xref: /netbsd-src/usr.bin/compress/zopen.c (revision 23c8222edbfb0f0932d88a8351d3a0cf817dfb9e)
1 /*	$NetBSD: zopen.c,v 1.8 2003/08/07 11:13:29 agc 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. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  */
35 
36 #if defined(LIBC_SCCS) && !defined(lint)
37 #if 0
38 static char sccsid[] = "@(#)zopen.c	8.1 (Berkeley) 6/27/93";
39 #else
40 static char rcsid[] = "$NetBSD: zopen.c,v 1.8 2003/08/07 11:13:29 agc Exp $";
41 #endif
42 #endif /* LIBC_SCCS and not lint */
43 
44 /*-
45  * fcompress.c - File compression ala IEEE Computer, June 1984.
46  *
47  * Compress authors:
48  *		Spencer W. Thomas	(decvax!utah-cs!thomas)
49  *		Jim McKie		(decvax!mcvax!jim)
50  *		Steve Davies		(decvax!vax135!petsd!peora!srd)
51  *		Ken Turkowski		(decvax!decwrl!turtlevax!ken)
52  *		James A. Woods		(decvax!ihnp4!ames!jaw)
53  *		Joe Orost		(decvax!vax135!petsd!joe)
54  *
55  * Cleaned up and converted to library returning I/O streams by
56  * Diomidis Spinellis <dds@doc.ic.ac.uk>.
57  *
58  * zopen(filename, mode, bits)
59  *	Returns a FILE * that can be used for read or write.  The modes
60  *	supported are only "r" and "w".  Seeking is not allowed.  On
61  *	reading the file is decompressed, on writing it is compressed.
62  *	The output is compatible with compress(1) with 16 bit tables.
63  *	Any file produced by compress(1) can be read.
64  */
65 
66 #include <sys/param.h>
67 #include <sys/stat.h>
68 
69 #include <ctype.h>
70 #include <errno.h>
71 #include <signal.h>
72 #include <stdio.h>
73 #include <stdlib.h>
74 #include <string.h>
75 #include <unistd.h>
76 
77 #define	BITS		16		/* Default bits. */
78 #define	HSIZE		69001		/* 95% occupancy */
79 
80 /* A code_int must be able to hold 2**BITS values of type int, and also -1. */
81 typedef long code_int;
82 typedef long count_int;
83 
84 typedef u_char char_type;
85 static char_type magic_header[] =
86 	{'\037', '\235'};		/* 1F 9D */
87 
88 #define	BIT_MASK	0x1f		/* Defines for third byte of header. */
89 #define	BLOCK_MASK	0x80
90 
91 /*
92  * Masks 0x40 and 0x20 are free.  I think 0x20 should mean that there is
93  * a fourth header byte (for expansion).
94  */
95 #define	INIT_BITS 9			/* Initial number of bits/code. */
96 
97 #define	MAXCODE(n_bits)	((1 << (n_bits)) - 1)
98 
99 struct s_zstate {
100 	FILE *zs_fp;			/* File stream for I/O */
101 	char zs_mode;			/* r or w */
102 	enum {
103 		S_START, S_MIDDLE, S_EOF
104 	} zs_state;			/* State of computation */
105 	int zs_n_bits;			/* Number of bits/code. */
106 	int zs_maxbits;			/* User settable max # bits/code. */
107 	code_int zs_maxcode;		/* Maximum code, given n_bits. */
108 	code_int zs_maxmaxcode;		/* Should NEVER generate this code. */
109 	count_int zs_htab [HSIZE];
110 	u_short zs_codetab [HSIZE];
111 	code_int zs_hsize;		/* For dynamic table sizing. */
112 	code_int zs_free_ent;		/* First unused entry. */
113 	/*
114 	 * Block compression parameters -- after all codes are used up,
115 	 * and compression rate changes, start over.
