![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to zdeflate.c CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [PGP] / pgp / src|
Return to zdeflate.c CVS log | Up to [PGP] / pgp / src |
1.1.1.2 ! root 1: /* ! 2: ! 3: Copyright (C) 1990-1992 Mark Adler, Richard B. Wales, Jean-loup Gailly, ! 4: Kai Uwe Rommel and Igor Mandrichenko. ! 5: Permission is granted to any individual or institution to use, copy, or ! 6: redistribute this software so long as all of the original files are included ! 7: unmodified, that it is not sold for profit, and that this copyright notice ! 8: is retained. ! 9: ! 10: */ ! 11: ! 12: /* ! 13: * deflate.c by Jean-loup Gailly. ! 14: * ! 15: * PURPOSE ! 16: * ! 17: * Identify new text as repetitions of old text within a fixed- ! 18: * length sliding window trailing behind the new text. ! 19: * ! 20: * DISCUSSION ! 21: * ! 22: * The "deflation" process depends on being able to identify portions ! 23: * of the input text which are identical to earlier input (within a ! 24: * sliding window trailing behind the input currently being processed). ! 25: * ! 26: * The most straightforward technique turns out to be the fastest for ! 27: * most input files: try all possible matches and select the longest. ! 28: * The key feature is of this algorithm is that insertion and deletions ! 29: * from the string dictionary are very simple and thus fast. Insertions ! 30: * and deletions are performed at each input character, whereas string ! 31: * matches are performed only when the previous match ends. So it is ! 32: * preferable to spend more time in matches to allow very fast string ! 33: * insertions and deletions. The matching algorithm for small strings ! 34: * is inspired from that of Rabin & Karp. A brute force approach is ! 35: * used to find longer strings when a small match has been found. ! 36: * A similar algorithm is used in comic (by Jan-Mark Wams) and freeze ! 37: * (by Leonid Broukhis). ! 38: * A previous version of this file used a more sophisticated algorithm ! 39: * (by Fiala and Greene) which is guaranteed to run in linear amortized ! 40: * time, but has a larger average cost and uses more memory. However ! 41: * the F&G algorithm may be faster for some highly redundant files if ! 42: * the parameter max_chain_length (described below) is too large. ! 43: * ! 44: * ACKNOWLEDGEMENTS ! 45: * ! 46: * The idea of lazy evaluation of matches is due to Jan-Mark Wams, and ! 47: * I found it in 'freeze' written by Leonid Broukhis. ! 48: * Thanks to many info-zippers for bug reports and testing. ! 49: * ! 50: * REFERENCES ! 51: * ! 52: * APPNOTE.TXT documentation file in PKZIP 2.0 distribution. ! 53: * ! 54: * A description of the Rabin and Karp algorithm is given in the book ! 55: * "Algorithms" by R. Sedgewick, Addison-Wesley, p252. ! 56: * ! 57: * Fiala,E.R., and Greene,D.H. ! 58: * Data Compression with Finite Windows, Comm.ACM, 32,4 (1989) 490-595 ! 59: * ! 60: * INTERFACE ! 61: * ! 62: * void lm_init (int pack_level, ush *flags) ! 63: * Initialize the "longest match" routines for a new file ! 64: * ! 65: * ulg deflate (void) ! 66: * Processes a new input file and return its compressed length. Sets ! 67: * the compressed length, crc, deflate flags and internal file ! 68: * attributes. ! 69: */ ! 70: ! 71: #include "zip.h" ! 72: ! 73: /* =========================================================================== ! 74: * Configuration parameters ! 75: */ ! 76: ! 77: /* Compile with MEDIUM_MEM to reduce the memory requirements or ! 