xref: /freebsd/usr.bin/compress/zopen.c (revision afe61c15161c324a7af299a9b8457aba5afc92db)
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 
75 #define	BITS		16		/* Default bits. */
76 #define	HSIZE		69001		/* 95% occupancy */
77 
78 /* A code_int must be able to hold 2**BITS values of type int, and also -1. */
79 typedef long code_int;
80 typedef long count_int;
81 
82 typedef u_char char_type;
83 static char_type magic_header[] =
84 	{'\037', '\235'};		/* 1F 9D */
85 
86 #define	BIT_MASK	0x1f		/* Defines for third byte of header. */
87 #define	BLOCK_MASK	0x80
88 
89 /*
90  * Masks 0x40 and 0x20 are free.  I think 0x20 should mean that there is
91  * a fourth header byte (for expansion).
92  */
93 #define	INIT_BITS 9			/* Initial number of bits/code. */
94 
95 #define	MAXCODE(n_bits)	((1 << (n_bits)) - 1)
96 
97 struct s_zstate {
98 	FILE *zs_fp;			/* File stream for I/O */
99 	char zs_mode;			/* r or w */
100 	enum {
101 		S_START, S_MIDDLE, S_EOF
102 	} zs_state;			/* State of computation */
103 	int zs_n_bits;			/* Number of bits/code. */
104 	int zs_maxbits;			/* User settable max # bits/code. */
105 	code_int zs_maxcode;		/* Maximum code, given n_bits. */
106 	code_int zs_maxmaxcode;		/* Should NEVER generate this code. */
107 	count_int zs_htab [HSIZE];
108 	u_short zs_codetab [HSIZE];
109 	code_int zs_hsize;		/* For dynamic table sizing. */
110 	code_int zs_free_ent;		/* First unused entry. */
111 	/*
112 	 * Block compression parameters -- after all codes are used up,
113 	 * and compression rate changes, start over.
114 	 */
115 	int zs_block_compress;
116 	int zs_clear_flg;
117 	long zs_ratio;
118 	count_int zs_checkpoint;
119 	int zs_offset;
120 	long zs_in_count;		/* Length of input. */
121 	long zs_bytes_out;		/* Length of compressed output. */
122 	long zs_out_count;		/* # of codes output (for debugging). */
123 	char_type zs_buf[BITS];
124 	union {
125 		struct {
126 			long zs_fcode;
127 			code_int zs_ent;
128 			code_int zs_hsize_reg;
129 			int zs_hshift;
130 		} w;			/* Write paramenters */
131 		struct {
132 			char_type *zs_stackp;
133 			int zs_finchar;
134 			code_int zs_code, zs_oldcode, zs_incode;
135 			int zs_roffset, zs_size;
136 			char_type zs_gbuf[BITS];
137 		} r;			/* Read parameters */
138 	} u;
139 };
140 
141 /* Definitions to retain old variable names */
142 #define	fp		zs->zs_fp
143 #define	zmode		zs->zs_mode
144 #define	state		zs->zs_state
145 #define	n_bits		zs->zs_n_bits
146 #define	maxbits		zs->zs_maxbits
147 #define	maxcode		zs->zs_maxcode
148 #define	maxmaxcode	zs->zs_maxmaxcode
149 #define	htab		zs->zs_htab
150 #define	codetab		zs->zs_codetab
151 #define	hsize		zs->zs_hsize
152 #define	free_ent	zs->zs_free_ent
153 #define	block_compress	zs->zs_block_compress
154 #define	clear_flg	zs->zs_clear_flg
155 #define	ratio		zs->zs_ratio
156 #define	checkpoint	zs->zs_checkpoint
157 #define	offset		zs->zs_offset
158 #define	in_count	zs->zs_in_count
159 #define	bytes_out	zs->zs_bytes_out
160 #define	out_count	zs->zs_out_count
161 #define	buf		zs->zs_buf
162 #define	fcode		zs->u.w.zs_fcode
163 #define	hsize_reg	zs->u.w.zs_hsize_reg
164 #define	ent		zs->u.w.zs_ent
165 #define	hshift		zs->u.w.zs_hshift
166 #define	stackp		zs->u.r.zs_stackp
167 #define	finchar		zs->u.r.zs_finchar
168 #define	code		zs->u.r.zs_code
169 #define	oldcode		zs->u.r.zs_oldcode
170 #define	incode		zs->u.r.zs_incode
171 #define	roffset		zs->u.r.zs_roffset
172 #define	size		zs->u.r.zs_size
173 #define	gbuf		zs->u.r.zs_gbuf
174 
175 /*
176  * To save much memory, we overlay the table used by compress() with those
177  * used by decompress().  The tab_prefix table is the same size and type as
178  * the codetab.  The tab_suffix table needs 2**BITS characters.  We get this
179  * from the beginning of htab.  The output stack uses the rest of htab, and
180  * contains characters.  There is plenty of room for any possible stack
181  * (stack used to be 8000 characters).
