xref: /netbsd-src/usr.bin/compress/zopen.c (revision 2a399c6883d870daece976daec6ffa7bb7f934ce)
1 /*	$NetBSD: zopen.c,v 1.6 1997/09/15 10:58:39 lukem 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 1.6 1997/09/15 10:58:39 lukem 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 __P((struct s_zstate *));
207 static void	cl_hash __P((struct s_zstate *, count_int));
208 static code_int	getcode __P((struct s_zstate *));
209 static int	output __P((struct s_zstate *, code_int));
210 static int	zclose __P((void *));
211 FILE	       *zopen __P((const char *, const char *, int));
212 static int	zread __P((void *, char *, int));
213 static int	zwrite __P((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(cookie, wbp, num)
243 	void *cookie;
244 	const char *wbp;
245 	int num;
246 {
247 	code_int i;
248 	int c, disp;
249 	struct s_zstate *zs;
250 	const u_char *bp;
251 	u_char tmp;
252 	int count;
253 
254 	if (num == 0)
255 		return (0);
256 
257 	zs = cookie;
258 	count = num;
259 	bp = (u_char *)wbp;
260 	if (state == S_MIDDLE)
261 		goto middle;
262 	state = S_MIDDLE;
263 
264 	maxmaxcode = 1L << maxbits;
265 	if (fwrite(magic_header,
266 	    sizeof(char), sizeof(magic_header), fp) != sizeof(magic_header))
267 		return (-1);
268 	tmp = (u_char)(maxbits | block_compress);
269 	if (fwrite(&tmp, sizeof(char), sizeof(tmp), fp) != sizeof(tmp))
270 		return (-1);
271 
272 	offset = 0;
273 	bytes_out = 3;		/* Includes 3-byte header mojo. */
274 	out_count = 0;
275 	clear_flg = 0;
276 	ratio = 0;
277 	in_count = 1;
278 	checkpoint = CHECK_GAP;
279 	maxcode = MAXCODE(n_bits = INIT_BITS);
280 	free_ent = ((block_compress) ? FIRST : 256);
281 
282 	ent = *bp++;
283 	--count;
284 
285 	hshift = 0;
286 	for (fcode = (long)hsize; fcode < 65536L; fcode *= 2L)
287 		hshift++;
288 	hshift = 8 - hshift;	/* Set hash code range bound. */
289 
290 	hsize_reg = hsize;
291 	cl_hash(zs, (count_int)hsize_reg);	/* Clear hash table. */
292 
293 middle:	for (i = 0; count--;) {
294 		c = *bp++;
295 		in_count++;
296 		fcode = (long)(((long)c << maxbits) + ent);
297 		i = ((c << hshift) ^ ent);	/* Xor hashing. */
298 
299 		if (htabof(i) == fcode) {
300 			ent = codetabof(i);
301 			continue;
302 		} else if ((long)htabof(i) < 0)	/* Empty slot. */
303 			goto nomatch;
304 		disp = hsize_reg - i;	/* Secondary hash (after G. Knott). */
305 		if (i == 0)
306 			disp = 1;
307 probe:		if ((i -= disp) < 0)
308 			i += hsize_reg;
309 
310 		if (htabof(i) == fcode) {
311 			ent = codetabof(i);
312 			continue;
313 		}
314 		if ((long)htabof(i) >= 0)
315 			goto probe;
316 nomatch:	if (output(zs, (code_int) ent) == -1)
317 			return (-1);
318 		out_count++;
319 		ent = c;
320 		if (free_ent < maxmaxcode) {
321 			codetabof(i) = free_ent++;	/* code -> hashtable */
322 			htabof(i) = fcode;
323 		} else if ((count_int)in_count >=
324 		    checkpoint && block_compress) {
325 			if (cl_block(zs) == -1)
326 				return (-1);
327 		}
328 	}
329 	return (num);
330 }
331 
332 static int
333 zclose(cookie)
334 	void *cookie;
335 {
336 	struct s_zstate *zs;
337 	int rval;
338 
339 	zs = cookie;
340 	if (zmode == 'w') {		/* Put out the final code. */
341 		if (output(zs, (code_int) ent) == -1) {
342 			(void)fclose(fp);
343 			free(zs);
344 			return (-1);
345 		}
346 		out_count++;
347 		if (output(zs, (code_int) - 1) == -1) {
348 			(void)fclose(fp);
349 			free(zs);
350 			return (-1);
351 		}
352 	}
353 	rval = fclose(fp) == EOF ? -1 : 0;
354 	free(zs);
355 	return (rval);
356 }
357 
358 /*-
359  * Output the given code.