78: * with SMALL_MEM to use as little memory as possible. ! 79: * Warning: defining these symbols affects MATCH_BUFSIZE and HASH_BITS ! 80: * (see below) and thus affects the compression ratio. The compressed output ! 81: * is still correct, and might even be smaller in some cases. ! 82: */ ! 83: ! 84: #ifdef SMALL_MEM ! 85: # define HASH_BITS 13 /* Number of bits used to hash strings */ ! 86: #else ! 87: #ifdef MEDIUM_MEM ! 88: # define HASH_BITS 14 ! 89: #else ! 90: # define HASH_BITS 15 ! 91: /* For portability to 16 bit machines, do not use values above 15. */ ! 92: #endif ! 93: #endif ! 94: ! 95: #define HASH_SIZE (unsigned)(1<<HASH_BITS) ! 96: #define HASH_MASK (HASH_SIZE-1) ! 97: #define WMASK (WSIZE-1) ! 98: /* HASH_SIZE and WSIZE must be powers of two */ ! 99: ! 100: #define NIL 0 ! 101: /* Tail of hash chains */ ! 102: ! 103: #define FAST 4 ! 104: #define SLOW 2 ! 105: /* speed options for the general purpose bit flag */ ! 106: ! 107: #ifndef TOO_FAR ! 108: # define TOO_FAR 4096 ! 109: #endif ! 110: /* Matches of length 3 are discarded if their distance exceeds TOO_FAR */ ! 111: ! 112: /* =========================================================================== ! 113: * Local data used by the "longest match" routines. ! 114: */ ! 115: ! 116: typedef ush Pos; ! 117: typedef unsigned IPos; ! 118: /* A Pos is an index in the character window. We use short instead of int to ! 119: * save space in the various tables. IPos is used only for parameter passing. ! 120: */ ! 121: ! 122: #ifndef DYN_ALLOC ! 123: uch far window[2L*WSIZE]; ! 124: /* Sliding window. Input bytes are read into the second half of the window, ! 125: * and move to the first half later to keep a dictionary of at least WSIZE ! 126: * bytes. With this organization, matches are limited to a distance of ! 127: * WSIZE-MAX_MATCH bytes, but this ensures that IO is always ! 128: * performed with a length multiple of the block size. Also, it limits ! 129: * the window size to 64K, which is quite useful on MSDOS. ! 130: * To do: limit the window size to WSIZE+BSZ if SMALL_MEM (the code would ! 131: * be less efficient since the data would have to be copied WSIZE/BSZ times) ! 132: */ ! 133: Pos far prev[WSIZE]; ! 134: /* Link to older string with same hash index. To limit the size of this ! 135: * array to 64K, this link is maintained only for the last 32K strings. ! 136: * An index in this array is thus a window index modulo 32K. ! 137: */ ! 138: Pos far head[HASH_SIZE]; ! 139: /* Heads of the hash chains or NIL */ ! 140: #else ! 141: uch far * near window = NULL; ! 142: Pos far * near prev = NULL; ! 143: Pos far * near head; ! 144: #endif ! 145: ! 146: long block_start; ! 147: /* window position at the beginning of the current output block. Gets ! 148: * negative when the window is moved backwards. ! 149: */ ! 150: ! 151: local unsigned near ins_h; /* hash index of string to be inserted */ ! 152: ! 153: #define H_SHIFT ((HASH_BITS+MIN_MATCH-1)/MIN_MATCH) ! 154: /* Number of bits by which ins_h and del_h must be shifted at each ! 155: * input step. It must be such that after MIN_MATCH steps, the oldest ! 156: * byte no longer takes part in the hash key, that is: ! 157: * H_SHIFT * MIN_MATCH >= HASH_BITS ! 158: */ ! 159: ! 160: unsigned int near prev_length; ! 161: /* Length of the best match at previous step. Matches not greater than this ! 162: * are discarded. This is used in the lazy match evaluation. ! 163: */ ! 164: ! 165: unsigned near strstart; /* start of string to insert */ ! 