; /*error: must be power of two*/ - - if(nicematch > MAX_SUPPORTED_DEFLATE_LENGTH) nicematch = MAX_SUPPORTED_DEFLATE_LENGTH; - - for(pos = inpos; pos < insize; ++pos) { - size_t wpos = pos & (windowsize - 1); /*position for in 'circular' hash buffers*/ - unsigned chainlength = 0; - - hashval = getHash(in, insize, pos); - - if(usezeros && hashval == 0) { - if(numzeros == 0) numzeros = countZeros(in, insize, pos); - else if(pos + numzeros > insize || in[pos + numzeros - 1] != 0) --numzeros; - } else { - numzeros = 0; - } - - updateHashChain(hash, wpos, hashval, numzeros); - - /*the length and offset found for the current position*/ - length = 0; - offset = 0; - - hashpos = hash->chain[wpos]; - - lastptr = &in[insize < pos + MAX_SUPPORTED_DEFLATE_LENGTH ? insize : pos + MAX_SUPPORTED_DEFLATE_LENGTH]; - - /*search for the longest string*/ - prev_offset = 0; - for(;;) { - if(chainlength++ >= maxchainlength) break; - current_offset = (unsigned)(hashpos <= wpos ? wpos - hashpos : wpos - hashpos + windowsize); - - if(current_offset < prev_offset) break; /*stop when went completely around the circular buffer*/ - prev_offset = current_offset; - if(current_offset > 0) { - /*test the next characters*/ - foreptr = &in[pos]; - backptr = &in[pos - current_offset]; - - /*common case in PNGs is lots of zeros. Quickly skip over them as a speedup*/ - if(numzeros >= 3) { - unsigned skip = hash->zeros[hashpos]; - if(skip > numzeros) skip = numzeros; - backptr += skip; - foreptr += skip; - } - - while(foreptr != lastptr && *backptr == *foreptr) /*maximum supported length by deflate is max length*/ { - ++backptr; - ++foreptr; - } - current_length = (unsigned)(foreptr - &in[pos]); - - if(current_length > length) { - length = current_length; /*the longest length*/ - offset = current_offset; /*the offset that is related to this longest length*/ - /*jump out once a length of max length is found (speed gain). This also jumps - out if length is MAX_SUPPORTED_DEFLATE_LENGTH*/ - if(current_length >= nicematch) break; - } - } - - if(hashpos == hash->chain[hashpos]) break; - - if(numzeros >= 3 && length > numzeros) { - hashpos = hash->chainz[hashpos]; - if(hash->zeros[hashpos] != numzeros) break; - } else { - hashpos = hash->chain[hashpos]; - /*outdated hash value, happens if particular value was not encountered in whole last window*/ - if(hash->val[hashpos] != (int)hashval) break; - } - } - - if(lazymatching) { - if(!lazy && length >= 3 && length <= maxlazymatch && length < MAX_SUPPORTED_DEFLATE_LENGTH) { - lazy = 1; - lazylength = length; - lazyoffset = offset; - continue; /*try the next byte*/ - } - if(lazy) { - lazy = 0; - if(pos == 0) ERROR_BREAK(81); - if(length > lazylength + 1) { - /*push the previous character as literal*/ - if(!uivector_push_back(out, in[pos - 1])) ERROR_BREAK(83 /*alloc fail*/); - } else { - length = lazylength; - offset = lazyoffset; - hash->head[hashval] = -1; /*the same hashchain update will be done, this ensures no wrong alteration*/ - hash->headz[numzeros] = -1; /*idem*/ - --pos; - } - } - } - if(length >= 3 && offset > windowsize) ERROR_BREAK(86 /*too big (or overflown negative) offset*/); - - /*encode it as length/distance pair or literal value*/ - if(length < 3) /*only lengths of 3 or higher are supported as length/distance pair*/ { - if(!uivector_push_back(out, in[pos])) ERROR_BREAK(83 /*alloc fail*/); - } else if(length < minmatch || (length == 3 && offset > 4096)) { - /*compensate for the fact that longer offsets have more extra bits, a - length of only 3 may be not worth it then*/ - if(!uivector_push_back(out, in[pos])) ERROR_BREAK(83 /*alloc fail*/); - } else { - addLengthDistance(out, length, offset); - for(i = 1; i < length; ++i) { - ++pos; - wpos = pos & (windowsize - 1); - hashval = getHash(in, insize, pos); - if(usezeros && hashval == 0) { - if(numzeros == 0) numzeros = countZeros(in, insize, pos); - else if(pos + numzeros > insize || in[pos + numzeros - 1] != 0) --numzeros; - } else { - numzeros = 0; - } - updateHashChain(hash, wpos, hashval, numzeros); - } - } - } /*end of the loop through each character of input*/ - - return error; -} - -/* /////////////////////////////////////////////////////////////////////////// */ - -static unsigned deflateNoCompression(ucvector* out, const unsigned char* data, size_t datasize) { - /*non compressed deflate block data: 1 bit BFINAL,2 bits BTYPE,(5 bits): it jumps to start of next byte, - 2 bytes LEN, 2 bytes NLEN, LEN bytes literal DATA*/ - - size_t i, numdeflateblocks = (datasize + 65534u) / 65535u; - unsigned datapos = 0; - for(i = 0; i != numdeflateblocks; ++i) { - unsigned BFINAL, BTYPE, LEN, NLEN; - unsigned char firstbyte; - size_t pos = out->size; - - BFINAL = (i == numdeflateblocks - 1); - BTYPE = 0; - - LEN = 65535; - if(datasize - datapos < 65535u) LEN = (unsigned)datasize - datapos; - NLEN = 65535 - LEN; - - if(!ucvector_resize(out, out->size + LEN + 5)) return 83; /*alloc fail*/ - - firstbyte = (unsigned char)(BFINAL + ((BTYPE & 1u) << 1u) + ((BTYPE & 2u) << 1u)); - out->data[pos + 0] = firstbyte; - out->data[pos + 1] = (unsigned char)(LEN & 255); - out->data[pos + 2] = (unsigned char)(LEN >> 8u); - out->data[pos + 3] = (unsigned char)(NLEN & 255); - out->data[pos + 4] = (unsigned char)(NLEN >> 8u); - lodepng_memcpy(out->data + pos + 5, data + datapos, LEN); - datapos += LEN; - } - - return 0; -} - -/* -write the lz77-encoded data, which has lit, len and dist codes, to compressed stream using huffman trees. -tree_ll: the tree for lit and len codes. -tree_d: the tree for distance codes. -*/ -static void writeLZ77data(LodePNGBitWriter* writer, const uivector* lz77_encoded, - const HuffmanTree* tree_ll, const HuffmanTree* tree_d) { - size_t i = 0; - for(i = 0; i != lz77_encoded->size; ++i) { - unsigned val = lz77_encoded->data[i]; - writeBitsReversed(writer, tree_ll->codes[val], tree_ll->lengths[val]); - if(val > 256) /*for a length code, 3 more things have to be added*/ { - unsigned length_index = val - FIRST_LENGTH_CODE_INDEX; - unsigned n_length_extra_bits = LENGTHEXTRA[length_index]; - unsigned length_extra_bits = lz77_encoded->data[++i]; - - unsigned distance_code = lz77_encoded->data[++i]; - - unsigned distance_index = distance_code; - unsigned n_distance_extra_bits = DISTANCEEXTRA[distance_index]; - unsigned distance_extra_bits = lz77_encoded->data[++i]; - - writeBits(writer, length_extra_bits, n_length_extra_bits); - writeBitsReversed(writer, tree_d->codes[distance_code], tree_d->lengths[distance_code]); - writeBits(writer, distance_extra_bits, n_distance_extra_bits); - } - } -} - -/*Deflate for a block of type "dynamic", that is, with freely, optimally, created huffman trees*/ -static unsigned deflateDynamic(LodePNGBitWriter* writer, Hash* hash, - const unsigned char* data, size_t datapos, size_t dataend, - const LodePNGCompressSettings* settings, unsigned final) { - unsigned error = 0; - - /* - A block is compressed as follows: The PNG data is lz77 encoded, resulting in - literal bytes and length/distance pairs. This is then huffman compressed with - two huffman trees. One huffman tree is used for the lit and len values ("ll"), - another huffman tree is used for the dist values ("d"). These two trees are - stored using their code lengths, and to compress even more these code lengths - are also run-length encoded and huffman compressed. This gives a huffman tree - of code lengths "cl". The code lengths used to describe this third tree are - the code length code lengths ("clcl"). - */ - - /*The lz77 encoded data, represented with integers since there will also be length and distance codes in it*/ - uivector lz77_encoded; - HuffmanTree tree_ll; /*tree for lit,len values*/ - HuffmanTree tree_d; /*tree for distance codes*/ - HuffmanTree tree_cl; /*tree for encoding the code lengths representing tree_ll and tree_d*/ - unsigned* frequencies_ll = 0; /*frequency of lit,len codes*/ - unsigned* frequencies_d = 0; /*frequency of dist codes*/ - unsigned* frequencies_cl = 0; /*frequency of code length codes*/ - unsigned* bitlen_lld = 0; /*lit,len,dist code lengths (int bits), literally (without repeat codes).*/ - unsigned* bitlen_lld_e = 0; /*bitlen_lld encoded with repeat codes (this is a rudimentary run length compression)*/ - size_t datasize = dataend - datapos; - - /* - If we could call "bitlen_cl" the the code length code lengths ("clcl"), that is the bit lengths of codes to represent - tree_cl in CLCL_ORDER, then due to the huffman compression of huffman tree representations ("two levels"), there are - some analogies: - bitlen_lld is to tree_cl what data is to tree_ll and tree_d. - bitlen_lld_e is to bitlen_lld what lz77_encoded is to data. - bitlen_cl is to bitlen_lld_e what bitlen_lld is to lz77_encoded. - */ - - unsigned BFINAL = final; - size_t i; - size_t numcodes_ll, numcodes_d, numcodes_lld, numcodes_lld_e, numcodes_cl; - unsigned HLIT, HDIST, HCLEN; - - uivector_init(&lz77_encoded); - HuffmanTree_init(&tree_ll); - HuffmanTree_init(&tree_d); - HuffmanTree_init(&tree_cl); - /* could fit on stack, but >1KB is on the larger side so allocate instead */ - frequencies_ll = (unsigned*)lodepng_malloc(286 * sizeof(*frequencies_ll)); - frequencies_d = (unsigned*)lodepng_malloc(30 * sizeof(*frequencies_d)); - frequencies_cl = (unsigned*)lodepng_malloc(NUM_CODE_LENGTH_CODES * sizeof(*frequencies_cl)); - - if(!frequencies_ll || !frequencies_d || !frequencies_cl) error = 83; /*alloc fail*/ - - /*This while loop never loops due to a break at the end, it is here to - allow breaking out of it to the cleanup phase on error conditions.*/ - while(!error) { - lodepng_memset(frequencies_ll, 0, 286 * sizeof(*frequencies_ll)); - lodepng_memset(frequencies_d, 0, 30 * sizeof(*frequencies_d)); - lodepng_memset(frequencies_cl, 0, NUM_CODE_LENGTH_CODES * sizeof(*frequencies_cl)); - - if(settings->use_lz77) { - error = encodeLZ77(&lz77_encoded, hash, data, datapos, dataend, settings->windowsize, - settings->minmatch, settings->nicematch, settings->lazymatching); - if(error) break; - } else { - if(!uivector_resize(&lz77_encoded, datasize)) ERROR_BREAK(83 /*alloc fail*/); - for(i = datapos; i < dataend; ++i) lz77_encoded.data[i - datapos] = data[i]; /*no LZ77, but still will be Huffman compressed*/ - } - - /*Count the frequencies of lit, len and dist codes*/ - for(i = 0; i != lz77_encoded.size; ++i) { - unsigned symbol = lz77_encoded.data[i]; - ++frequencies_ll[symbol]; - if(symbol > 256) { - unsigned dist = lz77_encoded.data[i + 2]; - ++frequencies_d[dist]; - i += 3; - } - } - frequencies_ll[256] = 1; /*there will be exactly 1 end code, at the end of the block*/ - - /*Make both huffman trees, one for the lit and len codes, one for the dist codes*/ - error = HuffmanTree_makeFromFrequencies(&tree_ll, frequencies_ll, 257, 286, 15); - if(error) break; - /*2, not 1, is chosen for mincodes: some buggy PNG decoders require at least 2 symbols in the dist tree*/ - error = HuffmanTree_makeFromFrequencies(&tree_d, frequencies_d, 2, 30, 15); - if(error) break; - - numcodes_ll = LODEPNG_MIN(tree_ll.numcodes, 286); - numcodes_d = LODEPNG_MIN(tree_d.numcodes, 30); - /*store the code lengths of both generated trees in bitlen_lld*/ - numcodes_lld = numcodes_ll + numcodes_d; - bitlen_lld = (unsigned*)lodepng_malloc(numcodes_lld * sizeof(*bitlen_lld)); - /*numcodes_lld_e never needs more size than bitlen_lld*/ - bitlen_lld_e = (unsigned*)lodepng_malloc(numcodes_lld * sizeof(*bitlen_lld_e)); - if(!bitlen_lld || !bitlen_lld_e) ERROR_BREAK(83); /*alloc fail*/ - numcodes_lld_e = 0; - - for(i = 0; i != numcodes_ll; ++i) bitlen_lld[i] = tree_ll.lengths[i]; - for(i = 0; i != numcodes_d; ++i) bitlen_lld[numcodes_ll + i] = tree_d.lengths[i]; - - /*run-length compress bitlen_ldd into bitlen_lld_e by using repeat codes 16 (copy length 3-6 times), - 17 (3-10 zeroes), 18 (11-138 zeroes)*/ - for(i = 0; i != numcodes_lld; ++i) { - unsigned j = 0; /*amount of repetitions*/ - while(i + j + 1 < numcodes_lld && bitlen_lld[i + j + 1] == bitlen_lld[i]) ++j; - - if(bitlen_lld[i] == 0 && j >= 2) /*repeat code for zeroes*/ { - ++j; /*include the first zero*/ - if(j <= 10) /*repeat code 17 supports max 10 zeroes*/ { - bitlen_lld_e[numcodes_lld_e++] = 17; - bitlen_lld_e[numcodes_lld_e++] = j - 3; - } else /*repeat code 18 supports max 138 zeroes*/ { - if(j > 138) j = 138; - bitlen_lld_e[numcodes_lld_e++] = 18; - bitlen_lld_e[numcodes_lld_e++] = j - 11; - } - i += (j - 1); - } else if(j >= 3) /*repeat code for value other than zero*/ { - size_t k; - unsigned num = j / 6u, rest = j % 6u; - bitlen_lld_e[numcodes_lld_e++] = bitlen_lld[i]; - for(k = 0; k < num; ++k) { - bitlen_lld_e[numcodes_lld_e++] = 16; - bitlen_lld_e[numcodes_lld_e++] = 6 - 3; - } - if(rest >= 3) { - bitlen_lld_e[numcodes_lld_e++] = 16; - bitlen_lld_e[numcodes_lld_e++] = rest - 3; - } - else j -= rest; - i += j; - } else /*too short to benefit from repeat code*/ { - bitlen_lld_e[numcodes_lld_e++] = bitlen_lld[i]; - } - } - - /*generate tree_cl, the huffmantree of huffmantrees*/ - for(i = 0; i != numcodes_lld_e; ++i) { - ++frequencies_cl[bitlen_lld_e[i]]; - /*after a repeat code come the bits that specify the number of repetitions, - those don't need to be in the frequencies_cl calculation*/ - if(bitlen_lld_e[i] >= 16) ++i; - } - - error = HuffmanTree_makeFromFrequencies(&tree_cl, frequencies_cl, - NUM_CODE_LENGTH_CODES, NUM_CODE_LENGTH_CODES, 7); - if(error) break; - - /*compute amount of code-length-code-lengths to output*/ - numcodes_cl = NUM_CODE_LENGTH_CODES; - /*trim zeros at the end (using CLCL_ORDER), but minimum size must be 4 (see HCLEN below)*/ - while(numcodes_cl > 4u && tree_cl.lengths[CLCL_ORDER[numcodes_cl - 1u]] == 0) { - numcodes_cl--; - } - - /* - Write everything into the output - - After the BFINAL and BTYPE, the dynamic block consists out of the following: - - 5 bits HLIT, 5 bits HDIST, 4 bits HCLEN - - (HCLEN+4)*3 bits code lengths of code length alphabet - - HLIT + 257 code lengths of lit/length alphabet (encoded using the code length - alphabet, + possible repetition codes 16, 17, 18) - - HDIST + 1 code lengths of distance alphabet (encoded using the code length - alphabet, + possible repetition codes 16, 17, 18) - - compressed data - - 256 (end code) - */ - - /*Write block type*/ - writeBits(writer, BFINAL, 1); - writeBits(writer, 0, 1); /*first bit of BTYPE "dynamic"*/ - writeBits(writer, 1, 1); /*second bit of BTYPE "dynamic"*/ - - /*write the HLIT, HDIST and HCLEN values*/ - /*all three sizes take trimmed ending zeroes into account, done either by HuffmanTree_makeFromFrequencies - or in the loop for numcodes_cl above, which saves space. */ - HLIT = (unsigned)(numcodes_ll - 257); - HDIST = (unsigned)(numcodes_d - 1); - HCLEN = (unsigned)(numcodes_cl - 4); - writeBits(writer, HLIT, 5); - writeBits(writer, HDIST, 5); - writeBits(writer, HCLEN, 4); - - /*write the code lengths of the code length alphabet ("bitlen_cl")*/ - for(i = 0; i != numcodes_cl; ++i) writeBits(writer, tree_cl.lengths[CLCL_ORDER[i]], 3); - - /*write the lengths of the lit/len AND the dist alphabet*/ - for(i = 0; i != numcodes_lld_e; ++i) { - writeBitsReversed(writer, tree_cl.codes[bitlen_lld_e[i]], tree_cl.lengths[bitlen_lld_e[i]]); - /*extra bits of repeat codes*/ - if(bitlen_lld_e[i] == 16) writeBits(writer, bitlen_lld_e[++i], 2); - else if(bitlen_lld_e[i] == 17) writeBits(writer, bitlen_lld_e[++i], 3); - else if(bitlen_lld_e[i] == 18) writeBits(writer, bitlen_lld_e[++i], 7); - } - - /*write the compressed data symbols*/ - writeLZ77data(writer, &lz77_encoded, &tree_ll, &tree_d); - /*error: the length of the end code 256 must be larger than 0*/ - if(tree_ll.lengths[256] == 0) ERROR_BREAK(64); - - /*write the end code*/ - writeBitsReversed(writer, tree_ll.codes[256], tree_ll.lengths[256]); - - break; /*end of error-while*/ - } - - /*cleanup*/ - uivector_cleanup(&lz77_encoded); - HuffmanTree_cleanup(&tree_ll); - HuffmanTree_cleanup(&tree_d); - HuffmanTree_cleanup(&tree_cl); - lodepng_free(frequencies_ll); - lodepng_free(frequencies_d); - lodepng_free(frequencies_cl); - lodepng_free(bitlen_lld); - lodepng_free(bitlen_lld_e); - - return error; -} - -static unsigned deflateFixed(LodePNGBitWriter* writer, Hash* hash, - const unsigned char* data, - size_t datapos, size_t dataend, - const LodePNGCompressSettings* settings, unsigned final) { - HuffmanTree tree_ll; /*tree for literal values and length codes*/ - HuffmanTree tree_d; /*tree for distance codes*/ - - unsigned BFINAL = final; - unsigned error = 0; - size_t i; - - HuffmanTree_init(&tree_ll); - HuffmanTree_init(&tree_d); - - error = generateFixedLitLenTree(&tree_ll); - if(!error) error = generateFixedDistanceTree(&tree_d); - - if(!error) { - writeBits(writer, BFINAL, 1); - writeBits(writer, 1, 1); /*first bit of BTYPE*/ - writeBits(writer, 0, 1); /*second bit of BTYPE*/ - - if(settings->use_lz77) /*LZ77 encoded*/ { - uivector lz77_encoded; - uivector_init(&lz77_encoded); - error = encodeLZ77(&lz77_encoded, hash, data, datapos, dataend, settings->windowsize, - settings->minmatch, settings->nicematch, settings->lazymatching); - if(!error) writeLZ77data(writer, &lz77_encoded, &tree_ll, &tree_d); - uivector_cleanup(&lz77_encoded); - } else /*no LZ77, but still will be Huffman compressed*/ { - for(i = datapos; i < dataend; ++i) { - writeBitsReversed(writer, tree_ll.codes[data[i]], tree_ll.lengths[data[i]]); - } - } - /*add END code*/ - if(!error) writeBitsReversed(writer,tree_ll.codes[256], tree_ll.lengths[256]); - } - - /*cleanup*/ - HuffmanTree_cleanup(&tree_ll); - HuffmanTree_cleanup(&tree_d); - - return error; -} - -static unsigned lodepng_deflatev(ucvector* out, const unsigned char* in, size_t insize, - const LodePNGCompressSettings* settings) { - unsigned error = 0; - size_t i, blocksize, numdeflateblocks; - Hash hash; - LodePNGBitWriter writer; - - LodePNGBitWriter_init(&writer, out); - - if(settings->btype > 2) return 61; - else if(settings->btype == 0) return deflateNoCompression(out, in, insize); - else if(settings->btype == 1) blocksize = insize; - else /*if(settings->btype == 2)*/ { - /*on PNGs, deflate blocks of 65-262k seem to give most dense encoding*/ - blocksize = insize / 8u + 8; - if(blocksize < 65536) blocksize = 65536; - if(blocksize > 262144) blocksize = 262144; - } - - numdeflateblocks = (insize + blocksize - 1) / blocksize; - if(numdeflateblocks == 0) numdeflateblocks = 1; - - error = hash_init(&hash, settings->windowsize); - - if(!error) { - for(i = 0; i != numdeflateblocks && !error; ++i) { - unsigned final = (i == numdeflateblocks - 1); - size_t start = i * blocksize; - size_t end = start + blocksize; - if(end > insize) end = insize; - - if(settings->btype == 1) error = deflateFixed(&writer, &hash, in, start, end, settings, final); - else if(settings->btype == 2) error = deflateDynamic(&writer, &hash, in, start, end, settings, final); - } - } - - hash_cleanup(&hash); - - return error; -} - -unsigned lodepng_deflate(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGCompressSettings* settings) { - ucvector v = ucvector_init(*out, *outsize); - unsigned error = lodepng_deflatev(&v, in, insize, settings); - *out = v.data; - *outsize = v.size; - return error; -} - -static unsigned deflate(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGCompressSettings* settings) { - if(settings->custom_deflate) { - unsigned error = settings->custom_deflate(out, outsize, in, insize, settings); - /*the custom deflate is allowed to have its own error codes, however, we translate it to code 111*/ - return error ? 111 : 0; - } else { - return lodepng_deflate(out, outsize, in, insize, settings); - } -} - -#endif /*LODEPNG_COMPILE_DECODER*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Adler32 / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -static unsigned update_adler32(unsigned adler, const unsigned char* data, unsigned len) { - unsigned s1 = adler & 0xffffu; - unsigned s2 = (adler >> 16u) & 0xffffu; - - while(len != 0u) { - unsigned i; - /*at least 5552 sums can be done before the sums overflow, saving a lot of module divisions*/ - unsigned amount = len > 5552u ? 5552u : len; - len -= amount; - for(i = 0; i != amount; ++i) { - s1 += (*data++); - s2 += s1; - } - s1 %= 65521u; - s2 %= 65521u; - } - - return (s2 << 16u) | s1; -} - -/*Return the adler32 of the bytes data[0..len-1]*/ -static unsigned adler32(const unsigned char* data, unsigned len) { - return update_adler32(1u, data, len); -} - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Zlib / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_DECODER - -static unsigned lodepng_zlib_decompressv(ucvector* out, - const unsigned char* in, size_t insize, - const LodePNGDecompressSettings* settings) { - unsigned error = 0; - unsigned CM, CINFO, FDICT; - - if(insize < 2) return 53; /*error, size of zlib data too small*/ - /*read information from zlib header*/ - if((in[0] * 256 + in[1]) % 31 != 0) { - /*error: 256 * in[0] + in[1] must be a multiple of 31, the FCHECK value is supposed to be made that way*/ - return 24; - } - - CM = in[0] & 15; - CINFO = (in[0] >> 4) & 15; - /*FCHECK = in[1] & 31;*/ /*FCHECK is already tested above*/ - FDICT = (in[1] >> 5) & 1; - /*FLEVEL = (in[1] >> 6) & 3;*/ /*FLEVEL is not used here*/ - - if(CM != 8 || CINFO > 7) { - /*error: only compression method 8: inflate with sliding window of 32k is supported by the PNG spec*/ - return 25; - } - if(FDICT != 0) { - /*error: the specification of PNG says about the zlib stream: - "The additional flags shall not specify a preset dictionary."*/ - return 26; - } - - error = inflatev(out, in + 2, insize - 2, settings); - if(error) return error; - - if(!settings->ignore_adler32) { - unsigned ADLER32 = lodepng_read32bitInt(&in[insize - 4]); - unsigned checksum = adler32(out->data, (unsigned)(out->size)); - if(checksum != ADLER32) return 58; /*error, adler checksum not correct, data must be corrupted*/ - } - - return 0; /*no error*/ -} - - -unsigned lodepng_zlib_decompress(unsigned char** out, size_t* outsize, const unsigned char* in, - size_t insize, const LodePNGDecompressSettings* settings) { - ucvector v = ucvector_init(*out, *outsize); - unsigned error = lodepng_zlib_decompressv(&v, in, insize, settings); - *out = v.data; - *outsize = v.size; - return error; -} - -/*expected_size is expected output size, to avoid intermediate allocations. Set to 0 if not known. */ -static unsigned zlib_decompress(unsigned char** out, size_t* outsize, size_t expected_size, - const unsigned char* in, size_t insize, const LodePNGDecompressSettings* settings) { - unsigned error; - if(settings->custom_zlib) { - error = settings->custom_zlib(out, outsize, in, insize, settings); - if(error) { - /*the custom zlib is allowed to have its own error codes, however, we translate it to code 110*/ - error = 110; - /*if there's a max output size, and the custom zlib returned error, then indicate that error instead*/ - if(settings->max_output_size && *outsize > settings->max_output_size) error = 109; - } - } else { - ucvector v = ucvector_init(*out, *outsize); - if(expected_size) { - /*reserve the memory to avoid intermediate reallocations*/ - ucvector_resize(&v, *outsize + expected_size); - v.size = *outsize; - } - error = lodepng_zlib_decompressv(&v, in, insize, settings); - *out = v.data; - *outsize = v.size; - } - return error; -} - -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_ENCODER - -unsigned lodepng_zlib_compress(unsigned char** out, size_t* outsize, const unsigned char* in, - size_t insize, const LodePNGCompressSettings* settings) { - size_t i; - unsigned error; - unsigned char* deflatedata = 0; - size_t deflatesize = 0; - - error = deflate(&deflatedata, &deflatesize, in, insize, settings); - - *out = NULL; - *outsize = 0; - if(!error) { - *outsize = deflatesize + 6; - *out = (unsigned char*)lodepng_malloc(*outsize); - if(!*out) error = 83; /*alloc fail*/ - } - - if(!error) { - unsigned ADLER32 = adler32(in, (unsigned)insize); - /*zlib data: 1 byte CMF (CM+CINFO), 1 byte FLG, deflate data, 4 byte ADLER32 checksum of the Decompressed data*/ - unsigned CMF = 120; /*0b01111000: CM 8, CINFO 7. With CINFO 7, any window size up to 32768 can be used.*/ - unsigned FLEVEL = 0; - unsigned FDICT = 0; - unsigned CMFFLG = 256 * CMF + FDICT * 32 + FLEVEL * 64; - unsigned FCHECK = 31 - CMFFLG % 31; - CMFFLG += FCHECK; - - (*out)[0] = (unsigned char)(CMFFLG >> 8); - (*out)[1] = (unsigned char)(CMFFLG & 255); - for(i = 0; i != deflatesize; ++i) (*out)[i + 2] = deflatedata[i]; - lodepng_set32bitInt(&(*out)[*outsize - 4], ADLER32); - } - - lodepng_free(deflatedata); - return error; -} - -/* compress using the default or custom zlib function */ -static unsigned zlib_compress(unsigned char** out, size_t* outsize, const unsigned char* in, - size_t insize, const LodePNGCompressSettings* settings) { - if(settings->custom_zlib) { - unsigned error = settings->custom_zlib(out, outsize, in, insize, settings); - /*the custom zlib is allowed to have its own error codes, however, we translate it to code 111*/ - return error ? 111 : 0; - } else { - return lodepng_zlib_compress(out, outsize, in, insize, settings); - } -} - -#endif /*LODEPNG_COMPILE_ENCODER*/ - -#else /*no LODEPNG_COMPILE_ZLIB*/ - -#ifdef LODEPNG_COMPILE_DECODER -static unsigned zlib_decompress(unsigned char** out, size_t* outsize, size_t expected_size, - const unsigned char* in, size_t insize, const LodePNGDecompressSettings* settings) { - if(!settings->custom_zlib) return 87; /*no custom zlib function provided */ - (void)expected_size; - return settings->custom_zlib(out, outsize, in, insize, settings); -} -#endif /*LODEPNG_COMPILE_DECODER*/ -#ifdef LODEPNG_COMPILE_ENCODER -static unsigned zlib_compress(unsigned char** out, size_t* outsize, const unsigned char* in, - size_t insize, const LodePNGCompressSettings* settings) { - if(!settings->custom_zlib) return 87; /*no custom zlib function provided */ - return settings->custom_zlib(out, outsize, in, insize, settings); -} -#endif /*LODEPNG_COMPILE_ENCODER*/ - -#endif /*LODEPNG_COMPILE_ZLIB*/ - -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_ENCODER - -/*this is a good tradeoff between speed and compression ratio*/ -#define DEFAULT_WINDOWSIZE 2048 - -void lodepng_compress_settings_init(LodePNGCompressSettings* settings) { - /*compress with dynamic huffman tree (not in the mathematical sense, just not the predefined one)*/ - settings->btype = 2; - settings->use_lz77 = 1; - settings->windowsize = DEFAULT_WINDOWSIZE; - settings->minmatch = 3; - settings->nicematch = 128; - settings->lazymatching = 1; - - settings->custom_zlib = 0; - settings->custom_deflate = 0; - settings->custom_context = 0; -} - -const LodePNGCompressSettings lodepng_default_compress_settings = {2, 1, DEFAULT_WINDOWSIZE, 3, 128, 1, 0, 0, 0}; - - -#endif /*LODEPNG_COMPILE_ENCODER*/ - -#ifdef LODEPNG_COMPILE_DECODER - -void lodepng_decompress_settings_init(LodePNGDecompressSettings* settings) { - settings->ignore_adler32 = 0; - settings->ignore_nlen = 0; - settings->max_output_size = 0; - - settings->custom_zlib = 0; - settings->custom_inflate = 0; - settings->custom_context = 0; -} - -const LodePNGDecompressSettings lodepng_default_decompress_settings = {0, 0, 0, 0, 0, 0}; - -#endif /*LODEPNG_COMPILE_DECODER*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* // End of Zlib related code. Begin of PNG related code. // */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_PNG - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / CRC32 / */ -/* ////////////////////////////////////////////////////////////////////////// */ - - -#ifdef LODEPNG_COMPILE_CRC - -static const unsigned lodepng_crc32_table0[256] = { - 0x00000000u, 0x77073096u, 0xee0e612cu, 0x990951bau, 0x076dc419u, 0x706af48fu, 0xe963a535u, 0x9e6495a3u, - 0x0edb8832u, 0x79dcb8a4u, 0xe0d5e91eu, 0x97d2d988u, 0x09b64c2bu, 0x7eb17cbdu, 0xe7b82d07u, 0x90bf1d91u, - 0x1db71064u, 0x6ab020f2u, 0xf3b97148u, 0x84be41deu, 0x1adad47du, 0x6ddde4ebu, 0xf4d4b551u, 0x83d385c7u, - 0x136c9856u, 0x646ba8c0u, 0xfd62f97au, 0x8a65c9ecu, 0x14015c4fu, 0x63066cd9u, 0xfa0f3d63u, 0x8d080df5u, - 0x3b6e20c8u, 0x4c69105eu, 0xd56041e4u, 0xa2677172u, 0x3c03e4d1u, 0x4b04d447u, 0xd20d85fdu, 0xa50ab56bu, - 0x35b5a8fau, 0x42b2986cu, 0xdbbbc9d6u, 0xacbcf940u, 0x32d86ce3u, 0x45df5c75u, 0xdcd60dcfu, 0xabd13d59u, - 0x26d930acu, 0x51de003au, 0xc8d75180u, 0xbfd06116u, 0x21b4f4b5u, 0x56b3c423u, 0xcfba9599u, 0xb8bda50fu, - 0x2802b89eu, 0x5f058808u, 0xc60cd9b2u, 0xb10be924u, 0x2f6f7c87u, 0x58684c11u, 0xc1611dabu, 0xb6662d3du, - 0x76dc4190u, 0x01db7106u, 0x98d220bcu, 0xefd5102au, 0x71b18589u, 0x06b6b51fu, 0x9fbfe4a5u, 0xe8b8d433u, - 0x7807c9a2u, 0x0f00f934u, 0x9609a88eu, 0xe10e9818u, 0x7f6a0dbbu, 0x086d3d2du, 0x91646c97u, 0xe6635c01u, - 0x6b6b51f4u, 0x1c6c6162u, 0x856530d8u, 0xf262004eu, 0x6c0695edu, 0x1b01a57bu, 0x8208f4c1u, 0xf50fc457u, - 0x65b0d9c6u, 0x12b7e950u, 0x8bbeb8eau, 0xfcb9887cu, 0x62dd1ddfu, 0x15da2d49u, 0x8cd37cf3u, 0xfbd44c65u, - 0x4db26158u, 0x3ab551ceu, 0xa3bc0074u, 0xd4bb30e2u, 0x4adfa541u, 0x3dd895d7u, 0xa4d1c46du, 0xd3d6f4fbu, - 0x4369e96au, 0x346ed9fcu, 0xad678846u, 0xda60b8d0u, 0x44042d73u, 0x33031de5u, 0xaa0a4c5fu, 0xdd0d7cc9u, - 0x5005713cu, 0x270241aau, 0xbe0b1010u, 0xc90c2086u, 0x5768b525u, 0x206f85b3u, 0xb966d409u, 0xce61e49fu, - 0x5edef90eu, 0x29d9c998u, 0xb0d09822u, 0xc7d7a8b4u, 0x59b33d17u, 0x2eb40d81u, 0xb7bd5c3bu, 0xc0ba6cadu, - 0xedb88320u, 0x9abfb3b6u, 0x03b6e20cu, 0x74b1d29au, 0xead54739u, 0x9dd277afu, 0x04db2615u, 0x73dc1683u, - 0xe3630b12u, 0x94643b84u, 0x0d6d6a3eu, 0x7a6a5aa8u, 0xe40ecf0bu, 0x9309ff9du, 0x0a00ae27u, 0x7d079eb1u, - 0xf00f9344u, 0x8708a3d2u, 0x1e01f268u, 0x6906c2feu, 0xf762575du, 0x806567cbu, 0x196c3671u, 0x6e6b06e7u, - 0xfed41b76u, 0x89d32be0u, 0x10da7a5au, 0x67dd4accu, 0xf9b9df6fu, 0x8ebeeff9u, 0x17b7be43u, 0x60b08ed5u, - 0xd6d6a3e8u, 0xa1d1937eu, 0x38d8c2c4u, 0x4fdff252u, 0xd1bb67f1u, 0xa6bc5767u, 0x3fb506ddu, 0x48b2364bu, - 0xd80d2bdau, 0xaf0a1b4cu, 0x36034af6u, 0x41047a60u, 0xdf60efc3u, 0xa867df55u, 0x316e8eefu, 0x4669be79u, - 0xcb61b38cu, 0xbc66831au, 0x256fd2a0u, 0x5268e236u, 0xcc0c7795u, 0xbb0b4703u, 0x220216b9u, 0x5505262fu, - 0xc5ba3bbeu, 0xb2bd0b28u, 0x2bb45a92u, 0x5cb36a04u, 0xc2d7ffa7u, 0xb5d0cf31u, 0x2cd99e8bu, 0x5bdeae1du, - 0x9b64c2b0u, 0xec63f226u, 0x756aa39cu, 0x026d930au, 0x9c0906a9u, 0xeb0e363fu, 0x72076785u, 0x05005713u, - 0x95bf4a82u, 0xe2b87a14u, 0x7bb12baeu, 0x0cb61b38u, 0x92d28e9bu, 0xe5d5be0du, 0x7cdcefb7u, 0x0bdbdf21u, - 0x86d3d2d4u, 0xf1d4e242u, 0x68ddb3f8u, 0x1fda836eu, 0x81be16cdu, 0xf6b9265bu, 0x6fb077e1u, 0x18b74777u, - 0x88085ae6u, 0xff0f6a70u, 0x66063bcau, 0x11010b5cu, 0x8f659effu, 0xf862ae69u, 0x616bffd3u, 0x166ccf45u, - 0xa00ae278u, 0xd70dd2eeu, 0x4e048354u, 0x3903b3c2u, 0xa7672661u, 0xd06016f7u, 0x4969474du, 0x3e6e77dbu, - 0xaed16a4au, 0xd9d65adcu, 0x40df0b66u, 0x37d83bf0u, 0xa9bcae53u, 0xdebb9ec5u, 0x47b2cf7fu, 0x30b5ffe9u, - 0xbdbdf21cu, 0xcabac28au, 0x53b39330u, 0x24b4a3a6u, 0xbad03605u, 0xcdd70693u, 0x54de5729u, 0x23d967bfu, - 0xb3667a2eu, 0xc4614ab8u, 0x5d681b02u, 0x2a6f2b94u, 0xb40bbe37u, 0xc30c8ea1u, 0x5a05df1bu, 0x2d02ef8du -}; - -static const unsigned lodepng_crc32_table1[256] = { - 0x00000000u, 0x191b3141u, 0x32366282u, 0x2b2d53c3u, 0x646cc504u, 0x7d77f445u, 0x565aa786u, 0x4f4196c7u, - 0xc8d98a08u, 0xd1c2bb49u, 0xfaefe88au, 0xe3f4d9cbu, 0xacb54f0cu, 0xb5ae7e4du, 0x9e832d8eu, 0x87981ccfu, - 0x4ac21251u, 0x53d92310u, 0x78f470d3u, 0x61ef4192u, 0x2eaed755u, 0x37b5e614u, 0x1c98b5d7u, 0x05838496u, - 0x821b9859u, 0x9b00a918u, 0xb02dfadbu, 0xa936cb9au, 0xe6775d5du, 0xff6c6c1cu, 0xd4413fdfu, 0xcd5a0e9eu, - 0x958424a2u, 0x8c9f15e3u, 0xa7b24620u, 0xbea97761u, 0xf1e8e1a6u, 0xe8f3d0e7u, 0xc3de8324u, 0xdac5b265u, - 0x5d5daeaau, 0x44469febu, 0x6f6bcc28u, 0x7670fd69u, 0x39316baeu, 0x202a5aefu, 0x0b07092cu, 0x121c386du, - 0xdf4636f3u, 0xc65d07b2u, 0xed705471u, 0xf46b6530u, 0xbb2af3f7u, 0xa231c2b6u, 0x891c9175u, 0x9007a034u, - 0x179fbcfbu, 0x0e848dbau, 0x25a9de79u, 0x3cb2ef38u, 0x73f379ffu, 0x6ae848beu, 0x41c51b7du, 0x58de2a3cu, - 0xf0794f05u, 0xe9627e44u, 0xc24f2d87u, 0xdb541cc6u, 0x94158a01u, 0x8d0ebb40u, 0xa623e883u, 0xbf38d9c2u, - 0x38a0c50du, 0x21bbf44cu, 0x0a96a78fu, 0x138d96ceu, 0x5ccc0009u, 0x45d73148u, 0x6efa628bu, 0x77e153cau, - 0xbabb5d54u, 0xa3a06c15u, 0x888d3fd6u, 0x91960e97u, 0xded79850u, 0xc7cca911u, 0xece1fad2u, 0xf5facb93u, - 0x7262d75cu, 0x6b79e61du, 0x4054b5deu, 0x594f849fu, 0x160e1258u, 0x0f152319u, 0x243870dau, 0x3d23419bu, - 0x65fd6ba7u, 0x7ce65ae6u, 0x57cb0925u, 0x4ed03864u, 0x0191aea3u, 0x188a9fe2u, 0x33a7cc21u, 0x2abcfd60u, - 0xad24e1afu, 0xb43fd0eeu, 0x9f12832du, 0x8609b26cu, 0xc94824abu, 0xd05315eau, 0xfb7e4629u, 0xe2657768u, - 0x2f3f79f6u, 0x362448b7u, 0x1d091b74u, 0x04122a35u, 0x4b53bcf2u, 0x52488db3u, 0x7965de70u, 0x607eef31u, - 0xe7e6f3feu, 0xfefdc2bfu, 0xd5d0917cu, 0xcccba03du, 0x838a36fau, 0x9a9107bbu, 0xb1bc5478u, 0xa8a76539u, - 0x3b83984bu, 0x2298a90au, 0x09b5fac9u, 0x10aecb88u, 0x5fef5d4fu, 0x46f46c0eu, 0x6dd93fcdu, 0x74c20e8cu, - 0xf35a1243u, 0xea412302u, 0xc16c70c1u, 0xd8774180u, 0x9736d747u, 0x8e2de606u, 0xa500b5c5u, 0xbc1b8484u, - 0x71418a1au, 0x685abb5bu, 0x4377e898u, 0x5a6cd9d9u, 0x152d4f1eu, 0x0c367e5fu, 0x271b2d9cu, 0x3e001cddu, - 0xb9980012u, 0xa0833153u, 0x8bae6290u, 0x92b553d1u, 0xddf4c516u, 0xc4eff457u, 0xefc2a794u, 0xf6d996d5u, - 0xae07bce9u, 0xb71c8da8u, 0x9c31de6bu, 0x852aef2au, 0xca6b79edu, 0xd37048acu, 0xf85d1b6fu, 0xe1462a2eu, - 0x66de36e1u, 0x7fc507a0u, 0x54e85463u, 0x4df36522u, 0x02b2f3e5u, 0x1ba9c2a4u, 0x30849167u, 0x299fa026u, - 0xe4c5aeb8u, 0xfdde9ff9u, 0xd6f3cc3au, 0xcfe8fd7bu, 0x80a96bbcu, 0x99b25afdu, 0xb29f093eu, 0xab84387fu, - 0x2c1c24b0u, 0x350715f1u, 0x1e2a4632u, 0x07317773u, 0x4870e1b4u, 0x516bd0f5u, 0x7a468336u, 0x635db277u, - 0xcbfad74eu, 0xd2e1e60fu, 0xf9ccb5ccu, 0xe0d7848du, 0xaf96124au, 0xb68d230bu, 0x9da070c8u, 0x84bb4189u, - 0x03235d46u, 0x1a386c07u, 0x31153fc4u, 0x280e0e85u, 0x674f9842u, 0x7e54a903u, 0x5579fac0u, 0x4c62cb81u, - 0x8138c51fu, 0x9823f45eu, 0xb30ea79du, 0xaa1596dcu, 0xe554001bu, 0xfc4f315au, 0xd7626299u, 0xce7953d8u, - 0x49e14f17u, 0x50fa7e56u, 0x7bd72d95u, 0x62cc1cd4u, 0x2d8d8a13u, 0x3496bb52u, 0x1fbbe891u, 0x06a0d9d0u, - 0x5e7ef3ecu, 0x4765c2adu, 0x6c48916eu, 0x7553a02fu, 0x3a1236e8u, 0x230907a9u, 0x0824546au, 0x113f652bu, - 0x96a779e4u, 0x8fbc48a5u, 0xa4911b66u, 0xbd8a2a27u, 0xf2cbbce0u, 0xebd08da1u, 0xc0fdde62u, 0xd9e6ef23u, - 0x14bce1bdu, 0x0da7d0fcu, 0x268a833fu, 0x3f91b27eu, 0x70d024b9u, 0x69cb15f8u, 0x42e6463bu, 0x5bfd777au, - 0xdc656bb5u, 0xc57e5af4u, 0xee530937u, 0xf7483876u, 0xb809aeb1u, 0xa1129ff0u, 0x8a3fcc33u, 0x9324fd72u -}; - -static const unsigned lodepng_crc32_table2[256] = { - 0x00000000u, 0x01c26a37u, 0x0384d46eu, 0x0246be59u, 0x0709a8dcu, 0x06cbc2ebu, 0x048d7cb2u, 0x054f1685u, - 0x0e1351b8u, 0x0fd13b8fu, 0x0d9785d6u, 0x0c55efe1u, 0x091af964u, 0x08d89353u, 0x0a9e2d0au, 0x0b5c473du, - 0x1c26a370u, 0x1de4c947u, 0x1fa2771eu, 0x1e601d29u, 0x1b2f0bacu, 0x1aed619bu, 0x18abdfc2u, 0x1969b5f5u, - 0x1235f2c8u, 0x13f798ffu, 0x11b126a6u, 0x10734c91u, 0x153c5a14u, 0x14fe3023u, 0x16b88e7au, 0x177ae44du, - 0x384d46e0u, 0x398f2cd7u, 0x3bc9928eu, 0x3a0bf8b9u, 0x3f44ee3cu, 0x3e86840bu, 0x3cc03a52u, 0x3d025065u, - 0x365e1758u, 0x379c7d6fu, 0x35dac336u, 0x3418a901u, 0x3157bf84u, 0x3095d5b3u, 0x32d36beau, 0x331101ddu, - 0x246be590u, 0x25a98fa7u, 0x27ef31feu, 0x262d5bc9u, 0x23624d4cu, 0x22a0277bu, 0x20e69922u, 0x2124f315u, - 0x2a78b428u, 0x2bbade1fu, 0x29fc6046u, 0x283e0a71u, 0x2d711cf4u, 0x2cb376c3u, 0x2ef5c89au, 0x2f37a2adu, - 0x709a8dc0u, 0x7158e7f7u, 0x731e59aeu, 0x72dc3399u, 0x7793251cu, 0x76514f2bu, 0x7417f172u, 0x75d59b45u, - 0x7e89dc78u, 0x7f4bb64fu, 0x7d0d0816u, 0x7ccf6221u, 0x798074a4u, 0x78421e93u, 0x7a04a0cau, 0x7bc6cafdu, - 0x6cbc2eb0u, 0x6d7e4487u, 0x6f38fadeu, 0x6efa90e9u, 0x6bb5866cu, 0x6a77ec5bu, 0x68315202u, 0x69f33835u, - 0x62af7f08u, 0x636d153fu, 0x612bab66u, 0x60e9c151u, 0x65a6d7d4u, 0x6464bde3u, 0x662203bau, 0x67e0698du, - 0x48d7cb20u, 0x4915a117u, 0x4b531f4eu, 0x4a917579u, 0x4fde63fcu, 0x4e1c09cbu, 0x4c5ab792u, 0x4d98dda5u, - 0x46c49a98u, 0x4706f0afu, 0x45404ef6u, 0x448224c1u, 0x41cd3244u, 0x400f5873u, 0x4249e62au, 0x438b8c1du, - 0x54f16850u, 0x55330267u, 0x5775bc3eu, 0x56b7d609u, 0x53f8c08cu, 0x523aaabbu, 0x507c14e2u, 0x51be7ed5u, - 0x5ae239e8u, 0x5b2053dfu, 0x5966ed86u, 0x58a487b1u, 0x5deb9134u, 0x5c29fb03u, 0x5e6f455au, 0x5fad2f6du, - 0xe1351b80u, 0xe0f771b7u, 0xe2b1cfeeu, 0xe373a5d9u, 0xe63cb35cu, 0xe7fed96bu, 0xe5b86732u, 0xe47a0d05u, - 0xef264a38u, 0xeee4200fu, 0xeca29e56u, 0xed60f461u, 0xe82fe2e4u, 0xe9ed88d3u, 0xebab368au, 0xea695cbdu, - 0xfd13b8f0u, 0xfcd1d2c7u, 0xfe976c9eu, 0xff5506a9u, 0xfa1a102cu, 0xfbd87a1bu, 0xf99ec442u, 0xf85cae75u, - 0xf300e948u, 0xf2c2837fu, 0xf0843d26u, 0xf1465711u, 0xf4094194u, 0xf5cb2ba3u, 0xf78d95fau, 0xf64fffcdu, - 0xd9785d60u, 0xd8ba3757u, 0xdafc890eu, 0xdb3ee339u, 0xde71f5bcu, 0xdfb39f8bu, 0xddf521d2u, 0xdc374be5u, - 0xd76b0cd8u, 0xd6a966efu, 0xd4efd8b6u, 0xd52db281u, 0xd062a404u, 0xd1a0ce33u, 0xd3e6706au, 0xd2241a5du, - 0xc55efe10u, 0xc49c9427u, 0xc6da2a7eu, 0xc7184049u, 0xc25756ccu, 0xc3953cfbu, 0xc1d382a2u, 0xc011e895u, - 0xcb4dafa8u, 0xca8fc59fu, 0xc8c97bc6u, 0xc90b11f1u, 0xcc440774u, 0xcd866d43u, 0xcfc0d31au, 0xce02b92du, - 0x91af9640u, 0x906dfc77u, 0x922b422eu, 0x93e92819u, 0x96a63e9cu, 0x976454abu, 0x9522eaf2u, 0x94e080c5u, - 0x9fbcc7f8u, 0x9e7eadcfu, 0x9c381396u, 0x9dfa79a1u, 0x98b56f24u, 0x99770513u, 0x9b31bb4au, 0x9af3d17du, - 0x8d893530u, 0x8c4b5f07u, 0x8e0de15eu, 0x8fcf8b69u, 0x8a809decu, 0x8b42f7dbu, 0x89044982u, 0x88c623b5u, - 0x839a6488u, 0x82580ebfu, 0x801eb0e6u, 0x81dcdad1u, 0x8493cc54u, 0x8551a663u, 0x8717183au, 0x86d5720du, - 0xa9e2d0a0u, 0xa820ba97u, 0xaa6604ceu, 0xaba46ef9u, 0xaeeb787cu, 0xaf29124bu, 0xad6fac12u, 0xacadc625u, - 0xa7f18118u, 0xa633eb2fu, 0xa4755576u, 0xa5b73f41u, 0xa0f829c4u, 0xa13a43f3u, 0xa37cfdaau, 0xa2be979du, - 0xb5c473d0u, 0xb40619e7u, 0xb640a7beu, 0xb782cd89u, 0xb2cddb0cu, 0xb30fb13bu, 0xb1490f62u, 0xb08b6555u, - 0xbbd72268u, 0xba15485fu, 0xb853f606u, 0xb9919c31u, 0xbcde8ab4u, 0xbd1ce083u, 0xbf5a5edau, 0xbe9834edu -}; - -static const unsigned lodepng_crc32_table3[256] = { - 0x00000000u, 0xb8bc6765u, 0xaa09c88bu, 0x12b5afeeu, 0x8f629757u, 0x37def032u, 0x256b5fdcu, 0x9dd738b9u, - 0xc5b428efu, 0x7d084f8au, 0x6fbde064u, 0xd7018701u, 0x4ad6bfb8u, 0xf26ad8ddu, 0xe0df7733u, 0x58631056u, - 0x5019579fu, 0xe8a530fau, 0xfa109f14u, 0x42acf871u, 0xdf7bc0c8u, 0x67c7a7adu, 0x75720843u, 0xcdce6f26u, - 0x95ad7f70u, 0x2d111815u, 0x3fa4b7fbu, 0x8718d09eu, 0x1acfe827u, 0xa2738f42u, 0xb0c620acu, 0x087a47c9u, - 0xa032af3eu, 0x188ec85bu, 0x0a3b67b5u, 0xb28700d0u, 0x2f503869u, 0x97ec5f0cu, 0x8559f0e2u, 0x3de59787u, - 0x658687d1u, 0xdd3ae0b4u, 0xcf8f4f5au, 0x7733283fu, 0xeae41086u, 0x525877e3u, 0x40edd80du, 0xf851bf68u, - 0xf02bf8a1u, 0x48979fc4u, 0x5a22302au, 0xe29e574fu, 0x7f496ff6u, 0xc7f50893u, 0xd540a77du, 0x6dfcc018u, - 0x359fd04eu, 0x8d23b72bu, 0x9f9618c5u, 0x272a7fa0u, 0xbafd4719u, 0x0241207cu, 0x10f48f92u, 0xa848e8f7u, - 0x9b14583du, 0x23a83f58u, 0x311d90b6u, 0x89a1f7d3u, 0x1476cf6au, 0xaccaa80fu, 0xbe7f07e1u, 0x06c36084u, - 0x5ea070d2u, 0xe61c17b7u, 0xf4a9b859u, 0x4c15df3cu, 0xd1c2e785u, 0x697e80e0u, 0x7bcb2f0eu, 0xc377486bu, - 0xcb0d0fa2u, 0x73b168c7u, 0x6104c729u, 0xd9b8a04cu, 0x446f98f5u, 0xfcd3ff90u, 0xee66507eu, 0x56da371bu, - 0x0eb9274du, 0xb6054028u, 0xa4b0efc6u, 0x1c0c88a3u, 0x81dbb01au, 0x3967d77fu, 0x2bd27891u, 0x936e1ff4u, - 0x3b26f703u, 0x839a9066u, 0x912f3f88u, 0x299358edu, 0xb4446054u, 0x0cf80731u, 0x1e4da8dfu, 0xa6f1cfbau, - 0xfe92dfecu, 0x462eb889u, 0x549b1767u, 0xec277002u, 0x71f048bbu, 0xc94c2fdeu, 0xdbf98030u, 0x6345e755u, - 0x6b3fa09cu, 0xd383c7f9u, 0xc1366817u, 0x798a0f72u, 0xe45d37cbu, 0x5ce150aeu, 0x4e54ff40u, 0xf6e89825u, - 0xae8b8873u, 0x1637ef16u, 0x048240f8u, 0xbc3e279du, 0x21e91f24u, 0x99557841u, 0x8be0d7afu, 0x335cb0cau, - 0xed59b63bu, 0x55e5d15eu, 0x47507eb0u, 0xffec19d5u, 0x623b216cu, 0xda874609u, 0xc832e9e7u, 0x708e8e82u, - 0x28ed9ed4u, 0x9051f9b1u, 0x82e4565fu, 0x3a58313au, 0xa78f0983u, 0x1f336ee6u, 0x0d86c108u, 0xb53aa66du, - 0xbd40e1a4u, 0x05fc86c1u, 0x1749292fu, 0xaff54e4au, 0x322276f3u, 0x8a9e1196u, 0x982bbe78u, 0x2097d91du, - 0x78f4c94bu, 0xc048ae2eu, 0xd2fd01c0u, 0x6a4166a5u, 0xf7965e1cu, 0x4f2a3979u, 0x5d9f9697u, 0xe523f1f2u, - 0x4d6b1905u, 0xf5d77e60u, 0xe762d18eu, 0x5fdeb6ebu, 0xc2098e52u, 0x7ab5e937u, 0x680046d9u, 0xd0bc21bcu, - 0x88df31eau, 0x3063568fu, 0x22d6f961u, 0x9a6a9e04u, 0x07bda6bdu, 0xbf01c1d8u, 0xadb46e36u, 0x15080953u, - 0x1d724e9au, 0xa5ce29ffu, 0xb77b8611u, 0x0fc7e174u, 0x9210d9cdu, 0x2aacbea8u, 0x38191146u, 0x80a57623u, - 0xd8c66675u, 0x607a0110u, 0x72cfaefeu, 0xca73c99bu, 0x57a4f122u, 0xef189647u, 0xfdad39a9u, 0x45115eccu, - 0x764dee06u, 0xcef18963u, 0xdc44268du, 0x64f841e8u, 0xf92f7951u, 0x41931e34u, 0x5326b1dau, 0xeb9ad6bfu, - 0xb3f9c6e9u, 0x0b45a18cu, 0x19f00e62u, 0xa14c6907u, 0x3c9b51beu, 0x842736dbu, 0x96929935u, 0x2e2efe50u, - 0x2654b999u, 0x9ee8defcu, 0x8c5d7112u, 0x34e11677u, 0xa9362eceu, 0x118a49abu, 0x033fe645u, 0xbb838120u, - 0xe3e09176u, 0x5b5cf613u, 0x49e959fdu, 0xf1553e98u, 0x6c820621u, 0xd43e6144u, 0xc68bceaau, 0x7e37a9cfu, - 0xd67f4138u, 0x6ec3265du, 0x7c7689b3u, 0xc4caeed6u, 0x591dd66fu, 0xe1a1b10au, 0xf3141ee4u, 0x4ba87981u, - 0x13cb69d7u, 0xab770eb2u, 0xb9c2a15cu, 0x017ec639u, 0x9ca9fe80u, 0x241599e5u, 0x36a0360bu, 0x8e1c516eu, - 0x866616a7u, 0x3eda71c2u, 0x2c6fde2cu, 0x94d3b949u, 0x090481f0u, 0xb1b8e695u, 0xa30d497bu, 0x1bb12e1eu, - 0x43d23e48u, 0xfb6e592du, 0xe9dbf6c3u, 0x516791a6u, 0xccb0a91fu, 0x740cce7au, 0x66b96194u, 0xde0506f1u -}; - -static const unsigned lodepng_crc32_table4[256] = { - 0x00000000u, 0x3d6029b0u, 0x7ac05360u, 0x47a07ad0u, 0xf580a6c0u, 0xc8e08f70u, 0x8f40f5a0u, 0xb220dc10u, - 0x30704bc1u, 0x0d106271u, 0x4ab018a1u, 0x77d03111u, 0xc5f0ed01u, 0xf890c4b1u, 0xbf30be61u, 0x825097d1u, - 0x60e09782u, 0x5d80be32u, 0x1a20c4e2u, 0x2740ed52u, 0x95603142u, 0xa80018f2u, 0xefa06222u, 0xd2c04b92u, - 0x5090dc43u, 0x6df0f5f3u, 0x2a508f23u, 0x1730a693u, 0xa5107a83u, 0x98705333u, 0xdfd029e3u, 0xe2b00053u, - 0xc1c12f04u, 0xfca106b4u, 0xbb017c64u, 0x866155d4u, 0x344189c4u, 0x0921a074u, 0x4e81daa4u, 0x73e1f314u, - 0xf1b164c5u, 0xccd14d75u, 0x8b7137a5u, 0xb6111e15u, 0x0431c205u, 0x3951ebb5u, 0x7ef19165u, 0x4391b8d5u, - 0xa121b886u, 0x9c419136u, 0xdbe1ebe6u, 0xe681c256u, 0x54a11e46u, 0x69c137f6u, 0x2e614d26u, 0x13016496u, - 0x9151f347u, 0xac31daf7u, 0xeb91a027u, 0xd6f18997u, 0x64d15587u, 0x59b17c37u, 0x1e1106e7u, 0x23712f57u, - 0x58f35849u, 0x659371f9u, 0x22330b29u, 0x1f532299u, 0xad73fe89u, 0x9013d739u, 0xd7b3ade9u, 0xead38459u, - 0x68831388u, 0x55e33a38u, 0x124340e8u, 0x2f236958u, 0x9d03b548u, 0xa0639cf8u, 0xe7c3e628u, 0xdaa3cf98u, - 0x3813cfcbu, 0x0573e67bu, 0x42d39cabu, 0x7fb3b51bu, 0xcd93690bu, 0xf0f340bbu, 0xb7533a6bu, 0x8a3313dbu, - 0x0863840au, 0x3503adbau, 0x72a3d76au, 0x4fc3fedau, 0xfde322cau, 0xc0830b7au, 0x872371aau, 0xba43581au, - 0x9932774du, 0xa4525efdu, 0xe3f2242du, 0xde920d9du, 0x6cb2d18du, 0x51d2f83du, 0x167282edu, 0x2b12ab5du, - 0xa9423c8cu, 0x9422153cu, 0xd3826fecu, 0xeee2465cu, 0x5cc29a4cu, 0x61a2b3fcu, 0x2602c92cu, 0x1b62e09cu, - 0xf9d2e0cfu, 0xc4b2c97fu, 0x8312b3afu, 0xbe729a1fu, 0x0c52460fu, 0x31326fbfu, 0x7692156fu, 0x4bf23cdfu, - 0xc9a2ab0eu, 0xf4c282beu, 0xb362f86eu, 0x8e02d1deu, 0x3c220dceu, 0x0142247eu, 0x46e25eaeu, 0x7b82771eu, - 0xb1e6b092u, 0x8c869922u, 0xcb26e3f2u, 0xf646ca42u, 0x44661652u, 0x79063fe2u, 0x3ea64532u, 0x03c66c82u, - 0x8196fb53u, 0xbcf6d2e3u, 0xfb56a833u, 0xc6368183u, 0x74165d93u, 0x49767423u, 0x0ed60ef3u, 0x33b62743u, - 0xd1062710u, 0xec660ea0u, 0xabc67470u, 0x96a65dc0u, 0x248681d0u, 0x19e6a860u, 0x5e46d2b0u, 0x6326fb00u, - 0xe1766cd1u, 0xdc164561u, 0x9bb63fb1u, 0xa6d61601u, 0x14f6ca11u, 0x2996e3a1u, 0x6e369971u, 0x5356b0c1u, - 0x70279f96u, 0x4d47b626u, 0x0ae7ccf6u, 0x3787e546u, 0x85a73956u, 0xb8c710e6u, 0xff676a36u, 0xc2074386u, - 0x4057d457u, 0x7d37fde7u, 0x3a978737u, 0x07f7ae87u, 0xb5d77297u, 0x88b75b27u, 0xcf1721f7u, 0xf2770847u, - 0x10c70814u, 0x2da721a4u, 0x6a075b74u, 0x576772c4u, 0xe547aed4u, 0xd8278764u, 0x9f87fdb4u, 0xa2e7d404u, - 0x20b743d5u, 0x1dd76a65u, 0x5a7710b5u, 0x67173905u, 0xd537e515u, 0xe857cca5u, 0xaff7b675u, 0x92979fc5u, - 0xe915e8dbu, 0xd475c16bu, 0x93d5bbbbu, 0xaeb5920bu, 0x1c954e1bu, 0x21f567abu, 0x66551d7bu, 0x5b3534cbu, - 0xd965a31au, 0xe4058aaau, 0xa3a5f07au, 0x9ec5d9cau, 0x2ce505dau, 0x11852c6au, 0x562556bau, 0x6b457f0au, - 0x89f57f59u, 0xb49556e9u, 0xf3352c39u, 0xce550589u, 0x7c75d999u, 0x4115f029u, 0x06b58af9u, 0x3bd5a349u, - 0xb9853498u, 0x84e51d28u, 0xc34567f8u, 0xfe254e48u, 0x4c059258u, 0x7165bbe8u, 0x36c5c138u, 0x0ba5e888u, - 0x28d4c7dfu, 0x15b4ee6fu, 0x521494bfu, 0x6f74bd0fu, 0xdd54611fu, 0xe03448afu, 0xa794327fu, 0x9af41bcfu, - 0x18a48c1eu, 0x25c4a5aeu, 0x6264df7eu, 0x5f04f6ceu, 0xed242adeu, 0xd044036eu, 0x97e479beu, 0xaa84500eu, - 0x4834505du, 0x755479edu, 0x32f4033du, 0x0f942a8du, 0xbdb4f69du, 0x80d4df2du, 0xc774a5fdu, 0xfa148c4du, - 0x78441b9cu, 0x4524322cu, 0x028448fcu, 0x3fe4614cu, 0x8dc4bd5cu, 0xb0a494ecu, 0xf704ee3cu, 0xca64c78cu -}; - -static const unsigned lodepng_crc32_table5[256] = { - 0x00000000u, 0xcb5cd3a5u, 0x4dc8a10bu, 0x869472aeu, 0x9b914216u, 0x50cd91b3u, 0xd659e31du, 0x1d0530b8u, - 0xec53826du, 0x270f51c8u, 0xa19b2366u, 0x6ac7f0c3u, 0x77c2c07bu, 0xbc9e13deu, 0x3a0a6170u, 0xf156b2d5u, - 0x03d6029bu, 0xc88ad13eu, 0x4e1ea390u, 0x85427035u, 0x9847408du, 0x531b9328u, 0xd58fe186u, 0x1ed33223u, - 0xef8580f6u, 0x24d95353u, 0xa24d21fdu, 0x6911f258u, 0x7414c2e0u, 0xbf481145u, 0x39dc63ebu, 0xf280b04eu, - 0x07ac0536u, 0xccf0d693u, 0x4a64a43du, 0x81387798u, 0x9c3d4720u, 0x57619485u, 0xd1f5e62bu, 0x1aa9358eu, - 0xebff875bu, 0x20a354feu, 0xa6372650u, 0x6d6bf5f5u, 0x706ec54du, 0xbb3216e8u, 0x3da66446u, 0xf6fab7e3u, - 0x047a07adu, 0xcf26d408u, 0x49b2a6a6u, 0x82ee7503u, 0x9feb45bbu, 0x54b7961eu, 0xd223e4b0u, 0x197f3715u, - 0xe82985c0u, 0x23755665u, 0xa5e124cbu, 0x6ebdf76eu, 0x73b8c7d6u, 0xb8e41473u, 0x3e7066ddu, 0xf52cb578u, - 0x0f580a6cu, 0xc404d9c9u, 0x4290ab67u, 0x89cc78c2u, 0x94c9487au, 0x5f959bdfu, 0xd901e971u, 0x125d3ad4u, - 0xe30b8801u, 0x28575ba4u, 0xaec3290au, 0x659ffaafu, 0x789aca17u, 0xb3c619b2u, 0x35526b1cu, 0xfe0eb8b9u, - 0x0c8e08f7u, 0xc7d2db52u, 0x4146a9fcu, 0x8a1a7a59u, 0x971f4ae1u, 0x5c439944u, 0xdad7ebeau, 0x118b384fu, - 0xe0dd8a9au, 0x2b81593fu, 0xad152b91u, 0x6649f834u, 0x7b4cc88cu, 0xb0101b29u, 0x36846987u, 0xfdd8ba22u, - 0x08f40f5au, 0xc3a8dcffu, 0x453cae51u, 0x8e607df4u, 0x93654d4cu, 0x58399ee9u, 0xdeadec47u, 0x15f13fe2u, - 0xe4a78d37u, 0x2ffb5e92u, 0xa96f2c3cu, 0x6233ff99u, 0x7f36cf21u, 0xb46a1c84u, 0x32fe6e2au, 0xf9a2bd8fu, - 0x0b220dc1u, 0xc07ede64u, 0x46eaaccau, 0x8db67f6fu, 0x90b34fd7u, 0x5bef9c72u, 0xdd7beedcu, 0x16273d79u, - 0xe7718facu, 0x2c2d5c09u, 0xaab92ea7u, 0x61e5fd02u, 0x7ce0cdbau, 0xb7bc1e1fu, 0x31286cb1u, 0xfa74bf14u, - 0x1eb014d8u, 0xd5ecc77du, 0x5378b5d3u, 0x98246676u, 0x852156ceu, 0x4e7d856bu, 0xc8e9f7c5u, 0x03b52460u, - 0xf2e396b5u, 0x39bf4510u, 0xbf2b37beu, 0x7477e41bu, 0x6972d4a3u, 0xa22e0706u, 0x24ba75a8u, 0xefe6a60du, - 0x1d661643u, 0xd63ac5e6u, 0x50aeb748u, 0x9bf264edu, 0x86f75455u, 0x4dab87f0u, 0xcb3ff55eu, 0x006326fbu, - 0xf135942eu, 0x3a69478bu, 0xbcfd3525u, 0x77a1e680u, 0x6aa4d638u, 0xa1f8059du, 0x276c7733u, 0xec30a496u, - 0x191c11eeu, 0xd240c24bu, 0x54d4b0e5u, 0x9f886340u, 0x828d53f8u, 0x49d1805du, 0xcf45f2f3u, 0x04192156u, - 0xf54f9383u, 0x3e134026u, 0xb8873288u, 0x73dbe12du, 0x6eded195u, 0xa5820230u, 0x2316709eu, 0xe84aa33bu, - 0x1aca1375u, 0xd196c0d0u, 0x5702b27eu, 0x9c5e61dbu, 0x815b5163u, 0x4a0782c6u, 0xcc93f068u, 0x07cf23cdu, - 0xf6999118u, 0x3dc542bdu, 0xbb513013u, 0x700de3b6u, 0x6d08d30eu, 0xa65400abu, 0x20c07205u, 0xeb9ca1a0u, - 0x11e81eb4u, 0xdab4cd11u, 0x5c20bfbfu, 0x977c6c1au, 0x8a795ca2u, 0x41258f07u, 0xc7b1fda9u, 0x0ced2e0cu, - 0xfdbb9cd9u, 0x36e74f7cu, 0xb0733dd2u, 0x7b2fee77u, 0x662adecfu, 0xad760d6au, 0x2be27fc4u, 0xe0beac61u, - 0x123e1c2fu, 0xd962cf8au, 0x5ff6bd24u, 0x94aa6e81u, 0x89af5e39u, 0x42f38d9cu, 0xc467ff32u, 0x0f3b2c97u, - 0xfe6d9e42u, 0x35314de7u, 0xb3a53f49u, 0x78f9ececu, 0x65fcdc54u, 0xaea00ff1u, 0x28347d5fu, 0xe368aefau, - 0x16441b82u, 0xdd18c827u, 0x5b8cba89u, 0x90d0692cu, 0x8dd55994u, 0x46898a31u, 0xc01df89fu, 0x0b412b3au, - 0xfa1799efu, 0x314b4a4au, 0xb7df38e4u, 0x7c83eb41u, 0x6186dbf9u, 0xaada085cu, 0x2c4e7af2u, 0xe712a957u, - 0x15921919u, 0xdececabcu, 0x585ab812u, 0x93066bb7u, 0x8e035b0fu, 0x455f88aau, 0xc3cbfa04u, 0x089729a1u, - 0xf9c19b74u, 0x329d48d1u, 0xb4093a7fu, 0x7f55e9dau, 0x6250d962u, 0xa90c0ac7u, 0x2f987869u, 0xe4c4abccu -}; - -static const unsigned lodepng_crc32_table6[256] = { - 0x00000000u, 0xa6770bb4u, 0x979f1129u, 0x31e81a9du, 0xf44f2413u, 0x52382fa7u, 0x63d0353au, 0xc5a73e8eu, - 0x33ef4e67u, 0x959845d3u, 0xa4705f4eu, 0x020754fau, 0xc7a06a74u, 0x61d761c0u, 0x503f7b5du, 0xf64870e9u, - 0x67de9cceu, 0xc1a9977au, 0xf0418de7u, 0x56368653u, 0x9391b8ddu, 0x35e6b369u, 0x040ea9f4u, 0xa279a240u, - 0x5431d2a9u, 0xf246d91du, 0xc3aec380u, 0x65d9c834u, 0xa07ef6bau, 0x0609fd0eu, 0x37e1e793u, 0x9196ec27u, - 0xcfbd399cu, 0x69ca3228u, 0x582228b5u, 0xfe552301u, 0x3bf21d8fu, 0x9d85163bu, 0xac6d0ca6u, 0x0a1a0712u, - 0xfc5277fbu, 0x5a257c4fu, 0x6bcd66d2u, 0xcdba6d66u, 0x081d53e8u, 0xae6a585cu, 0x9f8242c1u, 0x39f54975u, - 0xa863a552u, 0x0e14aee6u, 0x3ffcb47bu, 0x998bbfcfu, 0x5c2c8141u, 0xfa5b8af5u, 0xcbb39068u, 0x6dc49bdcu, - 0x9b8ceb35u, 0x3dfbe081u, 0x0c13fa1cu, 0xaa64f1a8u, 0x6fc3cf26u, 0xc9b4c492u, 0xf85cde0fu, 0x5e2bd5bbu, - 0x440b7579u, 0xe27c7ecdu, 0xd3946450u, 0x75e36fe4u, 0xb044516au, 0x16335adeu, 0x27db4043u, 0x81ac4bf7u, - 0x77e43b1eu, 0xd19330aau, 0xe07b2a37u, 0x460c2183u, 0x83ab1f0du, 0x25dc14b9u, 0x14340e24u, 0xb2430590u, - 0x23d5e9b7u, 0x85a2e203u, 0xb44af89eu, 0x123df32au, 0xd79acda4u, 0x71edc610u, 0x4005dc8du, 0xe672d739u, - 0x103aa7d0u, 0xb64dac64u, 0x87a5b6f9u, 0x21d2bd4du, 0xe47583c3u, 0x42028877u, 0x73ea92eau, 0xd59d995eu, - 0x8bb64ce5u, 0x2dc14751u, 0x1c295dccu, 0xba5e5678u, 0x7ff968f6u, 0xd98e6342u, 0xe86679dfu, 0x4e11726bu, - 0xb8590282u, 0x1e2e0936u, 0x2fc613abu, 0x89b1181fu, 0x4c162691u, 0xea612d25u, 0xdb8937b8u, 0x7dfe3c0cu, - 0xec68d02bu, 0x4a1fdb9fu, 0x7bf7c102u, 0xdd80cab6u, 0x1827f438u, 0xbe50ff8cu, 0x8fb8e511u, 0x29cfeea5u, - 0xdf879e4cu, 0x79f095f8u, 0x48188f65u, 0xee6f84d1u, 0x2bc8ba5fu, 0x8dbfb1ebu, 0xbc57ab76u, 0x1a20a0c2u, - 0x8816eaf2u, 0x2e61e146u, 0x1f89fbdbu, 0xb9fef06fu, 0x7c59cee1u, 0xda2ec555u, 0xebc6dfc8u, 0x4db1d47cu, - 0xbbf9a495u, 0x1d8eaf21u, 0x2c66b5bcu, 0x8a11be08u, 0x4fb68086u, 0xe9c18b32u, 0xd82991afu, 0x7e5e9a1bu, - 0xefc8763cu, 0x49bf7d88u, 0x78576715u, 0xde206ca1u, 0x1b87522fu, 0xbdf0599bu, 0x8c184306u, 0x2a6f48b2u, - 0xdc27385bu, 0x7a5033efu, 0x4bb82972u, 0xedcf22c6u, 0x28681c48u, 0x8e1f17fcu, 0xbff70d61u, 0x198006d5u, - 0x47abd36eu, 0xe1dcd8dau, 0xd034c247u, 0x7643c9f3u, 0xb3e4f77du, 0x1593fcc9u, 0x247be654u, 0x820cede0u, - 0x74449d09u, 0xd23396bdu, 0xe3db8c20u, 0x45ac8794u, 0x800bb91au, 0x267cb2aeu, 0x1794a833u, 0xb1e3a387u, - 0x20754fa0u, 0x86024414u, 0xb7ea5e89u, 0x119d553du, 0xd43a6bb3u, 0x724d6007u, 0x43a57a9au, 0xe5d2712eu, - 0x139a01c7u, 0xb5ed0a73u, 0x840510eeu, 0x22721b5au, 0xe7d525d4u, 0x41a22e60u, 0x704a34fdu, 0xd63d3f49u, - 0xcc1d9f8bu, 0x6a6a943fu, 0x5b828ea2u, 0xfdf58516u, 0x3852bb98u, 0x9e25b02cu, 0xafcdaab1u, 0x09baa105u, - 0xfff2d1ecu, 0x5985da58u, 0x686dc0c5u, 0xce1acb71u, 0x0bbdf5ffu, 0xadcafe4bu, 0x9c22e4d6u, 0x3a55ef62u, - 0xabc30345u, 0x0db408f1u, 0x3c5c126cu, 0x9a2b19d8u, 0x5f8c2756u, 0xf9fb2ce2u, 0xc813367fu, 0x6e643dcbu, - 0x982c4d22u, 0x3e5b4696u, 0x0fb35c0bu, 0xa9c457bfu, 0x6c636931u, 0xca146285u, 0xfbfc7818u, 0x5d8b73acu, - 0x03a0a617u, 0xa5d7ada3u, 0x943fb73eu, 0x3248bc8au, 0xf7ef8204u, 0x519889b0u, 0x6070932du, 0xc6079899u, - 0x304fe870u, 0x9638e3c4u, 0xa7d0f959u, 0x01a7f2edu, 0xc400cc63u, 0x6277c7d7u, 0x539fdd4au, 0xf5e8d6feu, - 0x647e3ad9u, 0xc209316du, 0xf3e12bf0u, 0x55962044u, 0x90311ecau, 0x3646157eu, 0x07ae0fe3u, 0xa1d90457u, - 0x579174beu, 0xf1e67f0au, 0xc00e6597u, 0x66796e23u, 0xa3de50adu, 0x05a95b19u, 0x34414184u, 0x92364a30u -}; - -static const unsigned lodepng_crc32_table7[256] = { - 0x00000000u, 0xccaa009eu, 0x4225077du, 0x8e8f07e3u, 0x844a0efau, 0x48e00e64u, 0xc66f0987u, 0x0ac50919u, - 0xd3e51bb5u, 0x1f4f1b2bu, 0x91c01cc8u, 0x5d6a1c56u, 0x57af154fu, 0x9b0515d1u, 0x158a1232u, 0xd92012acu, - 0x7cbb312bu, 0xb01131b5u, 0x3e9e3656u, 0xf23436c8u, 0xf8f13fd1u, 0x345b3f4fu, 0xbad438acu, 0x767e3832u, - 0xaf5e2a9eu, 0x63f42a00u, 0xed7b2de3u, 0x21d12d7du, 0x2b142464u, 0xe7be24fau, 0x69312319u, 0xa59b2387u, - 0xf9766256u, 0x35dc62c8u, 0xbb53652bu, 0x77f965b5u, 0x7d3c6cacu, 0xb1966c32u, 0x3f196bd1u, 0xf3b36b4fu, - 0x2a9379e3u, 0xe639797du, 0x68b67e9eu, 0xa41c7e00u, 0xaed97719u, 0x62737787u, 0xecfc7064u, 0x205670fau, - 0x85cd537du, 0x496753e3u, 0xc7e85400u, 0x0b42549eu, 0x01875d87u, 0xcd2d5d19u, 0x43a25afau, 0x8f085a64u, - 0x562848c8u, 0x9a824856u, 0x140d4fb5u, 0xd8a74f2bu, 0xd2624632u, 0x1ec846acu, 0x9047414fu, 0x5ced41d1u, - 0x299dc2edu, 0xe537c273u, 0x6bb8c590u, 0xa712c50eu, 0xadd7cc17u, 0x617dcc89u, 0xeff2cb6au, 0x2358cbf4u, - 0xfa78d958u, 0x36d2d9c6u, 0xb85dde25u, 0x74f7debbu, 0x7e32d7a2u, 0xb298d73cu, 0x3c17d0dfu, 0xf0bdd041u, - 0x5526f3c6u, 0x998cf358u, 0x1703f4bbu, 0xdba9f425u, 0xd16cfd3cu, 0x1dc6fda2u, 0x9349fa41u, 0x5fe3fadfu, - 0x86c3e873u, 0x4a69e8edu, 0xc4e6ef0eu, 0x084cef90u, 0x0289e689u, 0xce23e617u, 0x40ace1f4u, 0x8c06e16au, - 0xd0eba0bbu, 0x1c41a025u, 0x92cea7c6u, 0x5e64a758u, 0x54a1ae41u, 0x980baedfu, 0x1684a93cu, 0xda2ea9a2u, - 0x030ebb0eu, 0xcfa4bb90u, 0x412bbc73u, 0x8d81bcedu, 0x8744b5f4u, 0x4beeb56au, 0xc561b289u, 0x09cbb217u, - 0xac509190u, 0x60fa910eu, 0xee7596edu, 0x22df9673u, 0x281a9f6au, 0xe4b09ff4u, 0x6a3f9817u, 0xa6959889u, - 0x7fb58a25u, 0xb31f8abbu, 0x3d908d58u, 0xf13a8dc6u, 0xfbff84dfu, 0x37558441u, 0xb9da83a2u, 0x7570833cu, - 0x533b85dau, 0x9f918544u, 0x111e82a7u, 0xddb48239u, 0xd7718b20u, 0x1bdb8bbeu, 0x95548c5du, 0x59fe8cc3u, - 0x80de9e6fu, 0x4c749ef1u, 0xc2fb9912u, 0x0e51998cu, 0x04949095u, 0xc83e900bu, 0x46b197e8u, 0x8a1b9776u, - 0x2f80b4f1u, 0xe32ab46fu, 0x6da5b38cu, 0xa10fb312u, 0xabcaba0bu, 0x6760ba95u, 0xe9efbd76u, 0x2545bde8u, - 0xfc65af44u, 0x30cfafdau, 0xbe40a839u, 0x72eaa8a7u, 0x782fa1beu, 0xb485a120u, 0x3a0aa6c3u, 0xf6a0a65du, - 0xaa4de78cu, 0x66e7e712u, 0xe868e0f1u, 0x24c2e06fu, 0x2e07e976u, 0xe2ade9e8u, 0x6c22ee0bu, 0xa088ee95u, - 0x79a8fc39u, 0xb502fca7u, 0x3b8dfb44u, 0xf727fbdau, 0xfde2f2c3u, 0x3148f25du, 0xbfc7f5beu, 0x736df520u, - 0xd6f6d6a7u, 0x1a5cd639u, 0x94d3d1dau, 0x5879d144u, 0x52bcd85du, 0x9e16d8c3u, 0x1099df20u, 0xdc33dfbeu, - 0x0513cd12u, 0xc9b9cd8cu, 0x4736ca6fu, 0x8b9ccaf1u, 0x8159c3e8u, 0x4df3c376u, 0xc37cc495u, 0x0fd6c40bu, - 0x7aa64737u, 0xb60c47a9u, 0x3883404au, 0xf42940d4u, 0xfeec49cdu, 0x32464953u, 0xbcc94eb0u, 0x70634e2eu, - 0xa9435c82u, 0x65e95c1cu, 0xeb665bffu, 0x27cc5b61u, 0x2d095278u, 0xe1a352e6u, 0x6f2c5505u, 0xa386559bu, - 0x061d761cu, 0xcab77682u, 0x44387161u, 0x889271ffu, 0x825778e6u, 0x4efd7878u, 0xc0727f9bu, 0x0cd87f05u, - 0xd5f86da9u, 0x19526d37u, 0x97dd6ad4u, 0x5b776a4au, 0x51b26353u, 0x9d1863cdu, 0x1397642eu, 0xdf3d64b0u, - 0x83d02561u, 0x4f7a25ffu, 0xc1f5221cu, 0x0d5f2282u, 0x079a2b9bu, 0xcb302b05u, 0x45bf2ce6u, 0x89152c78u, - 0x50353ed4u, 0x9c9f3e4au, 0x121039a9u, 0xdeba3937u, 0xd47f302eu, 0x18d530b0u, 0x965a3753u, 0x5af037cdu, - 0xff6b144au, 0x33c114d4u, 0xbd4e1337u, 0x71e413a9u, 0x7b211ab0u, 0xb78b1a2eu, 0x39041dcdu, 0xf5ae1d53u, - 0x2c8e0fffu, 0xe0240f61u, 0x6eab0882u, 0xa201081cu, 0xa8c40105u, 0x646e019bu, 0xeae10678u, 0x264b06e6u -}; - -/* Computes the cyclic redundancy check as used by PNG chunks*/ -unsigned lodepng_crc32(const unsigned char* data, size_t length) { - /*Using the Slicing by Eight algorithm*/ - unsigned r = 0xffffffffu; - while(length >= 8) { - r = lodepng_crc32_table7[(data[0] ^ (r & 0xffu))] ^ - lodepng_crc32_table6[(data[1] ^ ((r >> 8) & 0xffu))] ^ - lodepng_crc32_table5[(data[2] ^ ((r >> 16) & 0xffu))] ^ - lodepng_crc32_table4[(data[3] ^ ((r >> 24) & 0xffu))] ^ - lodepng_crc32_table3[data[4]] ^ - lodepng_crc32_table2[data[5]] ^ - lodepng_crc32_table1[data[6]] ^ - lodepng_crc32_table0[data[7]]; - data += 8; - length -= 8; - } - while(length--) { - r = lodepng_crc32_table0[(r ^ *data++) & 0xffu] ^ (r >> 8); - } - return r ^ 0xffffffffu; -} -#else /* LODEPNG_COMPILE_CRC */ -/*in this case, the function is only declared here, and must be defined externally -so that it will be linked in. - -Example implementation that uses a much smaller lookup table for memory constrained cases: - -unsigned lodepng_crc32(const unsigned char* data, size_t length) { - unsigned r = 0xffffffffu; - static const unsigned table[16] = { - 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac, 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c, - 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c, 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c - }; - while(length--) { - r = table[(r ^ *data) & 0xf] ^ (r >> 4); - r = table[(r ^ (*data >> 4)) & 0xf] ^ (r >> 4); - data++; - } - return r ^ 0xffffffffu; -} -*/ -unsigned lodepng_crc32(const unsigned char* data, size_t length); -#endif /* LODEPNG_COMPILE_CRC */ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Reading and writing PNG color channel bits / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -/* The color channel bits of less-than-8-bit pixels are read with the MSB of bytes first, -so LodePNGBitWriter and LodePNGBitReader can't be used for those. */ - -static unsigned char readBitFromReversedStream(size_t* bitpointer, const unsigned char* bitstream) { - unsigned char result = (unsigned char)((bitstream[(*bitpointer) >> 3] >> (7 - ((*bitpointer) & 0x7))) & 1); - ++(*bitpointer); - return result; -} - -/* TODO: make this faster */ -static unsigned readBitsFromReversedStream(size_t* bitpointer, const unsigned char* bitstream, size_t nbits) { - unsigned result = 0; - size_t i; - for(i = 0 ; i < nbits; ++i) { - result <<= 1u; - result |= (unsigned)readBitFromReversedStream(bitpointer, bitstream); - } - return result; -} - -static void setBitOfReversedStream(size_t* bitpointer, unsigned char* bitstream, unsigned char bit) { - /*the current bit in bitstream may be 0 or 1 for this to work*/ - if(bit == 0) bitstream[(*bitpointer) >> 3u] &= (unsigned char)(~(1u << (7u - ((*bitpointer) & 7u)))); - else bitstream[(*bitpointer) >> 3u] |= (1u << (7u - ((*bitpointer) & 7u))); - ++(*bitpointer); -} - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / PNG chunks / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -unsigned lodepng_chunk_length(const unsigned char* chunk) { - return lodepng_read32bitInt(chunk); -} - -void lodepng_chunk_type(char type[5], const unsigned char* chunk) { - unsigned i; - for(i = 0; i != 4; ++i) type[i] = (char)chunk[4 + i]; - type[4] = 0; /*null termination char*/ -} - -unsigned char lodepng_chunk_type_equals(const unsigned char* chunk, const char* type) { - if(lodepng_strlen(type) != 4) return 0; - return (chunk[4] == type[0] && chunk[5] == type[1] && chunk[6] == type[2] && chunk[7] == type[3]); -} - -unsigned char lodepng_chunk_ancillary(const unsigned char* chunk) { - return((chunk[4] & 32) != 0); -} - -unsigned char lodepng_chunk_private(const unsigned char* chunk) { - return((chunk[6] & 32) != 0); -} - -unsigned char lodepng_chunk_safetocopy(const unsigned char* chunk) { - return((chunk[7] & 32) != 0); -} - -unsigned char* lodepng_chunk_data(unsigned char* chunk) { - return &chunk[8]; -} - -const unsigned char* lodepng_chunk_data_const(const unsigned char* chunk) { - return &chunk[8]; -} - -unsigned lodepng_chunk_check_crc(const unsigned char* chunk) { - unsigned length = lodepng_chunk_length(chunk); - unsigned CRC = lodepng_read32bitInt(&chunk[length + 8]); - /*the CRC is taken of the data and the 4 chunk type letters, not the length*/ - unsigned checksum = lodepng_crc32(&chunk[4], length + 4); - if(CRC != checksum) return 1; - else return 0; -} - -void lodepng_chunk_generate_crc(unsigned char* chunk) { - unsigned length = lodepng_chunk_length(chunk); - unsigned CRC = lodepng_crc32(&chunk[4], length + 4); - lodepng_set32bitInt(chunk + 8 + length, CRC); -} - -unsigned char* lodepng_chunk_next(unsigned char* chunk, unsigned char* end) { - size_t available_size = (size_t)(end - chunk); - if(chunk >= end || available_size < 12) return end; /*too small to contain a chunk*/ - if(chunk[0] == 0x89 && chunk[1] == 0x50 && chunk[2] == 0x4e && chunk[3] == 0x47 - && chunk[4] == 0x0d && chunk[5] == 0x0a && chunk[6] == 0x1a && chunk[7] == 0x0a) { - /* Is PNG magic header at start of PNG file. Jump to first actual chunk. */ - return chunk + 8; - } else { - size_t total_chunk_length; - if(lodepng_addofl(lodepng_chunk_length(chunk), 12, &total_chunk_length)) return end; - if(total_chunk_length > available_size) return end; /*outside of range*/ - return chunk + total_chunk_length; - } -} - -const unsigned char* lodepng_chunk_next_const(const unsigned char* chunk, const unsigned char* end) { - size_t available_size = (size_t)(end - chunk); - if(chunk >= end || available_size < 12) return end; /*too small to contain a chunk*/ - if(chunk[0] == 0x89 && chunk[1] == 0x50 && chunk[2] == 0x4e && chunk[3] == 0x47 - && chunk[4] == 0x0d && chunk[5] == 0x0a && chunk[6] == 0x1a && chunk[7] == 0x0a) { - /* Is PNG magic header at start of PNG file. Jump to first actual chunk. */ - return chunk + 8; - } else { - size_t total_chunk_length; - if(lodepng_addofl(lodepng_chunk_length(chunk), 12, &total_chunk_length)) return end; - if(total_chunk_length > available_size) return end; /*outside of range*/ - return chunk + total_chunk_length; - } -} - -unsigned char* lodepng_chunk_find(unsigned char* chunk, unsigned char* end, const char type[5]) { - for(;;) { - if(chunk >= end || end - chunk < 12) return 0; /* past file end: chunk + 12 > end */ - if(lodepng_chunk_type_equals(chunk, type)) return chunk; - chunk = lodepng_chunk_next(chunk, end); - } -} - -const unsigned char* lodepng_chunk_find_const(const unsigned char* chunk, const unsigned char* end, const char type[5]) { - for(;;) { - if(chunk >= end || end - chunk < 12) return 0; /* past file end: chunk + 12 > end */ - if(lodepng_chunk_type_equals(chunk, type)) return chunk; - chunk = lodepng_chunk_next_const(chunk, end); - } -} - -unsigned lodepng_chunk_append(unsigned char** out, size_t* outsize, const unsigned char* chunk) { - unsigned i; - size_t total_chunk_length, new_length; - unsigned char *chunk_start, *new_buffer; - - if(lodepng_addofl(lodepng_chunk_length(chunk), 12, &total_chunk_length)) return 77; - if(lodepng_addofl(*outsize, total_chunk_length, &new_length)) return 77; - - new_buffer = (unsigned char*)lodepng_realloc(*out, new_length); - if(!new_buffer) return 83; /*alloc fail*/ - (*out) = new_buffer; - (*outsize) = new_length; - chunk_start = &(*out)[new_length - total_chunk_length]; - - for(i = 0; i != total_chunk_length; ++i) chunk_start[i] = chunk[i]; - - return 0; -} - -/*Sets length and name and allocates the space for data and crc but does not -set data or crc yet. Returns the start of the chunk in chunk. The start of -the data is at chunk + 8. To finalize chunk, add the data, then use -lodepng_chunk_generate_crc */ -static unsigned lodepng_chunk_init(unsigned char** chunk, - ucvector* out, - unsigned length, const char* type) { - size_t new_length = out->size; - if(lodepng_addofl(new_length, length, &new_length)) return 77; - if(lodepng_addofl(new_length, 12, &new_length)) return 77; - if(!ucvector_resize(out, new_length)) return 83; /*alloc fail*/ - *chunk = out->data + new_length - length - 12u; - - /*1: length*/ - lodepng_set32bitInt(*chunk, length); - - /*2: chunk name (4 letters)*/ - lodepng_memcpy(*chunk + 4, type, 4); - - return 0; -} - -/* like lodepng_chunk_create but with custom allocsize */ -static unsigned lodepng_chunk_createv(ucvector* out, - unsigned length, const char* type, const unsigned char* data) { - unsigned char* chunk; - CERROR_TRY_RETURN(lodepng_chunk_init(&chunk, out, length, type)); - - /*3: the data*/ - lodepng_memcpy(chunk + 8, data, length); - - /*4: CRC (of the chunkname characters and the data)*/ - lodepng_chunk_generate_crc(chunk); - - return 0; -} - -unsigned lodepng_chunk_create(unsigned char** out, size_t* outsize, - unsigned length, const char* type, const unsigned char* data) { - ucvector v = ucvector_init(*out, *outsize); - unsigned error = lodepng_chunk_createv(&v, length, type, data); - *out = v.data; - *outsize = v.size; - return error; -} - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Color types, channels, bits / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -/*checks if the colortype is valid and the bitdepth bd is allowed for this colortype. -Return value is a LodePNG error code.*/ -static unsigned checkColorValidity(LodePNGColorType colortype, unsigned bd) { - switch(colortype) { - case LCT_GREY: if(!(bd == 1 || bd == 2 || bd == 4 || bd == 8 || bd == 16)) return 37; break; - case LCT_RGB: if(!( bd == 8 || bd == 16)) return 37; break; - case LCT_PALETTE: if(!(bd == 1 || bd == 2 || bd == 4 || bd == 8 )) return 37; break; - case LCT_GREY_ALPHA: if(!( bd == 8 || bd == 16)) return 37; break; - case LCT_RGBA: if(!( bd == 8 || bd == 16)) return 37; break; - case LCT_MAX_OCTET_VALUE: return 31; /* invalid color type */ - default: return 31; /* invalid color type */ - } - return 0; /*allowed color type / bits combination*/ -} - -static unsigned getNumColorChannels(LodePNGColorType colortype) { - switch(colortype) { - case LCT_GREY: return 1; - case LCT_RGB: return 3; - case LCT_PALETTE: return 1; - case LCT_GREY_ALPHA: return 2; - case LCT_RGBA: return 4; - case LCT_MAX_OCTET_VALUE: return 0; /* invalid color type */ - default: return 0; /*invalid color type*/ - } -} - -static unsigned lodepng_get_bpp_lct(LodePNGColorType colortype, unsigned bitdepth) { - /*bits per pixel is amount of channels * bits per channel*/ - return getNumColorChannels(colortype) * bitdepth; -} - -/* ////////////////////////////////////////////////////////////////////////// */ - -void lodepng_color_mode_init(LodePNGColorMode* info) { - info->key_defined = 0; - info->key_r = info->key_g = info->key_b = 0; - info->colortype = LCT_RGBA; - info->bitdepth = 8; - info->palette = 0; - info->palettesize = 0; -} - -/*allocates palette memory if needed, and initializes all colors to black*/ -static void lodepng_color_mode_alloc_palette(LodePNGColorMode* info) { - size_t i; - /*if the palette is already allocated, it will have size 1024 so no reallocation needed in that case*/ - /*the palette must have room for up to 256 colors with 4 bytes each.*/ - if(!info->palette) info->palette = (unsigned char*)lodepng_malloc(1024); - if(!info->palette) return; /*alloc fail*/ - for(i = 0; i != 256; ++i) { - /*Initialize all unused colors with black, the value used for invalid palette indices. - This is an error according to the PNG spec, but common PNG decoders make it black instead. - That makes color conversion slightly faster due to no error handling needed.*/ - info->palette[i * 4 + 0] = 0; - info->palette[i * 4 + 1] = 0; - info->palette[i * 4 + 2] = 0; - info->palette[i * 4 + 3] = 255; - } -} - -void lodepng_color_mode_cleanup(LodePNGColorMode* info) { - lodepng_palette_clear(info); -} - -unsigned lodepng_color_mode_copy(LodePNGColorMode* dest, const LodePNGColorMode* source) { - lodepng_color_mode_cleanup(dest); - lodepng_memcpy(dest, source, sizeof(LodePNGColorMode)); - if(source->palette) { - dest->palette = (unsigned char*)lodepng_malloc(1024); - if(!dest->palette && source->palettesize) return 83; /*alloc fail*/ - lodepng_memcpy(dest->palette, source->palette, source->palettesize * 4); - } - return 0; -} - -LodePNGColorMode lodepng_color_mode_make(LodePNGColorType colortype, unsigned bitdepth) { - LodePNGColorMode result; - lodepng_color_mode_init(&result); - result.colortype = colortype; - result.bitdepth = bitdepth; - return result; -} - -static int lodepng_color_mode_equal(const LodePNGColorMode* a, const LodePNGColorMode* b) { - size_t i; - if(a->colortype != b->colortype) return 0; - if(a->bitdepth != b->bitdepth) return 0; - if(a->key_defined != b->key_defined) return 0; - if(a->key_defined) { - if(a->key_r != b->key_r) return 0; - if(a->key_g != b->key_g) return 0; - if(a->key_b != b->key_b) return 0; - } - if(a->palettesize != b->palettesize) return 0; - for(i = 0; i != a->palettesize * 4; ++i) { - if(a->palette[i] != b->palette[i]) return 0; - } - return 1; -} - -void lodepng_palette_clear(LodePNGColorMode* info) { - if(info->palette) lodepng_free(info->palette); - info->palette = 0; - info->palettesize = 0; -} - -unsigned lodepng_palette_add(LodePNGColorMode* info, - unsigned char r, unsigned char g, unsigned char b, unsigned char a) { - if(!info->palette) /*allocate palette if empty*/ { - lodepng_color_mode_alloc_palette(info); - if(!info->palette) return 83; /*alloc fail*/ - } - if(info->palettesize >= 256) { - return 108; /*too many palette values*/ - } - info->palette[4 * info->palettesize + 0] = r; - info->palette[4 * info->palettesize + 1] = g; - info->palette[4 * info->palettesize + 2] = b; - info->palette[4 * info->palettesize + 3] = a; - ++info->palettesize; - return 0; -} - -/*calculate bits per pixel out of colortype and bitdepth*/ -unsigned lodepng_get_bpp(const LodePNGColorMode* info) { - return lodepng_get_bpp_lct(info->colortype, info->bitdepth); -} - -unsigned lodepng_get_channels(const LodePNGColorMode* info) { - return getNumColorChannels(info->colortype); -} - -unsigned lodepng_is_greyscale_type(const LodePNGColorMode* info) { - return info->colortype == LCT_GREY || info->colortype == LCT_GREY_ALPHA; -} - -unsigned lodepng_is_alpha_type(const LodePNGColorMode* info) { - return (info->colortype & 4) != 0; /*4 or 6*/ -} - -unsigned lodepng_is_palette_type(const LodePNGColorMode* info) { - return info->colortype == LCT_PALETTE; -} - -unsigned lodepng_has_palette_alpha(const LodePNGColorMode* info) { - size_t i; - for(i = 0; i != info->palettesize; ++i) { - if(info->palette[i * 4 + 3] < 255) return 1; - } - return 0; -} - -unsigned lodepng_can_have_alpha(const LodePNGColorMode* info) { - return info->key_defined - || lodepng_is_alpha_type(info) - || lodepng_has_palette_alpha(info); -} - -static size_t lodepng_get_raw_size_lct(unsigned w, unsigned h, LodePNGColorType colortype, unsigned bitdepth) { - size_t bpp = lodepng_get_bpp_lct(colortype, bitdepth); - size_t n = (size_t)w * (size_t)h; - return ((n / 8u) * bpp) + ((n & 7u) * bpp + 7u) / 8u; -} - -size_t lodepng_get_raw_size(unsigned w, unsigned h, const LodePNGColorMode* color) { - return lodepng_get_raw_size_lct(w, h, color->colortype, color->bitdepth); -} - - -#ifdef LODEPNG_COMPILE_PNG - -/*in an idat chunk, each scanline is a multiple of 8 bits, unlike the lodepng output buffer, -and in addition has one extra byte per line: the filter byte. So this gives a larger -result than lodepng_get_raw_size. Set h to 1 to get the size of 1 row including filter byte. */ -static size_t lodepng_get_raw_size_idat(unsigned w, unsigned h, unsigned bpp) { - /* + 1 for the filter byte, and possibly plus padding bits per line. */ - /* Ignoring casts, the expression is equal to (w * bpp + 7) / 8 + 1, but avoids overflow of w * bpp */ - size_t line = ((size_t)(w / 8u) * bpp) + 1u + ((w & 7u) * bpp + 7u) / 8u; - return (size_t)h * line; -} - -#ifdef LODEPNG_COMPILE_DECODER -/*Safely checks whether size_t overflow can be caused due to amount of pixels. -This check is overcautious rather than precise. If this check indicates no overflow, -you can safely compute in a size_t (but not an unsigned): --(size_t)w * (size_t)h * 8 --amount of bytes in IDAT (including filter, padding and Adam7 bytes) --amount of bytes in raw color model -Returns 1 if overflow possible, 0 if not. -*/ -static int lodepng_pixel_overflow(unsigned w, unsigned h, - const LodePNGColorMode* pngcolor, const LodePNGColorMode* rawcolor) { - size_t bpp = LODEPNG_MAX(lodepng_get_bpp(pngcolor), lodepng_get_bpp(rawcolor)); - size_t numpixels, total; - size_t line; /* bytes per line in worst case */ - - if(lodepng_mulofl((size_t)w, (size_t)h, &numpixels)) return 1; - if(lodepng_mulofl(numpixels, 8, &total)) return 1; /* bit pointer with 8-bit color, or 8 bytes per channel color */ - - /* Bytes per scanline with the expression "(w / 8u) * bpp) + ((w & 7u) * bpp + 7u) / 8u" */ - if(lodepng_mulofl((size_t)(w / 8u), bpp, &line)) return 1; - if(lodepng_addofl(line, ((w & 7u) * bpp + 7u) / 8u, &line)) return 1; - - if(lodepng_addofl(line, 5, &line)) return 1; /* 5 bytes overhead per line: 1 filterbyte, 4 for Adam7 worst case */ - if(lodepng_mulofl(line, h, &total)) return 1; /* Total bytes in worst case */ - - return 0; /* no overflow */ -} -#endif /*LODEPNG_COMPILE_DECODER*/ -#endif /*LODEPNG_COMPILE_PNG*/ - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - -static void LodePNGUnknownChunks_init(LodePNGInfo* info) { - unsigned i; - for(i = 0; i != 3; ++i) info->unknown_chunks_data[i] = 0; - for(i = 0; i != 3; ++i) info->unknown_chunks_size[i] = 0; -} - -static void LodePNGUnknownChunks_cleanup(LodePNGInfo* info) { - unsigned i; - for(i = 0; i != 3; ++i) lodepng_free(info->unknown_chunks_data[i]); -} - -static unsigned LodePNGUnknownChunks_copy(LodePNGInfo* dest, const LodePNGInfo* src) { - unsigned i; - - LodePNGUnknownChunks_cleanup(dest); - - for(i = 0; i != 3; ++i) { - size_t j; - dest->unknown_chunks_size[i] = src->unknown_chunks_size[i]; - dest->unknown_chunks_data[i] = (unsigned char*)lodepng_malloc(src->unknown_chunks_size[i]); - if(!dest->unknown_chunks_data[i] && dest->unknown_chunks_size[i]) return 83; /*alloc fail*/ - for(j = 0; j < src->unknown_chunks_size[i]; ++j) { - dest->unknown_chunks_data[i][j] = src->unknown_chunks_data[i][j]; - } - } - - return 0; -} - -/******************************************************************************/ - -static void LodePNGText_init(LodePNGInfo* info) { - info->text_num = 0; - info->text_keys = NULL; - info->text_strings = NULL; -} - -static void LodePNGText_cleanup(LodePNGInfo* info) { - size_t i; - for(i = 0; i != info->text_num; ++i) { - string_cleanup(&info->text_keys[i]); - string_cleanup(&info->text_strings[i]); - } - lodepng_free(info->text_keys); - lodepng_free(info->text_strings); -} - -static unsigned LodePNGText_copy(LodePNGInfo* dest, const LodePNGInfo* source) { - size_t i = 0; - dest->text_keys = NULL; - dest->text_strings = NULL; - dest->text_num = 0; - for(i = 0; i != source->text_num; ++i) { - CERROR_TRY_RETURN(lodepng_add_text(dest, source->text_keys[i], source->text_strings[i])); - } - return 0; -} - -static unsigned lodepng_add_text_sized(LodePNGInfo* info, const char* key, const char* str, size_t size) { - char** new_keys = (char**)(lodepng_realloc(info->text_keys, sizeof(char*) * (info->text_num + 1))); - char** new_strings = (char**)(lodepng_realloc(info->text_strings, sizeof(char*) * (info->text_num + 1))); - - if(new_keys) info->text_keys = new_keys; - if(new_strings) info->text_strings = new_strings; - - if(!new_keys || !new_strings) return 83; /*alloc fail*/ - - ++info->text_num; - info->text_keys[info->text_num - 1] = alloc_string(key); - info->text_strings[info->text_num - 1] = alloc_string_sized(str, size); - if(!info->text_keys[info->text_num - 1] || !info->text_strings[info->text_num - 1]) return 83; /*alloc fail*/ - - return 0; -} - -unsigned lodepng_add_text(LodePNGInfo* info, const char* key, const char* str) { - return lodepng_add_text_sized(info, key, str, lodepng_strlen(str)); -} - -void lodepng_clear_text(LodePNGInfo* info) { - LodePNGText_cleanup(info); -} - -/******************************************************************************/ - -static void LodePNGIText_init(LodePNGInfo* info) { - info->itext_num = 0; - info->itext_keys = NULL; - info->itext_langtags = NULL; - info->itext_transkeys = NULL; - info->itext_strings = NULL; -} - -static void LodePNGIText_cleanup(LodePNGInfo* info) { - size_t i; - for(i = 0; i != info->itext_num; ++i) { - string_cleanup(&info->itext_keys[i]); - string_cleanup(&info->itext_langtags[i]); - string_cleanup(&info->itext_transkeys[i]); - string_cleanup(&info->itext_strings[i]); - } - lodepng_free(info->itext_keys); - lodepng_free(info->itext_langtags); - lodepng_free(info->itext_transkeys); - lodepng_free(info->itext_strings); -} - -static unsigned LodePNGIText_copy(LodePNGInfo* dest, const LodePNGInfo* source) { - size_t i = 0; - dest->itext_keys = NULL; - dest->itext_langtags = NULL; - dest->itext_transkeys = NULL; - dest->itext_strings = NULL; - dest->itext_num = 0; - for(i = 0; i != source->itext_num; ++i) { - CERROR_TRY_RETURN(lodepng_add_itext(dest, source->itext_keys[i], source->itext_langtags[i], - source->itext_transkeys[i], source->itext_strings[i])); - } - return 0; -} - -void lodepng_clear_itext(LodePNGInfo* info) { - LodePNGIText_cleanup(info); -} - -static unsigned lodepng_add_itext_sized(LodePNGInfo* info, const char* key, const char* langtag, - const char* transkey, const char* str, size_t size) { - char** new_keys = (char**)(lodepng_realloc(info->itext_keys, sizeof(char*) * (info->itext_num + 1))); - char** new_langtags = (char**)(lodepng_realloc(info->itext_langtags, sizeof(char*) * (info->itext_num + 1))); - char** new_transkeys = (char**)(lodepng_realloc(info->itext_transkeys, sizeof(char*) * (info->itext_num + 1))); - char** new_strings = (char**)(lodepng_realloc(info->itext_strings, sizeof(char*) * (info->itext_num + 1))); - - if(new_keys) info->itext_keys = new_keys; - if(new_langtags) info->itext_langtags = new_langtags; - if(new_transkeys) info->itext_transkeys = new_transkeys; - if(new_strings) info->itext_strings = new_strings; - - if(!new_keys || !new_langtags || !new_transkeys || !new_strings) return 83; /*alloc fail*/ - - ++info->itext_num; - - info->itext_keys[info->itext_num - 1] = alloc_string(key); - info->itext_langtags[info->itext_num - 1] = alloc_string(langtag); - info->itext_transkeys[info->itext_num - 1] = alloc_string(transkey); - info->itext_strings[info->itext_num - 1] = alloc_string_sized(str, size); - - return 0; -} - -unsigned lodepng_add_itext(LodePNGInfo* info, const char* key, const char* langtag, - const char* transkey, const char* str) { - return lodepng_add_itext_sized(info, key, langtag, transkey, str, lodepng_strlen(str)); -} - -/* same as set but does not delete */ -static unsigned lodepng_assign_icc(LodePNGInfo* info, const char* name, const unsigned char* profile, unsigned profile_size) { - if(profile_size == 0) return 100; /*invalid ICC profile size*/ - - info->iccp_name = alloc_string(name); - info->iccp_profile = (unsigned char*)lodepng_malloc(profile_size); - - if(!info->iccp_name || !info->iccp_profile) return 83; /*alloc fail*/ - - lodepng_memcpy(info->iccp_profile, profile, profile_size); - info->iccp_profile_size = profile_size; - - return 0; /*ok*/ -} - -unsigned lodepng_set_icc(LodePNGInfo* info, const char* name, const unsigned char* profile, unsigned profile_size) { - if(info->iccp_name) lodepng_clear_icc(info); - info->iccp_defined = 1; - - return lodepng_assign_icc(info, name, profile, profile_size); -} - -void lodepng_clear_icc(LodePNGInfo* info) { - string_cleanup(&info->iccp_name); - lodepng_free(info->iccp_profile); - info->iccp_profile = NULL; - info->iccp_profile_size = 0; - info->iccp_defined = 0; -} -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -void lodepng_info_init(LodePNGInfo* info) { - lodepng_color_mode_init(&info->color); - info->interlace_method = 0; - info->compression_method = 0; - info->filter_method = 0; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - info->background_defined = 0; - info->background_r = info->background_g = info->background_b = 0; - - LodePNGText_init(info); - LodePNGIText_init(info); - - info->time_defined = 0; - info->phys_defined = 0; - - info->gama_defined = 0; - info->chrm_defined = 0; - info->srgb_defined = 0; - info->iccp_defined = 0; - info->iccp_name = NULL; - info->iccp_profile = NULL; - - info->sbit_defined = 0; - info->sbit_r = info->sbit_g = info->sbit_b = info->sbit_a = 0; - - LodePNGUnknownChunks_init(info); -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -} - -void lodepng_info_cleanup(LodePNGInfo* info) { - lodepng_color_mode_cleanup(&info->color); -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - LodePNGText_cleanup(info); - LodePNGIText_cleanup(info); - - lodepng_clear_icc(info); - - LodePNGUnknownChunks_cleanup(info); -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -} - -unsigned lodepng_info_copy(LodePNGInfo* dest, const LodePNGInfo* source) { - lodepng_info_cleanup(dest); - lodepng_memcpy(dest, source, sizeof(LodePNGInfo)); - lodepng_color_mode_init(&dest->color); - CERROR_TRY_RETURN(lodepng_color_mode_copy(&dest->color, &source->color)); - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - CERROR_TRY_RETURN(LodePNGText_copy(dest, source)); - CERROR_TRY_RETURN(LodePNGIText_copy(dest, source)); - if(source->iccp_defined) { - CERROR_TRY_RETURN(lodepng_assign_icc(dest, source->iccp_name, source->iccp_profile, source->iccp_profile_size)); - } - - LodePNGUnknownChunks_init(dest); - CERROR_TRY_RETURN(LodePNGUnknownChunks_copy(dest, source)); -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - return 0; -} - -/* ////////////////////////////////////////////////////////////////////////// */ - -/*index: bitgroup index, bits: bitgroup size(1, 2 or 4), in: bitgroup value, out: octet array to add bits to*/ -static void addColorBits(unsigned char* out, size_t index, unsigned bits, unsigned in) { - unsigned m = bits == 1 ? 7 : bits == 2 ? 3 : 1; /*8 / bits - 1*/ - /*p = the partial index in the byte, e.g. with 4 palettebits it is 0 for first half or 1 for second half*/ - unsigned p = index & m; - in &= (1u << bits) - 1u; /*filter out any other bits of the input value*/ - in = in << (bits * (m - p)); - if(p == 0) out[index * bits / 8u] = in; - else out[index * bits / 8u] |= in; -} - -typedef struct ColorTree ColorTree; - -/* -One node of a color tree -This is the data structure used to count the number of unique colors and to get a palette -index for a color. It's like an octree, but because the alpha channel is used too, each -node has 16 instead of 8 children. -*/ -struct ColorTree { - ColorTree* children[16]; /*up to 16 pointers to ColorTree of next level*/ - int index; /*the payload. Only has a meaningful value if this is in the last level*/ -}; - -static void color_tree_init(ColorTree* tree) { - lodepng_memset(tree->children, 0, 16 * sizeof(*tree->children)); - tree->index = -1; -} - -static void color_tree_cleanup(ColorTree* tree) { - int i; - for(i = 0; i != 16; ++i) { - if(tree->children[i]) { - color_tree_cleanup(tree->children[i]); - lodepng_free(tree->children[i]); - } - } -} - -/*returns -1 if color not present, its index otherwise*/ -static int color_tree_get(ColorTree* tree, unsigned char r, unsigned char g, unsigned char b, unsigned char a) { - int bit = 0; - for(bit = 0; bit < 8; ++bit) { - int i = 8 * ((r >> bit) & 1) + 4 * ((g >> bit) & 1) + 2 * ((b >> bit) & 1) + 1 * ((a >> bit) & 1); - if(!tree->children[i]) return -1; - else tree = tree->children[i]; - } - return tree ? tree->index : -1; -} - -#ifdef LODEPNG_COMPILE_ENCODER -static int color_tree_has(ColorTree* tree, unsigned char r, unsigned char g, unsigned char b, unsigned char a) { - return color_tree_get(tree, r, g, b, a) >= 0; -} -#endif /*LODEPNG_COMPILE_ENCODER*/ - -/*color is not allowed to already exist. -Index should be >= 0 (it's signed to be compatible with using -1 for "doesn't exist") -Returns error code, or 0 if ok*/ -static unsigned color_tree_add(ColorTree* tree, - unsigned char r, unsigned char g, unsigned char b, unsigned char a, unsigned index) { - int bit; - for(bit = 0; bit < 8; ++bit) { - int i = 8 * ((r >> bit) & 1) + 4 * ((g >> bit) & 1) + 2 * ((b >> bit) & 1) + 1 * ((a >> bit) & 1); - if(!tree->children[i]) { - tree->children[i] = (ColorTree*)lodepng_malloc(sizeof(ColorTree)); - if(!tree->children[i]) return 83; /*alloc fail*/ - color_tree_init(tree->children[i]); - } - tree = tree->children[i]; - } - tree->index = (int)index; - return 0; -} - -/*put a pixel, given its RGBA color, into image of any color type*/ -static unsigned rgba8ToPixel(unsigned char* out, size_t i, - const LodePNGColorMode* mode, ColorTree* tree /*for palette*/, - unsigned char r, unsigned char g, unsigned char b, unsigned char a) { - if(mode->colortype == LCT_GREY) { - unsigned char gray = r; /*((unsigned short)r + g + b) / 3u;*/ - if(mode->bitdepth == 8) out[i] = gray; - else if(mode->bitdepth == 16) out[i * 2 + 0] = out[i * 2 + 1] = gray; - else { - /*take the most significant bits of gray*/ - gray = ((unsigned)gray >> (8u - mode->bitdepth)) & ((1u << mode->bitdepth) - 1u); - addColorBits(out, i, mode->bitdepth, gray); - } - } else if(mode->colortype == LCT_RGB) { - if(mode->bitdepth == 8) { - out[i * 3 + 0] = r; - out[i * 3 + 1] = g; - out[i * 3 + 2] = b; - } else { - out[i * 6 + 0] = out[i * 6 + 1] = r; - out[i * 6 + 2] = out[i * 6 + 3] = g; - out[i * 6 + 4] = out[i * 6 + 5] = b; - } - } else if(mode->colortype == LCT_PALETTE) { - int index = color_tree_get(tree, r, g, b, a); - if(index < 0) return 82; /*color not in palette*/ - if(mode->bitdepth == 8) out[i] = index; - else addColorBits(out, i, mode->bitdepth, (unsigned)index); - } else if(mode->colortype == LCT_GREY_ALPHA) { - unsigned char gray = r; /*((unsigned short)r + g + b) / 3u;*/ - if(mode->bitdepth == 8) { - out[i * 2 + 0] = gray; - out[i * 2 + 1] = a; - } else if(mode->bitdepth == 16) { - out[i * 4 + 0] = out[i * 4 + 1] = gray; - out[i * 4 + 2] = out[i * 4 + 3] = a; - } - } else if(mode->colortype == LCT_RGBA) { - if(mode->bitdepth == 8) { - out[i * 4 + 0] = r; - out[i * 4 + 1] = g; - out[i * 4 + 2] = b; - out[i * 4 + 3] = a; - } else { - out[i * 8 + 0] = out[i * 8 + 1] = r; - out[i * 8 + 2] = out[i * 8 + 3] = g; - out[i * 8 + 4] = out[i * 8 + 5] = b; - out[i * 8 + 6] = out[i * 8 + 7] = a; - } - } - - return 0; /*no error*/ -} - -/*put a pixel, given its RGBA16 color, into image of any color 16-bitdepth type*/ -static void rgba16ToPixel(unsigned char* out, size_t i, - const LodePNGColorMode* mode, - unsigned short r, unsigned short g, unsigned short b, unsigned short a) { - if(mode->colortype == LCT_GREY) { - unsigned short gray = r; /*((unsigned)r + g + b) / 3u;*/ - out[i * 2 + 0] = (gray >> 8) & 255; - out[i * 2 + 1] = gray & 255; - } else if(mode->colortype == LCT_RGB) { - out[i * 6 + 0] = (r >> 8) & 255; - out[i * 6 + 1] = r & 255; - out[i * 6 + 2] = (g >> 8) & 255; - out[i * 6 + 3] = g & 255; - out[i * 6 + 4] = (b >> 8) & 255; - out[i * 6 + 5] = b & 255; - } else if(mode->colortype == LCT_GREY_ALPHA) { - unsigned short gray = r; /*((unsigned)r + g + b) / 3u;*/ - out[i * 4 + 0] = (gray >> 8) & 255; - out[i * 4 + 1] = gray & 255; - out[i * 4 + 2] = (a >> 8) & 255; - out[i * 4 + 3] = a & 255; - } else if(mode->colortype == LCT_RGBA) { - out[i * 8 + 0] = (r >> 8) & 255; - out[i * 8 + 1] = r & 255; - out[i * 8 + 2] = (g >> 8) & 255; - out[i * 8 + 3] = g & 255; - out[i * 8 + 4] = (b >> 8) & 255; - out[i * 8 + 5] = b & 255; - out[i * 8 + 6] = (a >> 8) & 255; - out[i * 8 + 7] = a & 255; - } -} - -/*Get RGBA8 color of pixel with index i (y * width + x) from the raw image with given color type.*/ -static void getPixelColorRGBA8(unsigned char* r, unsigned char* g, - unsigned char* b, unsigned char* a, - const unsigned char* in, size_t i, - const LodePNGColorMode* mode) { - if(mode->colortype == LCT_GREY) { - if(mode->bitdepth == 8) { - *r = *g = *b = in[i]; - if(mode->key_defined && *r == mode->key_r) *a = 0; - else *a = 255; - } else if(mode->bitdepth == 16) { - *r = *g = *b = in[i * 2 + 0]; - if(mode->key_defined && 256U * in[i * 2 + 0] + in[i * 2 + 1] == mode->key_r) *a = 0; - else *a = 255; - } else { - unsigned highest = ((1U << mode->bitdepth) - 1U); /*highest possible value for this bit depth*/ - size_t j = i * mode->bitdepth; - unsigned value = readBitsFromReversedStream(&j, in, mode->bitdepth); - *r = *g = *b = (value * 255) / highest; - if(mode->key_defined && value == mode->key_r) *a = 0; - else *a = 255; - } - } else if(mode->colortype == LCT_RGB) { - if(mode->bitdepth == 8) { - *r = in[i * 3 + 0]; *g = in[i * 3 + 1]; *b = in[i * 3 + 2]; - if(mode->key_defined && *r == mode->key_r && *g == mode->key_g && *b == mode->key_b) *a = 0; - else *a = 255; - } else { - *r = in[i * 6 + 0]; - *g = in[i * 6 + 2]; - *b = in[i * 6 + 4]; - if(mode->key_defined && 256U * in[i * 6 + 0] + in[i * 6 + 1] == mode->key_r - && 256U * in[i * 6 + 2] + in[i * 6 + 3] == mode->key_g - && 256U * in[i * 6 + 4] + in[i * 6 + 5] == mode->key_b) *a = 0; - else *a = 255; - } - } else if(mode->colortype == LCT_PALETTE) { - unsigned index; - if(mode->bitdepth == 8) index = in[i]; - else { - size_t j = i * mode->bitdepth; - index = readBitsFromReversedStream(&j, in, mode->bitdepth); - } - /*out of bounds of palette not checked: see lodepng_color_mode_alloc_palette.*/ - *r = mode->palette[index * 4 + 0]; - *g = mode->palette[index * 4 + 1]; - *b = mode->palette[index * 4 + 2]; - *a = mode->palette[index * 4 + 3]; - } else if(mode->colortype == LCT_GREY_ALPHA) { - if(mode->bitdepth == 8) { - *r = *g = *b = in[i * 2 + 0]; - *a = in[i * 2 + 1]; - } else { - *r = *g = *b = in[i * 4 + 0]; - *a = in[i * 4 + 2]; - } - } else if(mode->colortype == LCT_RGBA) { - if(mode->bitdepth == 8) { - *r = in[i * 4 + 0]; - *g = in[i * 4 + 1]; - *b = in[i * 4 + 2]; - *a = in[i * 4 + 3]; - } else { - *r = in[i * 8 + 0]; - *g = in[i * 8 + 2]; - *b = in[i * 8 + 4]; - *a = in[i * 8 + 6]; - } - } -} - -/*Similar to getPixelColorRGBA8, but with all the for loops inside of the color -mode test cases, optimized to convert the colors much faster, when converting -to the common case of RGBA with 8 bit per channel. buffer must be RGBA with -enough memory.*/ -static void getPixelColorsRGBA8(unsigned char* LODEPNG_RESTRICT buffer, size_t numpixels, - const unsigned char* LODEPNG_RESTRICT in, - const LodePNGColorMode* mode) { - unsigned num_channels = 4; - size_t i; - if(mode->colortype == LCT_GREY) { - if(mode->bitdepth == 8) { - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - buffer[0] = buffer[1] = buffer[2] = in[i]; - buffer[3] = 255; - } - if(mode->key_defined) { - buffer -= numpixels * num_channels; - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - if(buffer[0] == mode->key_r) buffer[3] = 0; - } - } - } else if(mode->bitdepth == 16) { - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - buffer[0] = buffer[1] = buffer[2] = in[i * 2]; - buffer[3] = mode->key_defined && 256U * in[i * 2 + 0] + in[i * 2 + 1] == mode->key_r ? 0 : 255; - } - } else { - unsigned highest = ((1U << mode->bitdepth) - 1U); /*highest possible value for this bit depth*/ - size_t j = 0; - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - unsigned value = readBitsFromReversedStream(&j, in, mode->bitdepth); - buffer[0] = buffer[1] = buffer[2] = (value * 255) / highest; - buffer[3] = mode->key_defined && value == mode->key_r ? 0 : 255; - } - } - } else if(mode->colortype == LCT_RGB) { - if(mode->bitdepth == 8) { - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - lodepng_memcpy(buffer, &in[i * 3], 3); - buffer[3] = 255; - } - if(mode->key_defined) { - buffer -= numpixels * num_channels; - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - if(buffer[0] == mode->key_r && buffer[1]== mode->key_g && buffer[2] == mode->key_b) buffer[3] = 0; - } - } - } else { - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - buffer[0] = in[i * 6 + 0]; - buffer[1] = in[i * 6 + 2]; - buffer[2] = in[i * 6 + 4]; - buffer[3] = mode->key_defined - && 256U * in[i * 6 + 0] + in[i * 6 + 1] == mode->key_r - && 256U * in[i * 6 + 2] + in[i * 6 + 3] == mode->key_g - && 256U * in[i * 6 + 4] + in[i * 6 + 5] == mode->key_b ? 0 : 255; - } - } - } else if(mode->colortype == LCT_PALETTE) { - if(mode->bitdepth == 8) { - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - unsigned index = in[i]; - /*out of bounds of palette not checked: see lodepng_color_mode_alloc_palette.*/ - lodepng_memcpy(buffer, &mode->palette[index * 4], 4); - } - } else { - size_t j = 0; - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - unsigned index = readBitsFromReversedStream(&j, in, mode->bitdepth); - /*out of bounds of palette not checked: see lodepng_color_mode_alloc_palette.*/ - lodepng_memcpy(buffer, &mode->palette[index * 4], 4); - } - } - } else if(mode->colortype == LCT_GREY_ALPHA) { - if(mode->bitdepth == 8) { - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - buffer[0] = buffer[1] = buffer[2] = in[i * 2 + 0]; - buffer[3] = in[i * 2 + 1]; - } - } else { - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - buffer[0] = buffer[1] = buffer[2] = in[i * 4 + 0]; - buffer[3] = in[i * 4 + 2]; - } - } - } else if(mode->colortype == LCT_RGBA) { - if(mode->bitdepth == 8) { - lodepng_memcpy(buffer, in, numpixels * 4); - } else { - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - buffer[0] = in[i * 8 + 0]; - buffer[1] = in[i * 8 + 2]; - buffer[2] = in[i * 8 + 4]; - buffer[3] = in[i * 8 + 6]; - } - } - } -} - -/*Similar to getPixelColorsRGBA8, but with 3-channel RGB output.*/ -static void getPixelColorsRGB8(unsigned char* LODEPNG_RESTRICT buffer, size_t numpixels, - const unsigned char* LODEPNG_RESTRICT in, - const LodePNGColorMode* mode) { - const unsigned num_channels = 3; - size_t i; - if(mode->colortype == LCT_GREY) { - if(mode->bitdepth == 8) { - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - buffer[0] = buffer[1] = buffer[2] = in[i]; - } - } else if(mode->bitdepth == 16) { - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - buffer[0] = buffer[1] = buffer[2] = in[i * 2]; - } - } else { - unsigned highest = ((1U << mode->bitdepth) - 1U); /*highest possible value for this bit depth*/ - size_t j = 0; - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - unsigned value = readBitsFromReversedStream(&j, in, mode->bitdepth); - buffer[0] = buffer[1] = buffer[2] = (value * 255) / highest; - } - } - } else if(mode->colortype == LCT_RGB) { - if(mode->bitdepth == 8) { - lodepng_memcpy(buffer, in, numpixels * 3); - } else { - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - buffer[0] = in[i * 6 + 0]; - buffer[1] = in[i * 6 + 2]; - buffer[2] = in[i * 6 + 4]; - } - } - } else if(mode->colortype == LCT_PALETTE) { - if(mode->bitdepth == 8) { - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - unsigned index = in[i]; - /*out of bounds of palette not checked: see lodepng_color_mode_alloc_palette.*/ - lodepng_memcpy(buffer, &mode->palette[index * 4], 3); - } - } else { - size_t j = 0; - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - unsigned index = readBitsFromReversedStream(&j, in, mode->bitdepth); - /*out of bounds of palette not checked: see lodepng_color_mode_alloc_palette.*/ - lodepng_memcpy(buffer, &mode->palette[index * 4], 3); - } - } - } else if(mode->colortype == LCT_GREY_ALPHA) { - if(mode->bitdepth == 8) { - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - buffer[0] = buffer[1] = buffer[2] = in[i * 2 + 0]; - } - } else { - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - buffer[0] = buffer[1] = buffer[2] = in[i * 4 + 0]; - } - } - } else if(mode->colortype == LCT_RGBA) { - if(mode->bitdepth == 8) { - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - lodepng_memcpy(buffer, &in[i * 4], 3); - } - } else { - for(i = 0; i != numpixels; ++i, buffer += num_channels) { - buffer[0] = in[i * 8 + 0]; - buffer[1] = in[i * 8 + 2]; - buffer[2] = in[i * 8 + 4]; - } - } - } -} - -/*Get RGBA16 color of pixel with index i (y * width + x) from the raw image with -given color type, but the given color type must be 16-bit itself.*/ -static void getPixelColorRGBA16(unsigned short* r, unsigned short* g, unsigned short* b, unsigned short* a, - const unsigned char* in, size_t i, const LodePNGColorMode* mode) { - if(mode->colortype == LCT_GREY) { - *r = *g = *b = 256 * in[i * 2 + 0] + in[i * 2 + 1]; - if(mode->key_defined && 256U * in[i * 2 + 0] + in[i * 2 + 1] == mode->key_r) *a = 0; - else *a = 65535; - } else if(mode->colortype == LCT_RGB) { - *r = 256u * in[i * 6 + 0] + in[i * 6 + 1]; - *g = 256u * in[i * 6 + 2] + in[i * 6 + 3]; - *b = 256u * in[i * 6 + 4] + in[i * 6 + 5]; - if(mode->key_defined - && 256u * in[i * 6 + 0] + in[i * 6 + 1] == mode->key_r - && 256u * in[i * 6 + 2] + in[i * 6 + 3] == mode->key_g - && 256u * in[i * 6 + 4] + in[i * 6 + 5] == mode->key_b) *a = 0; - else *a = 65535; - } else if(mode->colortype == LCT_GREY_ALPHA) { - *r = *g = *b = 256u * in[i * 4 + 0] + in[i * 4 + 1]; - *a = 256u * in[i * 4 + 2] + in[i * 4 + 3]; - } else if(mode->colortype == LCT_RGBA) { - *r = 256u * in[i * 8 + 0] + in[i * 8 + 1]; - *g = 256u * in[i * 8 + 2] + in[i * 8 + 3]; - *b = 256u * in[i * 8 + 4] + in[i * 8 + 5]; - *a = 256u * in[i * 8 + 6] + in[i * 8 + 7]; - } -} - -unsigned lodepng_convert(unsigned char* out, const unsigned char* in, - const LodePNGColorMode* mode_out, const LodePNGColorMode* mode_in, - unsigned w, unsigned h) { - size_t i; - ColorTree tree; - size_t numpixels = (size_t)w * (size_t)h; - unsigned error = 0; - - if(mode_in->colortype == LCT_PALETTE && !mode_in->palette) { - return 107; /* error: must provide palette if input mode is palette */ - } - - if(lodepng_color_mode_equal(mode_out, mode_in)) { - size_t numbytes = lodepng_get_raw_size(w, h, mode_in); - lodepng_memcpy(out, in, numbytes); - return 0; - } - - if(mode_out->colortype == LCT_PALETTE) { - size_t palettesize = mode_out->palettesize; - const unsigned char* palette = mode_out->palette; - size_t palsize = (size_t)1u << mode_out->bitdepth; - /*if the user specified output palette but did not give the values, assume - they want the values of the input color type (assuming that one is palette). - Note that we never create a new palette ourselves.*/ - if(palettesize == 0) { - palettesize = mode_in->palettesize; - palette = mode_in->palette; - /*if the input was also palette with same bitdepth, then the color types are also - equal, so copy literally. This to preserve the exact indices that were in the PNG - even in case there are duplicate colors in the palette.*/ - if(mode_in->colortype == LCT_PALETTE && mode_in->bitdepth == mode_out->bitdepth) { - size_t numbytes = lodepng_get_raw_size(w, h, mode_in); - lodepng_memcpy(out, in, numbytes); - return 0; - } - } - if(palettesize < palsize) palsize = palettesize; - color_tree_init(&tree); - for(i = 0; i != palsize; ++i) { - const unsigned char* p = &palette[i * 4]; - error = color_tree_add(&tree, p[0], p[1], p[2], p[3], (unsigned)i); - if(error) break; - } - } - - if(!error) { - if(mode_in->bitdepth == 16 && mode_out->bitdepth == 16) { - for(i = 0; i != numpixels; ++i) { - unsigned short r = 0, g = 0, b = 0, a = 0; - getPixelColorRGBA16(&r, &g, &b, &a, in, i, mode_in); - rgba16ToPixel(out, i, mode_out, r, g, b, a); - } - } else if(mode_out->bitdepth == 8 && mode_out->colortype == LCT_RGBA) { - getPixelColorsRGBA8(out, numpixels, in, mode_in); - } else if(mode_out->bitdepth == 8 && mode_out->colortype == LCT_RGB) { - getPixelColorsRGB8(out, numpixels, in, mode_in); - } else { - unsigned char r = 0, g = 0, b = 0, a = 0; - for(i = 0; i != numpixels; ++i) { - getPixelColorRGBA8(&r, &g, &b, &a, in, i, mode_in); - error = rgba8ToPixel(out, i, mode_out, &tree, r, g, b, a); - if(error) break; - } - } - } - - if(mode_out->colortype == LCT_PALETTE) { - color_tree_cleanup(&tree); - } - - return error; -} - - -/* Converts a single rgb color without alpha from one type to another, color bits truncated to -their bitdepth. In case of single channel (gray or palette), only the r channel is used. Slow -function, do not use to process all pixels of an image. Alpha channel not supported on purpose: -this is for bKGD, supporting alpha may prevent it from finding a color in the palette, from the -specification it looks like bKGD should ignore the alpha values of the palette since it can use -any palette index but doesn't have an alpha channel. Idem with ignoring color key. */ -unsigned lodepng_convert_rgb( - unsigned* r_out, unsigned* g_out, unsigned* b_out, - unsigned r_in, unsigned g_in, unsigned b_in, - const LodePNGColorMode* mode_out, const LodePNGColorMode* mode_in) { - unsigned r = 0, g = 0, b = 0; - unsigned mul = 65535 / ((1u << mode_in->bitdepth) - 1u); /*65535, 21845, 4369, 257, 1*/ - unsigned shift = 16 - mode_out->bitdepth; - - if(mode_in->colortype == LCT_GREY || mode_in->colortype == LCT_GREY_ALPHA) { - r = g = b = r_in * mul; - } else if(mode_in->colortype == LCT_RGB || mode_in->colortype == LCT_RGBA) { - r = r_in * mul; - g = g_in * mul; - b = b_in * mul; - } else if(mode_in->colortype == LCT_PALETTE) { - if(r_in >= mode_in->palettesize) return 82; - r = mode_in->palette[r_in * 4 + 0] * 257u; - g = mode_in->palette[r_in * 4 + 1] * 257u; - b = mode_in->palette[r_in * 4 + 2] * 257u; - } else { - return 31; - } - - /* now convert to output format */ - if(mode_out->colortype == LCT_GREY || mode_out->colortype == LCT_GREY_ALPHA) { - *r_out = r >> shift ; - } else if(mode_out->colortype == LCT_RGB || mode_out->colortype == LCT_RGBA) { - *r_out = r >> shift ; - *g_out = g >> shift ; - *b_out = b >> shift ; - } else if(mode_out->colortype == LCT_PALETTE) { - unsigned i; - /* a 16-bit color cannot be in the palette */ - if((r >> 8) != (r & 255) || (g >> 8) != (g & 255) || (b >> 8) != (b & 255)) return 82; - for(i = 0; i < mode_out->palettesize; i++) { - unsigned j = i * 4; - if((r >> 8) == mode_out->palette[j + 0] && (g >> 8) == mode_out->palette[j + 1] && - (b >> 8) == mode_out->palette[j + 2]) { - *r_out = i; - return 0; - } - } - return 82; - } else { - return 31; - } - - return 0; -} - -#ifdef LODEPNG_COMPILE_ENCODER - -void lodepng_color_stats_init(LodePNGColorStats* stats) { - /*stats*/ - stats->colored = 0; - stats->key = 0; - stats->key_r = stats->key_g = stats->key_b = 0; - stats->alpha = 0; - stats->numcolors = 0; - stats->bits = 1; - stats->numpixels = 0; - /*settings*/ - stats->allow_palette = 1; - stats->allow_greyscale = 1; -} - -/*function used for debug purposes with C++*/ -/*void printColorStats(LodePNGColorStats* p) { - std::cout << "colored: " << (int)p->colored << ", "; - std::cout << "key: " << (int)p->key << ", "; - std::cout << "key_r: " << (int)p->key_r << ", "; - std::cout << "key_g: " << (int)p->key_g << ", "; - std::cout << "key_b: " << (int)p->key_b << ", "; - std::cout << "alpha: " << (int)p->alpha << ", "; - std::cout << "numcolors: " << (int)p->numcolors << ", "; - std::cout << "bits: " << (int)p->bits << std::endl; -}*/ - -/*Returns how many bits needed to represent given value (max 8 bit)*/ -static unsigned getValueRequiredBits(unsigned char value) { - if(value == 0 || value == 255) return 1; - /*The scaling of 2-bit and 4-bit values uses multiples of 85 and 17*/ - if(value % 17 == 0) return value % 85 == 0 ? 2 : 4; - return 8; -} - -/*stats must already have been inited. */ -unsigned lodepng_compute_color_stats(LodePNGColorStats* stats, - const unsigned char* in, unsigned w, unsigned h, - const LodePNGColorMode* mode_in) { - size_t i; - ColorTree tree; - size_t numpixels = (size_t)w * (size_t)h; - unsigned error = 0; - - /* mark things as done already if it would be impossible to have a more expensive case */ - unsigned colored_done = lodepng_is_greyscale_type(mode_in) ? 1 : 0; - unsigned alpha_done = lodepng_can_have_alpha(mode_in) ? 0 : 1; - unsigned numcolors_done = 0; - unsigned bpp = lodepng_get_bpp(mode_in); - unsigned bits_done = (stats->bits == 1 && bpp == 1) ? 1 : 0; - unsigned sixteen = 0; /* whether the input image is 16 bit */ - unsigned maxnumcolors = 257; - if(bpp <= 8) maxnumcolors = LODEPNG_MIN(257, stats->numcolors + (1u << bpp)); - - stats->numpixels += numpixels; - - /*if palette not allowed, no need to compute numcolors*/ - if(!stats->allow_palette) numcolors_done = 1; - - color_tree_init(&tree); - - /*If the stats was already filled in from previous data, fill its palette in tree - and mark things as done already if we know they are the most expensive case already*/ - if(stats->alpha) alpha_done = 1; - if(stats->colored) colored_done = 1; - if(stats->bits == 16) numcolors_done = 1; - if(stats->bits >= bpp) bits_done = 1; - if(stats->numcolors >= maxnumcolors) numcolors_done = 1; - - if(!numcolors_done) { - for(i = 0; i < stats->numcolors; i++) { - const unsigned char* color = &stats->palette[i * 4]; - error = color_tree_add(&tree, color[0], color[1], color[2], color[3], i); - if(error) goto cleanup; - } - } - - /*Check if the 16-bit input is truly 16-bit*/ - if(mode_in->bitdepth == 16 && !sixteen) { - unsigned short r = 0, g = 0, b = 0, a = 0; - for(i = 0; i != numpixels; ++i) { - getPixelColorRGBA16(&r, &g, &b, &a, in, i, mode_in); - if((r & 255) != ((r >> 8) & 255) || (g & 255) != ((g >> 8) & 255) || - (b & 255) != ((b >> 8) & 255) || (a & 255) != ((a >> 8) & 255)) /*first and second byte differ*/ { - stats->bits = 16; - sixteen = 1; - bits_done = 1; - numcolors_done = 1; /*counting colors no longer useful, palette doesn't support 16-bit*/ - break; - } - } - } - - if(sixteen) { - unsigned short r = 0, g = 0, b = 0, a = 0; - - for(i = 0; i != numpixels; ++i) { - getPixelColorRGBA16(&r, &g, &b, &a, in, i, mode_in); - - if(!colored_done && (r != g || r != b)) { - stats->colored = 1; - colored_done = 1; - } - - if(!alpha_done) { - unsigned matchkey = (r == stats->key_r && g == stats->key_g && b == stats->key_b); - if(a != 65535 && (a != 0 || (stats->key && !matchkey))) { - stats->alpha = 1; - stats->key = 0; - alpha_done = 1; - } else if(a == 0 && !stats->alpha && !stats->key) { - stats->key = 1; - stats->key_r = r; - stats->key_g = g; - stats->key_b = b; - } else if(a == 65535 && stats->key && matchkey) { - /* Color key cannot be used if an opaque pixel also has that RGB color. */ - stats->alpha = 1; - stats->key = 0; - alpha_done = 1; - } - } - if(alpha_done && numcolors_done && colored_done && bits_done) break; - } - - if(stats->key && !stats->alpha) { - for(i = 0; i != numpixels; ++i) { - getPixelColorRGBA16(&r, &g, &b, &a, in, i, mode_in); - if(a != 0 && r == stats->key_r && g == stats->key_g && b == stats->key_b) { - /* Color key cannot be used if an opaque pixel also has that RGB color. */ - stats->alpha = 1; - stats->key = 0; - alpha_done = 1; - } - } - } - } else /* < 16-bit */ { - unsigned char r = 0, g = 0, b = 0, a = 0; - for(i = 0; i != numpixels; ++i) { - getPixelColorRGBA8(&r, &g, &b, &a, in, i, mode_in); - - if(!bits_done && stats->bits < 8) { - /*only r is checked, < 8 bits is only relevant for grayscale*/ - unsigned bits = getValueRequiredBits(r); - if(bits > stats->bits) stats->bits = bits; - } - bits_done = (stats->bits >= bpp); - - if(!colored_done && (r != g || r != b)) { - stats->colored = 1; - colored_done = 1; - if(stats->bits < 8) stats->bits = 8; /*PNG has no colored modes with less than 8-bit per channel*/ - } - - if(!alpha_done) { - unsigned matchkey = (r == stats->key_r && g == stats->key_g && b == stats->key_b); - if(a != 255 && (a != 0 || (stats->key && !matchkey))) { - stats->alpha = 1; - stats->key = 0; - alpha_done = 1; - if(stats->bits < 8) stats->bits = 8; /*PNG has no alphachannel modes with less than 8-bit per channel*/ - } else if(a == 0 && !stats->alpha && !stats->key) { - stats->key = 1; - stats->key_r = r; - stats->key_g = g; - stats->key_b = b; - } else if(a == 255 && stats->key && matchkey) { - /* Color key cannot be used if an opaque pixel also has that RGB color. */ - stats->alpha = 1; - stats->key = 0; - alpha_done = 1; - if(stats->bits < 8) stats->bits = 8; /*PNG has no alphachannel modes with less than 8-bit per channel*/ - } - } - - if(!numcolors_done) { - if(!color_tree_has(&tree, r, g, b, a)) { - error = color_tree_add(&tree, r, g, b, a, stats->numcolors); - if(error) goto cleanup; - if(stats->numcolors < 256) { - unsigned char* p = stats->palette; - unsigned n = stats->numcolors; - p[n * 4 + 0] = r; - p[n * 4 + 1] = g; - p[n * 4 + 2] = b; - p[n * 4 + 3] = a; - } - ++stats->numcolors; - numcolors_done = stats->numcolors >= maxnumcolors; - } - } - - if(alpha_done && numcolors_done && colored_done && bits_done) break; - } - - if(stats->key && !stats->alpha) { - for(i = 0; i != numpixels; ++i) { - getPixelColorRGBA8(&r, &g, &b, &a, in, i, mode_in); - if(a != 0 && r == stats->key_r && g == stats->key_g && b == stats->key_b) { - /* Color key cannot be used if an opaque pixel also has that RGB color. */ - stats->alpha = 1; - stats->key = 0; - alpha_done = 1; - if(stats->bits < 8) stats->bits = 8; /*PNG has no alphachannel modes with less than 8-bit per channel*/ - } - } - } - - /*make the stats's key always 16-bit for consistency - repeat each byte twice*/ - stats->key_r += (stats->key_r << 8); - stats->key_g += (stats->key_g << 8); - stats->key_b += (stats->key_b << 8); - } - -cleanup: - color_tree_cleanup(&tree); - return error; -} - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS -/*Adds a single color to the color stats. The stats must already have been inited. The color must be given as 16-bit -(with 2 bytes repeating for 8-bit and 65535 for opaque alpha channel). This function is expensive, do not call it for -all pixels of an image but only for a few additional values. */ -static unsigned lodepng_color_stats_add(LodePNGColorStats* stats, - unsigned r, unsigned g, unsigned b, unsigned a) { - unsigned error = 0; - unsigned char image[8]; - LodePNGColorMode mode; - lodepng_color_mode_init(&mode); - image[0] = r >> 8; image[1] = r; image[2] = g >> 8; image[3] = g; - image[4] = b >> 8; image[5] = b; image[6] = a >> 8; image[7] = a; - mode.bitdepth = 16; - mode.colortype = LCT_RGBA; - error = lodepng_compute_color_stats(stats, image, 1, 1, &mode); - lodepng_color_mode_cleanup(&mode); - return error; -} -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -/*Computes a minimal PNG color model that can contain all colors as indicated by the stats. -The stats should be computed with lodepng_compute_color_stats. -mode_in is raw color profile of the image the stats were computed on, to copy palette order from when relevant. -Minimal PNG color model means the color type and bit depth that gives smallest amount of bits in the output image, -e.g. gray if only grayscale pixels, palette if less than 256 colors, color key if only single transparent color, ... -This is used if auto_convert is enabled (it is by default). -*/ -static unsigned auto_choose_color(LodePNGColorMode* mode_out, - const LodePNGColorMode* mode_in, - const LodePNGColorStats* stats) { - unsigned error = 0; - unsigned palettebits; - size_t i, n; - size_t numpixels = stats->numpixels; - unsigned palette_ok, gray_ok; - - unsigned alpha = stats->alpha; - unsigned key = stats->key; - unsigned bits = stats->bits; - - mode_out->key_defined = 0; - - if(key && numpixels <= 16) { - alpha = 1; /*too few pixels to justify tRNS chunk overhead*/ - key = 0; - if(bits < 8) bits = 8; /*PNG has no alphachannel modes with less than 8-bit per channel*/ - } - - gray_ok = !stats->colored; - if(!stats->allow_greyscale) gray_ok = 0; - if(!gray_ok && bits < 8) bits = 8; - - n = stats->numcolors; - palettebits = n <= 2 ? 1 : (n <= 4 ? 2 : (n <= 16 ? 4 : 8)); - palette_ok = n <= 256 && bits <= 8 && n != 0; /*n==0 means likely numcolors wasn't computed*/ - if(numpixels < n * 2) palette_ok = 0; /*don't add palette overhead if image has only a few pixels*/ - if(gray_ok && !alpha && bits <= palettebits) palette_ok = 0; /*gray is less overhead*/ - if(!stats->allow_palette) palette_ok = 0; - - if(palette_ok) { - const unsigned char* p = stats->palette; - lodepng_palette_clear(mode_out); /*remove potential earlier palette*/ - for(i = 0; i != stats->numcolors; ++i) { - error = lodepng_palette_add(mode_out, p[i * 4 + 0], p[i * 4 + 1], p[i * 4 + 2], p[i * 4 + 3]); - if(error) break; - } - - mode_out->colortype = LCT_PALETTE; - mode_out->bitdepth = palettebits; - - if(mode_in->colortype == LCT_PALETTE && mode_in->palettesize >= mode_out->palettesize - && mode_in->bitdepth == mode_out->bitdepth) { - /*If input should have same palette colors, keep original to preserve its order and prevent conversion*/ - lodepng_color_mode_cleanup(mode_out); /*clears palette, keeps the above set colortype and bitdepth fields as-is*/ - lodepng_color_mode_copy(mode_out, mode_in); - } - } else /*8-bit or 16-bit per channel*/ { - mode_out->bitdepth = bits; - mode_out->colortype = alpha ? (gray_ok ? LCT_GREY_ALPHA : LCT_RGBA) - : (gray_ok ? LCT_GREY : LCT_RGB); - if(key) { - unsigned mask = (1u << mode_out->bitdepth) - 1u; /*stats always uses 16-bit, mask converts it*/ - mode_out->key_r = stats->key_r & mask; - mode_out->key_g = stats->key_g & mask; - mode_out->key_b = stats->key_b & mask; - mode_out->key_defined = 1; - } - } - - return error; -} - -#endif /* #ifdef LODEPNG_COMPILE_ENCODER */ - -/*Paeth predictor, used by PNG filter type 4*/ -static unsigned char paethPredictor(unsigned char a, unsigned char b, unsigned char c) { - /* the subtractions of unsigned char cast it to a signed type. - With gcc, short is faster than int, with clang int is as fast (as of april 2023)*/ - short pa = (b - c) < 0 ? -(b - c) : (b - c); - short pb = (a - c) < 0 ? -(a - c) : (a - c); - /* writing it out like this compiles to something faster than introducing a temp variable*/ - short pc = (a + b - c - c) < 0 ? -(a + b - c - c) : (a + b - c - c); - /* return input value associated with smallest of pa, pb, pc (with certain priority if equal) */ - if(pb < pa) { a = b; pa = pb; } - return (pc < pa) ? c : a; -} - -/*shared values used by multiple Adam7 related functions*/ - -static const unsigned ADAM7_IX[7] = { 0, 4, 0, 2, 0, 1, 0 }; /*x start values*/ -static const unsigned ADAM7_IY[7] = { 0, 0, 4, 0, 2, 0, 1 }; /*y start values*/ -static const unsigned ADAM7_DX[7] = { 8, 8, 4, 4, 2, 2, 1 }; /*x delta values*/ -static const unsigned ADAM7_DY[7] = { 8, 8, 8, 4, 4, 2, 2 }; /*y delta values*/ - -/* -Outputs various dimensions and positions in the image related to the Adam7 reduced images. -passw: output containing the width of the 7 passes -passh: output containing the height of the 7 passes -filter_passstart: output containing the index of the start and end of each - reduced image with filter bytes -padded_passstart output containing the index of the start and end of each - reduced image when without filter bytes but with padded scanlines -passstart: output containing the index of the start and end of each reduced - image without padding between scanlines, but still padding between the images -w, h: width and height of non-interlaced image -bpp: bits per pixel -"padded" is only relevant if bpp is less than 8 and a scanline or image does not - end at a full byte -*/ -static void Adam7_getpassvalues(unsigned passw[7], unsigned passh[7], size_t filter_passstart[8], - size_t padded_passstart[8], size_t passstart[8], unsigned w, unsigned h, unsigned bpp) { - /*the passstart values have 8 values: the 8th one indicates the byte after the end of the 7th (= last) pass*/ - unsigned i; - - /*calculate width and height in pixels of each pass*/ - for(i = 0; i != 7; ++i) { - passw[i] = (w + ADAM7_DX[i] - ADAM7_IX[i] - 1) / ADAM7_DX[i]; - passh[i] = (h + ADAM7_DY[i] - ADAM7_IY[i] - 1) / ADAM7_DY[i]; - if(passw[i] == 0) passh[i] = 0; - if(passh[i] == 0) passw[i] = 0; - } - - filter_passstart[0] = padded_passstart[0] = passstart[0] = 0; - for(i = 0; i != 7; ++i) { - /*if passw[i] is 0, it's 0 bytes, not 1 (no filtertype-byte)*/ - filter_passstart[i + 1] = filter_passstart[i] - + ((passw[i] && passh[i]) ? passh[i] * (1u + (passw[i] * bpp + 7u) / 8u) : 0); - /*bits padded if needed to fill full byte at end of each scanline*/ - padded_passstart[i + 1] = padded_passstart[i] + passh[i] * ((passw[i] * bpp + 7u) / 8u); - /*only padded at end of reduced image*/ - passstart[i + 1] = passstart[i] + (passh[i] * passw[i] * bpp + 7u) / 8u; - } -} - -#ifdef LODEPNG_COMPILE_DECODER - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / PNG Decoder / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -/*read the information from the header and store it in the LodePNGInfo. return value is error*/ -unsigned lodepng_inspect(unsigned* w, unsigned* h, LodePNGState* state, - const unsigned char* in, size_t insize) { - unsigned width, height; - LodePNGInfo* info = &state->info_png; - if(insize == 0 || in == 0) { - CERROR_RETURN_ERROR(state->error, 48); /*error: the given data is empty*/ - } - if(insize < 33) { - CERROR_RETURN_ERROR(state->error, 27); /*error: the data length is smaller than the length of a PNG header*/ - } - - /*when decoding a new PNG image, make sure all parameters created after previous decoding are reset*/ - /* TODO: remove this. One should use a new LodePNGState for new sessions */ - lodepng_info_cleanup(info); - lodepng_info_init(info); - - if(in[0] != 137 || in[1] != 80 || in[2] != 78 || in[3] != 71 - || in[4] != 13 || in[5] != 10 || in[6] != 26 || in[7] != 10) { - CERROR_RETURN_ERROR(state->error, 28); /*error: the first 8 bytes are not the correct PNG signature*/ - } - if(lodepng_chunk_length(in + 8) != 13) { - CERROR_RETURN_ERROR(state->error, 94); /*error: header size must be 13 bytes*/ - } - if(!lodepng_chunk_type_equals(in + 8, "IHDR")) { - CERROR_RETURN_ERROR(state->error, 29); /*error: it doesn't start with a IHDR chunk!