1 | /* deflate.c -- compress data using the deflation algorithm |
---|
2 | * Copyright (C) 1995-2004 Jean-loup Gailly. |
---|
3 | * For conditions of distribution and use, see copyright notice in zlib.h |
---|
4 | */ |
---|
5 | |
---|
6 | /* |
---|
7 | * ALGORITHM |
---|
8 | * |
---|
9 | * The "deflation" process depends on being able to identify portions |
---|
10 | * of the input text which are identical to earlier input (within a |
---|
11 | * sliding window trailing behind the input currently being processed). |
---|
12 | * |
---|
13 | * The most straightforward technique turns out to be the fastest for |
---|
14 | * most input files: try all possible matches and select the longest. |
---|
15 | * The key feature of this algorithm is that insertions into the string |
---|
16 | * dictionary are very simple and thus fast, and deletions are avoided |
---|
17 | * completely. Insertions are performed at each input character, whereas |
---|
18 | * string matches are performed only when the previous match ends. So it |
---|
19 | * is preferable to spend more time in matches to allow very fast string |
---|
20 | * insertions and avoid deletions. The matching algorithm for small |
---|
21 | * strings is inspired from that of Rabin & Karp. A brute force approach |
---|
22 | * is used to find longer strings when a small match has been found. |
---|
23 | * A similar algorithm is used in comic (by Jan-Mark Wams) and freeze |
---|
24 | * (by Leonid Broukhis). |
---|
25 | * A previous version of this file used a more sophisticated algorithm |
---|
26 | * (by Fiala and Greene) which is guaranteed to run in linear amortized |
---|
27 | * time, but has a larger average cost, uses more memory and is patented. |
---|
28 | * However the F&G algorithm may be faster for some highly redundant |
---|
29 | * files if the parameter max_chain_length (described below) is too large. |
---|
30 | * |
---|
31 | * ACKNOWLEDGEMENTS |
---|
32 | * |
---|
33 | * The idea of lazy evaluation of matches is due to Jan-Mark Wams, and |
---|
34 | * I found it in 'freeze' written by Leonid Broukhis. |
---|
35 | * Thanks to many people for bug reports and testing. |
---|
36 | * |
---|
37 | * REFERENCES |
---|
38 | * |
---|
39 | * Deutsch, L.P.,"DEFLATE Compressed Data Format Specification". |
---|
40 | * Available in http://www.ietf.org/rfc/rfc1951.txt |
---|
41 | * |
---|
42 | * A description of the Rabin and Karp algorithm is given in the book |
---|
43 | * "Algorithms" by R. Sedgewick, Addison-Wesley, p252. |
---|
44 | * |
---|
45 | * Fiala,E.R., and Greene,D.H. |
---|
46 | * Data Compression with Finite Windows, Comm.ACM, 32,4 (1989) 490-595 |
---|
47 | * |
---|
48 | */ |
---|
49 | |
---|
50 | /* @(#) $Id$ */ |
---|
51 | |
---|
52 | #include "deflate.h" |
---|
53 | |
---|
54 | const char deflate_copyright[] = |
---|
55 | " deflate 1.2.2.2 Copyright 1995-2004 Jean-loup Gailly "; |
---|
56 | /* |
---|
57 | If you use the zlib library in a product, an acknowledgment is welcome |
---|
58 | in the documentation of your product. If for some reason you cannot |
---|
59 | include such an acknowledgment, I would appreciate that you keep this |
---|
60 | copyright string in the executable of your product. |
---|
61 | */ |
---|
62 | |
---|
63 | /* =========================================================================== |
---|
64 | * Function prototypes. |
---|
65 | */ |
---|
66 | typedef enum { |
---|
67 | need_more, /* block not completed, need more input or more output */ |
---|
68 | block_done, /* block flush performed */ |
---|
69 | finish_started, /* finish started, need only more output at next deflate */ |
---|
70 | finish_done /* finish done, accept no more input or output */ |
---|
71 | } block_state; |
---|
72 | |
---|
73 | typedef block_state (*compress_func) OF((deflate_state *s, int flush)); |
---|
74 | /* Compression function. Returns the block state after the call. */ |
---|
75 | |
---|
76 | local void fill_window OF((deflate_state *s)); |
---|
77 | local block_state deflate_stored OF((deflate_state *s, int flush)); |
---|
78 | local block_state deflate_fast OF((deflate_state *s, int flush)); |
---|
79 | #ifndef FASTEST |
---|
80 | local block_state deflate_slow OF((deflate_state *s, int flush)); |
---|
81 | #endif |
---|
82 | local void lm_init OF((deflate_state *s)); |
---|
83 | local void putShortMSB OF((deflate_state *s, uInt b)); |
---|
84 | local void flush_pending OF((z_streamp strm)); |
---|
85 | local int read_buf OF((z_streamp strm, Bytef *buf, unsigned size)); |
---|
86 | #ifndef FASTEST |
---|
87 | #ifdef ASMV |
---|
88 | void match_init OF((void)); /* asm code initialization */ |
---|
89 | uInt longest_match OF((deflate_state *s, IPos cur_match)); |
---|
90 | #else |
---|
91 | local uInt longest_match OF((deflate_state *s, IPos cur_match)); |
---|
92 | #endif |
---|
93 | #endif |
---|
94 | local uInt longest_match_fast OF((deflate_state *s, IPos cur_match)); |
---|
95 | |
---|
96 | #ifdef DEBUG |
---|
97 | local void check_match OF((deflate_state *s, IPos start, IPos match, |
---|
98 | int length)); |
---|
99 | #endif |
---|
100 | |
---|
101 | /* =========================================================================== |
---|
102 | * Local data |
---|
103 | */ |
---|
104 | |
---|
105 | #define NIL 0 |
---|
106 | /* Tail of hash chains */ |
---|
107 | |
---|
108 | #ifndef TOO_FAR |
---|
109 | # define TOO_FAR 4096 |
---|
110 | #endif |
---|
111 | /* Matches of length 3 are discarded if their distance exceeds TOO_FAR */ |
---|
112 | |
---|
113 | #define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1) |
---|
114 | /* Minimum amount of lookahead, except at the end of the input file. |
---|
115 | * See deflate.c for comments about the MIN_MATCH+1. |
---|
116 | */ |
---|
117 | |
---|
118 | /* Values for max_lazy_match, good_match and max_chain_length, depending on |
---|
119 | * the desired pack level (0..9). The values given below have been tuned to |
---|
120 | * exclude worst case performance for pathological files. Better values may be |
---|
121 | * found for specific files. |
---|
122 | */ |
---|
123 | typedef struct config_s { |
---|
124 | ush good_length; /* reduce lazy search above this match length */ |
---|
125 | ush max_lazy; /* do not perform lazy search above this match length */ |
---|
126 | ush nice_length; /* quit search above this match length */ |
---|
127 | ush max_chain; |
---|
128 | compress_func func; |
---|
129 | } config; |
---|
130 | |
---|
131 | #ifdef FASTEST |
---|
132 | local const config configuration_table[2] = { |
---|
133 | /* good lazy nice chain */ |
---|
134 | /* 0 */ {0, 0, 0, 0, deflate_stored}, /* store only */ |
---|
135 | /* 1 */ {4, 4, 8, 4, deflate_fast}}; /* max speed, no lazy matches */ |
---|
136 | #else |
---|
137 | local const config configuration_table[10] = { |
---|
138 | /* good lazy nice chain */ |
---|
139 | /* 0 */ {0, 0, 0, 0, deflate_stored}, /* store only */ |
---|
140 | /* 1 */ {4, 4, 8, 4, deflate_fast}, /* max speed, no lazy matches */ |
---|
141 | /* 2 */ {4, 5, 16, 8, deflate_fast}, |
---|
142 | /* 3 */ {4, 6, 32, 32, deflate_fast}, |
---|
143 | |
---|
144 | /* 4 */ {4, 4, 16, 16, deflate_slow}, /* lazy matches */ |
---|
145 | /* 5 */ {8, 16, 32, 32, deflate_slow}, |
---|
146 | /* 6 */ {8, 16, 128, 128, deflate_slow}, |
---|
147 | /* 7 */ {8, 32, 128, 256, deflate_slow}, |
---|
148 | /* 8 */ {32, 128, 258, 1024, deflate_slow}, |
---|
149 | /* 9 */ {32, 258, 258, 4096, deflate_slow}}; /* max compression */ |
---|
150 | #endif |
---|
151 | |
---|
152 | /* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4 |
---|
153 | * For deflate_fast() (levels <= 3) good is ignored and lazy has a different |
---|
154 | * meaning. |
---|
155 | */ |
---|
156 | |
---|
157 | #define EQUAL 0 |
---|
158 | /* result of memcmp for equal strings */ |
---|
159 | |
---|
160 | #ifndef NO_DUMMY_DECL |
---|
161 | struct static_tree_desc_s {int dummy;}; /* for buggy compilers */ |
---|
162 | #endif |
---|
163 | |
---|
164 | /* =========================================================================== |
---|
165 | * Update a hash value with the given input byte |
---|
166 | * IN assertion: all calls to to UPDATE_HASH are made with consecutive |
---|
167 | * input characters, so that a running hash key can be computed from the |
---|
168 | * previous key instead of complete recalculation each time. |
---|
169 | */ |
---|
170 | #define UPDATE_HASH(s,h,c) (h = (((h)<<s->hash_shift) ^ (c)) & s->hash_mask) |
---|
171 | |
---|
172 | |
---|
173 | /* =========================================================================== |
---|
174 | * Insert string str in the dictionary and set match_head to the previous head |
---|
175 | * of the hash chain (the most recent string with same hash key). Return |
---|
176 | * the previous length of the hash chain. |
---|
177 | * If this file is compiled with -DFASTEST, the compression level is forced |
---|
178 | * to 1, and no hash chains are maintained. |
---|
179 | * IN assertion: all calls to to INSERT_STRING are made with consecutive |
---|