182  */
183 
184 #define	htabof(i)	htab[i]
185 #define	codetabof(i)	codetab[i]
186 
187 #define	tab_prefixof(i)	codetabof(i)
188 #define	tab_suffixof(i)	((char_type *)(htab))[i]
189 #define	de_stack	((char_type *)&tab_suffixof(1 << BITS))
190 
191 #define	CHECK_GAP 10000		/* Ratio check interval. */
192 
193 /*
194  * the next two codes should not be changed lightly, as they must not
195  * lie within the contiguous general code space.
196  */
197 #define	FIRST	257		/* First free entry. */
198 #define	CLEAR	256		/* Table clear output code. */
199 
200 static int	cl_block __P((struct s_zstate *));
201 static void	cl_hash __P((struct s_zstate *, count_int));
202 static code_int	getcode __P((struct s_zstate *));
203 static int	output __P((struct s_zstate *, code_int));
204 static int	zclose __P((void *));
205 static int	zread __P((void *, char *, int));
206 static int	zwrite __P((void *, const char *, int));
207 
208 /*-
209  * Algorithm from "A Technique for High Performance Data Compression",
210  * Terry A. Welch, IEEE Computer Vol 17, No 6 (June 1984), pp 8-19.
211  *
212  * Algorithm:
213  * 	Modified Lempel-Ziv method (LZW).  Basically finds common
214  * substrings and replaces them with a variable size code.  This is
215  * deterministic, and can be done on the fly.  Thus, the decompression
216  * procedure needs no input table, but tracks the way the table was built.
217  */
218 
219 /*-
220  * compress write
221  *
222  * Algorithm:  use open addressing double hashing (no chaining) on the
223  * prefix code / next character combination.  We do a variant of Knuth's
224  * algorithm D (vol. 3, sec. 6.4) along with G. Knott's relatively-prime
225  * secondary probe.  Here, the modular division first probe is gives way
226  * to a faster exclusive-or manipulation.  Also do block compression with
227  * an adaptive reset, whereby the code table is cleared when the compression
228  * ratio decreases, but after the table fills.  The variable-length output
229  * codes are re-sized at this point, and a special CLEAR code is generated
230  * for the decompressor.  Late addition:  construct the table according to
231  * file size for noticeable speed improvement on small files.  Please direct
232  * questions about this implementation to ames!jaw.
