1 /* 2 * compress - File compression ala IEEE Computer, June 1984. 3 * 4 * Algorithm from "A Technique for High Performance Data Compression", 5 * Terry A. Welch, IEEE Computer Vol 17, No 6 (June 1984), pp 8-19. 6 * 7 * Usage: compress [-dfvc] [-b bits] [file ...] 8 * Inputs: 9 * -b: limit the max number of bits/code. 10 * -c: write output on stdout, don't remove original. 11 * -d: decompress instead. 12 * -f: Forces output file to be generated, even if one already 13 * exists, and even if no space is saved by compressing. 14 * If -f is not used, the user will be prompted if stdin is 15 * a tty, otherwise, the output file will not be overwritten. 16 * -v: Write compression statistics 17 * 18 * file ...: Files to be compressed. If none specified, stdin is used. 19 * Outputs: 20 * file.Z: Compressed form of file with same mode, owner, and utimes 21 * or stdout (if stdin used as input) 22 * 23 * Assumptions: 24 * When filenames are given, replaces with the compressed version 25 * (.Z suffix) only if the file decreases in size. 26 * Algorithm: 27 * Modified Lempel-Ziv method (LZW). Basically finds common 28 * substrings and replaces them with a variable size code. This is 29 * deterministic, and can be done on the fly. Thus, the decompression 30 * procedure needs no input table, but tracks the way the table was built. 31 32 * Authors: Spencer W. Thomas (decvax!harpo!utah-cs!utah-gr!thomas) 33 * Jim McKie (decvax!mcvax!jim) 34 * Steve Davies (decvax!vax135!petsd!peora!srd) 35 * Ken Turkowski (decvax!decwrl!turtlevax!ken) 36 * James A. Woods (decvax!ihnp4!ames!jaw) 37 * Joe Orost (decvax!vax135!petsd!joe) 38 */ 39 #define _PLAN9_SOURCE 40 #define _BSD_EXTENSION 41 #define _POSIX_SOURCE 42 43 #include <u.h> 44 #include <stdio.h> 45 #include <ctype.h> 46 #include <stdlib.h> 47 #include <unistd.h> 48 #include <string.h> 49 #include <signal.h> 50 #include <utime.h> 51 #include <sys/types.h> 52 #include <sys/stat.h> 53 54 #define min(a,b) ((a>b) ? b : a) 55 56 #define BITS 16 57 #define HSIZE 69001 /* 95% occupancy */ 58 59 /* 60 * a code_int must be able to hold 2**BITS values of type int, and also -1 61 */ 62 typedef long code_int; 63 typedef long count_int; 64 65 static char rcs_ident[] = "$Header: compress.c,v 4.0 85/07/30 12:50:00 joe Release $"; 66 67 uchar magic_header[] = { 0x1F, 0x9D }; /* 1F 9D */ 68 69 /* Defines for third byte of header */ 70 #define BIT_MASK 0x1f 71 #define BLOCK_MASK 0x80 72 /* Masks 0x40 and 0x20 are free. I think 0x20 should mean that there is 73 a fourth header byte (for expansion). 74 */ 75 #define INIT_BITS 9 /* initial number of bits/code */ 76 77 #define ARGVAL() (*++(*argv) || (--argc && *++argv)) 78 79 int n_bits; /* number of bits/code */ 80 int maxbits = BITS; /* user settable max # bits/code */ 81 code_int maxcode; /* maximum code, given n_bits */ 82 code_int maxmaxcode = 1 << BITS; /* should NEVER generate this code */ 83 84 #define MAXCODE(n_bits) ((1 << (n_bits)) - 1) 85 86 count_int htab[HSIZE]; 87 ushort codetab[HSIZE]; 88 89 #define htabof(i) htab[i] 90 #define codetabof(i) codetab[i] 91 92 code_int hsize = HSIZE; /* for dynamic table sizing */ 93 count_int fsize; 94 95 /* 96 * To save much memory, we overlay the table used by compress() with those 97 * used by decompress(). The tab_prefix table is the same size and type 98 * as the codetab. The tab_suffix table needs 2**BITS characters. We 99 * get this from the beginning of htab. The output stack uses the rest 100 * of htab, and contains characters. There is plenty of room for any 101 * possible stack (stack used to be 8000 characters). 102 */ 103 104 #define tab_prefixof(i) codetabof(i) 105 #define tab_suffixof(i) ((uchar *)(htab))[i] 106 #define de_stack ((uchar *)&tab_suffixof(1<<BITS)) 107 108 code_int free_ent = 0; /* first unused entry */ 109 int exit_stat = 0; 110 111 void cl_block(void); 112 void cl_hash(count_int); 113 void compress(void); 114 void copystat(char *, char *); 115 void decompress(void); 116 int foreground(void); 117 code_int getcode(void); 118 void onintr(int); 119 void oops(int); 120 void output(code_int); 121 void prratio(FILE *, long, long); 122 void version(void); 123 void writeerr(void); 124 125 void 126 Usage(void) 127 { 128 #ifdef DEBUG 129 fprintf(stderr,"Usage: compress [-cdfDV] [-b maxbits] [file ...]\n"); 130 #else 131 fprintf(stderr,"Usage: compress [-cdfvV] [-b maxbits] [file ...]