1 /* 2 * Copyright (c) 2008 Apple Inc. All rights reserved. 3 * 4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ 5 * 6 * This file contains Original Code and/or Modifications of Original Code 7 * as defined in and that are subject to the Apple Public Source License 8 * Version 2.0 (the 'License'). You may not use this file except in 9 * compliance with the License. The rights granted to you under the License 10 * may not be used to create, or enable the creation or redistribution of, 11 * unlawful or unlicensed copies of an Apple operating system, or to 12 * circumvent, violate, or enable the circumvention or violation of, any 13 * terms of an Apple operating system software license agreement. 14 * 15 * Please obtain a copy of the License at 16 * http://www.opensource.apple.com/apsl/ and read it before using this file. 17 * 18 * The Original Code and all software distributed under the License are 19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER 20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, 21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, 22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. 23 * Please see the License for the specific language governing rights and 24 * limitations under the License. 25 * 26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ 27 */ 28 /* deflate.h -- internal compression state 29 * Copyright (C) 1995-2004 Jean-loup Gailly 30 * For conditions of distribution and use, see copyright notice in zlib.h 31 */ 32 33 /* WARNING: this file should *not* be used by applications. It is 34 part of the implementation of the compression library and is 35 subject to change. Applications should only use zlib.h. 36 */ 37 38 /* @(#) $Id$ */ 39 40 #ifndef DEFLATE_H 41 #define DEFLATE_H 42 43 #include "zutil.h" 44 45 /* define NO_GZIP when compiling if you want to disable gzip header and 46 trailer creation by deflate(). NO_GZIP would be used to avoid linking in 47 the crc code when it is not needed. For shared libraries, gzip encoding 48 should be left enabled. */ 49 #ifndef NO_GZIP 50 # define GZIP 51 #endif 52 53 /* =========================================================================== 54 * Internal compression state. 55 */ 56 57 #define LENGTH_CODES 29 58 /* number of length codes, not counting the special END_BLOCK code */ 59 60 #define LITERALS 256 61 /* number of literal bytes 0..255 */ 62 63 #define L_CODES (LITERALS+1+LENGTH_CODES) 64 /* number of Literal or Length codes, including the END_BLOCK code */ 65 66 #define D_CODES 30 67 /* number of distance codes */ 68 69 #define BL_CODES 19 70 /* number of codes used to transfer the bit lengths */ 71 72 #define HEAP_SIZE (2*L_CODES+1) 73 /* maximum heap size */ 74 75 #define MAX_BITS 15 76 /* All codes must not exceed MAX_BITS bits */ 77 78 #define INIT_STATE 42 79 #define EXTRA_STATE 69 80 #define NAME_STATE 73 81 #define COMMENT_STATE 91 82 #define HCRC_STATE 103 83 #define BUSY_STATE 113 84 #define FINISH_STATE 666 85 /* Stream status */ 86 87 88 /* Data structure describing a single value and its code string. */ 89 typedef struct ct_data_s { 90 union { 91 ush freq; /* frequency count */ 92 ush code; /* bit string */ 93 } fc; 94 union { 95 ush dad; /* father node in Huffman tree */ 96 ush len; /* length of bit string */ 97 } dl; 98 } FAR ct_data; 99 100 #define Freq fc.freq 101 #define Code fc.code 102 #define Dad dl.dad 103 #define Len dl.len 104 105 typedef struct static_tree_desc_s static_tree_desc; 106 107 typedef struct tree_desc_s { 108 ct_data *dyn_tree; /* the dynamic tree */ 109 int max_code; /* largest code with non zero frequency */ 110 static_tree_desc *stat_desc; /* the corresponding static tree */ 111 } FAR tree_desc; 112 113 typedef ush Pos; 114 typedef Pos FAR Posf; 115 typedef unsigned IPos; 116 117 /* A Pos is an index in the character window. We use short instead of int to 118 * save space in the various tables. IPos is used only for parameter passing. 