]> git.saurik.com Git - apple/libc.git/blob - regex/regcomp.c
projects / apple / libc.git / blob
? search:


c26826d24ee3b7b57d1528581d4a71dbcb0fa0fa
[apple/libc.git] / regex / regcomp.c
1 /*
2 * Copyright (c) 1999 Apple Computer, Inc. All rights reserved.
3 *
4 * @APPLE_LICENSE_HEADER_START@
5 *
6 * Copyright (c) 1999-2003 Apple Computer, Inc. All Rights Reserved.
7 *
8 * This file contains Original Code and/or Modifications of Original Code
9 * as defined in and that are subject to the Apple Public Source License
10 * Version 2.0 (the 'License'). You may not use this file except in
11 * compliance with the License. Please obtain a copy of the License at
12 * http://www.opensource.apple.com/apsl/ and read it before using this
13 * file.
14 *
15 * The Original Code and all software distributed under the License are
16 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
17 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
18 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
20 * Please see the License for the specific language governing rights and
21 * limitations under the License.
22 *
23 * @APPLE_LICENSE_HEADER_END@
24 */
25 /*
26 * Copyright (c) 1992, 1993, 1994
27 * The Regents of the University of California. All rights reserved.
28 *
29 * This code is derived from software contributed to Berkeley by
30 * Henry Spencer.
31 *
32 * Redistribution and use in source and binary forms, with or without
33 * modification, are permitted provided that the following conditions
34 * are met:
35 * 1. Redistributions of source code must retain the above copyright
36 * notice, this list of conditions and the following disclaimer.
37 * 2. Redistributions in binary form must reproduce the above copyright
38 * notice, this list of conditions and the following disclaimer in the
39 * documentation and/or other materials provided with the distribution.
40 * 3. All advertising materials mentioning features or use of this software
41 * must display the following acknowledgement:
42 * This product includes software developed by the University of
43 * California, Berkeley and its contributors.
44 * 4. Neither the name of the University nor the names of its contributors
45 * may be used to endorse or promote products derived from this software
46 * without specific prior written permission.
47 *
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58 * SUCH DAMAGE.
59 */
60
61 #include <sys/types.h>
62 #include <stdio.h>
63 #include <string.h>
64 #include <ctype.h>
65 #include <limits.h>
66 #include <stdlib.h>
67 #include <regex.h>
68
69 #include "utils.h"
70 #include "regex2.h"
71
72 #include "cclass.h"
73 #include "cname.h"
74
75 /*
76 * parse structure, passed up and down to avoid global variables and
77 * other clumsinesses
78 */
79 struct parse {
80 char *next; /* next character in RE */
81 char *end; /* end of string (-> NUL normally) */
82 int error; /* has an error been seen? */
83 sop *strip; /* malloced strip */
84 sopno ssize; /* malloced strip size (allocated) */
85 sopno slen; /* malloced strip length (used) */
86 int ncsalloc; /* number of csets allocated */
87 struct re_guts *g;
88 # define NPAREN 10 /* we need to remember () 1-9 for back refs */
89 sopno pbegin[NPAREN]; /* -> ( ([0] unused) */
90 sopno pend[NPAREN]; /* -> ) ([0] unused) */
91 };
92
93 /* ========= begin header generated by ./mkh ========= */
94 #ifdef __cplusplus
95 extern "C" {
96 #endif
97
98 /* === regcomp.c === */
99 static void p_ere __P((struct parse *p, int stop));
100 static void p_ere_exp __P((struct parse *p));
101 static void p_str __P((struct parse *p));
102 static void p_bre __P((struct parse *p, int end1, int end2));
103 static int p_simp_re __P((struct parse *p, int starordinary));
104 static int p_count __P((struct parse *p));
105 static void p_bracket __P((struct parse *p));
106 static void p_b_term __P((struct parse *p, cset *cs));
107 static void p_b_cclass __P((struct parse *p, cset *cs));
108 static void p_b_eclass __P((struct parse *p, cset *cs));
109 static char p_b_symbol __P((struct parse *p));
110 static char p_b_coll_elem __P((struct parse *p, int endc));
111 static char othercase __P((int ch));
112 static void bothcases __P((struct parse *p, int ch));
113 static void ordinary __P((struct parse *p, int ch));
114 static void nonnewline __P((struct parse *p));
115 static void repeat __P((struct parse *p, sopno start, int from, int to));
116 static int seterr __P((struct parse *p, int e));
117 static cset *allocset __P((struct parse *p));
118 static void freeset __P((struct parse *p, cset *cs));
119 static int freezeset __P((struct parse *p, cset *cs));
120 static int firstch __P((struct parse *p, cset *cs));
121 static int nch __P((struct parse *p, cset *cs));
122 static void mcadd __P((struct parse *p, cset *cs, char *cp));
123 static void mcsub __P((cset *cs, char *cp));
124 static int mcin __P((cset *cs, char *cp));
125 static char *mcfind __P((cset *cs, char *cp));
126 static void mcinvert __P((struct parse *p, cset *cs));
127 static void mccase __P((struct parse *p, cset *cs));
128 static int isinsets __P((struct re_guts *g, int c));
129 static int samesets __P((struct re_guts *g, int c1, int c2));
130 static void categorize __P((struct parse *p, struct re_guts *g));
131 static sopno dupl __P((struct parse *p, sopno start, sopno finish));
132 static void doemit __P((struct parse *p, sop op, size_t opnd));
133 static void doinsert __P((struct parse *p, sop op, size_t opnd, sopno pos));
134 static void dofwd __P((struct parse *p, sopno pos, sop value));
135 static void enlarge __P((struct parse *p, sopno size));
136 static void stripsnug __P((struct parse *p, struct re_guts *g));
137 static void findmust __P((struct parse *p, struct re_guts *g));
138 static sopno pluscount __P((struct parse *p, struct re_guts *g));
139
140 #ifdef __cplusplus
141 }
142 #endif
143 /* ========= end header generated by ./mkh ========= */
144
145 static char nuls[10]; /* place to point scanner in event of error */
146
147 /*
148 * macros for use with parse structure
149 * BEWARE: these know that the parse structure is named `p' !!!
