xref: /netbsd-src/usr.bin/compress/zopen.c (revision 5e4c038a45edbc7d63b7c2daa76e29f88b64a4e3)
1 /*	$NetBSD: zopen.c,v 1.7 2002/05/26 22:25:38 wiz Exp $	*/
2 
3 /*-
4  * Copyright (c) 1985, 1986, 1992, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Diomidis Spinellis and James A. Woods, derived from original
9  * work by Spencer Thomas and Joseph Orost.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the University of
22  *	California, Berkeley and its contributors.
23  * 4. Neither the name of the University nor the names of its contributors
24  *    may be used to endorse or promote products derived from this software
25  *    without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
29  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
30  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
31  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
32  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
33  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
34  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
35  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
36  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37  * SUCH DAMAGE.
38  */
39 
40 #if defined(LIBC_SCCS) && !defined(lint)
41 #if 0
42 static char sccsid[] = "@(#)zopen.c	8.1 (Berkeley) 6/27/93";
43 #else
44 static char rcsid[] = "$NetBSD: zopen.c,v 1.7 2002/05/26 22:25:38 wiz Exp $";
45 #endif
46 #endif /* LIBC_SCCS and not lint */
47 
48 /*-
49  * fcompress.c - File compression ala IEEE Computer, June 1984.
50  *
51  * Compress authors:
52  *		Spencer W. Thomas	(decvax!utah-cs!thomas)
53  *		Jim McKie		(decvax!mcvax!jim)
54  *		Steve Davies		(decvax!vax135!petsd!peora!srd)
55  *		Ken Turkowski		(decvax!decwrl!turtlevax!ken)
56  *		James A. Woods		(decvax!ihnp4!ames!jaw)
57  *		Joe Orost		(decvax!vax135!petsd!joe)
58  *
59  * Cleaned up and converted to library returning I/O streams by
60  * Diomidis Spinellis <dds@doc.ic.ac.uk>.
61  *
62  * zopen(filename, mode, bits)
63  *	Returns a FILE * that can be used for read or write.  The modes
64  *	supported are only "r" and "w".  Seeking is not allowed.  On
65  *	reading the file is decompressed, on writing it is compressed.
66  *	The output is compatible with compress(1) with 16 bit tables.
67  *	Any file produced by compress(1) can be read.
68  */
69 
70 #include <sys/param.h>
71 #include <sys/stat.h>
72 
73 #include <ctype.h>
74 #include <errno.h>
75 #include <signal.h>
76 #include <stdio.h>
77 #include <stdlib.h>
78 #include <string.h>
79 #include <unistd.h>
80 
81 #define	BITS		16		/* Default bits. */
82 #define	HSIZE		69001		/* 95% occupancy */
83 
84 /* A code_int must be able to hold 2**BITS values of type int, and also -1. */
85 typedef long code_int;
86 typedef long count_int;
87 
88 typedef u_char char_type;
89 static char_type magic_header[] =
90 	{'\037', '\235'};		/* 1F 9D */
91 
92 #define	BIT_MASK	0x1f		/* Defines for third byte of header. */
93 #define	BLOCK_MASK	0x80
94 
95 /*
96  * Masks 0x40 and 0x20 are free.  I think 0x20 should mean that there is
97  * a fourth header byte (for expansion).
98  */
99 #define	INIT_BITS 9			/* Initial number of bits/code. */
100 
101 #define	MAXCODE(n_bits)	((1 << (n_bits)) - 1)
102 
103 struct s_zstate {
104 	FILE *zs_fp;			/* File stream for I/O */
105 	char zs_mode;			/* r or w */
106 	enum {
107 		S_START, S_MIDDLE, S_EOF
108 	} zs_state;			/* State of computation */
109 	int zs_n_bits;			/* Number of bits/code. */
110 	int zs_maxbits;			/* User settable max # bits/code. */
111 	code_int zs_maxcode;		/* Maximum code, given n_bits. */
112 	code_int zs_maxmaxcode;		/* Should NEVER generate this code. */
113 	count_int zs_htab [HSIZE];
114 	u_short zs_codetab [HSIZE];
115 	code_int zs_hsize;		/* For dynamic table sizing. */
116 	code_int zs_free_ent;		/* First unused entry. */
117 	/*
118 	 * Block compression parameters -- after all codes are used up,
119 	 * and compression rate changes, start over.