166: unsigned near match_start; /* start of matching string */ ! 167: local int near eofile; /* flag set at end of input file */ ! 168: local unsigned near lookahead; /* number of valid bytes ahead in window */ ! 169: ! 170: unsigned near max_chain_length; ! 171: /* To speed up deflation, hash chains are never searched beyond this length. ! 172: * A higher limit improves compression ratio but degrades the speed. ! 173: */ ! 174: ! 175: local unsigned int max_lazy_match; ! 176: /* Attempt to find a better match only when the current match is strictly ! 177: * smaller than this value. ! 178: */ ! 179: ! 180: int near good_match; ! 181: /* Use a faster search when the previous match is longer than this */ ! 182: ! 183: ! 184: /* Values for max_lazy_match, good_match and max_chain_length, depending on ! 185: * the desired pack level (0..9). The values given below have been tuned to ! 186: * exclude worst case performance for pathological files. Better values may be ! 187: * found for specific files. ! 188: */ ! 189: typedef struct config { ! 190: int good_length; ! 191: int max_lazy; ! 192: unsigned max_chain; ! 193: uch flag; ! 194: } config; ! 195: ! 196: local config configuration_table[10] = { ! 197: /* good lazy chain flag */ ! 198: /* 0 */ {0, 0, 0, 0}, /* store only */ ! 199: /* 1 */ {4, 4, 16, FAST}, /* maximum speed */ ! 200: /* 2 */ {6, 8, 16, 0}, ! 201: /* 3 */ {8, 16, 32, 0}, ! 202: /* 4 */ {8, 32, 64, 0}, ! 203: /* 5 */ {8, 64, 128, 0}, ! 204: /* 6 */ {8, 128, 256, 0}, ! 205: /* 7 */ {8, 128, 512, 0}, ! 206: /* 8 */ {32, 258, 1024, 0}, ! 207: /* 9 */ {32, 258, 4096, SLOW}}; /* maximum compression */ ! 208: ! 209: /* Note: the current code requires max_lazy >= MIN_MATCH and max_chain >= 4 ! 210: * but these restrictions can easily be removed at a small cost. ! 211: */ ! 212: ! 213: #define EQUAL 0 ! 214: /* result of memcmp for equal strings */ ! 215: ! 216: /* =========================================================================== ! 217: * Prototypes for local functions. Use asm version by default for ! 218: * MSDOS but not Unix. However the asm version version is recommended ! 219: * for 386 Unix. ! 220: */ ! 221: #ifdef ATARI_ST ! 222: # undef MSDOS /* avoid the processor specific parts */ ! 223: #endif ! 224: #if defined(MSDOS) && !defined(NO_ASM) && !defined(ASM) ! 225: # define ASM ! 226: #endif ! 227: ! 228: local void fill_window OF((void)); ! 229: int longest_match OF((IPos cur_match)); ! 230: #ifdef ASM ! 231: void match_init OF((void)); /* asm code initialization */ ! 232: #endif ! 233: ! 234: #ifdef DEBUG ! 235: local void check_match OF((IPos start, IPos match, int length)); ! 236: #endif ! 237: ! 238: #define MIN(a,b) ((a) <= (b) ? (a) : (b)) ! 239: /* The arguments must not have side effects. */ ! 240: ! 241: /* =========================================================================== ! 242: * Update a hash value with the given input byte ! 243: * IN assertion: all calls to to UPDATE_HASH are made with consecutive ! 244: * input characters, so that a running hash key can be computed from the ! 245: * previous key instead of complete recalculation each time. ! 246: */ ! 247: #define UPDATE_HASH(h,c) (h = (((h)<<H_SHIFT) ^ (c)) & HASH_MASK) ! 248: ! 249: /* =========================================================================== ! 250: * Insert string s in the dictionary and set match_head to the previous head ! 251: * of the hash chain (the most recent string with same hash key). Return ! 