1 /* $Id: tif_lzw.c,v 1.45 2011-04-02 20:54:09 bfriesen Exp $ */
4 * Copyright (c) 1988-1997 Sam Leffler
5 * Copyright (c) 1991-1997 Silicon Graphics, Inc.
7 * Permission to use, copy, modify, distribute, and sell this software and
8 * its documentation for any purpose is hereby granted without fee, provided
9 * that (i) the above copyright notices and this permission notice appear in
10 * all copies of the software and related documentation, and (ii) the names of
11 * Sam Leffler and Silicon Graphics may not be used in any advertising or
12 * publicity relating to the software without the specific, prior written
13 * permission of Sam Leffler and Silicon Graphics.
15 * THE SOFTWARE IS PROVIDED "AS-IS" AND WITHOUT WARRANTY OF ANY KIND,
16 * EXPRESS, IMPLIED OR OTHERWISE, INCLUDING WITHOUT LIMITATION, ANY
17 * WARRANTY OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
19 * IN NO EVENT SHALL SAM LEFFLER OR SILICON GRAPHICS BE LIABLE FOR
20 * ANY SPECIAL, INCIDENTAL, INDIRECT OR CONSEQUENTIAL DAMAGES OF ANY KIND,
21 * OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
22 * WHETHER OR NOT ADVISED OF THE POSSIBILITY OF DAMAGE, AND ON ANY THEORY OF
23 * LIABILITY, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
31 * Rev 5.0 Lempel-Ziv & Welch Compression Support
33 * This code is derived from the compress program whose code is
34 * derived from software contributed to Berkeley by James A. Woods,
35 * derived from original work by Spencer Thomas and Joseph Orost.
37 * The original Berkeley copyright notice appears below in its entirety.
39 #include "tif_predict.h"
44 * NB: The 5.0 spec describes a different algorithm than Aldus
45 * implements. Specifically, Aldus does code length transitions
46 * one code earlier than should be done (for real LZW).
47 * Earlier versions of this library implemented the correct
48 * LZW algorithm, but emitted codes in a bit order opposite
49 * to the TIFF spec. Thus, to maintain compatibility w/ Aldus
50 * we interpret MSB-LSB ordered codes to be images written w/
51 * old versions of this library, but otherwise adhere to the
52 * Aldus "off by one" algorithm.
54 * Future revisions to the TIFF spec are expected to "clarify this issue".
56 #define LZW_COMPAT /* include backwards compatibility code */
58 * Each strip of data is supposed to be terminated by a CODE_EOI.
59 * If the following #define is included, the decoder will also
60 * check for end-of-strip w/o seeing this code. This makes the
61 * library more robust, but also slower.
63 #define LZW_CHECKEOS /* include checks for strips w/o EOI code */
65 #define MAXCODE(n) ((1L<<(n))-1)
67 * The TIFF spec specifies that encoded bit
68 * strings range from 9 to 12 bits.
70 #define BITS_MIN 9 /* start with 9 bits */
71 #define BITS_MAX 12 /* max of 12 bit strings */
72 /* predefined codes */
73 #define CODE_CLEAR 256 /* code to clear string table */
74 #define CODE_EOI 257 /* end-of-information code */
75 #define CODE_FIRST 258 /* first free code entry */
76 #define CODE_MAX MAXCODE(BITS_MAX)
77 #define HSIZE 9001L /* 91% occupancy */
80 /* NB: +1024 is for compatibility with old files */
81 #define CSIZE (MAXCODE(BITS_MAX)+1024L)
83 #define CSIZE (MAXCODE(BITS_MAX)+1L)
87 * State block for each open TIFF file using LZW
88 * compression/decompression. Note that the predictor
89 * state block must be first in this data structure.
92 TIFFPredictorState predict; /* predictor super class */
94 unsigned short nbits; /* # of bits/code */
95 unsigned short maxcode; /* maximum code for lzw_nbits */
96 unsigned short free_ent; /* next free entry in hash table */
97 long nextdata; /* next bits of i/o */
98 long nextbits; /* # of valid bits in lzw_nextdata */
100 int rw_mode; /* preserve rw_mode from init */
103 #define lzw_nbits base.nbits
104 #define lzw_maxcode base.maxcode
105 #define lzw_free_ent base.free_ent
106 #define lzw_nextdata base.nextdata
107 #define lzw_nextbits base.nextbits
110 * Encoding-specific state.