\n"); 132 #endif /* DEBUG */ 133 } 134 135 int debug = 0; 136 int nomagic = 0; /* Use a 3-byte magic number header, unless old file */ 137 int zcat_flg = 0; /* Write output on stdout, suppress messages */ 138 int quiet = 1; /* don't tell me about compression */ 139 140 /* 141 * block compression parameters -- after all codes are used up, 142 * and compression rate changes, start over. 143 */ 144 int block_compress = BLOCK_MASK; 145 int clear_flg = 0; 146 long ratio = 0; 147 #define CHECK_GAP 10000 /* ratio check interval */ 148 count_int checkpoint = CHECK_GAP; 149 /* 150 * the next two codes should not be changed lightly, as they must not 151 * lie within the contiguous general code space. 152 */ 153 #define FIRST 257 /* first free entry */ 154 #define CLEAR 256 /* table clear output code */ 155 156 int force = 0; 157 char ofname [100]; 158 #ifdef DEBUG 159 int verbose = 0; 160 #endif /* DEBUG */ 161 void (*bgnd_flag)(int); 162 163 int do_decomp = 0; 164 165 main(argc, argv) 166 int argc; 167 char **argv; 168 { 169 int overwrite = 0; /* Do not overwrite unless given -f flag */ 170 char tempname[512]; 171 char **filelist, **fileptr; 172 char *cp; 173 struct stat statbuf; 174 175 if ( (bgnd_flag = signal ( SIGINT, SIG_IGN )) != SIG_IGN ) { 176 signal(SIGINT, onintr); 177 signal(SIGSEGV, oops); 178 } 179 180 filelist = fileptr = (char **)(malloc(argc * sizeof(*argv))); 181 *filelist = NULL; 182 183 if((cp = strrchr(argv[0], '/')) != 0) 184 cp++; 185 else 186 cp = argv[0]; 187 if(strcmp(cp, "uncompress") == 0) 188 do_decomp = 1; 189 else if(strcmp(cp, "zcat") == 0) { 190 do_decomp = 1; 191 zcat_flg = 1; 192 } 193 194 /* 195 * Argument Processing 196 * All flags are optional. 197 * -C generate output compatible with compress 2.0. 198 * -D debug 199 * -V print Version; debug verbose 200 * -b maxbits maxbits. If -b is specified, then maxbits MUST be 201 * given also. 202 * -c cat all output to stdout 203 * -d do_decomp 204 * -f force overwrite of output file 205 * -n no header: useful to uncompress old files 206 * -v unquiet 207 * if a string is left, must be an input filename. 208 */ 209 for (argc--, argv++; argc > 0; argc--, argv++) { 210 if (**argv == '-') { /* A flag argument */ 211 while (*++(*argv)) { /* Process all flags in this arg */ 212 switch (**argv) { 213 case 'C': 214 block_compress = 0; 215 break; 216 #ifdef DEBUG 217 case 'D': 218 debug = 1; 219 break; 220 case 'V': 221 verbose = 1; 222 version(); 223 break; 224 #else 225 case 'V': 226 version(); 227 break; 228 #endif 229 case 'b': 230 if (!ARGVAL()) { 231 fprintf(stderr, "Missing maxbits\n"); 232 Usage(); 233 exit(1); 234 } 235 maxbits = atoi(*argv); 236 goto nextarg; 237 case 'c': 238 zcat_flg = 1; 239 break; 240 case 'd': 241 do_decomp = 1; 242 break; 243 case 'f': 244 case 'F': 245 overwrite = 1; 246 force = 1; 247 break; 248 case 'n': 249 nomagic = 1; 250 break; 251 case 'q': 252 quiet = 1; 253 break; 254 case 'v': 255 quiet = 0; 256 break; 257 default: 258 fprintf(stderr, "Unknown flag: '%c'; ", **argv); 259 Usage(); 260 exit(1); 261 } 262 } 263 } else { /* Input file name */ 264 *fileptr++ = *argv; /* Build input file list */ 265 *fileptr = NULL; 266 /* process nextarg; */ 267 } 268 nextarg: 269 continue; 270 } 271 272 if(maxbits < INIT_BITS) maxbits = INIT_BITS; 273 if (maxbits > BITS) maxbits = BITS; 274 maxmaxcode = 1 << maxbits; 275 276 if (*filelist != NULL) { 277 for (fileptr = filelist; *fileptr; fileptr++) { 278 exit_stat = 0; 279 if (do_decomp != 0) { /* DECOMPRESSION */ 280 /* Check for .Z suffix */ 281 if (strcmp(*fileptr + strlen(*fileptr) - 2, ".Z") != 0) { 282 /* No .Z: tack one on */ 283 strcpy(tempname, *fileptr); 284 strcat(tempname, ".Z"); 285 *fileptr = tempname; 286 } 287 /* Open input file */ 288 if ((freopen(*fileptr, "r", stdin)) == NULL) { 289 perror(*fileptr); 290 continue; 291 } 292 /* Check the magic number */ 293 if (nomagic == 0) { 294 if ((getchar() != (magic_header[0] & 0xFF)) 295 || (getchar() != (magic_header[1] & 0xFF))) { 296 fprintf(stderr, "%s: not in compressed format\n", 297 *fileptr); 298 continue; 299 } 300 maxbits = getchar(); /* set -b from file */ 301 block_compress = maxbits & BLOCK_MASK; 302 maxbits &= BIT_MASK; 303 maxmaxcode = 1 << maxbits; 304 if(maxbits > BITS) { 305 fprintf(stderr, 306 "%s: compressed with %d bits, can only handle %d bits\n", 307 *fileptr, maxbits, BITS); 308 continue; 309 } 310 } 311 /* Generate output filename */ 312 strcpy(ofname, *fileptr); 313 ofname[strlen(*fileptr) - 2] = '\0'; /* Strip off .Z */ 314 } else { /* COMPRESSION */ 315 if (strcmp(*fileptr + strlen(*fileptr) - 2, ".Z") == 0) { 316 fprintf(stderr, 317 "%s: already has .Z suffix -- no change\n", 318 *fileptr); 319 continue; 320 } 321 /* Open input file */ 322 if ((freopen(*fileptr, "r", stdin)) == NULL) { 323 perror(*fileptr); 324 continue; 325 } 326 (void) stat(*fileptr, &statbuf); 327 fsize = (long) statbuf.st_size; 328 /* 329 * tune hash table size for small files -- ad hoc, 330 * but the sizes match earlier #defines, which 331 * serve as upper bounds on the number of output codes. 