xref: /freebsd/usr.bin/compress/zopen.c (revision 2ad872c5794e4c26fdf6ed219ad3f09ca0d5304a)
1 /*-
2  * Copyright (c) 1985, 1986, 1992, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
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.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
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.
24  *
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
35  * SUCH DAMAGE.
36  */
37 
38 #if defined(LIBC_SCCS) && !defined(lint)
39 static char sccsid[] = "@(#)zopen.c	8.1 (Berkeley) 6/27/93";
40 #endif /* LIBC_SCCS and not lint */
41 
42 /*-
43  * fcompress.c - File compression ala IEEE Computer, June 1984.
44  *
45  * Compress authors:
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)
52  *
53  * Cleaned up and converted to library returning I/O streams by
54  * Diomidis Spinellis <dds@doc.ic.ac.uk>.
55  *
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.
62  */
63 
64 #include <sys/param.h>
65 #include <sys/stat.h>
66 
67 #include <ctype.h>
68 #include <errno.h>
69 #include <signal.h>
70 #include <stdio.h>
71 #include <stdlib.h>
72 #include <string.h>
73 #include <unistd.h>
74 #include "zopen.h"
75 
76 #define	BITS		16		/* Default bits. */
77 #define	HSIZE		69001		/* 95% occupancy */
78 
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;
82 
83 typedef u_char char_type;
84 static char_type magic_header[] =
85 	{'\037', '\235'};		/* 1F 9D */
86 
87 #define	BIT_MASK	0x1f		/* Defines for third byte of header. */
88 #define	BLOCK_MASK	0x80
89 
90 /*
91  * Masks 0x40 and 0x20 are free.  I think 0x20 should mean that there is
92  * a fourth header byte (for expansion).
93  */
94 #define	INIT_BITS 9			/* Initial number of bits/code. */
95 
96 #define	MAXCODE(n_bits)	((1 << (n_bits)) - 1)
97 
98 struct s_zstate {
99 	FILE *zs_fp;			/* File stream for I/O */
100 	char zs_mode;			/* r or w */
101 	enum {
102 		S_START, S_MIDDLE, S_EOF
103 	} zs_state;			/* State of computation */
104 	int zs_n_bits;			/* Number of bits/code. */
105 	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. */
112 	/*
113 	 * Block compression parameters -- after all codes are used up,
114 	 * and compression rate changes, start over.
115 	 */
116 	int zs_block_compress;
117 	int zs_clear_flg;
118 	long zs_ratio;
119 	count_int zs_checkpoint;
120 	int zs_offset;
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];
125 	union {
126 		struct {
127 			long zs_fcode;
128 			code_int zs_ent;
129 			code_int zs_hsize_reg;
130 			int zs_hshift;
131 		} w;			/* Write paramenters */
132 		struct {
133 			char_type *zs_stackp;
134 			int zs_finchar;
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 */
139 	} u;
140 };
141 
142 /* Definitions to retain old variable names */
143 #define	fp		zs->zs_fp
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
175 
176 /*
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).
183  */
184 
185 #define	htabof(i)	htab[i]
186 #define	codetabof(i)	codetab[i]
187 
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))
191 
192 #define	CHECK_GAP 10000		/* Ratio check interval. */
193 
194 /*
195  * the next two codes should not be changed lightly, as they must not
196  * lie within the contiguous general code space.
197  */
198 #define	FIRST	257		/* First free entry. */
199 #define	CLEAR	256		/* Table clear output code. */
200 
201 static int	cl_block __P((struct s_zstate *));
202 static void	cl_hash __P((struct s_zstate *, count_int));
203 static code_int	getcode __P((struct s_zstate *));
204 static int	output __P((struct s_zstate *, code_int));
205 static int	zclose __P((void *));
206 static int	zread __P((void *, char *, int));
207 static int	zwrite __P((void *, const char *, int));
208 
209 /*-
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.
212  *
213  * Algorithm:
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.
218  */
219 
220 /*-
221  * compress write
222  *
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.