233  */
234 static int
235 zwrite(cookie, wbp, num)
236 	void *cookie;
237 	const char *wbp;
238 	int num;
239 {
240 	register code_int i;
241 	register int c, disp;
242 	struct s_zstate *zs;
243 	const u_char *bp;
244 	u_char tmp;
245 	int count;
246 
247 	if (num == 0)
248 		return (0);
249 
250 	zs = cookie;
251 	count = num;
252 	bp = (u_char *)wbp;
253 	if (state == S_MIDDLE)
254 		goto middle;
255 	state = S_MIDDLE;
256 
257 	maxmaxcode = 1L << BITS;
258 	if (fwrite(magic_header,
259 	    sizeof(char), sizeof(magic_header), fp) != sizeof(magic_header))
260 		return (-1);
261 	tmp = (u_char)(BITS | block_compress);
262 	if (fwrite(&tmp, sizeof(char), sizeof(tmp), fp) != sizeof(tmp))
263 		return (-1);
264 
265 	offset = 0;
266 	bytes_out = 3;		/* Includes 3-byte header mojo. */
267 	out_count = 0;
268 	clear_flg = 0;
269 	ratio = 0;
270 	in_count = 1;
271 	checkpoint = CHECK_GAP;
272 	maxcode = MAXCODE(n_bits = INIT_BITS);
273 	free_ent = ((block_compress) ? FIRST : 256);
274 
275 	ent = *bp++;
276 	--count;
277 
278 	hshift = 0;
279 	for (fcode = (long)hsize; fcode < 65536L; fcode *= 2L)
280 		hshift++;
281 	hshift = 8 - hshift;	/* Set hash code range bound. */
282 
283 	hsize_reg = hsize;
284 	cl_hash(zs, (count_int)hsize_reg);	/* Clear hash table. */
285 
286 middle:	for (i = 0; count--;) {
287 		c = *bp++;
288 		in_count++;
289 		fcode = (long)(((long)c << maxbits) + ent);
290 		i = ((c << hshift) ^ ent);	/* Xor hashing. */
291 
292 		if (htabof(i) == fcode) {
293 			ent = codetabof(i);
294 			continue;
295 		} else if ((long)htabof(i) < 0)	/* Empty slot. */
296 			goto nomatch;
297 		disp = hsize_reg - i;	/* Secondary hash (after G. Knott). */
298 		if (i == 0)
299 			disp = 1;
300 probe:		if ((i -= disp) < 0)
301 			i += hsize_reg;
302 
303 		if (htabof(i) == fcode) {
304 			ent = codetabof(i);
305 			continue;
306 		}
307 		if ((long)htabof(i) >= 0)
308 			goto probe;
309 nomatch:	if (output(zs, (code_int) ent) == -1)
310 			return (-1);
311 		out_count++;
312 		ent = c;
313 		if (free_ent < maxmaxcode) {
314 			codetabof(i) = free_ent++;	/* code -> hashtable */
315 			htabof(i) = fcode;
316 		} else if ((count_int)in_count >=
317 		    checkpoint && block_compress) {
318 			if (cl_block(zs) == -1)
319 				return (-1);
320 		}
321 	}
322 	return (num);
323 }
324 
325 static int
326 zclose(cookie)
327 	void *cookie;
328 {
329 	struct s_zstate *zs;
330 	int rval;
331 
332 	zs = cookie;
333 	if (zmode == 'w') {		/* Put out the final code. */
334 		if (output(zs, (code_int) ent) == -1) {
335 			(void)fclose(fp);
336 			free(zs);
337 			return (-1);
338 		}
339 		out_count++;
340 		if (output(zs, (code_int) - 1) == -1) {
341 			(void)fclose(fp);
342 			free(zs);
343 			return (-1);
344 		}
345 	}
346 	rval = fclose(fp) == EOF ? -1 : 0;
347 	free(zs);
348 	return (rval);
349 }
350 
351 /*-
352  * Output the given code.
353  * Inputs:
354  * 	code:	A n_bits-bit integer.  If == -1, then EOF.  This assumes
355  *		that n_bits =< (long)wordsize - 1.
356  * Outputs:
357  * 	Outputs code to the file.
358  * Assumptions:
359  *	Chars are 8 bits long.
360  * Algorithm:
361  * 	Maintain a BITS character long buffer (so that 8 codes will
362  * fit in it exactly).  Use the VAX insv instruction to insert each
363  * code in turn.  When the buffer fills up empty it and start over.
364  */
365 
366 static char_type lmask[9] =
367 	{0xff, 0xfe, 0xfc, 0xf8, 0xf0, 0xe0, 0xc0, 0x80, 0x00};
368 static char_type rmask[9] =
369 	{0x00, 0x01, 0x03, 0x07, 0x0f, 0x1f, 0x3f, 0x7f, 0xff};
370 
371 static int
372 output(zs, ocode)
373 	struct s_zstate *zs;
374 	code_int ocode;
375 {
376 	register int bits, r_off;
377 	register char_type *bp;
378 
379 	r_off = offset;
380 	bits = n_bits;
381 	bp = buf;
382 	if (ocode >= 0) {
383 		/* Get to the first byte. */
384 		bp += (r_off >> 3);
385 		r_off &= 7;
386 		/*
387 		 * Since ocode is always >= 8 bits, only need to mask the first
388 		 * hunk on the left.