150 */
151 #define PEEK() (*p->next)
152 #define PEEK2() (*(p->next+1))
153 #define MORE() (p->next < p->end)
154 #define MORE2() (p->next+1 < p->end)
155 #define SEE(c) (MORE() && PEEK() == (c))
156 #define SEETWO(a, b) (MORE() && MORE2() && PEEK() == (a) && PEEK2() == (b))
157 #define EAT(c) ((SEE(c)) ? (NEXT(), 1) : 0)
158 #define EATTWO(a, b) ((SEETWO(a, b)) ? (NEXT2(), 1) : 0)
159 #define NEXT() (p->next++)
160 #define NEXT2() (p->next += 2)
161 #define NEXTn(n) (p->next += (n))
162 #define GETNEXT() (*p->next++)
163 #define SETERROR(e) seterr(p, (e))
164 #define REQUIRE(co, e) ((co) || SETERROR(e))
165 #define MUSTSEE(c, e) (REQUIRE(MORE() && PEEK() == (c), e))
166 #define MUSTEAT(c, e) (REQUIRE(MORE() && GETNEXT() == (c), e))
167 #define MUSTNOTSEE(c, e) (REQUIRE(!MORE() || PEEK() != (c), e))
168 #define EMIT(op, sopnd) doemit(p, (sop)(op), (size_t)(sopnd))
169 #define INSERT(op, pos) doinsert(p, (sop)(op), HERE()-(pos)+1, pos)
170 #define AHEAD(pos) dofwd(p, pos, HERE()-(pos))
171 #define ASTERN(sop, pos) EMIT(sop, HERE()-pos)
172 #define HERE() (p->slen)
173 #define THERE() (p->slen - 1)
174 #define THERETHERE() (p->slen - 2)
175 #define DROP(n) (p->slen -= (n))
176
177 #ifndef NDEBUG
178 static int never = 0; /* for use in asserts; shuts lint up */
179 #else
180 #define never 0 /* some <assert.h>s have bugs too */
181 #endif
182
183 /*
184 - regcomp - interface for parser and compilation
185 = extern int regcomp(regex_t *, const char *, int);
186 = #define REG_BASIC 0000
187 = #define REG_EXTENDED 0001
188 = #define REG_ICASE 0002
189 = #define REG_NOSUB 0004
190 = #define REG_NEWLINE 0010
191 = #define REG_NOSPEC 0020
192 = #define REG_PEND 0040
193 = #define REG_DUMP 0200
194 */
195 int /* 0 success, otherwise REG_something */
196 regcomp(preg, pattern, cflags)
197 regex_t *preg;
198 const char *pattern;
199 int cflags;
200 {
201 struct parse pa;
202 register struct re_guts *g;
203 register struct parse *p = &pa;
204 register int i;
205 register size_t len;
206 #ifdef REDEBUG
207 # define GOODFLAGS(f) (f)
208 #else
209 # define GOODFLAGS(f) ((f)&~REG_DUMP)
210 #endif
211
212 cflags = GOODFLAGS(cflags);
213 if ((cflags&REG_EXTENDED) && (cflags&REG_NOSPEC))
214 return(REG_INVARG);
215
216 if (cflags&REG_PEND) {
217 if (preg->re_endp < pattern)
218 return(REG_INVARG);
219 len = preg->re_endp - pattern;
220 } else
221 len = strlen((char *)pattern);
222
223 /* do the mallocs early so failure handling is easy */
224 g = (struct re_guts *)malloc(sizeof(struct re_guts) +
225 (NC-1)*sizeof(cat_t));
226 if (g == NULL)
227 return(REG_ESPACE);
228 p->ssize = len/(size_t)2*(size_t)3 + (size_t)1; /* ugh */
229 p->strip = (sop *)malloc(p->ssize * sizeof(sop));
230 p->slen = 0;
231 if (p->strip == NULL) {
232 free((char *)g);
233 return(REG_ESPACE);
234 }
235
236 /* set things up */
237 p->g = g;
238 p->next = (char *)pattern; /* convenience; we do not modify it */
239 p->end = p->next + len;
240 p->error = 0;
241 p->ncsalloc = 0;
242 for (i = 0; i < NPAREN; i++) {
243 p->pbegin[i] = 0;
244 p->pend[i] = 0;
245 }
246 g->csetsize = NC;
247 g->sets = NULL;
248 g->setbits = NULL;
249 g->ncsets = 0;
250 g->cflags = cflags;
251 g->iflags = 0;
252 g->nbol = 0;
253 g->neol = 0;
254 g->must = NULL;
255 g->mlen = 0;
256 g->nsub = 0;
257 g->ncategories = 1; /* category 0 is "everything else" */
258 g->categories = &g->catspace[-(CHAR_MIN)];
259 (void) memset((char *)g->catspace, 0, NC*sizeof(cat_t));
260 g->backrefs = 0;
261
262 /* do it */
263 EMIT(OEND, 0);
264 g->firststate = THERE();
265 if (cflags&REG_EXTENDED)
266 p_ere(p, OUT);
267 else if (cflags&REG_NOSPEC)
268 p_str(p);
269 else
270 p_bre(p, OUT, OUT);
271 EMIT(OEND, 0);
272 g->laststate = THERE();
273
274 /* tidy up loose ends and fill things in */
275 categorize(p, g);
276 stripsnug(p, g);
277 findmust(p, g);
278 g->nplus = pluscount(p, g);
279 g->magic = MAGIC2;
280 preg->re_nsub = g->nsub;
281 preg->re_g = g;
282 preg->re_magic = MAGIC1;
283 #ifndef REDEBUG
284 /* not debugging, so can't rely on the assert() in regexec() */
285 if (g->iflags&BAD)
286 SETERROR(REG_ASSERT);
287 #endif
288
289 /* win or lose, we're done */
290 if (p->error != 0) /* lose */
291 regfree(preg);
292 return(p->error);
293 }
294
295 /*
296 - p_ere - ERE parser top level, concatenation and alternation
297 == static void p_ere(register struct parse *p, int stop);
298 */
299 static void
300 p_ere(p, stop)
301 register struct parse *p;
302 int stop; /* character this ERE should end at */
303 {
304 register char c;
305 register sopno prevback;
306 register sopno prevfwd;
307 register sopno conc;
308 register int first = 1; /* is this the first alternative? */
309
310 for (;;) {
311 /* do a bunch of concatenated expressions */
312 conc = HERE();
313 while (MORE() && (c = PEEK()) != '|' && c != stop)
314 p_ere_exp(p);
315 REQUIRE(HERE() != conc, REG_EMPTY); /* require nonempty */
316
317 if (!EAT('|'))
318 break; /* NOTE BREAK OUT */
319
320 if (first) {
321 INSERT(OCH_, conc); /* offset is wrong */
322 prevfwd = conc;
323 prevback = conc;
324 first = 0;
325 }
326 ASTERN(OOR1, prevback);
327 prevback = THERE();
328 AHEAD(prevfwd); /* fix previous offset */
329 prevfwd = HERE();
330 EMIT(OOR2, 0); /* offset is very wrong */
331 }
332
333 if (!first) { /* tail-end fixups */
334 AHEAD(prevfwd);
335 ASTERN(O_CH, prevback);
336 }
337
338 assert(!MORE() || SEE(stop));
339 }
340
341 /*
342 - p_ere_exp - parse one subERE, an atom possibly followed by a repetition op
343 == static void p_ere_exp(register struct parse *p);
344 */
345 static void
346 p_ere_exp(p)
347 register struct parse *p;
348 {
349 register char c;
350 register sopno pos;
351 register int count;
352 register int count2;
353 register sopno subno;
354 int wascaret = 0;
355
356 assert(MORE()); /* caller should have ensured this */
357 c = GETNEXT();
358
359 pos = HERE();
360 switch (c) {
361 case '(':
362 REQUIRE(MORE(), REG_EPAREN);
363 p->g->nsub++;
364 subno = p->g->nsub;
365 if (subno < NPAREN)
366 p->pbegin[subno] = HERE();
367 EMIT(OLPAREN, subno);
368 if (!SEE(')'))
369 p_ere(p, ')');
370 if (subno < NPAREN) {
371 p->pend[subno] = HERE();
372 assert(p->pend[subno] != 0);
373 }
374 EMIT(ORPAREN, subno);
375 MUSTEAT(')', REG_EPAREN);
376 break;
377 #ifndef POSIX_MISTAKE
378 case ')': /* happens only if no current unmatched ( */
379 /*
380 * You may ask, why the ifndef? Because I didn't notice
381 * this until slightly too late for 1003.2, and none of the
382 * other 1003.2 regular-expression reviewers noticed it at
383 * all. So an unmatched ) is legal POSIX, at least until
384 * we can get it fixed.
