xref: /freebsd/usr.bin/compress/zopen.c (revision 54ab3ed82b639b593fb0fe6db845e38a9d18970e)
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 #include <sys/cdefs.h>
39 
40 __FBSDID("$FreeBSD$");
41 
42 #if defined(LIBC_SCCS) && !defined(lint)
43 static char sccsid[] = "@(#)zopen.c	8.1 (Berkeley) 6/27/93";
44 #endif /* LIBC_SCCS and not lint */
45 
46 /*-
47  * fcompress.c - File compression ala IEEE Computer, June 1984.
48  *
49  * Compress authors:
50  *		Spencer W. Thomas	(decvax!utah-cs!thomas)
51  *		Jim McKie		(decvax!mcvax!jim)
52  *		Steve Davies		(decvax!vax135!petsd!peora!srd)
53  *		Ken Turkowski		(decvax!decwrl!turtlevax!ken)
54  *		James A. Woods		(decvax!ihnp4!ames!jaw)
55  *		Joe Orost		(decvax!vax135!petsd!joe)
56  *
57  * Cleaned up and converted to library returning I/O streams by
58  * Diomidis Spinellis <dds@doc.ic.ac.uk>.
59  *
60  * zopen(filename, mode, bits)
61  *	Returns a FILE * that can be used for read or write.  The modes
62  *	supported are only "r" and "w".  Seeking is not allowed.  On
63  *	reading the file is decompressed, on writing it is compressed.
64  *	The output is compatible with compress(1) with 16 bit tables.
65  *	Any file produced by compress(1) can be read.
66  */
67 
68 #include <sys/param.h>
69 #include <sys/stat.h>
70 
71 #include <ctype.h>
72 #include <errno.h>
73 #include <signal.h>
74 #include <stdio.h>
75 #include <stdlib.h>
76 #include <string.h>
77 #include <unistd.h>
78 #include "zopen.h"
79 
80 #define	BITS		16		/* Default bits. */
81 #define	HSIZE		69001		/* 95% occupancy */
82 
83 /* A code_int must be able to hold 2**BITS values of type int, and also -1. */
84 typedef long code_int;
85 typedef long count_int;
86 
87 typedef u_char char_type;
88 static char_type magic_header[] =
89 	{'\037', '\235'};		/* 1F 9D */
90 
91 #define	BIT_MASK	0x1f		/* Defines for third byte of header. */
92 #define	BLOCK_MASK	0x80
93 
94 /*
95  * Masks 0x40 and 0x20 are free.  I think 0x20 should mean that there is
96  * a fourth header byte (for expansion).
97  */
98 #define	INIT_BITS 9			/* Initial number of bits/code. */
99 
100 #define	MAXCODE(n_bits)	((1 << (n_bits)) - 1)
101 
102 struct s_zstate {
103 	FILE *zs_fp;			/* File stream for I/O */
104 	char zs_mode;			/* r or w */
105 	enum {
106 		S_START, S_MIDDLE, S_EOF
107 	} zs_state;			/* State of computation */
108 	u_int zs_n_bits;		/* Number of bits/code. */
109 	u_int zs_maxbits;		/* User settable max # bits/code. */
110 	code_int zs_maxcode;		/* Maximum code, given n_bits. */
111 	code_int zs_maxmaxcode;		/* Should NEVER generate this code. */
112 	count_int zs_htab [HSIZE];
113 	u_short zs_codetab [HSIZE];
114 	code_int zs_hsize;		/* For dynamic table sizing. */
115 	code_int zs_free_ent;		/* First unused entry. */
116 	/*
117 	 * Block compression parameters -- after all codes are used up,
118 	 * and compression rate changes, start over.