112 typedef uint16 hcode_t; /* codes fit in 16 bits */
119 * Decoding-specific state.
121 typedef struct code_ent {
122 struct code_ent *next;
123 unsigned short length; /* string len, including this token */
124 unsigned char value; /* data value */
125 unsigned char firstchar; /* first token of string */
128 typedef int (*decodeFunc)(TIFF*, uint8*, tmsize_t, uint16);
133 /* Decoding specific data */
134 long dec_nbitsmask; /* lzw_nbits 1 bits, right adjusted */
135 long dec_restart; /* restart count */
137 uint64 dec_bitsleft; /* available bits in raw data */
139 decodeFunc dec_decode; /* regular or backwards compatible */
140 code_t* dec_codep; /* current recognized code */
141 code_t* dec_oldcodep; /* previously recognized code */
142 code_t* dec_free_entp; /* next free entry */
143 code_t* dec_maxcodep; /* max available entry */
144 code_t* dec_codetab; /* kept separate for small machines */
146 /* Encoding specific data */
147 int enc_oldcode; /* last code encountered */
148 long enc_checkpoint; /* point at which to clear table */
149 #define CHECK_GAP 10000 /* enc_ratio check interval */
150 long enc_ratio; /* current compression ratio */
151 long enc_incount; /* (input) data bytes encoded */
152 long enc_outcount; /* encoded (output) bytes */
153 uint8* enc_rawlimit; /* bound on tif_rawdata buffer */
154 hash_t* enc_hashtab; /* kept separate for small machines */
157 #define LZWState(tif) ((LZWBaseState*) (tif)->tif_data)
158 #define DecoderState(tif) ((LZWCodecState*) LZWState(tif))
159 #define EncoderState(tif) ((LZWCodecState*) LZWState(tif))
161 static int LZWDecode(TIFF* tif, uint8* op0, tmsize_t occ0, uint16 s);
163 static int LZWDecodeCompat(TIFF* tif, uint8* op0, tmsize_t occ0, uint16 s);
165 static void cl_hash(LZWCodecState*);
173 * This check shouldn't be necessary because each
174 * strip is suppose to be terminated with CODE_EOI.
176 #define NextCode(_tif, _sp, _bp, _code, _get) { \
177 if ((_sp)->dec_bitsleft < (uint64)nbits) { \
178 TIFFWarningExt(_tif->tif_clientdata, module, \
179 "LZWDecode: Strip %d not terminated with EOI code", \
180 _tif->tif_curstrip); \
183 _get(_sp,_bp,_code); \
184 (_sp)->dec_bitsleft -= nbits; \
188 #define NextCode(tif, sp, bp, code, get) get(sp, bp, code)
192 LZWFixupTags(TIFF* tif)
199 LZWSetupDecode(TIFF* tif)
201 static const char module[] = "LZWSetupDecode";
202 LZWCodecState* sp = DecoderState(tif);
208 * Allocate state block so tag methods have storage to record
211 tif->tif_data = (uint8*) _TIFFmalloc(sizeof(LZWCodecState));
212 if (tif->tif_data == NULL)
214 TIFFErrorExt(tif->tif_clientdata, module, "No space for LZW state block");
218 DecoderState(tif)->dec_codetab = NULL;
219 DecoderState(tif)->dec_decode = NULL;
222 * Setup predictor setup.
224 (void) TIFFPredictorInit(tif);
226 sp = DecoderState(tif);
231 if (sp->dec_codetab == NULL) {
232 sp->dec_codetab = (code_t*)_TIFFmalloc(CSIZE*sizeof (code_t));
233 if (sp->dec_codetab == NULL) {
234 TIFFErrorExt(tif->tif_clientdata, module,
235 "No space for LZW code table");
239 * Pre-load the table.
