sen/skcblitz
Archived
1
0
Fork 0
Code Activity
This repository has been archived on 2025-09-02. You can view files and clone it, but you cannot make any changes to it's state, such as pushing and creating new issues, pull requests or comments.
skcblitz/stb_timage.h
Sen c9e78fd810
first and last commit
2025-09-02 14:43:36 +02:00

2112 lines
76 KiB
C
Raw Permalink Blame History

// stb_image 2.28_tiny + stb_image_write 1.16_tiny - modified by sen - public domain image loader + writer - http://nothings.org/stb
// #ifndef STBI_STDMEM
#define STBI_MALLOC(s) mem_alloc(s, MEM_IMAGE)
#define STBI_REALLOC(p, s) mem_realloc(p, s, MEM_IMAGE)
#define STBI_FREE(p) mem_free(p)
// #else
// #define STBI_MALLOC(s) malloc(s)
// #define STBI_REALLOC(p, s) realloc(p, s)
// #define STBI_FREE(p) free(p)
// #endif
#define STBI_MEMMOVE(a,b,sz) memmove(a,b,sz)
#define STBI_MAX_DIMENSIONS (1 << 24)
// fast-way is faster to check than jpeg huffman, but slow way is slower
#define STBI__ZFAST_BITS 9 // accelerate all cases in default tables
#define STBI__ZFAST_MASK ((1 << STBI__ZFAST_BITS) - 1)
#define STBI__ZNSYMS 288 // number of symbols in literal/length alphabet
// stbi__err - error
// stbi__errpuc - error returning pointer to byte
#define stbi__errpuc(x,y) ((byte *)(ulong) (stbi__err(x,y)?NULL:NULL))
#define STBI__BYTECAST(x) ((byte) ((x) & 255)) // truncate int to byte without warnings
#define STBI__PNG_TYPE(a,b,c,d) (((unsigned) (a) << 24) + ((unsigned) (b) << 16) + ((unsigned) (c) << 8) + (unsigned) (d))
#define stbiw__wpng4(o,a,b,c,d) ((o)[0]=STBI__BYTECAST(a),(o)[1]=STBI__BYTECAST(b),(o)[2]=STBI__BYTECAST(c),(o)[3]=STBI__BYTECAST(d),(o)+=4)
#define stbiw__wp32(data,v) stbiw__wpng4(data, (v)>>24,(v)>>16,(v)>>8,(v));
#define stbiw__wptag(data,s) stbiw__wpng4(data, s[0],s[1],s[2],s[3])
// stretchy buffer; stbiw__sbpush() == vector<>::push_back() -- stbiw__sbcount() == vector<>::size()
#define stbiw__sbraw(a) ((int *) (void *) (a) - 2)
#define stbiw__sbm(a) stbiw__sbraw(a)[0]
#define stbiw__sbn(a) stbiw__sbraw(a)[1]
#define stbiw__sbneedgrow(a,n) ((a)==0 || stbiw__sbn(a)+n >= stbiw__sbm(a))
#define stbiw__sbmaybegrow(a,n) (stbiw__sbneedgrow(a,(n)) ? stbiw__sbgrow(a,n) : 0)
#define stbiw__sbgrow(a,n) stbiw__sbgrowf((void **) &(a), (n), sizeof(*(a)))
#define stbiw__sbpush(a, v) (stbiw__sbmaybegrow(a,1), (a)[stbiw__sbn(a)++] = (v))
#define stbiw__sbcount(a) ((a) ? stbiw__sbn(a) : 0)
#define stbiw__sbfree(a) ((a) ? STBI_FREE(stbiw__sbraw(a)),0 : 0)
#define stbiw__zlib_flush() (out = stbiw__zlib_flushf(out, &bitbuf, &bitcount))
#define stbiw__zlib_add(code,codebits) \
(bitbuf |= (code) << bitcount, bitcount += (codebits), stbiw__zlib_flush())
#define stbiw__zlib_huffa(b,c) stbiw__zlib_add(stbiw__zlib_bitrev(b,c),c)
// default huffman tables
#define stbiw__zlib_huff1(n) stbiw__zlib_huffa(0x30 + (n), 8)
#define stbiw__zlib_huff2(n) stbiw__zlib_huffa(0x190 + (n)-144, 9)
#define stbiw__zlib_huff3(n) stbiw__zlib_huffa(0 + (n)-256,7)
#define stbiw__zlib_huff4(n) stbiw__zlib_huffa(0xc0 + (n)-280,8)
#define stbiw__zlib_huff(n) ((n) <= 143 ? stbiw__zlib_huff1(n) : (n) <= 255 ? stbiw__zlib_huff2(n) : (n) <= 279 ? stbiw__zlib_huff3(n) : stbiw__zlib_huff4(n))
#define stbiw__zlib_huffb(n) ((n) <= 143 ? stbiw__zlib_huff1(n) : stbiw__zlib_huff2(n))
#define stbiw__ZHASH 16384
typedef struct
{
int (*read) (void *user,char *data,int size); // fill 'data' with 'size' bytes. return number of bytes actually read
void (*skip) (void *user,int n); // skip the next 'n' bytes, or 'unget' the last -n bytes if negative
int (*eof) (void *user); // returns nonzero if we are at end of file/data
} stbi_io_callbacks;
// stbi__context structure is our basic context used by all images, so it
// contains all the IO context, plus some basic image information
typedef struct
{
uint img_x, img_y;
int img_n, img_out_n;
stbi_io_callbacks io;
void *io_user_data;
int read_from_callbacks;
int buflen;
byte buffer_start[128];
int callback_already_read;
byte *img_buffer, *img_buffer_end;
byte *img_buffer_original, *img_buffer_original_end;
} stbi__context;
typedef struct
{
int bits_per_channel;
int num_channels;
} stbi__result_info;
// zlib-style huffman encoding
// (jpegs packs from left, zlib from right, so can't share code)
typedef struct
{
ushort fast[1 << STBI__ZFAST_BITS];
ushort firstcode[16];
int maxcode[17];
ushort firstsymbol[16];
byte size[STBI__ZNSYMS];
ushort value[STBI__ZNSYMS];
} stbi__zhuffman;
typedef struct
{
byte *zbuffer, *zbuffer_end;
int num_bits;
uint code_buffer;
char *zout;
char *zout_start;
char *zout_end;
int z_expandable;
stbi__zhuffman z_length, z_distance;
} stbi__zbuf;
typedef struct
{
uint length;
uint type;
} stbi__pngchunk;
typedef struct
{
stbi__context *s;
byte *idata, *expanded, *out;
int depth;
} stbi__png;
static const char *stbi__g_failure_reason;
static const char *stbi__g_failure_desc;
static int stbi__unpremultiply_on_load = 0;
static int stbi__de_iphone_flag = 0;
static int stbi_write_png_compression_level = 8;
static int stbi_write_force_png_filter = -1;
// get a VERY brief reason for failure
static const char *stbi_failure_reason(void)
{
return stbi__g_failure_reason;
}
// get a VERY brief reason for failure (user message)
static const char *stbi_failure_desc(void)
{
return stbi__g_failure_desc;
}
// for image formats that explicitly notate that they have premultiplied alpha,
// we just return the colors as stored in the file. set this flag to force
// unpremultiplication. results are undefined if the unpremultiply overflow.
static void stbi_set_unpremultiply_on_load(int should_unpremultiply)
{
stbi__unpremultiply_on_load = should_unpremultiply;
}
// indicate whether we should process iphone images back to canonical format,
// or just pass them through "as-is"
static void stbi_convert_iphone_png_to_rgb(int should_convert)
{
stbi__de_iphone_flag = should_convert;
}
static void stbi_set_png_compression_level(int png_compression_level)
{
stbi_write_png_compression_level = png_compression_level;
}
static void stbi_set_force_png_filter(int force_png_filter)
{
stbi_write_force_png_filter = force_png_filter;
}
static int stbi__err(const char *reason, const char *desc)
{
stbi__g_failure_reason = reason;
stbi__g_failure_desc = desc;
return 0;
}
// ZLIB client - used by PNG, available for other purposes
static char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen);
static char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header);
static char *stbi_zlib_decode_malloc(const char *buffer, int len, int *outlen);
static int stbi_zlib_decode_buffer(char *obuffer, int olen, const char *ibuffer, int ilen);
static char *stbi_zlib_decode_noheader_malloc(const char *buffer, int len, int *outlen);
static int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen);
static void stbi__refill_buffer(stbi__context *s);
// initialize a memory-decode context
static void stbi__start_mem(stbi__context *s, byte const *buffer, int len)
{
s->io.read = NULL;
s->read_from_callbacks = 0;
s->callback_already_read = 0;
s->img_buffer = s->img_buffer_original = (byte *) buffer;
s->img_buffer_end = s->img_buffer_original_end = (byte *) buffer+len;
}
// initialize a callback-based context
static void stbi__start_callbacks(stbi__context *s, stbi_io_callbacks *c, void *user)
{
s->io = *c;
s->io_user_data = user;
s->buflen = sizeof(s->buffer_start);
s->read_from_callbacks = 1;
s->callback_already_read = 0;
s->img_buffer = s->img_buffer_original = s->buffer_start;
stbi__refill_buffer(s);
s->img_buffer_original_end = s->img_buffer_end;
}
static int stbi__stdio_read(void *user, char *data, int size)
{
return (int) fread(data,1,size,(FILE*) user);
}
static void stbi__stdio_skip(void *user, int n)
{
int ch;
fseek((FILE*) user, n, SEEK_CUR);
ch = fgetc((FILE*) user); /* have to read a byte to reset feof()'s flag */
if (ch != EOF) {
ungetc(ch, (FILE *) user); /* push byte back onto stream if valid. */
}
}
static int stbi__stdio_eof(void *user)
{
return feof((FILE*) user) || ferror((FILE *) user);
}
static stbi_io_callbacks stbi__stdio_callbacks =
{
stbi__stdio_read,
stbi__stdio_skip,
stbi__stdio_eof,
};
static void stbi__start_file(stbi__context *s, FILE *f)
{
stbi__start_callbacks(s, &stbi__stdio_callbacks, (void *) f);
}
//static void stop_file(stbi__context *s) { }
static void stbi__rewind(stbi__context *s)
{
// conceptually rewind SHOULD rewind to the beginning of the stream,
// but we just rewind to the beginning of the initial buffer, because
// we only use it after doing 'test', which only ever looks at at most 92 bytes
s->img_buffer = s->img_buffer_original;
s->img_buffer_end = s->img_buffer_original_end;
}
static int stbi__png_test(stbi__context *s);
static void *stbi__png_load(stbi__context *s, int *x, int *y, stbi__result_info *ri);
static void *stbi__malloc(ulong size)
{
return STBI_MALLOC(size);
}
// stb_image uses ints pervasively, including for offset calculations.