119 	 */
120 	int zs_block_compress;
121 	int zs_clear_flg;
122 	long zs_ratio;
123 	count_int zs_checkpoint;
124 	u_int zs_offset;
125 	long zs_in_count;		/* Length of input. */
126 	long zs_bytes_out;		/* Length of compressed output. */
127 	long zs_out_count;		/* # of codes output (for debugging). */
128 	char_type zs_buf[BITS];
129 	union {
130 		struct {
131 			long zs_fcode;
132 			code_int zs_ent;
133 			code_int zs_hsize_reg;
134 			int zs_hshift;
135 		} w;			/* Write paramenters */
136 		struct {
137 			char_type *zs_stackp;
138 			int zs_finchar;
139 			code_int zs_code, zs_oldcode, zs_incode;
140 			int zs_roffset, zs_size;
141 			char_type zs_gbuf[BITS];
142 		} r;			/* Read parameters */
143 	} u;
144 };
145 
146 /* Definitions to retain old variable names */
147 #define	fp		zs->zs_fp
148 #define	zmode		zs->zs_mode
149 #define	state		zs->zs_state
150 #define	n_bits		zs->zs_n_bits
151 #define	maxbits		zs->zs_maxbits
152 #define	maxcode		zs->zs_maxcode
153 #define	maxmaxcode	zs->zs_maxmaxcode
154 #define	htab		zs->zs_htab
155 #define	codetab		zs->zs_codetab
156 #define	hsize		zs->zs_hsize
157 #define	free_ent	zs->zs_free_ent
158 #define	block_compress	zs->zs_block_compress
159 #define	clear_flg	zs->zs_clear_flg
160 #define	ratio		zs->zs_ratio
161 #define	checkpoint	zs->zs_checkpoint
162 #define	offset		zs->zs_offset
163 #define	in_count	zs->zs_in_count
164 #define	bytes_out	zs->zs_bytes_out
165 #define	out_count	zs->zs_out_count
166 #define	buf		zs->zs_buf
167 #define	fcode		zs->u.w.zs_fcode
168 #define	hsize_reg	zs->u.w.zs_hsize_reg
169 #define	ent		zs->u.w.zs_ent
170 #define	hshift		zs->u.w.zs_hshift
171 #define	stackp		zs->u.r.zs_stackp
172 #define	finchar		zs->u.r.zs_finchar
173 #define	code		zs->u.r.zs_code
174 #define	oldcode		zs->u.r.zs_oldcode
175 #define	incode		zs->u.r.zs_incode
176 #define	roffset		zs->u.r.zs_roffset
177 #define	size		zs->u.r.zs_size
178 #define	gbuf		zs->u.r.zs_gbuf
179 
180 /*
181  * To save much memory, we overlay the table used by compress() with those
182  * used by decompress().  The tab_prefix table is the same size and type as
183  * the codetab.  The tab_suffix table needs 2**BITS characters.  We get this
184  * from the beginning of htab.  The output stack uses the rest of htab, and
185  * contains characters.  There is plenty of room for any possible stack
186  * (stack used to be 8000 characters).
187  */
188 
189 #define	htabof(i)	htab[i]
190 #define	codetabof(i)	codetab[i]
191 
192 #define	tab_prefixof(i)	codetabof(i)
193 #define	tab_suffixof(i)	((char_type *)(htab))[i]
194 #define	de_stack	((char_type *)&tab_suffixof(1 << BITS))
195 
196 #define	CHECK_GAP 10000		/* Ratio check interval. */
197 
198 /*
199  * the next two codes should not be changed lightly, as they must not
200  * lie within the contiguous general code space.
201  */
202 #define	FIRST	257		/* First free entry. */
203 #define	CLEAR	256		/* Table clear output code. */
204 
205 static int	cl_block __P((struct s_zstate *));
206 static void	cl_hash __P((struct s_zstate *, count_int));
207 static code_int	getcode __P((struct s_zstate *));
208 static int	output __P((struct s_zstate *, code_int));
209 static int	zclose __P((void *));
210 static int	zread __P((void *, char *, int));
211 static int	zwrite __P((void *, const char *, int));
212 
213 /*-
214  * Algorithm from "A Technique for High Performance Data Compression",
215  * Terry A. Welch, IEEE Computer Vol 17, No 6 (June 1984), pp 8-19.
216  *
217  * Algorithm:
218  * 	Modified Lempel-Ziv method (LZW).  Basically finds common
219  * substrings and replaces them with a variable size code.  This is
220  * deterministic, and can be done on the fly.  Thus, the decompression
221  * procedure needs no input table, but tracks the way the table was built.
222  */
223 
224 /*-
225  * compress write
226  *
227  * Algorithm:  use open addressing double hashing (no chaining) on the
228  * prefix code / next character combination.  We do a variant of Knuth's
229  * algorithm D (vol. 3, sec. 6.4) along with G. Knott's relatively-prime
230  * secondary probe.  Here, the modular division first probe is gives way
231  * to a faster exclusive-or manipulation.  Also do block compression with
232  * an adaptive reset, whereby the code table is cleared when the compression
233  * ratio decreases, but after the table fills.  The variable-length output
234  * codes are re-sized at this point, and a special CLEAR code is generated
235  * for the decompressor.  Late addition:  construct the table according to
236  * file size for noticeable speed improvement on small files.  Please direct
237  * questions about this implementation to ames!jaw.