252: * the previous length of the hash chain. ! 253: * IN assertion: all calls to to INSERT_STRING are made with consecutive ! 254: * input characters and the first MIN_MATCH bytes of s are valid ! 255: * (except for the last MIN_MATCH-1 bytes of the input file). ! 256: */ ! 257: #define INSERT_STRING(s, match_head) \ ! 258: (UPDATE_HASH(ins_h, window[(s) + MIN_MATCH-1]), \ ! 259: prev[(s) & WMASK] = match_head = head[ins_h], \ ! 260: head[ins_h] = (s)) ! 261: ! 262: /* =========================================================================== ! 263: * Initialize the "longest match" routines for a new file ! 264: */ ! 265: void lm_init (pack_level, flags) ! 266: int pack_level; /* 0: store, 1: best speed, 9: best compression */ ! 267: ush *flags; /* general purpose bit flag */ ! 268: { ! 269: register unsigned j; ! 270: ! 271: if (pack_level < 1 || pack_level > 9) error("bad pack level"); ! 272: ! 273: /* Use dynamic allocation if compiler does not like big static arrays: */ ! 274: #ifdef DYN_ALLOC ! 275: if (window == NULL) { ! 276: window = (uch far*) fcalloc(WSIZE, 2*sizeof(uch)); ! 277: prev = (Pos far*) fcalloc(WSIZE, sizeof(Pos)); ! 278: head = (Pos far*) fcalloc(HASH_SIZE, sizeof(Pos)); ! 279: ! 280: if (window == NULL || prev == NULL || head == NULL) { ! 281: err(ZE_MEM, "window allocation"); ! 282: } ! 283: } ! 284: #endif /* DYN_ALLOC */ ! 285: #ifdef ASM ! 286: match_init(); /* initialize the asm code */ ! 287: #endif ! 288: /* Initialize the hash table. */ ! 289: for (j = 0; j < HASH_SIZE; j++) head[j] = NIL; ! 290: /* prev will be initialized on the fly */ ! 291: ! 292: /* Set the default configuration parameters: ! 293: */ ! 294: max_lazy_match = configuration_table[pack_level].max_lazy; ! 295: good_match = configuration_table[pack_level].good_length; ! 296: max_chain_length = configuration_table[pack_level].max_chain; ! 297: *flags |= configuration_table[pack_level].flag; ! 298: /* ??? reduce max_chain_length for binary files */ ! 299: ! 300: strstart = 0; ! 301: block_start = 0L; ! 302: ! 303: #if defined(MSDOS) && !defined(__32BIT__) ! 304: /* Can't read a 64K block under MSDOS */ ! 305: lookahead = read_buf((char*)window, (unsigned)WSIZE); ! 306: #else ! 307: lookahead = read_buf((char*)window, 2*WSIZE); ! 308: #endif ! 309: if (lookahead == 0 || lookahead == (unsigned)EOF) { ! 310: eofile = 1, lookahead = 0; ! 311: return; ! 312: } ! 313: eofile = 0; ! 314: /* Make sure that we always have enough lookahead. This is important ! 315: * if input comes from a device such as a tty. ! 316: */ ! 317: while (lookahead < MIN_LOOKAHEAD && !eofile) fill_window(); ! 318: ! 319: ins_h = 0; ! 320: for (j=0; j<MIN_MATCH-1; j++) UPDATE_HASH(ins_h, window[j]); ! 321: /* If lookahead < MIN_MATCH, ins_h is garbage, but this is ! 322: * not important since only literal bytes will be emitted. ! 323: */ ! 324: } ! 325: ! 326: /* =========================================================================== ! 327: * Set match_start to the longest match starting at the given string and ! 328: * return its length. Matches shorter or equal to prev_length are discarded, ! 329: * in which case the result is equal to prev_length and match_start is ! 330: * garbage. ! 331: * IN assertions: cur_match is the head of the hash chain for the current ! 332: * string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1 ! 333: */ ! 334: #ifndef ASM ! 335: /* For MSDOS, OS/2 and 386 Unix, an optimized version is in match.asm. The code ! 