diff --git a/src/fontstash.h b/src/fontstash.h index 7137709..44e7331 100644 --- a/src/fontstash.h +++ b/src/fontstash.h @@ -41,7 +41,7 @@ enum FONSalign { enum FONSerrorCode { // Font atlas is full. FONS_ATLAS_FULL = 1, - // Scratch memory used to render glyphs is full, requested size reported in 'val', you may need to bump up FONS_SCRATCH_BUF_SIZE. + // Scratch memory used to render glyphs is full, requested size reported in 'val', you may need to bump up FONS_SCRATCH_BUF_SIZE. FONS_SCRATCH_FULL = 2, // Calls to fonsPushState has created too large stack, if you need deep state stack bump up FONS_MAX_STATES. FONS_STATES_OVERFLOW = 3, @@ -90,7 +90,7 @@ void fonsDeleteInternal(FONScontext* s); void fonsSetErrorCallback(FONScontext* s, void (*callback)(void* uptr, int error, int val), void* uptr); // Returns current atlas size. void fonsGetAtlasSize(FONScontext* s, int* width, int* height); -// Expands the atlas size. +// Expands the atlas size. int fonsExpandAtlas(FONScontext* s, int width, int height); // Resets the whole stash. int fonsResetAtlas(FONScontext* stash, int width, int height); @@ -154,7 +154,7 @@ typedef struct FONSttFontImpl FONSttFontImpl; static FT_Library ftLibrary; int fons__tt_init(FONScontext *context) -{ +{ FT_Error ftError; FONS_NOTUSED(context); ftError = FT_Init_FreeType(&ftLibrary); @@ -304,7 +304,7 @@ int fons__tt_getGlyphKernAdvance(FONSttFontImpl *font, int glyph1, int glyph2) #endif #ifndef FONS_SCRATCH_BUF_SIZE -# define FONS_SCRATCH_BUF_SIZE 16000 +# define FONS_SCRATCH_BUF_SIZE 64000 #endif #ifndef FONS_HASH_LUT_SIZE # define FONS_HASH_LUT_SIZE 256 @@ -1403,7 +1403,7 @@ void fonsDrawDebug(FONScontext* stash, float x, float y) } float fonsTextBounds(FONScontext* stash, - float x, float y, + float x, float y, const char* str, const char* end, float* bounds) { @@ -1591,7 +1591,7 @@ int fonsExpandAtlas(FONScontext* stash, int width, int height) height = fons__maxi(height, stash->params.height); if (width == stash->params.width && height == stash->params.height) - return 1; + return 1; // Flush pending glyphs. fons__flush(stash); diff --git a/src/nanovg.c b/src/nanovg.c index 44cf25f..c640e85 100644 --- a/src/nanovg.c +++ b/src/nanovg.c @@ -18,6 +18,9 @@ #include #include +#include +#include + #include "nanovg.h" #define FONTSTASH_IMPLEMENTATION #include "fontstash.h" @@ -291,7 +294,7 @@ void nvgBeginFrame(NVGcontext* ctx, int windowWidth, int windowHeight, float dev nvgReset(ctx); nvg__setDevicePixelRatio(ctx, devicePixelRatio); - + ctx->params.renderViewport(ctx->params.userPtr, windowWidth, windowHeight); ctx->drawCallCount = 0; @@ -391,7 +394,7 @@ NVGcolor nvgLerpRGBA(NVGcolor c0, NVGcolor c1, float u) { cint.rgba[i] = c0.rgba[i] * oneminu + c1.rgba[i] * u; } - + return cint; } @@ -913,7 +916,7 @@ void nvgIntersectScissor(NVGcontext* ctx, float x, float y, float w, float h) } // Transform the current scissor rect into current transform space. - // If there is difference in rotation, this will be approximation. + // If there is difference in rotation, this will be approximation. memcpy(pxform, state->scissor.xform, sizeof(float)*6); ex = state->scissor.extent[0]; ey = state->scissor.extent[1]; @@ -1173,7 +1176,7 @@ static void nvg__tesselateBezier(NVGcontext* ctx, { float x12,y12,x23,y23,x34,y34,x123,y123,x234,y234,x1234,y1234; float dx,dy,d2,d3; - + if (level > 10) return; x12 = (x1+x2)*0.5f; @@ -1205,8 +1208,8 @@ static void nvg__tesselateBezier(NVGcontext* ctx, x1234 = (x123+x234)*0.5f; y1234 = (y123+y234)*0.5f; - nvg__tesselateBezier(ctx, x1,y1, x12,y12, x123,y123, x1234,y1234, level+1, 0); - nvg__tesselateBezier(ctx, x1234,y1234, x234,y234, x34,y34, x4,y4, level+1, type); + nvg__tesselateBezier(ctx, x1,y1, x12,y12, x123,y123, x1234,y1234, level+1, 0); + nvg__tesselateBezier(ctx, x1234,y1234, x234,y234, x34,y34, x4,y4, level+1, type); } static void nvg__flattenPaths(NVGcontext* ctx) @@ -1604,7 +1607,7 @@ static void nvg__calculateJoins(NVGcontext* ctx, float w, int lineJoin, float mi static int nvg__expandStroke(NVGcontext* ctx, float w, int lineCap, int lineJoin, float miterLimit) -{ +{ NVGpathCache* cache = ctx->cache; NVGvertex* verts; NVGvertex* dst; @@ -1868,7 +1871,7 @@ void nvgQuadTo(NVGcontext* ctx, float cx, float cy, float x, float y) { float x0 = ctx->commandx; float y0 = ctx->commandy; - float vals[] = { NVG_BEZIERTO, + float vals[] = { NVG_BEZIERTO, x0 + 2.0f/3.0f*(cx - x0), y0 + 2.0f/3.0f*(cy - y0), x + 2.0f/3.0f*(cx - x), y + 2.0f/3.0f*(cy - y), x, y }; @@ -1950,7 +1953,7 @@ void nvgArc(NVGcontext* ctx, float cx, float cy, float r, float a0, float a1, in float px = 0, py = 0, ptanx = 0, ptany = 0; float vals[3 + 5*7 + 100]; int i, ndivs, nvals; - int move = ctx->ncommands > 0 ? NVG_LINETO : NVG_MOVETO; + int move = ctx->ncommands > 0 ? NVG_LINETO : NVG_MOVETO; // Clamp angles da = a1 - a0; @@ -2343,7 +2346,7 @@ float nvgText(NVGcontext* ctx, float x, float y, const char* string, const char* } } - // TODO: add back-end bit to do this just once per frame. + // TODO: add back-end bit to do this just once per frame. nvg__flushTextTexture(ctx); nvg__renderText(ctx, verts, nverts); diff --git a/src/stb_image.h b/src/stb_image.h index 4da7ec6..e06f7a1 100644 --- a/src/stb_image.h +++ b/src/stb_image.h @@ -1,87 +1,236 @@ -/* stb_image - v1.39 - public domain JPEG/PNG reader - http://nothings.org/stb_image.c - when you control the images you're loading +/* stb_image - v2.10 - public domain image loader - http://nothings.org/stb_image.h 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 (no JPEG progressive) - PNG 8-bit-per-channel only + JPEG baseline & progressive (12 bpc/arithmetic not supported, same as stock IJG lib) + PNG 1/2/4/8-bit-per-channel (16 bpc not supported) TGA (not sure what subset, if a subset) BMP non-1bpp, non-RLE - PSD (composited view only, no extra channels) + 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) - - - stbi__jpeg_huff_decode from memory or through FILE (define STBI_NO_STDIO to remove code) - - stbi__jpeg_huff_decode from arbitrary I/O callbacks - - overridable dequantizing-IDCT, YCbCr-to-RGB conversion (define STBI_SIMD) - - Latest revisions: - 1.39 (2014-06-15) TGA optimization fix, multiple BMP fixes - 1.38 (2014-06-06) suppress MSVC run-time warnings, fix accidental rename of 'skip' - 1.37 (2014-06-04) remove duplicate typedef - 1.36 (2014-06-03) converted to header file, allow reading incorrect iphoned-images without iphone flag - 1.35 (2014-05-27) warnings, bugfixes, TGA optimization, etc - 1.34 (unknown ) warning fix - 1.33 (2011-07-14) minor fixes suggested by Dave Moore + 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. + + + Revision 2.00 release notes: + + - Progressive JPEG is now supported. + + - PPM and PGM binary formats are now supported, thanks to Ken Miller. + + - x86 platforms now make use of SSE2 SIMD instructions for + JPEG decoding, and ARM platforms can use NEON SIMD if requested. + This work was done by Fabian "ryg" Giesen. SSE2 is used by + default, but NEON must be enabled explicitly; see docs. + + With other JPEG optimizations included in this version, we see + 2x speedup on a JPEG on an x86 machine, and a 1.5x speedup + on a JPEG on an ARM machine, relative to previous versions of this + library. The same results will not obtain for all JPGs and for all + x86/ARM machines. (Note that progressive JPEGs are significantly + slower to decode than regular JPEGs.) This doesn't mean that this + is the fastest JPEG decoder in the land; rather, it brings it + closer to parity with standard libraries. If you want the fastest + decode, look elsewhere. (See "Philosophy" section of docs below.) + + See final bullet items below for more info on SIMD. + + - Added STBI_MALLOC, STBI_REALLOC, and STBI_FREE macros for replacing + the memory allocator. Unlike other STBI libraries, these macros don't + support a context parameter, so if you need to pass a context in to + the allocator, you'll have to store it in a global or a thread-local + variable. + + - Split existing STBI_NO_HDR flag into two flags, STBI_NO_HDR and + STBI_NO_LINEAR. + STBI_NO_HDR: suppress implementation of .hdr reader format + STBI_NO_LINEAR: suppress high-dynamic-range light-linear float API + + - 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) + + Note that you can define multiples of these, and you will get all + of them ("only x" and "only y" is interpreted to mean "only x&y"). + + - If you use STBI_NO_PNG (or _ONLY_ without PNG), and you still + want the zlib decoder to be available, #define STBI_SUPPORT_ZLIB + + - Compilation of all SIMD code can be suppressed with + #define STBI_NO_SIMD + It should not be necessary to disable SIMD unless you have issues + compiling (e.g. using an x86 compiler which doesn't support SSE + intrinsics or that doesn't support the method used to detect + SSE2 support at run-time), and even those can be reported as + bugs so I can refine the built-in compile-time checking to be + smarter. + + - The old STBI_SIMD system which allowed installing a user-defined + IDCT etc. has been removed. If you need this, don't upgrade. My + assumption is that almost nobody was doing this, and those who + were will find the built-in SIMD more satisfactory anyway. + + - RGB values computed for JPEG images are slightly different from + previous versions of stb_image. (This is due to using less + integer precision in SIMD.) The C code has been adjusted so + that the same RGB values will be computed regardless of whether + SIMD support is available, so your app should always produce + consistent results. But these results are slightly different from + previous versions. (Specifically, about 3% of available YCbCr values + will compute different RGB results from pre-1.49 versions by +-1; + most of the deviating values are one smaller in the G channel.) + + - If you must produce consistent results with previous versions of + stb_image, #define STBI_JPEG_OLD and you will get the same results + you used to; however, you will not get the SIMD speedups for + the YCbCr-to-RGB conversion step (although you should still see + significant JPEG speedup from the other changes). + + Please note that STBI_JPEG_OLD is a temporary feature; it will be + removed in future versions of the library. It is only intended for + near-term back-compatibility use. + + + Latest revision history: + 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 + 2.08 (2015-09-13) fix to 2.07 cleanup, reading RGB PSD as RGBA + 2.07 (2015-09-13) partial animated GIF support + limited 16-bit PSD support + minor bugs, code cleanup, and compiler warnings + 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) additional corruption checking + stbi_set_flip_vertically_on_load + fix NEON support; fix mingw support + 2.02 (2015-01-19) fix incorrect assert, fix warning + 2.01 (2015-01-17) fix various warnings + 2.00b (2014-12-25) fix STBI_MALLOC in progressive JPEG + 2.00 (2014-12-25) optimize JPEG, including x86 SSE2 & ARM NEON SIMD + progressive JPEG + PGM/PPM support + STBI_MALLOC,STBI_REALLOC,STBI_FREE + STBI_NO_*, STBI_ONLY_* + GIF bugfix + 1.48 (2014-12-14) fix incorrectly-named assert() + 1.47 (2014-12-14) 1/2/4-bit PNG support (both grayscale and paletted) + optimize PNG + fix bug in interlaced PNG with user-specified channel count See end of file for full revision history. - TODO: - stbi_info support for BMP,PSD,HDR,PIC - ============================ Contributors ========================= - - Image formats Bug fixes & warning fixes - Sean Barrett (jpeg, png, bmp) Marc LeBlanc - Nicolas Schulz (hdr, psd) Christpher Lloyd - Jonathan Dummer (tga) Dave Moore - Jean-Marc Lienher (gif) Won Chun - Tom Seddon (pic) the Horde3D community - Thatcher Ulrich (psd) Janez Zemva - Jonathan Blow - Laurent Gomila - Extensions, features Aruelien Pocheville - Jetro Lauha (stbi_info) Ryamond Barbiero - James "moose2000" Brown (iPhone PNG) David Woo - Ben "Disch" Wenger (io callbacks) Roy Eltham - Martin "SpartanJ" Golini Luke Graham - Thomas Ruf - John Bartholomew - Optimizations & bugfixes Ken Hamada - Fabian "ryg" Giesen Cort Stratton - Arseny Kapoulkine Blazej Dariusz Roszkowski - Thibault Reuille - If your name should be here but Paul Du Bois - isn't, let Sean know. Guillaume George - Jerry Jansson - Hayaki Saito + + 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) + urraka@github (animated gif) Junggon Kim (PNM comments) + Daniel Gibson (16-bit TGA) + + Optimizations & bugfixes + Fabian "ryg" Giesen + Arseny Kapoulkine + + Bug & warning fixes + Marc LeBlanc David Woo Guillaume George Martins Mozeiko + Christpher Lloyd Martin Golini Jerry Jansson Joseph Thomson + Dave Moore Roy Eltham Hayaki Saito Phil Jordan + Won Chun Luke Graham Johan Duparc Nathan Reed + the Horde3D community Thomas Ruf Ronny Chevalier Nick Verigakis + Janez Zemva John Bartholomew Michal Cichon svdijk@github + Jonathan Blow Ken Hamada Tero Hanninen Baldur Karlsson + Laurent Gomila Cort Stratton Sergio Gonzalez romigrou@github + Aruelien Pocheville Thibault Reuille Cass Everitt + Ryamond Barbiero Paul Du Bois Engin Manap + Blazej Dariusz Roszkowski + Michaelangel007@github + + +LICENSE + +This software is in the public domain. Where that dedication is not +recognized, you are granted a perpetual, irrevocable license to copy, +distribute, and modify this file as you see fit. + */ #ifndef STBI_INCLUDE_STB_IMAGE_H #define STBI_INCLUDE_STB_IMAGE_H +// DOCUMENTATION +// // Limitations: -// - no jpeg progressive support -// - non-HDR formats support 8-bit samples only (jpeg, png) -// - no delayed line count (jpeg) -- IJG doesn't support either +// - no 16-bit-per-channel PNG +// - no 12-bit-per-channel JPEG +// - no JPEGs with arithmetic coding // - no 1-bit BMP // - GIF always returns *comp=4 // -// Basic usage (see HDR discussion below): +// 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 ... +// // ... 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 @@ -94,14 +243,16 @@ // int req_comp -- 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. The pixel data consists of *y scanlines of *x pixels, +// 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 'req_comp' if req_comp is non-zero, or *comp otherwise. // If req_comp is non-zero, *comp has the number of components that _would_ // have been output otherwise. E.g. if you set req_comp to 4, you will always -// get RGBA output, but you can check *comp to easily see if it's opaque. +// get RGBA output, but you can check *comp 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: @@ -123,18 +274,66 @@ // // =========================================================================== // -// iPhone PNG support: +// Philosophy // -// By default we convert iphone-formatted PNGs back to RGB; nominally they -// would silently load as BGR, except the existing code should have just -// failed on such iPhone PNGs. But you can disable this conversion by -// by calling stbi_convert_iphone_png_to_rgb(0), in which case -// you will always just get the native iphone "format" through. +// stb libraries are designed with the following priorities: // -// 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). +// 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 +// make more explicit reasons why performance can't be emphasized. +// +// - Portable ("ease of use") +// - Small 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. +// +// The output of the JPEG decoder is slightly different from versions where +// SIMD support was introduced (that is, for versions before 1.49). The +// difference is only +-1 in the 8-bit RGB channels, and only on a small +// fraction of pixels. You can force the pre-1.49 behavior by defining +// STBI_JPEG_OLD, but this will disable some of the SIMD decoding path +// and hence cost some performance. +// +// 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. // // =========================================================================== // @@ -157,7 +356,7 @@ // (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: @@ -174,24 +373,22 @@ // // =========================================================================== // -// I/O callbacks +// iPhone PNG support: // -// 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. +// By default we convert iphone-formatted PNGs back to RGB, even though +// they are internally encoded differently. You can disable this conversion +// by by calling stbi_convert_iphone_png_to_rgb(0), in which case +// you will always just get the native iphone "format" through (which +// is BGR stored in RGB). +// +// 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). // -// 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). #ifndef STBI_NO_STDIO - -#if defined(_MSC_VER) && _MSC_VER >= 1400 -#define _CRT_SECURE_NO_WARNINGS // suppress warnings about fopen() -#pragma warning(push) -#pragma warning(disable:4996) // suppress even more warnings about fopen() -#endif #include #endif // STBI_NO_STDIO @@ -228,41 +425,43 @@ extern "C" { // load image by filename, open file, or memory buffer // -STBIDEF stbi_uc *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); - -#ifndef STBI_NO_STDIO -STBIDEF stbi_uc *stbi_load (char const *filename, int *x, int *y, int *comp, int req_comp); -STBIDEF stbi_uc *stbi_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); -// for stbi_load_from_file, file pointer is left pointing immediately after image -#endif - typedef struct { - int (*read) (void *user,char *data,int size); // fill 'data' with 'size' bytes. return number of bytes actually read + 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; -STBIDEF stbi_uc *stbi_load_from_callbacks (stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp); +STBIDEF stbi_uc *stbi_load (char const *filename, int *x, int *y, int *comp, int req_comp); +STBIDEF stbi_uc *stbi_load_from_memory (stbi_uc const *buffer, int len , int *x, int *y, int *comp, int 