385 */
386 SETERROR(REG_EPAREN);
387 break;
388 #endif
389 case '^':
390 EMIT(OBOL, 0);
391 p->g->iflags |= USEBOL;
392 p->g->nbol++;
393 wascaret = 1;
394 break;
395 case '$':
396 EMIT(OEOL, 0);
397 p->g->iflags |= USEEOL;
398 p->g->neol++;
399 break;
400 case '|':
401 SETERROR(REG_EMPTY);
402 break;
403 case '*':
404 case '+':
405 case '?':
406 SETERROR(REG_BADRPT);
407 break;
408 case '.':
409 if (p->g->cflags&REG_NEWLINE)
410 nonnewline(p);
411 else
412 EMIT(OANY, 0);
413 break;
414 case '[':
415 p_bracket(p);
416 break;
417 case '\\':
418 REQUIRE(MORE(), REG_EESCAPE);
419 c = GETNEXT();
420 ordinary(p, c);
421 break;
422 case '{': /* okay as ordinary except if digit follows */
423 REQUIRE(!MORE() || !isdigit(PEEK()), REG_BADRPT);
424 /* FALLTHROUGH */
425 default:
426 ordinary(p, c);
427 break;
428 }
429
430 if (!MORE())
431 return;
432 c = PEEK();
433 /* we call { a repetition if followed by a digit */
434 if (!( c == '*' || c == '+' || c == '?' ||
435 (c == '{' && MORE2() && isdigit(PEEK2())) ))
436 return; /* no repetition, we're done */
437 NEXT();
438
439 REQUIRE(!wascaret, REG_BADRPT);
440 switch (c) {
441 case '*': /* implemented as +? */
442 /* this case does not require the (y|) trick, noKLUDGE */
443 INSERT(OPLUS_, pos);
444 ASTERN(O_PLUS, pos);
445 INSERT(OQUEST_, pos);
446 ASTERN(O_QUEST, pos);
447 break;
448 case '+':
449 INSERT(OPLUS_, pos);
450 ASTERN(O_PLUS, pos);
451 break;
452 case '?':
453 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
454 INSERT(OCH_, pos); /* offset slightly wrong */
455 ASTERN(OOR1, pos); /* this one's right */
456 AHEAD(pos); /* fix the OCH_ */
457 EMIT(OOR2, 0); /* offset very wrong... */
458 AHEAD(THERE()); /* ...so fix it */
459 ASTERN(O_CH, THERETHERE());
460 break;
461 case '{':
462 count = p_count(p);
463 if (EAT(',')) {
464 if (isdigit(PEEK())) {
465 count2 = p_count(p);
466 REQUIRE(count <= count2, REG_BADBR);
467 } else /* single number with comma */
468 count2 = INFINITY;
469 } else /* just a single number */
470 count2 = count;
471 repeat(p, pos, count, count2);
472 if (!EAT('}')) { /* error heuristics */
473 while (MORE() && PEEK() != '}')
474 NEXT();
475 REQUIRE(MORE(), REG_EBRACE);
476 SETERROR(REG_BADBR);
477 }
478 break;
479 }
480
481 if (!MORE())
482 return;
483 c = PEEK();
484 if (!( c == '*' || c == '+' || c == '?' ||
485 (c == '{' && MORE2() && isdigit(PEEK2())) ) )
486 return;
487 SETERROR(REG_BADRPT);
488 }
489
490 /*
491 - p_str - string (no metacharacters) "parser"
492 == static void p_str(register struct parse *p);
493 */
494 static void
495 p_str(p)
496 register struct parse *p;
497 {
498 REQUIRE(MORE(), REG_EMPTY);
499 while (MORE())
500 ordinary(p, GETNEXT());
501 }
502
503 /*
504 - p_bre - BRE parser top level, anchoring and concatenation
505 == static void p_bre(register struct parse *p, register int end1, \
506 == register int end2);
507 * Giving end1 as OUT essentially eliminates the end1/end2 check.
508 *
509 * This implementation is a bit of a kludge, in that a trailing $ is first
510 * taken as an ordinary character and then revised to be an anchor. The
511 * only undesirable side effect is that '$' gets included as a character
512 * category in such cases. This is fairly harmless; not worth fixing.
513 * The amount of lookahead needed to avoid this kludge is excessive.
514 */
515 static void
516 p_bre(p, end1, end2)
517 register struct parse *p;
518 register int end1; /* first terminating character */
519 register int end2; /* second terminating character */
520 {
521 register sopno start = HERE();
522 register int first = 1; /* first subexpression? */
523 register int wasdollar = 0;
524
525 if (EAT('^')) {
526 EMIT(OBOL, 0);
527 p->g->iflags |= USEBOL;
528 p->g->nbol++;
529 }
530 while (MORE() && !SEETWO(end1, end2)) {
531 wasdollar = p_simp_re(p, first);
532 first = 0;
533 }
534 if (wasdollar) { /* oops, that was a trailing anchor */
535 DROP(1);
536 EMIT(OEOL, 0);
537 p->g->iflags |= USEEOL;
538 p->g->neol++;
539 }
540
541 REQUIRE(HERE() != start, REG_EMPTY); /* require nonempty */
542 }
543
544 /*
545 - p_simp_re - parse a simple RE, an atom possibly followed by a repetition
546 == static int p_simp_re(register struct parse *p, int starordinary);
547 */
548 static int /* was the simple RE an unbackslashed $? */
549 p_simp_re(p, starordinary)
550 register struct parse *p;
551 int starordinary; /* is a leading * an ordinary character? */
552 {
553 register int c;
554 register int count;
555 register int count2;
556 register sopno pos;
557 register int i;
558 register sopno subno;
559 # define BACKSL (1<<CHAR_BIT)
560
561 pos = HERE(); /* repetion op, if any, covers from here */
562
563 assert(MORE()); /* caller should have ensured this */
564 c = GETNEXT();
565 if (c == '\\') {
566 REQUIRE(MORE(), REG_EESCAPE);
567 c = BACKSL | (unsigned char)GETNEXT();
568 }
569 switch (c) {
570 case '.':
571 if (p->g->cflags&REG_NEWLINE)
572 nonnewline(p);
573 else
574 EMIT(OANY, 0);
575 break;
576 case '[':
577 p_bracket(p);
578 break;
579 case BACKSL|'{':
580 SETERROR(REG_BADRPT);
581 break;
582 case BACKSL|'(':
583 p->g->nsub++;
584 subno = p->g->nsub;