*/ - } - - /*read the values given in the header*/ - width = lodepng_read32bitInt(&in[16]); - height = lodepng_read32bitInt(&in[20]); - /*TODO: remove the undocumented feature that allows to give null pointers to width or height*/ - if(w) *w = width; - if(h) *h = height; - info->color.bitdepth = in[24]; - info->color.colortype = (LodePNGColorType)in[25]; - info->compression_method = in[26]; - info->filter_method = in[27]; - info->interlace_method = in[28]; - - /*errors returned only after the parsing so other values are still output*/ - - /*error: invalid image size*/ - if(width == 0 || height == 0) CERROR_RETURN_ERROR(state->error, 93); - /*error: invalid colortype or bitdepth combination*/ - state->error = checkColorValidity(info->color.colortype, info->color.bitdepth); - if(state->error) return state->error; - /*error: only compression method 0 is allowed in the specification*/ - if(info->compression_method != 0) CERROR_RETURN_ERROR(state->error, 32); - /*error: only filter method 0 is allowed in the specification*/ - if(info->filter_method != 0) CERROR_RETURN_ERROR(state->error, 33); - /*error: only interlace methods 0 and 1 exist in the specification*/ - if(info->interlace_method > 1) CERROR_RETURN_ERROR(state->error, 34); - - if(!state->decoder.ignore_crc) { - unsigned CRC = lodepng_read32bitInt(&in[29]); - unsigned checksum = lodepng_crc32(&in[12], 17); - if(CRC != checksum) { - CERROR_RETURN_ERROR(state->error, 57); /*invalid CRC*/ - } - } - - return state->error; -} - -static unsigned unfilterScanline(unsigned char* recon, const unsigned char* scanline, const unsigned char* precon, - size_t bytewidth, unsigned char filterType, size_t length) { - /* - For PNG filter method 0 - unfilter a PNG image scanline by scanline. when the pixels are smaller than 1 byte, - the filter works byte per byte (bytewidth = 1) - precon is the previous unfiltered scanline, recon the result, scanline the current one - the incoming scanlines do NOT include the filtertype byte, that one is given in the parameter filterType instead - recon and scanline MAY be the same memory address! precon must be disjoint. - */ - - size_t i; - switch(filterType) { - case 0: - for(i = 0; i != length; ++i) recon[i] = scanline[i]; - break; - case 1: { - size_t j = 0; - for(i = 0; i != bytewidth; ++i) recon[i] = scanline[i]; - for(i = bytewidth; i != length; ++i, ++j) recon[i] = scanline[i] + recon[j]; - break; - } - case 2: - if(precon) { - for(i = 0; i != length; ++i) recon[i] = scanline[i] + precon[i]; - } else { - for(i = 0; i != length; ++i) recon[i] = scanline[i]; - } - break; - case 3: - if(precon) { - size_t j = 0; - for(i = 0; i != bytewidth; ++i) recon[i] = scanline[i] + (precon[i] >> 1u); - /* Unroll independent paths of this predictor. A 6x and 8x version is also possible but that adds - too much code. Whether this speeds up anything depends on compiler and settings. */ - if(bytewidth >= 4) { - for(; i + 3 < length; i += 4, j += 4) { - unsigned char s0 = scanline[i + 0], s1 = scanline[i + 1], s2 = scanline[i + 2], s3 = scanline[i + 3]; - unsigned char r0 = recon[j + 0], r1 = recon[j + 1], r2 = recon[j + 2], r3 = recon[j + 3]; - unsigned char p0 = precon[i + 0], p1 = precon[i + 1], p2 = precon[i + 2], p3 = precon[i + 3]; - recon[i + 0] = s0 + ((r0 + p0) >> 1u); - recon[i + 1] = s1 + ((r1 + p1) >> 1u); - recon[i + 2] = s2 + ((r2 + p2) >> 1u); - recon[i + 3] = s3 + ((r3 + p3) >> 1u); - } - } else if(bytewidth >= 3) { - for(; i + 2 < length; i += 3, j += 3) { - unsigned char s0 = scanline[i + 0], s1 = scanline[i + 1], s2 = scanline[i + 2]; - unsigned char r0 = recon[j + 0], r1 = recon[j + 1], r2 = recon[j + 2]; - unsigned char p0 = precon[i + 0], p1 = precon[i + 1], p2 = precon[i + 2]; - recon[i + 0] = s0 + ((r0 + p0) >> 1u); - recon[i + 1] = s1 + ((r1 + p1) >> 1u); - recon[i + 2] = s2 + ((r2 + p2) >> 1u); - } - } else if(bytewidth >= 2) { - for(; i + 1 < length; i += 2, j += 2) { - unsigned char s0 = scanline[i + 0], s1 = scanline[i + 1]; - unsigned char r0 = recon[j + 0], r1 = recon[j + 1]; - unsigned char p0 = precon[i + 0], p1 = precon[i + 1]; - recon[i + 0] = s0 + ((r0 + p0) >> 1u); - recon[i + 1] = s1 + ((r1 + p1) >> 1u); - } - } - for(; i != length; ++i, ++j) recon[i] = scanline[i] + ((recon[j] + precon[i]) >> 1u); - } else { - size_t j = 0; - for(i = 0; i != bytewidth; ++i) recon[i] = scanline[i]; - for(i = bytewidth; i != length; ++i, ++j) recon[i] = scanline[i] + (recon[j] >> 1u); - } - break; - case 4: - if(precon) { - /* Unroll independent paths of this predictor. Whether this speeds up - anything depends on compiler and settings. */ - if(bytewidth == 8) { - unsigned char a0, b0 = 0, c0, d0 = 0, a1, b1 = 0, c1, d1 = 0; - unsigned char a2, b2 = 0, c2, d2 = 0, a3, b3 = 0, c3, d3 = 0; - unsigned char a4, b4 = 0, c4, d4 = 0, a5, b5 = 0, c5, d5 = 0; - unsigned char a6, b6 = 0, c6, d6 = 0, a7, b7 = 0, c7, d7 = 0; - for(i = 0; i + 7 < length; i += 8) { - c0 = b0; c1 = b1; c2 = b2; c3 = b3; - c4 = b4; c5 = b5; c6 = b6; c7 = b7; - b0 = precon[i + 0]; b1 = precon[i + 1]; b2 = precon[i + 2]; b3 = precon[i + 3]; - b4 = precon[i + 4]; b5 = precon[i + 5]; b6 = precon[i + 6]; b7 = precon[i + 7]; - a0 = d0; a1 = d1; a2 = d2; a3 = d3; - a4 = d4; a5 = d5; a6 = d6; a7 = d7; - d0 = scanline[i + 0] + paethPredictor(a0, b0, c0); - d1 = scanline[i + 1] + paethPredictor(a1, b1, c1); - d2 = scanline[i + 2] + paethPredictor(a2, b2, c2); - d3 = scanline[i + 3] + paethPredictor(a3, b3, c3); - d4 = scanline[i + 4] + paethPredictor(a4, b4, c4); - d5 = scanline[i + 5] + paethPredictor(a5, b5, c5); - d6 = scanline[i + 6] + paethPredictor(a6, b6, c6); - d7 = scanline[i + 7] + paethPredictor(a7, b7, c7); - recon[i + 0] = d0; recon[i + 1] = d1; recon[i + 2] = d2; recon[i + 3] = d3; - recon[i + 4] = d4; recon[i + 5] = d5; recon[i + 6] = d6; recon[i + 7] = d7; - } - } else if(bytewidth == 6) { - unsigned char a0, b0 = 0, c0, d0 = 0, a1, b1 = 0, c1, d1 = 0; - unsigned char a2, b2 = 0, c2, d2 = 0, a3, b3 = 0, c3, d3 = 0; - unsigned char a4, b4 = 0, c4, d4 = 0, a5, b5 = 0, c5, d5 = 0; - for(i = 0; i + 5 < length; i += 6) { - c0 = b0; c1 = b1; c2 = b2; - c3 = b3; c4 = b4; c5 = b5; - b0 = precon[i + 0]; b1 = precon[i + 1]; b2 = precon[i + 2]; - b3 = precon[i + 3]; b4 = precon[i + 4]; b5 = precon[i + 5]; - a0 = d0; a1 = d1; a2 = d2; - a3 = d3; a4 = d4; a5 = d5; - d0 = scanline[i + 0] + paethPredictor(a0, b0, c0); - d1 = scanline[i + 1] + paethPredictor(a1, b1, c1); - d2 = scanline[i + 2] + paethPredictor(a2, b2, c2); - d3 = scanline[i + 3] + paethPredictor(a3, b3, c3); - d4 = scanline[i + 4] + paethPredictor(a4, b4, c4); - d5 = scanline[i + 5] + paethPredictor(a5, b5, c5); - recon[i + 0] = d0; recon[i + 1] = d1; recon[i + 2] = d2; - recon[i + 3] = d3; recon[i + 4] = d4; recon[i + 5] = d5; - } - } else if(bytewidth == 4) { - unsigned char a0, b0 = 0, c0, d0 = 0, a1, b1 = 0, c1, d1 = 0; - unsigned char a2, b2 = 0, c2, d2 = 0, a3, b3 = 0, c3, d3 = 0; - for(i = 0; i + 3 < length; i += 4) { - c0 = b0; c1 = b1; c2 = b2; c3 = b3; - b0 = precon[i + 0]; b1 = precon[i + 1]; b2 = precon[i + 2]; b3 = precon[i + 3]; - a0 = d0; a1 = d1; a2 = d2; a3 = d3; - d0 = scanline[i + 0] + paethPredictor(a0, b0, c0); - d1 = scanline[i + 1] + paethPredictor(a1, b1, c1); - d2 = scanline[i + 2] + paethPredictor(a2, b2, c2); - d3 = scanline[i + 3] + paethPredictor(a3, b3, c3); - recon[i + 0] = d0; recon[i + 1] = d1; recon[i + 2] = d2; recon[i + 3] = d3; - } - } else if(bytewidth == 3) { - unsigned char a0, b0 = 0, c0, d0 = 0; - unsigned char a1, b1 = 0, c1, d1 = 0; - unsigned char a2, b2 = 0, c2, d2 = 0; - for(i = 0; i + 2 < length; i += 3) { - c0 = b0; c1 = b1; c2 = b2; - b0 = precon[i + 0]; b1 = precon[i + 1]; b2 = precon[i + 2]; - a0 = d0; a1 = d1; a2 = d2; - d0 = scanline[i + 0] + paethPredictor(a0, b0, c0); - d1 = scanline[i + 1] + paethPredictor(a1, b1, c1); - d2 = scanline[i + 2] + paethPredictor(a2, b2, c2); - recon[i + 0] = d0; recon[i + 1] = d1; recon[i + 2] = d2; - } - } else if(bytewidth == 2) { - unsigned char a0, b0 = 0, c0, d0 = 0; - unsigned char a1, b1 = 0, c1, d1 = 0; - for(i = 0; i + 1 < length; i += 2) { - c0 = b0; c1 = b1; - b0 = precon[i + 0]; - b1 = precon[i + 1]; - a0 = d0; a1 = d1; - d0 = scanline[i + 0] + paethPredictor(a0, b0, c0); - d1 = scanline[i + 1] + paethPredictor(a1, b1, c1); - recon[i + 0] = d0; - recon[i + 1] = d1; - } - } else if(bytewidth == 1) { - unsigned char a, b = 0, c, d = 0; - for(i = 0; i != length; ++i) { - c = b; - b = precon[i]; - a = d; - d = scanline[i] + paethPredictor(a, b, c); - recon[i] = d; - } - } else { - /* Normally not a possible case, but this would handle it correctly */ - for(i = 0; i != bytewidth; ++i) { - recon[i] = (scanline[i] + precon[i]); /*paethPredictor(0, precon[i], 0) is always precon[i]*/ - } - } - /* finish any remaining bytes */ - for(; i != length; ++i) { - recon[i] = (scanline[i] + paethPredictor(recon[i - bytewidth], precon[i], precon[i - bytewidth])); - } - } else { - size_t j = 0; - for(i = 0; i != bytewidth; ++i) { - recon[i] = scanline[i]; - } - for(i = bytewidth; i != length; ++i, ++j) { - /*paethPredictor(recon[i - bytewidth], 0, 0) is always recon[i - bytewidth]*/ - recon[i] = (scanline[i] + recon[j]); - } - } - break; - default: return 36; /*error: invalid filter type given*/ - } - return 0; -} - -static unsigned unfilter(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, unsigned bpp) { - /* - For PNG filter method 0 - this function unfilters a single image (e.g. without interlacing this is called once, with Adam7 seven times) - out must have enough bytes allocated already, in must have the scanlines + 1 filtertype byte per scanline - w and h are image dimensions or dimensions of reduced image, bpp is bits per pixel - in and out are allowed to be the same memory address (but aren't the same size since in has the extra filter bytes) - */ - - unsigned y; - unsigned char* prevline = 0; - - /*bytewidth is used for filtering, is 1 when bpp < 8, number of bytes per pixel otherwise*/ - size_t bytewidth = (bpp + 7u) / 8u; - /*the width of a scanline in bytes, not including the filter type*/ - size_t linebytes = lodepng_get_raw_size_idat(w, 1, bpp) - 1u; - - for(y = 0; y < h; ++y) { - size_t outindex = linebytes * y; - size_t inindex = (1 + linebytes) * y; /*the extra filterbyte added to each row*/ - unsigned char filterType = in[inindex]; - - CERROR_TRY_RETURN(unfilterScanline(&out[outindex], &in[inindex + 1], prevline, bytewidth, filterType, linebytes)); - - prevline = &out[outindex]; - } - - return 0; -} - -/* -in: Adam7 interlaced image, with no padding bits between scanlines, but between - reduced images so that each reduced image starts at a byte. -out: the same pixels, but re-ordered so that they're now a non-interlaced image with size w*h -bpp: bits per pixel -out has the following size in bits: w * h * bpp. -in is possibly bigger due to padding bits between reduced images. -out must be big enough AND must be 0 everywhere if bpp < 8 in the current implementation -(because that's likely a little bit faster) -NOTE: comments about padding bits are only relevant if bpp < 8 -*/ -static void Adam7_deinterlace(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, unsigned bpp) { - unsigned passw[7], passh[7]; - size_t filter_passstart[8], padded_passstart[8], passstart[8]; - unsigned i; - - Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); - - if(bpp >= 8) { - for(i = 0; i != 7; ++i) { - unsigned x, y, b; - size_t bytewidth = bpp / 8u; - for(y = 0; y < passh[i]; ++y) - for(x = 0; x < passw[i]; ++x) { - size_t pixelinstart = passstart[i] + (y * passw[i] + x) * bytewidth; - size_t pixeloutstart = ((ADAM7_IY[i] + (size_t)y * ADAM7_DY[i]) * (size_t)w - + ADAM7_IX[i] + (size_t)x * ADAM7_DX[i]) * bytewidth; - for(b = 0; b < bytewidth; ++b) { - out[pixeloutstart + b] = in[pixelinstart + b]; - } - } - } - } else /*bpp < 8: Adam7 with pixels < 8 bit is a bit trickier: with bit pointers*/ { - for(i = 0; i != 7; ++i) { - unsigned x, y, b; - unsigned ilinebits = bpp * passw[i]; - unsigned olinebits = bpp * w; - size_t obp, ibp; /*bit pointers (for out and in buffer)*/ - for(y = 0; y < passh[i]; ++y) - for(x = 0; x < passw[i]; ++x) { - ibp = (8 * passstart[i]) + (y * ilinebits + x * bpp); - obp = (ADAM7_IY[i] + (size_t)y * ADAM7_DY[i]) * olinebits + (ADAM7_IX[i] + (size_t)x * ADAM7_DX[i]) * bpp; - for(b = 0; b < bpp; ++b) { - unsigned char bit = readBitFromReversedStream(&ibp, in); - setBitOfReversedStream(&obp, out, bit); - } - } - } - } -} - -static void removePaddingBits(unsigned char* out, const unsigned char* in, - size_t olinebits, size_t ilinebits, unsigned h) { - /* - After filtering there are still padding bits if scanlines have non multiple of 8 bit amounts. They need - to be removed (except at last scanline of (Adam7-reduced) image) before working with pure image buffers - for the Adam7 code, the color convert code and the output to the user. - in and out are allowed to be the same buffer, in may also be higher but still overlapping; in must - have >= ilinebits*h bits, out must have >= olinebits*h bits, olinebits must be <= ilinebits - also used to move bits after earlier such operations happened, e.g. in a sequence of reduced images from Adam7 - only useful if (ilinebits - olinebits) is a value in the range 1..7 - */ - unsigned y; - size_t diff = ilinebits - olinebits; - size_t ibp = 0, obp = 0; /*input and output bit pointers*/ - for(y = 0; y < h; ++y) { - size_t x; - for(x = 0; x < olinebits; ++x) { - unsigned char bit = readBitFromReversedStream(&ibp, in); - setBitOfReversedStream(&obp, out, bit); - } - ibp += diff; - } -} - -/*out must be buffer big enough to contain full image, and in must contain the full decompressed data from -the IDAT chunks (with filter index bytes and possible padding bits) -return value is error*/ -static unsigned postProcessScanlines(unsigned char* out, unsigned char* in, - unsigned w, unsigned h, const LodePNGInfo* info_png) { - /* - This function converts the filtered-padded-interlaced data into pure 2D image buffer with the PNG's colortype. - Steps: - *) if no Adam7: 1) unfilter 2) remove padding bits (= possible extra bits per scanline if bpp < 8) - *) if adam7: 1) 7x unfilter 2) 7x remove padding bits 3) Adam7_deinterlace - NOTE: the in buffer will be overwritten with intermediate data! - */ - unsigned bpp = lodepng_get_bpp(&info_png->color); - if(bpp == 0) return 31; /*error: invalid colortype*/ - - if(info_png->interlace_method == 0) { - if(bpp < 8 && w * bpp != ((w * bpp + 7u) / 8u) * 8u) { - CERROR_TRY_RETURN(unfilter(in, in, w, h, bpp)); - removePaddingBits(out, in, w * bpp, ((w * bpp + 7u) / 8u) * 8u, h); - } - /*we can immediately filter into the out buffer, no other steps needed*/ - else CERROR_TRY_RETURN(unfilter(out, in, w, h, bpp)); - } else /*interlace_method is 1 (Adam7)*/ { - unsigned passw[7], passh[7]; size_t filter_passstart[8], padded_passstart[8], passstart[8]; - unsigned i; - - Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); - - for(i = 0; i != 7; ++i) { - CERROR_TRY_RETURN(unfilter(&in[padded_passstart[i]], &in[filter_passstart[i]], passw[i], passh[i], bpp)); - /*TODO: possible efficiency improvement: if in this reduced image the bits fit nicely in 1 scanline, - move bytes instead of bits or move not at all*/ - if(bpp < 8) { - /*remove padding bits in scanlines; after this there still may be padding - bits between the different reduced images: each reduced image still starts nicely at a byte*/ - removePaddingBits(&in[passstart[i]], &in[padded_passstart[i]], passw[i] * bpp, - ((passw[i] * bpp + 7u) / 8u) * 8u, passh[i]); - } - } - - Adam7_deinterlace(out, in, w, h, bpp); - } - - return 0; -} - -static unsigned readChunk_PLTE(LodePNGColorMode* color, const unsigned char* data, size_t chunkLength) { - unsigned pos = 0, i; - color->palettesize = chunkLength / 3u; - if(color->palettesize == 0 || color->palettesize > 256) return 38; /*error: palette too small or big*/ - lodepng_color_mode_alloc_palette(color); - if(!color->palette && color->palettesize) { - color->palettesize = 0; - return 83; /*alloc fail*/ - } - - for(i = 0; i != color->palettesize; ++i) { - color->palette[4 * i + 0] = data[pos++]; /*R*/ - color->palette[4 * i + 1] = data[pos++]; /*G*/ - color->palette[4 * i + 2] = data[pos++]; /*B*/ - color->palette[4 * i + 3] = 255; /*alpha*/ - } - - return 0; /* OK */ -} - -static unsigned readChunk_tRNS(LodePNGColorMode* color, const unsigned char* data, size_t chunkLength) { - unsigned i; - if(color->colortype == LCT_PALETTE) { - /*error: more alpha values given than there are palette entries*/ - if(chunkLength > color->palettesize) return 39; - - for(i = 0; i != chunkLength; ++i) color->palette[4 * i + 3] = data[i]; - } else if(color->colortype == LCT_GREY) { - /*error: this chunk must be 2 bytes for grayscale image*/ - if(chunkLength != 2) return 30; - - color->key_defined = 1; - color->key_r = color->key_g = color->key_b = 256u * data[0] + data[1]; - } else if(color->colortype == LCT_RGB) { - /*error: this chunk must be 6 bytes for RGB image*/ - if(chunkLength != 6) return 41; - - color->key_defined = 1; - color->key_r = 256u * data[0] + data[1]; - color->key_g = 256u * data[2] + data[3]; - color->key_b = 256u * data[4] + data[5]; - } - else return 42; /*error: tRNS chunk not allowed for other color models*/ - - return 0; /* OK */ -} - - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS -/*background color chunk (bKGD)*/ -static unsigned readChunk_bKGD(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) { - if(info->color.colortype == LCT_PALETTE) { - /*error: this chunk must be 1 byte for indexed color image*/ - if(chunkLength != 1) return 43; - - /*error: invalid palette index, or maybe this chunk appeared before PLTE*/ - if(data[0] >= info->color.palettesize) return 103; - - info->background_defined = 1; - info->background_r = info->background_g = info->background_b = data[0]; - } else if(info->color.colortype == LCT_GREY || info->color.colortype == LCT_GREY_ALPHA) { - /*error: this chunk must be 2 bytes for grayscale image*/ - if(chunkLength != 2) return 44; - - /*the values are truncated to bitdepth in the PNG file*/ - info->background_defined = 1; - info->background_r = info->background_g = info->background_b = 256u * data[0] + data[1]; - } else if(info->color.colortype == LCT_RGB || info->color.colortype == LCT_RGBA) { - /*error: this chunk must be 6 bytes for grayscale image*/ - if(chunkLength != 6) return 45; - - /*the values are truncated to bitdepth in the PNG file*/ - info->background_defined = 1; - info->background_r = 256u * data[0] + data[1]; - info->background_g = 256u * data[2] + data[3]; - info->background_b = 256u * data[4] + data[5]; - } - - return 0; /* OK */ -} - -/*text chunk (tEXt)*/ -static unsigned readChunk_tEXt(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) { - unsigned error = 0; - char *key = 0, *str = 0; - - while(!error) /*not really a while loop, only used to break on error*/ { - unsigned length, string2_begin; - - length = 0; - while(length < chunkLength && data[length] != 0) ++length; - /*even though it's not allowed by the standard, no error is thrown if - there's no null termination char, if the text is empty*/ - if(length < 1 || length > 79) CERROR_BREAK(error, 89); /*keyword too short or long*/ - - key = (char*)lodepng_malloc(length + 1); - if(!key) CERROR_BREAK(error, 83); /*alloc fail*/ - - lodepng_memcpy(key, data, length); - key[length] = 0; - - string2_begin = length + 1; /*skip keyword null terminator*/ - - length = (unsigned)(chunkLength < string2_begin ? 0 : chunkLength - string2_begin); - str = (char*)lodepng_malloc(length + 1); - if(!str) CERROR_BREAK(error, 83); /*alloc fail*/ - - lodepng_memcpy(str, data + string2_begin, length); - str[length] = 0; - - error = lodepng_add_text(info, key, str); - - break; - } - - lodepng_free(key); - lodepng_free(str); - - return error; -} - -/*compressed text chunk (zTXt)*/ -static unsigned readChunk_zTXt(LodePNGInfo* info, const LodePNGDecoderSettings* decoder, - const unsigned char* data, size_t chunkLength) { - unsigned error = 0; - - /*copy the object to change parameters in it*/ - LodePNGDecompressSettings zlibsettings = decoder->zlibsettings; - - unsigned length, string2_begin; - char *key = 0; - unsigned char* str = 0; - size_t size = 0; - - while(!error) /*not really a while loop, only used to break on error*/ { - for(length = 0; length < chunkLength && data[length] != 0; ++length) ; - if(length + 2 >= chunkLength) CERROR_BREAK(error, 75); /*no null termination, corrupt?*/ - if(length < 1 || length > 79) CERROR_BREAK(error, 89); /*keyword too short or long*/ - - key = (char*)lodepng_malloc(length + 1); - if(!key) CERROR_BREAK(error, 83); /*alloc fail*/ - - lodepng_memcpy(key, data, length); - key[length] = 0; - - if(data[length + 1] != 0) CERROR_BREAK(error, 72); /*the 0 byte indicating compression must be 0*/ - - string2_begin = length + 2; - if(string2_begin > chunkLength) CERROR_BREAK(error, 75); /*no null termination, corrupt?*/ - - length = (unsigned)chunkLength - string2_begin; - zlibsettings.max_output_size = decoder->max_text_size; - /*will fail if zlib error, e.g. if length is too small*/ - error = zlib_decompress(&str, &size, 0, &data[string2_begin], - length, &zlibsettings); - /*error: compressed text larger than decoder->max_text_size*/ - if(error && size > zlibsettings.max_output_size) error = 112; - if(error) break; - error = lodepng_add_text_sized(info, key, (char*)str, size); - break; - } - - lodepng_free(key); - lodepng_free(str); - - return error; -} - -/*international text chunk (iTXt)*/ -static unsigned readChunk_iTXt(LodePNGInfo* info, const LodePNGDecoderSettings* decoder, - const unsigned char* data, size_t chunkLength) { - unsigned error = 0; - unsigned i; - - /*copy the object to change parameters in it*/ - LodePNGDecompressSettings zlibsettings = decoder->zlibsettings; - - unsigned length, begin, compressed; - char *key = 0, *langtag = 0, *transkey = 0; - - while(!error) /*not really a while loop, only used to break on error*/ { - /*Quick check if the chunk length isn't too small. Even without check - it'd still fail with other error checks below if it's too short. This just gives a different error code.*/ - if(chunkLength < 5) CERROR_BREAK(error, 30); /*iTXt chunk too short*/ - - /*read the key*/ - for(length = 0; length < chunkLength && data[length] != 0; ++length) ; - if(length + 3 >= chunkLength) CERROR_BREAK(error, 75); /*no null termination char, corrupt?*/ - if(length < 1 || length > 79) CERROR_BREAK(error, 89); /*keyword too short or long*/ - - key = (char*)lodepng_malloc(length + 1); - if(!key) CERROR_BREAK(error, 83); /*alloc fail*/ - - lodepng_memcpy(key, data, length); - key[length] = 0; - - /*read the compression method*/ - compressed = data[length + 1]; - if(data[length + 2] != 0) CERROR_BREAK(error, 72); /*the 0 byte indicating compression must be 0*/ - - /*even though it's not allowed by the standard, no error is thrown if - there's no null termination char, if the text is empty for the next 3 texts*/ - - /*read the langtag*/ - begin = length + 3; - length = 0; - for(i = begin; i < chunkLength && data[i] != 0; ++i) ++length; - - langtag = (char*)lodepng_malloc(length + 1); - if(!langtag) CERROR_BREAK(error, 83); /*alloc fail*/ - - lodepng_memcpy(langtag, data + begin, length); - langtag[length] = 0; - - /*read the transkey*/ - begin += length + 1; - length = 0; - for(i = begin; i < chunkLength && data[i] != 0; ++i) ++length; - - transkey = (char*)lodepng_malloc(length + 1); - if(!transkey) CERROR_BREAK(error, 83); /*alloc fail*/ - - lodepng_memcpy(transkey, data + begin, length); - transkey[length] = 0; - - /*read the actual text*/ - begin += length + 1; - - length = (unsigned)chunkLength < begin ? 0 : (unsigned)chunkLength - begin; - - if(compressed) { - unsigned char* str = 0; - size_t size = 0; - zlibsettings.max_output_size = decoder->max_text_size; - /*will fail if zlib error, e.g. if length is too small*/ - error = zlib_decompress(&str, &size, 0, &data[begin], - length, &zlibsettings); - /*error: compressed text larger than decoder->max_text_size*/ - if(error && size > zlibsettings.max_output_size) error = 112; - if(!error) error = lodepng_add_itext_sized(info, key, langtag, transkey, (char*)str, size); - lodepng_free(str); - } else { - error = lodepng_add_itext_sized(info, key, langtag, transkey, (char*)(data + begin), length); - } - - break; - } - - lodepng_free(key); - lodepng_free(langtag); - lodepng_free(transkey); - - return error; -} - -static unsigned readChunk_tIME(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) { - if(chunkLength != 7) return 73; /*invalid tIME chunk size*/ - - info->time_defined = 1; - info->time.year = 256u * data[0] + data[1]; - info->time.month = data[2]; - info->time.day = data[3]; - info->time.hour = data[4]; - info->time.minute = data[5]; - info->time.second = data[6]; - - return 0; /* OK */ -} - -static unsigned readChunk_pHYs(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) { - if(chunkLength != 9) return 74; /*invalid pHYs chunk size*/ - - info->phys_defined = 1; - info->phys_x = 16777216u * data[0] + 65536u * data[1] + 256u * data[2] + data[3]; - info->phys_y = 16777216u * data[4] + 65536u * data[5] + 256u * data[6] + data[7]; - info->phys_unit = data[8]; - - return 0; /* OK */ -} - -static unsigned readChunk_gAMA(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) { - if(chunkLength != 4) return 96; /*invalid gAMA chunk size*/ - - info->gama_defined = 1; - info->gama_gamma = 16777216u * data[0] + 65536u * data[1] + 256u * data[2] + data[3]; - - return 0; /* OK */ -} - -static unsigned readChunk_cHRM(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) { - if(chunkLength != 32) return 97; /*invalid cHRM chunk size*/ - - info->chrm_defined = 1; - info->chrm_white_x = 16777216u * data[ 0] + 65536u * data[ 1] + 256u * data[ 2] + data[ 3]; - info->chrm_white_y = 16777216u * data[ 4] + 65536u * data[ 5] + 256u * data[ 6] + data[ 7]; - info->chrm_red_x = 16777216u * data[ 8] + 65536u * data[ 9] + 256u * data[10] + data[11]; - info->chrm_red_y = 16777216u * data[12] + 65536u * data[13] + 256u * data[14] + data[15]; - info->chrm_green_x = 16777216u * data[16] + 65536u * data[17] + 256u * data[18] + data[19]; - info->chrm_green_y = 16777216u * data[20] + 65536u * data[21] + 256u * data[22] + data[23]; - info->chrm_blue_x = 16777216u * data[24] + 65536u * data[25] + 256u * data[26] + data[27]; - info->chrm_blue_y = 16777216u * data[28] + 65536u * data[29] + 256u * data[30] + data[31]; - - return 0; /* OK */ -} - -static unsigned readChunk_sRGB(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) { - if(chunkLength != 1) return 98; /*invalid sRGB chunk size (this one is never ignored)*/ - - info->srgb_defined = 1; - info->srgb_intent = data[0]; - - return 0; /* OK */ -} - -static unsigned readChunk_iCCP(LodePNGInfo* info, const LodePNGDecoderSettings* decoder, - const unsigned char* data, size_t chunkLength) { - unsigned error = 0; - unsigned i; - size_t size = 0; - /*copy the object to change parameters in it*/ - LodePNGDecompressSettings zlibsettings = decoder->zlibsettings; - - unsigned length, string2_begin; - - info->iccp_defined = 1; - if(info->iccp_name) lodepng_clear_icc(info); - - for(length = 0; length < chunkLength && data[length] != 0; ++length) ; - if(length + 2 >= chunkLength) return 75; /*no null termination, corrupt?*/ - if(length < 1 || length > 79) return 89; /*keyword too short or long*/ - - info->iccp_name = (char*)lodepng_malloc(length + 1); - if(!info->iccp_name) return 83; /*alloc fail*/ - - info->iccp_name[length] = 0; - for(i = 0; i != length; ++i) info->iccp_name[i] = (char)data[i]; - - if(data[length + 1] != 0) return 72; /*the 0 byte indicating compression must be 0*/ - - string2_begin = length + 2; - if(string2_begin > chunkLength) return 75; /*no null termination, corrupt?*/ - - length = (unsigned)chunkLength - string2_begin; - zlibsettings.max_output_size = decoder->max_icc_size; - error = zlib_decompress(&info->iccp_profile, &size, 0, - &data[string2_begin], - length, &zlibsettings); - /*error: ICC profile larger than decoder->max_icc_size*/ - if(error && size > zlibsettings.max_output_size) error = 113; - info->iccp_profile_size = size; - if(!error && !info->iccp_profile_size) error = 100; /*invalid ICC profile size*/ - return error; -} - -/*significant bits chunk (sBIT)*/ -static unsigned readChunk_sBIT(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) { - unsigned bitdepth = (info->color.colortype == LCT_PALETTE) ? 8 : info->color.bitdepth; - if(info->color.colortype == LCT_GREY) { - /*error: this chunk must be 1 bytes for grayscale image*/ - if(chunkLength != 1) return 114; - if(data[0] == 0 || data[0] > bitdepth) return 115; - info->sbit_defined = 1; - info->sbit_r = info->sbit_g = info->sbit_b = data[0]; /*setting g and b is not required, but sensible*/ - } else if(info->color.colortype == LCT_RGB || info->color.colortype == LCT_PALETTE) { - /*error: this chunk must be 3 bytes for RGB and palette image*/ - if(chunkLength != 3) return 114; - if(data[0] == 0 || data[1] == 0 || data[2] == 0) return 115; - if(data[0] > bitdepth || data[1] > bitdepth || data[2] > bitdepth) return 115; - info->sbit_defined = 1; - info->sbit_r = data[0]; - info->sbit_g = data[1]; - info->sbit_b = data[2]; - } else if(info->color.colortype == LCT_GREY_ALPHA) { - /*error: this chunk must be 2 byte for grayscale with alpha image*/ - if(chunkLength != 2) return 114; - if(data[0] == 0 || data[1] == 0) return 115; - if(data[0] > bitdepth || data[1] > bitdepth) return 115; - info->sbit_defined = 1; - info->sbit_r = info->sbit_g = info->sbit_b = data[0]; /*setting g and b is not required, but sensible*/ - info->sbit_a = data[1]; - } else if(info->color.colortype == LCT_RGBA) { - /*error: this chunk must be 4 bytes for grayscale image*/ - if(chunkLength != 4) return 114; - if(data[0] == 0 || data[1] == 0 || data[2] == 0 || data[3] == 0) return 115; - if(data[0] > bitdepth || data[1] > bitdepth || data[2] > bitdepth || data[3] > bitdepth) return 115; - info->sbit_defined = 1; - info->sbit_r = data[0]; - info->sbit_g = data[1]; - info->sbit_b = data[2]; - info->sbit_a = data[3]; - } - - return 0; /* OK */ -} -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -unsigned lodepng_inspect_chunk(LodePNGState* state, size_t pos, - const unsigned char* in, size_t insize) { - const unsigned char* chunk = in + pos; - unsigned chunkLength; - const unsigned char* data; - unsigned unhandled = 0; - unsigned error = 0; - - if(pos + 4 > insize) return 30; - chunkLength = lodepng_chunk_length(chunk); - if(chunkLength > 2147483647) return 63; - data = lodepng_chunk_data_const(chunk); - if(chunkLength + 12 > insize - pos) return 30; - - if(lodepng_chunk_type_equals(chunk, "PLTE")) { - error = readChunk_PLTE(&state->info_png.color, data, chunkLength); - } else if(lodepng_chunk_type_equals(chunk, "tRNS")) { - error = readChunk_tRNS(&state->info_png.color, data, chunkLength); -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - } else if(lodepng_chunk_type_equals(chunk, "bKGD")) { - error = readChunk_bKGD(&state->info_png, data, chunkLength); - } else if(lodepng_chunk_type_equals(chunk, "tEXt")) { - error = readChunk_tEXt(&state->info_png, data, chunkLength); - } else if(lodepng_chunk_type_equals(chunk, "zTXt")) { - error = readChunk_zTXt(&state->info_png, &state->decoder, data, chunkLength); - } else if(lodepng_chunk_type_equals(chunk, "iTXt")) { - error = readChunk_iTXt(&state->info_png, &state->decoder, data, chunkLength); - } else if(lodepng_chunk_type_equals(chunk, "tIME")) { - error = readChunk_tIME(&state->info_png, data, chunkLength); - } else if(lodepng_chunk_type_equals(chunk, "pHYs")) { - error = readChunk_pHYs(&state->info_png, data, chunkLength); - } else if(lodepng_chunk_type_equals(chunk, "gAMA")) { - error = readChunk_gAMA(&state->info_png, data, chunkLength); - } else if(lodepng_chunk_type_equals(chunk, "cHRM")) { - error = readChunk_cHRM(&state->info_png, data, chunkLength); - } else if(lodepng_chunk_type_equals(chunk, "sRGB")) { - error = readChunk_sRGB(&state->info_png, data, chunkLength); - } else if(lodepng_chunk_type_equals(chunk, "iCCP")) { - error = readChunk_iCCP(&state->info_png, &state->decoder, data, chunkLength); - } else if(lodepng_chunk_type_equals(chunk, "sBIT")) { - error = readChunk_sBIT(&state->info_png, data, chunkLength); -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - } else { - /* unhandled chunk is ok (is not an error) */ - unhandled = 1; - } - - if(!error && !unhandled && !state->decoder.ignore_crc) { - if(lodepng_chunk_check_crc(chunk)) return 57; /*invalid CRC*/ - } - - return error; -} - -/*read a PNG, the result will be in the same color type as the PNG (hence "generic")*/ -static void decodeGeneric(unsigned char** out, unsigned* w, unsigned* h, - LodePNGState* state, - const unsigned char* in, size_t insize) { - unsigned char IEND = 0; - const unsigned char* chunk; /*points to beginning of next chunk*/ - unsigned char* idat; /*the data from idat chunks, zlib compressed*/ - size_t idatsize = 0; - unsigned char* scanlines = 0; - size_t scanlines_size = 0, expected_size = 0; - size_t outsize = 0; - - /*for unknown chunk order*/ - unsigned unknown = 0; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - unsigned critical_pos = 1; /*1 = after IHDR, 2 = after PLTE, 3 = after IDAT*/ -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - - - /* safe output values in case error happens */ - *out = 0; - *w = *h = 0; - - state->error = lodepng_inspect(w, h, state, in, insize); /*reads header and resets other parameters in state->info_png*/ - if(state->error) return; - - if(lodepng_pixel_overflow(*w, *h, &state->info_png.color, &state->info_raw)) { - CERROR_RETURN(state->error, 92); /*overflow possible due to amount of pixels*/ - } - - /*the input filesize is a safe upper bound for the sum of idat chunks size*/ - idat = (unsigned char*)lodepng_malloc(insize); - if(!idat) CERROR_RETURN(state->error, 83); /*alloc fail*/ - - chunk = &in[33]; /*first byte of the first chunk after the header*/ - - /*loop through the chunks, ignoring unknown chunks and stopping at IEND chunk. - IDAT data is put at the start of the in buffer*/ - while(!IEND && !state->error) { - unsigned chunkLength; - const unsigned char* data; /*the data in the chunk*/ - size_t pos = (size_t)(chunk - in); - - /*error: next chunk out of bounds of the in buffer*/ - if(chunk < in || pos + 12 > insize) { - if(state->decoder.ignore_end) break; /*other errors may still happen though*/ - CERROR_BREAK(state->error, 30); - } - - /*length of the data of the chunk, excluding the 12 bytes for length, chunk type and CRC*/ - chunkLength = lodepng_chunk_length(chunk); - /*error: chunk length larger than the max PNG chunk size*/ - if(chunkLength > 2147483647) { - if(state->decoder.ignore_end) break; /*other errors may still happen though*/ - CERROR_BREAK(state->error, 63); - } - - if(pos + (size_t)chunkLength + 12 > insize || pos + (size_t)chunkLength + 12 < pos) { - CERROR_BREAK(state->error, 64); /*error: size of the in buffer too small to contain next chunk (or int overflow)*/ - } - - data = lodepng_chunk_data_const(chunk); - - unknown = 0; - - /*IDAT chunk, containing compressed image data*/ - if(lodepng_chunk_type_equals(chunk, "IDAT")) { - size_t newsize; - if(lodepng_addofl(idatsize, chunkLength, &newsize)) CERROR_BREAK(state->error, 95); - if(newsize > insize) CERROR_BREAK(state->error, 95); - lodepng_memcpy(idat + idatsize, data, chunkLength); - idatsize += chunkLength; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - critical_pos = 3; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - } else if(lodepng_chunk_type_equals(chunk, "IEND")) { - /*IEND chunk*/ - IEND = 1; - } else if(lodepng_chunk_type_equals(chunk, "PLTE")) { - /*palette chunk (PLTE)*/ - state->error = readChunk_PLTE(&state->info_png.color, data, chunkLength); - if(state->error) break; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - critical_pos = 2; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - } else if(lodepng_chunk_type_equals(chunk, "tRNS")) { - /*palette transparency chunk (tRNS). Even though this one is an ancillary chunk , it is still compiled - in without 'LODEPNG_COMPILE_ANCILLARY_CHUNKS' because it contains essential color information that - affects the alpha channel of pixels. */ - state->error = readChunk_tRNS(&state->info_png.color, data, chunkLength); - if(state->error) break; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /*background color chunk (bKGD)*/ - } else if(lodepng_chunk_type_equals(chunk, "bKGD")) { - state->error = readChunk_bKGD(&state->info_png, data, chunkLength); - if(state->error) break; - } else if(lodepng_chunk_type_equals(chunk, "tEXt")) { - /*text chunk (tEXt)*/ - if(state->decoder.read_text_chunks) { - state->error = readChunk_tEXt(&state->info_png, data, chunkLength); - if(state->error) break; - } - } else if(lodepng_chunk_type_equals(chunk, "zTXt")) { - /*compressed text chunk (zTXt)*/ - if(state->decoder.read_text_chunks) { - state->error = readChunk_zTXt(&state->info_png, &state->decoder, data, chunkLength); - if(state->error) break; - } - } else if(lodepng_chunk_type_equals(chunk, "iTXt")) { - /*international text chunk (iTXt)*/ - if(state->decoder.read_text_chunks) { - state->error = readChunk_iTXt(&state->info_png, &state->decoder, data, chunkLength); - if(state->error) break; - } - } else if(lodepng_chunk_type_equals(chunk, "tIME")) { - state->error = readChunk_tIME(&state->info_png, data, chunkLength); - if(state->error) break; - } else if(lodepng_chunk_type_equals(chunk, "pHYs")) { - state->error = readChunk_pHYs(&state->info_png, data, chunkLength); - if(state->error) break; - } else if(lodepng_chunk_type_equals(chunk, "gAMA")) { - state->error = readChunk_gAMA(&state->info_png, data, chunkLength); - if(state->error) break; - } else if(lodepng_chunk_type_equals(chunk, "cHRM")) { - state->error = readChunk_cHRM(&state->info_png, data, chunkLength); - if(state->error) break; - } else if(lodepng_chunk_type_equals(chunk, "sRGB")) { - state->error = readChunk_sRGB(&state->info_png, data, chunkLength); - if(state->error) break; - } else if(lodepng_chunk_type_equals(chunk, "iCCP")) { - state->error = readChunk_iCCP(&state->info_png, &state->decoder, data, chunkLength); - if(state->error) break; - } else if(lodepng_chunk_type_equals(chunk, "sBIT")) { - state->error = readChunk_sBIT(&state->info_png, data, chunkLength); - if(state->error) break; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - } else /*it's not an implemented chunk type, so ignore it: skip over the data*/ { - /*error: unknown critical chunk (5th bit of first byte of chunk type is 0)*/ - if(!state->decoder.ignore_critical && !lodepng_chunk_ancillary(chunk)) { - CERROR_BREAK(state->error, 69); - } - - unknown = 1; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - if(state->decoder.remember_unknown_chunks) { - state->error = lodepng_chunk_append(&state->info_png.unknown_chunks_data[critical_pos - 1], - &state->info_png.unknown_chunks_size[critical_pos - 1], chunk); - if(state->error) break; - } -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - } - - if(!state->decoder.ignore_crc && !unknown) /*check CRC if wanted, only on known chunk types*/ { - if(lodepng_chunk_check_crc(chunk)) CERROR_BREAK(state->error, 57); /*invalid CRC*/ - } - - if(!IEND) chunk = lodepng_chunk_next_const(chunk, in + insize); - } - - if(!state->error && state->info_png.color.colortype == LCT_PALETTE && !state->info_png.color.palette) { - state->error = 106; /* error: PNG file must have PLTE chunk if color type is palette */ - } - - if(!state->error) { - /*predict output size, to allocate exact size for output buffer to avoid more dynamic allocation. - If the decompressed size does not match the prediction, the image must be corrupt.*/ - if(state->info_png.interlace_method == 0) { - size_t bpp = lodepng_get_bpp(&state->info_png.color); - expected_size = lodepng_get_raw_size_idat(*w, *h, bpp); - } else { - size_t bpp = lodepng_get_bpp(&state->info_png.color); - /*Adam-7 interlaced: expected size is the sum of the 7 sub-images sizes*/ - expected_size = 0; - expected_size += lodepng_get_raw_size_idat((*w + 7) >> 3, (*h + 7) >> 3, bpp); - if(*w > 4) expected_size += lodepng_get_raw_size_idat((*w + 3) >> 3, (*h + 7) >> 3, bpp); - expected_size += lodepng_get_raw_size_idat((*w + 3) >> 2, (*h + 3) >> 3, bpp); - if(*w > 2) expected_size += lodepng_get_raw_size_idat((*w + 1) >> 2, (*h + 3) >> 2, bpp); - expected_size += lodepng_get_raw_size_idat((*w + 1) >> 1, (*h + 1) >> 2, bpp); - if(*w > 1) expected_size += lodepng_get_raw_size_idat((*w + 0) >> 1, (*h + 1) >> 1, bpp); - expected_size += lodepng_get_raw_size_idat((*w + 0), (*h + 0) >> 1, bpp); - } - - state->error = zlib_decompress(&scanlines, &scanlines_size, expected_size, idat, idatsize, &state->decoder.zlibsettings); - } - if(!state->error && scanlines_size != expected_size) state->error = 91; /*decompressed size doesn't match prediction*/ - lodepng_free(idat); - - if(!state->error) { - outsize = lodepng_get_raw_size(*w, *h, &state->info_png.color); - *out = (unsigned char*)lodepng_malloc(outsize); - if(!*out) state->error = 83; /*alloc fail*/ - } - if(!state->error) { - lodepng_memset(*out, 0, outsize); - state->error = postProcessScanlines(*out, scanlines, *w, *h, &state->info_png); - } - lodepng_free(scanlines); -} - -unsigned lodepng_decode(unsigned char** out, unsigned* w, unsigned* h, - LodePNGState* state, - const unsigned char* in, size_t insize) { - *out = 0; - decodeGeneric(out, w, h, state, in, insize); - if(state->error) return state->error; - if(!state->decoder.color_convert || lodepng_color_mode_equal(&state->info_raw, &state->info_png.color)) { - /*same color type, no copying or converting of data needed*/ - /*store the info_png color settings on the info_raw so that the info_raw still reflects what colortype - the raw image has to the end user*/ - if(!state->decoder.color_convert) { - state->error = lodepng_color_mode_copy(&state->info_raw, &state->info_png.color); - if(state->error) return state->error; - } - } else { /*color conversion needed*/ - unsigned char* data = *out; - size_t outsize; - - /*TODO: check if this works according to the statement in the documentation: "The converter can convert - from grayscale input color type, to 8-bit grayscale or grayscale with alpha"*/ - if(!(state->info_raw.colortype == LCT_RGB || state->info_raw.colortype == LCT_RGBA) - && !(state->info_raw.bitdepth == 8)) { - return 56; /*unsupported color mode conversion*/ - } - - outsize = lodepng_get_raw_size(*w, *h, &state->info_raw); - *out = (unsigned char*)lodepng_malloc(outsize); - if(!(*out)) { - state->error = 83; /*alloc fail*/ - } - else state->error = lodepng_convert(*out, data, &state->info_raw, - &state->info_png.color, *w, *h); - lodepng_free(data); - } - return state->error; -} - -unsigned lodepng_decode_memory(unsigned char** out, unsigned* w, unsigned* h, const unsigned char* in, - size_t insize, LodePNGColorType colortype, unsigned bitdepth) { - unsigned error; - LodePNGState state; - lodepng_state_init(&state); - state.info_raw.colortype = colortype; - state.info_raw.bitdepth = bitdepth; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /*disable reading things that this function doesn't output*/ - state.decoder.read_text_chunks = 0; - state.decoder.remember_unknown_chunks = 0; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - error = lodepng_decode(out, w, h, &state, in, insize); - lodepng_state_cleanup(&state); - return error; -} - -unsigned lodepng_decode32(unsigned char** out, unsigned* w, unsigned* h, const unsigned char* in, size_t insize) { - return lodepng_decode_memory(out, w, h, in, insize, LCT_RGBA, 8); -} - -unsigned lodepng_decode24(unsigned char** out, unsigned* w, unsigned* h, const unsigned char* in, size_t insize) { - return lodepng_decode_memory(out, w, h, in, insize, LCT_RGB, 8); -} - -#ifdef LODEPNG_COMPILE_DISK -unsigned lodepng_decode_file(unsigned char** out, unsigned* w, unsigned* h, const char* filename, - LodePNGColorType colortype, unsigned bitdepth) { - unsigned char* buffer = 0; - size_t buffersize; - unsigned error; - /* safe output values in case error happens */ - *out = 0; - *w = *h = 0; - error = lodepng_load_file(&buffer, &buffersize, filename); - if(!error) error = lodepng_decode_memory(out, w, h, buffer, buffersize, colortype, bitdepth); - lodepng_free(buffer); - return error; -} - -unsigned lodepng_decode32_file(unsigned char** out, unsigned* w, unsigned* h, const char* filename) { - return lodepng_decode_file(out, w, h, filename, LCT_RGBA, 8); -} - -unsigned lodepng_decode24_file(unsigned char** out, unsigned* w, unsigned* h, const char* filename) { - return lodepng_decode_file(out, w, h, filename, LCT_RGB, 8); -} -#endif /*LODEPNG_COMPILE_DISK*/ - -void lodepng_decoder_settings_init(LodePNGDecoderSettings* settings) { - settings->color_convert = 1; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - settings->read_text_chunks = 1; - settings->remember_unknown_chunks = 0; - settings->max_text_size = 16777216; - settings->max_icc_size = 16777216; /* 16MB is much more than enough for any reasonable ICC profile */ -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - settings->ignore_crc = 0; - settings->ignore_critical = 0; - settings->ignore_end = 0; - lodepng_decompress_settings_init(&settings->zlibsettings); -} - -#endif /*LODEPNG_COMPILE_DECODER*/ - -#if defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) - -void lodepng_state_init(LodePNGState* state) { -#ifdef LODEPNG_COMPILE_DECODER - lodepng_decoder_settings_init(&state->decoder); -#endif /*LODEPNG_COMPILE_DECODER*/ -#ifdef LODEPNG_COMPILE_ENCODER - lodepng_encoder_settings_init(&state->encoder); -#endif /*LODEPNG_COMPILE_ENCODER*/ - lodepng_color_mode_init(&state->info_raw); - lodepng_info_init(&state->info_png); - state->error = 1; -} - -void lodepng_state_cleanup(LodePNGState* state) { - lodepng_color_mode_cleanup(&state->info_raw); - lodepng_info_cleanup(&state->info_png); -} - -void lodepng_state_copy(LodePNGState* dest, const LodePNGState* source) { - lodepng_state_cleanup(dest); - *dest = *source; - lodepng_color_mode_init(&dest->info_raw); - lodepng_info_init(&dest->info_png); - dest->error = lodepng_color_mode_copy(&dest->info_raw, &source->info_raw); if(dest->error) return; - dest->error = lodepng_info_copy(&dest->info_png, &source->info_png); if(dest->error) return; -} - -#endif /* defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) */ - -#ifdef LODEPNG_COMPILE_ENCODER - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / PNG Encoder / */ -/* ////////////////////////////////////////////////////////////////////////// */ - - -static unsigned writeSignature(ucvector* out) { - size_t pos = out->size; - const unsigned char signature[] = {137, 80, 78, 71, 13, 10, 26, 10}; - /*8 bytes PNG signature, aka the magic bytes*/ - if(!ucvector_resize(out, out->size + 8)) return 83; /*alloc fail*/ - lodepng_memcpy(out->data + pos, signature, 8); - return 0; -} - -static unsigned addChunk_IHDR(ucvector* out, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth, unsigned interlace_method) { - unsigned char *chunk, *data; - CERROR_TRY_RETURN(lodepng_chunk_init(&chunk, out, 13, "IHDR")); - data = chunk + 8; - - lodepng_set32bitInt(data + 0, w); /*width*/ - lodepng_set32bitInt(data + 4, h); /*height*/ - data[8] = (unsigned char)bitdepth; /*bit depth*/ - data[9] = (unsigned char)colortype; /*color type*/ - data[10] = 0; /*compression method*/ - data[11] = 0; /*filter method*/ - data[12] = interlace_method; /*interlace method*/ - - lodepng_chunk_generate_crc(chunk); - return 0; -} - -/* only adds the chunk if needed (there is a key or palette with alpha) */ -static unsigned addChunk_PLTE(ucvector* out, const LodePNGColorMode* info) { - unsigned char* chunk; - size_t i, j = 8; - - if(info->palettesize == 0 || info->palettesize > 256) { - return 68; /*invalid palette size, it is only allowed to be 1-256*/ - } - - CERROR_TRY_RETURN(lodepng_chunk_init(&chunk, out, info->palettesize * 3, "PLTE")); - - for(i = 0; i != info->palettesize; ++i) { - /*add all channels except alpha channel*/ - chunk[j++] = info->palette[i * 4 + 0]; - chunk[j++] = info->palette[i * 4 + 1]; - chunk[j++] = info->palette[i * 4 + 2]; - } - - lodepng_chunk_generate_crc(chunk); - return 0; -} - -static unsigned addChunk_tRNS(ucvector* out, const LodePNGColorMode* info) { - unsigned char* chunk = 0; - - if(info->colortype == LCT_PALETTE) { - size_t i, amount = info->palettesize; - /*the tail of palette values that all have 255 as alpha, does not have to be encoded*/ - for(i = info->palettesize; i != 0; --i) { - if(info->palette[4 * (i - 1) + 3] != 255) break; - --amount; - } - if(amount) { - CERROR_TRY_RETURN(lodepng_chunk_init(&chunk, out, amount, "tRNS")); - /*add the alpha channel values from the palette*/ - for(i = 0; i != amount; ++i) chunk[8 + i] = info->palette[4 * i + 3]; - } - } else if(info->colortype == LCT_GREY) { - if(info->key_defined) { - CERROR_TRY_RETURN(lodepng_chunk_init(&chunk, out, 2, "tRNS")); - chunk[8] = (unsigned char)(info->key_r >> 8); - chunk[9] = (unsigned char)(info->key_r & 255); - } - } else if(info->colortype == LCT_RGB) { - if(info->key_defined) { - CERROR_TRY_RETURN(lodepng_chunk_init(&chunk, out, 6, "tRNS")); - chunk[8] = (unsigned char)(info->key_r >> 8); - chunk[9] = (unsigned char)(info->key_r & 255); - chunk[10] = (unsigned char)(info->key_g >> 8); - chunk[11] = (unsigned char)(info->key_g & 255); - chunk[12] = (unsigned char)(info->key_b >> 8); - chunk[13] = (unsigned char)(info->key_b & 255); - } - } - - if(chunk) lodepng_chunk_generate_crc(chunk); - return 0; -} - -static unsigned addChunk_IDAT(ucvector* out, const unsigned char* data, size_t datasize, - LodePNGCompressSettings* zlibsettings) { - unsigned error = 0; - unsigned char* zlib = 0; - size_t zlibsize = 0; - - error = zlib_compress(&zlib, &zlibsize, data, datasize, zlibsettings); - if(!error) { - error = lodepng_chunk_createv(out, zlibsize, "IDAT", zlib); - } - lodepng_free(zlib); - return error; -} - -static unsigned addChunk_IEND(ucvector* out) { - return lodepng_chunk_createv(out, 0, "IEND", 0); -} - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - -static unsigned addChunk_tEXt(ucvector* out, const char* keyword, const char* textstring) { - unsigned char* chunk = 0; - size_t keysize = lodepng_strlen(keyword), textsize = lodepng_strlen(textstring); - size_t size = keysize + 1 + textsize; - if(keysize < 1 || keysize > 79) return 89; /*error: invalid keyword size*/ - CERROR_TRY_RETURN(lodepng_chunk_init(&chunk, out, size, "tEXt")); - lodepng_memcpy(chunk + 8, keyword, keysize); - chunk[8 + keysize] = 0; /*null termination char*/ - lodepng_memcpy(chunk + 9 + keysize, textstring, textsize); - lodepng_chunk_generate_crc(chunk); - return 0; -} - -static unsigned addChunk_zTXt(ucvector* out, const char* keyword, const char* textstring, - LodePNGCompressSettings* zlibsettings) { - unsigned error = 0; - unsigned char* chunk = 0; - unsigned char* compressed = 0; - size_t compressedsize = 0; - size_t textsize = lodepng_strlen(textstring); - size_t keysize = lodepng_strlen(keyword); - if(keysize < 1 || keysize > 79) return 89; /*error: invalid keyword size*/ - - error = zlib_compress(&compressed, &compressedsize, - (const unsigned char*)textstring, textsize, zlibsettings); - if(!error) { - size_t size = keysize + 2 + compressedsize; - error = lodepng_chunk_init(&chunk, out, size, "zTXt"); - } - if(!error) { - lodepng_memcpy(chunk + 8, keyword, keysize); - chunk[8 + keysize] = 0; /*null termination char*/ - chunk[9 + keysize] = 0; /*compression method: 0*/ - lodepng_memcpy(chunk + 10 + keysize, compressed, compressedsize); - lodepng_chunk_generate_crc(chunk); - } - - lodepng_free(compressed); - return error; -} - -static unsigned addChunk_iTXt(ucvector* out, unsigned compress, const char* keyword, const char* langtag, - const char* transkey, const char* textstring, LodePNGCompressSettings* zlibsettings) { - unsigned error = 0; - unsigned char* chunk = 0; - unsigned char* compressed = 0; - size_t compressedsize = 0; - size_t textsize = lodepng_strlen(textstring); - size_t keysize = lodepng_strlen(keyword), langsize = lodepng_strlen(langtag), transsize = lodepng_strlen(transkey); - - if(keysize < 1 || keysize > 79) return 89; /*error: invalid keyword size*/ - - if(compress) { - error = zlib_compress(&compressed, &compressedsize, - (const unsigned char*)textstring, textsize, zlibsettings); - } - if(!error) { - size_t size = keysize + 3 + langsize + 1 + transsize + 1 + (compress ? compressedsize : textsize); - error = lodepng_chunk_init(&chunk, out, size, "iTXt"); - } - if(!error) { - size_t pos = 8; - lodepng_memcpy(chunk + pos, keyword, keysize); - pos += keysize; - chunk[pos++] = 0; /*null termination char*/ - chunk[pos++] = (compress ? 1 : 0); /*compression flag*/ - chunk[pos++] = 0; /*compression method: 0*/ - lodepng_memcpy(chunk + pos, langtag, langsize); - pos += langsize; - chunk[pos++] = 0; /*null termination char*/ - lodepng_memcpy(chunk + pos, transkey, transsize); - pos += transsize; - chunk[pos++] = 0; /*null termination char*/ - if(compress) { - lodepng_memcpy(chunk + pos, compressed, compressedsize); - } else { - lodepng_memcpy(chunk + pos, textstring, textsize); - } - lodepng_chunk_generate_crc(chunk); - } - - lodepng_free(compressed); - return error; -} - -static unsigned addChunk_bKGD(ucvector* out, const LodePNGInfo* info) { - unsigned char* chunk = 0; - if(info->color.colortype == LCT_GREY || info->color.colortype == LCT_GREY_ALPHA) { - CERROR_TRY_RETURN(lodepng_chunk_init(&chunk, out, 2, "bKGD")); - chunk[8] = (unsigned char)(info->background_r >> 8); - chunk[9] = (unsigned char)(info->background_r & 255); - } else if(info->color.colortype == LCT_RGB || info->color.colortype == LCT_RGBA) { - CERROR_TRY_RETURN(lodepng_chunk_init(&chunk, out, 6, "bKGD")); - chunk[8] = (unsigned char)(info->background_r >> 8); - chunk[9] = (unsigned char)(info->background_r & 255); - chunk[10] = (unsigned char)(info->background_g >> 8); - chunk[11] = (unsigned char)(info->background_g & 255); - chunk[12] = (unsigned char)(info->background_b >> 8); - chunk[13] = (unsigned char)(info->background_b & 255); - } else if(info->color.colortype == LCT_PALETTE) { - CERROR_TRY_RETURN(lodepng_chunk_init(&chunk, out, 1, "bKGD")); - chunk[8] = (unsigned char)(info->background_r & 255); /*palette index*/ - } - if(chunk) lodepng_chunk_generate_crc(chunk); - return 0; -} - -static unsigned addChunk_tIME(ucvector* out, const LodePNGTime* time) { - unsigned char* chunk; - CERROR_TRY_RETURN(lodepng_chunk_init(&chunk, out, 7, "tIME")); - chunk[8] = (unsigned char)(time->year >> 8); - chunk[9] = (unsigned char)(time->year & 255); - chunk[10] = (unsigned char)time->month; - chunk[11] = (unsigned char)time->day; - chunk[12] = (unsigned char)time->hour; - chunk[13] = (unsigned char)time->minute; - chunk[14] = (unsigned char)time->second; - lodepng_chunk_generate_crc(chunk); - return 0; -} - -static unsigned addChunk_pHYs(ucvector* out, const LodePNGInfo* info) { - unsigned char* chunk; - CERROR_TRY_RETURN(lodepng_chunk_init(&chunk, out, 9, "pHYs")); - lodepng_set32bitInt(chunk + 8, info->phys_x); - lodepng_set32bitInt(chunk + 12, info->phys_y); - chunk[16] = info->phys_unit; - lodepng_chunk_generate_crc(chunk); - return 0; -} - -static unsigned addChunk_gAMA(ucvector* out, const LodePNGInfo* info) { - unsigned char* chunk; - CERROR_TRY_RETURN(lodepng_chunk_init(&chunk, out, 4, "gAMA")); - lodepng_set32bitInt(chunk + 8, info->gama_gamma); - lodepng_chunk_generate_crc(chunk); - return 0; -} - -static unsigned addChunk_cHRM(ucvector* out, const LodePNGInfo* info) { - unsigned char* chunk; - CERROR_TRY_RETURN(lodepng_chunk_init(&chunk, out, 32, "cHRM")); - lodepng_set32bitInt(chunk + 8, info->chrm_white_x); - lodepng_set32bitInt(chunk + 12, info->chrm_white_y); - lodepng_set32bitInt(chunk + 16, info->chrm_red_x); - lodepng_set32bitInt(chunk + 20, info->chrm_red_y); - lodepng_set32bitInt(chunk + 24, info->chrm_green_x); - lodepng_set32bitInt(chunk + 28, info->chrm_green_y); - lodepng_set32bitInt(chunk + 32, info->chrm_blue_x); - lodepng_set32bitInt(chunk + 36, info->chrm_blue_y); - lodepng_chunk_generate_crc(chunk); - return 0; -} - -static unsigned addChunk_sRGB(ucvector* out, const LodePNGInfo* info) { - unsigned char data = info->srgb_intent; - return lodepng_chunk_createv(out, 1, "sRGB", &data); -} - -static unsigned addChunk_iCCP(ucvector* out, const LodePNGInfo* info, LodePNGCompressSettings* zlibsettings) { - unsigned error = 0; - unsigned char* chunk = 0; - unsigned char* compressed = 0; - size_t compressedsize = 0; - size_t keysize = lodepng_strlen(info->iccp_name); - - if(keysize < 1 || keysize > 79) return 89; /*error: invalid keyword size*/ - error = zlib_compress(&compressed, &compressedsize, - info->iccp_profile, info->iccp_profile_size, zlibsettings); - if(!error) { - size_t size = keysize + 2 + compressedsize; - error = lodepng_chunk_init(&chunk, out, size, "iCCP"); - } - if(!error) { - lodepng_memcpy(chunk + 8, info->iccp_name, keysize); - chunk[8 + keysize] = 0; /*null termination char*/ - chunk[9 + keysize] = 0; /*compression method: 0*/ - lodepng_memcpy(chunk + 10 + keysize, compressed, compressedsize); - lodepng_chunk_generate_crc(chunk); - } - - lodepng_free(compressed); - return error; -} - -static unsigned addChunk_sBIT(ucvector* out, const LodePNGInfo* info) { - unsigned bitdepth = (info->color.colortype == LCT_PALETTE) ? 8 : info->color.bitdepth; - unsigned char* chunk = 0; - if(info->color.colortype == LCT_GREY) { - if(info->sbit_r == 0 || info->sbit_r > bitdepth) return 115; - CERROR_TRY_RETURN(lodepng_chunk_init(&chunk, out, 1, "sBIT")); - chunk[8] = info->sbit_r; - } else if(info->color.colortype == LCT_RGB || info->color.colortype == LCT_PALETTE) { - if(info->sbit_r == 0 || info->sbit_g == 0 || info->sbit_b == 0) return 115; - if(info->sbit_r > bitdepth || info->sbit_g > bitdepth || info->sbit_b > bitdepth) return 115; - CERROR_TRY_RETURN(lodepng_chunk_init(&chunk, out, 3, "sBIT")); - chunk[8] = info->sbit_r; - chunk[9] = info->sbit_g; - chunk[10] = info->sbit_b; - } else if(info->color.colortype == LCT_GREY_ALPHA) { - if(info->sbit_r == 0 || info->sbit_a == 0) return 115; - if(info->sbit_r > bitdepth || info->sbit_a > bitdepth) return 115; - CERROR_TRY_RETURN(lodepng_chunk_init(&chunk, out, 2, "sBIT")); - chunk[8] = info->sbit_r; - chunk[9] = info->sbit_a; - } else if(info->color.colortype == LCT_RGBA) { - if(info->sbit_r == 0 || info->sbit_g == 0 || info->sbit_b == 0 || info->sbit_a == 0 || - info->sbit_r > bitdepth || info->sbit_g > bitdepth || - info->sbit_b > bitdepth || info->sbit_a > bitdepth) { - return 115; - } - CERROR_TRY_RETURN(lodepng_chunk_init(&chunk, out, 4, "sBIT")); - chunk[8] = info->sbit_r; - chunk[9] = info->sbit_g; - chunk[10] = info->sbit_b; - chunk[11] = info->sbit_a; - } - if(chunk) lodepng_chunk_generate_crc(chunk); - return 0; -} - -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -static void filterScanline(unsigned char* out, const unsigned char* scanline, const unsigned char* prevline, - size_t length, size_t bytewidth, unsigned char filterType) { - size_t i; - switch(filterType) { - case 0: /*None*/ - for(i = 0; i != length; ++i) out[i] = scanline[i]; - break; - case 1: /*Sub*/ - for(i = 0; i != bytewidth; ++i) out[i] = scanline[i]; - for(i = bytewidth; i < length; ++i) out[i] = scanline[i] - scanline[i - bytewidth]; - break; - case 2: /*Up*/ - if(prevline) { - for(i = 0; i != length; ++i) out[i] = scanline[i] - prevline[i]; - } else { - for(i = 0; i != length; ++i) out[i] = scanline[i]; - } - break; - case 3: /*Average*/ - if(prevline) { - for(i = 0; i != bytewidth; ++i) out[i] = scanline[i] - (prevline[i] >> 1); - for(i = bytewidth; i < length; ++i) out[i] = scanline[i] - ((scanline[i - bytewidth] + prevline[i]) >> 1); - } else { - for(i = 0; i != bytewidth; ++i) out[i] = scanline[i]; - for(i = bytewidth; i < length; ++i) out[i] = scanline[i] - (scanline[i - bytewidth] >> 1); - } - break; - case 4: /*Paeth*/ - if(prevline) { - /*paethPredictor(0, prevline[i], 0) is always prevline[i]*/ - for(i = 0; i != bytewidth; ++i) out[i] = (scanline[i] - prevline[i]); - for(i = bytewidth; i < length; ++i) { - out[i] = (scanline[i] - paethPredictor(scanline[i - bytewidth], prevline[i], prevline[i - bytewidth])); - } - } else { - for(i = 0; i != bytewidth; ++i) out[i] = scanline[i]; - /*paethPredictor(scanline[i - bytewidth], 0, 0) is always scanline[i - bytewidth]*/ - for(i = bytewidth; i < length; ++i) out[i] = (scanline[i] - scanline[i - bytewidth]); - } - break; - default: return; /*invalid filter type given*/ - } -} - -/* integer binary logarithm, max return value is 31 */ -static size_t ilog2(size_t i) { - size_t result = 0; - if(i >= 65536) { result += 16; i >>= 16; } - if(i >= 256) { result += 8; i >>= 8; } - if(i >= 16) { result += 4; i >>= 4; } - if(i >= 4) { result += 2; i >>= 2; } - if(i >= 2) { result += 1; /*i >>= 1;*/ } - return result; -} - -/* integer approximation for i * log2(i), helper function for LFS_ENTROPY */ -static size_t ilog2i(size_t i) { - size_t l; - if(i == 0) return 0; - l = ilog2(i); - /* approximate i*log2(i): l is integer logarithm, ((i - (1u << l)) << 1u) - linearly approximates the missing fractional part multiplied by i */ - return i * l + ((i - (1u << l)) << 1u); -} - -static unsigned filter(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, - const LodePNGColorMode* color, const LodePNGEncoderSettings* settings) { - /* - For PNG filter method 0 - out must be a buffer with as size: h + (w * h * bpp + 7u) / 8u, because there are - the scanlines with 1 extra byte per scanline - */ - - unsigned bpp = lodepng_get_bpp(color); - /*the width of a scanline in bytes, not including the filter type*/ - size_t linebytes = lodepng_get_raw_size_idat(w, 1, bpp) - 1u; - - /*bytewidth is used for filtering, is 1 when bpp < 8, number of bytes per pixel otherwise*/ - size_t bytewidth = (bpp + 7u) / 8u; - const unsigned char* prevline = 0; - unsigned x, y; - unsigned error = 0; - LodePNGFilterStrategy strategy = settings->filter_strategy; - - /* - There is a heuristic called the minimum sum of absolute differences heuristic, suggested by the PNG standard: - * If the image type is Palette, or the bit depth is smaller than 8, then do not filter the image (i.e. - use fixed filtering, with the filter None). - * (The other case) If the image type is Grayscale or RGB (with or without Alpha), and the bit depth is - not smaller than 8, then use adaptive filtering heuristic as follows: independently for each row, apply - all five filters and select the filter that produces the smallest sum of absolute values per row. - This heuristic is used if filter strategy is LFS_MINSUM and filter_palette_zero is true. - - If filter_palette_zero is true and filter_strategy is not LFS_MINSUM, the above heuristic is followed, - but for "the other case", whatever strategy filter_strategy is set to instead of the minimum sum - heuristic is used. - */ - if(settings->filter_palette_zero && - (color->colortype == LCT_PALETTE || color->bitdepth < 8)) strategy = LFS_ZERO; - - if(bpp == 0) return 31; /*error: invalid color type*/ - - if(strategy >= LFS_ZERO && strategy <= LFS_FOUR) { - unsigned char type = (unsigned char)strategy; - for(y = 0; y != h; ++y) { - size_t outindex = (1 + linebytes) * y; /*the extra filterbyte added to each row*/ - size_t inindex = linebytes * y; - out[outindex] = type; /*filter type byte*/ - filterScanline(&out[outindex + 1], &in[inindex], prevline, linebytes, bytewidth, type); - prevline = &in[inindex]; - } - } else if(strategy == LFS_MINSUM) { - /*adaptive filtering*/ - unsigned char* attempt[5]; /*five filtering attempts, one for each filter type*/ - size_t smallest = 0; - unsigned char type, bestType = 0; - - for(type = 0; type != 5; ++type) { - attempt[type] = (unsigned char*)lodepng_malloc(linebytes); - if(!attempt[type]) error = 83; /*alloc fail*/ - } - - if(!error) { - for(y = 0; y != h; ++y) { - /*try the 5 filter types*/ - for(type = 0; type != 5; ++type) { - size_t sum = 0; - filterScanline(attempt[type], &in[y * linebytes], prevline, linebytes, bytewidth, type); - - /*calculate the sum of the result*/ - if(type == 0) { - for(x = 0; x != linebytes; ++x) sum += (unsigned char)(attempt[type][x]); - } else { - for(x = 0; x != linebytes; ++x) { - /*For differences, each byte should be treated as signed, values above 127 are negative - (converted to signed char). Filtertype 0 isn't a difference though, so use unsigned there. - This means filtertype 0 is almost never chosen, but that is justified.*/ - unsigned char s = attempt[type][x]; - sum += s < 128 ? s : (255U - s); - } - } - - /*check if this is smallest sum (or if type == 0 it's the first case so always store the values)*/ - if(type == 0 || sum < smallest) { - bestType = type; - smallest = sum; - } - } - - prevline = &in[y * linebytes]; - - /*now fill the out values*/ - out[y * (linebytes + 1)] = bestType; /*the first byte of a scanline will be the filter type*/ - for(x = 0; x != linebytes; ++x) out[y * (linebytes + 1) + 1 + x] = attempt[bestType][x]; - } - } - - for(type = 0; type != 5; ++type) lodepng_free(attempt[type]); - } else if(strategy == LFS_ENTROPY) { - unsigned char* attempt[5]; /*five filtering attempts, one for each filter type*/ - size_t bestSum = 0; - unsigned type, bestType = 0; - unsigned count[256]; - - for(type = 0; type != 5; ++type) { - attempt[type] = (unsigned char*)lodepng_malloc(linebytes); - if(!attempt[type]) error = 83; /*alloc fail*/ - } - - if(!error) { - for(y = 0; y != h; ++y) { - /*try the 5 filter types*/ - for(type = 0; type != 5; ++type) { - size_t sum = 0; - filterScanline(attempt[type], &in[y * linebytes], prevline, linebytes, bytewidth, type); - lodepng_memset(count, 0, 256 * sizeof(*count)); - for(x = 0; x != linebytes; ++x) ++count[attempt[type][x]]; - ++count[type]; /*the filter type itself is part of the scanline*/ - for(x = 0; x != 256; ++x) { - sum += ilog2i(count[x]); - } - /*check if this is smallest sum (or if type == 0 it's the first case so always store the values)*/ - if(type == 0 || sum > bestSum) { - bestType = type; - bestSum = sum; - } - } - - prevline = &in[y * linebytes]; - - /*now fill the out values*/ - out[y * (linebytes + 1)] = bestType; /*the first byte of a scanline will be the filter type*/ - for(x = 0; x != linebytes; ++x) out[y * (linebytes + 1) + 1 + x] = attempt[bestType][x]; - } - } - - for(type = 0; type != 5; ++type) lodepng_free(attempt[type]); - } else if(strategy == LFS_PREDEFINED) { - for(y = 0; y != h; ++y) { - size_t outindex = (1 + linebytes) * y; /*the extra filterbyte added to each row*/ - size_t inindex = linebytes * y; - unsigned char type = settings->predefined_filters[y]; - out[outindex] = type; /*filter type byte*/ - filterScanline(&out[outindex + 1], &in[inindex], prevline, linebytes, bytewidth, type); - prevline = &in[inindex]; - } - } else if(strategy == LFS_BRUTE_FORCE) { - /*brute force filter chooser. - deflate the scanline after every filter attempt to see which one deflates best. - This is very slow and gives only slightly smaller, sometimes even larger, result*/ - size_t size[5]; - unsigned char* attempt[5]; /*five filtering attempts, one for each filter type*/ - size_t smallest = 0; - unsigned type = 0, bestType = 0; - unsigned char* dummy; - LodePNGCompressSettings zlibsettings; - lodepng_memcpy(&zlibsettings, &settings->zlibsettings, sizeof(LodePNGCompressSettings)); - /*use fixed tree on the attempts so that the tree is not adapted to the filtertype on purpose, - to simulate the true case where the tree is the same for the whole image. Sometimes it gives - better result with dynamic tree anyway. Using the fixed tree sometimes gives worse, but in rare - cases better compression. It does make this a bit less slow, so it's worth doing this.*/ - zlibsettings.btype = 1; - /*a custom encoder likely doesn't read the btype setting and is optimized for complete PNG - images only, so disable it*/ - zlibsettings.custom_zlib = 0; - zlibsettings.custom_deflate = 0; - for(type = 0; type != 5; ++type) { - attempt[type] = (unsigned char*)lodepng_malloc(linebytes); - if(!attempt[type]) error = 83; /*alloc fail*/ - } - if(!error) { - for(y = 0; y != h; ++y) /*try the 5 filter types*/ { - for(type = 0; type != 5; ++type) { - unsigned testsize = (unsigned)linebytes; - /*if(testsize > 8) testsize /= 8;*/ /*it already works good enough by testing a part of the row*/ - - filterScanline(attempt[type], &in[y * linebytes], prevline, linebytes, bytewidth, type); - size[type] = 0; - dummy = 0; - zlib_compress(&dummy, &size[type], attempt[type], testsize, &zlibsettings); - lodepng_free(dummy); - /*check if this is smallest size (or if type == 0 it's the first case so always store the values)*/ - if(type == 0 || size[type] < smallest) { - bestType = type; - smallest = size[type]; - } - } - prevline = &in[y * linebytes]; - out[y * (linebytes + 1)] = bestType; /*the first byte of a scanline will be the filter type*/ - for(x = 0; x != linebytes; ++x) out[y * (linebytes + 1) + 1 + x] = attempt[bestType][x]; - } - } - for(type = 0; type != 5; ++type) lodepng_free(attempt[type]); - } - else return 88; /* unknown filter strategy */ - - return error; -} - -static void addPaddingBits(unsigned char* out, const unsigned char* in, - size_t olinebits, size_t ilinebits, unsigned h) { - /*The opposite of the removePaddingBits function - olinebits must be >= ilinebits*/ - unsigned y; - size_t diff = olinebits - ilinebits; - size_t obp = 0, ibp = 0; /*bit pointers*/ - for(y = 0; y != h; ++y) { - size_t x; - for(x = 0; x < ilinebits; ++x) { - unsigned char bit = readBitFromReversedStream(&ibp, in); - setBitOfReversedStream(&obp, out, bit); - } - /*obp += diff; --> no, fill in some value in the padding bits too, to avoid - "Use of uninitialised value of size ###" warning from valgrind*/ - for(x = 0; x != diff; ++x) setBitOfReversedStream(&obp, out, 0); - } -} - -/* -in: non-interlaced image with size w*h -out: the same pixels, but re-ordered according to PNG's Adam7 interlacing, with - no padding bits between scanlines, but between reduced images so that each - reduced image starts at a byte. -bpp: bits per pixel -there are no padding bits, not between scanlines, not between reduced images -in has the following size in bits: w * h * bpp. -out is possibly bigger due to padding bits between reduced images -NOTE: comments about padding bits are only relevant if bpp < 8 -*/ -static void Adam7_interlace(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, unsigned bpp) { - unsigned passw[7], passh[7]; - size_t filter_passstart[8], padded_passstart[8], passstart[8]; - unsigned i; - - Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); - - if(bpp >= 8) { - for(i = 0; i != 7; ++i) { - unsigned x, y, b; - size_t bytewidth = bpp / 8u; - for(y = 0; y < passh[i]; ++y) - for(x = 0; x < passw[i]; ++x) { - size_t pixelinstart = ((ADAM7_IY[i] + y * ADAM7_DY[i]) * w + ADAM7_IX[i] + x * ADAM7_DX[i]) * bytewidth; - size_t pixeloutstart = passstart[i] + (y * passw[i] + x) * bytewidth; - for(b = 0; b < bytewidth; ++b) { - out[pixeloutstart + b] = in[pixelinstart + b]; - } - } - } - } else /*bpp < 8: Adam7 with pixels < 8 bit is a bit trickier: with bit pointers*/ { - for(i = 0; i != 7; ++i) { - unsigned x, y, b; - unsigned ilinebits = bpp * passw[i]; - unsigned olinebits = bpp * w; - size_t obp, ibp; /*bit pointers (for out and in buffer)*/ - for(y = 0; y < passh[i]; ++y) - for(x = 0; x < passw[i]; ++x) { - ibp = (ADAM7_IY[i] + y * ADAM7_DY[i]) * olinebits + (ADAM7_IX[i] + x * ADAM7_DX[i]) * bpp; - obp = (8 * passstart[i]) + (y * ilinebits + x * bpp); - for(b = 0; b < bpp; ++b) { - unsigned char bit = readBitFromReversedStream(&ibp, in); - setBitOfReversedStream(&obp, out, bit); - } - } - } - } -} - -/*out must be buffer big enough to contain uncompressed IDAT chunk data, and in must contain the full image. -return value is error**/ -static unsigned preProcessScanlines(unsigned char** out, size_t* outsize, const unsigned char* in, - unsigned w, unsigned h, - const LodePNGInfo* info_png, const LodePNGEncoderSettings* settings) { - /* - This function converts the pure 2D image with the PNG's colortype, into filtered-padded-interlaced data. Steps: - *) if no Adam7: 1) add padding bits (= possible extra bits per scanline if bpp < 8) 2) filter - *) if adam7: 1) Adam7_interlace 2) 7x add padding bits 3) 7x filter - */ - unsigned bpp = lodepng_get_bpp(&info_png->color); - unsigned error = 0; - - if(info_png->interlace_method == 0) { - *outsize = h + (h * ((w * bpp + 7u) / 8u)); /*image size plus an extra byte per scanline + possible padding bits*/ - *out = (unsigned char*)lodepng_malloc(*outsize); - if(!(*out) && (*outsize)) error = 83; /*alloc fail*/ - - if(!error) { - /*non multiple of 8 bits per scanline, padding bits needed per scanline*/ - if(bpp < 8 && w * bpp != ((w * bpp + 7u) / 8u) * 8u) { - unsigned char* padded = (unsigned char*)lodepng_malloc(h * ((w * bpp + 7u) / 8u)); - if(!padded) error = 83; /*alloc fail*/ - if(!error) { - addPaddingBits(padded, in, ((w * bpp + 7u) / 8u) * 8u, w * bpp, h); - error = filter(*out, padded, w, h, &info_png->color, settings); - } - lodepng_free(padded); - } else { - /*we can immediately filter into the out buffer, no other steps needed*/ - error = filter(*out, in, w, h, &info_png->color, settings); - } - } - } else /*interlace_method is 1 (Adam7)*/ { - unsigned passw[7], passh[7]; - size_t filter_passstart[8], padded_passstart[8], passstart[8]; - unsigned char* adam7; - - Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); - - *outsize = filter_passstart[7]; /*image size plus an extra byte per scanline + possible padding bits*/ - *out = (unsigned char*)lodepng_malloc(*outsize); - if(!(*out)) error = 83; /*alloc fail*/ - - adam7 = (unsigned char*)lodepng_malloc(passstart[7]); - if(!adam7 && passstart[7]) error = 83; /*alloc fail*/ - - if(!error) { - unsigned i; - - Adam7_interlace(adam7, in, w, h, bpp); - for(i = 0; i != 7; ++i) { - if(bpp < 8) { - unsigned char* padded = (unsigned char*)lodepng_malloc(padded_passstart[i + 1] - padded_passstart[i]); - if(!padded) ERROR_BREAK(83); /*alloc fail*/ - addPaddingBits(padded, &adam7[passstart[i]], - ((passw[i] * bpp + 7u) / 8u) * 8u, passw[i] * bpp, passh[i]); - error = filter(&(*out)[filter_passstart[i]], padded, - passw[i], passh[i], &info_png->color, settings); - lodepng_free(padded); - } else { - error = filter(&(*out)[filter_passstart[i]], &adam7[padded_passstart[i]], - passw[i], passh[i], &info_png->color, settings); - } - - if(error) break; - } - } - - lodepng_free(adam7); - } - - return error; -} - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS -static unsigned addUnknownChunks(ucvector* out, unsigned char* data, size_t datasize) { - unsigned char* inchunk = data; - while((size_t)(inchunk - data) < datasize) { - CERROR_TRY_RETURN(lodepng_chunk_append(&out->data, &out->size, inchunk)); - out->allocsize = out->size; /*fix the allocsize again*/ - inchunk = lodepng_chunk_next(inchunk, data + datasize); - } - return 0; -} - -static unsigned isGrayICCProfile(const unsigned char* profile, unsigned size) { - /* - It is a gray profile if bytes 16-19 are "GRAY", rgb profile if bytes 16-19 - are "RGB ". We do not perform any full parsing of the ICC profile here, other - than check those 4 bytes to grayscale profile. Other than that, validity of - the profile is not checked. This is needed only because the PNG specification - requires using a non-gray color model if there is an ICC profile with "RGB " - (sadly limiting compression opportunities if the input data is grayscale RGB - data), and requires using a gray color model if it is "GRAY". - */ - if(size < 20) return 0; - return profile[16] == 'G' && profile[17] == 'R' && profile[18] == 'A' && profile[19] == 'Y'; -} - -static unsigned isRGBICCProfile(const unsigned char* profile, unsigned size) { - /* See comment in isGrayICCProfile*/ - if(size < 20) return 0; - return profile[16] == 'R' && profile[17] == 'G' && profile[18] == 'B' && profile[19] == ' '; -} -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -unsigned lodepng_encode(unsigned char** out, size_t* outsize, - const unsigned char* image, unsigned w, unsigned h, - LodePNGState* state) { - unsigned char* data = 0; /*uncompressed version of the IDAT chunk data*/ - size_t datasize = 0; - ucvector outv = ucvector_init(NULL, 0); - LodePNGInfo info; - const LodePNGInfo* info_png = &state->info_png; - LodePNGColorMode auto_color; - - lodepng_info_init(&info); - lodepng_color_mode_init(&auto_color); - - /*provide some proper output values if error will happen*/ - *out = 0; - *outsize = 0; - state->error = 0; - - /*check input values validity*/ - if((info_png->color.colortype == LCT_PALETTE || state->encoder.force_palette) - && (info_png->color.palettesize == 0 || info_png->color.palettesize > 256)) { - /*this error is returned even if auto_convert is enabled and thus encoder could - generate the palette by itself: while allowing this could be possible in theory, - it may complicate the code or edge cases, and always requiring to give a palette - when setting this color type is a simpler contract*/ - state->error = 68; /*invalid palette size, it is only allowed to be 1-256*/ - goto cleanup; - } - if(state->encoder.zlibsettings.btype > 2) { - state->error = 61; /*error: invalid btype*/ - goto cleanup; - } - if(info_png->interlace_method > 1) { - state->error = 71; /*error: invalid interlace mode*/ - goto cleanup; - } - state->error = checkColorValidity(info_png->color.colortype, info_png->color.bitdepth); - if(state->error) goto cleanup; /*error: invalid color type given*/ - state->error = checkColorValidity(state->info_raw.colortype, state->info_raw.bitdepth); - if(state->error) goto cleanup; /*error: invalid color type given*/ - - /* color convert and compute scanline filter types */ - lodepng_info_copy(&info, &state->info_png); - if(state->encoder.auto_convert) { - LodePNGColorStats stats; - unsigned allow_convert = 1; - lodepng_color_stats_init(&stats); -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - if(info_png->iccp_defined && - isGrayICCProfile(info_png->iccp_profile, info_png->iccp_profile_size)) { - /*the PNG specification does not allow to use palette with a GRAY ICC profile, even - if the palette has only gray colors, so disallow it.*/ - stats.allow_palette = 0; - } - if(info_png->iccp_defined && - isRGBICCProfile(info_png->iccp_profile, info_png->iccp_profile_size)) { - /*the PNG specification does not allow to use grayscale color with RGB ICC profile, so disallow gray.*/ - stats.allow_greyscale = 0; - } -#endif /* LODEPNG_COMPILE_ANCILLARY_CHUNKS */ - state->error = lodepng_compute_color_stats(&stats, image, w, h, &state->info_raw); - if(state->error) goto cleanup; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - if(info_png->background_defined) { - /*the background chunk's color must be taken into account as well*/ - unsigned r = 0, g = 0, b = 0; - LodePNGColorMode mode16 = lodepng_color_mode_make(LCT_RGB, 16); - lodepng_convert_rgb(&r, &g, &b, - info_png->background_r, info_png->background_g, info_png->background_b, &mode16, &info_png->color); - state->error = lodepng_color_stats_add(&stats, r, g, b, 65535); - if(state->error) goto cleanup; - } -#endif /* LODEPNG_COMPILE_ANCILLARY_CHUNKS */ - state->error = auto_choose_color(&auto_color, &state->info_raw, &stats); - if(state->error) goto cleanup; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - if(info_png->sbit_defined) { - /*if sbit is defined, due to strict requirements of which sbit values can be present for which color modes, - auto_convert can't be done in many cases. However, do support a few cases here. - TODO: more conversions may be possible, and it may also be possible to get a more appropriate color type out of - auto_choose_color if knowledge about sbit is used beforehand - */ - unsigned sbit_max = LODEPNG_MAX(LODEPNG_MAX(LODEPNG_MAX(info_png->sbit_r, info_png->sbit_g), - info_png->sbit_b), info_png->sbit_a); - unsigned equal = (!info_png->sbit_g || info_png->sbit_g == info_png->sbit_r) - && (!info_png->sbit_b || info_png->sbit_b == info_png->sbit_r) - && (!info_png->sbit_a || info_png->sbit_a == info_png->sbit_r); - allow_convert = 0; - if(info.color.colortype == LCT_PALETTE && - auto_color.colortype == LCT_PALETTE) { - /* input and output are palette, and in this case it may happen that palette data is - expected to be copied from info_raw into the info_png */ - allow_convert = 1; - } - /*going from 8-bit RGB to palette (or 16-bit as long as sbit_max <= 8) is possible - since both are 8-bit RGB for sBIT's purposes*/ - if(info.color.colortype == LCT_RGB && - auto_color.colortype == LCT_PALETTE && sbit_max <= 8) { - allow_convert = 1; - } - /*going from 8-bit RGBA to palette is also ok but only if sbit_a is exactly 8*/ - if(info.color.colortype == LCT_RGBA && auto_color.colortype == LCT_PALETTE && - info_png->sbit_a == 8 && sbit_max <= 8) { - allow_convert = 1; - } - /*going from 16-bit RGB(A) to 8-bit RGB(A) is ok if all sbit values are <= 8*/ - if((info.color.colortype == LCT_RGB || info.color.colortype == LCT_RGBA) && info.color.bitdepth == 16 && - auto_color.colortype == info.color.colortype && auto_color.bitdepth == 8 && - sbit_max <= 8) { - allow_convert = 1; - } - /*going to less channels is ok if all bit values are equal (all possible values in sbit, - as well as the chosen bitdepth of the result). Due to how auto_convert works, - we already know that auto_color.colortype has less than or equal amount of channels than - info.colortype. Palette is not used here. This conversion is not allowed if - info_png->sbit_r < auto_color.bitdepth, because specifically for alpha, non-presence of - an sbit value heavily implies that alpha's bit depth is equal to the PNG bit depth (rather - than the bit depths set in the r, g and b sbit values, by how the PNG specification describes - handling tRNS chunk case with sBIT), so be conservative here about ignoring user input.*/ - if(info.color.colortype != LCT_PALETTE && auto_color.colortype != LCT_PALETTE && - equal && info_png->sbit_r == auto_color.bitdepth) { - allow_convert = 1; - } - } -#endif - if(state->encoder.force_palette) { - if(info.color.colortype != LCT_GREY && info.color.colortype != LCT_GREY_ALPHA && - (auto_color.colortype == LCT_GREY || auto_color.colortype == LCT_GREY_ALPHA)) { - /*user speficially forced a PLTE palette, so cannot convert to grayscale types because - the PNG specification only allows writing a suggested palette in PLTE for truecolor types*/ - allow_convert = 0; - } - } - if(allow_convert) { - lodepng_color_mode_copy(&info.color, &auto_color); -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /*also convert the background chunk*/ - if(info_png->background_defined) { - if(lodepng_convert_rgb(&info.background_r, &info.background_g, &info.background_b, - info_png->background_r, info_png->background_g, info_png->background_b, &info.color, &info_png->color)) { - state->error = 104; - goto cleanup; - } - } -#endif /* LODEPNG_COMPILE_ANCILLARY_CHUNKS */ - } - } -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - if(info_png->iccp_defined) { - unsigned gray_icc = isGrayICCProfile(info_png->iccp_profile, info_png->iccp_profile_size); - unsigned rgb_icc = isRGBICCProfile(info_png->iccp_profile, info_png->iccp_profile_size); - unsigned gray_png = info.color.colortype == LCT_GREY || info.color.colortype == LCT_GREY_ALPHA; - if(!gray_icc && !rgb_icc) { - state->error = 100; /* Disallowed profile color type for PNG */ - goto cleanup; - } - if(gray_icc != gray_png) { - /*Not allowed to use RGB/RGBA/palette with GRAY ICC profile or vice versa, - or in case of auto_convert, it wasn't possible to find appropriate model*/ - state->error = state->encoder.auto_convert ? 102 : 101; - goto cleanup; - } - } -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - if(!lodepng_color_mode_equal(&state->info_raw, &info.color)) { - unsigned char* converted; - size_t size = ((size_t)w * (size_t)h * (size_t)lodepng_get_bpp(&info.color) + 7u) / 8u; - - converted = (unsigned char*)lodepng_malloc(size); - if(!converted && size) state->error = 83; /*alloc fail*/ - if(!state->error) { - state->error = lodepng_convert(converted, image, &info.color, &state->info_raw, w, h); - } - if(!state->error) { - state->error = preProcessScanlines(&data, &datasize, converted, w, h, &info, &state->encoder); - } - lodepng_free(converted); - if(state->error) goto cleanup; - } else { - state->error = preProcessScanlines(&data, &datasize, image, w, h, &info, &state->encoder); - if(state->error) goto cleanup; - } - - /* output all PNG chunks */ { -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - size_t i; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - /*write signature and chunks*/ - state->error = writeSignature(&outv); - if(state->error) goto cleanup; - /*IHDR*/ - state->error = addChunk_IHDR(&outv, w, h, info.color.colortype, info.color.bitdepth, info.interlace_method); - if(state->error) goto cleanup; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /*unknown chunks between IHDR and PLTE*/ - if(info.unknown_chunks_data[0]) { - state->error = addUnknownChunks(&outv, info.unknown_chunks_data[0], info.unknown_chunks_size[0]); - if(state->error) goto cleanup; - } - /*color profile chunks must come before PLTE */ - if(info.iccp_defined) { - state->error = addChunk_iCCP(&outv, &info, &state->encoder.zlibsettings); - if(state->error) goto cleanup; - } - if(info.srgb_defined) { - state->error = addChunk_sRGB(&outv, &info); - if(state->error) goto cleanup; - } - if(info.gama_defined) { - state->error = addChunk_gAMA(&outv, &info); - if(state->error) goto cleanup; - } - if(info.chrm_defined) { - state->error = addChunk_cHRM(&outv, &info); - if(state->error) goto cleanup; - } - if(info_png->sbit_defined) { - state->error = addChunk_sBIT(&outv, &info); - if(state->error) goto cleanup; - } -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - /*PLTE*/ - if(info.color.colortype == LCT_PALETTE) { - state->error = addChunk_PLTE(&outv, &info.color); - if(state->error) goto cleanup; - } - if(state->encoder.force_palette && (info.color.colortype == LCT_RGB || info.color.colortype == LCT_RGBA)) { - /*force_palette means: write suggested palette for truecolor in PLTE chunk*/ - state->error = addChunk_PLTE(&outv, &info.color); - if(state->error) goto cleanup; - } - /*tRNS (this will only add if when necessary) */ - state->error = addChunk_tRNS(&outv, &info.color); - if(state->error) goto cleanup; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /*bKGD (must come between PLTE and the IDAt chunks*/ - if(info.background_defined) { - state->error = addChunk_bKGD(&outv, &info); - if(state->error) goto cleanup; - } - /*pHYs (must come before the IDAT chunks)*/ - if(info.phys_defined) { - state->error = addChunk_pHYs(&outv, &info); - if(state->error) goto cleanup; - } - - /*unknown chunks between PLTE and IDAT*/ - if(info.unknown_chunks_data[1]) { - state->error = addUnknownChunks(&outv, info.unknown_chunks_data[1], info.unknown_chunks_size[1]); - if(state->error) goto cleanup; - } -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - /*IDAT (multiple IDAT chunks must be consecutive)*/ - state->error = addChunk_IDAT(&outv, data, datasize, &state->encoder.zlibsettings); - if(state->error) goto cleanup; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /*tIME*/ - if(info.time_defined) { - state->error = addChunk_tIME(&outv, &info.time); - if(state->error) goto cleanup; - } - /*tEXt and/or zTXt*/ - for(i = 0; i != info.text_num; ++i) { - if(lodepng_strlen(info.text_keys[i]) > 79) { - state->error = 66; /*text chunk too large*/ - goto cleanup; - } - if(lodepng_strlen(info.text_keys[i]) < 1) { - state->error = 67; /*text chunk too small*/ - goto cleanup; - } - if(state->encoder.text_compression) { - state->error = addChunk_zTXt(&outv, info.text_keys[i], info.text_strings[i], &state->encoder.zlibsettings); - if(state->error) goto cleanup; - } else { - state->error = addChunk_tEXt(&outv, info.text_keys[i], info.text_strings[i]); - if(state->error) goto cleanup; - } - } - /*LodePNG version id in text chunk*/ - if(state->encoder.add_id) { - unsigned already_added_id_text = 0; - for(i = 0; i != info.text_num; ++i) { - const char* k = info.text_keys[i]; - /* Could use strcmp, but we're not calling or reimplementing this C library function for this use only */ - if(k[0] == 'L' && k[1] == 'o' && k[2] == 'd' && k[3] == 'e' && - k[4] == 'P' && k[5] == 'N' && k[6] == 'G' && k[7] == '\0') { - already_added_id_text = 1; - break; - } - } - if(already_added_id_text == 0) { - state->error = addChunk_tEXt(&outv, "LodePNG", LODEPNG_VERSION_STRING); /*it's shorter as tEXt than as zTXt chunk*/ - if(state->error) goto cleanup; - } - } - /*iTXt*/ - for(i = 0; i != info.itext_num; ++i) { - if(lodepng_strlen(info.itext_keys[i]) > 79) { - state->error = 66; /*text chunk too large*/ - goto cleanup; - } - if(lodepng_strlen(info.itext_keys[i]) < 1) { - state->error = 67; /*text chunk too small*/ - goto cleanup; - } - state->error = addChunk_iTXt( - &outv, state->encoder.text_compression, - info.itext_keys[i], info.itext_langtags[i], info.itext_transkeys[i], info.itext_strings[i], - &state->encoder.zlibsettings); - if(state->error) goto cleanup; - } - - /*unknown chunks between IDAT and IEND*/ - if(info.unknown_chunks_data[2]) { - state->error = addUnknownChunks(&outv, info.unknown_chunks_data[2], info.unknown_chunks_size[2]); - if(state->error) goto cleanup; - } -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - state->error = addChunk_IEND(&outv); - if(state->error) goto cleanup; - } - -cleanup: - lodepng_info_cleanup(&info); - lodepng_free(data); - lodepng_color_mode_cleanup(&auto_color); - - /*instead of cleaning the vector up, give it to the output*/ - *out = outv.data; - *outsize = outv.size; - - return state->error; -} - -unsigned lodepng_encode_memory(unsigned char** out, size_t* outsize, const unsigned char* image, - unsigned w, unsigned h, LodePNGColorType colortype, unsigned bitdepth) { - unsigned error; - LodePNGState state; - lodepng_state_init(&state); - state.info_raw.colortype = colortype; - state.info_raw.bitdepth = bitdepth; - state.info_png.color.colortype = colortype; - state.info_png.color.bitdepth = bitdepth; - lodepng_encode(out, outsize, image, w, h, &state); - error = state.error; - lodepng_state_cleanup(&state); - return error; -} - -unsigned lodepng_encode32(unsigned char** out, size_t* outsize, const unsigned char* image, unsigned w, unsigned h) { - return lodepng_encode_memory(out, outsize, image, w, h, LCT_RGBA, 8); -} - -unsigned lodepng_encode24(unsigned char** out, size_t* outsize, const unsigned char* image, unsigned w, unsigned h) { - return lodepng_encode_memory(out, outsize, image, w, h, LCT_RGB, 8); -} - -#ifdef LODEPNG_COMPILE_DISK -unsigned lodepng_encode_file(const char* filename, const unsigned char* image, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth) { - unsigned char* buffer; - size_t buffersize; - unsigned error = lodepng_encode_memory(&buffer, &buffersize, image, w, h, colortype, bitdepth); - if(!error) error = lodepng_save_file(buffer, buffersize, filename); - lodepng_free(buffer); - return error; -} - -unsigned lodepng_encode32_file(const char* filename, const unsigned char* image, unsigned w, unsigned h) { - return lodepng_encode_file(filename, image, w, h, LCT_RGBA, 8); -} - -unsigned lodepng_encode24_file(const char* filename, const unsigned char* image, unsigned w, unsigned h) { - return lodepng_encode_file(filename, image, w, h, LCT_RGB, 8); -} -#endif /*LODEPNG_COMPILE_DISK*/ - -void lodepng_encoder_settings_init(LodePNGEncoderSettings* settings) { - lodepng_compress_settings_init(&settings->zlibsettings); - settings->filter_palette_zero = 1; - settings->filter_strategy = LFS_MINSUM; - settings->auto_convert = 1; - settings->force_palette = 0; - settings->predefined_filters = 0; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - settings->add_id = 0; - settings->text_compression = 1; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -} - -#endif /*LODEPNG_COMPILE_ENCODER*/ -#endif /*LODEPNG_COMPILE_PNG*/ - -#ifdef LODEPNG_COMPILE_ERROR_TEXT -/* -This returns the description of a numerical error code in English. This is also -the documentation of all the error codes. -*/ -const char* lodepng_error_text(unsigned code) { - switch(code) { - case 0: return "no error, everything went ok"; - case 1: return "nothing done yet"; /*the Encoder/Decoder has done nothing yet, error checking makes no sense yet*/ - case 10: return "end of input memory reached without huffman end code"; /*while huffman decoding*/ - case 11: return "error in code tree made it jump outside of huffman tree"; /*while huffman decoding*/ - case 13: return "problem while processing dynamic deflate block"; - case 14: return "problem while processing dynamic deflate block"; - case 15: return "problem while processing dynamic deflate block"; - /*this error could happen if there are only 0 or 1 symbols present in the huffman code:*/ - case 16: return "invalid code while processing dynamic deflate block"; - case 17: return "end of out buffer memory reached while inflating"; - case 18: return "invalid distance code while inflating"; - case 19: return "end of out buffer memory reached while inflating"; - case 20: return "invalid deflate block BTYPE encountered while decoding"; - case 21: return "NLEN is not ones complement of LEN in a deflate block"; - - /*end of out buffer memory reached while inflating: - This can happen if the inflated deflate data is longer than the amount of bytes required to fill up - all the pixels of the image, given the color depth and image dimensions. Something that doesn't - happen in a normal, well encoded, PNG image.*/ - case 22: return "end of out buffer memory reached while inflating"; - case 23: return "end of in buffer memory reached while inflating"; - case 24: return "invalid FCHECK in zlib header"; - case 25: return "invalid compression method in zlib header"; - case 26: return "FDICT encountered in zlib header while it's not used for PNG"; - case 27: return "PNG file is smaller than a PNG header"; - /*Checks the magic file header, the first 8 bytes of the PNG file*/ - case 28: return "incorrect PNG signature, it's no PNG or corrupted"; - case 29: return "first chunk is not the header chunk"; - case 30: return "chunk length too large, chunk broken off at end of file"; - case 31: return "illegal PNG color type or bpp"; - case 32: return "illegal PNG compression method"; - case 33: return "illegal PNG filter method"; - case 34: return "illegal PNG interlace method"; - case 35: return "chunk length of a chunk is too large or the chunk too small"; - case 36: return "illegal PNG filter type encountered"; - case 37: return "illegal bit depth for this color type given"; - case 38: return "the palette is too small or too big"; /*0, or more than 256 colors*/ - case 39: return "tRNS chunk before PLTE or has more entries than palette size"; - case 40: return "tRNS chunk has wrong size for grayscale image"; - case 41: return "tRNS chunk has wrong size for RGB image"; - case 42: return "tRNS chunk appeared while it was not allowed for this color type"; - case 43: return "bKGD chunk has wrong size for palette image"; - case 44: return "bKGD chunk has wrong size for grayscale image"; - case 45: return "bKGD chunk has wrong size for RGB image"; - case 48: return "empty input buffer given to decoder. Maybe caused by non-existing file?"; - case 49: return "jumped past memory while generating dynamic huffman tree"; - case 50: return "jumped past memory while generating dynamic huffman tree"; - case 51: return "jumped past memory while inflating huffman block"; - case 52: return "jumped past memory while inflating"; - case 53: return "size of zlib data too small"; - case 54: return "repeat symbol in tree while there was no value symbol yet"; - /*jumped past tree while generating huffman tree, this could be when the - tree will have more leaves than symbols after generating it out of the - given lengths. They call this an oversubscribed dynamic bit lengths tree in zlib.*/ - case 55: return "jumped past tree while generating huffman tree"; - case 56: return "given output image colortype or bitdepth not supported for color conversion"; - case 57: return "invalid CRC encountered (checking CRC can be disabled)"; - case 58: return "invalid ADLER32 encountered (checking ADLER32 can be disabled)"; - case 59: return "requested color conversion not supported"; - case 60: return "invalid window size given in the settings of the encoder (must be 0-32768)"; - case 61: return "invalid BTYPE given in the settings of the encoder (only 0, 1 and 2 are allowed)"; - /*LodePNG leaves the choice of RGB to grayscale conversion formula to the user.*/ - case 62: return "conversion from color to grayscale not supported"; - /*(2^31-1)*/ - case 63: return "length of a chunk too long, max allowed for PNG is 2147483647 bytes per chunk"; - /*this would result in the inability of a deflated block to ever contain an end code. It must be at least 1.*/ - case 64: return "the length of the END symbol 256 in the Huffman tree is 0"; - case 66: return "the length of a text chunk keyword given to the encoder is longer than the maximum of 79 bytes"; - case 67: return "the length of a text chunk keyword given to the encoder is smaller than the minimum of 1 byte"; - case 68: return "tried to encode a PLTE chunk with a palette that has less than 1 or more than 256 colors"; - case 69: return "unknown chunk type with 'critical' flag encountered by the decoder"; - case 71: return "invalid interlace mode given to encoder (must be 0 or 1)"; - case 72: return "while decoding, invalid compression method encountering in zTXt or iTXt chunk (it must be 0)"; - case 73: return "invalid tIME chunk size"; - case 74: return "invalid pHYs chunk size"; - /*length could be wrong, or data chopped off*/ - case 75: return "no null termination char found while decoding text chunk"; - case 76: return "iTXt chunk too short to contain required bytes"; - case 77: return "integer overflow in buffer size"; - case 78: return "failed to open file for reading"; /*file doesn't exist or couldn't be opened for reading*/ - case 79: return "failed to open file for writing"; - case 80: return "tried creating a tree of 0 symbols"; - case 81: return "lazy matching at pos 0 is impossible"; - case 82: return "color conversion to palette requested while a color isn't in palette, or index out of bounds"; - case 83: return "memory allocation failed"; - case 84: return "given image too small to contain all pixels to be encoded"; - case 86: return "impossible offset in lz77 encoding (internal bug)"; - case 87: return "must provide custom zlib function pointer if LODEPNG_COMPILE_ZLIB is not defined"; - case 88: return "invalid filter strategy given for LodePNGEncoderSettings.filter_strategy"; - case 89: return "text chunk keyword too short or long: must have size 1-79"; - /*the windowsize in the LodePNGCompressSettings. Requiring POT(==> & instead of %) makes encoding 12% faster.*/ - case 90: return "windowsize must be a power of two"; - case 91: return "invalid decompressed idat size"; - case 92: return "integer overflow due to too many pixels"; - case 93: return "zero width or height is invalid"; - case 94: return "header chunk must have a size of 13 bytes"; - case 95: return "integer overflow with combined idat chunk size"; - case 96: return "invalid gAMA chunk size"; - case 97: return "invalid cHRM chunk size"; - case 98: return "invalid sRGB chunk size"; - case 99: return "invalid sRGB rendering intent"; - case 100: return "invalid ICC profile color type, the PNG specification only allows RGB or GRAY"; - case 101: return "PNG specification does not allow RGB ICC profile on gray color types and vice versa"; - case 102: return "not allowed to set grayscale ICC profile with colored pixels by PNG specification"; - case 103: return "invalid palette index in bKGD chunk. Maybe it came before PLTE chunk?"; - case 104: return "invalid bKGD color while encoding (e.g. palette index out of range)"; - case 105: return "integer overflow of bitsize"; - case 106: return "PNG file must have PLTE chunk if color type is palette"; - case 107: return "color convert from palette mode requested without setting the palette data in it"; - case 108: return "tried to add more than 256 values to a palette"; - /*this limit can be configured in LodePNGDecompressSettings*/ - case 109: return "tried to decompress zlib or deflate data larger than desired max_output_size"; - case 110: return "custom zlib or inflate decompression failed"; - case 111: return "custom zlib or deflate compression failed"; - /*max text size limit can be configured in LodePNGDecoderSettings. This error prevents - unreasonable memory consumption when decoding due to impossibly large text sizes.*/ - case 112: return "compressed text unreasonably large"; - /*max ICC size limit can be configured in LodePNGDecoderSettings. This error prevents - unreasonable memory consumption when decoding due to impossibly large ICC profile*/ - case 113: return "ICC profile unreasonably large"; - case 114: return "sBIT chunk has wrong size for the color type of the image"; - case 115: return "sBIT value out of range"; - } - return "unknown error code"; -} -#endif /*LODEPNG_COMPILE_ERROR_TEXT*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* // C++ Wrapper // */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_CPP -namespace lodepng { - -#ifdef LODEPNG_COMPILE_DISK -unsigned load_file(std::vector<unsigned char>& buffer, const std::string& filename) { - long size = lodepng_filesize(filename.c_str()); - if(size < 0) return 78; - buffer.resize((size_t)size); - return size == 0 ? 0 : lodepng_buffer_file(&buffer[0], (size_t)size, filename.c_str()); -} - -/*write given buffer to the file, overwriting the file, it doesn't append to it.*/ -unsigned save_file(const std::vector<unsigned char>& buffer, const std::string& filename) { - return lodepng_save_file(buffer.empty() ? 0 : &buffer[0], buffer.size(), filename.c_str()); -} -#endif /* LODEPNG_COMPILE_DISK */ - -#ifdef LODEPNG_COMPILE_ZLIB -#ifdef LODEPNG_COMPILE_DECODER -unsigned decompress(std::vector<unsigned char>& out, const unsigned char* in, size_t insize, - const LodePNGDecompressSettings& settings) { - unsigned char* buffer = 0; - size_t buffersize = 0; - unsigned error = zlib_decompress(&buffer, &buffersize, 0, in, insize, &settings); - if(buffer) { - out.insert(out.end(), buffer, &buffer[buffersize]); - lodepng_free(buffer); - } - return error; -} - -unsigned decompress(std::vector<unsigned char>& out, const std::vector<unsigned char>& in, - const LodePNGDecompressSettings& settings) { - return decompress(out, in.empty() ? 0 : &in[0], in.size(), settings); -} -#endif /* LODEPNG_COMPILE_DECODER */ - -#ifdef LODEPNG_COMPILE_ENCODER -unsigned compress(std::vector<unsigned char>& out, const unsigned char* in, size_t insize, - const LodePNGCompressSettings& settings) { - unsigned char* buffer = 0; - size_t buffersize = 0; - unsigned error = zlib_compress(&buffer, &buffersize, in, insize, &settings); - if(buffer) { - out.insert(out.end(), buffer, &buffer[buffersize]); - lodepng_free(buffer); - } - return error; -} - -unsigned compress(std::vector<unsigned char>& out, const std::vector<unsigned char>& in, - const LodePNGCompressSettings& settings) { - return compress(out, in.empty() ? 0 : &in[0], in.size(), settings); -} -#endif /* LODEPNG_COMPILE_ENCODER */ -#endif /* LODEPNG_COMPILE_ZLIB */ - - -#ifdef LODEPNG_COMPILE_PNG - -State::State() { - lodepng_state_init(this); -} - -State::State(const State& other) { - lodepng_state_init(this); - lodepng_state_copy(this, &other); -} - -State::~State() { - lodepng_state_cleanup(this); -} - -State& State::operator=(const State& other) { - lodepng_state_copy(this, &other); - return *this; -} - -#ifdef LODEPNG_COMPILE_DECODER - -unsigned decode(std::vector<unsigned char>& out, unsigned& w, unsigned& h, const unsigned char* in, - size_t insize, LodePNGColorType colortype, unsigned bitdepth) { - unsigned char* buffer = 0; - unsigned error = lodepng_decode_memory(&buffer, &w, &h, in, insize, colortype, bitdepth); - if(buffer && !error) { - State state; - state.info_raw.colortype = colortype; - state.info_raw.bitdepth = bitdepth; - size_t buffersize = lodepng_get_raw_size(w, h, &state.info_raw); - out.insert(out.end(), buffer, &buffer[buffersize]); - } - lodepng_free(buffer); - return error; -} - -unsigned decode(std::vector<unsigned char>& out, unsigned& w, unsigned& h, - const std::vector<unsigned char>& in, LodePNGColorType colortype, unsigned bitdepth) { - return decode(out, w, h, in.empty() ? 0 : &in[0], (unsigned)in.size(), colortype, bitdepth); -} - -unsigned decode(std::vector<unsigned char>& out, unsigned& w, unsigned& h, - State& state, - const unsigned char* in, size_t insize) { - unsigned char* buffer = NULL; - unsigned error = lodepng_decode(&buffer, &w, &h, &state, in, insize); - if(buffer && !error) { - size_t buffersize = lodepng_get_raw_size(w, h, &state.info_raw); - out.insert(out.end(), buffer, &buffer[buffersize]); - } - lodepng_free(buffer); - return error; -} - -unsigned decode(std::vector<unsigned char>& out, unsigned& w, unsigned& h, - State& state, - const std::vector<unsigned char>& in) { - return decode(out, w, h, state, in.empty() ? 0 : &in[0], in.size()); -} - -#ifdef LODEPNG_COMPILE_DISK -unsigned decode(std::vector<unsigned char>& out, unsigned& w, unsigned& h, const std::string& filename, - LodePNGColorType colortype, unsigned bitdepth) { - std::vector<unsigned char> buffer; - /* safe output values in case error happens */ - w = h = 0; - unsigned error = load_file(buffer, filename); - if(error) return error; - return decode(out, w, h, buffer, colortype, bitdepth); -} -#endif /* LODEPNG_COMPILE_DECODER */ -#endif /* LODEPNG_COMPILE_DISK */ - -#ifdef LODEPNG_COMPILE_ENCODER -unsigned encode(std::vector<unsigned char>& out, const unsigned char* in, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth) { - unsigned char* buffer; - size_t buffersize; - unsigned error = lodepng_encode_memory(&buffer, &buffersize, in, w, h, colortype, bitdepth); - if(buffer) { - out.insert(out.end(), buffer, &buffer[buffersize]); - lodepng_free(buffer); - } - return error; -} - -unsigned encode(std::vector<unsigned char>& out, - const std::vector<unsigned char>& in, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth) { - if(lodepng_get_raw_size_lct(w, h, colortype, bitdepth) > in.size()) return 84; - return encode(out, in.empty() ? 0 : &in[0], w, h, colortype, bitdepth); -} - -unsigned encode(std::vector<unsigned char>& out, - const unsigned char* in, unsigned w, unsigned h, - State& state) { - unsigned char* buffer; - size_t buffersize; - unsigned error = lodepng_encode(&buffer, &buffersize, in, w, h, &state); - if(buffer) { - out.insert(out.end(), buffer, &buffer[buffersize]); - lodepng_free(buffer); - } - return error; -} - -unsigned encode(std::vector<unsigned char>& out, - const std::vector<unsigned char>& in, unsigned w, unsigned h, - State& state) { - if(lodepng_get_raw_size(w, h, &state.info_raw) > in.size()) return 84; - return encode(out, in.empty() ? 0 : &in[0], w, h, state); -} - -#ifdef LODEPNG_COMPILE_DISK -unsigned encode(const std::string& filename, - const unsigned char* in, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth) { - std::vector<unsigned char> buffer; - unsigned error = encode(buffer, in, w, h, colortype, bitdepth); - if(!error) error = save_file(buffer, filename); - return error; -} - -unsigned encode(const std::string& filename, - const std::vector<unsigned char>& in, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth) { - if(lodepng_get_raw_size_lct(w, h, colortype, bitdepth) > in.size()) return 84; - return encode(filename, in.empty() ? 0 : &in[0], w, h, colortype, bitdepth); -} -#endif /* LODEPNG_COMPILE_DISK */ -#endif /* LODEPNG_COMPILE_ENCODER */ -#endif /* LODEPNG_COMPILE_PNG */ -} /* namespace lodepng */ -#endif /*LODEPNG_COMPILE_CPP*/