332 */ 333 hsize = HSIZE; 334 if (fsize < (1 << 12)) 335 hsize = min(5003, HSIZE); 336 else if (fsize < (1 << 13)) 337 hsize = min(9001, HSIZE); 338 else if (fsize < (1 << 14)) 339 hsize = min (18013, HSIZE); 340 else if (fsize < (1 << 15)) 341 hsize = min (35023, HSIZE); 342 else if (fsize < 47000) 343 hsize = min (50021, HSIZE); 344 345 /* Generate output filename */ 346 strcpy(ofname, *fileptr); 347 #ifndef BSD4_2 348 if ((cp=strrchr(ofname,'/')) != NULL) 349 cp++; 350 else 351 cp = ofname; 352 /* 353 *** changed 12 to 25; should be NAMELEN-3, but I don't want 354 * to fight the headers. ehg 5 Nov 92 ** 355 */ 356 if (strlen(cp) > 25) { 357 fprintf(stderr, "%s: filename too long to tack on .Z\n", 358 cp); 359 continue; 360 } 361 #endif 362 strcat(ofname, ".Z"); 363 } 364 /* Check for overwrite of existing file */ 365 if (overwrite == 0 && zcat_flg == 0 && 366 stat(ofname, &statbuf) == 0) { 367 char response[2]; 368 369 response[0] = 'n'; 370 fprintf(stderr, "%s already exists;", ofname); 371 if (foreground()) { 372 fprintf(stderr, 373 " do you wish to overwrite %s (y or n)? ", 374 ofname); 375 fflush(stderr); 376 (void) read(2, response, 2); 377 while (response[1] != '\n') 378 if (read(2, response+1, 1) < 0) { 379 /* Ack! */ 380 perror("stderr"); 381 break; 382 } 383 } 384 if (response[0] != 'y') { 385 fprintf(stderr, "\tnot overwritten\n"); 386 continue; 387 } 388 } 389 if(zcat_flg == 0) { /* Open output file */ 390 if (freopen(ofname, "w", stdout) == NULL) { 391 perror(ofname); 392 continue; 393 } 394 if(!quiet) 395 fprintf(stderr, "%s: ", *fileptr); 396 } 397 398 /* Actually do the compression/decompression */ 399 if (do_decomp == 0) 400 compress(); 401 #ifndef DEBUG 402 else 403 decompress(); 404 #else 405 else if (debug == 0) 406 decompress(); 407 else 408 printcodes(); 409 if (verbose) 410 dump_tab(); 411 #endif /* DEBUG */ 412 if(zcat_flg == 0) { 413 copystat(*fileptr, ofname); /* Copy stats */ 414 if (exit_stat == 1 || !quiet) 415 putc('\n', stderr); 416 } 417 } 418 } else { /* Standard input */ 419 if (do_decomp == 0) { 420 compress(); 421 #ifdef DEBUG 422 if(verbose) 423 dump_tab(); 424 #endif 425 if(!quiet) 426 putc('\n', stderr); 427 } else { 428 /* Check the magic number */ 429 if (nomagic == 0) { 430 if ((getchar()!=(magic_header[0] & 0xFF)) 431 || (getchar()!=(magic_header[1] & 0xFF))) { 432 fprintf(stderr, "stdin: not in compressed format\n"); 433 exit(1); 434 } 435 maxbits = getchar(); /* set -b from file */ 436 block_compress = maxbits & BLOCK_MASK; 437 maxbits &= BIT_MASK; 438 maxmaxcode = 1 << maxbits; 439 fsize = 100000; /* assume stdin large for USERMEM */ 440 if(maxbits > BITS) { 441 fprintf(stderr, 442 "stdin: compressed with %d bits, can only handle %d bits\n", 443 maxbits, BITS); 444 exit(1); 445 } 446 } 447 #ifndef DEBUG 448 decompress(); 449 #else 450 if (debug == 0) 451 decompress(); 452 else 453 printcodes(); 454 if (verbose) 455 dump_tab(); 456 #endif /* DEBUG */ 457 } 458 } 459 exit(exit_stat); 460 return 0; 461 } 462 463 static int offset; 464 long in_count = 1; /* length of input */ 465 long bytes_out; /* length of compressed output */ 466 long out_count = 0; /* # of codes output (for debugging) */ 467 468 /* 469 * compress stdin to stdout 470 * 471 * Algorithm: use open addressing double hashing (no chaining) on the 472 * prefix code / next character combination. We do a variant of Knuth's 473 * algorithm D (vol. 3, sec. 6.4) along with G. Knott's relatively-prime 474 * secondary probe. Here, the modular division first probe is gives way 475 * to a faster exclusive-or manipulation. Also do block compression with 476 * an adaptive reset, whereby the code table is cleared when the compression 477 * ratio decreases, but after the table fills. The variable-length output 478 * codes are re-sized at this point, and a special CLEAR code