120 	 */
121 	int zs_block_compress;
122 	int zs_clear_flg;
123 	long zs_ratio;
124 	count_int zs_checkpoint;
125 	int zs_offset;
126 	long zs_in_count;		/* Length of input. */
127 	long zs_bytes_out;		/* Length of compressed output. */
128 	long zs_out_count;		/* # of codes output (for debugging). */
129 	char_type zs_buf[BITS];
130 	union {
131 		struct {
132 			long zs_fcode;
133 			code_int zs_ent;
134 			code_int zs_hsize_reg;
135 			int zs_hshift;
136 		} w;			/* Write paramenters */
137 		struct {
138 			char_type *zs_stackp;
139 			int zs_finchar;
140 			code_int zs_code, zs_oldcode, zs_incode;
141 			int zs_roffset, zs_size;
142 			char_type zs_gbuf[BITS];
143 		} r;			/* Read parameters */
144 	} u;
145 };
146 
147 /* Definitions to retain old variable names */
148 #define	fp		zs->zs_fp
149 #define	zmode		zs->zs_mode
150 #define	state		zs->zs_state
151 #define	n_bits		zs->zs_n_bits
152 #define	maxbits		zs->zs_maxbits
153 #define	maxcode		zs->zs_maxcode
154 #define	maxmaxcode	zs->zs_maxmaxcode
155 #define	htab		zs->zs_htab
156 #define	codetab		zs->zs_codetab
157 #define	hsize		zs->zs_hsize
158 #define	free_ent	zs->zs_free_ent
159 #define	block_compress	zs->zs_block_compress
160 #define	clear_flg	zs->zs_clear_flg
161 #define	ratio		zs->zs_ratio
162 #define	checkpoint	zs->zs_checkpoint
163 #define	offset		zs->zs_offset
164 #define	in_count	zs->zs_in_count
165 #define	bytes_out	zs->zs_bytes_out
166 #define	out_count	zs->zs_out_count
167 #define	buf		zs->zs_buf
168 #define	fcode		zs->u.w.zs_fcode
169 #define	hsize_reg	zs->u.w.zs_hsize_reg
170 #define	ent		zs->u.w.zs_ent
171 #define	hshift		zs->u.w.zs_hshift
172 #define	stackp		zs->u.r.zs_stackp
173 #define	finchar		zs->u.r.zs_finchar
174 #define	code		zs->u.r.zs_code
175 #define	oldcode		zs->u.r.zs_oldcode
176 #define	incode		zs->u.r.zs_incode
177 #define	roffset		zs->u.r.zs_roffset
178 #define	size		zs->u.r.zs_size
179 #define	gbuf		zs->u.r.zs_gbuf
180 
181 /*
182  * To save much memory, we overlay the table used by compress() with those
183  * used by decompress().  The tab_prefix table is the same size and type as
184  * the codetab.  The tab_suffix table needs 2**BITS characters.  We get this
185  * from the beginning of htab.  The output stack uses the rest of htab, and
186  * contains characters.  There is plenty of room for any possible stack
187  * (stack used to be 8000 characters).
188  */
189 
190 #define	htabof(i)	htab[i]
191 #define	codetabof(i)	codetab[i]
192 
193 #define	tab_prefixof(i)	codetabof(i)
194 #define	tab_suffixof(i)	((char_type *)(htab))[i]
195 #define	de_stack	((char_type *)&tab_suffixof(1 << BITS))
196 
197 #define	CHECK_GAP 10000		/* Ratio check interval. */
198 
199 /*
200  * the next two codes should not be changed lightly, as they must not
201  * lie within the contiguous general code space.