// therefore the largest decoded image size we can support with the
// current code, even on 64-bit targets, is INT_MAX. this is not a
// significant limitation for the intended use case.
//
// we do, however, need to make sure our size calculations don't
// overflow. hence a few helper functions for size calculations that
// multiply integers together, making sure that they're non-negative
// and no overflow occurs.
// return 1 if the sum is valid, 0 on overflow.
// negative terms are considered invalid.
static int stbi__addsizes_valid(int a, int b)
{
if (b < 0) return 0;
// now 0 <= b <= INT_MAX, hence also
// 0 <= INT_MAX - b <= INTMAX.
// And "a + b <= INT_MAX" (which might overflow) is the
// same as a <= INT_MAX - b (no overflow)
return a <= INT_MAX - b;
}
// returns 1 if the product is valid, 0 on overflow.
// negative factors are considered invalid.
static int stbi__mul2sizes_valid(int a, int b)
{
if (a < 0 || b < 0) return 0;
if (b == 0) return 1; // mul-by-0 is always safe
// portable way to check for no overflows in a*b
return a <= INT_MAX/b;
}
// returns 1 if "a*b + add" has no negative terms/factors and doesn't overflow
static int stbi__mad2sizes_valid(int a, int b, int add)
{
return stbi__mul2sizes_valid(a, b) && stbi__addsizes_valid(a*b, add);
}
// returns 1 if "a*b*c + add" has no negative terms/factors and doesn't overflow
static int stbi__mad3sizes_valid(int a, int b, int c, int add)
{
return stbi__mul2sizes_valid(a, b) && stbi__mul2sizes_valid(a*b, c) &&
stbi__addsizes_valid(a*b*c, add);
}
// mallocs with size overflow checking
static void *stbi__malloc_mad2(int a, int b, int add)
{
if (!stbi__mad2sizes_valid(a, b, add)) return NULL;
return stbi__malloc(a*b + add);
}
static void *stbi__malloc_mad3(int a, int b, int c, int add)
{
if (!stbi__mad3sizes_valid(a, b, c, add)) return NULL;
return stbi__malloc(a*b*c + add);
}
static void *stbi__load_main(stbi__context *s, int *x, int *y, stbi__result_info *ri)
{
memset(ri, 0, sizeof(*ri)); // make sure it's initialized if we add new fields
ri->bits_per_channel = 8; // default is 8 so most paths don't have to be changed
ri->num_channels = 0;
if (stbi__png_test(s)) return stbi__png_load(s,x,y, ri);
return stbi__errpuc("unknown image type", "Image not of any known type, or corrupt");
}
static byte *stbi__convert_16_to_8(ushort *orig, int w, int h)
{
int i;
int img_len = w * h * 4;
byte *reduced;
reduced = (byte *) stbi__malloc(img_len);
if (reduced == NULL) return stbi__errpuc("outofmem", "Out of memory");
for (i = 0; i < img_len; ++i)
reduced[i] = (byte)((orig[i] >> 8) & 0xFF); // top half of each byte is sufficient approx of 16->8 bit scaling
STBI_FREE(orig);
return reduced;
}
static void stbi__vertical_flip(void *image, int w, int h)
{
int row;
ulong bytes_per_row = (ulong)w * 4 * sizeof(byte);
byte temp[2048];
byte *bytes = (byte *)image;
for (row = 0; row < (h>>1); row++) {
byte *row0 = bytes + row*bytes_per_row;
byte *row1 = bytes + (h - row - 1)*bytes_per_row;
// swap row0 with row1
ulong bytes_left = bytes_per_row;
while (bytes_left) {
ulong bytes_copy = (bytes_left < sizeof(temp)) ? bytes_left : sizeof(temp);
memcpy(temp, row0, bytes_copy);
memcpy(row0, row1, bytes_copy);
memcpy(row1, temp, bytes_copy);
row0 += bytes_copy;
row1 += bytes_copy;
bytes_left -= bytes_copy;
}
}
}
static byte *stbi__load_and_postprocess_8bit(stbi__context *s, int *x, int *y, byte flip)
{
stbi__result_info ri;
void *result = stbi__load_main(s, x, y, &ri);
if (result == NULL)
return NULL;
// it is the responsibility of the loaders to make sure we get either 8 or 16 bit.
sys_assert(ri.bits_per_channel == 8 || ri.bits_per_channel == 16);
if (ri.bits_per_channel != 8) {
result = stbi__convert_16_to_8((ushort *) result, *x, *y);
ri.bits_per_channel = 8;
}
// @TODO: move stbi__convert_format to here
if (flip) {
stbi__vertical_flip(result, *x, *y);
}
return (byte *) result;
}
static byte *stbi_load(FILE *f, int *x, int *y, byte flip)
{
byte *result;
stbi__context s;
stbi__start_file(&s,f);
result = stbi__load_and_postprocess_8bit(&s,x,y, flip);
if (result) {
// need to 'unget' all the characters in the IO buffer
fseek(f, - (int) (s.img_buffer_end - s.img_buffer), SEEK_CUR);
}
return result;
}
static byte *stbi_load_mem(byte const *buffer, int len, int *x, int *y, byte flip)
{
stbi__context s;
stbi__start_mem(&s,buffer,len);
return stbi__load_and_postprocess_8bit(&s,x,y, flip);
}
//////////////////////////////////////////////////////////////////////////////
//
// Common code used by all image loaders
//
static void stbi__refill_buffer(stbi__context *s)
{
int n = (s->io.read)(s->io_user_data,(char*)s->buffer_start,s->buflen);
s->callback_already_read += (int) (s->img_buffer - s->img_buffer_original);
if (n == 0) {
// at end of file, treat same as if from memory, but need to handle case
// where s->img_buffer isn't pointing to safe memory, e.g. 0-byte file
s->read_from_callbacks = 0;
s->img_buffer = s->buffer_start;
s->img_buffer_end = s->buffer_start+1;
*s->img_buffer = 0;
} else {
s->img_buffer = s->buffer_start;
s->img_buffer_end = s->buffer_start + n;
}
}
static byte stbi__get8(stbi__context *s)
{
if (s->img_buffer < s->img_buffer_end)
return *s->img_buffer++;
if (s->read_from_callbacks) {
stbi__refill_buffer(s);
return *s->img_buffer++;
}
return 0;
}
static void stbi__skip(stbi__context *s, int n)
{
if (n == 0) return; // already there!