is generated 479 * for the decompressor. Late addition: construct the table according to 480 * file size for noticeable speed improvement on small files. Please direct 481 * questions about this implementation to ames!jaw. 482 */ 483 void 484 compress(void) 485 { 486 code_int ent, hsize_reg; 487 code_int i; 488 int c, disp, hshift; 489 long fcode; 490 491 if (nomagic == 0) { 492 putchar(magic_header[0]); 493 putchar(magic_header[1]); 494 putchar((char)(maxbits | block_compress)); 495 if(ferror(stdout)) 496 writeerr(); 497 } 498 offset = 0; 499 bytes_out = 3; /* includes 3-byte header mojo */ 500 out_count = 0; 501 clear_flg = 0; 502 ratio = 0; 503 in_count = 1; 504 checkpoint = CHECK_GAP; 505 maxcode = MAXCODE(n_bits = INIT_BITS); 506 free_ent = (block_compress? FIRST: 256); 507 508 ent = getchar (); 509 510 hshift = 0; 511 for (fcode = (long)hsize; fcode < 65536L; fcode *= 2) 512 hshift++; 513 hshift = 8 - hshift; /* set hash code range bound */ 514 515 hsize_reg = hsize; 516 cl_hash( (count_int) hsize_reg); /* clear hash table */ 517 518 while ((c = getchar()) != EOF) { 519 in_count++; 520 fcode = (long) (((long) c << maxbits) + ent); 521 i = ((c << hshift) ^ ent); /* xor hashing */ 522 523 if (htabof (i) == fcode) { 524 ent = codetabof(i); 525 continue; 526 } else if ((long)htabof(i) < 0 ) /* empty slot */ 527 goto nomatch; 528 disp = hsize_reg - i; /* secondary hash (after G. Knott) */ 529 if (i == 0) 530 disp = 1; 531 probe: 532 if ((i -= disp) < 0) 533 i += hsize_reg; 534 535 if (htabof (i) == fcode) { 536 ent = codetabof(i); 537 continue; 538 } 539 if ((long)htabof(i) > 0) 540 goto probe; 541 nomatch: 542 output((code_int)ent); 543 out_count++; 544 ent = c; 545 if (free_ent < maxmaxcode) { 546 codetabof(i) = free_ent++; /* code -> hashtable */ 547 htabof(i) = fcode; 548 } else if ((count_int)in_count >= checkpoint && block_compress) 549 cl_block (); 550 } 551 /* 552 * Put out the final code. 553 */ 554 output( (code_int)ent ); 555 out_count++; 556 output( (code_int)-1 ); 557 558 /* 559 * Print out stats on stderr 560 */ 561 if(zcat_flg == 0 && !quiet) { 562 #ifdef DEBUG 563 fprintf( stderr, 564 "%ld chars in, %ld codes (%ld bytes) out, compression factor: ", 565 in_count, out_count, bytes_out ); 566 prratio( stderr, in_count, bytes_out ); 567 fprintf( stderr, "\n"); 568 fprintf( stderr, "\tCompression as in compact: " ); 569 prratio( stderr, in_count-bytes_out, in_count ); 570 fprintf( stderr, "\n"); 571 fprintf( stderr, "\tLargest code (of last block) was %d (%d bits)\n", 572 free_ent - 1, n_bits ); 573 #else /* !DEBUG */ 574 fprintf( stderr, "Compression: " ); 575 prratio( stderr, in_count-bytes_out, in_count ); 576 #endif /* DEBUG */ 577 } 578 if(bytes_out > in_count) /* exit(2) if no savings */ 579 exit_stat = 2; 580 } 581 582 /* 583 * TAG( output ) 584 * 585 * Output the given code. 586 * Inputs: 587 * code: A n_bits-bit integer. If == -1, then EOF. This assumes 588 * that n_bits =< (long)wordsize - 1. 589 * Outputs: 590 * Outputs code to the file. 591 * Assumptions: 592 * Chars are 8 bits long. 593 * Algorithm: 594 * Maintain a BITS character long buffer (so that 8 codes will 595 * fit in it exactly). When the buffer fills up empty it and start over. 596 */ 597 598 static char buf[BITS]; 599 600 uchar lmask[9] = {0xff, 0xfe, 0xfc, 0xf8, 0xf0, 0xe0, 0xc0, 0x80, 0x00}; 601 uchar rmask[9] = {0x00, 0x01, 0x03, 0x07, 0x0f, 0x1f, 0x3f, 0x7f, 0xff}; 602 603 void 604 output( code ) 605 code_int code; 606 { 607 #ifdef DEBUG 608 static int col = 0; 609 #endif 610 int r_off = offset, bits= n_bits; 611 char *bp = buf; 612 613 #ifdef DEBUG 614 if (verbose) 615 fprintf(stderr, "%5d%c", code, 616 (col+=6) >= 74? (col = 0, '\n'): ' '); 617 #endif 618 if (code >= 0) { 619 /* 620 * byte/bit numbering on the VAX is simulated by the 621 * following code 622 */ 623 /* 624 * Get to the first byte. 625 */ 626 bp += (r_off >> 3); 627 r_off &= 7; 628 /* 629 * Since code is always >= 8 bits, only need to mask the first 630 * hunk on the left. 631 */ 632 *bp = (*bp & rmask[r_off]) | (code << r_off) & lmask[r_off]; 633 bp++; 634 bits -= 8 - r_off; 635 code >>= 8 - r_off; 636 /* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */ 637 if ( bits >= 8 ) { 638 *bp++ = code; 639 code >>= 8; 640 bits -= 8; 641 } 642 /* Last bits. */ 643 if(bits) 644 *bp = code; 645 646 offset += n_bits; 647 if ( offset == (n_bits << 3) ) { 648 bp = buf; 649 bits = n_bits; 650 bytes_out += bits; 651 do { 652 putchar(*bp++); 653 } while(--bits); 654 offset = 0; 655 } 656 657 /* 658 * If the next entry is going to be too big for the code size, 659 * then increase it, if possible. 