238  */
239 static int
240 zwrite(cookie, wbp, num)
241 	void *cookie;
242 	const char *wbp;
243 	int num;
244 {
245 	code_int i;
246 	int c, disp;
247 	struct s_zstate *zs;
248 	const u_char *bp;
249 	u_char tmp;
250 	int count;
251 
252 	if (num == 0)
253 		return (0);
254 
255 	zs = cookie;
256 	count = num;
257 	bp = wbp;
258 	if (state == S_MIDDLE)
259 		goto middle;
260 	state = S_MIDDLE;
261 
262 	maxmaxcode = 1L << maxbits;
263 	if (fwrite(magic_header,
264 	    sizeof(char), sizeof(magic_header), fp) != sizeof(magic_header))
265 		return (-1);
266 	tmp = (u_char)((maxbits) | block_compress);
267 	if (fwrite(&tmp, sizeof(char), sizeof(tmp), fp) != sizeof(tmp))
268 		return (-1);
269 
270 	offset = 0;
271 	bytes_out = 3;		/* Includes 3-byte header mojo. */
272 	out_count = 0;
273 	clear_flg = 0;
274 	ratio = 0;
275 	in_count = 1;
276 	checkpoint = CHECK_GAP;
277 	maxcode = MAXCODE(n_bits = INIT_BITS);
278 	free_ent = ((block_compress) ? FIRST : 256);
279 
280 	ent = *bp++;
281 	--count;
282 
283 	hshift = 0;
284 	for (fcode = (long)hsize; fcode < 65536L; fcode *= 2L)
285 		hshift++;
286 	hshift = 8 - hshift;	/* Set hash code range bound. */
287 
288 	hsize_reg = hsize;
289 	cl_hash(zs, (count_int)hsize_reg);	/* Clear hash table. */
290 
291 middle:	for (i = 0; count--;) {
292 		c = *bp++;
293 		in_count++;
294 		fcode = (long)(((long)c << maxbits) + ent);
295 		i = ((c << hshift) ^ ent);	/* Xor hashing. */
296 
297 		if (htabof(i) == fcode) {
298 			ent = codetabof(i);
299 			continue;
300 		} else if ((long)htabof(i) < 0)	/* Empty slot. */
301 			goto nomatch;
302 		disp = hsize_reg - i;	/* Secondary hash (after G. Knott). */
303 		if (i == 0)
304 			disp = 1;
305 probe:		if ((i -= disp) < 0)
306 			i += hsize_reg;
307 
308 		if (htabof(i) == fcode) {
309 			ent = codetabof(i);
310 			continue;
311 		}
312 		if ((long)htabof(i) >= 0)
313 			goto probe;
314 nomatch:	if (output(zs, (code_int) ent) == -1)
315 			return (-1);
316 		out_count++;
317 		ent = c;
318 		if (free_ent < maxmaxcode) {
319 			codetabof(i) = free_ent++;	/* code -> hashtable */
320 			htabof(i) = fcode;
321 		} else if ((count_int)in_count >=
322 		    checkpoint && block_compress) {
323 			if (cl_block(zs) == -1)
324 				return (-1);
325 		}
326 	}
327 	return (num);
328 }
329 
330 static int
331 zclose(cookie)
332 	void *cookie;
333 {
334 	struct s_zstate *zs;
335 	int rval;
336 
337 	zs = cookie;
338 	if (zmode == 'w') {		/* Put out the final code. */
339 		if (output(zs, (code_int) ent) == -1) {
340 			(void)fclose(fp);
341 			free(zs);
342 			return (-1);
343 		}
344 		out_count++;
345 		if (output(zs, (code_int) - 1) == -1) {
346 			(void)fclose(fp);
347 			free(zs);
348 			return (-1);
349 		}
350 	}
351 	rval = fclose(fp) == EOF ? -1 : 0;
352 	free(zs);
353 	return (rval);
354 }
355 
356 /*-
357  * Output the given code.
358  * Inputs:
359  * 	code:	A n_bits-bit integer.  If == -1, then EOF.  This assumes
360  *		that n_bits =< (long)wordsize - 1.
361  * Outputs:
362  * 	Outputs code to the file.