585 if (subno < NPAREN)
586 p->pbegin[subno] = HERE();
587 EMIT(OLPAREN, subno);
588 /* the MORE here is an error heuristic */
589 if (MORE() && !SEETWO('\\', ')'))
590 p_bre(p, '\\', ')');
591 if (subno < NPAREN) {
592 p->pend[subno] = HERE();
593 assert(p->pend[subno] != 0);
594 }
595 EMIT(ORPAREN, subno);
596 REQUIRE(EATTWO('\\', ')'), REG_EPAREN);
597 break;
598 case BACKSL|')': /* should not get here -- must be user */
599 case BACKSL|'}':
600 SETERROR(REG_EPAREN);
601 break;
602 case BACKSL|'1':
603 case BACKSL|'2':
604 case BACKSL|'3':
605 case BACKSL|'4':
606 case BACKSL|'5':
607 case BACKSL|'6':
608 case BACKSL|'7':
609 case BACKSL|'8':
610 case BACKSL|'9':
611 i = (c&~BACKSL) - '0';
612 assert(i < NPAREN);
613 if (p->pend[i] != 0) {
614 assert(i <= p->g->nsub);
615 EMIT(OBACK_, i);
616 assert(p->pbegin[i] != 0);
617 assert(OP(p->strip[p->pbegin[i]]) == OLPAREN);
618 assert(OP(p->strip[p->pend[i]]) == ORPAREN);
619 (void) dupl(p, p->pbegin[i]+1, p->pend[i]);
620 EMIT(O_BACK, i);
621 } else
622 SETERROR(REG_ESUBREG);
623 p->g->backrefs = 1;
624 break;
625 case '*':
626 REQUIRE(starordinary, REG_BADRPT);
627 /* FALLTHROUGH */
628 default:
629 ordinary(p, c &~ BACKSL);
630 break;
631 }
632
633 if (EAT('*')) { /* implemented as +? */
634 /* this case does not require the (y|) trick, noKLUDGE */
635 INSERT(OPLUS_, pos);
636 ASTERN(O_PLUS, pos);
637 INSERT(OQUEST_, pos);
638 ASTERN(O_QUEST, pos);
639 } else if (EATTWO('\\', '{')) {
640 count = p_count(p);
641 if (EAT(',')) {
642 if (MORE() && isdigit(PEEK())) {
643 count2 = p_count(p);
644 REQUIRE(count <= count2, REG_BADBR);
645 } else /* single number with comma */
646 count2 = INFINITY;
647 } else /* just a single number */
648 count2 = count;
649 repeat(p, pos, count, count2);
650 if (!EATTWO('\\', '}')) { /* error heuristics */
651 while (MORE() && !SEETWO('\\', '}'))
652 NEXT();
653 REQUIRE(MORE(), REG_EBRACE);
654 SETERROR(REG_BADBR);
655 }
656 } else if (c == (unsigned char)'$') /* $ (but not \$) ends it */
657 return(1);
658
659 return(0);
660 }
661
662 /*
663 - p_count - parse a repetition count
664 == static int p_count(register struct parse *p);
665 */
666 static int /* the value */
667 p_count(p)
668 register struct parse *p;
669 {
670 register int count = 0;
671 register int ndigits = 0;
672
673 while (MORE() && isdigit(PEEK()) && count <= DUPMAX) {
674 count = count*10 + (GETNEXT() - '0');
675 ndigits++;
676 }
677
678 REQUIRE(ndigits > 0 && count <= DUPMAX, REG_BADBR);
679 return(count);
680 }
681
682 /*
683 - p_bracket - parse a bracketed character list
684 == static void p_bracket(register struct parse *p);
685 *
686 * Note a significant property of this code: if the allocset() did SETERROR,
687 * no set operations are done.
688 */
689 static void
690 p_bracket(p)
691 register struct parse *p;
692 {
693 register char c;
694 register cset *cs = allocset(p);
695 register int invert = 0;
696
697 /* Dept of Truly Sickening Special-Case Kludges */
698 if (p->next + 5 < p->end && strncmp(p->next, "[:<:]]", 6) == 0) {
699 EMIT(OBOW, 0);
700 NEXTn(6);
701 return;
702 }
703 if (p->next + 5 < p->end && strncmp(p->next, "[:>:]]", 6) == 0) {
704 EMIT(OEOW, 0);
705 NEXTn(6);
706 return;
707 }
708
709 if (EAT('^'))
710 invert++; /* make note to invert set at end */
711 if (EAT(']'))
712 CHadd(cs, ']');
713 else if (EAT('-'))
714 CHadd(cs, '-');
715 while (MORE() && PEEK() != ']' && !SEETWO('-', ']'))
716 p_b_term(p, cs);
717 if (EAT('-'))
718 CHadd(cs, '-');
719 MUSTEAT(']', REG_EBRACK);
720
721 if (p->error != 0) /* don't mess things up further */
722 return;
723
724 if (p->g->cflags&REG_ICASE) {
725 register int i;
726 register int ci;
727
728 for (i = p->g->csetsize - 1; i >= 0; i--)
729 if (CHIN(cs, i) && isalpha(i)) {
730 ci = othercase(i);
731 if (ci != i)
732 CHadd(cs, ci);
733 }
734 if (cs->multis != NULL)
735 mccase(p, cs);
736 }
737 if (invert) {
738 register int i;
739
740 for (i = p->g->csetsize - 1; i >= 0; i--)
741 if (CHIN(cs, i))
742 CHsub(cs, i);
743 else
744 CHadd(cs, i);
745 if (p->g->cflags&REG_NEWLINE)
746 CHsub(cs, '\n');
747 if (cs->multis != NULL)
748 mcinvert(p, cs);
749 }
750
751 assert(cs->multis == NULL); /* xxx */
752
753 if (nch(p, cs) == 1) { /* optimize singleton sets */
754 ordinary(p, firstch(p, cs));
755 freeset(p, cs);
756 } else
757 EMIT(OANYOF, freezeset(p, cs));
758 }
759
760 /*
761 - p_b_term - parse one term of a bracketed character list
762 == static void p_b_term(register struct parse *p, register cset *cs);
763 */
764 static void
765 p_b_term(p, cs)
766 register struct parse *p;
767 register cset *cs;
768 {
769 register char c;
770 register char start, finish;
771 register int i;
772
773 /* classify what we've got */
774 switch ((MORE()) ? PEEK() : '\0') {
775 case '[':
776 c = (MORE2()) ? PEEK2() : '\0';
777 break;
778 case '-':
779 SETERROR(REG_ERANGE);
780 return; /* NOTE RETURN */
781 break;
782 default:
783 c = '\0';
784 break;
785 }
786
787 switch (c) {
788 case ':': /* character class */
789 NEXT2();
790 REQUIRE(MORE(), REG_EBRACK);
791 c = PEEK();
792 REQUIRE(c != '-' && c != ']', REG_ECTYPE);
793 p_b_cclass(p, cs);