234  */
235 static int
236 zwrite(cookie, wbp, num)
237 	void *cookie;
238 	const char *wbp;
239 	int num;
240 {
241 	register code_int i;
242 	register int c, disp;
243 	struct s_zstate *zs;
244 	const u_char *bp;
245 	u_char tmp;
246 	int count;
247 
248 	if (num == 0)
249 		return (0);
250 
251 	zs = cookie;
252 	count = num;
253 	bp = (u_char *)wbp;
254 	if (state == S_MIDDLE)
255 		goto middle;
256 	state = S_MIDDLE;
257 
258 	maxmaxcode = 1L << maxbits;
259 	if (fwrite(magic_header,
260 	    sizeof(char), sizeof(magic_header), fp) != sizeof(magic_header))
261 		return (-1);
262 	tmp = (u_char)((maxbits) | block_compress);
263 	if (fwrite(&tmp, sizeof(char), sizeof(tmp), fp) != sizeof(tmp))
264 		return (-1);
265 
266 	offset = 0;
267 	bytes_out = 3;		/* Includes 3-byte header mojo. */
268 	out_count = 0;
269 	clear_flg = 0;
270 	ratio = 0;
271 	in_count = 1;
272 	checkpoint = CHECK_GAP;
273 	maxcode = MAXCODE(n_bits = INIT_BITS);
274 	free_ent = ((block_compress) ? FIRST : 256);
275 
276 	ent = *bp++;
277 	--count;
278 
279 	hshift = 0;
280 	for (fcode = (long)hsize; fcode < 65536L; fcode *= 2L)
281 		hshift++;
282 	hshift = 8 - hshift;	/* Set hash code range bound. */
283 
284 	hsize_reg = hsize;
285 	cl_hash(zs, (count_int)hsize_reg);	/* Clear hash table. */
286 
287 middle:	for (i = 0; count--;) {
288 		c = *bp++;
289 		in_count++;
290 		fcode = (long)(((long)c << maxbits) + ent);
291 		i = ((c << hshift) ^ ent);	/* Xor hashing. */
292 
293 		if (htabof(i) == fcode) {
294 			ent = codetabof(i);
295 			continue;
296 		} else if ((long)htabof(i) < 0)	/* Empty slot. */
297 			goto nomatch;
298 		disp = hsize_reg - i;	/* Secondary hash (after G. Knott). */
299 		if (i == 0)
300 			disp = 1;
301 probe:		if ((i -= disp) < 0)
302 			i += hsize_reg;
303 
304 		if (htabof(i) == fcode) {
305 			ent = codetabof(i);
306 			continue;
307 		}
308 		if ((long)htabof(i) >= 0)
309 			goto probe;
310 nomatch:	if (output(zs, (code_int) ent) == -1)
311 			return (-1);
312 		out_count++;
313 		ent = c;
314 		if (free_ent < maxmaxcode) {
315 			codetabof(i) = free_ent++;	/* code -> hashtable */
316 			htabof(i) = fcode;
317 		} else if ((count_int)in_count >=
318 		    checkpoint && block_compress) {
319 			if (cl_block(zs) == -1)
320 				return (-1);
321 		}
322 	}
323 	return (num);
324 }
325 
326 static int
327 zclose(cookie)
328 	void *cookie;
329 {
330 	struct s_zstate *zs;
331 	int rval;
332 
333 	zs = cookie;
334 	if (zmode == 'w') {		/* Put out the final code. */
335 		if (output(zs, (code_int) ent) == -1) {
336 			(void)fclose(fp);
337 			free(zs);
338 			return (-1);
339 		}
340 		out_count++;
341 		if (output(zs, (code_int) - 1) == -1) {
342 			(void)fclose(fp);
343 			free(zs);
344 			return (-1);
345 		}
346 	}
347 	rval = fclose(fp) == EOF ? -1 : 0;
348 	free(zs);
349 	return (rval);
350 }
351 
352 /*-
353  * Output the given code.
354  * Inputs:
355  * 	code:	A n_bits-bit integer.  If == -1, then EOF.  This assumes
356  *		that n_bits =< (long)wordsize - 1.
357  * Outputs:
358  * 	Outputs code to the file.
359  * Assumptions:
360  *	Chars are 8 bits long.
361  * Algorithm:
362  * 	Maintain a BITS character long buffer (so that 8 codes will
363  * fit in it exactly).  Use the VAX insv instruction to insert each
364  * code in turn.  When the buffer fills up empty it and start over.