243 sp->dec_codetab[code].value = code;
244 sp->dec_codetab[code].firstchar = code;
245 sp->dec_codetab[code].length = 1;
246 sp->dec_codetab[code].next = NULL;
249 * Zero-out the unused entries
251 _TIFFmemset(&sp->dec_codetab[CODE_CLEAR], 0,
252 (CODE_FIRST - CODE_CLEAR) * sizeof (code_t));
258 * Setup state for decoding a strip.
261 LZWPreDecode(TIFF* tif, uint16 s)
263 static const char module[] = "LZWPreDecode";
264 LZWCodecState *sp = DecoderState(tif);
268 if( sp->dec_codetab == NULL )
270 tif->tif_setupdecode( tif );
274 * Check for old bit-reversed codes.
276 if (tif->tif_rawdata[0] == 0 && (tif->tif_rawdata[1] & 0x1)) {
278 if (!sp->dec_decode) {
279 TIFFWarningExt(tif->tif_clientdata, module,
280 "Old-style LZW codes, convert file");
282 * Override default decoding methods with
283 * ones that deal with the old coding.
284 * Otherwise the predictor versions set
285 * above will call the compatibility routines
286 * through the dec_decode method.
288 tif->tif_decoderow = LZWDecodeCompat;
289 tif->tif_decodestrip = LZWDecodeCompat;
290 tif->tif_decodetile = LZWDecodeCompat;
292 * If doing horizontal differencing, must
293 * re-setup the predictor logic since we
294 * switched the basic decoder methods...
296 (*tif->tif_setupdecode)(tif);
297 sp->dec_decode = LZWDecodeCompat;
299 sp->lzw_maxcode = MAXCODE(BITS_MIN);
300 #else /* !LZW_COMPAT */
301 if (!sp->dec_decode) {
302 TIFFErrorExt(tif->tif_clientdata, module,
303 "Old-style LZW codes not supported");
304 sp->dec_decode = LZWDecode;
307 #endif/* !LZW_COMPAT */
309 sp->lzw_maxcode = MAXCODE(BITS_MIN)-1;
310 sp->dec_decode = LZWDecode;
312 sp->lzw_nbits = BITS_MIN;
313 sp->lzw_nextbits = 0;
314 sp->lzw_nextdata = 0;
317 sp->dec_nbitsmask = MAXCODE(BITS_MIN);
319 sp->dec_bitsleft = ((uint64)tif->tif_rawcc) << 3;
321 sp->dec_free_entp = sp->dec_codetab + CODE_FIRST;
323 * Zero entries that are not yet filled in. We do
324 * this to guard against bogus input data that causes
325 * us to index into undefined entries. If you can
326 * come up with a way to safely bounds-check input codes
327 * while decoding then you can remove this operation.
329 _TIFFmemset(sp->dec_free_entp, 0, (CSIZE-CODE_FIRST)*sizeof (code_t));
330 sp->dec_oldcodep = &sp->dec_codetab[-1];
331 sp->dec_maxcodep = &sp->dec_codetab[sp->dec_nbitsmask-1];
336 * Decode a "hunk of data".
338 #define GetNextCode(sp, bp, code) { \
339 nextdata = (nextdata<<8) | *(bp)++; \
341 if (nextbits < nbits) { \
342 nextdata = (nextdata<<8) | *(bp)++; \
345 code = (hcode_t)((nextdata >> (nextbits-nbits)) & nbitsmask); \
350 codeLoop(TIFF* tif, const char* module)
352 TIFFErrorExt(tif->tif_clientdata, module,
353 "Bogus encoding, loop in the code table; scanline %d",
358 LZWDecode(TIFF* tif, uint8* op0, tmsize_t occ0, uint16 s)
360 static const char module[] = "LZWDecode";
361 LZWCodecState *sp = DecoderState(tif);
362 char *op = (char*) op0;
363 long occ = (long) occ0;
368 long nbits, nextbits, nextdata, nbitsmask;
369 code_t *codep, *free_entp, *maxcodep, *oldcodep;
373 assert(sp->dec_codetab != NULL);
376 Fail if value does not fit in long.
378 if ((tmsize_t) occ != occ0)
381 * Restart interrupted output operation.