660 */ 661 if ( free_ent > maxcode || (clear_flg > 0)) { 662 /* 663 * Write the whole buffer, because the input side won't 664 * discover the size increase until after it has read it. 665 */ 666 if ( offset > 0 ) { 667 if( fwrite( buf, 1, n_bits, stdout ) != n_bits) 668 writeerr(); 669 bytes_out += n_bits; 670 } 671 offset = 0; 672 673 if ( clear_flg ) { 674 maxcode = MAXCODE (n_bits = INIT_BITS); 675 clear_flg = 0; 676 } else { 677 n_bits++; 678 if ( n_bits == maxbits ) 679 maxcode = maxmaxcode; 680 else 681 maxcode = MAXCODE(n_bits); 682 } 683 #ifdef DEBUG 684 if ( debug ) { 685 fprintf(stderr, 686 "\nChange to %d bits\n", n_bits); 687 col = 0; 688 } 689 #endif 690 } 691 } else { 692 /* 693 * At EOF, write the rest of the buffer. 694 */ 695 if ( offset > 0 ) 696 fwrite( buf, 1, (offset + 7) / 8, stdout ); 697 bytes_out += (offset + 7) / 8; 698 offset = 0; 699 fflush( stdout ); 700 #ifdef DEBUG 701 if ( verbose ) 702 fprintf( stderr, "\n" ); 703 #endif 704 if( ferror( stdout ) ) 705 writeerr(); 706 } 707 } 708 709 /* 710 * Decompress stdin to stdout. This routine adapts to the codes in the 711 * file building the "string" table on-the-fly; requiring no table to 712 * be stored in the compressed file. The tables used herein are shared 713 * with those of the compress() routine. See the definitions above. 714 */ 715 void 716 decompress(void) 717 { 718 int finchar; 719 code_int code, oldcode, incode; 720 uchar *stackp; 721 722 /* 723 * As above, initialize the first 256 entries in the table. 724 */ 725 maxcode = MAXCODE(n_bits = INIT_BITS); 726 for (code = 255; code >= 0; code--) { 727 tab_prefixof(code) = 0; 728 tab_suffixof(code) = (uchar)code; 729 } 730 free_ent = (block_compress? FIRST: 256); 731 732 finchar = oldcode = getcode(); 733 if(oldcode == -1) /* EOF already? */ 734 return; /* Get out of here */ 735 putchar((char)finchar); /* first code must be 8 bits = char */ 736 if(ferror(stdout)) /* Crash if can't write */ 737 writeerr(); 738 stackp = de_stack; 739 740 while ((code = getcode()) > -1) { 741 if ((code == CLEAR) && block_compress) { 742 for (code = 255; code >= 0; code--) 743 tab_prefixof(code) = 0; 744 clear_flg = 1; 745 free_ent = FIRST - 1; 746 if ((code = getcode()) == -1) /* O, untimely death! */ 747 break; 748 } 749 incode = code; 750 /* 751 * Special case for KwKwK string. 752 */ 753 if (code >= free_ent) { 754 *stackp++ = finchar; 755 code = oldcode; 756 } 757 758 /* 759 * Generate output characters in reverse order 760 */ 761 while (code >= 256) { 762 *stackp++ = tab_suffixof(code); 763 code = tab_prefixof(code); 764 } 765 *stackp++ = finchar = tab_suffixof(code); 766 767 /* 768 * And put them out in forward order 769 */ 770 do { 771 putchar(*--stackp); 772 } while (stackp > de_stack); 773 774 /* 775 * Generate the new entry. 776 */ 777 if ( (code=free_ent) < maxmaxcode ) { 778 tab_prefixof(code) = (ushort)oldcode; 779 tab_suffixof(code) = finchar; 780 free_ent = code+1; 781 } 782 /* 783 * Remember previous code. 784 */ 785 oldcode = incode; 786 } 787 fflush(stdout); 788 if(ferror(stdout)) 789 writeerr(); 790 } 791 792 /* 793 * TAG( getcode ) 794 * 795 * Read one code from the standard input. If EOF, return -1. 796 * Inputs: 797 * stdin 798 * Outputs: 799 * code or -1 is returned. 800 */ 801 code_int 802 getcode() 803 { 804 int r_off, bits; 805 code_int code; 806 static int offset = 0, size = 0; 807 static uchar buf[BITS]; 808 uchar *bp = buf; 809 810 if ( clear_flg > 0 || offset >= size || free_ent > maxcode ) { 811 /* 812 * If the next entry will be too big for the current code 813 * size, then we must increase the size. This implies reading 814 * a new buffer full, too. 815 */ 816 if ( free_ent > maxcode ) { 817 n_bits++; 818 if ( n_bits == maxbits ) 819 maxcode = maxmaxcode; /* won't get any bigger now */ 820 else 821 maxcode = MAXCODE(n_bits); 822 } 823 if ( clear_flg > 0) { 824 maxcode = MAXCODE(n_bits = INIT_BITS); 825 clear_flg = 0; 826 } 827 size = fread(buf, 1, n_bits, stdin); 828 if (size <= 0) 829 return -1; /* end of file */ 830 offset = 0; 831 /* Round size down to integral number of codes */ 832 size = (size << 3) - (n_bits - 1); 833 } 834 r_off = offset; 835 bits = n_bits; 836 /* 837 * Get to the first byte. 838 */ 839 bp += (r_off >> 3); 840 r_off &= 7; 841 /* Get first part (low order bits) */ 842 code = (*bp++ >> r_off); 843 bits -= (8 - r_off); 844 r_off = 8 - r_off; /* now, offset into code word */ 845 /* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */ 846 if (bits >= 8) { 847 code |= *bp++ << r_off; 848 r_off += 8; 849 bits -= 8; 850 } 851 /* high order bits. */ 852 code |= (*bp & rmask[bits]) << r_off; 853 offset += n_bits; 854 return code; 855 } 856 857 #ifdef DEBUG 858 printcodes() 859 { 860 /* 861 * Just print out codes from input file. For debugging. 