if (n < 0) {
s->img_buffer = s->img_buffer_end;
return;
}
if (s->io.read) {
int blen = (int) (s->img_buffer_end - s->img_buffer);
if (blen < n) {
s->img_buffer = s->img_buffer_end;
(s->io.skip)(s->io_user_data, n - blen);
return;
}
}
s->img_buffer += n;
}
static int stbi__getn(stbi__context *s, byte *buffer, int n)
{
if (s->io.read) {
int blen = (int) (s->img_buffer_end - s->img_buffer);
if (blen < n) {
int res, count;
memcpy(buffer, s->img_buffer, blen);
count = (s->io.read)(s->io_user_data, (char*) buffer + blen, n - blen);
res = (count == (n-blen));
s->img_buffer = s->img_buffer_end;
return res;
}
}
if (s->img_buffer+n <= s->img_buffer_end) {
memcpy(buffer, s->img_buffer, n);
s->img_buffer += n;
return 1;
} else
return 0;
}
static int stbi__get16be(stbi__context *s)
{
int z = stbi__get8(s);
return (z << 8) + stbi__get8(s);
}
static uint stbi__get32be(stbi__context *s)
{
uint z = stbi__get16be(s);
return (z << 16) + stbi__get16be(s);
}
//////////////////////////////////////////////////////////////////////////////
//
// generic converter from built-in img_n to req_comp
// individual types do this automatically as much as possible (e.g. jpeg
// does all cases internally since it needs to colorspace convert anyway,
// and it never has alpha, so very few cases ). png can automatically
// interleave an alpha=255 channel, but falls back to this for other cases
//
// assume data buffer is malloced, so malloc a new one and free that one
// only failure mode is malloc failing
static byte stbi__compute_y(int r, int g, int b)
{
return (byte) (((r*77) + (g*150) + (29*b)) >> 8);
}
static byte *stbi__convert_format(byte *data, int img_n, uint x, uint y)
{
int i,j;
byte *good;
if (4 == img_n) return data;
good = (byte *) stbi__malloc_mad3(4, x, y, 0);
if (good == NULL) {
STBI_FREE(data);
return stbi__errpuc("outofmem", "Out of memory");
}
for (j=0; j < (int) y; ++j) {
byte *src = data + j * x * img_n ;
byte *dest = good + j * x * 4;
#define STBI__COMBO(a,b) ((a)*8+(b))
#define STBI__CASE(a,b) case STBI__COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b)
// convert source image with img_n components to one with 4 components;
// avoid switch per pixel, so use switch per scanline and massive macros
switch (STBI__COMBO(img_n, 4)) {
STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=255; } break;
STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0]; } break;
STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=255; } break;
STBI__CASE(2,1) { dest[0]=src[0]; } break;
STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0]; } break;
STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1]; } break;
STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=255; } break;
STBI__CASE(3,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); } break;
STBI__CASE(3,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = 255; } break;
STBI__CASE(4,1) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); } break;
STBI__CASE(4,2) { dest[0]=stbi__compute_y(src[0],src[1],src[2]); dest[1] = src[3]; } break;
STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2]; } break;
default: sys_assert(0); STBI_FREE(data); STBI_FREE(good); return stbi__errpuc("unsupported", "Unsupported format conversion");
}
#undef STBI__CASE
}
STBI_FREE(data);
return good;
}
static ushort stbi__compute_y_16(int r, int g, int b)
{
return (ushort) (((r*77) + (g*150) + (29*b)) >> 8);
}
static ushort *stbi__convert_format16(ushort *data, int img_n, uint x, uint y)
{
int i,j;
ushort *good;
if (4 == img_n) return data;
good = (ushort *) stbi__malloc(4 * x * y * 2);
if (good == NULL) {
STBI_FREE(data);
return (ushort *) stbi__errpuc("outofmem", "Out of memory");
}
for (j=0; j < (int) y; ++j) {
ushort *src = data + j * x * img_n ;
ushort *dest = good + j * x * 4;
#define STBI__COMBO(a,b) ((a)*8+(b))
#define STBI__CASE(a,b) case STBI__COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b)
// convert source image with img_n components to one with 4 components;
// avoid switch per pixel, so use switch per scanline and massive macros
switch (STBI__COMBO(img_n, 4)) {
STBI__CASE(1,2) { dest[0]=src[0]; dest[1]=0xffff; } break;
STBI__CASE(1,3) { dest[0]=dest[1]=dest[2]=src[0]; } break;
STBI__CASE(1,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=0xffff; } break;
STBI__CASE(2,1) { dest[0]=src[0]; } break;
STBI__CASE(2,3) { dest[0]=dest[1]=dest[2]=src[0]; } break;
STBI__CASE(2,4) { dest[0]=dest[1]=dest[2]=src[0]; dest[3]=src[1]; } break;
STBI__CASE(3,4) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2];dest[3]=0xffff; } break;
STBI__CASE(3,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); } break;
STBI__CASE(3,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = 0xffff; } break;
STBI__CASE(4,1) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); } break;
STBI__CASE(4,2) { dest[0]=stbi__compute_y_16(src[0],src[1],src[2]); dest[1] = src[3]; } break;
STBI__CASE(4,3) { dest[0]=src[0];dest[1]=src[1];dest[2]=src[2]; } break;
default: sys_assert(0); STBI_FREE(data); STBI_FREE(good); return (ushort*) stbi__errpuc("unsupported", "Unsupported format conversion");
}
#undef STBI__CASE
}
STBI_FREE(data);
return good;
}
// 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
static int stbi__bitreverse16(int n)
{
n = ((n & 0xAAAA) >> 1) | ((n & 0x5555) << 1);
n = ((n & 0xCCCC) >> 2) | ((n & 0x3333) << 2);
n = ((n & 0xF0F0) >> 4) | ((n & 0x0F0F) << 4);
n = ((n & 0xFF00) >> 8) | ((n & 0x00FF) << 8);
return n;
}
static int stbi__bit_reverse(int v, int bits)
{
sys_assert(bits <= 16);
// to bit reverse n bits, reverse 16 and shift
// e.g. 11 bits, bit reverse and shift away 5
return stbi__bitreverse16(v) >> (16-bits);
}
static int stbi__zbuild_huffman(stbi__zhuffman *z, const byte *sizelist, int num)
{
int i,k=0;
int code, next_code[16], sizes[17];
// DEFLATE spec for generating codes
memset(sizes, 0, sizeof(sizes));
memset(z->fast, 0, sizeof(z->fast));
for (i=0; i < num; ++i)
++sizes[sizelist[i]];
sizes[0] = 0;
for (i=1; i < 16; ++i)
if (sizes[i] > (1 << i))
return stbi__err("bad sizes", "Corrupt PNG");
code = 0;
for (i=1; i < 16; ++i) {
next_code[i] = code;
z->firstcode[i] = (ushort) code;
z->firstsymbol[i] = (ushort) k;
code = (code + sizes[i]);
if (sizes[i])
if (code-1 >= (1 << i)) return stbi__err("bad codelengths","Corrupt PNG");
z->maxcode[i] = code << (16-i); // preshift for inner loop
code <<= 1;
k += sizes[i];
}
z->maxcode[16] = 0x10000; // sentinel
for (i=0; i < num; ++i) {
int s = sizelist[i];
if (s) {
int c = next_code[s] - z->firstcode[s] + z->firstsymbol[s];
ushort fastv = (ushort) ((s << 9) | i);
z->size [c] = (byte ) s;
z->value[c] = (ushort) i;
if (s <= STBI__ZFAST_BITS) {
int j = stbi__bit_reverse(next_code[s],s);
while (j < (1 << STBI__ZFAST_BITS)) {
z->fast[j] = fastv;
j += (1 << s);
}
}
++next_code[s];
}
}
return 1;
}
// zlib-from-memory implementation for PNG reading
// because PNG allows splitting the zlib stream arbitrarily,
// and it's annoying structurally to have PNG call ZLIB call PNG,
// we require PNG read all the IDATs and combine them into a single
// memory buffer
static int stbi__zeof(stbi__zbuf *z)
{
return (z->zbuffer >= z->zbuffer_end);
}
static byte stbi__zget8(stbi__zbuf *z)
{
return stbi__zeof(z) ? 0 : *z->zbuffer++;
}
static void stbi__fill_bits(stbi__zbuf *z)
{
do {
if (z->code_buffer >= (1U << z->num_bits)) {
z->zbuffer = z->zbuffer_end; /* treat this as EOF so we fail. */
return;
}
z->code_buffer |= (uint) stbi__zget8(z) << z->num_bits;
z->num_bits += 8;
} while (z->num_bits <= 24);
}
static uint stbi__zreceive(stbi__zbuf *z, int n)
{
uint k;
if (z->num_bits < n) stbi__fill_bits(z);
k = z->code_buffer & ((1 << n) - 1);
z->code_buffer >>= n;
z->num_bits -= n;
return k;
}
static int stbi__zhuffman_decode_slowpath(stbi__zbuf *a, stbi__zhuffman *z)
{
int b,s,k;
// not resolved by fast table, so compute it the slow way
// use jpeg approach, which requires MSbits at top
k = stbi__bit_reverse(a->code_buffer, 16);
for (s=STBI__ZFAST_BITS+1; ; ++s)
if (k < z->maxcode[s])
break;
if (s >= 16) return -1; // invalid code!
// code size is s, so:
b = (k >> (16-s)) - z->firstcode[s] + z->firstsymbol[s];
if (b >= STBI__ZNSYMS) return -1; // some data was corrupt somewhere!
if (z->size[b] != s) return -1; // was originally an sys_assert, but report failure instead.
a->code_buffer >>= s;
a->num_bits -= s;
return z->value[b];
}
static int stbi__zhuffman_decode(stbi__zbuf *a, stbi__zhuffman *z)
{
int b,s;
if (a->num_bits < 16) {
if (stbi__zeof(a)) {
return -1; /* report error for unexpected end of data. */
}
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;
uint cur, limit; // , old_limit;
z->zout = zout;
if (!z->z_expandable) return stbi__err("output buffer limit","Corrupt PNG");
cur = (uint) (z->zout - z->zout_start);
limit = /* old_limit = */ (unsigned) (z->zout_end - z->zout_start);
if (UINT_MAX - cur < (unsigned) n) return stbi__err("outofmem", "Out of memory");
while (cur + n > limit) {
if(limit > UINT_MAX / 2) return stbi__err("outofmem", "Out of memory");
limit *= 2;
}
q = (char *) STBI_REALLOC(z->zout_start, limit);
// STBI_NOTUSED(old_limit);
if (q == NULL) return stbi__err("outofmem", "Out of memory");
z->zout_start = q;
z->zout = q + cur;
z->zout_end = q + limit;
return 1;
}
static const int stbi__zlength_base[31] = {
3,4,5,6,7,8,9,10,11,13,
15,17,19,23,27,31,35,43,51,59,
67,83,99,115,131,163,195,227,258,0,0 };
static const int stbi__zlength_extra[31]=
{ 0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0,0,0 };
static const int stbi__zdist_base[32] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193,
257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577,0,0};
static const int stbi__zdist_extra[32] =
{ 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13};
static int stbi__parse_huffman_block(stbi__zbuf *a)
{
char *zout = a->zout;
for(;;) {
int z = stbi__zhuffman_decode(a, &a->z_length);
if (z < 256) {
if (z < 0) return stbi__err("bad huffman code","Corrupt PNG"); // error in huffman codes
if (zout >= a->zout_end) {
if (!stbi__zexpand(a, zout, 1)) return 0;
zout = a->zout;
}
*zout++ = (char) z;
} else {
byte *p;
int len,dist;
if (z == 256) {
a->zout = zout;
return 1;
}
if (z >= 286) return stbi__err("bad huffman code","Corrupt PNG"); // per DEFLATE, length codes 286 and 287 must not appear in compressed data
z -= 257;
len = stbi__zlength_base[z];
if (stbi__zlength_extra[z]) len += stbi__zreceive(a, stbi__zlength_extra[z]);
z = stbi__zhuffman_decode(a, &a->z_distance);
if (z < 0 || z >= 30) return stbi__err("bad huffman code","Corrupt PNG"); // per DEFLATE, distance codes 30 and 31 must not appear in compressed data
dist = stbi__zdist_base[z];
if (stbi__zdist_extra[z]) dist += stbi__zreceive(a, stbi__zdist_extra[z]);
if (zout - a->zout_start < dist) return stbi__err("bad dist","Corrupt PNG");
if (zout + len > a->zout_end) {
if (!stbi__zexpand(a, zout, len)) return 0;
zout = a->zout;
}
p = (byte *) (zout - dist);
if (dist == 1) { // run of one byte; common in images.