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); -#ifndef STBI_NO_HDR - STBIDEF float *stbi_loadf_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); +#ifndef STBI_NO_STDIO +STBIDEF stbi_uc *stbi_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +// for stbi_load_from_file, file pointer is left pointing immediately after image +#endif + +#ifndef STBI_NO_LINEAR + STBIDEF float *stbi_loadf (char const *filename, int *x, int *y, int *comp, int req_comp); + STBIDEF float *stbi_loadf_from_memory (stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); + STBIDEF float *stbi_loadf_from_callbacks (stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp); #ifndef STBI_NO_STDIO - STBIDEF float *stbi_loadf (char const *filename, int *x, int *y, int *comp, int req_comp); STBIDEF float *stbi_loadf_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); #endif - - STBIDEF float *stbi_loadf_from_callbacks (stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp); +#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_HDR +#endif // STBI_NO_LINEAR -// stbi_is_hdr is always defined +// 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 @@ -273,7 +472,7 @@ STBIDEF int stbi_is_hdr_from_file(FILE *f); // get a VERY brief reason for failure // NOT THREADSAFE -STBIDEF const char *stbi_failure_reason (void); +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); @@ -299,6 +498,8 @@ STBIDEF void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultipl // 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); // ZLIB client - used by PNG, available for other purposes @@ -311,26 +512,6 @@ STBIDEF char *stbi_zlib_decode_noheader_malloc(const char *buffer, int len, int STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen); -// define faster low-level operations (typically SIMD support) -#ifdef STBI_SIMD -typedef void (*stbi_idct_8x8)(stbi_uc *out, int out_stride, short data[64], unsigned short *dequantize); -// compute an integer IDCT on "input" -// input[x] = data[x] * dequantize[x] -// write results to 'out': 64 samples, each run of 8 spaced by 'out_stride' -// CLAMP results to 0..255 -typedef void (*stbi_YCbCr_to_RGB_run)(stbi_uc *output, stbi_uc const *y, stbi_uc const *cb, stbi_uc const *cr, int count, int step); -// compute a conversion from YCbCr to RGB -// 'count' pixels -// write pixels to 'output'; each pixel is 'step' bytes (either 3 or 4; if 4, write '255' as 4th), order R,G,B -// y: Y input channel -// cb: Cb input channel; scale/biased to be 0..255 -// cr: Cr input channel; scale/biased to be 0..255 - -STBIDEF void stbi_install_idct(stbi_idct_8x8 func); -STBIDEF void stbi_install_YCbCr_to_RGB(stbi_YCbCr_to_RGB_run func); -#endif // STBI_SIMD - - #ifdef __cplusplus } #endif @@ -342,19 +523,62 @@ STBIDEF void stbi_install_YCbCr_to_RGB(stbi_YCbCr_to_RGB_run func); #ifdef STB_IMAGE_IMPLEMENTATION -#ifndef STBI_NO_HDR +#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 +#include // ptrdiff_t on osx +#include +#include + +#if !defined(STBI_NO_LINEAR) || !defined(STBI_NO_HDR) #include // ldexp -#include // strcmp, strtok #endif #ifndef STBI_NO_STDIO #include #endif -#include -#include + +#ifndef STBI_ASSERT #include -#include -#include // ptrdiff_t on osx +#define STBI_ASSERT(x) assert(x) +#endif + #ifndef _MSC_VER #ifdef __cplusplus @@ -399,6 +623,121 @@ typedef unsigned char validate_uint32[sizeof(stbi__uint32)==4 ? 1 : -1]; #define stbi_lrot(x,y) (((x) << (y)) | ((x) >> (32 - (y)))) #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(STBI__X64_TARGET)) && !defined(__SSE2__) && !defined(STBI_NO_SIMD) +// NOTE: not clear do we actually need this for the 64-bit path? +// gcc doesn't support sse2 intrinsics unless you compile with -msse2, +// (but compiling with -msse2 allows the compiler to use SSE2 everywhere; +// this is just broken and gcc are jerks for not fixing it properly +// http://www.virtualdub.org/blog/pivot/entry.php?id=363 ) +#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) +#define STBI_SSE2 +#include + +#ifdef _MSC_VER + +#if _MSC_VER >= 1400 // not VC6 +#include // __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 + +static int stbi__sse2_available() +{ + int info3 = stbi__cpuid3(); + return ((info3 >> 26) & 1) != 0; +} +#else // assume GCC-style if not VC++ +#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16))) + +static int stbi__sse2_available() +{ +#if defined(__GNUC__) && (__GNUC__ * 100 + __GNUC_MINOR__) >= 408 // GCC 4.8 or later + // GCC 4.8+ has a nice way to do this + return __builtin_cpu_supports("sse2"); +#else + // portable way to do this, preferably without using GCC inline ASM? + // just bail for now. + return 0; +#endif +} +#endif +#endif + +// ARM NEON +#if defined(STBI_NO_SIMD) && defined(STBI_NEON) +#undef STBI_NEON +#endif + +#ifdef STBI_NEON +#include +// assume GCC or Clang on ARM targets +#define STBI_SIMD_ALIGN(type, name) type name __attribute__((aligned(16))) +#endif + +#ifndef STBI_SIMD_ALIGN +#define STBI_SIMD_ALIGN(type, name) type name +#endif + /////////////////////////////////////////////// // // stbi__context struct and start_xxx functions @@ -409,7 +748,7 @@ typedef struct { stbi__uint32 img_x, img_y; int img_n, img_out_n; - + stbi_io_callbacks io; void *io_user_data; @@ -418,19 +757,19 @@ typedef struct stbi_uc buffer_start[128]; stbi_uc *img_buffer, *img_buffer_end; - stbi_uc *img_buffer_original; + stbi_uc *img_buffer_original, *img_buffer_original_end; } stbi__context; static void stbi__refill_buffer(stbi__context *s); -// initialize a memory-stbi__jpeg_huff_decode context +// 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->img_buffer = s->img_buffer_original = (stbi_uc *) buffer; - s->img_buffer_end = (stbi_uc *) buffer+len; + s->img_buffer_end = s->img_buffer_original_end = (stbi_uc *) buffer+len; } // initialize a callback-based context @@ -442,6 +781,7 @@ static void stbi__start_callbacks(stbi__context *s, stbi_io_callbacks *c, void * s->read_from_callbacks = 1; s->img_buffer_original = s->buffer_start; stbi__refill_buffer(s); + s->img_buffer_original_end = s->img_buffer_end; } #ifndef STBI_NO_STDIO @@ -483,31 +823,62 @@ static void stbi__rewind(stbi__context *s) // 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; } +#ifndef STBI_NO_JPEG static int stbi__jpeg_test(stbi__context *s); static stbi_uc *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *comp, int req_comp); 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 stbi_uc *stbi__png_load(stbi__context *s, int *x, int *y, int *comp, int req_comp); static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp); +#endif + +#ifndef STBI_NO_BMP static int stbi__bmp_test(stbi__context *s); static stbi_uc *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp); +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 stbi_uc *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp); 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 stbi_uc *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int req_comp); +static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp); +#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); +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 stbi_uc *stbi__pic_load(stbi__context *s, int *x, int *y, int *comp, int req_comp); +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 stbi_uc *stbi__gif_load(stbi__context *s, int *x, int *y, 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 stbi_uc *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp); +static int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp); +#endif // this is not threadsafe static const char *stbi__g_failure_reason; @@ -523,6 +894,11 @@ static int stbi__err(const char *str) return 0; } +static void *stbi__malloc(size_t size) +{ + return STBI_MALLOC(size); +} + // stbi__err - error // stbi__errpf - error returning pointer to float // stbi__errpuc - error returning pointer to unsigned char @@ -535,27 +911,52 @@ static int stbi__err(const char *str) #define stbi__err(x,y) stbi__err(x) #endif -#define stbi__errpf(x,y) ((float *) (stbi__err(x,y)?NULL:NULL)) -#define stbi__errpuc(x,y) ((unsigned char *) (stbi__err(x,y)?NULL:NULL)) +#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) { - free(retval_from_stbi_load); + STBI_FREE(retval_from_stbi_load); } -#ifndef STBI_NO_HDR +#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 unsigned char *stbi_load_main(stbi__context *s, int *x, int *y, int *comp, int req_comp) +static int stbi__vertically_flip_on_load = 0; + +STBIDEF void stbi_set_flip_vertically_on_load(int flag_true_if_should_flip) +{ + stbi__vertically_flip_on_load = flag_true_if_should_flip; +} + +static unsigned char *stbi__load_main(stbi__context *s, int *x, int *y, int *comp, int req_comp) { + #ifndef STBI_NO_JPEG if (stbi__jpeg_test(s)) return stbi__jpeg_load(s,x,y,comp,req_comp); + #endif + #ifndef STBI_NO_PNG if (stbi__png_test(s)) return stbi__png_load(s,x,y,comp,req_comp); + #endif + #ifndef STBI_NO_BMP if (stbi__bmp_test(s)) return stbi__bmp_load(s,x,y,comp,req_comp); + #endif + #ifndef STBI_NO_GIF if (stbi__gif_test(s)) return stbi__gif_load(s,x,y,comp,req_comp); + #endif + #ifndef STBI_NO_PSD if (stbi__psd_test(s)) return stbi__psd_load(s,x,y,comp,req_comp); + #endif + #ifndef STBI_NO_PIC if (stbi__pic_test(s)) return stbi__pic_load(s,x,y,comp,req_comp); + #endif + #ifndef STBI_NO_PNM + if (stbi__pnm_test(s)) return stbi__pnm_load(s,x,y,comp,req_comp); + #endif #ifndef STBI_NO_HDR if (stbi__hdr_test(s)) { @@ -564,16 +965,81 @@ static unsigned char *stbi_load_main(stbi__context *s, int *x, int *y, int *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); + #endif + return stbi__errpuc("unknown image type", "Image not of any known type, or corrupt"); } +static unsigned char *stbi__load_flip(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *result = stbi__load_main(s, x, y, comp, req_comp); + + if (stbi__vertically_flip_on_load && result != NULL) { + int w = *x, h = *y; + int depth = req_comp ? req_comp : *comp; + int row,col,z; + stbi_uc temp; + + // @OPTIMIZE: use a bigger temp buffer and memcpy multiple pixels at once + for (row = 0; row < (h>>1); row++) { + for (col = 0; col < w; col++) { + for (z = 0; z < depth; z++) { + temp = result[(row * w + col) * depth + z]; + result[(row * w + col) * depth + z] = result[((h - row - 1) * w + col) * depth + z]; + result[((h - row - 1) * w + col) * depth + z] = temp; + } + } + } + } + + return result; +} + +#ifndef STBI_NO_HDR +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 w = *x, h = *y; + int depth = req_comp ? req_comp : *comp; + int row,col,z; + float temp; + + // @OPTIMIZE: use a bigger temp buffer and memcpy multiple pixels at once + for (row = 0; row < (h>>1); row++) { + for (col = 0; col < w; col++) { + for (z = 0; z < depth; z++) { + temp = result[(row * w + col) * depth + z]; + result[(row * w + col) * depth + z] = result[((h - row - 1) * w + col) * depth + z]; + result[((h - row - 1) * w + col) * depth + z] = temp; + } + } + } + } +} +#endif + #ifndef STBI_NO_STDIO -STBIDEF unsigned char *stbi_load(char const *filename, int *x, int *y, int *comp, int req_comp) + +static FILE *stbi__fopen(char const *filename, char const *mode) { - FILE *f = fopen(filename, "rb"); + FILE *f; +#if 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); @@ -581,12 +1047,12 @@ STBIDEF unsigned char *stbi_load(char const *filename, int *x, int *y, int *comp return result; } -STBIDEF unsigned char *stbi_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +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_main(&s,x,y,comp,req_comp); + result = stbi__load_flip(&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); @@ -595,75 +1061,78 @@ STBIDEF unsigned char *stbi_load_from_file(FILE *f, int *x, int *y, int *comp, i } #endif //!STBI_NO_STDIO -STBIDEF unsigned char *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +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_main(&s,x,y,comp,req_comp); + return stbi__load_flip(&s,x,y,comp,req_comp); } -unsigned char *stbi_load_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int 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_main(&s,x,y,comp,req_comp); + return stbi__load_flip(&s,x,y,comp,req_comp); } -#ifndef STBI_NO_HDR - -float *stbi_loadf_main(stbi__context *s, int *x, int *y, int *comp, int req_comp) +#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)) - return stbi__hdr_load(s,x,y,comp,req_comp); + if (stbi__hdr_test(s)) { + float *hdr_data = stbi__hdr_load(s,x,y,comp,req_comp); + if (hdr_data) + stbi__float_postprocess(hdr_data,x,y,comp,req_comp); + return hdr_data; + } #endif - data = stbi_load_main(s, x, y, comp, req_comp); + data = stbi__load_flip(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"); } -float *stbi_loadf_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +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); + return stbi__loadf_main(&s,x,y,comp,req_comp); } -float *stbi_loadf_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int 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); + return stbi__loadf_main(&s,x,y,comp,req_comp); } #ifndef STBI_NO_STDIO -float *stbi_loadf(char const *filename, int *x, int *y, int *comp, int req_comp) +STBIDEF float *stbi_loadf(char const *filename, int *x, int *y, int *comp, int req_comp) { - FILE *f = fopen(filename, "rb"); 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; } -float *stbi_loadf_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +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); + return stbi__loadf_main(&s,x,y,comp,req_comp); } #endif // !STBI_NO_STDIO -#endif // !STBI_NO_HDR +#endif // !STBI_NO_LINEAR -// these is-hdr-or-not is defined independent of whether STBI_NO_HDR is -// defined, for API simplicity; if STBI_NO_HDR is defined, it always +// 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! -int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len) +STBIDEF int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len) { #ifndef STBI_NO_HDR stbi__context s; @@ -679,7 +1148,7 @@ int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len) #ifndef STBI_NO_STDIO STBIDEF int stbi_is_hdr (char const *filename) { - FILE *f = fopen(filename, "rb"); + FILE *f = stbi__fopen(filename, "rb"); int result=0; if (f) { result = stbi_is_hdr_from_file(f); @@ -695,6 +1164,7 @@ STBIDEF int stbi_is_hdr_from_file(FILE *f) stbi__start_file(&s,f); return stbi__hdr_test(&s); #else + STBI_NOTUSED(f); return 0; #endif } @@ -707,21 +1177,24 @@ STBIDEF int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void 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_HDR -static float stbi__h2l_gamma_i=1.0f/2.2f, stbi__h2l_scale_i=1.0f; +#ifndef STBI_NO_LINEAR static float stbi__l2h_gamma=2.2f, stbi__l2h_scale=1.0f; -void stbi_hdr_to_ldr_gamma(float gamma) { stbi__h2l_gamma_i = 1/gamma; } -void stbi_hdr_to_ldr_scale(float scale) { stbi__h2l_scale_i = 1/scale; } - -void stbi_ldr_to_hdr_gamma(float gamma) { stbi__l2h_gamma = gamma; } -void stbi_ldr_to_hdr_scale(float scale) { stbi__l2h_scale = scale; } +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; } + ////////////////////////////////////////////////////////////////////////////// // @@ -730,9 +1203,9 @@ void stbi_ldr_to_hdr_scale(float scale) { stbi__l2h_scale = scale; } enum { - SCAN_load=0, - SCAN_type, - SCAN_header + STBI__SCAN_load=0, + STBI__SCAN_type, + STBI__SCAN_header }; static void stbi__refill_buffer(stbi__context *s) @@ -740,7 +1213,7 @@ static void stbi__refill_buffer(stbi__context *s) int n = (s->io.read)(s->io_user_data,(char*)s->buffer_start,s->buflen); 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, stbi__err.g. 0-byte file + // 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; @@ -771,11 +1244,15 @@ stbi_inline static int stbi__at_eof(stbi__context *s) if (s->read_from_callbacks == 0) return 1; } - return s->img_buffer >= s->img_buffer_end; + return s->img_buffer >= s->img_buffer_end; } static void stbi__skip(stbi__context *s, int n) { + 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) { @@ -795,7 +1272,7 @@ static int stbi__getn(stbi__context *s, stbi_uc *buffer, int 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; @@ -823,22 +1300,31 @@ static stbi__uint32 stbi__get32be(stbi__context *s) return (z << 16) + stbi__get16be(s); } +#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); return z + (stbi__get16le(s) << 16); } +#endif + +#define STBI__BYTECAST(x) ((stbi_uc) ((x) & 255)) // truncate int to byte without warnings + ////////////////////////////////////////////////////////////////////////////// // // generic converter from built-in img_n to req_comp -// individual types do this automatically as much as possible (stbi__err.g. jpeg +// 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 @@ -857,11 +1343,11 @@ static unsigned char *stbi__convert_format(unsigned char *data, int img_n, int r unsigned char *good; if (req_comp == img_n) return data; - assert(req_comp >= 1 && req_comp <= 4); + STBI_ASSERT(req_comp >= 1 && req_comp <= 4); - good = (unsigned char *) malloc(req_comp * x * y); + good = (unsigned char *) stbi__malloc(req_comp * x * y); if (good == NULL) { - free(data); + STBI_FREE(data); return stbi__errpuc("outofmem", "Out of memory"); } @@ -886,21 +1372,21 @@ static unsigned char *stbi__convert_format(unsigned char *data, int img_n, int r CASE(4,1) dest[0]=stbi__compute_y(src[0],src[1],src[2]); break; CASE(4,2) dest[0]=stbi__compute_y(src[0],src[1],src[2]), dest[1] = src[3]; break; CASE(4,3) dest[0]=src[0],dest[1]=src[1],dest[2]=src[2]; break; - default: assert(0); + default: STBI_ASSERT(0); } #undef CASE } - free(data); + STBI_FREE(data); return good; } -#ifndef STBI_NO_HDR +#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 = (float *) malloc(x * y * comp * sizeof(float)); - if (output == NULL) { free(data); return stbi__errpf("outofmem", "Out of memory"); } + float *output = (float *) stbi__malloc(x * y * comp * sizeof(float)); + 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) { @@ -909,16 +1395,18 @@ static float *stbi__ldr_to_hdr(stbi_uc *data, int x, int y, int comp) } if (k < comp) output[i*comp + k] = data[i*comp+k]/255.0f; } - free(data); + 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 = (stbi_uc *) malloc(x * y * comp); - if (output == NULL) { free(data); return stbi__errpuc("outofmem", "Out of memory"); } + stbi_uc *output = (stbi_uc *) stbi__malloc(x * y * comp); + 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) { @@ -935,17 +1423,16 @@ static stbi_uc *stbi__hdr_to_ldr(float *data, int x, int y, int comp) output[i*comp + k] = (stbi_uc) stbi__float2int(z); } } - free(data); + STBI_FREE(data); return output; } #endif ////////////////////////////////////////////////////////////////////////////// // -// "baseline" JPEG/JFIF decoder (not actually fully baseline implementation) +// "baseline" JPEG/JFIF decoder // // simple implementation -// - channel subsampling of at most 2 in each dimension // - 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 @@ -959,13 +1446,10 @@ static stbi_uc *stbi__hdr_to_ldr(float *data, int x, int y, int comp) // - 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 -// - load http://nothings.org/remote/anemones.jpg 3 times on 2.8Ghz P4 -// stb_jpeg: 1.34 seconds (MSVC6, default release build) -// stb_jpeg: 1.06 seconds (MSVC6, processor = Pentium Pro) -// IJL11.dll: 1.08 seconds (compiled by intel) -// IJG 1998: 0.98 seconds (MSVC6, makefile provided by IJG) -// IJG 1998: 0.95 seconds (MSVC6, makefile + proc=PPro) + +#ifndef STBI_NO_JPEG // huffman decoding acceleration #define FAST_BITS 9 // larger handles more cases; smaller stomps less cache @@ -983,13 +1467,11 @@ typedef struct typedef struct { - #ifdef STBI_SIMD - unsigned short dequant2[4][64]; - #endif stbi__context *s; stbi__huffman huff_dc[4]; stbi__huffman huff_ac[4]; stbi_uc dequant[4][64]; + stbi__int16 fast_ac[4][1 << FAST_BITS]; // sizes for components, interleaved MCUs int img_h_max, img_v_max; @@ -1007,17 +1489,31 @@ typedef struct int x,y,w2,h2; stbi_uc *data; - void *raw_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 + 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 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) @@ -1061,6 +1557,33 @@ static int stbi__build_huffman(stbi__huffman *h, int *count) 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 += (-1 << 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 << 8) + (run << 4) + (len + magbits)); + } + } + } +} + static void stbi__grow_buffer_unsafe(stbi__jpeg *j) { do { @@ -1081,7 +1604,7 @@ static void stbi__grow_buffer_unsafe(stbi__jpeg *j) // (1 << n) - 1 static stbi__uint32 stbi__bmask[17]={0,1,3,7,15,31,63,127,255,511,1023,2047,4095,8191,16383,32767,65535}; -// stbi__jpeg_huff_decode a jpeg huffman value from the bitstream +// 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; @@ -1123,7 +1646,7 @@ stbi_inline static int stbi__jpeg_huff_decode(stbi__jpeg *j, stbi__huffman *h) // convert the huffman code to the symbol id c = ((j->code_buffer >> (32 - k)) & stbi__bmask[k]) + h->delta[k]; - assert((((j->code_buffer) >> (32 - h->size[c])) & stbi__bmask[h->size[c]]) == h->code[c]); + 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; @@ -1131,31 +1654,46 @@ stbi_inline static int stbi__jpeg_huff_decode(stbi__jpeg *j, stbi__huffman *h) return h->values[c]; } +// bias[n] = (-1<code_bits < n) stbi__grow_buffer_unsafe(j); - #if 1 + sgn = (stbi__int32)j->code_buffer >> 31; // sign bit is always in MSB k = stbi_lrot(j->code_buffer, n); + STBI_ASSERT(n >= 0 && n < (int) (sizeof(stbi__bmask)/sizeof(*stbi__bmask))); j->code_buffer = k & ~stbi__bmask[n]; k &= stbi__bmask[n]; j->code_bits -= n; - #else - k = (j->code_buffer >> (32 - n)) & stbi__bmask[n]; + return k + (stbi__jbias[n] & ~sgn); +} + +// 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); + k = stbi_lrot(j->code_buffer, n); + j->code_buffer = k & ~stbi__bmask[n]; + k &= stbi__bmask[n]; j->code_bits -= n; - j->code_buffer <<= n; - #endif - // the following test is probably a random branch that won't - // predict well. I tried to table accelerate it but failed. - // maybe it's compiling as a conditional move? - if (k < m) - return (-1 << n) + k + 1; - else - return k; + 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); + 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, @@ -1175,11 +1713,14 @@ static stbi_uc stbi__jpeg_dezigzag[64+15] = 63, 63, 63, 63, 63, 63, 63 }; -// stbi__jpeg_huff_decode one 64-entry block-- -static int stbi__jpeg_decode_block(stbi__jpeg *j, short data[64], stbi__huffman *hdc, stbi__huffman *hac, int b) +// 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_uc *dequant) { int diff,dc,k; - int t = stbi__jpeg_huff_decode(j, hdc); + int t; + + if (j->code_bits < 16) stbi__grow_buffer_unsafe(j); + t = stbi__jpeg_huff_decode(j, hdc); if (t < 0) return stbi__err("bad huffman code","Corrupt JPEG"); // 0 all the ac values now so we can do it 32-bits at a time @@ -1188,28 +1729,189 @@ static int stbi__jpeg_decode_block(stbi__jpeg *j, short data[64], stbi__huffman diff = t ? stbi__extend_receive(j, t) : 0; dc = j->img_comp[b].dc_pred + diff; j->img_comp[b].dc_pred = dc; - data[0] = (short) dc; + data[0] = (short) (dc * dequant[0]); - // stbi__jpeg_huff_decode AC components, see JPEG spec + // decode AC components, see JPEG spec k = 1; do { - int r,s; - 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; + 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 + 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 { - k += r; - // stbi__jpeg_huff_decode into unzigzag'd location - data[stbi__jpeg_dezigzag[k++]] = (short) stbi__extend_receive(j,s); + 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); + diff = t ? stbi__extend_receive(j, t) : 0; + + dc = j->img_comp[b].dc_pred + diff; + j->img_comp[b].dc_pred = dc; + data[0] = (short) (dc << 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 + j->code_buffer <<= s; + j->code_bits -= s; + zig = stbi__jpeg_dezigzag[k++]; + data[zig] = (short) ((r >> 8) << 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) << 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) { @@ -1221,21 +1923,21 @@ stbi_inline static stbi_uc stbi__clamp(int x) return (stbi_uc) x; } -#define stbi__f2f(x) (int) (((x) * 4096 + 0.5)) +#define stbi__f2f(x) ((int) (((x) * 4096 + 0.5))) #define stbi__fsh(x) ((x) << 12) // derived from jidctint -- DCT_ISLOW -#define STBI__IDCT_1D(s0,s1,s2,s3,s4,s5,s6,s7) \ +#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); \ + 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); \ + t0 = stbi__fsh(p2+p3); \ + t1 = stbi__fsh(p2-p3); \ x0 = t0+t3; \ x3 = t0-t3; \ x1 = t1+t2; \ @@ -1248,36 +1950,28 @@ stbi_inline static stbi_uc stbi__clamp(int x) 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); \ + 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; -#ifdef STBI_SIMD -typedef unsigned short stbi_dequantize_t; -#else -typedef stbi_uc stbi_dequantize_t; -#endif - -// .344 seconds on 3*anemones.jpg -static void stbi__idct_block(stbi_uc *out, int out_stride, short data[64], stbi_dequantize_t *dequantize) +static void stbi__idct_block(stbi_uc *out, int out_stride, short data[64]) { int i,val[64],*v=val; - stbi_dequantize_t *dq = dequantize; stbi_uc *o; short *d = data; // columns - for (i=0; i < 8; ++i,++d,++dq, ++v) { + 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) { @@ -1285,11 +1979,10 @@ static void stbi__idct_block(stbi_uc *out, int out_stride, short data[64], stbi_ // (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] * dq[0] << 2; + int dcterm = d[0] << 2; v[0] = v[8] = v[16] = v[24] = v[32] = v[40] = v[48] = v[56] = dcterm; } else { - STBI__IDCT_1D(d[ 0]*dq[ 0],d[ 8]*dq[ 8],d[16]*dq[16],d[24]*dq[24], - d[32]*dq[32],d[40]*dq[40],d[48]*dq[48],d[56]*dq[56]) + 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; @@ -1330,14 +2023,394 @@ static void stbi__idct_block(stbi_uc *out, int out_stride, short data[64], stbi_ } } -#ifdef STBI_SIMD -static stbi_idct_8x8 stbi__idct_installed = stbi__idct_block; +#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)); \ + } -STBIDEF void stbi_install_idct(stbi_idct_8x8 func) -{ - stbi__idct_installed = func; + // 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 + +#endif // STBI_NEON #define STBI__MARKER_none 0xff // if there's a pending marker from the entropy stream, return that @@ -1368,6 +2441,7 @@ static void stbi__jpeg_reset(stbi__jpeg *j) j->img_comp[0].dc_pred = j->img_comp[1].dc_pred = j->img_comp[2].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 } @@ -1375,73 +2449,151 @@ static void stbi__jpeg_reset(stbi__jpeg *j) static int stbi__parse_entropy_coded_data(stbi__jpeg *z) { stbi__jpeg_reset(z); - if (z->scan_n == 1) { - int i,j; - #ifdef STBI_SIMD - __declspec(align(16)) - #endif - 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) { - if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+z->img_comp[n].ha, n)) return 0; - #ifdef STBI_SIMD - stbi__idct_installed(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data, z->dequant2[z->img_comp[n].tq]); - #else - stbi__idct_block(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data, z->dequant[z->img_comp[n].tq]); - #endif - // 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); + 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 { // interleaved! - int i,j,k,x,y; - 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; - if (!stbi__jpeg_decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+z->img_comp[n].ha, n)) return 0; - #ifdef STBI_SIMD - stbi__idct_installed(z->img_comp[n].data+z->img_comp[n].w2*y2+x2, z->img_comp[n].w2, data, z->dequant2[z->img_comp[n].tq]); - #else - stbi__idct_block(z->img_comp[n].data+z->img_comp[n].w2*y2+x2, z->img_comp[n].w2, data, z->dequant[z->img_comp[n].tq]); - #endif + } 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); + } } - // 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 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; + } + } +} + +static void stbi__jpeg_dequantize(short *data, stbi_uc *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); } } } } - return 1; } static int stbi__process_marker(stbi__jpeg *z, int m) @@ -1451,9 +2603,6 @@ static int stbi__process_marker(stbi__jpeg *z, int m) case STBI__MARKER_none: // no marker found return stbi__err("expected marker","Corrupt JPEG"); - case 0xC2: // stbi__SOF - progressive - return stbi__err("progressive jpeg","JPEG format not supported (progressive)"); - 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); @@ -1469,10 +2618,6 @@ static int stbi__process_marker(stbi__jpeg *z, int m) 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__get8(z->s); - #ifdef STBI_SIMD - for (i=0; i < 64; ++i) - z->dequant2[t][i] = z->dequant[t][i]; - #endif L -= 65; } return L==0; @@ -1500,6 +2645,8 @@ static int stbi__process_marker(stbi__jpeg *z, int m) } 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; @@ -1512,28 +2659,42 @@ static int stbi__process_marker(stbi__jpeg *z, int m) return 0; } -// after we see stbi__SOS +// 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 stbi__SOS component count","Corrupt JPEG"); - if (Ls != 6+2*z->scan_n) return stbi__err("bad stbi__SOS len","Corrupt JPEG"); + 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; + 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; } - if (stbi__get8(z->s) != 0) return stbi__err("bad stbi__SOS","Corrupt JPEG"); - stbi__get8(z->s); // should be 63, but might be 0 - if (stbi__get8(z->s) != 0) return stbi__err("bad stbi__SOS","Corrupt JPEG"); + + { + 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; } @@ -1542,8 +2703,8 @@ 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 stbi__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 + 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 c = stbi__get8(s); @@ -1554,7 +2715,7 @@ static int stbi__process_frame_header(stbi__jpeg *z, int scan) z->img_comp[i].linebuf = NULL; } - if (Lf != 8+3*s->img_n) return stbi__err("bad stbi__SOF len","Corrupt JPEG"); + if (Lf != 8+3*s->img_n) return stbi__err("bad SOF len","Corrupt JPEG"); for (i=0; i < s->img_n; ++i) { z->img_comp[i].id = stbi__get8(s); @@ -1567,9 +2728,9 @@ static int stbi__process_frame_header(stbi__jpeg *z, int scan) z->img_comp[i].tq = stbi__get8(s); if (z->img_comp[i].tq > 3) return stbi__err("bad TQ","Corrupt JPEG"); } - if (scan != SCAN_load) return 1; + if (scan != STBI__SCAN_load) return 1; - if ((1 << 30) / s->img_x / s->img_n < s->img_y) return stbi__err("too large", "Image too large to stbi__jpeg_huff_decode"); + if ((1 << 30) / s->img_x / s->img_n < s->img_y) 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; @@ -1585,64 +2746,82 @@ static int stbi__process_frame_header(stbi__jpeg *z, int scan) 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 (stbi__err.g. for non-interleaved MCU) + // 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 stbi__jpeg_huff_decode + // to simplify generation, we'll allocate enough memory to decode // the bogus oversized data from using interleaved MCUs and their - // big blocks (stbi__err.g. a 16x16 iMCU on an image of width 33); we won't + // big blocks (e.g. a 16x16 iMCU on an image of width 33); we won't // discard the extra data until colorspace conversion 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].raw_data = malloc(z->img_comp[i].w2 * z->img_comp[i].h2+15); + z->img_comp[i].raw_data = stbi__malloc(z->img_comp[i].w2 * z->img_comp[i].h2+15); + if (z->img_comp[i].raw_data == NULL) { for(--i; i >= 0; --i) { - free(z->img_comp[i].raw_data); - z->img_comp[i].data = NULL; + STBI_FREE(z->img_comp[i].raw_data); + z->img_comp[i].raw_data = NULL; } return stbi__err("outofmem", "Out of memory"); } - // align blocks for installable-idct using mmx/sse + // align blocks for idct using mmx/sse z->img_comp[i].data = (stbi_uc*) (((size_t) z->img_comp[i].raw_data + 15) & ~15); z->img_comp[i].linebuf = NULL; + if (z->progressive) { + z->img_comp[i].coeff_w = (z->img_comp[i].w2 + 7) >> 3; + z->img_comp[i].coeff_h = (z->img_comp[i].h2 + 7) >> 3; + z->img_comp[i].raw_coeff = STBI_MALLOC(z->img_comp[i].coeff_w * z->img_comp[i].coeff_h * 64 * sizeof(short) + 15); + z->img_comp[i].coeff = (short*) (((size_t) z->img_comp[i].raw_coeff + 15) & ~15); + } else { + z->img_comp[i].coeff = 0; + z->img_comp[i].raw_coeff = 0; + } } return 1; } -// use comparisons since in some cases we handle more than one case (stbi__err.g. stbi__SOF) +// 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) +#define stbi__SOF(x) ((x) == 0xc0 || (x) == 0xc1 || (x) == 0xc2) #define stbi__SOS(x) ((x) == 0xda) -static int decode_jpeg_header(stbi__jpeg *z, int scan) +#define stbi__SOF_progressive(x) ((x) == 0xc2) + +static int stbi__decode_jpeg_header(stbi__jpeg *z, int scan) { int m; z->marker = STBI__MARKER_none; // initialize cached marker to empty m = stbi__get_marker(z); - if (!stbi__SOI(m)) return stbi__err("no stbi__SOI","Corrupt JPEG"); - if (scan == SCAN_type) return 1; + 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 stbi__SOF", "Corrupt JPEG"); + 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 int decode_jpeg_image(stbi__jpeg *j) +// 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 (!decode_jpeg_header(j, SCAN_load)) return 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)) { @@ -1656,7 +2835,7 @@ static int decode_jpeg_image(stbi__jpeg *j) j->marker = stbi__get8(j->s); break; } else if (x != 0) { - return 0; + return stbi__err("junk before marker", "Corrupt JPEG"); } } // if we reach eof without hitting a marker, stbi__get_marker() below will fail and we'll eventually return 0 @@ -1666,6 +2845,8 @@ static int decode_jpeg_image(stbi__jpeg *j) } m = stbi__get_marker(j); } + if (j->progressive) + stbi__jpeg_finish(j); return 1; } @@ -1749,6 +2930,123 @@ static stbi_uc *stbi__resample_row_hv_2(stbi_uc *out, stbi_uc *in_near, stbi_uc 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 @@ -1760,10 +3058,10 @@ static stbi_uc *stbi__resample_row_generic(stbi_uc *out, stbi_uc *in_near, stbi_ return out; } +#ifdef STBI_JPEG_OLD +// this is the same YCbCr-to-RGB calculation that stb_image has used +// historically before the algorithm changes in 1.49 #define float2fixed(x) ((int) ((x) * 65536 + 0.5)) - -// 0.38 seconds on 3*anemones.jpg (0.25 with processor = Pro) -// VC6 without processor=Pro is generating multiple LEAs per multiply! 