389 		 */
390 		*bp = (*bp & rmask[r_off]) | (ocode << r_off) & lmask[r_off];
391 		bp++;
392 		bits -= (8 - r_off);
393 		ocode >>= 8 - r_off;
394 		/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
395 		if (bits >= 8) {
396 			*bp++ = ocode;
397 			ocode >>= 8;
398 			bits -= 8;
399 		}
400 		/* Last bits. */
401 		if (bits)
402 			*bp = ocode;
403 		offset += n_bits;
404 		if (offset == (n_bits << 3)) {
405 			bp = buf;
406 			bits = n_bits;
407 			bytes_out += bits;
408 			if (fwrite(bp, sizeof(char), bits, fp) != bits)
409 				return (-1);
410 			bp += bits;
411 			bits = 0;
412 			offset = 0;
413 		}
414 		/*
415 		 * If the next entry is going to be too big for the ocode size,
416 		 * then increase it, if possible.
417 		 */
418 		if (free_ent > maxcode || (clear_flg > 0)) {
419 		       /*
420 			* Write the whole buffer, because the input side won't
421 			* discover the size increase until after it has read it.
422 			*/
423 			if (offset > 0) {
424 				if (fwrite(buf, 1, n_bits, fp) != n_bits)
425 					return (-1);
426 				bytes_out += n_bits;
427 			}
428 			offset = 0;
429 
430 			if (clear_flg) {
431 				maxcode = MAXCODE(n_bits = INIT_BITS);
432 				clear_flg = 0;
433 			} else {
434 				n_bits++;
435 				if (n_bits == maxbits)
436 					maxcode = maxmaxcode;
437 				else
438 					maxcode = MAXCODE(n_bits);
439 			}
440 		}
441 	} else {
442 		/* At EOF, write the rest of the buffer. */
443 		if (offset > 0) {
444 			offset = (offset + 7) / 8;
445 			if (fwrite(buf, 1, offset, fp) != offset)
446 				return (-1);
447 			bytes_out += offset;
448 		}
449 		offset = 0;
450 	}
451 	return (0);
452 }
453 
454 /*
455  * Decompress read.  This routine adapts to the codes in the file building
456  * the "string" table on-the-fly; requiring no table to be stored in the
457  * compressed file.  The tables used herein are shared with those of the
458  * compress() routine.  See the definitions above.
459  */
460 static int
461 zread(cookie, rbp, num)
462 	void *cookie;
463 	char *rbp;
464 	int num;
465 {
466 	register u_int count;
467 	struct s_zstate *zs;
468 	u_char *bp, header[3];
469 
470 	if (num == 0)
471 		return (0);
472 
473 	zs = cookie;
474 	count = num;
475 	bp = (u_char *)rbp;
476 	switch (state) {
477 	case S_START:
478 		state = S_MIDDLE;
479 		break;
480 	case S_MIDDLE:
481 		goto middle;
482 	case S_EOF:
483 		goto eof;
484 	}
485 
486 	/* Check the magic number */
487 	if (fread(header,
488 	    sizeof(char), sizeof(header), fp) != sizeof(header) ||
489 	    memcmp(header, magic_header, sizeof(magic_header)) != 0) {
490 		errno = EFTYPE;
491 		return (-1);
492 	}
493 	maxbits = header[2];	/* Set -b from file. */
494 	block_compress = maxbits & BLOCK_MASK;
495 	maxbits &= BIT_MASK;
496 	maxmaxcode = 1L << maxbits;
497 	if (maxbits > BITS) {
498 		errno = EFTYPE;
499 		return (-1);
500 	}
501 	/* As above, initialize the first 256 entries in the table. */
502 	maxcode = MAXCODE(n_bits = INIT_BITS);
503 	for (code = 255; code >= 0; code--) {
504 		tab_prefixof(code) = 0;
505 		tab_suffixof(code) = (char_type) code;