383 if (sp->dec_restart) {
386 codep = sp->dec_codep;
387 residue = codep->length - sp->dec_restart;
390 * Residue from previous decode is sufficient
391 * to satisfy decode request. Skip to the
392 * start of the decoded string, place decoded
393 * values in the output buffer, and return.
395 sp->dec_restart += occ;
398 } while (--residue > occ && codep);
402 *--tp = codep->value;
404 } while (--occ && codep);
409 * Residue satisfies only part of the decode request.
411 op += residue, occ -= residue;
419 } while (--residue && codep);
423 bp = (unsigned char *)tif->tif_rawcp;
424 nbits = sp->lzw_nbits;
425 nextdata = sp->lzw_nextdata;
426 nextbits = sp->lzw_nextbits;
427 nbitsmask = sp->dec_nbitsmask;
428 oldcodep = sp->dec_oldcodep;
429 free_entp = sp->dec_free_entp;
430 maxcodep = sp->dec_maxcodep;
433 NextCode(tif, sp, bp, code, GetNextCode);
434 if (code == CODE_EOI)
436 if (code == CODE_CLEAR) {
437 free_entp = sp->dec_codetab + CODE_FIRST;
438 _TIFFmemset(free_entp, 0,
439 (CSIZE - CODE_FIRST) * sizeof (code_t));
441 nbitsmask = MAXCODE(BITS_MIN);
442 maxcodep = sp->dec_codetab + nbitsmask-1;
443 NextCode(tif, sp, bp, code, GetNextCode);
444 if (code == CODE_EOI)
446 if (code >= CODE_CLEAR) {
447 TIFFErrorExt(tif->tif_clientdata, tif->tif_name,
448 "LZWDecode: Corrupted LZW table at scanline %d",
452 *op++ = (char)code, occ--;
453 oldcodep = sp->dec_codetab + code;
456 codep = sp->dec_codetab + code;
459 * Add the new entry to the code table.
461 if (free_entp < &sp->dec_codetab[0] ||
462 free_entp >= &sp->dec_codetab[CSIZE]) {
463 TIFFErrorExt(tif->tif_clientdata, module,
464 "Corrupted LZW table at scanline %d",
469 free_entp->next = oldcodep;
470 if (free_entp->next < &sp->dec_codetab[0] ||
471 free_entp->next >= &sp->dec_codetab[CSIZE]) {
472 TIFFErrorExt(tif->tif_clientdata, module,
473 "Corrupted LZW table at scanline %d",
477 free_entp->firstchar = free_entp->next->firstchar;
478 free_entp->length = free_entp->next->length+1;
479 free_entp->value = (codep < free_entp) ?
480 codep->firstchar : free_entp->firstchar;
481 if (++free_entp > maxcodep) {
482 if (++nbits > BITS_MAX) /* should not happen */
484 nbitsmask = MAXCODE(nbits);
485 maxcodep = sp->dec_codetab + nbitsmask-1;
490 * Code maps to a string, copy string
491 * value to output (written in reverse).
493 if(codep->length == 0) {
494 TIFFErrorExt(tif->tif_clientdata, module,
495 "Wrong length of decoded string: "
496 "data probably corrupted at scanline %d",
500 if (codep->length > occ) {
502 * String is too long for decode buffer,
503 * locate portion that will fit, copy to
504 * the decode buffer, and setup restart
505 * logic for the next decoding call.
507 sp->dec_codep = codep;
510 } while (codep && codep->length > occ);
512 sp->dec_restart = (long)occ;
515 *--tp = codep->value;
517 } while (--occ && codep);
519 codeLoop(tif, module);
531 } while (codep && tp > op);
533 codeLoop(tif, module);
537 op += len, occ -= len;
539 *op++ = (char)code, occ--;
542 tif->tif_rawcp = (uint8*) bp;
543 sp->lzw_nbits = (unsigned short) nbits;
544 sp->lzw_nextdata = nextdata;
545 sp->lzw_nextbits = nextbits;
546 sp->dec_nbitsmask = nbitsmask;
547 sp->dec_oldcodep = oldcodep;
548 sp->dec_free_entp = free_entp;
549 sp->dec_maxcodep = maxcodep;
552 #if defined(__WIN32__) && (defined(_MSC_VER) || defined(__MINGW32__))
553 TIFFErrorExt(tif->tif_clientdata, module,
554 "Not enough data at scanline %d (short %I64d bytes)",
555 tif->tif_row, (unsigned __int64) occ);
557 TIFFErrorExt(tif->tif_clientdata, module,
558 "Not enough data at scanline %d (short %llu bytes)",
559 tif->tif_row, (unsigned long long) occ);
568 * Decode a "hunk of data" for old images.
