2 * Copyright (c) 1985, 1986, 1992, 1993
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from software contributed to Berkeley by
6 * Diomidis Spinellis and James A. Woods, derived from original
7 * work by Spencer Thomas and Joseph Orost.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * @(#)zopen.c 8.1 (Berkeley) 6/27/93
34 * $FreeBSD: src/usr.bin/compress/zopen.c,v 1.5.6.1 2002/07/16 00:52:08 tjr Exp $
35 * $DragonFly: src/usr.bin/compress/zopen.c,v 1.3 2003/10/02 17:42:27 hmp Exp $
39 * fcompress.c - File compression ala IEEE Computer, June 1984.
42 * Spencer W. Thomas (decvax!utah-cs!thomas)
43 * Jim McKie (decvax!mcvax!jim)
44 * Steve Davies (decvax!vax135!petsd!peora!srd)
45 * Ken Turkowski (decvax!decwrl!turtlevax!ken)
46 * James A. Woods (decvax!ihnp4!ames!jaw)
47 * Joe Orost (decvax!vax135!petsd!joe)
49 * Cleaned up and converted to library returning I/O streams by
50 * Diomidis Spinellis <dds@doc.ic.ac.uk>.
52 * zopen(filename, mode, bits)
53 * Returns a FILE * that can be used for read or write. The modes
54 * supported are only "r" and "w". Seeking is not allowed. On
55 * reading the file is decompressed, on writing it is compressed.
56 * The output is compatible with compress(1) with 16 bit tables.
57 * Any file produced by compress(1) can be read.
60 #include <sys/param.h>
72 #define BITS 16 /* Default bits. */
73 #define HSIZE 69001 /* 95% occupancy */
75 /* A code_int must be able to hold 2**BITS values of type int, and also -1. */
76 typedef long code_int;
77 typedef long count_int;
79 typedef u_char char_type;
80 static char_type magic_header[] =
81 {'\037', '\235'}; /* 1F 9D */
83 #define BIT_MASK 0x1f /* Defines for third byte of header. */
84 #define BLOCK_MASK 0x80
87 * Masks 0x40 and 0x20 are free. I think 0x20 should mean that there is
88 * a fourth header byte (for expansion).
90 #define INIT_BITS 9 /* Initial number of bits/code. */
92 #define MAXCODE(n_bits) ((1 << (n_bits)) - 1)
95 FILE *zs_fp; /* File stream for I/O */
96 char zs_mode; /* r or w */
98 S_START, S_MIDDLE, S_EOF
99 } zs_state; /* State of computation */
100 u_int zs_n_bits; /* Number of bits/code. */
101 u_int zs_maxbits; /* User settable max # bits/code. */
102 code_int zs_maxcode; /* Maximum code, given n_bits. */
103 code_int zs_maxmaxcode; /* Should NEVER generate this code. */
104 count_int zs_htab [HSIZE];
105 u_short zs_codetab [HSIZE];
106 code_int zs_hsize; /* For dynamic table sizing. */
107 code_int zs_free_ent; /* First unused entry. */
109 * Block compression parameters -- after all codes are used up,
110 * and compression rate changes, start over.