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; @@ -1788,28 +3086,214 @@ static void stbi__YCbCr_to_RGB_row(stbi_uc *out, const stbi_uc *y, const stbi_uc out += step; } } +#else +// 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 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* float2fixed(1.40200f); + g = y_fixed + (cr*-float2fixed(0.71414f)) + ((cb*-float2fixed(0.34414f)) & 0xffff0000); + b = y_fixed + cb* 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 + +#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_SIMD -static stbi_YCbCr_to_RGB_run stbi__YCbCr_installed = stbi__YCbCr_to_RGB_row; +#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 -STBIDEF void stbi_install_YCbCr_to_RGB(stbi_YCbCr_to_RGB_run func) -{ - stbi__YCbCr_installed = func; + 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* float2fixed(1.40200f); + g = y_fixed + cr*-float2fixed(0.71414f) + ((cb*-float2fixed(0.34414f)) & 0xffff0000); + b = y_fixed + cb* 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; + #ifndef STBI_JPEG_OLD + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd; + #endif + j->resample_row_hv_2_kernel = stbi__resample_row_hv_2_simd; + } +#endif + +#ifdef STBI_NEON + j->idct_block_kernel = stbi__idct_simd; + #ifndef STBI_JPEG_OLD + j->YCbCr_to_RGB_kernel = stbi__YCbCr_to_RGB_simd; + #endif + 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) { int i; for (i=0; i < j->s->img_n; ++i) { - if (j->img_comp[i].data) { - free(j->img_comp[i].raw_data); + if (j->img_comp[i].raw_data) { + STBI_FREE(j->img_comp[i].raw_data); + j->img_comp[i].raw_data = NULL; j->img_comp[i].data = NULL; } + if (j->img_comp[i].raw_coeff) { + STBI_FREE(j->img_comp[i].raw_coeff); + j->img_comp[i].raw_coeff = 0; + j->img_comp[i].coeff = 0; + } if (j->img_comp[i].linebuf) { - free(j->img_comp[i].linebuf); + STBI_FREE(j->img_comp[i].linebuf); j->img_comp[i].linebuf = NULL; } } @@ -1820,7 +3304,7 @@ 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 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; @@ -1828,12 +3312,13 @@ typedef struct static stbi_uc *load_jpeg_image(stbi__jpeg *z, int *out_x, int *out_y, int *comp, int req_comp) { int n, decode_n; + 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"); - z->s->img_n = 0; - // load a jpeg image from whichever source - if (!decode_jpeg_image(z)) { stbi__cleanup_jpeg(z); return NULL; } + // 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; @@ -1857,7 +3342,7 @@ static stbi_uc *load_jpeg_image(stbi__jpeg *z, int *out_x, int *out_y, int *comp // allocate line buffer big enough for upsampling off the edges // with upsample factor of 4 - z->img_comp[k].linebuf = (stbi_uc *) malloc(z->s->img_x + 3); + 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; @@ -1870,12 +3355,12 @@ static stbi_uc *load_jpeg_image(stbi__jpeg *z, int *out_x, int *out_y, int *comp 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 = stbi__resample_row_hv_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 *) malloc(n * z->s->img_x * z->s->img_y + 1); + output = (stbi_uc *) stbi__malloc(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 @@ -1898,11 +3383,7 @@ static stbi_uc *load_jpeg_image(stbi__jpeg *z, int *out_x, int *out_y, int *comp if (n >= 3) { stbi_uc *y = coutput[0]; if (z->s->img_n == 3) { - #ifdef STBI_SIMD - stbi__YCbCr_installed(out, y, coutput[1], coutput[2], z->s->img_x, n); - #else - stbi__YCbCr_to_RGB_row(out, y, coutput[1], coutput[2], z->s->img_x, n); - #endif + 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]; @@ -1929,6 +3410,7 @@ static unsigned char *stbi__jpeg_load(stbi__context *s, int *x, int *y, int *com { stbi__jpeg j; j.s = s; + stbi__setup_jpeg(&j); return load_jpeg_image(&j, x,y,comp,req_comp); } @@ -1937,14 +3419,15 @@ static int stbi__jpeg_test(stbi__context *s) int r; stbi__jpeg j; j.s = s; - r = decode_jpeg_header(&j, SCAN_type); + stbi__setup_jpeg(&j); + r = stbi__decode_jpeg_header(&j, STBI__SCAN_type); stbi__rewind(s); return r; } static int stbi__jpeg_info_raw(stbi__jpeg *j, int *x, int *y, int *comp) { - if (!decode_jpeg_header(j, SCAN_header)) { + if (!stbi__decode_jpeg_header(j, STBI__SCAN_header)) { stbi__rewind( j->s ); return 0; } @@ -1960,14 +3443,17 @@ static int stbi__jpeg_info(stbi__context *s, int *x, int *y, int *comp) j.s = s; return stbi__jpeg_info_raw(&j, x, y, comp); } +#endif -// public domain zlib stbi__jpeg_huff_decode v0.2 Sean Barrett 2006-11-18 +// 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) @@ -1981,7 +3467,7 @@ typedef struct int maxcode[17]; stbi__uint16 firstsymbol[16]; stbi_uc size[288]; - stbi__uint16 value[288]; + stbi__uint16 value[288]; } stbi__zhuffman; stbi_inline static int stbi__bitreverse16(int n) @@ -1995,9 +3481,9 @@ stbi_inline static int stbi__bitreverse16(int n) stbi_inline static int stbi__bit_reverse(int v, int bits) { - assert(bits <= 16); + STBI_ASSERT(bits <= 16); // to bit reverse n bits, reverse 16 and shift - // stbi__err.g. 11 bits, bit reverse and shift away 5 + // e.g. 11 bits, bit reverse and shift away 5 return stbi__bitreverse16(v) >> (16-bits); } @@ -2008,12 +3494,13 @@ static int stbi__zbuild_huffman(stbi__zhuffman *z, stbi_uc *sizelist, int num) // DEFLATE spec for generating codes memset(sizes, 0, sizeof(sizes)); - memset(z->fast, 255, sizeof(z->fast)); - for (i=0; i < num; ++i) + 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) - assert(sizes[i] <= (1 << 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; @@ -2021,7 +3508,7 @@ static int stbi__zbuild_huffman(stbi__zhuffman *z, stbi_uc *sizelist, int num) z->firstsymbol[i] = (stbi__uint16) k; code = (code + sizes[i]); if (sizes[i]) - if (code-1 >= (1 << i)) return stbi__err("bad codelengths","Corrupt JPEG"); + 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]; @@ -2031,13 +3518,14 @@ static int stbi__zbuild_huffman(stbi__zhuffman *z, stbi_uc *sizelist, int num) 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 k = stbi__bit_reverse(next_code[s],s); - while (k < (1 << STBI__ZFAST_BITS)) { - z->fast[k] = (stbi__uint16) c; - k += (1 << s); + 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]; @@ -2075,8 +3563,8 @@ stbi_inline static stbi_uc stbi__zget8(stbi__zbuf *z) static void stbi__fill_bits(stbi__zbuf *z) { do { - assert(z->code_buffer < (1U << z->num_bits)); - z->code_buffer |= stbi__zget8(z) << z->num_bits; + STBI_ASSERT(z->code_buffer < (1U << z->num_bits)); + z->code_buffer |= (unsigned int) stbi__zget8(z) << z->num_bits; z->num_bits += 8; } while (z->num_bits <= 24); } @@ -2088,21 +3576,12 @@ stbi_inline static unsigned int stbi__zreceive(stbi__zbuf *z, int n) k = z->code_buffer & ((1 << n) - 1); z->code_buffer >>= n; z->num_bits -= n; - return k; + return k; } -stbi_inline static int stbi__zhuffman_decode(stbi__zbuf *a, stbi__zhuffman *z) +static int stbi__zhuffman_decode_slowpath(stbi__zbuf *a, stbi__zhuffman *z) { int b,s,k; - if (a->num_bits < 16) stbi__fill_bits(a); - b = z->fast[a->code_buffer & STBI__ZFAST_MASK]; - if (b < 0xffff) { - s = z->size[b]; - a->code_buffer >>= s; - a->num_bits -= s; - return z->value[b]; - } - // 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); @@ -2112,22 +3591,38 @@ stbi_inline static int stbi__zhuffman_decode(stbi__zbuf *a, stbi__zhuffman *z) if (s == 16) return -1; // invalid code! // code size is s, so: b = (k >> (16-s)) - z->firstcode[s] + z->firstsymbol[s]; - assert(z->size[b] == s); + STBI_ASSERT(z->size[b] == s); a->code_buffer >>= s; a->num_bits -= s; return z->value[b]; } -static int stbi__zexpand(stbi__zbuf *z, int n) // need to make room for n bytes +stbi_inline static int stbi__zhuffman_decode(stbi__zbuf *a, stbi__zhuffman *z) +{ + int b,s; + if (a->num_bits < 16) 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; - int cur, limit; + int cur, limit, old_limit; + z->zout = zout; if (!z->z_expandable) return stbi__err("output buffer limit","Corrupt PNG"); cur = (int) (z->zout - z->zout_start); - limit = (int) (z->zout_end - z->zout_start); + limit = old_limit = (int) (z->zout_end - z->zout_start); while (cur + n > limit) limit *= 2; - q = (char *) realloc(z->zout_start, limit); + 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; @@ -2140,7 +3635,7 @@ static int stbi__zlength_base[31] = { 15,17,19,23,27,31,35,43,51,59, 67,83,99,115,131,163,195,227,258,0,0 }; -static int stbi__zlength_extra[31]= +static 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 int stbi__zdist_base[32] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193, @@ -2151,16 +3646,23 @@ static int stbi__zdist_extra[32] = 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 (a->zout >= a->zout_end) if (!stbi__zexpand(a, 1)) return 0; - *a->zout++ = (char) z; + 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) return 1; + if (z == 256) { + a->zout = zout; + return 1; + } z -= 257; len = stbi__zlength_base[z]; if (stbi__zlength_extra[z]) len += stbi__zreceive(a, stbi__zlength_extra[z]); @@ -2168,11 +3670,18 @@ static int stbi__parse_huffman_block(stbi__zbuf *a) if (z < 0) return stbi__err("bad huffman code","Corrupt PNG"); dist = stbi__zdist_base[z]; if (stbi__zdist_extra[z]) dist += stbi__zreceive(a, stbi__zdist_extra[z]); - if (a->zout - a->zout_start < dist) return stbi__err("bad dist","Corrupt PNG"); - if (a->zout + len > a->zout_end) if (!stbi__zexpand(a, len)) return 0; - p = (stbi_uc *) (a->zout - dist); - while (len--) - *a->zout++ = *p++; + if (zout - a->zout_start < dist) return stbi__err("bad dist","Corrupt PNG"); + if (zout + len > a->zout_end) { + 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); } + } } } } @@ -2199,7 +3708,7 @@ static int stbi__compute_huffman_codes(stbi__zbuf *a) n = 0; while (n < hlit + hdist) { int c = stbi__zhuffman_decode(a, &z_codelength); - assert(c >= 0 && c < 19); + if (c < 0 || c >= 19) return stbi__err("bad codelengths", "Corrupt PNG"); if (c < 16) lencodes[n++] = (stbi_uc) c; else if (c == 16) { @@ -2211,7 +3720,7 @@ static int stbi__compute_huffman_codes(stbi__zbuf *a) memset(lencodes+n, 0, c); n += c; } else { - assert(c == 18); + STBI_ASSERT(c == 18); c = stbi__zreceive(a,7)+11; memset(lencodes+n, 0, c); n += c; @@ -2236,7 +3745,7 @@ static int stbi__parse_uncomperssed_block(stbi__zbuf *a) a->code_buffer >>= 8; a->num_bits -= 8; } - assert(a->num_bits == 0); + STBI_ASSERT(a->num_bits == 0); // now fill header the normal way while (k < 4) header[k++] = stbi__zget8(a); @@ -2245,7 +3754,7 @@ static int stbi__parse_uncomperssed_block(stbi__zbuf *a) 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, len)) return 0; + if (!stbi__zexpand(a, a->zout, len)) return 0; memcpy(a->zout, a->zbuffer, len); a->zbuffer += len; a->zout += len; @@ -2320,7 +3829,7 @@ static int stbi__do_zlib(stbi__zbuf *a, char *obuf, int olen, int exp, int parse STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen) { stbi__zbuf a; - char *p = (char *) malloc(initial_size); + char *p = (char *) stbi__malloc(initial_size); if (p == NULL) return NULL; a.zbuffer = (stbi_uc *) buffer; a.zbuffer_end = (stbi_uc *) buffer + len; @@ -2328,7 +3837,7 @@ STBIDEF char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int if (outlen) *outlen = (int) (a.zout - a.zout_start); return a.zout_start; } else { - free(a.zout_start); + STBI_FREE(a.zout_start); return NULL; } } @@ -2341,7 +3850,7 @@ STBIDEF char *stbi_zlib_decode_malloc(char const *buffer, int len, int *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 *) malloc(initial_size); + char *p = (char *) stbi__malloc(initial_size); if (p == NULL) return NULL; a.zbuffer = (stbi_uc *) buffer; a.zbuffer_end = (stbi_uc *) buffer + len; @@ -2349,7 +3858,7 @@ STBIDEF char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, i if (outlen) *outlen = (int) (a.zout - a.zout_start); return a.zout_start; } else { - free(a.zout_start); + STBI_FREE(a.zout_start); return NULL; } } @@ -2368,7 +3877,7 @@ STBIDEF int stbi_zlib_decode_buffer(char *obuffer, int olen, char const *ibuffer STBIDEF char *stbi_zlib_decode_noheader_malloc(char const *buffer, int len, int *outlen) { stbi__zbuf a; - char *p = (char *) malloc(16384); + char *p = (char *) stbi__malloc(16384); if (p == NULL) return NULL; a.zbuffer = (stbi_uc *) buffer; a.zbuffer_end = (stbi_uc *) buffer+len; @@ -2376,7 +3885,7 @@ STBIDEF char *stbi_zlib_decode_noheader_malloc(char const *buffer, int len, int if (outlen) *outlen = (int) (a.zout - a.zout_start); return a.zout_start; } else { - free(a.zout_start); + STBI_FREE(a.zout_start); return NULL; } } @@ -2391,6 +3900,7 @@ STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char else return -1; } +#endif // public domain "baseline" PNG decoder v0.10 Sean Barrett 2006-11-18 // simple implementation @@ -2402,15 +3912,13 @@ STBIDEF int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char // 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; -#define PNG_TYPE(a,b,c,d) (((a) << 24) + ((b) << 16) + ((c) << 8) + (d)) - static stbi__pngchunk stbi__get_chunk_header(stbi__context *s) { stbi__pngchunk c; @@ -2436,13 +3944,23 @@ typedef struct enum { - STBI__F_none=0, STBI__F_sub=1, STBI__F_up=2, STBI__F_avg=3, STBI__F_paeth=4, - STBI__F_avg_first, STBI__F_paeth_first + STBI__F_none=0, + STBI__F_sub=1, + STBI__F_up=2, + STBI__F_avg=3, + STBI__F_paeth=4, + // synthetic filters used for first scanline to avoid needing a dummy row of 0s + STBI__F_avg_first, + STBI__F_paeth_first }; static stbi_uc first_row_filter[5] = { - STBI__F_none, STBI__F_sub, STBI__F_none, STBI__F_avg_first, STBI__F_paeth_first + STBI__F_none, + STBI__F_sub, + STBI__F_none, + STBI__F_avg_first, + STBI__F_paeth_first }; static int stbi__paeth(int a, int b, int c) @@ -2456,32 +3974,50 @@ static int stbi__paeth(int a, int b, int c) return c; } -#define STBI__BYTECAST(x) ((stbi_uc) ((x) & 255)) // truncate int to byte without warnings +static stbi_uc stbi__depth_scale_table[9] = { 0, 0xff, 0x55, 0, 0x11, 0,0,0, 0x01 }; // 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) +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) { stbi__context *s = a->s; stbi__uint32 i,j,stride = x*out_n; + stbi__uint32 img_len, img_width_bytes; int k; int img_n = s->img_n; // copy it into a local for later - assert(out_n == s->img_n || out_n == s->img_n+1); - a->out = (stbi_uc *) malloc(x * y * out_n); + + STBI_ASSERT(out_n == s->img_n || out_n == s->img_n+1); + a->out = (stbi_uc *) stbi__malloc(x * y * out_n); // extra bytes to write off the end into if (!a->out) return stbi__err("outofmem", "Out of memory"); + + img_width_bytes = (((img_n * x * depth) + 7) >> 3); + img_len = (img_width_bytes + 1) * y; if (s->img_x == x && s->img_y == y) { - if (raw_len != (img_n * x + 1) * y) return stbi__err("not enough pixels","Corrupt PNG"); + if (raw_len != img_len) return stbi__err("not enough pixels","Corrupt PNG"); } else { // interlaced: - if (raw_len < (img_n * x + 1) * y) return stbi__err("not enough pixels","Corrupt PNG"); + if (raw_len < img_len) return stbi__err("not enough pixels","Corrupt PNG"); } + for (j=0; j < y; ++j) { stbi_uc *cur = a->out + stride*j; stbi_uc *prior = cur - stride; int filter = *raw++; - if (filter > 4) return stbi__err("invalid filter","Corrupt PNG"); + int filter_bytes = img_n; + int width = x; + if (filter > 4) + return stbi__err("invalid filter","Corrupt PNG"); + + if (depth < 8) { + STBI_ASSERT(img_width_bytes <= x); + cur += x*out_n - img_width_bytes; // store output to the rightmost img_len bytes, so we can decode in place + filter_bytes = 1; + width = img_width_bytes; + } + // if first row, use special filter that doesn't sample previous row if (j == 0) filter = first_row_filter[filter]; - // handle first pixel explicitly - for (k=0; k < img_n; ++k) { + + // handle first byte explicitly + for (k=0; k < filter_bytes; ++k) { switch (filter) { case STBI__F_none : cur[k] = raw[k]; break; case STBI__F_sub : cur[k] = raw[k]; break; @@ -2492,28 +4028,39 @@ static int stbi__create_png_image_raw(stbi__png *a, stbi_uc *raw, stbi__uint32 r case STBI__F_paeth_first: cur[k] = raw[k]; break; } } - if (img_n != out_n) cur[img_n] = 255; - raw += img_n; - cur += out_n; - prior += out_n; + + if (depth == 8) { + if (img_n != out_n) + cur[img_n] = 255; // first pixel + raw += img_n; + cur += out_n; + prior += out_n; + } else { + raw += 1; + cur += 1; + prior += 1; + } + // this is a little gross, so that we don't switch per-pixel or per-component - if (img_n == out_n) { + if (depth < 8 || img_n == out_n) { + int nk = (width - 1)*img_n; #define CASE(f) \ case f: \ - for (i=x-1; i >= 1; --i, raw+=img_n,cur+=img_n,prior+=img_n) \ - for (k=0; k < img_n; ++k) + for (k=0; k < nk; ++k) switch (filter) { - CASE(STBI__F_none) cur[k] = raw[k]; break; - CASE(STBI__F_sub) cur[k] = STBI__BYTECAST(raw[k] + cur[k-img_n]); break; + // "none" filter turns into a memcpy here; make that explicit. + case STBI__F_none: memcpy(cur, raw, nk); break; + CASE(STBI__F_sub) cur[k] = STBI__BYTECAST(raw[k] + cur[k-filter_bytes]); break; CASE(STBI__F_up) cur[k] = STBI__BYTECAST(raw[k] + prior[k]); break; - CASE(STBI__F_avg) cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k-img_n])>>1)); break; - CASE(STBI__F_paeth) cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-img_n],prior[k],prior[k-img_n])); break; - CASE(STBI__F_avg_first) cur[k] = STBI__BYTECAST(raw[k] + (cur[k-img_n] >> 1)); break; - CASE(STBI__F_paeth_first) cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-img_n],0,0)); break; + CASE(STBI__F_avg) cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k-filter_bytes])>>1)); break; + CASE(STBI__F_paeth) cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],prior[k],prior[k-filter_bytes])); break; + CASE(STBI__F_avg_first) cur[k] = STBI__BYTECAST(raw[k] + (cur[k-filter_bytes] >> 1)); break; + CASE(STBI__F_paeth_first) cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],0,0)); break; } #undef CASE + raw += nk; } else { - assert(img_n+1 == out_n); + STBI_ASSERT(img_n+1 == out_n); #define CASE(f) \ case f: \ for (i=x-1; i >= 1; --i, cur[img_n]=255,raw+=img_n,cur+=out_n,prior+=out_n) \ @@ -2530,18 +4077,93 @@ static int stbi__create_png_image_raw(stbi__png *a, stbi_uc *raw, stbi__uint32 r #undef CASE } } + + // we make a separate pass to expand bits to pixels; for performance, + // this could run two scanlines behind the above code, so it won't + // intefere with filtering but will still be in the cache. + if (depth < 8) { + for (j=0; j < y; ++j) { + stbi_uc *cur = a->out + stride*j; + stbi_uc *in = a->out + stride*j + x*out_n - img_width_bytes; + // unpack 1/2/4-bit into a 8-bit buffer. allows us to keep the common 8-bit path optimal at minimal cost for 1/2/4-bit + // png guarante byte alignment, if width is not multiple of 8/4/2 we'll decode dummy trailing data that will be skipped in the later loop + stbi_uc scale = (color == 0) ? stbi__depth_scale_table[depth] : 1; // scale grayscale values to 0..255 range + + // note that the final byte might overshoot and write more data than desired. + // we can allocate enough data that this never writes out of memory, but it + // could also overwrite the next scanline. can it overwrite non-empty data + // on the next scanline? yes, consider 1-pixel-wide scanlines with 1-bit-per-pixel. + // so we need to explicitly clamp the final ones + + if (depth == 4) { + for (k=x*img_n; k >= 2; k-=2, ++in) { + *cur++ = scale * ((*in >> 4) ); + *cur++ = scale * ((*in ) & 0x0f); + } + if (k > 0) *cur++ = scale * ((*in >> 4) ); + } else if (depth == 2) { + for (k=x*img_n; k >= 4; k-=4, ++in) { + *cur++ = scale * ((*in >> 6) ); + *cur++ = scale * ((*in >> 4) & 0x03); + *cur++ = scale * ((*in >> 2) & 0x03); + *cur++ = scale * ((*in ) & 0x03); + } + if (k > 0) *cur++ = scale * ((*in >> 6) ); + if (k > 1) *cur++ = scale * ((*in >> 4) & 0x03); + if (k > 2) *cur++ = scale * ((*in >> 2) & 0x03); + } else if (depth == 1) { + for (k=x*img_n; k >= 8; k-=8, ++in) { + *cur++ = scale * ((*in >> 7) ); + *cur++ = scale * ((*in >> 6) & 0x01); + *cur++ = scale * ((*in >> 5) & 0x01); + *cur++ = scale * ((*in >> 4) & 0x01); + *cur++ = scale * ((*in >> 3) & 0x01); + *cur++ = scale * ((*in >> 2) & 0x01); + *cur++ = scale * ((*in >> 1) & 0x01); + *cur++ = scale * ((*in ) & 0x01); + } + if (k > 0) *cur++ = scale * ((*in >> 7) ); + if (k > 1) *cur++ = scale * ((*in >> 6) & 0x01); + if (k > 2) *cur++ = scale * ((*in >> 5) & 0x01); + if (k > 3) *cur++ = scale * ((*in >> 4) & 0x01); + if (k > 4) *cur++ = scale * ((*in >> 3) & 0x01); + if (k > 5) *cur++ = scale * ((*in >> 2) & 0x01); + if (k > 6) *cur++ = scale * ((*in >> 1) & 0x01); + } + if (img_n != out_n) { + int q; + // insert alpha = 255 + cur = a->out + stride*j; + if (img_n == 1) { + for (q=x-1; q >= 0; --q) { + cur[q*2+1] = 255; + cur[q*2+0] = cur[q]; + } + } else { + STBI_ASSERT(img_n == 3); + for (q=x-1; q >= 0; --q) { + cur[q*4+3] = 255; + cur[q*4+2] = cur[q*3+2]; + cur[q*4+1] = cur[q*3+1]; + cur[q*4+0] = cur[q*3+0]; + } + } + } + } + } + return 1; } -static int stbi__create_png_image(stbi__png *a, stbi_uc *raw, stbi__uint32 raw_len, int out_n, int interlaced) +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) { stbi_uc *final; int p; if (!interlaced) - return stbi__create_png_image_raw(a, raw, raw_len, out_n, a->s->img_x, a->s->img_y); + 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 *) malloc(a->s->img_x * a->s->img_y * out_n); + final = (stbi_uc *) stbi__malloc(a->s->img_x * a->s->img_y * out_n); for (p=0; p < 7; ++p) { int xorig[] = { 0,4,0,2,0,1,0 }; int yorig[] = { 0,0,4,0,2,0,1 }; @@ -2552,17 +4174,22 @@ static int stbi__create_png_image(stbi__png *a, stbi_uc *raw, stbi__uint32 raw_l 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) { - if (!stbi__create_png_image_raw(a, raw, raw_len, out_n, x, y)) { - free(final); + 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) - memcpy(final + (j*yspc[p]+yorig[p])*a->s->img_x*out_n + (i*xspc[p]+xorig[p])*out_n, + 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_n + out_x*out_n, a->out + (j*x+i)*out_n, out_n); - free(a->out); - raw += (x*out_n+1)*y; - raw_len -= (x*out_n+1)*y; + } + } + STBI_FREE(a->out); + image_data += img_len; + image_data_len -= img_len; } } a->out = final; @@ -2578,7 +4205,7 @@ static int stbi__compute_transparency(stbi__png *z, stbi_uc tc[3], int out_n) // compute color-based transparency, assuming we've // already got 255 as the alpha value in the output - assert(out_n == 2 || out_n == 4); + STBI_ASSERT(out_n == 2 || out_n == 4); if (out_n == 2) { for (i=0; i < pixel_count; ++i) { @@ -2600,7 +4227,7 @@ static int stbi__expand_png_palette(stbi__png *a, stbi_uc *palette, int len, int stbi__uint32 i, pixel_count = a->s->img_x * a->s->img_y; stbi_uc *p, *temp_out, *orig = a->out; - p = (stbi_uc *) malloc(pixel_count * pal_img_n); + p = (stbi_uc *) stbi__malloc(pixel_count * pal_img_n); if (p == NULL) return stbi__err("outofmem", "Out of memory"); // between here and free(out) below, exitting would leak @@ -2624,7 +4251,7 @@ static int stbi__expand_png_palette(stbi__png *a, stbi_uc *palette, int len, int p += 4; } } - free(a->out); + STBI_FREE(a->out); a->out = temp_out; STBI_NOTUSED(len); @@ -2659,7 +4286,7 @@ static void stbi__de_iphone(stbi__png *z) p += 3; } } else { - assert(s->img_out_n == 4); + 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) { @@ -2672,7 +4299,7 @@ static void stbi__de_iphone(stbi__png *z) } else { p[0] = p[2]; p[2] = t; - } + } p += 4; } } else { @@ -2687,12 +4314,14 @@ static void stbi__de_iphone(stbi__png *z) } } +#define STBI__PNG_TYPE(a,b,c,d) (((a) << 24) + ((b) << 16) + ((c) << 8) + (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]; stbi__uint32 ioff=0, idata_limit=0, i, pal_len=0; - int first=1,k,interlace=0, is_iphone=0; + int first=1,k,interlace=0, color=0, depth=0, is_iphone=0; stbi__context *s = z->s; z->expanded = NULL; @@ -2701,23 +4330,23 @@ static int stbi__parse_png_file(stbi__png *z, int scan, int req_comp) if (!stbi__check_png_header(s)) return 0; - if (scan == SCAN_type) return 1; + if (scan == STBI__SCAN_type) return 1; for (;;) { stbi__pngchunk c = stbi__get_chunk_header(s); switch (c.type) { - case PNG_TYPE('C','g','B','I'): + case STBI__PNG_TYPE('C','g','B','I'): is_iphone = 1; stbi__skip(s, c.length); break; - case PNG_TYPE('I','H','D','R'): { - int depth,color,comp,filter; + 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); if (s->img_x > (1 << 24)) return stbi__err("too large","Very large image (corrupt?)"); s->img_y = stbi__get32be(s); if (s->img_y > (1 << 24)) return stbi__err("too large","Very large image (corrupt?)"); - depth = stbi__get8(s); if (depth != 8) return stbi__err("8bit only","PNG not supported: 8-bit only"); + depth = stbi__get8(s); if (depth != 1 && depth != 2 && depth != 4 && depth != 8) return stbi__err("1/2/4/8-bit only","PNG not supported: 1/2/4/8-bit only"); color = stbi__get8(s); if (color > 6) 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"); @@ -2726,8 +4355,8 @@ static int stbi__parse_png_file(stbi__png *z, int scan, int req_comp) 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 stbi__jpeg_huff_decode"); - if (scan == SCAN_header) return 1; + if ((1 << 30) / s->img_x / s->img_n < s->img_y) return stbi__err("too large", "Image too large to decode"); + if (scan == STBI__SCAN_header) return 1; } else { // if paletted, then pal_n is our final components, and // img_n is # components to decompress/filter. @@ -2738,7 +4367,7 @@ static int stbi__parse_png_file(stbi__png *z, int scan, int req_comp) break; } - case PNG_TYPE('P','L','T','E'): { + 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; @@ -2752,11 +4381,11 @@ static int stbi__parse_png_file(stbi__png *z, int scan, int req_comp) break; } - case PNG_TYPE('t','R','N','S'): { + 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 == SCAN_header) { s->img_n = 4; return 1; } + 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; @@ -2767,21 +4396,24 @@ static int stbi__parse_png_file(stbi__png *z, int scan, int req_comp) if (c.length != (stbi__uint32) s->img_n*2) return stbi__err("bad tRNS len","Corrupt PNG"); has_trans = 1; for (k=0; k < s->img_n; ++k) - tc[k] = (stbi_uc) (stbi__get16be(s) & 255); // non 8-bit images will be larger + tc[k] = (stbi_uc) (stbi__get16be(s) & 255) * stbi__depth_scale_table[depth]; // non 8-bit images will be larger } break; } - case PNG_TYPE('I','D','A','T'): { + 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 == SCAN_header) { s->img_n = pal_img_n; return 1; } + if (scan == STBI__SCAN_header) { s->img_n = pal_img_n; return 1; } + 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; - p = (stbi_uc *) realloc(z->idata, idata_limit); if (p == NULL) return stbi__err("outofmem", "Out of memory"); + 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"); @@ -2789,19 +4421,22 @@ static int stbi__parse_png_file(stbi__png *z, int scan, int req_comp) break; } - case PNG_TYPE('I','E','N','D'): { - stbi__uint32 raw_len; + 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 != SCAN_load) return 1; + if (scan != STBI__SCAN_load) return 1; if (z->idata == NULL) return stbi__err("no IDAT","Corrupt PNG"); - z->expanded = (stbi_uc *) stbi_zlib_decode_malloc_guesssize_headerflag((char *) z->idata, ioff, 16384, (int *) &raw_len, !is_iphone); + // initial guess for decoded data size to avoid unnecessary reallocs + bpl = (s->img_x * 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 - free(z->idata); z->idata = NULL; + 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, interlace)) return 0; + if (!stbi__create_png_image(z, z->expanded, raw_len, s->img_out_n, depth, color, interlace)) return 0; if (has_trans) if (!stbi__compute_transparency(z, tc, s->img_out_n)) return 0; if (is_iphone && stbi__de_iphone_flag && s->img_out_n > 2) @@ -2814,7 +4449,7 @@ static int stbi__parse_png_file(stbi__png *z, int scan, int req_comp) if (!stbi__expand_png_palette(z, palette, pal_len, s->img_out_n)) return 0; } - free(z->expanded); z->expanded = NULL; + STBI_FREE(z->expanded); z->expanded = NULL; return 1; } @@ -2844,7 +4479,7 @@ static unsigned char *stbi__do_png(stbi__png *p, int *x, int *y, int *n, int req { unsigned char *result=NULL; if (req_comp < 0 || req_comp > 4) return stbi__errpuc("bad req_comp", "Internal error"); - if (stbi__parse_png_file(p, SCAN_load, req_comp)) { + if (stbi__parse_png_file(p, STBI__SCAN_load, req_comp)) { result = p->out; p->out = NULL; if (req_comp && req_comp != p->s->img_out_n) { @@ -2854,11 +4489,11 @@ static unsigned char *stbi__do_png(stbi__png *p, int *x, int *y, int *n, int req } *x = p->s->img_x; *y = p->s->img_y; - if (n) *n = p->s->img_n; + if (n) *n = p->s->img_out_n; } - free(p->out); p->out = NULL; - free(p->expanded); p->expanded = NULL; - free(p->idata); p->idata = NULL; + STBI_FREE(p->out); p->out = NULL; + STBI_FREE(p->expanded); p->expanded = NULL; + STBI_FREE(p->idata); p->idata = NULL; return result; } @@ -2880,7 +4515,7 @@ static int stbi__png_test(stbi__context *s) static int stbi__png_info_raw(stbi__png *p, int *x, int *y, int *comp) { - if (!stbi__parse_png_file(p, SCAN_header, 0)) { + if (!stbi__parse_png_file(p, STBI__SCAN_header, 0)) { stbi__rewind( p->s ); return 0; } @@ -2896,8 +4531,11 @@ static int stbi__png_info(stbi__context *s, int *x, int *y, int *comp) p.s = s; return stbi__png_info_raw(&p, x, y, comp); } +#endif // Microsoft/Windows BMP image + +#ifndef STBI_NO_BMP static int stbi__bmp_test_raw(stbi__context *s) { int r; @@ -2961,19 +4599,22 @@ static int stbi__shiftsigned(int v, int shift, int bits) return result; } -static stbi_uc *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp) +typedef struct { - stbi_uc *out; - unsigned int mr=0,mg=0,mb=0,ma=0, fake_a=0; - stbi_uc pal[256][4]; - int psize=0,i,j,compress=0,width; - int bpp, flip_vertically, pad, target, offset, hsz; + int bpp, offset, hsz; + unsigned int mr,mg,mb,ma, all_a; +} stbi__bmp_data; + +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 - offset = stbi__get32le(s); - hsz = stbi__get32le(s); + info->offset = stbi__get32le(s); + info->hsz = hsz = stbi__get32le(s); + 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); @@ -2983,15 +4624,10 @@ static stbi_uc *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int s->img_y = stbi__get32le(s); } if (stbi__get16le(s) != 1) return stbi__errpuc("bad BMP", "bad BMP"); - bpp = stbi__get16le(s); - if (bpp == 1) return stbi__errpuc("monochrome", "BMP type not supported: 1-bit"); - flip_vertically = ((int) s->img_y) > 0; - s->img_y = abs((int) s->img_y); - if (hsz == 12) { - if (bpp < 24) - psize = (offset - 14 - 24) / 3; - } else { - compress = stbi__get32le(s); + info->bpp = stbi__get16le(s); + if (info->bpp == 1) return stbi__errpuc("monochrome", "BMP type not supported: 1-bit"); + if (hsz != 12) { + int compress = stbi__get32le(s); if (compress == 1 || compress == 2) return stbi__errpuc("BMP RLE", "BMP type not supported: RLE"); stbi__get32le(s); // discard sizeof stbi__get32le(s); // discard hres @@ -3005,27 +4641,26 @@ static stbi_uc *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int stbi__get32le(s); stbi__get32le(s); } - if (bpp == 16 || bpp == 32) { - mr = mg = mb = 0; + if (info->bpp == 16 || info->bpp == 32) { + info->mr = info->mg = info->mb = 0; if (compress == 0) { - if (bpp == 32) { - mr = 0xffu << 16; - mg = 0xffu << 8; - mb = 0xffu << 0; - ma = 0xffu << 24; - fake_a = 1; // @TODO: check for cases like alpha value is all 0 and switch it to 255 - STBI_NOTUSED(fake_a); + 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 { - mr = 31u << 10; - mg = 31u << 5; - mb = 31u << 0; + info->mr = 31u << 10; + info->mg = 31u << 5; + info->mb = 31u << 0; } } else if (compress == 3) { - mr = stbi__get32le(s); - mg = stbi__get32le(s); - mb = stbi__get32le(s); + info->mr = stbi__get32le(s); + info->mg = stbi__get32le(s); + info->mb = stbi__get32le(s); // not documented, but generated by photoshop and handled by mspaint - if (mr == mg && mg == mb) { + if (info->mr == info->mg && info->mg == info->mb) { // ?!?!? return stbi__errpuc("bad BMP", "bad BMP"); } @@ -3033,11 +4668,13 @@ static stbi_uc *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int return stbi__errpuc("bad BMP", "bad BMP"); } } else { - assert(hsz == 108 || hsz == 124); - mr = stbi__get32le(s); - mg = stbi__get32le(s); - mb = stbi__get32le(s); - ma = stbi__get32le(s); + 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); stbi__get32le(s); // discard color space for (i=0; i < 12; ++i) stbi__get32le(s); // discard color space parameters @@ -3048,35 +4685,68 @@ static stbi_uc *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int stbi__get32le(s); // discard reserved } } - if (bpp < 16) - psize = (offset - 14 - hsz) >> 2; } + return (void *) 1; +} + + +static stbi_uc *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + 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; + + 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); + + 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 - 14 - 24) / 3; + } else { + if (info.bpp < 16) + psize = (info.offset - 14 - info.hsz) >> 2; + } + s->img_n = ma ? 4 : 3; - if (req_comp && req_comp >= 3) // we can directly stbi__jpeg_huff_decode 3 or 4 + 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 - out = (stbi_uc *) malloc(target * s->img_x * s->img_y); + + out = (stbi_uc *) stbi__malloc(target * s->img_x * s->img_y); if (!out) return stbi__errpuc("outofmem", "Out of memory"); - if (bpp < 16) { + if (info.bpp < 16) { int z=0; - if (psize == 0 || psize > 256) { free(out); return stbi__errpuc("invalid", "Corrupt BMP"); } + 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 (hsz != 12) stbi__get8(s); + if (info.hsz != 12) stbi__get8(s); pal[i][3] = 255; } - stbi__skip(s, offset - 14 - hsz - psize * (hsz == 12 ? 3 : 4)); - if (bpp == 4) width = (s->img_x + 1) >> 1; - else if (bpp == 8) width = s->img_x; - else { free(out); return stbi__errpuc("bad bpp", "Corrupt BMP"); } + stbi__skip(s, info.offset - 14 - info.hsz - psize * (info.hsz == 12 ? 3 : 4)); + 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; 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 (bpp == 4) { + if (info.bpp == 4) { v2 = v & 15; v >>= 4; } @@ -3085,7 +4755,7 @@ static stbi_uc *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int out[z++] = pal[v][2]; if (target == 4) out[z++] = 255; if (i+1 == (int) s->img_x) break; - v = (bpp == 8) ? stbi__get8(s) : v2; + v = (info.bpp == 8) ? stbi__get8(s) : v2; out[z++] = pal[v][0]; out[z++] = pal[v][1]; out[z++] = pal[v][2]; @@ -3097,19 +4767,19 @@ static stbi_uc *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int 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, offset - 14 - hsz); - if (bpp == 24) width = 3 * s->img_x; - else if (bpp == 16) width = 2*s->img_x; + stbi__skip(s, info.offset - 14 - 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 (bpp == 24) { + if (info.bpp == 24) { easy = 1; - } else if (bpp == 32) { + } else if (info.bpp == 32) { if (mb == 0xff && mg == 0xff00 && mr == 0x00ff0000 && ma == 0xff000000) easy = 2; } if (!easy) { - if (!mr || !mg || !mb) { free(out); return stbi__errpuc("bad masks", "Corrupt BMP"); } + 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); @@ -3125,22 +4795,31 @@ static stbi_uc *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int 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 = (stbi__uint32) (bpp == 16 ? stbi__get16le(s) : stbi__get32le(s)); + stbi__uint32 v = (bpp == 16 ? (stbi__uint32) stbi__get16le(s) : stbi__get32le(s)); 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); - if (target == 4) out[z++] = STBI__BYTECAST(a); + 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) { @@ -3162,24 +4841,60 @@ static stbi_uc *stbi__bmp_load(stbi__context *s, int *x, int *y, int *comp, int 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; + // else: fall-through + case 15: if(is_rgb16) *is_rgb16 = 1; + return STBI_rgb; + case 24: // fall-through + 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; - int sz; + 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 - sz = stbi__get8(s); // color type - if( sz > 1 ) { + tga_colormap_type = stbi__get8(s); // colormap type + if( tga_colormap_type > 1 ) { stbi__rewind(s); return 0; // only RGB or indexed allowed } - sz = stbi__get8(s); // image type - // only RGB or grey allowed, +/- RLE - if ((sz != 1) && (sz != 2) && (sz != 3) && (sz != 9) && (sz != 10) && (sz != 11)) return 0; - stbi__skip(s,9); + 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); @@ -3190,44 +4905,80 @@ static int stbi__tga_info(stbi__context *s, int *x, int *y, int *comp) stbi__rewind(s); return 0; // test height } - sz = stbi__get8(s); // bits per pixel - // only RGB or RGBA or grey allowed - if ((sz != 8) && (sz != 16) && (sz != 24) && (sz != 32)) { - stbi__rewind(s); - return 0; + 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; } - tga_comp = sz; if (x) *x = tga_w; if (y) *y = tga_h; - if (comp) *comp = tga_comp / 8; + if (comp) *comp = tga_comp; return 1; // seems to have passed everything } static int stbi__tga_test(stbi__context *s) { - int res; - int sz; + int res = 0; + int sz, tga_color_type; stbi__get8(s); // discard Offset - sz = stbi__get8(s); // color type - if ( sz > 1 ) return 0; // only RGB or indexed allowed + 