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. All advertising materials mentioning features or use of this software
18 * must display the following acknowledgement:
19 * This product includes software developed by the University of
20 * California, Berkeley and its contributors.
21 * 4. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 * @(#)zopen.c 8.1 (Berkeley) 6/27/93
38 * $FreeBSD: src/usr.bin/compress/zopen.c,v 1.5.6.1 2002/07/16 00:52:08 tjr Exp $
39 * $DragonFly: src/usr.bin/compress/zopen.c,v 1.3 2003/10/02 17:42:27 hmp Exp $
43 * fcompress.c - File compression ala IEEE Computer, June 1984.
46 * Spencer W. Thomas (decvax!utah-cs!thomas)
47 * Jim McKie (decvax!mcvax!jim)
48 * Steve Davies (decvax!vax135!petsd!peora!srd)
49 * Ken Turkowski (decvax!decwrl!turtlevax!ken)
50 * James A. Woods (decvax!ihnp4!ames!jaw)
51 * Joe Orost (decvax!vax135!petsd!joe)
53 * Cleaned up and converted to library returning I/O streams by
54 * Diomidis Spinellis <dds@doc.ic.ac.uk>.
56 * zopen(filename, mode, bits)
57 * Returns a FILE * that can be used for read or write. The modes
58 * supported are only "r" and "w". Seeking is not allowed. On
59 * reading the file is decompressed, on writing it is compressed.
60 * The output is compatible with compress(1) with 16 bit tables.
61 * Any file produced by compress(1) can be read.
64 #include <sys/param.h>
76 #define BITS 16 /* Default bits. */
77 #define HSIZE 69001 /* 95% occupancy */
79 /* A code_int must be able to hold 2**BITS values of type int, and also -1. */
80 typedef long code_int
;
81 typedef long count_int
;
83 typedef u_char char_type
;
84 static char_type magic_header
[] =
85 {'\037', '\235'}; /* 1F 9D */
87 #define BIT_MASK 0x1f /* Defines for third byte of header. */
88 #define BLOCK_MASK 0x80
91 * Masks 0x40 and 0x20 are free. I think 0x20 should mean that there is
92 * a fourth header byte (for expansion).
94 #define INIT_BITS 9 /* Initial number of bits/code. */
96 #define MAXCODE(n_bits) ((1 << (n_bits)) - 1)
99 FILE *zs_fp
; /* File stream for I/O */
100 char zs_mode
; /* r or w */
102 S_START
, S_MIDDLE
, S_EOF
103 } zs_state
; /* State of computation */
104 u_int zs_n_bits
; /* Number of bits/code. */
105 u_int zs_maxbits
; /* User settable max # bits/code. */
106 code_int zs_maxcode
; /* Maximum code, given n_bits. */
107 code_int zs_maxmaxcode
; /* Should NEVER generate this code. */
108 count_int zs_htab
[HSIZE
];
109 u_short zs_codetab
[HSIZE
];
110 code_int zs_hsize
; /* For dynamic table sizing. */
111 code_int zs_free_ent
; /* First unused entry. */
113 * Block compression parameters -- after all codes are used up,
114 * and compression rate changes, start over.
116 int zs_block_compress
;
119 count_int zs_checkpoint
;
121 long zs_in_count
; /* Length of input. */
122 long zs_bytes_out
; /* Length of compressed output. */
123 long zs_out_count
; /* # of codes output (for debugging). */
124 char_type zs_buf
[BITS
];
129 code_int zs_hsize_reg
;
131 } w
; /* Write paramenters */
133 char_type
*zs_stackp
;
135 code_int zs_code
, zs_oldcode
, zs_incode
;
136 int zs_roffset
, zs_size
;
137 char_type zs_gbuf
[BITS
];
138 } r
; /* Read parameters */
142 /* Definitions to retain old variable names */
144 #define zmode zs->zs_mode
145 #define state zs->zs_state
146 #define n_bits zs->zs_n_bits
147 #define maxbits zs->zs_maxbits
148 #define maxcode zs->zs_maxcode
149 #define maxmaxcode zs->zs_maxmaxcode
150 #define htab zs->zs_htab
151 #define codetab zs->zs_codetab
152 #define hsize zs->zs_hsize
153 #define free_ent zs->zs_free_ent
154 #define block_compress zs->zs_block_compress
155 #define clear_flg zs->zs_clear_flg
156 #define ratio zs->zs_ratio
157 #define checkpoint zs->zs_checkpoint
158 #define offset zs->zs_offset
159 #define in_count zs->zs_in_count
160 #define bytes_out zs->zs_bytes_out
161 #define out_count zs->zs_out_count
162 #define buf zs->zs_buf
163 #define fcode zs->u.w.zs_fcode
164 #define hsize_reg zs->u.w.zs_hsize_reg
165 #define ent zs->u.w.zs_ent
166 #define hshift zs->u.w.zs_hshift
167 #define stackp zs->u.r.zs_stackp
168 #define finchar zs->u.r.zs_finchar
169 #define code zs->u.r.zs_code
170 #define oldcode zs->u.r.zs_oldcode
171 #define incode zs->u.r.zs_incode
172 #define roffset zs->u.r.zs_roffset
173 #define size zs->u.r.zs_size
174 #define gbuf zs->u.r.zs_gbuf
177 * To save much memory, we overlay the table used by compress() with those
178 * used by decompress(). The tab_prefix table is the same size and type as
179 * the codetab. The tab_suffix table needs 2**BITS characters. We get this
180 * from the beginning of htab. The output stack uses the rest of htab, and
181 * contains characters. There is plenty of room for any possible stack
182 * (stack used to be 8000 characters).