862 */ 863 code_int code; 864 int col = 0, bits; 865 866 bits = n_bits = INIT_BITS; 867 maxcode = MAXCODE(n_bits); 868 free_ent = ((block_compress) ? FIRST : 256 ); 869 while ( ( code = getcode() ) >= 0 ) { 870 if ( (code == CLEAR) && block_compress ) { 871 free_ent = FIRST - 1; 872 clear_flg = 1; 873 } 874 else if ( free_ent < maxmaxcode ) 875 free_ent++; 876 if ( bits != n_bits ) { 877 fprintf(stderr, "\nChange to %d bits\n", n_bits ); 878 bits = n_bits; 879 col = 0; 880 } 881 fprintf(stderr, "%5d%c", code, (col+=6) >= 74 ? (col = 0, '\n') : ' ' ); 882 } 883 putc( '\n', stderr ); 884 exit( 0 ); 885 } 886 887 code_int sorttab[1<<BITS]; /* sorted pointers into htab */ 888 889 #define STACK_SIZE 15000 890 891 dump_tab() /* dump string table */ 892 { 893 int i, first, c, ent; 894 int stack_top = STACK_SIZE; 895 896 if(do_decomp == 0) { /* compressing */ 897 int flag = 1; 898 899 for(i=0; i<hsize; i++) { /* build sort pointers */ 900 if((long)htabof(i) >= 0) { 901 sorttab[codetabof(i)] = i; 902 } 903 } 904 first = block_compress ? FIRST : 256; 905 for(i = first; i < free_ent; i++) { 906 fprintf(stderr, "%5d: \"", i); 907 de_stack[--stack_top] = '\n'; 908 de_stack[--stack_top] = '"'; 909 stack_top = in_stack((htabof(sorttab[i])>>maxbits)&0xff, 910 stack_top); 911 for(ent=htabof(sorttab[i]) & ((1<<maxbits)-1); 912 ent > 256; 913 ent=htabof(sorttab[ent]) & ((1<<maxbits)-1)) { 914 stack_top = in_stack(htabof(sorttab[ent]) >> maxbits, 915 stack_top); 916 } 917 stack_top = in_stack(ent, stack_top); 918 fwrite( &de_stack[stack_top], 1, STACK_SIZE-stack_top, stderr); 919 stack_top = STACK_SIZE; 920 } 921 } else if(!debug) { /* decompressing */ 922 923 for ( i = 0; i < free_ent; i++ ) { 924 ent = i; 925 c = tab_suffixof(ent); 926 if ( isascii(c) && isprint(c) ) 927 fprintf( stderr, "%5d: %5d/'%c' \"", 928 ent, tab_prefixof(ent), c ); 929 else 930 fprintf( stderr, "%5d: %5d/\\%03o \"", 931 ent, tab_prefixof(ent), c ); 932 de_stack[--stack_top] = '\n'; 933 de_stack[--stack_top] = '"'; 934 for ( ; ent != NULL; 935 ent = (ent >= FIRST ? tab_prefixof(ent) : NULL) ) { 936 stack_top = in_stack(tab_suffixof(ent), stack_top); 937 } 938 fwrite( &de_stack[stack_top], 1, STACK_SIZE - stack_top, stderr ); 939 stack_top = STACK_SIZE; 940 } 941 } 942 } 943 944 int 945 in_stack(int c, int stack_top) 946 { 947 if ( (isascii(c) && isprint(c) && c != '\\') || c == ' ' ) { 948 de_stack[--stack_top] = c; 949 } else { 950 switch( c ) { 951 case '\n': de_stack[--stack_top] = 'n'; break; 952 case '\t': de_stack[--stack_top] = 't'; break; 953 case '\b': de_stack[--stack_top] = 'b'; break; 954 case '\f': de_stack[--stack_top] = 'f'; break; 955 case '\r': de_stack[--stack_top] = 'r'; break; 956 case '\\': de_stack[--stack_top] = '\\'; break; 957 default: 958 de_stack[--stack_top] = '0' + c % 8; 959 de_stack[--stack_top] = '0' + (c / 8) % 8; 960 de_stack[--stack_top] = '0' + c / 64; 961 break; 962 } 963 de_stack[--stack_top] = '\\'; 964 } 965 return stack_top; 966 } 967 #endif /* DEBUG */ 968 969 void 970 writeerr(void) 971 { 972 perror(ofname); 973 unlink(ofname); 974 exit(1); 975 } 976 977 void 978 copystat(ifname, ofname) 979 char *ifname, *ofname; 980 { 981 int mode; 982 time_t timep[2]; /* should be struct utimbuf */ 983 struct stat statbuf; 984 985 fclose(stdout); 986 if (stat(ifname, &statbuf)) { /* Get stat on input file */ 987 perror(ifname); 988 return; 989 } 990 if (!S_ISREG(statbuf.st_mode)) { 991 if (quiet) 992 fprintf(stderr, "%s: ", ifname); 993 fprintf(stderr, " -- not a regular file: unchanged"); 994 exit_stat = 1; 995 } else if (exit_stat == 2 && !force) { 996 /* No compression: remove file.Z */ 997 if (!quiet) 998 fprintf(stderr, " -- file unchanged"); 999 } else { /* Successful Compression */ 1000 exit_stat = 0; 1001 mode = statbuf.st_mode & 0777; 1002 if (chmod(ofname, mode)) /* Copy modes */ 1003 perror(ofname); 1004 /* Copy