byte v = *p;
if (len) { do *zout++ = v; while (--len); }
} else {
if (len) { do *zout++ = *p++; while (--len); }
}
}
}
}
static int stbi__compute_huffman_codes(stbi__zbuf *a)
{
static const byte length_dezigzag[19] = { 16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15 };
stbi__zhuffman z_codelength;
byte lencodes[286+32+137];//padding for maximum single op
byte codelength_sizes[19];
int i,n;
int hlit = stbi__zreceive(a,5) + 257;
int hdist = stbi__zreceive(a,5) + 1;
int hclen = stbi__zreceive(a,4) + 4;
int ntot = hlit + hdist;
memset(codelength_sizes, 0, sizeof(codelength_sizes));
for (i=0; i < hclen; ++i) {
int s = stbi__zreceive(a,3);
codelength_sizes[length_dezigzag[i]] = (byte) s;
}
if (!stbi__zbuild_huffman(&z_codelength, codelength_sizes, 19)) return 0;
n = 0;
while (n < ntot) {
int c = stbi__zhuffman_decode(a, &z_codelength);
if (c < 0 || c >= 19) return stbi__err("bad codelengths", "Corrupt PNG");
if (c < 16)
lencodes[n++] = (byte) c;
else {
byte fill = 0;
if (c == 16) {
c = stbi__zreceive(a,2)+3;
if (n == 0) return stbi__err("bad codelengths", "Corrupt PNG");
fill = lencodes[n-1];
} else if (c == 17) {
c = stbi__zreceive(a,3)+3;
} else if (c == 18) {
c = stbi__zreceive(a,7)+11;
} else {
return stbi__err("bad codelengths", "Corrupt PNG");
}
if (ntot - n < c) return stbi__err("bad codelengths", "Corrupt PNG");
memset(lencodes+n, fill, c);
n += c;
}
}
if (n != ntot) return stbi__err("bad codelengths","Corrupt PNG");
if (!stbi__zbuild_huffman(&a->z_length, lencodes, hlit)) return 0;
if (!stbi__zbuild_huffman(&a->z_distance, lencodes+hlit, hdist)) return 0;
return 1;
}
static int stbi__parse_uncompressed_block(stbi__zbuf *a)
{
byte header[4];
int len,nlen,k;
if (a->num_bits & 7)
stbi__zreceive(a, a->num_bits & 7); // discard
// drain the bit-packed data into header
k = 0;
while (a->num_bits > 0) {
header[k++] = (byte) (a->code_buffer & 255); // suppress MSVC run-time check
a->code_buffer >>= 8;
a->num_bits -= 8;
}
if (a->num_bits < 0) return stbi__err("zlib corrupt","Corrupt PNG");
// now fill header the normal way
while (k < 4)
header[k++] = stbi__zget8(a);
len = header[1] * 256 + header[0];
nlen = header[3] * 256 + header[2];
if (nlen != (len ^ 0xffff)) return stbi__err("zlib corrupt","Corrupt PNG");
if (a->zbuffer + len > a->zbuffer_end) return stbi__err("read past buffer","Corrupt PNG");
if (a->zout + len > a->zout_end)
if (!stbi__zexpand(a, a->zout, len)) return 0;
memcpy(a->zout, a->zbuffer, len);
a->zbuffer += len;
a->zout += len;
return 1;
}
static int stbi__parse_zlib_header(stbi__zbuf *a)
{
int cmf = stbi__zget8(a);
int cm = cmf & 15;
/* int cinfo = cmf >> 4; */
int flg = stbi__zget8(a);
if (stbi__zeof(a)) return stbi__err("bad zlib header","Corrupt PNG"); // zlib spec
if ((cmf*256+flg) % 31 != 0) return stbi__err("bad zlib header","Corrupt PNG"); // zlib spec
if (flg & 32) return stbi__err("no preset dict","Corrupt PNG"); // preset dictionary not allowed in png
if (cm != 8) return stbi__err("bad compression","Corrupt PNG"); // DEFLATE required for png
// window = 1 << (8 + cinfo)... but who cares, we fully buffer output
return 1;
}
static const byte stbi__zdefault_length[STBI__ZNSYMS] =
{
8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,
8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,
9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,
9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,
9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9, 9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,
7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, 7,7,7,7,7,7,7,7,8,8,8,8,8,8,8,8
};
static const byte stbi__zdefault_distance[32] =
{
5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5
};
/*
Init algorithm:
{
int i; // use <= to match clearly with spec
for (i=0; i <= 143; ++i) stbi__zdefault_length[i] = 8;
for ( ; i <= 255; ++i) stbi__zdefault_length[i] = 9;
for ( ; i <= 279; ++i) stbi__zdefault_length[i] = 7;
for ( ; i <= 287; ++i) stbi__zdefault_length[i] = 8;
for (i=0; i <= 31; ++i) stbi__zdefault_distance[i] = 5;
}
*/
static int stbi__parse_zlib(stbi__zbuf *a, int parse_header)
{
int final, type;
if (parse_header)
if (!stbi__parse_zlib_header(a)) return 0;
a->num_bits = 0;
a->code_buffer = 0;
do {
final = stbi__zreceive(a,1);
type = stbi__zreceive(a,2);
if (type == 0) {
if (!stbi__parse_uncompressed_block(a)) return 0;
} else if (type == 3) {
return 0;
} else {
if (type == 1) {
// use fixed code lengths
if (!stbi__zbuild_huffman(&a->z_length , stbi__zdefault_length , STBI__ZNSYMS)) return 0;
if (!stbi__zbuild_huffman(&a->z_distance, stbi__zdefault_distance, 32)) return 0;
} else {
if (!stbi__compute_huffman_codes(a)) return 0;
}
if (!stbi__parse_huffman_block(a)) return 0;
}
} while (!final);
return 1;
}
static int stbi__do_zlib(stbi__zbuf *a, char *obuf, int olen, int exp, int parse_header)
{
a->zout_start = obuf;
a->zout = obuf;
a->zout_end = obuf + olen;
a->z_expandable = exp;
return stbi__parse_zlib(a, parse_header);
}
static char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen)
{
stbi__zbuf a;
char *p = (char *) stbi__malloc(initial_size);
if (p == NULL) return NULL;
a.zbuffer = (byte *) buffer;
a.zbuffer_end = (byte *) buffer + len;
if (stbi__do_zlib(&a, p, initial_size, 1, 1)) {
if (outlen) *outlen = (int) (a.zout - a.zout_start);
return a.zout_start;
} else {
STBI_FREE(a.zout_start);
return NULL;
}
}
static char *stbi_zlib_decode_malloc(char const *buffer, int len, int *outlen)
{
return stbi_zlib_decode_malloc_guesssize(buffer, len, 16384, outlen);
}
static char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header)
{
stbi__zbuf a;
char *p = (char *) stbi__malloc(initial_size);
if (p == NULL) return NULL;
a.zbuffer = (byte *) buffer;
a.zbuffer_end = (byte *) buffer + len;
if (stbi__do_zlib(&a, p, initial_size, 1, parse_header)) {
if (outlen) *outlen = (int) (a.zout - a.zout_start);
return a.zout_start;
} else {
STBI_FREE(a.zout_start);
return NULL;
}
}
static int stbi_zlib_decode_buffer(char *obuffer, int olen, char const *ibuffer, int ilen)
{
stbi__zbuf a;
a.zbuffer = (byte *) ibuffer;
a.zbuffer_end = (byte *) ibuffer + ilen;
if (stbi__do_zlib(&a, obuffer, olen, 0, 1))
return (int) (a.zout - a.zout_start);
else
return -1;
}
static char *stbi_zlib_decode_noheader_malloc(char const *buffer, int len, int *outlen)
{
stbi__zbuf a;
char *p = (char *) stbi__malloc(16384);
if (p == NULL) return NULL;
a.zbuffer = (byte *) buffer;
a.zbuffer_end = (byte *) buffer+len;
if (stbi__do_zlib(&a, p, 16384, 1, 0)) {
if (outlen) *outlen = (int) (a.zout - a.zout_start);
return a.zout_start;
} else {
STBI_FREE(a.zout_start);
return NULL;
}
}
static int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen)
{
stbi__zbuf a;
a.zbuffer = (byte *) ibuffer;
a.zbuffer_end = (byte *) ibuffer + ilen;
if (stbi__do_zlib(&a, obuffer, olen, 0, 0))
return (int) (a.zout - a.zout_start);
else
return -1;
}
// public domain "baseline" PNG decoder v0.10 Sean Barrett 2006-11-18
// simple implementation
// - only 8-bit samples
// - no CRC checking
// - allocates lots of intermediate memory
// - avoids problem of streaming data between subsystems
// - avoids explicit window management
// performance
// - uses stb_zlib, a PD zlib implementation with fast huffman decoding
static stbi__pngchunk stbi__get_chunk_header(stbi__context *s)
{
stbi__pngchunk c;
c.length = stbi__get32be(s);
c.type = stbi__get32be(s);
return c;
}
static int stbi__check_png_header(stbi__context *s)
{
static const byte png_sig[8] = { 137,80,78,71,13,10,26,10 };
int i;
for (i=0; i < 8; ++i)
if (stbi__get8(s) != png_sig[i]) return stbi__err("bad png sig","Not a PNG");
return 1;
}
enum {
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 byte first_row_filter[5] =
{
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)
{
int p = a + b - c;
int pa = abs(p-a);
int pb = abs(p-b);
int pc = abs(p-c);
if (pa <= pb && pa <= pc) return a;
if (pb <= pc) return b;
return c;
}
static const byte 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, byte *raw, uint raw_len, int out_n, uint x, uint y, int depth, int color)
{
int bytes = (depth == 16? 2 : 1);
stbi__context *s = a->s;
uint i,j,stride = x*out_n*bytes;
uint img_len, img_width_bytes;
int k;
int img_n = s->img_n; // copy it into a local for later
int output_bytes = out_n*bytes;
int filter_bytes = img_n*bytes;
int width = x;
sys_assert(out_n == s->img_n || out_n == s->img_n+1);
a->out = (byte *) stbi__malloc_mad3(x, y, output_bytes, 0); // extra bytes to write off the end into
if (!a->out) return stbi__err("outofmem", "Out of memory");
if (!stbi__mad3sizes_valid(img_n, x, depth, 7)) return stbi__err("too large", "Corrupt PNG");
img_width_bytes = (((img_n * x * depth) + 7) >> 3);
img_len = (img_width_bytes + 1) * y;
// we used to check for exact match between raw_len and img_len on non-interlaced PNGs,
// but issue #276 reported a PNG in the wild that had extra data at the end (all zeros),
// so just check for raw_len < img_len always.