180 | * input characters and the first MIN_MATCH bytes of str are valid |
---|
181 | * (except for the last MIN_MATCH-1 bytes of the input file). |
---|
182 | */ |
---|
183 | #ifdef FASTEST |
---|
184 | #define INSERT_STRING(s, str, match_head) \ |
---|
185 | (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \ |
---|
186 | match_head = s->head[s->ins_h], \ |
---|
187 | s->head[s->ins_h] = (Pos)(str)) |
---|
188 | #else |
---|
189 | #define INSERT_STRING(s, str, match_head) \ |
---|
190 | (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \ |
---|
191 | match_head = s->prev[(str) & s->w_mask] = s->head[s->ins_h], \ |
---|
192 | s->head[s->ins_h] = (Pos)(str)) |
---|
193 | #endif |
---|
194 | |
---|
195 | /* =========================================================================== |
---|
196 | * Initialize the hash table (avoiding 64K overflow for 16 bit systems). |
---|
197 | * prev[] will be initialized on the fly. |
---|
198 | */ |
---|
199 | #define CLEAR_HASH(s) \ |
---|
200 | s->head[s->hash_size-1] = NIL; \ |
---|
201 | zmemzero((Bytef *)s->head, (unsigned)(s->hash_size-1)*sizeof(*s->head)); |
---|
202 | |
---|
203 | /* ========================================================================= */ |
---|
204 | int ZEXPORT deflateInit_(strm, level, version, stream_size) |
---|
205 | z_streamp strm; |
---|
206 | int level; |
---|
207 | const char *version; |
---|
208 | int stream_size; |
---|
209 | { |
---|
210 | return deflateInit2_(strm, level, Z_DEFLATED, MAX_WBITS, DEF_MEM_LEVEL, |
---|
211 | Z_DEFAULT_STRATEGY, version, stream_size); |
---|
212 | /* To do: ignore strm->next_in if we use it as window */ |
---|
213 | } |
---|
214 | |
---|
215 | /* ========================================================================= */ |
---|
216 | int ZEXPORT deflateInit2_(strm, level, method, windowBits, memLevel, strategy, |
---|
217 | version, stream_size) |
---|
218 | z_streamp strm; |
---|
219 | int level; |
---|
220 | int method; |
---|
221 | int windowBits; |
---|
222 | int memLevel; |
---|
223 | int strategy; |
---|
224 | const char *version; |
---|
225 | int stream_size; |
---|
226 | { |
---|
227 | deflate_state *s; |
---|
228 | int wrap = 1; |
---|
229 | static const char my_version[] = ZLIB_VERSION; |
---|
230 | |
---|
231 | ushf *overlay; |
---|
232 | /* We overlay pending_buf and d_buf+l_buf. This works since the average |
---|
233 | * output size for (length,distance) codes is <= 24 bits. |
---|
234 | */ |
---|
235 | |
---|
236 | if (version == Z_NULL || version[0] != my_version[0] || |
---|
237 | stream_size != sizeof(z_stream)) { |
---|
238 | return Z_VERSION_ERROR; |
---|
239 | } |
---|
240 | if (strm == Z_NULL) return Z_STREAM_ERROR; |
---|
241 | |
---|
242 | strm->msg = Z_NULL; |
---|
243 | if (strm->zalloc == (alloc_func)0) { |
---|
244 | strm->zalloc = zcalloc; |
---|
245 | strm->opaque = (voidpf)0; |
---|
246 | } |
---|
247 | if (strm->zfree == (free_func)0) strm->zfree = zcfree; |
---|
248 | |
---|
249 | #ifdef FASTEST |
---|
250 | if (level != 0) level = 1; |
---|
251 | #else |
---|
252 | if (level == Z_DEFAULT_COMPRESSION) level = 6; |
---|
253 | #endif |
---|
254 | |
---|
255 | if (windowBits < 0) { /* suppress zlib wrapper */ |
---|
256 | wrap = 0; |
---|
257 | windowBits = -windowBits; |
---|
258 | } |
---|
259 | #ifdef GZIP |
---|
260 | else if (windowBits > 15) { |
---|
261 | wrap = 2; /* write gzip wrapper instead */ |
---|
262 | windowBits -= 16; |
---|
263 | } |
---|
264 | #endif |
---|
265 | if (memLevel < 1 || memLevel > MAX_MEM_LEVEL || method != Z_DEFLATED || |
---|
266 | windowBits < 8 || windowBits > 15 || level < 0 || level > 9 || |
---|
267 | strategy < 0 || strategy > Z_FIXED) { |
---|
268 | return Z_STREAM_ERROR; |
---|
269 | } |
---|
270 | if (windowBits == 8) windowBits = 9; /* until 256-byte window bug fixed */ |
---|
271 | s = (deflate_state *) ZALLOC(strm, 1, sizeof(deflate_state)); |
---|
272 | if (s == Z_NULL) return Z_MEM_ERROR; |
---|
273 | strm->state = (struct internal_state FAR *)s; |
---|
274 | s->strm = strm; |
---|
275 | |
---|
276 | s->wrap = wrap; |
---|
277 | s->gzhead = Z_NULL; |
---|
278 | s->w_bits = windowBits; |
---|
279 | s->w_size = 1 << s->w_bits; |
---|
280 | s->w_mask = s->w_size - 1; |
---|
281 | |
---|
282 | s->hash_bits = memLevel + 7; |
---|
283 | s->hash_size = 1 << s->hash_bits; |
---|
284 | s->hash_mask = s->hash_size - 1; |
---|
285 | s->hash_shift = ((s->hash_bits+MIN_MATCH-1)/MIN_MATCH); |
---|
286 | |
---|
287 | s->window = (Bytef *) ZALLOC(strm, s->w_size, 2*sizeof(Byte)); |
---|
288 | s->prev = (Posf *) ZALLOC(strm, s->w_size, sizeof(Pos)); |
---|
289 | s->head = (Posf *) ZALLOC(strm, s->hash_size, sizeof(Pos)); |
---|
290 | |
---|
291 | s->lit_bufsize = 1 << (memLevel + 6); /* 16K elements by default */ |
---|
292 | |
---|
293 | overlay = (ushf *) ZALLOC(strm, s->lit_bufsize, sizeof(ush)+2); |
---|
294 | s->pending_buf = (uchf *) overlay; |
---|
295 | s->pending_buf_size = (ulg)s->lit_bufsize * (sizeof(ush)+2L); |
---|
296 | |
---|
297 | if (s->window == Z_NULL || s->prev == Z_NULL || s->head == Z_NULL || |
---|
298 | s->pending_buf == Z_NULL) { |
---|
299 | s->status = FINISH_STATE; |
---|
300 | strm->msg = (char*)ERR_MSG(Z_MEM_ERROR); |
---|
301 | deflateEnd (strm); |
---|
302 | return Z_MEM_ERROR; |
---|
303 | } |
---|
304 | s->d_buf = overlay + s->lit_bufsize/sizeof(ush); |
---|
305 | s->l_buf = s->pending_buf + (1+sizeof(ush))*s->lit_bufsize; |
---|
306 | |
---|
307 | s->level = level; |
---|
308 | s->strategy = strategy; |
---|
309 | s->method = (Byte)method; |
---|
310 | |
---|
311 | return deflateReset(strm); |
---|
312 | } |
---|
313 | |
---|
314 | /* ========================================================================= */ |
---|
315 | int ZEXPORT deflateSetDictionary (strm, dictionary, dictLength) |
---|
316 | z_streamp strm; |
---|
317 | const Bytef *dictionary; |
---|
318 | uInt dictLength; |
---|
319 | { |
---|
320 | deflate_state *s; |
---|
321 | uInt length = dictLength; |
---|
322 | uInt n; |
---|
323 | IPos hash_head = 0; |
---|
324 | |
---|
325 | if (strm == Z_NULL || strm->state == Z_NULL || dictionary == Z_NULL || |
---|
326 | strm->state->wrap == 2 || |
---|
327 | (strm->state->wrap == 1 && strm->state->status != INIT_STATE)) |
---|
328 | return Z_STREAM_ERROR; |
---|
329 | |
---|
330 | s = strm->state; |
---|
331 | if (s->wrap) |
---|
332 | strm->adler = adler32(strm->adler, dictionary, dictLength); |
---|
333 | |
---|
334 | if (length < MIN_MATCH) return Z_OK; |
---|
335 | if (length > MAX_DIST(s)) { |
---|
336 | length = MAX_DIST(s); |
---|
337 | #ifndef USE_DICT_HEAD |
---|
338 | dictionary += dictLength - length; /* use the tail of the dictionary */ |
---|
339 | #endif |
---|
340 | } |
---|
341 | zmemcpy(s->window, dictionary, length); |
---|
342 | s->strstart = length; |
---|
343 | s->block_start = (long)length; |
---|
344 | |
---|
345 | /* Insert all strings in the hash table (except for the last two bytes). |
---|
346 | * s->lookahead stays null, so s->ins_h will be recomputed at the next |
---|
347 | * call of fill_window. |
---|
348 | */ |
---|
349 | s->ins_h = s->window[0]; |
---|
350 | UPDATE_HASH(s, s->ins_h, s->window[1]); |
---|
351 | for (n = 0; n <= length - MIN_MATCH; n++) { |
---|
352 | INSERT_STRING(s, n, hash_head); |
---|
353 | } |
---|
354 | if (hash_head) hash_head = 0; /* to make compiler happy */ |
---|
355 | return Z_OK; |
---|
356 | } |
---|
357 | |
---|
358 | /* ========================================================================= */ |
---|
359 | int ZEXPORT deflateReset (strm) |
---|
360 | z_streamp strm; |
---|
361 | { |
---|
362 | deflate_state *s; |
---|
363 | |
---|
364 | if (strm == Z_NULL || strm->state == Z_NULL || |
---|
365 | strm->zalloc == (alloc_func)0 || strm->zfree == (free_func)0) { |
---|
366 | return Z_STREAM_ERROR; |
---|
367 | } |
---|
368 | |
---|
369 | strm->total_in = strm->total_out = 0; |
---|
370 | strm->msg = Z_NULL; /* use zfree if we ever allocate msg dynamically */ |
---|
371 | strm->data_type = Z_UNKNOWN; |
---|
372 | |
---|
373 | s = (deflate_state *)strm->state; |
---|
374 | s->pending = 0; |
---|
375 | s->pending_out = s->pending_buf; |
---|
376 | |
---|
377 | if (s->wrap < 0) { |
---|
378 | s->wrap = -s->wrap; /* was made negative by deflate(..., Z_FINISH); */ |
---|
379 | } |
---|
380 | s->status = s->wrap ? INIT_STATE : BUSY_STATE; |
---|
381 | strm->adler = |
---|
382 | #ifdef GZIP |
---|
383 | s->wrap == 2 ? crc32(0L, Z_NULL, 0) : |
---|
384 | #endif |
---|
385 | adler32(0L, Z_NULL, 0); |
---|
386 | s->last_flush = Z_NO_FLUSH; |
---|
387 | |
---|
388 | _tr_init(s); |
---|
389 | lm_init(s); |
---|
390 | |
---|
391 | return Z_OK; |
---|
392 | } |
---|
393 | |
---|
394 | /* ========================================================================= */ |
---|
395 | int ZEXPORT deflateSetHeader (strm, head) |
---|
396 | z_streamp strm; |
---|
397 | gz_headerp head; |
---|
398 | { |
---|
399 | if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; |
---|
400 | if (strm->state->wrap != 2) return Z_STREAM_ERROR; |
---|