112 int zs_block_compress;
115 count_int zs_checkpoint;
117 long zs_in_count; /* Length of input. */
118 long zs_bytes_out; /* Length of compressed output. */
119 long zs_out_count; /* # of codes output (for debugging). */
120 char_type zs_buf[BITS];
125 code_int zs_hsize_reg;
127 } w; /* Write paramenters */
129 char_type *zs_stackp;
131 code_int zs_code, zs_oldcode, zs_incode;
132 int zs_roffset, zs_size;
133 char_type zs_gbuf[BITS];
134 } r; /* Read parameters */
138 /* Definitions to retain old variable names */
140 #define zmode zs->zs_mode
141 #define state zs->zs_state
142 #define n_bits zs->zs_n_bits
143 #define maxbits zs->zs_maxbits
144 #define maxcode zs->zs_maxcode
145 #define maxmaxcode zs->zs_maxmaxcode
146 #define htab zs->zs_htab
147 #define codetab zs->zs_codetab
148 #define hsize zs->zs_hsize
149 #define free_ent zs->zs_free_ent
150 #define block_compress zs->zs_block_compress
151 #define clear_flg zs->zs_clear_flg
152 #define ratio zs->zs_ratio
153 #define checkpoint zs->zs_checkpoint
154 #define offset zs->zs_offset
155 #define in_count zs->zs_in_count
156 #define bytes_out zs->zs_bytes_out
157 #define out_count zs->zs_out_count
158 #define buf zs->zs_buf
159 #define fcode zs->u.w.zs_fcode
160 #define hsize_reg zs->u.w.zs_hsize_reg
161 #define ent zs->u.w.zs_ent
162 #define hshift zs->u.w.zs_hshift
163 #define stackp zs->u.r.zs_stackp
164 #define finchar zs->u.r.zs_finchar
165 #define code zs->u.r.zs_code
166 #define oldcode zs->u.r.zs_oldcode
167 #define incode zs->u.r.zs_incode
168 #define roffset zs->u.r.zs_roffset
169 #define size zs->u.r.zs_size
170 #define gbuf zs->u.r.zs_gbuf
173 * To save much memory, we overlay the table used by compress() with those
174 * used by decompress(). The tab_prefix table is the same size and type as
175 * the codetab. The tab_suffix table needs 2**BITS characters. We get this
176 * from the beginning of htab. The output stack uses the rest of htab, and
177 * contains characters. There is plenty of room for any possible stack
178 * (stack used to be 8000 characters).
181 #define htabof(i) htab[i]
182 #define codetabof(i) codetab[i]
184 #define tab_prefixof(i) codetabof(i)
185 #define tab_suffixof(i) ((char_type *)(htab))[i]
186 #define de_stack ((char_type *)&tab_suffixof(1 << BITS))
188 #define CHECK_GAP 10000 /* Ratio check interval. */
191 * the next two codes should not be changed lightly, as they must not
192 * lie within the contiguous general code space.
194 #define FIRST 257 /* First free entry. */
195 #define CLEAR 256 /* Table clear output code. */
197 static int cl_block(struct s_zstate *);
198 static void cl_hash(struct s_zstate *, count_int);
199 static code_int getcode(struct s_zstate *);
200 static int output(struct s_zstate *, code_int);
201 static int zclose(void *);
202 static int zread(void *, char *, int);
203 static int zwrite(void *, const char *, int);
206 * Algorithm from "A Technique for High Performance Data Compression",
207 * Terry A. Welch, IEEE Computer Vol 17, No 6 (June 1984), pp 8-19.
210 * Modified Lempel-Ziv method (LZW). Basically finds common
211 * substrings and replaces them with a variable size code. This is
212 * deterministic, and can be done on the fly. Thus, the decompression
213 * procedure needs no input table, but tracks the way the table was built.
219 * Algorithm: use open addressing double hashing (no chaining) on the
220 * prefix code / next character combination. We do a variant of Knuth's
221 * algorithm D (vol. 3, sec. 6.4) along with G. Knott's relatively-prime
222 * secondary probe. Here, the modular division first probe is gives way
223 * to a faster exclusive-or manipulation. Also do block compression with
224 * an adaptive reset, whereby the code table is cleared when the compression
225 * ratio decreases, but after the table fills. The variable-length output
226 * codes are re-sized at this point, and a special CLEAR code is generated
227 * for the decompressor. Late addition: construct the table according to
228 * file size for noticeable speed improvement on small files. Please direct
229 * questions about this implementation to ames!jaw.