if (raw_len < img_len) return stbi__err("not enough pixels","Corrupt PNG");
for (j=0; j < y; ++j) {
byte *cur = a->out + stride*j;
byte *prior;
int filter = *raw++;
if (filter > 4)
return stbi__err("invalid filter","Corrupt PNG");
if (depth < 8) {
if (img_width_bytes > x) return stbi__err("invalid width","Corrupt PNG");
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;
}
prior = cur - stride; // bugfix: need to compute this after 'cur +=' computation above
// if first row, use special filter that doesn't sample previous row
if (j == 0) filter = first_row_filter[filter];
// 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;
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]>>1)); break;
case STBI__F_paeth : cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(0,prior[k],0)); break;
case STBI__F_avg_first : cur[k] = raw[k]; break;
case STBI__F_paeth_first: cur[k] = raw[k]; break;
}
}
if (depth == 8) {
if (img_n != out_n)
cur[img_n] = 255; // first pixel
raw += img_n;
cur += out_n;
prior += out_n;
} else if (depth == 16) {
if (img_n != out_n) {
cur[filter_bytes] = 255; // first pixel top byte
cur[filter_bytes+1] = 255; // first pixel bottom byte
}
raw += filter_bytes;
cur += output_bytes;
prior += output_bytes;
} else {
raw += 1;
cur += 1;
prior += 1;
}
// this is a little gross, so that we don't switch per-pixel or per-component
if (depth < 8 || img_n == out_n) {
int nk = (width - 1)*filter_bytes;
#define STBI__CASE(f) \
case f: \
for (k=0; k < nk; ++k)
switch (filter) {
// "none" filter turns into a memcpy here; make that explicit.
case STBI__F_none: memcpy(cur, raw, nk); break;
STBI__CASE(STBI__F_sub) { cur[k] = STBI__BYTECAST(raw[k] + cur[k-filter_bytes]); } break;
STBI__CASE(STBI__F_up) { cur[k] = STBI__BYTECAST(raw[k] + prior[k]); } break;
STBI__CASE(STBI__F_avg) { cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k-filter_bytes])>>1)); } break;
STBI__CASE(STBI__F_paeth) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],prior[k],prior[k-filter_bytes])); } break;
STBI__CASE(STBI__F_avg_first) { cur[k] = STBI__BYTECAST(raw[k] + (cur[k-filter_bytes] >> 1)); } break;
STBI__CASE(STBI__F_paeth_first) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k-filter_bytes],0,0)); } break;
}
#undef STBI__CASE
raw += nk;
} else {
sys_assert(img_n+1 == out_n);
#define STBI__CASE(f) \
case f: \
for (i=x-1; i >= 1; --i, cur[filter_bytes]=255,raw+=filter_bytes,cur+=output_bytes,prior+=output_bytes) \
for (k=0; k < filter_bytes; ++k)
switch (filter) {
STBI__CASE(STBI__F_none) { cur[k] = raw[k]; } break;
STBI__CASE(STBI__F_sub) { cur[k] = STBI__BYTECAST(raw[k] + cur[k- output_bytes]); } break;
STBI__CASE(STBI__F_up) { cur[k] = STBI__BYTECAST(raw[k] + prior[k]); } break;
STBI__CASE(STBI__F_avg) { cur[k] = STBI__BYTECAST(raw[k] + ((prior[k] + cur[k- output_bytes])>>1)); } break;
STBI__CASE(STBI__F_paeth) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k- output_bytes],prior[k],prior[k- output_bytes])); } break;
STBI__CASE(STBI__F_avg_first) { cur[k] = STBI__BYTECAST(raw[k] + (cur[k- output_bytes] >> 1)); } break;
STBI__CASE(STBI__F_paeth_first) { cur[k] = STBI__BYTECAST(raw[k] + stbi__paeth(cur[k- output_bytes],0,0)); } break;
}
#undef STBI__CASE
// the loop above sets the high byte of the pixels' alpha, but for
// 16 bit png files we also need the low byte set. we'll do that here.
if (depth == 16) {
cur = a->out + stride*j; // start at the beginning of the row again
for (i=0; i < x; ++i,cur+=output_bytes) {
cur[filter_bytes+1] = 255;
}
}
}
}
// 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) {
byte *cur = a->out + stride*j;
byte *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
byte 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 {
sys_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];
}
}
}
}
} else if (depth == 16) {
// force the image data from big-endian to platform-native.
// this is done in a separate pass due to the decoding relying
// on the data being untouched, but could probably be done
// per-line during decode if care is taken.
byte *cur = a->out;
ushort *cur16 = (ushort*)cur;
for(i=0; i < x*y*out_n; ++i,cur16++,cur+=2) {
*cur16 = (cur[0] << 8) | cur[1];
}
}
return 1;
}
static int stbi__create_png_image(stbi__png *a, byte *image_data, uint image_data_len, int out_n, int depth, int color, int interlaced)
{
int bytes = (depth == 16 ? 2 : 1);
int out_bytes = out_n * bytes;
byte *final;
int p;
if (!interlaced)
return stbi__create_png_image_raw(a, image_data, image_data_len, out_n, a->s->img_x, a->s->img_y, depth, color);
// de-interlacing
final = (byte *) stbi__malloc_mad3(a->s->img_x, a->s->img_y, out_bytes, 0);
if (!final) return stbi__err("outofmem", "Out of memory");
for (p=0; p < 7; ++p) {
int xorig[] = { 0,4,0,2,0,1,0 };
int yorig[] = { 0,0,4,0,2,0,1 };
int xspc[] = { 8,8,4,4,2,2,1 };
int yspc[] = { 8,8,8,4,4,2,2 };
int i,j,x,y;
// pass1_x[4] = 0, pass1_x[5] = 1, pass1_x[12] = 1
x = (a->s->img_x - xorig[p] + xspc[p]-1) / xspc[p];
y = (a->s->img_y - yorig[p] + yspc[p]-1) / yspc[p];
if (x && y) {
uint img_len = ((((a->s->img_n * x * depth) + 7) >> 3) + 1) * y;
if (!stbi__create_png_image_raw(a, image_data, image_data_len, out_n, x, y, depth, color)) {
STBI_FREE(final);
return 0;
}
for (j=0; j < y; ++j) {
for (i=0; i < x; ++i) {
int out_y = j*yspc[p]+yorig[p];
int out_x = i*xspc[p]+xorig[p];
memcpy(final + out_y*a->s->img_x*out_bytes + out_x*out_bytes,
a->out + (j*x+i)*out_bytes, out_bytes);
}
}
STBI_FREE(a->out);
image_data += img_len;
image_data_len -= img_len;
}
}
a->out = final;
return 1;
}
static int stbi__compute_transparency(stbi__png *z, byte tc[3], int out_n)
{
stbi__context *s = z->s;
uint i, pixel_count = s->img_x * s->img_y;
byte *p = z->out;
// compute color-based transparency, assuming we've
// already got 255 as the alpha value in the output
sys_assert(out_n == 2 || out_n == 4);
if (out_n == 2) {
for (i=0; i < pixel_count; ++i) {
p[1] = (p[0] == tc[0] ? 0 : 255);
p += 2;
}
} else {
for (i=0; i < pixel_count; ++i) {
if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2])
p[3] = 0;
p += 4;
}
}
return 1;
}
static int stbi__compute_transparency16(stbi__png *z, ushort tc[3], int out_n)
{
stbi__context *s = z->s;
uint i, pixel_count = s->img_x * s->img_y;
ushort *p = (ushort*) z->out;
// compute color-based transparency, assuming we've
// already got 65535 as the alpha value in the output
sys_assert(out_n == 2 || out_n == 4);
if (out_n == 2) {
for (i = 0; i < pixel_count; ++i) {
p[1] = (p[0] == tc[0] ? 0 : 65535);
p += 2;
}
} else {
for (i = 0; i < pixel_count; ++i) {
if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2])
p[3] = 0;
p += 4;
}
}
return 1;
}
static int stbi__expand_png_palette(stbi__png *a, byte *palette, int pal_img_n)
{
uint i, pixel_count = a->s->img_x * a->s->img_y;
byte *p, *temp_out, *orig = a->out;
p = (byte *) stbi__malloc_mad2(pixel_count, pal_img_n, 0);
if (p == NULL) return stbi__err("outofmem", "Out of memory");
// between here and free(out) below, exitting would leak
temp_out = p;
if (pal_img_n == 3) {
for (i=0; i < pixel_count; ++i) {
int n = orig[i]*4;
p[0] = palette[n ];
p[1] = palette[n+1];
p[2] = palette[n+2];
p += 3;
}
} else {
for (i=0; i < pixel_count; ++i) {
int n = orig[i]*4;
p[0] = palette[n ];
p[1] = palette[n+1];
p[2] = palette[n+2];
p[3] = palette[n+3];