116 	 */
117 	int zs_block_compress;
118 	int zs_clear_flg;
119 	long zs_ratio;
120 	count_int zs_checkpoint;
121 	int zs_offset;
122 	long zs_in_count;		/* Length of input. */
123 	long zs_bytes_out;		/* Length of compressed output. */
124 	long zs_out_count;		/* # of codes output (for debugging). */
125 	char_type zs_buf[BITS];
126 	union {
127 		struct {
128 			long zs_fcode;
129 			code_int zs_ent;
130 			code_int zs_hsize_reg;
131 			int zs_hshift;
132 		} w;			/* Write paramenters */
133 		struct {
134 			char_type *zs_stackp;
135 			int zs_finchar;
136 			code_int zs_code, zs_oldcode, zs_incode;
137 			int zs_roffset, zs_size;
138 			char_type zs_gbuf[BITS];
139 		} r;			/* Read parameters */
140 	} u;
141 };
142 
143 /* Definitions to retain old variable names */
144 #define	fp		zs->zs_fp
145 #define	zmode		zs->zs_mode
146 #define	state		zs->zs_state
147 #define	n_bits		zs->zs_n_bits
148 #define	maxbits		zs->zs_maxbits
149 #define	maxcode		zs->zs_maxcode
150 #define	maxmaxcode	zs->zs_maxmaxcode
151 #define	htab		zs->zs_htab
152 #define	codetab		zs->zs_codetab
153 #define	hsize		zs->zs_hsize
154 #define	free_ent	zs->zs_free_ent
155 #define	block_compress	zs->zs_block_compress
156 #define	clear_flg	zs->zs_clear_flg
157 #define	ratio		zs->zs_ratio
158 #define	checkpoint	zs->zs_checkpoint
159 #define	offset		zs->zs_offset
160 #define	in_count	zs->zs_in_count
161 #define	bytes_out	zs->zs_bytes_out
162 #define	out_count	zs->zs_out_count
163 #define	buf		zs->zs_buf
164 #define	fcode		zs->u.w.zs_fcode
165 #define	hsize_reg	zs->u.w.zs_hsize_reg
166 #define	ent		zs->u.w.zs_ent
167 #define	hshift		zs->u.w.zs_hshift
168 #define	stackp		zs->u.r.zs_stackp
169 #define	finchar		zs->u.r.zs_finchar
170 #define	code		zs->u.r.zs_code
171 #define	oldcode		zs->u.r.zs_oldcode
172 #define	incode		zs->u.r.zs_incode
173 #define	roffset		zs->u.r.zs_roffset
174 #define	size		zs->u.r.zs_size
175 #define	gbuf		zs->u.r.zs_gbuf
176 
177 /*
178  * To save much memory, we overlay the table used by compress() with those
179  * used by decompress().  The tab_prefix table is the same size and type as
180  * the codetab.  The tab_suffix table needs 2**BITS characters.  We get this
181  * from the beginning of htab.  The output stack uses the rest of htab, and
182  * contains characters.  There is plenty of room for any possible stack
183  * (stack used to be 8000 characters).
184  */
185 
186 #define	htabof(i)	htab[i]
187 #define	codetabof(i)	codetab[i]
188 
189 #define	tab_prefixof(i)	codetabof(i)
190 #define	tab_suffixof(i)	((char_type *)(htab))[i]
191 #define	de_stack	((char_type *)&tab_suffixof(1 << BITS))
192 
193 #define	CHECK_GAP 10000		/* Ratio check interval. */
194 
195 /*
196  * the next two codes should not be changed lightly, as they must not
197  * lie within the contiguous general code space.
198  */
199 #define	FIRST	257		/* First free entry. */
200 #define	CLEAR	256		/* Table clear output code. */
201 
202 static int	cl_block(struct s_zstate *);
203 static void	cl_hash(struct s_zstate *, count_int);
204 static code_int	getcode(struct s_zstate *);
205 static int	output(struct s_zstate *, code_int);
206 static int	zclose(void *);
207 FILE	       *zopen(const char *, const char *, int);
208 static int	zread(void *, char *, int);
209 static int	zwrite(void *, const char *, int);
210 
211 /*-
212  * Algorithm from "A Technique for High Performance Data Compression",
213  * Terry A. Welch, IEEE Computer Vol 17, No 6 (June 1984), pp 8-19.