336: * is functionally equivalent, so you can use the C version if desired. ! 337: */ ! 338: int longest_match(cur_match) ! 339: IPos cur_match; /* current match */ ! 340: { ! 341: unsigned chain_length = max_chain_length; /* max hash chain length */ ! 342: register uch far *scan = window + strstart; /* current string */ ! 343: register uch far *match = scan; /* matched string */ ! 344: register int len; /* length of current match */ ! 345: int best_len = prev_length; /* best match length so far */ ! 346: IPos limit = strstart > (IPos)MAX_DIST ? strstart - (IPos)MAX_DIST : NIL; ! 347: /* Stop when cur_match becomes <= limit. To simplify the code, ! 348: * we prevent matches with the string of window index 0. ! 349: */ ! 350: #ifdef UNALIGNED_OK ! 351: register ush scan_start = *(ush*)scan; ! 352: register ush scan_end = *(ush*)(scan+best_len-1); ! 353: #else ! 354: register uch scan_start = *scan; ! 355: register uch scan_end1 = scan[best_len-1]; ! 356: register uch scan_end = scan[best_len]; ! 357: #endif ! 358: ! 359: /* Do not waste too much time if we already have a good match: */ ! 360: if (prev_length >= good_match) { ! 361: chain_length >>= 2; ! 362: } ! 363: ! 364: do { ! 365: Assert(cur_match < strstart, "no future"); ! 366: match = window + cur_match; ! 367: ! 368: /* Skip to next match if the match length cannot increase ! 369: * or if the match length is less than 2: ! 370: */ ! 371: #if (defined(UNALIGNED_OK) && HASH_BITS >= 8) ! 372: /* This code assumes sizeof(unsigned short) == 2 and ! 373: * sizeof(unsigned long) == 4. Do not use UNALIGNED_OK if your ! 374: * compiler uses different sizes. ! 375: */ ! 376: if (*(ush*)(match+best_len-1) != scan_end || ! 377: *(ush*)match != scan_start) continue; ! 378: ! 379: len = MIN_MATCH - 4; ! 380: /* It is not necessary to compare scan[2] and match[2] since they are ! 381: * always equal when the other bytes match, given that the hash keys ! 382: * are equal and that HASH_BITS >= 8. ! 383: */ ! 384: do {} while ((len+=4) < MAX_MATCH-3 && ! 385: *(ulg*)(scan+len) == *(ulg*)(match+len)); ! 386: /* The funny do {} generates better code for most compilers */ ! 387: ! 388: if (*(ush*)(scan+len) == *(ush*)(match+len)) len += 2; ! 389: if (scan[len] == match[len]) len++; ! 390: ! 391: #else /* UNALIGNED_OK */ ! 392: if (match[best_len] != scan_end || ! 393: match[best_len-1] != scan_end1 || *match != scan_start) ! 394: continue; ! 395: /* It is not necessary to compare scan[1] and match[1] since they ! 396: * are always equal when the other bytes match, given that ! 397: * the hash keys are equal and that h_shift+8 <= HASH_BITS, ! 398: * that is, when the last byte is entirely included in the hash key. ! 399: * The condition is equivalent to ! 400: * (HASH_BITS+2)/3 + 8 <= HASH_BITS ! 401: * or: HASH_BITS >= 13 ! 402: * Also, we check for a match at best_len-1 to get rid quickly of ! 403: * the match with the suffix of the match made at the previous step, ! 404: * which is known to fail. ! 405: */ ! 406: #if HASH_BITS >= 13 ! 407: len = 1; ! 408: #else ! 409: len = 0; ! 410: #endif ! 411: do {} while (++len < MAX_MATCH && scan[len] == match[len]); ! 412: ! 413: #endif /* UNALIGNED_OK */ ! 414: ! 415: if (len > best_len) { ! 416: match_start = cur_match; ! 417: best_len = len; ! 418: if (len == MAX_MATCH) break; ! 419: #ifdef UNALIGNED_OK ! 420: scan_end = *(ush*)(scan+best_len-1); ! 421: #else ! 422: scan_end1 = scan[best_len-1]; ! 423: scan_end = scan[best_len]; ! 424: #endif ! 425: } ! 426: } while (--chain_length != 0 && ! 