--- a/semicongine/contrib/lodepng.h Sun Jul 28 23:18:05 2024 +0700 +++ /dev/null Thu Jan 01 00:00:00 1970 +0000 @@ -1,2089 +0,0 @@ -/* -LodePNG version 20230410 - -Copyright (c) 2005-2023 Lode Vandevenne - -This software is provided 'as-is', without any express or implied -warranty. In no event will the authors be held liable for any damages -arising from the use of this software. - -Permission is granted to anyone to use this software for any purpose, -including commercial applications, and to alter it and redistribute it -freely, subject to the following restrictions: - - 1. The origin of this software must not be misrepresented; you must not - claim that you wrote the original software. If you use this software - in a product, an acknowledgment in the product documentation would be - appreciated but is not required. - - 2. Altered source versions must be plainly marked as such, and must not be - misrepresented as being the original software. - - 3. This notice may not be removed or altered from any source - distribution. -*/ - -#ifndef LODEPNG_H -#define LODEPNG_H - -#include <string.h> /*for size_t*/ - -extern const char* LODEPNG_VERSION_STRING; - -/* -The following #defines are used to create code sections. They can be disabled -to disable code sections, which can give faster compile time and smaller binary. -The "NO_COMPILE" defines are designed to be used to pass as defines to the -compiler command to disable them without modifying this header, e.g. --DLODEPNG_NO_COMPILE_ZLIB for gcc or clang. -*/ -/*deflate & zlib. If disabled, you must specify alternative zlib functions in -the custom_zlib field of the compress and decompress settings*/ -#ifndef LODEPNG_NO_COMPILE_ZLIB -/*pass -DLODEPNG_NO_COMPILE_ZLIB to the compiler to disable this, or comment out LODEPNG_COMPILE_ZLIB below*/ -#define LODEPNG_COMPILE_ZLIB -#endif - -/*png encoder and png decoder*/ -#ifndef LODEPNG_NO_COMPILE_PNG -/*pass -DLODEPNG_NO_COMPILE_PNG to the compiler to disable this, or comment out LODEPNG_COMPILE_PNG below*/ -#define LODEPNG_COMPILE_PNG -#endif - -/*deflate&zlib decoder and png decoder*/ -#ifndef LODEPNG_NO_COMPILE_DECODER -/*pass -DLODEPNG_NO_COMPILE_DECODER to the compiler to disable this, or comment out LODEPNG_COMPILE_DECODER below*/ -#define LODEPNG_COMPILE_DECODER -#endif - -/*deflate&zlib encoder and png encoder*/ -#ifndef LODEPNG_NO_COMPILE_ENCODER -/*pass -DLODEPNG_NO_COMPILE_ENCODER to the compiler to disable this, or comment out LODEPNG_COMPILE_ENCODER below*/ -#define LODEPNG_COMPILE_ENCODER -#endif - -/*the optional built in harddisk file loading and saving functions*/ -#ifndef LODEPNG_NO_COMPILE_DISK -/*pass -DLODEPNG_NO_COMPILE_DISK to the compiler to disable this, or comment out LODEPNG_COMPILE_DISK below*/ -#define LODEPNG_COMPILE_DISK -#endif - -/*support for chunks other than IHDR, IDAT, PLTE, tRNS, IEND: ancillary and unknown chunks*/ -#ifndef LODEPNG_NO_COMPILE_ANCILLARY_CHUNKS -/*pass -DLODEPNG_NO_COMPILE_ANCILLARY_CHUNKS to the compiler to disable this, -or comment out LODEPNG_COMPILE_ANCILLARY_CHUNKS below*/ -#define LODEPNG_COMPILE_ANCILLARY_CHUNKS -#endif - -/*ability to convert error numerical codes to English text string*/ -#ifndef LODEPNG_NO_COMPILE_ERROR_TEXT -/*pass -DLODEPNG_NO_COMPILE_ERROR_TEXT to the compiler to disable this, -or comment out LODEPNG_COMPILE_ERROR_TEXT below*/ -#define LODEPNG_COMPILE_ERROR_TEXT -#endif - -/*Compile the default allocators (C's free, malloc and realloc). If you disable this, -you can define the functions lodepng_free, lodepng_malloc and lodepng_realloc in your -source files with custom allocators.*/ -#ifndef LODEPNG_NO_COMPILE_ALLOCATORS -/*pass -DLODEPNG_NO_COMPILE_ALLOCATORS to the compiler to disable the built-in ones, -or comment out LODEPNG_COMPILE_ALLOCATORS below*/ -#define LODEPNG_COMPILE_ALLOCATORS -#endif - -/*Disable built-in CRC function, in that case a custom implementation of -lodepng_crc32 must be defined externally so that it can be linked in. -The default built-in CRC code comes with 8KB of lookup tables, so for memory constrained environment you may want it -disabled and provide a much smaller implementation externally as said above. You can find such an example implementation -in a comment in the lodepng.c(pp) file in the 'else' case of the searchable LODEPNG_COMPILE_CRC section.*/ -#ifndef LODEPNG_NO_COMPILE_CRC -/*pass -DLODEPNG_NO_COMPILE_CRC to the compiler to disable the built-in one, -or comment out LODEPNG_COMPILE_CRC below*/ -#define LODEPNG_COMPILE_CRC -#endif - -/*compile the C++ version (you can disable the C++ wrapper here even when compiling for C++)*/ -#ifdef __cplusplus -#ifndef LODEPNG_NO_COMPILE_CPP -/*pass -DLODEPNG_NO_COMPILE_CPP to the compiler to disable C++ (not needed if a C-only compiler), -or comment out LODEPNG_COMPILE_CPP below*/ -#define LODEPNG_COMPILE_CPP -#endif -#endif - -#ifdef LODEPNG_COMPILE_CPP -#include <vector> -#include <string> -#endif /*LODEPNG_COMPILE_CPP*/ - -#ifdef LODEPNG_COMPILE_PNG -/*The PNG color types (also used for raw image).*/ -typedef enum LodePNGColorType { - LCT_GREY = 0, /*grayscale: 1,2,4,8,16 bit*/ - LCT_RGB = 2, /*RGB: 8,16 bit*/ - LCT_PALETTE = 3, /*palette: 1,2,4,8 bit*/ - LCT_GREY_ALPHA = 4, /*grayscale with alpha: 8,16 bit*/ - LCT_RGBA = 6, /*RGB with alpha: 8,16 bit*/ - /*LCT_MAX_OCTET_VALUE lets the compiler allow this enum to represent any invalid - byte value from 0 to 255 that could be present in an invalid PNG file header. Do - not use, compare with or set the name LCT_MAX_OCTET_VALUE, instead either use - the valid color type names above, or numeric values like 1 or 7 when checking for - particular disallowed color type byte values, or cast to integer to print it.*/ - LCT_MAX_OCTET_VALUE = 255 -} LodePNGColorType; - -#ifdef LODEPNG_COMPILE_DECODER -/* -Converts PNG data in memory to raw pixel data. -out: Output parameter. Pointer to buffer that will contain the raw pixel data. - After decoding, its size is w * h * (bytes per pixel) bytes larger than - initially. Bytes per pixel depends on colortype and bitdepth. - Must be freed after usage with free(*out). - Note: for 16-bit per channel colors, uses big endian format like PNG does. -w: Output parameter. Pointer to width of pixel data. -h: Output parameter. Pointer to height of pixel data. -in: Memory buffer with the PNG file. -insize: size of the in buffer. -colortype: the desired color type for the raw output image. See explanation on PNG color types. -bitdepth: the desired bit depth for the raw output image. See explanation on PNG color types. -Return value: LodePNG error code (0 means no error). -*/ -unsigned lodepng_decode_memory(unsigned char** out, unsigned* w, unsigned* h, - const unsigned char* in, size_t insize, - LodePNGColorType colortype, unsigned bitdepth); - -/*Same as lodepng_decode_memory, but always decodes to 32-bit RGBA raw image*/ -unsigned lodepng_decode32(unsigned char** out, unsigned* w, unsigned* h, - const unsigned char* in, size_t insize); - -/*Same as lodepng_decode_memory, but always decodes to 24-bit RGB raw image*/ -unsigned lodepng_decode24(unsigned char** out, unsigned* w, unsigned* h, - const unsigned char* in, size_t insize); - -#ifdef LODEPNG_COMPILE_DISK -/* -Load PNG from disk, from file with given name. -Same as the other decode functions, but instead takes a filename as input. - -NOTE: Wide-character filenames are not supported, you can use an external method -to handle such files and decode in-memory.*/ -unsigned lodepng_decode_file(unsigned char** out, unsigned* w, unsigned* h, - const char* filename, - LodePNGColorType colortype, unsigned bitdepth); - -/*Same as lodepng_decode_file, but always decodes to 32-bit RGBA raw image. - -NOTE: Wide-character filenames are not supported, you can use an external method -to handle such files and decode in-memory.*/ -unsigned lodepng_decode32_file(unsigned char** out, unsigned* w, unsigned* h, - const char* filename); - -/*Same as lodepng_decode_file, but always decodes to 24-bit RGB raw image. - -NOTE: Wide-character filenames are not supported, you can use an external method -to handle such files and decode in-memory.*/ -unsigned lodepng_decode24_file(unsigned char** out, unsigned* w, unsigned* h, - const char* filename); -#endif /*LODEPNG_COMPILE_DISK*/ -#endif /*LODEPNG_COMPILE_DECODER*/ - - -#ifdef LODEPNG_COMPILE_ENCODER -/* -Converts raw pixel data into a PNG image in memory. The colortype and bitdepth - of the output PNG image cannot be chosen, they are automatically determined - by the colortype, bitdepth and content of the input pixel data. - Note: for 16-bit per channel colors, needs big endian format like PNG does. -out: Output parameter. Pointer to buffer that will contain the PNG image data. - Must be freed after usage with free(*out). -outsize: Output parameter. Pointer to the size in bytes of the out buffer. -image: The raw pixel data to encode. The size of this buffer should be - w * h * (bytes per pixel), bytes per pixel depends on colortype and bitdepth. -w: width of the raw pixel data in pixels. -h: height of the raw pixel data in pixels. -colortype: the color type of the raw input image. See explanation on PNG color types. -bitdepth: the bit depth of the raw input image. See explanation on PNG color types. -Return value: LodePNG error code (0 means no error). -*/ -unsigned lodepng_encode_memory(unsigned char** out, size_t* outsize, - const unsigned char* image, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth); - -/*Same as lodepng_encode_memory, but always encodes from 32-bit RGBA raw image.*/ -unsigned lodepng_encode32(unsigned char** out, size_t* outsize, - const unsigned char* image, unsigned w, unsigned h); - -/*Same as lodepng_encode_memory, but always encodes from 24-bit RGB raw image.*/ -unsigned lodepng_encode24(unsigned char** out, size_t* outsize, - const unsigned char* image, unsigned w, unsigned h); - -#ifdef LODEPNG_COMPILE_DISK -/* -Converts raw pixel data into a PNG file on disk. -Same as the other encode functions, but instead takes a filename as output. - -NOTE: This overwrites existing files without warning! - -NOTE: Wide-character filenames are not supported, you can use an external method -to handle such files and encode in-memory.*/ -unsigned lodepng_encode_file(const char* filename, - const unsigned char* image, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth); - -/*Same as lodepng_encode_file, but always encodes from 32-bit RGBA raw image. - -NOTE: Wide-character filenames are not supported, you can use an external method -to handle such files and encode in-memory.*/ -unsigned lodepng_encode32_file(const char* filename, - const unsigned char* image, unsigned w, unsigned h); - -/*Same as lodepng_encode_file, but always encodes from 24-bit RGB raw image. - -NOTE: Wide-character filenames are not supported, you can use an external method -to handle such files and encode in-memory.*/ -unsigned lodepng_encode24_file(const char* filename, - const unsigned char* image, unsigned w, unsigned h); -#endif /*LODEPNG_COMPILE_DISK*/ -#endif /*LODEPNG_COMPILE_ENCODER*/ - - -#ifdef LODEPNG_COMPILE_CPP -namespace lodepng { -#ifdef LODEPNG_COMPILE_DECODER -/*Same as lodepng_decode_memory, but decodes to an std::vector. The colortype -is the format to output the pixels to. Default is RGBA 8-bit per channel.*/ -unsigned decode(std::vector<unsigned char>& out, unsigned& w, unsigned& h, - const unsigned char* in, size_t insize, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); -unsigned decode(std::vector<unsigned char>& out, unsigned& w, unsigned& h, - const std::vector<unsigned char>& in, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); -#ifdef LODEPNG_COMPILE_DISK -/* -Converts PNG file from disk to raw pixel data in memory. -Same as the other decode functions, but instead takes a filename as input. - -NOTE: Wide-character filenames are not supported, you can use an external method -to handle such files and decode in-memory. -*/ -unsigned decode(std::vector<unsigned char>& out, unsigned& w, unsigned& h, - const std::string& filename, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); -#endif /* LODEPNG_COMPILE_DISK */ -#endif /* LODEPNG_COMPILE_DECODER */ - -#ifdef LODEPNG_COMPILE_ENCODER -/*Same as lodepng_encode_memory, but encodes to an std::vector. colortype -is that of the raw input data. The output PNG color type will be auto chosen.*/ -unsigned encode(std::vector<unsigned char>& out, - const unsigned char* in, unsigned w, unsigned h, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); -unsigned encode(std::vector<unsigned char>& out, - const std::vector<unsigned char>& in, unsigned w, unsigned h, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); -#ifdef LODEPNG_COMPILE_DISK -/* -Converts 32-bit RGBA raw pixel data into a PNG file on disk. -Same as the other encode functions, but instead takes a filename as output. - -NOTE: This overwrites existing files without warning! - -NOTE: Wide-character filenames are not supported, you can use an external method -to handle such files and decode in-memory. -*/ -unsigned encode(const std::string& filename, - const unsigned char* in, unsigned w, unsigned h, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); -unsigned encode(const std::string& filename, - const std::vector<unsigned char>& in, unsigned w, unsigned h, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); -#endif /* LODEPNG_COMPILE_DISK */ -#endif /* LODEPNG_COMPILE_ENCODER */ -} /* namespace lodepng */ -#endif /*LODEPNG_COMPILE_CPP*/ -#endif /*LODEPNG_COMPILE_PNG*/ - -#ifdef LODEPNG_COMPILE_ERROR_TEXT -/*Returns an English description of the numerical error code.*/ -const char* lodepng_error_text(unsigned code); -#endif /*LODEPNG_COMPILE_ERROR_TEXT*/ - -#ifdef LODEPNG_COMPILE_DECODER -/*Settings for zlib decompression*/ -typedef struct LodePNGDecompressSettings LodePNGDecompressSettings; -struct LodePNGDecompressSettings { - /* Check LodePNGDecoderSettings for more ignorable errors such as ignore_crc */ - unsigned ignore_adler32; /*if 1, continue and don't give an error message if the Adler32 checksum is corrupted*/ - unsigned ignore_nlen; /*ignore complement of len checksum in uncompressed blocks*/ - - /*Maximum decompressed size, beyond this the decoder may (and is encouraged to) stop decoding, - return an error, output a data size > max_output_size and all the data up to that point. This is - not hard limit nor a guarantee, but can prevent excessive memory usage. This setting is - ignored by the PNG decoder, but is used by the deflate/zlib decoder and can be used by custom ones. - Set to 0 to impose no limit (the default).*/ - size_t max_output_size; - - /*use custom zlib decoder instead of built in one (default: null). - Should return 0 if success, any non-0 if error (numeric value not exposed).*/ - unsigned (*custom_zlib)(unsigned char**, size_t*, - const unsigned char*, size_t, - const LodePNGDecompressSettings*); - /*use custom deflate decoder instead of built in one (default: null) - if custom_zlib is not null, custom_inflate is ignored (the zlib format uses deflate). - Should return 0 if success, any non-0 if error (numeric value not exposed).*/ - unsigned (*custom_inflate)(unsigned char**, size_t*, - const unsigned char*, size_t, - const LodePNGDecompressSettings*); - - const void* custom_context; /*optional custom settings for custom functions*/ -}; - -extern const LodePNGDecompressSettings lodepng_default_decompress_settings; -void lodepng_decompress_settings_init(LodePNGDecompressSettings* settings); -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_ENCODER -/* -Settings for zlib compression. Tweaking these settings tweaks the balance -between speed and compression ratio. -*/ -typedef struct LodePNGCompressSettings LodePNGCompressSettings; -struct LodePNGCompressSettings /*deflate = compress*/ { - /*LZ77 related settings*/ - unsigned btype; /*the block type for LZ (0, 1, 2 or 3, see zlib standard). Should be 2 for proper compression.*/ - unsigned use_lz77; /*whether or not to use LZ77. Should be 1 for proper compression.*/ - unsigned windowsize; /*must be a power of two <= 32768. higher compresses more but is slower. Default value: 2048.*/ - unsigned minmatch; /*minimum lz77 length. 3 is normally best, 6 can be better for some PNGs. Default: 0*/ - unsigned nicematch; /*stop searching if >= this length found. Set to 258 for best compression. Default: 128*/ - unsigned lazymatching; /*use lazy matching: better compression but a bit slower. Default: true*/ - - /*use custom zlib encoder instead of built in one (default: null)*/ - unsigned (*custom_zlib)(unsigned char**, size_t*, - const unsigned char*, size_t, - const LodePNGCompressSettings*); - /*use custom deflate encoder instead of built in one (default: null) - if custom_zlib is used, custom_deflate is ignored since only the built in - zlib function will call custom_deflate*/ - unsigned (*custom_deflate)(unsigned char**, size_t*, - const unsigned char*, size_t, - const LodePNGCompressSettings*); - - const void* custom_context; /*optional custom settings for custom functions*/ -}; - -extern const LodePNGCompressSettings lodepng_default_compress_settings; -void lodepng_compress_settings_init(LodePNGCompressSettings* settings); -#endif /*LODEPNG_COMPILE_ENCODER*/ - -#ifdef LODEPNG_COMPILE_PNG -/* -Color mode of an image. Contains all information required to decode the pixel -bits to RGBA colors. This information is the same as used in the PNG file -format, and is used both for PNG and raw image data in LodePNG. -*/ -typedef struct LodePNGColorMode { - /*header (IHDR)*/ - LodePNGColorType colortype; /*color type, see PNG standard or documentation further in this header file*/ - unsigned bitdepth; /*bits per sample, see PNG standard or documentation further in this header file*/ - - /* - palette (PLTE and tRNS) - - Dynamically allocated with the colors of the palette, including alpha. - This field may not be allocated directly, use lodepng_color_mode_init first, - then lodepng_palette_add per color to correctly initialize it (to ensure size - of exactly 1024 bytes). - - The alpha channels must be set as well, set them to 255 for opaque images. - - When decoding, with the default settings you can ignore this palette, since - LodePNG already fills the palette colors in the pixels of the raw RGBA output, - but when decoding to the original PNG color mode it is needed to reconstruct - the colors. - - The palette is only supported for color type 3. - */ - unsigned char* palette; /*palette in RGBARGBA... order. Must be either 0, or when allocated must have 1024 bytes*/ - size_t palettesize; /*palette size in number of colors (amount of used bytes is 4 * palettesize)*/ - - /* - transparent color key (tRNS) - - This color uses the same bit depth as the bitdepth value in this struct, which can be 1-bit to 16-bit. - For grayscale PNGs, r, g and b will all 3 be set to the same. - - When decoding, by default you can ignore this information, since LodePNG sets - pixels with this key to transparent already in the raw RGBA output. - - The color key is only supported for color types 0 and 2. - */ - unsigned key_defined; /*is a transparent color key given? 0 = false, 1 = true*/ - unsigned key_r; /*red/grayscale component of color key*/ - unsigned key_g; /*green component of color key*/ - unsigned key_b; /*blue component of color key*/ -} LodePNGColorMode; - -/*init, cleanup and copy functions to use with this struct*/ -void lodepng_color_mode_init(LodePNGColorMode* info); -void lodepng_color_mode_cleanup(LodePNGColorMode* info); -/*return value is error code (0 means no error)*/ -unsigned lodepng_color_mode_copy(LodePNGColorMode* dest, const LodePNGColorMode* source); -/* Makes a temporary LodePNGColorMode that does not need cleanup (no palette) */ -LodePNGColorMode lodepng_color_mode_make(LodePNGColorType colortype, unsigned bitdepth); - -void lodepng_palette_clear(LodePNGColorMode* info); -/*add 1 color to the palette*/ -unsigned lodepng_palette_add(LodePNGColorMode* info, - unsigned char r, unsigned char g, unsigned char b, unsigned char a); - -/*get the total amount of bits per pixel, based on colortype and bitdepth in the struct*/ -unsigned lodepng_get_bpp(const LodePNGColorMode* info); -/*get the amount of color channels used, based on colortype in the struct. -If a palette is used, it counts as 1 channel.*/ -unsigned lodepng_get_channels(const LodePNGColorMode* info); -/*is it a grayscale type? (only colortype 0 or 4)*/ -unsigned lodepng_is_greyscale_type(const LodePNGColorMode* info); -/*has it got an alpha channel? (only colortype 2 or 6)*/ -unsigned lodepng_is_alpha_type(const LodePNGColorMode* info); -/*has it got a palette? (only colortype 3)*/ -unsigned lodepng_is_palette_type(const LodePNGColorMode* info); -/*only returns true if there is a palette and there is a value in the palette with alpha < 255. -Loops through the palette to check this.*/ -unsigned lodepng_has_palette_alpha(const LodePNGColorMode* info); -/* -Check if the given color info indicates the possibility of having non-opaque pixels in the PNG image. -Returns true if the image can have translucent or invisible pixels (it still be opaque if it doesn't use such pixels). -Returns false if the image can only have opaque pixels. -In detail, it returns true only if it's a color type with alpha, or has a palette with non-opaque values, -or if "key_defined" is true. -*/ -unsigned lodepng_can_have_alpha(const LodePNGColorMode* info); -/*Returns the byte size of a raw image buffer with given width, height and color mode*/ -size_t lodepng_get_raw_size(unsigned w, unsigned h, const LodePNGColorMode* color); - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS -/*The information of a Time chunk in PNG.*/ -typedef struct LodePNGTime { - unsigned year; /*2 bytes used (0-65535)*/ - unsigned month; /*1-12*/ - unsigned day; /*1-31*/ - unsigned hour; /*0-23*/ - unsigned minute; /*0-59*/ - unsigned second; /*0-60 (to allow for leap seconds)*/ -} LodePNGTime; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -/*Information about the PNG image, except pixels, width and height.*/ -typedef struct LodePNGInfo { - /*header (IHDR), palette (PLTE) and transparency (tRNS) chunks*/ - unsigned compression_method;/*compression method of the original file. Always 0.*/ - unsigned filter_method; /*filter method of the original file*/ - unsigned interlace_method; /*interlace method of the original file: 0=none, 1=Adam7*/ - LodePNGColorMode color; /*color type and bits, palette and transparency of the PNG file*/ - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /* - Suggested background color chunk (bKGD) - - This uses the same color mode and bit depth as the PNG (except no alpha channel), - with values truncated to the bit depth in the unsigned integer. - - For grayscale and palette PNGs, the value is stored in background_r. The values - in background_g and background_b are then unused. The decoder will set them - equal to background_r, the encoder ignores them in this case. - - When decoding, you may get these in a different color mode than the one you requested - for the raw pixels: the colortype and bitdepth defined by info_png.color, that is the - ones defined in the header of the PNG image, are used. - - When encoding with auto_convert, you must use the color model defined in info_png.color for - these values. The encoder normally ignores info_png.color when auto_convert is on, but will - use it to interpret these values (and convert copies of them to its chosen color model). - - When encoding, avoid setting this to an expensive color, such as a non-gray value - when the image is gray, or the compression will be worse since it will be forced to - write the PNG with a more expensive color mode (when auto_convert is on). - - The decoder does not use this background color to edit the color of pixels. This is a - completely optional metadata feature. - */ - unsigned background_defined; /*is a suggested background color given?*/ - unsigned background_r; /*red/gray/palette component of suggested background color*/ - unsigned background_g; /*green component of suggested background color*/ - unsigned background_b; /*blue component of suggested background color*/ - - /* - Non-international text chunks (tEXt and zTXt) - - The char** arrays each contain num strings. The actual messages are in - text_strings, while text_keys are keywords that give a short description what - the actual text represents, e.g. Title, Author, Description, or anything else. - - All the string fields below including strings, keys, names and language tags are null terminated. - The PNG specification uses null characters for the keys, names and tags, and forbids null - characters to appear in the main text which is why we can use null termination everywhere here. - - A keyword is minimum 1 character and maximum 79 characters long (plus the - additional null terminator). It's discouraged to use a single line length - longer than 79 characters for texts. - - Don't allocate these text buffers yourself. Use the init/cleanup functions - correctly and use lodepng_add_text and lodepng_clear_text. - - Standard text chunk keywords and strings are encoded using Latin-1. - */ - size_t text_num; /*the amount of texts in these char** buffers (there may be more texts in itext)*/ - char** text_keys; /*the keyword of a text chunk (e.g. "Comment")*/ - char** text_strings; /*the actual text*/ - - /* - International text chunks (iTXt) - Similar to the non-international text chunks, but with additional strings - "langtags" and "transkeys", and the following text encodings are used: - keys: Latin-1, langtags: ASCII, transkeys and strings: UTF-8. - keys must be 1-79 characters (plus the additional null terminator), the other - strings are any length. - */ - size_t itext_num; /*the amount of international texts in this PNG*/ - char** itext_keys; /*the English keyword of the text chunk (e.g. "Comment")*/ - char** itext_langtags; /*language tag for this text's language, ISO/IEC 646 string, e.g. ISO 639 language tag*/ - char** itext_transkeys; /*keyword translated to the international language - UTF-8 string*/ - char** itext_strings; /*the actual international text - UTF-8 string*/ - - /*time chunk (tIME)*/ - unsigned time_defined; /*set to 1 to make the encoder generate a tIME chunk*/ - LodePNGTime time; - - /*phys chunk (pHYs)*/ - unsigned phys_defined; /*if 0, there is no pHYs chunk and the values below are undefined, if 1 else there is one*/ - unsigned phys_x; /*pixels per unit in x direction*/ - unsigned phys_y; /*pixels per unit in y direction*/ - unsigned phys_unit; /*may be 0 (unknown unit) or 1 (metre)*/ - - /* - Color profile related chunks: gAMA, cHRM, sRGB, iCPP, sBIT - - LodePNG does not apply any color conversions on pixels in the encoder or decoder and does not interpret these color - profile values. It merely passes on the information. If you wish to use color profiles and convert colors, please - use these values with a color management library. - - See the PNG, ICC and sRGB specifications for more information about the meaning of these values. - */ - - /* gAMA chunk: optional, overridden by sRGB or iCCP if those are present. */ - unsigned gama_defined; /* Whether a gAMA chunk is present (0 = not present, 1 = present). */ - unsigned gama_gamma; /* Gamma exponent times 100000 */ - - /* cHRM chunk: optional, overridden by sRGB or iCCP if those are present. */ - unsigned chrm_defined; /* Whether a cHRM chunk is present (0 = not present, 1 = present). */ - unsigned chrm_white_x; /* White Point x times 100000 */ - unsigned chrm_white_y; /* White Point y times 100000 */ - unsigned chrm_red_x; /* Red x times 100000 */ - unsigned chrm_red_y; /* Red y times 100000 */ - unsigned chrm_green_x; /* Green x times 100000 */ - unsigned chrm_green_y; /* Green y times 100000 */ - unsigned chrm_blue_x; /* Blue x times 100000 */ - unsigned chrm_blue_y; /* Blue y times 100000 */ - - /* - sRGB chunk: optional. May not appear at the same time as iCCP. - If gAMA is also present gAMA must contain value 45455. - If cHRM is also present cHRM must contain respectively 31270,32900,64000,33000,30000,60000,15000,6000. - */ - unsigned srgb_defined; /* Whether an sRGB chunk is present (0 = not present, 1 = present). */ - unsigned srgb_intent; /* Rendering intent: 0=perceptual, 1=rel. colorimetric, 2=saturation, 3=abs. colorimetric */ - - /* - iCCP chunk: optional. May not appear at the same time as sRGB. - - LodePNG does not parse or use the ICC profile (except its color space header field for an edge case), a - separate library to handle the ICC data (not included in LodePNG) format is needed to use it for color - management and conversions. - - For encoding, if iCCP is present, gAMA and cHRM are recommended to be added as well with values that match the ICC - profile as closely as possible, if you wish to do this you should provide the correct values for gAMA and cHRM and - enable their '_defined' flags since LodePNG will not automatically compute them from the ICC profile. - - For encoding, the ICC profile is required by the PNG specification to be an "RGB" profile for non-gray - PNG color types and a "GRAY" profile for gray PNG color types. If you disable auto_convert, you must ensure - the ICC profile type matches your requested color type, else the encoder gives an error. If auto_convert is - enabled (the default), and the ICC profile is not a good match for the pixel data, this will result in an encoder - error if the pixel data has non-gray pixels for a GRAY profile, or a silent less-optimal compression of the pixel - data if the pixels could be encoded as grayscale but the ICC profile is RGB. - - To avoid this do not set an ICC profile in the image unless there is a good reason for it, and when doing so - make sure you compute it carefully to avoid the above problems. - */ - unsigned iccp_defined; /* Whether an iCCP chunk is present (0 = not present, 1 = present). */ - char* iccp_name; /* Null terminated string with profile name, 1-79 bytes */ - /* - The ICC profile in iccp_profile_size bytes. - Don't allocate this buffer yourself. Use the init/cleanup functions - correctly and use lodepng_set_icc and lodepng_clear_icc. - */ - unsigned char* iccp_profile; - unsigned iccp_profile_size; /* The size of iccp_profile in bytes */ - - /* - sBIT chunk: significant bits. Optional metadata, only set this if needed. - - If defined, these values give the bit depth of the original data. Since PNG only stores 1, 2, 4, 8 or 16-bit - per channel data, the significant bits value can be used to indicate the original encoded data has another - sample depth, such as 10 or 12. - - Encoders using this value, when storing the pixel data, should use the most significant bits - of the data to store the original bits, and use a good sample depth scaling method such as - "left bit replication" to fill in the least significant bits, rather than fill zeroes. - - Decoders using this value, if able to work with data that's e.g. 10-bit or 12-bit, should right - shift the data to go back to the original bit depth, but decoders are also allowed to ignore - sbit and work e.g. with the 8-bit or 16-bit data from the PNG directly, since thanks - to the encoder contract, the values encoded in PNG are in valid range for the PNG bit depth. - - For grayscale images, sbit_g and sbit_b are not used, and for images that don't use color - type RGBA or grayscale+alpha, sbit_a is not used (it's not used even for palette images with - translucent palette values, or images with color key). The values that are used must be - greater than zero and smaller than or equal to the PNG bit depth. - - The color type from the header in the PNG image defines these used and unused fields: if - decoding with a color mode conversion, such as always decoding to RGBA, this metadata still - only uses the color type of the original PNG, and may e.g. lack the alpha channel info - if the PNG was RGB. When encoding with auto_convert (as well as without), also always the - color model defined in info_png.color determines this. - - NOTE: enabling sbit can hurt compression, because the encoder can then not always use - auto_convert to choose a more optimal color mode for the data, because the PNG format has - strict requirements for the allowed sbit values in combination with color modes. - For example, setting these fields to 10-bit will force the encoder to keep using a 16-bit per channel - color mode, even if the pixel data would in fact fit in a more efficient 8-bit mode. - */ - unsigned sbit_defined; /*is significant bits given? if not, the values below are unused*/ - unsigned sbit_r; /*red or gray component of significant bits*/ - unsigned sbit_g; /*green component of significant bits*/ - unsigned sbit_b; /*blue component of significant bits*/ - unsigned sbit_a; /*alpha component of significant bits*/ - - /* End of color profile related chunks */ - - - /* - unknown chunks: chunks not known by LodePNG, passed on byte for byte. - - There are 3 buffers, one for each position in the PNG where unknown chunks can appear. - Each buffer contains all unknown chunks for that position consecutively. - The 3 positions are: - 0: between IHDR and PLTE, 1: between PLTE and IDAT, 2: between IDAT and IEND. - - For encoding, do not store critical chunks or known chunks that are enabled with a "_defined" flag - above in here, since the encoder will blindly follow this and could then encode an invalid PNG file - (such as one with two IHDR chunks or the disallowed combination of sRGB with iCCP). But do use - this if you wish to store an ancillary chunk that is not supported by LodePNG (such as sPLT or hIST), - or any non-standard PNG chunk. - - Do not allocate or traverse this data yourself. Use the chunk traversing functions declared - later, such as lodepng_chunk_next and lodepng_chunk_append, to read/write this struct. - */ - unsigned char* unknown_chunks_data[3]; - size_t unknown_chunks_size[3]; /*size in bytes of the unknown chunks, given for protection*/ -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -} LodePNGInfo; - -/*init, cleanup and copy functions to use with this struct*/ -void lodepng_info_init(LodePNGInfo* info); -void lodepng_info_cleanup(LodePNGInfo* info); -/*return value is error code (0 means no error)*/ -unsigned lodepng_info_copy(LodePNGInfo* dest, const LodePNGInfo* source); - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS -unsigned lodepng_add_text(LodePNGInfo* info, const char* key, const char* str); /*push back both texts at once*/ -void lodepng_clear_text(LodePNGInfo* info); /*use this to clear the texts again after you filled them in*/ - -unsigned lodepng_add_itext(LodePNGInfo* info, const char* key, const char* langtag, - const char* transkey, const char* str); /*push back the 4 texts of 1 chunk at once*/ -void lodepng_clear_itext(LodePNGInfo* info); /*use this to clear the itexts again after you filled them in*/ - -/*replaces if exists*/ -unsigned lodepng_set_icc(LodePNGInfo* info, const char* name, const unsigned char* profile, unsigned profile_size); -void lodepng_clear_icc(LodePNGInfo* info); /*use this to clear the texts again after you filled them in*/ -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -/* -Converts raw buffer from one color type to another color type, based on -LodePNGColorMode structs to describe the input and output color type. -See the reference manual at the end of this header file to see which color conversions are supported. -return value = LodePNG error code (0 if all went ok, an error if the conversion isn't supported) -The out buffer must have size (w * h * bpp + 7) / 8, where bpp is the bits per pixel -of the output color type (lodepng_get_bpp). -For < 8 bpp images, there should not be padding bits at the end of scanlines. -For 16-bit per channel colors, uses big endian format like PNG does. -Return value is LodePNG error code -*/ -unsigned lodepng_convert(unsigned char* out, const unsigned char* in, - const LodePNGColorMode* mode_out, const LodePNGColorMode* mode_in, - unsigned w, unsigned h); - -#ifdef LODEPNG_COMPILE_DECODER -/* -Settings for the decoder. This contains settings for the PNG and the Zlib -decoder, but not the Info settings from the Info structs. -*/ -typedef struct LodePNGDecoderSettings { - LodePNGDecompressSettings zlibsettings; /*in here is the setting to ignore Adler32 checksums*/ - - /* Check LodePNGDecompressSettings for more ignorable errors such as ignore_adler32 */ - unsigned ignore_crc; /*ignore CRC checksums*/ - unsigned ignore_critical; /*ignore unknown critical chunks*/ - unsigned ignore_end; /*ignore issues at end of file if possible (missing IEND chunk, too large chunk, ...)*/ - /* TODO: make a system involving warnings with levels and a strict mode instead. Other potentially recoverable - errors: srgb rendering intent value, size of content of ancillary chunks, more than 79 characters for some - strings, placement/combination rules for ancillary chunks, crc of unknown chunks, allowed characters - in string keys, etc... */ - - unsigned color_convert; /*whether to convert the PNG to the color type you want. Default: yes*/ - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - unsigned read_text_chunks; /*if false but remember_unknown_chunks is true, they're stored in the unknown chunks*/ - - /*store all bytes from unknown chunks in the LodePNGInfo (off by default, useful for a png editor)*/ - unsigned remember_unknown_chunks; - - /* maximum size for decompressed text chunks. If a text chunk's text is larger than this, an error is returned, - unless reading text chunks is disabled or this limit is set higher or disabled. Set to 0 to allow any size. - By default it is a value that prevents unreasonably large strings from hogging memory. */ - size_t max_text_size; - - /* maximum size for compressed ICC chunks. If the ICC profile is larger than this, an error will be returned. Set to - 0 to allow any size. By default this is a value that prevents ICC profiles that would be much larger than any - legitimate profile could be to hog memory. */ - size_t max_icc_size; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -} LodePNGDecoderSettings; - -void lodepng_decoder_settings_init(LodePNGDecoderSettings* settings); -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_ENCODER -/*automatically use color type with less bits per pixel if losslessly possible. Default: AUTO*/ -typedef enum LodePNGFilterStrategy { - /*every filter at zero*/ - LFS_ZERO = 0, - /*every filter at 1, 2, 3 or 4 (paeth), unlike LFS_ZERO not a good choice, but for testing*/ - LFS_ONE = 1, - LFS_TWO = 2, - LFS_THREE = 3, - LFS_FOUR = 4, - /*Use filter that gives minimum sum, as described in the official PNG filter heuristic.*/ - LFS_MINSUM, - /*Use the filter type that gives smallest Shannon entropy for this scanline. Depending - on the image, this is better or worse than minsum.*/ - LFS_ENTROPY, - /* - Brute-force-search PNG filters by compressing each filter for each scanline. - Experimental, very slow, and only rarely gives better compression than MINSUM. - */ - LFS_BRUTE_FORCE, - /*use predefined_filters buffer: you specify the filter type for each scanline*/ - LFS_PREDEFINED -} LodePNGFilterStrategy; - -/*Gives characteristics about the integer RGBA colors of the image (count, alpha channel usage, bit depth, ...), -which helps decide which color model to use for encoding. -Used internally by default if "auto_convert" is enabled. Public because it's useful for custom algorithms.*/ -typedef struct LodePNGColorStats { - unsigned colored; /*not grayscale*/ - unsigned key; /*image is not opaque and color key is possible instead of full alpha*/ - unsigned short key_r; /*key values, always as 16-bit, in 8-bit case the byte is duplicated, e.g. 65535 means 255*/ - unsigned short key_g; - unsigned short key_b; - unsigned alpha; /*image is not opaque and alpha channel or alpha palette required*/ - unsigned numcolors; /*amount of colors, up to 257. Not valid if bits == 16 or allow_palette is disabled.*/ - unsigned char palette[1024]; /*Remembers up to the first 256 RGBA colors, in no particular order, only valid when numcolors is valid*/ - unsigned bits; /*bits per channel (not for palette). 1,2 or 4 for grayscale only. 16 if 16-bit per channel required.*/ - size_t numpixels; - - /*user settings for computing/using the stats*/ - unsigned allow_palette; /*default 1. if 0, disallow choosing palette colortype in auto_choose_color, and don't count numcolors*/ - unsigned allow_greyscale; /*default 1. if 0, choose RGB or RGBA even if the image only has gray colors*/ -} LodePNGColorStats; - -void lodepng_color_stats_init(LodePNGColorStats* stats); - -/*Get a LodePNGColorStats of the image. The stats must already have been inited. -Returns error code (e.g. alloc fail) or 0 if ok.*/ -unsigned lodepng_compute_color_stats(LodePNGColorStats* stats, - const unsigned char* image, unsigned w, unsigned h, - const LodePNGColorMode* mode_in); - -/*Settings for the encoder.*/ -typedef struct LodePNGEncoderSettings { - LodePNGCompressSettings zlibsettings; /*settings for the zlib encoder, such as window size, ...*/ - - unsigned auto_convert; /*automatically choose output PNG color type. Default: true*/ - - /*If true, follows the official PNG heuristic: if the PNG uses a palette or lower than - 8 bit depth, set all filters to zero. Otherwise use the filter_strategy. Note that to - completely follow the official PNG heuristic, filter_palette_zero must be true and - filter_strategy must be LFS_MINSUM*/ - unsigned filter_palette_zero; - /*Which filter strategy to use when not using zeroes due to filter_palette_zero. - Set filter_palette_zero to 0 to ensure always using your chosen strategy. Default: LFS_MINSUM*/ - LodePNGFilterStrategy filter_strategy; - /*used if filter_strategy is LFS_PREDEFINED. In that case, this must point to a buffer with - the same length as the amount of scanlines in the image, and each value must <= 5. You - have to cleanup this buffer, LodePNG will never free it. Don't forget that filter_palette_zero - must be set to 0 to ensure this is also used on palette or low bitdepth images.*/ - const unsigned char* predefined_filters; - - /*force creating a PLTE chunk if colortype is 2 or 6 (= a suggested palette). - If colortype is 3, PLTE is always created. If color type is explicitely set - to a grayscale type (1 or 4), this is not done and is ignored. If enabling this, - a palette must be present in the info_png. - NOTE: enabling this may worsen compression if auto_convert is used to choose - optimal color mode, because it cannot use grayscale color modes in this case*/ - unsigned force_palette; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /*add LodePNG identifier and version as a text chunk, for debugging*/ - unsigned add_id; - /*encode text chunks as zTXt chunks instead of tEXt chunks, and use compression in iTXt chunks*/ - unsigned text_compression; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -} LodePNGEncoderSettings; - -void lodepng_encoder_settings_init(LodePNGEncoderSettings* settings); -#endif /*LODEPNG_COMPILE_ENCODER*/ - - -#if defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) -/*The settings, state and information for extended encoding and decoding.*/ -typedef struct LodePNGState { -#ifdef LODEPNG_COMPILE_DECODER - LodePNGDecoderSettings decoder; /*the decoding settings*/ -#endif /*LODEPNG_COMPILE_DECODER*/ -#ifdef LODEPNG_COMPILE_ENCODER - LodePNGEncoderSettings encoder; /*the encoding settings*/ -#endif /*LODEPNG_COMPILE_ENCODER*/ - LodePNGColorMode info_raw; /*specifies the format in which you would like to get the raw pixel buffer*/ - LodePNGInfo info_png; /*info of the PNG image obtained after decoding*/ - unsigned error; -} LodePNGState; - -/*init, cleanup and copy functions to use with this struct*/ -void lodepng_state_init(LodePNGState* state); -void lodepng_state_cleanup(LodePNGState* state); -void lodepng_state_copy(LodePNGState* dest, const LodePNGState* source); -#endif /* defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) */ - -#ifdef LODEPNG_COMPILE_DECODER -/* -Same as lodepng_decode_memory, but uses a LodePNGState to allow custom settings and -getting much more information about the PNG image and color mode. -*/ -unsigned lodepng_decode(unsigned char** out, unsigned* w, unsigned* h, - LodePNGState* state, - const unsigned char* in, size_t insize); - -/* -Read the PNG header, but not the actual data. This returns only the information -that is in the IHDR chunk of the PNG, such as width, height and color type. The -information is placed in the info_png field of the LodePNGState. -*/ -unsigned lodepng_inspect(unsigned* w, unsigned* h, - LodePNGState* state, - const unsigned char* in, size_t insize); -#endif /*LODEPNG_COMPILE_DECODER*/ - -/* -Reads one metadata chunk (other than IHDR, which is handled by lodepng_inspect) -of the PNG file and outputs what it read in the state. Returns error code on failure. -Use lodepng_inspect first with a new state, then e.g. lodepng_chunk_find_const -to find the desired chunk type, and if non null use lodepng_inspect_chunk (with -chunk_pointer - start_of_file as pos). -Supports most metadata chunks from the PNG standard (gAMA, bKGD, tEXt, ...). -Ignores unsupported, unknown, non-metadata or IHDR chunks (without error). -Requirements: &in[pos] must point to start of a chunk, must use regular -lodepng_inspect first since format of most other chunks depends on IHDR, and if -there is a PLTE chunk, that one must be inspected before tRNS or bKGD. -*/ -unsigned lodepng_inspect_chunk(LodePNGState* state, size_t pos, - const unsigned char* in, size_t insize); - -#ifdef LODEPNG_COMPILE_ENCODER -/*This function allocates the out buffer with standard malloc and stores the size in *outsize.*/ -unsigned lodepng_encode(unsigned char** out, size_t* outsize, - const unsigned char* image, unsigned w, unsigned h, - LodePNGState* state); -#endif /*LODEPNG_COMPILE_ENCODER*/ - -/* -The lodepng_chunk functions are normally not needed, except to traverse the -unknown chunks stored in the LodePNGInfo struct, or add new ones to it. -It also allows traversing the chunks of an encoded PNG file yourself. - -The chunk pointer always points to the beginning of the chunk itself, that is -the first byte of the 4 length bytes. - -In the PNG file format, chunks have the following format: --4 bytes length: length of the data of the chunk in bytes (chunk itself is 12 bytes longer) --4 bytes chunk type (ASCII a-z,A-Z only, see below) --length bytes of data (may be 0 bytes if length was 0) --4 bytes of CRC, computed on chunk name + data - -The first chunk starts at the 8th byte of the PNG file, the entire rest of the file -exists out of concatenated chunks with the above format. - -PNG standard chunk ASCII naming conventions: --First byte: uppercase = critical, lowercase = ancillary --Second byte: uppercase = public, lowercase = private --Third byte: must be uppercase --Fourth byte: uppercase = unsafe to copy, lowercase = safe to copy -*/ - -/* -Gets the length of the data of the chunk. Total chunk length has 12 bytes more. -There must be at least 4 bytes to read from. If the result value is too large, -it may be corrupt data. -*/ -unsigned lodepng_chunk_length(const unsigned char* chunk); - -/*puts the 4-byte type in null terminated string*/ -void lodepng_chunk_type(char type[5], const unsigned char* chunk); - -/*check if the type is the given type*/ -unsigned char lodepng_chunk_type_equals(const unsigned char* chunk, const char* type); - -/*0: it's one of the critical chunk types, 1: it's an ancillary chunk (see PNG standard)*/ -unsigned char lodepng_chunk_ancillary(const unsigned char* chunk); - -/*0: public, 1: private (see PNG standard)*/ -unsigned char lodepng_chunk_private(const unsigned char* chunk); - -/*0: the chunk is unsafe to copy, 1: the chunk is safe to copy (see PNG standard)*/ -unsigned char lodepng_chunk_safetocopy(const unsigned char* chunk); - -/*get pointer to the data of the chunk, where the input points to the header of the chunk*/ -unsigned char* lodepng_chunk_data(unsigned char* chunk); -const unsigned char* lodepng_chunk_data_const(const unsigned char* chunk); - -/*returns 0 if the crc is correct, 1 if it's incorrect (0 for OK as usual!)*/ -unsigned lodepng_chunk_check_crc(const unsigned char* chunk); - -/*generates the correct CRC from the data and puts it in the last 4 bytes of the chunk*/ -void lodepng_chunk_generate_crc(unsigned char* chunk); - -/* -Iterate to next chunks, allows iterating through all chunks of the PNG file. -Input must be at the beginning of a chunk (result of a previous lodepng_chunk_next call, -or the 8th byte of a PNG file which always has the first chunk), or alternatively may -point to the first byte of the PNG file (which is not a chunk but the magic header, the -function will then skip over it and return the first real chunk). -Will output pointer to the start of the next chunk, or at or beyond end of the file if there -is no more chunk after this or possibly if the chunk is corrupt. -Start this process at the 8th byte of the PNG file. -In a non-corrupt PNG file, the last chunk should have name "IEND". -*/ -unsigned char* lodepng_chunk_next(unsigned char* chunk, unsigned char* end); -const unsigned char* lodepng_chunk_next_const(const unsigned char* chunk, const unsigned char* end); - -/*Finds the first chunk with the given type in the range [chunk, end), or returns NULL if not found.*/ -unsigned char* lodepng_chunk_find(unsigned char* chunk, unsigned char* end, const char type[5]); -const unsigned char* lodepng_chunk_find_const(const unsigned char* chunk, const unsigned char* end, const char type[5]); - -/* -Appends chunk to the data in out. The given chunk should already have its chunk header. -The out variable and outsize are updated to reflect the new reallocated buffer. -Returns error code (0 if it went ok) -*/ -unsigned lodepng_chunk_append(unsigned char** out, size_t* outsize, const unsigned char* chunk); - -/* -Appends new chunk to out. The chunk to append is given by giving its length, type -and data separately. The type is a 4-letter string. -The out variable and outsize are updated to reflect the new reallocated buffer. -Returne error code (0 if it went ok) -*/ -unsigned lodepng_chunk_create(unsigned char** out, size_t* outsize, unsigned length, - const char* type, const unsigned char* data); - - -/*Calculate CRC32 of buffer*/ -unsigned lodepng_crc32(const unsigned char* buf, size_t len); -#endif /*LODEPNG_COMPILE_PNG*/ - - -#ifdef LODEPNG_COMPILE_ZLIB -/* -This zlib part can be used independently to zlib compress and decompress a -buffer. It cannot be used to create gzip files however, and it only supports the -part of zlib that is required for PNG, it does not support dictionaries. -*/ - -#ifdef LODEPNG_COMPILE_DECODER -/*Inflate a buffer. Inflate is the decompression step of deflate. Out buffer must be freed after use.*/ -unsigned lodepng_inflate(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGDecompressSettings* settings); - -/* -Decompresses Zlib data. Reallocates the out buffer and appends the data. The -data must be according to the zlib specification. -Either, *out must be NULL and *outsize must be 0, or, *out must be a valid -buffer and *outsize its size in bytes. out must be freed by user after usage. -*/ -unsigned lodepng_zlib_decompress(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGDecompressSettings* settings); -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_ENCODER -/* -Compresses data with Zlib. Reallocates the out buffer and appends the data. -Zlib adds a small header and trailer around the deflate data. -The data is output in the format of the zlib specification. -Either, *out must be NULL and *outsize must be 0, or, *out must be a valid -buffer and *outsize its size in bytes. out must be freed by user after usage. -*/ -unsigned lodepng_zlib_compress(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGCompressSettings* settings); - -/* -Find length-limited Huffman code for given frequencies. This function is in the -public interface only for tests, it's used internally by lodepng_deflate. -*/ -unsigned lodepng_huffman_code_lengths(unsigned* lengths, const unsigned* frequencies, - size_t numcodes, unsigned maxbitlen); - -/*Compress a buffer with deflate. See RFC 1951. Out buffer must be freed after use.*/ -unsigned lodepng_deflate(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGCompressSettings* settings); - -#endif /*LODEPNG_COMPILE_ENCODER*/ -#endif /*LODEPNG_COMPILE_ZLIB*/ - -#ifdef LODEPNG_COMPILE_DISK -/* -Load a file from disk into buffer. The function allocates the out buffer, and -after usage you should free it. -out: output parameter, contains pointer to loaded buffer. -outsize: output parameter, size of the allocated out buffer -filename: the path to the file to load -return value: error code (0 means ok) - -NOTE: Wide-character filenames are not supported, you can use an external method -to handle such files and decode in-memory. -*/ -unsigned lodepng_load_file(unsigned char** out, size_t* outsize, const char* filename); - -/* -Save a file from buffer to disk. Warning, if it exists, this function overwrites -the file without warning! -buffer: the buffer to write -buffersize: size of the buffer to write -filename: the path to the file to save to -return value: error code (0 means ok) - -NOTE: Wide-character filenames are not supported, you can use an external method -to handle such files and encode in-memory -*/ -unsigned lodepng_save_file(const unsigned char* buffer, size_t buffersize, const char* filename); -#endif /*LODEPNG_COMPILE_DISK*/ - -#ifdef LODEPNG_COMPILE_CPP -/* The LodePNG C++ wrapper uses std::vectors instead of manually allocated memory buffers. */ -namespace lodepng { -#ifdef LODEPNG_COMPILE_PNG -class State : public LodePNGState { - public: - State(); - State(const State& other); - ~State(); - State& operator=(const State& other); -}; - -#ifdef LODEPNG_COMPILE_DECODER -/* Same as other lodepng::decode, but using a State for more settings and information. */ -unsigned decode(std::vector<unsigned char>& out, unsigned& w, unsigned& h, - State& state, - const unsigned char* in, size_t insize); -unsigned decode(std::vector<unsigned char>& out, unsigned& w, unsigned& h, - State& state, - const std::vector<unsigned char>& in); -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_ENCODER -/* Same as other lodepng::encode, but using a State for more settings and information. */ -unsigned encode(std::vector<unsigned char>& out, - const unsigned char* in, unsigned w, unsigned h, - State& state); -unsigned encode(std::vector<unsigned char>& out, - const std::vector<unsigned char>& in, unsigned w, unsigned h, - State& state); -#endif /*LODEPNG_COMPILE_ENCODER*/ - -#ifdef LODEPNG_COMPILE_DISK -/* -Load a file from disk into an std::vector. -return value: error code (0 means ok) - -NOTE: Wide-character filenames are not supported, you can use an external method -to handle such files and decode in-memory -*/ -unsigned load_file(std::vector<unsigned char>& buffer, const std::string& filename); - -/* -Save the binary data in an std::vector to a file on disk. The file is overwritten -without warning. - -NOTE: Wide-character filenames are not supported, you can use an external method -to handle such files and encode in-memory -*/ -unsigned save_file(const std::vector<unsigned char>& buffer, const std::string& filename); -#endif /* LODEPNG_COMPILE_DISK */ -#endif /* LODEPNG_COMPILE_PNG */ - -#ifdef LODEPNG_COMPILE_ZLIB -#ifdef LODEPNG_COMPILE_DECODER -/* Zlib-decompress an unsigned char buffer */ -unsigned decompress(std::vector<unsigned char>& out, const unsigned char* in, size_t insize, - const LodePNGDecompressSettings& settings = lodepng_default_decompress_settings); - -/* Zlib-decompress an std::vector */ -unsigned decompress(std::vector<unsigned char>& out, const std::vector<unsigned char>& in, - const LodePNGDecompressSettings& settings = lodepng_default_decompress_settings); -#endif /* LODEPNG_COMPILE_DECODER */ - -#ifdef LODEPNG_COMPILE_ENCODER -/* Zlib-compress an unsigned char buffer */ -unsigned compress(std::vector<unsigned char>& out, const unsigned char* in, size_t insize, - const LodePNGCompressSettings& settings = lodepng_default_compress_settings); - -/* Zlib-compress an std::vector */ -unsigned compress(std::vector<unsigned char>& out, const std::vector<unsigned char>& in, - const LodePNGCompressSettings& settings = lodepng_default_compress_settings); -#endif /* LODEPNG_COMPILE_ENCODER */ -#endif /* LODEPNG_COMPILE_ZLIB */ -} /* namespace lodepng */ -#endif /*LODEPNG_COMPILE_CPP*/ - -/* -TODO: -[.] test if there are no memory leaks or security exploits - done a lot but needs to be checked often -[.] check compatibility with various compilers - done but needs to be redone for every newer version -[X] converting color to 16-bit per channel types -[X] support color profile chunk types (but never let them touch RGB values by default) -[ ] support all public PNG chunk types (almost done except sPLT and hIST) -[ ] make sure encoder generates no chunks with size > (2^31)-1 -[ ] partial decoding (stream processing) -[X] let the "isFullyOpaque" function check color keys and transparent palettes too -[X] better name for the variables "codes", "codesD", "codelengthcodes", "clcl" and "lldl" -[ ] allow treating some errors like warnings, when image is recoverable (e.g. 69, 57, 58) -[ ] make warnings like: oob palette, checksum fail, data after iend, wrong/unknown crit chunk, no null terminator in text, ... -[ ] error messages with line numbers (and version) -[ ] errors in state instead of as return code? -[ ] new errors/warnings like suspiciously big decompressed ztxt or iccp chunk -[ ] let the C++ wrapper catch exceptions coming from the standard library and return LodePNG error codes -[ ] allow user to provide custom color conversion functions, e.g. for premultiplied alpha, padding bits or not, ... -[ ] allow user to give data (void*) to custom allocator -[X] provide alternatives for C library functions not present on some platforms (memcpy, ...) -*/ - -#endif /*LODEPNG_H inclusion guard*/ - -/* -LodePNG Documentation ---------------------- - -0. table of contents --------------------- - - 1. about - 1.1. supported features - 1.2. features not supported - 2. C and C++ version - 3. security - 4. decoding - 5. encoding - 6. color conversions - 6.1. PNG color types - 6.2. color conversions - 6.3. padding bits - 6.4. A note about 16-bits per channel and endianness - 7. error values - 8. chunks and PNG editing - 9. compiler support - 10. examples - 10.1. decoder C++ example - 10.2. decoder C example - 11. state settings reference - 12. changes - 13. contact information - - -1. about --------- - -PNG is a file format to store raster images losslessly with good compression, -supporting different color types and alpha channel. - -LodePNG is a PNG codec according to the Portable Network Graphics (PNG) -Specification (Second Edition) - W3C Recommendation 10 November 2003. - -The specifications used are: - -*) Portable Network Graphics (PNG) Specification (Second Edition): - http://www.w3.org/TR/2003/REC-PNG-20031110 -*) RFC 1950 ZLIB Compressed Data Format version 3.3: - http://www.gzip.org/zlib/rfc-zlib.html -*) RFC 1951 DEFLATE Compressed Data Format Specification ver 1.3: - http://www.gzip.org/zlib/rfc-deflate.html - -The most recent version of LodePNG can currently be found at -http://lodev.org/lodepng/ - -LodePNG works both in C (ISO C90) and C++, with a C++ wrapper that adds -extra functionality. - -LodePNG exists out of two files: --lodepng.h: the header file for both C and C++ --lodepng.c(pp): give it the name lodepng.c or lodepng.cpp (or .cc) depending on your usage - -If you want to start using LodePNG right away without reading this doc, get the -examples from the LodePNG website to see how to use it in code, or check the -smaller examples in chapter 13 here. - -LodePNG is simple but only supports the basic requirements. To achieve -simplicity, the following design choices were made: There are no dependencies -on any external library. There are functions to decode and encode a PNG with -a single function call, and extended versions of these functions taking a -LodePNGState struct allowing to specify or get more information. By default -the colors of the raw image are always RGB or RGBA, no matter what color type -the PNG file uses. To read and write files, there are simple functions to -convert the files to/from buffers in memory. - -This all makes LodePNG suitable for loading textures in games, demos and small -programs, ... It's less suitable for full fledged image editors, loading PNGs -over network (it requires all the image data to be available before decoding can -begin), life-critical systems, ... - -1.1. supported features ------------------------ - -The following features are supported by the decoder: - -*) decoding of PNGs with any color type, bit depth and interlace mode, to a 24- or 32-bit color raw image, - or the same color type as the PNG -*) encoding of PNGs, from any raw image to 24- or 32-bit color, or the same color type as the raw image -*) Adam7 interlace and deinterlace for any color type -*) loading the image from harddisk or decoding it from a buffer from other sources than harddisk -*) support for alpha channels, including RGBA color model, translucent palettes and color keying -*) zlib decompression (inflate) -*) zlib compression (deflate) -*) CRC32 and ADLER32 checksums -*) colorimetric color profile conversions: currently experimentally available in lodepng_util.cpp only, - plus alternatively ability to pass on chroma/gamma/ICC profile information to other color management system. -*) handling of unknown chunks, allowing making a PNG editor that stores custom and unknown chunks. -*) the following chunks are supported by both encoder and decoder: - IHDR: header information - PLTE: color palette - IDAT: pixel data - IEND: the final chunk - tRNS: transparency for palettized images - tEXt: textual information - zTXt: compressed textual information - iTXt: international textual information - bKGD: suggested background color - pHYs: physical dimensions - tIME: modification time - cHRM: RGB chromaticities - gAMA: RGB gamma correction - iCCP: ICC color profile - sRGB: rendering intent - sBIT: significant bits - -1.2. features not supported ---------------------------- - -The following features are not (yet) supported: - -*) some features needed to make a conformant PNG-Editor might be still missing. -*) partial loading/stream processing. All data must be available and is processed in one call. -*) The hIST and sPLT public chunks are not (yet) supported but treated as unknown chunks - - -2. C and C++ version --------------------- - -The C version uses buffers allocated with alloc that you need to free() -yourself. You need to use init and cleanup functions for each struct whenever -using a struct from the C version to avoid exploits and memory leaks. - -The C++ version has extra functions with std::vectors in the interface and the -lodepng::State class which is a LodePNGState with constructor and destructor. - -These files work without modification for both C and C++ compilers because all -the additional C++ code is in "#ifdef __cplusplus" blocks that make C-compilers -ignore it, and the C code is made to compile both with strict ISO C90 and C++. - -To use the C++ version, you need to rename the source file to lodepng.cpp -(instead of lodepng.c), and compile it with a C++ compiler. - -To use the C version, you need to rename the source file to lodepng.c (instead -of lodepng.cpp), and compile it with a C compiler. - - -3. Security ------------ - -Even if carefully designed, it's always possible that LodePNG contains possible -exploits. If you discover one, please let me know, and it will be fixed. - -When using LodePNG, care has to be taken with the C version of LodePNG, as well -as the C-style structs when working with C++. The following conventions are used -for all C-style structs: - --if a struct has a corresponding init function, always call the init function when making a new one --if a struct has a corresponding cleanup function, call it before the struct disappears to avoid memory leaks --if a struct has a corresponding copy function, use the copy function instead of "=". - The destination must also be inited already. - - -4. Decoding ------------ - -Decoding converts a PNG compressed image to a raw pixel buffer. - -Most documentation on using the decoder is at its declarations in the header -above. For C, simple decoding can be done with functions such as -lodepng_decode32, and more advanced decoding can be done with the struct -LodePNGState and lodepng_decode. For C++, all decoding can be done with the -various lodepng::decode functions, and lodepng::State can be used for advanced -features. - -When using the LodePNGState, it uses the following fields for decoding: -*) LodePNGInfo info_png: it stores extra information about the PNG (the input) in here -*) LodePNGColorMode info_raw: here you can say what color mode of the raw image (the output) you want to get -*) LodePNGDecoderSettings decoder: you can specify a few extra settings for the decoder to use - -LodePNGInfo info_png --------------------- - -After decoding, this contains extra information of the PNG image, except the actual -pixels, width and height because these are already gotten directly from the decoder -functions. - -It contains for example the original color type of the PNG image, text comments, -suggested background color, etc... More details about the LodePNGInfo struct are -at its declaration documentation. - -LodePNGColorMode info_raw -------------------------- - -When decoding, here you can specify which color type you want -the resulting raw image to be. If this is different from the colortype of the -PNG, then the decoder will automatically convert the result. This conversion -always works, except if you want it to convert a color PNG to grayscale or to -a palette with missing colors. - -By default, 32-bit color is used for the result. - -LodePNGDecoderSettings decoder ------------------------------- - -The settings can be used to ignore the errors created by invalid CRC and Adler32 -chunks, and to disable the decoding of tEXt chunks. - -There's also a setting color_convert, true by default. If false, no conversion -is done, the resulting data will be as it was in the PNG (after decompression) -and you'll have to puzzle the colors of the pixels together yourself using the -color type information in the LodePNGInfo. - - -5. Encoding ------------ - -Encoding converts a raw pixel buffer to a PNG compressed image. - -Most documentation on using the encoder is at its declarations in the header -above. For C, simple encoding can be done with functions such as -lodepng_encode32, and more advanced decoding can be done with the struct -LodePNGState and lodepng_encode. For C++, all encoding can be done with the -various lodepng::encode functions, and lodepng::State can be used for advanced -features. - -Like the decoder, the encoder can also give errors. However it gives less errors -since the encoder input is trusted, the decoder input (a PNG image that could -be forged by anyone) is not trusted. - -When using the LodePNGState, it uses the following fields for encoding: -*) LodePNGInfo info_png: here you specify how you want the PNG (the output) to be. -*) LodePNGColorMode info_raw: here you say what color type of the raw image (the input) has -*) LodePNGEncoderSettings encoder: you can specify a few settings for the encoder to use - -LodePNGInfo info_png --------------------- - -When encoding, you use this the opposite way as when decoding: for encoding, -you fill in the values you want the PNG to have before encoding. By default it's -not needed to specify a color type for the PNG since it's automatically chosen, -but it's possible to choose it yourself given the right settings. - -The encoder will not always exactly match the LodePNGInfo struct you give, -it tries as close as possible. Some things are ignored by the encoder. The -encoder uses, for example, the following settings from it when applicable: -colortype and bitdepth, text chunks, time chunk, the color key, the palette, the -background color, the interlace method, unknown chunks, ... - -When encoding to a PNG with colortype 3, the encoder will generate a PLTE chunk. -If the palette contains any colors for which the alpha channel is not 255 (so -there are translucent colors in the palette), it'll add a tRNS chunk. - -LodePNGColorMode info_raw -------------------------- - -You specify the color type of the raw image that you give to the input here, -including a possible transparent color key and palette you happen to be using in -your raw image data. - -By default, 32-bit color is assumed, meaning your input has to be in RGBA -format with 4 bytes (unsigned chars) per pixel. - -LodePNGEncoderSettings encoder ------------------------------- - -The following settings are supported (some are in sub-structs): -*) auto_convert: when this option is enabled, the encoder will -automatically choose the smallest possible color mode (including color key) that -can encode the colors of all pixels without information loss. -*) btype: the block type for LZ77. 0 = uncompressed, 1 = fixed huffman tree, - 2 = dynamic huffman tree (best compression). Should be 2 for proper - compression. -*) use_lz77: whether or not to use LZ77 for compressed block types. Should be - true for proper compression. -*) windowsize: the window size used by the LZ77 encoder (1 - 32768). Has value - 2048 by default, but can be set to 32768 for better, but slow, compression. -*) force_palette: if colortype is 2 or 6, you can make the encoder write a PLTE - chunk if force_palette is true. This can used as suggested palette to convert - to by viewers that don't support more than 256 colors (if those still exist) -*) add_id: add text chunk "Encoder: LodePNG <version>" to the image. -*) text_compression: default 1. If 1, it'll store texts as zTXt instead of tEXt chunks. - zTXt chunks use zlib compression on the text. This gives a smaller result on - large texts but a larger result on small texts (such as a single program name). - It's all tEXt or all zTXt though, there's no separate setting per text yet. - - -6. color conversions --------------------- - -An important thing to note about LodePNG, is that the color type of the PNG, and -the color type of the raw image, are completely independent. By default, when -you decode a PNG, you get the result as a raw image in the color type you want, -no matter whether the PNG was encoded with a palette, grayscale or RGBA color. -And if you encode an image, by default LodePNG will automatically choose the PNG -color type that gives good compression based on the values of colors and amount -of colors in the image. It can be configured to let you control it instead as -well, though. - -To be able to do this, LodePNG does conversions from one color mode to another. -It can convert from almost any color type to any other color type, except the -following conversions: RGB to grayscale is not supported, and converting to a -palette when the palette doesn't have a required color is not supported. This is -not supported on purpose: this is information loss which requires a color -reduction algorithm that is beyond the scope of a PNG encoder (yes, RGB to gray -is easy, but there are multiple ways if you want to give some channels more -weight). - -By default, when decoding, you get the raw image in 32-bit RGBA or 24-bit RGB -color, no matter what color type the PNG has. And by default when encoding, -LodePNG automatically picks the best color model for the output PNG, and expects -the input image to be 32-bit RGBA or 24-bit RGB. So, unless you want to control -the color format of the images yourself, you can skip this chapter. - -6.1. PNG color types --------------------- - -A PNG image can have many color types, ranging from 1-bit color to 64-bit color, -as well as palettized color modes. After the zlib decompression and unfiltering -in the PNG image is done, the raw pixel data will have that color type and thus -a certain amount of bits per pixel. If you want the output raw image after -decoding to have another color type, a conversion is done by LodePNG. - -The PNG specification gives the following color types: - -0: grayscale, bit depths 1, 2, 4, 8, 16 -2: RGB, bit depths 8 and 16 -3: palette, bit depths 1, 2, 4 and 8 -4: grayscale with alpha, bit depths 8 and 16 -6: RGBA, bit depths 8 and 16 - -Bit depth is the amount of bits per pixel per color channel. So the total amount -of bits per pixel is: amount of channels * bitdepth. - -6.2. color conversions ----------------------- - -As explained in the sections about the encoder and decoder, you can specify -color types and bit depths in info_png and info_raw to change the default -behaviour. - -If, when decoding, you want the raw image to be something else than the default, -you need to set the color type and bit depth you want in the LodePNGColorMode, -or the parameters colortype and bitdepth of the simple decoding function. - -If, when encoding, you use another color type than the default in the raw input -image, you need to specify its color type and bit depth in the LodePNGColorMode -of the raw image, or use the parameters colortype and bitdepth of the simple -encoding function. - -If, when encoding, you don't want LodePNG to choose the output PNG color type -but control it yourself, you need to set auto_convert in the encoder settings -to false, and specify the color type you want in the LodePNGInfo of the -encoder (including palette: it can generate a palette if auto_convert is true, -otherwise not). - -If the input and output color type differ (whether user chosen or auto chosen), -LodePNG will do a color conversion, which follows the rules below, and may -sometimes result in an error. - -To avoid some confusion: --the decoder converts from PNG to raw image --the encoder converts from raw image to PNG --the colortype and bitdepth in LodePNGColorMode info_raw, are those of the raw image --the colortype and bitdepth in the color field of LodePNGInfo info_png, are those of the PNG --when encoding, the color type in LodePNGInfo is ignored if auto_convert - is enabled, it is automatically generated instead --when decoding, the color type in LodePNGInfo is set by the decoder to that of the original - PNG image, but it can be ignored since the raw image has the color type you requested instead --if the color type of the LodePNGColorMode and PNG image aren't the same, a conversion - between the color types is done if the color types are supported. If it is not - supported, an error is returned. If the types are the same, no conversion is done. --even though some conversions aren't supported, LodePNG supports loading PNGs from any - colortype and saving PNGs to any colortype, sometimes it just requires preparing - the raw image correctly before encoding. --both encoder and decoder use the same color converter. - -The function lodepng_convert does the color conversion. It is available in the -interface but normally isn't needed since the encoder and decoder already call -it. - -Non supported color conversions: --color to grayscale when non-gray pixels are present: no error is thrown, but -the result will look ugly because only the red channel is taken (it assumes all -three channels are the same in this case so ignores green and blue). The reason -no error is given is to allow converting from three-channel grayscale images to -one-channel even if there are numerical imprecisions. --anything to palette when the palette does not have an exact match for a from-color -in it: in this case an error is thrown - -Supported color conversions: --anything to 8-bit RGB, 8-bit RGBA, 16-bit RGB, 16-bit RGBA --any gray or gray+alpha, to gray or gray+alpha --anything to a palette, as long as the palette has the requested colors in it --removing alpha channel --higher to smaller bitdepth, and vice versa - -If you want no color conversion to be done (e.g. for speed or control): --In the encoder, you can make it save a PNG with any color type by giving the -raw color mode and LodePNGInfo the same color mode, and setting auto_convert to -false. --In the decoder, you can make it store the pixel data in the same color type -as the PNG has, by setting the color_convert setting to false. Settings in -info_raw are then ignored. - -6.3. padding bits ------------------ - -In the PNG file format, if a less than 8-bit per pixel color type is used and the scanlines -have a bit amount that isn't a multiple of 8, then padding bits are used so that each -scanline starts at a fresh byte. But that is NOT true for the LodePNG raw input and output. -The raw input image you give to the encoder, and the raw output image you get from the decoder -will NOT have these padding bits, e.g. in the case of a 1-bit image with a width -of 7 pixels, the first pixel of the second scanline will the 8th bit of the first byte, -not the first bit of a new byte. - -6.4. A note about 16-bits per channel and endianness ----------------------------------------------------- - -LodePNG uses unsigned char arrays for 16-bit per channel colors too, just like -for any other color format. The 16-bit values are stored in big endian (most -significant byte first) in these arrays. This is the opposite order of the -little endian used by x86 CPU's. - -LodePNG always uses big endian because the PNG file format does so internally. -Conversions to other formats than PNG uses internally are not supported by -LodePNG on purpose, there are myriads of formats, including endianness of 16-bit -colors, the order in which you store R, G, B and A, and so on. Supporting and -converting to/from all that is outside the scope of LodePNG. - -This may mean that, depending on your use case, you may want to convert the big -endian output of LodePNG to little endian with a for loop. This is certainly not -always needed, many applications and libraries support big endian 16-bit colors -anyway, but it means you cannot simply cast the unsigned char* buffer to an -unsigned short* buffer on x86 CPUs. - - -7. error values ---------------- - -All functions in LodePNG that return an error code, return 0 if everything went -OK, or a non-zero code if there was an error. - -The meaning of the LodePNG error values can be retrieved with the function -lodepng_error_text: given the numerical error code, it returns a description -of the error in English as a string. - -Check the implementation of lodepng_error_text to see the meaning of each code. - -It is not recommended to use the numerical values to programmatically make -different decisions based on error types as the numbers are not guaranteed to -stay backwards compatible. They are for human consumption only. Programmatically -only 0 or non-0 matter. - - -8. chunks and PNG editing -------------------------- - -If you want to add extra chunks to a PNG you encode, or use LodePNG for a PNG -editor that should follow the rules about handling of unknown chunks, or if your -program is able to read other types of chunks than the ones handled by LodePNG, -then that's possible with the chunk functions of LodePNG. - -A PNG chunk has the following layout: - -4 bytes length -4 bytes type name -length bytes data -4 bytes CRC - -8.1. iterating through chunks ------------------------------ - -If you have a buffer containing the PNG image data, then the first chunk (the -IHDR chunk) starts at byte number 8 of that buffer. The first 8 bytes are the -signature of the PNG and are not part of a chunk. But if you start at byte 8 -then you have a chunk, and can check the following things of it. - -NOTE: none of these functions check for memory buffer boundaries. To avoid -exploits, always make sure the buffer contains all the data of the chunks. -When using lodepng_chunk_next, make sure the returned value is within the -allocated memory. - -unsigned lodepng_chunk_length(const unsigned char* chunk): - -Get the length of the chunk's data. The total chunk length is this length + 12. - -void lodepng_chunk_type(char type[5], const unsigned char* chunk): -unsigned char lodepng_chunk_type_equals(const unsigned char* chunk, const char* type): - -Get the type of the chunk or compare if it's a certain type - -unsigned char lodepng_chunk_critical(const unsigned char* chunk): -unsigned char lodepng_chunk_private(const unsigned char* chunk): -unsigned char lodepng_chunk_safetocopy(const unsigned char* chunk): - -Check if the chunk is critical in the PNG standard (only IHDR, PLTE, IDAT and IEND are). -Check if the chunk is private (public chunks are part of the standard, private ones not). -Check if the chunk is safe to copy. If it's not, then, when modifying data in a critical -chunk, unsafe to copy chunks of the old image may NOT be saved in the new one if your -program doesn't handle that type of unknown chunk. - -unsigned char* lodepng_chunk_data(unsigned char* chunk): -const unsigned char* lodepng_chunk_data_const(const unsigned char* chunk): - -Get a pointer to the start of the data of the chunk. - -unsigned lodepng_chunk_check_crc(const unsigned char* chunk): -void lodepng_chunk_generate_crc(unsigned char* chunk): - -Check if the crc is correct or generate a correct one. - -unsigned char* lodepng_chunk_next(unsigned char* chunk): -const unsigned char* lodepng_chunk_next_const(const unsigned char* chunk): - -Iterate to the next chunk. This works if you have a buffer with consecutive chunks. Note that these -functions do no boundary checking of the allocated data whatsoever, so make sure there is enough -data available in the buffer to be able to go to the next chunk. - -unsigned lodepng_chunk_append(unsigned char** out, size_t* outsize, const unsigned char* chunk): -unsigned lodepng_chunk_create(unsigned char** out, size_t* outsize, unsigned length, - const char* type, const unsigned char* data): - -These functions are used to create new chunks that are appended to the data in *out that has -length *outsize. The append function appends an existing chunk to the new data. The create -function creates a new chunk with the given parameters and appends it. Type is the 4-letter -name of the chunk. - -8.2. chunks in info_png ------------------------ - -The LodePNGInfo struct contains fields with the unknown chunk in it. It has 3 -buffers (each with size) to contain 3 types of unknown chunks: -the ones that come before the PLTE chunk, the ones that come between the PLTE -and the IDAT chunks, and the ones that come after the IDAT chunks. -It's necessary to make the distinction between these 3 cases because the PNG -standard forces to keep the ordering of unknown chunks compared to the critical -chunks, but does not force any other ordering rules. - -info_png.unknown_chunks_data[0] is the chunks before PLTE -info_png.unknown_chunks_data[1] is the chunks after PLTE, before IDAT -info_png.unknown_chunks_data[2] is the chunks after IDAT - -The chunks in these 3 buffers can be iterated through and read by using the same -way described in the previous subchapter. - -When using the decoder to decode a PNG, you can make it store all unknown chunks -if you set the option settings.remember_unknown_chunks to 1. By default, this -option is off (0). - -The encoder will always encode unknown chunks that are stored in the info_png. -If you need it to add a particular chunk that isn't known by LodePNG, you can -use lodepng_chunk_append or lodepng_chunk_create to the chunk data in -info_png.unknown_chunks_data[x]. - -Chunks that are known by LodePNG should not be added in that way. E.g. to make -LodePNG add a bKGD chunk, set background_defined to true and add the correct -parameters there instead. - - -9. compiler support -------------------- - -No libraries other than the current standard C library are needed to compile -LodePNG. For the C++ version, only the standard C++ library is needed on top. -Add the files lodepng.c(pp) and lodepng.h to your project, include -lodepng.h where needed, and your program can read/write PNG files. - -It is compatible with C90 and up, and C++03 and up. - -If performance is important, use optimization when compiling! For both the -encoder and decoder, this makes a large difference. - -Make sure that LodePNG is compiled with the same compiler of the same version -and with the same settings as the rest of the program, or the interfaces with -std::vectors and std::strings in C++ can be incompatible. - -CHAR_BITS must be 8 or higher, because LodePNG uses unsigned chars for octets. - -*) gcc and g++ - -LodePNG is developed in gcc so this compiler is natively supported. It gives no -warnings with compiler options "-Wall -Wextra -pedantic -ansi", with gcc and g++ -version 4.7.1 on Linux, 32-bit and 64-bit. - -*) Clang - -Fully supported and warning-free. - -*) Mingw - -The Mingw compiler (a port of gcc for Windows) should be fully supported by -LodePNG. - -*) Visual Studio and Visual C++ Express Edition - -LodePNG should be warning-free with warning level W4. Two warnings were disabled -with pragmas though: warning 4244 about implicit conversions, and warning 4996 -where it wants to use a non-standard function fopen_s instead of the standard C -fopen. - -Visual Studio may want "stdafx.h" files to be included in each source file and -give an error "unexpected end of file while looking for precompiled header". -This is not standard C++ and will not be added to the stock LodePNG. You can -disable it for lodepng.cpp only by right clicking it, Properties, C/C++, -Precompiled Headers, and set it to Not Using Precompiled Headers there. - -NOTE: Modern versions of VS should be fully supported, but old versions, e.g. -VS6, are not guaranteed to work. - -*) Compilers on Macintosh - -LodePNG has been reported to work both with gcc and LLVM for Macintosh, both for -C and C++. - -*) Other Compilers - -If you encounter problems on any compilers, feel free to let me know and I may -try to fix it if the compiler is modern and standards compliant. - - -10. examples ------------- - -This decoder example shows the most basic usage of LodePNG. More complex -examples can be found on the LodePNG website. - -NOTE: these examples do not support wide-character filenames, you can use an -external method to handle such files and encode or decode in-memory - -10.1. decoder C++ example -------------------------- - -#include "lodepng.h" -#include <iostream> - -int main(int argc, char *argv[]) { - const char* filename = argc > 1 ? argv[1] : "test.png"; - - //load and decode - std::vector<unsigned char> image; - unsigned width, height; - unsigned error = lodepng::decode(image, width, height, filename); - - //if there's an error, display it - if(error) std::cout << "decoder error " << error << ": " << lodepng_error_text(error) << std::endl; - - //the pixels are now in the vector "image", 4 bytes per pixel, ordered RGBARGBA..., use it as texture, draw it, ... -} - -10.2. decoder C example ------------------------ - -#include "lodepng.h" - -int main(int argc, char *argv[]) { - unsigned error; - unsigned char* image; - size_t width, height; - const char* filename = argc > 1 ? argv[1] : "test.png"; - - error = lodepng_decode32_file(&image, &width, &height, filename); - - if(error) printf("decoder error %u: %s\n", error, lodepng_error_text(error)); - - / * use image here * / - - free(image); - return 0; -} - -11. state settings reference ----------------------------- - -A quick reference of some settings to set on the LodePNGState - -For decoding: - -state.decoder.zlibsettings.ignore_adler32: ignore ADLER32 checksums -state.decoder.zlibsettings.custom_...: use custom inflate function -state.decoder.ignore_crc: ignore CRC checksums -state.decoder.ignore_critical: ignore unknown critical chunks -state.decoder.ignore_end: ignore missing IEND chunk. May fail if this corruption causes other errors -state.decoder.color_convert: convert internal PNG color to chosen one -state.decoder.read_text_chunks: whether to read in text metadata chunks -state.decoder.remember_unknown_chunks: whether to read in unknown chunks -state.info_raw.colortype: desired color type for decoded image -state.info_raw.bitdepth: desired bit depth for decoded image -state.info_raw....: more color settings, see struct LodePNGColorMode -state.info_png....: no settings for decoder but ouput, see struct LodePNGInfo - -For encoding: - -state.encoder.zlibsettings.btype: disable compression by setting it to 0 -state.encoder.zlibsettings.use_lz77: use LZ77 in compression -state.encoder.zlibsettings.windowsize: tweak LZ77 windowsize -state.encoder.zlibsettings.minmatch: tweak min LZ77 length to match -state.encoder.zlibsettings.nicematch: tweak LZ77 match where to stop searching -state.encoder.zlibsettings.lazymatching: try one more LZ77 matching -state.encoder.zlibsettings.custom_...: use custom deflate function -state.encoder.auto_convert: choose optimal PNG color type, if 0 uses info_png -state.encoder.filter_palette_zero: PNG filter strategy for palette -state.encoder.filter_strategy: PNG filter strategy to encode with -state.encoder.force_palette: add palette even if not encoding to one -state.encoder.add_id: add LodePNG identifier and version as a text chunk -state.encoder.text_compression: use compressed text chunks for metadata -state.info_raw.colortype: color type of raw input image you provide -state.info_raw.bitdepth: bit depth of raw input image you provide -state.info_raw: more color settings, see struct LodePNGColorMode -state.info_png.color.colortype: desired color type if auto_convert is false -state.info_png.color.bitdepth: desired bit depth if auto_convert is false -state.info_png.color....: more color settings, see struct LodePNGColorMode -state.info_png....: more PNG related settings, see struct LodePNGInfo - - -12. changes ------------ - -The version number of LodePNG is the date of the change given in the format -yyyymmdd. - -Some changes aren't backwards compatible. Those are indicated with a (!) -symbol. - -Not all changes are listed here, the commit history in github lists more: -https://github.com/lvandeve/lodepng - -*) 10 apr 2023: faster CRC32 implementation, but with larger lookup table. -*) 13 jun 2022: added support for the sBIT chunk. -*) 09 jan 2022: minor decoder speed improvements. -*) 27 jun 2021: added warnings that file reading/writing functions don't support - wide-character filenames (support for this is not planned, opening files is - not the core part of PNG decoding/decoding and is platform dependent). -*) 17 okt 2020: prevent decoding too large text/icc chunks by default. -*) 06 mar 2020: simplified some of the dynamic memory allocations. -*) 12 jan 2020: (!) added 'end' argument to lodepng_chunk_next to allow correct - overflow checks. -*) 14 aug 2019: around 25% faster decoding thanks to huffman lookup tables. -*) 15 jun 2019: (!) auto_choose_color API changed (for bugfix: don't use palette - if gray ICC profile) and non-ICC LodePNGColorProfile renamed to - LodePNGColorStats. -*) 30 dec 2018: code style changes only: removed newlines before opening braces. -*) 10 sep 2018: added way to inspect metadata chunks without full decoding. -*) 19 aug 2018: (!) fixed color mode bKGD is encoded with and made it use - palette index in case of palette. -*) 10 aug 2018: (!) added support for gAMA, cHRM, sRGB and iCCP chunks. This - change is backwards compatible unless you relied on unknown_chunks for those. -*) 11 jun 2018: less restrictive check for pixel size integer overflow -*) 14 jan 2018: allow optionally ignoring a few more recoverable errors -*) 17 sep 2017: fix memory leak for some encoder input error cases -*) 27 nov 2016: grey+alpha auto color model detection bugfix -*) 18 apr 2016: Changed qsort to custom stable sort (for platforms w/o qsort). -*) 09 apr 2016: Fixed colorkey usage detection, and better file loading (within - the limits of pure C90). -*) 08 dec 2015: Made load_file function return error if file can't be opened. -*) 24 okt 2015: Bugfix with decoding to palette output. -*) 18 apr 2015: Boundary PM instead of just package-merge for faster encoding. -*) 24 aug 2014: Moved to github -*) 23 aug 2014: Reduced needless memory usage of decoder. -*) 28 jun 2014: Removed fix_png setting, always support palette OOB for - simplicity. Made ColorProfile public. -*) 09 jun 2014: Faster encoder by fixing hash bug and more zeros optimization. -*) 22 dec 2013: Power of two windowsize required for optimization. -*) 15 apr 2013: Fixed bug with LAC_ALPHA and color key. -*) 25 mar 2013: Added an optional feature to ignore some PNG errors (fix_png). -*) 11 mar 2013: (!) Bugfix with custom free. Changed from "my" to "lodepng_" - prefix for the custom allocators and made it possible with a new #define to - use custom ones in your project without needing to change lodepng's code. -*) 28 jan 2013: Bugfix with color key. -*) 27 okt 2012: Tweaks in text chunk keyword length error handling. -*) 8 okt 2012: (!) Added new filter strategy (entropy) and new auto color mode. - (no palette). Better deflate tree encoding. New compression tweak settings. - Faster color conversions while decoding. Some internal cleanups. -*) 23 sep 2012: Reduced warnings in Visual Studio a little bit. -*) 1 sep 2012: (!) Removed #define's for giving custom (de)compression functions - and made it work with function pointers instead. -*) 23 jun 2012: Added more filter strategies. Made it easier to use custom alloc - and free functions and toggle #defines from compiler flags. Small fixes. -*) 6 may 2012: (!) Made plugging in custom zlib/deflate functions more flexible. -*) 22 apr 2012: (!) Made interface more consistent, renaming a lot. Removed - redundant C++ codec classes. Reduced amount of structs. Everything changed, - but it is cleaner now imho and functionality remains the same. Also fixed - several bugs and shrunk the implementation code. Made new samples. -*) 6 nov 2011: (!) By default, the encoder now automatically chooses the best - PNG color model and bit depth, based on the amount and type of colors of the - raw image. For this, autoLeaveOutAlphaChannel replaced by auto_choose_color. -*) 9 okt 2011: simpler hash chain implementation for the encoder. -*) 8 sep 2011: lz77 encoder lazy matching instead of greedy matching. -*) 23 aug 2011: tweaked the zlib compression parameters after benchmarking. - A bug with the PNG filtertype heuristic was fixed, so that it chooses much - better ones (it's quite significant). A setting to do an experimental, slow, - brute force search for PNG filter types is added. -*) 17 aug 2011: (!) changed some C zlib related function names. -*) 16 aug 2011: made the code less wide (max 120 characters per line). -*) 17 apr 2011: code cleanup. Bugfixes. Convert low to 16-bit per sample colors. -*) 21 feb 2011: fixed compiling for C90. Fixed compiling with sections disabled. -*) 11 dec 2010: encoding is made faster, based on suggestion by Peter Eastman - to optimize long sequences of zeros. -*) 13 nov 2010: added LodePNG_InfoColor_hasPaletteAlpha and - LodePNG_InfoColor_canHaveAlpha functions for convenience. -*) 7 nov 2010: added LodePNG_error_text function to get error code description. -*) 30 okt 2010: made decoding slightly faster -*) 26 okt 2010: (!) changed some C function and struct names (more consistent). - Reorganized the documentation and the declaration order in the header. -*) 08 aug 2010: only changed some comments and external samples. -*) 05 jul 2010: fixed bug thanks to warnings in the new gcc version. -*) 14 mar 2010: fixed bug where too much memory was allocated for char buffers. -*) 02 sep 2008: fixed bug where it could create empty tree that linux apps could - read by ignoring the problem but windows apps couldn't. -*) 06 jun 2008: added more error checks for out of memory cases. -*) 26 apr 2008: added a few more checks here and there to ensure more safety. -*) 06 mar 2008: crash with encoding of strings fixed -*) 02 feb 2008: support for international text chunks added (iTXt) -*) 23 jan 2008: small cleanups, and #defines to divide code in sections -*) 20 jan 2008: support for unknown chunks allowing using LodePNG for an editor. -*) 18 jan 2008: support for tIME and pHYs chunks added to encoder and decoder. -*) 17 jan 2008: ability to encode and decode compressed zTXt chunks added - Also various fixes, such as in the deflate and the padding bits code. -*) 13 jan 2008: Added ability to encode Adam7-interlaced images. Improved - filtering code of encoder. -*) 07 jan 2008: (!) changed LodePNG to use ISO C90 instead of C++. A - C++ wrapper around this provides an interface almost identical to before. - Having LodePNG be pure ISO C90 makes it more portable. The C and C++ code - are together in these files but it works both for C and C++ compilers. -*) 29 dec 2007: (!) changed most integer types to unsigned int + other tweaks -*) 30 aug 2007: bug fixed which makes this Borland C++ compatible -*) 09 aug 2007: some VS2005 warnings removed again -*) 21 jul 2007: deflate code placed in new namespace separate from zlib code -*) 08 jun 2007: fixed bug with 2- and 4-bit color, and small interlaced images -*) 04 jun 2007: improved support for Visual Studio 2005: crash with accessing - invalid std::vector element [0] fixed, and level 3 and 4 warnings removed -*) 02 jun 2007: made the encoder add a tag with version by default -*) 27 may 2007: zlib and png code separated (but still in the same file), - simple encoder/decoder functions added for more simple usage cases -*) 19 may 2007: minor fixes, some code cleaning, new error added (error 69), - moved some examples from here to lodepng_examples.cpp -*) 12 may 2007: palette decoding bug fixed -*) 24 apr 2007: changed the license from BSD to the zlib license -*) 11 mar 2007: very simple addition: ability to encode bKGD chunks. -*) 04 mar 2007: (!) tEXt chunk related fixes, and support for encoding - palettized PNG images. Plus little interface change with palette and texts. -*) 03 mar 2007: Made it encode dynamic Huffman shorter with repeat codes. - Fixed a bug where the end code of a block had length 0 in the Huffman tree. -*) 26 feb 2007: Huffman compression with dynamic trees (BTYPE 2) now implemented - and supported by the encoder, resulting in smaller PNGs at the output. -*) 27 jan 2007: Made the Adler-32 test faster so that a timewaste is gone. -*) 24 jan 2007: gave encoder an error interface. Added color conversion from any - greyscale type to 8-bit greyscale with or without alpha. -*) 21 jan 2007: (!) Totally changed the interface. It allows more color types - to convert to and is more uniform. See the manual for how it works now. -*) 07 jan 2007: Some cleanup & fixes, and a few changes over the last days: - encode/decode custom tEXt chunks, separate classes for zlib & deflate, and - at last made the decoder give errors for incorrect Adler32 or Crc. -*) 01 jan 2007: Fixed bug with encoding PNGs with less than 8 bits per channel. -*) 29 dec 2006: Added support for encoding images without alpha channel, and - cleaned out code as well as making certain parts faster. -*) 28 dec 2006: Added "Settings" to the encoder. -*) 26 dec 2006: The encoder now does LZ77 encoding and produces much smaller files now. - Removed some code duplication in the decoder. Fixed little bug in an example. -*) 09 dec 2006: (!) Placed output parameters of public functions as first parameter. - Fixed a bug of the decoder with 16-bit per color. -*) 15 okt 2006: Changed documentation structure -*) 09 okt 2006: Encoder class added. It encodes a valid PNG image from the - given image buffer, however for now it's not compressed. -*) 08 sep 2006: (!) Changed to interface with a Decoder class -*) 30 jul 2006: (!) LodePNG_InfoPng , width and height are now retrieved in different - way. Renamed decodePNG to decodePNGGeneric. -*) 29 jul 2006: (!) Changed the interface: image info is now returned as a - struct of type LodePNG::LodePNG_Info, instead of a vector, which was a bit clumsy. -*) 28 jul 2006: Cleaned the code and added new error checks. - Corrected terminology "deflate" into "inflate". -*) 23 jun 2006: Added SDL example in the documentation in the header, this - example allows easy debugging by displaying the PNG and its transparency. -*) 22 jun 2006: (!) Changed way to obtain error value. Added - loadFile function for convenience. Made decodePNG32 faster. -*) 21 jun 2006: (!) Changed type of info vector to unsigned. - Changed position of palette in info vector. Fixed an important bug that - happened on PNGs with an uncompressed block. -*) 16 jun 2006: Internally changed unsigned into unsigned where - needed, and performed some optimizations. -*) 07 jun 2006: (!) Renamed functions to decodePNG and placed them - in LodePNG namespace. Changed the order of the parameters. Rewrote the - documentation in the header. Renamed files to lodepng.cpp and lodepng.h -*) 22 apr 2006: Optimized and improved some code -*) 07 sep 2005: (!) Changed to std::vector interface -*) 12 aug 2005: Initial release (C++, decoder only) - - -13. contact information ------------------------ - -Feel free to contact me with suggestions, problems, comments, ... concerning -LodePNG. If you encounter a PNG image that doesn't work properly with this -decoder, feel free to send it and I'll use it to find and fix the problem. - -My email address is (puzzle the account and domain together with an @ symbol): -Domain: gmail dot com. -Account: lode dot vandevenne. - - -Copyright (c) 2005-2022 Lode Vandevenne -*/
--- a/semicongine/gltf.nim Sun Jul 28 23:18:05 2024 +0700 +++ b/semicongine/gltf.nim Mon Jul 29 00:00:28 2024 +0700 @@ -150,7 +150,7 @@ if root["images"][imageIndex].hasKey("uri"): raise newException(Exception, "Unsupported feature: Cannot load images from external files") let imageType = root["images"][imageIndex]["mimeType"].getStr() - assert imageType == "image/png", "glTF loader currently only supports PNG, but found '" & imageType & "'" + # assert imageType == "image/png", "glTF loader currently only supports PNG, but found '" & imageType & "'" let bufferView = root["bufferViews"][root["images"][imageIndex][ "bufferView"].getInt()]
--- a/semicongine/image.nim Sun Jul 28 23:18:05 2024 +0700 +++ b/semicongine/image.nim Mon Jul 29 00:00:28 2024 +0700 @@ -1,3 +1,15 @@ +{.emit: "#define STB_IMAGE_STATIC".} +{.emit: "#define STB_IMAGE_IMPLEMENTATION".} +{.emit: "#include \"" & currentSourcePath.parentDir() & "/thirdparty/stb/stb_image.h\"".} + +proc stbi_load_from_memory( + buffer: ptr uint8, + len: cint, + x, y: ptr cint, + channels_in_file: ptr cint, + desired_channels: cint +): ptr uint8 {.importc, nodecl.} + type Gray* = TVec1[uint8] BGRA* = TVec4[uint8] @@ -18,18 +30,26 @@ proc LoadImageData*[T: PixelType](pngData: string|seq[uint8]): Image[T] = when T is Gray: - let pngType = 0.cint + let nChannels = 1.cint elif T is BGRA: - let pngType = 6.cint + let nChannels = 4.cint + + var w, h, c: cint - var w, h: cuint - var data: cstring - - if lodepng_decode_memory(out_data = addr(data), w = addr(w), h = addr(h), in_data = cast[cstring](pngData.ToCPointer), insize = csize_t(pngData.len), colorType = pngType, bitdepth = 8) != 0: + let data = stbi_load_from_memory( + buffer = cast[ptr uint8](pngData.ToCPointer), + len = pngData.len.cint, + x = addr(w), + y = addr(h), + channels_in_file = addr(c), + desired_channels = nChannels + ) + # if lodepng_decode_memory(out_data = addr(data), w = addr(w), h = addr(h), in_data = cast[cstring](pngData.ToCPointer), insize = csize_t(pngData.len), colorType = pngType, bitdepth = 8) != 0: + if data == nil: raise newException(Exception, "An error occured while loading PNG file") let imagesize = w * h * 4 - result = Image[T](width: w, height: h, data: newSeq[T](w * h)) + result = Image[T](width: w.uint32, height: h.uint32, data: newSeq[T](w * h)) copyMem(result.data.ToCPointer, data, imagesize) nativeFree(data) @@ -46,7 +66,11 @@ result = LoadImageData[T](loadResource_intern(path, package = package).readAll()) - +# stb_image.h has no encoding support, maybe check stb_image_write or similar +# +# proc lodepng_encode_memory(out_data: ptr cstring, outsize: ptr csize_t, image: cstring, w: cuint, h: cuint, colorType: cint, bitdepth: cuint): cuint {.importc.} +# +#[ proc toPNG[T: PixelType](image: Image[T]): seq[uint8] = when T is Gray: let pngType = 0 # hardcoded in lodepng.h @@ -79,3 +103,4 @@ let written = f.writeBytes(data, 0, data.len) assert written == data.len, &"There was an error while saving '{filename}': only {written} of {data.len} bytes were written" f.close() +]#
--- a/semicongine/resources.nim Sun Jul 28 23:18:05 2024 +0700 +++ b/semicongine/resources.nim Mon Jul 29 00:00:28 2024 +0700 @@ -147,21 +147,6 @@ proc LoadConfig*(path: string, package = DEFAULT_PACKAGE): Config = path.loadResource_intern(package = package).loadConfig(filename = path) -# PNG-images - -{.compile: currentSourcePath.parentDir() & "/contrib/lodepng.c".} -proc lodepng_decode_memory(out_data: ptr cstring, w: ptr cuint, h: ptr cuint, in_data: cstring, insize: csize_t, colorType: cint, bitdepth: cuint): cuint {.importc.} - -proc lodepng_encode_memory(out_data: ptr cstring, outsize: ptr csize_t, image: cstring, w: cuint, h: cuint, colorType: cint, bitdepth: cuint): cuint {.importc.} - -#[ -proc LoadMeshes*(path: string, defaultMaterial: MaterialType, package = DEFAULT_PACKAGE): seq[MeshTree] = - loadResource_intern(path, package = package).ReadglTF(defaultMaterial) - -proc LoadFirstMesh*(path: string, defaultMaterial: MaterialType, package = DEFAULT_PACKAGE): Mesh = - loadResource_intern(path, package = package).ReadglTF(defaultMaterial)[0].toSeq[0] -]# - proc Packages*(): seq[string] = modList_intern()
--- a/semicongine/text/font.nim Sun Jul 28 23:18:05 2024 +0700 +++ b/semicongine/text/font.nim Mon Jul 29 00:00:28 2024 +0700 @@ -1,6 +1,6 @@ {.emit: "#define STBTT_STATIC".} {.emit: "#define STB_TRUETYPE_IMPLEMENTATION".} -{.emit: "#include \"" & currentSourcePath.parentDir() & "/stb_truetype.h\"".} +{.emit: "#include \"" & currentSourcePath.parentDir().parentDir() & "/thirdparty/stb/stb_truetype.h\"".} type stbtt_fontinfo {.importc, incompleteStruct.} = object