214  *
215  * Algorithm:
216  * 	Modified Lempel-Ziv method (LZW).  Basically finds common
217  * substrings and replaces them with a variable size code.  This is
218  * deterministic, and can be done on the fly.  Thus, the decompression
219  * procedure needs no input table, but tracks the way the table was built.
220  */
221 
222 /*-
223  * compress write
224  *
225  * Algorithm:  use open addressing double hashing (no chaining) on the
226  * prefix code / next character combination.  We do a variant of Knuth's
227  * algorithm D (vol. 3, sec. 6.4) along with G. Knott's relatively-prime
228  * secondary probe.  Here, the modular division first probe is gives way
229  * to a faster exclusive-or manipulation.  Also do block compression with
230  * an adaptive reset, whereby the code table is cleared when the compression
231  * ratio decreases, but after the table fills.  The variable-length output
232  * codes are re-sized at this point, and a special CLEAR code is generated
233  * for the decompressor.  Late addition:  construct the table according to
234  * file size for noticeable speed improvement on small files.  Please direct
235  * questions about this implementation to ames!jaw.
236  */
237 static int
238 zwrite(void *cookie, const char *wbp, int num)
239 {
240 	code_int i;
241 	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 << maxbits;
258 	if (fwrite(magic_header,
259 	    sizeof(char), sizeof(magic_header), fp) != sizeof(magic_header))
260 		return (-1);
261 	tmp = (u_char)(maxbits | 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(void *cookie)
327 {
328 	struct s_zstate *zs;
329 	int rval;
330 
331 	zs = cookie;
332 	if (zmode == 'w') {		/* Put out the final code. */
333 		if (output(zs, (code_int) ent) == -1) {
334 			(void)fclose(fp);
335 			free(zs);
336 			return (-1);
337 		}
338 		out_count++;
339 		if (output(zs, (code_int) - 1) == -1) {
340 			(void)fclose(fp);
341 			free(zs);
342 			return (-1);
343 		}
344 	}
345 	rval = fclose(fp) == EOF ? -1 : 0;
346 	free(zs);
347 	return (rval);
348 }
349 
350 /*-
351  * Output the given code.
352  * Inputs:
353  * 	code:	A n_bits-bit integer.  If == -1, then EOF.  This assumes
354  *		that n_bits =< (long)wordsize - 1.
355  * Outputs:
356  * 	Outputs code to the file.
357  * Assumptions:
358  *	Chars are 8 bits long.
359  * Algorithm:
360  * 	Maintain a BITS character long buffer (so that 8 codes will
361  * fit in it exactly).  Use the VAX insv instruction to insert each
362  * code in turn.  When the buffer fills up empty it and start over.
363  */
364 
365 static char_type lmask[9] =
366 	{0xff, 0xfe, 0xfc, 0xf8, 0xf0, 0xe0, 0xc0, 0x80, 0x00};
367 static char_type rmask[9] =
368 	{0x00, 0x01, 0x03, 0x07, 0x0f, 0x1f, 0x3f, 0x7f, 0xff};
369 
370 static int
371 output(struct s_zstate *zs, code_int ocode)
372 {
373 	int bits, r_off;
374 	char_type *bp;
375 
376 	r_off = offset;
377 	bits = n_bits;
378 	bp = buf;
379 	if (ocode >= 0) {
380 		/* Get to the first byte. */
381 		bp += (r_off >> 3);
382 		r_off &= 7;
383 		/*
384 		 * Since ocode is always >= 8 bits, only need to mask the first
385 		 * hunk on the left.