427: (cur_match = prev[cur_match & WMASK]) > limit); ! 428: ! 429: return best_len; ! 430: } ! 431: #endif /* NO_ASM */ ! 432: ! 433: #ifdef DEBUG ! 434: /* =========================================================================== ! 435: * Check that the match at match_start is indeed a match. ! 436: */ ! 437: local void check_match(start, match, length) ! 438: IPos start, match; ! 439: int length; ! 440: { ! 441: /* check that the match is indeed a match */ ! 442: if (memcmp((char*)window + match, ! 443: (char*)window + start, length) != EQUAL) { ! 444: fprintf(stderr, ! 445: " start %d, match %d, length %d\n", ! 446: start, match, length); ! 447: error("invalid match"); ! 448: } ! 449: if (verbose > 1) { ! 450: fprintf(stderr,"\\[%d,%d]", start-match, length); ! 451: do { putc(window[start++], stderr); } while (--length != 0); ! 452: } ! 453: } ! 454: #else ! 455: # define check_match(start, match, length) ! 456: #endif ! 457: ! 458: /* =========================================================================== ! 459: * Fill the window when the lookahead becomes insufficient. ! 460: * Updates strstart and lookahead, and sets eofile if end of input file. ! 461: * IN assertion: lookahead < MIN_LOOKAHEAD && strstart + lookahead > 0 ! 462: * OUT assertions: at least one byte has been read, or eofile is set; ! 463: * file reads are performed for at least two bytes (required for the ! 464: * translate_eol option). ! 465: */ ! 466: local void fill_window() ! 467: { ! 468: register unsigned n, m; ! 469: unsigned more = (unsigned)((ulg)2*WSIZE - (ulg)lookahead - (ulg)strstart); ! 470: /* Amount of free space at the end of the window. */ ! 471: ! 472: /* If the window is full, move the upper half to the lower one to make ! 473: * room in the upper half. ! 474: */ ! 475: if (more == (unsigned)EOF) { ! 476: /* Very unlikely, but possible on 16 bit machine if strstart == 0 ! 477: * and lookahead == 1 (input done one byte at time) ! 478: */ ! 479: more--; ! 480: } else if (more <= 1) { ! 481: /* By the IN assertion, the window is not empty so we can't confuse ! 482: * more == 0 with more == 64K on a 16 bit machine. ! 483: */ ! 484: memcpy((char*)window, (char*)window+WSIZE, (unsigned)WSIZE); ! 485: match_start -= WSIZE; ! 486: strstart -= WSIZE; ! 487: /* strstart - WSIZE >= WSIZE - 1 - lookahead >= WSIZE - MIN_LOOKAHEAD ! 488: * so we now have strstart >= MAX_DIST: ! 489: */ ! 490: Assert (strstart >= MAX_DIST, "window slide too early"); ! 491: block_start -= (long) WSIZE; ! 492: ! 493: for (n = 0; n < HASH_SIZE; n++) { ! 494: m = head[n]; ! 495: head[n] = (Pos)(m >= WSIZE ? m-WSIZE : NIL); ! 496: } ! 497: for (n = 0; n < WSIZE; n++) { ! 498: m = prev[n]; ! 499: prev[n] = (Pos)(m >= WSIZE ? m-WSIZE : NIL); ! 500: /* If n is not on any hash chain, prev[n] is garbage but ! 501: * its value will never be used. ! 502: */ ! 503: } ! 504: more += WSIZE; ! 505: #ifdef ZIP ! 506: if (verbose) putc('.', stderr); ! 507: #endif ! 508: } ! 509: /* At this point, more >= 2 */ ! 510: n = read_buf((char*)window+strstart+lookahead, more); ! 511: if (n == 0 || n == (unsigned)EOF) { ! 512: eofile = 1; ! 513: } else { ! 514: lookahead += n; ! 515: } ! 516: } ! 517: ! 518: /* =========================================================================== ! 519: * Flush the current block, with given end-of-file flag. ! 520: * IN assertion: strstart is set to the end of the current match. ! 521: */ ! 522: #define FLUSH_BLOCK(eof) \ ! 