365  */
366 
367 static char_type lmask[9] =
368 	{0xff, 0xfe, 0xfc, 0xf8, 0xf0, 0xe0, 0xc0, 0x80, 0x00};
369 static char_type rmask[9] =
370 	{0x00, 0x01, 0x03, 0x07, 0x0f, 0x1f, 0x3f, 0x7f, 0xff};
371 
372 static int
373 output(zs, ocode)
374 	struct s_zstate *zs;
375 	code_int ocode;
376 {
377 	register int bits, r_off;
378 	register char_type *bp;
379 
380 	r_off = offset;
381 	bits = n_bits;
382 	bp = buf;
383 	if (ocode >= 0) {
384 		/* Get to the first byte. */
385 		bp += (r_off >> 3);
386 		r_off &= 7;
387 		/*
388 		 * Since ocode is always >= 8 bits, only need to mask the first
389 		 * hunk on the left.
390 		 */
391 		*bp = (*bp & rmask[r_off]) | ((ocode << r_off) & lmask[r_off]);
392 		bp++;
393 		bits -= (8 - r_off);
394 		ocode >>= 8 - r_off;
395 		/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
396 		if (bits >= 8) {
397 			*bp++ = ocode;
398 			ocode >>= 8;
399 			bits -= 8;
400 		}
401 		/* Last bits. */
402 		if (bits)
403 			*bp = ocode;
404 		offset += n_bits;
405 		if (offset == (n_bits << 3)) {
406 			bp = buf;
407 			bits = n_bits;
408 			bytes_out += bits;
409 			if (fwrite(bp, sizeof(char), bits, fp) != bits)
410 				return (-1);
411 			bp += bits;
412 			bits = 0;
413 			offset = 0;
414 		}
415 		/*
416 		 * If the next entry is going to be too big for the ocode size,
417 		 * then increase it, if possible.
418 		 */
419 		if (free_ent > maxcode || (clear_flg > 0)) {
420 		       /*
421 			* Write the whole buffer, because the input side won't
422 			* discover the size increase until after it has read it.
423 			*/
424 			if (offset > 0) {
425 				if (fwrite(buf, 1, n_bits, fp) != n_bits)
426 					return (-1);
427 				bytes_out += n_bits;
428 			}
429 			offset = 0;
430 
431 			if (clear_flg) {
432 				maxcode = MAXCODE(n_bits = INIT_BITS);
433 				clear_flg = 0;
434 			} else {
435 				n_bits++;
436 				if (n_bits == maxbits)
437 					maxcode = maxmaxcode;
438 				else
439 					maxcode = MAXCODE(n_bits);
440 			}
441 		}
442 	} else {
443 		/* At EOF, write the rest of the buffer. */
444 		if (offset > 0) {
445 			offset = (offset + 7) / 8;
446 			if (fwrite(buf, 1, offset, fp) != offset)
447 				return (-1);
448 			bytes_out += offset;
449 		}
450 		offset = 0;
451 	}
452 	return (0);
453 }
454 
455 /*
456  * Decompress read.  This routine adapts to the codes in the file building
457  * the "string" table on-the-fly; requiring no table to be stored in the
458  * compressed file.  The tables used herein are shared with those of the
459  * compress() routine.  See the definitions above.