202  */
203 #define	FIRST	257		/* First free entry. */
204 #define	CLEAR	256		/* Table clear output code. */
205 
206 static int	cl_block(struct s_zstate *);
207 static void	cl_hash(struct s_zstate *, count_int);
208 static code_int	getcode(struct s_zstate *);
209 static int	output(struct s_zstate *, code_int);
210 static int	zclose(void *);
211 FILE	       *zopen(const char *, const char *, int);
212 static int	zread(void *, char *, int);
213 static int	zwrite(void *, const char *, int);
214 
215 /*-
216  * Algorithm from "A Technique for High Performance Data Compression",
217  * Terry A. Welch, IEEE Computer Vol 17, No 6 (June 1984), pp 8-19.
218  *
219  * Algorithm:
220  * 	Modified Lempel-Ziv method (LZW).  Basically finds common
221  * substrings and replaces them with a variable size code.  This is
222  * deterministic, and can be done on the fly.  Thus, the decompression
223  * procedure needs no input table, but tracks the way the table was built.
224  */
225 
226 /*-
227  * compress write
228  *
229  * Algorithm:  use open addressing double hashing (no chaining) on the
230  * prefix code / next character combination.  We do a variant of Knuth's
231  * algorithm D (vol. 3, sec. 6.4) along with G. Knott's relatively-prime
232  * secondary probe.  Here, the modular division first probe is gives way
233  * to a faster exclusive-or manipulation.  Also do block compression with
234  * an adaptive reset, whereby the code table is cleared when the compression
235  * ratio decreases, but after the table fills.  The variable-length output
236  * codes are re-sized at this point, and a special CLEAR code is generated
237  * for the decompressor.  Late addition:  construct the table according to
238  * file size for noticeable speed improvement on small files.  Please direct
239  * questions about this implementation to ames!jaw.
240  */
241 static int
242 zwrite(void *cookie, const char *wbp, int num)
243 {
244 	code_int i;
245 	int c, disp;
246 	struct s_zstate *zs;
247 	const u_char *bp;
248 	u_char tmp;
249 	int count;
250 
251 	if (num == 0)
252 		return (0);
253 
254 	zs = cookie;
255 	count = num;
256 	bp = (u_char *)wbp;
257 	if (state == S_MIDDLE)
258 		goto middle;
259 	state = S_MIDDLE;
260 
261 	maxmaxcode = 1L << maxbits;
262 	if (fwrite(magic_header,
263 	    sizeof(char), sizeof(magic_header), fp) != sizeof(magic_header))
264 		return (-1);
265 	tmp = (u_char)(maxbits | block_compress);
266 	if (fwrite(&tmp, sizeof(char), sizeof(tmp), fp) != sizeof(tmp))
267 		return (-1);
268 
269 	offset = 0;
270 	bytes_out = 3;		/* Includes 3-byte header mojo. */
271 	out_count = 0;
272 	clear_flg = 0;
273 	ratio = 0;
274 	in_count = 1;
275 	checkpoint = CHECK_GAP;
276 	maxcode = MAXCODE(n_bits = INIT_BITS);
277 	free_ent = ((block_compress) ? FIRST : 256);
278 
279 	ent = *bp++;
280 	--count;
281 
282 	hshift = 0;
283 	for (fcode = (long)hsize; fcode < 65536L; fcode *= 2L)
284 		hshift++;
285 	hshift = 8 - hshift;	/* Set hash code range bound. */
286 
287 	hsize_reg = hsize;
288 	cl_hash(zs, (count_int)hsize_reg);	/* Clear hash table. */
289 
290 middle:	for (i = 0; count--;) {
291 		c = *bp++;