570 #define GetNextCodeCompat(sp, bp, code) { \
571 nextdata |= (unsigned long) *(bp)++ << nextbits; \
573 if (nextbits < nbits) { \
574 nextdata |= (unsigned long) *(bp)++ << nextbits;\
577 code = (hcode_t)(nextdata & nbitsmask); \
578 nextdata >>= nbits; \
583 LZWDecodeCompat(TIFF* tif, uint8* op0, tmsize_t occ0, uint16 s)
585 static const char module[] = "LZWDecodeCompat";
586 LZWCodecState *sp = DecoderState(tif);
587 char *op = (char*) op0;
588 long occ = (long) occ0;
592 long nextbits, nextdata, nbitsmask;
593 code_t *codep, *free_entp, *maxcodep, *oldcodep;
599 Fail if value does not fit in long.
601 if ((tmsize_t) occ != occ0)
605 * Restart interrupted output operation.
607 if (sp->dec_restart) {
610 codep = sp->dec_codep;
611 residue = codep->length - sp->dec_restart;
614 * Residue from previous decode is sufficient
615 * to satisfy decode request. Skip to the
616 * start of the decoded string, place decoded
617 * values in the output buffer, and return.
619 sp->dec_restart += occ;
622 } while (--residue > occ);
625 *--tp = codep->value;
631 * Residue satisfies only part of the decode request.
633 op += residue, occ -= residue;
636 *--tp = codep->value;
642 bp = (unsigned char *)tif->tif_rawcp;
643 nbits = sp->lzw_nbits;
644 nextdata = sp->lzw_nextdata;
645 nextbits = sp->lzw_nextbits;
646 nbitsmask = sp->dec_nbitsmask;
647 oldcodep = sp->dec_oldcodep;
648 free_entp = sp->dec_free_entp;
649 maxcodep = sp->dec_maxcodep;
652 NextCode(tif, sp, bp, code, GetNextCodeCompat);
653 if (code == CODE_EOI)
655 if (code == CODE_CLEAR) {
656 free_entp = sp->dec_codetab + CODE_FIRST;
657 _TIFFmemset(free_entp, 0,
658 (CSIZE - CODE_FIRST) * sizeof (code_t));
660 nbitsmask = MAXCODE(BITS_MIN);
661 maxcodep = sp->dec_codetab + nbitsmask;
662 NextCode(tif, sp, bp, code, GetNextCodeCompat);
663 if (code == CODE_EOI)
665 if (code >= CODE_CLEAR) {
666 TIFFErrorExt(tif->tif_clientdata, tif->tif_name,
667 "LZWDecode: Corrupted LZW table at scanline %d",
672 oldcodep = sp->dec_codetab + code;
675 codep = sp->dec_codetab + code;
678 * Add the new entry to the code table.
680 if (free_entp < &sp->dec_codetab[0] ||
681 free_entp >= &sp->dec_codetab[CSIZE]) {
682 TIFFErrorExt(tif->tif_clientdata, module,
683 "Corrupted LZW table at scanline %d", tif->tif_row);
687 free_entp->next = oldcodep;
688 if (free_entp->next < &sp->dec_codetab[0] ||
689 free_entp->next >= &sp->dec_codetab[CSIZE]) {
690 TIFFErrorExt(tif->tif_clientdata, module,
691 "Corrupted LZW table at scanline %d", tif->tif_row);
694 free_entp->firstchar = free_entp->next->firstchar;
695 free_entp->length = free_entp->next->length+1;
696 free_entp->value = (codep < free_entp) ?
697 codep->firstchar : free_entp->firstchar;
698 if (++free_entp > maxcodep) {
699 if (++nbits > BITS_MAX) /* should not happen */
701 nbitsmask = MAXCODE(nbits);
702 maxcodep = sp->dec_codetab + nbitsmask;
707 * Code maps to a string, copy string
708 * value to output (written in reverse).