ownership */ 1005 chown(ofname, statbuf.st_uid, statbuf.st_gid); 1006 timep[0] = statbuf.st_atime; 1007 timep[1] = statbuf.st_mtime; 1008 /* Update last accessed and modified times */ 1009 utime(ofname, (struct utimbuf *)timep); 1010 // if (unlink(ifname)) /* Remove input file */ 1011 // perror(ifname); 1012 return; /* success */ 1013 } 1014 1015 /* Unsuccessful return -- one of the tests failed */ 1016 if (unlink(ofname)) 1017 perror(ofname); 1018 } 1019 1020 /* 1021 * This routine returns 1 if we are running in the foreground and stderr 1022 * is a tty. 1023 */ 1024 int 1025 foreground(void) 1026 { 1027 if(bgnd_flag) /* background? */ 1028 return 0; 1029 else /* foreground */ 1030 return isatty(2); /* and stderr is a tty */ 1031 } 1032 1033 void 1034 onintr(int x) 1035 { 1036 USED(x); 1037 unlink(ofname); 1038 exit(1); 1039 } 1040 1041 void 1042 oops(int x) /* wild pointer -- assume bad input */ 1043 { 1044 USED(x); 1045 if (do_decomp == 1) 1046 fprintf(stderr, "uncompress: corrupt input\n"); 1047 unlink(ofname); 1048 exit(1); 1049 } 1050 1051 void 1052 cl_block(void) /* table clear for block compress */ 1053 { 1054 long rat; 1055 1056 checkpoint = in_count + CHECK_GAP; 1057 #ifdef DEBUG 1058 if ( debug ) { 1059 fprintf ( stderr, "count: %ld, ratio: ", in_count ); 1060 prratio ( stderr, in_count, bytes_out ); 1061 fprintf ( stderr, "\n"); 1062 } 1063 #endif /* DEBUG */ 1064 1065 if (in_count > 0x007fffff) { /* shift will overflow */ 1066 rat = bytes_out >> 8; 1067 if (rat == 0) /* Don't divide by zero */ 1068 rat = 0x7fffffff; 1069 else 1070 rat = in_count / rat; 1071 } else 1072 rat = (in_count << 8) / bytes_out; /* 8 fractional bits */ 1073 if (rat > ratio) 1074 ratio = rat; 1075 else { 1076 ratio = 0; 1077 #ifdef DEBUG 1078 if (verbose) 1079 dump_tab(); /* dump string table */ 1080 #endif 1081 cl_hash((count_int)hsize); 1082 free_ent = FIRST; 1083 clear_flg = 1; 1084 output((code_int)CLEAR); 1085 #ifdef DEBUG 1086 if (debug) 1087 fprintf(stderr, "clear\n"); 1088 #endif /* DEBUG */ 1089 } 1090 } 1091 1092 void 1093 cl_hash(count_int hsize) /* reset code table */ 1094 { 1095 count_int *htab_p = htab+hsize; 1096 long i; 1097 long m1 = -1; 1098 1099 i = hsize - 16; 1100 do { /* might use Sys V memset(3) here */ 1101 *(htab_p-16) = m1; 1102 *(htab_p-15) = m1; 1103 *(htab_p-14) = m1; 1104 *(htab_p-13) = m1; 1105 *(htab_p-12) = m1; 1106 *(htab_p-11) = m1; 1107 *(htab_p-10) = m1; 1108 *(htab_p-9) = m1; 1109 *(htab_p-8) = m1; 1110 *(htab_p-7) = m1; 1111 *(htab_p-6) = m1; 1112 *(htab_p-5) = m1; 1113 *(htab_p-4) = m1; 1114 *(htab_p-3) = m1; 1115 *(htab_p-2) = m1; 1116 *(htab_p-1) = m1; 1117 htab_p -= 16; 1118 } while ((i -= 16) >= 0); 1119 for ( i += 16; i > 0; i-- ) 1120 *--htab_p = m1; 1121 } 1122 1123 void 1124 prratio(stream, num, den) 1125 FILE *stream; 1126 long num, den; 1127 { 1128 int q; /* Doesn't need to be long */ 1129 1130 if(num > 214748L) /* 2147483647/10000 */ 1131 q = num / (den / 10000L); 1132 else 1133 q = 10000L * num / den; /* Long calculations, though */ 1134 if (q < 0) { 1135 putc('-', stream); 1136 q = -q; 1137 } 1138 fprintf(stream, "%d.%02d%%", q / 100, q % 100); 1139 } 1140 1141 void 1142 version(void) 1143 { 1144 fprintf(stderr, "%s\n", rcs_ident); 1145 fprintf(stderr, "Options: "); 1146 #ifdef DEBUG 1147 fprintf(stderr, "DEBUG, "); 1148 #endif 1149 #ifdef BSD4_2 1150 fprintf(stderr, "BSD4_2, "); 1151 #endif 1152 fprintf(stderr, "BITS = %d\n", BITS); 1153 } 1154 1155 /* 1156 * The revision-history novel: 1157 * 1158 * $Header: compress.c,v 4.0 85/07/30 12:50:00 joe Release $ 1159 * $Log: compress.c,v $ 1160 * Revision 4.0 85/07/30 12:50:00 joe 1161 * Removed ferror() calls in output routine on every output except first. 1162 * Prepared for release to the world. 1163 * 1164 * Revision 3.6 85/07/04 01:22:21 joe 1165 * Remove much wasted storage by overlaying hash table with the tables 1166 * used by decompress: tab_suffix[1<<BITS], stack[8000]. Updated USERMEM 1167 * computations. Fixed dump_tab() DEBUG routine. 1168 * 1169 * Revision 3.5 85/06/30 20:47:21 jaw 1170 * Change hash function to use exclusive-or. Rip out hash cache. These 1171 * speedups render the megamemory version defunct, for now. Make decoder 1172 * stack global. Parts of the RCS trunks 2.7, 2.6, and 2.1 no longer apply. 1173 * 1174 * Revision 3.4 85/06/27 12:00:00 ken 1175 * Get rid of all floating-point calculations by doing all compression ratio 1176 * calculations in fixed point. 