360  * Inputs:
361  * 	code:	A n_bits-bit integer.  If == -1, then EOF.  This assumes
362  *		that n_bits =< (long)wordsize - 1.
363  * Outputs:
364  * 	Outputs code to the file.
365  * Assumptions:
366  *	Chars are 8 bits long.
367  * Algorithm:
368  * 	Maintain a BITS character long buffer (so that 8 codes will
369  * fit in it exactly).  Use the VAX insv instruction to insert each
370  * code in turn.  When the buffer fills up empty it and start over.
371  */
372 
373 static char_type lmask[9] =
374 	{0xff, 0xfe, 0xfc, 0xf8, 0xf0, 0xe0, 0xc0, 0x80, 0x00};
375 static char_type rmask[9] =
376 	{0x00, 0x01, 0x03, 0x07, 0x0f, 0x1f, 0x3f, 0x7f, 0xff};
377 
378 static int
379 output(zs, ocode)
380 	struct s_zstate *zs;
381 	code_int ocode;
382 {
383 	int bits, r_off;
384 	char_type *bp;
385 
386 	r_off = offset;
387 	bits = n_bits;
388 	bp = buf;
389 	if (ocode >= 0) {
390 		/* Get to the first byte. */
391 		bp += (r_off >> 3);
392 		r_off &= 7;
393 		/*
394 		 * Since ocode is always >= 8 bits, only need to mask the first
395 		 * hunk on the left.
396 		 */
397 		*bp = (*bp & rmask[r_off]) | ((ocode << r_off) & lmask[r_off]);
398 		bp++;
399 		bits -= (8 - r_off);
400 		ocode >>= 8 - r_off;
401 		/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
402 		if (bits >= 8) {
403 			*bp++ = ocode;
404 			ocode >>= 8;
405 			bits -= 8;
406 		}
407 		/* Last bits. */
408 		if (bits)
409 			*bp = ocode;
410 		offset += n_bits;
411 		if (offset == (n_bits << 3)) {
412 			bp = buf;
413 			bits = n_bits;
414 			bytes_out += bits;
415 			if (fwrite(bp, sizeof(char), bits, fp) != bits)
416 				return (-1);
417 			bp += bits;
418 			bits = 0;
419 			offset = 0;
420 		}
421 		/*
422 		 * If the next entry is going to be too big for the ocode size,
423 		 * then increase it, if possible.
424 		 */
425 		if (free_ent > maxcode || (clear_flg > 0)) {
426 		       /*
427 			* Write the whole buffer, because the input side won't
428 			* discover the size increase until after it has read it.
429 			*/
430 			if (offset > 0) {
431 				if (fwrite(buf, 1, n_bits, fp) != n_bits)
432 					return (-1);
433 				bytes_out += n_bits;
434 			}
435 			offset = 0;
436 
437 			if (clear_flg) {
438 				maxcode = MAXCODE(n_bits = INIT_BITS);
439 				clear_flg = 0;
440 			} else {
441 				n_bits++;
442 				if (n_bits == maxbits)
443 					maxcode = maxmaxcode;
444 				else
445 					maxcode = MAXCODE(n_bits);
446 			}
447 		}
448 	} else {
449 		/* At EOF, write the rest of the buffer. */
450 		if (offset > 0) {
451 			offset = (offset + 7) / 8;
452 			if (fwrite(buf, 1, offset, fp) != offset)
453 				return (-1);
454 			bytes_out += offset;
455 		}
456 		offset = 0;
457 	}
458 	return (0);
459 }
460 
461 /*
462  * Decompress read.  This routine adapts to the codes in the file building
463  * the "string" table on-the-fly; requiring no table to be stored in the
464  * compressed file.  The tables used herein are shared with those of the
465  * compress() routine.  See the definitions above.