232 zwrite(void *cookie, const char *wbp, int num)
247 if (state == S_MIDDLE)
251 maxmaxcode = 1L << maxbits;
252 if (fwrite(magic_header,
253 sizeof(char), sizeof(magic_header), fp) != sizeof(magic_header))
255 tmp = (u_char)((maxbits) | block_compress);
256 if (fwrite(&tmp, sizeof(char), sizeof(tmp), fp) != sizeof(tmp))
260 bytes_out = 3; /* Includes 3-byte header mojo. */
265 checkpoint = CHECK_GAP;
266 maxcode = MAXCODE(n_bits = INIT_BITS);
267 free_ent = ((block_compress) ? FIRST : 256);
273 for (fcode = (long)hsize; fcode < 65536L; fcode *= 2L)
275 hshift = 8 - hshift; /* Set hash code range bound. */
278 cl_hash(zs, (count_int)hsize_reg); /* Clear hash table. */
280 middle: for (i = 0; count--;) {
283 fcode = (long)(((long)c << maxbits) + ent);
284 i = ((c << hshift) ^ ent); /* Xor hashing. */
286 if (htabof(i) == fcode) {
289 } else if ((long)htabof(i) < 0) /* Empty slot. */
291 disp = hsize_reg - i; /* Secondary hash (after G. Knott). */
294 probe: if ((i -= disp) < 0)
297 if (htabof(i) == fcode) {
301 if ((long)htabof(i) >= 0)
303 nomatch: if (output(zs, (code_int) ent) == -1)
307 if (free_ent < maxmaxcode) {
308 codetabof(i) = free_ent++; /* code -> hashtable */
310 } else if ((count_int)in_count >=
311 checkpoint && block_compress) {
312 if (cl_block(zs) == -1)
326 if (zmode == 'w') { /* Put out the final code. */
327 if (output(zs, (code_int) ent) == -1) {
333 if (output(zs, (code_int) - 1) == -1) {
339 rval = fclose(fp) == EOF ? -1 : 0;
345 * Output the given code.
347 * code: A n_bits-bit integer. If == -1, then EOF. This assumes
348 * that n_bits =< (long)wordsize - 1.
350 * Outputs code to the file.
352 * Chars are 8 bits long.
354 * Maintain a BITS character long buffer (so that 8 codes will
355 * fit in it exactly). Use the VAX insv instruction to insert each
356 * code in turn. When the buffer fills up empty it and start over.
359 static char_type lmask[9] =
360 {0xff, 0xfe, 0xfc, 0xf8, 0xf0, 0xe0, 0xc0, 0x80, 0x00};
361 static char_type rmask[9] =
362 {0x00, 0x01, 0x03, 0x07, 0x0f, 0x1f, 0x3f, 0x7f, 0xff};
365 output(struct s_zstate *zs, code_int ocode)
375 /* Get to the first byte. */
379 * Since ocode is always >= 8 bits, only need to mask the first
382 *bp = (*bp & rmask[r_off]) | ((ocode << r_off) & lmask[r_off]);
386 /* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
396 if (offset == (n_bits << 3)) {
400 if (fwrite(bp, sizeof(char), bits, fp) != bits)
407 * If the next entry is going to be too big for the ocode size,
408 * then increase it, if possible.
410 if (free_ent > maxcode || (clear_flg > 0)) {
412 * Write the whole buffer, because the input side won't
413 * discover the size increase until after it has read it.
416 if (fwrite(buf, 1, n_bits, fp) != n_bits)
423 maxcode = MAXCODE(n_bits = INIT_BITS);
427 if (n_bits == maxbits)
428 maxcode = maxmaxcode;
430 maxcode = MAXCODE(n_bits);
434 /* At EOF, write the rest of the buffer. */
436 offset = (offset + 7) / 8;
437 if (fwrite(buf, 1, offset, fp) != offset)
447 * Decompress read. This routine adapts to the codes in the file building
448 * the "string" table on-the-fly; requiring no table to be stored in the
449 * compressed file. The tables used herein are shared with those of the
450 * compress() routine. See the definitions above.