119 */ 120 121 typedef struct internal_state { 122 z_streamp strm; /* pointer back to this zlib stream */ 123 z_input_func zinput; 124 z_output_func zoutput; 125 int status; /* as the name implies */ 126 Bytef *pending_buf; /* output still pending */ 127 ulg pending_buf_size; /* size of pending_buf */ 128 Bytef *pending_out; /* next pending byte to output to the stream */ 129 uInt pending; /* nb of bytes in the pending buffer */ 130 int wrap; /* bit 0 true for zlib, bit 1 true for gzip */ 131 gz_headerp gzhead; /* gzip header information to write */ 132 uInt gzindex; /* where in extra, name, or comment */ 133 Byte method; /* STORED (for zip only) or DEFLATED */ 134 int last_flush; /* value of flush param for previous deflate call */ 135 136 /* used by deflate.c: */ 137 138 uInt w_size; /* LZ77 window size (32K by default) */ 139 uInt w_bits; /* log2(w_size) (8..16) */ 140 uInt w_mask; /* w_size - 1 */ 141 142 Bytef *window; 143 /* Sliding window. Input bytes are read into the second half of the window, 144 * and move to the first half later to keep a dictionary of at least wSize 145 * bytes. With this organization, matches are limited to a distance of 146 * wSize-MAX_MATCH bytes, but this ensures that IO is always 147 * performed with a length multiple of the block size. Also, it limits 148 * the window size to 64K, which is quite useful on MSDOS. 149 * To do: use the user input buffer as sliding window. 150 */ 151 152 ulg window_size; 153 /* Actual size of window: 2*wSize, except when the user input buffer 154 * is directly used as sliding window. 155 */ 156 157 Posf *prev; 158 /* Link to older string with same hash index. To limit the size of this 159 * array to 64K, this link is maintained only for the last 32K strings. 160 * An index in this array is thus a window index modulo 32K. 161 */ 162 163 Posf *head; /* Heads of the hash chains or NIL. */ 164 165 uInt ins_h; /* hash index of string to be inserted */ 166 uInt hash_size; /* number of elements in hash table */ 167 uInt hash_bits; /* log2(hash_size) */ 168 uInt hash_mask; /* hash_size-1 */ 169 170 uInt hash_shift; 171 /* Number of bits by which ins_h must be shifted at each input 172 * step. It must be such that after MIN_MATCH steps, the oldest 173 * byte no longer takes part in the hash key, that is: 174 * hash_shift * MIN_MATCH >= hash_bits 175 */ 176 177 long block_start; 178 /* Window position at the beginning of the current output block. Gets 179 * negative when the window is moved backwards. 180 */ 181 182 uInt match_length; /* length of best match */ 183 IPos prev_match; /* previous match */ 184 int match_available; /* set if previous match exists */ 185 uInt strstart; /* start of string to insert */ 186 uInt match_start; /* start of matching string */ 187 uInt lookahead; /* number of valid bytes ahead in window */ 188 189 uInt prev_length; 190 /* Length of the best match at previous step. Matches not greater than this 191 * are discarded. This is used in the lazy match evaluation. 192 */ 193 194 uInt max_chain_length; 195 /* To speed up deflation, hash chains are never searched beyond this 196 * length. A higher limit improves compression ratio but degrades the 197 * speed. 198 */ 199 200 uInt max_lazy_match; 201 /* Attempt to find a better match only when the current match is strictly 202 * smaller than this value. This mechanism is used only for compression 203 * levels >= 4. 204 */ 205 # define max_insert_length max_lazy_match 206 /* Insert new strings in the hash table only if the match length is not 207 * greater than this length. This saves time but degrades compression. 208 * max_insert_length is used only for compression levels <= 3. 209 */ 210 211 int level; /* compression level (1..9) */ 212 int strategy; /* favor or force Huffman coding*/ 213 214 uInt good_match; 215 /* Use a faster search when the previous match is longer than this */ 216 217 int nice_match; /* Stop searching when current match exceeds this */ 218 219 /* used by trees.c: */ 220 /* Didn't use ct_data typedef below to supress compiler warning */ 221 struct ct_data_s dyn_ltree[HEAP_SIZE]; /* literal and length tree */ 222 struct ct_data_s dyn_dtree[2*D_CODES+1]; /* distance tree */ 223 struct ct_data_s bl_tree[2*BL_CODES+1]; /* Huffman tree for bit lengths */ 224 225 struct tree_desc_s l_desc; /* desc. for literal tree */ 226 struct tree_desc_s d_desc; /* desc. for distance tree */ 227 struct tree_desc_s bl_desc; /* desc. for bit length tree */ 228 229 ush bl_count[MAX_BITS+1]; 230 /* number of codes at each bit length for an optimal tree */ 231 232 int heap[2*L_CODES+1]; /* heap used to build the Huffman trees */ 233 int heap_len; /* number of elements in the heap */ 234 int heap_max; /* element of largest frequency */ 235 /* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used. 236 * The same heap array is used to build all trees. 