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 ( (sz != 1) && (sz != 2) && (sz != 3) && (sz != 9) && (sz != 10) && (sz != 11) ) return 0; // only RGB or grey allowed, +/- RLE - stbi__get16be(s); // discard palette start - stbi__get16be(s); // discard palette length - stbi__get8(s); // discard bits per palette color entry - stbi__get16be(s); // discard x origin - stbi__get16be(s); // discard y origin - if ( stbi__get16be(s) < 1 ) return 0; // test width - if ( stbi__get16be(s) < 1 ) return 0; // test height + 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 ( (sz != 8) && (sz != 16) && (sz != 24) && (sz != 32) ) - res = 0; - else - res = 1; + 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 +void stbi__tga_read_rgb16(stbi__context *s, stbi_uc* out) +{ + stbi__uint16 px = 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] = (r * 255)/31; + out[1] = (g * 255)/31; + out[2] = (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 stbi_uc *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int req_comp) { // read in the TGA header stuff @@ -3243,8 +4994,9 @@ static stbi_uc *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int int tga_width = stbi__get16le(s); int tga_height = stbi__get16le(s); int tga_bits_per_pixel = stbi__get8(s); - int tga_comp = tga_bits_per_pixel / 8; + 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; @@ -3260,41 +5012,30 @@ static stbi_uc *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int tga_image_type -= 8; tga_is_RLE = 1; } - /* int tga_alpha_bits = tga_inverted & 15; */ tga_inverted = 1 - ((tga_inverted >> 5) & 1); - // error check - if ( //(tga_indexed) || - (tga_width < 1) || (tga_height < 1) || - (tga_image_type < 1) || (tga_image_type > 3) || - ((tga_bits_per_pixel != 8) && (tga_bits_per_pixel != 16) && - (tga_bits_per_pixel != 24) && (tga_bits_per_pixel != 32)) - ) - { - return NULL; // we don't report this as a bad TGA because we don't even know if it's TGA - } - // If I'm paletted, then I'll use the number of bits from the palette - if ( tga_indexed ) - { - tga_comp = tga_palette_bits / 8; - } + 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; - tga_data = (unsigned char*)malloc( tga_width * tga_height * tga_comp ); + tga_data = (unsigned char*)stbi__malloc( (size_t)tga_width * tga_height * tga_comp ); 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) { + if ( !tga_indexed && !tga_is_RLE && !tga_rgb16 ) { for (i=0; i < tga_height; ++i) { - int y = tga_inverted ? tga_height -i - 1 : i; - stbi_uc *tga_row = tga_data + y*tga_width*tga_comp; + 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 { @@ -3304,15 +5045,22 @@ static stbi_uc *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int // any data to skip? (offset usually = 0) stbi__skip(s, tga_palette_start ); // load the palette - tga_palette = (unsigned char*)malloc( tga_palette_len * tga_palette_bits / 8 ); + tga_palette = (unsigned char*)stbi__malloc( tga_palette_len * tga_comp ); if (!tga_palette) { - free(tga_data); + STBI_FREE(tga_data); return stbi__errpuc("outofmem", "Out of memory"); } - if (!stbi__getn(s, tga_palette, tga_palette_len * tga_palette_bits / 8 )) { - free(tga_data); - free(tga_palette); - return stbi__errpuc("bad palette", "Corrupt TGA"); + 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 @@ -3342,23 +5090,22 @@ static stbi_uc *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int // load however much data we did have if ( tga_indexed ) { - // read in 1 byte, then perform the lookup - int pal_idx = stbi__get8(s); - if ( pal_idx >= tga_palette_len ) - { - // invalid index + // 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_bits_per_pixel / 8; - for (j = 0; j*8 < tga_bits_per_pixel; ++j) - { + pal_idx *= tga_comp; + for (j = 0; j < tga_comp; ++j) { raw_data[j] = tga_palette[pal_idx+j]; } - } else - { + } 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*8 < tga_bits_per_pixel; ++j) - { + for (j = 0; j < tga_comp; ++j) { raw_data[j] = stbi__get8(s); } } @@ -3393,12 +5140,12 @@ static stbi_uc *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int // clear my palette, if I had one if ( tga_palette != NULL ) { - free( tga_palette ); + STBI_FREE( tga_palette ); } } - // swap RGB - if (tga_comp >= 3) + // 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) @@ -3421,10 +5168,12 @@ static stbi_uc *stbi__tga_load(stbi__context *s, int *x, int *y, int *comp, int // 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); @@ -3437,6 +5186,7 @@ static stbi_uc *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int int pixelCount; int channelCount, compression; int channel, i, count, len; + int bitdepth; int w,h; stbi_uc *out; @@ -3459,10 +5209,11 @@ static stbi_uc *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int // Read the rows and columns of the image. h = stbi__get32be(s); w = stbi__get32be(s); - + // Make sure the depth is 8 bits. - if (stbi__get16be(s) != 8) - return stbi__errpuc("unsupported bit depth", "PSD bit depth is not 8 bit"); + 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: @@ -3495,13 +5246,13 @@ static stbi_uc *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int return stbi__errpuc("bad compression", "PSD has an unknown compression format"); // Create the destination image. - out = (stbi_uc *) malloc(4 * w*h); + 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 @@ -3519,11 +5270,12 @@ static stbi_uc *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int // 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 = (channel == 3 ? 255 : 0), p += 4; + for (i = 0; i < pixelCount; i++, p += 4) + *p = (channel == 3 ? 255 : 0); } else { // Read the RLE data. count = 0; @@ -3557,23 +5309,30 @@ static stbi_uc *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int } } } - + } 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 value for each pixel in the image. - + // Read the data by channel. for (channel = 0; channel < 4; channel++) { stbi_uc *p; - + p = out + channel; - if (channel > channelCount) { + if (channel >= channelCount) { // Fill this channel with default data. - for (i = 0; i < pixelCount; i++) *p = channel == 3 ? 255 : 0, p += 4; + stbi_uc val = channel == 3 ? 255 : 0; + for (i = 0; i < pixelCount; i++, p += 4) + *p = val; } else { // Read the data. - for (i = 0; i < pixelCount; i++) - *p = stbi__get8(s), p += 4; + if (bitdepth == 16) { + 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); + } } } } @@ -3583,12 +5342,13 @@ static stbi_uc *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int if (out == NULL) return out; // stbi__convert_format frees input on failure } - if (comp) *comp = channelCount; + if (comp) *comp = 4; *y = h; *x = w; - + return out; } +#endif // ************************************************************************************************* // Softimage PIC loader @@ -3597,6 +5357,7 @@ static stbi_uc *stbi__psd_load(stbi__context *s, int *x, int *y, int *comp, int // 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; @@ -3729,7 +5490,6 @@ static stbi_uc *stbi__pic_load_core(stbi__context *s,int width,int height,int *c if (count >= 128) { // Repeated stbi_uc value[4]; - int i; if (count==128) count = stbi__get16be(s); @@ -3773,18 +5533,18 @@ static stbi_uc *stbi__pic_load(stbi__context *s,int *px,int *py,int *comp,int re x = stbi__get16be(s); y = stbi__get16be(s); if (stbi__at_eof(s)) return stbi__errpuc("bad file","file too short (pic header)"); - if ((1 << 28) / x < y) return stbi__errpuc("too large", "Image too large to stbi__jpeg_huff_decode"); + if ((1 << 28) / x < y) return stbi__errpuc("too large", "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 *) malloc(x*y*4); + result = (stbi_uc *) stbi__malloc(x*y*4); memset(result, 0xff, x*y*4); if (!stbi__pic_load_core(s,x,y,comp, result)) { - free(result); + STBI_FREE(result); result=0; } *px = x; @@ -3801,10 +5561,13 @@ static int stbi__pic_test(stbi__context *s) stbi__rewind(s); return r; } +#endif // ************************************************************************************************* // GIF loader -- public domain by Jean-Marc Lienher -- simplified/shrunk by stb -typedef struct + +#ifndef STBI_NO_GIF +typedef struct { stbi__int16 prefix; stbi_uc first; @@ -3814,8 +5577,8 @@ typedef struct typedef struct { int w,h; - stbi_uc *out; // output buffer (always 4 components) - int flags, bgindex, ratio, transparent, eflags; + stbi_uc *out, *old_out; // output buffer (always 4 components) + int flags, bgindex, ratio, transparent, eflags, delay; stbi_uc pal[256][4]; stbi_uc lpal[256][4]; stbi__gif_lzw codes[4096]; @@ -3852,8 +5615,8 @@ static void stbi__gif_parse_colortable(stbi__context *s, stbi_uc pal[256][4], in pal[i][2] = stbi__get8(s); pal[i][1] = stbi__get8(s); pal[i][0] = stbi__get8(s); - pal[i][3] = transp ? 0 : 255; - } + pal[i][3] = transp == i ? 0 : 255; + } } static int stbi__gif_header(stbi__context *s, stbi__gif *g, int *comp, int is_info) @@ -3864,8 +5627,8 @@ static int stbi__gif_header(stbi__context *s, stbi__gif *g, int *comp, int is_in 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"); - + 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); @@ -3886,7 +5649,7 @@ static int stbi__gif_header(stbi__context *s, stbi__gif *g, int *comp, int is_in static int stbi__gif_info_raw(stbi__context *s, int *x, int *y, int *comp) { - stbi__gif g; + stbi__gif g; if (!stbi__gif_header(s, &g, comp, 1)) { stbi__rewind( s ); return 0; @@ -3900,13 +5663,13 @@ static void stbi__out_gif_code(stbi__gif *g, stbi__uint16 code) { stbi_uc *p, *c; - // recurse to stbi__jpeg_huff_decode the prefixes, since the linked-list is backwards, + // 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; - + p = &g->out[g->cur_x + g->cur_y]; c = &g->color_table[g->codes[code].suffix * 4]; @@ -3933,22 +5696,23 @@ static void stbi__out_gif_code(stbi__gif *g, stbi__uint16 code) static stbi_uc *stbi__process_gif_raster(stbi__context *s, stbi__gif *g) { stbi_uc lzw_cs; - stbi__int32 len, code; + 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 (code = 0; code < clear; code++) { - g->codes[code].prefix = -1; - g->codes[code].first = (stbi_uc) code; - g->codes[code].suffix = (stbi_uc) code; + 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 @@ -3960,7 +5724,7 @@ static stbi_uc *stbi__process_gif_raster(stbi__context *s, stbi__gif *g) if (valid_bits < codesize) { if (len == 0) { len = stbi__get8(s); // start new block - if (len == 0) + if (len == 0) return g->out; } --len; @@ -4005,21 +5769,22 @@ static stbi_uc *stbi__process_gif_raster(stbi__context *s, stbi__gif *g) } else { return stbi__errpuc("illegal code in raster", "Corrupt GIF"); } - } + } } } -static void stbi__fill_gif_background(stbi__gif *g) +static void stbi__fill_gif_background(stbi__gif *g, int x0, int y0, int x1, int y1) { - int i; + int x, y; stbi_uc *c = g->pal[g->bgindex]; - // @OPTIMIZE: write a dword at a time - for (i = 0; i < g->w * g->h * 4; i += 4) { - stbi_uc *p = &g->out[i]; - p[0] = c[2]; - p[1] = c[1]; - p[2] = c[0]; - p[3] = c[3]; + for (y = y0; y < y1; y += 4 * g->w) { + for (x = x0; x < x1; x += 4) { + stbi_uc *p = &g->out[y + x]; + p[0] = c[2]; + p[1] = c[1]; + p[2] = c[0]; + p[3] = 0; + } } } @@ -4027,27 +5792,40 @@ static void stbi__fill_gif_background(stbi__gif *g) static stbi_uc *stbi__gif_load_next(stbi__context *s, stbi__gif *g, int *comp, int req_comp) { int i; - stbi_uc *old_out = 0; + stbi_uc *prev_out = 0; - if (g->out == 0) { - if (!stbi__gif_header(s, g, comp,0)) return 0; // stbi__g_failure_reason set by stbi__gif_header - g->out = (stbi_uc *) malloc(4 * g->w * g->h); - if (g->out == 0) return stbi__errpuc("outofmem", "Out of memory"); - stbi__fill_gif_background(g); - } else { - // animated-gif-only path - if (((g->eflags & 0x1C) >> 2) == 3) { - old_out = g->out; - g->out = (stbi_uc *) malloc(4 * g->w * g->h); - if (g->out == 0) return stbi__errpuc("outofmem", "Out of memory"); - memcpy(g->out, old_out, g->w*g->h*4); - } + if (g->out == 0 && !stbi__gif_header(s, g, comp,0)) + return 0; // stbi__g_failure_reason set by stbi__gif_header + + prev_out = g->out; + g->out = (stbi_uc *) stbi__malloc(4 * g->w * g->h); + if (g->out == 0) return stbi__errpuc("outofmem", "Out of memory"); + + switch ((g->eflags & 0x1C) >> 2) { + case 0: // unspecified (also always used on 1st frame) + stbi__fill_gif_background(g, 0, 0, 4 * g->w, 4 * g->w * g->h); + break; + case 1: // do not dispose + if (prev_out) memcpy(g->out, prev_out, 4 * g->w * g->h); + g->old_out = prev_out; + break; + case 2: // dispose to background + if (prev_out) memcpy(g->out, prev_out, 4 * g->w * g->h); + stbi__fill_gif_background(g, g->start_x, g->start_y, g->max_x, g->max_y); + break; + case 3: // dispose to previous + if (g->old_out) { + for (i = g->start_y; i < g->max_y; i += 4 * g->w) + memcpy(&g->out[i + g->start_x], &g->old_out[i + g->start_x], g->max_x - g->start_x); + } + break; } - + for (;;) { switch (stbi__get8(s)) { case 0x2C: /* Image Descriptor */ { + int prev_trans = -1; stbi__int32 x, y, w, h; stbi_uc *o; @@ -4078,21 +5856,22 @@ static stbi_uc *stbi__gif_load_next(stbi__context *s, stbi__gif *g, int *comp, i 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; + g->color_table = (stbi_uc *) g->lpal; } else if (g->flags & 0x80) { - for (i=0; i < 256; ++i) // @OPTIMIZE: stbi__jpeg_reset only the previous transparent - g->pal[i][3] = 255; - if (g->transparent >= 0 && (g->eflags & 0x01)) + if (g->transparent >= 0 && (g->eflags & 0x01)) { + prev_trans = g->pal[g->transparent][3]; g->pal[g->transparent][3] = 0; + } 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 == NULL) return NULL; - if (req_comp && req_comp != 4) - o = stbi__convert_format(o, 4, req_comp, g->w, g->h); + if (prev_trans != -1) + g->pal[g->transparent][3] = (stbi_uc) prev_trans; + return o; } @@ -4103,7 +5882,7 @@ static stbi_uc *stbi__gif_load_next(stbi__context *s, stbi__gif *g, int *comp, i len = stbi__get8(s); if (len == 4) { g->eflags = stbi__get8(s); - stbi__get16le(s); // delay + g->delay = stbi__get16le(s); g->transparent = stbi__get8(s); } else { stbi__skip(s, len); @@ -4116,25 +5895,32 @@ static stbi_uc *stbi__gif_load_next(stbi__context *s, stbi__gif *g, int *comp, i } case 0x3B: // gif stream termination code - return (stbi_uc *) 1; + return (stbi_uc *) s; // using '1' causes warning on some compilers default: return stbi__errpuc("unknown code", "Corrupt GIF"); } } + + STBI_NOTUSED(req_comp); } static stbi_uc *stbi__gif_load(stbi__context *s, int *x, int *y, int *comp, int req_comp) { stbi_uc *u = 0; - stbi__gif g={0}; + stbi__gif g; + memset(&g, 0, sizeof(g)); u = stbi__gif_load_next(s, &g, comp, req_comp); - if (u == (void *) 1) u = 0; // end of animated gif marker + if (u == (stbi_uc *) s) u = 0; // end of animated gif marker if (u) { *x = g.w; *y = g.h; + if (req_comp && req_comp != 4) + u = stbi__convert_format(u, 4, req_comp, g.w, g.h); } + else if (g.out) + STBI_FREE(g.out); return u; } @@ -4143,7 +5929,7 @@ 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 @@ -4232,7 +6018,7 @@ static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int re // Check identifier if (strcmp(stbi__hdr_gettoken(s,buffer), "#?RADIANCE") != 0) return stbi__errpf("not HDR", "Corrupt HDR image"); - + // Parse header for(;;) { token = stbi__hdr_gettoken(s,buffer); @@ -4260,7 +6046,7 @@ static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int re if (req_comp == 0) req_comp = 3; // Read data - hdr_data = (float *) malloc(height * width * req_comp * sizeof(float)); + hdr_data = (float *) stbi__malloc(height * width * req_comp * sizeof(float)); // Load image data // image data is stored as some number of sca @@ -4293,14 +6079,14 @@ static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int re stbi__hdr_convert(hdr_data, rgbe, req_comp); i = 1; j = 0; - free(scanline); + STBI_FREE(scanline); goto main_decode_loop; // yes, this makes no sense } len <<= 8; len |= stbi__get8(s); - if (len != width) { free(hdr_data); free(scanline); return stbi__errpf("invalid decoded scanline length", "corrupt HDR"); } - if (scanline == NULL) scanline = (stbi_uc *) malloc(width * 4); - + 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(width * 4); + for (k = 0; k < 4; ++k) { i = 0; while (i < width) { @@ -4321,7 +6107,7 @@ static float *stbi__hdr_load(stbi__context *s, int *x, int *y, int *comp, int re for (i=0; i < width; ++i) stbi__hdr_convert(hdr_data+(j*width + i)*req_comp, scanline + i*4, req_comp); } - free(scanline); + STBI_FREE(scanline); } return hdr_data; @@ -4333,7 +6119,7 @@ static int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp) char *token; int valid = 0; - if (strcmp(stbi__hdr_gettoken(s,buffer), "#?RADIANCE") != 0) { + if (stbi__hdr_test(s) == 0) { stbi__rewind( s ); return 0; } @@ -4367,34 +6153,25 @@ static int stbi__hdr_info(stbi__context *s, int *x, int *y, int *comp) } #endif // STBI_NO_HDR +#ifndef STBI_NO_BMP static int stbi__bmp_info(stbi__context *s, int *x, int *y, int *comp) { - int hsz; - if (stbi__get8(s) != 'B' || stbi__get8(s) != 'M') { - stbi__rewind( s ); - return 0; - } - stbi__skip(s,12); - hsz = stbi__get32le(s); - if (hsz != 12 && hsz != 40 && hsz != 56 && hsz != 108 && hsz != 124) { - stbi__rewind( s ); - return 0; - } - if (hsz == 12) { - *x = stbi__get16le(s); - *y = stbi__get16le(s); - } else { - *x = stbi__get32le(s); - *y = stbi__get32le(s); - } - if (stbi__get16le(s) != 1) { - stbi__rewind( s ); - return 0; - } - *comp = stbi__get16le(s) / 8; + void *p; + stbi__bmp_data info; + + info.all_a = 255; + p = stbi__bmp_parse_header(s, &info); + stbi__rewind( s ); + if (p == NULL) + return 0; + *x = s->img_x; + *y = s->img_y; + *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; @@ -4425,20 +6202,30 @@ static int stbi__psd_info(stbi__context *s, int *x, int *y, int *comp) *comp = 4; 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; stbi__pic_packet packets[10]; - stbi__skip(s, 92); + 