386 		 */
387 		*bp = (*bp & rmask[r_off]) | ((ocode << r_off) & lmask[r_off]);
388 		bp++;
389 		bits -= (8 - r_off);
390 		ocode >>= 8 - r_off;
391 		/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
392 		if (bits >= 8) {
393 			*bp++ = ocode;
394 			ocode >>= 8;
395 			bits -= 8;
396 		}
397 		/* Last bits. */
398 		if (bits)
399 			*bp = ocode;
400 		offset += n_bits;
401 		if (offset == (n_bits << 3)) {
402 			bp = buf;
403 			bits = n_bits;
404 			bytes_out += bits;
405 			if (fwrite(bp, sizeof(char), bits, fp) != bits)
406 				return (-1);
407 			bp += bits;
408 			bits = 0;
409 			offset = 0;
410 		}
411 		/*
412 		 * If the next entry is going to be too big for the ocode size,
413 		 * then increase it, if possible.
414 		 */
415 		if (free_ent > maxcode || (clear_flg > 0)) {
416 		       /*
417 			* Write the whole buffer, because the input side won't
418 			* discover the size increase until after it has read it.
419 			*/
420 			if (offset > 0) {
421 				if (fwrite(buf, 1, n_bits, fp) != n_bits)
422 					return (-1);
423 				bytes_out += n_bits;
424 			}
425 			offset = 0;
426 
427 			if (clear_flg) {
428 				maxcode = MAXCODE(n_bits = INIT_BITS);
429 				clear_flg = 0;
430 			} else {
431 				n_bits++;
432 				if (n_bits == maxbits)
433 					maxcode = maxmaxcode;
434 				else
435 					maxcode = MAXCODE(n_bits);
436 			}
437 		}
438 	} else {
439 		/* At EOF, write the rest of the buffer. */
440 		if (offset > 0) {
441 			offset = (offset + 7) / 8;
442 			if (fwrite(buf, 1, offset, fp) != offset)
443 				return (-1);
444 			bytes_out += offset;
445 		}
446 		offset = 0;
447 	}
448 	return (0);
449 }
450 
451 /*
452  * Decompress read.  This routine adapts to the codes in the file building
453  * the "string" table on-the-fly; requiring no table to be stored in the
454  * compressed file.  The tables used herein are shared with those of the
455  * compress() routine.  See the definitions above.
456  */
457 static int
458 zread(void *cookie, char *rbp, int num)
459 {
460 	u_int count;
461 	struct s_zstate *zs;
462 	u_char *bp, header[3];
463 
464 	if (num == 0)
465 		return (0);
466 
467 	zs = cookie;
468 	count = num;
469 	bp = (u_char *)rbp;
470 	switch (state) {
471 	case S_START:
472 		state = S_MIDDLE;
473 		break;
474 	case S_MIDDLE:
475 		goto middle;
476 	case S_EOF:
477 		goto eof;
478 	}
479 
480 	/* Check the magic number */
481 	if (fread(header,
482 	    sizeof(char), sizeof(header), fp) != sizeof(header) ||
483 	    memcmp(header, magic_header, sizeof(magic_header)) != 0) {
484 		errno = EFTYPE;
485 		return (-1);
486 	}
487 	maxbits = header[2];	/* Set -b from file. */
488 	block_compress = maxbits & BLOCK_MASK;
489 	maxbits &= BIT_MASK;
490 	maxmaxcode = 1L << maxbits;
491 	if (maxbits > BITS) {
492 		errno = EFTYPE;
493 		return (-1);
494 	}
495 	/* As above, initialize the first 256 entries in the table. */
496 	maxcode = MAXCODE(n_bits = INIT_BITS);
497 	for (code = 255; code >= 0; code--) {
498 		tab_prefixof(code) = 0;
499 		tab_suffixof(code) = (char_type) code;
