1 /*
2 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
3 *
4 * This code is free software; you can redistribute it and/or modify it
5 * under the terms of the GNU General Public License version 2 only, as
6 * published by the Free Software Foundation. Oracle designates this
7 * particular file as subject to the "Classpath" exception as provided
8 * by Oracle in the LICENSE file that accompanied this code.
9 *
10 * This code is distributed in the hope that it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
13 * version 2 for more details (a copy is included in the LICENSE file that
14 * accompanied this code).
15 *
16 * You should have received a copy of the GNU General Public License version
17 * 2 along with this work; if not, write to the Free Software Foundation,
18 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
19 *
20 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
21 * or visit www.oracle.com if you need additional information or have any
22 * questions.
23 */
24
25 /* deflate.h -- internal compression state
26 * Copyright (C) 1995-2010 Jean-loup Gailly
27 * For conditions of distribution and use, see copyright notice in zlib.h
28 */
29
30 /* WARNING: this file should *not* be used by applications. It is
31 part of the implementation of the compression library and is
32 subject to change. Applications should only use zlib.h.
33 */
34
35 /* @(#) $Id$ */
36
37 #ifndef DEFLATE_H
38 #define DEFLATE_H
39
40 #include "zutil.h"
41
42 /* define NO_GZIP when compiling if you want to disable gzip header and
43 trailer creation by deflate(). NO_GZIP would be used to avoid linking in
44 the crc code when it is not needed. For shared libraries, gzip encoding
45 should be left enabled. */
46 #ifndef NO_GZIP
47 # define GZIP
48 #endif
49
50 /* ===========================================================================
51 * Internal compression state.
52 */
53
54 #define LENGTH_CODES 29
55 /* number of length codes, not counting the special END_BLOCK code */
56
57 #define LITERALS 256
58 /* number of literal bytes 0..255 */
59
60 #define L_CODES (LITERALS+1+LENGTH_CODES)
61 /* number of Literal or Length codes, including the END_BLOCK code */
62
63 #define D_CODES 30
64 /* number of distance codes */
65
66 #define BL_CODES 19
67 /* number of codes used to transfer the bit lengths */
68
69 #define HEAP_SIZE (2*L_CODES+1)
70 /* maximum heap size */
71
72 #define MAX_BITS 15
73 /* All codes must not exceed MAX_BITS bits */
74
75 #define INIT_STATE 42
76 #define EXTRA_STATE 69
77 #define NAME_STATE 73
78 #define COMMENT_STATE 91
79 #define HCRC_STATE 103
80 #define BUSY_STATE 113
81 #define FINISH_STATE 666
82 /* Stream status */
83
84
85 /* Data structure describing a single value and its code string. */
86 typedef struct ct_data_s {
87 union {
88 ush freq; /* frequency count */
89 ush code; /* bit string */
90 } fc;
91 union {
92 ush dad; /* father node in Huffman tree */
93 ush len; /* length of bit string */
94 } dl;
95 } FAR ct_data;
96
97 #define Freq fc.freq
98 #define Code fc.code
99 #define Dad dl.dad
100 #define Len dl.len
101
102 typedef struct static_tree_desc_s static_tree_desc;
103
104 typedef struct tree_desc_s {
105 ct_data *dyn_tree; /* the dynamic tree */
106 int max_code; /* largest code with non zero frequency */
107 static_tree_desc *stat_desc; /* the corresponding static tree */
108 } FAR tree_desc;
109
110 typedef ush Pos;
111 typedef Pos FAR Posf;
112 typedef unsigned IPos;
113
114 /* A Pos is an index in the character window. We use short instead of int to
115 * save space in the various tables. IPos is used only for parameter passing.
116 */
117
118 typedef struct internal_state {
119 z_streamp strm; /* pointer back to this zlib stream */
120 int status; /* as the name implies */
121 Bytef *pending_buf; /* output still pending */
122 ulg pending_buf_size; /* size of pending_buf */
123 Bytef *pending_out; /* next pending byte to output to the stream */
124 uInt pending; /* nb of bytes in the pending buffer */
125 int wrap; /* bit 0 true for zlib, bit 1 true for gzip */
126 gz_headerp gzhead; /* gzip header information to write */
127 uInt gzindex; /* where in extra, name, or comment */
128 Byte method; /* STORED (for zip only) or DEFLATED */
129 int last_flush; /* value of flush param for previous deflate call */
130
131 /* used by deflate.c: */
132
133 uInt w_size; /* LZ77 window size (32K by default) */
134 uInt w_bits; /* log2(w_size) (8..16) */
135 uInt w_mask; /* w_size - 1 */
136
137 Bytef *window;
138 /* Sliding window. Input bytes are read into the second half of the window,
139 * and move to the first half later to keep a dictionary of at least wSize
140 * bytes. With this organization, matches are limited to a distance of
141 * wSize-MAX_MATCH bytes, but this ensures that IO is always
142 * performed with a length multiple of the block size. Also, it limits
143 * the window size to 64K, which is quite useful on MSDOS.