p += 4;
}
}
STBI_FREE(a->out);
a->out = temp_out;
// STBI_NOTUSED(len);
return 1;
}
static void stbi__de_iphone(stbi__png *z)
{
stbi__context *s = z->s;
uint i, pixel_count = s->img_x * s->img_y;
byte *p = z->out;
if (s->img_out_n == 3) { // convert bgr to rgb
for (i=0; i < pixel_count; ++i) {
byte t = p[0];
p[0] = p[2];
p[2] = t;
p += 3;
}
} else {
sys_assert(s->img_out_n == 4);
if (stbi__unpremultiply_on_load) {
// convert bgr to rgb and unpremultiply
for (i=0; i < pixel_count; ++i) {
byte a = p[3];
byte t = p[0];
if (a) {
byte half = a / 2;
p[0] = (p[2] * 255 + half) / a;
p[1] = (p[1] * 255 + half) / a;
p[2] = ( t * 255 + half) / a;
} else {
p[0] = p[2];
p[2] = t;
}
p += 4;
}
} else {
// convert bgr to rgb
for (i=0; i < pixel_count; ++i) {
byte t = p[0];
p[0] = p[2];
p[2] = t;
p += 4;
}
}
}
}
static int stbi__parse_png_file(stbi__png *z)
{
byte palette[1024], pal_img_n=0;
byte has_trans=0, tc[3]={0};
ushort tc16[3];
uint ioff=0, idata_limit=0, i, pal_len=0;
int first=1,k,interlace=0, color=0, is_iphone=0;
stbi__context *s = z->s;
z->expanded = NULL;
z->idata = NULL;
z->out = NULL;
if (!stbi__check_png_header(s)) return 0;
// if (scan == STBI__SCAN_type) return 1;
for (;;) {
stbi__pngchunk c = stbi__get_chunk_header(s);
switch (c.type) {
case STBI__PNG_TYPE('C','g','B','I'):
is_iphone = 1;
stbi__skip(s, c.length);
break;
case STBI__PNG_TYPE('I','H','D','R'): {
int comp,filter;
if (!first) return stbi__err("multiple IHDR","Corrupt PNG");
first = 0;
if (c.length != 13) return stbi__err("bad IHDR len","Corrupt PNG");
s->img_x = stbi__get32be(s);
s->img_y = stbi__get32be(s);
if (s->img_y > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)");
if (s->img_x > STBI_MAX_DIMENSIONS) return stbi__err("too large","Very large image (corrupt?)");
z->depth = stbi__get8(s); if (z->depth != 1 && z->depth != 2 && z->depth != 4 && z->depth != 8 && z->depth != 16) return stbi__err("1/2/4/8/16-bit only","PNG not supported: 1/2/4/8/16-bit only");
color = stbi__get8(s); if (color > 6) return stbi__err("bad ctype","Corrupt PNG");
if (color == 3 && z->depth == 16) return stbi__err("bad ctype","Corrupt PNG");
if (color == 3) pal_img_n = 3; else if (color & 1) return stbi__err("bad ctype","Corrupt PNG");
comp = stbi__get8(s); if (comp) return stbi__err("bad comp method","Corrupt PNG");
filter= stbi__get8(s); if (filter) return stbi__err("bad filter method","Corrupt PNG");
interlace = stbi__get8(s); if (interlace>1) return stbi__err("bad interlace method","Corrupt PNG");
if (!s->img_x || !s->img_y) return stbi__err("0-pixel image","Corrupt PNG");
if (!pal_img_n) {
s->img_n = (color & 2 ? 3 : 1) + (color & 4 ? 1 : 0);
if ((1 << 30) / s->img_x / s->img_n < s->img_y) return stbi__err("too large", "Image too large to decode");
} else {
// if paletted, then pal_n is our final components, and
// img_n is # components to decompress/filter.
s->img_n = 1;
if ((1 << 30) / s->img_x / 4 < s->img_y) return stbi__err("too large","Corrupt PNG");
}
// even with SCAN_header, have to scan to see if we have a tRNS
break;
}
case STBI__PNG_TYPE('P','L','T','E'): {
if (first) return stbi__err("first not IHDR", "Corrupt PNG");
if (c.length > 256*3) return stbi__err("invalid PLTE","Corrupt PNG");
pal_len = c.length / 3;
if (pal_len * 3 != c.length) return stbi__err("invalid PLTE","Corrupt PNG");
for (i=0; i < pal_len; ++i) {
palette[i*4+0] = stbi__get8(s);
palette[i*4+1] = stbi__get8(s);
palette[i*4+2] = stbi__get8(s);
palette[i*4+3] = 255;
}
break;
}
case STBI__PNG_TYPE('t','R','N','S'): {
if (first) return stbi__err("first not IHDR", "Corrupt PNG");
if (z->idata) return stbi__err("tRNS after IDAT","Corrupt PNG");
if (pal_img_n) {
// if (scan == STBI__SCAN_header) { s->img_n = 4; return 1; }
if (pal_len == 0) return stbi__err("tRNS before PLTE","Corrupt PNG");
if (c.length > pal_len) return stbi__err("bad tRNS len","Corrupt PNG");
pal_img_n = 4;
for (i=0; i < c.length; ++i)
palette[i*4+3] = stbi__get8(s);
} else {
if (!(s->img_n & 1)) return stbi__err("tRNS with alpha","Corrupt PNG");
if (c.length != (uint) s->img_n*2) return stbi__err("bad tRNS len","Corrupt PNG");
has_trans = 1;
// non-paletted with tRNS = constant alpha. if header-scanning, we can stop now.
// if (scan == STBI__SCAN_header) { ++s->img_n; return 1; }
if (z->depth == 16) {
for (k = 0; k < s->img_n; ++k) tc16[k] = (ushort)stbi__get16be(s); // copy the values as-is
} else {
for (k = 0; k < s->img_n; ++k) tc[k] = (byte)(stbi__get16be(s) & 255) * stbi__depth_scale_table[z->depth]; // non 8-bit images will be larger
}
}
break;
}
case STBI__PNG_TYPE('I','D','A','T'): {
if (first) return stbi__err("first not IHDR", "Corrupt PNG");
if (pal_img_n && !pal_len) return stbi__err("no PLTE","Corrupt PNG");
// if (scan == STBI__SCAN_header) {
// header scan definitely stops at first IDAT
// if (pal_img_n)
// s->img_n = pal_img_n;
// return 1;
// }
if (c.length > (1u << 30)) return stbi__err("IDAT size limit", "IDAT section larger than 2^30 bytes");
if ((int)(ioff + c.length) < (int)ioff) return 0;
if (ioff + c.length > idata_limit) {
// uint idata_limit_old = idata_limit;
byte *p;
if (idata_limit == 0) idata_limit = c.length > 4096 ? c.length : 4096;
while (ioff + c.length > idata_limit)
idata_limit *= 2;
// STBI_NOTUSED(idata_limit_old);
p = (byte *) STBI_REALLOC(z->idata, idata_limit); if (p == NULL) return stbi__err("outofmem", "Out of memory");
z->idata = p;
}
if (!stbi__getn(s, z->idata+ioff,c.length)) return stbi__err("outofdata","Corrupt PNG");
ioff += c.length;
break;
}
case STBI__PNG_TYPE('I','E','N','D'): {
uint raw_len, bpl;
if (first) return stbi__err("first not IHDR", "Corrupt PNG");
// if (scan != STBI__SCAN_load) return 1;
if (z->idata == NULL) return stbi__err("no IDAT","Corrupt PNG");
// initial guess for decoded data size to avoid unnecessary reallocs
bpl = (s->img_x * z->depth + 7) / 8; // bytes per line, per component
raw_len = bpl * s->img_y * s->img_n /* pixels */ + s->img_y /* filter mode per row */;
z->expanded = (byte *) stbi_zlib_decode_malloc_guesssize_headerflag((char *) z->idata, ioff, raw_len, (int *) &raw_len, !is_iphone);
if (z->expanded == NULL) return 0; // zlib should set error
STBI_FREE(z->idata); z->idata = NULL;
if ((4 == s->img_n+1 && !pal_img_n) || has_trans)
s->img_out_n = s->img_n+1;
else
s->img_out_n = s->img_n;
if (!stbi__create_png_image(z, z->expanded, raw_len, s->img_out_n, z->depth, color, interlace)) return 0;
if (has_trans) {
if (z->depth == 16) {
if (!stbi__compute_transparency16(z, tc16, s->img_out_n)) return 0;
} else {
if (!stbi__compute_transparency(z, tc, s->img_out_n)) return 0;
}
}
if (is_iphone && stbi__de_iphone_flag && s->img_out_n > 2)
stbi__de_iphone(z);
if (pal_img_n) {
// pal_img_n == 3 or 4
s->img_n = pal_img_n; // record the actual colors we had
// s->img_out_n = pal_img_n;
s->img_out_n = 4;
if (!stbi__expand_png_palette(z, palette, /* pal_len, */ s->img_out_n))
return 0;
} else if (has_trans) {
// non-paletted image with tRNS -> source image has (constant) alpha
++s->img_n;
}
STBI_FREE(z->expanded); z->expanded = NULL;
// end of PNG chunk, read and skip CRC
stbi__get32be(s);
return 1;
}
default:
// if critical, fail
if (first) return stbi__err("first not IHDR", "Corrupt PNG");
if ((c.type & (1 << 29)) == 0) {
// not threadsafe
static char invalid_chunk[] = "XXXX PNG chunk not known";