460  */
461 static int
462 zread(cookie, rbp, num)
463 	void *cookie;
464 	char *rbp;
465 	int num;
466 {
467 	register u_int count;
468 	struct s_zstate *zs;
469 	u_char *bp, header[3];
470 
471 	if (num == 0)
472 		return (0);
473 
474 	zs = cookie;
475 	count = num;
476 	bp = (u_char *)rbp;
477 	switch (state) {
478 	case S_START:
479 		state = S_MIDDLE;
480 		break;
481 	case S_MIDDLE:
482 		goto middle;
483 	case S_EOF:
484 		goto eof;
485 	}
486 
487 	/* Check the magic number */
488 	if (fread(header,
489 	    sizeof(char), sizeof(header), fp) != sizeof(header) ||
490 	    memcmp(header, magic_header, sizeof(magic_header)) != 0) {
491 		errno = EFTYPE;
492 		return (-1);
493 	}
494 	maxbits = header[2];	/* Set -b from file. */
495 	block_compress = maxbits & BLOCK_MASK;
496 	maxbits &= BIT_MASK;
497 	maxmaxcode = 1L << maxbits;
498 	if (maxbits > BITS) {
499 		errno = EFTYPE;
500 		return (-1);
501 	}
502 	/* As above, initialize the first 256 entries in the table. */
503 	maxcode = MAXCODE(n_bits = INIT_BITS);
504 	for (code = 255; code >= 0; code--) {
505 		tab_prefixof(code) = 0;
506 		tab_suffixof(code) = (char_type) code;
507 	}
508 	free_ent = block_compress ? FIRST : 256;
509 
510 	finchar = oldcode = getcode(zs);
511 	if (oldcode == -1)	/* EOF already? */
512 		return (0);	/* Get out of here */
513 
514 	/* First code must be 8 bits = char. */
515 	*bp++ = (u_char)finchar;
516 	count--;
517 	stackp = de_stack;
518 
519 	while ((code = getcode(zs)) > -1) {
520 
521 		if ((code == CLEAR) && block_compress) {
522 			for (code = 255; code >= 0; code--)
523 				tab_prefixof(code) = 0;
524 			clear_flg = 1;
525 			free_ent = FIRST - 1;
526 			if ((code = getcode(zs)) == -1)	/* O, untimely death! */
527 				break;
528 		}
529 		incode = code;
530 
531 		/* Special case for KwKwK string. */
532 		if (code >= free_ent) {
533 			*stackp++ = finchar;
534 			code = oldcode;
535 		}
536 
537 		/* Generate output characters in reverse order. */
538 		while (code >= 256) {
539 			*stackp++ = tab_suffixof(code);
540 			code = tab_prefixof(code);
541 		}
542 		*stackp++ = finchar = tab_suffixof(code);
543 
544 		/* And put them out in forward order.  */
545 middle:		do {
546 			if (count-- == 0)
547 				return (num);
548 			*bp++ = *--stackp;
549 		} while (stackp > de_stack);
550 
551 		/* Generate the new entry. */
552 		if ((code = free_ent) < maxmaxcode) {
553 			tab_prefixof(code) = (u_short) oldcode;
554 			tab_suffixof(code) = finchar;
555 			free_ent = code + 1;
556 		}
557 
558 		/* Remember previous code. */
559 		oldcode = incode;
560 	}
561 	state = S_EOF;
562 eof:	return (num - count);
563 }
564 
565 /*-
566  * Read one code from the standard input.  If EOF, return -1.
567  * Inputs:
568  * 	stdin
569  * Outputs:
570  * 	code or -1 is returned.
571  */
572 static code_int
573 getcode(zs)
574 	struct s_zstate *zs;
575 {
576 	register code_int gcode;
577 	register int r_off, bits;
578 	register char_type *bp;
579 
580 	bp = gbuf;
581 	if (clear_flg > 0 || roffset >= size || free_ent > maxcode) {
582 		/*
583 		 * If the next entry will be too big for the current gcode
584 		 * size, then we must increase the size.  This implies reading
585 		 * a new buffer full, too.
586 		 */
587 		if (free_ent > maxcode) {
588 			n_bits++;
589 			if (n_bits == maxbits)	/* Won't get any bigger now. */
590 				maxcode = maxmaxcode;
591 			else
592 				maxcode = MAXCODE(n_bits);
593 		}
594 		if (clear_flg > 0) {
595 			maxcode = MAXCODE(n_bits = INIT_BITS);
596 			clear_flg = 0;
597 		}
598 		size = fread(gbuf, 1, n_bits, fp);
599 		if (size <= 0)			/* End of file. */
600 			return (-1);