506 	}
507 	free_ent = block_compress ? FIRST : 256;
508 
509 	finchar = oldcode = getcode(zs);
510 	if (oldcode == -1)	/* EOF already? */
511 		return (0);	/* Get out of here */
512 
513 	/* First code must be 8 bits = char. */
514 	*bp++ = (u_char)finchar;
515 	count--;
516 	stackp = de_stack;
517 
518 	while ((code = getcode(zs)) > -1) {
519 
520 		if ((code == CLEAR) && block_compress) {
521 			for (code = 255; code >= 0; code--)
522 				tab_prefixof(code) = 0;
523 			clear_flg = 1;
524 			free_ent = FIRST - 1;
525 			if ((code = getcode(zs)) == -1)	/* O, untimely death! */
526 				break;
527 		}
528 		incode = code;
529 
530 		/* Special case for KwKwK string. */
531 		if (code >= free_ent) {
532 			*stackp++ = finchar;
533 			code = oldcode;
534 		}
535 
536 		/* Generate output characters in reverse order. */
537 		while (code >= 256) {
538 			*stackp++ = tab_suffixof(code);
539 			code = tab_prefixof(code);
540 		}
541 		*stackp++ = finchar = tab_suffixof(code);
542 
543 		/* And put them out in forward order.  */
544 middle:		do {
545 			if (count-- == 0)
546 				return (num);
547 			*bp++ = *--stackp;
548 		} while (stackp > de_stack);
549 
550 		/* Generate the new entry. */
551 		if ((code = free_ent) < maxmaxcode) {
552 			tab_prefixof(code) = (u_short) oldcode;
553 			tab_suffixof(code) = finchar;
554 			free_ent = code + 1;
555 		}
556 
557 		/* Remember previous code. */
558 		oldcode = incode;
559 	}
560 	state = S_EOF;
561 eof:	return (num - count);
562 }
563 
564 /*-
565  * Read one code from the standard input.  If EOF, return -1.
566  * Inputs:
567  * 	stdin
568  * Outputs:
569  * 	code or -1 is returned.
570  */
571 static code_int
572 getcode(zs)
573 	struct s_zstate *zs;
574 {
575 	register code_int gcode;
576 	register int r_off, bits;
577 	register char_type *bp;
578 
579 	bp = gbuf;
580 	if (clear_flg > 0 || roffset >= size || free_ent > maxcode) {
581 		/*
582 		 * If the next entry will be too big for the current gcode
583 		 * size, then we must increase the size.  This implies reading
584 		 * a new buffer full, too.
585 		 */
586 		if (free_ent > maxcode) {
587 			n_bits++;
588 			if (n_bits == maxbits)	/* Won't get any bigger now. */
589 				maxcode = maxmaxcode;
590 			else
591 				maxcode = MAXCODE(n_bits);
592 		}
593 		if (clear_flg > 0) {
594 			maxcode = MAXCODE(n_bits = INIT_BITS);
595 			clear_flg = 0;
596 		}
597 		size = fread(gbuf, 1, n_bits, fp);
598 		if (size <= 0)			/* End of file. */
599 			return (-1);
600 		roffset = 0;
601 		/* Round size down to integral number of codes. */
602 		size = (size << 3) - (n_bits - 1);
603 	}
604 	r_off = roffset;
605 	bits = n_bits;
606 
607 	/* Get to the first byte. */
608 	bp += (r_off >> 3);
609 	r_off &= 7;
610 
611 	/* Get first part (low order bits). */
612 	gcode = (*bp++ >> r_off);
613 	bits -= (8 - r_off);
614 	r_off = 8 - r_off;	/* Now, roffset into gcode word. */
615 
616 	/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
617 	if (bits >= 8) {
618 		gcode |= *bp++ << r_off;
619 		r_off += 8;
620 		bits -= 8;