401 | strm->state->gzhead = head; |
---|
402 | return Z_OK; |
---|
403 | } |
---|
404 | |
---|
405 | /* ========================================================================= */ |
---|
406 | int ZEXPORT deflatePrime (strm, bits, value) |
---|
407 | z_streamp strm; |
---|
408 | int bits; |
---|
409 | int value; |
---|
410 | { |
---|
411 | if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; |
---|
412 | strm->state->bi_valid = bits; |
---|
413 | strm->state->bi_buf = (ush)(value & ((1 << bits) - 1)); |
---|
414 | return Z_OK; |
---|
415 | } |
---|
416 | |
---|
417 | /* ========================================================================= */ |
---|
418 | int ZEXPORT deflateParams(strm, level, strategy) |
---|
419 | z_streamp strm; |
---|
420 | int level; |
---|
421 | int strategy; |
---|
422 | { |
---|
423 | deflate_state *s; |
---|
424 | compress_func func; |
---|
425 | int err = Z_OK; |
---|
426 | |
---|
427 | if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; |
---|
428 | s = strm->state; |
---|
429 | |
---|
430 | #ifdef FASTEST |
---|
431 | if (level != 0) level = 1; |
---|
432 | #else |
---|
433 | if (level == Z_DEFAULT_COMPRESSION) level = 6; |
---|
434 | #endif |
---|
435 | if (level < 0 || level > 9 || strategy < 0 || strategy > Z_FIXED) { |
---|
436 | return Z_STREAM_ERROR; |
---|
437 | } |
---|
438 | func = configuration_table[s->level].func; |
---|
439 | |
---|
440 | if (func != configuration_table[level].func && strm->total_in != 0) { |
---|
441 | /* Flush the last buffer: */ |
---|
442 | err = deflate(strm, Z_PARTIAL_FLUSH); |
---|
443 | } |
---|
444 | if (s->level != level) { |
---|
445 | s->level = level; |
---|
446 | s->max_lazy_match = configuration_table[level].max_lazy; |
---|
447 | s->good_match = configuration_table[level].good_length; |
---|
448 | s->nice_match = configuration_table[level].nice_length; |
---|
449 | s->max_chain_length = configuration_table[level].max_chain; |
---|
450 | } |
---|
451 | s->strategy = strategy; |
---|
452 | return err; |
---|
453 | } |
---|
454 | |
---|
455 | /* ========================================================================= |
---|
456 | * For the default windowBits of 15 and memLevel of 8, this function returns |
---|
457 | * a close to exact, as well as small, upper bound on the compressed size. |
---|
458 | * They are coded as constants here for a reason--if the #define's are |
---|
459 | * changed, then this function needs to be changed as well. The return |
---|
460 | * value for 15 and 8 only works for those exact settings. |
---|
461 | * |
---|
462 | * For any setting other than those defaults for windowBits and memLevel, |
---|
463 | * the value returned is a conservative worst case for the maximum expansion |
---|
464 | * resulting from using fixed blocks instead of stored blocks, which deflate |
---|
465 | * can emit on compressed data for some combinations of the parameters. |
---|
466 | * |
---|
467 | * This function could be more sophisticated to provide closer upper bounds |
---|
468 | * for every combination of windowBits and memLevel, as well as wrap. |
---|
469 | * But even the conservative upper bound of about 14% expansion does not |
---|
470 | * seem onerous for output buffer allocation. |
---|
471 | */ |
---|
472 | uLong ZEXPORT deflateBound(strm, sourceLen) |
---|
473 | z_streamp strm; |
---|
474 | uLong sourceLen; |
---|
475 | { |
---|
476 | deflate_state *s; |
---|
477 | uLong destLen; |
---|
478 | |
---|
479 | /* conservative upper bound */ |
---|
480 | destLen = sourceLen + |
---|
481 | ((sourceLen + 7) >> 3) + ((sourceLen + 63) >> 6) + 11; |
---|
482 | |
---|
483 | /* if can't get parameters, return conservative bound */ |
---|
484 | if (strm == Z_NULL || strm->state == Z_NULL) |
---|
485 | return destLen; |
---|
486 | |
---|
487 | /* if not default parameters, return conservative bound */ |
---|
488 | s = strm->state; |
---|
489 | if (s->w_bits != 15 || s->hash_bits != 8 + 7) |
---|
490 | return destLen; |
---|
491 | |
---|
492 | /* default settings: return tight bound for that case */ |
---|
493 | return compressBound(sourceLen); |
---|
494 | } |
---|
495 | |
---|
496 | /* ========================================================================= |
---|
497 | * Put a short in the pending buffer. The 16-bit value is put in MSB order. |
---|
498 | * IN assertion: the stream state is correct and there is enough room in |
---|
499 | * pending_buf. |
---|
500 | */ |
---|
501 | local void putShortMSB (s, b) |
---|
502 | deflate_state *s; |
---|
503 | uInt b; |
---|
504 | { |
---|
505 | put_byte(s, (Byte)(b >> 8)); |
---|
506 | put_byte(s, (Byte)(b & 0xff)); |
---|
507 | } |
---|
508 | |
---|
509 | /* ========================================================================= |
---|
510 | * Flush as much pending output as possible. All deflate() output goes |
---|
511 | * through this function so some applications may wish to modify it |
---|
512 | * to avoid allocating a large strm->next_out buffer and copying into it. |
---|
513 | * (See also read_buf()). |
---|
514 | */ |
---|
515 | local void flush_pending(strm) |
---|
516 | z_streamp strm; |
---|
517 | { |
---|
518 | unsigned len = strm->state->pending; |
---|
519 | |
---|
520 | if (len > strm->avail_out) len = strm->avail_out; |
---|
521 | if (len == 0) return; |
---|
522 | |
---|
523 | zmemcpy(strm->next_out, strm->state->pending_out, len); |
---|
524 | strm->next_out += len; |
---|
525 | strm->state->pending_out += len; |
---|
526 | strm->total_out += len; |
---|
527 | strm->avail_out -= len; |
---|
528 | strm->state->pending -= len; |
---|
529 | if (strm->state->pending == 0) { |
---|
530 | strm->state->pending_out = strm->state->pending_buf; |
---|
531 | } |
---|
532 | } |
---|
533 | |
---|
534 | /* ========================================================================= */ |
---|
535 | int ZEXPORT deflate (strm, flush) |
---|
536 | z_streamp strm; |
---|
537 | int flush; |
---|
538 | { |
---|
539 | int old_flush; /* value of flush param for previous deflate call */ |
---|
540 | deflate_state *s; |
---|
541 | |
---|
542 | if (strm == Z_NULL || strm->state == Z_NULL || |
---|
543 | flush > Z_FINISH || flush < 0) { |
---|
544 | return Z_STREAM_ERROR; |
---|
545 | } |
---|
546 | s = strm->state; |
---|
547 | |
---|
548 | if (strm->next_out == Z_NULL || |
---|
549 | (strm->next_in == Z_NULL && strm->avail_in != 0) || |
---|
550 | (s->status == FINISH_STATE && flush != Z_FINISH)) { |
---|
551 | ERR_RETURN(strm, Z_STREAM_ERROR); |
---|
552 | } |
---|
553 | if (strm->avail_out == 0) ERR_RETURN(strm, Z_BUF_ERROR); |
---|
554 | |
---|
555 | s->strm = strm; /* just in case */ |
---|
556 | old_flush = s->last_flush; |
---|
557 | s->last_flush = flush; |
---|
558 | |
---|
559 | /* Write the header */ |
---|
560 | if (s->status == INIT_STATE) { |
---|
561 | #ifdef GZIP |
---|
562 | if (s->wrap == 2) { |
---|
563 | strm->adler = crc32(0L, Z_NULL, 0); |
---|
564 | put_byte(s, 31); |
---|
565 | put_byte(s, 139); |
---|
566 | put_byte(s, 8); |
---|
567 | if (s->gzhead == NULL) { |
---|
568 | put_byte(s, 0); |
---|
569 | put_byte(s, 0); |
---|
570 | put_byte(s, 0); |
---|
571 | put_byte(s, 0); |
---|
572 | put_byte(s, 0); |
---|
573 | put_byte(s, s->level == 9 ? 2 : |
---|
574 | (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ? |
---|
575 | 4 : 0)); |
---|
576 | put_byte(s, OS_CODE); |
---|
577 | s->status = BUSY_STATE; |
---|
578 | } |
---|
579 | else { |
---|
580 | put_byte(s, (s->gzhead->text ? 1 : 0) + |
---|
581 | (s->gzhead->hcrc ? 2 : 0) + |
---|
582 | (s->gzhead->extra == Z_NULL ? 0 : 4) + |
---|
583 | (s->gzhead->name == Z_NULL ? 0 : 8) + |
---|
584 | (s->gzhead->comment == Z_NULL ? 0 : 16) |
---|
585 | ); |
---|
586 | put_byte(s, s->gzhead->time & 0xff); |
---|
587 | put_byte(s, (s->gzhead->time >> 8) & 0xff); |
---|
588 | put_byte(s, (s->gzhead->time >> 16) & 0xff); |
---|
589 | put_byte(s, (s->gzhead->time >> 24) & 0xff); |
---|
590 | put_byte(s, s->level == 9 ? 2 : |
---|
591 | (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ? |
---|
592 | 4 : 0)); |
---|
593 | put_byte(s, s->gzhead->os & 0xff); |
---|
594 | if (s->gzhead->extra != NULL) { |
---|
595 | put_byte(s, s->gzhead->extra_len & 0xff); |
---|
596 | put_byte(s, (s->gzhead->extra_len >> 8) & 0xff); |
---|
597 | } |
---|
598 | if (s->gzhead->hcrc) |
---|
599 | strm->adler = crc32(strm->adler, s->pending_buf, |
---|
600 | s->pending); |
---|
601 | s->gzindex = 0; |
---|
602 | s->status = EXTRA_STATE; |
---|
603 | } |
---|
604 | } |
---|
605 | else |
---|
606 | #endif |
---|
607 | { |
---|
608 | uInt header = (Z_DEFLATED + ((s->w_bits-8)<<4)) << 8; |
---|
609 | uInt level_flags; |
---|
610 | |
---|
611 | if (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2) |
---|
612 | level_flags = 0; |
---|
613 | else if (s->level < 6) |
---|
614 | level_flags = 1; |
---|
615 | else if (s->level == 6) |
---|
616 | level_flags = 2; |
---|
617 | else |
---|
618 | level_flags = 3; |
---|
619 | header |= (level_flags << 6); |
---|
620 | if (s->strstart != 0) header |= PRESET_DICT; |
---|