466  */
467 static int
468 zread(cookie, rbp, num)
469 	void *cookie;
470 	char *rbp;
471 	int num;
472 {
473 	u_int count;
474 	struct s_zstate *zs;
475 	u_char *bp, header[3];
476 
477 	if (num == 0)
478 		return (0);
479 
480 	zs = cookie;
481 	count = num;
482 	bp = (u_char *)rbp;
483 	switch (state) {
484 	case S_START:
485 		state = S_MIDDLE;
486 		break;
487 	case S_MIDDLE:
488 		goto middle;
489 	case S_EOF:
490 		goto eof;
491 	}
492 
493 	/* Check the magic number */
494 	if (fread(header,
495 	    sizeof(char), sizeof(header), fp) != sizeof(header) ||
496 	    memcmp(header, magic_header, sizeof(magic_header)) != 0) {
497 		errno = EFTYPE;
498 		return (-1);
499 	}
500 	maxbits = header[2];	/* Set -b from file. */
501 	block_compress = maxbits & BLOCK_MASK;
502 	maxbits &= BIT_MASK;
503 	maxmaxcode = 1L << maxbits;
504 	if (maxbits > BITS) {
505 		errno = EFTYPE;
506 		return (-1);
507 	}
508 	/* As above, initialize the first 256 entries in the table. */
509 	maxcode = MAXCODE(n_bits = INIT_BITS);
510 	for (code = 255; code >= 0; code--) {
511 		tab_prefixof(code) = 0;
512 		tab_suffixof(code) = (char_type) code;
513 	}
514 	free_ent = block_compress ? FIRST : 256;
515 
516 	finchar = oldcode = getcode(zs);
517 	if (oldcode == -1)	/* EOF already? */
518 		return (0);	/* Get out of here */
519 
520 	/* First code must be 8 bits = char. */
521 	*bp++ = (u_char)finchar;
522 	count--;
523 	stackp = de_stack;
524 
525 	while ((code = getcode(zs)) > -1) {
526 
527 		if ((code == CLEAR) && block_compress) {
528 			for (code = 255; code >= 0; code--)
529 				tab_prefixof(code) = 0;
530 			clear_flg = 1;
531 			free_ent = FIRST - 1;
532 			if ((code = getcode(zs)) == -1)	/* O, untimely death! */
533 				break;
534 		}
535 		incode = code;
536 
537 		/* Special case for KwKwK string. */
538 		if (code >= free_ent) {
539 			*stackp++ = finchar;
540 			code = oldcode;
541 		}
542 
543 		/* Generate output characters in reverse order. */
544 		while (code >= 256) {
545 			*stackp++ = tab_suffixof(code);
546 			code = tab_prefixof(code);
547 		}
548 		*stackp++ = finchar = tab_suffixof(code);
549 
550 		/* And put them out in forward order.  */
551 middle:		do {
552 			if (count-- == 0)
553 				return (num);
554 			*bp++ = *--stackp;
555 		} while (stackp > de_stack);
556 
557 		/* Generate the new entry. */
558 		if ((code = free_ent) < maxmaxcode) {
559 			tab_prefixof(code) = (u_short) oldcode;
560 			tab_suffixof(code) = finchar;
561 			free_ent = code + 1;
562 		}
563 
564 		/* Remember previous code. */
565 		oldcode = incode;
566 	}
567 	state = S_EOF;
568 eof:	return (num - count);
569 }
570 
571 /*-
572  * Read one code from the standard input.  If EOF, return -1.
573  * Inputs:
574  * 	stdin
575  * Outputs:
576  * 	code or -1 is returned.
577  */
578 static code_int
579 getcode(zs)
580 	struct s_zstate *zs;
581 {
582 	code_int gcode;
583 	int r_off, bits;
584 	char_type *bp;
585 
586 	bp = gbuf;
587 	if (clear_flg > 0 || roffset >= size || free_ent > maxcode) {
588 		/*
589 		 * If the next entry will be too big for the current gcode
590 		 * size, then we must increase the size.  This implies reading
591 		 * a new buffer full, too.