--- a/semicongine/text/stb_truetype.h Sun Jul 28 23:18:05 2024 +0700 +++ /dev/null Thu Jan 01 00:00:00 1970 +0000 @@ -1,5077 +0,0 @@ -// stb_truetype.h - v1.26 - public domain -// authored from 2009-2021 by Sean Barrett / RAD Game Tools -// -// ======================================================================= -// -// NO SECURITY GUARANTEE -- DO NOT USE THIS ON UNTRUSTED FONT FILES -// -// This library does no range checking of the offsets found in the file, -// meaning an attacker can use it to read arbitrary memory. -// -// ======================================================================= -// -// This library processes TrueType files: -// parse files -// extract glyph metrics -// extract glyph shapes -// render glyphs to one-channel bitmaps with antialiasing (box filter) -// render glyphs to one-channel SDF bitmaps (signed-distance field/function) -// -// Todo: -// non-MS cmaps -// crashproof on bad data -// hinting? (no longer patented) -// cleartype-style AA? -// optimize: use simple memory allocator for intermediates -// optimize: build edge-list directly from curves -// optimize: rasterize directly from curves? -// -// ADDITIONAL CONTRIBUTORS -// -// Mikko Mononen: compound shape support, more cmap formats -// Tor Andersson: kerning, subpixel rendering -// Dougall Johnson: OpenType / Type 2 font handling -// Daniel Ribeiro Maciel: basic GPOS-based kerning -// -// Misc other: -// Ryan Gordon -// Simon Glass -// github:IntellectualKitty -// Imanol Celaya -// Daniel Ribeiro Maciel -// -// Bug/warning reports/fixes: -// "Zer" on mollyrocket Fabian "ryg" Giesen github:NiLuJe -// Cass Everitt Martins Mozeiko github:aloucks -// stoiko (Haemimont Games) Cap Petschulat github:oyvindjam -// Brian Hook Omar Cornut github:vassvik -// Walter van Niftrik Ryan Griege -// David Gow Peter LaValle -// David Given Sergey Popov -// Ivan-Assen Ivanov Giumo X. Clanjor -// Anthony Pesch Higor Euripedes -// Johan Duparc Thomas Fields -// Hou Qiming Derek Vinyard -// Rob Loach Cort Stratton -// Kenney Phillis Jr. Brian Costabile -// Ken Voskuil (kaesve) -// -// VERSION HISTORY -// -// 1.26 (2021-08-28) fix broken rasterizer -// 1.25 (2021-07-11) many fixes -// 1.24 (2020-02-05) fix warning -// 1.23 (2020-02-02) query SVG data for glyphs; query whole kerning table (but only kern not GPOS) -// 1.22 (2019-08-11) minimize missing-glyph duplication; fix kerning if both 'GPOS' and 'kern' are defined -// 1.21 (2019-02-25) fix warning -// 1.20 (2019-02-07) PackFontRange skips missing codepoints; GetScaleFontVMetrics() -// 1.19 (2018-02-11) GPOS kerning, STBTT_fmod -// 1.18 (2018-01-29) add missing function -// 1.17 (2017-07-23) make more arguments const; doc fix -// 1.16 (2017-07-12) SDF support -// 1.15 (2017-03-03) make more arguments const -// 1.14 (2017-01-16) num-fonts-in-TTC function -// 1.13 (2017-01-02) support OpenType fonts, certain Apple fonts -// 1.12 (2016-10-25) suppress warnings about casting away const with -Wcast-qual -// 1.11 (2016-04-02) fix unused-variable warning -// 1.10 (2016-04-02) user-defined fabs(); rare memory leak; remove duplicate typedef -// 1.09 (2016-01-16) warning fix; avoid crash on outofmem; use allocation userdata properly -// 1.08 (2015-09-13) document stbtt_Rasterize(); fixes for vertical & horizontal edges -// 1.07 (2015-08-01) allow PackFontRanges to accept arrays of sparse codepoints; -// variant PackFontRanges to pack and render in separate phases; -// fix stbtt_GetFontOFfsetForIndex (never worked for non-0 input?); -// fixed an assert() bug in the new rasterizer -// replace assert() with STBTT_assert() in new rasterizer -// -// Full history can be found at the end of this file. -// -// LICENSE -// -// See end of file for license information. -// -// USAGE -// -// Include this file in whatever places need to refer to it. In ONE C/C++ -// file, write: -// #define STB_TRUETYPE_IMPLEMENTATION -// before the #include of this file. This expands out the actual -// implementation into that C/C++ file. -// -// To make the implementation private to the file that generates the implementation, -// #define STBTT_STATIC -// -// Simple 3D API (don't ship this, but it's fine for tools and quick start) -// stbtt_BakeFontBitmap() -- bake a font to a bitmap for use as texture -// stbtt_GetBakedQuad() -- compute quad to draw for a given char -// -// Improved 3D API (more shippable): -// #include "stb_rect_pack.h" -- optional, but you really want it -// stbtt_PackBegin() -// stbtt_PackSetOversampling() -- for improved quality on small fonts -// stbtt_PackFontRanges() -- pack and renders -// stbtt_PackEnd() -// stbtt_GetPackedQuad() -// -// "Load" a font file from a memory buffer (you have to keep the buffer loaded) -// stbtt_InitFont() -// stbtt_GetFontOffsetForIndex() -- indexing for TTC font collections -// stbtt_GetNumberOfFonts() -- number of fonts for TTC font collections -// -// Render a unicode codepoint to a bitmap -// stbtt_GetCodepointBitmap() -- allocates and returns a bitmap -// stbtt_MakeCodepointBitmap() -- renders into bitmap you provide -// stbtt_GetCodepointBitmapBox() -- how big the bitmap must be -// -// Character advance/positioning -// stbtt_GetCodepointHMetrics() -// stbtt_GetFontVMetrics() -// stbtt_GetFontVMetricsOS2() -// stbtt_GetCodepointKernAdvance() -// -// Starting with version 1.06, the rasterizer was replaced with a new, -// faster and generally-more-precise rasterizer. The new rasterizer more -// accurately measures pixel coverage for anti-aliasing, except in the case -// where multiple shapes overlap, in which case it overestimates the AA pixel -// coverage. Thus, anti-aliasing of intersecting shapes may look wrong. If -// this turns out to be a problem, you can re-enable the old rasterizer with -// #define STBTT_RASTERIZER_VERSION 1 -// which will incur about a 15% speed hit. -// -// ADDITIONAL DOCUMENTATION -// -// Immediately after this block comment are a series of sample programs. -// -// After the sample programs is the "header file" section. This section -// includes documentation for each API function. -// -// Some important concepts to understand to use this library: -// -// Codepoint -// Characters are defined by unicode codepoints, e.g. 65 is -// uppercase A, 231 is lowercase c with a cedilla, 0x7e30 is -// the hiragana for "ma". -// -// Glyph -// A visual character shape (every codepoint is rendered as -// some glyph) -// -// Glyph index -// A font-specific integer ID representing a glyph -// -// Baseline -// Glyph shapes are defined relative to a baseline, which is the -// bottom of uppercase characters. Characters extend both above -// and below the baseline. -// -// Current Point -// As you draw text to the screen, you keep track of a "current point" -// which is the origin of each character. The current point's vertical -// position is the baseline. Even "baked fonts" use this model. -// -// Vertical Font Metrics -// The vertical qualities of the font, used to vertically position -// and space the characters. See docs for stbtt_GetFontVMetrics. -// -// Font Size in Pixels or Points -// The preferred interface for specifying font sizes in stb_truetype -// is to specify how tall the font's vertical extent should be in pixels. -// If that sounds good enough, skip the next paragraph. -// -// Most font APIs instead use "points", which are a common typographic -// measurement for describing font size, defined as 72 points per inch. -// stb_truetype provides a point API for compatibility. However, true -// "per inch" conventions don't make much sense on computer displays -// since different monitors have different number of pixels per -// inch. For example, Windows traditionally uses a convention that -// there are 96 pixels per inch, thus making 'inch' measurements have -// nothing to do with inches, and thus effectively defining a point to -// be 1.333 pixels. Additionally, the TrueType font data provides -// an explicit scale factor to scale a given font's glyphs to points, -// but the author has observed that this scale factor is often wrong -// for non-commercial fonts, thus making fonts scaled in points -// according to the TrueType spec incoherently sized in practice. -// -// DETAILED USAGE: -// -// Scale: -// Select how high you want the font to be, in points or pixels. -// Call ScaleForPixelHeight or ScaleForMappingEmToPixels to compute -// a scale factor SF that will be used by all other functions. -// -// Baseline: -// You need to select a y-coordinate that is the baseline of where -// your text will appear. Call GetFontBoundingBox to get the baseline-relative -// bounding box for all characters. SF*-y0 will be the distance in pixels -// that the worst-case character could extend above the baseline, so if -// you want the top edge of characters to appear at the top of the -// screen where y=0, then you would set the baseline to SF*-y0. -// -// Current point: -// Set the current point where the first character will appear. The -// first character could extend left of the current point; this is font -// dependent. You can either choose a current point that is the leftmost -// point and hope, or add some padding, or check the bounding box or -// left-side-bearing of the first character to be displayed and set -// the current point based on that. -// -// Displaying a character: -// Compute the bounding box of the character. It will contain signed values -// relative to <current_point, baseline>. I.e. if it returns x0,y0,x1,y1, -// then the character should be displayed in the rectangle from -// <current_point+SF*x0, baseline+SF*y0> to <current_point+SF*x1,baseline+SF*y1). -// -// Advancing for the next character: -// Call GlyphHMetrics, and compute 'current_point += SF * advance'. -// -// -// ADVANCED USAGE -// -// Quality: -// -// - Use the functions with Subpixel at the end to allow your characters -// to have subpixel positioning. Since the font is anti-aliased, not -// hinted, this is very import for quality. (This is not possible with -// baked fonts.) -// -// - Kerning is now supported, and if you're supporting subpixel rendering -// then kerning is worth using to give your text a polished look. -// -// Performance: -// -// - Convert Unicode codepoints to glyph indexes and operate on the glyphs; -// if you don't do this, stb_truetype is forced to do the conversion on -// every call. -// -// - There are a lot of memory allocations. We should modify it to take -// a temp buffer and allocate from the temp buffer (without freeing), -// should help performance a lot. -// -// NOTES -// -// The system uses the raw data found in the .ttf file without changing it -// and without building auxiliary data structures. This is a bit inefficient -// on little-endian systems (the data is big-endian), but assuming you're -// caching the bitmaps or glyph shapes this shouldn't be a big deal. -// -// It appears to be very hard to programmatically determine what font a -// given file is in a general way. I provide an API for this, but I don't -// recommend it. -// -// -// PERFORMANCE MEASUREMENTS FOR 1.06: -// -// 32-bit 64-bit -// Previous release: 8.83 s 7.68 s -// Pool allocations: 7.72 s 6.34 s -// Inline sort : 6.54 s 5.65 s -// New rasterizer : 5.63 s 5.00 s - -////////////////////////////////////////////////////////////////////////////// -////////////////////////////////////////////////////////////////////////////// -//// -//// SAMPLE PROGRAMS -//// -// -// Incomplete text-in-3d-api example, which draws quads properly aligned to be lossless. -// See "tests/truetype_demo_win32.c" for a complete version. -#if 0 -#define STB_TRUETYPE_IMPLEMENTATION // force following include to generate implementation -#include "stb_truetype.h" - -unsigned char ttf_buffer[1<<20]; -unsigned char temp_bitmap[512*512]; - -stbtt_bakedchar cdata[96]; // ASCII 32..126 is 95 glyphs -GLuint ftex; - -void my_stbtt_initfont(void) -{ - fread(ttf_buffer, 1, 1<<20, fopen("c:/windows/fonts/times.ttf", "rb")); - stbtt_BakeFontBitmap(ttf_buffer,0, 32.0, temp_bitmap,512,512, 32,96, cdata); // no guarantee this fits! - // can free ttf_buffer at this point - glGenTextures(1, &ftex); - glBindTexture(GL_TEXTURE_2D, ftex); - glTexImage2D(GL_TEXTURE_2D, 0, GL_ALPHA, 512,512, 0, GL_ALPHA, GL_UNSIGNED_BYTE, temp_bitmap); - // can free temp_bitmap at this point - glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR); -} - -void my_stbtt_print(float x, float y, char *text) -{ - // assume orthographic projection with units = screen pixels, origin at top left - glEnable(GL_BLEND); - glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA); - glEnable(GL_TEXTURE_2D); - glBindTexture(GL_TEXTURE_2D, ftex); - glBegin(GL_QUADS); - while (*text) { - if (*text >= 32 && *text < 128) { - stbtt_aligned_quad q; - stbtt_GetBakedQuad(cdata, 512,512, *text-32, &x,&y,&q,1);//1=opengl & d3d10+,0=d3d9 - glTexCoord2f(q.s0,q.t0); glVertex2f(q.x0,q.y0); - glTexCoord2f(q.s1,q.t0); glVertex2f(q.x1,q.y0); - glTexCoord2f(q.s1,q.t1); glVertex2f(q.x1,q.y1); - glTexCoord2f(q.s0,q.t1); glVertex2f(q.x0,q.y1); - } - ++text; - } - glEnd(); -} -#endif -// -// -////////////////////////////////////////////////////////////////////////////// -// -// Complete program (this compiles): get a single bitmap, print as ASCII art -// -#if 0 -#include <stdio.h> -#define STB_TRUETYPE_IMPLEMENTATION // force following include to generate implementation -#include "stb_truetype.h" - -char ttf_buffer[1<<25]; - -int main(int argc, char **argv) -{ - stbtt_fontinfo font; - unsigned char *bitmap; - int w,h,i,j,c = (argc > 1 ? atoi(argv[1]) : 'a'), s = (argc > 2 ? atoi(argv[2]) : 20); - - fread(ttf_buffer, 1, 1<<25, fopen(argc > 3 ? argv[3] : "c:/windows/fonts/arialbd.ttf", "rb")); - - stbtt_InitFont(&font, ttf_buffer, stbtt_GetFontOffsetForIndex(ttf_buffer,0)); - bitmap = stbtt_GetCodepointBitmap(&font, 0,stbtt_ScaleForPixelHeight(&font, s), c, &w, &h, 0,0); - - for (j=0; j < h; ++j) { - for (i=0; i < w; ++i) - putchar(" .:ioVM@"[bitmap[j*w+i]>>5]); - putchar('\n'); - } - return 0; -} -#endif -// -// Output: -// -// .ii. -// @@@@@@. -// V@Mio@@o -// :i. V@V -// :oM@@M -// :@@@MM@M -// @@o o@M -// :@@. M@M -// @@@o@@@@ -// :M@@V:@@. -// -////////////////////////////////////////////////////////////////////////////// -// -// Complete program: print "Hello World!" banner, with bugs -// -#if 0 -char buffer[24<<20]; -unsigned char screen[20][79]; - -int main(int arg, char **argv) -{ - stbtt_fontinfo font; - int i,j,ascent,baseline,ch=0; - float scale, xpos=2; // leave a little padding in case the character extends left - char *text = "Heljo World!"; // intentionally misspelled to show 'lj' brokenness - - fread(buffer, 1, 1000000, fopen("c:/windows/fonts/arialbd.ttf", "rb")); - stbtt_InitFont(&font, buffer, 0); - - scale = stbtt_ScaleForPixelHeight(&font, 15); - stbtt_GetFontVMetrics(&font, &ascent,0,0); - baseline = (int) (ascent*scale); - - while (text[ch]) { - int advance,lsb,x0,y0,x1,y1; - float x_shift = xpos - (float) floor(xpos); - stbtt_GetCodepointHMetrics(&font, text[ch], &advance, &lsb); - stbtt_GetCodepointBitmapBoxSubpixel(&font, text[ch], scale,scale,x_shift,0, &x0,&y0,&x1,&y1); - stbtt_MakeCodepointBitmapSubpixel(&font, &screen[baseline + y0][(int) xpos + x0], x1-x0,y1-y0, 79, scale,scale,x_shift,0, text[ch]); - // note that this stomps the old data, so where character boxes overlap (e.g. 'lj') it's wrong - // because this API is really for baking character bitmaps into textures. if you want to render - // a sequence of characters, you really need to render each bitmap to a temp buffer, then - // "alpha blend" that into the working buffer - xpos += (advance * scale); - if (text[ch+1]) - xpos += scale*stbtt_GetCodepointKernAdvance(&font, text[ch],text[ch+1]); - ++ch; - } - - for (j=0; j < 20; ++j) { - for (i=0; i < 78; ++i) - putchar(" .:ioVM@"[screen[j][i]>>5]); - putchar('\n'); - } - - return 0; -} -#endif - - -////////////////////////////////////////////////////////////////////////////// -////////////////////////////////////////////////////////////////////////////// -//// -//// INTEGRATION WITH YOUR CODEBASE -//// -//// The following sections allow you to supply alternate definitions -//// of C library functions used by stb_truetype, e.g. if you don't -//// link with the C runtime library. - -#ifdef STB_TRUETYPE_IMPLEMENTATION - // #define your own (u)stbtt_int8/16/32 before including to override this - #ifndef stbtt_uint8 - typedef unsigned char stbtt_uint8; - typedef signed char stbtt_int8; - typedef unsigned short stbtt_uint16; - typedef signed short stbtt_int16; - typedef unsigned int stbtt_uint32; - typedef signed int stbtt_int32; - #endif - - typedef char stbtt__check_size32[sizeof(stbtt_int32)==4 ? 1 : -1]; - typedef char stbtt__check_size16[sizeof(stbtt_int16)==2 ? 1 : -1]; - - // e.g. #define your own STBTT_ifloor/STBTT_iceil() to avoid math.h - #ifndef STBTT_ifloor - #include <math.h> - #define STBTT_ifloor(x) ((int) floor(x)) - #define STBTT_iceil(x) ((int) ceil(x)) - #endif - - #ifndef STBTT_sqrt - #include <math.h> - #define STBTT_sqrt(x) sqrt(x) - #define STBTT_pow(x,y) pow(x,y) - #endif - - #ifndef STBTT_fmod - #include <math.h> - #define STBTT_fmod(x,y) fmod(x,y) - #endif - - #ifndef STBTT_cos - #include <math.h> - #define STBTT_cos(x) cos(x) - #define STBTT_acos(x) acos(x) - #endif - - #ifndef STBTT_fabs - #include <math.h> - #define STBTT_fabs(x) fabs(x) - #endif - - // #define your own functions "STBTT_malloc" / "STBTT_free" to avoid malloc.h - #ifndef STBTT_malloc - #include <stdlib.h> - #define STBTT_malloc(x,u) ((void)(u),malloc(x)) - #define STBTT_free(x,u) ((void)(u),free(x)) - #endif - - #ifndef STBTT_assert - #include <assert.h> - #define STBTT_assert(x) assert(x) - #endif - - #ifndef STBTT_strlen - #include <string.h> - #define STBTT_strlen(x) strlen(x) - #endif - - #ifndef STBTT_memcpy - #include <string.h> - #define STBTT_memcpy memcpy - #define STBTT_memset memset - #endif -#endif - -/////////////////////////////////////////////////////////////////////////////// -/////////////////////////////////////////////////////////////////////////////// -//// -//// INTERFACE -//// -//// - -#ifndef __STB_INCLUDE_STB_TRUETYPE_H__ -#define __STB_INCLUDE_STB_TRUETYPE_H__ - -#ifdef STBTT_STATIC -#define STBTT_DEF static -#else -#define STBTT_DEF extern -#endif - -#ifdef __cplusplus -extern "C" { -#endif - -// private structure -typedef struct -{ - unsigned char *data; - int cursor; - int size; -} stbtt__buf; - -////////////////////////////////////////////////////////////////////////////// -// -// TEXTURE BAKING API -// -// If you use this API, you only have to call two functions ever. -// - -typedef struct -{ - unsigned short x0,y0,x1,y1; // coordinates of bbox in bitmap - float xoff,yoff,xadvance; -} stbtt_bakedchar; - -STBTT_DEF int stbtt_BakeFontBitmap(const unsigned char *data, int offset, // font location (use offset=0 for plain .ttf) - float pixel_height, // height of font in pixels - unsigned char *pixels, int pw, int ph, // bitmap to be filled in - int first_char, int num_chars, // characters to bake - stbtt_bakedchar *chardata); // you allocate this, it's num_chars long -// if return is positive, the first unused row of the bitmap -// if return is negative, returns the negative of the number of characters that fit -// if return is 0, no characters fit and no rows were used -// This uses a very crappy packing. - -typedef struct -{ - float x0,y0,s0,t0; // top-left - float x1,y1,s1,t1; // bottom-right -} stbtt_aligned_quad; - -STBTT_DEF void stbtt_GetBakedQuad(const stbtt_bakedchar *chardata, int pw, int ph, // same data as above - int char_index, // character to display - float *xpos, float *ypos, // pointers to current position in screen pixel space - stbtt_aligned_quad *q, // output: quad to draw - int opengl_fillrule); // true if opengl fill rule; false if DX9 or earlier -// Call GetBakedQuad with char_index = 'character - first_char', and it -// creates the quad you need to draw and advances the current position. -// -// The coordinate system used assumes y increases downwards. -// -// Characters will extend both above and below the current position; -// see discussion of "BASELINE" above. -// -// It's inefficient; you might want to c&p it and optimize it. - -STBTT_DEF void stbtt_GetScaledFontVMetrics(const unsigned char *fontdata, int index, float size, float *ascent, float *descent, float *lineGap); -// Query the font vertical metrics without having to create a font first. - - -////////////////////////////////////////////////////////////////////////////// -// -// NEW TEXTURE BAKING API -// -// This provides options for packing multiple fonts into one atlas, not -// perfectly but better than nothing. - -typedef struct -{ - unsigned short x0,y0,x1,y1; // coordinates of bbox in bitmap - float xoff,yoff,xadvance; - float xoff2,yoff2; -} stbtt_packedchar; - -typedef struct stbtt_pack_context stbtt_pack_context; -typedef struct stbtt_fontinfo stbtt_fontinfo; -#ifndef STB_RECT_PACK_VERSION -typedef struct stbrp_rect stbrp_rect; -#endif - -STBTT_DEF int stbtt_PackBegin(stbtt_pack_context *spc, unsigned char *pixels, int width, int height, int stride_in_bytes, int padding, void *alloc_context); -// Initializes a packing context stored in the passed-in stbtt_pack_context. -// Future calls using this context will pack characters into the bitmap passed -// in here: a 1-channel bitmap that is width * height. stride_in_bytes is -// the distance from one row to the next (or 0 to mean they are packed tightly -// together). "padding" is the amount of padding to leave between each -// character (normally you want '1' for bitmaps you'll use as textures with -// bilinear filtering). -// -// Returns 0 on failure, 1 on success. - -STBTT_DEF void stbtt_PackEnd (stbtt_pack_context *spc); -// Cleans up the packing context and frees all memory. - -#define STBTT_POINT_SIZE(x) (-(x)) - -STBTT_DEF int stbtt_PackFontRange(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, float font_size, - int first_unicode_char_in_range, int num_chars_in_range, stbtt_packedchar *chardata_for_range); -// Creates character bitmaps from the font_index'th font found in fontdata (use -// font_index=0 if you don't know what that is). It creates num_chars_in_range -// bitmaps for characters with unicode values starting at first_unicode_char_in_range -// and increasing. Data for how to render them is stored in chardata_for_range; -// pass these to stbtt_GetPackedQuad to get back renderable quads. -// -// font_size is the full height of the character from ascender to descender, -// as computed by stbtt_ScaleForPixelHeight. To use a point size as computed -// by stbtt_ScaleForMappingEmToPixels, wrap the point size in STBTT_POINT_SIZE() -// and pass that result as 'font_size': -// ..., 20 , ... // font max minus min y is 20 pixels tall -// ..., STBTT_POINT_SIZE(20), ... // 'M' is 20 pixels tall - -typedef struct -{ - float font_size; - int first_unicode_codepoint_in_range; // if non-zero, then the chars are continuous, and this is the first codepoint - int *array_of_unicode_codepoints; // if non-zero, then this is an array of unicode codepoints - int num_chars; - stbtt_packedchar *chardata_for_range; // output - unsigned char h_oversample, v_oversample; // don't set these, they're used internally -} stbtt_pack_range; - -STBTT_DEF int stbtt_PackFontRanges(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, stbtt_pack_range *ranges, int num_ranges); -// Creates character bitmaps from multiple ranges of characters stored in -// ranges. This will usually create a better-packed bitmap than multiple -// calls to stbtt_PackFontRange. Note that you can call this multiple -// times within a single PackBegin/PackEnd. - -STBTT_DEF void stbtt_PackSetOversampling(stbtt_pack_context *spc, unsigned int h_oversample, unsigned int v_oversample); -// Oversampling a font increases the quality by allowing higher-quality subpixel -// positioning, and is especially valuable at smaller text sizes. -// -// This function sets the amount of oversampling for all following calls to -// stbtt_PackFontRange(s) or stbtt_PackFontRangesGatherRects for a given -// pack context. The default (no oversampling) is achieved by h_oversample=1 -// and v_oversample=1. The total number of pixels required is -// h_oversample*v_oversample larger than the default; for example, 2x2 -// oversampling requires 4x the storage of 1x1. For best results, render -// oversampled textures with bilinear filtering. Look at the readme in -// stb/tests/oversample for information about oversampled fonts -// -// To use with PackFontRangesGather etc., you must set it before calls -// call to PackFontRangesGatherRects. - -STBTT_DEF void stbtt_PackSetSkipMissingCodepoints(stbtt_pack_context *spc, int skip); -// If skip != 0, this tells stb_truetype to skip any codepoints for which -// there is no corresponding glyph. If skip=0, which is the default, then -// codepoints without a glyph recived the font's "missing character" glyph, -// typically an empty box by convention. - -STBTT_DEF void stbtt_GetPackedQuad(const stbtt_packedchar *chardata, int pw, int ph, // same data as above - int char_index, // character to display - float *xpos, float *ypos, // pointers to current position in screen pixel space - stbtt_aligned_quad *q, // output: quad to draw - int align_to_integer); - -STBTT_DEF int stbtt_PackFontRangesGatherRects(stbtt_pack_context *spc, const stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects); -STBTT_DEF void stbtt_PackFontRangesPackRects(stbtt_pack_context *spc, stbrp_rect *rects, int num_rects); -STBTT_DEF int stbtt_PackFontRangesRenderIntoRects(stbtt_pack_context *spc, const stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects); -// Calling these functions in sequence is roughly equivalent to calling -// stbtt_PackFontRanges(). If you more control over the packing of multiple -// fonts, or if you want to pack custom data into a font texture, take a look -// at the source to of stbtt_PackFontRanges() and create a custom version -// using these functions, e.g. call GatherRects multiple times, -// building up a single array of rects, then call PackRects once, -// then call RenderIntoRects repeatedly. This may result in a -// better packing than calling PackFontRanges multiple times -// (or it may not). - -// this is an opaque structure that you shouldn't mess with which holds -// all the context needed from PackBegin to PackEnd. -struct stbtt_pack_context { - void *user_allocator_context; - void *pack_info; - int width; - int height; - int stride_in_bytes; - int padding; - int skip_missing; - unsigned int h_oversample, v_oversample; - unsigned char *pixels; - void *nodes; -}; - -////////////////////////////////////////////////////////////////////////////// -// -// FONT LOADING -// -// - -STBTT_DEF int stbtt_GetNumberOfFonts(const unsigned char *data); -// This function will determine the number of fonts in a font file. TrueType -// collection (.ttc) files may contain multiple fonts, while TrueType font -// (.ttf) files only contain one font. The number of fonts can be used for -// indexing with the previous function where the index is between zero and one -// less than the total fonts. If an error occurs, -1 is returned. - -STBTT_DEF int stbtt_GetFontOffsetForIndex(const unsigned char *data, int index); -// Each .ttf/.ttc file may have more than one font. Each font has a sequential -// index number starting from 0. Call this function to get the font offset for -// a given index; it returns -1 if the index is out of range. A regular .ttf -// file will only define one font and it always be at offset 0, so it will -// return '0' for index 0, and -1 for all other indices. - -// The following structure is defined publicly so you can declare one on -// the stack or as a global or etc, but you should treat it as opaque. -struct stbtt_fontinfo -{ - void * userdata; - unsigned char * data; // pointer to .ttf file - int fontstart; // offset of start of font - - int numGlyphs; // number of glyphs, needed for range checking - - int loca,head,glyf,hhea,hmtx,kern,gpos,svg; // table locations as offset from start of .ttf - int index_map; // a cmap mapping for our chosen character encoding - int indexToLocFormat; // format needed to map from glyph index to glyph - - stbtt__buf cff; // cff font data - stbtt__buf charstrings; // the charstring index - stbtt__buf gsubrs; // global charstring subroutines index - stbtt__buf subrs; // private charstring subroutines index - stbtt__buf fontdicts; // array of font dicts - stbtt__buf fdselect; // map from glyph to fontdict -}; - -STBTT_DEF int stbtt_InitFont(stbtt_fontinfo *info, const unsigned char *data, int offset); -// Given an offset into the file that defines a font, this function builds -// the necessary cached info for the rest of the system. You must allocate -// the stbtt_fontinfo yourself, and stbtt_InitFont will fill it out. You don't -// need to do anything special to free it, because the contents are pure -// value data with no additional data structures. Returns 0 on failure. - - -////////////////////////////////////////////////////////////////////////////// -// -// CHARACTER TO GLYPH-INDEX CONVERSIOn - -STBTT_DEF int stbtt_FindGlyphIndex(const stbtt_fontinfo *info, int unicode_codepoint); -// If you're going to perform multiple operations on the same character -// and you want a speed-up, call this function with the character you're -// going to process, then use glyph-based functions instead of the -// codepoint-based functions. -// Returns 0 if the character codepoint is not defined in the font. - - -////////////////////////////////////////////////////////////////////////////// -// -// CHARACTER PROPERTIES -// - -STBTT_DEF float stbtt_ScaleForPixelHeight(const stbtt_fontinfo *info, float pixels); -// computes a scale factor to produce a font whose "height" is 'pixels' tall. -// Height is measured as the distance from the highest ascender to the lowest -// descender; in other words, it's equivalent to calling stbtt_GetFontVMetrics -// and computing: -// scale = pixels / (ascent - descent) -// so if you prefer to measure height by the ascent only, use a similar calculation. - -STBTT_DEF float stbtt_ScaleForMappingEmToPixels(const stbtt_fontinfo *info, float pixels); -// computes a scale factor to produce a font whose EM size is mapped to -// 'pixels' tall. This is probably what traditional APIs compute, but -// I'm not positive. - -STBTT_DEF void stbtt_GetFontVMetrics(const stbtt_fontinfo *info, int *ascent, int *descent, int *lineGap); -// ascent is the coordinate above the baseline the font extends; descent -// is the coordinate below the baseline the font extends (i.e. it is typically negative) -// lineGap is the spacing between one row's descent and the next row's ascent... -// so you should advance the vertical position by "*ascent - *descent + *lineGap" -// these are expressed in unscaled coordinates, so you must multiply by -// the scale factor for a given size - -STBTT_DEF int stbtt_GetFontVMetricsOS2(const stbtt_fontinfo *info, int *typoAscent, int *typoDescent, int *typoLineGap); -// analogous to GetFontVMetrics, but returns the "typographic" values from the OS/2 -// table (specific to MS/Windows TTF files). -// -// Returns 1 on success (table present), 0 on failure. - -STBTT_DEF void stbtt_GetFontBoundingBox(const stbtt_fontinfo *info, int *x0, int *y0, int *x1, int *y1); -// the bounding box around all possible characters - -STBTT_DEF void stbtt_GetCodepointHMetrics(const stbtt_fontinfo *info, int codepoint, int *advanceWidth, int *leftSideBearing); -// leftSideBearing is the offset from the current horizontal position to the left edge of the character -// advanceWidth is the offset from the current horizontal position to the next horizontal position -// these are expressed in unscaled coordinates - -STBTT_DEF int stbtt_GetCodepointKernAdvance(const stbtt_fontinfo *info, int ch1, int ch2); -// an additional amount to add to the 'advance' value between ch1 and ch2 - -STBTT_DEF int stbtt_GetCodepointBox(const stbtt_fontinfo *info, int codepoint, int *x0, int *y0, int *x1, int *y1); -// Gets the bounding box of the visible part of the glyph, in unscaled coordinates - -STBTT_DEF void stbtt_GetGlyphHMetrics(const stbtt_fontinfo *info, int glyph_index, int *advanceWidth, int *leftSideBearing); -STBTT_DEF int stbtt_GetGlyphKernAdvance(const stbtt_fontinfo *info, int glyph1, int glyph2); -STBTT_DEF int stbtt_GetGlyphBox(const stbtt_fontinfo *info, int glyph_index, int *x0, int *y0, int *x1, int *y1); -// as above, but takes one or more glyph indices for greater efficiency - -typedef struct stbtt_kerningentry -{ - int glyph1; // use stbtt_FindGlyphIndex - int glyph2; - int advance; -} stbtt_kerningentry; - -STBTT_DEF int stbtt_GetKerningTableLength(const stbtt_fontinfo *info); -STBTT_DEF int stbtt_GetKerningTable(const stbtt_fontinfo *info, stbtt_kerningentry* table, int table_length); -// Retrieves a complete list of all of the kerning pairs provided by the font -// stbtt_GetKerningTable never writes more than table_length entries and returns how many entries it did write. -// The table will be sorted by (a.glyph1 == b.glyph1)?(a.glyph2 < b.glyph2):(a.glyph1 < b.glyph1) - -////////////////////////////////////////////////////////////////////////////// -// -// GLYPH SHAPES (you probably don't need these, but they have to go before -// the bitmaps for C declaration-order reasons) -// - -#ifndef STBTT_vmove // you can predefine these to use different values (but why?) - enum { - STBTT_vmove=1, - STBTT_vline, - STBTT_vcurve, - STBTT_vcubic - }; -#endif - -#ifndef stbtt_vertex // you can predefine this to use different values - // (we share this with other code at RAD) - #define stbtt_vertex_type short // can't use stbtt_int16 because that's not visible in the header file - typedef struct - { - stbtt_vertex_type x,y,cx,cy,cx1,cy1; - unsigned char type,padding; - } stbtt_vertex; -#endif - -STBTT_DEF int stbtt_IsGlyphEmpty(const stbtt_fontinfo *info, int glyph_index); -// returns non-zero if nothing is drawn for this glyph - -STBTT_DEF int stbtt_GetCodepointShape(const stbtt_fontinfo *info, int unicode_codepoint, stbtt_vertex **vertices); -STBTT_DEF int stbtt_GetGlyphShape(const stbtt_fontinfo *info, int glyph_index, stbtt_vertex **vertices); -// returns # of vertices and fills *vertices with the pointer to them -// these are expressed in "unscaled" coordinates -// -// The shape is a series of contours. Each one starts with -// a STBTT_moveto, then consists of a series of mixed -// STBTT_lineto and STBTT_curveto segments. A lineto -// draws a line from previous endpoint to its x,y; a curveto -// draws a quadratic bezier from previous endpoint to -// its x,y, using cx,cy as the bezier control point. - -STBTT_DEF void stbtt_FreeShape(const stbtt_fontinfo *info, stbtt_vertex *vertices); -// frees the data allocated above - -STBTT_DEF unsigned char *stbtt_FindSVGDoc(const stbtt_fontinfo *info, int gl); -STBTT_DEF int stbtt_GetCodepointSVG(const stbtt_fontinfo *info, int unicode_codepoint, const char **svg); -STBTT_DEF int stbtt_GetGlyphSVG(const stbtt_fontinfo *info, int gl, const char **svg); -// fills svg with the character's SVG data. -// returns data size or 0 if SVG not found. - -////////////////////////////////////////////////////////////////////////////// -// -// BITMAP RENDERING -// - -STBTT_DEF void stbtt_FreeBitmap(unsigned char *bitmap, void *userdata); -// frees the bitmap allocated below - -STBTT_DEF unsigned char *stbtt_GetCodepointBitmap(const stbtt_fontinfo *info, float scale_x, float scale_y, int codepoint, int *width, int *height, int *xoff, int *yoff); -// allocates a large-enough single-channel 8bpp bitmap and renders the -// specified character/glyph at the specified scale into it, with -// antialiasing. 0 is no coverage (transparent), 255 is fully covered (opaque). -// *width & *height are filled out with the width & height of the bitmap, -// which is stored left-to-right, top-to-bottom. -// -// xoff/yoff are the offset it pixel space from the glyph origin to the top-left of the bitmap - -STBTT_DEF unsigned char *stbtt_GetCodepointBitmapSubpixel(const stbtt_fontinfo *info, float scale_x, float scale_y, float shift_x, float shift_y, int codepoint, int *width, int *height, int *xoff, int *yoff); -// the same as stbtt_GetCodepoitnBitmap, but you can specify a subpixel -// shift for the character - -STBTT_DEF void stbtt_MakeCodepointBitmap(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, int codepoint); -// the same as stbtt_GetCodepointBitmap, but you pass in storage for the bitmap -// in the form of 'output', with row spacing of 'out_stride' bytes. the bitmap -// is clipped to out_w/out_h bytes. Call stbtt_GetCodepointBitmapBox to get the -// width and height and positioning info for it first. - -STBTT_DEF void stbtt_MakeCodepointBitmapSubpixel(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int codepoint); -// same as stbtt_MakeCodepointBitmap, but you can specify a subpixel -// shift for the character - -STBTT_DEF void stbtt_MakeCodepointBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int oversample_x, int oversample_y, float *sub_x, float *sub_y, int codepoint); -// same as stbtt_MakeCodepointBitmapSubpixel, but prefiltering -// is performed (see stbtt_PackSetOversampling) - -STBTT_DEF void stbtt_GetCodepointBitmapBox(const stbtt_fontinfo *font, int codepoint, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1); -// get the bbox of the bitmap centered around the glyph origin; so the -// bitmap width is ix1-ix0, height is iy1-iy0, and location to place -// the bitmap top left is (leftSideBearing*scale,iy0). -// (Note that the bitmap uses y-increases-down, but the shape uses -// y-increases-up, so CodepointBitmapBox and CodepointBox are inverted.) - -STBTT_DEF void stbtt_GetCodepointBitmapBoxSubpixel(const stbtt_fontinfo *font, int codepoint, float scale_x, float scale_y, float shift_x, float shift_y, int *ix0, int *iy0, int *ix1, int *iy1); -// same as stbtt_GetCodepointBitmapBox, but you can specify a subpixel -// shift for the character - -// the following functions are equivalent to the above functions, but operate -// on glyph indices instead of Unicode codepoints (for efficiency) -STBTT_DEF unsigned char *stbtt_GetGlyphBitmap(const stbtt_fontinfo *info, float scale_x, float scale_y, int glyph, int *width, int *height, int *xoff, int *yoff); -STBTT_DEF unsigned char *stbtt_GetGlyphBitmapSubpixel(const stbtt_fontinfo *info, float scale_x, float scale_y, float shift_x, float shift_y, int glyph, int *width, int *height, int *xoff, int *yoff); -STBTT_DEF void stbtt_MakeGlyphBitmap(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, int glyph); -STBTT_DEF void stbtt_MakeGlyphBitmapSubpixel(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int glyph); -STBTT_DEF void stbtt_MakeGlyphBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int oversample_x, int oversample_y, float *sub_x, float *sub_y, int glyph); -STBTT_DEF void stbtt_GetGlyphBitmapBox(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1); -STBTT_DEF void stbtt_GetGlyphBitmapBoxSubpixel(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y,float shift_x, float shift_y, int *ix0, int *iy0, int *ix1, int *iy1); - - -// @TODO: don't expose this structure -typedef struct -{ - int w,h,stride; - unsigned char *pixels; -} stbtt__bitmap; - -// rasterize a shape with quadratic beziers into a bitmap -STBTT_DEF void stbtt_Rasterize(stbtt__bitmap *result, // 1-channel bitmap to draw into - float flatness_in_pixels, // allowable error of curve in pixels - stbtt_vertex *vertices, // array of vertices defining shape - int num_verts, // number of vertices in above array - float scale_x, float scale_y, // scale applied to input vertices - float shift_x, float shift_y, // translation applied to input vertices - int x_off, int y_off, // another translation applied to input - int invert, // if non-zero, vertically flip shape - void *userdata); // context for to STBTT_MALLOC - -////////////////////////////////////////////////////////////////////////////// -// -// Signed Distance Function (or Field) rendering - -STBTT_DEF void stbtt_FreeSDF(unsigned char *bitmap, void *userdata); -// frees the SDF bitmap allocated below - -STBTT_DEF unsigned char * stbtt_GetGlyphSDF(const stbtt_fontinfo *info, float scale, int glyph, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff); -STBTT_DEF unsigned char * stbtt_GetCodepointSDF(const stbtt_fontinfo *info, float scale, int codepoint, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff); -// These functions compute a discretized SDF field for a single character, suitable for storing -// in a single-channel texture, sampling with bilinear filtering, and testing against -// larger than some threshold to produce scalable fonts. -// info -- the font -// scale -- controls the size of the resulting SDF bitmap, same as it would be creating a regular bitmap -// glyph/codepoint -- the character to generate the SDF for -// padding -- extra "pixels" around the character which are filled with the distance to the character (not 0), -// which allows effects like bit outlines -// onedge_value -- value 0-255 to test the SDF against to reconstruct the character (i.e. the isocontour of the character) -// pixel_dist_scale -- what value the SDF should increase by when moving one SDF "pixel" away from the edge (on the 0..255 scale) -// if positive, > onedge_value is inside; if negative, < onedge_value is inside -// width,height -- output height & width of the SDF bitmap (including padding) -// xoff,yoff -- output origin of the character -// return value -- a 2D array of bytes 0..255, width*height in size -// -// pixel_dist_scale & onedge_value are a scale & bias that allows you to make -// optimal use of the limited 0..255 for your application, trading off precision -// and special effects. SDF values outside the range 0..255 are clamped to 0..255. -// -// Example: -// scale = stbtt_ScaleForPixelHeight(22) -// padding = 5 -// onedge_value = 180 -// pixel_dist_scale = 180/5.0 = 36.0 -// -// This will create an SDF bitmap in which the character is about 22 pixels -// high but the whole bitmap is about 22+5+5=32 pixels high. To produce a filled -// shape, sample the SDF at each pixel and fill the pixel if the SDF value -// is greater than or equal to 180/255. (You'll actually want to antialias, -// which is beyond the scope of this example.) Additionally, you can compute -// offset outlines (e.g. to stroke the character border inside & outside, -// or only outside). For example, to fill outside the character up to 3 SDF -// pixels, you would compare against (180-36.0*3)/255 = 72/255. The above -// choice of variables maps a range from 5 pixels outside the shape to -// 2 pixels inside the shape to 0..255; this is intended primarily for apply -// outside effects only (the interior range is needed to allow proper -// antialiasing of the font at *smaller* sizes) -// -// The function computes the SDF analytically at each SDF pixel, not by e.g. -// building a higher-res bitmap and approximating it. In theory the quality -// should be as high as possible for an SDF of this size & representation, but -// unclear if this is true in practice (perhaps building a higher-res bitmap -// and computing from that can allow drop-out prevention). -// -// The algorithm has not been optimized at all, so expect it to be slow -// if computing lots of characters or very large sizes. - - - -////////////////////////////////////////////////////////////////////////////// -// -// Finding the right font... -// -// You should really just solve this offline, keep your own tables -// of what font is what, and don't try to get it out of the .ttf file. -// That's because getting it out of the .ttf file is really hard, because -// the names in the file can appear in many possible encodings, in many -// possible languages, and e.g. if you need a case-insensitive comparison, -// the details of that depend on the encoding & language in a complex way -// (actually underspecified in truetype, but also gigantic). -// -// But you can use the provided functions in two possible ways: -// stbtt_FindMatchingFont() will use *case-sensitive* comparisons on -// unicode-encoded names to try to find the font you want; -// you can run this before calling stbtt_InitFont() -// -// stbtt_GetFontNameString() lets you get any of the various strings -// from the file yourself and do your own comparisons on them. -// You have to have called stbtt_InitFont() first. - - -STBTT_DEF int stbtt_FindMatchingFont(const unsigned char *fontdata, const char *name, int flags); -// returns the offset (not index) of the font that matches, or -1 if none -// if you use STBTT_MACSTYLE_DONTCARE, use a font name like "Arial Bold". -// if you use any other flag, use a font name like "Arial"; this checks -// the 'macStyle' header field; i don't know if fonts set this consistently -#define STBTT_MACSTYLE_DONTCARE 0 -#define STBTT_MACSTYLE_BOLD 1 -#define STBTT_MACSTYLE_ITALIC 2 -#define STBTT_MACSTYLE_UNDERSCORE 4 -#define STBTT_MACSTYLE_NONE 8 // <= not same as 0, this makes us check the bitfield is 0 - -STBTT_DEF int stbtt_CompareUTF8toUTF16_bigendian(const char *s1, int len1, const char *s2, int len2); -// returns 1/0 whether the first string interpreted as utf8 is identical to -// the second string interpreted as big-endian utf16... useful for strings from next func - -STBTT_DEF const char *stbtt_GetFontNameString(const stbtt_fontinfo *font, int *length, int platformID, int encodingID, int languageID, int nameID); -// returns the string (which may be big-endian double byte, e.g. for unicode) -// and puts the length in bytes in *length. -// -// some of the values for the IDs are below; for more see the truetype spec: -// http://developer.apple.com/textfonts/TTRefMan/RM06/Chap6name.html -// http://www.microsoft.com/typography/otspec/name.htm - -enum { // platformID - STBTT_PLATFORM_ID_UNICODE =0, - STBTT_PLATFORM_ID_MAC =1, - STBTT_PLATFORM_ID_ISO =2, - STBTT_PLATFORM_ID_MICROSOFT =3 -}; - -enum { // encodingID for STBTT_PLATFORM_ID_UNICODE - STBTT_UNICODE_EID_UNICODE_1_0 =0, - STBTT_UNICODE_EID_UNICODE_1_1 =1, - STBTT_UNICODE_EID_ISO_10646 =2, - STBTT_UNICODE_EID_UNICODE_2_0_BMP=3, - STBTT_UNICODE_EID_UNICODE_2_0_FULL=4 -}; - -enum { // encodingID for STBTT_PLATFORM_ID_MICROSOFT - STBTT_MS_EID_SYMBOL =0, - STBTT_MS_EID_UNICODE_BMP =1, - STBTT_MS_EID_SHIFTJIS =2, - STBTT_MS_EID_UNICODE_FULL =10 -}; - -enum { // encodingID for STBTT_PLATFORM_ID_MAC; same as Script Manager codes - STBTT_MAC_EID_ROMAN =0, STBTT_MAC_EID_ARABIC =4, - STBTT_MAC_EID_JAPANESE =1, STBTT_MAC_EID_HEBREW =5, - STBTT_MAC_EID_CHINESE_TRAD =2, STBTT_MAC_EID_GREEK =6, - STBTT_MAC_EID_KOREAN =3, STBTT_MAC_EID_RUSSIAN =7 -}; - -enum { // languageID for STBTT_PLATFORM_ID_MICROSOFT; same as LCID... - // problematic because there are e.g. 16 english LCIDs and 16 arabic LCIDs - STBTT_MS_LANG_ENGLISH =0x0409, STBTT_MS_LANG_ITALIAN =0x0410, - STBTT_MS_LANG_CHINESE =0x0804, STBTT_MS_LANG_JAPANESE =0x0411, - STBTT_MS_LANG_DUTCH =0x0413, STBTT_MS_LANG_KOREAN =0x0412, - STBTT_MS_LANG_FRENCH =0x040c, STBTT_MS_LANG_RUSSIAN =0x0419, - STBTT_MS_LANG_GERMAN =0x0407, STBTT_MS_LANG_SPANISH =0x0409, - STBTT_MS_LANG_HEBREW =0x040d, STBTT_MS_LANG_SWEDISH =0x041D -}; - -enum { // languageID for STBTT_PLATFORM_ID_MAC - STBTT_MAC_LANG_ENGLISH =0 , STBTT_MAC_LANG_JAPANESE =11, - STBTT_MAC_LANG_ARABIC =12, STBTT_MAC_LANG_KOREAN =23, - STBTT_MAC_LANG_DUTCH =4 , STBTT_MAC_LANG_RUSSIAN =32, - STBTT_MAC_LANG_FRENCH =1 , STBTT_MAC_LANG_SPANISH =6 , - STBTT_MAC_LANG_GERMAN =2 , STBTT_MAC_LANG_SWEDISH =5 , - STBTT_MAC_LANG_HEBREW =10, STBTT_MAC_LANG_CHINESE_SIMPLIFIED =33, - STBTT_MAC_LANG_ITALIAN =3 , STBTT_MAC_LANG_CHINESE_TRAD =19 -}; - -#ifdef __cplusplus -} -#endif - -#endif // __STB_INCLUDE_STB_TRUETYPE_H__ - -/////////////////////////////////////////////////////////////////////////////// -/////////////////////////////////////////////////////////////////////////////// -//// -//// IMPLEMENTATION -//// -//// - -#ifdef STB_TRUETYPE_IMPLEMENTATION - -#ifndef STBTT_MAX_OVERSAMPLE -#define STBTT_MAX_OVERSAMPLE 8 -#endif - -#if STBTT_MAX_OVERSAMPLE > 255 -#error "STBTT_MAX_OVERSAMPLE cannot be > 255" -#endif - -typedef int stbtt__test_oversample_pow2[(STBTT_MAX_OVERSAMPLE & (STBTT_MAX_OVERSAMPLE-1)) == 0 ? 1 : -1]; - -#ifndef STBTT_RASTERIZER_VERSION -#define STBTT_RASTERIZER_VERSION 2 -#endif - -#ifdef _MSC_VER -#define STBTT__NOTUSED(v) (void)(v) -#else -#define STBTT__NOTUSED(v) (void)sizeof(v) -#endif - -////////////////////////////////////////////////////////////////////////// -// -// stbtt__buf helpers to parse data from file -// - -static stbtt_uint8 stbtt__buf_get8(stbtt__buf *b) -{ - if (b->cursor >= b->size) - return 0; - return b->data[b->cursor++]; -} - -static stbtt_uint8 stbtt__buf_peek8(stbtt__buf *b) -{ - if (b->cursor >= b->size) - return 0; - return b->data[b->cursor]; -} - -static void stbtt__buf_seek(stbtt__buf *b, int o) -{ - STBTT_assert(!(o > b->size || o < 0)); - b->cursor = (o > b->size || o < 0) ? b->size : o; -} - -static void stbtt__buf_skip(stbtt__buf *b, int o) -{ - stbtt__buf_seek(b, b->cursor + o); -} - -static stbtt_uint32 stbtt__buf_get(stbtt__buf *b, int n) -{ - stbtt_uint32 v = 0; - int i; - STBTT_assert(n >= 1 && n <= 4); - for (i = 0; i < n; i++) - v = (v << 8) | stbtt__buf_get8(b); - return v; -} - -static stbtt__buf stbtt__new_buf(const void *p, size_t size) -{ - stbtt__buf r; - STBTT_assert(size < 0x40000000); - r.data = (stbtt_uint8*) p; - r.size = (int) size; - r.cursor = 0; - return r; -} - -#define stbtt__buf_get16(b) stbtt__buf_get((b), 2) -#define stbtt__buf_get32(b) stbtt__buf_get((b), 4) - -static stbtt__buf stbtt__buf_range(const stbtt__buf *b, int o, int s) -{ - stbtt__buf r = stbtt__new_buf(NULL, 0); - if (o < 0 || s < 0 || o > b->size || s > b->size - o) return r; - r.data = b->data + o; - r.size = s; - return r; -} - -static stbtt__buf stbtt__cff_get_index(stbtt__buf *b) -{ - int count, start, offsize; - start = b->cursor; - count = stbtt__buf_get16(b); - if (count) { - offsize = stbtt__buf_get8(b); - STBTT_assert(offsize >= 1 && offsize <= 4); - stbtt__buf_skip(b, offsize * count); - stbtt__buf_skip(b, stbtt__buf_get(b, offsize) - 1); - } - return stbtt__buf_range(b, start, b->cursor - start); -} - -static stbtt_uint32 stbtt__cff_int(stbtt__buf *b) -{ - int b0 = stbtt__buf_get8(b); - if (b0 >= 32 && b0 <= 246) return b0 - 139; - else if (b0 >= 247 && b0 <= 250) return (b0 - 247)*256 + stbtt__buf_get8(b) + 108; - else if (b0 >= 251 && b0 <= 254) return -(b0 - 251)*256 - stbtt__buf_get8(b) - 108; - else if (b0 == 28) return stbtt__buf_get16(b); - else if (b0 == 29) return stbtt__buf_get32(b); - STBTT_assert(0); - return 0; -} - -static void stbtt__cff_skip_operand(stbtt__buf *b) { - int v, b0 = stbtt__buf_peek8(b); - STBTT_assert(b0 >= 28); - if (b0 == 30) { - stbtt__buf_skip(b, 1); - while (b->cursor < b->size) { - v = stbtt__buf_get8(b); - if ((v & 0xF) == 0xF || (v >> 4) == 0xF) - break; - } - } else { - stbtt__cff_int(b); - } -} - -static stbtt__buf stbtt__dict_get(stbtt__buf *b, int key) -{ - stbtt__buf_seek(b, 0); - while (b->cursor < b->size) { - int start = b->cursor, end, op; - while (stbtt__buf_peek8(b) >= 28) - stbtt__cff_skip_operand(b); - end = b->cursor; - op = stbtt__buf_get8(b); - if (op == 12) op = stbtt__buf_get8(b) | 0x100; - if (op == key) return stbtt__buf_range(b, start, end-start); - } - return stbtt__buf_range(b, 0, 0); -} - -static void stbtt__dict_get_ints(stbtt__buf *b, int key, int outcount, stbtt_uint32 *out) -{ - int i; - stbtt__buf operands = stbtt__dict_get(b, key); - for (i = 0; i < outcount && operands.cursor < operands.size; i++) - out[i] = stbtt__cff_int(&operands); -} - -static int stbtt__cff_index_count(stbtt__buf *b) -{ - stbtt__buf_seek(b, 0); - return stbtt__buf_get16(b); -} - -static stbtt__buf stbtt__cff_index_get(stbtt__buf b, int i) -{ - int count, offsize, start, end; - stbtt__buf_seek(&b, 0); - count = stbtt__buf_get16(&b); - offsize = stbtt__buf_get8(&b); - STBTT_assert(i >= 0 && i < count); - STBTT_assert(offsize >= 1 && offsize <= 4); - stbtt__buf_skip(&b, i*offsize); - start = stbtt__buf_get(&b, offsize); - end = stbtt__buf_get(&b, offsize); - return stbtt__buf_range(&b, 2+(count+1)*offsize+start, end - start); -} - -////////////////////////////////////////////////////////////////////////// -// -// accessors to parse data from file -// - -// on platforms that don't allow misaligned reads, if we want to allow -// truetype fonts that aren't padded to alignment, define ALLOW_UNALIGNED_TRUETYPE - -#define ttBYTE(p) (* (stbtt_uint8 *) (p)) -#define ttCHAR(p) (* (stbtt_int8 *) (p)) -#define ttFixed(p) ttLONG(p) - -static stbtt_uint16 ttUSHORT(stbtt_uint8 *p) { return p[0]*256 + p[1]; } -static stbtt_int16 ttSHORT(stbtt_uint8 *p) { return p[0]*256 + p[1]; } -static stbtt_uint32 ttULONG(stbtt_uint8 *p) { return (p[0]<<24) + (p[1]<<16) + (p[2]<<8) + p[3]; } -static stbtt_int32 ttLONG(stbtt_uint8 *p) { return (p[0]<<24) + (p[1]<<16) + (p[2]<<8) + p[3]; } - -#define stbtt_tag4(p,c0,c1,c2,c3) ((p)[0] == (c0) && (p)[1] == (c1) && (p)[2] == (c2) && (p)[3] == (c3)) -#define stbtt_tag(p,str) stbtt_tag4(p,str[0],str[1],str[2],str[3]) - -static int stbtt__isfont(stbtt_uint8 *font) -{ - // check the version number - if (stbtt_tag4(font, '1',0,0,0)) return 1; // TrueType 1 - if (stbtt_tag(font, "typ1")) return 1; // TrueType with type 1 font -- we don't support this! - if (stbtt_tag(font, "OTTO")) return 1; // OpenType with CFF - if (stbtt_tag4(font, 0,1,0,0)) return 1; // OpenType 1.0 - if (stbtt_tag(font, "true")) return 1; // Apple specification for TrueType fonts - return 0; -} - -// @OPTIMIZE: binary search -static stbtt_uint32 stbtt__find_table(stbtt_uint8 *data, stbtt_uint32 fontstart, const char *tag) -{ - stbtt_int32 num_tables = ttUSHORT(data+fontstart+4); - stbtt_uint32 tabledir = fontstart + 12; - stbtt_int32 i; - for (i=0; i < num_tables; ++i) { - stbtt_uint32 loc = tabledir + 16*i; - if (stbtt_tag(data+loc+0, tag)) - return ttULONG(data+loc+8); - } - return 0; -} - -static int stbtt_GetFontOffsetForIndex_internal(unsigned char *font_collection, int index) -{ - // if it's just a font, there's only one valid index - if (stbtt__isfont(font_collection)) - return index == 0 ? 0 : -1; - - // check if it's a TTC - if (stbtt_tag(font_collection, "ttcf")) { - // version 1? - if (ttULONG(font_collection+4) == 0x00010000 || ttULONG(font_collection+4) == 0x00020000) { - stbtt_int32 n = ttLONG(font_collection+8); - if (index >= n) - return -1; - return ttULONG(font_collection+12+index*4); - } - } - return -1; -} - -static int stbtt_GetNumberOfFonts_internal(unsigned char *font_collection) -{ - // if it's just a font, there's only one valid font - if (stbtt__isfont(font_collection)) - return 1; - - // check if it's a TTC - if (stbtt_tag(font_collection, "ttcf")) { - // version 1? - if (ttULONG(font_collection+4) == 0x00010000 || ttULONG(font_collection+4) == 0x00020000) { - return ttLONG(font_collection+8); - } - } - return 0; -} - -static stbtt__buf stbtt__get_subrs(stbtt__buf cff, stbtt__buf fontdict) -{ - stbtt_uint32 subrsoff = 0, private_loc[2] = { 0, 0 }; - stbtt__buf pdict; - stbtt__dict_get_ints(&fontdict, 18, 2, private_loc); - if (!private_loc[1] || !private_loc[0]) return stbtt__new_buf(NULL, 0); - pdict = stbtt__buf_range(&cff, private_loc[1], private_loc[0]); - stbtt__dict_get_ints(&pdict, 19, 1, &subrsoff); - if (!subrsoff) return stbtt__new_buf(NULL, 0); - stbtt__buf_seek(&cff, private_loc[1]+subrsoff); - return stbtt__cff_get_index(&cff); -} - -// since most people won't use this, find this table the first time it's needed -static int stbtt__get_svg(stbtt_fontinfo *info) -{ - stbtt_uint32 t; - if (info->svg < 0) { - t = stbtt__find_table(info->data, info->fontstart, "SVG "); - if (t) { - stbtt_uint32 offset = ttULONG(info->data + t + 2); - info->svg = t + offset; - } else { - info->svg = 0; - } - } - return info->svg; -} - -static int stbtt_InitFont_internal(stbtt_fontinfo *info, unsigned char *data, int fontstart) -{ - stbtt_uint32 cmap, t; - stbtt_int32 i,numTables; - - info->data = data; - info->fontstart = fontstart; - info->cff = stbtt__new_buf(NULL, 0); - - cmap = stbtt__find_table(data, fontstart, "cmap"); // required - info->loca = stbtt__find_table(data, fontstart, "loca"); // required - info->head = stbtt__find_table(data, fontstart, "head"); // required - info->glyf = stbtt__find_table(data, fontstart, "glyf"); // required - info->hhea = stbtt__find_table(data, fontstart, "hhea"); // required - info->hmtx = stbtt__find_table(data, fontstart, "hmtx"); // required - info->kern = stbtt__find_table(data, fontstart, "kern"); // not required - info->gpos = stbtt__find_table(data, fontstart, "GPOS"); // not required - - if (!cmap || !info->head || !info->hhea || !info->hmtx) - return 0; - if (info->glyf) { - // required for truetype - if (!info->loca) return 0; - } else { - // initialization for CFF / Type2 fonts (OTF) - stbtt__buf b, topdict, topdictidx; - stbtt_uint32 cstype = 2, charstrings = 0, fdarrayoff = 0, fdselectoff = 0; - stbtt_uint32 cff; - - cff = stbtt__find_table(data, fontstart, "CFF "); - if (!cff) return 0; - - info->fontdicts = stbtt__new_buf(NULL, 0); - info->fdselect = stbtt__new_buf(NULL, 0); - - // @TODO this should use size from table (not 512MB) - info->cff = stbtt__new_buf(data+cff, 512*1024*1024); - b = info->cff; - - // read the header - stbtt__buf_skip(&b, 2); - stbtt__buf_seek(&b, stbtt__buf_get8(&b)); // hdrsize - - // @TODO the name INDEX could list multiple fonts, - // but we just use the first one. - stbtt__cff_get_index(&b); // name INDEX - topdictidx = stbtt__cff_get_index(&b); - topdict = stbtt__cff_index_get(topdictidx, 0); - stbtt__cff_get_index(&b); // string INDEX - info->gsubrs = stbtt__cff_get_index(&b); - - stbtt__dict_get_ints(&topdict, 17, 1, &charstrings); - stbtt__dict_get_ints(&topdict, 0x100 | 6, 1, &cstype); - stbtt__dict_get_ints(&topdict, 0x100 | 36, 1, &fdarrayoff); - stbtt__dict_get_ints(&topdict, 0x100 | 37, 1, &fdselectoff); - info->subrs = stbtt__get_subrs(b, topdict); - - // we only support Type 2 charstrings - if (cstype != 2) return 0; - if (charstrings == 0) return 0; - - if (fdarrayoff) { - // looks like a CID font - if (!fdselectoff) return 0; - stbtt__buf_seek(&b, fdarrayoff); - info->fontdicts = stbtt__cff_get_index(&b); - info->fdselect = stbtt__buf_range(&b, fdselectoff, b.size-fdselectoff); - } - - stbtt__buf_seek(&b, charstrings); - info->charstrings = stbtt__cff_get_index(&b); - } - - t = stbtt__find_table(data, fontstart, "maxp"); - if (t) - info->numGlyphs = ttUSHORT(data+t+4); - else - info->numGlyphs = 0xffff; - - info->svg = -1; - - // find a cmap encoding table we understand *now* to avoid searching - // later. (todo: could make this installable) - // the same regardless of glyph. - numTables = ttUSHORT(data + cmap + 2); - info->index_map = 0; - for (i=0; i < numTables; ++i) { - stbtt_uint32 encoding_record = cmap + 4 + 8 * i; - // find an encoding we understand: - switch(ttUSHORT(data+encoding_record)) { - case STBTT_PLATFORM_ID_MICROSOFT: - switch (ttUSHORT(data+encoding_record+2)) { - case STBTT_MS_EID_UNICODE_BMP: - case STBTT_MS_EID_UNICODE_FULL: - // MS/Unicode - info->index_map = cmap + ttULONG(data+encoding_record+4); - break; - } - break; - case STBTT_PLATFORM_ID_UNICODE: - // Mac/iOS has these - // all the encodingIDs are unicode, so we don't bother to check it - info->index_map = cmap + ttULONG(data+encoding_record+4); - break; - } - } - if (info->index_map == 0) - return 0; - - info->indexToLocFormat = ttUSHORT(data+info->head + 50); - return 1; -} - -STBTT_DEF int stbtt_FindGlyphIndex(const stbtt_fontinfo *info, int unicode_codepoint) -{ - stbtt_uint8 *data = info->data; - stbtt_uint32 index_map = info->index_map; - - stbtt_uint16 format = ttUSHORT(data + index_map + 0); - if (format == 0) { // apple byte encoding - stbtt_int32 bytes = ttUSHORT(data + index_map + 2); - if (unicode_codepoint < bytes-6) - return ttBYTE(data + index_map + 6 + unicode_codepoint); - return 0; - } else if (format == 6) { - stbtt_uint32 first = ttUSHORT(data + index_map + 6); - stbtt_uint32 count = ttUSHORT(data + index_map + 8); - if ((stbtt_uint32) unicode_codepoint >= first && (stbtt_uint32) unicode_codepoint < first+count) - return ttUSHORT(data + index_map + 10 + (unicode_codepoint - first)*2); - return 0; - } else if (format == 2) { - STBTT_assert(0); // @TODO: high-byte mapping for japanese/chinese/korean - return 0; - } else if (format == 4) { // standard mapping for windows fonts: binary search collection of ranges - stbtt_uint16 segcount = ttUSHORT(data+index_map+6) >> 1; - stbtt_uint16 searchRange = ttUSHORT(data+index_map+8) >> 1; - stbtt_uint16 entrySelector = ttUSHORT(data+index_map+10); - stbtt_uint16 rangeShift = ttUSHORT(data+index_map+12) >> 1; - - // do a binary search of the segments - stbtt_uint32 endCount = index_map + 14; - stbtt_uint32 search = endCount; - - if (unicode_codepoint > 0xffff) - return 0; - - // they lie from endCount .. endCount + segCount - // but searchRange is the nearest power of two, so... - if (unicode_codepoint >= ttUSHORT(data + search + rangeShift*2)) - search += rangeShift*2; - - // now decrement to bias correctly to find smallest - search -= 2; - while (entrySelector) { - stbtt_uint16 end; - searchRange >>= 1; - end = ttUSHORT(data + search + searchRange*2); - if (unicode_codepoint > end) - search += searchRange*2; - --entrySelector; - } - search += 2; - - { - stbtt_uint16 offset, start, last; - stbtt_uint16 item = (stbtt_uint16) ((search - endCount) >> 1); - - start = ttUSHORT(data + index_map + 14 + segcount*2 + 2 + 2*item); - last = ttUSHORT(data + endCount + 2*item); - if (unicode_codepoint < start || unicode_codepoint > last) - return 0; - - offset = ttUSHORT(data + index_map + 14 + segcount*6 + 2 + 2*item); - if (offset == 0) - return (stbtt_uint16) (unicode_codepoint + ttSHORT(data + index_map + 14 + segcount*4 + 2 + 2*item)); - - return ttUSHORT(data + offset + (unicode_codepoint-start)*2 + index_map + 14 + segcount*6 + 2 + 2*item); - } - } else if (format == 12 || format == 13) { - stbtt_uint32 ngroups = ttULONG(data+index_map+12); - stbtt_int32 low,high; - low = 0; high = (stbtt_int32)ngroups; - // Binary search the right group. - while (low < high) { - stbtt_int32 mid = low + ((high-low) >> 1); // rounds down, so low <= mid < high - stbtt_uint32 start_char = ttULONG(data+index_map+16+mid*12); - stbtt_uint32 end_char = ttULONG(data+index_map+16+mid*12+4); - if ((stbtt_uint32) unicode_codepoint < start_char) - high = mid; - else if ((stbtt_uint32) unicode_codepoint > end_char) - low = mid+1; - else { - stbtt_uint32 start_glyph = ttULONG(data+index_map+16+mid*12+8); - if (format == 12) - return start_glyph + unicode_codepoint-start_char; - else // format == 13 - return start_glyph; - } - } - return 0; // not found - } - // @TODO - STBTT_assert(0); - return 0; -} - -STBTT_DEF int stbtt_GetCodepointShape(const stbtt_fontinfo *info, int unicode_codepoint, stbtt_vertex **vertices) -{ - return stbtt_GetGlyphShape(info, stbtt_FindGlyphIndex(info, unicode_codepoint), vertices); -} - -static void stbtt_setvertex(stbtt_vertex *v, stbtt_uint8 type, stbtt_int32 x, stbtt_int32 y, stbtt_int32 cx, stbtt_int32 cy) -{ - v->type = type; - v->x = (stbtt_int16) x; - v->y = (stbtt_int16) y; - v->cx = (stbtt_int16) cx; - v->cy = (stbtt_int16) cy; -} - -static int stbtt__GetGlyfOffset(const stbtt_fontinfo *info, int glyph_index) -{ - int g1,g2; - - STBTT_assert(!info->cff.size); - - if (glyph_index >= info->numGlyphs) return -1; // glyph index out of range - if (info->indexToLocFormat >= 2) return -1; // unknown index->glyph map format - - if (info->indexToLocFormat == 0) { - g1 = info->glyf + ttUSHORT(info->data + info->loca + glyph_index * 2) * 2; - g2 = info->glyf + ttUSHORT(info->data + info->loca + glyph_index * 2 + 2) * 2; - } else { - g1 = info->glyf + ttULONG (info->data + info->loca + glyph_index * 4); - g2 = info->glyf + ttULONG (info->data + info->loca + glyph_index * 4 + 4); - } - - return g1==g2 ? -1 : g1; // if length is 0, return -1 -} - -static int stbtt__GetGlyphInfoT2(const stbtt_fontinfo *info, int glyph_index, int *x0, int *y0, int *x1, int *y1); - -STBTT_DEF int stbtt_GetGlyphBox(const stbtt_fontinfo *info, int glyph_index, int *x0, int *y0, int *x1, int *y1) -{ - if (info->cff.size) { - stbtt__GetGlyphInfoT2(info, glyph_index, x0, y0, x1, y1); - } else { - int g = stbtt__GetGlyfOffset(info, glyph_index); - if (g < 0) return 0; - - if (x0) *x0 = ttSHORT(info->data + g + 2); - if (y0) *y0 = ttSHORT(info->data + g + 4); - if (x1) *x1 = ttSHORT(info->data + g + 6); - if (y1) *y1 = ttSHORT(info->data + g + 8); - } - return 1; -} - -STBTT_DEF int stbtt_GetCodepointBox(const stbtt_fontinfo *info, int codepoint, int *x0, int *y0, int *x1, int *y1) -{ - return stbtt_GetGlyphBox(info, stbtt_FindGlyphIndex(info,codepoint), x0,y0,x1,y1); -} - -STBTT_DEF int stbtt_IsGlyphEmpty(const stbtt_fontinfo *info, int glyph_index) -{ - stbtt_int16 numberOfContours; - int g; - if (info->cff.size) - return stbtt__GetGlyphInfoT2(info, glyph_index, NULL, NULL, NULL, NULL) == 0; - g = stbtt__GetGlyfOffset(info, glyph_index); - if (g < 0) return 1; - numberOfContours = ttSHORT(info->data + g); - return numberOfContours == 0; -} - -static int stbtt__close_shape(stbtt_vertex *vertices, int num_vertices, int was_off, int start_off, - stbtt_int32 sx, stbtt_int32 sy, stbtt_int32 scx, stbtt_int32 scy, stbtt_int32 cx, stbtt_int32 cy) -{ - if (start_off) { - if (was_off) - stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve, (cx+scx)>>1, (cy+scy)>>1, cx,cy); - stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve, sx,sy,scx,scy); - } else { - if (was_off) - stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve,sx,sy,cx,cy); - else - stbtt_setvertex(&vertices[num_vertices++], STBTT_vline,sx,sy,0,0); - } - return num_vertices; -} - -static int stbtt__GetGlyphShapeTT(const stbtt_fontinfo *info, int glyph_index, stbtt_vertex **pvertices) -{ - stbtt_int16 numberOfContours; - stbtt_uint8 *endPtsOfContours; - stbtt_uint8 *data = info->data; - stbtt_vertex *vertices=0; - int num_vertices=0; - int g = stbtt__GetGlyfOffset(info, glyph_index); - - *pvertices = NULL; - - if (g < 0) return 0; - - numberOfContours = ttSHORT(data + g); - - if (numberOfContours > 0) { - stbtt_uint8 flags=0,flagcount; - stbtt_int32 ins, i,j=0,m,n, next_move, was_off=0, off, start_off=0; - stbtt_int32 x,y,cx,cy,sx,sy, scx,scy; - stbtt_uint8 *points; - endPtsOfContours = (data + g + 10); - ins = ttUSHORT(data + g + 10 + numberOfContours * 2); - points = data + g + 10 + numberOfContours * 2 + 2 + ins; - - n = 1+ttUSHORT(endPtsOfContours + numberOfContours*2-2); - - m = n + 2*numberOfContours; // a loose bound on how many vertices we might need - vertices = (stbtt_vertex *) STBTT_malloc(m * sizeof(vertices[0]), info->userdata); - if (vertices == 0) - return 0; - - next_move = 0; - flagcount=0; - - // in first pass, we load uninterpreted data into the allocated array - // above, shifted to the end of the array so we won't overwrite it when - // we create our final data starting from the front - - off = m - n; // starting offset for uninterpreted data, regardless of how m ends up being calculated - - // first load flags - - for (i=0; i < n; ++i) { - if (flagcount == 0) { - flags = *points++; - if (flags & 8) - flagcount = *points++; - } else - --flagcount; - vertices[off+i].type = flags; - } - - // now load x coordinates - x=0; - for (i=0; i < n; ++i) { - flags = vertices[off+i].type; - if (flags & 2) { - stbtt_int16 dx = *points++; - x += (flags & 16) ? dx : -dx; // ??? - } else { - if (!(flags & 16)) { - x = x + (stbtt_int16) (points[0]*256 + points[1]); - points += 2; - } - } - vertices[off+i].x = (stbtt_int16) x; - } - - // now load y coordinates - y=0; - for (i=0; i < n; ++i) { - flags = vertices[off+i].type; - if (flags & 4) { - stbtt_int16 dy = *points++; - y += (flags & 32) ? dy : -dy; // ??? - } else { - if (!(flags & 32)) { - y = y + (stbtt_int16) (points[0]*256 + points[1]); - points += 2; - } - } - vertices[off+i].y = (stbtt_int16) y; - } - - // now convert them to our format - num_vertices=0; - sx = sy = cx = cy = scx = scy = 0; - for (i=0; i < n; ++i) { - flags = vertices[off+i].type; - x = (stbtt_int16) vertices[off+i].x; - y = (stbtt_int16) vertices[off+i].y; - - if (next_move == i) { - if (i != 0) - num_vertices = stbtt__close_shape(vertices, num_vertices, was_off, start_off, sx,sy,scx,scy,cx,cy); - - // now start the new one - start_off = !(flags & 1); - if (start_off) { - // if we start off with an off-curve point, then when we need to find a point on the curve - // where we can start, and we need to save some state for when we wraparound. - scx = x; - scy = y; - if (!(vertices[off+i+1].type & 1)) { - // next point is also a curve point, so interpolate an on-point curve - sx = (x + (stbtt_int32) vertices[off+i+1].x) >> 1; - sy = (y + (stbtt_int32) vertices[off+i+1].y) >> 1; - } else { - // otherwise just use the next point as our start point - sx = (stbtt_int32) vertices[off+i+1].x; - sy = (stbtt_int32) vertices[off+i+1].y; - ++i; // we're using point i+1 as the starting point, so skip it - } - } else { - sx = x; - sy = y; - } - stbtt_setvertex(&vertices[num_vertices++], STBTT_vmove,sx,sy,0,0); - was_off = 0; - next_move = 1 + ttUSHORT(endPtsOfContours+j*2); - ++j; - } else { - if (!(flags & 1)) { // if it's a curve - if (was_off) // two off-curve control points in a row means interpolate an on-curve midpoint - stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve, (cx+x)>>1, (cy+y)>>1, cx, cy); - cx = x; - cy = y; - was_off = 1; - } else { - if (was_off) - stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve, x,y, cx, cy); - else - stbtt_setvertex(&vertices[num_vertices++], STBTT_vline, x,y,0,0); - was_off = 0; - } - } - } - num_vertices = stbtt__close_shape(vertices, num_vertices, was_off, start_off, sx,sy,scx,scy,cx,cy); - } else if (numberOfContours < 0) { - // Compound shapes. - int more = 1; - stbtt_uint8 *comp = data + g + 10; - num_vertices = 0; - vertices = 0; - while (more) { - stbtt_uint16 flags, gidx; - int comp_num_verts = 0, i; - stbtt_vertex *comp_verts = 0, *tmp = 0; - float mtx[6] = {1,0,0,1,0,0}, m, n; - - flags = ttSHORT(comp); comp+=2; - gidx = ttSHORT(comp); comp+=2; - - if (flags & 2) { // XY values - if (flags & 1) { // shorts - mtx[4] = ttSHORT(comp); comp+=2; - mtx[5] = ttSHORT(comp); comp+=2; - } else { - mtx[4] = ttCHAR(comp); comp+=1; - mtx[5] = ttCHAR(comp); comp+=1; - } - } - else { - // @TODO handle matching point - STBTT_assert(0); - } - if (flags & (1<<3)) { // WE_HAVE_A_SCALE - mtx[0] = mtx[3] = ttSHORT(comp)/16384.0f; comp+=2; - mtx[1] = mtx[2] = 0; - } else if (flags & (1<<6)) { // WE_HAVE_AN_X_AND_YSCALE - mtx[0] = ttSHORT(comp)/16384.0f; comp+=2; - mtx[1] = mtx[2] = 0; - mtx[3] = ttSHORT(comp)/16384.0f; comp+=2; - } else if (flags & (1<<7)) { // WE_HAVE_A_TWO_BY_TWO - mtx[0] = ttSHORT(comp)/16384.0f; comp+=2; - mtx[1] = ttSHORT(comp)/16384.0f; comp+=2; - mtx[2] = ttSHORT(comp)/16384.0f; comp+=2; - mtx[3] = ttSHORT(comp)/16384.0f; comp+=2; - } - - // Find transformation scales. - m = (float) STBTT_sqrt(mtx[0]*mtx[0] + mtx[1]*mtx[1]); - n = (float) STBTT_sqrt(mtx[2]*mtx[2] + mtx[3]*mtx[3]); - - // Get indexed glyph. - comp_num_verts = stbtt_GetGlyphShape(info, gidx, &comp_verts); - if (comp_num_verts > 0) { - // Transform vertices. - for (i = 0; i < comp_num_verts; ++i) { - stbtt_vertex* v = &comp_verts[i]; - stbtt_vertex_type x,y; - x=v->x; y=v->y; - v->x = (stbtt_vertex_type)(m * (mtx[0]*x + mtx[2]*y + mtx[4])); - v->y = (stbtt_vertex_type)(n * (mtx[1]*x + mtx[3]*y + mtx[5])); - x=v->cx; y=v->cy; - v->cx = (stbtt_vertex_type)(m * (mtx[0]*x + mtx[2]*y + mtx[4])); - v->cy = (stbtt_vertex_type)(n * (mtx[1]*x + mtx[3]*y + mtx[5])); - } - // Append vertices. - tmp = (stbtt_vertex*)STBTT_malloc((num_vertices+comp_num_verts)*sizeof(stbtt_vertex), info->userdata); - if (!tmp) { - if (vertices) STBTT_free(vertices, info->userdata); - if (comp_verts) STBTT_free(comp_verts, info->userdata); - return 0; - } - if (num_vertices > 0 && vertices) STBTT_memcpy(tmp, vertices, num_vertices*sizeof(stbtt_vertex)); - STBTT_memcpy(tmp+num_vertices, comp_verts, comp_num_verts*sizeof(stbtt_vertex)); - if (vertices) STBTT_free(vertices, info->userdata); - vertices = tmp; - STBTT_free(comp_verts, info->userdata); - num_vertices += comp_num_verts; - } - // More components ? - more = flags & (1<<5); - } - } else { - // numberOfCounters == 0, do nothing - } - - *pvertices = vertices; - return num_vertices; -} - -typedef struct -{ - int bounds; - int started; - float first_x, first_y; - float x, y; - stbtt_int32 min_x, max_x, min_y, max_y; - - stbtt_vertex *pvertices; - int num_vertices; -} stbtt__csctx; - -#define STBTT__CSCTX_INIT(bounds) {bounds,0, 0,0, 0,0, 0,0,0,0, NULL, 0} - -static void stbtt__track_vertex(stbtt__csctx *c, stbtt_int32 x, stbtt_int32 y) -{ - if (x > c->max_x || !c->started) c->max_x = x; - if (y > c->max_y || !c->started) c->max_y = y; - if (x < c->min_x || !c->started) c->min_x = x; - if (y < c->min_y || !c->started) c->min_y = y; - c->started = 1; -} - -static void stbtt__csctx_v(stbtt__csctx *c, stbtt_uint8 type, stbtt_int32 x, stbtt_int32 y, stbtt_int32 cx, stbtt_int32 cy, stbtt_int32 cx1, stbtt_int32 cy1) -{ - if (c->bounds) { - stbtt__track_vertex(c, x, y); - if (type == STBTT_vcubic) { - stbtt__track_vertex(c, cx, cy); - stbtt__track_vertex(c, cx1, cy1); - } - } else { - stbtt_setvertex(&c->pvertices[c->num_vertices], type, x, y, cx, cy); - c->pvertices[c->num_vertices].cx1 = (stbtt_int16) cx1; - c->pvertices[c->num_vertices].cy1 = (stbtt_int16) cy1; - } - c->num_vertices++; -} - -static void stbtt__csctx_close_shape(stbtt__csctx *ctx) -{ - if (ctx->first_x != ctx->x || ctx->first_y != ctx->y) - stbtt__csctx_v(ctx, STBTT_vline, (int)ctx->first_x, (int)ctx->first_y, 0, 0, 0, 0); -} - -static void stbtt__csctx_rmove_to(stbtt__csctx *ctx, float dx, float dy) -{ - stbtt__csctx_close_shape(ctx); - ctx->first_x = ctx->x = ctx->x + dx; - ctx->first_y = ctx->y = ctx->y + dy; - stbtt__csctx_v(ctx, STBTT_vmove, (int)ctx->x, (int)ctx->y, 0, 0, 0, 0); -} - -static void stbtt__csctx_rline_to(stbtt__csctx *ctx, float dx, float dy) -{ - ctx->x += dx; - ctx->y += dy; - stbtt__csctx_v(ctx, STBTT_vline, (int)ctx->x, (int)ctx->y, 0, 0, 0, 0); -} - -static void stbtt__csctx_rccurve_to(stbtt__csctx *ctx, float dx1, float dy1, float dx2, float dy2, float dx3, float dy3) -{ - float cx1 = ctx->x + dx1; - float cy1 = ctx->y + dy1; - float cx2 = cx1 + dx2; - float cy2 = cy1 + dy2; - ctx->x = cx2 + dx3; - ctx->y = cy2 + dy3; - stbtt__csctx_v(ctx, STBTT_vcubic, (int)ctx->x, (int)ctx->y, (int)cx1, (int)cy1, (int)cx2, (int)cy2); -} - -static stbtt__buf stbtt__get_subr(stbtt__buf idx, int n) -{ - int count = stbtt__cff_index_count(&idx); - int bias = 107; - if (count >= 33900) - bias = 32768; - else if (count >= 1240) - bias = 1131; - n += bias; - if (n < 0 || n >= count) - return stbtt__new_buf(NULL, 0); - return stbtt__cff_index_get(idx, n); -} - -static stbtt__buf stbtt__cid_get_glyph_subrs(const stbtt_fontinfo *info, int glyph_index) -{ - stbtt__buf fdselect = info->fdselect; - int nranges, start, end, v, fmt, fdselector = -1, i; - - stbtt__buf_seek(&fdselect, 0); - fmt = stbtt__buf_get8(&fdselect); - if (fmt == 0) { - // untested - stbtt__buf_skip(&fdselect, glyph_index); - fdselector = stbtt__buf_get8(&fdselect); - } else if (fmt == 3) { - nranges = stbtt__buf_get16(&fdselect); - start = stbtt__buf_get16(&fdselect); - for (i = 0; i < nranges; i++) { - v = stbtt__buf_get8(&fdselect); - end = stbtt__buf_get16(&fdselect); - if (glyph_index >= start && glyph_index < end) { - fdselector = v; - break; - } - start = end; - } - } - if (fdselector == -1) stbtt__new_buf(NULL, 0); - return stbtt__get_subrs(info->cff, stbtt__cff_index_get(info->fontdicts, fdselector)); -} - -static int stbtt__run_charstring(const stbtt_fontinfo *info, int glyph_index, stbtt__csctx *c) -{ - int in_header = 1, maskbits = 0, subr_stack_height = 0, sp = 0, v, i, b0; - int has_subrs = 0, clear_stack; - float s[48]; - stbtt__buf subr_stack[10], subrs = info->subrs, b; - float f; - -#define STBTT__CSERR(s) (0) - - // this currently ignores the initial width value, which isn't needed if we have hmtx - b = stbtt__cff_index_get(info->charstrings, glyph_index); - while (b.cursor < b.size) { - i = 0; - clear_stack = 1; - b0 = stbtt__buf_get8(&b); - switch (b0) { - // @TODO implement hinting - case 0x13: // hintmask - case 0x14: // cntrmask - if (in_header) - maskbits += (sp / 2); // implicit "vstem" - in_header = 0; - stbtt__buf_skip(&b, (maskbits + 7) / 8); - break; - - case 0x01: // hstem - case 0x03: // vstem - case 0x12: // hstemhm - case 0x17: // vstemhm - maskbits += (sp / 2); - break; - - case 0x15: // rmoveto - in_header = 0; - if (sp < 2) return STBTT__CSERR("rmoveto stack"); - stbtt__csctx_rmove_to(c, s[sp-2], s[sp-1]); - break; - case 0x04: // vmoveto - in_header = 0; - if (sp < 1) return STBTT__CSERR("vmoveto stack"); - stbtt__csctx_rmove_to(c, 0, s[sp-1]); - break; - case 0x16: // hmoveto - in_header = 0; - if (sp < 1) return STBTT__CSERR("hmoveto stack"); - stbtt__csctx_rmove_to(c, s[sp-1], 0); - break; - - case 0x05: // rlineto - if (sp < 2) return STBTT__CSERR("rlineto stack"); - for (; i + 1 < sp; i += 2) - stbtt__csctx_rline_to(c, s[i], s[i+1]); - break; - - // hlineto/vlineto and vhcurveto/hvcurveto alternate horizontal and vertical - // starting from a different place. - - case 0x07: // vlineto - if (sp < 1) return STBTT__CSERR("vlineto stack"); - goto vlineto; - case 0x06: // hlineto - if (sp < 1) return STBTT__CSERR("hlineto stack"); - for (;;) { - if (i >= sp) break; - stbtt__csctx_rline_to(c, s[i], 0); - i++; - vlineto: - if (i >= sp) break; - stbtt__csctx_rline_to(c, 0, s[i]); - i++; - } - break; - - case 0x1F: // hvcurveto - if (sp < 4) return STBTT__CSERR("hvcurveto stack"); - goto hvcurveto; - case 0x1E: // vhcurveto - if (sp < 4) return STBTT__CSERR("vhcurveto stack"); - for (;;) { - if (i + 3 >= sp) break; - stbtt__csctx_rccurve_to(c, 0, s[i], s[i+1], s[i+2], s[i+3], (sp - i == 5) ? s[i + 4] : 0.0f); - i += 4; - hvcurveto: - if (i + 3 >= sp) break; - stbtt__csctx_rccurve_to(c, s[i], 0, s[i+1], s[i+2], (sp - i == 5) ? s[i+4] : 0.0f, s[i+3]); - i += 4; - } - break; - - case 0x08: // rrcurveto - if (sp < 6) return STBTT__CSERR("rcurveline stack"); - for (; i + 5 < sp; i += 6) - stbtt__csctx_rccurve_to(c, s[i], s[i+1], s[i+2], s[i+3], s[i+4], s[i+5]); - break; - - case 0x18: // rcurveline - if (sp < 8) return STBTT__CSERR("rcurveline stack"); - for (; i + 5 < sp - 2; i += 6) - stbtt__csctx_rccurve_to(c, s[i], s[i+1], s[i+2], s[i+3], s[i+4], s[i+5]); - if (i + 1 >= sp) return STBTT__CSERR("rcurveline stack"); - stbtt__csctx_rline_to(c, s[i], s[i+1]); - break; - - case 0x19: // rlinecurve - if (sp < 8) return STBTT__CSERR("rlinecurve stack"); - for (; i + 1 < sp - 6; i += 2) - stbtt__csctx_rline_to(c, s[i], s[i+1]); - if (i + 5 >= sp) return STBTT__CSERR("rlinecurve stack"); - stbtt__csctx_rccurve_to(c, s[i], s[i+1], s[i+2], s[i+3], s[i+4], s[i+5]); - break; - - case 0x1A: // vvcurveto - case 0x1B: // hhcurveto - if (sp < 4) return STBTT__CSERR("(vv|hh)curveto stack"); - f = 0.0; - if (sp & 1) { f = s[i]; i++; } - for (; i + 3 < sp; i += 4) { - if (b0 == 0x1B) - stbtt__csctx_rccurve_to(c, s[i], f, s[i+1], s[i+2], s[i+3], 0.0); - else - stbtt__csctx_rccurve_to(c, f, s[i], s[i+1], s[i+2], 0.0, s[i+3]); - f = 0.0; - } - break; - - case 0x0A: // callsubr - if (!has_subrs) { - if (info->fdselect.size) - subrs = stbtt__cid_get_glyph_subrs(info, glyph_index); - has_subrs = 1; - } - // FALLTHROUGH - case 0x1D: // callgsubr - if (sp < 1) return STBTT__CSERR("call(g|)subr stack"); - v = (int) s[--sp]; - if (subr_stack_height >= 10) return STBTT__CSERR("recursion limit"); - subr_stack[subr_stack_height++] = b; - b = stbtt__get_subr(b0 == 0x0A ? subrs : info->gsubrs, v); - if (b.size == 0) return STBTT__CSERR("subr not found"); - b.cursor = 0; - clear_stack = 0; - break; - - case 0x0B: // return - if (subr_stack_height <= 0) return STBTT__CSERR("return outside subr"); - b = subr_stack[--subr_stack_height]; - clear_stack = 0; - break; - - case 0x0E: // endchar - stbtt__csctx_close_shape(c); - return 1; - - case 0x0C: { // two-byte escape - float dx1, dx2, dx3, dx4, dx5, dx6, dy1, dy2, dy3, dy4, dy5, dy6; - float dx, dy; - int b1 = stbtt__buf_get8(&b); - switch (b1) { - // @TODO These "flex" implementations ignore the flex-depth and resolution, - // and always draw beziers. - case 0x22: // hflex - if (sp < 7) return STBTT__CSERR("hflex stack"); - dx1 = s[0]; - dx2 = s[1]; - dy2 = s[2]; - dx3 = s[3]; - dx4 = s[4]; - dx5 = s[5]; - dx6 = s[6]; - stbtt__csctx_rccurve_to(c, dx1, 0, dx2, dy2, dx3, 0); - stbtt__csctx_rccurve_to(c, dx4, 0, dx5, -dy2, dx6, 0); - break; - - case 0x23: // flex - if (sp < 13) return STBTT__CSERR("flex stack"); - dx1 = s[0]; - dy1 = s[1]; - dx2 = s[2]; - dy2 = s[3]; - dx3 = s[4]; - dy3 = s[5]; - dx4 = s[6]; - dy4 = s[7]; - dx5 = s[8]; - dy5 = s[9]; - dx6 = s[10]; - dy6 = s[11]; - //fd is s[12] - stbtt__csctx_rccurve_to(c, dx1, dy1, dx2, dy2, dx3, dy3); - stbtt__csctx_rccurve_to(c, dx4, dy4, dx5, dy5, dx6, dy6); - break; - - case 0x24: // hflex1 - if (sp < 9) return STBTT__CSERR("hflex1 stack"); - dx1 = s[0]; - dy1 = s[1]; - dx2 = s[2]; - dy2 = s[3]; - dx3 = s[4]; - dx4 = s[5]; - dx5 = s[6]; - dy5 = s[7]; - dx6 = s[8]; - stbtt__csctx_rccurve_to(c, dx1, dy1, dx2, dy2, dx3, 0); - stbtt__csctx_rccurve_to(c, dx4, 0, dx5, dy5, dx6, -(dy1+dy2+dy5)); - break; - - case 0x25: // flex1 - if (sp < 11) return STBTT__CSERR("flex1 stack"); - dx1 = s[0]; - dy1 = s[1]; - dx2 = s[2]; - dy2 = s[3]; - dx3 = s[4]; - dy3 = s[5]; - dx4 = s[6]; - dy4 = s[7]; - dx5 = s[8]; - dy5 = s[9]; - dx6 = dy6 = s[10]; - dx = dx1+dx2+dx3+dx4+dx5; - dy = dy1+dy2+dy3+dy4+dy5; - if (STBTT_fabs(dx) > STBTT_fabs(dy)) - dy6 = -dy; - else - dx6 = -dx; - stbtt__csctx_rccurve_to(c, dx1, dy1, dx2, dy2, dx3, dy3); - stbtt__csctx_rccurve_to(c, dx4, dy4, dx5, dy5, dx6, dy6); - break; - - default: - return STBTT__CSERR("unimplemented"); - } - } break; - - default: - if (b0 != 255 && b0 != 28 && b0 < 32) - return STBTT__CSERR("reserved operator"); - - // push immediate - if (b0 == 255) { - f = (float)(stbtt_int32)stbtt__buf_get32(&b) / 0x10000; - } else { - stbtt__buf_skip(&b, -1); - f = (float)(stbtt_int16)stbtt__cff_int(&b); - } - if (sp >= 48) return STBTT__CSERR("push stack overflow"); - s[sp++] = f; - clear_stack = 0; - break; - } - if (clear_stack) sp = 0; - } - return STBTT__CSERR("no endchar"); - -#undef STBTT__CSERR -} - -static int stbtt__GetGlyphShapeT2(const stbtt_fontinfo *info, int glyph_index, stbtt_vertex **pvertices) -{ - // runs the charstring twice, once to count and once to output (to avoid realloc) - stbtt__csctx count_ctx = STBTT__CSCTX_INIT(1); - stbtt__csctx output_ctx = STBTT__CSCTX_INIT(0); - if (stbtt__run_charstring(info, glyph_index, &count_ctx)) { - *pvertices = (stbtt_vertex*)STBTT_malloc(count_ctx.num_vertices*sizeof(stbtt_vertex), info->userdata); - output_ctx.pvertices = *pvertices; - if (stbtt__run_charstring(info, glyph_index, &output_ctx)) { - STBTT_assert(output_ctx.num_vertices == count_ctx.num_vertices); - return output_ctx.num_vertices; - } - } - *pvertices = NULL; - return 0; -} - -static int stbtt__GetGlyphInfoT2(const stbtt_fontinfo *info, int glyph_index, int *x0, int *y0, int *x1, int *y1) -{ - stbtt__csctx c = STBTT__CSCTX_INIT(1); - int r = stbtt__run_charstring(info, glyph_index, &c); - if (x0) *x0 = r ? c.min_x : 0; - if (y0) *y0 = r ? c.min_y : 0; - if (x1) *x1 = r ? c.max_x : 0; - if (y1) *y1 = r ? c.max_y : 0; - return r ? c.num_vertices : 0; -} - -STBTT_DEF int stbtt_GetGlyphShape(const stbtt_fontinfo *info, int glyph_index, stbtt_vertex **pvertices) -{ - if (!info->cff.size) - return stbtt__GetGlyphShapeTT(info, glyph_index, pvertices); - else - return stbtt__GetGlyphShapeT2(info, glyph_index, pvertices); -} - -STBTT_DEF void stbtt_GetGlyphHMetrics(const stbtt_fontinfo *info, int glyph_index, int *advanceWidth, int *leftSideBearing) -{ - stbtt_uint16 numOfLongHorMetrics = ttUSHORT(info->data+info->hhea + 34); - if (glyph_index < numOfLongHorMetrics) { - if (advanceWidth) *advanceWidth = ttSHORT(info->data + info->hmtx + 4*glyph_index); - if (leftSideBearing) *leftSideBearing = ttSHORT(info->data + info->hmtx + 4*glyph_index + 2); - } else { - if (advanceWidth) *advanceWidth = ttSHORT(info->data + info->hmtx + 4*(numOfLongHorMetrics-1)); - if (leftSideBearing) *leftSideBearing = ttSHORT(info->data + info->hmtx + 4*numOfLongHorMetrics + 2*(glyph_index - numOfLongHorMetrics)); - } -} - -STBTT_DEF int stbtt_GetKerningTableLength(const stbtt_fontinfo *info) -{ - stbtt_uint8 *data = info->data + info->kern; - - // we only look at the first table. it must be 'horizontal' and format 0. - if (!info->kern) - return 0; - if (ttUSHORT(data+2) < 1) // number of tables, need at least 1 - return 0; - if (ttUSHORT(data+8) != 1) // horizontal flag must be set in format - return 0; - - return ttUSHORT(data+10); -} - -STBTT_DEF int stbtt_GetKerningTable(const stbtt_fontinfo *info, stbtt_kerningentry* table, int table_length) -{ - stbtt_uint8 *data = info->data + info->kern; - int k, length; - - // we only look at the first table. it must be 'horizontal' and format 0. - if (!info->kern) - return 0; - if (ttUSHORT(data+2) < 1) // number of tables, need at least 1 - return 0; - if (ttUSHORT(data+8) != 1) // horizontal flag must be set in format - return 0; - - length = ttUSHORT(data+10); - if (table_length < length) - length = table_length; - - for (k = 0; k < length; k++) - { - table[k].glyph1 = ttUSHORT(data+18+(k*6)); - table[k].glyph2 = ttUSHORT(data+20+(k*6)); - table[k].advance = ttSHORT(data+22+(k*6)); - } - - return length; -} - -static int stbtt__GetGlyphKernInfoAdvance(const stbtt_fontinfo *info, int glyph1, int glyph2) -{ - stbtt_uint8 *data = info->data + info->kern; - stbtt_uint32 needle, straw; - int l, r, m; - - // we only look at the first table. it must be 'horizontal' and format 0. - if (!info->kern) - return 0; - if (ttUSHORT(data+2) < 1) // number of tables, need at least 1 - return 0; - if (ttUSHORT(data+8) != 1) // horizontal flag must be set in format - return 0; - - l = 0; - r = ttUSHORT(data+10) - 1; - needle = glyph1 << 16 | glyph2; - while (l <= r) { - m = (l + r) >> 1; - straw = ttULONG(data+18+(m*6)); // note: unaligned read - if (needle < straw) - r = m - 1; - else if (needle > straw) - l = m + 1; - else - return ttSHORT(data+22+(m*6)); - } - return 0; -} - -static stbtt_int32 stbtt__GetCoverageIndex(stbtt_uint8 *coverageTable, int glyph) -{ - stbtt_uint16 coverageFormat = ttUSHORT(coverageTable); - switch (coverageFormat) { - case 1: { - stbtt_uint16 glyphCount = ttUSHORT(coverageTable + 2); - - // Binary search. - stbtt_int32 l=0, r=glyphCount-1, m; - int straw, needle=glyph; - while (l <= r) { - stbtt_uint8 *glyphArray = coverageTable + 4; - stbtt_uint16 glyphID; - m = (l + r) >> 1; - glyphID = ttUSHORT(glyphArray + 2 * m); - straw = glyphID; - if (needle < straw) - r = m - 1; - else if (needle > straw) - l = m + 1; - else { - return m; - } - } - break; - } - - case 2: { - stbtt_uint16 rangeCount = ttUSHORT(coverageTable + 2); - stbtt_uint8 *rangeArray = coverageTable + 4; - - // Binary search. - stbtt_int32 l=0, r=rangeCount-1, m; - int strawStart, strawEnd, needle=glyph; - while (l <= r) { - stbtt_uint8 *rangeRecord; - m = (l + r) >> 1; - rangeRecord = rangeArray + 6 * m; - strawStart = ttUSHORT(rangeRecord); - strawEnd = ttUSHORT(rangeRecord + 2); - if (needle < strawStart) - r = m - 1; - else if (needle > strawEnd) - l = m + 1; - else { - stbtt_uint16 startCoverageIndex = ttUSHORT(rangeRecord + 4); - return startCoverageIndex + glyph - strawStart; - } - } - break; - } - - default: return -1; // unsupported - } - - return -1; -} - -static stbtt_int32 stbtt__GetGlyphClass(stbtt_uint8 *classDefTable, int glyph) -{ - stbtt_uint16 classDefFormat = ttUSHORT(classDefTable); - switch (classDefFormat) - { - case 1: { - stbtt_uint16 startGlyphID = ttUSHORT(classDefTable + 2); - stbtt_uint16 glyphCount = ttUSHORT(classDefTable + 4); - stbtt_uint8 *classDef1ValueArray = classDefTable + 6; - - if (glyph >= startGlyphID && glyph < startGlyphID + glyphCount) - return (stbtt_int32)ttUSHORT(classDef1ValueArray + 2 * (glyph - startGlyphID)); - break; - } - - case 2: { - stbtt_uint16 classRangeCount = ttUSHORT(classDefTable + 2); - stbtt_uint8 *classRangeRecords = classDefTable + 4; - - // Binary search. - stbtt_int32 l=0, r=classRangeCount-1, m; - int strawStart, strawEnd, needle=glyph; - while (l <= r) { - stbtt_uint8 *classRangeRecord; - m = (l + r) >> 1; - classRangeRecord = classRangeRecords + 6 * m; - strawStart = ttUSHORT(classRangeRecord); - strawEnd = ttUSHORT(classRangeRecord + 2); - if (needle < strawStart) - r = m - 1; - else if (needle > strawEnd) - l = m + 1; - else - return (stbtt_int32)ttUSHORT(classRangeRecord + 4); - } - break; - } - - default: - return -1; // Unsupported definition type, return an error. - } - - // "All glyphs not assigned to a class fall into class 0". (OpenType spec) - return 0; -} - -// Define to STBTT_assert(x) if you want to break on unimplemented formats. -#define STBTT_GPOS_TODO_assert(x) - -static stbtt_int32 stbtt__GetGlyphGPOSInfoAdvance(const stbtt_fontinfo *info, int glyph1, int glyph2) -{ - stbtt_uint16 lookupListOffset; - stbtt_uint8 *lookupList; - stbtt_uint16 lookupCount; - stbtt_uint8 *data; - stbtt_int32 i, sti; - - if (!info->gpos) return 0; - - data = info->data + info->gpos; - - if (ttUSHORT(data+0) != 1) return 0; // Major version 1 - if (ttUSHORT(data+2) != 0) return 0; // Minor version 0 - - lookupListOffset = ttUSHORT(data+8); - lookupList = data + lookupListOffset; - lookupCount = ttUSHORT(lookupList); - - for (i=0; i<lookupCount; ++i) { - stbtt_uint16 lookupOffset = ttUSHORT(lookupList + 2 + 2 * i); - stbtt_uint8 *lookupTable = lookupList + lookupOffset; - - stbtt_uint16 lookupType = ttUSHORT(lookupTable); - stbtt_uint16 subTableCount = ttUSHORT(lookupTable + 4); - stbtt_uint8 *subTableOffsets = lookupTable + 6; - if (lookupType != 2) // Pair Adjustment Positioning Subtable - continue; - - for (sti=0; sti<subTableCount; sti++) { - stbtt_uint16 subtableOffset = ttUSHORT(subTableOffsets + 2 * sti); - stbtt_uint8 *table = lookupTable + subtableOffset; - stbtt_uint16 posFormat = ttUSHORT(table); - stbtt_uint16 coverageOffset = ttUSHORT(table + 2); - stbtt_int32 coverageIndex = stbtt__GetCoverageIndex(table + coverageOffset, glyph1); - if (coverageIndex == -1) continue; - - switch (posFormat) { - case 1: { - stbtt_int32 l, r, m; - int straw, needle; - stbtt_uint16 valueFormat1 = ttUSHORT(table + 4); - stbtt_uint16 valueFormat2 = ttUSHORT(table + 6); - if (valueFormat1 == 4 && valueFormat2 == 0) { // Support more formats? - stbtt_int32 valueRecordPairSizeInBytes = 2; - stbtt_uint16 pairSetCount = ttUSHORT(table + 8); - stbtt_uint16 pairPosOffset = ttUSHORT(table + 10 + 2 * coverageIndex); - stbtt_uint8 *pairValueTable = table + pairPosOffset; - stbtt_uint16 pairValueCount = ttUSHORT(pairValueTable); - stbtt_uint8 *pairValueArray = pairValueTable + 2; - - if (coverageIndex >= pairSetCount) return 0; - - needle=glyph2; - r=pairValueCount-1; - l=0; - - // Binary search. - while (l <= r) { - stbtt_uint16 secondGlyph; - stbtt_uint8 *pairValue; - m = (l + r) >> 1; - pairValue = pairValueArray + (2 + valueRecordPairSizeInBytes) * m; - secondGlyph = ttUSHORT(pairValue); - straw = secondGlyph; - if (needle < straw) - r = m - 1; - else if (needle > straw) - l = m + 1; - else { - stbtt_int16 xAdvance = ttSHORT(pairValue + 2); - return xAdvance; - } - } - } else - return 0; - break; - } - - case 2: { - stbtt_uint16 valueFormat1 = ttUSHORT(table + 4); - stbtt_uint16 valueFormat2 = ttUSHORT(table + 6); - if (valueFormat1 == 4 && valueFormat2 == 0) { // Support more formats? - stbtt_uint16 classDef1Offset = ttUSHORT(table + 8); - stbtt_uint16 classDef2Offset = ttUSHORT(table + 10); - int glyph1class = stbtt__GetGlyphClass(table + classDef1Offset, glyph1); - int glyph2class = stbtt__GetGlyphClass(table + classDef2Offset, glyph2); - - stbtt_uint16 class1Count = ttUSHORT(table + 12); - stbtt_uint16 class2Count = ttUSHORT(table + 14); - stbtt_uint8 *class1Records, *class2Records; - stbtt_int16 xAdvance; - - if (glyph1class < 0 || glyph1class >= class1Count) return 0; // malformed - if (glyph2class < 0 || glyph2class >= class2Count) return 0; // malformed - - class1Records = table + 16; - class2Records = class1Records + 2 * (glyph1class * class2Count); - xAdvance = ttSHORT(class2Records + 2 * glyph2class); - return xAdvance; - } else - return 0; - break; - } - - default: - return 0; // Unsupported position format - } - } - } - - return 0; -} - -STBTT_DEF int stbtt_GetGlyphKernAdvance(const stbtt_fontinfo *info, int g1, int g2) -{ - int xAdvance = 0; - - if (info->gpos) - xAdvance += stbtt__GetGlyphGPOSInfoAdvance(info, g1, g2); - else if (info->kern) - xAdvance += stbtt__GetGlyphKernInfoAdvance(info, g1, g2); - - return xAdvance; -} - -STBTT_DEF int stbtt_GetCodepointKernAdvance(const stbtt_fontinfo *info, int ch1, int ch2) -{ - if (!info->kern && !info->gpos) // if no kerning table, don't waste time looking up both codepoint->glyphs - return 0; - return stbtt_GetGlyphKernAdvance(info, stbtt_FindGlyphIndex(info,ch1), stbtt_FindGlyphIndex(info,ch2)); -} - -STBTT_DEF void stbtt_GetCodepointHMetrics(const stbtt_fontinfo *info, int codepoint, int *advanceWidth, int *leftSideBearing) -{ - stbtt_GetGlyphHMetrics(info, stbtt_FindGlyphIndex(info,codepoint), advanceWidth, leftSideBearing); -} - -STBTT_DEF void stbtt_GetFontVMetrics(const stbtt_fontinfo *info, int *ascent, int *descent, int *lineGap) -{ - if (ascent ) *ascent = ttSHORT(info->data+info->hhea + 4); - if (descent) *descent = ttSHORT(info->data+info->hhea + 6); - if (lineGap) *lineGap = ttSHORT(info->data+info->hhea + 8); -} - -STBTT_DEF int stbtt_GetFontVMetricsOS2(const stbtt_fontinfo *info, int *typoAscent, int *typoDescent, int *typoLineGap) -{ - int tab = stbtt__find_table(info->data, info->fontstart, "OS/2"); - if (!tab) - return 0; - if (typoAscent ) *typoAscent = ttSHORT(info->data+tab + 68); - if (typoDescent) *typoDescent = ttSHORT(info->data+tab + 70); - if (typoLineGap) *typoLineGap = ttSHORT(info->data+tab + 72); - return 1; -} - -STBTT_DEF void stbtt_GetFontBoundingBox(const stbtt_fontinfo *info, int *x0, int *y0, int *x1, int *y1) -{ - *x0 = ttSHORT(info->data + info->head + 36); - *y0 = ttSHORT(info->data + info->head + 38); - *x1 = ttSHORT(info->data + info->head + 40); - *y1 = ttSHORT(info->data + info->head + 42); -} - -STBTT_DEF float stbtt_ScaleForPixelHeight(const stbtt_fontinfo *info, float height) -{ - int fheight = ttSHORT(info->data + info->hhea + 4) - ttSHORT(info->data + info->hhea + 6); - return (float) height / fheight; -} - -STBTT_DEF float stbtt_ScaleForMappingEmToPixels(const stbtt_fontinfo *info, float pixels) -{ - int unitsPerEm = ttUSHORT(info->data + info->head + 18); - return pixels / unitsPerEm; -} - -STBTT_DEF void stbtt_FreeShape(const stbtt_fontinfo *info, stbtt_vertex *v) -{ - STBTT_free(v, info->userdata); -} - -STBTT_DEF stbtt_uint8 *stbtt_FindSVGDoc(const stbtt_fontinfo *info, int gl) -{ - int i; - stbtt_uint8 *data = info->data; - stbtt_uint8 *svg_doc_list = data + stbtt__get_svg((stbtt_fontinfo *) info); - - int numEntries = ttUSHORT(svg_doc_list); - stbtt_uint8 *svg_docs = svg_doc_list + 2; - - for(i=0; i<numEntries; i++) { - stbtt_uint8 *svg_doc = svg_docs + (12 * i); - if ((gl >= ttUSHORT(svg_doc)) && (gl <= ttUSHORT(svg_doc + 2))) - return svg_doc; - } - return 0; -} - -STBTT_DEF int stbtt_GetGlyphSVG(const stbtt_fontinfo *info, int gl, const char **svg) -{ - stbtt_uint8 *data = info->data; - stbtt_uint8 *svg_doc; - - if (info->svg == 0) - return 0; - - svg_doc = stbtt_FindSVGDoc(info, gl); - if (svg_doc != NULL) { - *svg = (char *) data + info->svg + ttULONG(svg_doc + 4); - return ttULONG(svg_doc + 8); - } else { - return 0; - } -} - -STBTT_DEF int stbtt_GetCodepointSVG(const stbtt_fontinfo *info, int unicode_codepoint, const char **svg) -{ - return stbtt_GetGlyphSVG(info, stbtt_FindGlyphIndex(info, unicode_codepoint), svg); -} - -////////////////////////////////////////////////////////////////////////////// -// -// antialiasing software rasterizer -// - -STBTT_DEF void stbtt_GetGlyphBitmapBoxSubpixel(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y,float shift_x, float shift_y, int *ix0, int *iy0, int *ix1, int *iy1) -{ - int x0=0,y0=0,x1,y1; // =0 suppresses compiler warning - if (!stbtt_GetGlyphBox(font, glyph, &x0,&y0,&x1,&y1)) { - // e.g. space character - if (ix0) *ix0 = 0; - if (iy0) *iy0 = 0; - if (ix1) *ix1 = 0; - if (iy1) *iy1 = 0; - } else { - // move to integral bboxes (treating pixels as little squares, what pixels get touched)? - if (ix0) *ix0 = STBTT_ifloor( x0 * scale_x + shift_x); - if (iy0) *iy0 = STBTT_ifloor(-y1 * scale_y + shift_y); - if (ix1) *ix1 = STBTT_iceil ( x1 * scale_x + shift_x); - if (iy1) *iy1 = STBTT_iceil (-y0 * scale_y + shift_y); - } -} - -STBTT_DEF void stbtt_GetGlyphBitmapBox(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1) -{ - stbtt_GetGlyphBitmapBoxSubpixel(font, glyph, scale_x, scale_y,0.0f,0.0f, ix0, iy0, ix1, iy1); -} - -STBTT_DEF void stbtt_GetCodepointBitmapBoxSubpixel(const stbtt_fontinfo *font, int codepoint, float scale_x, float scale_y, float shift_x, float shift_y, int *ix0, int *iy0, int *ix1, int *iy1) -{ - stbtt_GetGlyphBitmapBoxSubpixel(font, stbtt_FindGlyphIndex(font,codepoint), scale_x, scale_y,shift_x,shift_y, ix0,iy0,ix1,iy1); -} - -STBTT_DEF void stbtt_GetCodepointBitmapBox(const stbtt_fontinfo *font, int codepoint, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1) -{ - stbtt_GetCodepointBitmapBoxSubpixel(font, codepoint, scale_x, scale_y,0.0f,0.0f, ix0,iy0,ix1,iy1); -} - -////////////////////////////////////////////////////////////////////////////// -// -// Rasterizer - -typedef struct stbtt__hheap_chunk -{ - struct stbtt__hheap_chunk *next; -} stbtt__hheap_chunk; - -typedef struct stbtt__hheap -{ - struct stbtt__hheap_chunk *head; - void *first_free; - int num_remaining_in_head_chunk; -} stbtt__hheap; - -static void *stbtt__hheap_alloc(stbtt__hheap *hh, size_t size, void *userdata) -{ - if (hh->first_free) { - void *p = hh->first_free; - hh->first_free = * (void **) p; - return p; - } else { - if (hh->num_remaining_in_head_chunk == 0) { - int count = (size < 32 ? 2000 : size < 128 ? 800 : 100); - stbtt__hheap_chunk *c = (stbtt__hheap_chunk *) STBTT_malloc(sizeof(stbtt__hheap_chunk) + size * count, userdata); - if (c == NULL) - return NULL; - c->next = hh->head; - hh->head = c; - hh->num_remaining_in_head_chunk = count; - } - --hh->num_remaining_in_head_chunk; - return (char *) (hh->head) + sizeof(stbtt__hheap_chunk) + size * hh->num_remaining_in_head_chunk; - } -} - -static void stbtt__hheap_free(stbtt__hheap *hh, void *p) -{ - *(void **) p = hh->first_free; - hh->first_free = p; -} - -static void stbtt__hheap_cleanup(stbtt__hheap *hh, void *userdata) -{ - stbtt__hheap_chunk *c = hh->head; - while (c) { - stbtt__hheap_chunk *n = c->next; - STBTT_free(c, userdata); - c = n; - } -} - -typedef struct stbtt__edge { - float x0,y0, x1,y1; - int invert; -} stbtt__edge; - - -typedef struct stbtt__active_edge -{ - struct stbtt__active_edge *next; - #if STBTT_RASTERIZER_VERSION==1 - int x,dx; - float ey; - int direction; - #elif STBTT_RASTERIZER_VERSION==2 - float fx,fdx,fdy; - float direction; - float sy; - float ey; - #else - #error "Unrecognized value of STBTT_RASTERIZER_VERSION" - #endif -} stbtt__active_edge; - -#if STBTT_RASTERIZER_VERSION == 1 -#define STBTT_FIXSHIFT 10 -#define STBTT_FIX (1 << STBTT_FIXSHIFT) -#define STBTT_FIXMASK (STBTT_FIX-1) - -static stbtt__active_edge *stbtt__new_active(stbtt__hheap *hh, stbtt__edge *e, int off_x, float start_point, void *userdata) -{ - stbtt__active_edge *z = (stbtt__active_edge *) stbtt__hheap_alloc(hh, sizeof(*z), userdata); - float dxdy = (e->x1 - e->x0) / (e->y1 - e->y0); - STBTT_assert(z != NULL); - if (!z) return z; - - // round dx down to avoid overshooting - if (dxdy < 0) - z->dx = -STBTT_ifloor(STBTT_FIX * -dxdy); - else - z->dx = STBTT_ifloor(STBTT_FIX * dxdy); - - z->x = STBTT_ifloor(STBTT_FIX * e->x0 + z->dx * (start_point - e->y0)); // use z->dx so when we offset later it's by the same amount - z->x -= off_x * STBTT_FIX; - - z->ey = e->y1; - z->next = 0; - z->direction = e->invert ? 1 : -1; - return z; -} -#elif STBTT_RASTERIZER_VERSION == 2 -static stbtt__active_edge *stbtt__new_active(stbtt__hheap *hh, stbtt__edge *e, int off_x, float start_point, void *userdata) -{ - stbtt__active_edge *z = (stbtt__active_edge *) stbtt__hheap_alloc(hh, sizeof(*z), userdata); - float dxdy = (e->x1 - e->x0) / (e->y1 - e->y0); - STBTT_assert(z != NULL); - //STBTT_assert(e->y0 <= start_point); - if (!z) return z; - z->fdx = dxdy; - z->fdy = dxdy != 0.0f ? (1.0f/dxdy) : 0.0f; - z->fx = e->x0 + dxdy * (start_point - e->y0); - z->fx -= off_x; - z->direction = e->invert ? 1.0f : -1.0f; - z->sy = e->y0; - z->ey = e->y1; - z->next = 0; - return z; -} -#else -#error "Unrecognized value of STBTT_RASTERIZER_VERSION" -#endif - -#if STBTT_RASTERIZER_VERSION == 1 -// note: this routine clips fills that extend off the edges... ideally this -// wouldn't happen, but it could happen if the truetype glyph bounding boxes -// are wrong, or if the user supplies a too-small bitmap -static void stbtt__fill_active_edges(unsigned char *scanline, int len, stbtt__active_edge *e, int max_weight) -{ - // non-zero winding fill - int x0=0, w=0; - - while (e) { - if (w == 0) { - // if we're currently at zero, we need to record the edge start point - x0 = e->x; w += e->direction; - } else { - int x1 = e->x; w += e->direction; - // if we went to zero, we need to draw - if (w == 0) { - int i = x0 >> STBTT_FIXSHIFT; - int j = x1 >> STBTT_FIXSHIFT; - - if (i < len && j >= 0) { - if (i == j) { - // x0,x1 are the same pixel, so compute combined coverage - scanline[i] = scanline[i] + (stbtt_uint8) ((x1 - x0) * max_weight >> STBTT_FIXSHIFT); - } else { - if (i >= 0) // add antialiasing for x0 - scanline[i] = scanline[i] + (stbtt_uint8) (((STBTT_FIX - (x0 & STBTT_FIXMASK)) * max_weight) >> STBTT_FIXSHIFT); - else - i = -1; // clip - - if (j < len) // add antialiasing for x1 - scanline[j] = scanline[j] + (stbtt_uint8) (((x1 & STBTT_FIXMASK) * max_weight) >> STBTT_FIXSHIFT); - else - j = len; // clip - - for (++i; i < j; ++i) // fill pixels between x0 and x1 - scanline[i] = scanline[i] + (stbtt_uint8) max_weight; - } - } - } - } - - e = e->next; - } -} - -static void stbtt__rasterize_sorted_edges(stbtt__bitmap *result, stbtt__edge *e, int n, int vsubsample, int off_x, int off_y, void *userdata) -{ - stbtt__hheap hh = { 0, 0, 0 }; - stbtt__active_edge *active = NULL; - int y,j=0; - int max_weight = (255 / vsubsample); // weight per vertical scanline - int s; // vertical subsample index - unsigned char scanline_data[512], *scanline; - - if (result->w > 512) - scanline = (unsigned char *) STBTT_malloc(result->w, userdata); - else - scanline = scanline_data; - - y = off_y * vsubsample; - e[n].y0 = (off_y + result->h) * (float) vsubsample + 1; - - while (j < result->h) { - STBTT_memset(scanline, 0, result->w); - for (s=0; s < vsubsample; ++s) { - // find center of pixel for this scanline - float scan_y = y + 0.5f; - stbtt__active_edge **step = &active; - - // update all active edges; - // remove all active edges that terminate before the center of this scanline - while (*step) { - stbtt__active_edge * z = *step; - if (z->ey <= scan_y) { - *step = z->next; // delete from list - STBTT_assert(z->direction); - z->direction = 0; - stbtt__hheap_free(&hh, z); - } else { - z->x += z->dx; // advance to position for current scanline - step = &((*step)->next); // advance through list - } - } - - // resort the list if needed - for(;;) { - int changed=0; - step = &active; - while (*step && (*step)->next) { - if ((*step)->x > (*step)->next->x) { - stbtt__active_edge *t = *step; - stbtt__active_edge *q = t->next; - - t->next = q->next; - q->next = t; - *step = q; - changed = 1; - } - step = &(*step)->next; - } - if (!changed) break; - } - - // insert all edges that start before the center of this scanline -- omit ones that also end on this scanline - while (e->y0 <= scan_y) { - if (e->y1 > scan_y) { - stbtt__active_edge *z = stbtt__new_active(&hh, e, off_x, scan_y, userdata); - if (z != NULL) { - // find insertion point - if (active == NULL) - active = z; - else if (z->x < active->x) { - // insert at front - z->next = active; - active = z; - } else { - // find thing to insert AFTER - stbtt__active_edge *p = active; - while (p->next && p->next->x < z->x) - p = p->next; - // at this point, p->next->x is NOT < z->x - z->next = p->next; - p->next = z; - } - } - } - ++e; - } - - // now process all active edges in XOR fashion - if (active) - stbtt__fill_active_edges(scanline, result->w, active, max_weight); - - ++y; - } - STBTT_memcpy(result->pixels + j * result->stride, scanline, result->w); - ++j; - } - - stbtt__hheap_cleanup(&hh, userdata); - - if (scanline != scanline_data) - STBTT_free(scanline, userdata); -} - -#elif STBTT_RASTERIZER_VERSION == 2 - -// the edge passed in here does not cross the vertical line at x or the vertical line at x+1 -// (i.e. it has already been clipped to those) -static void stbtt__handle_clipped_edge(float *scanline, int x, stbtt__active_edge *e, float x0, float y0, float x1, float y1) -{ - if (y0 == y1) return; - STBTT_assert(y0 < y1); - STBTT_assert(e->sy <= e->ey); - if (y0 > e->ey) return; - if (y1 < e->sy) return; - if (y0 < e->sy) { - x0 += (x1-x0) * (e->sy - y0) / (y1-y0); - y0 = e->sy; - } - if (y1 > e->ey) { - x1 += (x1-x0) * (e->ey - y1) / (y1-y0); - y1 = e->ey; - } - - if (x0 == x) - STBTT_assert(x1 <= x+1); - else if (x0 == x+1) - STBTT_assert(x1 >= x); - else if (x0 <= x) - STBTT_assert(x1 <= x); - else if (x0 >= x+1) - STBTT_assert(x1 >= x+1); - else - STBTT_assert(x1 >= x && x1 <= x+1); - - if (x0 <= x && x1 <= x) - scanline[x] += e->direction * (y1-y0); - else if (x0 >= x+1 && x1 >= x+1) - ; - else { - STBTT_assert(x0 >= x && x0 <= x+1 && x1 >= x && x1 <= x+1); - scanline[x] += e->direction * (y1-y0) * (1-((x0-x)+(x1-x))/2); // coverage = 1 - average x position - } -} - -static float stbtt__sized_trapezoid_area(float height, float top_width, float bottom_width) -{ - STBTT_assert(top_width >= 0); - STBTT_assert(bottom_width >= 0); - return (top_width + bottom_width) / 2.0f * height; -} - -static float stbtt__position_trapezoid_area(float height, float tx0, float tx1, float bx0, float bx1) -{ - return stbtt__sized_trapezoid_area(height, tx1 - tx0, bx1 - bx0); -} - -static float stbtt__sized_triangle_area(float height, float width) -{ - return height * width / 2; -} - -static void stbtt__fill_active_edges_new(float *scanline, float *scanline_fill, int len, stbtt__active_edge *e, float y_top) -{ - float y_bottom = y_top+1; - - while (e) { - // brute force every pixel - - // compute intersection points with top & bottom - STBTT_assert(e->ey >= y_top); - - if (e->fdx == 0) { - float x0 = e->fx; - if (x0 < len) { - if (x0 >= 0) { - stbtt__handle_clipped_edge(scanline,(int) x0,e, x0,y_top, x0,y_bottom); - stbtt__handle_clipped_edge(scanline_fill-1,(int) x0+1,e, x0,y_top, x0,y_bottom); - } else { - stbtt__handle_clipped_edge(scanline_fill-1,0,e, x0,y_top, x0,y_bottom); - } - } - } else { - float x0 = e->fx; - float dx = e->fdx; - float xb = x0 + dx; - float x_top, x_bottom; - float sy0,sy1; - float dy = e->fdy; - STBTT_assert(e->sy <= y_bottom && e->ey >= y_top); - - // compute endpoints of line segment clipped to this scanline (if the - // line segment starts on this scanline. x0 is the intersection of the - // line with y_top, but that may be off the line segment. - if (e->sy > y_top) { - x_top = x0 + dx * (e->sy - y_top); - sy0 = e->sy; - } else { - x_top = x0; - sy0 = y_top; - } - if (e->ey < y_bottom) { - x_bottom = x0 + dx * (e->ey - y_top); - sy1 = e->ey; - } else { - x_bottom = xb; - sy1 = y_bottom; - } - - if (x_top >= 0 && x_bottom >= 0 && x_top < len && x_bottom < len) { - // from here on, we don't have to range check x values - - if ((int) x_top == (int) x_bottom) { - float height; - // simple case, only spans one pixel - int x = (int) x_top; - height = (sy1 - sy0) * e->direction; - STBTT_assert(x >= 0 && x < len); - scanline[x] += stbtt__position_trapezoid_area(height, x_top, x+1.0f, x_bottom, x+1.0f); - scanline_fill[x] += height; // everything right of this pixel is filled - } else { - int x,x1,x2; - float y_crossing, y_final, step, sign, area; - // covers 2+ pixels - if (x_top > x_bottom) { - // flip scanline vertically; signed area is the same - float t; - sy0 = y_bottom - (sy0 - y_top); - sy1 = y_bottom - (sy1 - y_top); - t = sy0, sy0 = sy1, sy1 = t; - t = x_bottom, x_bottom = x_top, x_top = t; - dx = -dx; - dy = -dy; - t = x0, x0 = xb, xb = t; - } - STBTT_assert(dy >= 0); - STBTT_assert(dx >= 0); - - x1 = (int) x_top; - x2 = (int) x_bottom; - // compute intersection with y axis at x1+1 - y_crossing = y_top + dy * (x1+1 - x0); - - // compute intersection with y axis at x2 - y_final = y_top + dy * (x2 - x0); - - // x1 x_top x2 x_bottom - // y_top +------|-----+------------+------------+--------|---+------------+ - // | | | | | | - // | | | | | | - // sy0 | Txxxxx|............|............|............|............| - // y_crossing | *xxxxx.......|............|............|............| - // | | xxxxx..|............|............|............| - // | | /- xx*xxxx........|............|............| - // | | dy < | xxxxxx..|............|............| - // y_final | | \- | xx*xxx.........|............| - // sy1 | | | | xxxxxB...|............| - // | | | | | | - // | | | | | | - // y_bottom +------------+------------+------------+------------+------------+ - // - // goal is to measure the area covered by '.' in each pixel - - // if x2 is right at the right edge of x1, y_crossing can blow up, github #1057 - // @TODO: maybe test against sy1 rather than y_bottom? - if (y_crossing > y_bottom) - y_crossing = y_bottom; - - sign = e->direction; - - // area of the rectangle covered from sy0..y_crossing - area = sign * (y_crossing-sy0); - - // area of the triangle (x_top,sy0), (x1+1,sy0), (x1+1,y_crossing) - scanline[x1] += stbtt__sized_triangle_area(area, x1+1 - x_top); - - // check if final y_crossing is blown up; no test case for this - if (y_final > y_bottom) { - y_final = y_bottom; - dy = (y_final - y_crossing ) / (x2 - (x1+1)); // if denom=0, y_final = y_crossing, so y_final <= y_bottom - } - - // in second pixel, area covered by line segment found in first pixel - // is always a rectangle 1 wide * the height of that line segment; this - // is exactly what the variable 'area' stores. it also gets a contribution - // from the line segment within it. the THIRD pixel will get the first - // pixel's rectangle contribution, the second pixel's rectangle contribution, - // and its own contribution. the 'own contribution' is the same in every pixel except - // the leftmost and rightmost, a trapezoid that slides down in each pixel. - // the second pixel's contribution to the third pixel will be the - // rectangle 1 wide times the height change in the second pixel, which is dy. - - step = sign * dy * 1; // dy is dy/dx, change in y for every 1 change in x, - // which multiplied by 1-pixel-width is how much pixel area changes for each step in x - // so the area advances by 'step' every time - - for (x = x1+1; x < x2; ++x) { - scanline[x] += area + step/2; // area of trapezoid is 1*step/2 - area += step; - } - STBTT_assert(STBTT_fabs(area) <= 1.01f); // accumulated error from area += step unless we round step down - STBTT_assert(sy1 > y_final-0.01f); - - // area covered in the last pixel is the rectangle from all the pixels to the left, - // plus the trapezoid filled by the line segment in this pixel all the way to the right edge - scanline[x2] += area + sign * stbtt__position_trapezoid_area(sy1-y_final, (float) x2, x2+1.0f, x_bottom, x2+1.0f); - - // the rest of the line is filled based on the total height of the line segment in this pixel - scanline_fill[x2] += sign * (sy1-sy0); - } - } else { - // if edge goes outside of box we're drawing, we require - // clipping logic. since this does not match the intended use - // of this library, we use a different, very slow brute - // force implementation - // note though that this does happen some of the time because - // x_top and x_bottom can be extrapolated at the top & bottom of - // the shape and actually lie outside the bounding box - int x; - for (x=0; x < len; ++x) { - // cases: - // - // there can be up to two intersections with the pixel. any intersection - // with left or right edges can be handled by splitting into two (or three) - // regions. intersections with top & bottom do not necessitate case-wise logic. - // - // the old way of doing this found the intersections with the left & right edges, - // then used some simple logic to produce up to three segments in sorted order - // from top-to-bottom. however, this had a problem: if an x edge was epsilon - // across the x border, then the corresponding y position might not be distinct - // from the other y segment, and it might ignored as an empty segment. to avoid - // that, we need to explicitly produce segments based on x positions. - - // rename variables to clearly-defined pairs - float y0 = y_top; - float x1 = (float) (x); - float x2 = (float) (x+1); - float x3 = xb; - float y3 = y_bottom; - - // x = e->x + e->dx * (y-y_top) - // (y-y_top) = (x - e->x) / e->dx - // y = (x - e->x) / e->dx + y_top - float y1 = (x - x0) / dx + y_top; - float y2 = (x+1 - x0) / dx + y_top; - - if (x0 < x1 && x3 > x2) { // three segments descending down-right - stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x1,y1); - stbtt__handle_clipped_edge(scanline,x,e, x1,y1, x2,y2); - stbtt__handle_clipped_edge(scanline,x,e, x2,y2, x3,y3); - } else if (x3 < x1 && x0 > x2) { // three segments descending down-left - stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x2,y2); - stbtt__handle_clipped_edge(scanline,x,e, x2,y2, x1,y1); - stbtt__handle_clipped_edge(scanline,x,e, x1,y1, x3,y3); - } else if (x0 < x1 && x3 > x1) { // two segments across x, down-right - stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x1,y1); - stbtt__handle_clipped_edge(scanline,x,e, x1,y1, x3,y3); - } else if (x3 < x1 && x0 > x1) { // two segments across x, down-left - stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x1,y1); - stbtt__handle_clipped_edge(scanline,x,e, x1,y1, x3,y3); - } else if (x0 < x2 && x3 > x2) { // two segments across x+1, down-right - stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x2,y2); - stbtt__handle_clipped_edge(scanline,x,e, x2,y2, x3,y3); - } else if (x3 < x2 && x0 > x2) { // two segments across x+1, down-left - stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x2,y2); - stbtt__handle_clipped_edge(scanline,x,e, x2,y2, x3,y3); - } else { // one segment - stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x3,y3); - } - } - } - } - e = e->next; - } -} - -// directly AA rasterize edges w/o supersampling -static void stbtt__rasterize_sorted_edges(stbtt__bitmap *result, stbtt__edge *e, int n, int vsubsample, int off_x, int off_y, void *userdata) -{ - stbtt__hheap hh = { 0, 0, 0 }; - stbtt__active_edge *active = NULL; - int y,j=0, i; - float scanline_data[129], *scanline, *scanline2; - - STBTT__NOTUSED(vsubsample); - - if (result->w > 64) - scanline = (float *) STBTT_malloc((result->w*2+1) * sizeof(float), userdata); - else - scanline = scanline_data; - - scanline2 = scanline + result->w; - - y = off_y; - e[n].y0 = (float) (off_y + result->h) + 1; - - while (j < result->h) { - // find center of pixel for this scanline - float scan_y_top = y + 0.0f; - float scan_y_bottom = y + 1.0f; - stbtt__active_edge **step = &active; - - STBTT_memset(scanline , 0, result->w*sizeof(scanline[0])); - STBTT_memset(scanline2, 0, (result->w+1)*sizeof(scanline[0])); - - // update all active edges; - // remove all active edges that terminate before the top of this scanline - while (*step) { - stbtt__active_edge * z = *step; - if (z->ey <= scan_y_top) { - *step = z->next; // delete from list - STBTT_assert(z->direction); - z->direction = 0; - stbtt__hheap_free(&hh, z); - } else { - step = &((*step)->next); // advance through list - } - } - - // insert all edges that start before the bottom of this scanline - while (e->y0 <= scan_y_bottom) { - if (e->y0 != e->y1) { - stbtt__active_edge *z = stbtt__new_active(&hh, e, off_x, scan_y_top, userdata); - if (z != NULL) { - if (j == 0 && off_y != 0) { - if (z->ey < scan_y_top) { - // this can happen due to subpixel positioning and some kind of fp rounding error i think - z->ey = scan_y_top; - } - } - STBTT_assert(z->ey >= scan_y_top); // if we get really unlucky a tiny bit of an edge can be out of bounds - // insert at front - z->next = active; - active = z; - } - } - ++e; - } - - // now process all active edges - if (active) - stbtt__fill_active_edges_new(scanline, scanline2+1, result->w, active, scan_y_top); - - { - float sum = 0; - for (i=0; i < result->w; ++i) { - float k; - int m; - sum += scanline2[i]; - k = scanline[i] + sum; - k = (float) STBTT_fabs(k)*255 + 0.5f; - m = (int) k; - if (m > 255) m = 255; - result->pixels[j*result->stride + i] = (unsigned char) m; - } - } - // advance all the edges - step = &active; - while (*step) { - stbtt__active_edge *z = *step; - z->fx += z->fdx; // advance to position for current scanline - step = &((*step)->next); // advance through list - } - - ++y; - ++j; - } - - stbtt__hheap_cleanup(&hh, userdata); - - if (scanline != scanline_data) - STBTT_free(scanline, userdata); -} -#else -#error "Unrecognized value of STBTT_RASTERIZER_VERSION" -#endif - -#define STBTT__COMPARE(a,b) ((a)->y0 < (b)->y0) - -static void stbtt__sort_edges_ins_sort(stbtt__edge *p, int n) -{ - int i,j; - for (i=1; i < n; ++i) { - stbtt__edge t = p[i], *a = &t; - j = i; - while (j > 0) { - stbtt__edge *b = &p[j-1]; - int c = STBTT__COMPARE(a,b); - if (!c) break; - p[j] = p[j-1]; - --j; - } - if (i != j) - p[j] = t; - } -} - -static void stbtt__sort_edges_quicksort(stbtt__edge *p, int n) -{ - /* threshold for transitioning to insertion sort */ - while (n > 12) { - stbtt__edge t; - int c01,c12,c,m,i,j; - - /* compute median of three */ - m = n >> 1; - c01 = STBTT__COMPARE(&p[0],&p[m]); - c12 = STBTT__COMPARE(&p[m],&p[n-1]); - /* if 0 >= mid >= end, or 0 < mid < end, then use mid */ - if (c01 != c12) { - /* otherwise, we'll need to swap something else to middle */ - int z; - c = STBTT__COMPARE(&p[0],&p[n-1]); - /* 0>mid && mid<n: 0>n => n; 0<n => 0 */ - /* 0<mid && mid>n: 0>n => 0; 0<n => n */ - z = (c == c12) ? 0 : n-1; - t = p[z]; - p[z] = p[m]; - p[m] = t; - } - /* now p[m] is the median-of-three */ - /* swap it to the beginning so it won't move around */ - t = p[0]; - p[0] = p[m]; - p[m] = t; - - /* partition loop */ - i=1; - j=n-1; - for(;;) { - /* handling of equality is crucial here */ - /* for sentinels & efficiency with duplicates */ - for (;;++i) { - if (!STBTT__COMPARE(&p[i], &p[0])) break; - } - for (;;--j) { - if (!STBTT__COMPARE(&p[0], &p[j])) break; - } - /* make sure we haven't crossed */ - if (i >= j) break; - t = p[i]; - p[i] = p[j]; - p[j] = t; - - ++i; - --j; - } - /* recurse on smaller side, iterate on larger */ - if (j < (n-i)) { - stbtt__sort_edges_quicksort(p,j); - p = p+i; - n = n-i; - } else { - stbtt__sort_edges_quicksort(p+i, n-i); - n = j; - } - } -} - -static void stbtt__sort_edges(stbtt__edge *p, int n) -{ - stbtt__sort_edges_quicksort(p, n); - stbtt__sort_edges_ins_sort(p, n); -} - -typedef struct -{ - float x,y; -} stbtt__point; - -static void stbtt__rasterize(stbtt__bitmap *result, stbtt__point *pts, int *wcount, int windings, float scale_x, float scale_y, float shift_x, float shift_y, int off_x, int off_y, int invert, void *userdata) -{ - float y_scale_inv = invert ? -scale_y : scale_y; - stbtt__edge *e; - int n,i,j,k,m; -#if STBTT_RASTERIZER_VERSION == 1 - int vsubsample = result->h < 8 ? 15 : 5; -#elif STBTT_RASTERIZER_VERSION == 2 - int vsubsample = 1; -#else - #error "Unrecognized value of STBTT_RASTERIZER_VERSION" -#endif - // vsubsample should divide 255 evenly; otherwise we won't reach full opacity - - // now we have to blow out the windings into explicit edge lists - n = 0; - for (i=0; i < windings; ++i) - n += wcount[i]; - - e = (stbtt__edge *) STBTT_malloc(sizeof(*e) * (n+1), userdata); // add an extra one as a sentinel - if (e == 0) return; - n = 0; - - m=0; - for (i=0; i < windings; ++i) { - stbtt__point *p = pts + m; - m += wcount[i]; - j = wcount[i]-1; - for (k=0; k < wcount[i]; j=k++) { - int a=k,b=j; - // skip the edge if horizontal - if (p[j].y == p[k].y) - continue; - // add edge from j to k to the list - e[n].invert = 0; - if (invert ? p[j].y > p[k].y : p[j].y < p[k].y) { - e[n].invert = 1; - a=j,b=k; - } - e[n].x0 = p[a].x * scale_x + shift_x; - e[n].y0 = (p[a].y * y_scale_inv + shift_y) * vsubsample; - e[n].x1 = p[b].x * scale_x + shift_x; - e[n].y1 = (p[b].y * y_scale_inv + shift_y) * vsubsample; - ++n; - } - } - - // now sort the edges by their highest point (should snap to integer, and then by x) - //STBTT_sort(e, n, sizeof(e[0]), stbtt__edge_compare); - stbtt__sort_edges(e, n); - - // now, traverse the scanlines and find the intersections on each scanline, use xor winding rule - stbtt__rasterize_sorted_edges(result, e, n, vsubsample, off_x, off_y, userdata); - - STBTT_free(e, userdata); -} - -static void stbtt__add_point(stbtt__point *points, int n, float x, float y) -{ - if (!points) return; // during first pass, it's unallocated - points[n].x = x; - points[n].y = y; -} - -// tessellate until threshold p is happy... @TODO warped to compensate for non-linear stretching -static int stbtt__tesselate_curve(stbtt__point *points, int *num_points, float x0, float y0, float x1, float y1, float x2, float y2, float objspace_flatness_squared, int n) -{ - // midpoint - float mx = (x0 + 2*x1 + x2)/4; - float my = (y0 + 2*y1 + y2)/4; - // versus directly drawn line - float dx = (x0+x2)/2 - mx; - float dy = (y0+y2)/2 - my; - if (n > 16) // 65536 segments on one curve better be enough! - return 1; - if (dx*dx+dy*dy > objspace_flatness_squared) { // half-pixel error allowed... need to be smaller if AA - stbtt__tesselate_curve(points, num_points, x0,y0, (x0+x1)/2.0f,(y0+y1)/2.0f, mx,my, objspace_flatness_squared,n+1); - stbtt__tesselate_curve(points, num_points, mx,my, (x1+x2)/2.0f,(y1+y2)/2.0f, x2,y2, objspace_flatness_squared,n+1); - } else { - stbtt__add_point(points, *num_points,x2,y2); - *num_points = *num_points+1; - } - return 1; -} - -static void stbtt__tesselate_cubic(stbtt__point *points, int *num_points, float x0, float y0, float x1, float y1, float x2, float y2, float x3, float y3, float objspace_flatness_squared, int n) -{ - // @TODO this "flatness" calculation is just made-up nonsense that seems to work well enough - float dx0 = x1-x0; - float dy0 = y1-y0; - float dx1 = x2-x1; - float dy1 = y2-y1; - float dx2 = x3-x2; - float dy2 = y3-y2; - float dx = x3-x0; - float dy = y3-y0; - float longlen = (float) (STBTT_sqrt(dx0*dx0+dy0*dy0)+STBTT_sqrt(dx1*dx1+dy1*dy1)+STBTT_sqrt(dx2*dx2+dy2*dy2)); - float shortlen = (float) STBTT_sqrt(dx*dx+dy*dy); - float flatness_squared = longlen*longlen-shortlen*shortlen; - - if (n > 16) // 65536 segments on one curve better be enough! - return; - - if (flatness_squared > objspace_flatness_squared) { - float x01 = (x0+x1)/2; - float y01 = (y0+y1)/2; - float x12 = (x1+x2)/2; - float y12 = (y1+y2)/2; - float x23 = (x2+x3)/2; - float y23 = (y2+y3)/2; - - float xa = (x01+x12)/2; - float ya = (y01+y12)/2; - float xb = (x12+x23)/2; - float yb = (y12+y23)/2; - - float mx = (xa+xb)/2; - float my = (ya+yb)/2; - - stbtt__tesselate_cubic(points, num_points, x0,y0, x01,y01, xa,ya, mx,my, objspace_flatness_squared,n+1); - stbtt__tesselate_cubic(points, num_points, mx,my, xb,yb, x23,y23, x3,y3, objspace_flatness_squared,n+1); - } else { - stbtt__add_point(points, *num_points,x3,y3); - *num_points = *num_points+1; - } -} - -// returns number of contours -static stbtt__point *stbtt_FlattenCurves(stbtt_vertex *vertices, int num_verts, float objspace_flatness, int **contour_lengths, int *num_contours, void *userdata) -{ - stbtt__point *points=0; - int num_points=0; - - float objspace_flatness_squared = objspace_flatness * objspace_flatness; - int i,n=0,start=0, pass; - - // count how many "moves" there are to get the contour count - for (i=0; i < num_verts; ++i) - if (vertices[i].type == STBTT_vmove) - ++n; - - *num_contours = n; - if (n == 0) return 0; - - *contour_lengths = (int *) STBTT_malloc(sizeof(**contour_lengths) * n, userdata); - - if (*contour_lengths == 0) { - *num_contours = 0; - return 0; - } - - // make two passes through the points so we don't need to realloc - for (pass=0; pass < 2; ++pass) { - float x=0,y=0; - if (pass == 1) { - points = (stbtt__point *) STBTT_malloc(num_points * sizeof(points[0]), userdata); - if (points == NULL) goto error; - } - num_points = 0; - n= -1; - for (i=0; i < num_verts; ++i) { - switch (vertices[i].type) { - case STBTT_vmove: - // start the next contour - if (n >= 0) - (*contour_lengths)[n] = num_points - start; - ++n; - start = num_points; - - x = vertices[i].x, y = vertices[i].y; - stbtt__add_point(points, num_points++, x,y); - break; - case STBTT_vline: - x = vertices[i].x, y = vertices[i].y; - stbtt__add_point(points, num_points++, x, y); - break; - case STBTT_vcurve: - stbtt__tesselate_curve(points, &num_points, x,y, - vertices[i].cx, vertices[i].cy, - vertices[i].x, vertices[i].y, - objspace_flatness_squared, 0); - x = vertices[i].x, y = vertices[i].y; - break; - case STBTT_vcubic: - stbtt__tesselate_cubic(points, &num_points, x,y, - vertices[i].cx, vertices[i].cy, - vertices[i].cx1, vertices[i].cy1, - vertices[i].x, vertices[i].y, - objspace_flatness_squared, 0); - x = vertices[i].x, y = vertices[i].y; - break; - } - } - (*contour_lengths)[n] = num_points - start; - } - - return points; -error: - STBTT_free(points, userdata); - STBTT_free(*contour_lengths, userdata); - *contour_lengths = 0; - *num_contours = 0; - return NULL; -} - -STBTT_DEF void stbtt_Rasterize(stbtt__bitmap *result, float flatness_in_pixels, stbtt_vertex *vertices, int num_verts, float scale_x, float scale_y, float shift_x, float shift_y, int x_off, int y_off, int invert, void *userdata) -{ - float scale = scale_x > scale_y ? scale_y : scale_x; - int winding_count = 0; - int *winding_lengths = NULL; - stbtt__point *windings = stbtt_FlattenCurves(vertices, num_verts, flatness_in_pixels / scale, &winding_lengths, &winding_count, userdata); - if (windings) { - stbtt__rasterize(result, windings, winding_lengths, winding_count, scale_x, scale_y, shift_x, shift_y, x_off, y_off, invert, userdata); - STBTT_free(winding_lengths, userdata); - STBTT_free(windings, userdata); - } -} - -STBTT_DEF void stbtt_FreeBitmap(unsigned char *bitmap, void *userdata) -{ - STBTT_free(bitmap, userdata); -} - -STBTT_DEF unsigned char *stbtt_GetGlyphBitmapSubpixel(const stbtt_fontinfo *info, float scale_x, float scale_y, float shift_x, float shift_y, int glyph, int *width, int *height, int *xoff, int *yoff) -{ - int ix0,iy0,ix1,iy1; - stbtt__bitmap gbm; - stbtt_vertex *vertices; - int num_verts = stbtt_GetGlyphShape(info, glyph, &vertices); - - if (scale_x == 0) scale_x = scale_y; - if (scale_y == 0) { - if (scale_x == 0) { - STBTT_free(vertices, info->userdata); - return NULL; - } - scale_y = scale_x; - } - - stbtt_GetGlyphBitmapBoxSubpixel(info, glyph, scale_x, scale_y, shift_x, shift_y, &ix0,&iy0,&ix1,&iy1); - - // now we get the size - gbm.w = (ix1 - ix0); - gbm.h = (iy1 - iy0); - gbm.pixels = NULL; // in case we error - - if (width ) *width = gbm.w; - if (height) *height = gbm.h; - if (xoff ) *xoff = ix0; - if (yoff ) *yoff = iy0; - - if (gbm.w && gbm.h) { - gbm.pixels = (unsigned char *) STBTT_malloc(gbm.w * gbm.h, info->userdata); - if (gbm.pixels) { - gbm.stride = gbm.w; - - stbtt_Rasterize(&gbm, 0.35f, vertices, num_verts, scale_x, scale_y, shift_x, shift_y, ix0, iy0, 1, info->userdata); - } - } - STBTT_free(vertices, info->userdata); - return gbm.pixels; -} - -STBTT_DEF unsigned char *stbtt_GetGlyphBitmap(const stbtt_fontinfo *info, float scale_x, float scale_y, int glyph, int *width, int *height, int *xoff, int *yoff) -{ - return stbtt_GetGlyphBitmapSubpixel(info, scale_x, scale_y, 0.0f, 0.0f, glyph, width, height, xoff, yoff); -} - -STBTT_DEF void stbtt_MakeGlyphBitmapSubpixel(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int glyph) -{ - int ix0,iy0; - stbtt_vertex *vertices; - int num_verts = stbtt_GetGlyphShape(info, glyph, &vertices); - stbtt__bitmap gbm; - - stbtt_GetGlyphBitmapBoxSubpixel(info, glyph, scale_x, scale_y, shift_x, shift_y, &ix0,&iy0,0,0); - gbm.pixels = output; - gbm.w = out_w; - gbm.h = out_h; - gbm.stride = out_stride; - - if (gbm.w && gbm.h) - stbtt_Rasterize(&gbm, 0.35f, vertices, num_verts, scale_x, scale_y, shift_x, shift_y, ix0,iy0, 1, info->userdata); - - STBTT_free(vertices, info->userdata); -} - -STBTT_DEF void stbtt_MakeGlyphBitmap(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, int glyph) -{ - stbtt_MakeGlyphBitmapSubpixel(info, output, out_w, out_h, out_stride, scale_x, scale_y, 0.0f,0.0f, glyph); -} - -STBTT_DEF unsigned char *stbtt_GetCodepointBitmapSubpixel(const stbtt_fontinfo *info, float scale_x, float scale_y, float shift_x, float shift_y, int codepoint, int *width, int *height, int *xoff, int *yoff) -{ - return stbtt_GetGlyphBitmapSubpixel(info, scale_x, scale_y,shift_x,shift_y, stbtt_FindGlyphIndex(info,codepoint), width,height,xoff,yoff); -} - -STBTT_DEF void stbtt_MakeCodepointBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int oversample_x, int oversample_y, float *sub_x, float *sub_y, int codepoint) -{ - stbtt_MakeGlyphBitmapSubpixelPrefilter(info, output, out_w, out_h, out_stride, scale_x, scale_y, shift_x, shift_y, oversample_x, oversample_y, sub_x, sub_y, stbtt_FindGlyphIndex(info,codepoint)); -} - -STBTT_DEF void stbtt_MakeCodepointBitmapSubpixel(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int codepoint) -{ - stbtt_MakeGlyphBitmapSubpixel(info, output, out_w, out_h, out_stride, scale_x, scale_y, shift_x, shift_y, stbtt_FindGlyphIndex(info,codepoint)); -} - -STBTT_DEF unsigned char *stbtt_GetCodepointBitmap(const stbtt_fontinfo *info, float scale_x, float scale_y, int codepoint, int *width, int *height, int *xoff, int *yoff) -{ - return stbtt_GetCodepointBitmapSubpixel(info, scale_x, scale_y, 0.0f,0.0f, codepoint, width,height,xoff,yoff); -} - -STBTT_DEF void stbtt_MakeCodepointBitmap(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, int codepoint) -{ - stbtt_MakeCodepointBitmapSubpixel(info, output, out_w, out_h, out_stride, scale_x, scale_y, 0.0f,0.0f, codepoint); -} - -////////////////////////////////////////////////////////////////////////////// -// -// bitmap baking -// -// This is SUPER-CRAPPY packing to keep source code small - -static int stbtt_BakeFontBitmap_internal(unsigned char *data, int offset, // font location (use offset=0 for plain .ttf) - float pixel_height, // height of font in pixels - unsigned char *pixels, int pw, int ph, // bitmap to be filled in - int first_char, int num_chars, // characters to bake - stbtt_bakedchar *chardata) -{ - float scale; - int x,y,bottom_y, i; - stbtt_fontinfo f; - f.userdata = NULL; - if (!stbtt_InitFont(&f, data, offset)) - return -1; - STBTT_memset(pixels, 0, pw*ph); // background of 0 around pixels - x=y=1; - bottom_y = 1; - - scale = stbtt_ScaleForPixelHeight(&f, pixel_height); - - for (i=0; i < num_chars; ++i) { - int advance, lsb, x0,y0,x1,y1,gw,gh; - int g = stbtt_FindGlyphIndex(&f, first_char + i); - stbtt_GetGlyphHMetrics(&f, g, &advance, &lsb); - stbtt_GetGlyphBitmapBox(&f, g, scale,scale, &x0,&y0,&x1,&y1); - gw = x1-x0; - gh = y1-y0; - if (x + gw + 1 >= pw) - y = bottom_y, x = 1; // advance to next row - if (y + gh + 1 >= ph) // check if it fits vertically AFTER potentially moving to next row - return -i; - STBTT_assert(x+gw < pw); - STBTT_assert(y+gh < ph); - stbtt_MakeGlyphBitmap(&f, pixels+x+y*pw, gw,gh,pw, scale,scale, g); - chardata[i].x0 = (stbtt_int16) x; - chardata[i].y0 = (stbtt_int16) y; - chardata[i].x1 = (stbtt_int16) (x + gw); - chardata[i].y1 = (stbtt_int16) (y + gh); - chardata[i].xadvance = scale * advance; - chardata[i].xoff = (float) x0; - chardata[i].yoff = (float) y0; - x = x + gw + 1; - if (y+gh+1 > bottom_y) - bottom_y = y+gh+1; - } - return bottom_y; -} - -STBTT_DEF void stbtt_GetBakedQuad(const stbtt_bakedchar *chardata, int pw, int ph, int char_index, float *xpos, float *ypos, stbtt_aligned_quad *q, int opengl_fillrule) -{ - float d3d_bias = opengl_fillrule ? 