621 | header += 31 - (header % 31); |
---|
622 | |
---|
623 | s->status = BUSY_STATE; |
---|
624 | putShortMSB(s, header); |
---|
625 | |
---|
626 | /* Save the adler32 of the preset dictionary: */ |
---|
627 | if (s->strstart != 0) { |
---|
628 | putShortMSB(s, (uInt)(strm->adler >> 16)); |
---|
629 | putShortMSB(s, (uInt)(strm->adler & 0xffff)); |
---|
630 | } |
---|
631 | strm->adler = adler32(0L, Z_NULL, 0); |
---|
632 | } |
---|
633 | } |
---|
634 | #ifdef GZIP |
---|
635 | if (s->status == EXTRA_STATE) { |
---|
636 | if (s->gzhead->extra != NULL) { |
---|
637 | int beg = s->pending; /* start of bytes to update crc */ |
---|
638 | |
---|
639 | while (s->gzindex < (s->gzhead->extra_len & 0xffff)) { |
---|
640 | if (s->pending == s->pending_buf_size) { |
---|
641 | if (s->gzhead->hcrc && s->pending > beg) |
---|
642 | strm->adler = crc32(strm->adler, s->pending_buf + beg, |
---|
643 | s->pending - beg); |
---|
644 | flush_pending(strm); |
---|
645 | beg = s->pending; |
---|
646 | if (s->pending == s->pending_buf_size) |
---|
647 | break; |
---|
648 | } |
---|
649 | put_byte(s, s->gzhead->extra[s->gzindex]); |
---|
650 | s->gzindex++; |
---|
651 | } |
---|
652 | if (s->gzhead->hcrc && s->pending > beg) |
---|
653 | strm->adler = crc32(strm->adler, s->pending_buf + beg, |
---|
654 | s->pending - beg); |
---|
655 | if (s->gzindex == s->gzhead->extra_len) { |
---|
656 | s->gzindex = 0; |
---|
657 | s->status = NAME_STATE; |
---|
658 | } |
---|
659 | } |
---|
660 | else |
---|
661 | s->status = NAME_STATE; |
---|
662 | } |
---|
663 | if (s->status == NAME_STATE) { |
---|
664 | if (s->gzhead->name != NULL) { |
---|
665 | int beg = s->pending; /* start of bytes to update crc */ |
---|
666 | int val; |
---|
667 | |
---|
668 | do { |
---|
669 | if (s->pending == s->pending_buf_size) { |
---|
670 | if (s->gzhead->hcrc && s->pending > beg) |
---|
671 | strm->adler = crc32(strm->adler, s->pending_buf + beg, |
---|
672 | s->pending - beg); |
---|
673 | flush_pending(strm); |
---|
674 | beg = s->pending; |
---|
675 | if (s->pending == s->pending_buf_size) { |
---|
676 | val = 1; |
---|
677 | break; |
---|
678 | } |
---|
679 | } |
---|
680 | val = s->gzhead->name[s->gzindex++]; |
---|
681 | put_byte(s, val); |
---|
682 | } while (val != 0); |
---|
683 | if (s->gzhead->hcrc && s->pending > beg) |
---|
684 | strm->adler = crc32(strm->adler, s->pending_buf + beg, |
---|
685 | s->pending - beg); |
---|
686 | if (val == 0) { |
---|
687 | s->gzindex = 0; |
---|
688 | s->status = COMMENT_STATE; |
---|
689 | } |
---|
690 | } |
---|
691 | else |
---|
692 | s->status = COMMENT_STATE; |
---|
693 | } |
---|
694 | if (s->status == COMMENT_STATE) { |
---|
695 | if (s->gzhead->comment != NULL) { |
---|
696 | int beg = s->pending; /* start of bytes to update crc */ |
---|
697 | int val; |
---|
698 | |
---|
699 | do { |
---|
700 | if (s->pending == s->pending_buf_size) { |
---|
701 | if (s->gzhead->hcrc && s->pending > beg) |
---|
702 | strm->adler = crc32(strm->adler, s->pending_buf + beg, |
---|
703 | s->pending - beg); |
---|
704 | flush_pending(strm); |
---|
705 | beg = s->pending; |
---|
706 | if (s->pending == s->pending_buf_size) { |
---|
707 | val = 1; |
---|
708 | break; |
---|
709 | } |
---|
710 | } |
---|
711 | val = s->gzhead->comment[s->gzindex++]; |
---|
712 | put_byte(s, val); |
---|
713 | } while (val != 0); |
---|
714 | if (s->gzhead->hcrc && s->pending > beg) |
---|
715 | strm->adler = crc32(strm->adler, s->pending_buf + beg, |
---|
716 | s->pending - beg); |
---|
717 | if (val == 0) |
---|
718 | s->status = HCRC_STATE; |
---|
719 | } |
---|
720 | else |
---|
721 | s->status = HCRC_STATE; |
---|
722 | } |
---|
723 | if (s->status == HCRC_STATE) { |
---|
724 | if (s->gzhead->hcrc) { |
---|
725 | if (s->pending + 2 > s->pending_buf_size) |
---|
726 | flush_pending(strm); |
---|
727 | if (s->pending + 2 <= s->pending_buf_size) { |
---|
728 | put_byte(s, strm->adler & 0xff); |
---|
729 | put_byte(s, (strm->adler >> 8) & 0xff); |
---|
730 | strm->adler = crc32(0L, Z_NULL, 0); |
---|
731 | s->status = BUSY_STATE; |
---|
732 | } |
---|
733 | } |
---|
734 | else |
---|
735 | s->status = BUSY_STATE; |
---|
736 | } |
---|
737 | #endif |
---|
738 | |
---|
739 | /* Flush as much pending output as possible */ |
---|
740 | if (s->pending != 0) { |
---|
741 | flush_pending(strm); |
---|
742 | if (strm->avail_out == 0) { |
---|
743 | /* Since avail_out is 0, deflate will be called again with |
---|
744 | * more output space, but possibly with both pending and |
---|
745 | * avail_in equal to zero. There won't be anything to do, |
---|
746 | * but this is not an error situation so make sure we |
---|
747 | * return OK instead of BUF_ERROR at next call of deflate: |
---|
748 | */ |
---|
749 | s->last_flush = -1; |
---|
750 | return Z_OK; |
---|
751 | } |
---|
752 | |
---|
753 | /* Make sure there is something to do and avoid duplicate consecutive |
---|
754 | * flushes. For repeated and useless calls with Z_FINISH, we keep |
---|
755 | * returning Z_STREAM_END instead of Z_BUF_ERROR. |
---|
756 | */ |
---|
757 | } else if (strm->avail_in == 0 && flush <= old_flush && |
---|
758 | flush != Z_FINISH) { |
---|
759 | ERR_RETURN(strm, Z_BUF_ERROR); |
---|
760 | } |
---|
761 | |
---|
762 | /* User must not provide more input after the first FINISH: */ |
---|
763 | if (s->status == FINISH_STATE && strm->avail_in != 0) { |
---|
764 | ERR_RETURN(strm, Z_BUF_ERROR); |
---|
765 | } |
---|
766 | |
---|
767 | /* Start a new block or continue the current one. |
---|
768 | */ |
---|
769 | if (strm->avail_in != 0 || s->lookahead != 0 || |
---|
770 | (flush != Z_NO_FLUSH && s->status != FINISH_STATE)) { |
---|
771 | block_state bstate; |
---|
772 | |
---|
773 | bstate = (*(configuration_table[s->level].func))(s, flush); |
---|
774 | |
---|
775 | if (bstate == finish_started || bstate == finish_done) { |
---|
776 | s->status = FINISH_STATE; |
---|
777 | } |
---|
778 | if (bstate == need_more || bstate == finish_started) { |
---|
779 | if (strm->avail_out == 0) { |
---|
780 | s->last_flush = -1; /* avoid BUF_ERROR next call, see above */ |
---|
781 | } |
---|
782 | return Z_OK; |
---|
783 | /* If flush != Z_NO_FLUSH && avail_out == 0, the next call |
---|
784 | * of deflate should use the same flush parameter to make sure |
---|
785 | * that the flush is complete. So we don't have to output an |
---|
786 | * empty block here, this will be done at next call. This also |
---|
787 | * ensures that for a very small output buffer, we emit at most |
---|
788 | * one empty block. |
---|
789 | */ |
---|
790 | } |
---|
791 | if (bstate == block_done) { |
---|
792 | if (flush == Z_PARTIAL_FLUSH) { |
---|
793 | _tr_align(s); |
---|
794 | } else { /* FULL_FLUSH or SYNC_FLUSH */ |
---|
795 | _tr_stored_block(s, (char*)0, 0L, 0); |
---|
796 | /* For a full flush, this empty block will be recognized |
---|
797 | * as a special marker by inflate_sync(). |
---|
798 | */ |
---|
799 | if (flush == Z_FULL_FLUSH) { |
---|
800 | CLEAR_HASH(s); /* forget history */ |
---|
801 | } |
---|
802 | } |
---|
803 | flush_pending(strm); |
---|
804 | if (strm->avail_out == 0) { |
---|
805 | s->last_flush = -1; /* avoid BUF_ERROR at next call, see above */ |
---|
806 | return Z_OK; |
---|
807 | } |
---|
808 | } |
---|
809 | } |
---|
810 | Assert(strm->avail_out > 0, "bug2"); |
---|
811 | |
---|
812 | if (flush != Z_FINISH) return Z_OK; |
---|
813 | if (s->wrap <= 0) return Z_STREAM_END; |
---|
814 | |
---|
815 | /* Write the trailer */ |
---|
816 | #ifdef GZIP |
---|
817 | if (s->wrap == 2) { |
---|
818 | put_byte(s, (Byte)(strm->adler & 0xff)); |
---|
819 | put_byte(s, (Byte)((strm->adler >> 8) & 0xff)); |
---|
820 | put_byte(s, (Byte)((strm->adler >> 16) & 0xff)); |
---|
821 | put_byte(s, (Byte)((strm->adler >> 24) & 0xff)); |
---|
822 | put_byte(s, (Byte)(strm->total_in & 0xff)); |
---|
823 | put_byte(s, (Byte)((strm->total_in >> 8) & 0xff)); |
---|
824 | put_byte(s, (Byte)((strm->total_in >> 16) & 0xff)); |
---|
825 | put_byte(s, (Byte)((strm->total_in >> 24) & 0xff)); |
---|
826 | } |
---|
827 | else |
---|
828 | #endif |
---|
829 | { |
---|
830 | putShortMSB(s, (uInt)(strm->adler >> 16)); |
---|
831 | putShortMSB(s, (uInt)(strm->adler & 0xffff)); |
---|
832 | } |
---|
833 | flush_pending(strm); |
---|
834 | /* If avail_out is zero, the application will call deflate again |
---|
835 | * to flush the rest. |
---|
836 | */ |
---|
837 | if (s->wrap > 0) s->wrap = -s->wrap; /* write the trailer only once! */ |
---|
838 | return s->pending != 0 ? Z_OK : Z_STREAM_END; |
---|
839 | } |
---|
840 | |
---|
841 | /* ========================================================================= */ |
---|
842 | int ZEXPORT deflateEnd (strm) |
---|
843 | z_streamp strm; |
---|
844 | { |
---|
845 | int status; |
---|
846 | |
---|
847 | if (strm == Z_NULL || strm->state == Z_NULL) return Z_STREAM_ERROR; |
---|
848 | |
---|
849 | status = strm->state->status; |
---|
850 | if (status != INIT_STATE && |
---|
851 | status != EXTRA_STATE && |
---|
852 | status != NAME_STATE && |
---|
853 | status != COMMENT_STATE && |