794 REQUIRE(MORE(), REG_EBRACK);
795 REQUIRE(EATTWO(':', ']'), REG_ECTYPE);
796 break;
797 case '=': /* equivalence class */
798 NEXT2();
799 REQUIRE(MORE(), REG_EBRACK);
800 c = PEEK();
801 REQUIRE(c != '-' && c != ']', REG_ECOLLATE);
802 p_b_eclass(p, cs);
803 REQUIRE(MORE(), REG_EBRACK);
804 REQUIRE(EATTWO('=', ']'), REG_ECOLLATE);
805 break;
806 default: /* symbol, ordinary character, or range */
807 /* xxx revision needed for multichar stuff */
808 start = p_b_symbol(p);
809 if (SEE('-') && MORE2() && PEEK2() != ']') {
810 /* range */
811 NEXT();
812 if (EAT('-'))
813 finish = '-';
814 else
815 finish = p_b_symbol(p);
816 } else
817 finish = start;
818 /* xxx what about signed chars here... */
819 REQUIRE(start <= finish, REG_ERANGE);
820 for (i = start; i <= finish; i++)
821 CHadd(cs, i);
822 break;
823 }
824 }
825
826 /*
827 - p_b_cclass - parse a character-class name and deal with it
828 == static void p_b_cclass(register struct parse *p, register cset *cs);
829 */
830 static void
831 p_b_cclass(p, cs)
832 register struct parse *p;
833 register cset *cs;
834 {
835 register char *sp = p->next;
836 register struct cclass *cp;
837 register size_t len;
838 register char *u;
839 register char c;
840
841 while (MORE() && isalpha(PEEK()))
842 NEXT();
843 len = p->next - sp;
844 for (cp = cclasses; cp->name != NULL; cp++)
845 if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
846 break;
847 if (cp->name == NULL) {
848 /* oops, didn't find it */
849 SETERROR(REG_ECTYPE);
850 return;
851 }
852
853 u = cp->chars;
854 while ((c = *u++) != '\0')
855 CHadd(cs, c);
856 for (u = cp->multis; *u != '\0'; u += strlen(u) + 1)
857 MCadd(p, cs, u);
858 }
859
860 /*
861 - p_b_eclass - parse an equivalence-class name and deal with it
862 == static void p_b_eclass(register struct parse *p, register cset *cs);
863 *
864 * This implementation is incomplete. xxx
865 */
866 static void
867 p_b_eclass(p, cs)
868 register struct parse *p;
869 register cset *cs;
870 {
871 register char c;
872
873 c = p_b_coll_elem(p, '=');
874 CHadd(cs, c);
875 }
876
877 /*
878 - p_b_symbol - parse a character or [..]ed multicharacter collating symbol
879 == static char p_b_symbol(register struct parse *p);
880 */
881 static char /* value of symbol */
882 p_b_symbol(p)
883 register struct parse *p;
884 {
885 register char value;
886
887 REQUIRE(MORE(), REG_EBRACK);
888 if (!EATTWO('[', '.'))
889 return(GETNEXT());
890
891 /* collating symbol */
892 value = p_b_coll_elem(p, '.');
893 REQUIRE(EATTWO('.', ']'), REG_ECOLLATE);
894 return(value);
895 }
896
897 /*
898 - p_b_coll_elem - parse a collating-element name and look it up
899 == static char p_b_coll_elem(register struct parse *p, int endc);
900 */
901 static char /* value of collating element */
902 p_b_coll_elem(p, endc)
903 register struct parse *p;
904 int endc; /* name ended by endc,']' */
905 {
906 register char *sp = p->next;
907 register struct cname *cp;
908 register int len;
909 register char c;
910
911 while (MORE() && !SEETWO(endc, ']'))
912 NEXT();
913 if (!MORE()) {
914 SETERROR(REG_EBRACK);
915 return(0);
916 }
917 len = p->next - sp;
918 for (cp = cnames; cp->name != NULL; cp++)
919 if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
920 return(cp->code); /* known name */
921 if (len == 1)
922 return(*sp); /* single character */
923 SETERROR(REG_ECOLLATE); /* neither */
924 return(0);
925 }
926
927 /*
928 - othercase - return the case counterpart of an alphabetic
929 == static char othercase(int ch);
930 */
931 static char /* if no counterpart, return ch */
932 othercase(ch)
933 int ch;
934 {
935 assert(isalpha(ch));
936 if (isupper(ch))
937 return(tolower(ch));
938 else if (islower(ch))
939 return(toupper(ch));
940 else /* peculiar, but could happen */
941 return(ch);
942 }
943
944 /*
945 - bothcases - emit a dualcase version of a two-case character
946 == static void bothcases(register struct parse *p, int ch);
947 *
948 * Boy, is this implementation ever a kludge...
949 */
950 static void
951 bothcases(p, ch)
952 register struct parse *p;
953 int ch;
954 {
955 register char *oldnext = p->next;
956 register char *oldend = p->end;
957 char bracket[3];
958
959 assert(othercase(ch) != ch); /* p_bracket() would recurse */
960 p->next = bracket;
961 p->end = bracket+2;
962 bracket[0] = ch;
963 bracket[1] = ']';
964 bracket[2] = '\0';
965 p_bracket(p);
966 assert(p->next == bracket+2);
967 p->next = oldnext;
968 p->end = oldend;
969 }
970
971 /*
972 - ordinary - emit an ordinary character
973 == static void ordinary(register struct parse *p, register int ch);
974 */
975 static void
976 ordinary(p, ch)
977 register struct parse *p;
978 register int ch;
979 {
980 register cat_t *cap = p->g->categories;
981
982 if ((p->g->cflags&REG_ICASE) && isalpha(ch) && othercase(ch) != ch)
983 bothcases(p, ch);
984 else {
985 EMIT(OCHAR, (unsigned char)ch);
986 if (cap[ch] == 0)
987 cap[ch] = p->g->ncategories++;
988 }
989 }
990
991 /*
992 - nonnewline - emit REG_NEWLINE version of OANY
993 == static void nonnewline(register struct parse *p);
994 *
995 * Boy, is this implementation ever a kludge...