363  * Assumptions:
364  *	Chars are 8 bits long.
365  * Algorithm:
366  * 	Maintain a BITS character long buffer (so that 8 codes will
367  * fit in it exactly).  Use the VAX insv instruction to insert each
368  * code in turn.  When the buffer fills up empty it and start over.
369  */
370 
371 static char_type lmask[9] =
372 	{0xff, 0xfe, 0xfc, 0xf8, 0xf0, 0xe0, 0xc0, 0x80, 0x00};
373 static char_type rmask[9] =
374 	{0x00, 0x01, 0x03, 0x07, 0x0f, 0x1f, 0x3f, 0x7f, 0xff};
375 
376 static int
377 output(zs, ocode)
378 	struct s_zstate *zs;
379 	code_int ocode;
380 {
381 	int r_off;
382 	u_int bits;
383 	char_type *bp;
384 
385 	r_off = offset;
386 	bits = n_bits;
387 	bp = buf;
388 	if (ocode >= 0) {
389 		/* Get to the first byte. */
390 		bp += (r_off >> 3);
391 		r_off &= 7;
392 		/*
393 		 * Since ocode is always >= 8 bits, only need to mask the first
394 		 * hunk on the left.
395 		 */
396 		*bp = (*bp & rmask[r_off]) | ((ocode << r_off) & lmask[r_off]);
397 		bp++;
398 		bits -= (8 - r_off);
399 		ocode >>= 8 - r_off;
400 		/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
401 		if (bits >= 8) {
402 			*bp++ = ocode;
403 			ocode >>= 8;
404 			bits -= 8;
405 		}
406 		/* Last bits. */
407 		if (bits)
408 			*bp = ocode;
409 		offset += n_bits;
410 		if (offset == (n_bits << 3)) {
411 			bp = buf;
412 			bits = n_bits;
413 			bytes_out += bits;
414 			if (fwrite(bp, sizeof(char), bits, fp) != bits)
415 				return (-1);
416 			bp += bits;
417 			bits = 0;
418 			offset = 0;
419 		}
420 		/*
421 		 * If the next entry is going to be too big for the ocode size,
422 		 * then increase it, if possible.
423 		 */
424 		if (free_ent > maxcode || (clear_flg > 0)) {
425 		       /*
426 			* Write the whole buffer, because the input side won't
427 			* discover the size increase until after it has read it.
428 			*/
429 			if (offset > 0) {
430 				if (fwrite(buf, 1, n_bits, fp) != n_bits)
431 					return (-1);
432 				bytes_out += n_bits;
433 			}
434 			offset = 0;
435 
436 			if (clear_flg) {
437 				maxcode = MAXCODE(n_bits = INIT_BITS);
438 				clear_flg = 0;
439 			} else {
440 				n_bits++;
441 				if (n_bits == maxbits)
442 					maxcode = maxmaxcode;
443 				else
444 					maxcode = MAXCODE(n_bits);
445 			}
446 		}
447 	} else {
448 		/* At EOF, write the rest of the buffer. */
449 		if (offset > 0) {
450 			offset = (offset + 7) / 8;
451 			if (fwrite(buf, 1, offset, fp) != offset)
452 				return (-1);
453 			bytes_out += offset;
454 		}
455 		offset = 0;
456 	}
457 	return (0);
458 }
459 
460 /*
461  * Decompress read.  This routine adapts to the codes in the file building
462  * the "string" table on-the-fly; requiring no table to be stored in the
463  * compressed file.  The tables used herein are shared with those of the
464  * compress() routine.  See the definitions above.