710 if(codep->length == 0) {
711 TIFFErrorExt(tif->tif_clientdata, module,
712 "Wrong length of decoded "
713 "string: data probably corrupted at scanline %d",
717 if (codep->length > occ) {
719 * String is too long for decode buffer,
720 * locate portion that will fit, copy to
721 * the decode buffer, and setup restart
722 * logic for the next decoding call.
724 sp->dec_codep = codep;
727 } while (codep->length > occ);
728 sp->dec_restart = occ;
731 *--tp = codep->value;
736 assert(occ >= codep->length);
737 op += codep->length, occ -= codep->length;
740 *--tp = codep->value;
741 } while( (codep = codep->next) != NULL );
746 tif->tif_rawcp = (uint8*) bp;
747 sp->lzw_nbits = nbits;
748 sp->lzw_nextdata = nextdata;
749 sp->lzw_nextbits = nextbits;
750 sp->dec_nbitsmask = nbitsmask;
751 sp->dec_oldcodep = oldcodep;
752 sp->dec_free_entp = free_entp;
753 sp->dec_maxcodep = maxcodep;
756 #if defined(__WIN32__) && (defined(_MSC_VER) || defined(__MINGW32__))
757 TIFFErrorExt(tif->tif_clientdata, module,
758 "Not enough data at scanline %d (short %I64d bytes)",
759 tif->tif_row, (unsigned __int64) occ);
761 TIFFErrorExt(tif->tif_clientdata, module,
762 "Not enough data at scanline %d (short %llu bytes)",
763 tif->tif_row, (unsigned long long) occ);
769 #endif /* LZW_COMPAT */
776 LZWSetupEncode(TIFF* tif)
778 static const char module[] = "LZWSetupEncode";
779 LZWCodecState* sp = EncoderState(tif);
782 sp->enc_hashtab = (hash_t*) _TIFFmalloc(HSIZE*sizeof (hash_t));
783 if (sp->enc_hashtab == NULL) {
784 TIFFErrorExt(tif->tif_clientdata, module,
785 "No space for LZW hash table");
792 * Reset encoding state at the start of a strip.
795 LZWPreEncode(TIFF* tif, uint16 s)
797 LZWCodecState *sp = EncoderState(tif);
802 if( sp->enc_hashtab == NULL )
804 tif->tif_setupencode( tif );
807 sp->lzw_nbits = BITS_MIN;
808 sp->lzw_maxcode = MAXCODE(BITS_MIN);
809 sp->lzw_free_ent = CODE_FIRST;
810 sp->lzw_nextbits = 0;
811 sp->lzw_nextdata = 0;
812 sp->enc_checkpoint = CHECK_GAP;
815 sp->enc_outcount = 0;
817 * The 4 here insures there is space for 2 max-sized
818 * codes in LZWEncode and LZWPostDecode.
820 sp->enc_rawlimit = tif->tif_rawdata + tif->tif_rawdatasize-1 - 4;
821 cl_hash(sp); /* clear hash table */
822 sp->enc_oldcode = (hcode_t) -1; /* generates CODE_CLEAR in LZWEncode */
826 #define CALCRATIO(sp, rat) { \
827 if (incount > 0x007fffff) { /* NB: shift will overflow */\
828 rat = outcount >> 8; \
829 rat = (rat == 0 ? 0x7fffffff : incount/rat); \
831 rat = (incount<<8) / outcount; \
833 #define PutNextCode(op, c) { \
834 nextdata = (nextdata << nbits) | c; \
836 *op++ = (unsigned char)(nextdata >> (nextbits-8)); \
838 if (nextbits >= 8) { \
839 *op++ = (unsigned char)(nextdata >> (nextbits-8)); \
846 * Encode a chunk of pixels.
848 * Uses an open addressing double hashing (no chaining) on the
849 * prefix code/next character combination. We do a variant of
850 * Knuth's algorithm D (vol. 3, sec. 6.4) along with G. Knott's
851 * relatively-prime secondary probe. Here, the modular division
852 * first probe is gives way to a faster exclusive-or manipulation.