500 	}
501 	free_ent = block_compress ? FIRST : 256;
502 
503 	finchar = oldcode = getcode(zs);
504 	if (oldcode == -1)	/* EOF already? */
505 		return (0);	/* Get out of here */
506 
507 	/* First code must be 8 bits = char. */
508 	*bp++ = (u_char)finchar;
509 	count--;
510 	stackp = de_stack;
511 
512 	while ((code = getcode(zs)) > -1) {
513 
514 		if ((code == CLEAR) && block_compress) {
515 			for (code = 255; code >= 0; code--)
516 				tab_prefixof(code) = 0;
517 			clear_flg = 1;
518 			free_ent = FIRST - 1;
519 			if ((code = getcode(zs)) == -1)	/* O, untimely death! */
520 				break;
521 		}
522 		incode = code;
523 
524 		/* Special case for KwKwK string. */
525 		if (code >= free_ent) {
526 			*stackp++ = finchar;
527 			code = oldcode;
528 		}
529 
530 		/* Generate output characters in reverse order. */
531 		while (code >= 256) {
532 			*stackp++ = tab_suffixof(code);
533 			code = tab_prefixof(code);
534 		}
535 		*stackp++ = finchar = tab_suffixof(code);
536 
537 		/* And put them out in forward order.  */
538 middle:		do {
539 			if (count-- == 0)
540 				return (num);
541 			*bp++ = *--stackp;
542 		} while (stackp > de_stack);
543 
544 		/* Generate the new entry. */
545 		if ((code = free_ent) < maxmaxcode) {
546 			tab_prefixof(code) = (u_short) oldcode;
547 			tab_suffixof(code) = finchar;
548 			free_ent = code + 1;
549 		}
550 
551 		/* Remember previous code. */
552 		oldcode = incode;
553 	}
554 	state = S_EOF;
555 eof:	return (num - count);
556 }
557 
558 /*-
559  * Read one code from the standard input.  If EOF, return -1.
560  * Inputs:
561  * 	stdin
562  * Outputs:
563  * 	code or -1 is returned.
564  */
565 static code_int
566 getcode(struct s_zstate *zs)
567 {
568 	code_int gcode;
569 	int r_off, bits;
570 	char_type *bp;
571 
572 	bp = gbuf;
573 	if (clear_flg > 0 || roffset >= size || free_ent > maxcode) {
574 		/*
575 		 * If the next entry will be too big for the current gcode
576 		 * size, then we must increase the size.  This implies reading
577 		 * a new buffer full, too.
578 		 */
579 		if (free_ent > maxcode) {
580 			n_bits++;
581 			if (n_bits == maxbits)	/* Won't get any bigger now. */
582 				maxcode = maxmaxcode;
583 			else
584 				maxcode = MAXCODE(n_bits);
585 		}
586 		if (clear_flg > 0) {
587 			maxcode = MAXCODE(n_bits = INIT_BITS);
588 			clear_flg = 0;
589 		}
590 		size = fread(gbuf, 1, n_bits, fp);
591 		if (size <= 0)			/* End of file. */
592 			return (-1);
593 		roffset = 0;
594 		/* Round size down to integral number of codes. */
595 		size = (size << 3) - (n_bits - 1);
596 	}
597 	r_off = roffset;
598 	bits = n_bits;
599 
600 	/* Get to the first byte. */
601 	bp += (r_off >> 3);
602 	r_off &= 7;
603 
604 	/* Get first part (low order bits). */
605 	gcode = (*bp++ >> r_off);
606 	bits -= (8 - r_off);
607 	r_off = 8 - r_off;	/* Now, roffset into gcode word. */
608 
609 	/* Get any 8 bit parts in the middle (<=1 for up to 16 bits). */
610 	if (bits >= 8) {
611 		gcode |= *bp++ << r_off;
612 		r_off += 8;