465  */
466 static int
467 zread(cookie, rbp, num)
468 	void *cookie;
469 	char *rbp;
470 	int num;
471 {
472 	u_int count;
473 	struct s_zstate *zs;
474 	u_char *bp, header[3];
475 
476 	if (num == 0)
477 		return (0);
478 
479 	zs = cookie;
480 	count = num;
481 	bp = (u_char *)rbp;
482 	switch (state) {
483 	case S_START:
484 		state = S_MIDDLE;
485 		break;
486 	case S_MIDDLE:
487 		goto middle;
488 	case S_EOF:
489 		goto eof;
490 	}
491 
492 	/* Check the magic number */
493 	if (fread(header,
494 	    sizeof(char), sizeof(header), fp) != sizeof(header) ||
495 	    memcmp(header, magic_header, sizeof(magic_header)) != 0) {
496 		errno = EFTYPE;
497 		return (-1);
498 	}
499 	maxbits = header[2];	/* Set -b from file. */
500 	block_compress = maxbits & BLOCK_MASK;
501 	maxbits &= BIT_MASK;
502 	maxmaxcode = 1L << maxbits;
503 	if (maxbits > BITS) {
504 		errno = EFTYPE;
505 		return (-1);
506 	}
507 	/* As above, initialize the first 256 entries in the table. */
508 	maxcode = MAXCODE(n_bits = INIT_BITS);
509 	for (code = 255; code >= 0; code--) {
510 		tab_prefixof(code) = 0;
511 		tab_suffixof(code) = (char_type) code;
512 	}
513 	free_ent = block_compress ? FIRST : 256;
514 
515 	finchar = oldcode = getcode(zs);
516 	if (oldcode == -1)	/* EOF already? */
517 		return (0);	/* Get out of here */
518 
519 	/* First code must be 8 bits = char. */
520 	*bp++ = (u_char)finchar;
521 	count--;
522 	stackp = de_stack;
523 
524 	while ((code = getcode(zs)) > -1) {
525 
526 		if ((code == CLEAR) && block_compress) {
527 			for (code = 255; code >= 0; code--)
528 				tab_prefixof(code) = 0;
529 			clear_flg = 1;
530 			free_ent = FIRST - 1;
531 			if ((code = getcode(zs)) == -1)	/* O, untimely death! */
532 				break;
533 		}
534 		incode = code;
535 
536 		/* Special case for KwKwK string. */
537 		if (code >= free_ent) {
538 			*stackp++ = finchar;
539 			code = oldcode;
540 		}
541 
542 		/* Generate output characters in reverse order. */
543 		while (code >= 256) {
544 			*stackp++ = tab_suffixof(code);
545 			code = tab_prefixof(code);
546 		}
547 		*stackp++ = finchar = tab_suffixof(code);
548 
549 		/* And put them out in forward order.  */
550 middle:		do {
551 			if (count-- == 0)
552 				return (num);
553 			*bp++ = *--stackp;
554 		} while (stackp > de_stack);
555 
556 		/* Generate the new entry. */
557 		if ((code = free_ent) < maxmaxcode) {
558 			tab_prefixof(code) = (u_short) oldcode;
559 			tab_suffixof(code) = finchar;
560 			free_ent = code + 1;
561 		}
562 
563 		/* Remember previous code. */
564 		oldcode = incode;
565 	}
566 	state = S_EOF;
567 eof:	return (num - count);
568 }
569 
570 /*-
571  * Read one code from the standard input.  If EOF, return -1.
572  * Inputs:
573  * 	stdin
574  * Outputs:
575  * 	code or -1 is returned.
576  */
577 static code_int
578 getcode(zs)
579 	struct s_zstate *zs;
580 {
581 	code_int gcode;
582 	int r_off, bits;
583 	char_type *bp;
584 
585 	bp = gbuf;
586 	if (clear_flg > 0 || roffset >= size || free_ent > maxcode) {
587 		/*
588 		 * If the next entry will be too big for the current gcode
589 		 * size, then we must increase the size.  This implies reading
590 		 * a new buffer full, too.