237 */ 238 239 uch depth[2*L_CODES+1]; 240 /* Depth of each subtree used as tie breaker for trees of equal frequency 241 */ 242 243 uchf *l_buf; /* buffer for literals or lengths */ 244 245 uInt lit_bufsize; 246 /* Size of match buffer for literals/lengths. There are 4 reasons for 247 * limiting lit_bufsize to 64K: 248 * - frequencies can be kept in 16 bit counters 249 * - if compression is not successful for the first block, all input 250 * data is still in the window so we can still emit a stored block even 251 * when input comes from standard input. (This can also be done for 252 * all blocks if lit_bufsize is not greater than 32K.) 253 * - if compression is not successful for a file smaller than 64K, we can 254 * even emit a stored file instead of a stored block (saving 5 bytes). 255 * This is applicable only for zip (not gzip or zlib). 256 * - creating new Huffman trees less frequently may not provide fast 257 * adaptation to changes in the input data statistics. (Take for 258 * example a binary file with poorly compressible code followed by 259 * a highly compressible string table.) Smaller buffer sizes give 260 * fast adaptation but have of course the overhead of transmitting 261 * trees more frequently. 262 * - I can't count above 4 263 */ 264 265 uInt last_lit; /* running index in l_buf */ 266 267 ushf *d_buf; 268 /* Buffer for distances. To simplify the code, d_buf and l_buf have 269 * the same number of elements. To use different lengths, an extra flag 270 * array would be necessary. 271 */ 272 273 ulg opt_len; /* bit length of current block with optimal trees */ 274 ulg static_len; /* bit length of current block with static trees */ 275 uInt matches; /* number of string matches in current block */ 276 int last_eob_len; /* bit length of EOB code for last block */ 277 278 #ifdef DEBUG 279 ulg compressed_len; /* total bit length of compressed file mod 2^32 */ 280 ulg bits_sent; /* bit length of compressed data sent mod 2^32 */ 281 #endif 282 283 ush bi_buf; 284 /* Output buffer. bits are inserted starting at the bottom (least 285 * significant bits). 286 */ 287 int bi_valid; 288 /* Number of valid bits in bi_buf. All bits above the last valid bit 289 * are always zero. 290 */ 291 292 } FAR deflate_state; 293 294 /* Output a byte on the stream. 295 * IN assertion: there is enough room in pending_buf. 296 */ 297 #define put_byte(s, c) {s->pending_buf[s->pending++] = (c);} 298 299 300 #define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1) 301 /* Minimum amount of lookahead, except at the end of the input file. 302 * See deflate.c for comments about the MIN_MATCH+1. 303 */ 304 305 #define MAX_DIST(s) ((s)->w_size-MIN_LOOKAHEAD) 306 /* In order to simplify the code, particularly on 16 bit machines, match 307 * distances are limited to MAX_DIST instead of WSIZE. 308 */ 309 310 /* in trees.c */ 311 void _tr_init OF((deflate_state *s)); 312 int _tr_tally OF((deflate_state *s, unsigned dist, unsigned lc)); 313 void _tr_flush_block OF((deflate_state *s, charf *buf, ulg stored_len, 314 int eof)); 315 void _tr_align OF((deflate_state *s)); 316 void _tr_stored_block OF((deflate_state *s, charf *buf, ulg stored_len, 317 int eof)); 318 319 #define d_code(dist) \ 320 ((dist) < 256 ? _dist_code[dist] : _dist_code[256+((dist)>>7)]) 321 /* Mapping from a distance to a distance code. dist is the distance - 1 and 322 * must not have side effects. _dist_code[256] and _dist_code[257] are never 323 * used. 324 */ 325 326 #ifndef DEBUG 327 /* Inline versions of _tr_tally for speed: */ 328 329 #if defined(GEN_TREES_H) || !defined(STDC) 330 extern uch _length_code[]; 331 extern uch _dist_code[]; 332 #else 333 extern const uch _length_code[]; 334 extern const uch _dist_code[]; 335 #endif 336 337 # define _tr_tally_lit(s, c, flush) \ 338 { uch cc = (c); \ 339 s->d_buf[s->last_lit] = 0; \ 340 s->l_buf[s->last_lit++] = cc; \ 341 s->dyn_ltree[cc].Freq++; \ 342 flush = (s->last_lit == s->lit_bufsize-1); \ 343 } 344 # define _tr_tally_dist(s, distance, length, flush) \ 345 { uch len = (uch)(length); \ 346 ush dist = (ush)(distance); \ 347 s->d_buf[s->last_lit] = dist; \ 348 s->l_buf[s->last_lit++] = len; \ 349 dist--; \ 350 s->dyn_ltree[_length_code[len]+LITERALS+1].Freq++; \ 351 s->dyn_dtree[d_code(dist)].Freq++; \ 352 flush = (s->last_lit == s->lit_bufsize-1); \ 353 } 354 #else 355 # define _tr_tally_lit(s, c, flush) flush = _tr_tally(s, 0, c) 356 # define _tr_tally_dist(s, distance, length, flush) \ 357 flush = _tr_tally(s, distance, length) 358 #endif 359 360 #endif /* DEFLATE_H */ 361