292 		in_count++;
293 		fcode = (long)(((long)c << maxbits) + ent);
294 		i = ((c << hshift) ^ ent);	/* Xor hashing. */
295 
296 		if (htabof(i) == fcode) {
297 			ent = codetabof(i);
298 			continue;
299 		} else if ((long)htabof(i) < 0)	/* Empty slot. */
300 			goto nomatch;
301 		disp = hsize_reg - i;	/* Secondary hash (after G. Knott). */
302 		if (i == 0)
303 			disp = 1;
304 probe:		if ((i -= disp) < 0)
305 			i += hsize_reg;
306 
307 		if (htabof(i) == fcode) {
308 			ent = codetabof(i);
309 			continue;
310 		}
311 		if ((long)htabof(i) >= 0)
312 			goto probe;
313 nomatch:	if (output(zs, (code_int) ent) == -1)
314 			return (-1);
315 		out_count++;
316 		ent = c;
317 		if (free_ent < maxmaxcode) {
318 			codetabof(i) = free_ent++;	/* code -> hashtable */
319 			htabof(i) = fcode;
320 		} else if ((count_int)in_count >=
321 		    checkpoint && block_compress) {
322 			if (cl_block(zs) == -1)
323 				return (-1);
324 		}
325 	}
326 	return (num);
327 }
328 
329 static int
330 zclose(void *cookie)
331 {
332 	struct s_zstate *zs;
333 	int rval;
334 
335 	zs = cookie;
336 	if (zmode == 'w') {		/* Put out the final code. */
337 		if (output(zs, (code_int) ent) == -1) {
338 			(void)fclose(fp);
339 			free(zs);
340 			return (-1);
341 		}
342 		out_count++;
343 		if (output(zs, (code_int) - 1) == -1) {
344 			(void)fclose(fp);
345 			free(zs);
346 			return (-1);
347 		}
348 	}
349 	rval = fclose(fp) == EOF ? -1 : 0;
350 	free(zs);
351 	return (rval);
352 }
353 
354 /*-
355  * Output the given code.
356  * Inputs:
357  * 	code:	A n_bits-bit integer.  If == -1, then EOF.  This assumes
358  *		that n_bits =< (long)wordsize - 1.
359  * Outputs:
360  * 	Outputs code to the file.
361  * Assumptions:
362  *	Chars are 8 bits long.
363  * Algorithm:
364  * 	Maintain a BITS character long buffer (so that 8 codes will
365  * fit in it exactly).  Use the VAX insv instruction to insert each
366  * code in turn.  When the buffer fills up empty it and start over.
367  */
368 
369 static char_type lmask[9] =
370 	{0xff, 0xfe, 0xfc, 0xf8, 0xf0, 0xe0, 0xc0, 0x80, 0x00};
371 static char_type rmask[9] =
372 	{0x00, 0x01, 0x03, 0x07, 0x0f, 0x1f, 0x3f, 0x7f, 0xff};
373 
374 static int
375 output(struct s_zstate *zs, code_int ocode)
376 {
377 	int bits, r_off;
378 	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(void *cookie, char *rbp, int num)
463 {
464 	u_int count;
465 	struct s_zstate *zs;
466 	u_char *bp, header[3];
467 
468 	if (num == 0)
469 		return (0);
470 
471 	zs = cookie;
472 	count = num;
473 	bp = (u_char *)rbp;
474 	switch (state) {
475 	case S_START:
476 		state = S_MIDDLE;
477 		break;
478 	case S_MIDDLE:
479 		goto middle;
480 	case S_EOF:
481 		goto eof;
482 	}
483 
484 	/* Check the magic number */
485 	if (fread(header,
486 	    sizeof(char), sizeof(header), fp) != sizeof(header) ||
487 	    memcmp(header, magic_header, sizeof(magic_header)) != 0) {
488 		errno = EFTYPE;
489 		return (-1);
490 	}
491 	maxbits = header[2];	/* Set -b from file. */
492 	block_compress = maxbits & BLOCK_MASK;
493 	maxbits &= BIT_MASK;
494 	maxmaxcode = 1L << maxbits;
495 	if (maxbits > BITS) {