853 * Also do block compression with an adaptive reset, whereby the
854 * code table is cleared when the compression ratio decreases,
855 * but after the table fills. The variable-length output codes
856 * are re-sized at this point, and a CODE_CLEAR is generated
860 LZWEncode(TIFF* tif, uint8* bp, tmsize_t cc, uint16 s)
862 register LZWCodecState *sp = EncoderState(tif);
868 long incount, outcount, checkpoint;
869 long nextdata, nextbits;
870 int free_ent, maxcode, nbits;
878 assert(sp->enc_hashtab != NULL);
883 incount = sp->enc_incount;
884 outcount = sp->enc_outcount;
885 checkpoint = sp->enc_checkpoint;
886 nextdata = sp->lzw_nextdata;
887 nextbits = sp->lzw_nextbits;
888 free_ent = sp->lzw_free_ent;
889 maxcode = sp->lzw_maxcode;
890 nbits = sp->lzw_nbits;
892 limit = sp->enc_rawlimit;
893 ent = sp->enc_oldcode;
895 if (ent == (hcode_t) -1 && cc > 0) {
897 * NB: This is safe because it can only happen
898 * at the start of a strip where we know there
899 * is space in the data buffer.
901 PutNextCode(op, CODE_CLEAR);
902 ent = *bp++; cc--; incount++;
905 c = *bp++; cc--; incount++;
906 fcode = ((long)c << BITS_MAX) + ent;
907 h = (c << HSHIFT) ^ ent; /* xor hashing */
910 * Check hash index for an overflow.
915 hp = &sp->enc_hashtab[h];
916 if (hp->hash == fcode) {
922 * Primary hash failed, check secondary hash.
929 * Avoid pointer arithmetic 'cuz of
930 * wraparound problems with segments.
934 hp = &sp->enc_hashtab[h];
935 if (hp->hash == fcode) {
939 } while (hp->hash >= 0);
942 * New entry, emit code and add to table.
945 * Verify there is space in the buffer for the code
946 * and any potential Clear code that might be emitted
947 * below. The value of limit is setup so that there
948 * are at least 4 bytes free--room for 2 codes.
951 tif->tif_rawcc = (tmsize_t)(op - tif->tif_rawdata);
953 op = tif->tif_rawdata;
955 PutNextCode(op, ent);
957 hp->code = free_ent++;
959 if (free_ent == CODE_MAX-1) {
960 /* table is full, emit clear code and reset */
965 free_ent = CODE_FIRST;
966 PutNextCode(op, CODE_CLEAR);
968 maxcode = MAXCODE(BITS_MIN);
971 * If the next entry is going to be too big for
972 * the code size, then increase it, if possible.
974 if (free_ent > maxcode) {
976 assert(nbits <= BITS_MAX);
977 maxcode = (int) MAXCODE(nbits);
978 } else if (incount >= checkpoint) {
981 * Check compression ratio and, if things seem
982 * to be slipping, clear the hash table and
983 * reset state. The compression ratio is a
984 * 24+8-bit fractional number.
986 checkpoint = incount+CHECK_GAP;
988 if (rat <= sp->enc_ratio) {
993 free_ent = CODE_FIRST;
994 PutNextCode(op, CODE_CLEAR);
996 maxcode = MAXCODE(BITS_MIN);
1006 * Restore global state.
1008 sp->enc_incount = incount;
1009 sp->enc_outcount = outcount;
1010 sp->enc_checkpoint = checkpoint;
1011 sp->enc_oldcode = ent;
1012 sp->lzw_nextdata = nextdata;
1013 sp->lzw_nextbits = nextbits;
1014 sp->lzw_free_ent = free_ent;
1015 sp->lzw_maxcode = maxcode;
1016 sp->lzw_nbits = nbits;
1017 tif->tif_rawcp = op;
1022 * Finish off an encoded strip by flushing the last
1023 * string and tacking on an End Of Information code.