144 * To do: use the user input buffer as sliding window.
145 */
146
147 ulg window_size;
148 /* Actual size of window: 2*wSize, except when the user input buffer
149 * is directly used as sliding window.
150 */
151
152 Posf *prev;
153 /* Link to older string with same hash index. To limit the size of this
154 * array to 64K, this link is maintained only for the last 32K strings.
155 * An index in this array is thus a window index modulo 32K.
156 */
157
158 Posf *head; /* Heads of the hash chains or NIL. */
159
160 uInt ins_h; /* hash index of string to be inserted */
161 uInt hash_size; /* number of elements in hash table */
162 uInt hash_bits; /* log2(hash_size) */
163 uInt hash_mask; /* hash_size-1 */
164
165 uInt hash_shift;
166 /* Number of bits by which ins_h must be shifted at each input
167 * step. It must be such that after MIN_MATCH steps, the oldest
168 * byte no longer takes part in the hash key, that is:
169 * hash_shift * MIN_MATCH >= hash_bits
170 */
171
172 long block_start;
173 /* Window position at the beginning of the current output block. Gets
174 * negative when the window is moved backwards.
175 */
176
177 uInt match_length; /* length of best match */
178 IPos prev_match; /* previous match */
179 int match_available; /* set if previous match exists */
180 uInt strstart; /* start of string to insert */
181 uInt match_start; /* start of matching string */
182 uInt lookahead; /* number of valid bytes ahead in window */
183
184 uInt prev_length;
185 /* Length of the best match at previous step. Matches not greater than this
186 * are discarded. This is used in the lazy match evaluation.
187 */
188
189 uInt max_chain_length;
190 /* To speed up deflation, hash chains are never searched beyond this
191 * length. A higher limit improves compression ratio but degrades the
192 * speed.
193 */
194
195 uInt max_lazy_match;
196 /* Attempt to find a better match only when the current match is strictly
197 * smaller than this value. This mechanism is used only for compression
198 * levels >= 4.
199 */
200 # define max_insert_length max_lazy_match
201 /* Insert new strings in the hash table only if the match length is not
202 * greater than this length. This saves time but degrades compression.
203 * max_insert_length is used only for compression levels <= 3.
204 */
205
206 int level; /* compression level (1..9) */
207 int strategy; /* favor or force Huffman coding*/
208
209 uInt good_match;
210 /* Use a faster search when the previous match is longer than this */
211
212 int nice_match; /* Stop searching when current match exceeds this */
213
214 /* used by trees.c: */
215 /* Didn't use ct_data typedef below to supress compiler warning */
216 struct ct_data_s dyn_ltree[HEAP_SIZE]; /* literal and length tree */
217 struct ct_data_s dyn_dtree[2*D_CODES+1]; /* distance tree */
218 struct ct_data_s bl_tree[2*BL_CODES+1]; /* Huffman tree for bit lengths */
219
220 struct tree_desc_s l_desc; /* desc. for literal tree */
221 struct tree_desc_s d_desc; /* desc. for distance tree */
222 struct tree_desc_s bl_desc; /* desc. for bit length tree */
223
224 ush bl_count[MAX_BITS+1];
225 /* number of codes at each bit length for an optimal tree */
226
227 int heap[2*L_CODES+1]; /* heap used to build the Huffman trees */
228 int heap_len; /* number of elements in the heap */
229 int heap_max; /* element of largest frequency */
230 /* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used.
231 * The same heap array is used to build all trees.
232 */
233
234 uch depth[2*L_CODES+1];
235 /* Depth of each subtree used as tie breaker for trees of equal frequency
236 */
237
238 uchf *l_buf; /* buffer for literals or lengths */
239
240 uInt lit_bufsize;
241 /* Size of match buffer for literals/lengths. There are 4 reasons for
242 * limiting lit_bufsize to 64K:
243 * - frequencies can be kept in 16 bit counters
244 * - if compression is not successful for the first block, all input
245 * data is still in the window so we can still emit a stored block even
246 * when input comes from standard input. (This can also be done for
247 * all blocks if lit_bufsize is not greater than 32K.)
248 * - if compression is not successful for a file smaller than 64K, we can
249 * even emit a stored file instead of a stored block (saving 5 bytes).
250 * This is applicable only for zip (not gzip or zlib).