496 		errno = EFTYPE;
497 		return (-1);
498 	}
499 	/* As above, initialize the first 256 entries in the table. */
500 	maxcode = MAXCODE(n_bits = INIT_BITS);
501 	for (code = 255; code >= 0; code--) {
502 		tab_prefixof(code) = 0;
503 		tab_suffixof(code) = (char_type) code;
504 	}
505 	free_ent = block_compress ? FIRST : 256;
506 
507 	finchar = oldcode = getcode(zs);
508 	if (oldcode == -1)	/* EOF already? */
509 		return (0);	/* Get out of here */
510 
511 	/* First code must be 8 bits = char. */
512 	*bp++ = (u_char)finchar;
513 	count--;
514 	stackp = de_stack;
515 
516 	while ((code = getcode(zs)) > -1) {
517 
518 		if ((code == CLEAR) && block_compress) {
519 			for (code = 255; code >= 0; code--)
520 				tab_prefixof(code) = 0;
521 			clear_flg = 1;
522 			free_ent = FIRST - 1;
523 			if ((code = getcode(zs)) == -1)	/* O, untimely death! */
524 				break;
525 		}
526 		incode = code;
527 
528 		/* Special case for KwKwK string. */
529 		if (code >= free_ent) {
530 			*stackp++ = finchar;
531 			code = oldcode;
532 		}
533 
534 		/* Generate output characters in reverse order. */
535 		while (code >= 256) {
536 			*stackp++ = tab_suffixof(code);
537 			code = tab_prefixof(code);
538 		}
539 		*stackp++ = finchar = tab_suffixof(code);
540 
541 		/* And put them out in forward order.  */
542 middle:		do {
543 			if (count-- == 0)
544 				return (num);
545 			*bp++ = *--stackp;
546 		} while (stackp > de_stack);
547 
548 		/* Generate the new entry. */
549 		if ((code = free_ent) < maxmaxcode) {
550 			tab_prefixof(code) = (u_short) oldcode;
551 			tab_suffixof(code) = finchar;
552 			free_ent = code + 1;
553 		}
554 
555 		/* Remember previous code. */
556 		oldcode = incode;
557 	}
558 	state = S_EOF;
559 eof:	return (num - count);
560 }
561 
562 /*-
563  * Read one code from the standard input.  If EOF, return -1.
564  * Inputs:
565  * 	stdin
566  * Outputs:
567  * 	code or -1 is returned.
568  */
569 static code_int
570 getcode(struct s_zstate *zs)
571 {
572 	code_int gcode;
573 	int r_off, bits;
574 	char_type *bp;
575 
576 	bp = gbuf;
577 	if (clear_flg > 0 || roffset >= size || free_ent > maxcode) {
578 		/*
579 		 * If the next entry will be too big for the current gcode
580 		 * size, then we must increase the size.  This implies reading
581 		 * a new buffer full, too.
582 		 */
583 		if (free_ent > maxcode) {
584 			n_bits++;
585 			if (n_bits == maxbits)	/* Won't get any bigger now. */
586 				maxcode = maxmaxcode;
587 			else
588 				maxcode = MAXCODE(n_bits);
589 		}
590 		if (clear_flg > 0) {
591 			maxcode = MAXCODE(n_bits = INIT_BITS);
592 			clear_flg = 0;
593 		}
594 		size = fread(gbuf, 1, n_bits, fp);
595 		if (size <= 0)			/* End of file. */
596 			return (-1);
597 		roffset = 0;
598 		/* Round size down to integral number of codes. */
599 		size = (size << 3) - (n_bits - 1);
600 	}
601 	r_off = roffset;
602 	bits = n_bits;
603 
604 	/* Get to the first byte. */
605 	bp += (r_off >> 3);
606 	r_off &= 7;
607 
608 	/* Get first part (low order bits). */
609 	gcode = (*bp++ >> r_off);
610 	bits -= (8 - r_off);
611 	r_off = 8 - r_off;	/* Now, roffset into gcode word. */
612 