---|
854 | status != HCRC_STATE && |
---|
855 | status != BUSY_STATE && |
---|
856 | status != FINISH_STATE) { |
---|
857 | return Z_STREAM_ERROR; |
---|
858 | } |
---|
859 | |
---|
860 | /* Deallocate in reverse order of allocations: */ |
---|
861 | TRY_FREE(strm, strm->state->pending_buf); |
---|
862 | TRY_FREE(strm, strm->state->head); |
---|
863 | TRY_FREE(strm, strm->state->prev); |
---|
864 | TRY_FREE(strm, strm->state->window); |
---|
865 | |
---|
866 | ZFREE(strm, strm->state); |
---|
867 | strm->state = Z_NULL; |
---|
868 | |
---|
869 | return status == BUSY_STATE ? Z_DATA_ERROR : Z_OK; |
---|
870 | } |
---|
871 | |
---|
872 | /* ========================================================================= |
---|
873 | * Copy the source state to the destination state. |
---|
874 | * To simplify the source, this is not supported for 16-bit MSDOS (which |
---|
875 | * doesn't have enough memory anyway to duplicate compression states). |
---|
876 | */ |
---|
877 | int ZEXPORT deflateCopy (dest, source) |
---|
878 | z_streamp dest; |
---|
879 | z_streamp source; |
---|
880 | { |
---|
881 | #ifdef MAXSEG_64K |
---|
882 | return Z_STREAM_ERROR; |
---|
883 | #else |
---|
884 | deflate_state *ds; |
---|
885 | deflate_state *ss; |
---|
886 | ushf *overlay; |
---|
887 | |
---|
888 | |
---|
889 | if (source == Z_NULL || dest == Z_NULL || source->state == Z_NULL) { |
---|
890 | return Z_STREAM_ERROR; |
---|
891 | } |
---|
892 | |
---|
893 | ss = source->state; |
---|
894 | |
---|
895 | zmemcpy(dest, source, sizeof(z_stream)); |
---|
896 | |
---|
897 | ds = (deflate_state *) ZALLOC(dest, 1, sizeof(deflate_state)); |
---|
898 | if (ds == Z_NULL) return Z_MEM_ERROR; |
---|
899 | dest->state = (struct internal_state FAR *) ds; |
---|
900 | zmemcpy(ds, ss, sizeof(deflate_state)); |
---|
901 | ds->strm = dest; |
---|
902 | |
---|
903 | ds->window = (Bytef *) ZALLOC(dest, ds->w_size, 2*sizeof(Byte)); |
---|
904 | ds->prev = (Posf *) ZALLOC(dest, ds->w_size, sizeof(Pos)); |
---|
905 | ds->head = (Posf *) ZALLOC(dest, ds->hash_size, sizeof(Pos)); |
---|
906 | overlay = (ushf *) ZALLOC(dest, ds->lit_bufsize, sizeof(ush)+2); |
---|
907 | ds->pending_buf = (uchf *) overlay; |
---|
908 | |
---|
909 | if (ds->window == Z_NULL || ds->prev == Z_NULL || ds->head == Z_NULL || |
---|
910 | ds->pending_buf == Z_NULL) { |
---|
911 | deflateEnd (dest); |
---|
912 | return Z_MEM_ERROR; |
---|
913 | } |
---|
914 | /* following zmemcpy do not work for 16-bit MSDOS */ |
---|
915 | zmemcpy(ds->window, ss->window, ds->w_size * 2 * sizeof(Byte)); |
---|
916 | zmemcpy(ds->prev, ss->prev, ds->w_size * sizeof(Pos)); |
---|
917 | zmemcpy(ds->head, ss->head, ds->hash_size * sizeof(Pos)); |
---|
918 | zmemcpy(ds->pending_buf, ss->pending_buf, (uInt)ds->pending_buf_size); |
---|
919 | |
---|
920 | ds->pending_out = ds->pending_buf + (ss->pending_out - ss->pending_buf); |
---|
921 | ds->d_buf = overlay + ds->lit_bufsize/sizeof(ush); |
---|
922 | ds->l_buf = ds->pending_buf + (1+sizeof(ush))*ds->lit_bufsize; |
---|
923 | |
---|
924 | ds->l_desc.dyn_tree = ds->dyn_ltree; |
---|
925 | ds->d_desc.dyn_tree = ds->dyn_dtree; |
---|
926 | ds->bl_desc.dyn_tree = ds->bl_tree; |
---|
927 | |
---|
928 | return Z_OK; |
---|
929 | #endif /* MAXSEG_64K */ |
---|
930 | } |
---|
931 | |
---|
932 | /* =========================================================================== |
---|
933 | * Read a new buffer from the current input stream, update the adler32 |
---|
934 | * and total number of bytes read. All deflate() input goes through |
---|
935 | * this function so some applications may wish to modify it to avoid |
---|
936 | * allocating a large strm->next_in buffer and copying from it. |
---|
937 | * (See also flush_pending()). |
---|
938 | */ |
---|
939 | local int read_buf(strm, buf, size) |
---|
940 | z_streamp strm; |
---|
941 | Bytef *buf; |
---|
942 | unsigned size; |
---|
943 | { |
---|
944 | unsigned len = strm->avail_in; |
---|
945 | |
---|
946 | if (len > size) len = size; |
---|
947 | if (len == 0) return 0; |
---|
948 | |
---|
949 | strm->avail_in -= len; |
---|
950 | |
---|
951 | if (strm->state->wrap == 1) { |
---|
952 | strm->adler = adler32(strm->adler, strm->next_in, len); |
---|
953 | } |
---|
954 | #ifdef GZIP |
---|
955 | else if (strm->state->wrap == 2) { |
---|
956 | strm->adler = crc32(strm->adler, strm->next_in, len); |
---|
957 | } |
---|
958 | #endif |
---|
959 | zmemcpy(buf, strm->next_in, len); |
---|
960 | strm->next_in += len; |
---|
961 | strm->total_in += len; |
---|
962 | |
---|
963 | return (int)len; |
---|
964 | } |
---|
965 | |
---|
966 | /* =========================================================================== |
---|
967 | * Initialize the "longest match" routines for a new zlib stream |
---|
968 | */ |
---|
969 | local void lm_init (s) |
---|
970 | deflate_state *s; |
---|
971 | { |
---|
972 | s->window_size = (ulg)2L*s->w_size; |
---|
973 | |
---|
974 | CLEAR_HASH(s); |
---|
975 | |
---|
976 | /* Set the default configuration parameters: |
---|
977 | */ |
---|
978 | s->max_lazy_match = configuration_table[s->level].max_lazy; |
---|
979 | s->good_match = configuration_table[s->level].good_length; |
---|
980 | s->nice_match = configuration_table[s->level].nice_length; |
---|
981 | s->max_chain_length = configuration_table[s->level].max_chain; |
---|
982 | |
---|
983 | s->strstart = 0; |
---|
984 | s->block_start = 0L; |
---|
985 | s->lookahead = 0; |
---|
986 | s->match_length = s->prev_length = MIN_MATCH-1; |
---|
987 | s->match_available = 0; |
---|
988 | s->ins_h = 0; |
---|
989 | #ifdef ASMV |
---|
990 | match_init(); /* initialize the asm code */ |
---|
991 | #endif |
---|
992 | } |
---|
993 | |
---|
994 | #ifndef FASTEST |
---|
995 | /* =========================================================================== |
---|
996 | * Set match_start to the longest match starting at the given string and |
---|
997 | * return its length. Matches shorter or equal to prev_length are discarded, |
---|
998 | * in which case the result is equal to prev_length and match_start is |
---|
999 | * garbage. |
---|
1000 | * IN assertions: cur_match is the head of the hash chain for the current |
---|
1001 | * string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1 |
---|
1002 | * OUT assertion: the match length is not greater than s->lookahead. |
---|
1003 | */ |
---|
1004 | #ifndef ASMV |
---|
1005 | /* For 80x86 and 680x0, an optimized version will be provided in match.asm or |
---|
1006 | * match.S. The code will be functionally equivalent. |
---|
1007 | */ |
---|
1008 | local uInt longest_match(s, cur_match) |
---|
1009 | deflate_state *s; |
---|
1010 | IPos cur_match; /* current match */ |
---|
1011 | { |
---|
1012 | unsigned chain_length = s->max_chain_length;/* max hash chain length */ |
---|
1013 | register Bytef *scan = s->window + s->strstart; /* current string */ |
---|
1014 | register Bytef *match; /* matched string */ |
---|
1015 | register int len; /* length of current match */ |
---|
1016 | int best_len = s->prev_length; /* best match length so far */ |
---|
1017 | int nice_match = s->nice_match; /* stop if match long enough */ |
---|
1018 | IPos limit = s->strstart > (IPos)MAX_DIST(s) ? |
---|
1019 | s->strstart - (IPos)MAX_DIST(s) : NIL; |
---|
1020 | /* Stop when cur_match becomes <= limit. To simplify the code, |
---|
1021 | * we prevent matches with the string of window index 0. |
---|
1022 | */ |
---|
1023 | Posf *prev = s->prev; |
---|
1024 | uInt wmask = s->w_mask; |
---|
1025 | |
---|
1026 | #ifdef UNALIGNED_OK |
---|
1027 | /* Compare two bytes at a time. Note: this is not always beneficial. |
---|
1028 | * Try with and without -DUNALIGNED_OK to check. |
---|
1029 | */ |
---|
1030 | register Bytef *strend = s->window + s->strstart + MAX_MATCH - 1; |
---|
1031 | register ush scan_start = *(ushf*)scan; |
---|
1032 | register ush scan_end = *(ushf*)(scan+best_len-1); |
---|
1033 | #else |
---|
1034 | register Bytef *strend = s->window + s->strstart + MAX_MATCH; |
---|
1035 | register Byte scan_end1 = scan[best_len-1]; |
---|
1036 | register Byte scan_end = scan[best_len]; |
---|
1037 | #endif |
---|
1038 | |
---|
1039 | /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16. |
---|
1040 | * It is easy to get rid of this optimization if necessary. |
---|
1041 | */ |
---|
1042 | Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever"); |
---|
1043 | |
---|
1044 | /* Do not waste too much time if we already have a good match: */ |
---|
1045 | if (s->prev_length >= s->good_match) { |
---|
1046 | chain_length >>= 2; |
---|
1047 | } |
---|
1048 | /* Do not look for matches beyond the end of the input. This is necessary |
---|
1049 | * to make deflate deterministic. |
---|
1050 | */ |
---|
1051 | if ((uInt)nice_match > s->lookahead) nice_match = s->lookahead; |
---|
1052 | |
---|
1053 | Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead"); |
---|
1054 | |
---|
1055 | do { |
---|
1056 | Assert(cur_match < s->strstart, "no future"); |
---|
1057 | match = s->window + cur_match; |
---|
1058 | |
---|