invalid_chunk[0] = STBI__BYTECAST(c.type >> 24);
invalid_chunk[1] = STBI__BYTECAST(c.type >> 16);
invalid_chunk[2] = STBI__BYTECAST(c.type >> 8);
invalid_chunk[3] = STBI__BYTECAST(c.type >> 0);
return stbi__err(invalid_chunk, "PNG not supported: unknown PNG chunk type");
}
stbi__skip(s, c.length);
break;
}
// end of PNG chunk, read and skip CRC
stbi__get32be(s);
}
}
static void *stbi__do_png(stbi__png *p, int *x, int *y, stbi__result_info *ri)
{
void *result=NULL;
if (stbi__parse_png_file(p)) {
if (p->depth <= 8)
ri->bits_per_channel = 8;
else if (p->depth == 16)
ri->bits_per_channel = 16;
else
return stbi__errpuc("bad bits_per_channel", "PNG not supported: unsupported color depth");
result = p->out;
p->out = NULL;
if (4 != p->s->img_out_n) {
if (ri->bits_per_channel == 8)
result = stbi__convert_format((byte *) result, p->s->img_out_n, p->s->img_x, p->s->img_y);
else
result = stbi__convert_format16((ushort *) result, p->s->img_out_n, p->s->img_x, p->s->img_y);
p->s->img_out_n = 4;
if (result == NULL) return result;
}
*x = p->s->img_x;
*y = p->s->img_y;
}
STBI_FREE(p->out); p->out = NULL;
STBI_FREE(p->expanded); p->expanded = NULL;
STBI_FREE(p->idata); p->idata = NULL;
return result;
}
static void *stbi__png_load(stbi__context *s, int *x, int *y, stbi__result_info *ri)
{
stbi__png p;
p.s = s;
return stbi__do_png(&p, x,y, ri);
}
static int stbi__png_test(stbi__context *s)
{
int r;
r = stbi__check_png_header(s);
stbi__rewind(s);
return r;
}
//////////////////////////////////////////////////////////////////////////////
//
// PNG writer
//
static void *stbiw__sbgrowf(void **arr, int increment, int itemsize)
{
int m = *arr ? 2*stbiw__sbm(*arr)+increment : increment+1;
void *p = STBI_REALLOC(*arr ? stbiw__sbraw(*arr) : 0, /* *arr ? (stbiw__sbm(*arr)*itemsize + sizeof(int)*2) : 0, */ itemsize * m + sizeof(int)*2);
sys_assert(p);
if (p) {
if (!*arr) ((int *) p)[1] = 0;
*arr = (void *) ((int *) p + 2);
stbiw__sbm(*arr) = m;
}
return *arr;
}
static byte *stbiw__zlib_flushf(byte *data, uint *bitbuffer, int *bitcount)
{
while (*bitcount >= 8) {
stbiw__sbpush(data, STBI__BYTECAST(*bitbuffer));
*bitbuffer >>= 8;
*bitcount -= 8;
}
return data;
}
static int stbiw__zlib_bitrev(int code, int codebits)
{
int res=0;
while (codebits--) {
res = (res << 1) | (code & 1);
code >>= 1;
}
return res;
}
static uint stbiw__zlib_countm(byte *a, byte *b, int limit)
{
int i;
for (i=0; i < limit && i < 258; ++i)
if (a[i] != b[i]) break;
return i;
}
static uint stbiw__zhash(byte *data)
{
uint hash = data[0] + (data[1] << 8) + (data[2] << 16);
hash ^= hash << 3;
hash += hash >> 5;
hash ^= hash << 4;
hash += hash >> 17;
hash ^= hash << 25;
hash += hash >> 6;
return hash;
}
static byte * stbi_zlib_compress(byte *data, int data_len, int *out_len, int quality)
{
static ushort lengthc[] = { 3,4,5,6,7,8,9,10,11,13,15,17,19,23,27,31,35,43,51,59,67,83,99,115,131,163,195,227,258, 259 };
static byte lengtheb[]= { 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 };
static ushort distc[] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193,257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577, 32768 };
static byte disteb[] = { 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13 };
uint bitbuf=0;
int i,j, bitcount=0;
byte *out = NULL;
byte ***hash_table = (byte***) STBI_MALLOC(stbiw__ZHASH * sizeof(byte**));
if (hash_table == NULL)
return NULL;
if (quality < 5) quality = 5;
stbiw__sbpush(out, 0x78); // DEFLATE 32K window
stbiw__sbpush(out, 0x5e); // FLEVEL = 1
stbiw__zlib_add(1,1); // BFINAL = 1
stbiw__zlib_add(1,2); // BTYPE = 1 -- fixed huffman
for (i=0; i < stbiw__ZHASH; ++i)
hash_table[i] = NULL;
i=0;
while (i < data_len-3) {
// hash next 3 bytes of data to be compressed
int h = stbiw__zhash(data+i)&(stbiw__ZHASH-1), best=3;
byte *bestloc = 0;
byte **hlist = hash_table[h];
int n = stbiw__sbcount(hlist);
for (j=0; j < n; ++j) {
if (hlist[j]-data > i-32768) { // if entry lies within window
int d = stbiw__zlib_countm(hlist[j], data+i, data_len-i);
if (d >= best) { best=d; bestloc=hlist[j]; }
}
}
// when hash table entry is too long, delete half the entries
if (hash_table[h] && stbiw__sbn(hash_table[h]) == 2*quality) {
STBI_MEMMOVE(hash_table[h], hash_table[h]+quality, sizeof(hash_table[h][0])*quality);
stbiw__sbn(hash_table[h]) = quality;
}
stbiw__sbpush(hash_table[h],data+i);
if (bestloc) {
// "lazy matching" - check match at *next* byte, and if it's better, do cur byte as literal
h = stbiw__zhash(data+i+1)&(stbiw__ZHASH-1);
hlist = hash_table[h];
n = stbiw__sbcount(hlist);
for (j=0; j < n; ++j) {
if (hlist[j]-data > i-32767) {
int e = stbiw__zlib_countm(hlist[j], data+i+1, data_len-i-1);
if (e > best) { // if next match is better, bail on current match
bestloc = NULL;
break;
}
}
}
}
if (bestloc) {
int d = (int) (data+i - bestloc); // distance back
sys_assert(d <= 32767 && best <= 258);
for (j=0; best > lengthc[j+1]-1; ++j);
stbiw__zlib_huff(j+257);
if (lengtheb[j]) stbiw__zlib_add(best - lengthc[j], lengtheb[j]);
for (j=0; d > distc[j+1]-1; ++j);
stbiw__zlib_add(stbiw__zlib_bitrev(j,5),5);
if (disteb[j]) stbiw__zlib_add(d - distc[j], disteb[j]);
i += best;
} else {
stbiw__zlib_huffb(data[i]);
++i;
}
}
// write out final bytes
for (;i < data_len; ++i)
stbiw__zlib_huffb(data[i]);
stbiw__zlib_huff(256); // end of block
// pad with 0 bits to byte boundary
while (bitcount)
stbiw__zlib_add(0,1);
for (i=0; i < stbiw__ZHASH; ++i)
(void) stbiw__sbfree(hash_table[i]);
STBI_FREE(hash_table);
// store uncompressed instead if compression was worse
if (stbiw__sbn(out) > data_len + 2 + ((data_len+32766)/32767)*5) {
stbiw__sbn(out) = 2; // truncate to DEFLATE 32K window and FLEVEL = 1
for (j = 0; j < data_len;) {
int blocklen = data_len - j;
if (blocklen > 32767) blocklen = 32767;
stbiw__sbpush(out, data_len - j == blocklen); // BFINAL = ?, BTYPE = 0 -- no compression
stbiw__sbpush(out, STBI__BYTECAST(blocklen)); // LEN
stbiw__sbpush(out, STBI__BYTECAST(blocklen >> 8));
stbiw__sbpush(out, STBI__BYTECAST(~blocklen)); // NLEN
stbiw__sbpush(out, STBI__BYTECAST(~blocklen >> 8));
memcpy(out+stbiw__sbn(out), data+j, blocklen);
stbiw__sbn(out) += blocklen;
j += blocklen;
}
}
{
// compute adler32 on input
uint s1=1, s2=0;
int blocklen = (int) (data_len % 5552);
j=0;
while (j < data_len) {
for (i=0; i < blocklen; ++i) { s1 += data[j+i]; s2 += s1; }
s1 %= 65521; s2 %= 65521;
j += blocklen;
blocklen = 5552;
}
stbiw__sbpush(out, STBI__BYTECAST(s2 >> 8));
stbiw__sbpush(out, STBI__BYTECAST(s2));
stbiw__sbpush(out, STBI__BYTECAST(s1 >> 8));
stbiw__sbpush(out, STBI__BYTECAST(s1));
}
*out_len = stbiw__sbn(out);
// make returned pointer freeable
STBI_MEMMOVE(stbiw__sbraw(out), out, *out_len);
return (byte *) stbiw__sbraw(out);
}
static uint stbiw__crc32(byte *buffer, int len)
{
static uint crc_table[256] =
{
0x00000000, 0x77073096, 0xEE0E612C, 0x990951BA, 0x076DC419, 0x706AF48F, 0xE963A535, 0x9E6495A3,
0x0eDB8832, 0x79DCB8A4, 0xE0D5E91E, 0x97D2D988, 0x09B64C2B, 0x7EB17CBD, 0xE7B82D07, 0x90BF1D91,
0x1DB71064, 0x6AB020F2, 0xF3B97148, 0x84BE41DE, 0x1ADAD47D, 0x6DDDE4EB, 0xF4D4B551, 0x83D385C7,
0x136C9856, 0x646BA8C0, 0xFD62F97A, 0x8A65C9EC, 0x14015C4F, 0x63066CD9, 0xFA0F3D63, 0x8D080DF5,
0x3B6E20C8, 0x4C69105E, 0xD56041E4, 0xA2677172, 0x3C03E4D1, 0x4B04D447, 0xD20D85FD, 0xA50AB56B,