601 		roffset = 0;
602 		/* Round size down to integral number of codes. */
603 		size = (size << 3) - (n_bits - 1);
604 	}
605 	r_off = roffset;
606 	bits = n_bits;
607 
608 	/* Get to the first byte. */
609 	bp += (r_off >> 3);
610 	r_off &= 7;
611 
612 	/* Get first part (low order bits). */
613 	gcode = (*bp++ >> r_off);
614 	bits -= (8 - r_off);
615 	r_off = 8 - r_off;	/* Now, roffset into gcode word. */
616 
617 	/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
618 	if (bits >= 8) {
619 		gcode |= *bp++ << r_off;
620 		r_off += 8;
621 		bits -= 8;
622 	}
623 
624 	/* High order bits. */
625 	gcode |= (*bp & rmask[bits]) << r_off;
626 	roffset += n_bits;
627 
628 	return (gcode);
629 }
630 
631 static int
632 cl_block(zs)			/* Table clear for block compress. */
633 	struct s_zstate *zs;
634 {
635 	register long rat;
636 
637 	checkpoint = in_count + CHECK_GAP;
638 
639 	if (in_count > 0x007fffff) {	/* Shift will overflow. */
640 		rat = bytes_out >> 8;
641 		if (rat == 0)		/* Don't divide by zero. */
642 			rat = 0x7fffffff;
643 		else
644 			rat = in_count / rat;
645 	} else
646 		rat = (in_count << 8) / bytes_out;	/* 8 fractional bits. */
647 	if (rat > ratio)
648 		ratio = rat;
649 	else {
650 		ratio = 0;
651 		cl_hash(zs, (count_int) hsize);
652 		free_ent = FIRST;
653 		clear_flg = 1;
654 		if (output(zs, (code_int) CLEAR) == -1)
655 			return (-1);
656 	}
657 	return (0);
658 }
659 
660 static void
661 cl_hash(zs, cl_hsize)			/* Reset code table. */
662 	struct s_zstate *zs;
663 	register count_int cl_hsize;
664 {
665 	register count_int *htab_p;
666 	register long i, m1;
667 
668 	m1 = -1;
669 	htab_p = htab + cl_hsize;
670 	i = cl_hsize - 16;
671 	do {			/* Might use Sys V memset(3) here. */
672 		*(htab_p - 16) = m1;
673 		*(htab_p - 15) = m1;
674 		*(htab_p - 14) = m1;
675 		*(htab_p - 13) = m1;
676 		*(htab_p - 12) = m1;
677 		*(htab_p - 11) = m1;
678 		*(htab_p - 10) = m1;
679 		*(htab_p - 9) = m1;
680 		*(htab_p - 8) = m1;
681 		*(htab_p - 7) = m1;
682 		*(htab_p - 6) = m1;
683 		*(htab_p - 5) = m1;
684 		*(htab_p - 4) = m1;
685 		*(htab_p - 3) = m1;
686 		*(htab_p - 2) = m1;
687 		*(htab_p - 1) = m1;
688 		htab_p -= 16;
689 	} while ((i -= 16) >= 0);
690 	for (i += 16; i > 0; i--)
691 		*--htab_p = m1;
692 }
693 
694 FILE *
695 zopen(fname, mode, bits)
696 	const char *fname, *mode;
697 	int bits;
698 {
699 	struct s_zstate *zs;
700 
701 	if ((mode[0] != 'r' && mode[0] != 'w') || mode[1] != '\0' ||
702 	    bits < 0 || bits > BITS) {
703 		errno = EINVAL;
704 		return (NULL);
705 	}
706 
707 	if ((zs = calloc(1, sizeof(struct s_zstate))) == NULL)
708 		return (NULL);
709 
710 	maxbits = bits ? bits : BITS;	/* User settable max # bits/code. */
711 	maxmaxcode = 1L << maxbits;	/* Should NEVER generate this code. */
712 	hsize = HSIZE;			/* For dynamic table sizing. */
713 	free_ent = 0;			/* First unused entry. */
714 	block_compress = BLOCK_MASK;
715 	clear_flg = 0;
716 	ratio = 0;
717 	checkpoint = CHECK_GAP;
718 	in_count = 1;			/* Length of input. */
719 	out_count = 0;			/* # of codes output (for debugging). */
720 	state = S_START;
721 	roffset = 0;
722 	size = 0;
723 
724 	/*
725 	 * Layering compress on top of stdio in order to provide buffering,
726 	 * and ensure that reads and write work with the data specified.
727 	 */
728 	if ((fp = fopen(fname, mode)) == NULL) {
729 		free(zs);
730 		return (NULL);
731 	}
732 	switch (*mode) {
733 	case 'r':
734 		zmode = 'r';
735 		return (funopen(zs, zread, NULL, NULL, zclose));
736 	case 'w':
737 		zmode = 'w';
738 		return (funopen(zs, NULL, zwrite, NULL, zclose));
739 	}
740 	/* NOTREACHED */
741 	return (NULL);
742 }
743