1177 * 1178 * Revision 3.3 85/06/24 21:53:24 joe 1179 * Incorporate portability suggestion for M_XENIX. Got rid of text on #else 1180 * and #endif lines. Cleaned up #ifdefs for vax and interdata. 1181 * 1182 * Revision 3.2 85/06/06 21:53:24 jaw 1183 * Incorporate portability suggestions for Z8000, IBM PC/XT from mailing list. 1184 * Default to "quiet" output (no compression statistics). 1185 * 1186 * Revision 3.1 85/05/12 18:56:13 jaw 1187 * Integrate decompress() stack speedups (from early pointer mods by McKie). 1188 * Repair multi-file USERMEM gaffe. Unify 'force' flags to mimic semantics 1189 * of SVR2 'pack'. Streamline block-compress table clear logic. Increase 1190 * output byte count by magic number size. 1191 * 1192 * Revision 3.0 84/11/27 11:50:00 petsd!joe 1193 * Set HSIZE depending on BITS. Set BITS depending on USERMEM. Unrolled 1194 * loops in clear routines. Added "-C" flag for 2.0 compatibility. Used 1195 * unsigned compares on Perkin-Elmer. Fixed foreground check. 1196 * 1197 * Revision 2.7 84/11/16 19:35:39 ames!jaw 1198 * Cache common hash codes based on input statistics; this improves 1199 * performance for low-density raster images. Pass on #ifdef bundle 1200 * from Turkowski. 1201 * 1202 * Revision 2.6 84/11/05 19:18:21 ames!jaw 1203 * Vary size of hash tables to reduce time for small files. 1204 * Tune PDP-11 hash function. 1205 * 1206 * Revision 2.5 84/10/30 20:15:14 ames!jaw 1207 * Junk chaining; replace with the simpler (and, on the VAX, faster) 1208 * double hashing, discussed within. Make block compression standard. 1209 * 1210 * Revision 2.4 84/10/16 11:11:11 ames!jaw 1211 * Introduce adaptive reset for block compression, to boost the rate 1212 * another several percent. (See mailing list notes.) 1213 * 1214 * Revision 2.3 84/09/22 22:00:00 petsd!joe 1215 * Implemented "-B" block compress. Implemented REVERSE sorting of tab_next. 1216 * Bug fix for last bits. Changed fwrite to putchar loop everywhere. 1217 * 1218 * Revision 2.2 84/09/18 14:12:21 ames!jaw 1219 * Fold in news changes, small machine typedef from thomas, 1220 * #ifdef interdata from joe. 1221 * 1222 * Revision 2.1 84/09/10 12:34:56 ames!jaw 1223 * Configured fast table lookup for 32-bit machines. 1224 * This cuts user time in half for b <= FBITS, and is useful for news batching 1225 * from VAX to PDP sites. Also sped up decompress() [fwrite->putc] and 1226 * added signal catcher [plus beef in writeerr()] to delete effluvia. 1227 * 1228 * Revision 2.0 84/08/28 22:00:00 petsd!joe 1229 * Add check for foreground before prompting user. Insert maxbits into 1230 * compressed file. Force file being uncompressed to end with ".Z". 1231 * Added "-c" flag and "zcat". Prepared for release. 1232 * 1233 * Revision 1.10 84/08/24 18:28:00 turtlevax!ken 1234 * Will only compress regular files (no directories), added a magic number 1235 * header (plus an undocumented -n flag to handle old files without headers), 1236 * added -f flag to force overwriting of possibly existing destination file, 1237 * otherwise the user is prompted for a response. Will tack on a .Z to a 1238 * filename if it doesn't have one when decompressing. Will only replace 1239 * file if it was compressed. 1240 * 1241 * Revision 1.9 84/08/16 17:28:00 turtlevax!ken 1242 * Removed scanargs(), getopt(), added .Z extension and unlimited number of 1243 * filenames to compress. Flags may be clustered (-Ddvb12) or separated 1244 * (-D -d -v -b 12), or combination thereof. Modes and other status is 1245 * copied with copystat(). -O bug for 4.2 seems to have disappeared with 1246 * 1.8. 1247 * 1248 * Revision 1.8 84/08/09 23:15:00 joe 1249 * Made it compatible with vax version, installed jim's fixes/enhancements 1250 * 1251 * Revision 1.6 84/08/01 22:08:00 joe 1252 * Sped up algorithm significantly by sorting the compress chain. 1253 * 1254 * Revision 1.5 84/07/13 13:11:00 srd 1255 * Added C version of vax asm routines. Changed structure to arrays to 1256 * save much memory. Do unsigned compares where possible (faster on 1257 * Perkin-Elmer) 1258 * 1259 * Revision 1.4 84/07/05 03:11:11 thomas 1260 * Clean up the code a little and lint it. (Lint complains about all 1261 * the regs used in the asm, but I'm not going to "fix" this.) 1262 * 1263 * Revision 1.3 84/07/05 02:06:54 thomas 1264 * Minor fixes. 1265 * 1266 * Revision 1.2 84/07/05 00:27:27 thomas 1267 * Add variable bit length output. 1268 */ 1269