1026 LZWPostEncode(TIFF* tif)
1028 register LZWCodecState *sp = EncoderState(tif);
1029 uint8* op = tif->tif_rawcp;
1030 long nextbits = sp->lzw_nextbits;
1031 long nextdata = sp->lzw_nextdata;
1032 long outcount = sp->enc_outcount;
1033 int nbits = sp->lzw_nbits;
1035 if (op > sp->enc_rawlimit) {
1036 tif->tif_rawcc = (tmsize_t)(op - tif->tif_rawdata);
1037 TIFFFlushData1(tif);
1038 op = tif->tif_rawdata;
1040 if (sp->enc_oldcode != (hcode_t) -1) {
1041 PutNextCode(op, sp->enc_oldcode);
1042 sp->enc_oldcode = (hcode_t) -1;
1044 PutNextCode(op, CODE_EOI);
1046 *op++ = (unsigned char)(nextdata << (8-nextbits));
1047 tif->tif_rawcc = (tmsize_t)(op - tif->tif_rawdata);
1052 * Reset encoding hash table.
1055 cl_hash(LZWCodecState* sp)
1057 register hash_t *hp = &sp->enc_hashtab[HSIZE-1];
1058 register long i = HSIZE-8;
1072 for (i += 8; i > 0; i--, hp--)
1077 LZWCleanup(TIFF* tif)
1079 (void)TIFFPredictorCleanup(tif);
1081 assert(tif->tif_data != 0);
1083 if (DecoderState(tif)->dec_codetab)
1084 _TIFFfree(DecoderState(tif)->dec_codetab);
1086 if (EncoderState(tif)->enc_hashtab)
1087 _TIFFfree(EncoderState(tif)->enc_hashtab);
1089 _TIFFfree(tif->tif_data);
1090 tif->tif_data = NULL;
1092 _TIFFSetDefaultCompressionState(tif);
1096 TIFFInitLZW(TIFF* tif, int scheme)
1098 static const char module[] = "TIFFInitLZW";
1099 assert(scheme == COMPRESSION_LZW);
1101 * Allocate state block so tag methods have storage to record values.
1103 tif->tif_data = (uint8*) _TIFFmalloc(sizeof (LZWCodecState));
1104 if (tif->tif_data == NULL)
1106 DecoderState(tif)->dec_codetab = NULL;
1107 DecoderState(tif)->dec_decode = NULL;
1108 EncoderState(tif)->enc_hashtab = NULL;
1109 LZWState(tif)->rw_mode = tif->tif_mode;
1112 * Install codec methods.
1114 tif->tif_fixuptags = LZWFixupTags;
1115 tif->tif_setupdecode = LZWSetupDecode;
1116 tif->tif_predecode = LZWPreDecode;
1117 tif->tif_decoderow = LZWDecode;
1118 tif->tif_decodestrip = LZWDecode;
1119 tif->tif_decodetile = LZWDecode;
1120 tif->tif_setupencode = LZWSetupEncode;
1121 tif->tif_preencode = LZWPreEncode;
1122 tif->tif_postencode = LZWPostEncode;
1123 tif->tif_encoderow = LZWEncode;
1124 tif->tif_encodestrip = LZWEncode;
1125 tif->tif_encodetile = LZWEncode;
1126 tif->tif_cleanup = LZWCleanup;
1128 * Setup predictor setup.
1130 (void) TIFFPredictorInit(tif);
1133 TIFFErrorExt(tif->tif_clientdata, module,
1134 "No space for LZW state block");
1139 * Copyright (c) 1985, 1986 The Regents of the University of California.
1140 * All rights reserved.
1142 * This code is derived from software contributed to Berkeley by
1143 * James A. Woods, derived from original work by Spencer Thomas
1146 * Redistribution and use in source and binary forms are permitted
1147 * provided that the above copyright notice and this paragraph are
1148 * duplicated in all such forms and that any documentation,
1149 * advertising materials, and other materials related to such
1150 * distribution and use acknowledge that the software was developed
1151 * by the University of California, Berkeley. The name of the
1152 * University may not be used to endorse or promote products derived
1153 * from this software without specific prior written permission.
1154 * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR
1155 * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED
1156 * WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR A PARTICULAR PURPOSE.
1158 #endif /* LZW_SUPPORT */
1160 /* vim: set ts=8 sts=8 sw=8 noet: */