613 	/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
614 	if (bits >= 8) {
615 		gcode |= *bp++ << r_off;
616 		r_off += 8;
617 		bits -= 8;
618 	}
619 
620 	/* High order bits. */
621 	gcode |= (*bp & rmask[bits]) << r_off;
622 	roffset += n_bits;
623 
624 	return (gcode);
625 }
626 
627 static int
628 cl_block(struct s_zstate *zs)		/* Table clear for block compress. */
629 {
630 	long rat;
631 
632 	checkpoint = in_count + CHECK_GAP;
633 
634 	if (in_count > 0x007fffff) {	/* Shift will overflow. */
635 		rat = bytes_out >> 8;
636 		if (rat == 0)		/* Don't divide by zero. */
637 			rat = 0x7fffffff;
638 		else
639 			rat = in_count / rat;
640 	} else
641 		rat = (in_count << 8) / bytes_out;	/* 8 fractional bits. */
642 	if (rat > ratio)
643 		ratio = rat;
644 	else {
645 		ratio = 0;
646 		cl_hash(zs, (count_int) hsize);
647 		free_ent = FIRST;
648 		clear_flg = 1;
649 		if (output(zs, (code_int) CLEAR) == -1)
650 			return (-1);
651 	}
652 	return (0);
653 }
654 
655 static void
656 cl_hash(struct s_zstate *zs, count_int cl_hsize)	/* Reset code table. */
657 {
658 	count_int *htab_p;
659 	long i, m1;
660 
661 	m1 = -1;
662 	htab_p = htab + cl_hsize;
663 	i = cl_hsize - 16;
664 	do {			/* Might use Sys V memset(3) here. */
665 		*(htab_p - 16) = m1;
666 		*(htab_p - 15) = m1;
667 		*(htab_p - 14) = m1;
668 		*(htab_p - 13) = m1;
669 		*(htab_p - 12) = m1;
670 		*(htab_p - 11) = m1;
671 		*(htab_p - 10) = m1;
672 		*(htab_p - 9) = m1;
673 		*(htab_p - 8) = m1;
674 		*(htab_p - 7) = m1;
675 		*(htab_p - 6) = m1;
676 		*(htab_p - 5) = m1;
677 		*(htab_p - 4) = m1;
678 		*(htab_p - 3) = m1;
679 		*(htab_p - 2) = m1;
680 		*(htab_p - 1) = m1;
681 		htab_p -= 16;
682 	} while ((i -= 16) >= 0);
683 	for (i += 16; i > 0; i--)
684 		*--htab_p = m1;
685 }
686 
687 FILE *
688 zopen(const char *fname, const char *mode, int bits)
689 {
690 	struct s_zstate *zs;
691 
692 	if ((mode[0] != 'r' && mode[0] != 'w') || mode[1] != '\0' ||
693 	    bits < 0 || bits > BITS) {
694 		errno = EINVAL;
695 		return (NULL);
696 	}
697 
698 	if ((zs = calloc(1, sizeof(struct s_zstate))) == NULL)
699 		return (NULL);
700 
701 	maxbits = bits ? bits : BITS;	/* User settable max # bits/code. */
702 	maxmaxcode = 1 << maxbits;	/* Should NEVER generate this code. */
703 	hsize = HSIZE;			/* For dynamic table sizing. */
704 	free_ent = 0;			/* First unused entry. */
705 	block_compress = BLOCK_MASK;
706 	clear_flg = 0;
707 	ratio = 0;
708 	checkpoint = CHECK_GAP;
709 	in_count = 1;			/* Length of input. */
710 	out_count = 0;			/* # of codes output (for debugging). */
711 	state = S_START;
712 	roffset = 0;
713 	size = 0;
714 
715 	/*
716 	 * Layering compress on top of stdio in order to provide buffering,
717 	 * and ensure that reads and write work with the data specified.
718 	 */
719 	if ((fp = fopen(fname, mode)) == NULL) {
720 		free(zs);
721 		return (NULL);
722 	}
723 	switch (*mode) {
724 	case 'r':
725 		zmode = 'r';
726 		return (funopen(zs, zread, NULL, NULL, zclose));
727 	case 'w':
728 		zmode = 'w';
729 		return (funopen(zs, NULL, zwrite, NULL, zclose));
730 	}
731 	/* NOTREACHED */
732 	return (NULL);
733 }
734