1059 | /* Skip to next match if the match length cannot increase |
---|
1060 | * or if the match length is less than 2. Note that the checks below |
---|
1061 | * for insufficient lookahead only occur occasionally for performance |
---|
1062 | * reasons. Therefore uninitialized memory will be accessed, and |
---|
1063 | * conditional jumps will be made that depend on those values. |
---|
1064 | * However the length of the match is limited to the lookahead, so |
---|
1065 | * the output of deflate is not affected by the uninitialized values. |
---|
1066 | */ |
---|
1067 | #if (defined(UNALIGNED_OK) && MAX_MATCH == 258) |
---|
1068 | /* This code assumes sizeof(unsigned short) == 2. Do not use |
---|
1069 | * UNALIGNED_OK if your compiler uses a different size. |
---|
1070 | */ |
---|
1071 | if (*(ushf*)(match+best_len-1) != scan_end || |
---|
1072 | *(ushf*)match != scan_start) continue; |
---|
1073 | |
---|
1074 | /* It is not necessary to compare scan[2] and match[2] since they are |
---|
1075 | * always equal when the other bytes match, given that the hash keys |
---|
1076 | * are equal and that HASH_BITS >= 8. Compare 2 bytes at a time at |
---|
1077 | * strstart+3, +5, ... up to strstart+257. We check for insufficient |
---|
1078 | * lookahead only every 4th comparison; the 128th check will be made |
---|
1079 | * at strstart+257. If MAX_MATCH-2 is not a multiple of 8, it is |
---|
1080 | * necessary to put more guard bytes at the end of the window, or |
---|
1081 | * to check more often for insufficient lookahead. |
---|
1082 | */ |
---|
1083 | Assert(scan[2] == match[2], "scan[2]?"); |
---|
1084 | scan++, match++; |
---|
1085 | do { |
---|
1086 | } while (*(ushf*)(scan+=2) == *(ushf*)(match+=2) && |
---|
1087 | *(ushf*)(scan+=2) == *(ushf*)(match+=2) && |
---|
1088 | *(ushf*)(scan+=2) == *(ushf*)(match+=2) && |
---|
1089 | *(ushf*)(scan+=2) == *(ushf*)(match+=2) && |
---|
1090 | scan < strend); |
---|
1091 | /* The funny "do {}" generates better code on most compilers */ |
---|
1092 | |
---|
1093 | /* Here, scan <= window+strstart+257 */ |
---|
1094 | Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan"); |
---|
1095 | if (*scan == *match) scan++; |
---|
1096 | |
---|
1097 | len = (MAX_MATCH - 1) - (int)(strend-scan); |
---|
1098 | scan = strend - (MAX_MATCH-1); |
---|
1099 | |
---|
1100 | #else /* UNALIGNED_OK */ |
---|
1101 | |
---|
1102 | if (match[best_len] != scan_end || |
---|
1103 | match[best_len-1] != scan_end1 || |
---|
1104 | *match != *scan || |
---|
1105 | *++match != scan[1]) continue; |
---|
1106 | |
---|
1107 | /* The check at best_len-1 can be removed because it will be made |
---|
1108 | * again later. (This heuristic is not always a win.) |
---|
1109 | * It is not necessary to compare scan[2] and match[2] since they |
---|
1110 | * are always equal when the other bytes match, given that |
---|
1111 | * the hash keys are equal and that HASH_BITS >= 8. |
---|
1112 | */ |
---|
1113 | scan += 2, match++; |
---|
1114 | Assert(*scan == *match, "match[2]?"); |
---|
1115 | |
---|
1116 | /* We check for insufficient lookahead only every 8th comparison; |
---|
1117 | * the 256th check will be made at strstart+258. |
---|
1118 | */ |
---|
1119 | do { |
---|
1120 | } while (*++scan == *++match && *++scan == *++match && |
---|
1121 | *++scan == *++match && *++scan == *++match && |
---|
1122 | *++scan == *++match && *++scan == *++match && |
---|
1123 | *++scan == *++match && *++scan == *++match && |
---|
1124 | scan < strend); |
---|
1125 | |
---|
1126 | Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan"); |
---|
1127 | |
---|
1128 | len = MAX_MATCH - (int)(strend - scan); |
---|
1129 | scan = strend - MAX_MATCH; |
---|
1130 | |
---|
1131 | #endif /* UNALIGNED_OK */ |
---|
1132 | |
---|
1133 | if (len > best_len) { |
---|
1134 | s->match_start = cur_match; |
---|
1135 | best_len = len; |
---|
1136 | if (len >= nice_match) break; |
---|
1137 | #ifdef UNALIGNED_OK |
---|
1138 | scan_end = *(ushf*)(scan+best_len-1); |
---|
1139 | #else |
---|
1140 | scan_end1 = scan[best_len-1]; |
---|
1141 | scan_end = scan[best_len]; |
---|
1142 | #endif |
---|
1143 | } |
---|
1144 | } while ((cur_match = prev[cur_match & wmask]) > limit |
---|
1145 | && --chain_length != 0); |
---|
1146 | |
---|
1147 | if ((uInt)best_len <= s->lookahead) return (uInt)best_len; |
---|
1148 | return s->lookahead; |
---|
1149 | } |
---|
1150 | #endif /* ASMV */ |
---|
1151 | #endif /* FASTEST */ |
---|
1152 | |
---|
1153 | /* --------------------------------------------------------------------------- |
---|
1154 | * Optimized version for level == 1 or strategy == Z_RLE only |
---|
1155 | */ |
---|
1156 | local uInt longest_match_fast(s, cur_match) |
---|
1157 | deflate_state *s; |
---|
1158 | IPos cur_match; /* current match */ |
---|
1159 | { |
---|
1160 | register Bytef *scan = s->window + s->strstart; /* current string */ |
---|
1161 | register Bytef *match; /* matched string */ |
---|
1162 | register int len; /* length of current match */ |
---|
1163 | register Bytef *strend = s->window + s->strstart + MAX_MATCH; |
---|
1164 | |
---|
1165 | /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16. |
---|
1166 | * It is easy to get rid of this optimization if necessary. |
---|
1167 | */ |
---|
1168 | Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever"); |
---|
1169 | |
---|
1170 | Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead"); |
---|
1171 | |
---|
1172 | Assert(cur_match < s->strstart, "no future"); |
---|
1173 | |
---|
1174 | match = s->window + cur_match; |
---|
1175 | |
---|
1176 | /* Return failure if the match length is less than 2: |
---|
1177 | */ |
---|
1178 | if (match[0] != scan[0] || match[1] != scan[1]) return MIN_MATCH-1; |
---|
1179 | |
---|
1180 | /* The check at best_len-1 can be removed because it will be made |
---|
1181 | * again later. (This heuristic is not always a win.) |
---|
1182 | * It is not necessary to compare scan[2] and match[2] since they |
---|
1183 | * are always equal when the other bytes match, given that |
---|
1184 | * the hash keys are equal and that HASH_BITS >= 8. |
---|
1185 | */ |
---|
1186 | scan += 2, match += 2; |
---|
1187 | Assert(*scan == *match, "match[2]?"); |
---|
1188 | |
---|
1189 | /* We check for insufficient lookahead only every 8th comparison; |
---|
1190 | * the 256th check will be made at strstart+258. |
---|
1191 | */ |
---|
1192 | do { |
---|
1193 | } while (*++scan == *++match && *++scan == *++match && |
---|
1194 | *++scan == *++match && *++scan == *++match && |
---|
1195 | *++scan == *++match && *++scan == *++match && |
---|
1196 | *++scan == *++match && *++scan == *++match && |
---|
1197 | scan < strend); |
---|
1198 | |
---|
1199 | Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan"); |
---|
1200 | |
---|
1201 | len = MAX_MATCH - (int)(strend - scan); |
---|
1202 | |
---|
1203 | if (len < MIN_MATCH) return MIN_MATCH - 1; |
---|
1204 | |
---|
1205 | s->match_start = cur_match; |
---|
1206 | return (uInt)len <= s->lookahead ? (uInt)len : s->lookahead; |
---|
1207 | } |
---|
1208 | |
---|
1209 | #ifdef DEBUG |
---|
1210 | /* =========================================================================== |
---|
1211 | * Check that the match at match_start is indeed a match. |
---|
1212 | */ |
---|
1213 | local void check_match(s, start, match, length) |
---|
1214 | deflate_state *s; |
---|
1215 | IPos start, match; |
---|
1216 | int length; |
---|
1217 | { |
---|
1218 | /* check that the match is indeed a match */ |
---|
1219 | if (zmemcmp(s->window + match, |
---|
1220 | s->window + start, length) != EQUAL) { |
---|
1221 | fprintf(stderr, " start %u, match %u, length %d\n", |
---|
1222 | start, match, length); |
---|
1223 | do { |
---|
1224 | fprintf(stderr, "%c%c", s->window[match++], s->window[start++]); |
---|
1225 | } while (--length != 0); |
---|
1226 | z_error("invalid match"); |
---|
1227 | } |
---|
1228 | if (z_verbose > 1) { |
---|
1229 | fprintf(stderr,"\\[%d,%d]", start-match, length); |
---|
1230 | do { putc(s->window[start++], stderr); } while (--length != 0); |
---|
1231 | } |
---|
1232 | } |
---|
1233 | #else |
---|
1234 | # define check_match(s, start, match, length) |
---|
1235 | #endif /* DEBUG */ |
---|
1236 | |
---|
1237 | /* =========================================================================== |
---|
1238 | * Fill the window when the lookahead becomes insufficient. |
---|
1239 | * Updates strstart and lookahead. |
---|
1240 | * |
---|
1241 | * IN assertion: lookahead < MIN_LOOKAHEAD |
---|
1242 | * OUT assertions: strstart <= window_size-MIN_LOOKAHEAD |
---|
1243 | * At least one byte has been read, or avail_in == 0; reads are |
---|
1244 | * performed for at least two bytes (required for the zip translate_eol |
---|
1245 | * option -- not supported here). |
---|
1246 | */ |
---|
1247 | local void fill_window(s) |
---|
1248 | deflate_state *s; |
---|
1249 | { |
---|
1250 | register unsigned n, m; |
---|
1251 | register Posf *p; |
---|
1252 | unsigned more; /* Amount of free space at the end of the window. */ |