251 * - creating new Huffman trees less frequently may not provide fast
252 * adaptation to changes in the input data statistics. (Take for
253 * example a binary file with poorly compressible code followed by
254 * a highly compressible string table.) Smaller buffer sizes give
255 * fast adaptation but have of course the overhead of transmitting
256 * trees more frequently.
257 * - I can't count above 4
258 */
259
260 uInt last_lit; /* running index in l_buf */
261
262 ushf *d_buf;
263 /* Buffer for distances. To simplify the code, d_buf and l_buf have
264 * the same number of elements. To use different lengths, an extra flag
265 * array would be necessary.
266 */
267
268 ulg opt_len; /* bit length of current block with optimal trees */
269 ulg static_len; /* bit length of current block with static trees */
270 uInt matches; /* number of string matches in current block */
271 int last_eob_len; /* bit length of EOB code for last block */
272
273 #ifdef DEBUG
274 ulg compressed_len; /* total bit length of compressed file mod 2^32 */
275 ulg bits_sent; /* bit length of compressed data sent mod 2^32 */
276 #endif
277
278 ush bi_buf;
279 /* Output buffer. bits are inserted starting at the bottom (least
280 * significant bits).
281 */
282 int bi_valid;
283 /* Number of valid bits in bi_buf. All bits above the last valid bit
284 * are always zero.
285 */
286
287 ulg high_water;
288 /* High water mark offset in window for initialized bytes -- bytes above
289 * this are set to zero in order to avoid memory check warnings when
290 * longest match routines access bytes past the input. This is then
291 * updated to the new high water mark.
292 */
293
294 } FAR deflate_state;
295
296 /* Output a byte on the stream.
297 * IN assertion: there is enough room in pending_buf.
298 */
299 #define put_byte(s, c) {s->pending_buf[s->pending++] = (c);}
300
301
302 #define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
303 /* Minimum amount of lookahead, except at the end of the input file.
304 * See deflate.c for comments about the MIN_MATCH+1.
305 */
306
307 #define MAX_DIST(s) ((s)->w_size-MIN_LOOKAHEAD)
308 /* In order to simplify the code, particularly on 16 bit machines, match
309 * distances are limited to MAX_DIST instead of WSIZE.
310 */
311
312 #define WIN_INIT MAX_MATCH
313 /* Number of bytes after end of data in window to initialize in order to avoid
314 memory checker errors from longest match routines */
315
316 /* in trees.c */
317 void ZLIB_INTERNAL _tr_init OF((deflate_state *s));
318 int ZLIB_INTERNAL _tr_tally OF((deflate_state *s, unsigned dist, unsigned lc));
319 void ZLIB_INTERNAL _tr_flush_block OF((deflate_state *s, charf *buf,
320 ulg stored_len, int last));
321 void ZLIB_INTERNAL _tr_align OF((deflate_state *s));
322 void ZLIB_INTERNAL _tr_stored_block OF((deflate_state *s, charf *buf,
323 ulg stored_len, int last));
324
325 #define d_code(dist) \
326 ((dist) < 256 ? _dist_code[dist] : _dist_code[256+((dist)>>7)])
327 /* Mapping from a distance to a distance code. dist is the distance - 1 and
328 * must not have side effects. _dist_code[256] and _dist_code[257] are never
329 * used.
330 */
331
332 #ifndef DEBUG
333 /* Inline versions of _tr_tally for speed: */
334
335 #if defined(GEN_TREES_H) || !defined(STDC)
336 extern uch ZLIB_INTERNAL _length_code[];
337 extern uch ZLIB_INTERNAL _dist_code[];
338 #else
339 extern const uch ZLIB_INTERNAL _length_code[];
340 extern const uch ZLIB_INTERNAL _dist_code[];
341 #endif
342
343 # define _tr_tally_lit(s, c, flush) \
344 { uch cc = (c); \
345 s->d_buf[s->last_lit] = 0; \
346 s->l_buf[s->last_lit++] = cc; \
347 s->dyn_ltree[cc].Freq++; \
348 flush = (s->last_lit == s->lit_bufsize-1); \
349 }
350 # define _tr_tally_dist(s, distance, length, flush) \
351 { uch len = (length); \
352 ush dist = (distance); \
353 s->d_buf[s->last_lit] = dist; \
354 s->l_buf[s->last_lit++] = len; \
355 dist--; \
356 s->dyn_ltree[_length_code[len]+LITERALS+1].Freq++; \
357 s->dyn_dtree[d_code(dist)].Freq++; \
358 flush = (s->last_lit == s->lit_bufsize-1); \
359 }
360 #else
361 # define _tr_tally_lit(s, c, flush) flush = _tr_tally(s, 0, c)
362 # define _tr_tally_dist(s, distance, length, flush) \
363 flush = _tr_tally(s, distance, length)
364 #endif
365
366 #endif /* DEFLATE_H */