185 #define htabof(i) htab[i]
186 #define codetabof(i) codetab[i]
188 #define tab_prefixof(i) codetabof(i)
189 #define tab_suffixof(i) ((char_type *)(htab))[i]
190 #define de_stack ((char_type *)&tab_suffixof(1 << BITS))
192 #define CHECK_GAP 10000 /* Ratio check interval. */
195 * the next two codes should not be changed lightly, as they must not
196 * lie within the contiguous general code space.
198 #define FIRST 257 /* First free entry. */
199 #define CLEAR 256 /* Table clear output code. */
201 static int cl_block(struct s_zstate
*);
202 static void cl_hash(struct s_zstate
*, count_int
);
203 static code_int
getcode(struct s_zstate
*);
204 static int output(struct s_zstate
*, code_int
);
205 static int zclose(void *);
206 static int zread(void *, char *, int);
207 static int zwrite(void *, const char *, int);
210 * Algorithm from "A Technique for High Performance Data Compression",
211 * Terry A. Welch, IEEE Computer Vol 17, No 6 (June 1984), pp 8-19.
214 * Modified Lempel-Ziv method (LZW). Basically finds common
215 * substrings and replaces them with a variable size code. This is
216 * deterministic, and can be done on the fly. Thus, the decompression
217 * procedure needs no input table, but tracks the way the table was built.
223 * Algorithm: use open addressing double hashing (no chaining) on the
224 * prefix code / next character combination. We do a variant of Knuth's
225 * algorithm D (vol. 3, sec. 6.4) along with G. Knott's relatively-prime
226 * secondary probe. Here, the modular division first probe is gives way
227 * to a faster exclusive-or manipulation. Also do block compression with
228 * an adaptive reset, whereby the code table is cleared when the compression
229 * ratio decreases, but after the table fills. The variable-length output
230 * codes are re-sized at this point, and a special CLEAR code is generated
231 * for the decompressor. Late addition: construct the table according to
232 * file size for noticeable speed improvement on small files. Please direct
233 * questions about this implementation to ames!jaw.
236 zwrite(void *cookie
, const char *wbp
, int num
)
251 if (state
== S_MIDDLE
)
255 maxmaxcode
= 1L << maxbits
;
256 if (fwrite(magic_header
,
257 sizeof(char), sizeof(magic_header
), fp
) != sizeof(magic_header
))
259 tmp
= (u_char
)((maxbits
) | block_compress
);
260 if (fwrite(&tmp
, sizeof(char), sizeof(tmp
), fp
) != sizeof(tmp
))
264 bytes_out
= 3; /* Includes 3-byte header mojo. */
269 checkpoint
= CHECK_GAP
;
270 maxcode
= MAXCODE(n_bits
= INIT_BITS
);
271 free_ent
= ((block_compress
) ? FIRST
: 256);
277 for (fcode
= (long)hsize
; fcode
< 65536L; fcode
*= 2L)
279 hshift
= 8 - hshift
; /* Set hash code range bound. */
282 cl_hash(zs
, (count_int
)hsize_reg
); /* Clear hash table. */
284 middle
: for (i
= 0; count
--;) {
287 fcode
= (long)(((long)c
<< maxbits
) + ent
);
288 i
= ((c
<< hshift
) ^ ent
); /* Xor hashing. */
290 if (htabof(i
) == fcode
) {
293 } else if ((long)htabof(i
) < 0) /* Empty slot. */
295 disp
= hsize_reg
- i
; /* Secondary hash (after G. Knott). */
298 probe
: if ((i
-= disp
) < 0)
301 if (htabof(i
) == fcode
) {
305 if ((long)htabof(i
) >= 0)
307 nomatch
: if (output(zs
, (code_int
) ent
) == -1)
311 if (free_ent
< maxmaxcode
) {
312 codetabof(i
) = free_ent
++; /* code -> hashtable */
314 } else if ((count_int
)in_count
>=
315 checkpoint
&& block_compress
) {
316 if (cl_block(zs
) == -1)
330 if (zmode
== 'w') { /* Put out the final code. */
331 if (output(zs
, (code_int
) ent
) == -1) {
337 if (output(zs
, (code_int
) - 1) == -1) {
343 rval
= fclose(fp
) == EOF
? -1 : 0;
349 * Output the given code.