523: flush_block(block_start >= 0L ? (char*)&window[block_start] : (char*)NULL,\ ! 524: (long)strstart - block_start, (eof)) ! 525: ! 526: /* =========================================================================== ! 527: * Processes a new input file and return its compressed length. ! 528: */ ! 529: #ifdef NO_LAZY ! 530: ulg deflate() ! 531: { ! 532: IPos hash_head; /* head of the hash chain */ ! 533: int flush; /* set if current block must be flushed */ ! 534: unsigned match_length = 0; /* length of best match */ ! 535: ! 536: prev_length = MIN_MATCH-1; ! 537: while (lookahead != 0) { ! 538: /* Insert the string window[strstart .. strstart+2] in the ! 539: * dictionary, and set hash_head to the head of the hash chain: ! 540: */ ! 541: INSERT_STRING(strstart, hash_head); ! 542: ! 543: /* Find the longest match, discarding those <= prev_length. ! 544: * At this point we have always match_length < MIN_MATCH ! 545: */ ! 546: if (hash_head != NIL && strstart - hash_head <= MAX_DIST) { ! 547: /* To simplify the code, we prevent matches with the string ! 548: * of window index 0 (in particular we have to avoid a match ! 549: * of the string with itself at the start of the input file). ! 550: */ ! 551: match_length = longest_match (hash_head); ! 552: /* longest_match() sets match_start */ ! 553: if (match_length > lookahead) match_length = lookahead; ! 554: } ! 555: if (match_length >= MIN_MATCH) { ! 556: check_match(strstart, match_start, match_length); ! 557: ! 558: flush = ct_tally(strstart-match_start, match_length - MIN_MATCH); ! 559: ! 560: lookahead -= match_length; ! 561: match_length--; /* string at strstart already in hash table */ ! 562: do { ! 563: strstart++; ! 564: INSERT_STRING(strstart, hash_head); ! 565: /* strstart never exceeds WSIZE-MAX_MATCH, so there are ! 566: * always MIN_MATCH bytes ahead. If lookahead < MIN_MATCH ! 567: * these bytes are garbage, but it does not matter since the ! 568: * next lookahead bytes will always be emitted as literals. ! 569: */ ! 570: } while (--match_length != 0); ! 571: } else { ! 572: /* No match, output a literal byte */ ! 573: flush = ct_tally (0, window[strstart]); ! 574: lookahead--; ! 575: } ! 576: strstart++; ! 577: if (flush) FLUSH_BLOCK(0), block_start = strstart; ! 578: ! 579: /* Make sure that we always have enough lookahead, except ! 580: * at the end of the input file. We need MAX_MATCH bytes ! 581: * for the next match, plus MIN_MATCH bytes to insert the ! 582: * string following the next match. ! 583: */ ! 584: while (lookahead < MIN_LOOKAHEAD && !eofile) fill_window(); ! 585: ! 586: } ! 587: return FLUSH_BLOCK(1); /* eof */ ! 588: } ! 589: #else /* LAZY */ ! 590: /* =========================================================================== ! 591: * Same as above, but achieves better compression. We use a lazy ! 592: * evaluation for matches: a match is finally adopted only if there is ! 593: * no better match at the next window position. ! 594: */ ! 595: ulg deflate() ! 596: { ! 597: IPos hash_head; /* head of hash chain */ ! 598: IPos prev_match; /* previous match */ ! 599: int flush; /* set if current block must be flushed */ ! 600: int match_available = 0; /* set if previous match exists */ ! 601: register unsigned match_length = MIN_MATCH-1; /* length of best match */ ! 602: #ifdef DEBUG ! 603: extern ulg isize; /* byte length of input file, for debug only */ ! 604: #endif ! 605: ! 606: /* Process the input block. */ ! 607: while (lookahead != 0) { ! 608: /* Insert the string window[strstart .. strstart+2] in the ! 