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)) return 0; + if (stbi__at_eof(s)) { + stbi__rewind( s); + return 0; + } if ( (*x) != 0 && (1 << 28) / (*x) < (*y)) { - stbi__rewind( s ); - return 0; + stbi__rewind( s ); + return 0; } stbi__skip(s, 8); @@ -4470,35 +6257,171 @@ static int stbi__pic_info(stbi__context *s, int *x, int *y, int *comp) 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) +// Does not support 16-bit-per-channel + +#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 stbi_uc *stbi__pnm_load(stbi__context *s, int *x, int *y, int *comp, int req_comp) +{ + stbi_uc *out; + if (!stbi__pnm_info(s, (int *)&s->img_x, (int *)&s->img_y, (int *)&s->img_n)) + return 0; + *x = s->img_x; + *y = s->img_y; + *comp = s->img_n; + + out = (stbi_uc *) stbi__malloc(s->img_n * s->img_x * s->img_y); + if (!out) return stbi__errpuc("outofmem", "Out of memory"); + stbi__getn(s, out, s->img_n * s->img_x * s->img_y); + + if (req_comp && req_comp != s->img_n) { + 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); + } + + return value; +} + +static int stbi__pnm_info(stbi__context *s, int *x, int *y, int *comp) +{ + int maxv; + char c, p, t; + + 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 + stbi__pnm_skip_whitespace(s, &c); + + *y = stbi__pnm_getinteger(s, &c); // read height + stbi__pnm_skip_whitespace(s, &c); + + maxv = stbi__pnm_getinteger(s, &c); // read max value + + if (maxv > 255) + return stbi__err("max value > 255", "PPM image not 8-bit"); + else + return 1; +} +#endif static int stbi__info_main(stbi__context *s, int *x, int *y, int *comp) { - if (stbi__jpeg_info(s, x, y, comp)) - return 1; - if (stbi__png_info(s, x, y, comp)) - return 1; - if (stbi__gif_info(s, x, y, comp)) - return 1; - if (stbi__bmp_info(s, x, y, comp)) - return 1; - if (stbi__psd_info(s, x, y, comp)) - return 1; - if (stbi__pic_info(s, x, y, comp)) - return 1; + #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; + 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"); } #ifndef STBI_NO_STDIO STBIDEF int stbi_info(char const *filename, int *x, int *y, int *comp) { - FILE *f = fopen(filename, "rb"); + 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); @@ -4534,114 +6457,158 @@ STBIDEF int stbi_info_from_callbacks(stbi_io_callbacks const *c, void *user, int #endif // STB_IMAGE_IMPLEMENTATION -#if !defined(STBI_NO_STDIO) && defined(_MSC_VER) && _MSC_VER >= 1400 -#pragma warning(pop) -#endif - - /* revision history: - 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 first released version + 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 */ diff --git a/src/stb_truetype.h b/src/stb_truetype.h index 570f236..bfb1841 100644 --- a/src/stb_truetype.h +++ b/src/stb_truetype.h @@ -1,2081 +1,3249 @@ -// stb_truetype.h - v0.8 - public domain -// authored from 2009-2013 by Sean Barrett / RAD Game Tools -// -// This library processes TrueType files: -// parse files -// extract glyph metrics -// extract glyph shapes -// render glyphs to one-channel bitmaps with antialiasing (box filter) -// -// 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 -// -// Bug/warning reports: -// "Zer" on mollyrocket (with fix) -// Cass Everitt -// stoiko (Haemimont Games) -// Brian Hook -// Walter van Niftrik -// David Gow -// David Given -// Ivan-Assen Ivanov -// Anthony Pesch -// -// VERSION HISTORY -// -// 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 -// -// LICENSE -// -// This software is in the public domain. Where that dedication is not -// recognized, you are granted a perpetual, irrevokable license to copy -// and modify this file as you see fit. -// -// USAGE -// -// Include this file in whatever places neeed 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. -// -// Simple 3D API (don't ship this, but it's fine for tools and quick start, -// and you can cut and paste from it to move to more advanced) -// stbtt_BakeFontBitmap() -- bake a font to a bitmap for use as texture -// stbtt_GetBakedQuad() -- compute quad to draw for a given char -// -// "Load" a font file from a memory buffer (you have to keep the buffer loaded) -// stbtt_InitFont() -// stbtt_GetFontOffsetForIndex() -- use 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_GetCodepointKernAdvance() -// -// 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 they 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. -// -// 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. -// -// -// SOURCE STATISTICS (based on v0.6c, 2050 LOC) -// -// Documentation & header file 520 LOC \___ 660 LOC documentation -// Sample code 140 LOC / -// Truetype parsing 620 LOC ---- 620 LOC TrueType -// Software rasterization 240 LOC \ . -// Curve tesselation 120 LOC \__ 550 LOC Bitmap creation -// Bitmap management 100 LOC / -// Baked bitmap interface 70 LOC / -// Font name matching & access 150 LOC ---- 150 -// C runtime library abstraction 60 LOC ---- 60 - - -////////////////////////////////////////////////////////////////////////////// -////////////////////////////////////////////////////////////////////////////// -//// -//// SAMPLE PROGRAMS -//// -// -// Incomplete text-in-3d-api example, which draws quads properly aligned to be lossless -// -#if 0 -#define STB_TRUETYPE_IMPLEMENTATION // force following include to generate implementation -#include "stb_truetype.h" - -char ttf_buffer[1<<20]; -unsigned char temp_bitmap[512*512]; - -stbtt_bakedchar cdata[96]; // ASCII 32..126 is 95 glyphs -GLstbtt_uint ftex; - -void my_stbtt_initfont(void) -{ - fread(ttf_buffer, 1, 1<<20, fopen("c:/windows/fonts/times.ttf", "rb")); - stbtt_BakeFontBitmap(data,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 - 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.t1); glVertex2f(q.x0,q.y0); - glTexCoord2f(q.s1,q.t1); glVertex2f(q.x1,q.y0); - glTexCoord2f(q.s1,q.t0); glVertex2f(q.x1,q.y1); - glTexCoord2f(q.s0,q.t0); glVertex2f(q.x0,q.y1); - } - ++text; - } - glEnd(); -} -#endif -// -// -////////////////////////////////////////////////////////////////////////////// -// -// Complete program (this compiles): get a single bitmap, print as ASCII art -// -#if 0 -#include -#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!"; - - 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. - -#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]; - - // #define your own STBTT_sort() to override this to avoid qsort - #ifndef STBTT_sort - #include - #define STBTT_sort(data,num_items,item_size,compare_func) qsort(data,num_items,item_size,compare_func) - #endif - - // #define your own STBTT_ifloor/STBTT_iceil() to avoid math.h - #ifndef STBTT_ifloor - #include - #define STBTT_ifloor(x) ((int) floor(x)) - #define STBTT_iceil(x) ((int) ceil(x)) - #endif - - #ifndef STBTT_sqrt - #include - #define STBTT_sqrt(x) sqrt(x) - #endif - - // #define your own functions "STBTT_malloc" / "STBTT_free" to avoid malloc.h - #ifndef STBTT_malloc - #include - #define STBTT_malloc(x,u) ((void)(u),malloc(x)) - #define STBTT_free(x,u) free(x) - #endif - - #ifndef STBTT_assert - #include - #define STBTT_assert(x) assert(x) - #endif - - #ifndef STBTT_strlen - #include - #define STBTT_strlen(x) strlen(x) - #endif - - #ifndef STBTT_memcpy - #include - #define STBTT_memcpy memcpy - #define STBTT_memset memset - #endif -#endif - -/////////////////////////////////////////////////////////////////////////////// -/////////////////////////////////////////////////////////////////////////////// -//// -//// INTERFACE -//// -//// - -#ifndef __STB_INCLUDE_STB_TRUETYPE_H__ -#define __STB_INCLUDE_STB_TRUETYPE_H__ - -#ifdef __cplusplus -extern "C" { -#endif - -////////////////////////////////////////////////////////////////////////////// -// -// 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; - -extern 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; - -extern void stbtt_GetBakedQuad(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. - - -////////////////////////////////////////////////////////////////////////////// -// -// FONT LOADING -// -// - -extern 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. You can just skip -// this step if you know it's that kind of font. - - -// The following structure is defined publically so you can declare one on -// the stack or as a global or etc, but you should treat it as opaque. -typedef 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; // 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_fontinfo; - -extern 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 - -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. - - -////////////////////////////////////////////////////////////////////////////// -// -// CHARACTER PROPERTIES -// - -extern 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. - -extern 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. - -extern 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 - -extern void stbtt_GetFontBoundingBox(const stbtt_fontinfo *info, int *x0, int *y0, int *x1, int *y1); -// the bounding box around all possible characters - -extern 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 - -extern int stbtt_GetCodepointKernAdvance(const stbtt_fontinfo *info, int ch1, int ch2); -// an additional amount to add to the 'advance' value between ch1 and ch2 - -extern 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 - -extern void stbtt_GetGlyphHMetrics(const stbtt_fontinfo *info, int glyph_index, int *advanceWidth, int *leftSideBearing); -extern int stbtt_GetGlyphKernAdvance(const stbtt_fontinfo *info, int glyph1, int glyph2); -extern 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 - - -////////////////////////////////////////////////////////////////////////////// -// -// 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 - }; -#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; - unsigned char type,padding; - } stbtt_vertex; -#endif - -extern int stbtt_IsGlyphEmpty(const stbtt_fontinfo *info, int glyph_index); -// returns non-zero if nothing is drawn for this glyph - -extern int stbtt_GetCodepointShape(const stbtt_fontinfo *info, int unicode_codepoint, stbtt_vertex **vertices); -extern 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 countours. 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. - -extern void stbtt_FreeShape(const stbtt_fontinfo *info, stbtt_vertex *vertices); -// frees the data allocated above - -////////////////////////////////////////////////////////////////////////////// -// -// BITMAP RENDERING -// - -extern void stbtt_FreeBitmap(unsigned char *bitmap, void *userdata); -// frees the bitmap allocated below - -extern 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 - -extern 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 - -extern 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. - -extern 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 - -extern 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.) - -extern 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) -extern unsigned char *stbtt_GetGlyphBitmap(const stbtt_fontinfo *info, float scale_x, float scale_y, int glyph, int *width, int *height, int *xoff, int *yoff); -extern 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); -extern 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); -extern 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); -extern void stbtt_GetGlyphBitmapBox(const stbtt_fontinfo *font, int glyph, float scale_x, float scale_y, int *ix0, int *iy0, int *ix1, int *iy1); -extern 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; - -extern 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); - -////////////////////////////////////////////////////////////////////////////// -// -// 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. - - -extern 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 - -extern 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 - -extern 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 - -////////////////////////////////////////////////////////////////////////// -// -// 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) - -#if defined(STB_TRUETYPE_BIGENDIAN) && !defined(ALLOW_UNALIGNED_TRUETYPE) - - #define ttUSHORT(p) (* (stbtt_uint16 *) (p)) - #define ttSHORT(p) (* (stbtt_int16 *) (p)) - #define ttULONG(p) (* (stbtt_uint32 *) (p)) - #define ttLONG(p) (* (stbtt_int32 *) (p)) - -#else - - stbtt_uint16 ttUSHORT(const stbtt_uint8 *p) { return p[0]*256 + p[1]; } - stbtt_int16 ttSHORT(const stbtt_uint8 *p) { return p[0]*256 + p[1]; } - stbtt_uint32 ttULONG(const stbtt_uint8 *p) { return (p[0]<<24) + (p[1]<<16) + (p[2]<<8) + p[3]; } - stbtt_int32 ttLONG(const stbtt_uint8 *p) { return (p[0]<<24) + (p[1]<<16) + (p[2]<<8) + p[3]; } - -#endif - -#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(const 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 - 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; -} - -int stbtt_GetFontOffsetForIndex(const 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*14); - } - } - return -1; -} - -int stbtt_InitFont(stbtt_fontinfo *info, const unsigned char *data2, int fontstart) -{ - stbtt_uint8 *data = (stbtt_uint8 *) data2; - stbtt_uint32 cmap, t; - stbtt_int32 i,numTables; - - info->data = data; - info->fontstart = fontstart; - - 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 - if (!cmap || !info->loca || !info->head || !info->glyf || !info->hhea || !info->hmtx) - return 0; - - t = stbtt__find_table(data, fontstart, "maxp"); - if (t) - info->numGlyphs = ttUSHORT(data+t+4); - else - info->numGlyphs = 0xffff; - - // 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; - } - } - if (info->index_map == 0) - return 0; - - info->indexToLocFormat = ttUSHORT(data+info->head + 50); - return 1; -} - -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; - stbtt_uint16 item, offset, start, end; - - // 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) { - searchRange >>= 1; - start = ttUSHORT(data + search + searchRange*2 + segcount*2 + 2); - end = ttUSHORT(data + search + searchRange*2); - if (unicode_codepoint > end) - search += searchRange*2; - --entrySelector; - } - search += 2; - - item = (stbtt_uint16) ((search - endCount) >> 1); - - STBTT_assert(unicode_codepoint <= ttUSHORT(data + endCount + 2*item)); - start = ttUSHORT(data + index_map + 14 + segcount*2 + 2 + 2*item); - end = ttUSHORT(data + index_map + 14 + 2 + 2*item); - if (unicode_codepoint < start) - 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; -} - -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; - - 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 -} - -int stbtt_GetGlyphBox(const stbtt_fontinfo *info, int glyph_index, int *x0, int *y0, int *x1, int *y1) -{ - 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; -} - -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); -} - -int stbtt_IsGlyphEmpty(const stbtt_fontinfo *info, int glyph_index) -{ - stbtt_int16 numberOfContours; - int 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; -} - -int stbtt_GetGlyphShape(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 == -1) { - // 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) 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 if (numberOfContours < 0) { - // @TODO other compound variations? - STBTT_assert(0); - } else { - // numberOfCounters == 0, do nothing - } - - *pvertices = vertices; - return num_vertices; -} - -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)); - } -} - -int stbtt_GetGlyphKernAdvance(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; -} - -int stbtt_GetCodepointKernAdvance(const stbtt_fontinfo *info, int ch1, int ch2) -{ - if (!info->kern) // 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)); -} - -void stbtt_GetCodepointHMetrics(const stbtt_fontinfo *info, int codepoint, int *advanceWidth, int *leftSideBearing) -{ - stbtt_GetGlyphHMetrics(info, stbtt_FindGlyphIndex(info,codepoint), advanceWidth, leftSideBearing); -} - -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); -} - -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); -} - -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; -} - -float stbtt_ScaleForMappingEmToPixels(const stbtt_fontinfo *info, float pixels) -{ - int unitsPerEm = ttUSHORT(info->data + info->head + 18); - return pixels / unitsPerEm; -} - -void stbtt_FreeShape(const stbtt_fontinfo *info, stbtt_vertex *v) -{ - STBTT_free(v, info->userdata); -} - -////////////////////////////////////////////////////////////////////////////// -// -// antialiasing software rasterizer -// - -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,y0,x1,y1; - if (!stbtt_GetGlyphBox(font, glyph, &x0,&y0,&x1,&y1)) - x0=y0=x1=y1=0; // e.g. space character - // now 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_iceil (y1 * scale_y + shift_y); - if (ix1) *ix1 = STBTT_iceil (x1 * scale_x + shift_x); - if (iy1) *iy1 = -STBTT_ifloor(y0 * scale_y + shift_y); -} -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); -} - -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); -} - -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); -} - -typedef struct stbtt__edge { - float x0,y0, x1,y1; - int invert; -} stbtt__edge; - -typedef struct stbtt__active_edge -{ - int x,dx; - float ey; - struct stbtt__active_edge *next; - int valid; -} stbtt__active_edge; - -#define FIXSHIFT 10 -#define FIX (1 << FIXSHIFT) -#define FIXMASK (FIX-1) - -static stbtt__active_edge *new_active(stbtt__edge *e, int off_x, float start_point, void *userdata) -{ - stbtt__active_edge *z = (stbtt__active_edge *) STBTT_malloc(sizeof(*z), userdata); // @TODO: make a pool of these!!! - float dxdy = (e->x1 - e->x0) / (e->y1 - e->y0); - STBTT_assert(e->y0 <= start_point); - if (!z) return z; - // round dx down to avoid going too far - if (dxdy < 0) - z->dx = -STBTT_ifloor(FIX * -dxdy); - else - z->dx = STBTT_ifloor(FIX * dxdy); - z->x = STBTT_ifloor(FIX * (e->x0 + dxdy * (start_point - e->y0))); - z->x -= off_x * FIX; - z->ey = e->y1; - z->next = 0; - z->valid = e->invert ? 