453 zread(void *cookie, char *rbp, int num)
457 u_char *bp, header[3];
475 /* Check the magic number */
477 sizeof(char), sizeof(header), fp) != sizeof(header) ||
478 memcmp(header, magic_header, sizeof(magic_header)) != 0) {
482 maxbits = header[2]; /* Set -b from file. */
483 block_compress = maxbits & BLOCK_MASK;
485 maxmaxcode = 1L << maxbits;
486 if (maxbits > BITS) {
490 /* As above, initialize the first 256 entries in the table. */
491 maxcode = MAXCODE(n_bits = INIT_BITS);
492 for (code = 255; code >= 0; code--) {
493 tab_prefixof(code) = 0;
494 tab_suffixof(code) = (char_type) code;
496 free_ent = block_compress ? FIRST : 256;
498 finchar = oldcode = getcode(zs);
499 if (oldcode == -1) /* EOF already? */
500 return (0); /* Get out of here */
502 /* First code must be 8 bits = char. */
503 *bp++ = (u_char)finchar;
507 while ((code = getcode(zs)) > -1) {
509 if ((code == CLEAR) && block_compress) {
510 for (code = 255; code >= 0; code--)
511 tab_prefixof(code) = 0;
513 free_ent = FIRST - 1;
514 if ((code = getcode(zs)) == -1) /* O, untimely death! */
519 /* Special case for KwKwK string. */
520 if (code >= free_ent) {
525 /* Generate output characters in reverse order. */
526 while (code >= 256) {
527 *stackp++ = tab_suffixof(code);
528 code = tab_prefixof(code);
530 *stackp++ = finchar = tab_suffixof(code);
532 /* And put them out in forward order. */
537 } while (stackp > de_stack);
539 /* Generate the new entry. */
540 if ((code = free_ent) < maxmaxcode) {
541 tab_prefixof(code) = (u_short) oldcode;
542 tab_suffixof(code) = finchar;
546 /* Remember previous code. */
550 eof: return (num - count);
554 * Read one code from the standard input. If EOF, return -1.
558 * code or -1 is returned.
561 getcode(struct s_zstate *zs)
568 if (clear_flg > 0 || roffset >= size || free_ent > maxcode) {
570 * If the next entry will be too big for the current gcode
571 * size, then we must increase the size. This implies reading
572 * a new buffer full, too.
574 if (free_ent > maxcode) {
576 if (n_bits == maxbits) /* Won't get any bigger now. */
577 maxcode = maxmaxcode;
579 maxcode = MAXCODE(n_bits);
582 maxcode = MAXCODE(n_bits = INIT_BITS);
585 size = fread(gbuf, 1, n_bits, fp);
586 if (size <= 0) /* End of file. */
589 /* Round size down to integral number of codes. */
590 size = (size << 3) - (n_bits - 1);
595 /* Get to the first byte. */
599 /* Get first part (low order bits). */
600 gcode = (*bp++ >> r_off);
602 r_off = 8 - r_off; /* Now, roffset into gcode word. */
604 /* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
606 gcode |= *bp++ << r_off;
611 /* High order bits. */
612 gcode |= (*bp & rmask[bits]) << r_off;
619 cl_block(struct s_zstate *zs) /* Table clear for block compress. */
623 checkpoint = in_count + CHECK_GAP;
625 if (in_count > 0x007fffff) { /* Shift will overflow. */
626 rat = bytes_out >> 8;
627 if (rat == 0) /* Don't divide by zero. */
630 rat = in_count / rat;
632 rat = (in_count << 8) / bytes_out; /* 8 fractional bits. */
637 cl_hash(zs, (count_int) hsize);
640 if (output(zs, (code_int) CLEAR) == -1)
647 cl_hash(struct s_zstate *zs, count_int cl_hsize) /* Reset code table. */
653 htab_p = htab + cl_hsize;
655 do { /* Might use Sys V memset(3) here. */
673 } while ((i -= 16) >= 0);
674 for (i += 16; i > 0; i--)
679 zopen(const char *fname, const char *mode, int bits)
683 if ((mode[0] != 'r' && mode[0] != 'w') || mode[1] != '\0' ||
684 bits < 0 || bits > BITS) {
689 if ((zs = calloc(1, sizeof(struct s_zstate))) == NULL)
692 maxbits = bits ? bits : BITS; /* User settable max # bits/code. */
693 maxmaxcode = 1L << maxbits; /* Should NEVER generate this code. */
694 hsize = HSIZE; /* For dynamic table sizing. */
695 free_ent = 0; /* First unused entry. */
696 block_compress = BLOCK_MASK;
699 checkpoint = CHECK_GAP;
700 in_count = 1; /* Length of input. */
701 out_count = 0; /* # of codes output (for debugging). */
707 * Layering compress on top of stdio in order to provide buffering,
708 * and ensure that reads and write work with the data specified.
710 if ((fp = fopen(fname, mode)) == NULL) {
717 return (funopen(zs, zread, NULL, NULL, zclose));
720 return (funopen(zs, NULL, zwrite, NULL, zclose));