996 */
997 static void
998 nonnewline(p)
999 register struct parse *p;
1000 {
1001 register char *oldnext = p->next;
1002 register char *oldend = p->end;
1003 char bracket[4];
1004
1005 p->next = bracket;
1006 p->end = bracket+3;
1007 bracket[0] = '^';
1008 bracket[1] = '\n';
1009 bracket[2] = ']';
1010 bracket[3] = '\0';
1011 p_bracket(p);
1012 assert(p->next == bracket+3);
1013 p->next = oldnext;
1014 p->end = oldend;
1015 }
1016
1017 /*
1018 - repeat - generate code for a bounded repetition, recursively if needed
1019 == static void repeat(register struct parse *p, sopno start, int from, int to);
1020 */
1021 static void
1022 repeat(p, start, from, to)
1023 register struct parse *p;
1024 sopno start; /* operand from here to end of strip */
1025 int from; /* repeated from this number */
1026 int to; /* to this number of times (maybe INFINITY) */
1027 {
1028 register sopno finish = HERE();
1029 # define N 2
1030 # define INF 3
1031 # define REP(f, t) ((f)*8 + (t))
1032 # define MAP(n) (((n) <= 1) ? (n) : ((n) == INFINITY) ? INF : N)
1033 register sopno copy;
1034
1035 if (p->error != 0) /* head off possible runaway recursion */
1036 return;
1037
1038 assert(from <= to);
1039
1040 switch (REP(MAP(from), MAP(to))) {
1041 case REP(0, 0): /* must be user doing this */
1042 DROP(finish-start); /* drop the operand */
1043 break;
1044 case REP(0, 1): /* as x{1,1}? */
1045 case REP(0, N): /* as x{1,n}? */
1046 case REP(0, INF): /* as x{1,}? */
1047 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1048 INSERT(OCH_, start); /* offset is wrong... */
1049 repeat(p, start+1, 1, to);
1050 ASTERN(OOR1, start);
1051 AHEAD(start); /* ... fix it */
1052 EMIT(OOR2, 0);
1053 AHEAD(THERE());
1054 ASTERN(O_CH, THERETHERE());
1055 break;
1056 case REP(1, 1): /* trivial case */
1057 /* done */
1058 break;
1059 case REP(1, N): /* as x?x{1,n-1} */
1060 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1061 INSERT(OCH_, start);
1062 ASTERN(OOR1, start);
1063 AHEAD(start);
1064 EMIT(OOR2, 0); /* offset very wrong... */
1065 AHEAD(THERE()); /* ...so fix it */
1066 ASTERN(O_CH, THERETHERE());
1067 copy = dupl(p, start+1, finish+1);
1068 assert(copy == finish+4);
1069 repeat(p, copy, 1, to-1);
1070 break;
1071 case REP(1, INF): /* as x+ */
1072 INSERT(OPLUS_, start);
1073 ASTERN(O_PLUS, start);
1074 break;
1075 case REP(N, N): /* as xx{m-1,n-1} */
1076 copy = dupl(p, start, finish);
1077 repeat(p, copy, from-1, to-1);
1078 break;
1079 case REP(N, INF): /* as xx{n-1,INF} */
1080 copy = dupl(p, start, finish);
1081 repeat(p, copy, from-1, to);
1082 break;
1083 default: /* "can't happen" */
1084 SETERROR(REG_ASSERT); /* just in case */
1085 break;
1086 }
1087 }
1088
1089 /*
1090 - seterr - set an error condition
1091 == static int seterr(register struct parse *p, int e);
1092 */
1093 static int /* useless but makes type checking happy */
1094 seterr(p, e)
1095 register struct parse *p;
1096 int e;
1097 {
1098 if (p->error == 0) /* keep earliest error condition */
1099 p->error = e;
1100 p->next = nuls; /* try to bring things to a halt */
1101 p->end = nuls;
1102 return(0); /* make the return value well-defined */
1103 }
1104
1105 /*
1106 - allocset - allocate a set of characters for []
1107 == static cset *allocset(register struct parse *p);
1108 */
1109 static cset *
1110 allocset(p)
1111 register struct parse *p;
1112 {
1113 register int no = p->g->ncsets++;
1114 register size_t nc;
1115 register size_t nbytes;
1116 register cset *cs;
1117 register size_t css = (size_t)p->g->csetsize;
1118 register int i;
1119
1120 if (no >= p->ncsalloc) { /* need another column of space */
1121 p->ncsalloc += CHAR_BIT;
1122 nc = p->ncsalloc;
1123 assert(nc % CHAR_BIT == 0);
1124 nbytes = nc / CHAR_BIT * css;
1125 if (p->g->sets == NULL)
1126 p->g->sets = (cset *)malloc(nc * sizeof(cset));
1127 else
1128 p->g->sets = (cset *)realloc((char *)p->g->sets,
1129 nc * sizeof(cset));
1130 if (p->g->setbits == NULL)
1131 p->g->setbits = (uch *)malloc(nbytes);
1132 else {
1133 p->g->setbits = (uch *)realloc((char *)p->g->setbits,
1134 nbytes);
1135 /* xxx this isn't right if setbits is now NULL */
1136 for (i = 0; i < no; i++)
1137 p->g->sets[i].ptr = p->g->setbits + css*(i/CHAR_BIT);
1138 }
1139 if (p->g->sets != NULL && p->g->setbits != NULL)
1140 (void) memset((char *)p->g->setbits + (nbytes - css),
1141 0, css);
1142 else {
1143 no = 0;
1144 SETERROR(REG_ESPACE);
1145 /* caller's responsibility not to do set ops */
1146 }
1147 }
1148
1149 assert(p->g->sets != NULL); /* xxx */
1150 cs = &p->g->sets[no];
1151 cs->ptr = p->g->setbits + css*((no)/CHAR_BIT);
1152 cs->mask = 1 << ((no) % CHAR_BIT);
1153 cs->hash = 0;
1154 cs->smultis = 0;
1155 cs->multis = NULL;
1156
1157 return(cs);
1158 }
1159
1160 /*
1161 - freeset - free a now-unused set
1162 == static void freeset(register struct parse *p, register cset *cs);
1163 */
1164 static void
1165 freeset(p, cs)
1166 register struct parse *p;
1167 register cset *cs;
1168 {
1169 register int i;
1170 register cset *top = &p->g->sets[p->g->ncsets];
1171 register size_t css = (size_t)p->g->csetsize;
1172
1173 for (i = 0; i < css; i++)
1174 CHsub(cs, i);
1175 if (cs == top-1) /* recover only the easy case */
1176 p->g->ncsets--;
1177 }
1178
1179 /*
1180 - freezeset - final processing on a set of characters
1181 == static int freezeset(register struct parse *p, register cset *cs);
1182 *
1183 * The main task here is merging identical sets. This is usually a waste
1184 * of time (although the hash code minimizes the overhead), but can win
1185 * big if REG_ICASE is being used. REG_ICASE, by the way, is why the hash
1186 * is done using addition rather than xor -- all ASCII [aA] sets xor to
1187 * the same value!