0 : -0.5f; - float ipw = 1.0f / pw, iph = 1.0f / ph; - const stbtt_bakedchar *b = chardata + char_index; - int round_x = STBTT_ifloor((*xpos + b->xoff) + 0.5f); - int round_y = STBTT_ifloor((*ypos + b->yoff) + 0.5f); - - q->x0 = round_x + d3d_bias; - q->y0 = round_y + d3d_bias; - q->x1 = round_x + b->x1 - b->x0 + d3d_bias; - q->y1 = round_y + b->y1 - b->y0 + d3d_bias; - - q->s0 = b->x0 * ipw; - q->t0 = b->y0 * iph; - q->s1 = b->x1 * ipw; - q->t1 = b->y1 * iph; - - *xpos += b->xadvance; -} - -////////////////////////////////////////////////////////////////////////////// -// -// rectangle packing replacement routines if you don't have stb_rect_pack.h -// - -#ifndef STB_RECT_PACK_VERSION - -typedef int stbrp_coord; - -//////////////////////////////////////////////////////////////////////////////////// -// // -// // -// COMPILER WARNING ?!?!? // -// // -// // -// if you get a compile warning due to these symbols being defined more than // -// once, move #include "stb_rect_pack.h" before #include "stb_truetype.h" // -// // -//////////////////////////////////////////////////////////////////////////////////// - -typedef struct -{ - int width,height; - int x,y,bottom_y; -} stbrp_context; - -typedef struct -{ - unsigned char x; -} stbrp_node; - -struct stbrp_rect -{ - stbrp_coord x,y; - int id,w,h,was_packed; -}; - -static void stbrp_init_target(stbrp_context *con, int pw, int ph, stbrp_node *nodes, int num_nodes) -{ - con->width = pw; - con->height = ph; - con->x = 0; - con->y = 0; - con->bottom_y = 0; - STBTT__NOTUSED(nodes); - STBTT__NOTUSED(num_nodes); -} - -static void stbrp_pack_rects(stbrp_context *con, stbrp_rect *rects, int num_rects) -{ - int i; - for (i=0; i < num_rects; ++i) { - if (con->x + rects[i].w > con->width) { - con->x = 0; - con->y = con->bottom_y; - } - if (con->y + rects[i].h > con->height) - break; - rects[i].x = con->x; - rects[i].y = con->y; - rects[i].was_packed = 1; - con->x += rects[i].w; - if (con->y + rects[i].h > con->bottom_y) - con->bottom_y = con->y + rects[i].h; - } - for ( ; i < num_rects; ++i) - rects[i].was_packed = 0; -} -#endif - -////////////////////////////////////////////////////////////////////////////// -// -// bitmap baking -// -// This is SUPER-AWESOME (tm Ryan Gordon) packing using stb_rect_pack.h. If -// stb_rect_pack.h isn't available, it uses the BakeFontBitmap strategy. - -STBTT_DEF int stbtt_PackBegin(stbtt_pack_context *spc, unsigned char *pixels, int pw, int ph, int stride_in_bytes, int padding, void *alloc_context) -{ - stbrp_context *context = (stbrp_context *) STBTT_malloc(sizeof(*context) ,alloc_context); - int num_nodes = pw - padding; - stbrp_node *nodes = (stbrp_node *) STBTT_malloc(sizeof(*nodes ) * num_nodes,alloc_context); - - if (context == NULL || nodes == NULL) { - if (context != NULL) STBTT_free(context, alloc_context); - if (nodes != NULL) STBTT_free(nodes , alloc_context); - return 0; - } - - spc->user_allocator_context = alloc_context; - spc->width = pw; - spc->height = ph; - spc->pixels = pixels; - spc->pack_info = context; - spc->nodes = nodes; - spc->padding = padding; - spc->stride_in_bytes = stride_in_bytes != 0 ? stride_in_bytes : pw; - spc->h_oversample = 1; - spc->v_oversample = 1; - spc->skip_missing = 0; - - stbrp_init_target(context, pw-padding, ph-padding, nodes, num_nodes); - - if (pixels) - STBTT_memset(pixels, 0, pw*ph); // background of 0 around pixels - - return 1; -} - -STBTT_DEF void stbtt_PackEnd (stbtt_pack_context *spc) -{ - STBTT_free(spc->nodes , spc->user_allocator_context); - STBTT_free(spc->pack_info, spc->user_allocator_context); -} - -STBTT_DEF void stbtt_PackSetOversampling(stbtt_pack_context *spc, unsigned int h_oversample, unsigned int v_oversample) -{ - STBTT_assert(h_oversample <= STBTT_MAX_OVERSAMPLE); - STBTT_assert(v_oversample <= STBTT_MAX_OVERSAMPLE); - if (h_oversample <= STBTT_MAX_OVERSAMPLE) - spc->h_oversample = h_oversample; - if (v_oversample <= STBTT_MAX_OVERSAMPLE) - spc->v_oversample = v_oversample; -} - -STBTT_DEF void stbtt_PackSetSkipMissingCodepoints(stbtt_pack_context *spc, int skip) -{ - spc->skip_missing = skip; -} - -#define STBTT__OVER_MASK (STBTT_MAX_OVERSAMPLE-1) - -static void stbtt__h_prefilter(unsigned char *pixels, int w, int h, int stride_in_bytes, unsigned int kernel_width) -{ - unsigned char buffer[STBTT_MAX_OVERSAMPLE]; - int safe_w = w - kernel_width; - int j; - STBTT_memset(buffer, 0, STBTT_MAX_OVERSAMPLE); // suppress bogus warning from VS2013 -analyze - for (j=0; j < h; ++j) { - int i; - unsigned int total; - STBTT_memset(buffer, 0, kernel_width); - - total = 0; - - // make kernel_width a constant in common cases so compiler can optimize out the divide - switch (kernel_width) { - case 2: - for (i=0; i <= safe_w; ++i) { - total += pixels[i] - buffer[i & STBTT__OVER_MASK]; - buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i]; - pixels[i] = (unsigned char) (total / 2); - } - break; - case 3: - for (i=0; i <= safe_w; ++i) { - total += pixels[i] - buffer[i & STBTT__OVER_MASK]; - buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i]; - pixels[i] = (unsigned char) (total / 3); - } - break; - case 4: - for (i=0; i <= safe_w; ++i) { - total += pixels[i] - buffer[i & STBTT__OVER_MASK]; - buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i]; - pixels[i] = (unsigned char) (total / 4); - } - break; - case 5: - for (i=0; i <= safe_w; ++i) { - total += pixels[i] - buffer[i & STBTT__OVER_MASK]; - buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i]; - pixels[i] = (unsigned char) (total / 5); - } - break; - default: - for (i=0; i <= safe_w; ++i) { - total += pixels[i] - buffer[i & STBTT__OVER_MASK]; - buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i]; - pixels[i] = (unsigned char) (total / kernel_width); - } - break; - } - - for (; i < w; ++i) { - STBTT_assert(pixels[i] == 0); - total -= buffer[i & STBTT__OVER_MASK]; - pixels[i] = (unsigned char) (total / kernel_width); - } - - pixels += stride_in_bytes; - } -} - -static void stbtt__v_prefilter(unsigned char *pixels, int w, int h, int stride_in_bytes, unsigned int kernel_width) -{ - unsigned char buffer[STBTT_MAX_OVERSAMPLE]; - int safe_h = h - kernel_width; - int j; - STBTT_memset(buffer, 0, STBTT_MAX_OVERSAMPLE); // suppress bogus warning from VS2013 -analyze - for (j=0; j < w; ++j) { - int i; - unsigned int total; - STBTT_memset(buffer, 0, kernel_width); - - total = 0; - - // make kernel_width a constant in common cases so compiler can optimize out the divide - switch (kernel_width) { - case 2: - for (i=0; i <= safe_h; ++i) { - total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK]; - buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes]; - pixels[i*stride_in_bytes] = (unsigned char) (total / 2); - } - break; - case 3: - for (i=0; i <= safe_h; ++i) { - total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK]; - buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes]; - pixels[i*stride_in_bytes] = (unsigned char) (total / 3); - } - break; - case 4: - for (i=0; i <= safe_h; ++i) { - total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK]; - buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes]; - pixels[i*stride_in_bytes] = (unsigned char) (total / 4); - } - break; - case 5: - for (i=0; i <= safe_h; ++i) { - total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK]; - buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes]; - pixels[i*stride_in_bytes] = (unsigned char) (total / 5); - } - break; - default: - for (i=0; i <= safe_h; ++i) { - total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK]; - buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes]; - pixels[i*stride_in_bytes] = (unsigned char) (total / kernel_width); - } - break; - } - - for (; i < h; ++i) { - STBTT_assert(pixels[i*stride_in_bytes] == 0); - total -= buffer[i & STBTT__OVER_MASK]; - pixels[i*stride_in_bytes] = (unsigned char) (total / kernel_width); - } - - pixels += 1; - } -} - -static float stbtt__oversample_shift(int oversample) -{ - if (!oversample) - return 0.0f; - - // The prefilter is a box filter of width "oversample", - // which shifts phase by (oversample - 1)/2 pixels in - // oversampled space. We want to shift in the opposite - // direction to counter this. - return (float)-(oversample - 1) / (2.0f * (float)oversample); -} - -// rects array must be big enough to accommodate all characters in the given ranges -STBTT_DEF int stbtt_PackFontRangesGatherRects(stbtt_pack_context *spc, const stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects) -{ - int i,j,k; - int missing_glyph_added = 0; - - k=0; - for (i=0; i < num_ranges; ++i) { - float fh = ranges[i].font_size; - float scale = fh > 0 ? stbtt_ScaleForPixelHeight(info, fh) : stbtt_ScaleForMappingEmToPixels(info, -fh); - ranges[i].h_oversample = (unsigned char) spc->h_oversample; - ranges[i].v_oversample = (unsigned char) spc->v_oversample; - for (j=0; j < ranges[i].num_chars; ++j) { - int x0,y0,x1,y1; - int codepoint = ranges[i].array_of_unicode_codepoints == NULL ? ranges[i].first_unicode_codepoint_in_range + j : ranges[i].array_of_unicode_codepoints[j]; - int glyph = stbtt_FindGlyphIndex(info, codepoint); - if (glyph == 0 && (spc->skip_missing || missing_glyph_added)) { - rects[k].w = rects[k].h = 0; - } else { - stbtt_GetGlyphBitmapBoxSubpixel(info,glyph, - scale * spc->h_oversample, - scale * spc->v_oversample, - 0,0, - &x0,&y0,&x1,&y1); - rects[k].w = (stbrp_coord) (x1-x0 + spc->padding + spc->h_oversample-1); - rects[k].h = (stbrp_coord) (y1-y0 + spc->padding + spc->v_oversample-1); - if (glyph == 0) - missing_glyph_added = 1; - } - ++k; - } - } - - return k; -} - -STBTT_DEF void stbtt_MakeGlyphBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int prefilter_x, int prefilter_y, float *sub_x, float *sub_y, int glyph) -{ - stbtt_MakeGlyphBitmapSubpixel(info, - output, - out_w - (prefilter_x - 1), - out_h - (prefilter_y - 1), - out_stride, - scale_x, - scale_y, - shift_x, - shift_y, - glyph); - - if (prefilter_x > 1) - stbtt__h_prefilter(output, out_w, out_h, out_stride, prefilter_x); - - if (prefilter_y > 1) - stbtt__v_prefilter(output, out_w, out_h, out_stride, prefilter_y); - - *sub_x = stbtt__oversample_shift(prefilter_x); - *sub_y = stbtt__oversample_shift(prefilter_y); -} - -// rects array must be big enough to accommodate all characters in the given ranges -STBTT_DEF int stbtt_PackFontRangesRenderIntoRects(stbtt_pack_context *spc, const stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects) -{ - int i,j,k, missing_glyph = -1, return_value = 1; - - // save current values - int old_h_over = spc->h_oversample; - int old_v_over = spc->v_oversample; - - k = 0; - for (i=0; i < num_ranges; ++i) { - float fh = ranges[i].font_size; - float scale = fh > 0 ? stbtt_ScaleForPixelHeight(info, fh) : stbtt_ScaleForMappingEmToPixels(info, -fh); - float recip_h,recip_v,sub_x,sub_y; - spc->h_oversample = ranges[i].h_oversample; - spc->v_oversample = ranges[i].v_oversample; - recip_h = 1.0f / spc->h_oversample; - recip_v = 1.0f / spc->v_oversample; - sub_x = stbtt__oversample_shift(spc->h_oversample); - sub_y = stbtt__oversample_shift(spc->v_oversample); - for (j=0; j < ranges[i].num_chars; ++j) { - stbrp_rect *r = &rects[k]; - if (r->was_packed && r->w != 0 && r->h != 0) { - stbtt_packedchar *bc = &ranges[i].chardata_for_range[j]; - int advance, lsb, x0,y0,x1,y1; - int codepoint = ranges[i].array_of_unicode_codepoints == NULL ? ranges[i].first_unicode_codepoint_in_range + j : ranges[i].array_of_unicode_codepoints[j]; - int glyph = stbtt_FindGlyphIndex(info, codepoint); - stbrp_coord pad = (stbrp_coord) spc->padding; - - // pad on left and top - r->x += pad; - r->y += pad; - r->w -= pad; - r->h -= pad; - stbtt_GetGlyphHMetrics(info, glyph, &advance, &lsb); - stbtt_GetGlyphBitmapBox(info, glyph, - scale * spc->h_oversample, - scale * spc->v_oversample, - &x0,&y0,&x1,&y1); - stbtt_MakeGlyphBitmapSubpixel(info, - spc->pixels + r->x + r->y*spc->stride_in_bytes, - r->w - spc->h_oversample+1, - r->h - spc->v_oversample+1, - spc->stride_in_bytes, - scale * spc->h_oversample, - scale * spc->v_oversample, - 0,0, - glyph); - - if (spc->h_oversample > 1) - stbtt__h_prefilter(spc->pixels + r->x + r->y*spc->stride_in_bytes, - r->w, r->h, spc->stride_in_bytes, - spc->h_oversample); - - if (spc->v_oversample > 1) - stbtt__v_prefilter(spc->pixels + r->x + r->y*spc->stride_in_bytes, - r->w, r->h, spc->stride_in_bytes, - spc->v_oversample); - - bc->x0 = (stbtt_int16) r->x; - bc->y0 = (stbtt_int16) r->y; - bc->x1 = (stbtt_int16) (r->x + r->w); - bc->y1 = (stbtt_int16) (r->y + r->h); - bc->xadvance = scale * advance; - bc->xoff = (float) x0 * recip_h + sub_x; - bc->yoff = (float) y0 * recip_v + sub_y; - bc->xoff2 = (x0 + r->w) * recip_h + sub_x; - bc->yoff2 = (y0 + r->h) * recip_v + sub_y; - - if (glyph == 0) - missing_glyph = j; - } else if (spc->skip_missing) { - return_value = 0; - } else if (r->was_packed && r->w == 0 && r->h == 0 && missing_glyph >= 0) { - ranges[i].chardata_for_range[j] = ranges[i].chardata_for_range[missing_glyph]; - } else { - return_value = 0; // if any fail, report failure - } - - ++k; - } - } - - // restore original values - spc->h_oversample = old_h_over; - spc->v_oversample = old_v_over; - - return return_value; -} - -STBTT_DEF void stbtt_PackFontRangesPackRects(stbtt_pack_context *spc, stbrp_rect *rects, int num_rects) -{ - stbrp_pack_rects((stbrp_context *) spc->pack_info, rects, num_rects); -} - -STBTT_DEF int stbtt_PackFontRanges(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, stbtt_pack_range *ranges, int num_ranges) -{ - stbtt_fontinfo info; - int i,j,n, return_value = 1; - //stbrp_context *context = (stbrp_context *) spc->pack_info; - stbrp_rect *rects; - - // flag all characters as NOT packed - for (i=0; i < num_ranges; ++i) - for (j=0; j < ranges[i].num_chars; ++j) - ranges[i].chardata_for_range[j].x0 = - ranges[i].chardata_for_range[j].y0 = - ranges[i].chardata_for_range[j].x1 = - ranges[i].chardata_for_range[j].y1 = 0; - - n = 0; - for (i=0; i < num_ranges; ++i) - n += ranges[i].num_chars; - - rects = (stbrp_rect *) STBTT_malloc(sizeof(*rects) * n, spc->user_allocator_context); - if (rects == NULL) - return 0; - - info.userdata = spc->user_allocator_context; - stbtt_InitFont(&info, fontdata, stbtt_GetFontOffsetForIndex(fontdata,font_index)); - - n = stbtt_PackFontRangesGatherRects(spc, &info, ranges, num_ranges, rects); - - stbtt_PackFontRangesPackRects(spc, rects, n); - - return_value = stbtt_PackFontRangesRenderIntoRects(spc, &info, ranges, num_ranges, rects); - - STBTT_free(rects, spc->user_allocator_context); - return return_value; -} - -STBTT_DEF int stbtt_PackFontRange(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, float font_size, - int first_unicode_codepoint_in_range, int num_chars_in_range, stbtt_packedchar *chardata_for_range) -{ - stbtt_pack_range range; - range.first_unicode_codepoint_in_range = first_unicode_codepoint_in_range; - range.array_of_unicode_codepoints = NULL; - range.num_chars = num_chars_in_range; - range.chardata_for_range = chardata_for_range; - range.font_size = font_size; - return stbtt_PackFontRanges(spc, fontdata, font_index, &range, 1); -} - -STBTT_DEF void stbtt_GetScaledFontVMetrics(const unsigned char *fontdata, int index, float size, float *ascent, float *descent, float *lineGap) -{ - int i_ascent, i_descent, i_lineGap; - float scale; - stbtt_fontinfo info; - stbtt_InitFont(&info, fontdata, stbtt_GetFontOffsetForIndex(fontdata, index)); - scale = size > 0 ? stbtt_ScaleForPixelHeight(&info, size) : stbtt_ScaleForMappingEmToPixels(&info, -size); - stbtt_GetFontVMetrics(&info, &i_ascent, &i_descent, &i_lineGap); - *ascent = (float) i_ascent * scale; - *descent = (float) i_descent * scale; - *lineGap = (float) i_lineGap * scale; -} - -STBTT_DEF void stbtt_GetPackedQuad(const stbtt_packedchar *chardata, int pw, int ph, int char_index, float *xpos, float *ypos, stbtt_aligned_quad *q, int align_to_integer) -{ - float ipw = 1.0f / pw, iph = 1.0f / ph; - const stbtt_packedchar *b = chardata + char_index; - - if (align_to_integer) { - float x = (float) STBTT_ifloor((*xpos + b->xoff) + 0.5f); - float y = (float) STBTT_ifloor((*ypos + b->yoff) + 0.5f); - q->x0 = x; - q->y0 = y; - q->x1 = x + b->xoff2 - b->xoff; - q->y1 = y + b->yoff2 - b->yoff; - } else { - q->x0 = *xpos + b->xoff; - q->y0 = *ypos + b->yoff; - q->x1 = *xpos + b->xoff2; - q->y1 = *ypos + b->yoff2; - } - - q->s0 = b->x0 * ipw; - q->t0 = b->y0 * iph; - q->s1 = b->x1 * ipw; - q->t1 = b->y1 * iph; - - *xpos += b->xadvance; -} - -////////////////////////////////////////////////////////////////////////////// -// -// sdf computation -// - -#define STBTT_min(a,b) ((a) < (b) ? (a) : (b)) -#define STBTT_max(a,b) ((a) < (b) ? (b) : (a)) - -static int stbtt__ray_intersect_bezier(float orig[2], float ray[2], float q0[2], float q1[2], float q2[2], float hits[2][2]) -{ - float q0perp = q0[1]*ray[0] - q0[0]*ray[1]; - float q1perp = q1[1]*ray[0] - q1[0]*ray[1]; - float q2perp = q2[1]*ray[0] - q2[0]*ray[1]; - float roperp = orig[1]*ray[0] - orig[0]*ray[1]; - - float a = q0perp - 2*q1perp + q2perp; - float b = q1perp - q0perp; - float c = q0perp - roperp; - - float s0 = 0., s1 = 0.; - int num_s = 0; - - if (a != 0.0) { - float discr = b*b - a*c; - if (discr > 0.0) { - float rcpna = -1 / a; - float d = (float) STBTT_sqrt(discr); - s0 = (b+d) * rcpna; - s1 = (b-d) * rcpna; - if (s0 >= 0.0 && s0 <= 1.0) - num_s = 1; - if (d > 0.0 && s1 >= 0.0 && s1 <= 1.0) { - if (num_s == 0) s0 = s1; - ++num_s; - } - } - } else { - // 2*b*s + c = 0 - // s = -c / (2*b) - s0 = c / (-2 * b); - if (s0 >= 0.0 && s0 <= 1.0) - num_s = 1; - } - - if (num_s == 0) - return 0; - else { - float rcp_len2 = 1 / (ray[0]*ray[0] + ray[1]*ray[1]); - float rayn_x = ray[0] * rcp_len2, rayn_y = ray[1] * rcp_len2; - - float q0d = q0[0]*rayn_x + q0[1]*rayn_y; - float q1d = q1[0]*rayn_x + q1[1]*rayn_y; - float q2d = q2[0]*rayn_x + q2[1]*rayn_y; - float rod = orig[0]*rayn_x + orig[1]*rayn_y; - - float q10d = q1d - q0d; - float q20d = q2d - q0d; - float q0rd = q0d - rod; - - hits[0][0] = q0rd + s0*(2.0f - 2.0f*s0)*q10d + s0*s0*q20d; - hits[0][1] = a*s0+b; - - if (num_s > 1) { - hits[1][0] = q0rd + s1*(2.0f - 2.0f*s1)*q10d + s1*s1*q20d; - hits[1][1] = a*s1+b; - return 2; - } else { - return 1; - } - } -} - -static int equal(float *a, float *b) -{ - return (a[0] == b[0] && a[1] == b[1]); -} - -static int stbtt__compute_crossings_x(float x, float y, int nverts, stbtt_vertex *verts) -{ - int i; - float orig[2], ray[2] = { 1, 0 }; - float y_frac; - int winding = 0; - - // make sure y never passes through a vertex of the shape - y_frac = (float) STBTT_fmod(y, 1.0f); - if (y_frac < 0.01f) - y += 0.01f; - else if (y_frac > 0.99f) - y -= 0.01f; - - orig[0] = x; - orig[1] = y; - - // test a ray from (-infinity,y) to (x,y) - for (i=0; i < nverts; ++i) { - if (verts[i].type == STBTT_vline) { - int x0 = (int) verts[i-1].x, y0 = (int) verts[i-1].y; - int x1 = (int) verts[i ].x, y1 = (int) verts[i ].y; - if (y > STBTT_min(y0,y1) && y < STBTT_max(y0,y1) && x > STBTT_min(x0,x1)) { - float x_inter = (y - y0) / (y1 - y0) * (x1-x0) + x0; - if (x_inter < x) - winding += (y0 < y1) ? 1 : -1; - } - } - if (verts[i].type == STBTT_vcurve) { - int x0 = (int) verts[i-1].x , y0 = (int) verts[i-1].y ; - int x1 = (int) verts[i ].cx, y1 = (int) verts[i ].cy; - int x2 = (int) verts[i ].x , y2 = (int) verts[i ].y ; - int ax = STBTT_min(x0,STBTT_min(x1,x2)), ay = STBTT_min(y0,STBTT_min(y1,y2)); - int by = STBTT_max(y0,STBTT_max(y1,y2)); - if (y > ay && y < by && x > ax) { - float q0[2],q1[2],q2[2]; - float hits[2][2]; - q0[0] = (float)x0; - q0[1] = (float)y0; - q1[0] = (float)x1; - q1[1] = (float)y1; - q2[0] = (float)x2; - q2[1] = (float)y2; - if (equal(q0,q1) || equal(q1,q2)) { - x0 = (int)verts[i-1].x; - y0 = (int)verts[i-1].y; - x1 = (int)verts[i ].x; - y1 = (int)verts[i ].y; - if (y > STBTT_min(y0,y1) && y < STBTT_max(y0,y1) && x > STBTT_min(x0,x1)) { - float x_inter = (y - y0) / (y1 - y0) * (x1-x0) + x0; - if (x_inter < x) - winding += (y0 < y1) ? 1 : -1; - } - } else { - int num_hits = stbtt__ray_intersect_bezier(orig, ray, q0, q1, q2, hits); - if (num_hits >= 1) - if (hits[0][0] < 0) - winding += (hits[0][1] < 0 ? -1 : 1); - if (num_hits >= 2) - if (hits[1][0] < 0) - winding += (hits[1][1] < 0 ? -1 : 1); - } - } - } - } - return winding; -} - -static float stbtt__cuberoot( float x ) -{ - if (x<0) - return -(float) STBTT_pow(-x,1.0f/3.0f); - else - return (float) STBTT_pow( x,1.0f/3.0f); -} - -// x^3 + a*x^2 + b*x + c = 0 -static int stbtt__solve_cubic(float a, float b, float c, float* r) -{ - float s = -a / 3; - float p = b - a*a / 3; - float q = a * (2*a*a - 9*b) / 27 + c; - float p3 = p*p*p; - float d = q*q + 4*p3 / 27; - if (d >= 0) { - float z = (float) STBTT_sqrt(d); - float u = (-q + z) / 2; - float v = (-q - z) / 2; - u = stbtt__cuberoot(u); - v = stbtt__cuberoot(v); - r[0] = s + u + v; - return 1; - } else { - float u = (float) STBTT_sqrt(-p/3); - float v = (float) STBTT_acos(-STBTT_sqrt(-27/p3) * q / 2) / 3; // p3 must be negative, since d is negative - float m = (float) STBTT_cos(v); - float n = (float) STBTT_cos(v-3.141592/2)*1.732050808f; - r[0] = s + u * 2 * m; - r[1] = s - u * (m + n); - r[2] = s - u * (m - n); - - //STBTT_assert( STBTT_fabs(((r[0]+a)*r[0]+b)*r[0]+c) < 0.05f); // these asserts may not be safe at all scales, though they're in bezier t parameter units so maybe? - //STBTT_assert( STBTT_fabs(((r[1]+a)*r[1]+b)*r[1]+c) < 0.05f); - //STBTT_assert( STBTT_fabs(((r[2]+a)*r[2]+b)*r[2]+c) < 0.05f); - return 3; - } -} - -STBTT_DEF unsigned char * stbtt_GetGlyphSDF(const stbtt_fontinfo *info, float scale, int glyph, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff) -{ - float scale_x = scale, scale_y = scale; - int ix0,iy0,ix1,iy1; - int w,h; - unsigned char *data; - - if (scale == 0) return NULL; - - stbtt_GetGlyphBitmapBoxSubpixel(info, glyph, scale, scale, 0.0f,0.0f, &ix0,&iy0,&ix1,&iy1); - - // if empty, return NULL - if (ix0 == ix1 || iy0 == iy1) - return NULL; - - ix0 -= padding; - iy0 -= padding; - ix1 += padding; - iy1 += padding; - - w = (ix1 - ix0); - h = (iy1 - iy0); - - if (width ) *width = w; - if (height) *height = h; - if (xoff ) *xoff = ix0; - if (yoff ) *yoff = iy0; - - // invert for y-downwards bitmaps - scale_y = -scale_y; - - { - int x,y,i,j; - float *precompute; - stbtt_vertex *verts; - int num_verts = stbtt_GetGlyphShape(info, glyph, &verts); - data = (unsigned char *) STBTT_malloc(w * h, info->userdata); - precompute = (float *) STBTT_malloc(num_verts * sizeof(float), info->userdata); - - for (i=0,j=num_verts-1; i < num_verts; j=i++) { - if (verts[i].type == STBTT_vline) { - float x0 = verts[i].x*scale_x, y0 = verts[i].y*scale_y; - float x1 = verts[j].x*scale_x, y1 = verts[j].y*scale_y; - float dist = (float) STBTT_sqrt((x1-x0)*(x1-x0) + (y1-y0)*(y1-y0)); - precompute[i] = (dist == 0) ? 0.0f : 1.0f / dist; - } else if (verts[i].type == STBTT_vcurve) { - float x2 = verts[j].x *scale_x, y2 = verts[j].y *scale_y; - float x1 = verts[i].cx*scale_x, y1 = verts[i].cy*scale_y; - float x0 = verts[i].x *scale_x, y0 = verts[i].y *scale_y; - float bx = x0 - 2*x1 + x2, by = y0 - 2*y1 + y2; - float len2 = bx*bx + by*by; - if (len2 != 0.0f) - precompute[i] = 1.0f / (bx*bx + by*by); - else - precompute[i] = 0.0f; - } else - precompute[i] = 0.0f; - } - - for (y=iy0; y < iy1; ++y) { - for (x=ix0; x < ix1; ++x) { - float val; - float min_dist = 999999.0f; - float sx = (float) x + 0.5f; - float sy = (float) y + 0.5f; - float x_gspace = (sx / scale_x); - float y_gspace = (sy / scale_y); - - int winding = stbtt__compute_crossings_x(x_gspace, y_gspace, num_verts, verts); // @OPTIMIZE: this could just be a rasterization, but needs to be line vs. non-tesselated curves so a new path - - for (i=0; i < num_verts; ++i) { - float x0 = verts[i].x*scale_x, y0 = verts[i].y*scale_y; - - if (verts[i].type == STBTT_vline && precompute[i] != 0.0f) { - float x1 = verts[i-1].x*scale_x, y1 = verts[i-1].y*scale_y; - - float dist,dist2 = (x0-sx)*(x0-sx) + (y0-sy)*(y0-sy); - if (dist2 < min_dist*min_dist) - min_dist = (float) STBTT_sqrt(dist2); - - // coarse culling against bbox - //if (sx > STBTT_min(x0,x1)-min_dist && sx < STBTT_max(x0,x1)+min_dist && - // sy > STBTT_min(y0,y1)-min_dist && sy < STBTT_max(y0,y1)+min_dist) - dist = (float) STBTT_fabs((x1-x0)*(y0-sy) - (y1-y0)*(x0-sx)) * precompute[i]; - STBTT_assert(i != 0); - if (dist < min_dist) { - // check position along line - // x' = x0 + t*(x1-x0), y' = y0 + t*(y1-y0) - // minimize (x'-sx)*(x'-sx)+(y'-sy)*(y'-sy) - float dx = x1-x0, dy = y1-y0; - float px = x0-sx, py = y0-sy; - // minimize (px+t*dx)^2 + (py+t*dy)^2 = px*px + 2*px*dx*t + t^2*dx*dx + py*py + 2*py*dy*t + t^2*dy*dy - // derivative: 2*px*dx + 2*py*dy + (2*dx*dx+2*dy*dy)*t, set to 0 and solve - float t = -(px*dx + py*dy) / (dx*dx + dy*dy); - if (t >= 0.0f && t <= 1.0f) - min_dist = dist; - } - } else if (verts[i].type == STBTT_vcurve) { - float x2 = verts[i-1].x *scale_x, y2 = verts[i-1].y *scale_y; - float x1 = verts[i ].cx*scale_x, y1 = verts[i ].cy*scale_y; - float box_x0 = STBTT_min(STBTT_min(x0,x1),x2); - float box_y0 = STBTT_min(STBTT_min(y0,y1),y2); - float box_x1 = STBTT_max(STBTT_max(x0,x1),x2); - float box_y1 = STBTT_max(STBTT_max(y0,y1),y2); - // coarse culling against bbox to avoid computing cubic unnecessarily - if (sx > box_x0-min_dist && sx < box_x1+min_dist && sy > box_y0-min_dist && sy < box_y1+min_dist) { - int num=0; - float ax = x1-x0, ay = y1-y0; - float bx = x0 - 2*x1 + x2, by = y0 - 2*y1 + y2; - float mx = x0 - sx, my = y0 - sy; - float res[3] = {0.f,0.f,0.f}; - float px,py,t,it,dist2; - float a_inv = precompute[i]; - if (a_inv == 0.0) { // if a_inv is 0, it's 2nd degree so use quadratic formula - float a = 3*(ax*bx + ay*by); - float b = 2*(ax*ax + ay*ay) + (mx*bx+my*by); - float c = mx*ax+my*ay; - if (a == 0.0) { // if a is 0, it's linear - if (b != 0.0) { - res[num++] = -c/b; - } - } else { - float discriminant = b*b - 4*a*c; - if (discriminant < 0) - num = 0; - else { - float root = (float) STBTT_sqrt(discriminant); - res[0] = (-b - root)/(2*a); - res[1] = (-b + root)/(2*a); - num = 2; // don't bother distinguishing 1-solution case, as code below will still work - } - } - } else { - float b = 3*(ax*bx + ay*by) * a_inv; // could precompute this as it doesn't depend on sample point - float c = (2*(ax*ax + ay*ay) + (mx*bx+my*by)) * a_inv; - float d = (mx*ax+my*ay) * a_inv; - num = stbtt__solve_cubic(b, c, d, res); - } - dist2 = (x0-sx)*(x0-sx) + (y0-sy)*(y0-sy); - if (dist2 < min_dist*min_dist) - min_dist = (float) STBTT_sqrt(dist2); - - if (num >= 1 && res[0] >= 0.0f && res[0] <= 1.0f) { - t = res[0], it = 1.0f - t; - px = it*it*x0 + 2*t*it*x1 + t*t*x2; - py = it*it*y0 + 2*t*it*y1 + t*t*y2; - dist2 = (px-sx)*(px-sx) + (py-sy)*(py-sy); - if (dist2 < min_dist * min_dist) - min_dist = (float) STBTT_sqrt(dist2); - } - if (num >= 2 && res[1] >= 0.0f && res[1] <= 1.0f) { - t = res[1], it = 1.0f - t; - px = it*it*x0 + 2*t*it*x1 + t*t*x2; - py = it*it*y0 + 2*t*it*y1 + t*t*y2; - dist2 = (px-sx)*(px-sx) + (py-sy)*(py-sy); - if (dist2 < min_dist * min_dist) - min_dist = (float) STBTT_sqrt(dist2); - } - if (num >= 3 && res[2] >= 0.0f && res[2] <= 1.0f) { - t = res[2], it = 1.0f - t; - px = it*it*x0 + 2*t*it*x1 + t*t*x2; - py = it*it*y0 + 2*t*it*y1 + t*t*y2; - dist2 = (px-sx)*(px-sx) + (py-sy)*(py-sy); - if (dist2 < min_dist * min_dist) - min_dist = (float) STBTT_sqrt(dist2); - } - } - } - } - if (winding == 0) - min_dist = -min_dist; // if outside the shape, value is negative - val = onedge_value + pixel_dist_scale * min_dist; - if (val < 0) - val = 0; - else if (val > 255) - val = 255; - data[(y-iy0)*w+(x-ix0)] = (unsigned char) val; - } - } - STBTT_free(precompute, info->userdata); - STBTT_free(verts, info->userdata); - } - return data; -} - -STBTT_DEF unsigned char * stbtt_GetCodepointSDF(const stbtt_fontinfo *info, float scale, int codepoint, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff) -{ - return stbtt_GetGlyphSDF(info, scale, stbtt_FindGlyphIndex(info, codepoint), padding, onedge_value, pixel_dist_scale, width, height, xoff, yoff); -} - -STBTT_DEF void stbtt_FreeSDF(unsigned char *bitmap, void *userdata) -{ - STBTT_free(bitmap, userdata); -} - -////////////////////////////////////////////////////////////////////////////// -// -// font name matching -- recommended not to use this -// - -// check if a utf8 string contains a prefix which is the utf16 string; if so return length of matching utf8 string -static stbtt_int32 stbtt__CompareUTF8toUTF16_bigendian_prefix(stbtt_uint8 *s1, stbtt_int32 len1, stbtt_uint8 *s2, stbtt_int32 len2) -{ - stbtt_int32 i=0; - - // convert utf16 to utf8 and compare the results while converting - while (len2) { - stbtt_uint16 ch = s2[0]*256 + s2[1]; - if (ch < 0x80) { - if (i >= len1) return -1; - if (s1[i++] != ch) return -1; - } else if (ch < 0x800) { - if (i+1 >= len1) return -1; - if (s1[i++] != 0xc0 + (ch >> 6)) return -1; - if (s1[i++] != 0x80 + (ch & 0x3f)) return -1; - } else if (ch >= 0xd800 && ch < 0xdc00) { - stbtt_uint32 c; - stbtt_uint16 ch2 = s2[2]*256 + s2[3]; - if (i+3 >= len1) return -1; - c = ((ch - 0xd800) << 10) + (ch2 - 0xdc00) + 0x10000; - if (s1[i++] != 0xf0 + (c >> 18)) return -1; - if (s1[i++] != 0x80 + ((c >> 12) & 0x3f)) return -1; - if (s1[i++] != 0x80 + ((c >> 6) & 0x3f)) return -1; - if (s1[i++] != 0x80 + ((c ) & 0x3f)) return -1; - s2 += 2; // plus another 2 below - len2 -= 2; - } else if (ch >= 0xdc00 && ch < 0xe000) { - return -1; - } else { - if (i+2 >= len1) return -1; - if (s1[i++] != 0xe0 + (ch >> 12)) return -1; - if (s1[i++] != 0x80 + ((ch >> 6) & 0x3f)) return -1; - if (s1[i++] != 0x80 + ((ch ) & 0x3f)) return -1; - } - s2 += 2; - len2 -= 2; - } - return i; -} - -static int stbtt_CompareUTF8toUTF16_bigendian_internal(char *s1, int len1, char *s2, int len2) -{ - return len1 == stbtt__CompareUTF8toUTF16_bigendian_prefix((stbtt_uint8*) s1, len1, (stbtt_uint8*) s2, len2); -} - -// returns results in whatever encoding you request... but note that 2-byte encodings -// will be BIG-ENDIAN... use stbtt_CompareUTF8toUTF16_bigendian() to compare -STBTT_DEF const char *stbtt_GetFontNameString(const stbtt_fontinfo *font, int *length, int platformID, int encodingID, int languageID, int nameID) -{ - stbtt_int32 i,count,stringOffset; - stbtt_uint8 *fc = font->data; - stbtt_uint32 offset = font->fontstart; - stbtt_uint32 nm = stbtt__find_table(fc, offset, "name"); - if (!nm) return NULL; - - count = ttUSHORT(fc+nm+2); - stringOffset = nm + ttUSHORT(fc+nm+4); - for (i=0; i < count; ++i) { - stbtt_uint32 loc = nm + 6 + 12 * i; - if (platformID == ttUSHORT(fc+loc+0) && encodingID == ttUSHORT(fc+loc+2) - && languageID == ttUSHORT(fc+loc+4) && nameID == ttUSHORT(fc+loc+6)) { - *length = ttUSHORT(fc+loc+8); - return (const char *) (fc+stringOffset+ttUSHORT(fc+loc+10)); - } - } - return NULL; -} - -static int stbtt__matchpair(stbtt_uint8 *fc, stbtt_uint32 nm, stbtt_uint8 *name, stbtt_int32 nlen, stbtt_int32 target_id, stbtt_int32 next_id) -{ - stbtt_int32 i; - stbtt_int32 count = ttUSHORT(fc+nm+2); - stbtt_int32 stringOffset = nm + ttUSHORT(fc+nm+4); - - for (i=0; i < count; ++i) { - stbtt_uint32 loc = nm + 6 + 12 * i; - stbtt_int32 id = ttUSHORT(fc+loc+6); - if (id == target_id) { - // find the encoding - stbtt_int32 platform = ttUSHORT(fc+loc+0), encoding = ttUSHORT(fc+loc+2), language = ttUSHORT(fc+loc+4); - - // is this a Unicode encoding? - if (platform == 0 || (platform == 3 && encoding == 1) || (platform == 3 && encoding == 10)) { - stbtt_int32 slen = ttUSHORT(fc+loc+8); - stbtt_int32 off = ttUSHORT(fc+loc+10); - - // check if there's a prefix match - stbtt_int32 matchlen = stbtt__CompareUTF8toUTF16_bigendian_prefix(name, nlen, fc+stringOffset+off,slen); - if (matchlen >= 0) { - // check for target_id+1 immediately following, with same encoding & language - if (i+1 < count && ttUSHORT(fc+loc+12+6) == next_id && ttUSHORT(fc+loc+12) == platform && ttUSHORT(fc+loc+12+2) == encoding && ttUSHORT(fc+loc+12+4) == language) { - slen = ttUSHORT(fc+loc+12+8); - off = ttUSHORT(fc+loc+12+10); - if (slen == 0) { - if (matchlen == nlen) - return 1; - } else if (matchlen < nlen && name[matchlen] == ' ') { - ++matchlen; - if (stbtt_CompareUTF8toUTF16_bigendian_internal((char*) (name+matchlen), nlen-matchlen, (char*)(fc+stringOffset+off),slen)) - return 1; - } - } else { - // if nothing immediately following - if (matchlen == nlen) - return 1; - } - } - } - - // @TODO handle other encodings - } - } - return 0; -} - -static int stbtt__matches(stbtt_uint8 *fc, stbtt_uint32 offset, stbtt_uint8 *name, stbtt_int32 flags) -{ - stbtt_int32 nlen = (stbtt_int32) STBTT_strlen((char *) name); - stbtt_uint32 nm,hd; - if (!stbtt__isfont(fc+offset)) return 0; - - // check italics/bold/underline flags in macStyle... - if (flags) { - hd = stbtt__find_table(fc, offset, "head"); - if ((ttUSHORT(fc+hd+44) & 7) != (flags & 7)) return 0; - } - - nm = stbtt__find_table(fc, offset, "name"); - if (!nm) return 0; - - if (flags) { - // if we checked the macStyle flags, then just check the family and ignore the subfamily - if (stbtt__matchpair(fc, nm, name, nlen, 16, -1)) return 1; - if (stbtt__matchpair(fc, nm, name, nlen, 1, -1)) return 1; - if (stbtt__matchpair(fc, nm, name, nlen, 3, -1)) return 1; - } else { - if (stbtt__matchpair(fc, nm, name, nlen, 16, 17)) return 1; - if (stbtt__matchpair(fc, nm, name, nlen, 1, 2)) return 1; - if (stbtt__matchpair(fc, nm, name, nlen, 3, -1)) return 1; - } - - return 0; -} - -static int stbtt_FindMatchingFont_internal(unsigned char *font_collection, char *name_utf8, stbtt_int32 flags) -{ - stbtt_int32 i; - for (i=0;;++i) { - stbtt_int32 off = stbtt_GetFontOffsetForIndex(font_collection, i); - if (off < 0) return off; - if (stbtt__matches((stbtt_uint8 *) font_collection, off, (stbtt_uint8*) name_utf8, flags)) - return off; - } -} - -#if defined(__GNUC__) || defined(__clang__) -#pragma GCC diagnostic push -#pragma GCC diagnostic ignored "-Wcast-qual" -#endif - -STBTT_DEF int stbtt_BakeFontBitmap(const unsigned char *data, int offset, - float pixel_height, unsigned char *pixels, int pw, int ph, - int first_char, int num_chars, stbtt_bakedchar *chardata) -{ - return stbtt_BakeFontBitmap_internal((unsigned char *) data, offset, pixel_height, pixels, pw, ph, first_char, num_chars, chardata); -} - -STBTT_DEF int stbtt_GetFontOffsetForIndex(const unsigned char *data, int index) -{ - return stbtt_GetFontOffsetForIndex_internal((unsigned char *) data, index); -} - -STBTT_DEF int stbtt_GetNumberOfFonts(const unsigned char *data) -{ - return stbtt_GetNumberOfFonts_internal((unsigned char *) data); -} - -STBTT_DEF int stbtt_InitFont(stbtt_fontinfo *info, const unsigned char *data, int offset) -{ - return stbtt_InitFont_internal(info, (unsigned char *) data, offset); -} - -STBTT_DEF int stbtt_FindMatchingFont(const unsigned char *fontdata, const char *name, int flags) -{ - return stbtt_FindMatchingFont_internal((unsigned char *) fontdata, (char *) name, flags); -} - -STBTT_DEF int stbtt_CompareUTF8toUTF16_bigendian(const char *s1, int len1, const char *s2, int len2) -{ - return stbtt_CompareUTF8toUTF16_bigendian_internal((char *) s1, len1, (char *) s2, len2); -} - -#if defined(__GNUC__) || defined(__clang__) -#pragma GCC diagnostic pop -#endif - -#endif // STB_TRUETYPE_IMPLEMENTATION - - -// FULL VERSION HISTORY -// -// 1.25 (2021-07-11) many fixes -// 1.24 (2020-02-05) fix warning -// 1.23 (2020-02-02) query SVG data for glyphs; query whole kerning table (but only kern not GPOS) -// 1.22 (2019-08-11) minimize missing-glyph duplication; fix kerning if both 'GPOS' and 'kern' are defined -// 1.21 (2019-02-25) fix warning -// 1.20 (2019-02-07) PackFontRange skips missing codepoints; GetScaleFontVMetrics() -// 1.19 (2018-02-11) OpenType GPOS kerning (horizontal only), STBTT_fmod -// 1.18 (2018-01-29) add missing function -// 1.17 (2017-07-23) make more arguments const; doc fix -// 1.16 (2017-07-12) SDF support -// 1.15 (2017-03-03) make more arguments const -// 1.14 (2017-01-16) num-fonts-in-TTC function -// 1.13 (2017-01-02) support OpenType fonts, certain Apple fonts -// 1.12 (2016-10-25) suppress warnings about casting away const with -Wcast-qual -// 1.11 (2016-04-02) fix unused-variable warning -// 1.10 (2016-04-02) allow user-defined fabs() replacement -// fix memory leak if fontsize=0.0 -// fix warning from duplicate typedef -// 1.09 (2016-01-16) warning fix; avoid crash on outofmem; use alloc userdata for PackFontRanges -// 1.08 (2015-09-13) document stbtt_Rasterize(); fixes for vertical & horizontal edges -// 1.07 (2015-08-01) allow PackFontRanges to accept arrays of sparse codepoints; -// allow PackFontRanges to pack and render in separate phases; -// fix stbtt_GetFontOFfsetForIndex (never worked for non-0 input?); -// fixed an assert() bug in the new rasterizer -// replace assert() with STBTT_assert() in new rasterizer -// 1.06 (2015-07-14) performance improvements (~35% faster on x86 and x64 on test machine) -// also more precise AA rasterizer, except if shapes overlap -// remove need for STBTT_sort -// 1.05 (2015-04-15) fix misplaced definitions for STBTT_STATIC -// 1.04 (2015-04-15) typo in example -// 1.03 (2015-04-12) STBTT_STATIC, fix memory leak in new packing, various fixes -// 1.02 (2014-12-10) fix various warnings & compile issues w/ stb_rect_pack, C++ -// 1.01 (2014-12-08) fix subpixel position when oversampling to exactly match -// non-oversampled; STBTT_POINT_SIZE for packed case only -// 1.00 (2014-12-06) add new PackBegin etc. API, w/ support for oversampling -// 0.99 (2014-09-18) fix multiple bugs with subpixel rendering (ryg) -// 0.9 (2014-08-07) support certain mac/iOS fonts without an MS platformID -// 0.8b (2014-07-07) fix a warning -// 0.8 (2014-05-25) fix a few more warnings -// 0.7 (2013-09-25) bugfix: subpixel glyph bug fixed in 0.5 had come back -// 0.6c (2012-07-24) improve documentation -// 0.6b (2012-07-20) fix a few more warnings -// 0.6 (2012-07-17) fix warnings; added stbtt_ScaleForMappingEmToPixels, -// stbtt_GetFontBoundingBox, stbtt_IsGlyphEmpty -// 0.5 (2011-12-09) bugfixes: -// subpixel glyph renderer computed wrong bounding box -// first vertex of shape can be off-curve (FreeSans) -// 0.4b (2011-12-03) fixed an error in the font baking example -// 0.4 (2011-12-01) kerning, subpixel rendering (tor) -// bugfixes for: -// codepoint-to-glyph conversion using table fmt=12 -// codepoint-to-glyph conversion using table fmt=4 -// stbtt_GetBakedQuad with non-square texture (Zer) -// updated Hello World! sample to use kerning and subpixel -// fixed some warnings -// 0.3 (2009-06-24) cmap fmt=12, compound shapes (MM) -// userdata, malloc-from-userdata, non-zero fill (stb) -// 0.2 (2009-03-11) Fix unsigned/signed char warnings -// 0.1 (2009-03-09) First public release -// - -/* ------------------------------------------------------------------------------- -This software is available under 2 licenses -- choose whichever you prefer. ------------------------------------------------------------------------------- -ALTERNATIVE A - MIT License -Copyright (c) 2017 Sean Barrett -Permission is hereby granted, free of charge, to any person obtaining a copy of -this software and associated documentation files (the "Software"), to deal in -the Software without restriction, including without limitation the rights to -use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies -of the Software, and to permit persons to whom the Software is furnished to do -so, subject to the following conditions: -The above copyright notice and this permission notice shall be included in all -copies or substantial portions of the Software. -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER -LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, -OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE -SOFTWARE. ------------------------------------------------------------------------------- -ALTERNATIVE B - Public Domain (www.unlicense.org) -This is free and unencumbered software released into the public domain. -Anyone is free to copy, modify, publish, use, compile, sell, or distribute this -software, either in source code form or as a compiled binary, for any purpose, -commercial or non-commercial, and by any means. -In jurisdictions that recognize copyright laws, the author or authors of this -software dedicate any and all copyright interest in the software to the public -domain. We make this dedication for the benefit of the public at large and to -the detriment of our heirs and successors. We intend this dedication to be an -overt act of relinquishment in perpetuity of all present and future rights to -this software under copyright law. -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE -AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN -ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION -WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. ------------------------------------------------------------------------------- -*/
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/semicongine/thirdparty/stb/stb_image.h Mon Jul 29 00:00:28 2024 +0700 @@ -0,0 +1,7988 @@ +/* stb_image - v2.30 - public domain image loader - http://nothings.org/stb + no warranty implied; use at your own risk + + Do this: + #define STB_IMAGE_IMPLEMENTATION + before you include this file in *one* C or C++ file to create the implementation. + + // i.e. it should look like this: + #include ... + #include ... + #include ... + #define STB_IMAGE_IMPLEMENTATION + #include "stb_image.h" + + You can #define STBI_ASSERT(x) before the #include to avoid using assert.h. + And #define STBI_MALLOC, STBI_REALLOC, and STBI_FREE to avoid using malloc,realloc,free + + + QUICK NOTES: + Primarily of interest to game developers and other people who can + avoid problematic images and only need the trivial interface + + JPEG baseline & progressive (12 bpc/arithmetic not supported, same as stock IJG lib) + PNG 1/2/4/8/16-bit-per-channel + + TGA (not sure what subset, if a subset) + BMP non-1bpp, non-RLE + PSD (composited view only, no extra channels, 8/16 bit-per-channel) + + GIF (*comp always reports as 4-channel) + HDR (radiance rgbE format) + PIC (Softimage PIC) + PNM (PPM and PGM binary only) + + Animated GIF still needs a proper API, but here's one way to do it: + http://gist.github.com/urraka/685d9a6340b26b830d49 + + - decode from memory or through FILE (define STBI_NO_STDIO to remove code) + - decode from arbitrary I/O callbacks + - SIMD acceleration on x86/x64 (SSE2) and ARM (NEON) + + Full documentation under "DOCUMENTATION" below. + + +LICENSE + + See end of file for license information. + +RECENT REVISION HISTORY: + + 2.30 (2024-05-31) avoid erroneous gcc warning + 2.29 (2023-05-xx) optimizations + 2.28 (2023-01-29) many error fixes, security errors, just tons of stuff + 2.27 (2021-07-11) document stbi_info better, 16-bit PNM support, bug fixes + 2.26 (2020-07-13) many minor fixes + 2.25 (2020-02-02) fix warnings + 2.24 (2020-02-02) fix warnings; thread-local failure_reason and flip_vertically + 2.23 (2019-08-11) fix clang static analysis warning + 2.22 (2019-03-04) gif fixes, fix warnings + 2.21 (2019-02-25) fix typo in comment + 2.20 (2019-02-07) support utf8 filenames in Windows; fix warnings and platform ifdefs + 2.19 (2018-02-11) fix warning + 2.18 (2018-01-30) fix warnings + 2.17 (2018-01-29) bugfix, 1-bit BMP, 16-bitness query, fix warnings + 2.16 (2017-07-23) all functions have 16-bit variants; optimizations; bugfixes + 2.15 (2017-03-18) fix png-1,2,4; all Imagenet JPGs; no runtime SSE detection on GCC + 2.14 (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs + 2.13 (2016-12-04) experimental 16-bit API, only for PNG so far; fixes + 2.12 (2016-04-02) fix typo in 2.11 PSD fix that caused crashes + 2.11 (2016-04-02) 16-bit PNGS; enable SSE2 in non-gcc x64 + RGB-format JPEG; remove white matting in PSD; + allocate large structures on the stack; + correct channel count for PNG & BMP + 2.10 (2016-01-22) avoid warning introduced in 2.09 + 2.09 (2016-01-16) 16-bit TGA; comments in PNM files; STBI_REALLOC_SIZED + + See end of file for full revision history. + + + ============================ Contributors ========================= + + Image formats Extensions, features + Sean Barrett (jpeg, png, bmp) Jetro Lauha (stbi_info) + Nicolas Schulz (hdr, psd) Martin "SpartanJ" Golini (stbi_info) + Jonathan Dummer (tga) James "moose2000" Brown (iPhone PNG) + Jean-Marc Lienher (gif) Ben "Disch" Wenger (io callbacks) + Tom Seddon (pic) Omar Cornut (1/2/4-bit PNG) + Thatcher Ulrich (psd) Nicolas Guillemot (vertical flip) + Ken Miller (pgm, ppm) Richard Mitton (16-bit PSD) + github:urraka (animated gif) Junggon Kim (PNM comments) + Christopher Forseth (animated gif) Daniel Gibson (16-bit TGA) + socks-the-fox (16-bit PNG) + Jeremy Sawicki (handle all ImageNet JPGs) + Optimizations & bugfixes Mikhail Morozov (1-bit BMP) + Fabian "ryg" Giesen Anael Seghezzi (is-16-bit query) + Arseny Kapoulkine Simon Breuss (16-bit PNM) + John-Mark Allen + Carmelo J Fdez-Aguera + + Bug & warning fixes + Marc LeBlanc David Woo Guillaume George Martins Mozeiko + Christpher Lloyd Jerry Jansson Joseph Thomson Blazej Dariusz Roszkowski + Phil Jordan Dave Moore Roy Eltham + Hayaki Saito Nathan Reed Won Chun + Luke Graham Johan Duparc Nick Verigakis the Horde3D community + Thomas Ruf Ronny Chevalier github:rlyeh + Janez Zemva John Bartholomew Michal Cichon github:romigrou + Jonathan Blow Ken Hamada Tero Hanninen github:svdijk + Eugene Golushkov Laurent Gomila Cort Stratton github:snagar + Aruelien Pocheville Sergio Gonzalez Thibault Reuille github:Zelex + Cass Everitt Ryamond Barbiero github:grim210 + Paul Du Bois Engin Manap Aldo Culquicondor github:sammyhw + Philipp Wiesemann Dale Weiler Oriol Ferrer Mesia github:phprus + Josh Tobin Neil Bickford Matthew Gregan github:poppolopoppo + Julian Raschke Gregory Mullen Christian Floisand github:darealshinji + Baldur Karlsson Kevin Schmidt JR Smith github:Michaelangel007 + Brad Weinberger Matvey Cherevko github:mosra + Luca Sas Alexander Veselov Zack Middleton [reserved] + Ryan C. Gordon [reserved] [reserved] + DO NOT ADD YOUR NAME HERE + + Jacko Dirks + + To add your name to the credits, pick a random blank space in the middle and fill it. + 80% of merge conflicts on stb PRs are due to people adding their name at the end + of the credits. +*/ + +#ifndef STBI_INCLUDE_STB_IMAGE_H +#define STBI_INCLUDE_STB_IMAGE_H + +// DOCUMENTATION +// +// Limitations: +// - no 12-bit-per-channel JPEG +// - no JPEGs with arithmetic coding +// - GIF always returns *comp=4 +// +// Basic usage (see HDR discussion below for HDR usage): +// int x,y,n; +// unsigned char *data = stbi_load(filename, &x, &y, &n, 0); +// // ... process data if not NULL ... +// // ... x = width, y = height, n = # 8-bit components per pixel ... +// // ... replace '0' with '1'..'4' to force that many components per pixel +// // ... but 'n' will always be the number that it would have been if you said 0 +// stbi_image_free(data); +// +// Standard parameters: +// int *x -- outputs image width in pixels +// int *y -- outputs image height in pixels +// int *channels_in_file -- outputs # of image components in image file +// int desired_channels -- if non-zero, # of image components requested in result +// +// The return value from an image loader is an 'unsigned char *' which points +// to the pixel data, or NULL on an allocation failure or if the image is +// corrupt or invalid. The pixel data consists of *y scanlines of *x pixels, +// with each pixel consisting of N interleaved 8-bit components; the first +// pixel pointed to is top-left-most in the image. There is no padding between +// image scanlines or between pixels, regardless of format. The number of +// components N is 'desired_channels' if desired_channels is non-zero, or +// *channels_in_file otherwise. If desired_channels is non-zero, +// *channels_in_file has the number of components that _would_ have been +// output otherwise. E.g. if you set desired_channels to 4, you will always +// get RGBA output, but you can check *channels_in_file to see if it's trivially +// opaque because e.g. there were only 3 channels in the source image. +// +// An output image with N components has the following components interleaved +// in this order in each pixel: +// +// N=#comp components +// 1 grey +// 2 grey, alpha +// 3 red, green, blue +// 4 red, green, blue, alpha +// +// If image loading fails for any reason, the return value will be NULL, +// and *x, *y, *channels_in_file will be unchanged. The function +// stbi_failure_reason() can be queried for an extremely brief, end-user +// unfriendly explanation of why the load failed. Define STBI_NO_FAILURE_STRINGS +// to avoid compiling these strings at all, and STBI_FAILURE_USERMSG to get slightly +// more user-friendly ones. +// +// Paletted PNG, BMP, GIF, and PIC images are automatically depalettized. +// +// To query the width, height and component count of an image without having to +// decode the full file, you can use the stbi_info family of functions: +// +// int x,y,n,ok; +// ok = stbi_info(filename, &x, &y, &n); +// // returns ok=1 and sets x, y, n if image is a supported format, +// // 0 otherwise. +// +// Note that stb_image pervasively uses ints in its public API for sizes, +// including sizes of memory buffers. This is now part of the API and thus +// hard to change without causing breakage. As a result, the various image +// loaders all have certain limits on image size; these differ somewhat +// by format but generally boil down to either just under 2GB or just under +// 1GB. When the decoded image would be larger than this, stb_image decoding +// will fail. +// +// Additionally, stb_image will reject image files that have any of their +// dimensions set to a larger value than the configurable STBI_MAX_DIMENSIONS, +// which defaults to 2**24 = 16777216 pixels. Due to the above memory limit, +// the only way to have an image with such dimensions load correctly +// is for it to have a rather extreme aspect ratio. Either way, the +// assumption here is that such larger images are likely to be malformed +// or malicious. If you do need to load an image with individual dimensions +// larger than that, and it still fits in the overall size limit, you can +// #define STBI_MAX_DIMENSIONS on your own to be something larger. +// +// =========================================================================== +// +// UNICODE: +// +// If compiling for Windows and you wish to use Unicode filenames, compile +// with +// #define STBI_WINDOWS_UTF8 +// and pass utf8-encoded filenames. Call stbi_convert_wchar_to_utf8 to convert +// Windows wchar_t filenames to utf8. +// +// =========================================================================== +// +// Philosophy +// +// stb libraries are designed with the following priorities: +// +// 1. easy to use +// 2. easy to maintain +// 3. good performance +// +// Sometimes I let "good performance" creep up in priority over "easy to maintain", +// and for best performance I may provide less-easy-to-use APIs that give higher +// performance, in addition to the easy-to-use ones. Nevertheless, it's important +// to keep in mind that from the standpoint of you, a client of this library, +// all you care about is #1 and #3, and stb libraries DO NOT emphasize #3 above all. +// +// Some secondary priorities arise directly from the first two, some of which +// provide more explicit reasons why performance can't be emphasized. +// +// - Portable ("ease of use") +// - Small source code footprint ("easy to maintain") +// - No dependencies ("ease of use") +// +// =========================================================================== +// +// I/O callbacks +// +// I/O callbacks allow you to read from arbitrary sources, like packaged +// files or some other source. Data read from callbacks are processed +// through a small internal buffer (currently 128 bytes) to try to reduce +// overhead. +// +// The three functions you must define are "read" (reads some bytes of data), +// "skip" (skips some bytes of data), "eof" (reports if the stream is at the end). +// +// =========================================================================== +// +// SIMD support +// +// The JPEG decoder will try to automatically use SIMD kernels on x86 when +// supported by the compiler. For ARM Neon support, you must explicitly +// request it. +// +// (The old do-it-yourself SIMD API is no longer supported in the current +// code.) +// +// On x86, SSE2 will automatically be used when available based on a run-time +// test; if not, the generic C versions are used as a fall-back. On ARM targets, +// the typical path is to have separate builds for NEON and non-NEON devices +// (at least this is true for iOS and Android). Therefore, the NEON support is +// toggled by a build flag: define STBI_NEON to get NEON loops. +// +// If for some reason you do not want to use any of SIMD code, or if +// you have issues compiling it, you can disable it entirely by +// defining STBI_NO_SIMD. +// +// =========================================================================== +// +// HDR image support (disable by defining STBI_NO_HDR) +// +// stb_image supports loading HDR images in general, and currently the Radiance +// .HDR file format specifically. You can still load any file through the existing +// interface; if you attempt to load an HDR file, it will be automatically remapped +// to LDR, assuming gamma 2.2 and an arbitrary scale factor defaulting to 1; +// both of these constants can be reconfigured through this interface: +// +// stbi_hdr_to_ldr_gamma(2.2f); +// stbi_hdr_to_ldr_scale(1.0f); +// +// (note, do not use _inverse_ constants; stbi_image will invert them +// appropriately). +// +// Additionally, there is a new, parallel interface for loading files as +// (linear) floats to preserve the full dynamic range: +// +// float *data = stbi_loadf(filename, &x, &y, &n, 0); +// +// If you load LDR images through this interface, those images will +// be promoted to floating point values, run through the inverse of +// constants corresponding to the above: +// +// stbi_ldr_to_hdr_scale(1.0f); +// stbi_ldr_to_hdr_gamma(2.2f); +// +// Finally, given a filename (or an open file or memory block--see header +// file for details) containing image data, you can query for the "most +// appropriate" interface to use (that is, whether the image is HDR or +// not), using: +// +// stbi_is_hdr(char *filename); +// +// =========================================================================== +// +// iPhone PNG support: +// +// We optionally support converting iPhone-formatted PNGs (which store +// premultiplied BGRA) back to RGB, even though they're internally encoded +// differently. To enable this conversion, call +// stbi_convert_iphone_png_to_rgb(1). +// +// Call stbi_set_unpremultiply_on_load(1) as well to force a divide per +// pixel to remove any premultiplied alpha *only* if the image file explicitly +// says there's premultiplied data (currently only happens in iPhone images, +// and only if iPhone convert-to-rgb processing is on). +// +// =========================================================================== +// +// ADDITIONAL CONFIGURATION +// +// - You can suppress implementation of any of the decoders to reduce +// your code footprint by #defining one or more of the following +// symbols before creating the implementation. +// +// STBI_NO_JPEG +// STBI_NO_PNG +// STBI_NO_BMP +// STBI_NO_PSD +// STBI_NO_TGA +// STBI_NO_GIF +// STBI_NO_HDR +// STBI_NO_PIC +// STBI_NO_PNM (.ppm and .pgm) +// +// - You can request *only* certain decoders and suppress all other ones +// (this will be more forward-compatible, as addition of new decoders +// doesn't require you to disable them explicitly): +// +// STBI_ONLY_JPEG +// STBI_ONLY_PNG +// STBI_ONLY_BMP +// STBI_ONLY_PSD +// STBI_ONLY_TGA +// STBI_ONLY_GIF +// STBI_ONLY_HDR +// STBI_ONLY_PIC +// STBI_ONLY_PNM (.ppm and .pgm) +// +// - If you use STBI_NO_PNG (or _ONLY_ without PNG), and you still +// want the zlib decoder to be available, #define STBI_SUPPORT_ZLIB +// +// - If you define STBI_MAX_DIMENSIONS, stb_image will reject images greater +// than that size (in either width or height) without further processing. +// This is to let programs in the wild set an upper bound to prevent +// denial-of-service attacks on untrusted data, as one could generate a +// valid image of gigantic dimensions and force stb_image to allocate a +// huge block of memory and spend disproportionate time decoding it. By +// default this is set to (1 << 24), which is 16777216, but that's still +// very big. + +#ifndef STBI_NO_STDIO +#include <stdio.h> +#endif // STBI_NO_STDIO + +#define STBI_VERSION 1 + +enum +{ + STBI_default = 0, // only used for desired_channels + + STBI_grey = 1, + STBI_grey_alpha = 2, + STBI_rgb = 3, + STBI_rgb_alpha = 4 +}; + +#include <stdlib.h> +typedef unsigned char stbi_uc; +typedef unsigned short stbi_us; + +#ifdef __cplusplus +extern "C" { +#endif + +#ifndef STBIDEF +#ifdef STB_IMAGE_STATIC +#define STBIDEF static +#else +#define STBIDEF extern +#endif +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// PRIMARY API - works on images of any type +// + +// +// load image by filename, open file, or memory buffer +// + +typedef struct +{ + int (*read) (void *user,char *data,int size); // fill 'data' with 'size' bytes. return number of bytes actually read + void (*skip) (void *user,int n); // skip the next 'n' bytes, or 'unget' the last -n bytes if negative + int (*eof) (void *user); // returns nonzero if we are at end of file/data +} stbi_io_callbacks; + +//////////////////////////////////// +// +// 8-bits-per-channel interface +// + +STBIDEF stbi_uc *stbi_load_from_memory (stbi_uc const *buffer, int len , int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk , void *user, int *x, int *y, int *channels_in_file, int desired_channels); + +#ifndef STBI_NO_STDIO +STBIDEF stbi_uc *stbi_load (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_uc *stbi_load_from_file (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels); +// for stbi_load_from_file, file pointer is left pointing immediately after image +#endif + +#ifndef STBI_NO_GIF +STBIDEF stbi_uc *stbi_load_gif_from_memory(stbi_uc const *buffer, int len, int **delays, int *x, int *y, int *z, int *comp, int req_comp); +#endif + +#ifdef STBI_WINDOWS_UTF8 +STBIDEF int stbi_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input); +#endif + +//////////////////////////////////// +// +// 16-bits-per-channel interface +// + +STBIDEF stbi_us *stbi_load_16_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels); + +#ifndef STBI_NO_STDIO +STBIDEF stbi_us *stbi_load_16 (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels); +STBIDEF stbi_us *stbi_load_from_file_16(FILE *f, int *x, int *y, int *channels_in_file, int desired_channels); +#endif + +//////////////////////////////////// +// +// float-per-channel interface +// +#ifndef STBI_NO_LINEAR + STBIDEF float *stbi_loadf_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels); + STBIDEF float *stbi_loadf_from_callbacks (stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels); + + #ifndef STBI_NO_STDIO + STBIDEF float *stbi_loadf (char const *filename, int *x, int *y, int *channels_in_file, int desired_channels); + STBIDEF float *stbi_loadf_from_file (FILE *f, int *x, int *y, int *channels_in_file, int desired_channels); + #endif +#endif + +#ifndef STBI_NO_HDR + STBIDEF void stbi_hdr_to_ldr_gamma(float gamma); + STBIDEF void stbi_hdr_to_ldr_scale(float scale); +#endif // STBI_NO_HDR + +#ifndef STBI_NO_LINEAR + STBIDEF void stbi_ldr_to_hdr_gamma(float gamma); + STBIDEF void stbi_ldr_to_hdr_scale(float scale); +#endif // STBI_NO_LINEAR + +// stbi_is_hdr is always defined, but always returns false if STBI_NO_HDR +STBIDEF int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user); +STBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len); +#ifndef STBI_NO_STDIO +STBIDEF int stbi_is_hdr (char const *filename); +STBIDEF int stbi_is_hdr_from_file(FILE *f); +#endif // STBI_NO_STDIO + + +// get a VERY brief reason for failure +// on most compilers (and ALL modern mainstream compilers) this is threadsafe +STBIDEF const char *stbi_failure_reason (void); + +// free the loaded image -- this is just free() +STBIDEF void stbi_image_free (void *retval_from_stbi_load); + +// get image dimensions & components without fully decoding +STBIDEF int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp); +STBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp); +STBIDEF int stbi_is_16_bit_from_memory(stbi_uc const *buffer, int len); +STBIDEF int stbi_is_16_bit_from_callbacks(stbi_io_callbacks const *clbk, void *user); + +#ifndef STBI_NO_STDIO +STBIDEF int stbi_info (char const *filename, int *x, int *y, int *comp); +STBIDEF int stbi_info_from_file (FILE *f, int *x, int *y, int *comp); +STBIDEF int stbi_is_16_bit (char const *filename); +STBIDEF int stbi_is_16_bit_from_file(FILE *f); +#endif + + + +// for image formats that explicitly notate that they have premultiplied alpha, +// we just return the colors as stored in the file. set this flag to force +// unpremultiplication. results are undefined if the unpremultiply overflow. +STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply); + +// indicate whether we should process iphone images back to canonical format, +// or just pass them through "as-is" +STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert); + +// flip the image vertically, so the first pixel in the output array is the bottom left +STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip); + +// as above, but only applies to images loaded on the thread that calls the function +// this function is only available if your compiler supports thread-local variables; +// calling it will fail to link if your compiler doesn't +STBIDEF void stbi_set_unpremultiply_on_load_thread(int flag_true_if_should_unpremultiply); +STBIDEF void stbi_convert_iphone_png_to_rgb_thread(int flag_true_if_should_convert); +STBIDEF void stbi_set_flip_vertically_on_load_thread(int flag_true_if_should_flip); + +// ZLIB client - used by PNG, available for other purposes + +STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen); +STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header); +STBIDEF char *stbi_zlib_decode_malloc(const char *buffer, int len, int *outlen); +STBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, const char *ibuffer, int ilen); + +STBIDEF char *stbi_zlib_decode_noheader_malloc(const char *buffer, int len, int *outlen); +STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen); + + +#ifdef __cplusplus +} +#endif + +// +// +//// end header file ///////////////////////////////////////////////////// +#endif // STBI_INCLUDE_STB_IMAGE_H + +#ifdef STB_IMAGE_IMPLEMENTATION + +#if defined(STBI_ONLY_JPEG) || defined(STBI_ONLY_PNG) || defined(STBI_ONLY_BMP) \ + || defined(STBI_ONLY_TGA) || defined(STBI_ONLY_GIF) || defined(STBI_ONLY_PSD) \ + || defined(STBI_ONLY_HDR) || defined(STBI_ONLY_PIC) || defined(STBI_ONLY_PNM) \ + || defined(STBI_ONLY_ZLIB) + #ifndef STBI_ONLY_JPEG + #define STBI_NO_JPEG + #endif + #ifndef STBI_ONLY_PNG + #define STBI_NO_PNG + #endif + #ifndef STBI_ONLY_BMP + #define STBI_NO_BMP + #endif + #ifndef STBI_ONLY_PSD + #define STBI_NO_PSD + #endif + #ifndef STBI_ONLY_TGA + #define STBI_NO_TGA + #endif + #ifndef STBI_ONLY_GIF + #define STBI_NO_GIF + #endif + #ifndef STBI_ONLY_HDR + #define STBI_NO_HDR + #endif + #ifndef STBI_ONLY_PIC + #define STBI_NO_PIC + #endif + #ifndef STBI_ONLY_PNM + #define STBI_NO_PNM + #endif +#endif + +#if defined(STBI_NO_PNG) && !defined(STBI_SUPPORT_ZLIB) && !defined(STBI_NO_ZLIB) +#define STBI_NO_ZLIB +#endif + + +#include <stdarg.h> +#include <stddef.h> // ptrdiff_t on osx +#include <stdlib.h> +#include <string.h> +#include <limits.h> + +#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) +#include <math.h> // ldexp, pow +#endif + +#ifndef STBI_NO_STDIO +#include <stdio.h> +#endif + +#ifndef STBI_ASSERT +#include <assert.h> +#define STBI_ASSERT(x) assert(x) +#endif + +#ifdef __cplusplus +#define STBI_EXTERN extern "C" +#else +#define STBI_EXTERN extern +#endif + + +#ifndef _MSC_VER + #ifdef __cplusplus + #define stbi_inline inline + #else + #define stbi_inline + #endif +#else + #define stbi_inline __forceinline +#endif + +#ifndef STBI_NO_THREAD_LOCALS + #if defined(__cplusplus) && __cplusplus >= 201103L + #define STBI_THREAD_LOCAL thread_local + #elif defined(__GNUC__) && __GNUC__ < 5 + #define STBI_THREAD_LOCAL __thread + #elif defined(_MSC_VER) + #define STBI_THREAD_LOCAL __declspec(thread) + #elif defined (__STDC_VERSION__) && __STDC_VERSION__ >= 201112L && !defined(__STDC_NO_THREADS__) + #define STBI_THREAD_LOCAL _Thread_local + #endif + + #ifndef STBI_THREAD_LOCAL + #if defined(__GNUC__) + #define STBI_THREAD_LOCAL __thread + #endif + #endif +#endif + +#if defined(_MSC_VER) || defined(__SYMBIAN32__) +typedef unsigned short stbi__uint16; +typedef signed short stbi__int16; +typedef unsigned int stbi__uint32; +typedef signed int stbi__int32; +#else +#include <stdint.h> +typedef uint16_t stbi__uint16; +typedef int16_t stbi__int16; +typedef uint32_t stbi__uint32; +typedef int32_t stbi__int32; +#endif + +// should produce compiler error if size is wrong +typedef unsigned char validate_uint32[sizeof(stbi__uint32)==4 ? 1 : -1]; + +#ifdef _MSC_VER +#define STBI_NOTUSED(v) (void)(v) +#else +#define STBI_NOTUSED(v) (void)sizeof(v) +#endif + +#ifdef _MSC_VER +#define STBI_HAS_LROTL +#endif + +#ifdef STBI_HAS_LROTL + #define stbi_lrot(x,y) _lrotl(x,y) +#else + #define stbi_lrot(x,y) (((x) << (y)) | ((x) >> (-(y) & 31))) +#endif + +#if defined(STBI_MALLOC) && defined(STBI_FREE) && (defined(STBI_REALLOC) || defined(STBI_REALLOC_SIZED)) +// ok +#elif !defined(STBI_MALLOC) && !defined(STBI_FREE) && !defined(STBI_REALLOC) && !defined(STBI_REALLOC_SIZED) +// ok +#else +#error "Must define all or none of STBI_MALLOC, STBI_FREE, and STBI_REALLOC (or STBI_REALLOC_SIZED)." +#endif + +#ifndef STBI_MALLOC +#define STBI_MALLOC(sz) malloc(sz) +#define STBI_REALLOC(p,newsz) realloc(p,newsz) +#define STBI_FREE(p) free(p) +#endif + +#ifndef STBI_REALLOC_SIZED +#define STBI_REALLOC_SIZED(p,oldsz,newsz) STBI_REALLOC(p,newsz) +#endif + +// x86/x64 detection +#if defined(__x86_64__) || defined(_M_X64) +#define STBI__X64_TARGET +#elif defined(__i386) || defined(_M_IX86) +#define STBI__X86_TARGET +#endif + +#if defined(__GNUC__) && defined(STBI__X86_TARGET) && !defined(__SSE2__) && !defined(STBI_NO_SIMD) +// gcc doesn't support sse2 intrinsics unless you compile with -msse2, +// which in turn means it gets to use SSE2 everywhere. This is unfortunate, +// but previous attempts to provide the SSE2 functions with runtime +// detection caused numerous issues. The way architecture extensions are +// exposed in GCC/Clang is, sadly, not really suited for one-file libs. +// New behavior: if compiled with -msse2, we use SSE2 without any +// detection; if not, we don't use it at all. +#define STBI_NO_SIMD +#endif + +#if defined(__MINGW32__) && defined(STBI__X86_TARGET) && !defined(STBI_MINGW_ENABLE_SSE2) && !defined(STBI_NO_SIMD) +// Note that __MINGW32__ doesn't actually mean 32-bit, so we have to avoid STBI__X64_TARGET +// +// 32-bit MinGW wants ESP to be 16-byte aligned, but this is not in the +// Windows ABI and VC++ as well as Windows DLLs don't maintain that invariant. +// As a result, enabling SSE2 on 32-bit MinGW is dangerous when not +// simultaneously enabling "-mstackrealign". +// +// See https://github.com/nothings/stb/issues/81 for more information. +// +// So default to no SSE2 on 32-bit MinGW. If you've read this far and added +// -mstackrealign to your build settings, feel free to #define STBI_MINGW_ENABLE_SSE2. +#define STBI_NO_SIMD +#endif + +#if !defined(STBI_NO_SIMD) && (defined(STBI__X86_TARGET) || defined(STBI__X64_TARGET)) +#define STBI_SSE2 +#include <emmintrin.h> + +#ifdef _MSC_VER + +#if _MSC_VER >= 1400 // not VC6 +#include <intrin.h> // __cpuid +static int stbi__cpuid3(void) +{ + int info[4]; + __cpuid(info,1); + return info[3]; +} +#else +static int stbi__cpuid3(void) +{ + int res; + __asm { + mov eax,1 + cpuid + mov res,edx + } + return res; +} +#endif + +#define STBI_SIMD_ALIGN(type, name) __declspec(align(16)) type name + +#if !defined(STBI_NO_JPEG) && defined(STBI_SSE2) +static int stbi__sse2_available(void) +{ + int info3 = stbi__cpuid3(); + return ((info3 >> 26) & 1) != 0; +} +#endif + +#else // assume GCC-style if not VC++ +#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16))) + +#if !defined(STBI_NO_JPEG) && defined(STBI_SSE2) +static int stbi__sse2_available(void) +{ + // If we're even attempting to compile this on GCC/Clang, that means + // -msse2 is on, which means the compiler is allowed to use SSE2 + // instructions at will, and so are we. + return 1; +} +#endif + +#endif +#endif + +// ARM NEON +#if defined(STBI_NO_SIMD) && defined(STBI_NEON) +#undef STBI_NEON +#endif + +#ifdef STBI_NEON +#include <arm_neon.h> +#ifdef _MSC_VER +#define STBI_SIMD_ALIGN(type, name) __declspec(align(16)) type name +#else +#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16))) +#endif +#endif + +#ifndef STBI_SIMD_ALIGN +#define STBI_SIMD_ALIGN(type, name) type name +#endif + +#ifndef STBI_MAX_DIMENSIONS +#define STBI_MAX_DIMENSIONS (1 << 24) +#endif + +/////////////////////////////////////////////// +// +// stbi__context struct and start_xxx functions + +// stbi__context structure is our basic context used by all images, so it +// contains all the IO context, plus some basic image information +typedef struct +{ + stbi__uint32 img_x, img_y; + int img_n, img_out_n; + + stbi_io_callbacks io; + void *io_user_data; + + int read_from_callbacks; + int buflen; + stbi_uc buffer_start[128]; + int callback_already_read; + + stbi_uc *img_buffer, *img_buffer_end; + stbi_uc *img_buffer_original, *img_buffer_original_end; +} stbi__context; + + +static void stbi__refill_buffer(stbi__context *s); + +// initialize a memory-decode context +static void stbi__start_mem(stbi__context *s, stbi_uc const *buffer, int len) +{ + s->io.read = NULL; + s->read_from_callbacks = 0; + s->callback_already_read = 0; + s->img_buffer = s->img_buffer_original = (stbi_uc *) buffer; + s->img_buffer_end = s->img_buffer_original_end = (stbi_uc *) buffer+len; +} + +// initialize a callback-based context +static void stbi__start_callbacks(stbi__context *s, stbi_io_callbacks *c, void *user) +{ + s->io = *c; + s->io_user_data = user; + s->buflen = sizeof(s->buffer_start); + s->read_from_callbacks = 1; + s->callback_already_read = 0; + s->img_buffer = s->img_buffer_original = s->buffer_start; + stbi__refill_buffer(s); + s->img_buffer_original_end = s->img_buffer_end; +} + +#ifndef STBI_NO_STDIO + +static int stbi__stdio_read(void *user, char *data, int size) +{ + return (int) fread(data,1,size,(FILE*) user); +} + +static void stbi__stdio_skip(void *user, int n) +{ + int ch; + fseek((FILE*) user, n, SEEK_CUR); + ch = fgetc((FILE*) user); /* have to read a byte to reset feof()'s flag */ + if (ch != EOF) { + ungetc(ch, (FILE *) user); /* push byte back onto stream if valid. */ + } +} + +static int stbi__stdio_eof(void *user) +{ + return feof((FILE*) user) || ferror((FILE *) user); +} + +static stbi_io_callbacks stbi__stdio_callbacks = +{ + stbi__stdio_read, + stbi__stdio_skip, + stbi__stdio_eof, +}; + +static void stbi__start_file(stbi__context *s, FILE *f) +{ + stbi__start_callbacks(s, &stbi__stdio_callbacks, (void *) f); +} + +//static void stop_file(stbi__context *s) { } + +#endif // !STBI_NO_STDIO + +static void stbi__rewind(stbi__context *s) +{ + // conceptually rewind SHOULD rewind to the beginning of the stream, + // but we just rewind to the beginning of the initial buffer, because + // we only use it after doing 'test', which only ever looks at at most 92 bytes + s->img_buffer = s->img_buffer_original; + s->img_buffer_end = s->img_buffer_original_end; +} + +enum +{ + STBI_ORDER_RGB, + STBI_ORDER_BGR +}; + +typedef struct +{ + int bits_per_channel; + int num_channels; + int channel_order; +} stbi__result_info; + +#ifndef STBI_NO_JPEG +static int stbi__jpeg_test(stbi__context *s); +static void *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PNG +static int stbi__png_test(stbi__context *s); +static void *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp); +static int stbi__png_is16(stbi__context *s); +#endif + +#ifndef STBI_NO_BMP +static int stbi__bmp_test(stbi__context *s); +static void *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_TGA +static int stbi__tga_test(stbi__context *s); +static void *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PSD +static int stbi__psd_test(stbi__context *s); +static void *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc); +static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp); +static int stbi__psd_is16(stbi__context *s); +#endif + +#ifndef STBI_NO_HDR +static int stbi__hdr_test(stbi__context *s); +static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PIC +static int stbi__pic_test(stbi__context *s); +static void *stbi__pic_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_GIF +static int stbi__gif_test(stbi__context *s); +static void *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static void *stbi__load_gif_main(stbi__context *s, int **delays, int *x, int *y, int *z, int *comp, int req_comp); +static int stbi__gif_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_PNM +static int stbi__pnm_test(stbi__context *s); +static void *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri); +static int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp); +static int stbi__pnm_is16(stbi__context *s); +#endif + +static +#ifdef STBI_THREAD_LOCAL +STBI_THREAD_LOCAL +#endif +const char *stbi__g_failure_reason; + +STBIDEF const char *stbi_failure_reason(void) +{ + return stbi__g_failure_reason; +} + +#ifndef STBI_NO_FAILURE_STRINGS +static int stbi__err(const char *str) +{ + stbi__g_failure_reason = str; + return 0; +} +#endif + +static void *stbi__malloc(size_t size) +{ + return STBI_MALLOC(size); +} + +// stb_image uses ints pervasively, including for offset calculations. +// therefore the largest decoded image size we can support with the +// current code, even on 64-bit targets, is INT_MAX. this is not a +// significant limitation for the intended use case. +// +// we do, however, need to make sure our size calculations don't +// overflow. hence a few helper functions for size calculations that +// multiply integers together, making sure that they're non-negative +// and no overflow occurs. + +// return 1 if the sum is valid, 0 on overflow. +// negative terms are considered invalid. +static int stbi__addsizes_valid(int a, int b) +{ + if (b < 0) return 0; + // now 0 <= b <= INT_MAX, hence also + // 0 <= INT_MAX - b <= INTMAX. + // And "a + b <= INT_MAX" (which might overflow) is the + // same as a <= INT_MAX - b (no overflow) + return a <= INT_MAX - b; +} + +// returns 1 if the product is valid, 0 on overflow. +// negative factors are considered invalid. +static int stbi__mul2sizes_valid(int a, int b) +{ + if (a < 0 || b < 0) return 0; + if (b == 0) return 1; // mul-by-0 is always safe + // portable way to check for no overflows in a*b + return a <= INT_MAX/b; +} + +#if !defined(STBI_NO_JPEG) || !defined(STBI_NO_PNG) || !defined(STBI_NO_TGA) || !defined(STBI_NO_HDR) +// returns 1 if "a*b + add" has no negative terms/factors and doesn't overflow +static int stbi__mad2sizes_valid(int a, int b, int add) +{ + return stbi__mul2sizes_valid(a, b) && stbi__addsizes_valid(a*b, add); +} +#endif + +// returns 1 if "a*b*c + add" has no negative terms/factors and doesn't overflow +static int stbi__mad3sizes_valid(int a, int b, int c, int add) +{ + return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) && + stbi__addsizes_valid(a*b*c, add); +} + +// returns 1 if "a*b*c*d + add" has no negative terms/factors and doesn't overflow +#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) || !defined(STBI_NO_PNM) +static int stbi__mad4sizes_valid(int a, int b, int c, int d, int add) +{ + return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) && + stbi__mul2sizes_valid(a*b*c, d) && stbi__addsizes_valid(a*b*c*d, add); +} +#endif + +#if !defined(STBI_NO_JPEG) || !defined(STBI_NO_PNG) || !defined(STBI_NO_TGA) || !defined(STBI_NO_HDR) +// mallocs with size overflow checking +static void *stbi__malloc_mad2(int a, int b, int add) +{ + if (!stbi__mad2sizes_valid(a, b, add)) return NULL; + return stbi__malloc(a*b + add); +} +#endif + +static void *stbi__malloc_mad3(int a, int b, int c, int add) +{ + if (!stbi__mad3sizes_valid(a, b, c, add)) return NULL; + return stbi__malloc(a*b*c + add); +} + +#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) || !defined(STBI_NO_PNM) +static void *stbi__malloc_mad4(int a, int b, int c, int d, int add) +{ + if (!stbi__mad4sizes_valid(a, b, c, d, add)) return NULL; + return stbi__malloc(a*b*c*d + add); +} +#endif + +// returns 1 if the sum of two signed ints is valid (between -2^31 and 2^31-1 inclusive), 0 on overflow. +static int stbi__addints_valid(int a, int b) +{ + if ((a >= 0) != (b >= 0)) return 1; // a and b have different signs, so no overflow + if (a < 0 && b < 0) return a >= INT_MIN - b; // same as a + b >= INT_MIN; INT_MIN - b cannot overflow since b < 0. + return a <= INT_MAX - b; +} + +// returns 1 if the product of two ints fits in a signed short, 0 on overflow. +static int stbi__mul2shorts_valid(int a, int b) +{ + if (b == 0 || b == -1) return 1; // multiplication by 0 is always 0; check for -1 so SHRT_MIN/b doesn't overflow + if ((a >= 0) == (b >= 0)) return a <= SHRT_MAX/b; // product is positive, so similar to mul2sizes_valid + if (b < 0) return a <= SHRT_MIN / b; // same as a * b >= SHRT_MIN + return a >= SHRT_MIN / b; +} + +// stbi__err - error +// stbi__errpf - error returning pointer to float +// stbi__errpuc - error returning pointer to unsigned char + +#ifdef STBI_NO_FAILURE_STRINGS + #define stbi__err(x,y) 0 +#elif defined(STBI_FAILURE_USERMSG) + #define stbi__err(x,y) stbi__err(y) +#else + #define stbi__err(x,y) stbi__err(x) +#endif + +#define stbi__errpf(x,y) ((float *)(size_t) (stbi__err(x,y)?NULL:NULL)) +#define stbi__errpuc(x,y) ((unsigned char *)(size_t) (stbi__err(x,y)?NULL:NULL)) + +STBIDEF void stbi_image_free(void *retval_from_stbi_load) +{ + STBI_FREE(retval_from_stbi_load); +} + +#ifndef STBI_NO_LINEAR +static float *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp); +#endif + +#ifndef STBI_NO_HDR +static stbi_uc *stbi__hdr_to_ldr(float *data, int x, int y, int comp); +#endif + +static int stbi__vertically_flip_on_load_global = 0; + +STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip) +{ + stbi__vertically_flip_on_load_global = flag_true_if_should_flip; +} + +#ifndef STBI_THREAD_LOCAL +#define stbi__vertically_flip_on_load stbi__vertically_flip_on_load_global +#else +static STBI_THREAD_LOCAL int stbi__vertically_flip_on_load_local, stbi__vertically_flip_on_load_set; + +STBIDEF void stbi_set_flip_vertically_on_load_thread(int flag_true_if_should_flip) +{ + stbi__vertically_flip_on_load_local = flag_true_if_should_flip; + stbi__vertically_flip_on_load_set = 1; +} + +#define stbi__vertically_flip_on_load (stbi__vertically_flip_on_load_set \ + ? stbi__vertically_flip_on_load_local \ + : stbi__vertically_flip_on_load_global) +#endif // STBI_THREAD_LOCAL + +static void *stbi__load_main(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc) +{ + memset(ri, 0, sizeof(*ri)); // make sure it's initialized if we add new fields + ri->bits_per_channel = 8; // default is 8 so most paths don't have to be changed + ri->channel_order = STBI_ORDER_RGB; // all current input & output are this, but this is here so we can add BGR order + ri->num_channels = 0; + + // test the formats with a very explicit header first (at least a FOURCC + // or distinctive magic number first) + #ifndef STBI_NO_PNG + if (stbi__png_test(s)) return stbi__png_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_BMP + if (stbi__bmp_test(s)) return stbi__bmp_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_GIF + if (stbi__gif_test(s)) return stbi__gif_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_PSD + if (stbi__psd_test(s)) return stbi__psd_load(s,x,y,comp,req_comp, ri, bpc); + #else + STBI_NOTUSED(bpc); + #endif + #ifndef STBI_NO_PIC + if (stbi__pic_test(s)) return stbi__pic_load(s,x,y,comp,req_comp, ri); + #endif + + // then the formats that can end up attempting to load with just 1 or 2 + // bytes matching expectations; these are prone to false positives, so + // try them later + #ifndef STBI_NO_JPEG + if (stbi__jpeg_test(s)) return stbi__jpeg_load(s,x,y,comp,req_comp, ri); + #endif + #ifndef STBI_NO_PNM + if (stbi__pnm_test(s)) return stbi__pnm_load(s,x,y,comp,req_comp, ri); + #endif + + #ifndef STBI_NO_HDR + if (stbi__hdr_test(s)) { + float *hdr = stbi__hdr_load(s, x,y,comp,req_comp, ri); + return stbi__hdr_to_ldr(hdr, *x, *y, req_comp ? req_comp : *comp); + } + #endif + + #ifndef STBI_NO_TGA + // test tga last because it's a crappy test! + if (stbi__tga_test(s)) + return stbi__tga_load(s,x,y,comp,req_comp, ri); + #endif + + return stbi__errpuc("unknown image type", "Image not of any known type, or corrupt"); +} + +static stbi_uc *stbi__convert_16_to_8(stbi__uint16 *orig, int w, int h, int channels) +{ + int i; + int img_len = w * h * channels; + stbi_uc *reduced; + + reduced = (stbi_uc *) stbi__malloc(img_len); + if (reduced == NULL) return stbi__errpuc("outofmem", "Out of memory"); + + for (i = 0; i < img_len; ++i) + reduced[i] = (stbi_uc)((orig[i] >> 8) & 0xFF); // top half of each byte is sufficient approx of 16->8 bit scaling + + STBI_FREE(orig); + return reduced; +} + +static stbi__uint16 *stbi__convert_8_to_16(stbi_uc *orig, int w, int h, int channels) +{ + int i; + int img_len = w * h * channels; + stbi__uint16 *enlarged; + + enlarged = (stbi__uint16 *) stbi__malloc(img_len*2); + if (enlarged == NULL) return (stbi__uint16 *) stbi__errpuc("outofmem", "Out of memory"); + + for (i = 0; i < img_len; ++i) + enlarged[i] = (stbi__uint16)((orig[i] << 8) + orig[i]); // replicate to high and low byte, maps 0->0, 255->0xffff + + STBI_FREE(orig); + return enlarged; +} + +static void stbi__vertical_flip(void *image, int w, int h, int bytes_per_pixel) +{ + int row; + size_t bytes_per_row = (size_t)w * bytes_per_pixel; + stbi_uc temp[2048]; + stbi_uc *bytes = (stbi_uc *)image; + + for (row = 0; row < (h>>1); row++) { + stbi_uc *row0 = bytes + row*bytes_per_row; + stbi_uc *row1 = bytes + (h - row - 1)*bytes_per_row; + // swap row0 with row1 + size_t bytes_left = bytes_per_row; + while (bytes_left) { + size_t bytes_copy = (bytes_left < sizeof(temp)) ? bytes_left : sizeof(temp); + memcpy(temp, row0, bytes_copy); + memcpy(row0, row1, bytes_copy); + memcpy(row1, temp, bytes_copy); + row0 += bytes_copy; + row1 += bytes_copy; + bytes_left -= bytes_copy; + } + } +} + +#ifndef STBI_NO_GIF +static void stbi__vertical_flip_slices(void *image, int w, int h, int z, int bytes_per_pixel) +{ + int slice; + int slice_size = w * h * bytes_per_pixel; + + stbi_uc *bytes = (stbi_uc *)image; + for (slice = 0; slice < z; ++slice) { + stbi__vertical_flip(bytes, w, h, bytes_per_pixel); + bytes += slice_size; + } +} +#endif + +static unsigned char *stbi__load_and_postprocess_8bit(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + stbi__result_info ri; + void *result = stbi__load_main(s, x, y, comp, req_comp, &ri, 8); + + if (result == NULL) + return NULL; + + // it is the responsibility of the loaders to make sure we get either 8 or 16 bit. + STBI_ASSERT(ri.bits_per_channel == 8 || ri.bits_per_channel == 16); + + if (ri.bits_per_channel != 8) { + result = stbi__convert_16_to_8((stbi__uint16 *) result, *x, *y, req_comp == 0 ? *comp : req_comp); + ri.bits_per_channel = 8; + } + + // @TODO: move stbi__convert_format to here + + if (stbi__vertically_flip_on_load) { + int channels = req_comp ? req_comp : *comp; + stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi_uc)); + } + + return (unsigned char *) result; +} + +static stbi__uint16 *stbi__load_and_postprocess_16bit(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + stbi__result_info ri; + void *result = stbi__load_main(s, x, y, comp, req_comp, &ri, 16); + + if (result == NULL) + return NULL; + + // it is the responsibility of the loaders to make sure we get either 8 or 16 bit. + STBI_ASSERT(ri.bits_per_channel == 8 || ri.bits_per_channel == 16); + + if (ri.bits_per_channel != 16) { + result = stbi__convert_8_to_16((stbi_uc *) result, *x, *y, req_comp == 0 ? *comp : req_comp); + ri.bits_per_channel = 16; + } + + // @TODO: move stbi__convert_format16 to here + // @TODO: special case RGB-to-Y (and RGBA-to-YA) for 8-bit-to-16-bit case to keep more precision + + if (stbi__vertically_flip_on_load) { + int channels = req_comp ? req_comp : *comp; + stbi__vertical_flip(result, *x, *y, channels * sizeof(stbi__uint16)); + } + + return (stbi__uint16 *) result; +} + +#if !defined(STBI_NO_HDR) && !defined(STBI_NO_LINEAR) +static void stbi__float_postprocess(float *result, int *x, int *y, int *comp, int req_comp) +{ + if (stbi__vertically_flip_on_load && result != NULL) { + int channels = req_comp ? req_comp : *comp; + stbi__vertical_flip(result, *x, *y, channels * sizeof(float)); + } +} +#endif + +#ifndef STBI_NO_STDIO + +#if defined(_WIN32) && defined(STBI_WINDOWS_UTF8) +STBI_EXTERN __declspec(dllimport) int __stdcall MultiByteToWideChar(unsigned int cp, unsigned long flags, const char *str, int cbmb, wchar_t *widestr, int cchwide); +STBI_EXTERN __declspec(dllimport) int __stdcall WideCharToMultiByte(unsigned int cp, unsigned long flags, const wchar_t *widestr, int cchwide, char *str, int cbmb, const char *defchar, int *used_default); +#endif + +#if defined(_WIN32) && defined(STBI_WINDOWS_UTF8) +STBIDEF int stbi_convert_wchar_to_utf8(char *buffer, size_t bufferlen, const wchar_t* input) +{ + return WideCharToMultiByte(65001 /* UTF8 */, 0, input, -1, buffer, (int) bufferlen, NULL, NULL); +} +#endif + +static FILE *stbi__fopen(char const *filename, char const *mode) +{ + FILE *f; +#if defined(_WIN32) && defined(STBI_WINDOWS_UTF8) + wchar_t wMode[64]; + wchar_t wFilename[1024]; + if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, filename, -1, wFilename, sizeof(wFilename)/sizeof(*wFilename))) + return 0; + + if (0 == MultiByteToWideChar(65001 /* UTF8 */, 0, mode, -1, wMode, sizeof(wMode)/sizeof(*wMode))) + return 0; + +#if defined(_MSC_VER) && _MSC_VER >= 1400 + if (0 != _wfopen_s(&f, wFilename, wMode)) + f = 0; +#else + f = _wfopen(wFilename, wMode); +#endif + +#elif defined(_MSC_VER) && _MSC_VER >= 1400 + if (0 != fopen_s(&f, filename, mode)) + f=0; +#else + f = fopen(filename, mode); +#endif + return f; +} + + +STBIDEF stbi_uc *stbi_load(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = stbi__fopen(filename, "rb"); + unsigned char *result; + if (!f) return stbi__errpuc("can't fopen", "Unable to open file"); + result = stbi_load_from_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} + +STBIDEF stbi_uc *stbi_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *result; + stbi__context s; + stbi__start_file(&s,f); + result = stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp); + if (result) { + // need to 'unget' all the characters in the IO buffer + fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR); + } + return result; +} + +STBIDEF stbi__uint16 *stbi_load_from_file_16(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi__uint16 *result; + stbi__context s; + stbi__start_file(&s,f); + result = stbi__load_and_postprocess_16bit(&s,x,y,comp,req_comp); + if (result) { + // need to 'unget' all the characters in the IO buffer + fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR); + } + return result; +} + +STBIDEF stbi_us *stbi_load_16(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = stbi__fopen(filename, "rb"); + stbi__uint16 *result; + if (!f) return (stbi_us *) stbi__errpuc("can't fopen", "Unable to open file"); + result = stbi_load_from_file_16(f,x,y,comp,req_comp); + fclose(f); + return result; +} + + +#endif //!STBI_NO_STDIO + +STBIDEF stbi_us *stbi_load_16_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *channels_in_file, int desired_channels) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels); +} + +STBIDEF stbi_us *stbi_load_16_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *channels_in_file, int desired_channels) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *)clbk, user); + return stbi__load_and_postprocess_16bit(&s,x,y,channels_in_file,desired_channels); +} + +STBIDEF stbi_uc *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp); +} + +STBIDEF stbi_uc *stbi_load_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user); + return stbi__load_and_postprocess_8bit(&s,x,y,comp,req_comp); +} + +#ifndef STBI_NO_GIF +STBIDEF stbi_uc *stbi_load_gif_from_memory(stbi_uc const *buffer, int len, int **delays, int *x, int *y, int *z, int *comp, int req_comp) +{ + unsigned char *result; + stbi__context s; + stbi__start_mem(&s,buffer,len); + + result = (unsigned char*) stbi__load_gif_main(&s, delays, x, y, z, comp, req_comp); + if (stbi__vertically_flip_on_load) { + stbi__vertical_flip_slices( result, *x, *y, *z, *comp ); + } + + return result; +} +#endif + +#ifndef STBI_NO_LINEAR +static float *stbi__loadf_main(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *data; + #ifndef STBI_NO_HDR + if (stbi__hdr_test(s)) { + stbi__result_info ri; + float *hdr_data = stbi__hdr_load(s,x,y,comp,req_comp, &ri); + if (hdr_data) + stbi__float_postprocess(hdr_data,x,y,comp,req_comp); + return hdr_data; + } + #endif + data = stbi__load_and_postprocess_8bit(s, x, y, comp, req_comp); + if (data) + return stbi__ldr_to_hdr(data, *x, *y, req_comp ? req_comp : *comp); + return stbi__errpf("unknown image type", "Image not of any known type, or corrupt"); +} + +STBIDEF float *stbi_loadf_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__loadf_main(&s,x,y,comp,req_comp); +} + +STBIDEF float *stbi_loadf_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user); + return stbi__loadf_main(&s,x,y,comp,req_comp); +} + +#ifndef STBI_NO_STDIO +STBIDEF float *stbi_loadf(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + float *result; + FILE *f = stbi__fopen(filename, "rb"); + if (!f) return stbi__errpf("can't fopen", "Unable to open file"); + result = stbi_loadf_from_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} + +STBIDEF float *stbi_loadf_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi__context s; + stbi__start_file(&s,f); + return stbi__loadf_main(&s,x,y,comp,req_comp); +} +#endif // !STBI_NO_STDIO + +#endif // !STBI_NO_LINEAR + +// these is-hdr-or-not is defined independent of whether STBI_NO_LINEAR is +// defined, for API simplicity; if STBI_NO_LINEAR is defined, it always +// reports false! + +STBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len) +{ + #ifndef STBI_NO_HDR + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__hdr_test(&s); + #else + STBI_NOTUSED(buffer); + STBI_NOTUSED(len); + return 0; + #endif +} + +#ifndef STBI_NO_STDIO +STBIDEF int stbi_is_hdr (char const *filename) +{ + FILE *f = stbi__fopen(filename, "rb"); + int result=0; + if (f) { + result = stbi_is_hdr_from_file(f); + fclose(f); + } + return result; +} + +STBIDEF int stbi_is_hdr_from_file(FILE *f) +{ + #ifndef STBI_NO_HDR + long pos = ftell(f); + int res; + stbi__context s; + stbi__start_file(&s,f); + res = stbi__hdr_test(&s); + fseek(f, pos, SEEK_SET); + return res; + #else + STBI_NOTUSED(f); + return 0; + #endif +} +#endif // !STBI_NO_STDIO + +STBIDEF int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user) +{ + #ifndef STBI_NO_HDR + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) clbk, user); + return stbi__hdr_test(&s); + #else + STBI_NOTUSED(clbk); + STBI_NOTUSED(user); + return 0; + #endif +} + +#ifndef STBI_NO_LINEAR +static float stbi__l2h_gamma=2.2f, stbi__l2h_scale=1.0f; + +STBIDEF void stbi_ldr_to_hdr_gamma(float gamma) { stbi__l2h_gamma = gamma; } +STBIDEF void stbi_ldr_to_hdr_scale(float scale) { stbi__l2h_scale = scale; } +#endif + +static float stbi__h2l_gamma_i=1.0f/2.2f, stbi__h2l_scale_i=1.0f; + +STBIDEF void stbi_hdr_to_ldr_gamma(float gamma) { stbi__h2l_gamma_i = 1/gamma; } +STBIDEF void stbi_hdr_to_ldr_scale(float scale) { stbi__h2l_scale_i = 1/scale; } + + +////////////////////////////////////////////////////////////////////////////// +// +// Common code used by all image loaders +// + +enum +{ + STBI__SCAN_load=0, + STBI__SCAN_type, + STBI__SCAN_header +}; + +static void stbi__refill_buffer(stbi__context *s) +{ + int n = (s->io.read)(s->io_user_data,(char*)s->buffer_start,s->buflen); + s->callback_already_read += (int) (s->img_buffer - s->img_buffer_original); + if (n == 0) { + // at end of file, treat same as if from memory, but need to handle case + // where s->img_buffer isn't pointing to safe memory, e.g. 0-byte file + s->read_from_callbacks = 0; + s->img_buffer = s->buffer_start; + s->img_buffer_end = s->buffer_start+1; + *s->img_buffer = 0; + } else { + s->img_buffer = s->buffer_start; + s->img_buffer_end = s->buffer_start + n; + } +} + +stbi_inline static stbi_uc stbi__get8(stbi__context *s) +{ + if (s->img_buffer < s->img_buffer_end) + return *s->img_buffer++; + if (s->read_from_callbacks) { + stbi__refill_buffer(s); + return *s->img_buffer++; + } + return 0; +} + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_HDR) && defined(STBI_NO_PIC) && defined(STBI_NO_PNM) +// nothing +#else +stbi_inline static int stbi__at_eof(stbi__context *s) +{ + if (s->io.read) { + if (!(s->io.eof)(s->io_user_data)) return 0; + // if feof() is true, check if buffer = end + // special case: we've only got the special 0 character at the end + if (s->read_from_callbacks == 0) return 1; + } + + return s->img_buffer >= s->img_buffer_end; +} +#endif + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_PNG) && defined(STBI_NO_BMP) && defined(STBI_NO_PSD) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) && defined(STBI_NO_PIC) +// nothing +#else +static void stbi__skip(stbi__context *s, int n) +{ + if (n == 0) return; // already there! + if (n < 0) { + s->img_buffer = s->img_buffer_end; + return; + } + if (s->io.read) { + int blen = (int) (s->img_buffer_end - s->img_buffer); + if (blen < n) { + s->img_buffer = s->img_buffer_end; + (s->io.skip)(s->io_user_data, n - blen); + return; + } + } + s->img_buffer += n; +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_TGA) && defined(STBI_NO_HDR) && defined(STBI_NO_PNM) +// nothing +#else +static int stbi__getn(stbi__context *s, stbi_uc *buffer, int n) +{ + if (s->io.read) { + int blen = (int) (s->img_buffer_end - s->img_buffer); + if (blen < n) { + int res, count; + + memcpy(buffer, s->img_buffer, blen); + + count = (s->io.read)(s->io_user_data, (char*) buffer + blen, n - blen); + res = (count == (n-blen)); + s->img_buffer = s->img_buffer_end; + return res; + } + } + + if (s->img_buffer+n <= s->img_buffer_end) { + memcpy(buffer, s->img_buffer, n); + s->img_buffer += n; + return 1; + } else + return 0; +} +#endif + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_PNG) && defined(STBI_NO_PSD) && defined(STBI_NO_PIC) +// nothing +#else +static int stbi__get16be(stbi__context *s) +{ + int z = stbi__get8(s); + return (z << 8) + stbi__get8(s); +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_PSD) && defined(STBI_NO_PIC) +// nothing +#else +static stbi__uint32 stbi__get32be(stbi__context *s) +{ + stbi__uint32 z = stbi__get16be(s); + return (z << 16) + stbi__get16be(s); +} +#endif + +#if defined(STBI_NO_BMP) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) +// nothing +#else +static int stbi__get16le(stbi__context *s) +{ + int z = stbi__get8(s); + return z + (stbi__get8(s) << 8); +} +#endif + +#ifndef STBI_NO_BMP +static stbi__uint32 stbi__get32le(stbi__context *s) +{ + stbi__uint32 z = stbi__get16le(s); + z += (stbi__uint32)stbi__get16le(s) << 16; + return z; +} +#endif + +#define STBI__BYTECAST(x) ((stbi_uc) ((x) & 255)) // truncate int to byte without warnings + +#if defined(STBI_NO_JPEG) && defined(STBI_NO_PNG) && defined(STBI_NO_BMP) && defined(STBI_NO_PSD) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) && defined(STBI_NO_PIC) && defined(STBI_NO_PNM) +// nothing +#else +////////////////////////////////////////////////////////////////////////////// +// +// generic converter from built-in img_n to req_comp +// individual types do this automatically as much as possible (e.g. jpeg +// does all cases internally since it needs to colorspace convert anyway, +// and it never has alpha, so very few cases ). png can automatically +// interleave an alpha=255 channel, but falls back to this for other cases +// +// assume data buffer is malloced, so malloc a new one and free that one +// only failure mode is malloc failing + +static stbi_uc stbi__compute_y(int r, int g, int b) +{ + return (stbi_uc) (((r*77) + (g*150) + (29*b)) >> 8); +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_BMP) && defined(STBI_NO_PSD) && defined(STBI_NO_TGA) && defined(STBI_NO_GIF) && defined(STBI_NO_PIC) && defined(STBI_NO_PNM) +// nothing +#else +static unsigned char *stbi__convert_format(unsigned char *data, int img_n, int req_comp, unsigned int x, unsigned int y) +{ + int i,j; + unsigned char *good; + + if (req_comp == img_n) return data; + STBI_ASSERT(req_comp >= 1 && req_comp <= 4); + + good = (unsigned char *) stbi__malloc_mad3(req_comp, x, y, 0); + if (good == NULL) { + STBI_FREE(data); + return stbi__errpuc("outofmem", "Out of memory"); + } + + for (j=0; j < (int) y; ++j) { + unsigned char *src = data + j * x * img_n ; + unsigned char *dest = good + j * x * req_comp; + + #define STBI__COMBO(a,b) ((a)*8+(b)) + #define STBI__CASE(a,b) case STBI__COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b) + // convert source image with img_n components to one with req_comp components; + // avoid switch per pixel, so use switch per scanline and massive macros + switch (STBI__COMBO(img_n, req_comp)) { + STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=255; } break; + STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=255; } break; + STBI__CASE(2,1) { dest[0]=src[0]; } break; + STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1]; } break; + STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=255; } break; + STBI__CASE(3,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); } break; + STBI__CASE(3,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = 255; } break; + STBI__CASE(4,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); } break; + STBI__CASE(4,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = src[3]; } break; + STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2]; } break; + default: STBI_ASSERT(0); STBI_FREE(data); STBI_FREE(good); return stbi__errpuc("unsupported", "Unsupported format conversion"); + } + #undef STBI__CASE + } + + STBI_FREE(data); + return good; +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_PSD) +// nothing +#else +static stbi__uint16 stbi__compute_y_16(int r, int g, int b) +{ + return (stbi__uint16) (((r*77) + (g*150) + (29*b)) >> 8); +} +#endif + +#if defined(STBI_NO_PNG) && defined(STBI_NO_PSD) +// nothing +#else +static stbi__uint16 *stbi__convert_format16(stbi__uint16 *data, int img_n, int req_comp, unsigned int x, unsigned int y) +{ + int i,j; + stbi__uint16 *good; + + if (req_comp == img_n) return data; + STBI_ASSERT(req_comp >= 1 && req_comp <= 4); + + good = (stbi__uint16 *) stbi__malloc(req_comp * x * y * 2); + if (good == NULL) { + STBI_FREE(data); + return (stbi__uint16 *) stbi__errpuc("outofmem", "Out of memory"); + } + + for (j=0; j < (int) y; ++j) { + stbi__uint16 *src = data + j * x * img_n ; + stbi__uint16 *dest = good + j * x * req_comp; + + #define STBI__COMBO(a,b) ((a)*8+(b)) + #define STBI__CASE(a,b) case STBI__COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b) + // convert source image with img_n components to one with req_comp components; + // avoid switch per pixel, so use switch per scanline and massive macros + switch (STBI__COMBO(img_n, req_comp)) { + STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=0xffff; } break; + STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=0xffff; } break; + STBI__CASE(2,1) { dest[0]=src[0]; } break; + STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0]; } break; + STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1]; } break; + STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=0xffff; } break; + STBI__CASE(3,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); } break; + STBI__CASE(3,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = 0xffff; } break; + STBI__CASE(4,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); } break; + STBI__CASE(4,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = src[3]; } break; + STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2]; } break; + default: STBI_ASSERT(0); STBI_FREE(data); STBI_FREE(good); return (stbi__uint16*) stbi__errpuc("unsupported", "Unsupported format conversion"); + } + #undef STBI__CASE + } + + STBI_FREE(data); + return good; +} +#endif + +#ifndef STBI_NO_LINEAR +static float *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp) +{ + int i,k,n; + float *output; + if (!data) return NULL; + output = (float *) stbi__malloc_mad4(x, y, comp, sizeof(float), 0); + if (output == NULL) { STBI_FREE(data); return stbi__errpf("outofmem", "Out of memory"); } + // compute number of non-alpha components + if (comp & 1) n = comp; else n = comp-1; + for (i=0; i < x*y; ++i) { + for (k=0; k < n; ++k) { + output[i*comp + k] = (float) (pow(data[i*comp+k]/255.0f, stbi__l2h_gamma) * stbi__l2h_scale); + } + } + if (n < comp) { + for (i=0; i < x*y; ++i) { + output[i*comp + n] = data[i*comp + n]/255.0f; + } + } + STBI_FREE(data); + return output; +} +#endif + +#ifndef STBI_NO_HDR +#define stbi__float2int(x) ((int) (x)) +static stbi_uc *stbi__hdr_to_ldr(float *data, int x, int y, int comp) +{ + int i,k,n; + stbi_uc *output; + if (!data) return NULL; + output = (stbi_uc *) stbi__malloc_mad3(x, y, comp, 0); + if (output == NULL) { STBI_FREE(data); return stbi__errpuc("outofmem", "Out of memory"); } + // compute number of non-alpha components + if (comp & 1) n = comp; else n = comp-1; + for (i=0; i < x*y; ++i) { + for (k=0; k < n; ++k) { + float z = (float) pow(data[i*comp+k]*stbi__h2l_scale_i, stbi__h2l_gamma_i) * 255 + 0.5f; + if (z < 0) z = 0; + if (z > 255) z = 255; + output[i*comp + k] = (stbi_uc) stbi__float2int(z); + } + if (k < comp) { + float z = data[i*comp+k] * 255 + 0.5f; + if (z < 0) z = 0; + if (z > 255) z = 255; + output[i*comp + k] = (stbi_uc) stbi__float2int(z); + } + } + STBI_FREE(data); + return output; +} +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// "baseline" JPEG/JFIF decoder +// +// simple implementation +// - doesn't support delayed output of y-dimension +// - simple interface (only one output format: 8-bit interleaved RGB) +// - doesn't try to recover corrupt jpegs +// - doesn't allow partial loading, loading multiple at once +// - still fast on x86 (copying globals into locals doesn't help x86) +// - allocates lots of intermediate memory (full size of all components) +// - non-interleaved case requires this anyway +// - allows good upsampling (see next) +// high-quality +// - upsampled channels are bilinearly interpolated, even across blocks +// - quality integer IDCT derived from IJG's 'slow' +// performance +// - fast huffman; reasonable integer IDCT +// - some SIMD kernels for common paths on targets with SSE2/NEON +// - uses a lot of intermediate memory, could cache poorly + +#ifndef STBI_NO_JPEG + +// huffman decoding acceleration +#define FAST_BITS 9 // larger handles more cases; smaller stomps less cache + +typedef struct +{ + stbi_uc fast[1 << FAST_BITS]; + // weirdly, repacking this into AoS is a 10% speed loss, instead of a win + stbi__uint16 code[256]; + stbi_uc values[256]; + stbi_uc size[257]; + unsigned int maxcode[18]; + int delta[17]; // old 'firstsymbol' - old 'firstcode' +} stbi__huffman; + +typedef struct +{ + stbi__context *s; + stbi__huffman huff_dc[4]; + stbi__huffman huff_ac[4]; + stbi__uint16 dequant[4][64]; + stbi__int16 fast_ac[4][1 << FAST_BITS]; + +// sizes for components, interleaved MCUs + int img_h_max, img_v_max; + int img_mcu_x, img_mcu_y; + int img_mcu_w, img_mcu_h; + +// definition of jpeg image component + struct + { + int id; + int h,v; + int tq; + int hd,ha; + int dc_pred; + + int x,y,w2,h2; + stbi_uc *data; + void *raw_data, *raw_coeff; + stbi_uc *linebuf; + short *coeff; // progressive only + int coeff_w, coeff_h; // number of 8x8 coefficient blocks + } img_comp[4]; + + stbi__uint32 code_buffer; // jpeg entropy-coded buffer + int code_bits; // number of valid bits + unsigned char marker; // marker seen while filling entropy buffer + int nomore; // flag if we saw a marker so must stop + + int progressive; + int spec_start; + int spec_end; + int succ_high; + int succ_low; + int eob_run; + int jfif; + int app14_color_transform; // Adobe APP14 tag + int rgb; + + int scan_n, order[4]; + int restart_interval, todo; + +// kernels + void (*idct_block_kernel)(stbi_uc *out, int out_stride, short data[64]); + void (*YCbCr_to_RGB_kernel)(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step); + stbi_uc *(*resample_row_hv_2_kernel)(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs); +} stbi__jpeg; + +static int stbi__build_huffman(stbi__huffman *h, int *count) +{ + int i,j,k=0; + unsigned int code; + // build size list for each symbol (from JPEG spec) + for (i=0; i < 16; ++i) { + for (j=0; j < count[i]; ++j) { + h->size[k++] = (stbi_uc) (i+1); + if(k >= 257) return stbi__err("bad size list","Corrupt JPEG"); + } + } + h->size[k] = 0; + + // compute actual symbols (from jpeg spec) + code = 0; + k = 0; + for(j=1; j <= 16; ++j) { + // compute delta to add to code to compute symbol id + h->delta[j] = k - code; + if (h->size[k] == j) { + while (h->size[k] == j) + h->code[k++] = (stbi__uint16) (code++); + if (code-1 >= (1u << j)) return stbi__err("bad code lengths","Corrupt JPEG"); + } + // compute largest code + 1 for this size, preshifted as needed later + h->maxcode[j] = code << (16-j); + code <<= 1; + } + h->maxcode[j] = 0xffffffff; + + // build non-spec acceleration table; 255 is flag for not-accelerated + memset(h->fast, 255, 1 << FAST_BITS); + for (i=0; i < k; ++i) { + int s = h->size[i]; + if (s <= FAST_BITS) { + int c = h->code[i] << (FAST_BITS-s); + int m = 1 << (FAST_BITS-s); + for (j=0; j < m; ++j) { + h->fast[c+j] = (stbi_uc) i; + } + } + } + return 1; +} + +// build a table that decodes both magnitude and value of small ACs in +// one go. +static void stbi__build_fast_ac(stbi__int16 *fast_ac, stbi__huffman *h) +{ + int i; + for (i=0; i < (1 << FAST_BITS); ++i) { + stbi_uc fast = h->fast[i]; + fast_ac[i] = 0; + if (fast < 255) { + int rs = h->values[fast]; + int run = (rs >> 4) & 15; + int magbits = rs & 15; + int len = h->size[fast]; + + if (magbits && len + magbits <= FAST_BITS) { + // magnitude code followed by receive_extend code + int k = ((i << len) & ((1 << FAST_BITS) - 1)) >> (FAST_BITS - magbits); + int m = 1 << (magbits - 1); + if (k < m) k += (~0U << magbits) + 1; + // if the result is small enough, we can fit it in fast_ac table + if (k >= -128 && k <= 127) + fast_ac[i] = (stbi__int16) ((k * 256) + (run * 16) + (len + magbits)); + } + } + } +} + +static void stbi__grow_buffer_unsafe(stbi__jpeg *j) +{ + do { + unsigned int b = j->nomore ? 