613 		bits -= 8;
614 	}
615 
616 	/* High order bits. */
617 	gcode |= (*bp & rmask[bits]) << r_off;
618 	roffset += n_bits;
619 
620 	return (gcode);
621 }
622 
623 static int
624 cl_block(struct s_zstate *zs)		/* Table clear for block compress. */
625 {
626 	long rat;
627 
628 	checkpoint = in_count + CHECK_GAP;
629 
630 	if (in_count > 0x007fffff) {	/* Shift will overflow. */
631 		rat = bytes_out >> 8;
632 		if (rat == 0)		/* Don't divide by zero. */
633 			rat = 0x7fffffff;
634 		else
635 			rat = in_count / rat;
636 	} else
637 		rat = (in_count << 8) / bytes_out;	/* 8 fractional bits. */
638 	if (rat > ratio)
639 		ratio = rat;
640 	else {
641 		ratio = 0;
642 		cl_hash(zs, (count_int) hsize);
643 		free_ent = FIRST;
644 		clear_flg = 1;
645 		if (output(zs, (code_int) CLEAR) == -1)
646 			return (-1);
647 	}
648 	return (0);
649 }
650 
651 static void
652 cl_hash(struct s_zstate *zs, count_int cl_hsize)	/* Reset code table. */
653 {
654 	count_int *htab_p;
655 	long i, m1;
656 
657 	m1 = -1;
658 	htab_p = htab + cl_hsize;
659 	i = cl_hsize - 16;
660 	do {			/* Might use Sys V memset(3) here. */
661 		*(htab_p - 16) = m1;
662 		*(htab_p - 15) = m1;
663 		*(htab_p - 14) = m1;
664 		*(htab_p - 13) = m1;
665 		*(htab_p - 12) = m1;
666 		*(htab_p - 11) = m1;
667 		*(htab_p - 10) = m1;
668 		*(htab_p - 9) = m1;
669 		*(htab_p - 8) = m1;
670 		*(htab_p - 7) = m1;
671 		*(htab_p - 6) = m1;
672 		*(htab_p - 5) = m1;
673 		*(htab_p - 4) = m1;
674 		*(htab_p - 3) = m1;
675 		*(htab_p - 2) = m1;
676 		*(htab_p - 1) = m1;
677 		htab_p -= 16;
678 	} while ((i -= 16) >= 0);
679 	for (i += 16; i > 0; i--)
680 		*--htab_p = m1;
681 }
682 
683 FILE *
684 zopen(const char *fname, const char *mode, int bits)
685 {
686 	struct s_zstate *zs;
687 
688 	if ((mode[0] != 'r' && mode[0] != 'w') || mode[1] != '\0' ||
689 	    bits < 0 || bits > BITS) {
690 		errno = EINVAL;
691 		return (NULL);
692 	}
693 
694 	if ((zs = calloc(1, sizeof(struct s_zstate))) == NULL)
695 		return (NULL);
696 
697 	maxbits = bits ? bits : BITS;	/* User settable max # bits/code. */
698 	maxmaxcode = 1 << maxbits;	/* Should NEVER generate this code. */
699 	hsize = HSIZE;			/* For dynamic table sizing. */
700 	free_ent = 0;			/* First unused entry. */
701 	block_compress = BLOCK_MASK;
702 	clear_flg = 0;
703 	ratio = 0;
704 	checkpoint = CHECK_GAP;
705 	in_count = 1;			/* Length of input. */
706 	out_count = 0;			/* # of codes output (for debugging). */
707 	state = S_START;
708 	roffset = 0;
709 	size = 0;
710 
711 	/*
712 	 * Layering compress on top of stdio in order to provide buffering,
713 	 * and ensure that reads and write work with the data specified.
714 	 */
715 	if ((fp = fopen(fname, mode)) == NULL) {
716 		free(zs);
717 		return (NULL);
718 	}
719 	switch (*mode) {
720 	case 'r':
721 		zmode = 'r';
722 		return (funopen(zs, zread, NULL, NULL, zclose));
723 	case 'w':
724 		zmode = 'w';
725 		return (funopen(zs, NULL, zwrite, NULL, zclose));
726 	}
727 	/* NOTREACHED */
728 	return (NULL);
729 }
730