1188 */
1189 static int /* set number */
1190 freezeset(p, cs)
1191 register struct parse *p;
1192 register cset *cs;
1193 {
1194 register uch h = cs->hash;
1195 register int i;
1196 register cset *top = &p->g->sets[p->g->ncsets];
1197 register cset *cs2;
1198 register size_t css = (size_t)p->g->csetsize;
1199
1200 /* look for an earlier one which is the same */
1201 for (cs2 = &p->g->sets[0]; cs2 < top; cs2++)
1202 if (cs2->hash == h && cs2 != cs) {
1203 /* maybe */
1204 for (i = 0; i < css; i++)
1205 if (!!CHIN(cs2, i) != !!CHIN(cs, i))
1206 break; /* no */
1207 if (i == css)
1208 break; /* yes */
1209 }
1210
1211 if (cs2 < top) { /* found one */
1212 freeset(p, cs);
1213 cs = cs2;
1214 }
1215
1216 return((int)(cs - p->g->sets));
1217 }
1218
1219 /*
1220 - firstch - return first character in a set (which must have at least one)
1221 == static int firstch(register struct parse *p, register cset *cs);
1222 */
1223 static int /* character; there is no "none" value */
1224 firstch(p, cs)
1225 register struct parse *p;
1226 register cset *cs;
1227 {
1228 register int i;
1229 register size_t css = (size_t)p->g->csetsize;
1230
1231 for (i = 0; i < css; i++)
1232 if (CHIN(cs, i))
1233 return((char)i);
1234 assert(never);
1235 return(0); /* arbitrary */
1236 }
1237
1238 /*
1239 - nch - number of characters in a set
1240 == static int nch(register struct parse *p, register cset *cs);
1241 */
1242 static int
1243 nch(p, cs)
1244 register struct parse *p;
1245 register cset *cs;
1246 {
1247 register int i;
1248 register size_t css = (size_t)p->g->csetsize;
1249 register int n = 0;
1250
1251 for (i = 0; i < css; i++)
1252 if (CHIN(cs, i))
1253 n++;
1254 return(n);
1255 }
1256
1257 /*
1258 - mcadd - add a collating element to a cset
1259 == static void mcadd(register struct parse *p, register cset *cs, \
1260 == register char *cp);
1261 */
1262 static void
1263 mcadd(p, cs, cp)
1264 register struct parse *p;
1265 register cset *cs;
1266 register char *cp;
1267 {
1268 register size_t oldend = cs->smultis;
1269
1270 cs->smultis += strlen(cp) + 1;
1271 if (cs->multis == NULL)
1272 cs->multis = malloc(cs->smultis);
1273 else
1274 cs->multis = realloc(cs->multis, cs->smultis);
1275 if (cs->multis == NULL) {
1276 SETERROR(REG_ESPACE);
1277 return;
1278 }
1279
1280 (void) strcpy(cs->multis + oldend - 1, cp);
1281 cs->multis[cs->smultis - 1] = '\0';
1282 }
1283
1284 /*
1285 - mcsub - subtract a collating element from a cset
1286 == static void mcsub(register cset *cs, register char *cp);
1287 */
1288 static void
1289 mcsub(cs, cp)
1290 register cset *cs;
1291 register char *cp;
1292 {
1293 register char *fp = mcfind(cs, cp);
1294 register size_t len = strlen(fp);
1295
1296 assert(fp != NULL);
1297 (void) memmove(fp, fp + len + 1,
1298 cs->smultis - (fp + len + 1 - cs->multis));
1299 cs->smultis -= len;
1300
1301 if (cs->smultis == 0) {
1302 free(cs->multis);
1303 cs->multis = NULL;
1304 return;
1305 }
1306
1307 cs->multis = realloc(cs->multis, cs->smultis);
1308 assert(cs->multis != NULL);
1309 }
1310
1311 /*
1312 - mcin - is a collating element in a cset?
1313 == static int mcin(register cset *cs, register char *cp);
1314 */
1315 static int
1316 mcin(cs, cp)
1317 register cset *cs;
1318 register char *cp;
1319 {
1320 return(mcfind(cs, cp) != NULL);
1321 }
1322
1323 /*
1324 - mcfind - find a collating element in a cset
1325 == static char *mcfind(register cset *cs, register char *cp);
1326 */
1327 static char *
1328 mcfind(cs, cp)
1329 register cset *cs;
1330 register char *cp;
1331 {
1332 register char *p;
1333
1334 if (cs->multis == NULL)
1335 return(NULL);
1336 for (p = cs->multis; *p != '\0'; p += strlen(p) + 1)
1337 if (strcmp(cp, p) == 0)
1338 return(p);
1339 return(NULL);
1340 }
1341
1342 /*
1343 - mcinvert - invert the list of collating elements in a cset
1344 == static void mcinvert(register struct parse *p, register cset *cs);
1345 *
1346 * This would have to know the set of possibilities. Implementation
1347 * is deferred.
1348 */
1349 static void
1350 mcinvert(p, cs)
1351 register struct parse *p;
1352 register cset *cs;
1353 {
1354 assert(cs->multis == NULL); /* xxx */
1355 }
1356
1357 /*
1358 - mccase - add case counterparts of the list of collating elements in a cset
1359 == static void mccase(register struct parse *p, register cset *cs);
1360 *
1361 * This would have to know the set of possibilities. Implementation
1362 * is deferred.
1363 */
1364 static void
1365 mccase(p, cs)
1366 register struct parse *p;
1367 register cset *cs;
1368 {
1369 assert(cs->multis == NULL); /* xxx */
1370 }
1371
1372 /*
1373 - isinsets - is this character in any sets?
1374 == static int isinsets(register struct re_guts *g, int c);
1375 */
1376 static int /* predicate */
1377 isinsets(g, c)
1378 register struct re_guts *g;
1379 int c;
1380 {
1381 register uch *col;
1382 register int i;
1383 register int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1384 register unsigned uc = (unsigned char)c;
1385
1386 for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1387 if (col[uc] != 0)
1388 return(1);
1389 return(0);
1390 }
1391
1392 /*
1393 - samesets - are these two characters in exactly the same sets?
1394 == static int samesets(register struct re_guts *g, int c1, int c2);
1395 */
1396 static int /* predicate */
1397 samesets(g, c1, c2)
1398 register struct re_guts *g;
1399 int c1;
1400 int c2;
1401 {
1402 register uch *col;
1403 register int i;
1404 register int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1405 register unsigned uc1 = (unsigned char)c1;
1406 register unsigned uc2 = (unsigned char)c2;
1407
1408 for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1409 if (col[uc1] != col[uc2])
1410 return(0);
1411 return(1);
1412 }
1413
1414 /*
1415 - categorize - sort out character categories
1416 == static void categorize(struct parse *p, register struct re_guts *g);
1417 */
1418 static void
1419 categorize(p, g)
1420 struct parse *p;
1421 register struct re_guts *g;
1422 {
1423 register cat_t *cats = g->categories;
1424 register int c;
1425 register int c2;
1426 register cat_t cat;
1427
1428 /* avoid making error situations worse */
1429 if (p->error != 0)
1430 return;
1431
1432 for (c = CHAR_MIN; c <= CHAR_MAX; c++)
1433 if (cats[c] == 0 && isinsets(g, c)) {
1434 cat = g->ncategories++;
1435 cats[c] = cat;
1436 for (c2 = c+1; c2 <= CHAR_MAX; c2++)
1437 if (cats[c2] == 0 && samesets(g, c, c2))
1438 cats[c2] = cat;
1439 }
1440 }
1441
1442 /*
1443 - dupl - emit a duplicate of a bunch of sops
1444 == static sopno dupl(register struct parse *p, sopno start, sopno finish);
1445 */
1446 static sopno /* start of duplicate */
1447 dupl(p, start, finish)
1448 register struct parse *p;
1449 sopno start; /* from here */
1450 sopno finish; /* to this less one */
1451 {
1452 register sopno ret = HERE();
1453 register sopno len = finish - start;
1454
1455 assert(finish >= start);
1456 if (len == 0)
1457 return(ret);
1458 enlarge(p, p->ssize + len); /* this many unexpected additions */
1459 assert(p->ssize >= p->slen + len);
1460 (void) memcpy((char *)(p->strip + p->slen),
1461 (char *)(p->strip + start), (size_t)len*sizeof(sop));
1462 p->slen += len;
1463 return(ret);
1464 }
1465
1466 /*
1467 - doemit - emit a strip operator
1468 == static void doemit(register struct parse *p, sop op, size_t opnd);
1469 *
1470 * It might seem better to implement this as a macro with a function as
1471 * hard-case backup, but it's just too big and messy unless there are
1472 * some changes to the data structures. Maybe later.