0 : stbi__get8(j->s); + if (b == 0xff) { + int c = stbi__get8(j->s); + while (c == 0xff) c = stbi__get8(j->s); // consume fill bytes + if (c != 0) { + j->marker = (unsigned char) c; + j->nomore = 1; + return; + } + } + j->code_buffer |= b << (24 - j->code_bits); + j->code_bits += 8; + } while (j->code_bits <= 24); +} + +// (1 << n) - 1 +static const stbi__uint32 stbi__bmask[17]={0,1,3,7,15,31,63,127,255,511,1023,2047,4095,8191,16383,32767,65535}; + +// decode a jpeg huffman value from the bitstream +stbi_inline static int stbi__jpeg_huff_decode(stbi__jpeg *j, stbi__huffman *h) +{ + unsigned int temp; + int c,k; + + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + + // look at the top FAST_BITS and determine what symbol ID it is, + // if the code is <= FAST_BITS + c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1); + k = h->fast[c]; + if (k < 255) { + int s = h->size[k]; + if (s > j->code_bits) + return -1; + j->code_buffer <<= s; + j->code_bits -= s; + return h->values[k]; + } + + // naive test is to shift the code_buffer down so k bits are + // valid, then test against maxcode. To speed this up, we've + // preshifted maxcode left so that it has (16-k) 0s at the + // end; in other words, regardless of the number of bits, it + // wants to be compared against something shifted to have 16; + // that way we don't need to shift inside the loop. + temp = j->code_buffer >> 16; + for (k=FAST_BITS+1 ; ; ++k) + if (temp < h->maxcode[k]) + break; + if (k == 17) { + // error! code not found + j->code_bits -= 16; + return -1; + } + + if (k > j->code_bits) + return -1; + + // convert the huffman code to the symbol id + c = ((j->code_buffer >> (32 - k)) & stbi__bmask[k]) + h->delta[k]; + if(c < 0 || c >= 256) // symbol id out of bounds! + return -1; + STBI_ASSERT((((j->code_buffer) >> (32 - h->size[c])) & stbi__bmask[h->size[c]]) == h->code[c]); + + // convert the id to a symbol + j->code_bits -= k; + j->code_buffer <<= k; + return h->values[c]; +} + +// bias[n] = (-1<<n) + 1 +static const int stbi__jbias[16] = {0,-1,-3,-7,-15,-31,-63,-127,-255,-511,-1023,-2047,-4095,-8191,-16383,-32767}; + +// combined JPEG 'receive' and JPEG 'extend', since baseline +// always extends everything it receives. +stbi_inline static int stbi__extend_receive(stbi__jpeg *j, int n) +{ + unsigned int k; + int sgn; + if (j->code_bits < n) stbi__grow_buffer_unsafe(j); + if (j->code_bits < n) return 0; // ran out of bits from stream, return 0s intead of continuing + + sgn = j->code_buffer >> 31; // sign bit always in MSB; 0 if MSB clear (positive), 1 if MSB set (negative) + k = stbi_lrot(j->code_buffer, n); + j->code_buffer = k & ~stbi__bmask[n]; + k &= stbi__bmask[n]; + j->code_bits -= n; + return k + (stbi__jbias[n] & (sgn - 1)); +} + +// get some unsigned bits +stbi_inline static int stbi__jpeg_get_bits(stbi__jpeg *j, int n) +{ + unsigned int k; + if (j->code_bits < n) stbi__grow_buffer_unsafe(j); + if (j->code_bits < n) return 0; // ran out of bits from stream, return 0s intead of continuing + k = stbi_lrot(j->code_buffer, n); + j->code_buffer = k & ~stbi__bmask[n]; + k &= stbi__bmask[n]; + j->code_bits -= n; + return k; +} + +stbi_inline static int stbi__jpeg_get_bit(stbi__jpeg *j) +{ + unsigned int k; + if (j->code_bits < 1) stbi__grow_buffer_unsafe(j); + if (j->code_bits < 1) return 0; // ran out of bits from stream, return 0s intead of continuing + k = j->code_buffer; + j->code_buffer <<= 1; + --j->code_bits; + return k & 0x80000000; +} + +// given a value that's at position X in the zigzag stream, +// where does it appear in the 8x8 matrix coded as row-major? +static const stbi_uc stbi__jpeg_dezigzag[64+15] = +{ + 0, 1, 8, 16, 9, 2, 3, 10, + 17, 24, 32, 25, 18, 11, 4, 5, + 12, 19, 26, 33, 40, 48, 41, 34, + 27, 20, 13, 6, 7, 14, 21, 28, + 35, 42, 49, 56, 57, 50, 43, 36, + 29, 22, 15, 23, 30, 37, 44, 51, + 58, 59, 52, 45, 38, 31, 39, 46, + 53, 60, 61, 54, 47, 55, 62, 63, + // let corrupt input sample past end + 63, 63, 63, 63, 63, 63, 63, 63, + 63, 63, 63, 63, 63, 63, 63 +}; + +// decode one 64-entry block-- +static int stbi__jpeg_decode_block(stbi__jpeg *j, short data[64], stbi__huffman *hdc, stbi__huffman *hac, stbi__int16 *fac, int b, stbi__uint16 *dequant) +{ + int diff,dc,k; + int t; + + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + t = stbi__jpeg_huff_decode(j, hdc); + if (t < 0 || t > 15) return stbi__err("bad huffman code","Corrupt JPEG"); + + // 0 all the ac values now so we can do it 32-bits at a time + memset(data,0,64*sizeof(data[0])); + + diff = t ? stbi__extend_receive(j, t) : 0; + if (!stbi__addints_valid(j->img_comp[b].dc_pred, diff)) return stbi__err("bad delta","Corrupt JPEG"); + dc = j->img_comp[b].dc_pred + diff; + j->img_comp[b].dc_pred = dc; + if (!stbi__mul2shorts_valid(dc, dequant[0])) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + data[0] = (short) (dc * dequant[0]); + + // decode AC components, see JPEG spec + k = 1; + do { + unsigned int zig; + int c,r,s; + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1); + r = fac[c]; + if (r) { // fast-AC path + k += (r >> 4) & 15; // run + s = r & 15; // combined length + if (s > j->code_bits) return stbi__err("bad huffman code", "Combined length longer than code bits available"); + j->code_buffer <<= s; + j->code_bits -= s; + // decode into unzigzag'd location + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) ((r >> 8) * dequant[zig]); + } else { + int rs = stbi__jpeg_huff_decode(j, hac); + if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (rs != 0xf0) break; // end block + k += 16; + } else { + k += r; + // decode into unzigzag'd location + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) (stbi__extend_receive(j,s) * dequant[zig]); + } + } + } while (k < 64); + return 1; +} + +static int stbi__jpeg_decode_block_prog_dc(stbi__jpeg *j, short data[64], stbi__huffman *hdc, int b) +{ + int diff,dc; + int t; + if (j->spec_end != 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + + if (j->succ_high == 0) { + // first scan for DC coefficient, must be first + memset(data,0,64*sizeof(data[0])); // 0 all the ac values now + t = stbi__jpeg_huff_decode(j, hdc); + if (t < 0 || t > 15) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + diff = t ? stbi__extend_receive(j, t) : 0; + + if (!stbi__addints_valid(j->img_comp[b].dc_pred, diff)) return stbi__err("bad delta", "Corrupt JPEG"); + dc = j->img_comp[b].dc_pred + diff; + j->img_comp[b].dc_pred = dc; + if (!stbi__mul2shorts_valid(dc, 1 << j->succ_low)) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + data[0] = (short) (dc * (1 << j->succ_low)); + } else { + // refinement scan for DC coefficient + if (stbi__jpeg_get_bit(j)) + data[0] += (short) (1 << j->succ_low); + } + return 1; +} + +// @OPTIMIZE: store non-zigzagged during the decode passes, +// and only de-zigzag when dequantizing +static int stbi__jpeg_decode_block_prog_ac(stbi__jpeg *j, short data[64], stbi__huffman *hac, stbi__int16 *fac) +{ + int k; + if (j->spec_start == 0) return stbi__err("can't merge dc and ac", "Corrupt JPEG"); + + if (j->succ_high == 0) { + int shift = j->succ_low; + + if (j->eob_run) { + --j->eob_run; + return 1; + } + + k = j->spec_start; + do { + unsigned int zig; + int c,r,s; + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1); + r = fac[c]; + if (r) { // fast-AC path + k += (r >> 4) & 15; // run + s = r & 15; // combined length + if (s > j->code_bits) return stbi__err("bad huffman code", "Combined length longer than code bits available"); + j->code_buffer <<= s; + j->code_bits -= s; + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) ((r >> 8) * (1 << shift)); + } else { + int rs = stbi__jpeg_huff_decode(j, hac); + if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (r < 15) { + j->eob_run = (1 << r); + if (r) + j->eob_run += stbi__jpeg_get_bits(j, r); + --j->eob_run; + break; + } + k += 16; + } else { + k += r; + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) (stbi__extend_receive(j,s) * (1 << shift)); + } + } + } while (k <= j->spec_end); + } else { + // refinement scan for these AC coefficients + + short bit = (short) (1 << j->succ_low); + + if (j->eob_run) { + --j->eob_run; + for (k = j->spec_start; k <= j->spec_end; ++k) { + short *p = &data[stbi__jpeg_dezigzag[k]]; + if (*p != 0) + if (stbi__jpeg_get_bit(j)) + if ((*p & bit)==0) { + if (*p > 0) + *p += bit; + else + *p -= bit; + } + } + } else { + k = j->spec_start; + do { + int r,s; + int rs = stbi__jpeg_huff_decode(j, hac); // @OPTIMIZE see if we can use the fast path here, advance-by-r is so slow, eh + if (rs < 0) return stbi__err("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (r < 15) { + j->eob_run = (1 << r) - 1; + if (r) + j->eob_run += stbi__jpeg_get_bits(j, r); + r = 64; // force end of block + } else { + // r=15 s=0 should write 16 0s, so we just do + // a run of 15 0s and then write s (which is 0), + // so we don't have to do anything special here + } + } else { + if (s != 1) return stbi__err("bad huffman code", "Corrupt JPEG"); + // sign bit + if (stbi__jpeg_get_bit(j)) + s = bit; + else + s = -bit; + } + + // advance by r + while (k <= j->spec_end) { + short *p = &data[stbi__jpeg_dezigzag[k++]]; + if (*p != 0) { + if (stbi__jpeg_get_bit(j)) + if ((*p & bit)==0) { + if (*p > 0) + *p += bit; + else + *p -= bit; + } + } else { + if (r == 0) { + *p = (short) s; + break; + } + --r; + } + } + } while (k <= j->spec_end); + } + } + return 1; +} + +// take a -128..127 value and stbi__clamp it and convert to 0..255 +stbi_inline static stbi_uc stbi__clamp(int x) +{ + // trick to use a single test to catch both cases + if ((unsigned int) x > 255) { + if (x < 0) return 0; + if (x > 255) return 255; + } + return (stbi_uc) x; +} + +#define stbi__f2f(x) ((int) (((x) * 4096 + 0.5))) +#define stbi__fsh(x) ((x) * 4096) + +// derived from jidctint -- DCT_ISLOW +#define STBI__IDCT_1D(s0,s1,s2,s3,s4,s5,s6,s7) \ + int t0,t1,t2,t3,p1,p2,p3,p4,p5,x0,x1,x2,x3; \ + p2 = s2; \ + p3 = s6; \ + p1 = (p2+p3) * stbi__f2f(0.5411961f); \ + t2 = p1 + p3*stbi__f2f(-1.847759065f); \ + t3 = p1 + p2*stbi__f2f( 0.765366865f); \ + p2 = s0; \ + p3 = s4; \ + t0 = stbi__fsh(p2+p3); \ + t1 = stbi__fsh(p2-p3); \ + x0 = t0+t3; \ + x3 = t0-t3; \ + x1 = t1+t2; \ + x2 = t1-t2; \ + t0 = s7; \ + t1 = s5; \ + t2 = s3; \ + t3 = s1; \ + p3 = t0+t2; \ + p4 = t1+t3; \ + p1 = t0+t3; \ + p2 = t1+t2; \ + p5 = (p3+p4)*stbi__f2f( 1.175875602f); \ + t0 = t0*stbi__f2f( 0.298631336f); \ + t1 = t1*stbi__f2f( 2.053119869f); \ + t2 = t2*stbi__f2f( 3.072711026f); \ + t3 = t3*stbi__f2f( 1.501321110f); \ + p1 = p5 + p1*stbi__f2f(-0.899976223f); \ + p2 = p5 + p2*stbi__f2f(-2.562915447f); \ + p3 = p3*stbi__f2f(-1.961570560f); \ + p4 = p4*stbi__f2f(-0.390180644f); \ + t3 += p1+p4; \ + t2 += p2+p3; \ + t1 += p2+p4; \ + t0 += p1+p3; + +static void stbi__idct_block(stbi_uc *out, int out_stride, short data[64]) +{ + int i,val[64],*v=val; + stbi_uc *o; + short *d = data; + + // columns + for (i=0; i < 8; ++i,++d, ++v) { + // if all zeroes, shortcut -- this avoids dequantizing 0s and IDCTing + if (d[ 8]==0 && d[16]==0 && d[24]==0 && d[32]==0 + && d[40]==0 && d[48]==0 && d[56]==0) { + // no shortcut 0 seconds + // (1|2|3|4|5|6|7)==0 0 seconds + // all separate -0.047 seconds + // 1 && 2|3 && 4|5 && 6|7: -0.047 seconds + int dcterm = d[0]*4; + v[0] = v[8] = v[16] = v[24] = v[32] = v[40] = v[48] = v[56] = dcterm; + } else { + STBI__IDCT_1D(d[ 0],d[ 8],d[16],d[24],d[32],d[40],d[48],d[56]) + // constants scaled things up by 1<<12; let's bring them back + // down, but keep 2 extra bits of precision + x0 += 512; x1 += 512; x2 += 512; x3 += 512; + v[ 0] = (x0+t3) >> 10; + v[56] = (x0-t3) >> 10; + v[ 8] = (x1+t2) >> 10; + v[48] = (x1-t2) >> 10; + v[16] = (x2+t1) >> 10; + v[40] = (x2-t1) >> 10; + v[24] = (x3+t0) >> 10; + v[32] = (x3-t0) >> 10; + } + } + + for (i=0, v=val, o=out; i < 8; ++i,v+=8,o+=out_stride) { + // no fast case since the first 1D IDCT spread components out + STBI__IDCT_1D(v[0],v[1],v[2],v[3],v[4],v[5],v[6],v[7]) + // constants scaled things up by 1<<12, plus we had 1<<2 from first + // loop, plus horizontal and vertical each scale by sqrt(8) so together + // we've got an extra 1<<3, so 1<<17 total we need to remove. + // so we want to round that, which means adding 0.5 * 1<<17, + // aka 65536. Also, we'll end up with -128 to 127 that we want + // to encode as 0..255 by adding 128, so we'll add that before the shift + x0 += 65536 + (128<<17); + x1 += 65536 + (128<<17); + x2 += 65536 + (128<<17); + x3 += 65536 + (128<<17); + // tried computing the shifts into temps, or'ing the temps to see + // if any were out of range, but that was slower + o[0] = stbi__clamp((x0+t3) >> 17); + o[7] = stbi__clamp((x0-t3) >> 17); + o[1] = stbi__clamp((x1+t2) >> 17); + o[6] = stbi__clamp((x1-t2) >> 17); + o[2] = stbi__clamp((x2+t1) >> 17); + o[5] = stbi__clamp((x2-t1) >> 17); + o[3] = stbi__clamp((x3+t0) >> 17); + o[4] = stbi__clamp((x3-t0) >> 17); + } +} + +#ifdef STBI_SSE2 +// sse2 integer IDCT. not the fastest possible implementation but it +// produces bit-identical results to the generic C version so it's +// fully "transparent". +static void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64]) +{ + // This is constructed to match our regular (generic) integer IDCT exactly. + __m128i row0, row1, row2, row3, row4, row5, row6, row7; + __m128i tmp; + + // dot product constant: even elems=x, odd elems=y + #define dct_const(x,y) _mm_setr_epi16((x),(y),(x),(y),(x),(y),(x),(y)) + + // out(0) = c0[even]*x + c0[odd]*y (c0, x, y 16-bit, out 32-bit) + // out(1) = c1[even]*x + c1[odd]*y + #define dct_rot(out0,out1, x,y,c0,c1) \ + __m128i c0##lo = _mm_unpacklo_epi16((x),(y)); \ + __m128i c0##hi = _mm_unpackhi_epi16((x),(y)); \ + __m128i out0##_l = _mm_madd_epi16(c0##lo, c0); \ + __m128i out0##_h = _mm_madd_epi16(c0##hi, c0); \ + __m128i out1##_l = _mm_madd_epi16(c0##lo, c1); \ + __m128i out1##_h = _mm_madd_epi16(c0##hi, c1) + + // out = in << 12 (in 16-bit, out 32-bit) + #define dct_widen(out, in) \ + __m128i out##_l = _mm_srai_epi32(_mm_unpacklo_epi16(_mm_setzero_si128(), (in)), 4); \ + __m128i out##_h = _mm_srai_epi32(_mm_unpackhi_epi16(_mm_setzero_si128(), (in)), 4) + + // wide add + #define dct_wadd(out, a, b) \ + __m128i out##_l = _mm_add_epi32(a##_l, b##_l); \ + __m128i out##_h = _mm_add_epi32(a##_h, b##_h) + + // wide sub + #define dct_wsub(out, a, b) \ + __m128i out##_l = _mm_sub_epi32(a##_l, b##_l); \ + __m128i out##_h = _mm_sub_epi32(a##_h, b##_h) + + // butterfly a/b, add bias, then shift by "s" and pack + #define dct_bfly32o(out0, out1, a,b,bias,s) \ + { \ + __m128i abiased_l = _mm_add_epi32(a##_l, bias); \ + __m128i abiased_h = _mm_add_epi32(a##_h, bias); \ + dct_wadd(sum, abiased, b); \ + dct_wsub(dif, abiased, b); \ + out0 = _mm_packs_epi32(_mm_srai_epi32(sum_l, s), _mm_srai_epi32(sum_h, s)); \ + out1 = _mm_packs_epi32(_mm_srai_epi32(dif_l, s), _mm_srai_epi32(dif_h, s)); \ + } + + // 8-bit interleave step (for transposes) + #define dct_interleave8(a, b) \ + tmp = a; \ + a = _mm_unpacklo_epi8(a, b); \ + b = _mm_unpackhi_epi8(tmp, b) + + // 16-bit interleave step (for transposes) + #define dct_interleave16(a, b) \ + tmp = a; \ + a = _mm_unpacklo_epi16(a, b); \ + b = _mm_unpackhi_epi16(tmp, b) + + #define dct_pass(bias,shift) \ + { \ + /* even part */ \ + dct_rot(t2e,t3e, row2,row6, rot0_0,rot0_1); \ + __m128i sum04 = _mm_add_epi16(row0, row4); \ + __m128i dif04 = _mm_sub_epi16(row0, row4); \ + dct_widen(t0e, sum04); \ + dct_widen(t1e, dif04); \ + dct_wadd(x0, t0e, t3e); \ + dct_wsub(x3, t0e, t3e); \ + dct_wadd(x1, t1e, t2e); \ + dct_wsub(x2, t1e, t2e); \ + /* odd part */ \ + dct_rot(y0o,y2o, row7,row3, rot2_0,rot2_1); \ + dct_rot(y1o,y3o, row5,row1, rot3_0,rot3_1); \ + __m128i sum17 = _mm_add_epi16(row1, row7); \ + __m128i sum35 = _mm_add_epi16(row3, row5); \ + dct_rot(y4o,y5o, sum17,sum35, rot1_0,rot1_1); \ + dct_wadd(x4, y0o, y4o); \ + dct_wadd(x5, y1o, y5o); \ + dct_wadd(x6, y2o, y5o); \ + dct_wadd(x7, y3o, y4o); \ + dct_bfly32o(row0,row7, x0,x7,bias,shift); \ + dct_bfly32o(row1,row6, x1,x6,bias,shift); \ + dct_bfly32o(row2,row5, x2,x5,bias,shift); \ + dct_bfly32o(row3,row4, x3,x4,bias,shift); \ + } + + __m128i rot0_0 = dct_const(stbi__f2f(0.5411961f), stbi__f2f(0.5411961f) + stbi__f2f(-1.847759065f)); + __m128i rot0_1 = dct_const(stbi__f2f(0.5411961f) + stbi__f2f( 0.765366865f), stbi__f2f(0.5411961f)); + __m128i rot1_0 = dct_const(stbi__f2f(1.175875602f) + stbi__f2f(-0.899976223f), stbi__f2f(1.175875602f)); + __m128i rot1_1 = dct_const(stbi__f2f(1.175875602f), stbi__f2f(1.175875602f) + stbi__f2f(-2.562915447f)); + __m128i rot2_0 = dct_const(stbi__f2f(-1.961570560f) + stbi__f2f( 0.298631336f), stbi__f2f(-1.961570560f)); + __m128i rot2_1 = dct_const(stbi__f2f(-1.961570560f), stbi__f2f(-1.961570560f) + stbi__f2f( 3.072711026f)); + __m128i rot3_0 = dct_const(stbi__f2f(-0.390180644f) + stbi__f2f( 2.053119869f), stbi__f2f(-0.390180644f)); + __m128i rot3_1 = dct_const(stbi__f2f(-0.390180644f), stbi__f2f(-0.390180644f) + stbi__f2f( 1.501321110f)); + + // rounding biases in column/row passes, see stbi__idct_block for explanation. + __m128i bias_0 = _mm_set1_epi32(512); + __m128i bias_1 = _mm_set1_epi32(65536 + (128<<17)); + + // load + row0 = _mm_load_si128((const __m128i *) (data + 0*8)); + row1 = _mm_load_si128((const __m128i *) (data + 1*8)); + row2 = _mm_load_si128((const __m128i *) (data + 2*8)); + row3 = _mm_load_si128((const __m128i *) (data + 3*8)); + row4 = _mm_load_si128((const __m128i *) (data + 4*8)); + row5 = _mm_load_si128((const __m128i *) (data + 5*8)); + row6 = _mm_load_si128((const __m128i *) (data + 6*8)); + row7 = _mm_load_si128((const __m128i *) (data + 7*8)); + + // column pass + dct_pass(bias_0, 10); + + { + // 16bit 8x8 transpose pass 1 + dct_interleave16(row0, row4); + dct_interleave16(row1, row5); + dct_interleave16(row2, row6); + dct_interleave16(row3, row7); + + // transpose pass 2 + dct_interleave16(row0, row2); + dct_interleave16(row1, row3); + dct_interleave16(row4, row6); + dct_interleave16(row5, row7); + + // transpose pass 3 + dct_interleave16(row0, row1); + dct_interleave16(row2, row3); + dct_interleave16(row4, row5); + dct_interleave16(row6, row7); + } + + // row pass + dct_pass(bias_1, 17); + + { + // pack + __m128i p0 = _mm_packus_epi16(row0, row1); // a0a1a2a3...a7b0b1b2b3...b7 + __m128i p1 = _mm_packus_epi16(row2, row3); + __m128i p2 = _mm_packus_epi16(row4, row5); + __m128i p3 = _mm_packus_epi16(row6, row7); + + // 8bit 8x8 transpose pass 1 + dct_interleave8(p0, p2); // a0e0a1e1... + dct_interleave8(p1, p3); // c0g0c1g1... + + // transpose pass 2 + dct_interleave8(p0, p1); // a0c0e0g0... + dct_interleave8(p2, p3); // b0d0f0h0... + + // transpose pass 3 + dct_interleave8(p0, p2); // a0b0c0d0... + dct_interleave8(p1, p3); // a4b4c4d4... + + // store + _mm_storel_epi64((__m128i *) out, p0); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p0, 0x4e)); out += out_stride; + _mm_storel_epi64((__m128i *) out, p2); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p2, 0x4e)); out += out_stride; + _mm_storel_epi64((__m128i *) out, p1); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p1, 0x4e)); out += out_stride; + _mm_storel_epi64((__m128i *) out, p3); out += out_stride; + _mm_storel_epi64((__m128i *) out, _mm_shuffle_epi32(p3, 0x4e)); + } + +#undef dct_const +#undef dct_rot +#undef dct_widen +#undef dct_wadd +#undef dct_wsub +#undef dct_bfly32o +#undef dct_interleave8 +#undef dct_interleave16 +#undef dct_pass +} + +#endif // STBI_SSE2 + +#ifdef STBI_NEON + +// NEON integer IDCT. should produce bit-identical +// results to the generic C version. +static void stbi__idct_simd(stbi_uc *out, int out_stride, short data[64]) +{ + int16x8_t row0, row1, row2, row3, row4, row5, row6, row7; + + int16x4_t rot0_0 = vdup_n_s16(stbi__f2f(0.5411961f)); + int16x4_t rot0_1 = vdup_n_s16(stbi__f2f(-1.847759065f)); + int16x4_t rot0_2 = vdup_n_s16(stbi__f2f( 0.765366865f)); + int16x4_t rot1_0 = vdup_n_s16(stbi__f2f( 1.175875602f)); + int16x4_t rot1_1 = vdup_n_s16(stbi__f2f(-0.899976223f)); + int16x4_t rot1_2 = vdup_n_s16(stbi__f2f(-2.562915447f)); + int16x4_t rot2_0 = vdup_n_s16(stbi__f2f(-1.961570560f)); + int16x4_t rot2_1 = vdup_n_s16(stbi__f2f(-0.390180644f)); + int16x4_t rot3_0 = vdup_n_s16(stbi__f2f( 0.298631336f)); + int16x4_t rot3_1 = vdup_n_s16(stbi__f2f( 2.053119869f)); + int16x4_t rot3_2 = vdup_n_s16(stbi__f2f( 3.072711026f)); + int16x4_t rot3_3 = vdup_n_s16(stbi__f2f( 1.501321110f)); + +#define dct_long_mul(out, inq, coeff) \ + int32x4_t out##_l = vmull_s16(vget_low_s16(inq), coeff); \ + int32x4_t out##_h = vmull_s16(vget_high_s16(inq), coeff) + +#define dct_long_mac(out, acc, inq, coeff) \ + int32x4_t out##_l = vmlal_s16(acc##_l, vget_low_s16(inq), coeff); \ + int32x4_t out##_h = vmlal_s16(acc##_h, vget_high_s16(inq), coeff) + +#define dct_widen(out, inq) \ + int32x4_t out##_l = vshll_n_s16(vget_low_s16(inq), 12); \ + int32x4_t out##_h = vshll_n_s16(vget_high_s16(inq), 12) + +// wide add +#define dct_wadd(out, a, b) \ + int32x4_t out##_l = vaddq_s32(a##_l, b##_l); \ + int32x4_t out##_h = vaddq_s32(a##_h, b##_h) + +// wide sub +#define dct_wsub(out, a, b) \ + int32x4_t out##_l = vsubq_s32(a##_l, b##_l); \ + int32x4_t out##_h = vsubq_s32(a##_h, b##_h) + +// butterfly a/b, then shift using "shiftop" by "s" and pack +#define dct_bfly32o(out0,out1, a,b,shiftop,s) \ + { \ + dct_wadd(sum, a, b); \ + dct_wsub(dif, a, b); \ + out0 = vcombine_s16(shiftop(sum_l, s), shiftop(sum_h, s)); \ + out1 = vcombine_s16(shiftop(dif_l, s), shiftop(dif_h, s)); \ + } + +#define dct_pass(shiftop, shift) \ + { \ + /* even part */ \ + int16x8_t sum26 = vaddq_s16(row2, row6); \ + dct_long_mul(p1e, sum26, rot0_0); \ + dct_long_mac(t2e, p1e, row6, rot0_1); \ + dct_long_mac(t3e, p1e, row2, rot0_2); \ + int16x8_t sum04 = vaddq_s16(row0, row4); \ + int16x8_t dif04 = vsubq_s16(row0, row4); \ + dct_widen(t0e, sum04); \ + dct_widen(t1e, dif04); \ + dct_wadd(x0, t0e, t3e); \ + dct_wsub(x3, t0e, t3e); \ + dct_wadd(x1, t1e, t2e); \ + dct_wsub(x2, t1e, t2e); \ + /* odd part */ \ + int16x8_t sum15 = vaddq_s16(row1, row5); \ + int16x8_t sum17 = vaddq_s16(row1, row7); \ + int16x8_t sum35 = vaddq_s16(row3, row5); \ + int16x8_t sum37 = vaddq_s16(row3, row7); \ + int16x8_t sumodd = vaddq_s16(sum17, sum35); \ + dct_long_mul(p5o, sumodd, rot1_0); \ + dct_long_mac(p1o, p5o, sum17, rot1_1); \ + dct_long_mac(p2o, p5o, sum35, rot1_2); \ + dct_long_mul(p3o, sum37, rot2_0); \ + dct_long_mul(p4o, sum15, rot2_1); \ + dct_wadd(sump13o, p1o, p3o); \ + dct_wadd(sump24o, p2o, p4o); \ + dct_wadd(sump23o, p2o, p3o); \ + dct_wadd(sump14o, p1o, p4o); \ + dct_long_mac(x4, sump13o, row7, rot3_0); \ + dct_long_mac(x5, sump24o, row5, rot3_1); \ + dct_long_mac(x6, sump23o, row3, rot3_2); \ + dct_long_mac(x7, sump14o, row1, rot3_3); \ + dct_bfly32o(row0,row7, x0,x7,shiftop,shift); \ + dct_bfly32o(row1,row6, x1,x6,shiftop,shift); \ + dct_bfly32o(row2,row5, x2,x5,shiftop,shift); \ + dct_bfly32o(row3,row4, x3,x4,shiftop,shift); \ + } + + // load + row0 = vld1q_s16(data + 0*8); + row1 = vld1q_s16(data + 1*8); + row2 = vld1q_s16(data + 2*8); + row3 = vld1q_s16(data + 3*8); + row4 = vld1q_s16(data + 4*8); + row5 = vld1q_s16(data + 5*8); + row6 = vld1q_s16(data + 6*8); + row7 = vld1q_s16(data + 7*8); + + // add DC bias + row0 = vaddq_s16(row0, vsetq_lane_s16(1024, vdupq_n_s16(0), 0)); + + // column pass + dct_pass(vrshrn_n_s32, 10); + + // 16bit 8x8 transpose + { +// these three map to a single VTRN.16, VTRN.32, and VSWP, respectively. +// whether compilers actually get this is another story, sadly. +#define dct_trn16(x, y) { int16x8x2_t t = vtrnq_s16(x, y); x = t.val[0]; y = t.val[1]; } +#define dct_trn32(x, y) { int32x4x2_t t = vtrnq_s32(vreinterpretq_s32_s16(x), vreinterpretq_s32_s16(y)); x = vreinterpretq_s16_s32(t.val[0]); y = vreinterpretq_s16_s32(t.val[1]); } +#define dct_trn64(x, y) { int16x8_t x0 = x; int16x8_t y0 = y; x = vcombine_s16(vget_low_s16(x0), vget_low_s16(y0)); y = vcombine_s16(vget_high_s16(x0), vget_high_s16(y0)); } + + // pass 1 + dct_trn16(row0, row1); // a0b0a2b2a4b4a6b6 + dct_trn16(row2, row3); + dct_trn16(row4, row5); + dct_trn16(row6, row7); + + // pass 2 + dct_trn32(row0, row2); // a0b0c0d0a4b4c4d4 + dct_trn32(row1, row3); + dct_trn32(row4, row6); + dct_trn32(row5, row7); + + // pass 3 + dct_trn64(row0, row4); // a0b0c0d0e0f0g0h0 + dct_trn64(row1, row5); + dct_trn64(row2, row6); + dct_trn64(row3, row7); + +#undef dct_trn16 +#undef dct_trn32 +#undef dct_trn64 + } + + // row pass + // vrshrn_n_s32 only supports shifts up to 16, we need + // 17. so do a non-rounding shift of 16 first then follow + // up with a rounding shift by 1. + dct_pass(vshrn_n_s32, 16); + + { + // pack and round + uint8x8_t p0 = vqrshrun_n_s16(row0, 1); + uint8x8_t p1 = vqrshrun_n_s16(row1, 1); + uint8x8_t p2 = vqrshrun_n_s16(row2, 1); + uint8x8_t p3 = vqrshrun_n_s16(row3, 1); + uint8x8_t p4 = vqrshrun_n_s16(row4, 1); + uint8x8_t p5 = vqrshrun_n_s16(row5, 1); + uint8x8_t p6 = vqrshrun_n_s16(row6, 1); + uint8x8_t p7 = vqrshrun_n_s16(row7, 1); + + // again, these can translate into one instruction, but often don't. +#define dct_trn8_8(x, y) { uint8x8x2_t t = vtrn_u8(x, y); x = t.val[0]; y = t.val[1]; } +#define dct_trn8_16(x, y) { uint16x4x2_t t = vtrn_u16(vreinterpret_u16_u8(x), vreinterpret_u16_u8(y)); x = vreinterpret_u8_u16(t.val[0]); y = vreinterpret_u8_u16(t.val[1]); } +#define dct_trn8_32(x, y) { uint32x2x2_t t = vtrn_u32(vreinterpret_u32_u8(x), vreinterpret_u32_u8(y)); x = vreinterpret_u8_u32(t.val[0]); y = vreinterpret_u8_u32(t.val[1]); } + + // sadly can't use interleaved stores here since we only write + // 8 bytes to each scan line! + + // 8x8 8-bit transpose pass 1 + dct_trn8_8(p0, p1); + dct_trn8_8(p2, p3); + dct_trn8_8(p4, p5); + dct_trn8_8(p6, p7); + + // pass 2 + dct_trn8_16(p0, p2); + dct_trn8_16(p1, p3); + dct_trn8_16(p4, p6); + dct_trn8_16(p5, p7); + + // pass 3 + dct_trn8_32(p0, p4); + dct_trn8_32(p1, p5); + dct_trn8_32(p2, p6); + dct_trn8_32(p3, p7); + + // store + vst1_u8(out, p0); out += out_stride; + vst1_u8(out, p1); out += out_stride; + vst1_u8(out, p2); out += out_stride; + vst1_u8(out, p3); out += out_stride; + vst1_u8(out, p4); out += out_stride; + vst1_u8(out, p5); out += out_stride; + vst1_u8(out, p6); out += out_stride; + vst1_u8(out, p7); + +#undef dct_trn8_8 +#undef dct_trn8_16 +#undef dct_trn8_32 + } + +#undef dct_long_mul +#undef dct_long_mac +#undef dct_widen +#undef dct_wadd +#undef dct_wsub +#undef dct_bfly32o +#undef dct_pass +} + +#endif // STBI_NEON + +#define STBI__MARKER_none 0xff +// if there's a pending marker from the entropy stream, return that +// otherwise, fetch from the stream and get a marker. if there's no +// marker, return 0xff, which is never a valid marker value +static stbi_uc stbi__get_marker(stbi__jpeg *j) +{ + stbi_uc x; + if (j->marker != STBI__MARKER_none) { x = j->marker; j->marker = STBI__MARKER_none; return x; } + x = stbi__get8(j->s); + if (x != 0xff) return STBI__MARKER_none; + while (x == 0xff) + x = stbi__get8(j->s); // consume repeated 0xff fill bytes + return x; +} + +// in each scan, we'll have scan_n components, and the order +// of the components is specified by order[] +#define STBI__RESTART(x) ((x) >= 0xd0 && (x) <= 0xd7) + +// after a restart interval, stbi__jpeg_reset the entropy decoder and +// the dc prediction +static void stbi__jpeg_reset(stbi__jpeg *j) +{ + j->code_bits = 0; + j->code_buffer = 0; + j->nomore = 0; + j->img_comp[0].dc_pred = j->img_comp[1].dc_pred = j->img_comp[2].dc_pred = j->img_comp[3].dc_pred = 0; + j->marker = STBI__MARKER_none; + j->todo = j->restart_interval ? j->restart_interval : 0x7fffffff; + j->eob_run = 0; + // no more than 1<<31 MCUs if no restart_interal? that's plenty safe, + // since we don't even allow 1<<30 pixels +} + +static int stbi__parse_entropy_coded_data(stbi__jpeg *z) +{ + stbi__jpeg_reset(z); + if (!z->progressive) { + if (z->scan_n == 1) { + int i,j; + STBI_SIMD_ALIGN(short, data[64]); + int n = z->order[0]; + // non-interleaved data, we just need to process one block at a time, + // in trivial scanline order + // number of blocks to do just depends on how many actual "pixels" this + // component has, independent of interleaved MCU blocking and such + int w = (z->img_comp[n].x+7) >> 3; + int h = (z->img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + int ha = z->img_comp[n].ha; + if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0; + z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data); + // every data block is an MCU, so countdown the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + // if it's NOT a restart, then just bail, so we get corrupt data + // rather than no data + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } else { // interleaved + int i,j,k,x,y; + STBI_SIMD_ALIGN(short, data[64]); + for (j=0; j < z->img_mcu_y; ++j) { + for (i=0; i < z->img_mcu_x; ++i) { + // scan an interleaved mcu... process scan_n components in order + for (k=0; k < z->scan_n; ++k) { + int n = z->order[k]; + // scan out an mcu's worth of this component; that's just determined + // by the basic H and V specified for the component + for (y=0; y < z->img_comp[n].v; ++y) { + for (x=0; x < z->img_comp[n].h; ++x) { + int x2 = (i*z->img_comp[n].h + x)*8; + int y2 = (j*z->img_comp[n].v + y)*8; + int ha = z->img_comp[n].ha; + if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+ha, z->fast_ac[ha], n, z->dequant[z->img_comp[n].tq])) return 0; + z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*y2+x2, z->img_comp[n].w2, data); + } + } + } + // after all interleaved components, that's an interleaved MCU, + // so now count down the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } + } else { + if (z->scan_n == 1) { + int i,j; + int n = z->order[0]; + // non-interleaved data, we just need to process one block at a time, + // in trivial scanline order + // number of blocks to do just depends on how many actual "pixels" this + // component has, independent of interleaved MCU blocking and such + int w = (z->img_comp[n].x+7) >> 3; + int h = (z->img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w); + if (z->spec_start == 0) { + if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n)) + return 0; + } else { + int ha = z->img_comp[n].ha; + if (!stbi__jpeg_decode_block_prog_ac(z, data, &z->huff_ac[ha], z->fast_ac[ha])) + return 0; + } + // every data block is an MCU, so countdown the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } else { // interleaved + int i,j,k,x,y; + for (j=0; j < z->img_mcu_y; ++j) { + for (i=0; i < z->img_mcu_x; ++i) { + // scan an interleaved mcu... process scan_n components in order + for (k=0; k < z->scan_n; ++k) { + int n = z->order[k]; + // scan out an mcu's worth of this component; that's just determined + // by the basic H and V specified for the component + for (y=0; y < z->img_comp[n].v; ++y) { + for (x=0; x < z->img_comp[n].h; ++x) { + int x2 = (i*z->img_comp[n].h + x); + int y2 = (j*z->img_comp[n].v + y); + short *data = z->img_comp[n].coeff + 64 * (x2 + y2 * z->img_comp[n].coeff_w); + if (!stbi__jpeg_decode_block_prog_dc(z, data, &z->huff_dc[z->img_comp[n].hd], n)) + return 0; + } + } + } + // after all interleaved components, that's an interleaved MCU, + // so now count down the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) stbi__grow_buffer_unsafe(z); + if (!STBI__RESTART(z->marker)) return 1; + stbi__jpeg_reset(z); + } + } + } + return 1; + } + } +} + +static void stbi__jpeg_dequantize(short *data, stbi__uint16 *dequant) +{ + int i; + for (i=0; i < 64; ++i) + data[i] *= dequant[i]; +} + +static void stbi__jpeg_finish(stbi__jpeg *z) +{ + if (z->progressive) { + // dequantize and idct the data + int i,j,n; + for (n=0; n < z->s->img_n; ++n) { + int w = (z->img_comp[n].x+7) >> 3; + int h = (z->img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + short *data = z->img_comp[n].coeff + 64 * (i + j * z->img_comp[n].coeff_w); + stbi__jpeg_dequantize(data, z->dequant[z->img_comp[n].tq]); + z->idct_block_kernel(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data); + } + } + } + } +} + +static int stbi__process_marker(stbi__jpeg *z, int m) +{ + int L; + switch (m) { + case STBI__MARKER_none: // no marker found + return stbi__err("expected marker","Corrupt JPEG"); + + case 0xDD: // DRI - specify restart interval + if (stbi__get16be(z->s) != 4) return stbi__err("bad DRI len","Corrupt JPEG"); + z->restart_interval = stbi__get16be(z->s); + return 1; + + case 0xDB: // DQT - define quantization table + L = stbi__get16be(z->s)-2; + while (L > 0) { + int q = stbi__get8(z->s); + int p = q >> 4, sixteen = (p != 0); + int t = q & 15,i; + if (p != 0 && p != 1) return stbi__err("bad DQT type","Corrupt JPEG"); + if (t > 3) return stbi__err("bad DQT table","Corrupt JPEG"); + + for (i=0; i < 64; ++i) + z->dequant[t][stbi__jpeg_dezigzag[i]] = (stbi__uint16)(sixteen ? stbi__get16be(z->s) : stbi__get8(z->s)); + L -= (sixteen ? 129 : 65); + } + return L==0; + + case 0xC4: // DHT - define huffman table + L = stbi__get16be(z->s)-2; + while (L > 0) { + stbi_uc *v; + int sizes[16],i,n=0; + int q = stbi__get8(z->s); + int tc = q >> 4; + int th = q & 15; + if (tc > 1 || th > 3) return stbi__err("bad DHT header","Corrupt JPEG"); + for (i=0; i < 16; ++i) { + sizes[i] = stbi__get8(z->s); + n += sizes[i]; + } + if(n > 256) return stbi__err("bad DHT header","Corrupt JPEG"); // Loop over i < n would write past end of values! + L -= 17; + if (tc == 0) { + if (!stbi__build_huffman(z->huff_dc+th, sizes)) return 0; + v = z->huff_dc[th].values; + } else { + if (!stbi__build_huffman(z->huff_ac+th, sizes)) return 0; + v = z->huff_ac[th].values; + } + for (i=0; i < n; ++i) + v[i] = stbi__get8(z->s); + if (tc != 0) + stbi__build_fast_ac(z->fast_ac[th], z->huff_ac + th); + L -= n; + } + return L==0; + } + + // check for comment block or APP blocks + if ((m >= 0xE0 && m <= 0xEF) || m == 0xFE) { + L = stbi__get16be(z->s); + if (L < 2) { + if (m == 0xFE) + return stbi__err("bad COM len","Corrupt JPEG"); + else + return stbi__err("bad APP len","Corrupt JPEG"); + } + L -= 2; + + if (m == 0xE0 && L >= 5) { // JFIF APP0 segment + static const unsigned char tag[5] = {'J','F','I','F','\0'}; + int ok = 1; + int i; + for (i=0; i < 5; ++i) + if (stbi__get8(z->s) != tag[i]) + ok = 0; + L -= 5; + if (ok) + z->jfif = 1; + } else if (m == 0xEE && L >= 12) { // Adobe APP14 segment + static const unsigned char tag[6] = {'A','d','o','b','e','\0'}; + int ok = 1; + int i; + for (i=0; i < 6; ++i) + if (stbi__get8(z->s) != tag[i]) + ok = 0; + L -= 6; + if (ok) { + stbi__get8(z->s); // version + stbi__get16be(z->s); // flags0 + stbi__get16be(z->s); // flags1 + z->app14_color_transform = stbi__get8(z->s); // color transform + L -= 6; + } + } + + stbi__skip(z->s, L); + return 1; + } + + return stbi__err("unknown marker","Corrupt JPEG"); +} + +// after we see SOS +static int stbi__process_scan_header(stbi__jpeg *z) +{ + int i; + int Ls = stbi__get16be(z->s); + z->scan_n = stbi__get8(z->s); + if (z->scan_n < 1 || z->scan_n > 4 || z->scan_n > (int) z->s->img_n) return stbi__err("bad SOS component count","Corrupt JPEG"); + if (Ls != 6+2*z->scan_n) return stbi__err("bad SOS len","Corrupt JPEG"); + for (i=0; i < z->scan_n; ++i) { + int id = stbi__get8(z->s), which; + int q = stbi__get8(z->s); + for (which = 0; which < z->s->img_n; ++which) + if (z->img_comp[which].id == id) + break; + if (which == z->s->img_n) return 0; // no match + z->img_comp[which].hd = q >> 4; if (z->img_comp[which].hd > 3) return stbi__err("bad DC huff","Corrupt JPEG"); + z->img_comp[which].ha = q & 15; if (z->img_comp[which].ha > 3) return stbi__err("bad AC huff","Corrupt JPEG"); + z->order[i] = which; + } + + { + int aa; + z->spec_start = stbi__get8(z->s); + z->spec_end = stbi__get8(z->s); // should be 63, but might be 0 + aa = stbi__get8(z->s); + z->succ_high = (aa >> 4); + z->succ_low = (aa & 15); + if (z->progressive) { + if (z->spec_start > 63 || z->spec_end > 63 || z->spec_start > z->spec_end || z->succ_high > 13 || z->succ_low > 13) + return stbi__err("bad SOS", "Corrupt JPEG"); + } else { + if (z->spec_start != 0) return stbi__err("bad SOS","Corrupt JPEG"); + if (z->succ_high != 0 || z->succ_low != 0) return stbi__err("bad SOS","Corrupt JPEG"); + z->spec_end = 63; + } + } + + return 1; +} + +static int stbi__free_jpeg_components(stbi__jpeg *z, int ncomp, int why) +{ + int i; + for (i=0; i < ncomp; ++i) { + if (z->img_comp[i].raw_data) { + STBI_FREE(z->img_comp[i].raw_data); + z->img_comp[i].raw_data = NULL; + z->img_comp[i].data = NULL; + } + if (z->img_comp[i].raw_coeff) { + STBI_FREE(z->img_comp[i].raw_coeff); + z->img_comp[i].raw_coeff = 0; + z->img_comp[i].coeff = 0; + } + if (z->img_comp[i].linebuf) { + STBI_FREE(z->img_comp[i].linebuf); + z->img_comp[i].linebuf = NULL; + } + } + return why; +} + +static int stbi__process_frame_header(stbi__jpeg *z, int scan) +{ + stbi__context *s = z->s; + int Lf,p,i,q, h_max=1,v_max=1,c; + Lf = stbi__get16be(s); if (Lf < 11) return stbi__err("bad SOF len","Corrupt JPEG"); // JPEG + p = stbi__get8(s); if (p != 8) return stbi__err("only 8-bit","JPEG format not supported: 8-bit only"); // JPEG baseline + s->img_y = stbi__get16be(s); if (s->img_y == 0) return stbi__err("no header height", "JPEG format not supported: delayed height"); // Legal, but we don't handle it--but neither does IJG + s->img_x = stbi__get16be(s); if (s->img_x == 0) return stbi__err("0 width","Corrupt JPEG"); // JPEG requires + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + c = stbi__get8(s); + if (c != 3 && c != 1 && c != 4) return stbi__err("bad component count","Corrupt JPEG"); + s->img_n = c; + for (i=0; i < c; ++i) { + z->img_comp[i].data = NULL; + z->img_comp[i].linebuf = NULL; + } + + if (Lf != 8+3*s->img_n) return stbi__err("bad SOF len","Corrupt JPEG"); + + z->rgb = 0; + for (i=0; i < s->img_n; ++i) { + static const unsigned char rgb[3] = { 'R', 'G', 'B' }; + z->img_comp[i].id = stbi__get8(s); + if (s->img_n == 3 && z->img_comp[i].id == rgb[i]) + ++z->rgb; + q = stbi__get8(s); + z->img_comp[i].h = (q >> 4); if (!z->img_comp[i].h || z->img_comp[i].h > 4) return stbi__err("bad H","Corrupt JPEG"); + z->img_comp[i].v = q & 15; if (!z->img_comp[i].v || z->img_comp[i].v > 4) return stbi__err("bad V","Corrupt JPEG"); + z->img_comp[i].tq = stbi__get8(s); if (z->img_comp[i].tq > 3) return stbi__err("bad TQ","Corrupt JPEG"); + } + + if (scan != STBI__SCAN_load) return 1; + + if (!stbi__mad3sizes_valid(s->img_x, s->img_y, s->img_n, 0)) return stbi__err("too large", "Image too large to decode"); + + for (i=0; i < s->img_n; ++i) { + if (z->img_comp[i].h > h_max) h_max = z->img_comp[i].h; + if (z->img_comp[i].v > v_max) v_max = z->img_comp[i].v; + } + + // check that plane subsampling factors are integer ratios; our resamplers can't deal with fractional ratios + // and I've never seen a non-corrupted JPEG file actually use them + for (i=0; i < s->img_n; ++i) { + if (h_max % z->img_comp[i].h != 0) return stbi__err("bad H","Corrupt JPEG"); + if (v_max % z->img_comp[i].v != 0) return stbi__err("bad V","Corrupt JPEG"); + } + + // compute interleaved mcu info + z->img_h_max = h_max; + z->img_v_max = v_max; + z->img_mcu_w = h_max * 8; + z->img_mcu_h = v_max * 8; + // these sizes can't be more than 17 bits + z->img_mcu_x = (s->img_x + z->img_mcu_w-1) / z->img_mcu_w; + z->img_mcu_y = (s->img_y + z->img_mcu_h-1) / z->img_mcu_h; + + for (i=0; i < s->img_n; ++i) { + // number of effective pixels (e.g. for non-interleaved MCU) + z->img_comp[i].x = (s->img_x * z->img_comp[i].h + h_max-1) / h_max; + z->img_comp[i].y = (s->img_y * z->img_comp[i].v + v_max-1) / v_max; + // to simplify generation, we'll allocate enough memory to decode + // the bogus oversized data from using interleaved MCUs and their + // big blocks (e.g. a 16x16 iMCU on an image of width 33); we won't + // discard the extra data until colorspace conversion + // + // img_mcu_x, img_mcu_y: <=17 bits; comp[i].h and .v are <=4 (checked earlier) + // so these muls can't overflow with 32-bit ints (which we require) + z->img_comp[i].w2 = z->img_mcu_x * z->img_comp[i].h * 8; + z->img_comp[i].h2 = z->img_mcu_y * z->img_comp[i].v * 8; + z->img_comp[i].coeff = 0; + z->img_comp[i].raw_coeff = 0; + z->img_comp[i].linebuf = NULL; + z->img_comp[i].raw_data = stbi__malloc_mad2(z->img_comp[i].w2, z->img_comp[i].h2, 15); + if (z->img_comp[i].raw_data == NULL) + return stbi__free_jpeg_components(z, i+1, stbi__err("outofmem", "Out of memory")); + // align blocks for idct using mmx/sse + z->img_comp[i].data = (stbi_uc*) (((size_t) z->img_comp[i].raw_data + 15) & ~15); + if (z->progressive) { + // w2, h2 are multiples of 8 (see above) + z->img_comp[i].coeff_w = z->img_comp[i].w2 / 8; + z->img_comp[i].coeff_h = z->img_comp[i].h2 / 8; + z->img_comp[i].raw_coeff = stbi__malloc_mad3(z->img_comp[i].w2, z->img_comp[i].h2, sizeof(short), 15); + if (z->img_comp[i].raw_coeff == NULL) + return stbi__free_jpeg_components(z, i+1, stbi__err("outofmem", "Out of memory")); + z->img_comp[i].coeff = (short*) (((size_t) z->img_comp[i].raw_coeff + 15) & ~15); + } + } + + return 1; +} + +// use comparisons since in some cases we handle more than one case (e.g. SOF) +#define stbi__DNL(x) ((x) == 0xdc) +#define stbi__SOI(x) ((x) == 0xd8) +#define stbi__EOI(x) ((x) == 0xd9) +#define stbi__SOF(x) ((x) == 0xc0 || (x) == 0xc1 || (x) == 0xc2) +#define stbi__SOS(x) ((x) == 0xda) + +#define stbi__SOF_progressive(x) ((x) == 0xc2) + +static int stbi__decode_jpeg_header(stbi__jpeg *z, int scan) +{ + int m; + z->jfif = 0; + z->app14_color_transform = -1; // valid values are 0,1,2 + z->marker = STBI__MARKER_none; // initialize cached marker to empty + m = stbi__get_marker(z); + if (!stbi__SOI(m)) return stbi__err("no SOI","Corrupt JPEG"); + if (scan == STBI__SCAN_type) return 1; + m = stbi__get_marker(z); + while (!stbi__SOF(m)) { + if (!stbi__process_marker(z,m)) return 0; + m = stbi__get_marker(z); + while (m == STBI__MARKER_none) { + // some files have extra padding after their blocks, so ok, we'll scan + if (stbi__at_eof(z->s)) return stbi__err("no SOF", "Corrupt JPEG"); + m = stbi__get_marker(z); + } + } + z->progressive = stbi__SOF_progressive(m); + if (!stbi__process_frame_header(z, scan)) return 0; + return 1; +} + +static stbi_uc stbi__skip_jpeg_junk_at_end(stbi__jpeg *j) +{ + // some JPEGs have junk at end, skip over it but if we find what looks + // like a valid marker, resume there + while (!stbi__at_eof(j->s)) { + stbi_uc x = stbi__get8(j->s); + while (x == 0xff) { // might be a marker + if (stbi__at_eof(j->s)) return STBI__MARKER_none; + x = stbi__get8(j->s); + if (x != 0x00 && x != 0xff) { + // not a stuffed zero or lead-in to another marker, looks + // like an actual marker, return it + return x; + } + // stuffed zero has x=0 now which ends the loop, meaning we go + // back to regular scan loop. + // repeated 0xff keeps trying to read the next byte of the marker. + } + } + return STBI__MARKER_none; +} + +// decode image to YCbCr format +static int stbi__decode_jpeg_image(stbi__jpeg *j) +{ + int m; + for (m = 0; m < 4; m++) { + j->img_comp[m].raw_data = NULL; + j->img_comp[m].raw_coeff = NULL; + } + j->restart_interval = 0; + if (!stbi__decode_jpeg_header(j, STBI__SCAN_load)) return 0; + m = stbi__get_marker(j); + while (!stbi__EOI(m)) { + if (stbi__SOS(m)) { + if (!stbi__process_scan_header(j)) return 0; + if (!stbi__parse_entropy_coded_data(j)) return 0; + if (j->marker == STBI__MARKER_none ) { + j->marker = stbi__skip_jpeg_junk_at_end(j); + // if we reach eof without hitting a marker, stbi__get_marker() below will fail and we'll eventually return 0 + } + m = stbi__get_marker(j); + if (STBI__RESTART(m)) + m = stbi__get_marker(j); + } else if (stbi__DNL(m)) { + int Ld = stbi__get16be(j->s); + stbi__uint32 NL = stbi__get16be(j->s); + if (Ld != 4) return stbi__err("bad DNL len", "Corrupt JPEG"); + if (NL != j->s->img_y) return stbi__err("bad DNL height", "Corrupt JPEG"); + m = stbi__get_marker(j); + } else { + if (!stbi__process_marker(j, m)) return 1; + m = stbi__get_marker(j); + } + } + if (j->progressive) + stbi__jpeg_finish(j); + return 1; +} + +// static jfif-centered resampling (across block boundaries) + +typedef stbi_uc *(*resample_row_func)(stbi_uc *out, stbi_uc *in0, stbi_uc *in1, + int w, int hs); + +#define stbi__div4(x) ((stbi_uc) ((x) >> 2)) + +static stbi_uc *resample_row_1(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + STBI_NOTUSED(out); + STBI_NOTUSED(in_far); + STBI_NOTUSED(w); + STBI_NOTUSED(hs); + return in_near; +} + +static stbi_uc* stbi__resample_row_v_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate two samples vertically for every one in input + int i; + STBI_NOTUSED(hs); + for (i=0; i < w; ++i) + out[i] = stbi__div4(3*in_near[i] + in_far[i] + 2); + return out; +} + +static stbi_uc* stbi__resample_row_h_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate two samples horizontally for every one in input + int i; + stbi_uc *input = in_near; + + if (w == 1) { + // if only one sample, can't do any interpolation + out[0] = out[1] = input[0]; + return out; + } + + out[0] = input[0]; + out[1] = stbi__div4(input[0]*3 + input[1] + 2); + for (i=1; i < w-1; ++i) { + int n = 3*input[i]+2; + out[i*2+0] = stbi__div4(n+input[i-1]); + out[i*2+1] = stbi__div4(n+input[i+1]); + } + out[i*2+0] = stbi__div4(input[w-2]*3 + input[w-1] + 2); + out[i*2+1] = input[w-1]; + + STBI_NOTUSED(in_far); + STBI_NOTUSED(hs); + + return out; +} + +#define stbi__div16(x) ((stbi_uc) ((x) >> 4)) + +static stbi_uc *stbi__resample_row_hv_2(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate 2x2 samples for every one in input + int i,t0,t1; + if (w == 1) { + out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2); + return out; + } + + t1 = 3*in_near[0] + in_far[0]; + out[0] = stbi__div4(t1+2); + for (i=1; i < w; ++i) { + t0 = t1; + t1 = 3*in_near[i]+in_far[i]; + out[i*2-1] = stbi__div16(3*t0 + t1 + 8); + out[i*2 ] = stbi__div16(3*t1 + t0 + 8); + } + out[w*2-1] = stbi__div4(t1+2); + + STBI_NOTUSED(hs); + + return out; +} + +#if defined(STBI_SSE2) || defined(STBI_NEON) +static stbi_uc *stbi__resample_row_hv_2_simd(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // need to generate 2x2 samples for every one in input + int i=0,t0,t1; + + if (w == 1) { + out[0] = out[1] = stbi__div4(3*in_near[0] + in_far[0] + 2); + return out; + } + + t1 = 3*in_near[0] + in_far[0]; + // process groups of 8 pixels for as long as we can. + // note we can't handle the last pixel in a row in this loop + // because we need to handle the filter boundary conditions. + for (; i < ((w-1) & ~7); i += 8) { +#if defined(STBI_SSE2) + // load and perform the vertical filtering pass + // this uses 3*x + y = 4*x + (y - x) + __m128i zero = _mm_setzero_si128(); + __m128i farb = _mm_loadl_epi64((__m128i *) (in_far + i)); + __m128i nearb = _mm_loadl_epi64((__m128i *) (in_near + i)); + __m128i farw = _mm_unpacklo_epi8(farb, zero); + __m128i nearw = _mm_unpacklo_epi8(nearb, zero); + __m128i diff = _mm_sub_epi16(farw, nearw); + __m128i nears = _mm_slli_epi16(nearw, 2); + __m128i curr = _mm_add_epi16(nears, diff); // current row + + // horizontal filter works the same based on shifted vers of current + // row. "prev" is current row shifted right by 1 pixel; we need to + // insert the previous pixel value (from t1). + // "next" is current row shifted left by 1 pixel, with first pixel + // of next block of 8 pixels added in. + __m128i prv0 = _mm_slli_si128(curr, 2); + __m128i nxt0 = _mm_srli_si128(curr, 2); + __m128i prev = _mm_insert_epi16(prv0, t1, 0); + __m128i next = _mm_insert_epi16(nxt0, 3*in_near[i+8] + in_far[i+8], 7); + + // horizontal filter, polyphase implementation since it's convenient: + // even pixels = 3*cur + prev = cur*4 + (prev - cur) + // odd pixels = 3*cur + next = cur*4 + (next - cur) + // note the shared term. + __m128i bias = _mm_set1_epi16(8); + __m128i curs = _mm_slli_epi16(curr, 2); + __m128i prvd = _mm_sub_epi16(prev, curr); + __m128i nxtd = _mm_sub_epi16(next, curr); + __m128i curb = _mm_add_epi16(curs, bias); + __m128i even = _mm_add_epi16(prvd, curb); + __m128i odd = _mm_add_epi16(nxtd, curb); + + // interleave even and odd pixels, then undo scaling. + __m128i int0 = _mm_unpacklo_epi16(even, odd); + __m128i int1 = _mm_unpackhi_epi16(even, odd); + __m128i de0 = _mm_srli_epi16(int0, 4); + __m128i de1 = _mm_srli_epi16(int1, 4); + + // pack and write output + __m128i outv = _mm_packus_epi16(de0, de1); + _mm_storeu_si128((__m128i *) (out + i*2), outv); +#elif defined(STBI_NEON) + // load and perform the vertical filtering pass + // this uses 3*x + y = 4*x + (y - x) + uint8x8_t farb = vld1_u8(in_far + i); + uint8x8_t nearb = vld1_u8(in_near + i); + int16x8_t diff = vreinterpretq_s16_u16(vsubl_u8(farb, nearb)); + int16x8_t nears = vreinterpretq_s16_u16(vshll_n_u8(nearb, 2)); + int16x8_t curr = vaddq_s16(nears, diff); // current row + + // horizontal filter works the same based on shifted vers of current + // row. "prev" is current row shifted right by 1 pixel; we need to + // insert the previous pixel value (from t1). + // "next" is current row shifted left by 1 pixel, with first pixel + // of next block of 8 pixels added in. + int16x8_t prv0 = vextq_s16(curr, curr, 7); + int16x8_t nxt0 = vextq_s16(curr, curr, 1); + int16x8_t prev = vsetq_lane_s16(t1, prv0, 0); + int16x8_t next = vsetq_lane_s16(3*in_near[i+8] + in_far[i+8], nxt0, 7); + + // horizontal filter, polyphase implementation since it's convenient: + // even pixels = 3*cur + prev = cur*4 + (prev - cur) + // odd pixels = 3*cur + next = cur*4 + (next - cur) + // note the shared term. + int16x8_t curs = vshlq_n_s16(curr, 2); + int16x8_t prvd = vsubq_s16(prev, curr); + int16x8_t nxtd = vsubq_s16(next, curr); + int16x8_t even = vaddq_s16(curs, prvd); + int16x8_t odd = vaddq_s16(curs, nxtd); + + // undo scaling and round, then store with even/odd phases interleaved + uint8x8x2_t o; + o.val[0] = vqrshrun_n_s16(even, 4); + o.val[1] = vqrshrun_n_s16(odd, 4); + vst2_u8(out + i*2, o); +#endif + + // "previous" value for next iter + t1 = 3*in_near[i+7] + in_far[i+7]; + } + + t0 = t1; + t1 = 3*in_near[i] + in_far[i]; + out[i*2] = stbi__div16(3*t1 + t0 + 8); + + for (++i; i < w; ++i) { + t0 = t1; + t1 = 3*in_near[i]+in_far[i]; + out[i*2-1] = stbi__div16(3*t0 + t1 + 8); + out[i*2 ] = stbi__div16(3*t1 + t0 + 8); + } + out[w*2-1] = stbi__div4(t1+2); + + STBI_NOTUSED(hs); + + return out; +} +#endif + +static stbi_uc *stbi__resample_row_generic(stbi_uc *out, stbi_uc *in_near, stbi_uc *in_far, int w, int hs) +{ + // resample with nearest-neighbor + int i,j; + STBI_NOTUSED(in_far); + for (i=0; i < w; ++i) + for (j=0; j < hs; ++j) + out[i*hs+j] = in_near[i]; + return out; +} + +// this is a reduced-precision calculation of YCbCr-to-RGB introduced +// to make sure the code produces the same results in both SIMD and scalar +#define stbi__float2fixed(x) (((int) ((x) * 4096.0f + 0.5f)) << 8) +static void stbi__YCbCr_to_RGB_row(stbi_uc *out, const stbi_uc *y, const stbi_uc *pcb, const stbi_uc *pcr, int count, int step) +{ + int i; + for (i=0; i < count; ++i) { + int y_fixed = (y[i] << 20) + (1<<19); // rounding + int r,g,b; + int cr = pcr[i] - 128; + int cb = pcb[i] - 128; + r = y_fixed + cr* stbi__float2fixed(1.40200f); + g = y_fixed + (cr*-stbi__float2fixed(0.71414f)) + ((cb*-stbi__float2fixed(0.34414f)) & 0xffff0000); + b = y_fixed + cb* stbi__float2fixed(1.77200f); + r >>= 20; + g >>= 20; + b >>= 20; + if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; } + if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; } + if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; } + out[0] = (stbi_uc)r; + out[1] = (stbi_uc)g; + out[2] = (stbi_uc)b; + out[3] = 255; + out += step; + } +} + +#if defined(STBI_SSE2) || defined(STBI_NEON) +static void stbi__YCbCr_to_RGB_simd(stbi_uc *out, stbi_uc const *y, stbi_uc const *pcb, stbi_uc const *pcr, int count, int step) +{ + int i = 0; + +#ifdef STBI_SSE2 + // step == 3 is pretty ugly on the final interleave, and i'm not convinced + // it's useful in practice (you wouldn't use it for textures, for example). + // so just accelerate step == 4 case. + if (step == 4) { + // this is a fairly straightforward implementation and not super-optimized. + __m128i signflip = _mm_set1_epi8(-0x80); + __m128i cr_const0 = _mm_set1_epi16( (short) ( 1.40200f*4096.0f+0.5f)); + __m128i cr_const1 = _mm_set1_epi16( - (short) ( 0.71414f*4096.0f+0.5f)); + __m128i cb_const0 = _mm_set1_epi16( - (short) ( 0.34414f*4096.0f+0.5f)); + __m128i cb_const1 = _mm_set1_epi16( (short) ( 1.77200f*4096.0f+0.5f)); + __m128i y_bias = _mm_set1_epi8((char) (unsigned char) 128); + __m128i xw = _mm_set1_epi16(255); // alpha channel + + for (; i+7 < count; i += 8) { + // load + __m128i y_bytes = _mm_loadl_epi64((__m128i *) (y+i)); + __m128i cr_bytes = _mm_loadl_epi64((__m128i *) (pcr+i)); + __m128i cb_bytes = _mm_loadl_epi64((__m128i *) (pcb+i)); + __m128i cr_biased = _mm_xor_si128(cr_bytes, signflip); // -128 + __m128i cb_biased = _mm_xor_si128(cb_bytes, signflip); // -128 + + // unpack to short (and left-shift cr, cb by 8) + __m128i yw = _mm_unpacklo_epi8(y_bias, y_bytes); + __m128i crw = _mm_unpacklo_epi8(_mm_setzero_si128(), cr_biased); + __m128i cbw = _mm_unpacklo_epi8(_mm_setzero_si128(), cb_biased); + + // color transform + __m128i yws = _mm_srli_epi16(yw, 4); + __m128i cr0 = _mm_mulhi_epi16(cr_const0, crw); + __m128i cb0 = _mm_mulhi_epi16(cb_const0, cbw); + __m128i cb1 = _mm_mulhi_epi16(cbw, cb_const1); + __m128i cr1 = _mm_mulhi_epi16(crw, cr_const1); + __m128i rws = _mm_add_epi16(cr0, yws); + __m128i gwt = _mm_add_epi16(cb0, yws); + __m128i bws = _mm_add_epi16(yws, cb1); + __m128i gws = _mm_add_epi16(gwt, cr1); + + // descale + __m128i rw = _mm_srai_epi16(rws, 4); + __m128i bw = _mm_srai_epi16(bws, 4); + __m128i gw = _mm_srai_epi16(gws, 4); + + // back to byte, set up for transpose + __m128i brb = _mm_packus_epi16(rw, bw); + __m128i gxb = _mm_packus_epi16(gw, xw); + + // transpose to interleave channels + __m128i t0 = _mm_unpacklo_epi8(brb, gxb); + __m128i t1 = _mm_unpackhi_epi8(brb, gxb); + __m128i o0 = _mm_unpacklo_epi16(t0, t1); + __m128i o1 = _mm_unpackhi_epi16(t0, t1); + + // store + _mm_storeu_si128((__m128i *) (out + 0), o0); + _mm_storeu_si128((__m128i *) (out + 16), o1); + out += 32; + } + } +#endif + +#ifdef STBI_NEON + // in this version, step=3 support would be easy to add. but is there demand? + if (step == 4) { + // this is a fairly straightforward implementation and not super-optimized. + uint8x8_t signflip = vdup_n_u8(0x80); + int16x8_t cr_const0 = vdupq_n_s16( (short) ( 1.40200f*4096.0f+0.5f)); + int16x8_t cr_const1 = vdupq_n_s16( - (short) ( 0.71414f*4096.0f+0.5f)); + int16x8_t cb_const0 = vdupq_n_s16( - (short) ( 0.34414f*4096.0f+0.5f)); + int16x8_t cb_const1 = vdupq_n_s16( (short) ( 1.77200f*4096.0f+0.5f)); + + for (; i+7 < count; i += 8) { + // load + uint8x8_t y_bytes = vld1_u8(y + i); + uint8x8_t cr_bytes = vld1_u8(pcr + i); + uint8x8_t cb_bytes = vld1_u8(pcb + i); + int8x8_t cr_biased = vreinterpret_s8_u8(vsub_u8(cr_bytes, signflip)); + int8x8_t cb_biased = vreinterpret_s8_u8(vsub_u8(cb_bytes, signflip)); + + // expand to s16 + int16x8_t yws = vreinterpretq_s16_u16(vshll_n_u8(y_bytes, 4)); + int16x8_t crw = vshll_n_s8(cr_biased, 7); + int16x8_t cbw = vshll_n_s8(cb_biased, 7); + + // color transform + int16x8_t cr0 = vqdmulhq_s16(crw, cr_const0); + int16x8_t cb0 = vqdmulhq_s16(cbw, cb_const0); + int16x8_t cr1 = vqdmulhq_s16(crw, cr_const1); + int16x8_t cb1 = vqdmulhq_s16(cbw, cb_const1); + int16x8_t rws = vaddq_s16(yws, cr0); + int16x8_t gws = vaddq_s16(vaddq_s16(yws, cb0), cr1); + int16x8_t bws = vaddq_s16(yws, cb1); + + // undo scaling, round, convert to byte + uint8x8x4_t o; + o.val[0] = vqrshrun_n_s16(rws, 4); + o.val[1] = vqrshrun_n_s16(gws, 4); + o.val[2] = vqrshrun_n_s16(bws, 4); + o.val[3] = vdup_n_u8(255); + + // store, interleaving r/g/b/a + vst4_u8(out, o); + out += 8*4; + } + } +#endif + + for (; i < count; ++i) { + int y_fixed = (y[i] << 20) + (1<<19); // rounding + int r,g,b; + int cr = pcr[i] - 128; + int cb = pcb[i] - 128; + r = y_fixed + cr* stbi__float2fixed(1.40200f); + g = y_fixed + cr*-stbi__float2fixed(0.71414f) + ((cb*-stbi__float2fixed(0.34414f)) & 0xffff0000); + b = y_fixed + cb* stbi__float2fixed(1.77200f); + r >>= 20; + g >>= 20; + b >>= 20; + if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; } + if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; } + if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; } + out[0] = (stbi_uc)r; + out[1] = (stbi_uc)g; + out[2] = (stbi_uc)b; + out[3] = 255; + out += step; + } +} +#endif + +// set up the kernels +static void stbi__setup_jpeg(stbi__jpeg *j) +{ + j->idct_block_kernel = stbi__idct_block; + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_row; + j->resample_row_hv_2_kernel = stbi__resample_row_hv_2; + +#ifdef STBI_SSE2 + if (stbi__sse2_available()) { + j->idct_block_kernel = stbi__idct_simd; + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd; + j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd; + } +#endif + +#ifdef STBI_NEON + j->idct_block_kernel = stbi__idct_simd; + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd; + j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd; +#endif +} + +// clean up the temporary component buffers +static void stbi__cleanup_jpeg(stbi__jpeg *j) +{ + stbi__free_jpeg_components(j, j->s->img_n, 0); +} + +typedef struct +{ + resample_row_func resample; + stbi_uc *line0,*line1; + int hs,vs; // expansion factor in each axis + int w_lores; // horizontal pixels pre-expansion + int ystep; // how far through vertical expansion we are + int ypos; // which pre-expansion row we're on +} stbi__resample; + +// fast 0..255 * 0..255 => 0..255 rounded multiplication +static stbi_uc stbi__blinn_8x8(stbi_uc x, stbi_uc y) +{ + unsigned int t = x*y + 128; + return (stbi_uc) ((t + (t >>8)) >> 8); +} + +static stbi_uc *load_jpeg_image(stbi__jpeg *z, int *out_x, int *out_y, int *comp, int req_comp) +{ + int n, decode_n, is_rgb; + z->s->img_n = 0; // make stbi__cleanup_jpeg safe + + // validate req_comp + if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error"); + + // load a jpeg image from whichever source, but leave in YCbCr format + if (!stbi__decode_jpeg_image(z)) { stbi__cleanup_jpeg(z); return NULL; } + + // determine actual number of components to generate + n = req_comp ? req_comp : z->s->img_n >= 3 ? 3 : 1; + + is_rgb = z->s->img_n == 3 && (z->rgb == 3 || (z->app14_color_transform == 0 && !z->jfif)); + + if (z->s->img_n == 3 && n < 3 && !is_rgb) + decode_n = 1; + else + decode_n = z->s->img_n; + + // nothing to do if no components requested; check this now to avoid + // accessing uninitialized coutput[0] later + if (decode_n <= 0) { stbi__cleanup_jpeg(z); return NULL; } + + // resample and color-convert + { + int k; + unsigned int i,j; + stbi_uc *output; + stbi_uc *coutput[4] = { NULL, NULL, NULL, NULL }; + + stbi__resample res_comp[4]; + + for (k=0; k < decode_n; ++k) { + stbi__resample *r = &res_comp[k]; + + // allocate line buffer big enough for upsampling off the edges + // with upsample factor of 4 + z->img_comp[k].linebuf = (stbi_uc *) stbi__malloc(z->s->img_x + 3); + if (!z->img_comp[k].linebuf) { stbi__cleanup_jpeg(z); return stbi__errpuc("outofmem", "Out of memory"); } + + r->hs = z->img_h_max / z->img_comp[k].h; + r->vs = z->img_v_max / z->img_comp[k].v; + r->ystep = r->vs >> 1; + r->w_lores = (z->s->img_x + r->hs-1) / r->hs; + r->ypos = 0; + r->line0 = r->line1 = z->img_comp[k].data; + + if (r->hs == 1 && r->vs == 1) r->resample = resample_row_1; + else if (r->hs == 1 && r->vs == 2) r->resample = stbi__resample_row_v_2; + else if (r->hs == 2 && r->vs == 1) r->resample = stbi__resample_row_h_2; + else if (r->hs == 2 && r->vs == 2) r->resample = z->resample_row_hv_2_kernel; + else r->resample = stbi__resample_row_generic; + } + + // can't error after this so, this is safe + output = (stbi_uc *) stbi__malloc_mad3(n, z->s->img_x, z->s->img_y, 1); + if (!output) { stbi__cleanup_jpeg(z); return stbi__errpuc("outofmem", "Out of memory"); } + + // now go ahead and resample + for (j=0; j < z->s->img_y; ++j) { + stbi_uc *out = output + n * z->s->img_x * j; + for (k=0; k < decode_n; ++k) { + stbi__resample *r = &res_comp[k]; + int y_bot = r->ystep >= (r->vs >> 1); + coutput[k] = r->resample(z->img_comp[k].linebuf, + y_bot ? r->line1 : r->line0, + y_bot ? r->line0 : r->line1, + r->w_lores, r->hs); + if (++r->ystep >= r->vs) { + r->ystep = 0; + r->line0 = r->line1; + if (++r->ypos < z->img_comp[k].y) + r->line1 += z->img_comp[k].w2; + } + } + if (n >= 3) { + stbi_uc *y = coutput[0]; + if (z->s->img_n == 3) { + if (is_rgb) { + for (i=0; i < z->s->img_x; ++i) { + out[0] = y[i]; + out[1] = coutput[1][i]; + out[2] = coutput[2][i]; + out[3] = 255; + out += n; + } + } else { + z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n); + } + } else if (z->s->img_n == 4) { + if (z->app14_color_transform == 0) { // CMYK + for (i=0; i < z->s->img_x; ++i) { + stbi_uc m = coutput[3][i]; + out[0] = stbi__blinn_8x8(coutput[0][i], m); + out[1] = stbi__blinn_8x8(coutput[1][i], m); + out[2] = stbi__blinn_8x8(coutput[2][i], m); + out[3] = 255; + out += n; + } + } else if (z->app14_color_transform == 2) { // YCCK + z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n); + for (i=0; i < z->s->img_x; ++i) { + stbi_uc m = coutput[3][i]; + out[0] = stbi__blinn_8x8(255 - out[0], m); + out[1] = stbi__blinn_8x8(255 - out[1], m); + out[2] = stbi__blinn_8x8(255 - out[2], m); + out += n; + } + } else { // YCbCr + alpha? Ignore the fourth channel for now + z->YCbCr_to_RGB_kernel(out, y, coutput[1], coutput[2], z->s->img_x, n); + } + } else + for (i=0; i < z->s->img_x; ++i) { + out[0] = out[1] = out[2] = y[i]; + out[3] = 255; // not used if n==3 + out += n; + } + } else { + if (is_rgb) { + if (n == 1) + for (i=0; i < z->s->img_x; ++i) + *out++ = stbi__compute_y(coutput[0][i], coutput[1][i], coutput[2][i]); + else { + for (i=0; i < z->s->img_x; ++i, out += 2) { + out[0] = stbi__compute_y(coutput[0][i], coutput[1][i], coutput[2][i]); + out[1] = 255; + } + } + } else if (z->s->img_n == 4 && z->app14_color_transform == 0) { + for (i=0; i < z->s->img_x; ++i) { + stbi_uc m = coutput[3][i]; + stbi_uc r = stbi__blinn_8x8(coutput[0][i], m); + stbi_uc g = stbi__blinn_8x8(coutput[1][i], m); + stbi_uc b = stbi__blinn_8x8(coutput[2][i], m); + out[0] = stbi__compute_y(r, g, b); + out[1] = 255; + out += n; + } + } else if (z->s->img_n == 4 && z->app14_color_transform == 2) { + for (i=0; i < z->s->img_x; ++i) { + out[0] = stbi__blinn_8x8(255 - coutput[0][i], coutput[3][i]); + out[1] = 255; + out += n; + } + } else { + stbi_uc *y = coutput[0]; + if (n == 1) + for (i=0; i < z->s->img_x; ++i) out[i] = y[i]; + else + for (i=0; i < z->s->img_x; ++i) { *out++ = y[i]; *out++ = 255; } + } + } + } + stbi__cleanup_jpeg(z); + *out_x = z->s->img_x; + *out_y = z->s->img_y; + if (comp) *comp = z->s->img_n >= 3 ? 3 : 1; // report original components, not output + return output; + } +} + +static void *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + unsigned char* result; + stbi__jpeg* j = (stbi__jpeg*) stbi__malloc(sizeof(stbi__jpeg)); + if (!j) return stbi__errpuc("outofmem", "Out of memory"); + memset(j, 0, sizeof(stbi__jpeg)); + STBI_NOTUSED(ri); + j->s = s; + stbi__setup_jpeg(j); + result = load_jpeg_image(j, x,y,comp,req_comp); + STBI_FREE(j); + return result; +} + +static int stbi__jpeg_test(stbi__context *s) +{ + int r; + stbi__jpeg* j = (stbi__jpeg*)stbi__malloc(sizeof(stbi__jpeg)); + if (!j) return stbi__err("outofmem", "Out of memory"); + memset(j, 0, sizeof(stbi__jpeg)); + j->s = s; + stbi__setup_jpeg(j); + r = stbi__decode_jpeg_header(j, STBI__SCAN_type); + stbi__rewind(s); + STBI_FREE(j); + return r; +} + +static int stbi__jpeg_info_raw(stbi__jpeg *j, int *x, int *y, int *comp) +{ + if (!stbi__decode_jpeg_header(j, STBI__SCAN_header)) { + stbi__rewind( j->s ); + return 0; + } + if (x) *x = j->s->img_x; + if (y) *y = j->s->img_y; + if (comp) *comp = j->s->img_n >= 3 ? 3 : 1; + return 1; +} + +static int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp) +{ + int result; + stbi__jpeg* j = (stbi__jpeg*) (stbi__malloc(sizeof(stbi__jpeg))); + if (!j) return stbi__err("outofmem", "Out of memory"); + memset(j, 0, sizeof(stbi__jpeg)); + j->s = s; + result = stbi__jpeg_info_raw(j, x, y, comp); + STBI_FREE(j); + return result; +} +#endif + +// public domain zlib decode v0.2 Sean Barrett 2006-11-18 +// simple implementation +// - all input must be provided in an upfront buffer +// - all output is written to a single output buffer (can malloc/realloc) +// performance +// - fast huffman + +#ifndef STBI_NO_ZLIB + +// fast-way is faster to check than jpeg huffman, but slow way is slower +#define STBI__ZFAST_BITS 9 // accelerate all cases in default tables +#define STBI__ZFAST_MASK ((1 << STBI__ZFAST_BITS) - 1) +#define STBI__ZNSYMS 288 // number of symbols in literal/length alphabet + +// zlib-style huffman encoding +// (jpegs packs from left, zlib from right, so can't share code) +typedef struct +{ + stbi__uint16 fast[1 << STBI__ZFAST_BITS]; + stbi__uint16 firstcode[16]; + int maxcode[17]; + stbi__uint16 firstsymbol[16]; + stbi_uc size[STBI__ZNSYMS]; + stbi__uint16 value[STBI__ZNSYMS]; +} stbi__zhuffman; + +stbi_inline static int stbi__bitreverse16(int n) +{ + n = ((n & 0xAAAA) >> 1) | ((n & 0x5555) << 1); + n = ((n & 0xCCCC) >> 2) | ((n & 0x3333) << 2); + n = ((n & 0xF0F0) >> 4) | ((n & 0x0F0F) << 4); + n = ((n & 0xFF00) >> 8) | ((n & 0x00FF) << 8); + return n; +} + +stbi_inline static int stbi__bit_reverse(int v, int bits) +{ + STBI_ASSERT(bits <= 16); + // to bit reverse n bits, reverse 16 and shift + // e.g. 11 bits, bit reverse and shift away 5 + return stbi__bitreverse16(v) >> (16-bits); +} + +static int stbi__zbuild_huffman(stbi__zhuffman *z, const stbi_uc *sizelist, int num) +{ + int i,k=0; + int code, next_code[16], sizes[17]; + + // DEFLATE spec for generating codes + memset(sizes, 0, sizeof(sizes)); + memset(z->fast, 0, sizeof(z->fast)); + for (i=0; i < num; ++i) + ++sizes[sizelist[i]]; + sizes[0] = 0; + for (i=1; i < 16; ++i) + if (sizes[i] > (1 << i)) + return stbi__err("bad sizes", "Corrupt PNG"); + code = 0; + for (i=1; i < 16; ++i) { + next_code[i] = code; + z->firstcode[i] = (stbi__uint16) code; + z->firstsymbol[i] = (stbi__uint16) k; + code = (code + sizes[i]); + if (sizes[i]) + if (code-1 >= (1 << i)) return stbi__err("bad codelengths","Corrupt PNG"); + z->maxcode[i] = code << (16-i); // preshift for inner loop + code <<= 1; + k += sizes[i]; + } + z->maxcode[16] = 0x10000; // sentinel + for (i=0; i < num; ++i) { + int s = sizelist[i]; + if (s) { + int c = next_code[s] - z->firstcode[s] + z->firstsymbol[s]; + stbi__uint16 fastv = (stbi__uint16) ((s << 9) | i); + z->size [c] = (stbi_uc ) s; + z->value[c] = (stbi__uint16) i; + if (s <= STBI__ZFAST_BITS) { + int j = stbi__bit_reverse(next_code[s],s); + while (j < (1 << STBI__ZFAST_BITS)) { + z->fast[j] = fastv; + j += (1 << s); + } + } + ++next_code[s]; + } + } + return 1; +} + +// zlib-from-memory implementation for PNG reading +// because PNG allows splitting the zlib stream arbitrarily, +// and it's annoying structurally to have PNG call ZLIB call PNG, +// we require PNG read all the IDATs and combine them into a single +// memory buffer + +typedef struct +{ + stbi_uc *zbuffer, *zbuffer_end; + int num_bits; + int hit_zeof_once; + stbi__uint32 code_buffer; + + char *zout; + char *zout_start; + char *zout_end; + int z_expandable; + + stbi__zhuffman z_length, z_distance; +} stbi__zbuf; + +stbi_inline static int stbi__zeof(stbi__zbuf *z) +{ + return (z->zbuffer >= z->zbuffer_end); +} + +stbi_inline static stbi_uc stbi__zget8(stbi__zbuf *z) +{ + return stbi__zeof(z) ? 0 : *z->zbuffer++; +} + +static void stbi__fill_bits(stbi__zbuf *z) +{ + do { + if (z->code_buffer >= (1U << z->num_bits)) { + z->zbuffer = z->zbuffer_end; /* treat this as EOF so we fail. */ + return; + } + z->code_buffer |= (unsigned int) stbi__zget8(z) << z->num_bits; + z->num_bits += 8; + } while (z->num_bits <= 24); +} + +stbi_inline static unsigned int stbi__zreceive(stbi__zbuf *z, int n) +{ + unsigned int k; + if (z->num_bits < n) stbi__fill_bits(z); + k = z->code_buffer & ((1 << n) - 1); + z->code_buffer >>= n; + z->num_bits -= n; + return k; +} + +static int stbi__zhuffman_decode_slowpath(stbi__zbuf *a, stbi__zhuffman *z) +{ + int b,s,k; + // not resolved by fast table, so compute it the slow way + // use jpeg approach, which requires MSbits at top + k = stbi__bit_reverse(a->code_buffer, 16); + for (s=STBI__ZFAST_BITS+1; ; ++s) + if (k < z->maxcode[s]) + break; + if (s >= 16) return -1; // invalid code! + // code size is s, so: + b = (k >> (16-s)) - z->firstcode[s] + z->firstsymbol[s]; + if (b >= STBI__ZNSYMS) return -1; // some data was corrupt somewhere! + if (z->size[b] != s) return -1; // was originally an assert, but report failure instead. + a->code_buffer >>= s; + a->num_bits -= s; + return z->value[b]; +} + +stbi_inline static int stbi__zhuffman_decode(stbi__zbuf *a, stbi__zhuffman *z) +{ + int b,s; + if (a->num_bits < 16) { + if (stbi__zeof(a)) { + if (!a->hit_zeof_once) { + // This is the first time we hit eof, insert 16 extra padding btis + // to allow us to keep going; if we actually consume any of them + // though, that is invalid data. This is caught later. + a->hit_zeof_once = 1; + a->num_bits += 16; // add 16 implicit zero bits + } else { + // We already inserted our extra 16 padding bits and are again + // out, this stream is actually prematurely terminated. + return -1; + } + } else { + stbi__fill_bits(a); + } + } + b = z->fast[a->code_buffer & STBI__ZFAST_MASK]; + if (b) { + s = b >> 9; + a->code_buffer >>= s; + a->num_bits -= s; + return b & 511; + } + return stbi__zhuffman_decode_slowpath(a, z); +} + +static int stbi__zexpand(stbi__zbuf *z, char *zout, int n) // need to make room for n bytes +{ + char *q; + unsigned int cur, limit, old_limit; + z->zout = zout; + if (!z->z_expandable) return stbi__err("output buffer limit","Corrupt PNG"); + cur = (unsigned int) (z->zout - z->zout_start); + limit = old_limit = (unsigned) (z->zout_end - z->zout_start); + if (UINT_MAX - cur < (unsigned) n) return stbi__err("outofmem", "Out of memory"); + while (cur + n > limit) { + if(limit > UINT_MAX / 2) return stbi__err("outofmem", "Out of memory"); + limit *= 2; + } + q = (char *) STBI_REALLOC_SIZED(z->zout_start, old_limit, limit); + STBI_NOTUSED(old_limit); + if (q == NULL) return stbi__err("outofmem", "Out of memory"); + z->zout_start = q; + z->zout = q + cur; + z->zout_end = q + limit; + return 1; +} + +static const int stbi__zlength_base[31] = { + 3,4,5,6,7,8,9,10,11,13, + 15,17,19,23,27,31,35,43,51,59, + 67,83,99,115,131,163,195,227,258,0,0 }; + +static const int stbi__zlength_extra[31]= +{ 0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0,0,0 }; + +static const int stbi__zdist_base[32] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193, +257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577,0,0}; + +static const int stbi__zdist_extra[32] = +{ 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13}; + +static int stbi__parse_huffman_block(stbi__zbuf *a) +{ + char *zout = a->zout; + for(;;) { + int z = stbi__zhuffman_decode(a, &a->z_length); + if (z < 256) { + if (z < 0) return stbi__err("bad huffman code","Corrupt PNG"); // error in huffman codes + if (zout >= a->zout_end) { + if (!stbi__zexpand(a, zout, 1)) return 0; + zout = a->zout; + } + *zout++ = (char) z; + } else { + stbi_uc *p; + int len,dist; + if (z == 256) { + a->zout = zout; + if (a->hit_zeof_once && a->num_bits < 16) { + // The first time we hit zeof, we inserted 16 extra zero bits into our bit + // buffer so the decoder can just do its speculative decoding. But if we + // actually consumed any of those bits (which is the case when num_bits < 16), + // the stream actually read past the end so it is malformed. + return stbi__err("unexpected end","Corrupt PNG"); + } + return 1; + } + if (z >= 286) return stbi__err("bad huffman code","Corrupt PNG"); // per DEFLATE, length codes 286 and 287 must not appear in compressed data + z -= 257; + len = stbi__zlength_base[z]; + if (stbi__zlength_extra[z]) len += stbi__zreceive(a, stbi__zlength_extra[z]); + z = stbi__zhuffman_decode(a, &a->z_distance); + if (z < 0 || z >= 30) return stbi__err("bad huffman code","Corrupt PNG"); // per DEFLATE, distance codes 30 and 31 must not appear in compressed data + dist = stbi__zdist_base[z]; + if (stbi__zdist_extra[z]) dist += stbi__zreceive(a, stbi__zdist_extra[z]); + if (zout - a->zout_start < dist) return stbi__err("bad dist","Corrupt PNG"); + if (len > a->zout_end - zout) { + if (!stbi__zexpand(a, zout, len)) return 0; + zout = a->zout; + } + p = (stbi_uc *) (zout - dist); + if (dist == 1) { // run of one byte; common in images. + stbi_uc v = *p; + if (len) { do *zout++ = v; while (--len); } + } else { + if (len) { do *zout++ = *p++; while (--len); } + } + } + } +} + +static int stbi__compute_huffman_codes(stbi__zbuf *a) +{ + static const stbi_uc length_dezigzag[19] = { 16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15 }; + stbi__zhuffman z_codelength; + stbi_uc lencodes[286+32+137];//padding for maximum single op + stbi_uc codelength_sizes[19]; + int i,n; + + int hlit = stbi__zreceive(a,5) + 257; + int hdist = stbi__zreceive(a,5) + 1; + int hclen = stbi__zreceive(a,4) + 4; + int ntot = hlit + hdist; + + memset(codelength_sizes, 0, sizeof(codelength_sizes)); + for (i=0; i < hclen; ++i) { + int s = stbi__zreceive(a,3); + codelength_sizes[length_dezigzag[i]] = (stbi_uc) s; + } + if (!stbi__zbuild_huffman(&z_codelength, codelength_sizes, 19)) return 0; + + n = 0; + while (n < ntot) { + int c = stbi__zhuffman_decode(a, &z_codelength); + if (c < 0 || c >= 19) return stbi__err("bad codelengths", "Corrupt PNG"); + if (c < 16) + lencodes[n++] = (stbi_uc) c; + else { + stbi_uc fill = 0; + if (c == 16) { + c = stbi__zreceive(a,2)+3; + if (n == 0) return stbi__err("bad codelengths", "Corrupt PNG"); + fill = lencodes[n-1]; + } else if (c == 17) { + c = stbi__zreceive(a,3)+3; + } else if (c == 18) { + c = stbi__zreceive(a,7)+11; + } else { + return stbi__err("bad codelengths", "Corrupt PNG"); + } + if (ntot - n < c) return stbi__err("bad codelengths", "Corrupt PNG"); + memset(lencodes+n, fill, c); + n += c; + } + } + if (n != ntot) return stbi__err("bad codelengths","Corrupt PNG"); + if (!stbi__zbuild_huffman(&a->z_length, lencodes, hlit)) return 0; + if (!stbi__zbuild_huffman(&a->z_distance, lencodes+hlit, hdist)) return 0; + return 1; +} + +static int stbi__parse_uncompressed_block(stbi__zbuf *a) +{ + stbi_uc header[4]; + int len,nlen,k; + if (a->num_bits & 7) + stbi__zreceive(a, a->num_bits & 7); // discard + // drain the bit-packed data into header + k = 0; + while (a->num_bits > 0) { + header[k++] = (stbi_uc) (a->code_buffer & 255); // suppress MSVC run-time check + a->code_buffer >>= 8; + a->num_bits -= 8; + } + if (a->num_bits < 0) return stbi__err("zlib corrupt","Corrupt PNG"); + // now fill header the normal way + while (k < 4) + header[k++] = stbi__zget8(a); + len = header[1] * 256 + header[0]; + nlen = header[3] * 256 + header[2]; + if (nlen != (len ^ 0xffff)) return stbi__err("zlib corrupt","Corrupt PNG"); + if (a->zbuffer + len > a->zbuffer_end) return stbi__err("read past buffer","Corrupt PNG"); + if (a->zout + len > a->zout_end) + if (!stbi__zexpand(a, a->zout, len)) return 0; + memcpy(a->zout, a->zbuffer, len); + a->zbuffer += len; + a->zout += len; + return 1; +} + +static int stbi__parse_zlib_header(stbi__zbuf *a) +{ + int cmf = stbi__zget8(a); + int cm = cmf & 15; + /* int cinfo = cmf >> 4; */ + int flg = stbi__zget8(a); + if (stbi__zeof(a)) return stbi__err("bad zlib header","Corrupt PNG"); // zlib spec + if ((cmf*256+flg) % 31 != 0) return stbi__err("bad zlib header","Corrupt PNG"); // zlib spec + if (flg & 32) return stbi__err("no preset dict","Corrupt PNG"); // preset dictionary not allowed in png + if (cm != 8) return stbi__err("bad compression","Corrupt PNG"); // DEFLATE required for png + // window = 1 << (8 + cinfo)... but who cares, we fully buffer output + return 1; +} + +static const stbi_uc stbi__zdefault_length[STBI__ZNSYMS] = +{ + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, + 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, 7,7,7,7,7,7,7,7,8,8,8,8,8,8,8,8 +}; +static const stbi_uc stbi__zdefault_distance[32] = +{ + 5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5 +}; +/* +Init algorithm: +{ + int i; // use <= to match clearly with spec + for (i=0; i <= 143; ++i) stbi__zdefault_length[i] = 8; + for ( ; i <= 255; ++i) stbi__zdefault_length[i] = 9; + for ( ; i <= 279; ++i) stbi__zdefault_length[i] = 7; + for ( ; i <= 287; ++i) stbi__zdefault_length[i] = 8; + + for (i=0; i <= 31; ++i) stbi__zdefault_distance[i] = 5; +} +*/ + +static int stbi__parse_zlib(stbi__zbuf *a, int parse_header) +{ + int final, type; + if (parse_header) + if (!stbi__parse_zlib_header(a)) return 0; + a->num_bits = 0; + a->code_buffer = 0; + a->hit_zeof_once = 0; + do { + final = stbi__zreceive(a,1); + type = stbi__zreceive(a,2); + if (type == 0) { + if (!stbi__parse_uncompressed_block(a)) return 0; + } else if (type == 3) { + return 0; + } else { + if (type == 1) { + // use fixed code lengths + if (!stbi__zbuild_huffman(&a->z_length , stbi__zdefault_length , STBI__ZNSYMS)) return 0; + if (!stbi__zbuild_huffman(&a->z_distance, stbi__zdefault_distance, 32)) return 0; + } else { + if (!stbi__compute_huffman_codes(a)) return 0; + } + if (!stbi__parse_huffman_block(a)) return 0; + } + } while (!final); + return 1; +} + +static int stbi__do_zlib(stbi__zbuf *a, char *obuf, int olen, int exp, int parse_header) +{ + a->zout_start = obuf; + a->zout = obuf; + a->zout_end = obuf + olen; + a->z_expandable = exp; + + return stbi__parse_zlib(a, parse_header); +} + +STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen) +{ + stbi__zbuf a; + char *p = (char *) stbi__malloc(initial_size); + if (p == NULL) return NULL; + a.zbuffer = (stbi_uc *) buffer; + a.zbuffer_end = (stbi_uc *) buffer + len; + if (stbi__do_zlib(&a, p, initial_size, 1, 1)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + STBI_FREE(a.zout_start); + return NULL; + } +} + +STBIDEF char *stbi_zlib_decode_malloc(char const *buffer, int len, int *outlen) +{ + return stbi_zlib_decode_malloc_guesssize(buffer, len, 16384, outlen); +} + +STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header) +{ + stbi__zbuf a; + char *p = (char *) stbi__malloc(initial_size); + if (p == NULL) return NULL; + a.zbuffer = (stbi_uc *) buffer; + a.zbuffer_end = (stbi_uc *) buffer + len; + if (stbi__do_zlib(&a, p, initial_size, 1, parse_header)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + STBI_FREE(a.zout_start); + return NULL; + } +} + +STBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, char const *ibuffer, int ilen) +{ + stbi__zbuf a; + a.zbuffer = (stbi_uc *) ibuffer; + a.zbuffer_end = (stbi_uc *) ibuffer + ilen; + if (stbi__do_zlib(&a, obuffer, olen, 0, 1)) + return (int) (a.zout - a.zout_start); + else + return -1; +} + +STBIDEF char *stbi_zlib_decode_noheader_malloc(char const *buffer, int len, int *outlen) +{ + stbi__zbuf a; + char *p = (char *) stbi__malloc(16384); + if (p == NULL) return NULL; + a.zbuffer = (stbi_uc *) buffer; + a.zbuffer_end = (stbi_uc *) buffer+len; + if (stbi__do_zlib(&a, p, 16384, 1, 0)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + STBI_FREE(a.zout_start); + return NULL; + } +} + +STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen) +{ + stbi__zbuf a; + a.zbuffer = (stbi_uc *) ibuffer; + a.zbuffer_end = (stbi_uc *) ibuffer + ilen; + if (stbi__do_zlib(&a, obuffer, olen, 0, 0)) + return (int) (a.zout - a.zout_start); + else + return -1; +} +#endif + +// public domain "baseline" PNG decoder v0.10 Sean Barrett 2006-11-18 +// simple implementation +// - only 8-bit samples +// - no CRC checking +// - allocates lots of intermediate memory +// - avoids problem of streaming data between subsystems +// - avoids explicit window management +// performance +// - uses stb_zlib, a PD zlib implementation with fast huffman decoding + +#ifndef STBI_NO_PNG +typedef struct +{ + stbi__uint32 length; + stbi__uint32 type; +} stbi__pngchunk; + +static stbi__pngchunk stbi__get_chunk_header(stbi__context *s) +{ + stbi__pngchunk c; + c.length = stbi__get32be(s); + c.type = stbi__get32be(s); + return c; +} + +static int stbi__check_png_header(stbi__context *s) +{ + static const stbi_uc png_sig[8] = { 137,80,78,71,13,10,26,10 }; + int i; + for (i=0; i < 8; ++i) + if (stbi__get8(s) != png_sig[i]) return stbi__err("bad png sig","Not a PNG"); + return 1; +} + +typedef struct +{ + stbi__context *s; + stbi_uc *idata, *expanded, *out; + int depth; +} stbi__png; + + +enum { + STBI__F_none=0, + STBI__F_sub=1, + STBI__F_up=2, + STBI__F_avg=3, + STBI__F_paeth=4, + // synthetic filter used for first scanline to avoid needing a dummy row of 0s + STBI__F_avg_first +}; + +static stbi_uc first_row_filter[5] = +{ + STBI__F_none, + STBI__F_sub, + STBI__F_none, + STBI__F_avg_first, + STBI__F_sub // Paeth with b=c=0 turns out to be equivalent to sub +}; + +static int stbi__paeth(int a, int b, int c) +{ + // This formulation looks very different from the reference in the PNG spec, but is + // actually equivalent and has favorable data dependencies and admits straightforward + // generation of branch-free code, which helps performance significantly. + int thresh = c*3 - (a + b); + int lo = a < b ? a : b; + int hi = a < b ? b : a; + int t0 = (hi <= thresh) ? lo : c; + int t1 = (thresh <= lo) ? hi : t0; + return t1; +} + +static const stbi_uc stbi__depth_scale_table[9] = { 0, 0xff, 0x55, 0, 0x11, 0,0,0, 0x01 }; + +// adds an extra all-255 alpha channel +// dest == src is legal +// img_n must be 1 or 3 +static void stbi__create_png_alpha_expand8(stbi_uc *dest, stbi_uc *src, stbi__uint32 x, int img_n) +{ + int i; + // must process data backwards since we allow dest==src + if (img_n == 1) { + for (i=x-1; i >= 0; --i) { + dest[i*2+1] = 255; + dest[i*2+0] = src[i]; + } + } else { + STBI_ASSERT(img_n == 3); + for (i=x-1; i >= 0; --i) { + dest[i*4+3] = 255; + dest[i*4+2] = src[i*3+2]; + dest[i*4+1] = src[i*3+1]; + dest[i*4+0] = src[i*3+0]; + } + } +} + +// create the png data from post-deflated data +static int stbi__create_png_image_raw(stbi__png *a, stbi_uc *raw, stbi__uint32 raw_len, int out_n, stbi__uint32 x, stbi__uint32 y, int depth, int color) +{ + int bytes = (depth == 16 ? 2 : 1); + stbi__context *s = a->s; + stbi__uint32 i,j,stride = x*out_n*bytes; + stbi__uint32 img_len, img_width_bytes; + stbi_uc *filter_buf; + int all_ok = 1; + int k; + int img_n = s->img_n; // copy it into a local for later + + int output_bytes = out_n*bytes; + int filter_bytes = img_n*bytes; + int width = x; + + STBI_ASSERT(out_n == s->img_n || out_n == s->img_n+1); + a->out = (stbi_uc *) stbi__malloc_mad3(x, y, output_bytes, 0); // extra bytes to write off the end into + if (!a->out) return stbi__err("outofmem", "Out of memory"); + + // note: error exits here don't need to clean up a->out individually, + // stbi__do_png always does on error. + if (!stbi__mad3sizes_valid(img_n, x, depth, 7)) return stbi__err("too large", "Corrupt PNG"); + img_width_bytes = (((img_n * x * depth) + 7) >> 3); + if (!stbi__mad2sizes_valid(img_width_bytes, y, img_width_bytes)) return stbi__err("too large", "Corrupt PNG"); + img_len = (img_width_bytes + 1) * y; + + // we used to check for exact match between raw_len and img_len on non-interlaced PNGs, + // but issue #276 reported a PNG in the wild that had extra data at the end (all zeros), + // so just check for raw_len < img_len always. + if (raw_len < img_len) return stbi__err("not enough pixels","Corrupt PNG"); + + // Allocate two scan lines worth of filter workspace buffer. + filter_buf = (stbi_uc *) stbi__malloc_mad2(img_width_bytes, 2, 0); + if (!filter_buf) return stbi__err("outofmem", "Out of memory"); + + // Filtering for low-bit-depth images + if (depth < 8) { + filter_bytes = 1; + width = img_width_bytes; + } + + for (j=0; j < y; ++j) { + // cur/prior filter buffers alternate + stbi_uc *cur = filter_buf + (j & 1)*img_width_bytes; + stbi_uc *prior = filter_buf + (~j & 1)*img_width_bytes; + stbi_uc *dest = a->out + stride*j; + int nk = width * filter_bytes; + int filter = *raw++; + + // check filter type + if (filter > 4) { + all_ok = stbi__err("invalid filter","Corrupt PNG"); + break; + } + + // if first row, use special filter that doesn't sample previous row + if (j == 0) filter = first_row_filter[filter]; + + // perform actual filtering + switch (filter) { + case STBI__F_none: + memcpy(cur, raw, nk); + break; + case STBI__F_sub: + memcpy(cur, raw, filter_bytes); + for (k = filter_bytes; k < nk; ++k) + cur[k] = STBI__BYTECAST(raw[k] + cur[k-filter_bytes]); + break; + case STBI__F_up: + for (k = 0; k < nk; ++k) + cur[k] = STBI__BYTECAST(raw[k] + prior[k]); + break; + case STBI__F_avg: + for (k = 0; k < filter_bytes; ++k) + cur[k] = STBI__BYTECAST(raw[k] + (prior[k]>>1)); + for (k = filter_bytes; k < nk; ++k) + cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k-filter_bytes])>>1)); + break; + case STBI__F_paeth: + for (k = 0; k < filter_bytes; ++k) + cur[k] = STBI__BYTECAST(raw[k] + prior[k]); // prior[k] == stbi__paeth(0,prior[k],0) + for (k = filter_bytes; k < nk; ++k) + cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes], prior[k], prior[k-filter_bytes])); + break; + case STBI__F_avg_first: + memcpy(cur, raw, filter_bytes); + for (k = filter_bytes; k < nk; ++k) + cur[k] = STBI__BYTECAST(raw[k] + (cur[k-filter_bytes] >> 1)); + break; + } + + raw += nk; + + // expand decoded bits in cur to dest, also adding an extra alpha channel if desired + if (depth < 8) { + stbi_uc scale = (color == 0) ? stbi__depth_scale_table[depth] : 1; // scale grayscale values to 0..255 range + stbi_uc *in = cur; + stbi_uc *out = dest; + stbi_uc inb = 0; + stbi__uint32 nsmp = x*img_n; + + // expand bits to bytes first + if (depth == 4) { + for (i=0; i < nsmp; ++i) { + if ((i & 1) == 0) inb = *in++; + *out++ = scale * (inb >> 4); + inb <<= 4; + } + } else if (depth == 2) { + for (i=0; i < nsmp; ++i) { + if ((i & 3) == 0) inb = *in++; + *out++ = scale * (inb >> 6); + inb <<= 2; + } + } else { + STBI_ASSERT(depth == 1); + for (i=0; i < nsmp; ++i) { + if ((i & 7) == 0) inb = *in++; + *out++ = scale * (inb >> 7); + inb <<= 1; + } + } + + // insert alpha=255 values if desired + if (img_n != out_n) + stbi__create_png_alpha_expand8(dest, dest, x, img_n); + } else if (depth == 8) { + if (img_n == out_n) + memcpy(dest, cur, x*img_n); + else + stbi__create_png_alpha_expand8(dest, cur, x, img_n); + } else if (depth == 16) { + // convert the image data from big-endian to platform-native + stbi__uint16 *dest16 = (stbi__uint16*)dest; + stbi__uint32 nsmp = x*img_n; + + if (img_n == out_n) { + for (i = 0; i < nsmp; ++i, ++dest16, cur += 2) + *dest16 = (cur[0] << 8) | cur[1]; + } else { + STBI_ASSERT(img_n+1 == out_n); + if (img_n == 1) { + for (i = 0; i < x; ++i, dest16 += 2, cur += 2) { + dest16[0] = (cur[0] << 8) | cur[1]; + dest16[1] = 0xffff; + } + } else { + STBI_ASSERT(img_n == 3); + for (i = 0; i < x; ++i, dest16 += 4, cur += 6) { + dest16[0] = (cur[0] << 8) | cur[1]; + dest16[1] = (cur[2] << 8) | cur[3]; + dest16[2] = (cur[4] << 8) | cur[5]; + dest16[3] = 0xffff; + } + } + } + } + } + + STBI_FREE(filter_buf); + if (!all_ok) return 0; + + return 1; +} + +static int stbi__create_png_image(stbi__png *a, stbi_uc *image_data, stbi__uint32 image_data_len, int out_n, int depth, int color, int interlaced) +{ + int bytes = (depth == 16 ? 2 : 1); + int out_bytes = out_n * bytes; + stbi_uc *final; + int p; + if (!interlaced) + return stbi__create_png_image_raw(a, image_data, image_data_len, out_n, a->s->img_x, a->s->img_y, depth, color); + + // de-interlacing + final = (stbi_uc *) stbi__malloc_mad3(a->s->img_x, a->s->img_y, out_bytes, 0); + if (!final) return stbi__err("outofmem", "Out of memory"); + for (p=0; p < 7; ++p) { + int xorig[] = { 0,4,0,2,0,1,0 }; + int yorig[] = { 0,0,4,0,2,0,1 }; + int xspc[] = { 8,8,4,4,2,2,1 }; + int yspc[] = { 8,8,8,4,4,2,2 }; + int i,j,x,y; + // pass1_x[4] = 0, pass1_x[5] = 1, pass1_x[12] = 1 + x = (a->s->img_x - xorig[p] + xspc[p]-1) / xspc[p]; + y = (a->s->img_y - yorig[p] + yspc[p]-1) / yspc[p]; + if (x && y) { + stbi__uint32 img_len = ((((a->s->img_n * x * depth) + 7) >> 3) + 1) * y; + if (!stbi__create_png_image_raw(a, image_data, image_data_len, out_n, x, y, depth, color)) { + STBI_FREE(final); + return 0; + } + for (j=0; j < y; ++j) { + for (i=0; i < x; ++i) { + int out_y = j*yspc[p]+yorig[p]; + int out_x = i*xspc[p]+xorig[p]; + memcpy(final + out_y*a->s->img_x*out_bytes + out_x*out_bytes, + a->out + (j*x+i)*out_bytes, out_bytes); + } + } + STBI_FREE(a->out); + image_data += img_len; + image_data_len -= img_len; + } + } + a->out = final; + + return 1; +} + +static int stbi__compute_transparency(stbi__png *z, stbi_uc tc[3], int out_n) +{ + stbi__context *s = z->s; + stbi__uint32 i, pixel_count = s->img_x * s->img_y; + stbi_uc *p = z->out; + + // compute color-based transparency, assuming we've + // already got 255 as the alpha value in the output + STBI_ASSERT(out_n == 2 || out_n == 4); + + if (out_n == 2) { + for (i=0; i < pixel_count; ++i) { + p[1] = (p[0] == tc[0] ? 0 : 255); + p += 2; + } + } else { + for (i=0; i < pixel_count; ++i) { + if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2]) + p[3] = 0; + p += 4; + } + } + return 1; +} + +static int stbi__compute_transparency16(stbi__png *z, stbi__uint16 tc[3], int out_n) +{ + stbi__context *s = z->s; + stbi__uint32 i, pixel_count = s->img_x * s->img_y; + stbi__uint16 *p = (stbi__uint16*) z->out; + + // compute color-based transparency, assuming we've + // already got 65535 as the alpha value in the output + STBI_ASSERT(out_n == 2 || out_n == 4); + + if (out_n == 2) { + for (i = 0; i < pixel_count; ++i) { + p[1] = (p[0] == tc[0] ? 0 : 65535); + p += 2; + } + } else { + for (i = 0; i < pixel_count; ++i) { + if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2]) + p[3] = 0; + p += 4; + } + } + return 1; +} + +static int stbi__expand_png_palette(stbi__png *a, stbi_uc *palette, int len, int pal_img_n) +{ + stbi__uint32 i, pixel_count = a->s->img_x * a->s->img_y; + stbi_uc *p, *temp_out, *orig = a->out; + + p = (stbi_uc *) stbi__malloc_mad2(pixel_count, pal_img_n, 0); + if (p == NULL) return stbi__err("outofmem", "Out of memory"); + + // between here and free(out) below, exitting would leak + temp_out = p; + + if (pal_img_n == 3) { + for (i=0; i < pixel_count; ++i) { + int n = orig[i]*4; + p[0] = palette[n ]; + p[1] = palette[n+1]; + p[2] = palette[n+2]; + p += 3; + } + } else { + for (i=0; i < pixel_count; ++i) { + int n = orig[i]*4; + p[0] = palette[n ]; + p[1] = palette[n+1]; + p[2] = palette[n+2]; + p[3] = palette[n+3]; + p += 4; + } + } + STBI_FREE(a->out); + a->out = temp_out; + + STBI_NOTUSED(len); + + return 1; +} + +static int stbi__unpremultiply_on_load_global = 0; +static int stbi__de_iphone_flag_global = 0; + +STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply) +{ + stbi__unpremultiply_on_load_global = flag_true_if_should_unpremultiply; +} + +STBIDEF void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert) +{ + stbi__de_iphone_flag_global = flag_true_if_should_convert; +} + +#ifndef STBI_THREAD_LOCAL +#define stbi__unpremultiply_on_load stbi__unpremultiply_on_load_global +#define stbi__de_iphone_flag stbi__de_iphone_flag_global +#else +static STBI_THREAD_LOCAL int stbi__unpremultiply_on_load_local, stbi__unpremultiply_on_load_set; +static STBI_THREAD_LOCAL int stbi__de_iphone_flag_local, stbi__de_iphone_flag_set; + +STBIDEF void stbi_set_unpremultiply_on_load_thread(int flag_true_if_should_unpremultiply) +{ + stbi__unpremultiply_on_load_local = flag_true_if_should_unpremultiply; + stbi__unpremultiply_on_load_set = 1; +} + +STBIDEF void stbi_convert_iphone_png_to_rgb_thread(int flag_true_if_should_convert) +{ + stbi__de_iphone_flag_local = flag_true_if_should_convert; + stbi__de_iphone_flag_set = 1; +} + +#define stbi__unpremultiply_on_load (stbi__unpremultiply_on_load_set \ + ? stbi__unpremultiply_on_load_local \ + : stbi__unpremultiply_on_load_global) +#define stbi__de_iphone_flag (stbi__de_iphone_flag_set \ + ? stbi__de_iphone_flag_local \ + : stbi__de_iphone_flag_global) +#endif // STBI_THREAD_LOCAL + +static void stbi__de_iphone(stbi__png *z) +{ + stbi__context *s = z->s; + stbi__uint32 i, pixel_count = s->img_x * s->img_y; + stbi_uc *p = z->out; + + if (s->img_out_n == 3) { // convert bgr to rgb + for (i=0; i < pixel_count; ++i) { + stbi_uc t = p[0]; + p[0] = p[2]; + p[2] = t; + p += 3; + } + } else { + STBI_ASSERT(s->img_out_n == 4); + if (stbi__unpremultiply_on_load) { + // convert bgr to rgb and unpremultiply + for (i=0; i < pixel_count; ++i) { + stbi_uc a = p[3]; + stbi_uc t = p[0]; + if (a) { + stbi_uc half = a / 2; + p[0] = (p[2] * 255 + half) / a; + p[1] = (p[1] * 255 + half) / a; + p[2] = ( t * 255 + half) / a; + } else { + p[0] = p[2]; + p[2] = t; + } + p += 4; + } + } else { + // convert bgr to rgb + for (i=0; i < pixel_count; ++i) { + stbi_uc t = p[0]; + p[0] = p[2]; + p[2] = t; + p += 4; + } + } + } +} + +#define STBI__PNG_TYPE(a,b,c,d) (((unsigned) (a) << 24) + ((unsigned) (b) << 16) + ((unsigned) (c) << 8) + (unsigned) (d)) + +static int stbi__parse_png_file(stbi__png *z, int scan, int req_comp) +{ + stbi_uc palette[1024], pal_img_n=0; + stbi_uc has_trans=0, tc[3]={0}; + stbi__uint16 tc16[3]; + stbi__uint32 ioff=0, idata_limit=0, i, pal_len=0; + int first=1,k,interlace=0, color=0, is_iphone=0; + stbi__context *s = z->s; + + z->expanded = NULL; + z->idata = NULL; + z->out = NULL; + + if (!stbi__check_png_header(s)) return 0; + + if (scan == STBI__SCAN_type) return 1; + + for (;;) { + stbi__pngchunk c = stbi__get_chunk_header(s); + switch (c.type) { + case STBI__PNG_TYPE('C','g','B','I'): + is_iphone = 1; + stbi__skip(s, c.length); + break; + case STBI__PNG_TYPE('I','H','D','R'): { + int comp,filter; + if (!first) return stbi__err("multiple IHDR","Corrupt PNG"); + first = 0; + if (c.length != 13) return stbi__err("bad IHDR len","Corrupt PNG"); + s->img_x = stbi__get32be(s); + s->img_y = stbi__get32be(s); + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + z->depth = stbi__get8(s); if (z->depth != 1 && z->depth != 2 && z->depth != 4 && z->depth != 8 && z->depth != 16) return stbi__err("1/2/4/8/16-bit only","PNG not supported: 1/2/4/8/16-bit only"); + color = stbi__get8(s); if (color > 6) return stbi__err("bad ctype","Corrupt PNG"); + if (color == 3 && z->depth == 16) return stbi__err("bad ctype","Corrupt PNG"); + if (color == 3) pal_img_n = 3; else if (color & 1) return stbi__err("bad ctype","Corrupt PNG"); + comp = stbi__get8(s); if (comp) return stbi__err("bad comp method","Corrupt PNG"); + filter= stbi__get8(s); if (filter) return stbi__err("bad filter method","Corrupt PNG"); + interlace = stbi__get8(s); if (interlace>1) return stbi__err("bad interlace method","Corrupt PNG"); + if (!s->img_x || !s->img_y) return stbi__err("0-pixel image","Corrupt PNG"); + if (!pal_img_n) { + s->img_n = (color & 2 ? 3 : 1) + (color & 4 ? 1 : 0); + if ((1 << 30) / s->img_x / s->img_n < s->img_y) return stbi__err("too large", "Image too large to decode"); + } else { + // if paletted, then pal_n is our final components, and + // img_n is # components to decompress/filter. + s->img_n = 1; + if ((1 << 30) / s->img_x / 4 < s->img_y) return stbi__err("too large","Corrupt PNG"); + } + // even with SCAN_header, have to scan to see if we have a tRNS + break; + } + + case STBI__PNG_TYPE('P','L','T','E'): { + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (c.length > 256*3) return stbi__err("invalid PLTE","Corrupt PNG"); + pal_len = c.length / 3; + if (pal_len * 3 != c.length) return stbi__err("invalid PLTE","Corrupt PNG"); + for (i=0; i < pal_len; ++i) { + palette[i*4+0] = stbi__get8(s); + palette[i*4+1] = stbi__get8(s); + palette[i*4+2] = stbi__get8(s); + palette[i*4+3] = 255; + } + break; + } + + case STBI__PNG_TYPE('t','R','N','S'): { + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (z->idata) return stbi__err("tRNS after IDAT","Corrupt PNG"); + if (pal_img_n) { + if (scan == STBI__SCAN_header) { s->img_n = 4; return 1; } + if (pal_len == 0) return stbi__err("tRNS before PLTE","Corrupt PNG"); + if (c.length > pal_len) return stbi__err("bad tRNS len","Corrupt PNG"); + pal_img_n = 4; + for (i=0; i < c.length; ++i) + palette[i*4+3] = stbi__get8(s); + } else { + if (!(s->img_n & 1)) return stbi__err("tRNS with alpha","Corrupt PNG"); + if (c.length != (stbi__uint32) s->img_n*2) return stbi__err("bad tRNS len","Corrupt PNG"); + has_trans = 1; + // non-paletted with tRNS = constant alpha. if header-scanning, we can stop now. + if (scan == STBI__SCAN_header) { ++s->img_n; return 1; } + if (z->depth == 16) { + for (k = 0; k < s->img_n && k < 3; ++k) // extra loop test to suppress false GCC warning + tc16[k] = (stbi__uint16)stbi__get16be(s); // copy the values as-is + } else { + for (k = 0; k < s->img_n && k < 3; ++k) + tc[k] = (stbi_uc)(stbi__get16be(s) & 255) * stbi__depth_scale_table[z->depth]; // non 8-bit images will be larger + } + } + break; + } + + case STBI__PNG_TYPE('I','D','A','T'): { + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (pal_img_n && !pal_len) return stbi__err("no PLTE","Corrupt PNG"); + if (scan == STBI__SCAN_header) { + // header scan definitely stops at first IDAT + if (pal_img_n) + s->img_n = pal_img_n; + return 1; + } + if (c.length > (1u << 30)) return stbi__err("IDAT size limit", "IDAT section larger than 2^30 bytes"); + if ((int)(ioff + c.length) < (int)ioff) return 0; + if (ioff + c.length > idata_limit) { + stbi__uint32 idata_limit_old = idata_limit; + stbi_uc *p; + if (idata_limit == 0) idata_limit = c.length > 4096 ? c.length : 4096; + while (ioff + c.length > idata_limit) + idata_limit *= 2; + STBI_NOTUSED(idata_limit_old); + p = (stbi_uc *) STBI_REALLOC_SIZED(z->idata, idata_limit_old, idata_limit); if (p == NULL) return stbi__err("outofmem", "Out of memory"); + z->idata = p; + } + if (!stbi__getn(s, z->idata+ioff,c.length)) return stbi__err("outofdata","Corrupt PNG"); + ioff += c.length; + break; + } + + case STBI__PNG_TYPE('I','E','N','D'): { + stbi__uint32 raw_len, bpl; + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if (scan != STBI__SCAN_load) return 1; + if (z->idata == NULL) return stbi__err("no IDAT","Corrupt PNG"); + // initial guess for decoded data size to avoid unnecessary reallocs + bpl = (s->img_x * z->depth + 7) / 8; // bytes per line, per component + raw_len = bpl * s->img_y * s->img_n /* pixels */ + s->img_y /* filter mode per row */; + z->expanded = (stbi_uc *) stbi_zlib_decode_malloc_guesssize_headerflag((char *) z->idata, ioff, raw_len, (int *) &raw_len, !is_iphone); + if (z->expanded == NULL) return 0; // zlib should set error + STBI_FREE(z->idata); z->idata = NULL; + if ((req_comp == s->img_n+1 && req_comp != 3 && !pal_img_n) || has_trans) + s->img_out_n = s->img_n+1; + else + s->img_out_n = s->img_n; + if (!stbi__create_png_image(z, z->expanded, raw_len, s->img_out_n, z->depth, color, interlace)) return 0; + if (has_trans) { + if (z->depth == 16) { + if (!stbi__compute_transparency16(z, tc16, s->img_out_n)) return 0; + } else { + if (!stbi__compute_transparency(z, tc, s->img_out_n)) return 0; + } + } + if (is_iphone && stbi__de_iphone_flag && s->img_out_n > 2) + stbi__de_iphone(z); + if (pal_img_n) { + // pal_img_n == 3 or 4 + s->img_n = pal_img_n; // record the actual colors we had + s->img_out_n = pal_img_n; + if (req_comp >= 3) s->img_out_n = req_comp; + if (!stbi__expand_png_palette(z, palette, pal_len, s->img_out_n)) + return 0; + } else if (has_trans) { + // non-paletted image with tRNS -> source image has (constant) alpha + ++s->img_n; + } + STBI_FREE(z->expanded); z->expanded = NULL; + // end of PNG chunk, read and skip CRC + stbi__get32be(s); + return 1; + } + + default: + // if critical, fail + if (first) return stbi__err("first not IHDR", "Corrupt PNG"); + if ((c.type & (1 << 29)) == 0) { + #ifndef STBI_NO_FAILURE_STRINGS + // not threadsafe + static char invalid_chunk[] = "XXXX PNG chunk not known"; + invalid_chunk[0] = STBI__BYTECAST(c.type >> 24); + invalid_chunk[1] = STBI__BYTECAST(c.type >> 16); + invalid_chunk[2] = STBI__BYTECAST(c.type >> 8); + invalid_chunk[3] = STBI__BYTECAST(c.type >> 0); + #endif + return stbi__err(invalid_chunk, "PNG not supported: unknown PNG chunk type"); + } + stbi__skip(s, c.length); + break; + } + // end of PNG chunk, read and skip CRC + stbi__get32be(s); + } +} + +static void *stbi__do_png(stbi__png *p, int *x, int *y, int *n, int req_comp, stbi__result_info *ri) +{ + void *result=NULL; + if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error"); + if (stbi__parse_png_file(p, STBI__SCAN_load, req_comp)) { + if (p->depth <= 8) + ri->bits_per_channel = 8; + else if (p->depth == 16) + ri->bits_per_channel = 16; + else + return stbi__errpuc("bad bits_per_channel", "PNG not supported: unsupported color depth"); + result = p->out; + p->out = NULL; + if (req_comp && req_comp != p->s->img_out_n) { + if (ri->bits_per_channel == 8) + result = stbi__convert_format((unsigned char *) result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y); + else + result = stbi__convert_format16((stbi__uint16 *) result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y); + p->s->img_out_n = req_comp; + if (result == NULL) return result; + } + *x = p->s->img_x; + *y = p->s->img_y; + if (n) *n = p->s->img_n; + } + STBI_FREE(p->out); p->out = NULL; + STBI_FREE(p->expanded); p->expanded = NULL; + STBI_FREE(p->idata); p->idata = NULL; + + return result; +} + +static void *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi__png p; + p.s = s; + return stbi__do_png(&p, x,y,comp,req_comp, ri); +} + +static int stbi__png_test(stbi__context *s) +{ + int r; + r = stbi__check_png_header(s); + stbi__rewind(s); + return r; +} + +static int stbi__png_info_raw(stbi__png *p, int *x, int *y, int *comp) +{ + if (!stbi__parse_png_file(p, STBI__SCAN_header, 0)) { + stbi__rewind( p->s ); + return 0; + } + if (x) *x = p->s->img_x; + if (y) *y = p->s->img_y; + if (comp) *comp = p->s->img_n; + return 1; +} + +static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp) +{ + stbi__png p; + p.s = s; + return stbi__png_info_raw(&p, x, y, comp); +} + +static int stbi__png_is16(stbi__context *s) +{ + stbi__png p; + p.s = s; + if (!stbi__png_info_raw(&p, NULL, NULL, NULL)) + return 0; + if (p.depth != 16) { + stbi__rewind(p.s); + return 0; + } + return 1; +} +#endif + +// Microsoft/Windows BMP image + +#ifndef STBI_NO_BMP +static int stbi__bmp_test_raw(stbi__context *s) +{ + int r; + int sz; + if (stbi__get8(s) != 'B') return 0; + if (stbi__get8(s) != 'M') return 0; + stbi__get32le(s); // discard filesize + stbi__get16le(s); // discard reserved + stbi__get16le(s); // discard reserved + stbi__get32le(s); // discard data offset + sz = stbi__get32le(s); + r = (sz == 12 || sz == 40 || sz == 56 || sz == 108 || sz == 124); + return r; +} + +static int stbi__bmp_test(stbi__context *s) +{ + int r = stbi__bmp_test_raw(s); + stbi__rewind(s); + return r; +} + + +// returns 0..31 for the highest set bit +static int stbi__high_bit(unsigned int z) +{ + int n=0; + if (z == 0) return -1; + if (z >= 0x10000) { n += 16; z >>= 16; } + if (z >= 0x00100) { n += 8; z >>= 8; } + if (z >= 0x00010) { n += 4; z >>= 4; } + if (z >= 0x00004) { n += 2; z >>= 2; } + if (z >= 0x00002) { n += 1;/* >>= 1;*/ } + return n; +} + +static int stbi__bitcount(unsigned int a) +{ + a = (a & 0x55555555) + ((a >> 1) & 0x55555555); // max 2 + a = (a & 0x33333333) + ((a >> 2) & 0x33333333); // max 4 + a = (a + (a >> 4)) & 0x0f0f0f0f; // max 8 per 4, now 8 bits + a = (a + (a >> 8)); // max 16 per 8 bits + a = (a + (a >> 16)); // max 32 per 8 bits + return a & 0xff; +} + +// extract an arbitrarily-aligned N-bit value (N=bits) +// from v, and then make it 8-bits long and fractionally +// extend it to full full range. +static int stbi__shiftsigned(unsigned int v, int shift, int bits) +{ + static unsigned int mul_table[9] = { + 0, + 0xff/*0b11111111*/, 0x55/*0b01010101*/, 0x49/*0b01001001*/, 0x11/*0b00010001*/, + 0x21/*0b00100001*/, 0x41/*0b01000001*/, 0x81/*0b10000001*/, 0x01/*0b00000001*/, + }; + static unsigned int shift_table[9] = { + 0, 0,0,1,0,2,4,6,0, + }; + if (shift < 0) + v <<= -shift; + else + v >>= shift; + STBI_ASSERT(v < 256); + v >>= (8-bits); + STBI_ASSERT(bits >= 0 && bits <= 8); + return (int) ((unsigned) v * mul_table[bits]) >> shift_table[bits]; +} + +typedef struct +{ + int bpp, offset, hsz; + unsigned int mr,mg,mb,ma, all_a; + int extra_read; +} stbi__bmp_data; + +static int stbi__bmp_set_mask_defaults(stbi__bmp_data *info, int compress) +{ + // BI_BITFIELDS specifies masks explicitly, don't override + if (compress == 3) + return 1; + + if (compress == 0) { + if (info->bpp == 16) { + info->mr = 31u << 10; + info->mg = 31u << 5; + info->mb = 31u << 0; + } else if (info->bpp == 32) { + info->mr = 0xffu << 16; + info->mg = 0xffu << 8; + info->mb = 0xffu << 0; + info->ma = 0xffu << 24; + info->all_a = 0; // if all_a is 0 at end, then we loaded alpha channel but it was all 0 + } else { + // otherwise, use defaults, which is all-0 + info->mr = info->mg = info->mb = info->ma = 0; + } + return 1; + } + return 0; // error +} + +static void *stbi__bmp_parse_header(stbi__context *s, stbi__bmp_data *info) +{ + int hsz; + if (stbi__get8(s) != 'B' || stbi__get8(s) != 'M') return stbi__errpuc("not BMP", "Corrupt BMP"); + stbi__get32le(s); // discard filesize + stbi__get16le(s); // discard reserved + stbi__get16le(s); // discard reserved + info->offset = stbi__get32le(s); + info->hsz = hsz = stbi__get32le(s); + info->mr = info->mg = info->mb = info->ma = 0; + info->extra_read = 14; + + if (info->offset < 0) return stbi__errpuc("bad BMP", "bad BMP"); + + if (hsz != 12 && hsz != 40 && hsz != 56 && hsz != 108 && hsz != 124) return stbi__errpuc("unknown BMP", "BMP type not supported: unknown"); + if (hsz == 12) { + s->img_x = stbi__get16le(s); + s->img_y = stbi__get16le(s); + } else { + s->img_x = stbi__get32le(s); + s->img_y = stbi__get32le(s); + } + if (stbi__get16le(s) != 1) return stbi__errpuc("bad BMP", "bad BMP"); + info->bpp = stbi__get16le(s); + if (hsz != 12) { + int compress = stbi__get32le(s); + if (compress == 1 || compress == 2) return stbi__errpuc("BMP RLE", "BMP type not supported: RLE"); + if (compress >= 4) return stbi__errpuc("BMP JPEG/PNG", "BMP type not supported: unsupported compression"); // this includes PNG/JPEG modes + if (compress == 3 && info->bpp != 16 && info->bpp != 32) return stbi__errpuc("bad BMP", "bad BMP"); // bitfields requires 16 or 32 bits/pixel + stbi__get32le(s); // discard sizeof + stbi__get32le(s); // discard hres + stbi__get32le(s); // discard vres + stbi__get32le(s); // discard colorsused + stbi__get32le(s); // discard max important + if (hsz == 40 || hsz == 56) { + if (hsz == 56) { + stbi__get32le(s); + stbi__get32le(s); + stbi__get32le(s); + stbi__get32le(s); + } + if (info->bpp == 16 || info->bpp == 32) { + if (compress == 0) { + stbi__bmp_set_mask_defaults(info, compress); + } else if (compress == 3) { + info->mr = stbi__get32le(s); + info->mg = stbi__get32le(s); + info->mb = stbi__get32le(s); + info->extra_read += 12; + // not documented, but generated by photoshop and handled by mspaint + if (info->mr == info->mg && info->mg == info->mb) { + // ?!?!? + return stbi__errpuc("bad BMP", "bad BMP"); + } + } else + return stbi__errpuc("bad BMP", "bad BMP"); + } + } else { + // V4/V5 header + int i; + if (hsz != 108 && hsz != 124) + return stbi__errpuc("bad BMP", "bad BMP"); + info->mr = stbi__get32le(s); + info->mg = stbi__get32le(s); + info->mb = stbi__get32le(s); + info->ma = stbi__get32le(s); + if (compress != 3) // override mr/mg/mb unless in BI_BITFIELDS mode, as per docs + stbi__bmp_set_mask_defaults(info, compress); + stbi__get32le(s); // discard color space + for (i=0; i < 12; ++i) + stbi__get32le(s); // discard color space parameters + if (hsz == 124) { + stbi__get32le(s); // discard rendering intent + stbi__get32le(s); // discard offset of profile data + stbi__get32le(s); // discard size of profile data + stbi__get32le(s); // discard reserved + } + } + } + return (void *) 1; +} + + +static void *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi_uc *out; + unsigned int mr=0,mg=0,mb=0,ma=0, all_a; + stbi_uc pal[256][4]; + int psize=0,i,j,width; + int flip_vertically, pad, target; + stbi__bmp_data info; + STBI_NOTUSED(ri); + + info.all_a = 255; + if (stbi__bmp_parse_header(s, &info) == NULL) + return NULL; // error code already set + + flip_vertically = ((int) s->img_y) > 0; + s->img_y = abs((int) s->img_y); + + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + mr = info.mr; + mg = info.mg; + mb = info.mb; + ma = info.ma; + all_a = info.all_a; + + if (info.hsz == 12) { + if (info.bpp < 24) + psize = (info.offset - info.extra_read - 24) / 3; + } else { + if (info.bpp < 16) + psize = (info.offset - info.extra_read - info.hsz) >> 2; + } + if (psize == 0) { + // accept some number of extra bytes after the header, but if the offset points either to before + // the header ends or implies a large amount of extra data, reject the file as malformed + int bytes_read_so_far = s->callback_already_read + (int)(s->img_buffer - s->img_buffer_original); + int header_limit = 1024; // max we actually read is below 256 bytes currently. + int extra_data_limit = 256*4; // what ordinarily goes here is a palette; 256 entries*4 bytes is its max size. + if (bytes_read_so_far <= 0 || bytes_read_so_far > header_limit) { + return stbi__errpuc("bad header", "Corrupt BMP"); + } + // we established that bytes_read_so_far is positive and sensible. + // the first half of this test rejects offsets that are either too small positives, or + // negative, and guarantees that info.offset >= bytes_read_so_far > 0. this in turn + // ensures the number computed in the second half of the test can't overflow. + if (info.offset < bytes_read_so_far || info.offset - bytes_read_so_far > extra_data_limit) { + return stbi__errpuc("bad offset", "Corrupt BMP"); + } else { + stbi__skip(s, info.offset - bytes_read_so_far); + } + } + + if (info.bpp == 24 && ma == 0xff000000) + s->img_n = 3; + else + s->img_n = ma ? 4 : 3; + if (req_comp && req_comp >= 3) // we can directly decode 3 or 4 + target = req_comp; + else + target = s->img_n; // if they want monochrome, we'll post-convert + + // sanity-check size + if (!stbi__mad3sizes_valid(target, s->img_x, s->img_y, 0)) + return stbi__errpuc("too large", "Corrupt BMP"); + + out = (stbi_uc *) stbi__malloc_mad3(target, s->img_x, s->img_y, 0); + if (!out) return stbi__errpuc("outofmem", "Out of memory"); + if (info.bpp < 16) { + int z=0; + if (psize == 0 || psize > 256) { STBI_FREE(out); return stbi__errpuc("invalid", "Corrupt BMP"); } + for (i=0; i < psize; ++i) { + pal[i][2] = stbi__get8(s); + pal[i][1] = stbi__get8(s); + pal[i][0] = stbi__get8(s); + if (info.hsz != 12) stbi__get8(s); + pal[i][3] = 255; + } + stbi__skip(s, info.offset - info.extra_read - info.hsz - psize * (info.hsz == 12 ? 