591 		 */
592 		if (free_ent > maxcode) {
593 			n_bits++;
594 			if (n_bits == maxbits)	/* Won't get any bigger now. */
595 				maxcode = maxmaxcode;
596 			else
597 				maxcode = MAXCODE(n_bits);
598 		}
599 		if (clear_flg > 0) {
600 			maxcode = MAXCODE(n_bits = INIT_BITS);
601 			clear_flg = 0;
602 		}
603 		size = fread(gbuf, 1, n_bits, fp);
604 		if (size <= 0)			/* End of file. */
605 			return (-1);
606 		roffset = 0;
607 		/* Round size down to integral number of codes. */
608 		size = (size << 3) - (n_bits - 1);
609 	}
610 	r_off = roffset;
611 	bits = n_bits;
612 
613 	/* Get to the first byte. */
614 	bp += (r_off >> 3);
615 	r_off &= 7;
616 
617 	/* Get first part (low order bits). */
618 	gcode = (*bp++ >> r_off);
619 	bits -= (8 - r_off);
620 	r_off = 8 - r_off;	/* Now, roffset into gcode word. */
621 
622 	/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
623 	if (bits >= 8) {
624 		gcode |= *bp++ << r_off;
625 		r_off += 8;
626 		bits -= 8;
627 	}
628 
629 	/* High order bits. */
630 	gcode |= (*bp & rmask[bits]) << r_off;
631 	roffset += n_bits;
632 
633 	return (gcode);
634 }
635 
636 static int
637 cl_block(zs)			/* Table clear for block compress. */
638 	struct s_zstate *zs;
639 {
640 	long rat;
641 
642 	checkpoint = in_count + CHECK_GAP;
643 
644 	if (in_count > 0x007fffff) {	/* Shift will overflow. */
645 		rat = bytes_out >> 8;
646 		if (rat == 0)		/* Don't divide by zero. */
647 			rat = 0x7fffffff;
648 		else
649 			rat = in_count / rat;
650 	} else
651 		rat = (in_count << 8) / bytes_out;	/* 8 fractional bits. */
652 	if (rat > ratio)
653 		ratio = rat;
654 	else {
655 		ratio = 0;
656 		cl_hash(zs, (count_int) hsize);
657 		free_ent = FIRST;
658 		clear_flg = 1;
659 		if (output(zs, (code_int) CLEAR) == -1)
660 			return (-1);
661 	}
662 	return (0);
663 }
664 
665 static void
666 cl_hash(zs, cl_hsize)			/* Reset code table. */
667 	struct s_zstate *zs;
668 	count_int cl_hsize;
669 {
670 	count_int *htab_p;
671 	long i, m1;
672 
673 	m1 = -1;
674 	htab_p = htab + cl_hsize;
675 	i = cl_hsize - 16;
676 	do {			/* Might use Sys V memset(3) here. */
677 		*(htab_p - 16) = m1;
678 		*(htab_p - 15) = m1;
679 		*(htab_p - 14) = m1;
680 		*(htab_p - 13) = m1;
681 		*(htab_p - 12) = m1;
682 		*(htab_p - 11) = m1;
683 		*(htab_p - 10) = m1;
684 		*(htab_p - 9) = m1;
685 		*(htab_p - 8) = m1;
686 		*(htab_p - 7) = m1;
687 		*(htab_p - 6) = m1;
688 		*(htab_p - 5) = m1;
689 		*(htab_p - 4) = m1;
690 		*(htab_p - 3) = m1;
691 		*(htab_p - 2) = m1;
692 		*(htab_p - 1) = m1;
693 		htab_p -= 16;
694 	} while ((i -= 16) >= 0);
695 	for (i += 16; i > 0; i--)
696 		*--htab_p = m1;
697 }
698 
699 FILE *
700 zopen(fname, mode, bits)
701 	const char *fname, *mode;
702 	int bits;
703 {
704 	struct s_zstate *zs;
705 
706 	if ((mode[0] != 'r' && mode[0] != 'w') || mode[1] != '\0' ||
707 	    bits < 0 || bits > BITS) {
708 		errno = EINVAL;
709 		return (NULL);
710 	}
711 
712 	if ((zs = calloc(1, sizeof(struct s_zstate))) == NULL)
713 		return (NULL);
714 
715 	maxbits = bits ? bits : BITS;	/* User settable max # bits/code. */
716 	maxmaxcode = 1L << maxbits;	/* Should NEVER generate this code. */
717 	hsize = HSIZE;			/* For dynamic table sizing. */
718 	free_ent = 0;			/* First unused entry. */
719 	block_compress = BLOCK_MASK;
720 	clear_flg = 0;
721 	ratio = 0;
722 	checkpoint = CHECK_GAP;
723 	in_count = 1;			/* Length of input. */
724 	out_count = 0;			/* # of codes output (for debugging). */
725 	state = S_START;
726 	roffset = 0;
727 	size = 0;
728 
729 	/*
730 	 * Layering compress on top of stdio in order to provide buffering,
731 	 * and ensure that reads and write work with the data specified.
732 	 */
733 	if ((fp = fopen(fname, mode)) == NULL) {
734 		free(zs);
735 		return (NULL);
736 	}
737 	switch (*mode) {
738 	case 'r':
739 		zmode = 'r';
740 		return (funopen(zs, zread, NULL, NULL, zclose));
741 	case 'w':
742 		zmode = 'w';
743 		return (funopen(zs, NULL, zwrite, NULL, zclose));
744 	}
745 	/* NOTREACHED */
746 	return (NULL);
747 }
748