---|
1253 | uInt wsize = s->w_size; |
---|
1254 | |
---|
1255 | do { |
---|
1256 | more = (unsigned)(s->window_size -(ulg)s->lookahead -(ulg)s->strstart); |
---|
1257 | |
---|
1258 | /* Deal with !@#$% 64K limit: */ |
---|
1259 | if (sizeof(int) <= 2) { |
---|
1260 | if (more == 0 && s->strstart == 0 && s->lookahead == 0) { |
---|
1261 | more = wsize; |
---|
1262 | |
---|
1263 | } else if (more == (unsigned)(-1)) { |
---|
1264 | /* Very unlikely, but possible on 16 bit machine if |
---|
1265 | * strstart == 0 && lookahead == 1 (input done a byte at time) |
---|
1266 | */ |
---|
1267 | more--; |
---|
1268 | } |
---|
1269 | } |
---|
1270 | |
---|
1271 | /* If the window is almost full and there is insufficient lookahead, |
---|
1272 | * move the upper half to the lower one to make room in the upper half. |
---|
1273 | */ |
---|
1274 | if (s->strstart >= wsize+MAX_DIST(s)) { |
---|
1275 | |
---|
1276 | zmemcpy(s->window, s->window+wsize, (unsigned)wsize); |
---|
1277 | s->match_start -= wsize; |
---|
1278 | s->strstart -= wsize; /* we now have strstart >= MAX_DIST */ |
---|
1279 | s->block_start -= (long) wsize; |
---|
1280 | |
---|
1281 | /* Slide the hash table (could be avoided with 32 bit values |
---|
1282 | at the expense of memory usage). We slide even when level == 0 |
---|
1283 | to keep the hash table consistent if we switch back to level > 0 |
---|
1284 | later. (Using level 0 permanently is not an optimal usage of |
---|
1285 | zlib, so we don't care about this pathological case.) |
---|
1286 | */ |
---|
1287 | n = s->hash_size; |
---|
1288 | p = &s->head[n]; |
---|
1289 | do { |
---|
1290 | m = *--p; |
---|
1291 | *p = (Pos)(m >= wsize ? m-wsize : NIL); |
---|
1292 | } while (--n); |
---|
1293 | |
---|
1294 | n = wsize; |
---|
1295 | #ifndef FASTEST |
---|
1296 | p = &s->prev[n]; |
---|
1297 | do { |
---|
1298 | m = *--p; |
---|
1299 | *p = (Pos)(m >= wsize ? m-wsize : NIL); |
---|
1300 | /* If n is not on any hash chain, prev[n] is garbage but |
---|
1301 | * its value will never be used. |
---|
1302 | */ |
---|
1303 | } while (--n); |
---|
1304 | #endif |
---|
1305 | more += wsize; |
---|
1306 | } |
---|
1307 | if (s->strm->avail_in == 0) return; |
---|
1308 | |
---|
1309 | /* If there was no sliding: |
---|
1310 | * strstart <= WSIZE+MAX_DIST-1 && lookahead <= MIN_LOOKAHEAD - 1 && |
---|
1311 | * more == window_size - lookahead - strstart |
---|
1312 | * => more >= window_size - (MIN_LOOKAHEAD-1 + WSIZE + MAX_DIST-1) |
---|
1313 | * => more >= window_size - 2*WSIZE + 2 |
---|
1314 | * In the BIG_MEM or MMAP case (not yet supported), |
---|
1315 | * window_size == input_size + MIN_LOOKAHEAD && |
---|
1316 | * strstart + s->lookahead <= input_size => more >= MIN_LOOKAHEAD. |
---|
1317 | * Otherwise, window_size == 2*WSIZE so more >= 2. |
---|
1318 | * If there was sliding, more >= WSIZE. So in all cases, more >= 2. |
---|
1319 | */ |
---|
1320 | Assert(more >= 2, "more < 2"); |
---|
1321 | |
---|
1322 | n = read_buf(s->strm, s->window + s->strstart + s->lookahead, more); |
---|
1323 | s->lookahead += n; |
---|
1324 | |
---|
1325 | /* Initialize the hash value now that we have some input: */ |
---|
1326 | if (s->lookahead >= MIN_MATCH) { |
---|
1327 | s->ins_h = s->window[s->strstart]; |
---|
1328 | UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]); |
---|
1329 | #if MIN_MATCH != 3 |
---|
1330 | Call UPDATE_HASH() MIN_MATCH-3 more times |
---|
1331 | #endif |
---|
1332 | } |
---|
1333 | /* If the whole input has less than MIN_MATCH bytes, ins_h is garbage, |
---|
1334 | * but this is not important since only literal bytes will be emitted. |
---|
1335 | */ |
---|
1336 | |
---|
1337 | } while (s->lookahead < MIN_LOOKAHEAD && s->strm->avail_in != 0); |
---|
1338 | } |
---|
1339 | |
---|
1340 | /* =========================================================================== |
---|
1341 | * Flush the current block, with given end-of-file flag. |
---|
1342 | * IN assertion: strstart is set to the end of the current match. |
---|
1343 | */ |
---|
1344 | #define FLUSH_BLOCK_ONLY(s, eof) { \ |
---|
1345 | _tr_flush_block(s, (s->block_start >= 0L ? \ |
---|
1346 | (charf *)&s->window[(unsigned)s->block_start] : \ |
---|
1347 | (charf *)Z_NULL), \ |
---|
1348 | (ulg)((long)s->strstart - s->block_start), \ |
---|
1349 | (eof)); \ |
---|
1350 | s->block_start = s->strstart; \ |
---|
1351 | flush_pending(s->strm); \ |
---|
1352 | Tracev((stderr,"[FLUSH]")); \ |
---|
1353 | } |
---|
1354 | |
---|
1355 | /* Same but force premature exit if necessary. */ |
---|
1356 | #define FLUSH_BLOCK(s, eof) { \ |
---|
1357 | FLUSH_BLOCK_ONLY(s, eof); \ |
---|
1358 | if (s->strm->avail_out == 0) return (eof) ? finish_started : need_more; \ |
---|
1359 | } |
---|
1360 | |
---|
1361 | /* =========================================================================== |
---|
1362 | * Copy without compression as much as possible from the input stream, return |
---|
1363 | * the current block state. |
---|
1364 | * This function does not insert new strings in the dictionary since |
---|
1365 | * uncompressible data is probably not useful. This function is used |
---|
1366 | * only for the level=0 compression option. |
---|
1367 | * NOTE: this function should be optimized to avoid extra copying from |
---|
1368 | * window to pending_buf. |
---|
1369 | */ |
---|
1370 | local block_state deflate_stored(s, flush) |
---|
1371 | deflate_state *s; |
---|
1372 | int flush; |
---|
1373 | { |
---|
1374 | /* Stored blocks are limited to 0xffff bytes, pending_buf is limited |
---|
1375 | * to pending_buf_size, and each stored block has a 5 byte header: |
---|
1376 | */ |
---|
1377 | ulg max_block_size = 0xffff; |
---|
1378 | ulg max_start; |
---|
1379 | |
---|
1380 | if (max_block_size > s->pending_buf_size - 5) { |
---|
1381 | max_block_size = s->pending_buf_size - 5; |
---|
1382 | } |
---|
1383 | |
---|
1384 | /* Copy as much as possible from input to output: */ |
---|
1385 | for (;;) { |
---|
1386 | /* Fill the window as much as possible: */ |
---|
1387 | if (s->lookahead <= 1) { |
---|
1388 | |
---|
1389 | Assert(s->strstart < s->w_size+MAX_DIST(s) || |
---|
1390 | s->block_start >= (long)s->w_size, "slide too late"); |
---|
1391 | |
---|
1392 | fill_window(s); |
---|
1393 | if (s->lookahead == 0 && flush == Z_NO_FLUSH) return need_more; |
---|
1394 | |
---|
1395 | if (s->lookahead == 0) break; /* flush the current block */ |
---|
1396 | } |
---|
1397 | Assert(s->block_start >= 0L, "block gone"); |
---|
1398 | |
---|
1399 | s->strstart += s->lookahead; |
---|
1400 | s->lookahead = 0; |
---|
1401 | |
---|
1402 | /* Emit a stored block if pending_buf will be full: */ |
---|
1403 | max_start = s->block_start + max_block_size; |
---|
1404 | if (s->strstart == 0 || (ulg)s->strstart >= max_start) { |
---|
1405 | /* strstart == 0 is possible when wraparound on 16-bit machine */ |
---|
1406 | s->lookahead = (uInt)(s->strstart - max_start); |
---|
1407 | s->strstart = (uInt)max_start; |
---|
1408 | FLUSH_BLOCK(s, 0); |
---|
1409 | } |
---|
1410 | /* Flush if we may have to slide, otherwise block_start may become |
---|
1411 | * negative and the data will be gone: |
---|
1412 | */ |
---|
1413 | if (s->strstart - (uInt)s->block_start >= MAX_DIST(s)) { |
---|
1414 | FLUSH_BLOCK(s, 0); |
---|
1415 | } |
---|
1416 | } |
---|
1417 | FLUSH_BLOCK(s, flush == Z_FINISH); |
---|
1418 | return flush == Z_FINISH ? finish_done : block_done; |
---|
1419 | } |
---|
1420 | |
---|
1421 | /* =========================================================================== |
---|
1422 | * Compress as much as possible from the input stream, return the current |
---|
1423 | * block state. |
---|
1424 | * This function does not perform lazy evaluation of matches and inserts |
---|
1425 | * new strings in the dictionary only for unmatched strings or for short |
---|
1426 | * matches. It is used only for the fast compression options. |
---|
1427 | */ |
---|
1428 | local block_state deflate_fast(s, flush) |
---|
1429 | deflate_state *s; |
---|
1430 | int flush; |
---|
1431 | { |
---|
1432 | IPos hash_head = NIL; /* head of the hash chain */ |
---|
1433 | int bflush; /* set if current block must be flushed */ |
---|
1434 | |
---|
1435 | for (;;) { |
---|
1436 | /* Make sure that we always have enough lookahead, except |
---|
1437 | * at the end of the input file. We need MAX_MATCH bytes |
---|
1438 | * for the next match, plus MIN_MATCH bytes to insert the |
---|
1439 | * string following the next match. |
---|
1440 | */ |
---|
1441 | if (s->lookahead < MIN_LOOKAHEAD) { |
---|
1442 | fill_window(s); |
---|
1443 | if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) { |
---|
1444 | return need_more; |
---|
1445 | } |
---|
1446 | if (s->lookahead == 0) break; /* flush the current block */ |
---|
1447 | } |
---|
1448 | |
---|
1449 | /* Insert the string window[strstart .. strstart+2] in the |
---|
1450 | * dictionary, and set hash_head to the head of the hash chain: |
---|
1451 | */ |
---|
1452 | if (s->lookahead >= MIN_MATCH) { |
---|
1453 | INSERT_STRING(s, s->strstart, hash_head); |