3 : 4)); + if (info.bpp == 1) width = (s->img_x + 7) >> 3; + else if (info.bpp == 4) width = (s->img_x + 1) >> 1; + else if (info.bpp == 8) width = s->img_x; + else { STBI_FREE(out); return stbi__errpuc("bad bpp", "Corrupt BMP"); } + pad = (-width)&3; + if (info.bpp == 1) { + for (j=0; j < (int) s->img_y; ++j) { + int bit_offset = 7, v = stbi__get8(s); + for (i=0; i < (int) s->img_x; ++i) { + int color = (v>>bit_offset)&0x1; + out[z++] = pal[color][0]; + out[z++] = pal[color][1]; + out[z++] = pal[color][2]; + if (target == 4) out[z++] = 255; + if (i+1 == (int) s->img_x) break; + if((--bit_offset) < 0) { + bit_offset = 7; + v = stbi__get8(s); + } + } + stbi__skip(s, pad); + } + } else { + for (j=0; j < (int) s->img_y; ++j) { + for (i=0; i < (int) s->img_x; i += 2) { + int v=stbi__get8(s),v2=0; + if (info.bpp == 4) { + v2 = v & 15; + v >>= 4; + } + out[z++] = pal[v][0]; + out[z++] = pal[v][1]; + out[z++] = pal[v][2]; + if (target == 4) out[z++] = 255; + if (i+1 == (int) s->img_x) break; + v = (info.bpp == 8) ? stbi__get8(s) : v2; + out[z++] = pal[v][0]; + out[z++] = pal[v][1]; + out[z++] = pal[v][2]; + if (target == 4) out[z++] = 255; + } + stbi__skip(s, pad); + } + } + } else { + int rshift=0,gshift=0,bshift=0,ashift=0,rcount=0,gcount=0,bcount=0,acount=0; + int z = 0; + int easy=0; + stbi__skip(s, info.offset - info.extra_read - info.hsz); + if (info.bpp == 24) width = 3 * s->img_x; + else if (info.bpp == 16) width = 2*s->img_x; + else /* bpp = 32 and pad = 0 */ width=0; + pad = (-width) & 3; + if (info.bpp == 24) { + easy = 1; + } else if (info.bpp == 32) { + if (mb == 0xff && mg == 0xff00 && mr == 0x00ff0000 && ma == 0xff000000) + easy = 2; + } + if (!easy) { + if (!mr || !mg || !mb) { STBI_FREE(out); return stbi__errpuc("bad masks", "Corrupt BMP"); } + // right shift amt to put high bit in position #7 + rshift = stbi__high_bit(mr)-7; rcount = stbi__bitcount(mr); + gshift = stbi__high_bit(mg)-7; gcount = stbi__bitcount(mg); + bshift = stbi__high_bit(mb)-7; bcount = stbi__bitcount(mb); + ashift = stbi__high_bit(ma)-7; acount = stbi__bitcount(ma); + if (rcount > 8 || gcount > 8 || bcount > 8 || acount > 8) { STBI_FREE(out); return stbi__errpuc("bad masks", "Corrupt BMP"); } + } + for (j=0; j < (int) s->img_y; ++j) { + if (easy) { + for (i=0; i < (int) s->img_x; ++i) { + unsigned char a; + out[z+2] = stbi__get8(s); + out[z+1] = stbi__get8(s); + out[z+0] = stbi__get8(s); + z += 3; + a = (easy == 2 ? stbi__get8(s) : 255); + all_a |= a; + if (target == 4) out[z++] = a; + } + } else { + int bpp = info.bpp; + for (i=0; i < (int) s->img_x; ++i) { + stbi__uint32 v = (bpp == 16 ? (stbi__uint32) stbi__get16le(s) : stbi__get32le(s)); + unsigned int a; + out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mr, rshift, rcount)); + out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mg, gshift, gcount)); + out[z++] = STBI__BYTECAST(stbi__shiftsigned(v & mb, bshift, bcount)); + a = (ma ? stbi__shiftsigned(v & ma, ashift, acount) : 255); + all_a |= a; + if (target == 4) out[z++] = STBI__BYTECAST(a); + } + } + stbi__skip(s, pad); + } + } + + // if alpha channel is all 0s, replace with all 255s + if (target == 4 && all_a == 0) + for (i=4*s->img_x*s->img_y-1; i >= 0; i -= 4) + out[i] = 255; + + if (flip_vertically) { + stbi_uc t; + for (j=0; j < (int) s->img_y>>1; ++j) { + stbi_uc *p1 = out + j *s->img_x*target; + stbi_uc *p2 = out + (s->img_y-1-j)*s->img_x*target; + for (i=0; i < (int) s->img_x*target; ++i) { + t = p1[i]; p1[i] = p2[i]; p2[i] = t; + } + } + } + + if (req_comp && req_comp != target) { + out = stbi__convert_format(out, target, req_comp, s->img_x, s->img_y); + if (out == NULL) return out; // stbi__convert_format frees input on failure + } + + *x = s->img_x; + *y = s->img_y; + if (comp) *comp = s->img_n; + return out; +} +#endif + +// Targa Truevision - TGA +// by Jonathan Dummer +#ifndef STBI_NO_TGA +// returns STBI_rgb or whatever, 0 on error +static int stbi__tga_get_comp(int bits_per_pixel, int is_grey, int* is_rgb16) +{ + // only RGB or RGBA (incl. 16bit) or grey allowed + if (is_rgb16) *is_rgb16 = 0; + switch(bits_per_pixel) { + case 8: return STBI_grey; + case 16: if(is_grey) return STBI_grey_alpha; + // fallthrough + case 15: if(is_rgb16) *is_rgb16 = 1; + return STBI_rgb; + case 24: // fallthrough + case 32: return bits_per_pixel/8; + default: return 0; + } +} + +static int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp) +{ + int tga_w, tga_h, tga_comp, tga_image_type, tga_bits_per_pixel, tga_colormap_bpp; + int sz, tga_colormap_type; + stbi__get8(s); // discard Offset + tga_colormap_type = stbi__get8(s); // colormap type + if( tga_colormap_type > 1 ) { + stbi__rewind(s); + return 0; // only RGB or indexed allowed + } + tga_image_type = stbi__get8(s); // image type + if ( tga_colormap_type == 1 ) { // colormapped (paletted) image + if (tga_image_type != 1 && tga_image_type != 9) { + stbi__rewind(s); + return 0; + } + stbi__skip(s,4); // skip index of first colormap entry and number of entries + sz = stbi__get8(s); // check bits per palette color entry + if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) { + stbi__rewind(s); + return 0; + } + stbi__skip(s,4); // skip image x and y origin + tga_colormap_bpp = sz; + } else { // "normal" image w/o colormap - only RGB or grey allowed, +/- RLE + if ( (tga_image_type != 2) && (tga_image_type != 3) && (tga_image_type != 10) && (tga_image_type != 11) ) { + stbi__rewind(s); + return 0; // only RGB or grey allowed, +/- RLE + } + stbi__skip(s,9); // skip colormap specification and image x/y origin + tga_colormap_bpp = 0; + } + tga_w = stbi__get16le(s); + if( tga_w < 1 ) { + stbi__rewind(s); + return 0; // test width + } + tga_h = stbi__get16le(s); + if( tga_h < 1 ) { + stbi__rewind(s); + return 0; // test height + } + tga_bits_per_pixel = stbi__get8(s); // bits per pixel + stbi__get8(s); // ignore alpha bits + if (tga_colormap_bpp != 0) { + if((tga_bits_per_pixel != 8) && (tga_bits_per_pixel != 16)) { + // when using a colormap, tga_bits_per_pixel is the size of the indexes + // I don't think anything but 8 or 16bit indexes makes sense + stbi__rewind(s); + return 0; + } + tga_comp = stbi__tga_get_comp(tga_colormap_bpp, 0, NULL); + } else { + tga_comp = stbi__tga_get_comp(tga_bits_per_pixel, (tga_image_type == 3) || (tga_image_type == 11), NULL); + } + if(!tga_comp) { + stbi__rewind(s); + return 0; + } + if (x) *x = tga_w; + if (y) *y = tga_h; + if (comp) *comp = tga_comp; + return 1; // seems to have passed everything +} + +static int stbi__tga_test(stbi__context *s) +{ + int res = 0; + int sz, tga_color_type; + stbi__get8(s); // discard Offset + tga_color_type = stbi__get8(s); // color type + if ( tga_color_type > 1 ) goto errorEnd; // only RGB or indexed allowed + sz = stbi__get8(s); // image type + if ( tga_color_type == 1 ) { // colormapped (paletted) image + if (sz != 1 && sz != 9) goto errorEnd; // colortype 1 demands image type 1 or 9 + stbi__skip(s,4); // skip index of first colormap entry and number of entries + sz = stbi__get8(s); // check bits per palette color entry + if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) goto errorEnd; + stbi__skip(s,4); // skip image x and y origin + } else { // "normal" image w/o colormap + if ( (sz != 2) && (sz != 3) && (sz != 10) && (sz != 11) ) goto errorEnd; // only RGB or grey allowed, +/- RLE + stbi__skip(s,9); // skip colormap specification and image x/y origin + } + if ( stbi__get16le(s) < 1 ) goto errorEnd; // test width + if ( stbi__get16le(s) < 1 ) goto errorEnd; // test height + sz = stbi__get8(s); // bits per pixel + if ( (tga_color_type == 1) && (sz != 8) && (sz != 16) ) goto errorEnd; // for colormapped images, bpp is size of an index + if ( (sz != 8) && (sz != 15) && (sz != 16) && (sz != 24) && (sz != 32) ) goto errorEnd; + + res = 1; // if we got this far, everything's good and we can return 1 instead of 0 + +errorEnd: + stbi__rewind(s); + return res; +} + +// read 16bit value and convert to 24bit RGB +static void stbi__tga_read_rgb16(stbi__context *s, stbi_uc* out) +{ + stbi__uint16 px = (stbi__uint16)stbi__get16le(s); + stbi__uint16 fiveBitMask = 31; + // we have 3 channels with 5bits each + int r = (px >> 10) & fiveBitMask; + int g = (px >> 5) & fiveBitMask; + int b = px & fiveBitMask; + // Note that this saves the data in RGB(A) order, so it doesn't need to be swapped later + out[0] = (stbi_uc)((r * 255)/31); + out[1] = (stbi_uc)((g * 255)/31); + out[2] = (stbi_uc)((b * 255)/31); + + // some people claim that the most significant bit might be used for alpha + // (possibly if an alpha-bit is set in the "image descriptor byte") + // but that only made 16bit test images completely translucent.. + // so let's treat all 15 and 16bit TGAs as RGB with no alpha. +} + +static void *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + // read in the TGA header stuff + int tga_offset = stbi__get8(s); + int tga_indexed = stbi__get8(s); + int tga_image_type = stbi__get8(s); + int tga_is_RLE = 0; + int tga_palette_start = stbi__get16le(s); + int tga_palette_len = stbi__get16le(s); + int tga_palette_bits = stbi__get8(s); + int tga_x_origin = stbi__get16le(s); + int tga_y_origin = stbi__get16le(s); + int tga_width = stbi__get16le(s); + int tga_height = stbi__get16le(s); + int tga_bits_per_pixel = stbi__get8(s); + int tga_comp, tga_rgb16=0; + int tga_inverted = stbi__get8(s); + // int tga_alpha_bits = tga_inverted & 15; // the 4 lowest bits - unused (useless?) + // image data + unsigned char *tga_data; + unsigned char *tga_palette = NULL; + int i, j; + unsigned char raw_data[4] = {0}; + int RLE_count = 0; + int RLE_repeating = 0; + int read_next_pixel = 1; + STBI_NOTUSED(ri); + STBI_NOTUSED(tga_x_origin); // @TODO + STBI_NOTUSED(tga_y_origin); // @TODO + + if (tga_height > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (tga_width > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + // do a tiny bit of precessing + if ( tga_image_type >= 8 ) + { + tga_image_type -= 8; + tga_is_RLE = 1; + } + tga_inverted = 1 - ((tga_inverted >> 5) & 1); + + // If I'm paletted, then I'll use the number of bits from the palette + if ( tga_indexed ) tga_comp = stbi__tga_get_comp(tga_palette_bits, 0, &tga_rgb16); + else tga_comp = stbi__tga_get_comp(tga_bits_per_pixel, (tga_image_type == 3), &tga_rgb16); + + if(!tga_comp) // shouldn't really happen, stbi__tga_test() should have ensured basic consistency + return stbi__errpuc("bad format", "Can't find out TGA pixelformat"); + + // tga info + *x = tga_width; + *y = tga_height; + if (comp) *comp = tga_comp; + + if (!stbi__mad3sizes_valid(tga_width, tga_height, tga_comp, 0)) + return stbi__errpuc("too large", "Corrupt TGA"); + + tga_data = (unsigned char*)stbi__malloc_mad3(tga_width, tga_height, tga_comp, 0); + if (!tga_data) return stbi__errpuc("outofmem", "Out of memory"); + + // skip to the data's starting position (offset usually = 0) + stbi__skip(s, tga_offset ); + + if ( !tga_indexed && !tga_is_RLE && !tga_rgb16 ) { + for (i=0; i < tga_height; ++i) { + int row = tga_inverted ? tga_height -i - 1 : i; + stbi_uc *tga_row = tga_data + row*tga_width*tga_comp; + stbi__getn(s, tga_row, tga_width * tga_comp); + } + } else { + // do I need to load a palette? + if ( tga_indexed) + { + if (tga_palette_len == 0) { /* you have to have at least one entry! */ + STBI_FREE(tga_data); + return stbi__errpuc("bad palette", "Corrupt TGA"); + } + + // any data to skip? (offset usually = 0) + stbi__skip(s, tga_palette_start ); + // load the palette + tga_palette = (unsigned char*)stbi__malloc_mad2(tga_palette_len, tga_comp, 0); + if (!tga_palette) { + STBI_FREE(tga_data); + return stbi__errpuc("outofmem", "Out of memory"); + } + if (tga_rgb16) { + stbi_uc *pal_entry = tga_palette; + STBI_ASSERT(tga_comp == STBI_rgb); + for (i=0; i < tga_palette_len; ++i) { + stbi__tga_read_rgb16(s, pal_entry); + pal_entry += tga_comp; + } + } else if (!stbi__getn(s, tga_palette, tga_palette_len * tga_comp)) { + STBI_FREE(tga_data); + STBI_FREE(tga_palette); + return stbi__errpuc("bad palette", "Corrupt TGA"); + } + } + // load the data + for (i=0; i < tga_width * tga_height; ++i) + { + // if I'm in RLE mode, do I need to get a RLE stbi__pngchunk? + if ( tga_is_RLE ) + { + if ( RLE_count == 0 ) + { + // yep, get the next byte as a RLE command + int RLE_cmd = stbi__get8(s); + RLE_count = 1 + (RLE_cmd & 127); + RLE_repeating = RLE_cmd >> 7; + read_next_pixel = 1; + } else if ( !RLE_repeating ) + { + read_next_pixel = 1; + } + } else + { + read_next_pixel = 1; + } + // OK, if I need to read a pixel, do it now + if ( read_next_pixel ) + { + // load however much data we did have + if ( tga_indexed ) + { + // read in index, then perform the lookup + int pal_idx = (tga_bits_per_pixel == 8) ? stbi__get8(s) : stbi__get16le(s); + if ( pal_idx >= tga_palette_len ) { + // invalid index + pal_idx = 0; + } + pal_idx *= tga_comp; + for (j = 0; j < tga_comp; ++j) { + raw_data[j] = tga_palette[pal_idx+j]; + } + } else if(tga_rgb16) { + STBI_ASSERT(tga_comp == STBI_rgb); + stbi__tga_read_rgb16(s, raw_data); + } else { + // read in the data raw + for (j = 0; j < tga_comp; ++j) { + raw_data[j] = stbi__get8(s); + } + } + // clear the reading flag for the next pixel + read_next_pixel = 0; + } // end of reading a pixel + + // copy data + for (j = 0; j < tga_comp; ++j) + tga_data[i*tga_comp+j] = raw_data[j]; + + // in case we're in RLE mode, keep counting down + --RLE_count; + } + // do I need to invert the image? + if ( tga_inverted ) + { + for (j = 0; j*2 < tga_height; ++j) + { + int index1 = j * tga_width * tga_comp; + int index2 = (tga_height - 1 - j) * tga_width * tga_comp; + for (i = tga_width * tga_comp; i > 0; --i) + { + unsigned char temp = tga_data[index1]; + tga_data[index1] = tga_data[index2]; + tga_data[index2] = temp; + ++index1; + ++index2; + } + } + } + // clear my palette, if I had one + if ( tga_palette != NULL ) + { + STBI_FREE( tga_palette ); + } + } + + // swap RGB - if the source data was RGB16, it already is in the right order + if (tga_comp >= 3 && !tga_rgb16) + { + unsigned char* tga_pixel = tga_data; + for (i=0; i < tga_width * tga_height; ++i) + { + unsigned char temp = tga_pixel[0]; + tga_pixel[0] = tga_pixel[2]; + tga_pixel[2] = temp; + tga_pixel += tga_comp; + } + } + + // convert to target component count + if (req_comp && req_comp != tga_comp) + tga_data = stbi__convert_format(tga_data, tga_comp, req_comp, tga_width, tga_height); + + // the things I do to get rid of an error message, and yet keep + // Microsoft's C compilers happy... [8^( + tga_palette_start = tga_palette_len = tga_palette_bits = + tga_x_origin = tga_y_origin = 0; + STBI_NOTUSED(tga_palette_start); + // OK, done + return tga_data; +} +#endif + +// ************************************************************************************************* +// Photoshop PSD loader -- PD by Thatcher Ulrich, integration by Nicolas Schulz, tweaked by STB + +#ifndef STBI_NO_PSD +static int stbi__psd_test(stbi__context *s) +{ + int r = (stbi__get32be(s) == 0x38425053); + stbi__rewind(s); + return r; +} + +static int stbi__psd_decode_rle(stbi__context *s, stbi_uc *p, int pixelCount) +{ + int count, nleft, len; + + count = 0; + while ((nleft = pixelCount - count) > 0) { + len = stbi__get8(s); + if (len == 128) { + // No-op. + } else if (len < 128) { + // Copy next len+1 bytes literally. + len++; + if (len > nleft) return 0; // corrupt data + count += len; + while (len) { + *p = stbi__get8(s); + p += 4; + len--; + } + } else if (len > 128) { + stbi_uc val; + // Next -len+1 bytes in the dest are replicated from next source byte. + // (Interpret len as a negative 8-bit int.) + len = 257 - len; + if (len > nleft) return 0; // corrupt data + val = stbi__get8(s); + count += len; + while (len) { + *p = val; + p += 4; + len--; + } + } + } + + return 1; +} + +static void *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri, int bpc) +{ + int pixelCount; + int channelCount, compression; + int channel, i; + int bitdepth; + int w,h; + stbi_uc *out; + STBI_NOTUSED(ri); + + // Check identifier + if (stbi__get32be(s) != 0x38425053) // "8BPS" + return stbi__errpuc("not PSD", "Corrupt PSD image"); + + // Check file type version. + if (stbi__get16be(s) != 1) + return stbi__errpuc("wrong version", "Unsupported version of PSD image"); + + // Skip 6 reserved bytes. + stbi__skip(s, 6 ); + + // Read the number of channels (R, G, B, A, etc). + channelCount = stbi__get16be(s); + if (channelCount < 0 || channelCount > 16) + return stbi__errpuc("wrong channel count", "Unsupported number of channels in PSD image"); + + // Read the rows and columns of the image. + h = stbi__get32be(s); + w = stbi__get32be(s); + + if (h > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (w > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + // Make sure the depth is 8 bits. + bitdepth = stbi__get16be(s); + if (bitdepth != 8 && bitdepth != 16) + return stbi__errpuc("unsupported bit depth", "PSD bit depth is not 8 or 16 bit"); + + // Make sure the color mode is RGB. + // Valid options are: + // 0: Bitmap + // 1: Grayscale + // 2: Indexed color + // 3: RGB color + // 4: CMYK color + // 7: Multichannel + // 8: Duotone + // 9: Lab color + if (stbi__get16be(s) != 3) + return stbi__errpuc("wrong color format", "PSD is not in RGB color format"); + + // Skip the Mode Data. (It's the palette for indexed color; other info for other modes.) + stbi__skip(s,stbi__get32be(s) ); + + // Skip the image resources. (resolution, pen tool paths, etc) + stbi__skip(s, stbi__get32be(s) ); + + // Skip the reserved data. + stbi__skip(s, stbi__get32be(s) ); + + // Find out if the data is compressed. + // Known values: + // 0: no compression + // 1: RLE compressed + compression = stbi__get16be(s); + if (compression > 1) + return stbi__errpuc("bad compression", "PSD has an unknown compression format"); + + // Check size + if (!stbi__mad3sizes_valid(4, w, h, 0)) + return stbi__errpuc("too large", "Corrupt PSD"); + + // Create the destination image. + + if (!compression && bitdepth == 16 && bpc == 16) { + out = (stbi_uc *) stbi__malloc_mad3(8, w, h, 0); + ri->bits_per_channel = 16; + } else + out = (stbi_uc *) stbi__malloc(4 * w*h); + + if (!out) return stbi__errpuc("outofmem", "Out of memory"); + pixelCount = w*h; + + // Initialize the data to zero. + //memset( out, 0, pixelCount * 4 ); + + // Finally, the image data. + if (compression) { + // RLE as used by .PSD and .TIFF + // Loop until you get the number of unpacked bytes you are expecting: + // Read the next source byte into n. + // If n is between 0 and 127 inclusive, copy the next n+1 bytes literally. + // Else if n is between -127 and -1 inclusive, copy the next byte -n+1 times. + // Else if n is 128, noop. + // Endloop + + // The RLE-compressed data is preceded by a 2-byte data count for each row in the data, + // which we're going to just skip. + stbi__skip(s, h * channelCount * 2 ); + + // Read the RLE data by channel. + for (channel = 0; channel < 4; channel++) { + stbi_uc *p; + + p = out+channel; + if (channel >= channelCount) { + // Fill this channel with default data. + for (i = 0; i < pixelCount; i++, p += 4) + *p = (channel == 3 ? 255 : 0); + } else { + // Read the RLE data. + if (!stbi__psd_decode_rle(s, p, pixelCount)) { + STBI_FREE(out); + return stbi__errpuc("corrupt", "bad RLE data"); + } + } + } + + } else { + // We're at the raw image data. It's each channel in order (Red, Green, Blue, Alpha, ...) + // where each channel consists of an 8-bit (or 16-bit) value for each pixel in the image. + + // Read the data by channel. + for (channel = 0; channel < 4; channel++) { + if (channel >= channelCount) { + // Fill this channel with default data. + if (bitdepth == 16 && bpc == 16) { + stbi__uint16 *q = ((stbi__uint16 *) out) + channel; + stbi__uint16 val = channel == 3 ? 65535 : 0; + for (i = 0; i < pixelCount; i++, q += 4) + *q = val; + } else { + stbi_uc *p = out+channel; + stbi_uc val = channel == 3 ? 255 : 0; + for (i = 0; i < pixelCount; i++, p += 4) + *p = val; + } + } else { + if (ri->bits_per_channel == 16) { // output bpc + stbi__uint16 *q = ((stbi__uint16 *) out) + channel; + for (i = 0; i < pixelCount; i++, q += 4) + *q = (stbi__uint16) stbi__get16be(s); + } else { + stbi_uc *p = out+channel; + if (bitdepth == 16) { // input bpc + for (i = 0; i < pixelCount; i++, p += 4) + *p = (stbi_uc) (stbi__get16be(s) >> 8); + } else { + for (i = 0; i < pixelCount; i++, p += 4) + *p = stbi__get8(s); + } + } + } + } + } + + // remove weird white matte from PSD + if (channelCount >= 4) { + if (ri->bits_per_channel == 16) { + for (i=0; i < w*h; ++i) { + stbi__uint16 *pixel = (stbi__uint16 *) out + 4*i; + if (pixel[3] != 0 && pixel[3] != 65535) { + float a = pixel[3] / 65535.0f; + float ra = 1.0f / a; + float inv_a = 65535.0f * (1 - ra); + pixel[0] = (stbi__uint16) (pixel[0]*ra + inv_a); + pixel[1] = (stbi__uint16) (pixel[1]*ra + inv_a); + pixel[2] = (stbi__uint16) (pixel[2]*ra + inv_a); + } + } + } else { + for (i=0; i < w*h; ++i) { + unsigned char *pixel = out + 4*i; + if (pixel[3] != 0 && pixel[3] != 255) { + float a = pixel[3] / 255.0f; + float ra = 1.0f / a; + float inv_a = 255.0f * (1 - ra); + pixel[0] = (unsigned char) (pixel[0]*ra + inv_a); + pixel[1] = (unsigned char) (pixel[1]*ra + inv_a); + pixel[2] = (unsigned char) (pixel[2]*ra + inv_a); + } + } + } + } + + // convert to desired output format + if (req_comp && req_comp != 4) { + if (ri->bits_per_channel == 16) + out = (stbi_uc *) stbi__convert_format16((stbi__uint16 *) out, 4, req_comp, w, h); + else + out = stbi__convert_format(out, 4, req_comp, w, h); + if (out == NULL) return out; // stbi__convert_format frees input on failure + } + + if (comp) *comp = 4; + *y = h; + *x = w; + + return out; +} +#endif + +// ************************************************************************************************* +// Softimage PIC loader +// by Tom Seddon +// +// See http://softimage.wiki.softimage.com/index.php/INFO:_PIC_file_format +// See http://ozviz.wasp.uwa.edu.au/~pbourke/dataformats/softimagepic/ + +#ifndef STBI_NO_PIC +static int stbi__pic_is4(stbi__context *s,const char *str) +{ + int i; + for (i=0; i<4; ++i) + if (stbi__get8(s) != (stbi_uc)str[i]) + return 0; + + return 1; +} + +static int stbi__pic_test_core(stbi__context *s) +{ + int i; + + if (!stbi__pic_is4(s,"\x53\x80\xF6\x34")) + return 0; + + for(i=0;i<84;++i) + stbi__get8(s); + + if (!stbi__pic_is4(s,"PICT")) + return 0; + + return 1; +} + +typedef struct +{ + stbi_uc size,type,channel; +} stbi__pic_packet; + +static stbi_uc *stbi__readval(stbi__context *s, int channel, stbi_uc *dest) +{ + int mask=0x80, i; + + for (i=0; i<4; ++i, mask>>=1) { + if (channel & mask) { + if (stbi__at_eof(s)) return stbi__errpuc("bad file","PIC file too short"); + dest[i]=stbi__get8(s); + } + } + + return dest; +} + +static void stbi__copyval(int channel,stbi_uc *dest,const stbi_uc *src) +{ + int mask=0x80,i; + + for (i=0;i<4; ++i, mask>>=1) + if (channel&mask) + dest[i]=src[i]; +} + +static stbi_uc *stbi__pic_load_core(stbi__context *s,int width,int height,int *comp, stbi_uc *result) +{ + int act_comp=0,num_packets=0,y,chained; + stbi__pic_packet packets[10]; + + // this will (should...) cater for even some bizarre stuff like having data + // for the same channel in multiple packets. + do { + stbi__pic_packet *packet; + + if (num_packets==sizeof(packets)/sizeof(packets[0])) + return stbi__errpuc("bad format","too many packets"); + + packet = &packets[num_packets++]; + + chained = stbi__get8(s); + packet->size = stbi__get8(s); + packet->type = stbi__get8(s); + packet->channel = stbi__get8(s); + + act_comp |= packet->channel; + + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (reading packets)"); + if (packet->size != 8) return stbi__errpuc("bad format","packet isn't 8bpp"); + } while (chained); + + *comp = (act_comp & 0x10 ? 4 : 3); // has alpha channel? + + for(y=0; y<height; ++y) { + int packet_idx; + + for(packet_idx=0; packet_idx < num_packets; ++packet_idx) { + stbi__pic_packet *packet = &packets[packet_idx]; + stbi_uc *dest = result+y*width*4; + + switch (packet->type) { + default: + return stbi__errpuc("bad format","packet has bad compression type"); + + case 0: {//uncompressed + int x; + + for(x=0;x<width;++x, dest+=4) + if (!stbi__readval(s,packet->channel,dest)) + return 0; + break; + } + + case 1://Pure RLE + { + int left=width, i; + + while (left>0) { + stbi_uc count,value[4]; + + count=stbi__get8(s); + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (pure read count)"); + + if (count > left) + count = (stbi_uc) left; + + if (!stbi__readval(s,packet->channel,value)) return 0; + + for(i=0; i<count; ++i,dest+=4) + stbi__copyval(packet->channel,dest,value); + left -= count; + } + } + break; + + case 2: {//Mixed RLE + int left=width; + while (left>0) { + int count = stbi__get8(s), i; + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (mixed read count)"); + + if (count >= 128) { // Repeated + stbi_uc value[4]; + + if (count==128) + count = stbi__get16be(s); + else + count -= 127; + if (count > left) + return stbi__errpuc("bad file","scanline overrun"); + + if (!stbi__readval(s,packet->channel,value)) + return 0; + + for(i=0;i<count;++i, dest += 4) + stbi__copyval(packet->channel,dest,value); + } else { // Raw + ++count; + if (count>left) return stbi__errpuc("bad file","scanline overrun"); + + for(i=0;i<count;++i, dest+=4) + if (!stbi__readval(s,packet->channel,dest)) + return 0; + } + left-=count; + } + break; + } + } + } + } + + return result; +} + +static void *stbi__pic_load(stbi__context *s,int *px,int *py,int *comp,int req_comp, stbi__result_info *ri) +{ + stbi_uc *result; + int i, x,y, internal_comp; + STBI_NOTUSED(ri); + + if (!comp) comp = &internal_comp; + + for (i=0; i<92; ++i) + stbi__get8(s); + + x = stbi__get16be(s); + y = stbi__get16be(s); + + if (y > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (x > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (pic header)"); + if (!stbi__mad3sizes_valid(x, y, 4, 0)) return stbi__errpuc("too large", "PIC image too large to decode"); + + stbi__get32be(s); //skip `ratio' + stbi__get16be(s); //skip `fields' + stbi__get16be(s); //skip `pad' + + // intermediate buffer is RGBA + result = (stbi_uc *) stbi__malloc_mad3(x, y, 4, 0); + if (!result) return stbi__errpuc("outofmem", "Out of memory"); + memset(result, 0xff, x*y*4); + + if (!stbi__pic_load_core(s,x,y,comp, result)) { + STBI_FREE(result); + result=0; + } + *px = x; + *py = y; + if (req_comp == 0) req_comp = *comp; + result=stbi__convert_format(result,4,req_comp,x,y); + + return result; +} + +static int stbi__pic_test(stbi__context *s) +{ + int r = stbi__pic_test_core(s); + stbi__rewind(s); + return r; +} +#endif + +// ************************************************************************************************* +// GIF loader -- public domain by Jean-Marc Lienher -- simplified/shrunk by stb + +#ifndef STBI_NO_GIF +typedef struct +{ + stbi__int16 prefix; + stbi_uc first; + stbi_uc suffix; +} stbi__gif_lzw; + +typedef struct +{ + int w,h; + stbi_uc *out; // output buffer (always 4 components) + stbi_uc *background; // The current "background" as far as a gif is concerned + stbi_uc *history; + int flags, bgindex, ratio, transparent, eflags; + stbi_uc pal[256][4]; + stbi_uc lpal[256][4]; + stbi__gif_lzw codes[8192]; + stbi_uc *color_table; + int parse, step; + int lflags; + int start_x, start_y; + int max_x, max_y; + int cur_x, cur_y; + int line_size; + int delay; +} stbi__gif; + +static int stbi__gif_test_raw(stbi__context *s) +{ + int sz; + if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8') return 0; + sz = stbi__get8(s); + if (sz != '9' && sz != '7') return 0; + if (stbi__get8(s) != 'a') return 0; + return 1; +} + +static int stbi__gif_test(stbi__context *s) +{ + int r = stbi__gif_test_raw(s); + stbi__rewind(s); + return r; +} + +static void stbi__gif_parse_colortable(stbi__context *s, stbi_uc pal[256][4], int num_entries, int transp) +{ + int i; + for (i=0; i < num_entries; ++i) { + pal[i][2] = stbi__get8(s); + pal[i][1] = stbi__get8(s); + pal[i][0] = stbi__get8(s); + pal[i][3] = transp == i ? 0 : 255; + } +} + +static int stbi__gif_header(stbi__context *s, stbi__gif *g, int *comp, int is_info) +{ + stbi_uc version; + if (stbi__get8(s) != 'G' || stbi__get8(s) != 'I' || stbi__get8(s) != 'F' || stbi__get8(s) != '8') + return stbi__err("not GIF", "Corrupt GIF"); + + version = stbi__get8(s); + if (version != '7' && version != '9') return stbi__err("not GIF", "Corrupt GIF"); + if (stbi__get8(s) != 'a') return stbi__err("not GIF", "Corrupt GIF"); + + stbi__g_failure_reason = ""; + g->w = stbi__get16le(s); + g->h = stbi__get16le(s); + g->flags = stbi__get8(s); + g->bgindex = stbi__get8(s); + g->ratio = stbi__get8(s); + g->transparent = -1; + + if (g->w > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + if (g->h > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)"); + + if (comp != 0) *comp = 4; // can't actually tell whether it's 3 or 4 until we parse the comments + + if (is_info) return 1; + + if (g->flags & 0x80) + stbi__gif_parse_colortable(s,g->pal, 2 << (g->flags & 7), -1); + + return 1; +} + +static int stbi__gif_info_raw(stbi__context *s, int *x, int *y, int *comp) +{ + stbi__gif* g = (stbi__gif*) stbi__malloc(sizeof(stbi__gif)); + if (!g) return stbi__err("outofmem", "Out of memory"); + if (!stbi__gif_header(s, g, comp, 1)) { + STBI_FREE(g); + stbi__rewind( s ); + return 0; + } + if (x) *x = g->w; + if (y) *y = g->h; + STBI_FREE(g); + return 1; +} + +static void stbi__out_gif_code(stbi__gif *g, stbi__uint16 code) +{ + stbi_uc *p, *c; + int idx; + + // recurse to decode the prefixes, since the linked-list is backwards, + // and working backwards through an interleaved image would be nasty + if (g->codes[code].prefix >= 0) + stbi__out_gif_code(g, g->codes[code].prefix); + + if (g->cur_y >= g->max_y) return; + + idx = g->cur_x + g->cur_y; + p = &g->out[idx]; + g->history[idx / 4] = 1; + + c = &g->color_table[g->codes[code].suffix * 4]; + if (c[3] > 128) { // don't render transparent pixels; + p[0] = c[2]; + p[1] = c[1]; + p[2] = c[0]; + p[3] = c[3]; + } + g->cur_x += 4; + + if (g->cur_x >= g->max_x) { + g->cur_x = g->start_x; + g->cur_y += g->step; + + while (g->cur_y >= g->max_y && g->parse > 0) { + g->step = (1 << g->parse) * g->line_size; + g->cur_y = g->start_y + (g->step >> 1); + --g->parse; + } + } +} + +static stbi_uc *stbi__process_gif_raster(stbi__context *s, stbi__gif *g) +{ + stbi_uc lzw_cs; + stbi__int32 len, init_code; + stbi__uint32 first; + stbi__int32 codesize, codemask, avail, oldcode, bits, valid_bits, clear; + stbi__gif_lzw *p; + + lzw_cs = stbi__get8(s); + if (lzw_cs > 12) return NULL; + clear = 1 << lzw_cs; + first = 1; + codesize = lzw_cs + 1; + codemask = (1 << codesize) - 1; + bits = 0; + valid_bits = 0; + for (init_code = 0; init_code < clear; init_code++) { + g->codes[init_code].prefix = -1; + g->codes[init_code].first = (stbi_uc) init_code; + g->codes[init_code].suffix = (stbi_uc) init_code; + } + + // support no starting clear code + avail = clear+2; + oldcode = -1; + + len = 0; + for(;;) { + if (valid_bits < codesize) { + if (len == 0) { + len = stbi__get8(s); // start new block + if (len == 0) + return g->out; + } + --len; + bits |= (stbi__int32) stbi__get8(s) << valid_bits; + valid_bits += 8; + } else { + stbi__int32 code = bits & codemask; + bits >>= codesize; + valid_bits -= codesize; + // @OPTIMIZE: is there some way we can accelerate the non-clear path? + if (code == clear) { // clear code + codesize = lzw_cs + 1; + codemask = (1 << codesize) - 1; + avail = clear + 2; + oldcode = -1; + first = 0; + } else if (code == clear + 1) { // end of stream code + stbi__skip(s, len); + while ((len = stbi__get8(s)) > 0) + stbi__skip(s,len); + return g->out; + } else if (code <= avail) { + if (first) { + return stbi__errpuc("no clear code", "Corrupt GIF"); + } + + if (oldcode >= 0) { + p = &g->codes[avail++]; + if (avail > 8192) { + return stbi__errpuc("too many codes", "Corrupt GIF"); + } + + p->prefix = (stbi__int16) oldcode; + p->first = g->codes[oldcode].first; + p->suffix = (code == avail) ? p->first : g->codes[code].first; + } else if (code == avail) + return stbi__errpuc("illegal code in raster", "Corrupt GIF"); + + stbi__out_gif_code(g, (stbi__uint16) code); + + if ((avail & codemask) == 0 && avail <= 0x0FFF) { + codesize++; + codemask = (1 << codesize) - 1; + } + + oldcode = code; + } else { + return stbi__errpuc("illegal code in raster", "Corrupt GIF"); + } + } + } +} + +// this function is designed to support animated gifs, although stb_image doesn't support it +// two back is the image from two frames ago, used for a very specific disposal format +static stbi_uc *stbi__gif_load_next(stbi__context *s, stbi__gif *g, int *comp, int req_comp, stbi_uc *two_back) +{ + int dispose; + int first_frame; + int pi; + int pcount; + STBI_NOTUSED(req_comp); + + // on first frame, any non-written pixels get the background colour (non-transparent) + first_frame = 0; + if (g->out == 0) { + if (!stbi__gif_header(s, g, comp,0)) return 0; // stbi__g_failure_reason set by stbi__gif_header + if (!stbi__mad3sizes_valid(4, g->w, g->h, 0)) + return stbi__errpuc("too large", "GIF image is too large"); + pcount = g->w * g->h; + g->out = (stbi_uc *) stbi__malloc(4 * pcount); + g->background = (stbi_uc *) stbi__malloc(4 * pcount); + g->history = (stbi_uc *) stbi__malloc(pcount); + if (!g->out || !g->background || !g->history) + return stbi__errpuc("outofmem", "Out of memory"); + + // image is treated as "transparent" at the start - ie, nothing overwrites the current background; + // background colour is only used for pixels that are not rendered first frame, after that "background" + // color refers to the color that was there the previous frame. + memset(g->out, 0x00, 4 * pcount); + memset(g->background, 0x00, 4 * pcount); // state of the background (starts transparent) + memset(g->history, 0x00, pcount); // pixels that were affected previous frame + first_frame = 1; + } else { + // second frame - how do we dispose of the previous one? + dispose = (g->eflags & 0x1C) >> 2; + pcount = g->w * g->h; + + if ((dispose == 3) && (two_back == 0)) { + dispose = 2; // if I don't have an image to revert back to, default to the old background + } + + if (dispose == 3) { // use previous graphic + for (pi = 0; pi < pcount; ++pi) { + if (g->history[pi]) { + memcpy( &g->out[pi * 4], &two_back[pi * 4], 4 ); + } + } + } else if (dispose == 2) { + // restore what was changed last frame to background before that frame; + for (pi = 0; pi < pcount; ++pi) { + if (g->history[pi]) { + memcpy( &g->out[pi * 4], &g->background[pi * 4], 4 ); + } + } + } else { + // This is a non-disposal case eithe way, so just + // leave the pixels as is, and they will become the new background + // 1: do not dispose + // 0: not specified. + } + + // background is what out is after the undoing of the previou frame; + memcpy( g->background, g->out, 4 * g->w * g->h ); + } + + // clear my history; + memset( g->history, 0x00, g->w * g->h ); // pixels that were affected previous frame + + for (;;) { + int tag = stbi__get8(s); + switch (tag) { + case 0x2C: /* Image Descriptor */ + { + stbi__int32 x, y, w, h; + stbi_uc *o; + + x = stbi__get16le(s); + y = stbi__get16le(s); + w = stbi__get16le(s); + h = stbi__get16le(s); + if (((x + w) > (g->w)) || ((y + h) > (g->h))) + return stbi__errpuc("bad Image Descriptor", "Corrupt GIF"); + + g->line_size = g->w * 4; + g->start_x = x * 4; + g->start_y = y * g->line_size; + g->max_x = g->start_x + w * 4; + g->max_y = g->start_y + h * g->line_size; + g->cur_x = g->start_x; + g->cur_y = g->start_y; + + // if the width of the specified rectangle is 0, that means + // we may not see *any* pixels or the image is malformed; + // to make sure this is caught, move the current y down to + // max_y (which is what out_gif_code checks). + if (w == 0) + g->cur_y = g->max_y; + + g->lflags = stbi__get8(s); + + if (g->lflags & 0x40) { + g->step = 8 * g->line_size; // first interlaced spacing + g->parse = 3; + } else { + g->step = g->line_size; + g->parse = 0; + } + + if (g->lflags & 0x80) { + stbi__gif_parse_colortable(s,g->lpal, 2 << (g->lflags & 7), g->eflags & 0x01 ? g->transparent : -1); + g->color_table = (stbi_uc *) g->lpal; + } else if (g->flags & 0x80) { + g->color_table = (stbi_uc *) g->pal; + } else + return stbi__errpuc("missing color table", "Corrupt GIF"); + + o = stbi__process_gif_raster(s, g); + if (!o) return NULL; + + // if this was the first frame, + pcount = g->w * g->h; + if (first_frame && (g->bgindex > 0)) { + // if first frame, any pixel not drawn to gets the background color + for (pi = 0; pi < pcount; ++pi) { + if (g->history[pi] == 0) { + g->pal[g->bgindex][3] = 255; // just in case it was made transparent, undo that; It will be reset next frame if need be; + memcpy( &g->out[pi * 4], &g->pal[g->bgindex], 4 ); + } + } + } + + return o; + } + + case 0x21: // Comment Extension. + { + int len; + int ext = stbi__get8(s); + if (ext == 0xF9) { // Graphic Control Extension. + len = stbi__get8(s); + if (len == 4) { + g->eflags = stbi__get8(s); + g->delay = 10 * stbi__get16le(s); // delay - 1/100th of a second, saving as 1/1000ths. + + // unset old transparent + if (g->transparent >= 0) { + g->pal[g->transparent][3] = 255; + } + if (g->eflags & 0x01) { + g->transparent = stbi__get8(s); + if (g->transparent >= 0) { + g->pal[g->transparent][3] = 0; + } + } else { + // don't need transparent + stbi__skip(s, 1); + g->transparent = -1; + } + } else { + stbi__skip(s, len); + break; + } + } + while ((len = stbi__get8(s)) != 0) { + stbi__skip(s, len); + } + break; + } + + case 0x3B: // gif stream termination code + return (stbi_uc *) s; // using '1' causes warning on some compilers + + default: + return stbi__errpuc("unknown code", "Corrupt GIF"); + } + } +} + +static void *stbi__load_gif_main_outofmem(stbi__gif *g, stbi_uc *out, int **delays) +{ + STBI_FREE(g->out); + STBI_FREE(g->history); + STBI_FREE(g->background); + + if (out) STBI_FREE(out); + if (delays && *delays) STBI_FREE(*delays); + return stbi__errpuc("outofmem", "Out of memory"); +} + +static void *stbi__load_gif_main(stbi__context *s, int **delays, int *x, int *y, int *z, int *comp, int req_comp) +{ + if (stbi__gif_test(s)) { + int layers = 0; + stbi_uc *u = 0; + stbi_uc *out = 0; + stbi_uc *two_back = 0; + stbi__gif g; + int stride; + int out_size = 0; + int delays_size = 0; + + STBI_NOTUSED(out_size); + STBI_NOTUSED(delays_size); + + memset(&g, 0, sizeof(g)); + if (delays) { + *delays = 0; + } + + do { + u = stbi__gif_load_next(s, &g, comp, req_comp, two_back); + if (u == (stbi_uc *) s) u = 0; // end of animated gif marker + + if (u) { + *x = g.w; + *y = g.h; + ++layers; + stride = g.w * g.h * 4; + + if (out) { + void *tmp = (stbi_uc*) STBI_REALLOC_SIZED( out, out_size, layers * stride ); + if (!tmp) + return stbi__load_gif_main_outofmem(&g, out, delays); + else { + out = (stbi_uc*) tmp; + out_size = layers * stride; + } + + if (delays) { + int *new_delays = (int*) STBI_REALLOC_SIZED( *delays, delays_size, sizeof(int) * layers ); + if (!new_delays) + return stbi__load_gif_main_outofmem(&g, out, delays); + *delays = new_delays; + delays_size = layers * sizeof(int); + } + } else { + out = (stbi_uc*)stbi__malloc( layers * stride ); + if (!out) + return stbi__load_gif_main_outofmem(&g, out, delays); + out_size = layers * stride; + if (delays) { + *delays = (int*) stbi__malloc( layers * sizeof(int) ); + if (!*delays) + return stbi__load_gif_main_outofmem(&g, out, delays); + delays_size = layers * sizeof(int); + } + } + memcpy( out + ((layers - 1) * stride), u, stride ); + if (layers >= 2) { + two_back = out - 2 * stride; + } + + if (delays) { + (*delays)[layers - 1U] = g.delay; + } + } + } while (u != 0); + + // free temp buffer; + STBI_FREE(g.out); + STBI_FREE(g.history); + STBI_FREE(g.background); + + // do the final conversion after loading everything; + if (req_comp && req_comp != 4) + out = stbi__convert_format(out, 4, req_comp, layers * g.w, g.h); + + *z = layers; + return out; + } else { + return stbi__errpuc("not GIF", "Image was not as a gif type."); + } +} + +static void *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi_uc *u = 0; + stbi__gif g; + memset(&g, 0, sizeof(g)); + STBI_NOTUSED(ri); + + u = stbi__gif_load_next(s, &g, comp, req_comp, 0); + if (u == (stbi_uc *) s) u = 0; // end of animated gif marker + if (u) { + *x = g.w; + *y = g.h; + + // moved conversion to after successful load so that the same + // can be done for multiple frames. + if (req_comp && req_comp != 4) + u = stbi__convert_format(u, 4, req_comp, g.w, g.h); + } else if (g.out) { + // if there was an error and we allocated an image buffer, free it! + STBI_FREE(g.out); + } + + // free buffers needed for multiple frame loading; + STBI_FREE(g.history); + STBI_FREE(g.background); + + return u; +} + +static int stbi__gif_info(stbi__context *s, int *x, int *y, int *comp) +{ + return stbi__gif_info_raw(s,x,y,comp); +} +#endif + +// ************************************************************************************************* +// Radiance RGBE HDR loader +// originally by Nicolas Schulz +#ifndef STBI_NO_HDR +static int stbi__hdr_test_core(stbi__context *s, const char *signature) +{ + int i; + for (i=0; signature[i]; ++i) + if (stbi__get8(s) != signature[i]) + return 0; + stbi__rewind(s); + return 1; +} + +static int stbi__hdr_test(stbi__context* s) +{ + int r = stbi__hdr_test_core(s, "#?RADIANCE\n"); + stbi__rewind(s); + if(!r) { + r = stbi__hdr_test_core(s, "#?RGBE\n"); + stbi__rewind(s); + } + return r; +} + +#define STBI__HDR_BUFLEN 1024 +static char *stbi__hdr_gettoken(stbi__context *z, char *buffer) +{ + int len=0; + char c = '\0'; + + c = (char) stbi__get8(z); + + while (!stbi__at_eof(z) && c != '\n') { + buffer[len++] = c; + if (len == STBI__HDR_BUFLEN-1) { + // flush to end of line + while (!stbi__at_eof(z) && stbi__get8(z) != '\n') + ; + break; + } + c = (char) stbi__get8(z); + } + + buffer[len] = 0; + return buffer; +} + +static void stbi__hdr_convert(float *output, stbi_uc *input, int req_comp) +{ + if ( input[3] != 0 ) { + float f1; + // Exponent + f1 = (float) ldexp(1.0f, input[3] - (int)(128 + 8)); + if (req_comp <= 2) + output[0] = (input[0] + input[1] + input[2]) * f1 / 3; + else { + output[0] = input[0] * f1; + output[1] = input[1] * f1; + output[2] = input[2] * f1; + } + if (req_comp == 2) output[1] = 1; + if (req_comp == 4) output[3] = 1; + } else { + switch (req_comp) { + case 4: output[3] = 1; /* fallthrough */ + case 3: output[0] = output[1] = output[2] = 0; + break; + case 2: output[1] = 1; /* fallthrough */ + case 1: output[0] = 0; + break; + } + } +} + +static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + char buffer[STBI__HDR_BUFLEN]; + char *token; + int valid = 0; + int width, height; + stbi_uc *scanline; + float *hdr_data; + int len; + unsigned char count, value; + int i, j, k, c1,c2, z; + const char *headerToken; + STBI_NOTUSED(ri); + + // Check identifier + headerToken = stbi__hdr_gettoken(s,buffer); + if (strcmp(headerToken, "#?RADIANCE") != 0 && strcmp(headerToken, "#?RGBE") != 0) + return stbi__errpf("not HDR", "Corrupt HDR image"); + + // Parse header + for(;;) { + token = stbi__hdr_gettoken(s,buffer); + if (token[0] == 0) break; + if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1; + } + + if (!valid) return stbi__errpf("unsupported format", "Unsupported HDR format"); + + // Parse width and height + // can't use sscanf() if we're not using stdio! + token = stbi__hdr_gettoken(s,buffer); + if (strncmp(token, "-Y ", 3)) return stbi__errpf("unsupported data layout", "Unsupported HDR format"); + token += 3; + height = (int) strtol(token, &token, 10); + while (*token == ' ') ++token; + if (strncmp(token, "+X ", 3)) return stbi__errpf("unsupported data layout", "Unsupported HDR format"); + token += 3; + width = (int) strtol(token, NULL, 10); + + if (height > STBI_MAX_DIMENSIONS) return stbi__errpf("too large","Very large image (corrupt?)"); + if (width > STBI_MAX_DIMENSIONS) return stbi__errpf("too large","Very large image (corrupt?)"); + + *x = width; + *y = height; + + if (comp) *comp = 3; + if (req_comp == 0) req_comp = 3; + + if (!stbi__mad4sizes_valid(width, height, req_comp, sizeof(float), 0)) + return stbi__errpf("too large", "HDR image is too large"); + + // Read data + hdr_data = (float *) stbi__malloc_mad4(width, height, req_comp, sizeof(float), 0); + if (!hdr_data) + return stbi__errpf("outofmem", "Out of memory"); + + // Load image data + // image data is stored as some number of sca + if ( width < 8 || width >= 32768) { + // Read flat data + for (j=0; j < height; ++j) { + for (i=0; i < width; ++i) { + stbi_uc rgbe[4]; + main_decode_loop: + stbi__getn(s, rgbe, 4); + stbi__hdr_convert(hdr_data + j * width * req_comp + i * req_comp, rgbe, req_comp); + } + } + } else { + // Read RLE-encoded data + scanline = NULL; + + for (j = 0; j < height; ++j) { + c1 = stbi__get8(s); + c2 = stbi__get8(s); + len = stbi__get8(s); + if (c1 != 2 || c2 != 2 || (len & 0x80)) { + // not run-length encoded, so we have to actually use THIS data as a decoded + // pixel (note this can't be a valid pixel--one of RGB must be >= 128) + stbi_uc rgbe[4]; + rgbe[0] = (stbi_uc) c1; + rgbe[1] = (stbi_uc) c2; + rgbe[2] = (stbi_uc) len; + rgbe[3] = (stbi_uc) stbi__get8(s); + stbi__hdr_convert(hdr_data, rgbe, req_comp); + i = 1; + j = 0; + STBI_FREE(scanline); + goto main_decode_loop; // yes, this makes no sense + } + len <<= 8; + len |= stbi__get8(s); + if (len != width) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("invalid decoded scanline length", "corrupt HDR"); } + if (scanline == NULL) { + scanline = (stbi_uc *) stbi__malloc_mad2(width, 4, 0); + if (!scanline) { + STBI_FREE(hdr_data); + return stbi__errpf("outofmem", "Out of memory"); + } + } + + for (k = 0; k < 4; ++k) { + int nleft; + i = 0; + while ((nleft = width - i) > 0) { + count = stbi__get8(s); + if (count > 128) { + // Run + value = stbi__get8(s); + count -= 128; + if ((count == 0) || (count > nleft)) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("corrupt", "bad RLE data in HDR"); } + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = value; + } else { + // Dump + if ((count == 0) || (count > nleft)) { STBI_FREE(hdr_data); STBI_FREE(scanline); return stbi__errpf("corrupt", "bad RLE data in HDR"); } + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = stbi__get8(s); + } + } + } + for (i=0; i < width; ++i) + stbi__hdr_convert(hdr_data+(j*width + i)*req_comp, scanline + i*4, req_comp); + } + if (scanline) + STBI_FREE(scanline); + } + + return hdr_data; +} + +static int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp) +{ + char buffer[STBI__HDR_BUFLEN]; + char *token; + int valid = 0; + int dummy; + + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + + if (stbi__hdr_test(s) == 0) { + stbi__rewind( s ); + return 0; + } + + for(;;) { + token = stbi__hdr_gettoken(s,buffer); + if (token[0] == 0) break; + if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1; + } + + if (!valid) { + stbi__rewind( s ); + return 0; + } + token = stbi__hdr_gettoken(s,buffer); + if (strncmp(token, "-Y ", 3)) { + stbi__rewind( s ); + return 0; + } + token += 3; + *y = (int) strtol(token, &token, 10); + while (*token == ' ') ++token; + if (strncmp(token, "+X ", 3)) { + stbi__rewind( s ); + return 0; + } + token += 3; + *x = (int) strtol(token, NULL, 10); + *comp = 3; + return 1; +} +#endif // STBI_NO_HDR + +#ifndef STBI_NO_BMP +static int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp) +{ + void *p; + stbi__bmp_data info; + + info.all_a = 255; + p = stbi__bmp_parse_header(s, &info); + if (p == NULL) { + stbi__rewind( s ); + return 0; + } + if (x) *x = s->img_x; + if (y) *y = s->img_y; + if (comp) { + if (info.bpp == 24 && info.ma == 0xff000000) + *comp = 3; + else + *comp = info.ma ? 4 : 3; + } + return 1; +} +#endif + +#ifndef STBI_NO_PSD +static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp) +{ + int channelCount, dummy, depth; + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + if (stbi__get32be(s) != 0x38425053) { + stbi__rewind( s ); + return 0; + } + if (stbi__get16be(s) != 1) { + stbi__rewind( s ); + return 0; + } + stbi__skip(s, 6); + channelCount = stbi__get16be(s); + if (channelCount < 0 || channelCount > 16) { + stbi__rewind( s ); + return 0; + } + *y = stbi__get32be(s); + *x = stbi__get32be(s); + depth = stbi__get16be(s); + if (depth != 8 && depth != 16) { + stbi__rewind( s ); + return 0; + } + if (stbi__get16be(s) != 3) { + stbi__rewind( s ); + return 0; + } + *comp = 4; + return 1; +} + +static int stbi__psd_is16(stbi__context *s) +{ + int channelCount, depth; + if (stbi__get32be(s) != 0x38425053) { + stbi__rewind( s ); + return 0; + } + if (stbi__get16be(s) != 1) { + stbi__rewind( s ); + return 0; + } + stbi__skip(s, 6); + channelCount = stbi__get16be(s); + if (channelCount < 0 || channelCount > 16) { + stbi__rewind( s ); + return 0; + } + STBI_NOTUSED(stbi__get32be(s)); + STBI_NOTUSED(stbi__get32be(s)); + depth = stbi__get16be(s); + if (depth != 16) { + stbi__rewind( s ); + return 0; + } + return 1; +} +#endif + +#ifndef STBI_NO_PIC +static int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp) +{ + int act_comp=0,num_packets=0,chained,dummy; + stbi__pic_packet packets[10]; + + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + + if (!stbi__pic_is4(s,"\x53\x80\xF6\x34")) { + stbi__rewind(s); + return 0; + } + + stbi__skip(s, 88); + + *x = stbi__get16be(s); + *y = stbi__get16be(s); + if (stbi__at_eof(s)) { + stbi__rewind( s); + return 0; + } + if ( (*x) != 0 && (1 << 28) / (*x) < (*y)) { + stbi__rewind( s ); + return 0; + } + + stbi__skip(s, 8); + + do { + stbi__pic_packet *packet; + + if (num_packets==sizeof(packets)/sizeof(packets[0])) + return 0; + + packet = &packets[num_packets++]; + chained = stbi__get8(s); + packet->size = stbi__get8(s); + packet->type = stbi__get8(s); + packet->channel = stbi__get8(s); + act_comp |= packet->channel; + + if (stbi__at_eof(s)) { + stbi__rewind( s ); + return 0; + } + if (packet->size != 8) { + stbi__rewind( s ); + return 0; + } + } while (chained); + + *comp = (act_comp & 0x10 ? 4 : 3); + + return 1; +} +#endif + +// ************************************************************************************************* +// Portable Gray Map and Portable Pixel Map loader +// by Ken Miller +// +// PGM: http://netpbm.sourceforge.net/doc/pgm.html +// PPM: http://netpbm.sourceforge.net/doc/ppm.html +// +// Known limitations: +// Does not support comments in the header section +// Does not support ASCII image data (formats P2 and P3) + +#ifndef STBI_NO_PNM + +static int stbi__pnm_test(stbi__context *s) +{ + char p, t; + p = (char) stbi__get8(s); + t = (char) stbi__get8(s); + if (p != 'P' || (t != '5' && t != '6')) { + stbi__rewind( s ); + return 0; + } + return 1; +} + +static void *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp, stbi__result_info *ri) +{ + stbi_uc *out; + STBI_NOTUSED(ri); + + ri->bits_per_channel = stbi__pnm_info(s, (int *)&s->img_x, (int *)&s->img_y, (int *)&s->img_n); + if (ri->bits_per_channel == 0) + return 0; + + if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__errpuc("too large","Very large image (corrupt?)"); + + *x = s->img_x; + *y = s->img_y; + if (comp) *comp = s->img_n; + + if (!stbi__mad4sizes_valid(s->img_n, s->img_x, s->img_y, ri->bits_per_channel / 8, 0)) + return stbi__errpuc("too large", "PNM too large"); + + out = (stbi_uc *) stbi__malloc_mad4(s->img_n, s->img_x, s->img_y, ri->bits_per_channel / 8, 0); + if (!out) return stbi__errpuc("outofmem", "Out of memory"); + if (!stbi__getn(s, out, s->img_n * s->img_x * s->img_y * (ri->bits_per_channel / 8))) { + STBI_FREE(out); + return stbi__errpuc("bad PNM", "PNM file truncated"); + } + + if (req_comp && req_comp != s->img_n) { + if (ri->bits_per_channel == 16) { + out = (stbi_uc *) stbi__convert_format16((stbi__uint16 *) out, s->img_n, req_comp, s->img_x, s->img_y); + } else { + out = stbi__convert_format(out, s->img_n, req_comp, s->img_x, s->img_y); + } + if (out == NULL) return out; // stbi__convert_format frees input on failure + } + return out; +} + +static int stbi__pnm_isspace(char c) +{ + return c == ' ' || c == '\t' || c == '\n' || c == '\v' || c == '\f' || c == '\r'; +} + +static void stbi__pnm_skip_whitespace(stbi__context *s, char *c) +{ + for (;;) { + while (!stbi__at_eof(s) && stbi__pnm_isspace(*c)) + *c = (char) stbi__get8(s); + + if (stbi__at_eof(s) || *c != '#') + break; + + while (!stbi__at_eof(s) && *c != '\n' && *c != '\r' ) + *c = (char) stbi__get8(s); + } +} + +static int stbi__pnm_isdigit(char c) +{ + return c >= '0' && c <= '9'; +} + +static int stbi__pnm_getinteger(stbi__context *s, char *c) +{ + int value = 0; + + while (!stbi__at_eof(s) && stbi__pnm_isdigit(*c)) { + value = value*10 + (*c - '0'); + *c = (char) stbi__get8(s); + if((value > 214748364) || (value == 214748364 && *c > '7')) + return stbi__err("integer parse overflow", "Parsing an integer in the PPM header overflowed a 32-bit int"); + } + + return value; +} + +static int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp) +{ + int maxv, dummy; + char c, p, t; + + if (!x) x = &dummy; + if (!y) y = &dummy; + if (!comp) comp = &dummy; + + stbi__rewind(s); + + // Get identifier + p = (char) stbi__get8(s); + t = (char) stbi__get8(s); + if (p != 'P' || (t != '5' && t != '6')) { + stbi__rewind(s); + return 0; + } + + *comp = (t == '6') ? 3 : 1; // '5' is 1-component .pgm; '6' is 3-component .ppm + + c = (char) stbi__get8(s); + stbi__pnm_skip_whitespace(s, &c); + + *x = stbi__pnm_getinteger(s, &c); // read width + if(*x == 0) + return stbi__err("invalid width", "PPM image header had zero or overflowing width"); + stbi__pnm_skip_whitespace(s, &c); + + *y = stbi__pnm_getinteger(s, &c); // read height + if (*y == 0) + return stbi__err("invalid width", "PPM image header had zero or overflowing width"); + stbi__pnm_skip_whitespace(s, &c); + + maxv = stbi__pnm_getinteger(s, &c); // read max value + if (maxv > 65535) + return stbi__err("max value > 65535", "PPM image supports only 8-bit and 16-bit images"); + else if (maxv > 255) + return 16; + else + return 8; +} + +static int stbi__pnm_is16(stbi__context *s) +{ + if (stbi__pnm_info(s, NULL, NULL, NULL) == 16) + return 1; + return 0; +} +#endif + +static int stbi__info_main(stbi__context *s, int *x, int *y, int *comp) +{ + #ifndef STBI_NO_JPEG + if (stbi__jpeg_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PNG + if (stbi__png_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_GIF + if (stbi__gif_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_BMP + if (stbi__bmp_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PSD + if (stbi__psd_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PIC + if (stbi__pic_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_PNM + if (stbi__pnm_info(s, x, y, comp)) return 1; + #endif + + #ifndef STBI_NO_HDR + if (stbi__hdr_info(s, x, y, comp)) return 1; + #endif + + // test tga last because it's a crappy test! + #ifndef STBI_NO_TGA + if (stbi__tga_info(s, x, y, comp)) + return 1; + #endif + return stbi__err("unknown image type", "Image not of any known type, or corrupt"); +} + +static int stbi__is_16_main(stbi__context *s) +{ + #ifndef STBI_NO_PNG + if (stbi__png_is16(s)) return 1; + #endif + + #ifndef STBI_NO_PSD + if (stbi__psd_is16(s)) return 1; + #endif + + #ifndef STBI_NO_PNM + if (stbi__pnm_is16(s)) return 1; + #endif + return 0; +} + +#ifndef STBI_NO_STDIO +STBIDEF int stbi_info(char const *filename, int *x, int *y, int *comp) +{ + FILE *f = stbi__fopen(filename, "rb"); + int result; + if (!f) return stbi__err("can't fopen", "Unable to open file"); + result = stbi_info_from_file(f, x, y, comp); + fclose(f); + return result; +} + +STBIDEF int stbi_info_from_file(FILE *f, int *x, int *y, int *comp) +{ + int r; + stbi__context s; + long pos = ftell(f); + stbi__start_file(&s, f); + r = stbi__info_main(&s,x,y,comp); + fseek(f,pos,SEEK_SET); + return r; +} + +STBIDEF int stbi_is_16_bit(char const *filename) +{ + FILE *f = stbi__fopen(filename, "rb"); + int result; + if (!f) return stbi__err("can't fopen", "Unable to open file"); + result = stbi_is_16_bit_from_file(f); + fclose(f); + return result; +} + +STBIDEF int stbi_is_16_bit_from_file(FILE *f) +{ + int r; + stbi__context s; + long pos = ftell(f); + stbi__start_file(&s, f); + r = stbi__is_16_main(&s); + fseek(f,pos,SEEK_SET); + return r; +} +#endif // !STBI_NO_STDIO + +STBIDEF int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__info_main(&s,x,y,comp); +} + +STBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *c, void *user, int *x, int *y, int *comp) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user); + return stbi__info_main(&s,x,y,comp); +} + +STBIDEF int stbi_is_16_bit_from_memory(stbi_uc const *buffer, int len) +{ + stbi__context s; + stbi__start_mem(&s,buffer,len); + return stbi__is_16_main(&s); +} + +STBIDEF int stbi_is_16_bit_from_callbacks(stbi_io_callbacks const *c, void *user) +{ + stbi__context s; + stbi__start_callbacks(&s, (stbi_io_callbacks *) c, user); + return stbi__is_16_main(&s); +} + +#endif // STB_IMAGE_IMPLEMENTATION + +/* + revision history: + 2.20 (2019-02-07) support utf8 filenames in Windows; fix warnings and platform ifdefs + 2.19 (2018-02-11) fix warning + 2.18 (2018-01-30) fix warnings + 2.17 (2018-01-29) change sbti__shiftsigned to avoid clang -O2 bug + 1-bit BMP + *_is_16_bit api + avoid warnings + 2.16 (2017-07-23) all functions have 16-bit variants; + STBI_NO_STDIO works again; + compilation fixes; + fix rounding in unpremultiply; + optimize vertical flip; + disable raw_len validation; + documentation fixes + 2.15 (2017-03-18) fix png-1,2,4 bug; now all Imagenet JPGs decode; + warning fixes; disable run-time SSE detection on gcc; + uniform handling of optional "return" values; + thread-safe initialization of zlib tables + 2.14 (2017-03-03) remove deprecated STBI_JPEG_OLD; fixes for Imagenet JPGs + 2.13 (2016-11-29) add 16-bit API, only supported for PNG right now + 2.12 (2016-04-02) fix typo in 2.11 PSD fix that caused crashes + 2.11 (2016-04-02) allocate large structures on the stack + remove white matting for transparent PSD + fix reported channel count for PNG & BMP + re-enable SSE2 in non-gcc 64-bit + support RGB-formatted JPEG + read 16-bit PNGs (only as 8-bit) + 2.10 (2016-01-22) avoid warning introduced in 2.09 by STBI_REALLOC_SIZED + 2.09 (2016-01-16) allow comments in PNM files + 16-bit-per-pixel TGA (not bit-per-component) + info() for TGA could break due to .hdr handling + info() for BMP to shares code instead of sloppy parse + can use STBI_REALLOC_SIZED if allocator doesn't support realloc + code cleanup + 2.08 (2015-09-13) fix to 2.07 cleanup, reading RGB PSD as RGBA + 2.07 (2015-09-13) fix compiler warnings + partial animated GIF support + limited 16-bpc PSD support + #ifdef unused functions + bug with < 92 byte PIC,PNM,HDR,TGA + 2.06 (2015-04-19) fix bug where PSD returns wrong '*comp' value + 2.05 (2015-04-19) fix bug in progressive JPEG handling, fix warning + 2.04 (2015-04-15) try to re-enable SIMD on MinGW 64-bit + 2.03 (2015-04-12) extra corruption checking (mmozeiko) + stbi_set_flip_vertically_on_load (nguillemot) + fix NEON support; fix mingw support + 2.02 (2015-01-19) fix incorrect assert, fix warning + 2.01 (2015-01-17) fix various warnings; suppress SIMD on gcc 32-bit without -msse2 + 2.00b (2014-12-25) fix STBI_MALLOC in progressive JPEG + 2.00 (2014-12-25) optimize JPG, including x86 SSE2 & NEON SIMD (ryg) + progressive JPEG (stb) + PGM/PPM support (Ken Miller) + STBI_MALLOC,STBI_REALLOC,STBI_FREE + GIF bugfix -- seemingly never worked + STBI_NO_*, STBI_ONLY_* + 1.48 (2014-12-14) fix incorrectly-named assert() + 1.47 (2014-12-14) 1/2/4-bit PNG support, both direct and paletted (Omar Cornut & stb) + optimize PNG (ryg) + fix bug in interlaced PNG with user-specified channel count (stb) + 1.46 (2014-08-26) + fix broken tRNS chunk (colorkey-style transparency) in non-paletted PNG + 1.45 (2014-08-16) + fix MSVC-ARM internal compiler error by wrapping malloc + 1.44 (2014-08-07) + various warning fixes from Ronny Chevalier + 1.43 (2014-07-15) + fix MSVC-only compiler problem in code changed in 1.42 + 1.42 (2014-07-09) + don't define _CRT_SECURE_NO_WARNINGS (affects user code) + fixes to stbi__cleanup_jpeg path + added STBI_ASSERT to avoid requiring assert.h + 1.41 (2014-06-25) + fix search&replace from 1.36 that messed up comments/error messages + 1.40 (2014-06-22) + fix gcc struct-initialization warning + 1.39 (2014-06-15) + fix to TGA optimization when req_comp != number of components in TGA; + fix to GIF loading because BMP wasn't rewinding (whoops, no GIFs in my test suite) + add support for BMP version 5 (more ignored fields) + 1.38 (2014-06-06) + suppress MSVC warnings on integer casts truncating values + fix accidental rename of 'skip' field of I/O + 1.37 (2014-06-04) + remove duplicate typedef + 1.36 (2014-06-03) + convert to header file single-file library + if de-iphone isn't set, load iphone images color-swapped instead of returning NULL + 1.35 (2014-05-27) + various warnings + fix broken STBI_SIMD path + fix bug where stbi_load_from_file no longer left file pointer in correct place + fix broken non-easy path for 32-bit BMP (possibly never used) + TGA optimization by Arseny Kapoulkine + 1.34 (unknown) + use STBI_NOTUSED in stbi__resample_row_generic(), fix one more leak in tga failure case + 1.33 (2011-07-14) + make stbi_is_hdr work in STBI_NO_HDR (as specified), minor compiler-friendly improvements + 1.32 (2011-07-13) + support for "info" function for all supported filetypes (SpartanJ) + 1.31 (2011-06-20) + a few more leak fixes, bug in PNG handling (SpartanJ) + 1.30 (2011-06-11) + added ability to load files via callbacks to accomidate custom input streams (Ben Wenger) + removed deprecated format-specific test/load functions + removed support for installable file formats (stbi_loader) -- would have been broken for IO callbacks anyway + error cases in bmp and tga give messages and don't leak (Raymond Barbiero, grisha) + fix inefficiency in decoding 32-bit BMP (David Woo) + 1.29 (2010-08-16) + various warning fixes from Aurelien Pocheville + 1.28 (2010-08-01) + fix bug in GIF palette transparency (SpartanJ) + 1.27 (2010-08-01) + cast-to-stbi_uc to fix warnings + 1.26 (2010-07-24) + fix bug in file buffering for PNG reported by SpartanJ + 1.25 (2010-07-17) + refix trans_data warning (Won Chun) + 1.24 (2010-07-12) + perf improvements reading from files on platforms with lock-heavy fgetc() + minor perf improvements for jpeg + deprecated type-specific functions so we'll get feedback if they're needed + attempt to fix trans_data warning (Won Chun) + 1.23 fixed bug in iPhone support + 1.22 (2010-07-10) + removed image *writing* support + stbi_info support from Jetro Lauha + GIF support from Jean-Marc Lienher + iPhone PNG-extensions from James Brown + warning-fixes from Nicolas Schulz and Janez Zemva (i.stbi__err. Janez (U+017D)emva) + 1.21 fix use of 'stbi_uc' in header (reported by jon blow) + 1.20 added support for Softimage PIC, by Tom Seddon + 1.19 bug in interlaced PNG corruption check (found by ryg) + 1.18 (2008-08-02) + fix a threading bug (local mutable static) + 1.17 support interlaced PNG + 1.16 major bugfix - stbi__convert_format converted one too many pixels + 1.15 initialize some fields for thread safety + 1.14 fix threadsafe conversion bug + header-file-only version (#define STBI_HEADER_FILE_ONLY before including) + 1.13 threadsafe + 1.12 const qualifiers in the API + 1.11 Support installable IDCT, colorspace conversion routines + 1.10 Fixes for 64-bit (don't use "unsigned long") + optimized upsampling by Fabian "ryg" Giesen + 1.09 Fix format-conversion for PSD code (bad global variables!) + 1.08 Thatcher Ulrich's PSD code integrated by Nicolas Schulz + 1.07 attempt to fix C++ warning/errors again + 1.06 attempt to fix C++ warning/errors again + 1.05 fix TGA loading to return correct *comp and use good luminance calc + 1.04 default float alpha is 1, not 255; use 'void *' for stbi_image_free + 1.03 bugfixes to STBI_NO_STDIO, STBI_NO_HDR + 1.02 support for (subset of) HDR files, float interface for preferred access to them + 1.01 fix bug: possible bug in handling right-side up bmps... not sure + fix bug: the stbi__bmp_load() and stbi__tga_load() functions didn't work at all + 1.00 interface to zlib that skips zlib header + 0.99 correct handling of alpha in palette + 0.98 TGA loader by lonesock; dynamically add loaders (untested) + 0.97 jpeg errors on too large a file; also catch another malloc failure + 0.96 fix detection of invalid v value - particleman@mollyrocket forum + 0.95 during header scan, seek to markers in case of padding + 0.94 STBI_NO_STDIO to disable stdio usage; rename all #defines the same + 0.93 handle jpegtran output; verbose errors + 0.92 read 4,8,16,24,32-bit BMP files of several formats + 0.91 output 24-bit Windows 3.0 BMP files + 0.90 fix a few more warnings; bump version number to approach 1.0 + 0.61 bugfixes due to Marc LeBlanc, Christopher Lloyd + 0.60 fix compiling as c++ + 0.59 fix warnings: merge Dave Moore's -Wall fixes + 0.58 fix bug: zlib uncompressed mode len/nlen was wrong endian + 0.57 fix bug: jpg last huffman symbol before marker was >9 bits but less than 16 available + 0.56 fix bug: zlib uncompressed mode len vs. nlen + 0.55 fix bug: restart_interval not initialized to 0 + 0.54 allow NULL for 'int *comp' + 0.53 fix bug in png 3->4; speedup png decoding + 0.52 png handles req_comp=3,4 directly; minor cleanup; jpeg comments + 0.51 obey req_comp requests, 1-component jpegs return as 1-component, + on 'test' only check type, not whether we support this variant + 0.50 (2006-11-19) + first released version +*/ + + +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2017 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/semicongine/thirdparty/stb/stb_truetype.h Mon Jul 29 00:00:28 2024 +0700 @@ -0,0 +1,5077 @@ +// stb_truetype.h - v1.26 - public domain +// authored from 2009-2021 by Sean Barrett / RAD Game Tools +// +// ======================================================================= +// +// NO SECURITY GUARANTEE -- DO NOT USE THIS ON UNTRUSTED FONT FILES +// +// This library does no range checking of the offsets found in the file, +// meaning an attacker can use it to read arbitrary memory. +// +// ======================================================================= +// +// This library processes TrueType files: +// parse files +// extract glyph metrics +// extract glyph shapes +// render glyphs to one-channel bitmaps with antialiasing (box filter) +// render glyphs to one-channel SDF bitmaps (signed-distance field/function) +// +// Todo: +// non-MS cmaps +// crashproof on bad data +// hinting? (no longer patented) +// cleartype-style AA? +// optimize: use simple memory allocator for intermediates +// optimize: build edge-list directly from curves +// optimize: rasterize directly from curves? +// +// ADDITIONAL CONTRIBUTORS +// +// Mikko Mononen: compound shape support, more cmap formats +// Tor Andersson: kerning, subpixel rendering +// Dougall Johnson: OpenType / Type 2 font handling +// Daniel Ribeiro Maciel: basic GPOS-based kerning +// +// Misc other: +// Ryan Gordon +// Simon Glass +// github:IntellectualKitty +// Imanol Celaya +// Daniel Ribeiro Maciel +// +// Bug/warning reports/fixes: +// "Zer" on mollyrocket Fabian "ryg" Giesen github:NiLuJe +// Cass Everitt Martins Mozeiko github:aloucks +// stoiko (Haemimont Games) Cap Petschulat github:oyvindjam +// Brian Hook Omar Cornut github:vassvik +// Walter van Niftrik Ryan Griege +// David Gow Peter LaValle +// David Given Sergey Popov +// Ivan-Assen Ivanov Giumo X. Clanjor +// Anthony Pesch Higor Euripedes +// Johan Duparc Thomas Fields +// Hou Qiming Derek Vinyard +// Rob Loach Cort Stratton +// Kenney Phillis Jr. Brian Costabile +// Ken Voskuil (kaesve) +// +// VERSION HISTORY +// +// 1.26 (2021-08-28) fix broken rasterizer +// 1.25 (2021-07-11) many fixes +// 1.24 (2020-02-05) fix warning +// 1.23 (2020-02-02) query SVG data for glyphs; query whole kerning table (but only kern not GPOS) +// 1.22 (2019-08-11) minimize missing-glyph duplication; fix kerning if both 'GPOS' and 'kern' are defined +// 1.21 (2019-02-25) fix warning +// 1.20 (2019-02-07) PackFontRange skips missing codepoints; GetScaleFontVMetrics() +// 1.19 (2018-02-11) GPOS kerning, STBTT_fmod +// 1.18 (2018-01-29) add missing function +// 1.17 (2017-07-23) make more arguments const; doc fix +// 1.16 (2017-07-12) SDF support +// 1.15 (2017-03-03) make more arguments const +// 1.14 (2017-01-16) num-fonts-in-TTC function +// 1.13 (2017-01-02) support OpenType fonts, certain Apple fonts +// 1.12 (2016-10-25) suppress warnings about casting away const with -Wcast-qual +// 1.11 (2016-04-02) fix unused-variable warning +// 1.10 (2016-04-02) user-defined fabs(); rare memory leak; remove duplicate typedef +// 1.09 (2016-01-16) warning fix; avoid crash on outofmem; use allocation userdata properly +// 1.08 (2015-09-13) document stbtt_Rasterize(); fixes for vertical & horizontal edges +// 1.07 (2015-08-01) allow PackFontRanges to accept arrays of sparse codepoints; +// variant PackFontRanges to pack and render in separate phases; +// fix stbtt_GetFontOFfsetForIndex (never worked for non-0 input?); +// fixed an assert() bug in the new rasterizer +// replace assert() with STBTT_assert() in new rasterizer +// +// Full history can be found at the end of this file. +// +// LICENSE +// +// See end of file for license information. +// +// USAGE +// +// Include this file in whatever places need to refer to it. In ONE C/C++ +// file, write: +// #define STB_TRUETYPE_IMPLEMENTATION +// before the #include of this file. This expands out the actual +// implementation into that C/C++ file. +// +// To make the implementation private to the file that generates the implementation, +// #define STBTT_STATIC +// +// Simple 3D API (don't ship this, but it's fine for tools and quick start) +// stbtt_BakeFontBitmap() -- bake a font to a bitmap for use as texture +// stbtt_GetBakedQuad() -- compute quad to draw for a given char +// +// Improved 3D API (more shippable): +// #include "stb_rect_pack.h" -- optional, but you really want it +// stbtt_PackBegin() +// stbtt_PackSetOversampling() -- for improved quality on small fonts +// stbtt_PackFontRanges() -- pack and renders +// stbtt_PackEnd() +// stbtt_GetPackedQuad() +// +// "Load" a font file from a memory buffer (you have to keep the buffer loaded) +// stbtt_InitFont() +// stbtt_GetFontOffsetForIndex() -- indexing for TTC font collections +// stbtt_GetNumberOfFonts() -- number of fonts for TTC font collections +// +// Render a unicode codepoint to a bitmap +// stbtt_GetCodepointBitmap() -- allocates and returns a bitmap +// stbtt_MakeCodepointBitmap() -- renders into bitmap you provide +// stbtt_GetCodepointBitmapBox() -- how big the bitmap must be +// +// Character advance/positioning +// stbtt_GetCodepointHMetrics() +// stbtt_GetFontVMetrics() +// stbtt_GetFontVMetricsOS2() +// stbtt_GetCodepointKernAdvance() +// +// Starting with version 1.06, the rasterizer was replaced with a new, +// faster and generally-more-precise rasterizer. The new rasterizer more +// accurately measures pixel coverage for anti-aliasing, except in the case +// where multiple shapes overlap, in which case it overestimates the AA pixel +// coverage. Thus, anti-aliasing of intersecting shapes may look wrong. If +// this turns out to be a problem, you can re-enable the old rasterizer with +// #define STBTT_RASTERIZER_VERSION 1 +// which will incur about a 15% speed hit. +// +// ADDITIONAL DOCUMENTATION +// +// Immediately after this block comment are a series of sample programs. +// +// After the sample programs is the "header file" section. This section +// includes documentation for each API function. +// +// Some important concepts to understand to use this library: +// +// Codepoint +// Characters are defined by unicode codepoints, e.g. 65 is +// uppercase A, 231 is lowercase c with a cedilla, 0x7e30 is +// the hiragana for "ma". +// +// Glyph +// A visual character shape (every codepoint is rendered as +// some glyph) +// +// Glyph index +// A font-specific integer ID representing a glyph +// +// Baseline +// Glyph shapes are defined relative to a baseline, which is the +// bottom of uppercase characters. Characters extend both above +// and below the baseline. +// +// Current Point +// As you draw text to the screen, you keep track of a "current point" +// which is the origin of each character. The current point's vertical +// position is the baseline. Even "baked fonts" use this model. +// +// Vertical Font Metrics +// The vertical qualities of the font, used to vertically position +// and space the characters. See docs for stbtt_GetFontVMetrics. +// +// Font Size in Pixels or Points +// The preferred interface for specifying font sizes in stb_truetype +// is to specify how tall the font's vertical extent should be in pixels. +// If that sounds good enough, skip the next paragraph. +// +// Most font APIs instead use "points", which are a common typographic +// measurement for describing font size, defined as 72 points per inch. +// stb_truetype provides a point API for compatibility. However, true +// "per inch" conventions don't make much sense on computer displays +// since different monitors have different number of pixels per +// inch. For example, Windows traditionally uses a convention that +// there are 96 pixels per inch, thus making 'inch' measurements have +// nothing to do with inches, and thus effectively defining a point to +// be 1.333 pixels. Additionally, the TrueType font data provides +// an explicit scale factor to scale a given font's glyphs to points, +// but the author has observed that this scale factor is often wrong +// for non-commercial fonts, thus making fonts scaled in points +// according to the TrueType spec incoherently sized in practice. +// +// DETAILED USAGE: +// +// Scale: +// Select how high you want the font to be, in points or pixels. +// Call ScaleForPixelHeight or ScaleForMappingEmToPixels to compute +// a scale factor SF that will be used by all other functions. +// +// Baseline: +// You need to select a y-coordinate that is the baseline of where +// your text will appear. Call GetFontBoundingBox to get the baseline-relative +// bounding box for all characters. SF*-y0 will be the distance in pixels +// that the worst-case character could extend above the baseline, so if +// you want the top edge of characters to appear at the top of the +// screen where y=0, then you would set the baseline to SF*-y0. +// +// Current point: +// Set the current point where the first character will appear. The +// first character could extend left of the current point; this is font +// dependent. You can either choose a current point that is the leftmost +// point and hope, or add some padding, or check the bounding box or +// left-side-bearing of the first character to be displayed and set +// the current point based on that. +// +// Displaying a character: +// Compute the bounding box of the character. It will contain signed values +// relative to <current_point, baseline>. I.e. if it returns x0,y0,x1,y1, +// then the character should be displayed in the rectangle from +// <current_point+SF*x0, baseline+SF*y0> to <current_point+SF*x1,baseline+SF*y1). +// +// Advancing for the next character: +// Call GlyphHMetrics, and compute 'current_point += SF * advance'. +// +// +// ADVANCED USAGE +// +// Quality: +// +// - Use the functions with Subpixel at the end to allow your characters +// to have subpixel positioning. Since the font is anti-aliased, not +// hinted, this is very import for quality. (This is not possible with +// baked fonts.) +// +// - Kerning is now supported, and if you're supporting subpixel rendering +// then kerning is worth using to give your text a polished look. +// +// Performance: +// +// - Convert Unicode codepoints to glyph indexes and operate on the glyphs; +// if you don't do this, stb_truetype is forced to do the conversion on +// every call. +// +// - There are a lot of memory allocations. We should modify it to take +// a temp buffer and allocate from the temp buffer (without freeing), +// should help performance a lot. +// +// NOTES +// +// The system uses the raw data found in the .ttf file without changing it +// and without building auxiliary data structures. This is a bit inefficient +// on little-endian systems (the data is big-endian), but assuming you're +// caching the bitmaps or glyph shapes this shouldn't be a big deal. +// +// It appears to be very hard to programmatically determine what font a +// given file is in a general way. I provide an API for this, but I don't +// recommend it. +// +// +// PERFORMANCE MEASUREMENTS FOR 1.06: +// +// 32-bit 64-bit +// Previous release: 8.83 s 7.68 s +// Pool allocations: 7.72 s 6.34 s +// Inline sort : 6.54 s 5.65 s +// New rasterizer : 5.63 s 5.00 s + +////////////////////////////////////////////////////////////////////////////// +////////////////////////////////////////////////////////////////////////////// +//// +//// SAMPLE PROGRAMS +//// +// +// Incomplete text-in-3d-api example, which draws quads properly aligned to be lossless. +// See "tests/truetype_demo_win32.c" for a complete version. +#if 0 +#define STB_TRUETYPE_IMPLEMENTATION // force following include to generate implementation +#include "stb_truetype.h" + +unsigned char ttf_buffer[1<<20]; +unsigned char temp_bitmap[512*512]; + +stbtt_bakedchar cdata[96]; // ASCII 32..126 is 95 glyphs +GLuint ftex; + +void my_stbtt_initfont(void) +{ + fread(ttf_buffer, 1, 1<<20, fopen("c:/windows/fonts/times.ttf", "rb")); + stbtt_BakeFontBitmap(ttf_buffer,0, 32.0, temp_bitmap,512,512, 32,96, cdata); // no guarantee this fits! + // can free ttf_buffer at this point + glGenTextures(1, &ftex); + glBindTexture(GL_TEXTURE_2D, ftex); + glTexImage2D(GL_TEXTURE_2D, 0, GL_ALPHA, 512,512, 0, GL_ALPHA, GL_UNSIGNED_BYTE, temp_bitmap); + // can free temp_bitmap at this point + glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR); +} + +void my_stbtt_print(float x, float y, char *text) +{ + // assume orthographic projection with units = screen pixels, origin at top left + glEnable(GL_BLEND); + glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA); + glEnable(GL_TEXTURE_2D); + glBindTexture(GL_TEXTURE_2D, ftex); + glBegin(GL_QUADS); + while (*text) { + if (*text >= 32 && *text < 128) { + stbtt_aligned_quad q; + stbtt_GetBakedQuad(cdata, 512,512, *text-32, &x,&y,&q,1);//1=opengl & d3d10+,0=d3d9 + glTexCoord2f(q.s0,q.t0); glVertex2f(q.x0,q.y0); + glTexCoord2f(q.s1,q.t0); glVertex2f(q.x1,q.y0); + glTexCoord2f(q.s1,q.t1); glVertex2f(q.x1,q.y1); + glTexCoord2f(q.s0,q.t1); glVertex2f(q.x0,q.y1); + } + ++text; + } + glEnd(); +} +#endif +// +// +////////////////////////////////////////////////////////////////////////////// +// +// Complete program (this compiles): get a single bitmap, print as ASCII art +// +#if 0 +#include <stdio.h> +#define STB_TRUETYPE_IMPLEMENTATION // force following include to generate implementation +#include "stb_truetype.h" + +char ttf_buffer[1<<25]; + +int main(int argc, char **argv) +{ + stbtt_fontinfo font; + unsigned char *bitmap; + int w,h,i,j,c = (argc > 1 ? atoi(argv[1]) : 'a'), s = (argc > 2 ? atoi(argv[2]) : 20); + + fread(ttf_buffer, 1, 1<<25, fopen(argc > 3 ? argv[3] : "c:/windows/fonts/arialbd.ttf", "rb")); + + stbtt_InitFont(&font, ttf_buffer, stbtt_GetFontOffsetForIndex(ttf_buffer,0)); + bitmap = stbtt_GetCodepointBitmap(&font, 0,stbtt_ScaleForPixelHeight(&font, s), c, &w, &h, 0,0); + + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) + putchar(" .:ioVM@"[bitmap[j*w+i]>>5]); + putchar('\n'); + } + return 0; +} +#endif +// +// Output: +// +// .ii. +// @@@@@@. +// V@Mio@@o +// :i. V@V +// :oM@@M +// :@@@MM@M +// @@o o@M +// :@@. M@M +// @@@o@@@@ +// :M@@V:@@. +// +////////////////////////////////////////////////////////////////////////////// +// +// Complete program: print "Hello World!" banner, with bugs +// +#if 0 +char buffer[24<<20]; +unsigned char screen[20][79]; + +int main(int arg, char **argv) +{ + stbtt_fontinfo font; + int i,j,ascent,baseline,ch=0; + float scale, xpos=2; // leave a little padding in case the character extends left + char *text = "Heljo World!"; // intentionally misspelled to show 'lj' brokenness + + fread(buffer, 1, 1000000, fopen("c:/windows/fonts/arialbd.ttf", "rb")); + stbtt_InitFont(&font, buffer, 0); + + scale = stbtt_ScaleForPixelHeight(&font, 15); + stbtt_GetFontVMetrics(&font, &ascent,0,0); + baseline = (int) (ascent*scale); + + while (text[ch]) { + int advance,lsb,x0,y0,x1,y1; + float x_shift = xpos - (float) floor(xpos); + stbtt_GetCodepointHMetrics(&font, text[ch], &advance, &lsb); + stbtt_GetCodepointBitmapBoxSubpixel(&font, text[ch], scale,scale,x_shift,0, &x0,&y0,&x1,&y1); + stbtt_MakeCodepointBitmapSubpixel(&font, &screen[baseline + y0][(int) xpos + x0], x1-x0,y1-y0, 79, scale,scale,x_shift,0, text[ch]); + // note that this stomps the old data, so where character boxes overlap (e.g. 'lj') it's wrong + // because this API is really for baking character bitmaps into textures. if you want to render + // a sequence of characters, you really need to render each bitmap to a temp buffer, then + // "alpha blend" that into the working buffer + xpos += (advance * scale); + if (text[ch+1]) + xpos += scale*stbtt_GetCodepointKernAdvance(&font, text[ch],text[ch+1]); + ++ch; + } + + for (j=0; j < 20; ++j) { + for (i=0; i < 78; ++i) + putchar(" .:ioVM@"[screen[j][i]>>5]); + putchar('\n'); + } + + return 0; +} +#endif + + +////////////////////////////////////////////////////////////////////////////// +////////////////////////////////////////////////////////////////////////////// +//// +//// INTEGRATION WITH YOUR CODEBASE +//// +//// The following sections allow you to supply alternate definitions +//// of C library functions used by stb_truetype, e.g. if you don't +//// link with the C runtime library. + +#ifdef STB_TRUETYPE_IMPLEMENTATION + // #define your own (u)stbtt_int8/16/32 before including to override this + #ifndef stbtt_uint8 + typedef unsigned char stbtt_uint8; + typedef signed char stbtt_int8; + typedef unsigned short stbtt_uint16; + typedef signed short stbtt_int16; + typedef unsigned int stbtt_uint32; + typedef signed int stbtt_int32; + #endif + + typedef char stbtt__check_size32[sizeof(stbtt_int32)==4 ? 1 : -1]; + typedef char stbtt__check_size16[sizeof(stbtt_int16)==2 ? 1 : -1]; + + // e.g. #define your own STBTT_ifloor/STBTT_iceil() to avoid math.h + #ifndef STBTT_ifloor + #include <math.h> + #define STBTT_ifloor(x) ((int) floor(x)) + #define STBTT_iceil(x) ((int) ceil(x)) + #endif + + #ifndef STBTT_sqrt + #include <math.h> + #define STBTT_sqrt(x) sqrt(x) + #define STBTT_pow(x,y) pow(x,y) + #endif + + #ifndef STBTT_fmod + #include <math.h> + #define STBTT_fmod(x,y) fmod(x,y) + #endif + + #ifndef STBTT_cos + #include <math.h> + #define STBTT_cos(x) cos(x) + #define STBTT_acos(x) acos(x) + #endif + + #ifndef STBTT_fabs + #include <math.h> + #define STBTT_fabs(x) fabs(x) + #endif + + // #define your own functions "STBTT_malloc" / "STBTT_free" to avoid malloc.h + #ifndef STBTT_malloc + #include <stdlib.h> + #define STBTT_malloc(x,u) ((void)(u),malloc(x)) + #define STBTT_free(x,u) ((void)(u),free(x)) + #endif + + #ifndef STBTT_assert + #include <assert.h> + #define STBTT_assert(x) assert(x) + #endif + + #ifndef STBTT_strlen + #include <string.h> + #define STBTT_strlen(x) strlen(x) + #endif + + #ifndef STBTT_memcpy + #include <string.h> + #define STBTT_memcpy memcpy + #define STBTT_memset memset + #endif +#endif + +/////////////////////////////////////////////////////////////////////////////// +/////////////////////////////////////////////////////////////////////////////// +//// +//// INTERFACE +//// +//// + +#ifndef __STB_INCLUDE_STB_TRUETYPE_H__ +#define __STB_INCLUDE_STB_TRUETYPE_H__ + +#ifdef STBTT_STATIC +#define STBTT_DEF static +#else +#define STBTT_DEF extern +#endif + +#ifdef __cplusplus +extern "C" { +#endif + +// private structure +typedef struct +{ + unsigned char *data; + int cursor; + int size; +} stbtt__buf; + +////////////////////////////////////////////////////////////////////////////// +// +// TEXTURE BAKING API +// +// If you use this API, you only have to call two functions ever. +// + +typedef struct +{ + unsigned short x0,y0,x1,y1; // coordinates of bbox in bitmap + float xoff,yoff,xadvance; +} stbtt_bakedchar; + +STBTT_DEF int stbtt_BakeFontBitmap(const unsigned char *data, int offset, // font location (use offset=0 for plain .ttf) + float pixel_height, // height of font in pixels + unsigned char *pixels, int pw, int ph, // bitmap to be filled in + int first_char, int num_chars, // characters to bake + stbtt_bakedchar *chardata); // you allocate this, it's num_chars long +// if return is positive, the first unused row of the bitmap +// if return is negative, returns the negative of the number of characters that fit +// if return is 0, no characters fit and no rows were used +// This uses a very crappy packing. + +typedef struct +{ + float x0,y0,s0,t0; // top-left + float x1,y1,s1,t1; // bottom-right +} stbtt_aligned_quad; + +STBTT_DEF void stbtt_GetBakedQuad(const stbtt_bakedchar *chardata, int pw, int ph, // same data as above + int char_index, // character to display + float *xpos, float *ypos, // pointers to current position in screen pixel space + stbtt_aligned_quad *q, // output: quad to draw + int opengl_fillrule); // true if opengl fill rule; false if DX9 or earlier +// Call GetBakedQuad with char_index = 'character - first_char', and it +// creates the quad you need to draw and advances the current position. +// +// The coordinate system used assumes y increases downwards. +// +// Characters will extend both above and below the current position; +// see discussion of "BASELINE" above. +// +// It's inefficient; you might want to c&p it and optimize it. + +STBTT_DEF void stbtt_GetScaledFontVMetrics(const unsigned char *fontdata, int index, float size, float *ascent, float *descent, float *lineGap); +// Query the font vertical metrics without having to create a font first. + + +////////////////////////////////////////////////////////////////////////////// +// +// NEW TEXTURE BAKING API +// +// This provides options for packing multiple fonts into one atlas, not +// perfectly but better than nothing. + +typedef struct +{ + unsigned short x0,y0,x1,y1; // coordinates of bbox in bitmap + float xoff,yoff,xadvance; + float xoff2,yoff2; +} stbtt_packedchar; + +typedef struct stbtt_pack_context stbtt_pack_context; +typedef struct stbtt_fontinfo stbtt_fontinfo; +#ifndef STB_RECT_PACK_VERSION +typedef struct stbrp_rect stbrp_rect; +#endif + +STBTT_DEF int stbtt_PackBegin(stbtt_pack_context *spc, unsigned char *pixels, int width, int height, int stride_in_bytes, int padding, void *alloc_context); +// Initializes a packing context stored in the passed-in stbtt_pack_context. +// Future calls using this context will pack characters into the bitmap passed +// in here: a 1-channel bitmap that is width * height. stride_in_bytes is +// the distance from one row to the next (or 0 to mean they are packed tightly +// together). "padding" is the amount of padding to leave between each +// character (normally you want '1' for bitmaps you'll use as textures with +// bilinear filtering). +// +// Returns 0 on failure, 1 on success. + +STBTT_DEF void stbtt_PackEnd (stbtt_pack_context *spc); +// Cleans up the packing context and frees all memory. + +#define STBTT_POINT_SIZE(x) (-(x)) + +STBTT_DEF int stbtt_PackFontRange(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, float font_size, + int first_unicode_char_in_range, int num_chars_in_range, stbtt_packedchar *chardata_for_range); +// Creates character bitmaps from the font_index'th font found in fontdata (use +// font_index=0 if you don't know what that is). It creates num_chars_in_range +// bitmaps for characters with unicode values starting at first_unicode_char_in_range +// and increasing. Data for how to render them is stored in chardata_for_range; +// pass these to stbtt_GetPackedQuad to get back renderable quads. +// +// font_size is the full height of the character from ascender to descender, +// as computed by stbtt_ScaleForPixelHeight. To use a point size as computed +// by stbtt_ScaleForMappingEmToPixels, wrap the point size in STBTT_POINT_SIZE() +// and pass that result as 'font_size': +// ..., 20 , ... // font max minus min y is 20 pixels tall +// ..., STBTT_POINT_SIZE(20), ... // 'M' is 20 pixels tall + +typedef struct +{ + float font_size; + int first_unicode_codepoint_in_range; // if non-zero, then the chars are continuous, and this is the first codepoint + int *array_of_unicode_codepoints; // if non-zero, then this is an array of unicode codepoints + int num_chars; + stbtt_packedchar *chardata_for_range; // output + unsigned char h_oversample, v_oversample; // don't set these, they're used internally +} stbtt_pack_range; + +STBTT_DEF int stbtt_PackFontRanges(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, stbtt_pack_range *ranges, int num_ranges); +// Creates character bitmaps from multiple ranges of characters stored in +// ranges. This will usually create a better-packed bitmap than multiple +// calls to stbtt_PackFontRange. Note that you can call this multiple +// times within a single PackBegin/PackEnd. + +STBTT_DEF void stbtt_PackSetOversampling(stbtt_pack_context *spc, unsigned int h_oversample, unsigned int v_oversample); +// Oversampling a font increases the quality by allowing higher-quality subpixel +// positioning, and is especially valuable at smaller text sizes. +// +// This function sets the amount of oversampling for all following calls to +// stbtt_PackFontRange(s) or stbtt_PackFontRangesGatherRects for a given +// pack context. The default (no oversampling) is achieved by h_oversample=1 +// and v_oversample=1. The total number of pixels required is +// h_oversample*v_oversample larger than the default; for example, 2x2 +// oversampling requires 4x the storage of 1x1. For best results, render +// oversampled textures with bilinear filtering. Look at the readme in +// stb/tests/oversample for information about oversampled fonts +// +// To use with PackFontRangesGather etc., you must set it before calls +// call to PackFontRangesGatherRects. + +STBTT_DEF void stbtt_PackSetSkipMissingCodepoints(stbtt_pack_context *spc, int skip); +// If skip != 0, this tells stb_truetype to skip any codepoints for which +// there is no corresponding glyph. If skip=0, which is the default, then +// codepoints without a glyph recived the font's "missing character" glyph, +// typically an empty box by convention. + +STBTT_DEF void stbtt_GetPackedQuad(const stbtt_packedchar *chardata, int pw, int ph, // same data as above + int char_index, // character to display + float *xpos, float *ypos, // pointers to current position in screen pixel space + stbtt_aligned_quad *q, // output: quad to draw + int align_to_integer); + +STBTT_DEF int stbtt_PackFontRangesGatherRects(stbtt_pack_context *spc, const stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects); +STBTT_DEF void stbtt_PackFontRangesPackRects(stbtt_pack_context *spc, stbrp_rect *rects, int num_rects); +STBTT_DEF int stbtt_PackFontRangesRenderIntoRects(stbtt_pack_context *spc, const stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects); +// Calling these functions in sequence is roughly equivalent to calling +// stbtt_PackFontRanges(). If you more control over the packing of multiple +// fonts, or if you want to pack custom data into a font texture, take a look +// at the source to of stbtt_PackFontRanges() and create a custom version +// using these functions, e.g. call GatherRects multiple times, +// building up a single array of rects, then call PackRects once, +// then call RenderIntoRects repeatedly. This may result in a +// better packing than calling PackFontRanges multiple times +// (or it may not). + +// this is an opaque structure that you shouldn't mess with which holds +// all the context needed from PackBegin to PackEnd. +struct stbtt_pack_context { + void *user_allocator_context; + void *pack_info; + int width; + int height; + int stride_in_bytes; + int padding; + int skip_missing; + unsigned int h_oversample, v_oversample; + unsigned char *pixels; + void *nodes; +}; + +////////////////////////////////////////////////////////////////////////////// +// +// FONT LOADING +// +// + +STBTT_DEF int stbtt_GetNumberOfFonts(const unsigned char *data); +// This function will determine the number of fonts in a font file. TrueType +// collection (.ttc) files may contain multiple fonts, while TrueType font +// (.ttf) files only contain one font. The number of fonts can be used for +// indexing with the previous function where the index is between zero and one +// less than the total fonts. If an error occurs, -1 is returned. + +STBTT_DEF int stbtt_GetFontOffsetForIndex(const unsigned char *data, int index); +// Each .ttf/.ttc file may have more than one font. Each font has a sequential +// index number starting from 0. Call this function to get the font offset for +// a given index; it returns -1 if the index is out of range. A regular .ttf +// file will only define one font and it always be at offset 0, so it will +// return '0' for index 0, and -1 for all other indices. + +// The following structure is defined publicly so you can declare one on +// the stack or as a global or etc, but you should treat it as opaque. +struct stbtt_fontinfo +{ + void * userdata; + unsigned char * data; // pointer to .ttf file + int fontstart; // offset of start of font + + int numGlyphs; // number of glyphs, needed for range checking + + int loca,head,glyf,hhea,hmtx,kern,gpos,svg; // table locations as offset from start of .ttf + int index_map; // a cmap mapping for our chosen character encoding + int indexToLocFormat; // format needed to map from glyph index to glyph + + stbtt__buf cff; // cff font data + stbtt__buf charstrings; // the charstring index + stbtt__buf gsubrs; // global charstring subroutines index + stbtt__buf subrs; // private charstring subroutines index + stbtt__buf fontdicts; // array of font dicts + stbtt__buf fdselect; // map from glyph to fontdict +}; + +STBTT_DEF int stbtt_InitFont(stbtt_fontinfo *info, const unsigned char *data, int offset); +// Given an offset into the file that defines a font, this function builds +// the necessary cached info for the rest of the system. You must allocate +// the stbtt_fontinfo yourself, and stbtt_InitFont will fill it out. You don't +// need to do anything special to free it, because the contents are pure +// value data with no additional data structures. Returns 0 on failure. + + +////////////////////////////////////////////////////////////////////////////// +// +// CHARACTER TO GLYPH-INDEX CONVERSIOn + +STBTT_DEF int stbtt_FindGlyphIndex(const stbtt_fontinfo *info, int unicode_codepoint); +// If you're going to perform multiple operations on the same character +// and you want a speed-up, call this function with the character you're +// going to process, then use glyph-based functions instead of the +// codepoint-based functions. +// Returns 0 if the character codepoint is not defined in the font. + + +////////////////////////////////////////////////////////////////////////////// +// +// CHARACTER PROPERTIES +// + +STBTT_DEF float stbtt_ScaleForPixelHeight(const stbtt_fontinfo *info, float pixels); +// computes a scale factor to produce a font whose "height" is 'pixels' tall. +// Height is measured as the distance from the highest ascender to the lowest +// descender; in other words, it's equivalent to calling stbtt_GetFontVMetrics +// and computing: +// scale = pixels / (ascent - descent) +// so if you prefer to measure height by the ascent only, use a similar calculation. + +STBTT_DEF float stbtt_ScaleForMappingEmToPixels(const stbtt_fontinfo *info, float pixels); +// computes a scale factor to produce a font whose EM size is mapped to +// 'pixels' tall. This is probably what traditional APIs compute, but +// I'm not positive. + +STBTT_DEF void stbtt_GetFontVMetrics(const stbtt_fontinfo *info, int *ascent, int *descent, int *lineGap); +// ascent is the coordinate above the baseline the font extends; descent +// is the coordinate below the baseline the font extends (i.e. it is typically negative) +// lineGap is the spacing between one row's descent and the next row's ascent... +// so you should advance the vertical position by "*ascent - *descent + *lineGap" +// these are expressed in unscaled coordinates, so you must multiply by +// the scale factor for a given size + +STBTT_DEF int stbtt_GetFontVMetricsOS2(const stbtt_fontinfo *info, int *typoAscent, int *typoDescent, int *typoLineGap); +// analogous to GetFontVMetrics, but returns the "typographic" values from the OS/2 +// table (specific to MS/Windows TTF files). +// +// Returns 1 on success (table present), 0 on failure. + +STBTT_DEF void stbtt_GetFontBoundingBox(const stbtt_fontinfo *info, int *x0, int *y0, int *x1, int *y1); +// the bounding box around all possible characters + +STBTT_DEF void stbtt_GetCodepointHMetrics(const stbtt_fontinfo *info, int codepoint, int *advanceWidth, int *leftSideBearing); +// leftSideBearing is the offset from the current horizontal position to the left edge of the character +// advanceWidth is the offset from the current horizontal position to the next horizontal position +// these are expressed in unscaled coordinates + +STBTT_DEF int stbtt_GetCodepointKernAdvance(const stbtt_fontinfo *info, int ch1, int ch2); +// an additional amount to add to the 'advance' value between ch1 and ch2 + +STBTT_DEF int stbtt_GetCodepointBox(const stbtt_fontinfo *info, int codepoint, int *x0, int *y0, int *x1, int *y1); +// Gets the bounding box of the visible part of the glyph, in unscaled coordinates + +STBTT_DEF void stbtt_GetGlyphHMetrics(const stbtt_fontinfo *info, int glyph_index, int *advanceWidth, int *leftSideBearing); +STBTT_DEF int stbtt_GetGlyphKernAdvance(const stbtt_fontinfo *info, int glyph1, int glyph2); +STBTT_DEF int stbtt_GetGlyphBox(const stbtt_fontinfo *info, int glyph_index, int *x0, int *y0, int *x1, int *y1); +// as above, but takes one or more glyph indices for greater efficiency + +typedef struct stbtt_kerningentry +{ + int glyph1; // use stbtt_FindGlyphIndex + int glyph2; + int advance; +} stbtt_kerningentry; + +STBTT_DEF int stbtt_GetKerningTableLength(const stbtt_fontinfo *info); +STBTT_DEF int stbtt_GetKerningTable(const stbtt_fontinfo *info, stbtt_kerningentry* table, int table_length); +// Retrieves a complete list of all of the kerning pairs provided by the font +// stbtt_GetKerningTable never writes more than table_length entries and returns how many entries it did write. +// The table will be sorted by (a.glyph1 == b.glyph1)?(a.glyph2 < b.glyph2):(a.glyph1 < b.glyph1) + +////////////////////////////////////////////////////////////////////////////// +// +// GLYPH SHAPES (you probably don't need these, but they have to go before +// the bitmaps for C declaration-order reasons) +// + +#ifndef STBTT_vmove // you can predefine these to use different values (but why?) + enum { + STBTT_vmove=1, + STBTT_vline, + STBTT_vcurve, + STBTT_vcubic + }; +#endif + +#ifndef stbtt_vertex // you can predefine this to use different values + // (we share this with other code at RAD) + #define stbtt_vertex_type short // can't use stbtt_int16 because that's not visible in the header file + typedef struct + { + stbtt_vertex_type x,y,cx,cy,cx1,cy1; + unsigned char type,padding; + } stbtt_vertex; +#endif + +STBTT_DEF int stbtt_IsGlyphEmpty(const stbtt_fontinfo *info, int glyph_index); +// returns non-zero if nothing is drawn for this glyph + +STBTT_DEF int stbtt_GetCodepointShape(const stbtt_fontinfo *info, int unicode_codepoint, stbtt_vertex **vertices); +STBTT_DEF int stbtt_GetGlyphShape(const stbtt_fontinfo *info, int glyph_index, stbtt_vertex **vertices); +// returns # of vertices and fills *vertices with the pointer to them +// these are expressed in "unscaled" coordinates +// +// The shape is a series of contours. Each one starts with +// a STBTT_moveto, then consists of a series of mixed +// STBTT_lineto and STBTT_curveto segments. A lineto +// draws a line from previous endpoint to its x,y; a curveto +// draws a quadratic bezier from previous endpoint to +// its x,y, using cx,cy as the bezier control point. + +STBTT_DEF void stbtt_FreeShape(const stbtt_fontinfo *info, stbtt_vertex *vertices); +// frees the data allocated above + +STBTT_DEF unsigned char *stbtt_FindSVGDoc(const stbtt_fontinfo *info, int gl); +STBTT_DEF int stbtt_GetCodepointSVG(const stbtt_fontinfo *info, int unicode_codepoint, const char **svg); +STBTT_DEF int stbtt_GetGlyphSVG(const stbtt_fontinfo *info, int gl, const char **svg); +// fills svg with the character's SVG data. +// returns data size or 0 if SVG not found. + +////////////////////////////////////////////////////////////////////////////// +// +// BITMAP RENDERING +// + +STBTT_DEF void stbtt_FreeBitmap(unsigned char *bitmap, void *userdata); +// frees the bitmap allocated below + +STBTT_DEF unsigned char *stbtt_GetCodepointBitmap(const stbtt_fontinfo *info, float scale_x, float scale_y, int codepoint, int *width, int *height, int *xoff, int *yoff); +// allocates a large-enough single-channel 8bpp bitmap and renders the +// specified character/glyph at the specified scale into it, with +// antialiasing. 0 is no coverage (transparent), 255 is fully covered (opaque). +// *width & *height are filled out with the width & height of the bitmap, +// which is stored left-to-right, top-to-bottom. +// +// xoff/yoff are the offset it pixel space from the glyph origin to the top-left of the bitmap + +STBTT_DEF unsigned char *stbtt_GetCodepointBitmapSubpixel(const stbtt_fontinfo *info, float scale_x, float scale_y, float shift_x, float shift_y, int codepoint, int *width, int *height, int *xoff, int *yoff); +// the same as stbtt_GetCodepoitnBitmap, but you can specify a subpixel +// shift for the character + +STBTT_DEF void stbtt_MakeCodepointBitmap(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, int codepoint); +// the same as stbtt_GetCodepointBitmap, but you pass in storage for the bitmap +// in the form of 'output', with row spacing of 'out_stride' bytes. the bitmap +// is clipped to out_w/out_h bytes. Call stbtt_GetCodepointBitmapBox to get the +// width and height and positioning info for it first. + +STBTT_DEF void stbtt_MakeCodepointBitmapSubpixel(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int codepoint); +// same as stbtt_MakeCodepointBitmap, but you can specify a subpixel +// shift for the character + +STBTT_DEF void stbtt_MakeCodepointBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int oversample_x, int oversample_y, float *sub_x, float *sub_y, int codepoint); +// same as stbtt_MakeCodepointBitmapSubpixel, but prefiltering +// is performed (see stbtt_PackSetOversampling) + +STBTT_DEF void stbtt_GetCodepointBitmapBox(const stbtt_fontinfo *font, int codepoint, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1); +// get the bbox of the bitmap centered around the glyph origin; so the +// bitmap width is ix1-ix0, height is iy1-iy0, and location to place +// the bitmap top left is (leftSideBearing*scale,iy0). +// (Note that the bitmap uses y-increases-down, but the shape uses +// y-increases-up, so CodepointBitmapBox and CodepointBox are inverted.) + +STBTT_DEF void stbtt_GetCodepointBitmapBoxSubpixel(const stbtt_fontinfo *font, int codepoint, float scale_x, float scale_y, float shift_x, float shift_y, int *ix0, int *iy0, int *ix1, int *iy1); +// same as stbtt_GetCodepointBitmapBox, but you can specify a subpixel +// shift for the character + +// the following functions are equivalent to the above functions, but operate +// on glyph indices instead of Unicode codepoints (for efficiency) +STBTT_DEF unsigned char *stbtt_GetGlyphBitmap(const stbtt_fontinfo *info, float scale_x, float scale_y, int glyph, int *width, int *height, int *xoff, int *yoff); +STBTT_DEF unsigned char *stbtt_GetGlyphBitmapSubpixel(const stbtt_fontinfo *info, float scale_x, float scale_y, float shift_x, float shift_y, int glyph, int *width, int *height, int *xoff, int *yoff); +STBTT_DEF void stbtt_MakeGlyphBitmap(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, int glyph); +STBTT_DEF void stbtt_MakeGlyphBitmapSubpixel(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int glyph); +STBTT_DEF void stbtt_MakeGlyphBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int oversample_x, int oversample_y, float *sub_x, float *sub_y, int glyph); +STBTT_DEF void stbtt_GetGlyphBitmapBox(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1); +STBTT_DEF void stbtt_GetGlyphBitmapBoxSubpixel(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y,float shift_x, float shift_y, int *ix0, int *iy0, int *ix1, int *iy1); + + +// @TODO: don't expose this structure +typedef struct +{ + int w,h,stride; + unsigned char *pixels; +} stbtt__bitmap; + +// rasterize a shape with quadratic beziers into a bitmap +STBTT_DEF void stbtt_Rasterize(stbtt__bitmap *result, // 1-channel bitmap to draw into + float flatness_in_pixels, // allowable error of curve in pixels + stbtt_vertex *vertices, // array of vertices defining shape + int num_verts, // number of vertices in above array + float scale_x, float scale_y, // scale applied to input vertices + float shift_x, float shift_y, // translation applied to input vertices + int x_off, int y_off, // another translation applied to input + int invert, // if non-zero, vertically flip shape + void *userdata); // context for to STBTT_MALLOC + +////////////////////////////////////////////////////////////////////////////// +// +// Signed Distance Function (or Field) rendering + +STBTT_DEF void stbtt_FreeSDF(unsigned char *bitmap, void *userdata); +// frees the SDF bitmap allocated below + +STBTT_DEF unsigned char * stbtt_GetGlyphSDF(const stbtt_fontinfo *info, float scale, int glyph, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff); +STBTT_DEF unsigned char * stbtt_GetCodepointSDF(const stbtt_fontinfo *info, float scale, int codepoint, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff); +// These functions compute a discretized SDF field for a single character, suitable for storing +// in a single-channel texture, sampling with bilinear filtering, and testing against +// larger than some threshold to produce scalable fonts. +// info -- the font +// scale -- controls the size of the resulting SDF bitmap, same as it would be creating a regular bitmap +// glyph/codepoint -- the character to generate the SDF for +// padding -- extra "pixels" around the character which are filled with the distance to the character (not 0), +// which allows effects like bit outlines +// onedge_value -- value 0-255 to test the SDF against to reconstruct the character (i.e. the isocontour of the character) +// pixel_dist_scale -- what value the SDF should increase by when moving one SDF "pixel" away from the edge (on the 0..255 scale) +// if positive, > onedge_value is inside; if negative, < onedge_value is inside +// width,height -- output height & width of the SDF bitmap (including padding) +// xoff,yoff -- output origin of the character +// return value -- a 2D array of bytes 0..255, width*height in size +// +// pixel_dist_scale & onedge_value are a scale & bias that allows you to make +// optimal use of the limited 0..255 for your application, trading off precision +// and special effects. SDF values outside the range 0..255 are clamped to 0..255. +// +// Example: +// scale = stbtt_ScaleForPixelHeight(22) +// padding = 5 +// onedge_value = 180 +// pixel_dist_scale = 180/5.0 = 36.0 +// +// This will create an SDF bitmap in which the character is about 22 pixels +// high but the whole bitmap is about 22+5+5=32 pixels high. To produce a filled +// shape, sample the SDF at each pixel and fill the pixel if the SDF value +// is greater than or equal to 180/255. (You'll actually want to antialias, +// which is beyond the scope of this example.) Additionally, you can compute +// offset outlines (e.g. to stroke the character border inside & outside, +// or only outside). For example, to fill outside the character up to 3 SDF +// pixels, you would compare against (180-36.0*3)/255 = 72/255. The above +// choice of variables maps a range from 5 pixels outside the shape to +// 2 pixels inside the shape to 0..255; this is intended primarily for apply +// outside effects only (the interior range is needed to allow proper +// antialiasing of the font at *smaller* sizes) +// +// The function computes the SDF analytically at each SDF pixel, not by e.g. +// building a higher-res bitmap and approximating it. In theory the quality +// should be as high as possible for an SDF of this size & representation, but +// unclear if this is true in practice (perhaps building a higher-res bitmap +// and computing from that can allow drop-out prevention). +// +// The algorithm has not been optimized at all, so expect it to be slow +// if computing lots of characters or very large sizes. + + + +////////////////////////////////////////////////////////////////////////////// +// +// Finding the right font... +// +// You should really just solve this offline, keep your own tables +// of what font is what, and don't try to get it out of the .ttf file. +// That's because getting it out of the .ttf file is really hard, because +// the names in the file can appear in many possible encodings, in many +// possible languages, and e.g. if you need a case-insensitive comparison, +// the details of that depend on the encoding & language in a complex way +// (actually underspecified in truetype, but also gigantic). +// +// But you can use the provided functions in two possible ways: +// stbtt_FindMatchingFont() will use *case-sensitive* comparisons on +// unicode-encoded names to try to find the font you want; +// you can run this before calling stbtt_InitFont() +// +// stbtt_GetFontNameString() lets you get any of the various strings +// from the file yourself and do your own comparisons on them. +// You have to have called stbtt_InitFont() first. + + +STBTT_DEF int stbtt_FindMatchingFont(const unsigned char *fontdata, const char *name, int flags); +// returns the offset (not index) of the font that matches, or -1 if none +// if you use STBTT_MACSTYLE_DONTCARE, use a font name like "Arial Bold". +// if you use any other flag, use a font name like "Arial"; this checks +// the 'macStyle' header field; i don't know if fonts set this consistently +#define STBTT_MACSTYLE_DONTCARE 0 +#define STBTT_MACSTYLE_BOLD 1 +#define STBTT_MACSTYLE_ITALIC 2 +#define STBTT_MACSTYLE_UNDERSCORE 4 +#define STBTT_MACSTYLE_NONE 8 // <= not same as 0, this makes us check the bitfield is 0 + +STBTT_DEF int stbtt_CompareUTF8toUTF16_bigendian(const char *s1, int len1, const char *s2, int len2); +// returns 1/0 whether the first string interpreted as utf8 is identical to +// the second string interpreted as big-endian utf16... useful for strings from next func + +STBTT_DEF const char *stbtt_GetFontNameString(const stbtt_fontinfo *font, int *length, int platformID, int encodingID, int languageID, int nameID); +// returns the string (which may be big-endian double byte, e.g. for unicode) +// and puts the length in bytes in *length. +// +// some of the values for the IDs are below; for more see the truetype spec: +// http://developer.apple.com/textfonts/TTRefMan/RM06/Chap6name.html +// http://www.microsoft.com/typography/otspec/name.htm + +enum { // platformID + STBTT_PLATFORM_ID_UNICODE =0, + STBTT_PLATFORM_ID_MAC =1, + STBTT_PLATFORM_ID_ISO =2, + STBTT_PLATFORM_ID_MICROSOFT =3 +}; + +enum { // encodingID for STBTT_PLATFORM_ID_UNICODE + STBTT_UNICODE_EID_UNICODE_1_0 =0, + STBTT_UNICODE_EID_UNICODE_1_1 =1, + STBTT_UNICODE_EID_ISO_10646 =2, + STBTT_UNICODE_EID_UNICODE_2_0_BMP=3, + STBTT_UNICODE_EID_UNICODE_2_0_FULL=4 +}; + +enum { // encodingID for STBTT_PLATFORM_ID_MICROSOFT + STBTT_MS_EID_SYMBOL =0, + STBTT_MS_EID_UNICODE_BMP =1, + STBTT_MS_EID_SHIFTJIS =2, + STBTT_MS_EID_UNICODE_FULL =10 +}; + +enum { // encodingID for STBTT_PLATFORM_ID_MAC; same as Script Manager codes + STBTT_MAC_EID_ROMAN =0, STBTT_MAC_EID_ARABIC =4, + STBTT_MAC_EID_JAPANESE =1, STBTT_MAC_EID_HEBREW =5, + STBTT_MAC_EID_CHINESE_TRAD =2, STBTT_MAC_EID_GREEK =6, + STBTT_MAC_EID_KOREAN =3, STBTT_MAC_EID_RUSSIAN =7 +}; + +enum { // languageID for STBTT_PLATFORM_ID_MICROSOFT; same as LCID... + // problematic because there are e.g. 16 english LCIDs and 16 arabic LCIDs + STBTT_MS_LANG_ENGLISH =0x0409, STBTT_MS_LANG_ITALIAN =0x0410, + STBTT_MS_LANG_CHINESE =0x0804, STBTT_MS_LANG_JAPANESE =0x0411, + STBTT_MS_LANG_DUTCH =0x0413, STBTT_MS_LANG_KOREAN =0x0412, + STBTT_MS_LANG_FRENCH =0x040c, STBTT_MS_LANG_RUSSIAN =0x0419, + STBTT_MS_LANG_GERMAN =0x0407, STBTT_MS_LANG_SPANISH =0x0409, + STBTT_MS_LANG_HEBREW =0x040d, STBTT_MS_LANG_SWEDISH =0x041D +}; + +enum { // languageID for STBTT_PLATFORM_ID_MAC + STBTT_MAC_LANG_ENGLISH =0 , STBTT_MAC_LANG_JAPANESE =11, + STBTT_MAC_LANG_ARABIC =12, STBTT_MAC_LANG_KOREAN =23, + STBTT_MAC_LANG_DUTCH =4 , STBTT_MAC_LANG_RUSSIAN =32, + STBTT_MAC_LANG_FRENCH =1 , STBTT_MAC_LANG_SPANISH =6 , + STBTT_MAC_LANG_GERMAN =2 , STBTT_MAC_LANG_SWEDISH =5 , + STBTT_MAC_LANG_HEBREW =10, STBTT_MAC_LANG_CHINESE_SIMPLIFIED =33, + STBTT_MAC_LANG_ITALIAN =3 , STBTT_MAC_LANG_CHINESE_TRAD =19 +}; + +#ifdef __cplusplus +} +#endif + +#endif // __STB_INCLUDE_STB_TRUETYPE_H__ + +/////////////////////////////////////////////////////////////////////////////// +/////////////////////////////////////////////////////////////////////////////// +//// +//// IMPLEMENTATION +//// +//// + +#ifdef STB_TRUETYPE_IMPLEMENTATION + +#ifndef STBTT_MAX_OVERSAMPLE +#define STBTT_MAX_OVERSAMPLE 8 +#endif + +#if STBTT_MAX_OVERSAMPLE > 255 +#error "STBTT_MAX_OVERSAMPLE cannot be > 255" +#endif + +typedef int stbtt__test_oversample_pow2[(STBTT_MAX_OVERSAMPLE & (STBTT_MAX_OVERSAMPLE-1)) == 0 ? 1 : -1]; + +#ifndef STBTT_RASTERIZER_VERSION +#define STBTT_RASTERIZER_VERSION 2 +#endif + +#ifdef _MSC_VER +#define STBTT__NOTUSED(v) (void)(v) +#else +#define STBTT__NOTUSED(v) (void)sizeof(v) +#endif + +////////////////////////////////////////////////////////////////////////// +// +// stbtt__buf helpers to parse data from file +// + +static stbtt_uint8 stbtt__buf_get8(stbtt__buf *b) +{ + if (b->cursor >= b->size) + return 0; + return b->data[b->cursor++]; +} + +static stbtt_uint8 stbtt__buf_peek8(stbtt__buf *b) +{ + if (b->cursor >= b->size) + return 0; + return b->data[b->cursor]; +} + +static void stbtt__buf_seek(stbtt__buf *b, int o) +{ + STBTT_assert(!(o > b->size || o < 0)); + b->cursor = (o > b->size || o < 0) ? b->size : o; +} + +static void stbtt__buf_skip(stbtt__buf *b, int o) +{ + stbtt__buf_seek(b, b->cursor + o); +} + +static stbtt_uint32 stbtt__buf_get(stbtt__buf *b, int n) +{ + stbtt_uint32 v = 0; + int i; + STBTT_assert(n >= 1 && n <= 4); + for (i = 0; i < n; i++) + v = (v << 8) | stbtt__buf_get8(b); + return v; +} + +static stbtt__buf stbtt__new_buf(const void *p, size_t size) +{ + stbtt__buf r; + STBTT_assert(size < 0x40000000); + r.data = (stbtt_uint8*) p; + r.size = (int) size; + r.cursor = 0; + return r; +} + +#define stbtt__buf_get16(b) stbtt__buf_get((b), 2) +#define stbtt__buf_get32(b) stbtt__buf_get((b), 4) + +static stbtt__buf stbtt__buf_range(const stbtt__buf *b, int o, int s) +{ + stbtt__buf r = stbtt__new_buf(NULL, 0); + if (o < 0 || s < 0 || o > b->size || s > b->size - o) return r; + r.data = b->data + o; + r.size = s; + return r; +} + +static stbtt__buf stbtt__cff_get_index(stbtt__buf *b) +{ + int count, start, offsize; + start = b->cursor; + count = stbtt__buf_get16(b); + if (count) { + offsize = stbtt__buf_get8(b); + STBTT_assert(offsize >= 1 && offsize <= 4); + stbtt__buf_skip(b, offsize * count); + stbtt__buf_skip(b, stbtt__buf_get(b, offsize) - 1); + } + return stbtt__buf_range(b, start, b->cursor - start); +} + +static stbtt_uint32 stbtt__cff_int(stbtt__buf *b) +{ + int b0 = stbtt__buf_get8(b); + if (b0 >= 32 && b0 <= 246) return b0 - 139; + else if (b0 >= 247 && b0 <= 250) return (b0 - 247)*256 + stbtt__buf_get8(b) + 108; + else if (b0 >= 251 && b0 <= 254) return -(b0 - 251)*256 - stbtt__buf_get8(b) - 108; + else if (b0 == 28) return stbtt__buf_get16(b); + else if (b0 == 29) return stbtt__buf_get32(b); + STBTT_assert(0); + return 0; +} + +static void stbtt__cff_skip_operand(stbtt__buf *b) { + int v, b0 = stbtt__buf_peek8(b); + STBTT_assert(b0 >= 28); + if (b0 == 30) { + stbtt__buf_skip(b, 1); + while (b->cursor < b->size) { + v = stbtt__buf_get8(b); + if ((v & 0xF) == 0xF || (v >> 4) == 0xF) + break; + } + } else { + stbtt__cff_int(b); + } +} + +static stbtt__buf stbtt__dict_get(stbtt__buf *b, int key) +{ + stbtt__buf_seek(b, 0); + while (b->cursor < b->size) { + int start = b->cursor, end, op; + while (stbtt__buf_peek8(b) >= 28) + stbtt__cff_skip_operand(b); + end = b->cursor; + op = stbtt__buf_get8(b); + if (op == 12) op = stbtt__buf_get8(b) | 0x100; + if (op == key) return stbtt__buf_range(b, start, end-start); + } + return stbtt__buf_range(b, 0, 0); +} + +static void stbtt__dict_get_ints(stbtt__buf *b, int key, int outcount, stbtt_uint32 *out) +{ + int i; + stbtt__buf operands = stbtt__dict_get(b, key); + for (i = 0; i < outcount && operands.cursor < operands.size; i++) + out[i] = stbtt__cff_int(&operands); +} + +static int stbtt__cff_index_count(stbtt__buf *b) +{ + stbtt__buf_seek(b, 0); + return stbtt__buf_get16(b); +} + +static stbtt__buf stbtt__cff_index_get(stbtt__buf b, int i) +{ + int count, offsize, start, end; + stbtt__buf_seek(&b, 0); + count = stbtt__buf_get16(&b); + offsize = stbtt__buf_get8(&b); + STBTT_assert(i >= 0 && i < count); + STBTT_assert(offsize >= 1 && offsize <= 4); + stbtt__buf_skip(&b, i*offsize); + start = stbtt__buf_get(&b, offsize); + end = stbtt__buf_get(&b, offsize); + return stbtt__buf_range(&b, 2+(count+1)*offsize+start, end - start); +} + +////////////////////////////////////////////////////////////////////////// +// +// accessors to parse data from file +// + +// on platforms that don't allow misaligned reads, if we want to allow +// truetype fonts that aren't padded to alignment, define ALLOW_UNALIGNED_TRUETYPE + +#define ttBYTE(p) (* (stbtt_uint8 *) (p)) +#define ttCHAR(p) (* (stbtt_int8 *) (p)) +#define ttFixed(p) ttLONG(p) + +static stbtt_uint16 ttUSHORT(stbtt_uint8 *p) { return p[0]*256 + p[1]; } +static stbtt_int16 ttSHORT(stbtt_uint8 *p) { return p[0]*256 + p[1]; } +static stbtt_uint32 ttULONG(stbtt_uint8 *p) { return (p[0]<<24) + (p[1]<<16) + (p[2]<<8) + p[3]; } +static stbtt_int32 ttLONG(stbtt_uint8 *p) { return (p[0]<<24) + (p[1]<<16) + (p[2]<<8) + p[3]; } + +#define stbtt_tag4(p,c0,c1,c2,c3) ((p)[0] == (c0) && (p)[1] == (c1) && (p)[2] == (c2) && (p)[3] == (c3)) +#define stbtt_tag(p,str) stbtt_tag4(p,str[0],str[1],str[2],str[3]) + +static int stbtt__isfont(stbtt_uint8 *font) +{ + // check the version number + if (stbtt_tag4(font, '1',0,0,0)) return 1; // TrueType 1 + if (stbtt_tag(font, "typ1")) return 1; // TrueType with type 1 font -- we don't support this! + if (stbtt_tag(font, "OTTO")) return 1; // OpenType with CFF + if (stbtt_tag4(font, 0,1,0,0)) return 1; // OpenType 1.0 + if (stbtt_tag(font, "true")) return 1; // Apple specification for TrueType fonts + return 0; +} + +// @OPTIMIZE: binary search +static stbtt_uint32 stbtt__find_table(stbtt_uint8 *data, stbtt_uint32 fontstart, const char *tag) +{ + stbtt_int32 num_tables = ttUSHORT(data+fontstart+4); + stbtt_uint32 tabledir = fontstart + 12; + stbtt_int32 i; + for (i=0; i < num_tables; ++i) { + stbtt_uint32 loc = tabledir + 16*i; + if (stbtt_tag(data+loc+0, tag)) + return ttULONG(data+loc+8); + } + return 0; +} + +static int stbtt_GetFontOffsetForIndex_internal(unsigned char *font_collection, int index) +{ + // if it's just a font, there's only one valid index + if (stbtt__isfont(font_collection)) + return index == 0 ? 