1473 */
1474 static void
1475 doemit(p, op, opnd)
1476 register struct parse *p;
1477 sop op;
1478 size_t opnd;
1479 {
1480 /* avoid making error situations worse */
1481 if (p->error != 0)
1482 return;
1483
1484 /* deal with oversize operands ("can't happen", more or less) */
1485 assert(opnd < 1<<OPSHIFT);
1486
1487 /* deal with undersized strip */
1488 if (p->slen >= p->ssize)
1489 enlarge(p, (p->ssize+1) / 2 * 3); /* +50% */
1490 assert(p->slen < p->ssize);
1491
1492 /* finally, it's all reduced to the easy case */
1493 p->strip[p->slen++] = SOP(op, opnd);
1494 }
1495
1496 /*
1497 - doinsert - insert a sop into the strip
1498 == static void doinsert(register struct parse *p, sop op, size_t opnd, sopno pos);
1499 */
1500 static void
1501 doinsert(p, op, opnd, pos)
1502 register struct parse *p;
1503 sop op;
1504 size_t opnd;
1505 sopno pos;
1506 {
1507 register sopno sn;
1508 register sop s;
1509 register int i;
1510
1511 /* avoid making error situations worse */
1512 if (p->error != 0)
1513 return;
1514
1515 sn = HERE();
1516 EMIT(op, opnd); /* do checks, ensure space */
1517 assert(HERE() == sn+1);
1518 s = p->strip[sn];
1519
1520 /* adjust paren pointers */
1521 assert(pos > 0);
1522 for (i = 1; i < NPAREN; i++) {
1523 if (p->pbegin[i] >= pos) {
1524 p->pbegin[i]++;
1525 }
1526 if (p->pend[i] >= pos) {
1527 p->pend[i]++;
1528 }
1529 }
1530
1531 memmove((char *)&p->strip[pos+1], (char *)&p->strip[pos],
1532 (HERE()-pos-1)*sizeof(sop));
1533 p->strip[pos] = s;
1534 }
1535
1536 /*
1537 - dofwd - complete a forward reference
1538 == static void dofwd(register struct parse *p, sopno pos, sop value);
1539 */
1540 static void
1541 dofwd(p, pos, value)
1542 register struct parse *p;
1543 register sopno pos;
1544 sop value;
1545 {
1546 /* avoid making error situations worse */
1547 if (p->error != 0)
1548 return;
1549
1550 assert(value < 1<<OPSHIFT);
1551 p->strip[pos] = OP(p->strip[pos]) | value;
1552 }
1553
1554 /*
1555 - enlarge - enlarge the strip
1556 == static void enlarge(register struct parse *p, sopno size);
1557 */
1558 static void
1559 enlarge(p, size)
1560 register struct parse *p;
1561 register sopno size;
1562 {
1563 register sop *sp;
1564
1565 if (p->ssize >= size)
1566 return;
1567
1568 sp = (sop *)realloc(p->strip, size*sizeof(sop));
1569 if (sp == NULL) {
1570 SETERROR(REG_ESPACE);
1571 return;
1572 }
1573 p->strip = sp;
1574 p->ssize = size;
1575 }
1576
1577 /*
1578 - stripsnug - compact the strip
1579 == static void stripsnug(register struct parse *p, register struct re_guts *g);
1580 */
1581 static void
1582 stripsnug(p, g)
1583 register struct parse *p;
1584 register struct re_guts *g;
1585 {
1586 g->nstates = p->slen;
1587 g->strip = (sop *)realloc((char *)p->strip, p->slen * sizeof(sop));
1588 if (g->strip == NULL) {
1589 SETERROR(REG_ESPACE);
1590 g->strip = p->strip;
1591 }
1592 }
1593
1594 /*
1595 - findmust - fill in must and mlen with longest mandatory literal string
1596 == static void findmust(register struct parse *p, register struct re_guts *g);
1597 *
1598 * This algorithm could do fancy things like analyzing the operands of |
1599 * for common subsequences. Someday. This code is simple and finds most
1600 * of the interesting cases.
1601 *
1602 * Note that must and mlen got initialized during setup.
1603 */
1604 static void
1605 findmust(p, g)
1606 struct parse *p;
1607 register struct re_guts *g;
1608 {
1609 register sop *scan;
1610 sop *start;
1611 register sop *newstart;
1612 register sopno newlen;
1613 register sop s;
1614 register char *cp;
1615 register sopno i;
1616
1617 /* avoid making error situations worse */
1618 if (p->error != 0)
1619 return;
1620
1621 /* find the longest OCHAR sequence in strip */
1622 newlen = 0;
1623 scan = g->strip + 1;
1624 do {
1625 s = *scan++;
1626 switch (OP(s)) {
1627 case OCHAR: /* sequence member */
1628 if (newlen == 0) /* new sequence */
1629 newstart = scan - 1;
1630 newlen++;
1631 break;
1632 case OPLUS_: /* things that don't break one */
1633 case OLPAREN:
1634 case ORPAREN:
1635 break;
1636 case OQUEST_: /* things that must be skipped */
1637 case OCH_:
1638 scan--;
1639 do {
1640 scan += OPND(s);
1641 s = *scan;
1642 /* assert() interferes w debug printouts */
1643 if (OP(s) != O_QUEST && OP(s) != O_CH &&
1644 OP(s) != OOR2) {
1645 g->iflags |= BAD;
1646 return;
1647 }
1648 } while (OP(s) != O_QUEST && OP(s) != O_CH);
1649 /* fallthrough */
1650 default: /* things that break a sequence */
1651 if (newlen > g->mlen) { /* ends one */
1652 start = newstart;
1653 g->mlen = newlen;
1654 }
1655 newlen = 0;
1656 break;
1657 }
1658 } while (OP(s) != OEND);
1659
1660 if (g->mlen == 0) /* there isn't one */
1661 return;
1662
1663 /* turn it into a character string */
1664 g->must = malloc((size_t)g->mlen + 1);
1665 if (g->must == NULL) { /* argh; just forget it */
1666 g->mlen = 0;
1667 return;
1668 }
1669 cp = g->must;
1670 scan = start;
1671 for (i = g->mlen; i > 0; i--) {
1672 while (OP(s = *scan++) != OCHAR)
1673 continue;
1674 assert(cp < g->must + g->mlen);
1675 *cp++ = (char)OPND(s);
1676 }
1677 assert(cp == g->must + g->mlen);
1678 *cp++ = '\0'; /* just on general principles */
1679 }
1680
1681 /*
1682 - pluscount - count + nesting
1683 == static sopno pluscount(register struct parse *p, register struct re_guts *g);
1684 */
1685 static sopno /* nesting depth */
1686 pluscount(p, g)
1687 struct parse *p;
1688 register struct re_guts *g;
1689 {
1690 register sop *scan;
1691 register sop s;
1692 register sopno plusnest = 0;
1693 register sopno maxnest = 0;
1694
1695 if (p->error != 0)
1696 return(0); /* there may not be an OEND */
1697
1698 scan = g->strip + 1;
1699 do {
1700 s = *scan++;
1701 switch (OP(s)) {
1702 case OPLUS_:
1703 plusnest++;
1704 break;
1705 case O_PLUS:
1706 if (plusnest > maxnest)
1707 maxnest = plusnest;
1708 plusnest--;
1709 break;
1710 }
1711 } while (OP(s) != OEND);
1712 if (plusnest != 0)
1713 g->iflags |= BAD;
1714 return(maxnest);
1715 }
mirror of libc