351 * code: A n_bits-bit integer. If == -1, then EOF. This assumes
352 * that n_bits =< (long)wordsize - 1.
354 * Outputs code to the file.
356 * Chars are 8 bits long.
358 * Maintain a BITS character long buffer (so that 8 codes will
359 * fit in it exactly). Use the VAX insv instruction to insert each
360 * code in turn. When the buffer fills up empty it and start over.
363 static char_type lmask
[9] =
364 {0xff, 0xfe, 0xfc, 0xf8, 0xf0, 0xe0, 0xc0, 0x80, 0x00};
365 static char_type rmask
[9] =
366 {0x00, 0x01, 0x03, 0x07, 0x0f, 0x1f, 0x3f, 0x7f, 0xff};
369 output(struct s_zstate
*zs
, code_int ocode
)
379 /* Get to the first byte. */
383 * Since ocode is always >= 8 bits, only need to mask the first
386 *bp
= (*bp
& rmask
[r_off
]) | ((ocode
<< r_off
) & lmask
[r_off
]);
390 /* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
400 if (offset
== (n_bits
<< 3)) {
404 if (fwrite(bp
, sizeof(char), bits
, fp
) != bits
)
411 * If the next entry is going to be too big for the ocode size,
412 * then increase it, if possible.
414 if (free_ent
> maxcode
|| (clear_flg
> 0)) {
416 * Write the whole buffer, because the input side won't
417 * discover the size increase until after it has read it.
420 if (fwrite(buf
, 1, n_bits
, fp
) != n_bits
)
427 maxcode
= MAXCODE(n_bits
= INIT_BITS
);
431 if (n_bits
== maxbits
)
432 maxcode
= maxmaxcode
;
434 maxcode
= MAXCODE(n_bits
);
438 /* At EOF, write the rest of the buffer. */
440 offset
= (offset
+ 7) / 8;
441 if (fwrite(buf
, 1, offset
, fp
) != offset
)
451 * Decompress read. This routine adapts to the codes in the file building
452 * the "string" table on-the-fly; requiring no table to be stored in the
453 * compressed file. The tables used herein are shared with those of the
454 * compress() routine. See the definitions above.
457 zread(void *cookie
, char *rbp
, int num
)
461 u_char
*bp
, header
[3];
479 /* Check the magic number */
481 sizeof(char), sizeof(header
), fp
) != sizeof(header
) ||
482 memcmp(header
, magic_header
, sizeof(magic_header
)) != 0) {
486 maxbits
= header
[2]; /* Set -b from file. */
487 block_compress
= maxbits
& BLOCK_MASK
;
489 maxmaxcode
= 1L << maxbits
;
490 if (maxbits
> BITS
) {
494 /* As above, initialize the first 256 entries in the table. */
495 maxcode
= MAXCODE(n_bits
= INIT_BITS
);
496 for (code
= 255; code
>= 0; code
--) {
497 tab_prefixof(code
) = 0;
498 tab_suffixof(code
) = (char_type
) code
;
500 free_ent
= block_compress
? FIRST
: 256;
502 finchar
= oldcode
= getcode(zs
);
503 if (oldcode
== -1) /* EOF already? */
504 return (0); /* Get out of here */
506 /* First code must be 8 bits = char. */
507 *bp
++ = (u_char
)finchar
;
511 while ((code
= getcode(zs
)) > -1) {
513 if ((code
== CLEAR
) && block_compress
) {
514 for (code
= 255; code
>= 0; code
--)
515 tab_prefixof(code
) = 0;
517 free_ent
= FIRST
- 1;
518 if ((code
= getcode(zs
)) == -1) /* O, untimely death! */
523 /* Special case for KwKwK string. */
524 if (code
>= free_ent
) {
529 /* Generate output characters in reverse order. */
530 while (code
>= 256) {
531 *stackp
++ = tab_suffixof(code
);
532 code
= tab_prefixof(code
);
534 *stackp
++ = finchar
= tab_suffixof(code
);
536 /* And put them out in forward order. */
541 } while (stackp
> de_stack
);
543 /* Generate the new entry. */
544 if ((code
= free_ent
) < maxmaxcode
) {
545 tab_prefixof(code
) = (u_short
) oldcode
;
546 tab_suffixof(code
) = finchar
;
550 /* Remember previous code. */
554 eof
: return (num
- count
);
558 * Read one code from the standard input. If EOF, return -1.