0x35B5A8FA, 0x42B2986C, 0xDBBBC9D6, 0xACBCF940, 0x32D86CE3, 0x45DF5C75, 0xDCD60DCF, 0xABD13D59,
0x26D930AC, 0x51DE003A, 0xC8D75180, 0xBFD06116, 0x21B4F4B5, 0x56B3C423, 0xCFBA9599, 0xB8BDA50F,
0x2802B89E, 0x5F058808, 0xC60CD9B2, 0xB10BE924, 0x2F6F7C87, 0x58684C11, 0xC1611DAB, 0xB6662D3D,
0x76DC4190, 0x01DB7106, 0x98D220BC, 0xEFD5102A, 0x71B18589, 0x06B6B51F, 0x9FBFE4A5, 0xE8B8D433,
0x7807C9A2, 0x0F00F934, 0x9609A88E, 0xE10E9818, 0x7F6A0DBB, 0x086D3D2D, 0x91646C97, 0xE6635C01,
0x6B6B51F4, 0x1C6C6162, 0x856530D8, 0xF262004E, 0x6C0695ED, 0x1B01A57B, 0x8208F4C1, 0xF50FC457,
0x65B0D9C6, 0x12B7E950, 0x8BBEB8EA, 0xFCB9887C, 0x62DD1DDF, 0x15DA2D49, 0x8CD37CF3, 0xFBD44C65,
0x4DB26158, 0x3AB551CE, 0xA3BC0074, 0xD4BB30E2, 0x4ADFA541, 0x3DD895D7, 0xA4D1C46D, 0xD3D6F4FB,
0x4369E96A, 0x346ED9FC, 0xAD678846, 0xDA60B8D0, 0x44042D73, 0x33031DE5, 0xAA0A4C5F, 0xDD0D7CC9,
0x5005713C, 0x270241AA, 0xBE0B1010, 0xC90C2086, 0x5768B525, 0x206F85B3, 0xB966D409, 0xCE61E49F,
0x5EDEF90E, 0x29D9C998, 0xB0D09822, 0xC7D7A8B4, 0x59B33D17, 0x2EB40D81, 0xB7BD5C3B, 0xC0BA6CAD,
0xEDB88320, 0x9ABFB3B6, 0x03B6E20C, 0x74B1D29A, 0xEAD54739, 0x9DD277AF, 0x04DB2615, 0x73DC1683,
0xE3630B12, 0x94643B84, 0x0D6D6A3E, 0x7A6A5AA8, 0xE40ECF0B, 0x9309FF9D, 0x0A00AE27, 0x7D079EB1,
0xF00F9344, 0x8708A3D2, 0x1E01F268, 0x6906C2FE, 0xF762575D, 0x806567CB, 0x196C3671, 0x6E6B06E7,
0xFED41B76, 0x89D32BE0, 0x10DA7A5A, 0x67DD4ACC, 0xF9B9DF6F, 0x8EBEEFF9, 0x17B7BE43, 0x60B08ED5,
0xD6D6A3E8, 0xA1D1937E, 0x38D8C2C4, 0x4FDFF252, 0xD1BB67F1, 0xA6BC5767, 0x3FB506DD, 0x48B2364B,
0xD80D2BDA, 0xAF0A1B4C, 0x36034AF6, 0x41047A60, 0xDF60EFC3, 0xA867DF55, 0x316E8EEF, 0x4669BE79,
0xCB61B38C, 0xBC66831A, 0x256FD2A0, 0x5268E236, 0xCC0C7795, 0xBB0B4703, 0x220216B9, 0x5505262F,
0xC5BA3BBE, 0xB2BD0B28, 0x2BB45A92, 0x5CB36A04, 0xC2D7FFA7, 0xB5D0CF31, 0x2CD99E8B, 0x5BDEAE1D,
0x9B64C2B0, 0xEC63F226, 0x756AA39C, 0x026D930A, 0x9C0906A9, 0xEB0E363F, 0x72076785, 0x05005713,
0x95BF4A82, 0xE2B87A14, 0x7BB12BAE, 0x0CB61B38, 0x92D28E9B, 0xE5D5BE0D, 0x7CDCEFB7, 0x0BDBDF21,
0x86D3D2D4, 0xF1D4E242, 0x68DDB3F8, 0x1FDA836E, 0x81BE16CD, 0xF6B9265B, 0x6FB077E1, 0x18B74777,
0x88085AE6, 0xFF0F6A70, 0x66063BCA, 0x11010B5C, 0x8F659EFF, 0xF862AE69, 0x616BFFD3, 0x166CCF45,
0xA00AE278, 0xD70DD2EE, 0x4E048354, 0x3903B3C2, 0xA7672661, 0xD06016F7, 0x4969474D, 0x3E6E77DB,
0xAED16A4A, 0xD9D65ADC, 0x40DF0B66, 0x37D83BF0, 0xA9BCAE53, 0xDEBB9EC5, 0x47B2CF7F, 0x30B5FFE9,
0xBDBDF21C, 0xCABAC28A, 0x53B39330, 0x24B4A3A6, 0xBAD03605, 0xCDD70693, 0x54DE5729, 0x23D967BF,
0xB3667A2E, 0xC4614AB8, 0x5D681B02, 0x2A6F2B94, 0xB40BBE37, 0xC30C8EA1, 0x5A05DF1B, 0x2D02EF8D
};
uint crc = ~0u;
int i;
for (i=0; i < len; ++i)
crc = (crc >> 8) ^ crc_table[buffer[i] ^ (crc & 0xff)];
return ~crc;
}
static void stbiw__wpcrc(byte **data, int len)
{
uint crc = stbiw__crc32(*data - len - 4, len+4);
stbiw__wp32(*data, crc);
}
static byte stbiw__paeth(int a, int b, int c)
{
int p = a + b - c, pa = abs(p-a), pb = abs(p-b), pc = abs(p-c);
if (pa <= pb && pa <= pc) return STBI__BYTECAST(a);
if (pb <= pc) return STBI__BYTECAST(b);
return STBI__BYTECAST(c);
}
// @OPTIMIZE: provide an option that always forces left-predict or paeth predict
static void stbiw__encode_png_line(byte *pixels, int stride_bytes, int width, int height, int y, int n, int filter_type, char *line_buffer, byte flip)
{
static int mapping[] = { 0,1,2,3,4 };
static int firstmap[] = { 0,1,0,5,6 };
int *mymap = (y != 0) ? mapping : firstmap;
int i;
int type = mymap[filter_type];
byte *z = pixels + stride_bytes * (flip ? height-1-y : y);
int signed_stride = flip ? -stride_bytes : stride_bytes;
if (type==0) {
memcpy(line_buffer, z, width*n);
return;
}
// first loop isn't optimized since it's just one pixel
for (i = 0; i < n; ++i) {
switch (type) {
case 1: line_buffer[i] = z[i]; break;
case 2: line_buffer[i] = z[i] - z[i-signed_stride]; break;
case 3: line_buffer[i] = z[i] - (z[i-signed_stride]>>1); break;
case 4: line_buffer[i] = (char) (z[i] - stbiw__paeth(0,z[i-signed_stride],0)); break;
case 5: line_buffer[i] = z[i]; break;
case 6: line_buffer[i] = z[i]; break;
}
}
switch (type) {
case 1: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - z[i-n]; break;
case 2: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - z[i-signed_stride]; break;
case 3: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - ((z[i-n] + z[i-signed_stride])>>1); break;
case 4: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - stbiw__paeth(z[i-n], z[i-signed_stride], z[i-signed_stride-n]); break;
case 5: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - (z[i-n]>>1); break;
case 6: for (i=n; i < width*n; ++i) line_buffer[i] = z[i] - stbiw__paeth(z[i-n], 0,0); break;
}
}
// For PNG, "stride_in_bytes" is the distance in bytes from the first byte of
// a row of pixels to the first byte of the next row of pixels.
static byte *stbi_write_png_to_mem(const byte *pixels, int stride_bytes, int x, int y, int n, int *out_len, byte flip)
{
int force_filter = stbi_write_force_png_filter;
int ctype[5] = { -1, 0, 4, 2, 6 };
byte sig[8] = { 137,80,78,71,13,10,26,10 };
byte *out,*o, *filt, *zlib;
char *line_buffer;
int j,zlen;
if (stride_bytes == 0)
stride_bytes = x * n;
if (force_filter >= 5) {
force_filter = -1;
}
filt = (byte *) STBI_MALLOC((x*n+1) * y); if (!filt) return 0;
line_buffer = (char *) STBI_MALLOC(x * n); if (!line_buffer) { STBI_FREE(filt); return 0; }
for (j=0; j < y; ++j) {
int filter_type;
if (force_filter > -1) {
filter_type = force_filter;
stbiw__encode_png_line((byte*)(pixels), stride_bytes, x, y, j, n, force_filter, line_buffer, flip);
} else { // Estimate the best filter by running through all of them:
int best_filter = 0, best_filter_val = 0x7fffffff, est, i;
for (filter_type = 0; filter_type < 5; filter_type++) {
stbiw__encode_png_line((byte*)(pixels), stride_bytes, x, y, j, n, filter_type, line_buffer, flip);
// Estimate the entropy of the line using this filter; the less, the better.
est = 0;
for (i = 0; i < x*n; ++i) {
est += abs((char) line_buffer[i]);
}
if (est < best_filter_val) {
best_filter_val = est;
best_filter = filter_type;
}
}
if (filter_type != best_filter) { // If the last iteration already got us the best filter, don't redo it
stbiw__encode_png_line((byte*)(pixels), stride_bytes, x, y, j, n, best_filter, line_buffer, flip);
filter_type = best_filter;
}
}
// when we get here, filter_type contains the filter type, and line_buffer contains the data
filt[j*(x*n+1)] = (byte) filter_type;
STBI_MEMMOVE(filt+j*(x*n+1)+1, line_buffer, x*n);
}
STBI_FREE(line_buffer);
zlib = stbi_zlib_compress(filt, y*( x*n+1), &zlen, stbi_write_png_compression_level);
STBI_FREE(filt);
if (!zlib) return 0;
// each tag requires 12 bytes of overhead
out = (byte *) STBI_MALLOC(8 + 12+13 + 12+zlen + 12);
if (!out) return 0;
*out_len = 8 + 12+13 + 12+zlen + 12;
o=out;
STBI_MEMMOVE(o,sig,8); o+= 8;
stbiw__wp32(o, 13); // header length
stbiw__wptag(o, "IHDR");
stbiw__wp32(o, x);
stbiw__wp32(o, y);
*o++ = 8;
*o++ = STBI__BYTECAST(ctype[n]);
*o++ = 0;
*o++ = 0;
*o++ = 0;
stbiw__wpcrc(&o,13);
stbiw__wp32(o, zlen);
stbiw__wptag(o, "IDAT");
STBI_MEMMOVE(o, zlib, zlen);
o += zlen;
STBI_FREE(zlib);
stbiw__wpcrc(&o, zlen);
stbiw__wp32(o,0);
stbiw__wptag(o, "IEND");
stbiw__wpcrc(&o,0);
sys_assert(o == out + *out_len);
return out;
}
View git blame Copy permalink