1 : -1; - return z; -} - -// 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->valid; - } else { - int x1 = e->x; w += e->valid; - // if we went to zero, we need to draw - if (w == 0) { - int i = x0 >> FIXSHIFT; - int j = x1 >> 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 >> FIXSHIFT); - } else { - if (i >= 0) // add antialiasing for x0 - scanline[i] = scanline[i] + (stbtt_uint8) (((FIX - (x0 & FIXMASK)) * max_weight) >> FIXSHIFT); - else - i = -1; // clip - - if (j < len) // add antialiasing for x1 - scanline[j] = scanline[j] + (stbtt_uint8) (((x1 & FIXMASK) * max_weight) >> 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__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->valid); - z->valid = 0; - STBTT_free(z, userdata); - } 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 = new_active(e, off_x, scan_y, userdata); - // 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; - } - - while (active) { - stbtt__active_edge *z = active; - active = active->next; - STBTT_free(z, userdata); - } - - if (scanline != scanline_data) - STBTT_free(scanline, userdata); -} - -static int stbtt__edge_compare(const void *p, const void *q) -{ - stbtt__edge *a = (stbtt__edge *) p; - stbtt__edge *b = (stbtt__edge *) q; - - if (a->y0 < b->y0) return -1; - if (a->y0 > b->y0) return 1; - return 0; -} - -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; - int vsubsample = result->h < 8 ? 15 : 5; - // 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 * vsubsample + shift_y; - e[n].x1 = p[b].x * scale_x + shift_x; - e[n].y1 = p[b].y * y_scale_inv * vsubsample + shift_y; - ++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); - - // 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; -} - -// tesselate until threshhold 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; -} - -// returns number of contours -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; - } - } - (*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; -} - -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, *winding_lengths; - 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); - } -} - -void stbtt_FreeBitmap(unsigned char *bitmap, void *userdata) -{ - STBTT_free(bitmap, userdata); -} - -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) 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; -} - -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); -} - -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); -} - -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); -} - -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); -} - -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)); -} - -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); -} - -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 - -extern 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) -{ - float scale; - int x,y,bottom_y, i; - stbtt_fontinfo f; - stbtt_InitFont(&f, data, offset); - 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 + 2; - if (y+gh+2 > bottom_y) - bottom_y = y+gh+2; - } - return bottom_y; -} - -void stbtt_GetBakedQuad(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; - stbtt_bakedchar *b = chardata + char_index; - int round_x = STBTT_ifloor((*xpos + b->xoff) + 0.5); - int round_y = STBTT_ifloor((*ypos + b->yoff) + 0.5); - - 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; -} - -////////////////////////////////////////////////////////////////////////////// -// -// 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(const stbtt_uint8 *s1, stbtt_int32 len1, const 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; -} - -int stbtt_CompareUTF8toUTF16_bigendian(const char *s1, int len1, const char *s2, int len2) -{ - return len1 == stbtt__CompareUTF8toUTF16_bigendian_prefix((const stbtt_uint8*) s1, len1, (const 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 -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((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; -} - -int stbtt_FindMatchingFont(const unsigned char *font_collection, const 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; - } -} - -#endif // STB_TRUETYPE_IMPLEMENTATION +// stb_truetype.h - v1.09 - public domain +// authored from 2009-2015 by Sean Barrett / RAD Game Tools +// +// This library processes TrueType files: +// parse files +// extract glyph metrics +// extract glyph shapes +// render glyphs to one-channel bitmaps with antialiasing (box filter) +// +// 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 +// +// Bug/warning reports/fixes: +// "Zer" on mollyrocket (with fix) +// Cass Everitt +// stoiko (Haemimont Games) +// Brian Hook +// Walter van Niftrik +// David Gow +// David Given +// Ivan-Assen Ivanov +// Anthony Pesch +// Johan Duparc +// Hou Qiming +// Fabian "ryg" Giesen +// Martins Mozeiko +// Cap Petschulat +// Omar Cornut +// github:aloucks +// Peter LaValle +// Sergey Popov +// Giumo X. Clanjor +// Higor Euripedes +// Thomas Fields +// Derek Vinyard +// +// Misc other: +// Ryan Gordon +// +// VERSION HISTORY +// +// 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 +// 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 +// +// Full history can be found at the end of this file. +// +// LICENSE +// +// This software is in the public domain. Where that dedication is not +// recognized, you are granted a perpetual, irrevocable license to copy, +// distribute, and modify this file as you see fit. +// +// USAGE +// +// Include this file in whatever places neeed 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_PackSetOversample() -- 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() -- use 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_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 they 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. +// +// 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. +// +// +// SOURCE STATISTICS (based on v0.6c, 2050 LOC) +// +// Documentation & header file 520 LOC \___ 660 LOC documentation +// Sample code 140 LOC / +// Truetype parsing 620 LOC ---- 620 LOC TrueType +// Software rasterization 240 LOC \ . +// Curve tesselation 120 LOC \__ 550 LOC Bitmap creation +// Bitmap management 100 LOC / +// Baked bitmap interface 70 LOC / +// Font name matching & access 150 LOC ---- 150 +// C runtime library abstraction 60 LOC ---- 60 +// +// +// 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 +// +#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_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.t1); glVertex2f(q.x0,q.y0); + glTexCoord2f(q.s1,q.t1); glVertex2f(q.x1,q.y0); + glTexCoord2f(q.s1,q.t0); glVertex2f(q.x1,q.y1); + glTexCoord2f(q.s0,q.t0); glVertex2f(q.x0,q.y1); + } + ++text; + } + glEnd(); +} +#endif +// +// +////////////////////////////////////////////////////////////////////////////// +// +// Complete program (this compiles): get a single bitmap, print as ASCII art +// +#if 0 +#include +#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. + +#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]; + + // #define your own STBTT_ifloor/STBTT_iceil() to avoid math.h + #ifndef STBTT_ifloor + #include + #define STBTT_ifloor(x) ((int) floor(x)) + #define STBTT_iceil(x) ((int) ceil(x)) + #endif + + #ifndef STBTT_sqrt + #include + #define STBTT_sqrt(x) sqrt(x) + #endif + + // #define your own functions "STBTT_malloc" / "STBTT_free" to avoid malloc.h + #ifndef STBTT_malloc + #include + #define STBTT_malloc(x,u) ((void)(u),malloc(x)) + #define STBTT_free(x,u) ((void)(u),free(x)) + #endif + + #ifndef STBTT_assert + #include + #define STBTT_assert(x) assert(x) + #endif + + #ifndef STBTT_strlen + #include + #define STBTT_strlen(x) strlen(x) + #endif + + #ifndef STBTT_memcpy + #include + #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 + +////////////////////////////////////////////////////////////////////////////// +// +// 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(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. + + + +////////////////////////////////////////////////////////////////////////////// +// +// 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 weight x 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, 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, 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_GetPackedQuad(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, 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, 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; + unsigned int h_oversample, v_oversample; + unsigned char *pixels; + void *nodes; +}; + +////////////////////////////////////////////////////////////////////////////// +// +// FONT LOADING +// +// + +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. You can just skip +// this step if you know it's that kind of font. + + +// The following structure is defined publically so you can declare one on +// the stack or as a global or etc, but you should treat it as opaque. +typedef 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; // 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_fontinfo; + +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. + + +////////////////////////////////////////////////////////////////////////////// +// +// 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 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 + + +////////////////////////////////////////////////////////////////////////////// +// +// 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 + }; +#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; + 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 countours. 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 + +////////////////////////////////////////////////////////////////////////////// +// +// 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_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_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 + +////////////////////////////////////////////////////////////////////////////// +// +// 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 + +////////////////////////////////////////////////////////////////////////// +// +// 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) + +#if defined(STB_TRUETYPE_BIGENDIAN) && !defined(ALLOW_UNALIGNED_TRUETYPE) + + #define ttUSHORT(p) (* (stbtt_uint16 *) (p)) + #define ttSHORT(p) (* (stbtt_int16 *) (p)) + #define ttULONG(p) (* (stbtt_uint32 *) (p)) + #define ttLONG(p) (* (stbtt_int32 *) (p)) + +#else + + static stbtt_uint16 ttUSHORT(const stbtt_uint8 *p) { return p[0]*256 + p[1]; } + static stbtt_int16 ttSHORT(const stbtt_uint8 *p) { return p[0]*256 + p[1]; } + static stbtt_uint32 ttULONG(const stbtt_uint8 *p) { return (p[0]<<24) + (p[1]<<16) + (p[2]<<8) + p[3]; } + static stbtt_int32 ttLONG(const stbtt_uint8 *p) { return (p[0]<<24) + (p[1]<<16) + (p[2]<<8) + p[3]; } + +#endif + +#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(const 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 + 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; +} + +STBTT_DEF int stbtt_GetFontOffsetForIndex(const 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; +} + +STBTT_DEF int stbtt_InitFont(stbtt_fontinfo *info, const unsigned char *data2, int fontstart) +{ + stbtt_uint8 *data = (stbtt_uint8 *) data2; + stbtt_uint32 cmap, t; + stbtt_int32 i,numTables; + + info->data = data; + info->fontstart = fontstart; + + 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 + if (!cmap || !info->loca || !info->head || !info->glyf || !info->hhea || !info->hmtx) + return 0; + + t = stbtt__find_table(data, fontstart, "maxp"); + if (t) + info->numGlyphs = ttUSHORT(data+t+4); + else + info->numGlyphs = 0xffff; + + // 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; + stbtt_uint16 item = (stbtt_uint16) ((search - endCount) >> 1); + + STBTT_assert(unicode_codepoint <= ttUSHORT(data + endCount + 2*item)); + start = ttUSHORT(data + index_map + 14 + segcount*2 + 2 + 2*item); + if (unicode_codepoint < start) + 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; + + 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 +} + +STBTT_DEF int stbtt_GetGlyphBox(const stbtt_fontinfo *info, int glyph_index, int *x0, int *y0, int *x1, int *y1) +{ + 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 = 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; +} + +STBTT_DEF int stbtt_GetGlyphShape(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 == -1) { + // 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) 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 if (numberOfContours < 0) { + // @TODO other compound variations? + STBTT_assert(0); + } else { + // numberOfCounters == 0, do nothing + } + + *pvertices = vertices; + return num_vertices; +} + +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_GetGlyphKernAdvance(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; +} + +STBTT_DEF int stbtt_GetCodepointKernAdvance(const stbtt_fontinfo *info, int ch1, int ch2) +{ + if (!info->kern) // 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 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); +} + +////////////////////////////////////////////////////////////////////////////// +// +// 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) + 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 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; + STBTT_assert(x >= 0 && x < len); + scanline[x] += e->direction * (1-((x_top - x) + (x_bottom-x))/2) * height; + scanline_fill[x] += e->direction * height; // everything right of this pixel is filled + } else { + int x,x1,x2; + float y_crossing, 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; + } + + x1 = (int) x_top; + x2 = (int) x_bottom; + // compute intersection with y axis at x1+1 + y_crossing = (x1+1 - x0) * dy + y_top; + + sign = e->direction; + // area of the rectangle covered from y0..y_crossing + area = sign * (y_crossing-sy0); + // area of the triangle (x_top,y0), (x+1,y0), (x+1,y_crossing) + scanline[x1] += area * (1-((x_top - x1)+(x1+1-x1))/2); + + step = sign * dy; + for (x = x1+1; x < x2; ++x) { + scanline[x] += area + step/2; + area += step; + } + y_crossing += dy * (x2 - (x1+1)); + + STBTT_assert(fabs(area) <= 1.01f); + + scanline[x2] += area + sign * (1-((x2-x2)+(x_bottom-x2))/2) * (sy1-y_crossing); + + 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 + 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 clear pairs + float y0 = y_top; + float x1 = (float) (x); + float x2 = (float) (x+1); + float x3 = xb; + float y3 = y_bottom; + float y1,y2; + + // x = e->x + e->dx * (y-y_top) + // (y-y_top) = (x - e->x) / e->dx + // y = (x - e->x) / e->dx + y_top + y1 = (x - x0) / dx + y_top; + 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; + + 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) { + STBTT_assert(z->ey >= scan_y_top); + // 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) 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) +{ + /* threshhold 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 && midn => n; 0 0 */ + /* 0n: 0>n => 0; 0 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; +} + +// tesselate until threshhold 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; +} + +// 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; + } + } + (*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, *winding_lengths; + 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) 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_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 + +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) +{ + 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(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; + 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 +#ifdef _MSC_VER +#define STBTT__NOTUSED(v) (void)(v) +#else +#define STBTT__NOTUSED(v) (void)sizeof(v) +#endif + +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; + + 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; +} + +#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, stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects) +{ + int i,j,k; + + 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); + 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); + ++k; + } + } + + return k; +} + +// rects array must be big enough to accommodate all characters in the given ranges +STBTT_DEF int stbtt_PackFontRangesRenderIntoRects(stbtt_pack_context *spc, stbtt_fontinfo *info, stbtt_pack_range *ranges, int num_ranges, stbrp_rect *rects) +{ + int i,j,k, 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) { + 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; + } 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, 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, 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_GetPackedQuad(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; + 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; +} + + +////////////////////////////////////////////////////////////////////////////// +// +// 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(const stbtt_uint8 *s1, stbtt_int32 len1, const 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; +} + +STBTT_DEF int stbtt_CompareUTF8toUTF16_bigendian(const char *s1, int len1, const char *s2, int len2) +{ + return len1 == stbtt__CompareUTF8toUTF16_bigendian_prefix((const stbtt_uint8*) s1, len1, (const 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((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; +} + +STBTT_DEF int stbtt_FindMatchingFont(const unsigned char *font_collection, const 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; + } +} + +#endif // STB_TRUETYPE_IMPLEMENTATION + + +// FULL VERSION HISTORY +// +// 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 +//