592 		 */
593 		if (free_ent > maxcode) {
594 			n_bits++;
595 			if (n_bits == maxbits)	/* Won't get any bigger now. */
596 				maxcode = maxmaxcode;
597 			else
598 				maxcode = MAXCODE(n_bits);
599 		}
600 		if (clear_flg > 0) {
601 			maxcode = MAXCODE(n_bits = INIT_BITS);
602 			clear_flg = 0;
603 		}
604 		size = fread(gbuf, 1, n_bits, fp);
605 		if (size <= 0)			/* End of file. */
606 			return (-1);
607 		roffset = 0;
608 		/* Round size down to integral number of codes. */
609 		size = (size << 3) - (n_bits - 1);
610 	}
611 	r_off = roffset;
612 	bits = n_bits;
613 
614 	/* Get to the first byte. */
615 	bp += (r_off >> 3);
616 	r_off &= 7;
617 
618 	/* Get first part (low order bits). */
619 	gcode = (*bp++ >> r_off);
620 	bits -= (8 - r_off);
621 	r_off = 8 - r_off;	/* Now, roffset into gcode word. */
622 
623 	/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
624 	if (bits >= 8) {
625 		gcode |= *bp++ << r_off;
626 		r_off += 8;
627 		bits -= 8;
628 	}
629 
630 	/* High order bits. */
631 	gcode |= (*bp & rmask[bits]) << r_off;
632 	roffset += n_bits;
633 
634 	return (gcode);
635 }
636 
637 static int
638 cl_block(zs)			/* Table clear for block compress. */
639 	struct s_zstate *zs;
640 {
641 	long rat;
642 
643 	checkpoint = in_count + CHECK_GAP;
644 
645 	if (in_count > 0x007fffff) {	/* Shift will overflow. */
646 		rat = bytes_out >> 8;
647 		if (rat == 0)		/* Don't divide by zero. */
648 			rat = 0x7fffffff;
649 		else
650 			rat = in_count / rat;
651 	} else
652 		rat = (in_count << 8) / bytes_out;	/* 8 fractional bits. */
653 	if (rat > ratio)
654 		ratio = rat;
655 	else {
656 		ratio = 0;
657 		cl_hash(zs, (count_int) hsize);
658 		free_ent = FIRST;
659 		clear_flg = 1;
660 		if (output(zs, (code_int) CLEAR) == -1)
661 			return (-1);
662 	}
663 	return (0);
664 }
665 
666 static void
667 cl_hash(zs, cl_hsize)			/* Reset code table. */
668 	struct s_zstate *zs;
669 	count_int cl_hsize;
670 {
671 	count_int *htab_p;
672 	long i, m1;
673 
674 	m1 = -1;
675 	htab_p = htab + cl_hsize;
676 	i = cl_hsize - 16;
677 	do {			/* Might use Sys V memset(3) here. */
678 		*(htab_p - 16) = m1;
679 		*(htab_p - 15) = m1;
680 		*(htab_p - 14) = m1;
681 		*(htab_p - 13) = m1;
682 		*(htab_p - 12) = m1;
683 		*(htab_p - 11) = m1;
684 		*(htab_p - 10) = m1;
685 		*(htab_p - 9) = m1;
686 		*(htab_p - 8) = m1;
687 		*(htab_p - 7) = m1;
688 		*(htab_p - 6) = m1;
689 		*(htab_p - 5) = m1;
690 		*(htab_p - 4) = m1;
691 		*(htab_p - 3) = m1;
692 		*(htab_p - 2) = m1;
693 		*(htab_p - 1) = m1;
694 		htab_p -= 16;
695 	} while ((i -= 16) >= 0);
696 	for (i += 16; i > 0; i--)
697 		*--htab_p = m1;
698 }
699 
700 FILE *
701 zopen(fname, mode, bits)
702 	const char *fname, *mode;
703 	int bits;
704 {
705 	struct s_zstate *zs;
706 
707 	if ((mode[0] != 'r' && mode[0] != 'w') || mode[1] != '\0' ||
708 	    bits < 0 || bits > BITS) {
709 		errno = EINVAL;
710 		return (NULL);
711 	}
712 
713 	if ((zs = calloc(1, sizeof(struct s_zstate))) == NULL)
714 		return (NULL);
715 
716 	maxbits = bits ? bits : BITS;	/* User settable max # bits/code. */
717 	maxmaxcode = 1 << maxbits;	/* Should NEVER generate this code. */
718 	hsize = HSIZE;			/* For dynamic table sizing. */
719 	free_ent = 0;			/* First unused entry. */
720 	block_compress = BLOCK_MASK;
721 	clear_flg = 0;
722 	ratio = 0;
723 	checkpoint = CHECK_GAP;
724 	in_count = 1;			/* Length of input. */
725 	out_count = 0;			/* # of codes output (for debugging). */
726 	state = S_START;
727 	roffset = 0;
728 	size = 0;
729 
730 	/*
731 	 * Layering compress on top of stdio in order to provide buffering,
732 	 * and ensure that reads and write work with the data specified.
733 	 */
734 	if ((fp = fopen(fname, mode)) == NULL) {
735 		free(zs);
736 		return (NULL);
737 	}
738 	switch (*mode) {
739 	case 'r':
740 		zmode = 'r';
741 		return (funopen(zs, zread, NULL, NULL, zclose));
742 	case 'w':
743 		zmode = 'w';
744 		return (funopen(zs, NULL, zwrite, NULL, zclose));
745 	}
746 	/* NOTREACHED */
747 	return (NULL);
748 }
749