562 * code or -1 is returned.
565 getcode(struct s_zstate
*zs
)
572 if (clear_flg
> 0 || roffset
>= size
|| free_ent
> maxcode
) {
574 * If the next entry will be too big for the current gcode
575 * size, then we must increase the size. This implies reading
576 * a new buffer full, too.
578 if (free_ent
> maxcode
) {
580 if (n_bits
== maxbits
) /* Won't get any bigger now. */
581 maxcode
= maxmaxcode
;
583 maxcode
= MAXCODE(n_bits
);
586 maxcode
= MAXCODE(n_bits
= INIT_BITS
);
589 size
= fread(gbuf
, 1, n_bits
, fp
);
590 if (size
<= 0) /* End of file. */
593 /* Round size down to integral number of codes. */
594 size
= (size
<< 3) - (n_bits
- 1);
599 /* Get to the first byte. */
603 /* Get first part (low order bits). */
604 gcode
= (*bp
++ >> r_off
);
606 r_off
= 8 - r_off
; /* Now, roffset into gcode word. */
608 /* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
610 gcode
|= *bp
++ << r_off
;
615 /* High order bits. */
616 gcode
|= (*bp
& rmask
[bits
]) << r_off
;
623 cl_block(struct s_zstate
*zs
) /* Table clear for block compress. */
627 checkpoint
= in_count
+ CHECK_GAP
;
629 if (in_count
> 0x007fffff) { /* Shift will overflow. */
630 rat
= bytes_out
>> 8;
631 if (rat
== 0) /* Don't divide by zero. */
634 rat
= in_count
/ rat
;
636 rat
= (in_count
<< 8) / bytes_out
; /* 8 fractional bits. */
641 cl_hash(zs
, (count_int
) hsize
);
644 if (output(zs
, (code_int
) CLEAR
) == -1)
651 cl_hash(struct s_zstate
*zs
, count_int cl_hsize
) /* Reset code table. */
657 htab_p
= htab
+ cl_hsize
;
659 do { /* Might use Sys V memset(3) here. */
677 } while ((i
-= 16) >= 0);
678 for (i
+= 16; i
> 0; i
--)
683 zopen(const char *fname
, const char *mode
, int bits
)
687 if ((mode
[0] != 'r' && mode
[0] != 'w') || mode
[1] != '\0' ||
688 bits
< 0 || bits
> BITS
) {
693 if ((zs
= calloc(1, sizeof(struct s_zstate
))) == NULL
)
696 maxbits
= bits
? bits
: BITS
; /* User settable max # bits/code. */
697 maxmaxcode
= 1L << maxbits
; /* Should NEVER generate this code. */
698 hsize
= HSIZE
; /* For dynamic table sizing. */
699 free_ent
= 0; /* First unused entry. */
700 block_compress
= BLOCK_MASK
;
703 checkpoint
= CHECK_GAP
;
704 in_count
= 1; /* Length of input. */
705 out_count
= 0; /* # of codes output (for debugging). */
711 * Layering compress on top of stdio in order to provide buffering,
712 * and ensure that reads and write work with the data specified.
714 if ((fp
= fopen(fname
, mode
)) == NULL
) {
721 return (funopen(zs
, zread
, NULL
, NULL
, zclose
));
724 return (funopen(zs
, NULL
, zwrite
, NULL
, zclose
));