0 : -1; + + // check if it's a TTC + if (stbtt_tag(font_collection, "ttcf")) { + // version 1? + if (ttULONG(font_collection+4) == 0x00010000 || ttULONG(font_collection+4) == 0x00020000) { + stbtt_int32 n = ttLONG(font_collection+8); + if (index >= n) + return -1; + return ttULONG(font_collection+12+index*4); + } + } + return -1; +} + +static int stbtt_GetNumberOfFonts_internal(unsigned char *font_collection) +{ + // if it's just a font, there's only one valid font + if (stbtt__isfont(font_collection)) + return 1; + + // check if it's a TTC + if (stbtt_tag(font_collection, "ttcf")) { + // version 1? + if (ttULONG(font_collection+4) == 0x00010000 || ttULONG(font_collection+4) == 0x00020000) { + return ttLONG(font_collection+8); + } + } + return 0; +} + +static stbtt__buf stbtt__get_subrs(stbtt__buf cff, stbtt__buf fontdict) +{ + stbtt_uint32 subrsoff = 0, private_loc[2] = { 0, 0 }; + stbtt__buf pdict; + stbtt__dict_get_ints(&fontdict, 18, 2, private_loc); + if (!private_loc[1] || !private_loc[0]) return stbtt__new_buf(NULL, 0); + pdict = stbtt__buf_range(&cff, private_loc[1], private_loc[0]); + stbtt__dict_get_ints(&pdict, 19, 1, &subrsoff); + if (!subrsoff) return stbtt__new_buf(NULL, 0); + stbtt__buf_seek(&cff, private_loc[1]+subrsoff); + return stbtt__cff_get_index(&cff); +} + +// since most people won't use this, find this table the first time it's needed +static int stbtt__get_svg(stbtt_fontinfo *info) +{ + stbtt_uint32 t; + if (info->svg < 0) { + t = stbtt__find_table(info->data, info->fontstart, "SVG "); + if (t) { + stbtt_uint32 offset = ttULONG(info->data + t + 2); + info->svg = t + offset; + } else { + info->svg = 0; + } + } + return info->svg; +} + +static int stbtt_InitFont_internal(stbtt_fontinfo *info, unsigned char *data, int fontstart) +{ + stbtt_uint32 cmap, t; + stbtt_int32 i,numTables; + + info->data = data; + info->fontstart = fontstart; + info->cff = stbtt__new_buf(NULL, 0); + + cmap = stbtt__find_table(data, fontstart, "cmap"); // required + info->loca = stbtt__find_table(data, fontstart, "loca"); // required + info->head = stbtt__find_table(data, fontstart, "head"); // required + info->glyf = stbtt__find_table(data, fontstart, "glyf"); // required + info->hhea = stbtt__find_table(data, fontstart, "hhea"); // required + info->hmtx = stbtt__find_table(data, fontstart, "hmtx"); // required + info->kern = stbtt__find_table(data, fontstart, "kern"); // not required + info->gpos = stbtt__find_table(data, fontstart, "GPOS"); // not required + + if (!cmap || !info->head || !info->hhea || !info->hmtx) + return 0; + if (info->glyf) { + // required for truetype + if (!info->loca) return 0; + } else { + // initialization for CFF / Type2 fonts (OTF) + stbtt__buf b, topdict, topdictidx; + stbtt_uint32 cstype = 2, charstrings = 0, fdarrayoff = 0, fdselectoff = 0; + stbtt_uint32 cff; + + cff = stbtt__find_table(data, fontstart, "CFF "); + if (!cff) return 0; + + info->fontdicts = stbtt__new_buf(NULL, 0); + info->fdselect = stbtt__new_buf(NULL, 0); + + // @TODO this should use size from table (not 512MB) + info->cff = stbtt__new_buf(data+cff, 512*1024*1024); + b = info->cff; + + // read the header + stbtt__buf_skip(&b, 2); + stbtt__buf_seek(&b, stbtt__buf_get8(&b)); // hdrsize + + // @TODO the name INDEX could list multiple fonts, + // but we just use the first one. + stbtt__cff_get_index(&b); // name INDEX + topdictidx = stbtt__cff_get_index(&b); + topdict = stbtt__cff_index_get(topdictidx, 0); + stbtt__cff_get_index(&b); // string INDEX + info->gsubrs = stbtt__cff_get_index(&b); + + stbtt__dict_get_ints(&topdict, 17, 1, &charstrings); + stbtt__dict_get_ints(&topdict, 0x100 | 6, 1, &cstype); + stbtt__dict_get_ints(&topdict, 0x100 | 36, 1, &fdarrayoff); + stbtt__dict_get_ints(&topdict, 0x100 | 37, 1, &fdselectoff); + info->subrs = stbtt__get_subrs(b, topdict); + + // we only support Type 2 charstrings + if (cstype != 2) return 0; + if (charstrings == 0) return 0; + + if (fdarrayoff) { + // looks like a CID font + if (!fdselectoff) return 0; + stbtt__buf_seek(&b, fdarrayoff); + info->fontdicts = stbtt__cff_get_index(&b); + info->fdselect = stbtt__buf_range(&b, fdselectoff, b.size-fdselectoff); + } + + stbtt__buf_seek(&b, charstrings); + info->charstrings = stbtt__cff_get_index(&b); + } + + t = stbtt__find_table(data, fontstart, "maxp"); + if (t) + info->numGlyphs = ttUSHORT(data+t+4); + else + info->numGlyphs = 0xffff; + + info->svg = -1; + + // find a cmap encoding table we understand *now* to avoid searching + // later. (todo: could make this installable) + // the same regardless of glyph. + numTables = ttUSHORT(data + cmap + 2); + info->index_map = 0; + for (i=0; i < numTables; ++i) { + stbtt_uint32 encoding_record = cmap + 4 + 8 * i; + // find an encoding we understand: + switch(ttUSHORT(data+encoding_record)) { + case STBTT_PLATFORM_ID_MICROSOFT: + switch (ttUSHORT(data+encoding_record+2)) { + case STBTT_MS_EID_UNICODE_BMP: + case STBTT_MS_EID_UNICODE_FULL: + // MS/Unicode + info->index_map = cmap + ttULONG(data+encoding_record+4); + break; + } + break; + case STBTT_PLATFORM_ID_UNICODE: + // Mac/iOS has these + // all the encodingIDs are unicode, so we don't bother to check it + info->index_map = cmap + ttULONG(data+encoding_record+4); + break; + } + } + if (info->index_map == 0) + return 0; + + info->indexToLocFormat = ttUSHORT(data+info->head + 50); + return 1; +} + +STBTT_DEF int stbtt_FindGlyphIndex(const stbtt_fontinfo *info, int unicode_codepoint) +{ + stbtt_uint8 *data = info->data; + stbtt_uint32 index_map = info->index_map; + + stbtt_uint16 format = ttUSHORT(data + index_map + 0); + if (format == 0) { // apple byte encoding + stbtt_int32 bytes = ttUSHORT(data + index_map + 2); + if (unicode_codepoint < bytes-6) + return ttBYTE(data + index_map + 6 + unicode_codepoint); + return 0; + } else if (format == 6) { + stbtt_uint32 first = ttUSHORT(data + index_map + 6); + stbtt_uint32 count = ttUSHORT(data + index_map + 8); + if ((stbtt_uint32) unicode_codepoint >= first && (stbtt_uint32) unicode_codepoint < first+count) + return ttUSHORT(data + index_map + 10 + (unicode_codepoint - first)*2); + return 0; + } else if (format == 2) { + STBTT_assert(0); // @TODO: high-byte mapping for japanese/chinese/korean + return 0; + } else if (format == 4) { // standard mapping for windows fonts: binary search collection of ranges + stbtt_uint16 segcount = ttUSHORT(data+index_map+6) >> 1; + stbtt_uint16 searchRange = ttUSHORT(data+index_map+8) >> 1; + stbtt_uint16 entrySelector = ttUSHORT(data+index_map+10); + stbtt_uint16 rangeShift = ttUSHORT(data+index_map+12) >> 1; + + // do a binary search of the segments + stbtt_uint32 endCount = index_map + 14; + stbtt_uint32 search = endCount; + + if (unicode_codepoint > 0xffff) + return 0; + + // they lie from endCount .. endCount + segCount + // but searchRange is the nearest power of two, so... + if (unicode_codepoint >= ttUSHORT(data + search + rangeShift*2)) + search += rangeShift*2; + + // now decrement to bias correctly to find smallest + search -= 2; + while (entrySelector) { + stbtt_uint16 end; + searchRange >>= 1; + end = ttUSHORT(data + search + searchRange*2); + if (unicode_codepoint > end) + search += searchRange*2; + --entrySelector; + } + search += 2; + + { + stbtt_uint16 offset, start, last; + stbtt_uint16 item = (stbtt_uint16) ((search - endCount) >> 1); + + start = ttUSHORT(data + index_map + 14 + segcount*2 + 2 + 2*item); + last = ttUSHORT(data + endCount + 2*item); + if (unicode_codepoint < start || unicode_codepoint > last) + return 0; + + offset = ttUSHORT(data + index_map + 14 + segcount*6 + 2 + 2*item); + if (offset == 0) + return (stbtt_uint16) (unicode_codepoint + ttSHORT(data + index_map + 14 + segcount*4 + 2 + 2*item)); + + return ttUSHORT(data + offset + (unicode_codepoint-start)*2 + index_map + 14 + segcount*6 + 2 + 2*item); + } + } else if (format == 12 || format == 13) { + stbtt_uint32 ngroups = ttULONG(data+index_map+12); + stbtt_int32 low,high; + low = 0; high = (stbtt_int32)ngroups; + // Binary search the right group. + while (low < high) { + stbtt_int32 mid = low + ((high-low) >> 1); // rounds down, so low <= mid < high + stbtt_uint32 start_char = ttULONG(data+index_map+16+mid*12); + stbtt_uint32 end_char = ttULONG(data+index_map+16+mid*12+4); + if ((stbtt_uint32) unicode_codepoint < start_char) + high = mid; + else if ((stbtt_uint32) unicode_codepoint > end_char) + low = mid+1; + else { + stbtt_uint32 start_glyph = ttULONG(data+index_map+16+mid*12+8); + if (format == 12) + return start_glyph + unicode_codepoint-start_char; + else // format == 13 + return start_glyph; + } + } + return 0; // not found + } + // @TODO + STBTT_assert(0); + return 0; +} + +STBTT_DEF int stbtt_GetCodepointShape(const stbtt_fontinfo *info, int unicode_codepoint, stbtt_vertex **vertices) +{ + return stbtt_GetGlyphShape(info, stbtt_FindGlyphIndex(info, unicode_codepoint), vertices); +} + +static void stbtt_setvertex(stbtt_vertex *v, stbtt_uint8 type, stbtt_int32 x, stbtt_int32 y, stbtt_int32 cx, stbtt_int32 cy) +{ + v->type = type; + v->x = (stbtt_int16) x; + v->y = (stbtt_int16) y; + v->cx = (stbtt_int16) cx; + v->cy = (stbtt_int16) cy; +} + +static int stbtt__GetGlyfOffset(const stbtt_fontinfo *info, int glyph_index) +{ + int g1,g2; + + STBTT_assert(!info->cff.size); + + if (glyph_index >= info->numGlyphs) return -1; // glyph index out of range + if (info->indexToLocFormat >= 2) return -1; // unknown index->glyph map format + + if (info->indexToLocFormat == 0) { + g1 = info->glyf + ttUSHORT(info->data + info->loca + glyph_index * 2) * 2; + g2 = info->glyf + ttUSHORT(info->data + info->loca + glyph_index * 2 + 2) * 2; + } else { + g1 = info->glyf + ttULONG (info->data + info->loca + glyph_index * 4); + g2 = info->glyf + ttULONG (info->data + info->loca + glyph_index * 4 + 4); + } + + return g1==g2 ? -1 : g1; // if length is 0, return -1 +} + +static int stbtt__GetGlyphInfoT2(const stbtt_fontinfo *info, int glyph_index, int *x0, int *y0, int *x1, int *y1); + +STBTT_DEF int stbtt_GetGlyphBox(const stbtt_fontinfo *info, int glyph_index, int *x0, int *y0, int *x1, int *y1) +{ + if (info->cff.size) { + stbtt__GetGlyphInfoT2(info, glyph_index, x0, y0, x1, y1); + } else { + int g = stbtt__GetGlyfOffset(info, glyph_index); + if (g < 0) return 0; + + if (x0) *x0 = ttSHORT(info->data + g + 2); + if (y0) *y0 = ttSHORT(info->data + g + 4); + if (x1) *x1 = ttSHORT(info->data + g + 6); + if (y1) *y1 = ttSHORT(info->data + g + 8); + } + return 1; +} + +STBTT_DEF int stbtt_GetCodepointBox(const stbtt_fontinfo *info, int codepoint, int *x0, int *y0, int *x1, int *y1) +{ + return stbtt_GetGlyphBox(info, stbtt_FindGlyphIndex(info,codepoint), x0,y0,x1,y1); +} + +STBTT_DEF int stbtt_IsGlyphEmpty(const stbtt_fontinfo *info, int glyph_index) +{ + stbtt_int16 numberOfContours; + int g; + if (info->cff.size) + return stbtt__GetGlyphInfoT2(info, glyph_index, NULL, NULL, NULL, NULL) == 0; + g = stbtt__GetGlyfOffset(info, glyph_index); + if (g < 0) return 1; + numberOfContours = ttSHORT(info->data + g); + return numberOfContours == 0; +} + +static int stbtt__close_shape(stbtt_vertex *vertices, int num_vertices, int was_off, int start_off, + stbtt_int32 sx, stbtt_int32 sy, stbtt_int32 scx, stbtt_int32 scy, stbtt_int32 cx, stbtt_int32 cy) +{ + if (start_off) { + if (was_off) + stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve, (cx+scx)>>1, (cy+scy)>>1, cx,cy); + stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve, sx,sy,scx,scy); + } else { + if (was_off) + stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve,sx,sy,cx,cy); + else + stbtt_setvertex(&vertices[num_vertices++], STBTT_vline,sx,sy,0,0); + } + return num_vertices; +} + +static int stbtt__GetGlyphShapeTT(const stbtt_fontinfo *info, int glyph_index, stbtt_vertex **pvertices) +{ + stbtt_int16 numberOfContours; + stbtt_uint8 *endPtsOfContours; + stbtt_uint8 *data = info->data; + stbtt_vertex *vertices=0; + int num_vertices=0; + int g = stbtt__GetGlyfOffset(info, glyph_index); + + *pvertices = NULL; + + if (g < 0) return 0; + + numberOfContours = ttSHORT(data + g); + + if (numberOfContours > 0) { + stbtt_uint8 flags=0,flagcount; + stbtt_int32 ins, i,j=0,m,n, next_move, was_off=0, off, start_off=0; + stbtt_int32 x,y,cx,cy,sx,sy, scx,scy; + stbtt_uint8 *points; + endPtsOfContours = (data + g + 10); + ins = ttUSHORT(data + g + 10 + numberOfContours * 2); + points = data + g + 10 + numberOfContours * 2 + 2 + ins; + + n = 1+ttUSHORT(endPtsOfContours + numberOfContours*2-2); + + m = n + 2*numberOfContours; // a loose bound on how many vertices we might need + vertices = (stbtt_vertex *) STBTT_malloc(m * sizeof(vertices[0]), info->userdata); + if (vertices == 0) + return 0; + + next_move = 0; + flagcount=0; + + // in first pass, we load uninterpreted data into the allocated array + // above, shifted to the end of the array so we won't overwrite it when + // we create our final data starting from the front + + off = m - n; // starting offset for uninterpreted data, regardless of how m ends up being calculated + + // first load flags + + for (i=0; i < n; ++i) { + if (flagcount == 0) { + flags = *points++; + if (flags & 8) + flagcount = *points++; + } else + --flagcount; + vertices[off+i].type = flags; + } + + // now load x coordinates + x=0; + for (i=0; i < n; ++i) { + flags = vertices[off+i].type; + if (flags & 2) { + stbtt_int16 dx = *points++; + x += (flags & 16) ? dx : -dx; // ??? + } else { + if (!(flags & 16)) { + x = x + (stbtt_int16) (points[0]*256 + points[1]); + points += 2; + } + } + vertices[off+i].x = (stbtt_int16) x; + } + + // now load y coordinates + y=0; + for (i=0; i < n; ++i) { + flags = vertices[off+i].type; + if (flags & 4) { + stbtt_int16 dy = *points++; + y += (flags & 32) ? dy : -dy; // ??? + } else { + if (!(flags & 32)) { + y = y + (stbtt_int16) (points[0]*256 + points[1]); + points += 2; + } + } + vertices[off+i].y = (stbtt_int16) y; + } + + // now convert them to our format + num_vertices=0; + sx = sy = cx = cy = scx = scy = 0; + for (i=0; i < n; ++i) { + flags = vertices[off+i].type; + x = (stbtt_int16) vertices[off+i].x; + y = (stbtt_int16) vertices[off+i].y; + + if (next_move == i) { + if (i != 0) + num_vertices = stbtt__close_shape(vertices, num_vertices, was_off, start_off, sx,sy,scx,scy,cx,cy); + + // now start the new one + start_off = !(flags & 1); + if (start_off) { + // if we start off with an off-curve point, then when we need to find a point on the curve + // where we can start, and we need to save some state for when we wraparound. + scx = x; + scy = y; + if (!(vertices[off+i+1].type & 1)) { + // next point is also a curve point, so interpolate an on-point curve + sx = (x + (stbtt_int32) vertices[off+i+1].x) >> 1; + sy = (y + (stbtt_int32) vertices[off+i+1].y) >> 1; + } else { + // otherwise just use the next point as our start point + sx = (stbtt_int32) vertices[off+i+1].x; + sy = (stbtt_int32) vertices[off+i+1].y; + ++i; // we're using point i+1 as the starting point, so skip it + } + } else { + sx = x; + sy = y; + } + stbtt_setvertex(&vertices[num_vertices++], STBTT_vmove,sx,sy,0,0); + was_off = 0; + next_move = 1 + ttUSHORT(endPtsOfContours+j*2); + ++j; + } else { + if (!(flags & 1)) { // if it's a curve + if (was_off) // two off-curve control points in a row means interpolate an on-curve midpoint + stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve, (cx+x)>>1, (cy+y)>>1, cx, cy); + cx = x; + cy = y; + was_off = 1; + } else { + if (was_off) + stbtt_setvertex(&vertices[num_vertices++], STBTT_vcurve, x,y, cx, cy); + else + stbtt_setvertex(&vertices[num_vertices++], STBTT_vline, x,y,0,0); + was_off = 0; + } + } + } + num_vertices = stbtt__close_shape(vertices, num_vertices, was_off, start_off, sx,sy,scx,scy,cx,cy); + } else if (numberOfContours < 0) { + // Compound shapes. + int more = 1; + stbtt_uint8 *comp = data + g + 10; + num_vertices = 0; + vertices = 0; + while (more) { + stbtt_uint16 flags, gidx; + int comp_num_verts = 0, i; + stbtt_vertex *comp_verts = 0, *tmp = 0; + float mtx[6] = {1,0,0,1,0,0}, m, n; + + flags = ttSHORT(comp); comp+=2; + gidx = ttSHORT(comp); comp+=2; + + if (flags & 2) { // XY values + if (flags & 1) { // shorts + mtx[4] = ttSHORT(comp); comp+=2; + mtx[5] = ttSHORT(comp); comp+=2; + } else { + mtx[4] = ttCHAR(comp); comp+=1; + mtx[5] = ttCHAR(comp); comp+=1; + } + } + else { + // @TODO handle matching point + STBTT_assert(0); + } + if (flags & (1<<3)) { // WE_HAVE_A_SCALE + mtx[0] = mtx[3] = ttSHORT(comp)/16384.0f; comp+=2; + mtx[1] = mtx[2] = 0; + } else if (flags & (1<<6)) { // WE_HAVE_AN_X_AND_YSCALE + mtx[0] = ttSHORT(comp)/16384.0f; comp+=2; + mtx[1] = mtx[2] = 0; + mtx[3] = ttSHORT(comp)/16384.0f; comp+=2; + } else if (flags & (1<<7)) { // WE_HAVE_A_TWO_BY_TWO + mtx[0] = ttSHORT(comp)/16384.0f; comp+=2; + mtx[1] = ttSHORT(comp)/16384.0f; comp+=2; + mtx[2] = ttSHORT(comp)/16384.0f; comp+=2; + mtx[3] = ttSHORT(comp)/16384.0f; comp+=2; + } + + // Find transformation scales. + m = (float) STBTT_sqrt(mtx[0]*mtx[0] + mtx[1]*mtx[1]); + n = (float) STBTT_sqrt(mtx[2]*mtx[2] + mtx[3]*mtx[3]); + + // Get indexed glyph. + comp_num_verts = stbtt_GetGlyphShape(info, gidx, &comp_verts); + if (comp_num_verts > 0) { + // Transform vertices. + for (i = 0; i < comp_num_verts; ++i) { + stbtt_vertex* v = &comp_verts[i]; + stbtt_vertex_type x,y; + x=v->x; y=v->y; + v->x = (stbtt_vertex_type)(m * (mtx[0]*x + mtx[2]*y + mtx[4])); + v->y = (stbtt_vertex_type)(n * (mtx[1]*x + mtx[3]*y + mtx[5])); + x=v->cx; y=v->cy; + v->cx = (stbtt_vertex_type)(m * (mtx[0]*x + mtx[2]*y + mtx[4])); + v->cy = (stbtt_vertex_type)(n * (mtx[1]*x + mtx[3]*y + mtx[5])); + } + // Append vertices. + tmp = (stbtt_vertex*)STBTT_malloc((num_vertices+comp_num_verts)*sizeof(stbtt_vertex), info->userdata); + if (!tmp) { + if (vertices) STBTT_free(vertices, info->userdata); + if (comp_verts) STBTT_free(comp_verts, info->userdata); + return 0; + } + if (num_vertices > 0 && vertices) STBTT_memcpy(tmp, vertices, num_vertices*sizeof(stbtt_vertex)); + STBTT_memcpy(tmp+num_vertices, comp_verts, comp_num_verts*sizeof(stbtt_vertex)); + if (vertices) STBTT_free(vertices, info->userdata); + vertices = tmp; + STBTT_free(comp_verts, info->userdata); + num_vertices += comp_num_verts; + } + // More components ? + more = flags & (1<<5); + } + } else { + // numberOfCounters == 0, do nothing + } + + *pvertices = vertices; + return num_vertices; +} + +typedef struct +{ + int bounds; + int started; + float first_x, first_y; + float x, y; + stbtt_int32 min_x, max_x, min_y, max_y; + + stbtt_vertex *pvertices; + int num_vertices; +} stbtt__csctx; + +#define STBTT__CSCTX_INIT(bounds) {bounds,0, 0,0, 0,0, 0,0,0,0, NULL, 0} + +static void stbtt__track_vertex(stbtt__csctx *c, stbtt_int32 x, stbtt_int32 y) +{ + if (x > c->max_x || !c->started) c->max_x = x; + if (y > c->max_y || !c->started) c->max_y = y; + if (x < c->min_x || !c->started) c->min_x = x; + if (y < c->min_y || !c->started) c->min_y = y; + c->started = 1; +} + +static void stbtt__csctx_v(stbtt__csctx *c, stbtt_uint8 type, stbtt_int32 x, stbtt_int32 y, stbtt_int32 cx, stbtt_int32 cy, stbtt_int32 cx1, stbtt_int32 cy1) +{ + if (c->bounds) { + stbtt__track_vertex(c, x, y); + if (type == STBTT_vcubic) { + stbtt__track_vertex(c, cx, cy); + stbtt__track_vertex(c, cx1, cy1); + } + } else { + stbtt_setvertex(&c->pvertices[c->num_vertices], type, x, y, cx, cy); + c->pvertices[c->num_vertices].cx1 = (stbtt_int16) cx1; + c->pvertices[c->num_vertices].cy1 = (stbtt_int16) cy1; + } + c->num_vertices++; +} + +static void stbtt__csctx_close_shape(stbtt__csctx *ctx) +{ + if (ctx->first_x != ctx->x || ctx->first_y != ctx->y) + stbtt__csctx_v(ctx, STBTT_vline, (int)ctx->first_x, (int)ctx->first_y, 0, 0, 0, 0); +} + +static void stbtt__csctx_rmove_to(stbtt__csctx *ctx, float dx, float dy) +{ + stbtt__csctx_close_shape(ctx); + ctx->first_x = ctx->x = ctx->x + dx; + ctx->first_y = ctx->y = ctx->y + dy; + stbtt__csctx_v(ctx, STBTT_vmove, (int)ctx->x, (int)ctx->y, 0, 0, 0, 0); +} + +static void stbtt__csctx_rline_to(stbtt__csctx *ctx, float dx, float dy) +{ + ctx->x += dx; + ctx->y += dy; + stbtt__csctx_v(ctx, STBTT_vline, (int)ctx->x, (int)ctx->y, 0, 0, 0, 0); +} + +static void stbtt__csctx_rccurve_to(stbtt__csctx *ctx, float dx1, float dy1, float dx2, float dy2, float dx3, float dy3) +{ + float cx1 = ctx->x + dx1; + float cy1 = ctx->y + dy1; + float cx2 = cx1 + dx2; + float cy2 = cy1 + dy2; + ctx->x = cx2 + dx3; + ctx->y = cy2 + dy3; + stbtt__csctx_v(ctx, STBTT_vcubic, (int)ctx->x, (int)ctx->y, (int)cx1, (int)cy1, (int)cx2, (int)cy2); +} + +static stbtt__buf stbtt__get_subr(stbtt__buf idx, int n) +{ + int count = stbtt__cff_index_count(&idx); + int bias = 107; + if (count >= 33900) + bias = 32768; + else if (count >= 1240) + bias = 1131; + n += bias; + if (n < 0 || n >= count) + return stbtt__new_buf(NULL, 0); + return stbtt__cff_index_get(idx, n); +} + +static stbtt__buf stbtt__cid_get_glyph_subrs(const stbtt_fontinfo *info, int glyph_index) +{ + stbtt__buf fdselect = info->fdselect; + int nranges, start, end, v, fmt, fdselector = -1, i; + + stbtt__buf_seek(&fdselect, 0); + fmt = stbtt__buf_get8(&fdselect); + if (fmt == 0) { + // untested + stbtt__buf_skip(&fdselect, glyph_index); + fdselector = stbtt__buf_get8(&fdselect); + } else if (fmt == 3) { + nranges = stbtt__buf_get16(&fdselect); + start = stbtt__buf_get16(&fdselect); + for (i = 0; i < nranges; i++) { + v = stbtt__buf_get8(&fdselect); + end = stbtt__buf_get16(&fdselect); + if (glyph_index >= start && glyph_index < end) { + fdselector = v; + break; + } + start = end; + } + } + if (fdselector == -1) stbtt__new_buf(NULL, 0); + return stbtt__get_subrs(info->cff, stbtt__cff_index_get(info->fontdicts, fdselector)); +} + +static int stbtt__run_charstring(const stbtt_fontinfo *info, int glyph_index, stbtt__csctx *c) +{ + int in_header = 1, maskbits = 0, subr_stack_height = 0, sp = 0, v, i, b0; + int has_subrs = 0, clear_stack; + float s[48]; + stbtt__buf subr_stack[10], subrs = info->subrs, b; + float f; + +#define STBTT__CSERR(s) (0) + + // this currently ignores the initial width value, which isn't needed if we have hmtx + b = stbtt__cff_index_get(info->charstrings, glyph_index); + while (b.cursor < b.size) { + i = 0; + clear_stack = 1; + b0 = stbtt__buf_get8(&b); + switch (b0) { + // @TODO implement hinting + case 0x13: // hintmask + case 0x14: // cntrmask + if (in_header) + maskbits += (sp / 2); // implicit "vstem" + in_header = 0; + stbtt__buf_skip(&b, (maskbits + 7) / 8); + break; + + case 0x01: // hstem + case 0x03: // vstem + case 0x12: // hstemhm + case 0x17: // vstemhm + maskbits += (sp / 2); + break; + + case 0x15: // rmoveto + in_header = 0; + if (sp < 2) return STBTT__CSERR("rmoveto stack"); + stbtt__csctx_rmove_to(c, s[sp-2], s[sp-1]); + break; + case 0x04: // vmoveto + in_header = 0; + if (sp < 1) return STBTT__CSERR("vmoveto stack"); + stbtt__csctx_rmove_to(c, 0, s[sp-1]); + break; + case 0x16: // hmoveto + in_header = 0; + if (sp < 1) return STBTT__CSERR("hmoveto stack"); + stbtt__csctx_rmove_to(c, s[sp-1], 0); + break; + + case 0x05: // rlineto + if (sp < 2) return STBTT__CSERR("rlineto stack"); + for (; i + 1 < sp; i += 2) + stbtt__csctx_rline_to(c, s[i], s[i+1]); + break; + + // hlineto/vlineto and vhcurveto/hvcurveto alternate horizontal and vertical + // starting from a different place. + + case 0x07: // vlineto + if (sp < 1) return STBTT__CSERR("vlineto stack"); + goto vlineto; + case 0x06: // hlineto + if (sp < 1) return STBTT__CSERR("hlineto stack"); + for (;;) { + if (i >= sp) break; + stbtt__csctx_rline_to(c, s[i], 0); + i++; + vlineto: + if (i >= sp) break; + stbtt__csctx_rline_to(c, 0, s[i]); + i++; + } + break; + + case 0x1F: // hvcurveto + if (sp < 4) return STBTT__CSERR("hvcurveto stack"); + goto hvcurveto; + case 0x1E: // vhcurveto + if (sp < 4) return STBTT__CSERR("vhcurveto stack"); + for (;;) { + if (i + 3 >= sp) break; + stbtt__csctx_rccurve_to(c, 0, s[i], s[i+1], s[i+2], s[i+3], (sp - i == 5) ? s[i + 4] : 0.0f); + i += 4; + hvcurveto: + if (i + 3 >= sp) break; + stbtt__csctx_rccurve_to(c, s[i], 0, s[i+1], s[i+2], (sp - i == 5) ? s[i+4] : 0.0f, s[i+3]); + i += 4; + } + break; + + case 0x08: // rrcurveto + if (sp < 6) return STBTT__CSERR("rcurveline stack"); + for (; i + 5 < sp; i += 6) + stbtt__csctx_rccurve_to(c, s[i], s[i+1], s[i+2], s[i+3], s[i+4], s[i+5]); + break; + + case 0x18: // rcurveline + if (sp < 8) return STBTT__CSERR("rcurveline stack"); + for (; i + 5 < sp - 2; i += 6) + stbtt__csctx_rccurve_to(c, s[i], s[i+1], s[i+2], s[i+3], s[i+4], s[i+5]); + if (i + 1 >= sp) return STBTT__CSERR("rcurveline stack"); + stbtt__csctx_rline_to(c, s[i], s[i+1]); + break; + + case 0x19: // rlinecurve + if (sp < 8) return STBTT__CSERR("rlinecurve stack"); + for (; i + 1 < sp - 6; i += 2) + stbtt__csctx_rline_to(c, s[i], s[i+1]); + if (i + 5 >= sp) return STBTT__CSERR("rlinecurve stack"); + stbtt__csctx_rccurve_to(c, s[i], s[i+1], s[i+2], s[i+3], s[i+4], s[i+5]); + break; + + case 0x1A: // vvcurveto + case 0x1B: // hhcurveto + if (sp < 4) return STBTT__CSERR("(vv|hh)curveto stack"); + f = 0.0; + if (sp & 1) { f = s[i]; i++; } + for (; i + 3 < sp; i += 4) { + if (b0 == 0x1B) + stbtt__csctx_rccurve_to(c, s[i], f, s[i+1], s[i+2], s[i+3], 0.0); + else + stbtt__csctx_rccurve_to(c, f, s[i], s[i+1], s[i+2], 0.0, s[i+3]); + f = 0.0; + } + break; + + case 0x0A: // callsubr + if (!has_subrs) { + if (info->fdselect.size) + subrs = stbtt__cid_get_glyph_subrs(info, glyph_index); + has_subrs = 1; + } + // FALLTHROUGH + case 0x1D: // callgsubr + if (sp < 1) return STBTT__CSERR("call(g|)subr stack"); + v = (int) s[--sp]; + if (subr_stack_height >= 10) return STBTT__CSERR("recursion limit"); + subr_stack[subr_stack_height++] = b; + b = stbtt__get_subr(b0 == 0x0A ? subrs : info->gsubrs, v); + if (b.size == 0) return STBTT__CSERR("subr not found"); + b.cursor = 0; + clear_stack = 0; + break; + + case 0x0B: // return + if (subr_stack_height <= 0) return STBTT__CSERR("return outside subr"); + b = subr_stack[--subr_stack_height]; + clear_stack = 0; + break; + + case 0x0E: // endchar + stbtt__csctx_close_shape(c); + return 1; + + case 0x0C: { // two-byte escape + float dx1, dx2, dx3, dx4, dx5, dx6, dy1, dy2, dy3, dy4, dy5, dy6; + float dx, dy; + int b1 = stbtt__buf_get8(&b); + switch (b1) { + // @TODO These "flex" implementations ignore the flex-depth and resolution, + // and always draw beziers. + case 0x22: // hflex + if (sp < 7) return STBTT__CSERR("hflex stack"); + dx1 = s[0]; + dx2 = s[1]; + dy2 = s[2]; + dx3 = s[3]; + dx4 = s[4]; + dx5 = s[5]; + dx6 = s[6]; + stbtt__csctx_rccurve_to(c, dx1, 0, dx2, dy2, dx3, 0); + stbtt__csctx_rccurve_to(c, dx4, 0, dx5, -dy2, dx6, 0); + break; + + case 0x23: // flex + if (sp < 13) return STBTT__CSERR("flex stack"); + dx1 = s[0]; + dy1 = s[1]; + dx2 = s[2]; + dy2 = s[3]; + dx3 = s[4]; + dy3 = s[5]; + dx4 = s[6]; + dy4 = s[7]; + dx5 = s[8]; + dy5 = s[9]; + dx6 = s[10]; + dy6 = s[11]; + //fd is s[12] + stbtt__csctx_rccurve_to(c, dx1, dy1, dx2, dy2, dx3, dy3); + stbtt__csctx_rccurve_to(c, dx4, dy4, dx5, dy5, dx6, dy6); + break; + + case 0x24: // hflex1 + if (sp < 9) return STBTT__CSERR("hflex1 stack"); + dx1 = s[0]; + dy1 = s[1]; + dx2 = s[2]; + dy2 = s[3]; + dx3 = s[4]; + dx4 = s[5]; + dx5 = s[6]; + dy5 = s[7]; + dx6 = s[8]; + stbtt__csctx_rccurve_to(c, dx1, dy1, dx2, dy2, dx3, 0); + stbtt__csctx_rccurve_to(c, dx4, 0, dx5, dy5, dx6, -(dy1+dy2+dy5)); + break; + + case 0x25: // flex1 + if (sp < 11) return STBTT__CSERR("flex1 stack"); + dx1 = s[0]; + dy1 = s[1]; + dx2 = s[2]; + dy2 = s[3]; + dx3 = s[4]; + dy3 = s[5]; + dx4 = s[6]; + dy4 = s[7]; + dx5 = s[8]; + dy5 = s[9]; + dx6 = dy6 = s[10]; + dx = dx1+dx2+dx3+dx4+dx5; + dy = dy1+dy2+dy3+dy4+dy5; + if (STBTT_fabs(dx) > STBTT_fabs(dy)) + dy6 = -dy; + else + dx6 = -dx; + stbtt__csctx_rccurve_to(c, dx1, dy1, dx2, dy2, dx3, dy3); + stbtt__csctx_rccurve_to(c, dx4, dy4, dx5, dy5, dx6, dy6); + break; + + default: + return STBTT__CSERR("unimplemented"); + } + } break; + + default: + if (b0 != 255 && b0 != 28 && b0 < 32) + return STBTT__CSERR("reserved operator"); + + // push immediate + if (b0 == 255) { + f = (float)(stbtt_int32)stbtt__buf_get32(&b) / 0x10000; + } else { + stbtt__buf_skip(&b, -1); + f = (float)(stbtt_int16)stbtt__cff_int(&b); + } + if (sp >= 48) return STBTT__CSERR("push stack overflow"); + s[sp++] = f; + clear_stack = 0; + break; + } + if (clear_stack) sp = 0; + } + return STBTT__CSERR("no endchar"); + +#undef STBTT__CSERR +} + +static int stbtt__GetGlyphShapeT2(const stbtt_fontinfo *info, int glyph_index, stbtt_vertex **pvertices) +{ + // runs the charstring twice, once to count and once to output (to avoid realloc) + stbtt__csctx count_ctx = STBTT__CSCTX_INIT(1); + stbtt__csctx output_ctx = STBTT__CSCTX_INIT(0); + if (stbtt__run_charstring(info, glyph_index, &count_ctx)) { + *pvertices = (stbtt_vertex*)STBTT_malloc(count_ctx.num_vertices*sizeof(stbtt_vertex), info->userdata); + output_ctx.pvertices = *pvertices; + if (stbtt__run_charstring(info, glyph_index, &output_ctx)) { + STBTT_assert(output_ctx.num_vertices == count_ctx.num_vertices); + return output_ctx.num_vertices; + } + } + *pvertices = NULL; + return 0; +} + +static int stbtt__GetGlyphInfoT2(const stbtt_fontinfo *info, int glyph_index, int *x0, int *y0, int *x1, int *y1) +{ + stbtt__csctx c = STBTT__CSCTX_INIT(1); + int r = stbtt__run_charstring(info, glyph_index, &c); + if (x0) *x0 = r ? c.min_x : 0; + if (y0) *y0 = r ? c.min_y : 0; + if (x1) *x1 = r ? c.max_x : 0; + if (y1) *y1 = r ? c.max_y : 0; + return r ? c.num_vertices : 0; +} + +STBTT_DEF int stbtt_GetGlyphShape(const stbtt_fontinfo *info, int glyph_index, stbtt_vertex **pvertices) +{ + if (!info->cff.size) + return stbtt__GetGlyphShapeTT(info, glyph_index, pvertices); + else + return stbtt__GetGlyphShapeT2(info, glyph_index, pvertices); +} + +STBTT_DEF void stbtt_GetGlyphHMetrics(const stbtt_fontinfo *info, int glyph_index, int *advanceWidth, int *leftSideBearing) +{ + stbtt_uint16 numOfLongHorMetrics = ttUSHORT(info->data+info->hhea + 34); + if (glyph_index < numOfLongHorMetrics) { + if (advanceWidth) *advanceWidth = ttSHORT(info->data + info->hmtx + 4*glyph_index); + if (leftSideBearing) *leftSideBearing = ttSHORT(info->data + info->hmtx + 4*glyph_index + 2); + } else { + if (advanceWidth) *advanceWidth = ttSHORT(info->data + info->hmtx + 4*(numOfLongHorMetrics-1)); + if (leftSideBearing) *leftSideBearing = ttSHORT(info->data + info->hmtx + 4*numOfLongHorMetrics + 2*(glyph_index - numOfLongHorMetrics)); + } +} + +STBTT_DEF int stbtt_GetKerningTableLength(const stbtt_fontinfo *info) +{ + stbtt_uint8 *data = info->data + info->kern; + + // we only look at the first table. it must be 'horizontal' and format 0. + if (!info->kern) + return 0; + if (ttUSHORT(data+2) < 1) // number of tables, need at least 1 + return 0; + if (ttUSHORT(data+8) != 1) // horizontal flag must be set in format + return 0; + + return ttUSHORT(data+10); +} + +STBTT_DEF int stbtt_GetKerningTable(const stbtt_fontinfo *info, stbtt_kerningentry* table, int table_length) +{ + stbtt_uint8 *data = info->data + info->kern; + int k, length; + + // we only look at the first table. it must be 'horizontal' and format 0. + if (!info->kern) + return 0; + if (ttUSHORT(data+2) < 1) // number of tables, need at least 1 + return 0; + if (ttUSHORT(data+8) != 1) // horizontal flag must be set in format + return 0; + + length = ttUSHORT(data+10); + if (table_length < length) + length = table_length; + + for (k = 0; k < length; k++) + { + table[k].glyph1 = ttUSHORT(data+18+(k*6)); + table[k].glyph2 = ttUSHORT(data+20+(k*6)); + table[k].advance = ttSHORT(data+22+(k*6)); + } + + return length; +} + +static int stbtt__GetGlyphKernInfoAdvance(const stbtt_fontinfo *info, int glyph1, int glyph2) +{ + stbtt_uint8 *data = info->data + info->kern; + stbtt_uint32 needle, straw; + int l, r, m; + + // we only look at the first table. it must be 'horizontal' and format 0. + if (!info->kern) + return 0; + if (ttUSHORT(data+2) < 1) // number of tables, need at least 1 + return 0; + if (ttUSHORT(data+8) != 1) // horizontal flag must be set in format + return 0; + + l = 0; + r = ttUSHORT(data+10) - 1; + needle = glyph1 << 16 | glyph2; + while (l <= r) { + m = (l + r) >> 1; + straw = ttULONG(data+18+(m*6)); // note: unaligned read + if (needle < straw) + r = m - 1; + else if (needle > straw) + l = m + 1; + else + return ttSHORT(data+22+(m*6)); + } + return 0; +} + +static stbtt_int32 stbtt__GetCoverageIndex(stbtt_uint8 *coverageTable, int glyph) +{ + stbtt_uint16 coverageFormat = ttUSHORT(coverageTable); + switch (coverageFormat) { + case 1: { + stbtt_uint16 glyphCount = ttUSHORT(coverageTable + 2); + + // Binary search. + stbtt_int32 l=0, r=glyphCount-1, m; + int straw, needle=glyph; + while (l <= r) { + stbtt_uint8 *glyphArray = coverageTable + 4; + stbtt_uint16 glyphID; + m = (l + r) >> 1; + glyphID = ttUSHORT(glyphArray + 2 * m); + straw = glyphID; + if (needle < straw) + r = m - 1; + else if (needle > straw) + l = m + 1; + else { + return m; + } + } + break; + } + + case 2: { + stbtt_uint16 rangeCount = ttUSHORT(coverageTable + 2); + stbtt_uint8 *rangeArray = coverageTable + 4; + + // Binary search. + stbtt_int32 l=0, r=rangeCount-1, m; + int strawStart, strawEnd, needle=glyph; + while (l <= r) { + stbtt_uint8 *rangeRecord; + m = (l + r) >> 1; + rangeRecord = rangeArray + 6 * m; + strawStart = ttUSHORT(rangeRecord); + strawEnd = ttUSHORT(rangeRecord + 2); + if (needle < strawStart) + r = m - 1; + else if (needle > strawEnd) + l = m + 1; + else { + stbtt_uint16 startCoverageIndex = ttUSHORT(rangeRecord + 4); + return startCoverageIndex + glyph - strawStart; + } + } + break; + } + + default: return -1; // unsupported + } + + return -1; +} + +static stbtt_int32 stbtt__GetGlyphClass(stbtt_uint8 *classDefTable, int glyph) +{ + stbtt_uint16 classDefFormat = ttUSHORT(classDefTable); + switch (classDefFormat) + { + case 1: { + stbtt_uint16 startGlyphID = ttUSHORT(classDefTable + 2); + stbtt_uint16 glyphCount = ttUSHORT(classDefTable + 4); + stbtt_uint8 *classDef1ValueArray = classDefTable + 6; + + if (glyph >= startGlyphID && glyph < startGlyphID + glyphCount) + return (stbtt_int32)ttUSHORT(classDef1ValueArray + 2 * (glyph - startGlyphID)); + break; + } + + case 2: { + stbtt_uint16 classRangeCount = ttUSHORT(classDefTable + 2); + stbtt_uint8 *classRangeRecords = classDefTable + 4; + + // Binary search. + stbtt_int32 l=0, r=classRangeCount-1, m; + int strawStart, strawEnd, needle=glyph; + while (l <= r) { + stbtt_uint8 *classRangeRecord; + m = (l + r) >> 1; + classRangeRecord = classRangeRecords + 6 * m; + strawStart = ttUSHORT(classRangeRecord); + strawEnd = ttUSHORT(classRangeRecord + 2); + if (needle < strawStart) + r = m - 1; + else if (needle > strawEnd) + l = m + 1; + else + return (stbtt_int32)ttUSHORT(classRangeRecord + 4); + } + break; + } + + default: + return -1; // Unsupported definition type, return an error. + } + + // "All glyphs not assigned to a class fall into class 0". (OpenType spec) + return 0; +} + +// Define to STBTT_assert(x) if you want to break on unimplemented formats. +#define STBTT_GPOS_TODO_assert(x) + +static stbtt_int32 stbtt__GetGlyphGPOSInfoAdvance(const stbtt_fontinfo *info, int glyph1, int glyph2) +{ + stbtt_uint16 lookupListOffset; + stbtt_uint8 *lookupList; + stbtt_uint16 lookupCount; + stbtt_uint8 *data; + stbtt_int32 i, sti; + + if (!info->gpos) return 0; + + data = info->data + info->gpos; + + if (ttUSHORT(data+0) != 1) return 0; // Major version 1 + if (ttUSHORT(data+2) != 0) return 0; // Minor version 0 + + lookupListOffset = ttUSHORT(data+8); + lookupList = data + lookupListOffset; + lookupCount = ttUSHORT(lookupList); + + for (i=0; i<lookupCount; ++i) { + stbtt_uint16 lookupOffset = ttUSHORT(lookupList + 2 + 2 * i); + stbtt_uint8 *lookupTable = lookupList + lookupOffset; + + stbtt_uint16 lookupType = ttUSHORT(lookupTable); + stbtt_uint16 subTableCount = ttUSHORT(lookupTable + 4); + stbtt_uint8 *subTableOffsets = lookupTable + 6; + if (lookupType != 2) // Pair Adjustment Positioning Subtable + continue; + + for (sti=0; sti<subTableCount; sti++) { + stbtt_uint16 subtableOffset = ttUSHORT(subTableOffsets + 2 * sti); + stbtt_uint8 *table = lookupTable + subtableOffset; + stbtt_uint16 posFormat = ttUSHORT(table); + stbtt_uint16 coverageOffset = ttUSHORT(table + 2); + stbtt_int32 coverageIndex = stbtt__GetCoverageIndex(table + coverageOffset, glyph1); + if (coverageIndex == -1) continue; + + switch (posFormat) { + case 1: { + stbtt_int32 l, r, m; + int straw, needle; + stbtt_uint16 valueFormat1 = ttUSHORT(table + 4); + stbtt_uint16 valueFormat2 = ttUSHORT(table + 6); + if (valueFormat1 == 4 && valueFormat2 == 0) { // Support more formats? + stbtt_int32 valueRecordPairSizeInBytes = 2; + stbtt_uint16 pairSetCount = ttUSHORT(table + 8); + stbtt_uint16 pairPosOffset = ttUSHORT(table + 10 + 2 * coverageIndex); + stbtt_uint8 *pairValueTable = table + pairPosOffset; + stbtt_uint16 pairValueCount = ttUSHORT(pairValueTable); + stbtt_uint8 *pairValueArray = pairValueTable + 2; + + if (coverageIndex >= pairSetCount) return 0; + + needle=glyph2; + r=pairValueCount-1; + l=0; + + // Binary search. + while (l <= r) { + stbtt_uint16 secondGlyph; + stbtt_uint8 *pairValue; + m = (l + r) >> 1; + pairValue = pairValueArray + (2 + valueRecordPairSizeInBytes) * m; + secondGlyph = ttUSHORT(pairValue); + straw = secondGlyph; + if (needle < straw) + r = m - 1; + else if (needle > straw) + l = m + 1; + else { + stbtt_int16 xAdvance = ttSHORT(pairValue + 2); + return xAdvance; + } + } + } else + return 0; + break; + } + + case 2: { + stbtt_uint16 valueFormat1 = ttUSHORT(table + 4); + stbtt_uint16 valueFormat2 = ttUSHORT(table + 6); + if (valueFormat1 == 4 && valueFormat2 == 0) { // Support more formats? + stbtt_uint16 classDef1Offset = ttUSHORT(table + 8); + stbtt_uint16 classDef2Offset = ttUSHORT(table + 10); + int glyph1class = stbtt__GetGlyphClass(table + classDef1Offset, glyph1); + int glyph2class = stbtt__GetGlyphClass(table + classDef2Offset, glyph2); + + stbtt_uint16 class1Count = ttUSHORT(table + 12); + stbtt_uint16 class2Count = ttUSHORT(table + 14); + stbtt_uint8 *class1Records, *class2Records; + stbtt_int16 xAdvance; + + if (glyph1class < 0 || glyph1class >= class1Count) return 0; // malformed + if (glyph2class < 0 || glyph2class >= class2Count) return 0; // malformed + + class1Records = table + 16; + class2Records = class1Records + 2 * (glyph1class * class2Count); + xAdvance = ttSHORT(class2Records + 2 * glyph2class); + return xAdvance; + } else + return 0; + break; + } + + default: + return 0; // Unsupported position format + } + } + } + + return 0; +} + +STBTT_DEF int stbtt_GetGlyphKernAdvance(const stbtt_fontinfo *info, int g1, int g2) +{ + int xAdvance = 0; + + if (info->gpos) + xAdvance += stbtt__GetGlyphGPOSInfoAdvance(info, g1, g2); + else if (info->kern) + xAdvance += stbtt__GetGlyphKernInfoAdvance(info, g1, g2); + + return xAdvance; +} + +STBTT_DEF int stbtt_GetCodepointKernAdvance(const stbtt_fontinfo *info, int ch1, int ch2) +{ + if (!info->kern && !info->gpos) // if no kerning table, don't waste time looking up both codepoint->glyphs + return 0; + return stbtt_GetGlyphKernAdvance(info, stbtt_FindGlyphIndex(info,ch1), stbtt_FindGlyphIndex(info,ch2)); +} + +STBTT_DEF void stbtt_GetCodepointHMetrics(const stbtt_fontinfo *info, int codepoint, int *advanceWidth, int *leftSideBearing) +{ + stbtt_GetGlyphHMetrics(info, stbtt_FindGlyphIndex(info,codepoint), advanceWidth, leftSideBearing); +} + +STBTT_DEF void stbtt_GetFontVMetrics(const stbtt_fontinfo *info, int *ascent, int *descent, int *lineGap) +{ + if (ascent ) *ascent = ttSHORT(info->data+info->hhea + 4); + if (descent) *descent = ttSHORT(info->data+info->hhea + 6); + if (lineGap) *lineGap = ttSHORT(info->data+info->hhea + 8); +} + +STBTT_DEF int stbtt_GetFontVMetricsOS2(const stbtt_fontinfo *info, int *typoAscent, int *typoDescent, int *typoLineGap) +{ + int tab = stbtt__find_table(info->data, info->fontstart, "OS/2"); + if (!tab) + return 0; + if (typoAscent ) *typoAscent = ttSHORT(info->data+tab + 68); + if (typoDescent) *typoDescent = ttSHORT(info->data+tab + 70); + if (typoLineGap) *typoLineGap = ttSHORT(info->data+tab + 72); + return 1; +} + +STBTT_DEF void stbtt_GetFontBoundingBox(const stbtt_fontinfo *info, int *x0, int *y0, int *x1, int *y1) +{ + *x0 = ttSHORT(info->data + info->head + 36); + *y0 = ttSHORT(info->data + info->head + 38); + *x1 = ttSHORT(info->data + info->head + 40); + *y1 = ttSHORT(info->data + info->head + 42); +} + +STBTT_DEF float stbtt_ScaleForPixelHeight(const stbtt_fontinfo *info, float height) +{ + int fheight = ttSHORT(info->data + info->hhea + 4) - ttSHORT(info->data + info->hhea + 6); + return (float) height / fheight; +} + +STBTT_DEF float stbtt_ScaleForMappingEmToPixels(const stbtt_fontinfo *info, float pixels) +{ + int unitsPerEm = ttUSHORT(info->data + info->head + 18); + return pixels / unitsPerEm; +} + +STBTT_DEF void stbtt_FreeShape(const stbtt_fontinfo *info, stbtt_vertex *v) +{ + STBTT_free(v, info->userdata); +} + +STBTT_DEF stbtt_uint8 *stbtt_FindSVGDoc(const stbtt_fontinfo *info, int gl) +{ + int i; + stbtt_uint8 *data = info->data; + stbtt_uint8 *svg_doc_list = data + stbtt__get_svg((stbtt_fontinfo *) info); + + int numEntries = ttUSHORT(svg_doc_list); + stbtt_uint8 *svg_docs = svg_doc_list + 2; + + for(i=0; i<numEntries; i++) { + stbtt_uint8 *svg_doc = svg_docs + (12 * i); + if ((gl >= ttUSHORT(svg_doc)) && (gl <= ttUSHORT(svg_doc + 2))) + return svg_doc; + } + return 0; +} + +STBTT_DEF int stbtt_GetGlyphSVG(const stbtt_fontinfo *info, int gl, const char **svg) +{ + stbtt_uint8 *data = info->data; + stbtt_uint8 *svg_doc; + + if (info->svg == 0) + return 0; + + svg_doc = stbtt_FindSVGDoc(info, gl); + if (svg_doc != NULL) { + *svg = (char *) data + info->svg + ttULONG(svg_doc + 4); + return ttULONG(svg_doc + 8); + } else { + return 0; + } +} + +STBTT_DEF int stbtt_GetCodepointSVG(const stbtt_fontinfo *info, int unicode_codepoint, const char **svg) +{ + return stbtt_GetGlyphSVG(info, stbtt_FindGlyphIndex(info, unicode_codepoint), svg); +} + +////////////////////////////////////////////////////////////////////////////// +// +// antialiasing software rasterizer +// + +STBTT_DEF void stbtt_GetGlyphBitmapBoxSubpixel(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y,float shift_x, float shift_y, int *ix0, int *iy0, int *ix1, int *iy1) +{ + int x0=0,y0=0,x1,y1; // =0 suppresses compiler warning + if (!stbtt_GetGlyphBox(font, glyph, &x0,&y0,&x1,&y1)) { + // e.g. space character + if (ix0) *ix0 = 0; + if (iy0) *iy0 = 0; + if (ix1) *ix1 = 0; + if (iy1) *iy1 = 0; + } else { + // move to integral bboxes (treating pixels as little squares, what pixels get touched)? + if (ix0) *ix0 = STBTT_ifloor( x0 * scale_x + shift_x); + if (iy0) *iy0 = STBTT_ifloor(-y1 * scale_y + shift_y); + if (ix1) *ix1 = STBTT_iceil ( x1 * scale_x + shift_x); + if (iy1) *iy1 = STBTT_iceil (-y0 * scale_y + shift_y); + } +} + +STBTT_DEF void stbtt_GetGlyphBitmapBox(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1) +{ + stbtt_GetGlyphBitmapBoxSubpixel(font, glyph, scale_x, scale_y,0.0f,0.0f, ix0, iy0, ix1, iy1); +} + +STBTT_DEF void stbtt_GetCodepointBitmapBoxSubpixel(const stbtt_fontinfo *font, int codepoint, float scale_x, float scale_y, float shift_x, float shift_y, int *ix0, int *iy0, int *ix1, int *iy1) +{ + stbtt_GetGlyphBitmapBoxSubpixel(font, stbtt_FindGlyphIndex(font,codepoint), scale_x, scale_y,shift_x,shift_y, ix0,iy0,ix1,iy1); +} + +STBTT_DEF void stbtt_GetCodepointBitmapBox(const stbtt_fontinfo *font, int codepoint, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1) +{ + stbtt_GetCodepointBitmapBoxSubpixel(font, codepoint, scale_x, scale_y,0.0f,0.0f, ix0,iy0,ix1,iy1); +} + +////////////////////////////////////////////////////////////////////////////// +// +// Rasterizer + +typedef struct stbtt__hheap_chunk +{ + struct stbtt__hheap_chunk *next; +} stbtt__hheap_chunk; + +typedef struct stbtt__hheap +{ + struct stbtt__hheap_chunk *head; + void *first_free; + int num_remaining_in_head_chunk; +} stbtt__hheap; + +static void *stbtt__hheap_alloc(stbtt__hheap *hh, size_t size, void *userdata) +{ + if (hh->first_free) { + void *p = hh->first_free; + hh->first_free = * (void **) p; + return p; + } else { + if (hh->num_remaining_in_head_chunk == 0) { + int count = (size < 32 ? 2000 : size < 128 ? 800 : 100); + stbtt__hheap_chunk *c = (stbtt__hheap_chunk *) STBTT_malloc(sizeof(stbtt__hheap_chunk) + size * count, userdata); + if (c == NULL) + return NULL; + c->next = hh->head; + hh->head = c; + hh->num_remaining_in_head_chunk = count; + } + --hh->num_remaining_in_head_chunk; + return (char *) (hh->head) + sizeof(stbtt__hheap_chunk) + size * hh->num_remaining_in_head_chunk; + } +} + +static void stbtt__hheap_free(stbtt__hheap *hh, void *p) +{ + *(void **) p = hh->first_free; + hh->first_free = p; +} + +static void stbtt__hheap_cleanup(stbtt__hheap *hh, void *userdata) +{ + stbtt__hheap_chunk *c = hh->head; + while (c) { + stbtt__hheap_chunk *n = c->next; + STBTT_free(c, userdata); + c = n; + } +} + +typedef struct stbtt__edge { + float x0,y0, x1,y1; + int invert; +} stbtt__edge; + + +typedef struct stbtt__active_edge +{ + struct stbtt__active_edge *next; + #if STBTT_RASTERIZER_VERSION==1 + int x,dx; + float ey; + int direction; + #elif STBTT_RASTERIZER_VERSION==2 + float fx,fdx,fdy; + float direction; + float sy; + float ey; + #else + #error "Unrecognized value of STBTT_RASTERIZER_VERSION" + #endif +} stbtt__active_edge; + +#if STBTT_RASTERIZER_VERSION == 1 +#define STBTT_FIXSHIFT 10 +#define STBTT_FIX (1 << STBTT_FIXSHIFT) +#define STBTT_FIXMASK (STBTT_FIX-1) + +static stbtt__active_edge *stbtt__new_active(stbtt__hheap *hh, stbtt__edge *e, int off_x, float start_point, void *userdata) +{ + stbtt__active_edge *z = (stbtt__active_edge *) stbtt__hheap_alloc(hh, sizeof(*z), userdata); + float dxdy = (e->x1 - e->x0) / (e->y1 - e->y0); + STBTT_assert(z != NULL); + if (!z) return z; + + // round dx down to avoid overshooting + if (dxdy < 0) + z->dx = -STBTT_ifloor(STBTT_FIX * -dxdy); + else + z->dx = STBTT_ifloor(STBTT_FIX * dxdy); + + z->x = STBTT_ifloor(STBTT_FIX * e->x0 + z->dx * (start_point - e->y0)); // use z->dx so when we offset later it's by the same amount + z->x -= off_x * STBTT_FIX; + + z->ey = e->y1; + z->next = 0; + z->direction = e->invert ? 1 : -1; + return z; +} +#elif STBTT_RASTERIZER_VERSION == 2 +static stbtt__active_edge *stbtt__new_active(stbtt__hheap *hh, stbtt__edge *e, int off_x, float start_point, void *userdata) +{ + stbtt__active_edge *z = (stbtt__active_edge *) stbtt__hheap_alloc(hh, sizeof(*z), userdata); + float dxdy = (e->x1 - e->x0) / (e->y1 - e->y0); + STBTT_assert(z != NULL); + //STBTT_assert(e->y0 <= start_point); + if (!z) return z; + z->fdx = dxdy; + z->fdy = dxdy != 0.0f ? (1.0f/dxdy) : 0.0f; + z->fx = e->x0 + dxdy * (start_point - e->y0); + z->fx -= off_x; + z->direction = e->invert ? 1.0f : -1.0f; + z->sy = e->y0; + z->ey = e->y1; + z->next = 0; + return z; +} +#else +#error "Unrecognized value of STBTT_RASTERIZER_VERSION" +#endif + +#if STBTT_RASTERIZER_VERSION == 1 +// note: this routine clips fills that extend off the edges... ideally this +// wouldn't happen, but it could happen if the truetype glyph bounding boxes +// are wrong, or if the user supplies a too-small bitmap +static void stbtt__fill_active_edges(unsigned char *scanline, int len, stbtt__active_edge *e, int max_weight) +{ + // non-zero winding fill + int x0=0, w=0; + + while (e) { + if (w == 0) { + // if we're currently at zero, we need to record the edge start point + x0 = e->x; w += e->direction; + } else { + int x1 = e->x; w += e->direction; + // if we went to zero, we need to draw + if (w == 0) { + int i = x0 >> STBTT_FIXSHIFT; + int j = x1 >> STBTT_FIXSHIFT; + + if (i < len && j >= 0) { + if (i == j) { + // x0,x1 are the same pixel, so compute combined coverage + scanline[i] = scanline[i] + (stbtt_uint8) ((x1 - x0) * max_weight >> STBTT_FIXSHIFT); + } else { + if (i >= 0) // add antialiasing for x0 + scanline[i] = scanline[i] + (stbtt_uint8) (((STBTT_FIX - (x0 & STBTT_FIXMASK)) * max_weight) >> STBTT_FIXSHIFT); + else + i = -1; // clip + + if (j < len) // add antialiasing for x1 + scanline[j] = scanline[j] + (stbtt_uint8) (((x1 & STBTT_FIXMASK) * max_weight) >> STBTT_FIXSHIFT); + else + j = len; // clip + + for (++i; i < j; ++i) // fill pixels between x0 and x1 + scanline[i] = scanline[i] + (stbtt_uint8) max_weight; + } + } + } + } + + e = e->next; + } +} + +static void stbtt__rasterize_sorted_edges(stbtt__bitmap *result, stbtt__edge *e, int n, int vsubsample, int off_x, int off_y, void *userdata) +{ + stbtt__hheap hh = { 0, 0, 0 }; + stbtt__active_edge *active = NULL; + int y,j=0; + int max_weight = (255 / vsubsample); // weight per vertical scanline + int s; // vertical subsample index + unsigned char scanline_data[512], *scanline; + + if (result->w > 512) + scanline = (unsigned char *) STBTT_malloc(result->w, userdata); + else + scanline = scanline_data; + + y = off_y * vsubsample; + e[n].y0 = (off_y + result->h) * (float) vsubsample + 1; + + while (j < result->h) { + STBTT_memset(scanline, 0, result->w); + for (s=0; s < vsubsample; ++s) { + // find center of pixel for this scanline + float scan_y = y + 0.5f; + stbtt__active_edge **step = &active; + + // update all active edges; + // remove all active edges that terminate before the center of this scanline + while (*step) { + stbtt__active_edge * z = *step; + if (z->ey <= scan_y) { + *step = z->next; // delete from list + STBTT_assert(z->direction); + z->direction = 0; + stbtt__hheap_free(&hh, z); + } else { + z->x += z->dx; // advance to position for current scanline + step = &((*step)->next); // advance through list + } + } + + // resort the list if needed + for(;;) { + int changed=0; + step = &active; + while (*step && (*step)->next) { + if ((*step)->x > (*step)->next->x) { + stbtt__active_edge *t = *step; + stbtt__active_edge *q = t->next; + + t->next = q->next; + q->next = t; + *step = q; + changed = 1; + } + step = &(*step)->next; + } + if (!changed) break; + } + + // insert all edges that start before the center of this scanline -- omit ones that also end on this scanline + while (e->y0 <= scan_y) { + if (e->y1 > scan_y) { + stbtt__active_edge *z = stbtt__new_active(&hh, e, off_x, scan_y, userdata); + if (z != NULL) { + // find insertion point + if (active == NULL) + active = z; + else if (z->x < active->x) { + // insert at front + z->next = active; + active = z; + } else { + // find thing to insert AFTER + stbtt__active_edge *p = active; + while (p->next && p->next->x < z->x) + p = p->next; + // at this point, p->next->x is NOT < z->x + z->next = p->next; + p->next = z; + } + } + } + ++e; + } + + // now process all active edges in XOR fashion + if (active) + stbtt__fill_active_edges(scanline, result->w, active, max_weight); + + ++y; + } + STBTT_memcpy(result->pixels + j * result->stride, scanline, result->w); + ++j; + } + + stbtt__hheap_cleanup(&hh, userdata); + + if (scanline != scanline_data) + STBTT_free(scanline, userdata); +} + +#elif STBTT_RASTERIZER_VERSION == 2 + +// the edge passed in here does not cross the vertical line at x or the vertical line at x+1 +// (i.e. it has already been clipped to those) +static void stbtt__handle_clipped_edge(float *scanline, int x, stbtt__active_edge *e, float x0, float y0, float x1, float y1) +{ + if (y0 == y1) return; + STBTT_assert(y0 < y1); + STBTT_assert(e->sy <= e->ey); + if (y0 > e->ey) return; + if (y1 < e->sy) return; + if (y0 < e->sy) { + x0 += (x1-x0) * (e->sy - y0) / (y1-y0); + y0 = e->sy; + } + if (y1 > e->ey) { + x1 += (x1-x0) * (e->ey - y1) / (y1-y0); + y1 = e->ey; + } + + if (x0 == x) + STBTT_assert(x1 <= x+1); + else if (x0 == x+1) + STBTT_assert(x1 >= x); + else if (x0 <= x) + STBTT_assert(x1 <= x); + else if (x0 >= x+1) + STBTT_assert(x1 >= x+1); + else + STBTT_assert(x1 >= x && x1 <= x+1); + + if (x0 <= x && x1 <= x) + scanline[x] += e->direction * (y1-y0); + else if (x0 >= x+1 && x1 >= x+1) + ; + else { + STBTT_assert(x0 >= x && x0 <= x+1 && x1 >= x && x1 <= x+1); + scanline[x] += e->direction * (y1-y0) * (1-((x0-x)+(x1-x))/2); // coverage = 1 - average x position + } +} + +static float stbtt__sized_trapezoid_area(float height, float top_width, float bottom_width) +{ + STBTT_assert(top_width >= 0); + STBTT_assert(bottom_width >= 0); + return (top_width + bottom_width) / 2.0f * height; +} + +static float stbtt__position_trapezoid_area(float height, float tx0, float tx1, float bx0, float bx1) +{ + return stbtt__sized_trapezoid_area(height, tx1 - tx0, bx1 - bx0); +} + +static float stbtt__sized_triangle_area(float height, float width) +{ + return height * width / 2; +} + +static void stbtt__fill_active_edges_new(float *scanline, float *scanline_fill, int len, stbtt__active_edge *e, float y_top) +{ + float y_bottom = y_top+1; + + while (e) { + // brute force every pixel + + // compute intersection points with top & bottom + STBTT_assert(e->ey >= y_top); + + if (e->fdx == 0) { + float x0 = e->fx; + if (x0 < len) { + if (x0 >= 0) { + stbtt__handle_clipped_edge(scanline,(int) x0,e, x0,y_top, x0,y_bottom); + stbtt__handle_clipped_edge(scanline_fill-1,(int) x0+1,e, x0,y_top, x0,y_bottom); + } else { + stbtt__handle_clipped_edge(scanline_fill-1,0,e, x0,y_top, x0,y_bottom); + } + } + } else { + float x0 = e->fx; + float dx = e->fdx; + float xb = x0 + dx; + float x_top, x_bottom; + float sy0,sy1; + float dy = e->fdy; + STBTT_assert(e->sy <= y_bottom && e->ey >= y_top); + + // compute endpoints of line segment clipped to this scanline (if the + // line segment starts on this scanline. x0 is the intersection of the + // line with y_top, but that may be off the line segment. + if (e->sy > y_top) { + x_top = x0 + dx * (e->sy - y_top); + sy0 = e->sy; + } else { + x_top = x0; + sy0 = y_top; + } + if (e->ey < y_bottom) { + x_bottom = x0 + dx * (e->ey - y_top); + sy1 = e->ey; + } else { + x_bottom = xb; + sy1 = y_bottom; + } + + if (x_top >= 0 && x_bottom >= 0 && x_top < len && x_bottom < len) { + // from here on, we don't have to range check x values + + if ((int) x_top == (int) x_bottom) { + float height; + // simple case, only spans one pixel + int x = (int) x_top; + height = (sy1 - sy0) * e->direction; + STBTT_assert(x >= 0 && x < len); + scanline[x] += stbtt__position_trapezoid_area(height, x_top, x+1.0f, x_bottom, x+1.0f); + scanline_fill[x] += height; // everything right of this pixel is filled + } else { + int x,x1,x2; + float y_crossing, y_final, step, sign, area; + // covers 2+ pixels + if (x_top > x_bottom) { + // flip scanline vertically; signed area is the same + float t; + sy0 = y_bottom - (sy0 - y_top); + sy1 = y_bottom - (sy1 - y_top); + t = sy0, sy0 = sy1, sy1 = t; + t = x_bottom, x_bottom = x_top, x_top = t; + dx = -dx; + dy = -dy; + t = x0, x0 = xb, xb = t; + } + STBTT_assert(dy >= 0); + STBTT_assert(dx >= 0); + + x1 = (int) x_top; + x2 = (int) x_bottom; + // compute intersection with y axis at x1+1 + y_crossing = y_top + dy * (x1+1 - x0); + + // compute intersection with y axis at x2 + y_final = y_top + dy * (x2 - x0); + + // x1 x_top x2 x_bottom + // y_top +------|-----+------------+------------+--------|---+------------+ + // | | | | | | + // | | | | | | + // sy0 | Txxxxx|............|............|............|............| + // y_crossing | *xxxxx.......|............|............|............| + // | | xxxxx..|............|............|............| + // | | /- xx*xxxx........|............|............| + // | | dy < | xxxxxx..|............|............| + // y_final | | \- | xx*xxx.........|............| + // sy1 | | | | xxxxxB...|............| + // | | | | | | + // | | | | | | + // y_bottom +------------+------------+------------+------------+------------+ + // + // goal is to measure the area covered by '.' in each pixel + + // if x2 is right at the right edge of x1, y_crossing can blow up, github #1057 + // @TODO: maybe test against sy1 rather than y_bottom? + if (y_crossing > y_bottom) + y_crossing = y_bottom; + + sign = e->direction; + + // area of the rectangle covered from sy0..y_crossing + area = sign * (y_crossing-sy0); + + // area of the triangle (x_top,sy0), (x1+1,sy0), (x1+1,y_crossing) + scanline[x1] += stbtt__sized_triangle_area(area, x1+1 - x_top); + + // check if final y_crossing is blown up; no test case for this + if (y_final > y_bottom) { + y_final = y_bottom; + dy = (y_final - y_crossing ) / (x2 - (x1+1)); // if denom=0, y_final = y_crossing, so y_final <= y_bottom + } + + // in second pixel, area covered by line segment found in first pixel + // is always a rectangle 1 wide * the height of that line segment; this + // is exactly what the variable 'area' stores. it also gets a contribution + // from the line segment within it. the THIRD pixel will get the first + // pixel's rectangle contribution, the second pixel's rectangle contribution, + // and its own contribution. the 'own contribution' is the same in every pixel except + // the leftmost and rightmost, a trapezoid that slides down in each pixel. + // the second pixel's contribution to the third pixel will be the + // rectangle 1 wide times the height change in the second pixel, which is dy. + + step = sign * dy * 1; // dy is dy/dx, change in y for every 1 change in x, + // which multiplied by 1-pixel-width is how much pixel area changes for each step in x + // so the area advances by 'step' every time + + for (x = x1+1; x < x2; ++x) { + scanline[x] += area + step/2; // area of trapezoid is 1*step/2 + area += step; + } + STBTT_assert(STBTT_fabs(area) <= 1.01f); // accumulated error from area += step unless we round step down + STBTT_assert(sy1 > y_final-0.01f); + + // area covered in the last pixel is the rectangle from all the pixels to the left, + // plus the trapezoid filled by the line segment in this pixel all the way to the right edge + scanline[x2] += area + sign * stbtt__position_trapezoid_area(sy1-y_final, (float) x2, x2+1.0f, x_bottom, x2+1.0f); + + // the rest of the line is filled based on the total height of the line segment in this pixel + scanline_fill[x2] += sign * (sy1-sy0); + } + } else { + // if edge goes outside of box we're drawing, we require + // clipping logic. since this does not match the intended use + // of this library, we use a different, very slow brute + // force implementation + // note though that this does happen some of the time because + // x_top and x_bottom can be extrapolated at the top & bottom of + // the shape and actually lie outside the bounding box + int x; + for (x=0; x < len; ++x) { + // cases: + // + // there can be up to two intersections with the pixel. any intersection + // with left or right edges can be handled by splitting into two (or three) + // regions. intersections with top & bottom do not necessitate case-wise logic. + // + // the old way of doing this found the intersections with the left & right edges, + // then used some simple logic to produce up to three segments in sorted order + // from top-to-bottom. however, this had a problem: if an x edge was epsilon + // across the x border, then the corresponding y position might not be distinct + // from the other y segment, and it might ignored as an empty segment. to avoid + // that, we need to explicitly produce segments based on x positions. + + // rename variables to clearly-defined pairs + float y0 = y_top; + float x1 = (float) (x); + float x2 = (float) (x+1); + float x3 = xb; + float y3 = y_bottom; + + // x = e->x + e->dx * (y-y_top) + // (y-y_top) = (x - e->x) / e->dx + // y = (x - e->x) / e->dx + y_top + float y1 = (x - x0) / dx + y_top; + float y2 = (x+1 - x0) / dx + y_top; + + if (x0 < x1 && x3 > x2) { // three segments descending down-right + stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x1,y1); + stbtt__handle_clipped_edge(scanline,x,e, x1,y1, x2,y2); + stbtt__handle_clipped_edge(scanline,x,e, x2,y2, x3,y3); + } else if (x3 < x1 && x0 > x2) { // three segments descending down-left + stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x2,y2); + stbtt__handle_clipped_edge(scanline,x,e, x2,y2, x1,y1); + stbtt__handle_clipped_edge(scanline,x,e, x1,y1, x3,y3); + } else if (x0 < x1 && x3 > x1) { // two segments across x, down-right + stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x1,y1); + stbtt__handle_clipped_edge(scanline,x,e, x1,y1, x3,y3); + } else if (x3 < x1 && x0 > x1) { // two segments across x, down-left + stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x1,y1); + stbtt__handle_clipped_edge(scanline,x,e, x1,y1, x3,y3); + } else if (x0 < x2 && x3 > x2) { // two segments across x+1, down-right + stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x2,y2); + stbtt__handle_clipped_edge(scanline,x,e, x2,y2, x3,y3); + } else if (x3 < x2 && x0 > x2) { // two segments across x+1, down-left + stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x2,y2); + stbtt__handle_clipped_edge(scanline,x,e, x2,y2, x3,y3); + } else { // one segment + stbtt__handle_clipped_edge(scanline,x,e, x0,y0, x3,y3); + } + } + } + } + e = e->next; + } +} + +// directly AA rasterize edges w/o supersampling +static void stbtt__rasterize_sorted_edges(stbtt__bitmap *result, stbtt__edge *e, int n, int vsubsample, int off_x, int off_y, void *userdata) +{ + stbtt__hheap hh = { 0, 0, 0 }; + stbtt__active_edge *active = NULL; + int y,j=0, i; + float scanline_data[129], *scanline, *scanline2; + + STBTT__NOTUSED(vsubsample); + + if (result->w > 64) + scanline = (float *) STBTT_malloc((result->w*2+1) * sizeof(float), userdata); + else + scanline = scanline_data; + + scanline2 = scanline + result->w; + + y = off_y; + e[n].y0 = (float) (off_y + result->h) + 1; + + while (j < result->h) { + // find center of pixel for this scanline + float scan_y_top = y + 0.0f; + float scan_y_bottom = y + 1.0f; + stbtt__active_edge **step = &active; + + STBTT_memset(scanline , 0, result->w*sizeof(scanline[0])); + STBTT_memset(scanline2, 0, (result->w+1)*sizeof(scanline[0])); + + // update all active edges; + // remove all active edges that terminate before the top of this scanline + while (*step) { + stbtt__active_edge * z = *step; + if (z->ey <= scan_y_top) { + *step = z->next; // delete from list + STBTT_assert(z->direction); + z->direction = 0; + stbtt__hheap_free(&hh, z); + } else { + step = &((*step)->next); // advance through list + } + } + + // insert all edges that start before the bottom of this scanline + while (e->y0 <= scan_y_bottom) { + if (e->y0 != e->y1) { + stbtt__active_edge *z = stbtt__new_active(&hh, e, off_x, scan_y_top, userdata); + if (z != NULL) { + if (j == 0 && off_y != 0) { + if (z->ey < scan_y_top) { + // this can happen due to subpixel positioning and some kind of fp rounding error i think + z->ey = scan_y_top; + } + } + STBTT_assert(z->ey >= scan_y_top); // if we get really unlucky a tiny bit of an edge can be out of bounds + // insert at front + z->next = active; + active = z; + } + } + ++e; + } + + // now process all active edges + if (active) + stbtt__fill_active_edges_new(scanline, scanline2+1, result->w, active, scan_y_top); + + { + float sum = 0; + for (i=0; i < result->w; ++i) { + float k; + int m; + sum += scanline2[i]; + k = scanline[i] + sum; + k = (float) STBTT_fabs(k)*255 + 0.5f; + m = (int) k; + if (m > 255) m = 255; + result->pixels[j*result->stride + i] = (unsigned char) m; + } + } + // advance all the edges + step = &active; + while (*step) { + stbtt__active_edge *z = *step; + z->fx += z->fdx; // advance to position for current scanline + step = &((*step)->next); // advance through list + } + + ++y; + ++j; + } + + stbtt__hheap_cleanup(&hh, userdata); + + if (scanline != scanline_data) + STBTT_free(scanline, userdata); +} +#else +#error "Unrecognized value of STBTT_RASTERIZER_VERSION" +#endif + +#define STBTT__COMPARE(a,b) ((a)->y0 < (b)->y0) + +static void stbtt__sort_edges_ins_sort(stbtt__edge *p, int n) +{ + int i,j; + for (i=1; i < n; ++i) { + stbtt__edge t = p[i], *a = &t; + j = i; + while (j > 0) { + stbtt__edge *b = &p[j-1]; + int c = STBTT__COMPARE(a,b); + if (!c) break; + p[j] = p[j-1]; + --j; + } + if (i != j) + p[j] = t; + } +} + +static void stbtt__sort_edges_quicksort(stbtt__edge *p, int n) +{ + /* threshold for transitioning to insertion sort */ + while (n > 12) { + stbtt__edge t; + int c01,c12,c,m,i,j; + + /* compute median of three */ + m = n >> 1; + c01 = STBTT__COMPARE(&p[0],&p[m]); + c12 = STBTT__COMPARE(&p[m],&p[n-1]); + /* if 0 >= mid >= end, or 0 < mid < end, then use mid */ + if (c01 != c12) { + /* otherwise, we'll need to swap something else to middle */ + int z; + c = STBTT__COMPARE(&p[0],&p[n-1]); + /* 0>mid && mid<n: 0>n => n; 0<n => 0 */ + /* 0<mid && mid>n: 0>n => 0; 0<n => n */ + z = (c == c12) ? 0 : n-1; + t = p[z]; + p[z] = p[m]; + p[m] = t; + } + /* now p[m] is the median-of-three */ + /* swap it to the beginning so it won't move around */ + t = p[0]; + p[0] = p[m]; + p[m] = t; + + /* partition loop */ + i=1; + j=n-1; + for(;;) { + /* handling of equality is crucial here */ + /* for sentinels & efficiency with duplicates */ + for (;;++i) { + if (!STBTT__COMPARE(&p[i], &p[0])) break; + } + for (;;--j) { + if (!STBTT__COMPARE(&p[0], &p[j])) break; + } + /* make sure we haven't crossed */ + if (i >= j) break; + t = p[i]; + p[i] = p[j]; + p[j] = t; + + ++i; + --j; + } + /* recurse on smaller side, iterate on larger */ + if (j < (n-i)) { + stbtt__sort_edges_quicksort(p,j); + p = p+i; + n = n-i; + } else { + stbtt__sort_edges_quicksort(p+i, n-i); + n = j; + } + } +} + +static void stbtt__sort_edges(stbtt__edge *p, int n) +{ + stbtt__sort_edges_quicksort(p, n); + stbtt__sort_edges_ins_sort(p, n); +} + +typedef struct +{ + float x,y; +} stbtt__point; + +static void stbtt__rasterize(stbtt__bitmap *result, stbtt__point *pts, int *wcount, int windings, float scale_x, float scale_y, float shift_x, float shift_y, int off_x, int off_y, int invert, void *userdata) +{ + float y_scale_inv = invert ? -scale_y : scale_y; + stbtt__edge *e; + int n,i,j,k,m; +#if STBTT_RASTERIZER_VERSION == 1 + int vsubsample = result->h < 8 ? 15 : 5; +#elif STBTT_RASTERIZER_VERSION == 2 + int vsubsample = 1; +#else + #error "Unrecognized value of STBTT_RASTERIZER_VERSION" +#endif + // vsubsample should divide 255 evenly; otherwise we won't reach full opacity + + // now we have to blow out the windings into explicit edge lists + n = 0; + for (i=0; i < windings; ++i) + n += wcount[i]; + + e = (stbtt__edge *) STBTT_malloc(sizeof(*e) * (n+1), userdata); // add an extra one as a sentinel + if (e == 0) return; + n = 0; + + m=0; + for (i=0; i < windings; ++i) { + stbtt__point *p = pts + m; + m += wcount[i]; + j = wcount[i]-1; + for (k=0; k < wcount[i]; j=k++) { + int a=k,b=j; + // skip the edge if horizontal + if (p[j].y == p[k].y) + continue; + // add edge from j to k to the list + e[n].invert = 0; + if (invert ? p[j].y > p[k].y : p[j].y < p[k].y) { + e[n].invert = 1; + a=j,b=k; + } + e[n].x0 = p[a].x * scale_x + shift_x; + e[n].y0 = (p[a].y * y_scale_inv + shift_y) * vsubsample; + e[n].x1 = p[b].x * scale_x + shift_x; + e[n].y1 = (p[b].y * y_scale_inv + shift_y) * vsubsample; + ++n; + } + } + + // now sort the edges by their highest point (should snap to integer, and then by x) + //STBTT_sort(e, n, sizeof(e[0]), stbtt__edge_compare); + stbtt__sort_edges(e, n); + + // now, traverse the scanlines and find the intersections on each scanline, use xor winding rule + stbtt__rasterize_sorted_edges(result, e, n, vsubsample, off_x, off_y, userdata); + + STBTT_free(e, userdata); +} + +static void stbtt__add_point(stbtt__point *points, int n, float x, float y) +{ + if (!points) return; // during first pass, it's unallocated + points[n].x = x; + points[n].y = y; +} + +// tessellate until threshold p is happy... @TODO warped to compensate for non-linear stretching +static int stbtt__tesselate_curve(stbtt__point *points, int *num_points, float x0, float y0, float x1, float y1, float x2, float y2, float objspace_flatness_squared, int n) +{ + // midpoint + float mx = (x0 + 2*x1 + x2)/4; + float my = (y0 + 2*y1 + y2)/4; + // versus directly drawn line + float dx = (x0+x2)/2 - mx; + float dy = (y0+y2)/2 - my; + if (n > 16) // 65536 segments on one curve better be enough! + return 1; + if (dx*dx+dy*dy > objspace_flatness_squared) { // half-pixel error allowed... need to be smaller if AA + stbtt__tesselate_curve(points, num_points, x0,y0, (x0+x1)/2.0f,(y0+y1)/2.0f, mx,my, objspace_flatness_squared,n+1); + stbtt__tesselate_curve(points, num_points, mx,my, (x1+x2)/2.0f,(y1+y2)/2.0f, x2,y2, objspace_flatness_squared,n+1); + } else { + stbtt__add_point(points, *num_points,x2,y2); + *num_points = *num_points+1; + } + return 1; +} + +static void stbtt__tesselate_cubic(stbtt__point *points, int *num_points, float x0, float y0, float x1, float y1, float x2, float y2, float x3, float y3, float objspace_flatness_squared, int n) +{ + // @TODO this "flatness" calculation is just made-up nonsense that seems to work well enough + float dx0 = x1-x0; + float dy0 = y1-y0; + float dx1 = x2-x1; + float dy1 = y2-y1; + float dx2 = x3-x2; + float dy2 = y3-y2; + float dx = x3-x0; + float dy = y3-y0; + float longlen = (float) (STBTT_sqrt(dx0*dx0+dy0*dy0)+STBTT_sqrt(dx1*dx1+dy1*dy1)+STBTT_sqrt(dx2*dx2+dy2*dy2)); + float shortlen = (float) STBTT_sqrt(dx*dx+dy*dy); + float flatness_squared = longlen*longlen-shortlen*shortlen; + + if (n > 16) // 65536 segments on one curve better be enough! + return; + + if (flatness_squared > objspace_flatness_squared) { + float x01 = (x0+x1)/2; + float y01 = (y0+y1)/2; + float x12 = (x1+x2)/2; + float y12 = (y1+y2)/2; + float x23 = (x2+x3)/2; + float y23 = (y2+y3)/2; + + float xa = (x01+x12)/2; + float ya = (y01+y12)/2; + float xb = (x12+x23)/2; + float yb = (y12+y23)/2; + + float mx = (xa+xb)/2; + float my = (ya+yb)/2; + + stbtt__tesselate_cubic(points, num_points, x0,y0, x01,y01, xa,ya, mx,my, objspace_flatness_squared,n+1); + stbtt__tesselate_cubic(points, num_points, mx,my, xb,yb, x23,y23, x3,y3, objspace_flatness_squared,n+1); + } else { + stbtt__add_point(points, *num_points,x3,y3); + *num_points = *num_points+1; + } +} + +// returns number of contours +static stbtt__point *stbtt_FlattenCurves(stbtt_vertex *vertices, int num_verts, float objspace_flatness, int **contour_lengths, int *num_contours, void *userdata) +{ + stbtt__point *points=0; + int num_points=0; + + float objspace_flatness_squared = objspace_flatness * objspace_flatness; + int i,n=0,start=0, pass; + + // count how many "moves" there are to get the contour count + for (i=0; i < num_verts; ++i) + if (vertices[i].type == STBTT_vmove) + ++n; + + *num_contours = n; + if (n == 0) return 0; + + *contour_lengths = (int *) STBTT_malloc(sizeof(**contour_lengths) * n, userdata); + + if (*contour_lengths == 0) { + *num_contours = 0; + return 0; + } + + // make two passes through the points so we don't need to realloc + for (pass=0; pass < 2; ++pass) { + float x=0,y=0; + if (pass == 1) { + points = (stbtt__point *) STBTT_malloc(num_points * sizeof(points[0]), userdata); + if (points == NULL) goto error; + } + num_points = 0; + n= -1; + for (i=0; i < num_verts; ++i) { + switch (vertices[i].type) { + case STBTT_vmove: + // start the next contour + if (n >= 0) + (*contour_lengths)[n] = num_points - start; + ++n; + start = num_points; + + x = vertices[i].x, y = vertices[i].y; + stbtt__add_point(points, num_points++, x,y); + break; + case STBTT_vline: + x = vertices[i].x, y = vertices[i].y; + stbtt__add_point(points, num_points++, x, y); + break; + case STBTT_vcurve: + stbtt__tesselate_curve(points, &num_points, x,y, + vertices[i].cx, vertices[i].cy, + vertices[i].x, vertices[i].y, + objspace_flatness_squared, 0); + x = vertices[i].x, y = vertices[i].y; + break; + case STBTT_vcubic: + stbtt__tesselate_cubic(points, &num_points, x,y, + vertices[i].cx, vertices[i].cy, + vertices[i].cx1, vertices[i].cy1, + vertices[i].x, vertices[i].y, + objspace_flatness_squared, 0); + x = vertices[i].x, y = vertices[i].y; + break; + } + } + (*contour_lengths)[n] = num_points - start; + } + + return points; +error: + STBTT_free(points, userdata); + STBTT_free(*contour_lengths, userdata); + *contour_lengths = 0; + *num_contours = 0; + return NULL; +} + +STBTT_DEF void stbtt_Rasterize(stbtt__bitmap *result, float flatness_in_pixels, stbtt_vertex *vertices, int num_verts, float scale_x, float scale_y, float shift_x, float shift_y, int x_off, int y_off, int invert, void *userdata) +{ + float scale = scale_x > scale_y ? scale_y : scale_x; + int winding_count = 0; + int *winding_lengths = NULL; + stbtt__point *windings = stbtt_FlattenCurves(vertices, num_verts, flatness_in_pixels / scale, &winding_lengths, &winding_count, userdata); + if (windings) { + stbtt__rasterize(result, windings, winding_lengths, winding_count, scale_x, scale_y, shift_x, shift_y, x_off, y_off, invert, userdata); + STBTT_free(winding_lengths, userdata); + STBTT_free(windings, userdata); + } +} + +STBTT_DEF void stbtt_FreeBitmap(unsigned char *bitmap, void *userdata) +{ + STBTT_free(bitmap, userdata); +} + +STBTT_DEF unsigned char *stbtt_GetGlyphBitmapSubpixel(const stbtt_fontinfo *info, float scale_x, float scale_y, float shift_x, float shift_y, int glyph, int *width, int *height, int *xoff, int *yoff) +{ + int ix0,iy0,ix1,iy1; + stbtt__bitmap gbm; + stbtt_vertex *vertices; + int num_verts = stbtt_GetGlyphShape(info, glyph, &vertices); + + if (scale_x == 0) scale_x = scale_y; + if (scale_y == 0) { + if (scale_x == 0) { + STBTT_free(vertices, info->userdata); + return NULL; + } + scale_y = scale_x; + } + + stbtt_GetGlyphBitmapBoxSubpixel(info, glyph, scale_x, scale_y, shift_x, shift_y, &ix0,&iy0,&ix1,&iy1); + + // now we get the size + gbm.w = (ix1 - ix0); + gbm.h = (iy1 - iy0); + gbm.pixels = NULL; // in case we error + + if (width ) *width = gbm.w; + if (height) *height = gbm.h; + if (xoff ) *xoff = ix0; + if (yoff ) *yoff = iy0; + + if (gbm.w && gbm.h) { + gbm.pixels = (unsigned char *) STBTT_malloc(gbm.w * gbm.h, info->userdata); + if (gbm.pixels) { + gbm.stride = gbm.w; + + stbtt_Rasterize(&gbm, 0.35f, vertices, num_verts, scale_x, scale_y, shift_x, shift_y, ix0, iy0, 1, info->userdata); + } + } + STBTT_free(vertices, info->userdata); + return gbm.pixels; +} + +STBTT_DEF unsigned char *stbtt_GetGlyphBitmap(const stbtt_fontinfo *info, float scale_x, float scale_y, int glyph, int *width, int *height, int *xoff, int *yoff) +{ + return stbtt_GetGlyphBitmapSubpixel(info, scale_x, scale_y, 0.0f, 0.0f, glyph, width, height, xoff, yoff); +} + +STBTT_DEF void stbtt_MakeGlyphBitmapSubpixel(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int glyph) +{ + int ix0,iy0; + stbtt_vertex *vertices; + int num_verts = stbtt_GetGlyphShape(info, glyph, &vertices); + stbtt__bitmap gbm; + + stbtt_GetGlyphBitmapBoxSubpixel(info, glyph, scale_x, scale_y, shift_x, shift_y, &ix0,&iy0,0,0); + gbm.pixels = output; + gbm.w = out_w; + gbm.h = out_h; + gbm.stride = out_stride; + + if (gbm.w && gbm.h) + stbtt_Rasterize(&gbm, 0.35f, vertices, num_verts, scale_x, scale_y, shift_x, shift_y, ix0,iy0, 1, info->userdata); + + STBTT_free(vertices, info->userdata); +} + +STBTT_DEF void stbtt_MakeGlyphBitmap(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, int glyph) +{ + stbtt_MakeGlyphBitmapSubpixel(info, output, out_w, out_h, out_stride, scale_x, scale_y, 0.0f,0.0f, glyph); +} + +STBTT_DEF unsigned char *stbtt_GetCodepointBitmapSubpixel(const stbtt_fontinfo *info, float scale_x, float scale_y, float shift_x, float shift_y, int codepoint, int *width, int *height, int *xoff, int *yoff) +{ + return stbtt_GetGlyphBitmapSubpixel(info, scale_x, scale_y,shift_x,shift_y, stbtt_FindGlyphIndex(info,codepoint), width,height,xoff,yoff); +} + +STBTT_DEF void stbtt_MakeCodepointBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int oversample_x, int oversample_y, float *sub_x, float *sub_y, int codepoint) +{ + stbtt_MakeGlyphBitmapSubpixelPrefilter(info, output, out_w, out_h, out_stride, scale_x, scale_y, shift_x, shift_y, oversample_x, oversample_y, sub_x, sub_y, stbtt_FindGlyphIndex(info,codepoint)); +} + +STBTT_DEF void stbtt_MakeCodepointBitmapSubpixel(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int codepoint) +{ + stbtt_MakeGlyphBitmapSubpixel(info, output, out_w, out_h, out_stride, scale_x, scale_y, shift_x, shift_y, stbtt_FindGlyphIndex(info,codepoint)); +} + +STBTT_DEF unsigned char *stbtt_GetCodepointBitmap(const stbtt_fontinfo *info, float scale_x, float scale_y, int codepoint, int *width, int *height, int *xoff, int *yoff) +{ + return stbtt_GetCodepointBitmapSubpixel(info, scale_x, scale_y, 0.0f,0.0f, codepoint, width,height,xoff,yoff); +} + +STBTT_DEF void stbtt_MakeCodepointBitmap(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, int codepoint) +{ + stbtt_MakeCodepointBitmapSubpixel(info, output, out_w, out_h, out_stride, scale_x, scale_y, 0.0f,0.0f, codepoint); +} + +////////////////////////////////////////////////////////////////////////////// +// +// bitmap baking +// +// This is SUPER-CRAPPY packing to keep source code small + +static int stbtt_BakeFontBitmap_internal(unsigned char *data, int offset, // font location (use offset=0 for plain .ttf) + float pixel_height, // height of font in pixels + unsigned char *pixels, int pw, int ph, // bitmap to be filled in + int first_char, int num_chars, // characters to bake + stbtt_bakedchar *chardata) +{ + float scale; + int x,y,bottom_y, i; + stbtt_fontinfo f; + f.userdata = NULL; + if (!stbtt_InitFont(&f, data, offset)) + return -1; + STBTT_memset(pixels, 0, pw*ph); // background of 0 around pixels + x=y=1; + bottom_y = 1; + + scale = stbtt_ScaleForPixelHeight(&f, pixel_height); + + for (i=0; i < num_chars; ++i) { + int advance, lsb, x0,y0,x1,y1,gw,gh; + int g = stbtt_FindGlyphIndex(&f, first_char + i); + stbtt_GetGlyphHMetrics(&f, g, &advance, &lsb); + stbtt_GetGlyphBitmapBox(&f, g, scale,scale, &x0,&y0,&x1,&y1); + gw = x1-x0; + gh = y1-y0; + if (x + gw + 1 >= pw) + y = bottom_y, x = 1; // advance to next row + if (y + gh + 1 >= ph) // check if it fits vertically AFTER potentially moving to next row + return -i; + STBTT_assert(x+gw < pw); + STBTT_assert(y+gh < ph); + stbtt_MakeGlyphBitmap(&f, pixels+x+y*pw, gw,gh,pw, scale,scale, g); + chardata[i].x0 = (stbtt_int16) x; + chardata[i].y0 = (stbtt_int16) y; + chardata[i].x1 = (stbtt_int16) (x + gw); + chardata[i].y1 = (stbtt_int16) (y + gh); + chardata[i].xadvance = scale * advance; + chardata[i].xoff = (float) x0; + chardata[i].yoff = (float) y0; + x = x + gw + 1; + if (y+gh+1 > bottom_y) + bottom_y = y+gh+1; + } + return bottom_y; +} + +STBTT_DEF void stbtt_GetBakedQuad(const stbtt_bakedchar *chardata, int pw, int ph, int char_index, float *xpos, float *ypos, stbtt_aligned_quad *q, int opengl_fillrule) +{ + float d3d_bias = opengl_fillrule ? 0 : -0.5f; + float ipw = 1.0f / pw, iph = 1.0f / ph; + const stbtt_bakedchar *b = chardata + char_index; + int round_x = STBTT_ifloor((*xpos + b->xoff) + 0.5f); + int round_y = STBTT_ifloor((*ypos + b->yoff) + 0.5f); + + q->x0 = round_x + d3d_bias; + q->y0 = round_y + d3d_bias; + q->x1 = round_x + b->x1 - b->x0 + d3d_bias; + q->y1 = round_y + b->y1 - b->y0 + d3d_bias; + + q->s0 = b->x0 * ipw; + q->t0 = b->y0 * iph; + q->s1 = b->x1 * ipw; + q->t1 = b->y1 * iph; + + *xpos += b->xadvance; +} + +////////////////////////////////////////////////////////////////////////////// +// +// rectangle packing replacement routines if you don't have stb_rect_pack.h +// + +#ifndef STB_RECT_PACK_VERSION + +typedef int stbrp_coord; + +//////////////////////////////////////////////////////////////////////////////////// +// // +// // +// COMPILER WARNING ?!?!? // +// // +// // +// if you get a compile warning due to these symbols being defined more than // +// once, move #include "stb_rect_pack.h" before #include "stb_truetype.h" // +// // +//////////////////////////////////////////////////////////////////////////////////// + +typedef struct +{ + int width,height; + int x,y,bottom_y; +} stbrp_context; + +typedef struct +{ + unsigned char x; +} stbrp_node; + +struct stbrp_rect +{ + stbrp_coord x,y; + int id,w,h,was_packed; +}; + +static void stbrp_init_target(stbrp_context *con, int pw, int ph, stbrp_node *nodes, int num_nodes) +{ + con->width = pw; + con->height = ph; + con->x = 0; + con->y = 0; + con->bottom_y = 0; + STBTT__NOTUSED(nodes); + STBTT__NOTUSED(num_nodes); +} + +static void stbrp_pack_rects(stbrp_context *con, stbrp_rect *rects, int num_rects) +{ + int i; + for (i=0; i < num_rects; ++i) { + if (con->x + rects[i].w > con->width) { + con->x = 0; + con->y = con->bottom_y; + } + if (con->y + rects[i].h > con->height) + break; + rects[i].x = con->x; + rects[i].y = con->y; + rects[i].was_packed = 1; + con->x += rects[i].w; + if (con->y + rects[i].h > con->bottom_y) + con->bottom_y = con->y + rects[i].h; + } + for ( ; i < num_rects; ++i) + rects[i].was_packed = 0; +} +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// bitmap baking +// +// This is SUPER-AWESOME (tm Ryan Gordon) packing using stb_rect_pack.h. If +// stb_rect_pack.h isn't available, it uses the BakeFontBitmap strategy. + +STBTT_DEF int stbtt_PackBegin(stbtt_pack_context *spc, unsigned char *pixels, int pw, int ph, int stride_in_bytes, int padding, void *alloc_context) +{ + stbrp_context *context = (stbrp_context *) STBTT_malloc(sizeof(*context) ,alloc_context); + int num_nodes = pw - padding; + stbrp_node *nodes = (stbrp_node *) STBTT_malloc(sizeof(*nodes ) * num_nodes,alloc_context); + + if (context == NULL || nodes == NULL) { + if (context != NULL) STBTT_free(context, alloc_context); + if (nodes != NULL) STBTT_free(nodes , alloc_context); + return 0; + } + + spc->user_allocator_context = alloc_context; + spc->width = pw; + spc->height = ph; + spc->pixels = pixels; + spc->pack_info = context; + spc->nodes = nodes; + spc->padding = padding; + spc->stride_in_bytes = stride_in_bytes != 0 ? stride_in_bytes : pw; + spc->h_oversample = 1; + spc->v_oversample = 1; + spc->skip_missing = 0; + + stbrp_init_target(context, pw-padding, ph-padding, nodes, num_nodes); + + if (pixels) + STBTT_memset(pixels, 0, pw*ph); // background of 0 around pixels + + return 1; +} + +STBTT_DEF void stbtt_PackEnd (stbtt_pack_context *spc) +{ + STBTT_free(spc->nodes , spc->user_allocator_context); + STBTT_free(spc->pack_info, spc->user_allocator_context); +} + +STBTT_DEF void stbtt_PackSetOversampling(stbtt_pack_context *spc, unsigned int h_oversample, unsigned int v_oversample) +{ + STBTT_assert(h_oversample <= STBTT_MAX_OVERSAMPLE); + STBTT_assert(v_oversample <= STBTT_MAX_OVERSAMPLE); + if (h_oversample <= STBTT_MAX_OVERSAMPLE) + spc->h_oversample = h_oversample; + if (v_oversample <= STBTT_MAX_OVERSAMPLE) + spc->v_oversample = v_oversample; +} + +STBTT_DEF void stbtt_PackSetSkipMissingCodepoints(stbtt_pack_context *spc, int skip) +{ + spc->skip_missing = skip; +} + +#define STBTT__OVER_MASK (STBTT_MAX_OVERSAMPLE-1) + +static void stbtt__h_prefilter(unsigned char *pixels, int w, int h, int stride_in_bytes, unsigned int kernel_width) +{ + unsigned char buffer[STBTT_MAX_OVERSAMPLE]; + int safe_w = w - kernel_width; + int j; + STBTT_memset(buffer, 0, STBTT_MAX_OVERSAMPLE); // suppress bogus warning from VS2013 -analyze + for (j=0; j < h; ++j) { + int i; + unsigned int total; + STBTT_memset(buffer, 0, kernel_width); + + total = 0; + + // make kernel_width a constant in common cases so compiler can optimize out the divide + switch (kernel_width) { + case 2: + for (i=0; i <= safe_w; ++i) { + total += pixels[i] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i]; + pixels[i] = (unsigned char) (total / 2); + } + break; + case 3: + for (i=0; i <= safe_w; ++i) { + total += pixels[i] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i]; + pixels[i] = (unsigned char) (total / 3); + } + break; + case 4: + for (i=0; i <= safe_w; ++i) { + total += pixels[i] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i]; + pixels[i] = (unsigned char) (total / 4); + } + break; + case 5: + for (i=0; i <= safe_w; ++i) { + total += pixels[i] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i]; + pixels[i] = (unsigned char) (total / 5); + } + break; + default: + for (i=0; i <= safe_w; ++i) { + total += pixels[i] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i]; + pixels[i] = (unsigned char) (total / kernel_width); + } + break; + } + + for (; i < w; ++i) { + STBTT_assert(pixels[i] == 0); + total -= buffer[i & STBTT__OVER_MASK]; + pixels[i] = (unsigned char) (total / kernel_width); + } + + pixels += stride_in_bytes; + } +} + +static void stbtt__v_prefilter(unsigned char *pixels, int w, int h, int stride_in_bytes, unsigned int kernel_width) +{ + unsigned char buffer[STBTT_MAX_OVERSAMPLE]; + int safe_h = h - kernel_width; + int j; + STBTT_memset(buffer, 0, STBTT_MAX_OVERSAMPLE); // suppress bogus warning from VS2013 -analyze + for (j=0; j < w; ++j) { + int i; + unsigned int total; + STBTT_memset(buffer, 0, kernel_width); + + total = 0; + + // make kernel_width a constant in common cases so compiler can optimize out the divide + switch (kernel_width) { + case 2: + for (i=0; i <= safe_h; ++i) { + total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes]; + pixels[i*stride_in_bytes] = (unsigned char) (total / 2); + } + break; + case 3: + for (i=0; i <= safe_h; ++i) { + total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes]; + pixels[i*stride_in_bytes] = (unsigned char) (total / 3); + } + break; + case 4: + for (i=0; i <= safe_h; ++i) { + total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes]; + pixels[i*stride_in_bytes] = (unsigned char) (total / 4); + } + break; + case 5: + for (i=0; i <= safe_h; ++i) { + total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes]; + pixels[i*stride_in_bytes] = (unsigned char) (total / 5); + } + break; + default: + for (i=0; i <= safe_h; ++i) { + total += pixels[i*stride_in_bytes] - buffer[i & STBTT__OVER_MASK]; + buffer[(i+kernel_width) & STBTT__OVER_MASK] = pixels[i*stride_in_bytes]; + pixels[i*stride_in_bytes] = (unsigned char) (total / kernel_width); + } + break; + } + + for (; i < h; ++i) { + STBTT_assert(pixels[i*stride_in_bytes] == 0); + total -= buffer[i & STBTT__OVER_MASK]; + pixels[i*stride_in_bytes] = (unsigned char) (total / kernel_width); + } + + pixels += 1; + } +} + +static float stbtt__oversample_shift(int oversample) +{ + if (!oversample) + return 0.0f; + + // The prefilter is a box filter of width "oversample", + // which shifts phase by (oversample - 1)/2 pixels in + // oversampled space. We want to shift in the opposite + // direction to counter this. + return (float)-(oversample - 1) / (2.0f * (float)oversample); +} + +// rects array must be big enough to accommodate all characters in the given ranges +STBTT_DEF int stbtt_PackFontRangesGatherRects(stbtt_pack_context *spc, const stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects) +{ + int i,j,k; + int missing_glyph_added = 0; + + k=0; + for (i=0; i < num_ranges; ++i) { + float fh = ranges[i].font_size; + float scale = fh > 0 ? stbtt_ScaleForPixelHeight(info, fh) : stbtt_ScaleForMappingEmToPixels(info, -fh); + ranges[i].h_oversample = (unsigned char) spc->h_oversample; + ranges[i].v_oversample = (unsigned char) spc->v_oversample; + for (j=0; j < ranges[i].num_chars; ++j) { + int x0,y0,x1,y1; + int codepoint = ranges[i].array_of_unicode_codepoints == NULL ? ranges[i].first_unicode_codepoint_in_range + j : ranges[i].array_of_unicode_codepoints[j]; + int glyph = stbtt_FindGlyphIndex(info, codepoint); + if (glyph == 0 && (spc->skip_missing || missing_glyph_added)) { + rects[k].w = rects[k].h = 0; + } else { + stbtt_GetGlyphBitmapBoxSubpixel(info,glyph, + scale * spc->h_oversample, + scale * spc->v_oversample, + 0,0, + &x0,&y0,&x1,&y1); + rects[k].w = (stbrp_coord) (x1-x0 + spc->padding + spc->h_oversample-1); + rects[k].h = (stbrp_coord) (y1-y0 + spc->padding + spc->v_oversample-1); + if (glyph == 0) + missing_glyph_added = 1; + } + ++k; + } + } + + return k; +} + +STBTT_DEF void stbtt_MakeGlyphBitmapSubpixelPrefilter(const stbtt_fontinfo *info, unsigned char *output, int out_w, int out_h, int out_stride, float scale_x, float scale_y, float shift_x, float shift_y, int prefilter_x, int prefilter_y, float *sub_x, float *sub_y, int glyph) +{ + stbtt_MakeGlyphBitmapSubpixel(info, + output, + out_w - (prefilter_x - 1), + out_h - (prefilter_y - 1), + out_stride, + scale_x, + scale_y, + shift_x, + shift_y, + glyph); + + if (prefilter_x > 1) + stbtt__h_prefilter(output, out_w, out_h, out_stride, prefilter_x); + + if (prefilter_y > 1) + stbtt__v_prefilter(output, out_w, out_h, out_stride, prefilter_y); + + *sub_x = stbtt__oversample_shift(prefilter_x); + *sub_y = stbtt__oversample_shift(prefilter_y); +} + +// rects array must be big enough to accommodate all characters in the given ranges +STBTT_DEF int stbtt_PackFontRangesRenderIntoRects(stbtt_pack_context *spc, const stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects) +{ + int i,j,k, missing_glyph = -1, return_value = 1; + + // save current values + int old_h_over = spc->h_oversample; + int old_v_over = spc->v_oversample; + + k = 0; + for (i=0; i < num_ranges; ++i) { + float fh = ranges[i].font_size; + float scale = fh > 0 ? stbtt_ScaleForPixelHeight(info, fh) : stbtt_ScaleForMappingEmToPixels(info, -fh); + float recip_h,recip_v,sub_x,sub_y; + spc->h_oversample = ranges[i].h_oversample; + spc->v_oversample = ranges[i].v_oversample; + recip_h = 1.0f / spc->h_oversample; + recip_v = 1.0f / spc->v_oversample; + sub_x = stbtt__oversample_shift(spc->h_oversample); + sub_y = stbtt__oversample_shift(spc->v_oversample); + for (j=0; j < ranges[i].num_chars; ++j) { + stbrp_rect *r = &rects[k]; + if (r->was_packed && r->w != 0 && r->h != 0) { + stbtt_packedchar *bc = &ranges[i].chardata_for_range[j]; + int advance, lsb, x0,y0,x1,y1; + int codepoint = ranges[i].array_of_unicode_codepoints == NULL ? ranges[i].first_unicode_codepoint_in_range + j : ranges[i].array_of_unicode_codepoints[j]; + int glyph = stbtt_FindGlyphIndex(info, codepoint); + stbrp_coord pad = (stbrp_coord) spc->padding; + + // pad on left and top + r->x += pad; + r->y += pad; + r->w -= pad; + r->h -= pad; + stbtt_GetGlyphHMetrics(info, glyph, &advance, &lsb); + stbtt_GetGlyphBitmapBox(info, glyph, + scale * spc->h_oversample, + scale * spc->v_oversample, + &x0,&y0,&x1,&y1); + stbtt_MakeGlyphBitmapSubpixel(info, + spc->pixels + r->x + r->y*spc->stride_in_bytes, + r->w - spc->h_oversample+1, + r->h - spc->v_oversample+1, + spc->stride_in_bytes, + scale * spc->h_oversample, + scale * spc->v_oversample, + 0,0, + glyph); + + if (spc->h_oversample > 1) + stbtt__h_prefilter(spc->pixels + r->x + r->y*spc->stride_in_bytes, + r->w, r->h, spc->stride_in_bytes, + spc->h_oversample); + + if (spc->v_oversample > 1) + stbtt__v_prefilter(spc->pixels + r->x + r->y*spc->stride_in_bytes, + r->w, r->h, spc->stride_in_bytes, + spc->v_oversample); + + bc->x0 = (stbtt_int16) r->x; + bc->y0 = (stbtt_int16) r->y; + bc->x1 = (stbtt_int16) (r->x + r->w); + bc->y1 = (stbtt_int16) (r->y + r->h); + bc->xadvance = scale * advance; + bc->xoff = (float) x0 * recip_h + sub_x; + bc->yoff = (float) y0 * recip_v + sub_y; + bc->xoff2 = (x0 + r->w) * recip_h + sub_x; + bc->yoff2 = (y0 + r->h) * recip_v + sub_y; + + if (glyph == 0) + missing_glyph = j; + } else if (spc->skip_missing) { + return_value = 0; + } else if (r->was_packed && r->w == 0 && r->h == 0 && missing_glyph >= 0) { + ranges[i].chardata_for_range[j] = ranges[i].chardata_for_range[missing_glyph]; + } else { + return_value = 0; // if any fail, report failure + } + + ++k; + } + } + + // restore original values + spc->h_oversample = old_h_over; + spc->v_oversample = old_v_over; + + return return_value; +} + +STBTT_DEF void stbtt_PackFontRangesPackRects(stbtt_pack_context *spc, stbrp_rect *rects, int num_rects) +{ + stbrp_pack_rects((stbrp_context *) spc->pack_info, rects, num_rects); +} + +STBTT_DEF int stbtt_PackFontRanges(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, stbtt_pack_range *ranges, int num_ranges) +{ + stbtt_fontinfo info; + int i,j,n, return_value = 1; + //stbrp_context *context = (stbrp_context *) spc->pack_info; + stbrp_rect *rects; + + // flag all characters as NOT packed + for (i=0; i < num_ranges; ++i) + for (j=0; j < ranges[i].num_chars; ++j) + ranges[i].chardata_for_range[j].x0 = + ranges[i].chardata_for_range[j].y0 = + ranges[i].chardata_for_range[j].x1 = + ranges[i].chardata_for_range[j].y1 = 0; + + n = 0; + for (i=0; i < num_ranges; ++i) + n += ranges[i].num_chars; + + rects = (stbrp_rect *) STBTT_malloc(sizeof(*rects) * n, spc->user_allocator_context); + if (rects == NULL) + return 0; + + info.userdata = spc->user_allocator_context; + stbtt_InitFont(&info, fontdata, stbtt_GetFontOffsetForIndex(fontdata,font_index)); + + n = stbtt_PackFontRangesGatherRects(spc, &info, ranges, num_ranges, rects); + + stbtt_PackFontRangesPackRects(spc, rects, n); + + return_value = stbtt_PackFontRangesRenderIntoRects(spc, &info, ranges, num_ranges, rects); + + STBTT_free(rects, spc->user_allocator_context); + return return_value; +} + +STBTT_DEF int stbtt_PackFontRange(stbtt_pack_context *spc, const unsigned char *fontdata, int font_index, float font_size, + int first_unicode_codepoint_in_range, int num_chars_in_range, stbtt_packedchar *chardata_for_range) +{ + stbtt_pack_range range; + range.first_unicode_codepoint_in_range = first_unicode_codepoint_in_range; + range.array_of_unicode_codepoints = NULL; + range.num_chars = num_chars_in_range; + range.chardata_for_range = chardata_for_range; + range.font_size = font_size; + return stbtt_PackFontRanges(spc, fontdata, font_index, &range, 1); +} + +STBTT_DEF void stbtt_GetScaledFontVMetrics(const unsigned char *fontdata, int index, float size, float *ascent, float *descent, float *lineGap) +{ + int i_ascent, i_descent, i_lineGap; + float scale; + stbtt_fontinfo info; + stbtt_InitFont(&info, fontdata, stbtt_GetFontOffsetForIndex(fontdata, index)); + scale = size > 0 ? stbtt_ScaleForPixelHeight(&info, size) : stbtt_ScaleForMappingEmToPixels(&info, -size); + stbtt_GetFontVMetrics(&info, &i_ascent, &i_descent, &i_lineGap); + *ascent = (float) i_ascent * scale; + *descent = (float) i_descent * scale; + *lineGap = (float) i_lineGap * scale; +} + +STBTT_DEF void stbtt_GetPackedQuad(const stbtt_packedchar *chardata, int pw, int ph, int char_index, float *xpos, float *ypos, stbtt_aligned_quad *q, int align_to_integer) +{ + float ipw = 1.0f / pw, iph = 1.0f / ph; + const stbtt_packedchar *b = chardata + char_index; + + if (align_to_integer) { + float x = (float) STBTT_ifloor((*xpos + b->xoff) + 0.5f); + float y = (float) STBTT_ifloor((*ypos + b->yoff) + 0.5f); + q->x0 = x; + q->y0 = y; + q->x1 = x + b->xoff2 - b->xoff; + q->y1 = y + b->yoff2 - b->yoff; + } else { + q->x0 = *xpos + b->xoff; + q->y0 = *ypos + b->yoff; + q->x1 = *xpos + b->xoff2; + q->y1 = *ypos + b->yoff2; + } + + q->s0 = b->x0 * ipw; + q->t0 = b->y0 * iph; + q->s1 = b->x1 * ipw; + q->t1 = b->y1 * iph; + + *xpos += b->xadvance; +} + +////////////////////////////////////////////////////////////////////////////// +// +// sdf computation +// + +#define STBTT_min(a,b) ((a) < (b) ? (a) : (b)) +#define STBTT_max(a,b) ((a) < (b) ? (b) : (a)) + +static int stbtt__ray_intersect_bezier(float orig[2], float ray[2], float q0[2], float q1[2], float q2[2], float hits[2][2]) +{ + float q0perp = q0[1]*ray[0] - q0[0]*ray[1]; + float q1perp = q1[1]*ray[0] - q1[0]*ray[1]; + float q2perp = q2[1]*ray[0] - q2[0]*ray[1]; + float roperp = orig[1]*ray[0] - orig[0]*ray[1]; + + float a = q0perp - 2*q1perp + q2perp; + float b = q1perp - q0perp; + float c = q0perp - roperp; + + float s0 = 0., s1 = 0.; + int num_s = 0; + + if (a != 0.0) { + float discr = b*b - a*c; + if (discr > 0.0) { + float rcpna = -1 / a; + float d = (float) STBTT_sqrt(discr); + s0 = (b+d) * rcpna; + s1 = (b-d) * rcpna; + if (s0 >= 0.0 && s0 <= 1.0) + num_s = 1; + if (d > 0.0 && s1 >= 0.0 && s1 <= 1.0) { + if (num_s == 0) s0 = s1; + ++num_s; + } + } + } else { + // 2*b*s + c = 0 + // s = -c / (2*b) + s0 = c / (-2 * b); + if (s0 >= 0.0 && s0 <= 1.0) + num_s = 1; + } + + if (num_s == 0) + return 0; + else { + float rcp_len2 = 1 / (ray[0]*ray[0] + ray[1]*ray[1]); + float rayn_x = ray[0] * rcp_len2, rayn_y = ray[1] * rcp_len2; + + float q0d = q0[0]*rayn_x + q0[1]*rayn_y; + float q1d = q1[0]*rayn_x + q1[1]*rayn_y; + float q2d = q2[0]*rayn_x + q2[1]*rayn_y; + float rod = orig[0]*rayn_x + orig[1]*rayn_y; + + float q10d = q1d - q0d; + float q20d = q2d - q0d; + float q0rd = q0d - rod; + + hits[0][0] = q0rd + s0*(2.0f - 2.0f*s0)*q10d + s0*s0*q20d; + hits[0][1] = a*s0+b; + + if (num_s > 1) { + hits[1][0] = q0rd + s1*(2.0f - 2.0f*s1)*q10d + s1*s1*q20d; + hits[1][1] = a*s1+b; + return 2; + } else { + return 1; + } + } +} + +static int equal(float *a, float *b) +{ + return (a[0] == b[0] && a[1] == b[1]); +} + +static int stbtt__compute_crossings_x(float x, float y, int nverts, stbtt_vertex *verts) +{ + int i; + float orig[2], ray[2] = { 1, 0 }; + float y_frac; + int winding = 0; + + // make sure y never passes through a vertex of the shape + y_frac = (float) STBTT_fmod(y, 1.0f); + if (y_frac < 0.01f) + y += 0.01f; + else if (y_frac > 0.99f) + y -= 0.01f; + + orig[0] = x; + orig[1] = y; + + // test a ray from (-infinity,y) to (x,y) + for (i=0; i < nverts; ++i) { + if (verts[i].type == STBTT_vline) { + int x0 = (int) verts[i-1].x, y0 = (int) verts[i-1].y; + int x1 = (int) verts[i ].x, y1 = (int) verts[i ].y; + if (y > STBTT_min(y0,y1) && y < STBTT_max(y0,y1) && x > STBTT_min(x0,x1)) { + float x_inter = (y - y0) / (y1 - y0) * (x1-x0) + x0; + if (x_inter < x) + winding += (y0 < y1) ? 1 : -1; + } + } + if (verts[i].type == STBTT_vcurve) { + int x0 = (int) verts[i-1].x , y0 = (int) verts[i-1].y ; + int x1 = (int) verts[i ].cx, y1 = (int) verts[i ].cy; + int x2 = (int) verts[i ].x , y2 = (int) verts[i ].y ; + int ax = STBTT_min(x0,STBTT_min(x1,x2)), ay = STBTT_min(y0,STBTT_min(y1,y2)); + int by = STBTT_max(y0,STBTT_max(y1,y2)); + if (y > ay && y < by && x > ax) { + float q0[2],q1[2],q2[2]; + float hits[2][2]; + q0[0] = (float)x0; + q0[1] = (float)y0; + q1[0] = (float)x1; + q1[1] = (float)y1; + q2[0] = (float)x2; + q2[1] = (float)y2; + if (equal(q0,q1) || equal(q1,q2)) { + x0 = (int)verts[i-1].x; + y0 = (int)verts[i-1].y; + x1 = (int)verts[i ].x; + y1 = (int)verts[i ].y; + if (y > STBTT_min(y0,y1) && y < STBTT_max(y0,y1) && x > STBTT_min(x0,x1)) { + float x_inter = (y - y0) / (y1 - y0) * (x1-x0) + x0; + if (x_inter < x) + winding += (y0 < y1) ? 1 : -1; + } + } else { + int num_hits = stbtt__ray_intersect_bezier(orig, ray, q0, q1, q2, hits); + if (num_hits >= 1) + if (hits[0][0] < 0) + winding += (hits[0][1] < 0 ? -1 : 1); + if (num_hits >= 2) + if (hits[1][0] < 0) + winding += (hits[1][1] < 0 ? -1 : 1); + } + } + } + } + return winding; +} + +static float stbtt__cuberoot( float x ) +{ + if (x<0) + return -(float) STBTT_pow(-x,1.0f/3.0f); + else + return (float) STBTT_pow( x,1.0f/3.0f); +} + +// x^3 + a*x^2 + b*x + c = 0 +static int stbtt__solve_cubic(float a, float b, float c, float* r) +{ + float s = -a / 3; + float p = b - a*a / 3; + float q = a * (2*a*a - 9*b) / 27 + c; + float p3 = p*p*p; + float d = q*q + 4*p3 / 27; + if (d >= 0) { + float z = (float) STBTT_sqrt(d); + float u = (-q + z) / 2; + float v = (-q - z) / 2; + u = stbtt__cuberoot(u); + v = stbtt__cuberoot(v); + r[0] = s + u + v; + return 1; + } else { + float u = (float) STBTT_sqrt(-p/3); + float v = (float) STBTT_acos(-STBTT_sqrt(-27/p3) * q / 2) / 3; // p3 must be negative, since d is negative + float m = (float) STBTT_cos(v); + float n = (float) STBTT_cos(v-3.141592/2)*1.732050808f; + r[0] = s + u * 2 * m; + r[1] = s - u * (m + n); + r[2] = s - u * (m - n); + + //STBTT_assert( STBTT_fabs(((r[0]+a)*r[0]+b)*r[0]+c) < 0.05f); // these asserts may not be safe at all scales, though they're in bezier t parameter units so maybe? + //STBTT_assert( STBTT_fabs(((r[1]+a)*r[1]+b)*r[1]+c) < 0.05f); + //STBTT_assert( STBTT_fabs(((r[2]+a)*r[2]+b)*r[2]+c) < 0.05f); + return 3; + } +} + +STBTT_DEF unsigned char * stbtt_GetGlyphSDF(const stbtt_fontinfo *info, float scale, int glyph, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff) +{ + float scale_x = scale, scale_y = scale; + int ix0,iy0,ix1,iy1; + int w,h; + unsigned char *data; + + if (scale == 0) return NULL; + + stbtt_GetGlyphBitmapBoxSubpixel(info, glyph, scale, scale, 0.0f,0.0f, &ix0,&iy0,&ix1,&iy1); + + // if empty, return NULL + if (ix0 == ix1 || iy0 == iy1) + return NULL; + + ix0 -= padding; + iy0 -= padding; + ix1 += padding; + iy1 += padding; + + w = (ix1 - ix0); + h = (iy1 - iy0); + + if (width ) *width = w; + if (height) *height = h; + if (xoff ) *xoff = ix0; + if (yoff ) *yoff = iy0; + + // invert for y-downwards bitmaps + scale_y = -scale_y; + + { + int x,y,i,j; + float *precompute; + stbtt_vertex *verts; + int num_verts = stbtt_GetGlyphShape(info, glyph, &verts); + data = (unsigned char *) STBTT_malloc(w * h, info->userdata); + precompute = (float *) STBTT_malloc(num_verts * sizeof(float), info->userdata); + + for (i=0,j=num_verts-1; i < num_verts; j=i++) { + if (verts[i].type == STBTT_vline) { + float x0 = verts[i].x*scale_x, y0 = verts[i].y*scale_y; + float x1 = verts[j].x*scale_x, y1 = verts[j].y*scale_y; + float dist = (float) STBTT_sqrt((x1-x0)*(x1-x0) + (y1-y0)*(y1-y0)); + precompute[i] = (dist == 0) ? 0.0f : 1.0f / dist; + } else if (verts[i].type == STBTT_vcurve) { + float x2 = verts[j].x *scale_x, y2 = verts[j].y *scale_y; + float x1 = verts[i].cx*scale_x, y1 = verts[i].cy*scale_y; + float x0 = verts[i].x *scale_x, y0 = verts[i].y *scale_y; + float bx = x0 - 2*x1 + x2, by = y0 - 2*y1 + y2; + float len2 = bx*bx + by*by; + if (len2 != 0.0f) + precompute[i] = 1.0f / (bx*bx + by*by); + else + precompute[i] = 0.0f; + } else + precompute[i] = 0.0f; + } + + for (y=iy0; y < iy1; ++y) { + for (x=ix0; x < ix1; ++x) { + float val; + float min_dist = 999999.0f; + float sx = (float) x + 0.5f; + float sy = (float) y + 0.5f; + float x_gspace = (sx / scale_x); + float y_gspace = (sy / scale_y); + + int winding = stbtt__compute_crossings_x(x_gspace, y_gspace, num_verts, verts); // @OPTIMIZE: this could just be a rasterization, but needs to be line vs. non-tesselated curves so a new path + + for (i=0; i < num_verts; ++i) { + float x0 = verts[i].x*scale_x, y0 = verts[i].y*scale_y; + + if (verts[i].type == STBTT_vline && precompute[i] != 0.0f) { + float x1 = verts[i-1].x*scale_x, y1 = verts[i-1].y*scale_y; + + float dist,dist2 = (x0-sx)*(x0-sx) + (y0-sy)*(y0-sy); + if (dist2 < min_dist*min_dist) + min_dist = (float) STBTT_sqrt(dist2); + + // coarse culling against bbox + //if (sx > STBTT_min(x0,x1)-min_dist && sx < STBTT_max(x0,x1)+min_dist && + // sy > STBTT_min(y0,y1)-min_dist && sy < STBTT_max(y0,y1)+min_dist) + dist = (float) STBTT_fabs((x1-x0)*(y0-sy) - (y1-y0)*(x0-sx)) * precompute[i]; + STBTT_assert(i != 0); + if (dist < min_dist) { + // check position along line + // x' = x0 + t*(x1-x0), y' = y0 + t*(y1-y0) + // minimize (x'-sx)*(x'-sx)+(y'-sy)*(y'-sy) + float dx = x1-x0, dy = y1-y0; + float px = x0-sx, py = y0-sy; + // minimize (px+t*dx)^2 + (py+t*dy)^2 = px*px + 2*px*dx*t + t^2*dx*dx + py*py + 2*py*dy*t + t^2*dy*dy + // derivative: 2*px*dx + 2*py*dy + (2*dx*dx+2*dy*dy)*t, set to 0 and solve + float t = -(px*dx + py*dy) / (dx*dx + dy*dy); + if (t >= 0.0f && t <= 1.0f) + min_dist = dist; + } + } else if (verts[i].type == STBTT_vcurve) { + float x2 = verts[i-1].x *scale_x, y2 = verts[i-1].y *scale_y; + float x1 = verts[i ].cx*scale_x, y1 = verts[i ].cy*scale_y; + float box_x0 = STBTT_min(STBTT_min(x0,x1),x2); + float box_y0 = STBTT_min(STBTT_min(y0,y1),y2); + float box_x1 = STBTT_max(STBTT_max(x0,x1),x2); + float box_y1 = STBTT_max(STBTT_max(y0,y1),y2); + // coarse culling against bbox to avoid computing cubic unnecessarily + if (sx > box_x0-min_dist && sx < box_x1+min_dist && sy > box_y0-min_dist && sy < box_y1+min_dist) { + int num=0; + float ax = x1-x0, ay = y1-y0; + float bx = x0 - 2*x1 + x2, by = y0 - 2*y1 + y2; + float mx = x0 - sx, my = y0 - sy; + float res[3] = {0.f,0.f,0.f}; + float px,py,t,it,dist2; + float a_inv = precompute[i]; + if (a_inv == 0.0) { // if a_inv is 0, it's 2nd degree so use quadratic formula + float a = 3*(ax*bx + ay*by); + float b = 2*(ax*ax + ay*ay) + (mx*bx+my*by); + float c = mx*ax+my*ay; + if (a == 0.0) { // if a is 0, it's linear + if (b != 0.0) { + res[num++] = -c/b; + } + } else { + float discriminant = b*b - 4*a*c; + if (discriminant < 0) + num = 0; + else { + float root = (float) STBTT_sqrt(discriminant); + res[0] = (-b - root)/(2*a); + res[1] = (-b + root)/(2*a); + num = 2; // don't bother distinguishing 1-solution case, as code below will still work + } + } + } else { + float b = 3*(ax*bx + ay*by) * a_inv; // could precompute this as it doesn't depend on sample point + float c = (2*(ax*ax + ay*ay) + (mx*bx+my*by)) * a_inv; + float d = (mx*ax+my*ay) * a_inv; + num = stbtt__solve_cubic(b, c, d, res); + } + dist2 = (x0-sx)*(x0-sx) + (y0-sy)*(y0-sy); + if (dist2 < min_dist*min_dist) + min_dist = (float) STBTT_sqrt(dist2); + + if (num >= 1 && res[0] >= 0.0f && res[0] <= 1.0f) { + t = res[0], it = 1.0f - t; + px = it*it*x0 + 2*t*it*x1 + t*t*x2; + py = it*it*y0 + 2*t*it*y1 + t*t*y2; + dist2 = (px-sx)*(px-sx) + (py-sy)*(py-sy); + if (dist2 < min_dist * min_dist) + min_dist = (float) STBTT_sqrt(dist2); + } + if (num >= 2 && res[1] >= 0.0f && res[1] <= 1.0f) { + t = res[1], it = 1.0f - t; + px = it*it*x0 + 2*t*it*x1 + t*t*x2; + py = it*it*y0 + 2*t*it*y1 + t*t*y2; + dist2 = (px-sx)*(px-sx) + (py-sy)*(py-sy); + if (dist2 < min_dist * min_dist) + min_dist = (float) STBTT_sqrt(dist2); + } + if (num >= 3 && res[2] >= 0.0f && res[2] <= 1.0f) { + t = res[2], it = 1.0f - t; + px = it*it*x0 + 2*t*it*x1 + t*t*x2; + py = it*it*y0 + 2*t*it*y1 + t*t*y2; + dist2 = (px-sx)*(px-sx) + (py-sy)*(py-sy); + if (dist2 < min_dist * min_dist) + min_dist = (float) STBTT_sqrt(dist2); + } + } + } + } + if (winding == 0) + min_dist = -min_dist; // if outside the shape, value is negative + val = onedge_value + pixel_dist_scale * min_dist; + if (val < 0) + val = 0; + else if (val > 255) + val = 255; + data[(y-iy0)*w+(x-ix0)] = (unsigned char) val; + } + } + STBTT_free(precompute, info->userdata); + STBTT_free(verts, info->userdata); + } + return data; +} + +STBTT_DEF unsigned char * stbtt_GetCodepointSDF(const stbtt_fontinfo *info, float scale, int codepoint, int padding, unsigned char onedge_value, float pixel_dist_scale, int *width, int *height, int *xoff, int *yoff) +{ + return stbtt_GetGlyphSDF(info, scale, stbtt_FindGlyphIndex(info, codepoint), padding, onedge_value, pixel_dist_scale, width, height, xoff, yoff); +} + +STBTT_DEF void stbtt_FreeSDF(unsigned char *bitmap, void *userdata) +{ + STBTT_free(bitmap, userdata); +} + +////////////////////////////////////////////////////////////////////////////// +// +// font name matching -- recommended not to use this +// + +// check if a utf8 string contains a prefix which is the utf16 string; if so return length of matching utf8 string +static stbtt_int32 stbtt__CompareUTF8toUTF16_bigendian_prefix(stbtt_uint8 *s1, stbtt_int32 len1, stbtt_uint8 *s2, stbtt_int32 len2) +{ + stbtt_int32 i=0; + + // convert utf16 to utf8 and compare the results while converting + while (len2) { + stbtt_uint16 ch = s2[0]*256 + s2[1]; + if (ch < 0x80) { + if (i >= len1) return -1; + if (s1[i++] != ch) return -1; + } else if (ch < 0x800) { + if (i+1 >= len1) return -1; + if (s1[i++] != 0xc0 + (ch >> 6)) return -1; + if (s1[i++] != 0x80 + (ch & 0x3f)) return -1; + } else if (ch >= 0xd800 && ch < 0xdc00) { + stbtt_uint32 c; + stbtt_uint16 ch2 = s2[2]*256 + s2[3]; + if (i+3 >= len1) return -1; + c = ((ch - 0xd800) << 10) + (ch2 - 0xdc00) + 0x10000; + if (s1[i++] != 0xf0 + (c >> 18)) return -1; + if (s1[i++] != 0x80 + ((c >> 12) & 0x3f)) return -1; + if (s1[i++] != 0x80 + ((c >> 6) & 0x3f)) return -1; + if (s1[i++] != 0x80 + ((c ) & 0x3f)) return -1; + s2 += 2; // plus another 2 below + len2 -= 2; + } else if (ch >= 0xdc00 && ch < 0xe000) { + return -1; + } else { + if (i+2 >= len1) return -1; + if (s1[i++] != 0xe0 + (ch >> 12)) return -1; + if (s1[i++] != 0x80 + ((ch >> 6) & 0x3f)) return -1; + if (s1[i++] != 0x80 + ((ch ) & 0x3f)) return -1; + } + s2 += 2; + len2 -= 2; + } + return i; +} + +static int stbtt_CompareUTF8toUTF16_bigendian_internal(char *s1, int len1, char *s2, int len2) +{ + return len1 == stbtt__CompareUTF8toUTF16_bigendian_prefix((stbtt_uint8*) s1, len1, (stbtt_uint8*) s2, len2); +} + +// returns results in whatever encoding you request... but note that 2-byte encodings +// will be BIG-ENDIAN... use stbtt_CompareUTF8toUTF16_bigendian() to compare +STBTT_DEF const char *stbtt_GetFontNameString(const stbtt_fontinfo *font, int *length, int platformID, int encodingID, int languageID, int nameID) +{ + stbtt_int32 i,count,stringOffset; + stbtt_uint8 *fc = font->data; + stbtt_uint32 offset = font->fontstart; + stbtt_uint32 nm = stbtt__find_table(fc, offset, "name"); + if (!nm) return NULL; + + count = ttUSHORT(fc+nm+2); + stringOffset = nm + ttUSHORT(fc+nm+4); + for (i=0; i < count; ++i) { + stbtt_uint32 loc = nm + 6 + 12 * i; + if (platformID == ttUSHORT(fc+loc+0) && encodingID == ttUSHORT(fc+loc+2) + && languageID == ttUSHORT(fc+loc+4) && nameID == ttUSHORT(fc+loc+6)) { + *length = ttUSHORT(fc+loc+8); + return (const char *) (fc+stringOffset+ttUSHORT(fc+loc+10)); + } + } + return NULL; +} + +static int stbtt__matchpair(stbtt_uint8 *fc, stbtt_uint32 nm, stbtt_uint8 *name, stbtt_int32 nlen, stbtt_int32 target_id, stbtt_int32 next_id) +{ + stbtt_int32 i; + stbtt_int32 count = ttUSHORT(fc+nm+2); + stbtt_int32 stringOffset = nm + ttUSHORT(fc+nm+4); + + for (i=0; i < count; ++i) { + stbtt_uint32 loc = nm + 6 + 12 * i; + stbtt_int32 id = ttUSHORT(fc+loc+6); + if (id == target_id) { + // find the encoding + stbtt_int32 platform = ttUSHORT(fc+loc+0), encoding = ttUSHORT(fc+loc+2), language = ttUSHORT(fc+loc+4); + + // is this a Unicode encoding? + if (platform == 0 || (platform == 3 && encoding == 1) || (platform == 3 && encoding == 10)) { + stbtt_int32 slen = ttUSHORT(fc+loc+8); + stbtt_int32 off = ttUSHORT(fc+loc+10); + + // check if there's a prefix match + stbtt_int32 matchlen = stbtt__CompareUTF8toUTF16_bigendian_prefix(name, nlen, fc+stringOffset+off,slen); + if (matchlen >= 0) { + // check for target_id+1 immediately following, with same encoding & language + if (i+1 < count && ttUSHORT(fc+loc+12+6) == next_id && ttUSHORT(fc+loc+12) == platform && ttUSHORT(fc+loc+12+2) == encoding && ttUSHORT(fc+loc+12+4) == language) { + slen = ttUSHORT(fc+loc+12+8); + off = ttUSHORT(fc+loc+12+10); + if (slen == 0) { + if (matchlen == nlen) + return 1; + } else if (matchlen < nlen && name[matchlen] == ' ') { + ++matchlen; + if (stbtt_CompareUTF8toUTF16_bigendian_internal((char*) (name+matchlen), nlen-matchlen, (char*)(fc+stringOffset+off),slen)) + return 1; + } + } else { + // if nothing immediately following + if (matchlen == nlen) + return 1; + } + } + } + + // @TODO handle other encodings + } + } + return 0; +} + +static int stbtt__matches(stbtt_uint8 *fc, stbtt_uint32 offset, stbtt_uint8 *name, stbtt_int32 flags) +{ + stbtt_int32 nlen = (stbtt_int32) STBTT_strlen((char *) name); + stbtt_uint32 nm,hd; + if (!stbtt__isfont(fc+offset)) return 0; + + // check italics/bold/underline flags in macStyle... + if (flags) { + hd = stbtt__find_table(fc, offset, "head"); + if ((ttUSHORT(fc+hd+44) & 7) != (flags & 7)) return 0; + } + + nm = stbtt__find_table(fc, offset, "name"); + if (!nm) return 0; + + if (flags) { + // if we checked the macStyle flags, then just check the family and ignore the subfamily + if (stbtt__matchpair(fc, nm, name, nlen, 16, -1)) return 1; + if (stbtt__matchpair(fc, nm, name, nlen, 1, -1)) return 1; + if (stbtt__matchpair(fc, nm, name, nlen, 3, -1)) return 1; + } else { + if (stbtt__matchpair(fc, nm, name, nlen, 16, 17)) return 1; + if (stbtt__matchpair(fc, nm, name, nlen, 1, 2)) return 1; + if (stbtt__matchpair(fc, nm, name, nlen, 3, -1)) return 1; + } + + return 0; +} + +static int stbtt_FindMatchingFont_internal(unsigned char *font_collection, char *name_utf8, stbtt_int32 flags) +{ + stbtt_int32 i; + for (i=0;;++i) { + stbtt_int32 off = stbtt_GetFontOffsetForIndex(font_collection, i); + if (off < 0) return off; + if (stbtt__matches((stbtt_uint8 *) font_collection, off, (stbtt_uint8*) name_utf8, flags)) + return off; + } +} + +#if defined(__GNUC__) || defined(__clang__) +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wcast-qual" +#endif + +STBTT_DEF int stbtt_BakeFontBitmap(const unsigned char *data, int offset, + float pixel_height, unsigned char *pixels, int pw, int ph, + int first_char, int num_chars, stbtt_bakedchar *chardata) +{ + return stbtt_BakeFontBitmap_internal((unsigned char *) data, offset, pixel_height, pixels, pw, ph, first_char, num_chars, chardata); +} + +STBTT_DEF int stbtt_GetFontOffsetForIndex(const unsigned char *data, int index) +{ + return stbtt_GetFontOffsetForIndex_internal((unsigned char *) data, index); +} + +STBTT_DEF int stbtt_GetNumberOfFonts(const unsigned char *data) +{ + return stbtt_GetNumberOfFonts_internal((unsigned char *) data); +} + +STBTT_DEF int stbtt_InitFont(stbtt_fontinfo *info, const unsigned char *data, int offset) +{ + return stbtt_InitFont_internal(info, (unsigned char *) data, offset); +} + +STBTT_DEF int stbtt_FindMatchingFont(const unsigned char *fontdata, const char *name, int flags) +{ + return stbtt_FindMatchingFont_internal((unsigned char *) fontdata, (char *) name, flags); +} + +STBTT_DEF int stbtt_CompareUTF8toUTF16_bigendian(const char *s1, int len1, const char *s2, int len2) +{ + return stbtt_CompareUTF8toUTF16_bigendian_internal((char *) s1, len1, (char *) s2, len2); +} + +#if defined(__GNUC__) || defined(__clang__) +#pragma GCC diagnostic pop +#endif + +#endif // STB_TRUETYPE_IMPLEMENTATION + + +// FULL VERSION HISTORY +// +// 1.25 (2021-07-11) many fixes +// 1.24 (2020-02-05) fix warning +// 1.23 (2020-02-02) query SVG data for glyphs; query whole kerning table (but only kern not GPOS) +// 1.22 (2019-08-11) minimize missing-glyph duplication; fix kerning if both 'GPOS' and 'kern' are defined +// 1.21 (2019-02-25) fix warning +// 1.20 (2019-02-07) PackFontRange skips missing codepoints; GetScaleFontVMetrics() +// 1.19 (2018-02-11) OpenType GPOS kerning (horizontal only), STBTT_fmod +// 1.18 (2018-01-29) add missing function +// 1.17 (2017-07-23) make more arguments const; doc fix +// 1.16 (2017-07-12) SDF support +// 1.15 (2017-03-03) make more arguments const +// 1.14 (2017-01-16) num-fonts-in-TTC function +// 1.13 (2017-01-02) support OpenType fonts, certain Apple fonts +// 1.12 (2016-10-25) suppress warnings about casting away const with -Wcast-qual +// 1.11 (2016-04-02) fix unused-variable warning +// 1.10 (2016-04-02) allow user-defined fabs() replacement +// fix memory leak if fontsize=0.0 +// fix warning from duplicate typedef +// 1.09 (2016-01-16) warning fix; avoid crash on outofmem; use alloc userdata for PackFontRanges +// 1.08 (2015-09-13) document stbtt_Rasterize(); fixes for vertical & horizontal edges +// 1.07 (2015-08-01) allow PackFontRanges to accept arrays of sparse codepoints; +// allow PackFontRanges to pack and render in separate phases; +// fix stbtt_GetFontOFfsetForIndex (never worked for non-0 input?); +// fixed an assert() bug in the new rasterizer +// replace assert() with STBTT_assert() in new rasterizer +// 1.06 (2015-07-14) performance improvements (~35% faster on x86 and x64 on test machine) +// also more precise AA rasterizer, except if shapes overlap +// remove need for STBTT_sort +// 1.05 (2015-04-15) fix misplaced definitions for STBTT_STATIC +// 1.04 (2015-04-15) typo in example +// 1.03 (2015-04-12) STBTT_STATIC, fix memory leak in new packing, various fixes +// 1.02 (2014-12-10) fix various warnings & compile issues w/ stb_rect_pack, C++ +// 1.01 (2014-12-08) fix subpixel position when oversampling to exactly match +// non-oversampled; STBTT_POINT_SIZE for packed case only +// 1.00 (2014-12-06) add new PackBegin etc. API, w/ support for oversampling +// 0.99 (2014-09-18) fix multiple bugs with subpixel rendering (ryg) +// 0.9 (2014-08-07) support certain mac/iOS fonts without an MS platformID +// 0.8b (2014-07-07) fix a warning +// 0.8 (2014-05-25) fix a few more warnings +// 0.7 (2013-09-25) bugfix: subpixel glyph bug fixed in 0.5 had come back +// 0.6c (2012-07-24) improve documentation +// 0.6b (2012-07-20) fix a few more warnings +// 0.6 (2012-07-17) fix warnings; added stbtt_ScaleForMappingEmToPixels, +// stbtt_GetFontBoundingBox, stbtt_IsGlyphEmpty +// 0.5 (2011-12-09) bugfixes: +// subpixel glyph renderer computed wrong bounding box +// first vertex of shape can be off-curve (FreeSans) +// 0.4b (2011-12-03) fixed an error in the font baking example +// 0.4 (2011-12-01) kerning, subpixel rendering (tor) +// bugfixes for: +// codepoint-to-glyph conversion using table fmt=12 +// codepoint-to-glyph conversion using table fmt=4 +// stbtt_GetBakedQuad with non-square texture (Zer) +// updated Hello World! sample to use kerning and subpixel +// fixed some warnings +// 0.3 (2009-06-24) cmap fmt=12, compound shapes (MM) +// userdata, malloc-from-userdata, non-zero fill (stb) +// 0.2 (2009-03-11) Fix unsigned/signed char warnings +// 0.1 (2009-03-09) First public release +// + +/* +------------------------------------------------------------------------------ +This software is available under 2 licenses -- choose whichever you prefer. +------------------------------------------------------------------------------ +ALTERNATIVE A - MIT License +Copyright (c) 2017 Sean Barrett +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. +------------------------------------------------------------------------------ +ALTERNATIVE B - Public Domain (www.unlicense.org) +This is free and unencumbered software released into the public domain. +Anyone is free to copy, modify, publish, use, compile, sell, or distribute this +software, either in source code form or as a compiled binary, for any purpose, +commercial or non-commercial, and by any means. +In jurisdictions that recognize copyright laws, the author or authors of this +software dedicate any and all copyright interest in the software to the public +domain. We make this dedication for the benefit of the public at large and to +the detriment of our heirs and successors. We intend this dedication to be an +overt act of relinquishment in perpetuity of all present and future rights to +this software under copyright law. +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN +ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +------------------------------------------------------------------------------ +*/
--- a/tests/test_gltf.nim Sun Jul 28 23:18:05 2024 +0700 +++ b/tests/test_gltf.nim Mon Jul 29 00:00:28 2024 +0700 @@ -29,7 +29,7 @@ emissive: Vec4f = NewVec4f(0, 0, 0, 0) # emissiveTexture: int32 = -1 MainDescriptors = object - materials: array[32, GPUValue[Material, UniformBuffer]] + materials: array[50, GPUValue[Material, UniformBuffer]] camera: GPUValue[Camera, UniformBufferMapped] Shader = object objectData {.PushConstantAttribute.}: ObjectData @@ -129,7 +129,7 @@ renderdata.AssignBuffers(primitive[0]) renderdata.AssignBuffers(descriptors) - var pipeline = CreatePipeline[Shader](renderPass = vulkan.swapchain.renderPass) + var pipeline = CreatePipeline[Shader](renderPass = vulkan.swapchain.renderPass, cullMode=[]) InitDescriptorSet(renderdata, pipeline.descriptorSetLayouts[0], descriptors) renderdata.FlushAllMemory()