609: * dictionary, and set hash_head to the head of the hash chain: ! 610: */ ! 611: INSERT_STRING(strstart, hash_head); ! 612: ! 613: /* Find the longest match, discarding those <= prev_length. ! 614: */ ! 615: prev_length = match_length, prev_match = match_start; ! 616: match_length = MIN_MATCH-1; ! 617: ! 618: if (hash_head != NIL && prev_length < max_lazy_match && ! 619: strstart - hash_head <= MAX_DIST) { ! 620: /* To simplify the code, we prevent matches with the string ! 621: * of window index 0 (in particular we have to avoid a match ! 622: * of the string with itself at the start of the input file). ! 623: */ ! 624: match_length = longest_match (hash_head); ! 625: /* longest_match() sets match_start */ ! 626: if (match_length > lookahead) match_length = lookahead; ! 627: /* Ignore a length 3 match if it is too distant: */ ! 628: if (match_length == MIN_MATCH && strstart-match_start > TOO_FAR){ ! 629: /* If prev_match is also MIN_MATCH, match_start is garbage ! 630: * but we will ignore the current match anyway. ! 631: */ ! 632: match_length--; ! 633: } ! 634: } ! 635: /* If there was a match at the previous step and the current ! 636: * match is not better, output the previous match: ! 637: */ ! 638: if (prev_length >= MIN_MATCH && match_length <= prev_length) { ! 639: ! 640: check_match(strstart-1, prev_match, prev_length); ! 641: ! 642: flush = ct_tally(strstart-1-prev_match, prev_length - MIN_MATCH); ! 643: ! 644: /* Insert in hash table all strings up to the end of the match. ! 645: * strstart-1 and strstart are already inserted. ! 646: */ ! 647: lookahead -= prev_length-1; ! 648: prev_length -= 2; ! 649: do { ! 650: strstart++; ! 651: INSERT_STRING(strstart, hash_head); ! 652: /* strstart never exceeds WSIZE-MAX_MATCH, so there are ! 653: * always MIN_MATCH bytes ahead. If lookahead < MIN_MATCH ! 654: * these bytes are garbage, but it does not matter since the ! 655: * next lookahead bytes will always be emitted as literals. ! 656: */ ! 657: } while (--prev_length != 0); ! 658: match_available = 0; ! 659: match_length = MIN_MATCH-1; ! 660: strstart++; ! 661: if (flush) FLUSH_BLOCK(0), block_start = strstart; ! 662: ! 663: } else if (match_available) { ! 664: /* If there was no match at the previous position, output a ! 665: * single literal. If there was a match but the current match ! 666: * is longer, truncate the previous match to a single literal. ! 667: */ ! 668: Tracevv((stderr,"%c",window[strstart-1])); ! 669: if (ct_tally (0, window[strstart-1])) { ! 670: FLUSH_BLOCK(0), block_start = strstart; ! 671: } ! 672: strstart++; ! 673: lookahead--; ! 674: } else { ! 675: /* There is no previous match to compare with, wait for ! 676: * the next step to decide. ! 677: */ ! 678: match_available = 1; ! 679: strstart++; ! 680: lookahead--; ! 681: } ! 682: #if 0 /* for pgp: disabled to allow compiling with -DDEBUG */ ! 683: Assert (strstart <= isize && lookahead <= isize, "a bit too far"); ! 684: #endif ! 685: ! 686: /* Make sure that we always have enough lookahead, except ! 687: * at the end of the input file. We need MAX_MATCH bytes ! 688: * for the next match, plus MIN_MATCH bytes to insert the ! 689: * string following the next match. ! 690: */ ! 691: while (lookahead < MIN_LOOKAHEAD && !eofile) fill_window(); ! 692: } ! 693: if (match_available) ct_tally (0, window[strstart-1]); ! 694: ! 695: return FLUSH_BLOCK(1); /* eof */ ! 696: } ! 697: #endif /* LAZY */
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.