---|
1454 | } |
---|
1455 | |
---|
1456 | /* Find the longest match, discarding those <= prev_length. |
---|
1457 | * At this point we have always match_length < MIN_MATCH |
---|
1458 | */ |
---|
1459 | if (hash_head != NIL && s->strstart - hash_head <= MAX_DIST(s)) { |
---|
1460 | /* To simplify the code, we prevent matches with the string |
---|
1461 | * of window index 0 (in particular we have to avoid a match |
---|
1462 | * of the string with itself at the start of the input file). |
---|
1463 | */ |
---|
1464 | #ifdef FASTEST |
---|
1465 | if ((s->strategy != Z_HUFFMAN_ONLY && s->strategy != Z_RLE) || |
---|
1466 | (s->strategy == Z_RLE && s->strstart - hash_head == 1)) { |
---|
1467 | s->match_length = longest_match_fast (s, hash_head); |
---|
1468 | } |
---|
1469 | #else |
---|
1470 | if (s->strategy != Z_HUFFMAN_ONLY && s->strategy != Z_RLE) { |
---|
1471 | s->match_length = longest_match (s, hash_head); |
---|
1472 | } else if (s->strategy == Z_RLE && s->strstart - hash_head == 1) { |
---|
1473 | s->match_length = longest_match_fast (s, hash_head); |
---|
1474 | } |
---|
1475 | #endif |
---|
1476 | /* longest_match() or longest_match_fast() sets match_start */ |
---|
1477 | } |
---|
1478 | if (s->match_length >= MIN_MATCH) { |
---|
1479 | check_match(s, s->strstart, s->match_start, s->match_length); |
---|
1480 | |
---|
1481 | _tr_tally_dist(s, s->strstart - s->match_start, |
---|
1482 | s->match_length - MIN_MATCH, bflush); |
---|
1483 | |
---|
1484 | s->lookahead -= s->match_length; |
---|
1485 | |
---|
1486 | /* Insert new strings in the hash table only if the match length |
---|
1487 | * is not too large. This saves time but degrades compression. |
---|
1488 | */ |
---|
1489 | #ifndef FASTEST |
---|
1490 | if (s->match_length <= s->max_insert_length && |
---|
1491 | s->lookahead >= MIN_MATCH) { |
---|
1492 | s->match_length--; /* string at strstart already in table */ |
---|
1493 | do { |
---|
1494 | s->strstart++; |
---|
1495 | INSERT_STRING(s, s->strstart, hash_head); |
---|
1496 | /* strstart never exceeds WSIZE-MAX_MATCH, so there are |
---|
1497 | * always MIN_MATCH bytes ahead. |
---|
1498 | */ |
---|
1499 | } while (--s->match_length != 0); |
---|
1500 | s->strstart++; |
---|
1501 | } else |
---|
1502 | #endif |
---|
1503 | { |
---|
1504 | s->strstart += s->match_length; |
---|
1505 | s->match_length = 0; |
---|
1506 | s->ins_h = s->window[s->strstart]; |
---|
1507 | UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]); |
---|
1508 | #if MIN_MATCH != 3 |
---|
1509 | Call UPDATE_HASH() MIN_MATCH-3 more times |
---|
1510 | #endif |
---|
1511 | /* If lookahead < MIN_MATCH, ins_h is garbage, but it does not |
---|
1512 | * matter since it will be recomputed at next deflate call. |
---|
1513 | */ |
---|
1514 | } |
---|
1515 | } else { |
---|
1516 | /* No match, output a literal byte */ |
---|
1517 | Tracevv((stderr,"%c", s->window[s->strstart])); |
---|
1518 | _tr_tally_lit (s, s->window[s->strstart], bflush); |
---|
1519 | s->lookahead--; |
---|
1520 | s->strstart++; |
---|
1521 | } |
---|
1522 | if (bflush) FLUSH_BLOCK(s, 0); |
---|
1523 | } |
---|
1524 | FLUSH_BLOCK(s, flush == Z_FINISH); |
---|
1525 | return flush == Z_FINISH ? finish_done : block_done; |
---|
1526 | } |
---|
1527 | |
---|
1528 | #ifndef FASTEST |
---|
1529 | /* =========================================================================== |
---|
1530 | * Same as above, but achieves better compression. We use a lazy |
---|
1531 | * evaluation for matches: a match is finally adopted only if there is |
---|
1532 | * no better match at the next window position. |
---|
1533 | */ |
---|
1534 | local block_state deflate_slow(s, flush) |
---|
1535 | deflate_state *s; |
---|
1536 | int flush; |
---|
1537 | { |
---|
1538 | IPos hash_head = NIL; /* head of hash chain */ |
---|
1539 | int bflush; /* set if current block must be flushed */ |
---|
1540 | |
---|
1541 | /* Process the input block. */ |
---|
1542 | for (;;) { |
---|
1543 | /* Make sure that we always have enough lookahead, except |
---|
1544 | * at the end of the input file. We need MAX_MATCH bytes |
---|
1545 | * for the next match, plus MIN_MATCH bytes to insert the |
---|
1546 | * string following the next match. |
---|
1547 | */ |
---|
1548 | if (s->lookahead < MIN_LOOKAHEAD) { |
---|
1549 | fill_window(s); |
---|
1550 | if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) { |
---|
1551 | return need_more; |
---|
1552 | } |
---|
1553 | if (s->lookahead == 0) break; /* flush the current block */ |
---|
1554 | } |
---|
1555 | |
---|
1556 | /* Insert the string window[strstart .. strstart+2] in the |
---|
1557 | * dictionary, and set hash_head to the head of the hash chain: |
---|
1558 | */ |
---|
1559 | if (s->lookahead >= MIN_MATCH) { |
---|
1560 | INSERT_STRING(s, s->strstart, hash_head); |
---|
1561 | } |
---|
1562 | |
---|
1563 | /* Find the longest match, discarding those <= prev_length. |
---|
1564 | */ |
---|
1565 | s->prev_length = s->match_length, s->prev_match = s->match_start; |
---|
1566 | s->match_length = MIN_MATCH-1; |
---|
1567 | |
---|
1568 | if (hash_head != NIL && s->prev_length < s->max_lazy_match && |
---|
1569 | s->strstart - hash_head <= MAX_DIST(s)) { |
---|
1570 | /* To simplify the code, we prevent matches with the string |
---|
1571 | * of window index 0 (in particular we have to avoid a match |
---|
1572 | * of the string with itself at the start of the input file). |
---|
1573 | */ |
---|
1574 | if (s->strategy != Z_HUFFMAN_ONLY && s->strategy != Z_RLE) { |
---|
1575 | s->match_length = longest_match (s, hash_head); |
---|
1576 | } else if (s->strategy == Z_RLE && s->strstart - hash_head == 1) { |
---|
1577 | s->match_length = longest_match_fast (s, hash_head); |
---|
1578 | } |
---|
1579 | /* longest_match() or longest_match_fast() sets match_start */ |
---|
1580 | |
---|
1581 | if (s->match_length <= 5 && (s->strategy == Z_FILTERED |
---|
1582 | #if TOO_FAR <= 32767 |
---|
1583 | || (s->match_length == MIN_MATCH && |
---|
1584 | s->strstart - s->match_start > TOO_FAR) |
---|
1585 | #endif |
---|
1586 | )) { |
---|
1587 | |
---|
1588 | /* If prev_match is also MIN_MATCH, match_start is garbage |
---|
1589 | * but we will ignore the current match anyway. |
---|
1590 | */ |
---|
1591 | s->match_length = MIN_MATCH-1; |
---|
1592 | } |
---|
1593 | } |
---|
1594 | /* If there was a match at the previous step and the current |
---|
1595 | * match is not better, output the previous match: |
---|
1596 | */ |
---|
1597 | if (s->prev_length >= MIN_MATCH && s->match_length <= s->prev_length) { |
---|
1598 | uInt max_insert = s->strstart + s->lookahead - MIN_MATCH; |
---|
1599 | /* Do not insert strings in hash table beyond this. */ |
---|
1600 | |
---|
1601 | check_match(s, s->strstart-1, s->prev_match, s->prev_length); |
---|
1602 | |
---|
1603 | _tr_tally_dist(s, s->strstart -1 - s->prev_match, |
---|
1604 | s->prev_length - MIN_MATCH, bflush); |
---|
1605 | |
---|
1606 | /* Insert in hash table all strings up to the end of the match. |
---|
1607 | * strstart-1 and strstart are already inserted. If there is not |
---|
1608 | * enough lookahead, the last two strings are not inserted in |
---|
1609 | * the hash table. |
---|
1610 | */ |
---|
1611 | s->lookahead -= s->prev_length-1; |
---|
1612 | s->prev_length -= 2; |
---|
1613 | do { |
---|
1614 | if (++s->strstart <= max_insert) { |
---|
1615 | INSERT_STRING(s, s->strstart, hash_head); |
---|
1616 | } |
---|
1617 | } while (--s->prev_length != 0); |
---|
1618 | s->match_available = 0; |
---|
1619 | s->match_length = MIN_MATCH-1; |
---|
1620 | s->strstart++; |
---|
1621 | |
---|
1622 | if (bflush) FLUSH_BLOCK(s, 0); |
---|
1623 | |
---|
1624 | } else if (s->match_available) { |
---|
1625 | /* If there was no match at the previous position, output a |
---|
1626 | * single literal. If there was a match but the current match |
---|
1627 | * is longer, truncate the previous match to a single literal. |
---|
1628 | */ |
---|
1629 | Tracevv((stderr,"%c", s->window[s->strstart-1])); |
---|
1630 | _tr_tally_lit(s, s->window[s->strstart-1], bflush); |
---|
1631 | if (bflush) { |
---|
1632 | FLUSH_BLOCK_ONLY(s, 0); |
---|
1633 | } |
---|
1634 | s->strstart++; |
---|
1635 | s->lookahead--; |
---|
1636 | if (s->strm->avail_out == 0) return need_more; |
---|
1637 | } else { |
---|
1638 | /* There is no previous match to compare with, wait for |
---|
1639 | * the next step to decide. |
---|
1640 | */ |
---|
1641 | s->match_available = 1; |
---|
1642 | s->strstart++; |
---|
1643 | s->lookahead--; |
---|
1644 | } |
---|
1645 | } |
---|
1646 | Assert (flush != Z_NO_FLUSH, "no flush?"); |
---|
1647 | if (s->match_available) { |
---|
1648 | Tracevv((stderr,"%c", s->window[s->strstart-1])); |
---|
1649 | _tr_tally_lit(s, s->window[s->strstart-1], bflush); |
---|
1650 | s->match_available = 0; |
---|
1651 | } |
---|
1652 | FLUSH_BLOCK(s, flush == Z_FINISH); |
---|
1653 | return flush == Z_FINISH ? finish_done : block_done; |
---|
1654 | } |
---|
1655 | #endif /* FASTEST */ |
---|