621 	}
622 
623 	/* High order bits. */
624 	gcode |= (*bp & rmask[bits]) << r_off;
625 	roffset += n_bits;
626 
627 	return (gcode);
628 }
629 
630 static int
631 cl_block(zs)			/* Table clear for block compress. */
632 	struct s_zstate *zs;
633 {
634 	register long rat;
635 
636 	checkpoint = in_count + CHECK_GAP;
637 
638 	if (in_count > 0x007fffff) {	/* Shift will overflow. */
639 		rat = bytes_out >> 8;
640 		if (rat == 0)		/* Don't divide by zero. */
641 			rat = 0x7fffffff;
642 		else
643 			rat = in_count / rat;
644 	} else
645 		rat = (in_count << 8) / bytes_out;	/* 8 fractional bits. */
646 	if (rat > ratio)
647 		ratio = rat;
648 	else {
649 		ratio = 0;
650 		cl_hash(zs, (count_int) hsize);
651 		free_ent = FIRST;
652 		clear_flg = 1;
653 		if (output(zs, (code_int) CLEAR) == -1)
654 			return (-1);
655 	}
656 	return (0);
657 }
658 
659 static void
660 cl_hash(zs, cl_hsize)			/* Reset code table. */
661 	struct s_zstate *zs;
662 	register count_int cl_hsize;
663 {
664 	register count_int *htab_p;
665 	register long i, m1;
666 
667 	m1 = -1;
668 	htab_p = htab + cl_hsize;
669 	i = cl_hsize - 16;
670 	do {			/* Might use Sys V memset(3) here. */
671 		*(htab_p - 16) = m1;
672 		*(htab_p - 15) = m1;
673 		*(htab_p - 14) = m1;
674 		*(htab_p - 13) = m1;
675 		*(htab_p - 12) = m1;
676 		*(htab_p - 11) = m1;
677 		*(htab_p - 10) = m1;
678 		*(htab_p - 9) = m1;
679 		*(htab_p - 8) = m1;
680 		*(htab_p - 7) = m1;
681 		*(htab_p - 6) = m1;
682 		*(htab_p - 5) = m1;
683 		*(htab_p - 4) = m1;
684 		*(htab_p - 3) = m1;
685 		*(htab_p - 2) = m1;
686 		*(htab_p - 1) = m1;
687 		htab_p -= 16;
688 	} while ((i -= 16) >= 0);
689 	for (i += 16; i > 0; i--)
690 		*--htab_p = m1;
691 }
692 
693 FILE *
694 zopen(fname, mode, bits)
695 	const char *fname, *mode;
696 	int bits;
697 {
698 	struct s_zstate *zs;
699 
700 	if (mode[0] != 'r' && mode[0] != 'w' || mode[1] != '\0' ||
701 	    bits < 0 || bits > BITS) {
702 		errno = EINVAL;
703 		return (NULL);
704 	}
705 
706 	if ((zs = calloc(1, sizeof(struct s_zstate))) == NULL)
707 		return (NULL);
708 
709 	maxbits = bits ? bits : BITS;	/* User settable max # bits/code. */
710 	maxmaxcode = 1 << BITS;		/* Should NEVER generate this code. */
711 	hsize = HSIZE;			/* For dynamic table sizing. */
712 	free_ent = 0;			/* First unused entry. */
713 	block_compress = BLOCK_MASK;
714 	clear_flg = 0;
715 	ratio = 0;
716 	checkpoint = CHECK_GAP;
717 	in_count = 1;			/* Length of input. */
718 	out_count = 0;			/* # of codes output (for debugging). */
719 	state = S_START;
720 	roffset = 0;
721 	size = 0;
722 
723 	/*
724 	 * Layering compress on top of stdio in order to provide buffering,
725 	 * and ensure that reads and write work with the data specified.
726 	 */
727 	if ((fp = fopen(fname, mode)) == NULL) {
728 		free(zs);
729 		return (NULL);
730 	}
731 	switch (*mode) {
732 	case 'r':
733 		zmode = 'r';
734 		return (funopen(zs, zread, NULL, NULL, zclose));
735 	case 'w':
736 		zmode = 'w';
737 		return (funopen(zs, NULL, zwrite, NULL, zclose));
738 	}
739 	/* NOTREACHED */
740 }
741