| blob | 0f27cfd4b5b8f0d4ed2b01bf222de5a02c3ffd9f |
1 /*-
2 * This code is derived from OpenBSD's libc/regex, original license follows:
3 *
4 * Copyright (c) 1992, 1993, 1994 Henry Spencer.
5 * Copyright (c) 1992, 1993, 1994
6 * The Regents of the University of California. All rights reserved.
7 *
8 * This code is derived from software contributed to Berkeley by
9 * Henry Spencer.
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 * @(#)engine.c 8.5 (Berkeley) 3/20/94
36 */
38 /*
39 * The matching engine and friends. This file is #included by regexec.c
40 * after suitable #defines of a variety of macros used herein, so that
41 * different state representations can be used without duplicating masses
42 * of code.
43 */
45 #ifdef SNAMES
46 #define matcher smatcher
47 #define fast sfast
48 #define slow sslow
49 #define dissect sdissect
50 #define backref sbackref
51 #define step sstep
52 #define print sprint
53 #define at sat
54 #define match smat
55 #define nope snope
56 #endif
57 #ifdef LNAMES
58 #define matcher lmatcher
59 #define fast lfast
60 #define slow lslow
61 #define dissect ldissect
62 #define backref lbackref
63 #define step lstep
64 #define print lprint
65 #define at lat
66 #define match lmat
67 #define nope lnope
68 #endif
70 /* another structure passed up and down to avoid zillions of parameters */
71 struct match {
72 struct re_guts *g;
73 int eflags;
74 llvm_regmatch_t *pmatch; /* [nsub+1] (0 element unused) */
75 char *offp; /* offsets work from here */
76 char *beginp; /* start of string -- virtual NUL precedes */
77 char *endp; /* end of string -- virtual NUL here */
78 char *coldp; /* can be no match starting before here */
79 char **lastpos; /* [nplus+1] */
80 STATEVARS;
81 states st; /* current states */
82 states fresh; /* states for a fresh start */
83 states tmp; /* temporary */
84 states empty; /* empty set of states */
85 };
87 static int matcher(struct re_guts *, char *, size_t, llvm_regmatch_t[], int);
88 static char *dissect(struct match *, char *, char *, sopno, sopno);
89 static char *backref(struct match *, char *, char *, sopno, sopno, sopno, int);
90 static char *fast(struct match *, char *, char *, sopno, sopno);
91 static char *slow(struct match *, char *, char *, sopno, sopno);
92 static states step(struct re_guts *, sopno, sopno, states, int, states);
93 #define MAX_RECURSION 100
94 #define BOL (OUT+1)
95 #define EOL (BOL+1)
96 #define BOLEOL (BOL+2)
97 #define NOTHING (BOL+3)
98 #define BOW (BOL+4)
99 #define EOW (BOL+5)
100 #define CODEMAX (BOL+5) /* highest code used */
101 #define NONCHAR(c) ((c) > CHAR_MAX)
102 #define NNONCHAR (CODEMAX-CHAR_MAX)
103 #ifdef REDEBUG
104 static void print(struct match *, char *, states, int, FILE *);
105 #endif
106 #ifdef REDEBUG
107 static void at(struct match *, char *, char *, char *, sopno, sopno);
108 #endif
109 #ifdef REDEBUG
110 static char *pchar(int);
111 #endif
113 #ifdef REDEBUG
114 #define SP(t, s, c) print(m, t, s, c, stdout)
115 #define AT(t, p1, p2, s1, s2) at(m, t, p1, p2, s1, s2)
116 #define NOTE(str) { if (m->eflags®_TRACE) (void)printf("=%s\n", (str)); }
117 static int nope = 0;
118 #else
119 #define SP(t, s, c) /* nothing */
120 #define AT(t, p1, p2, s1, s2) /* nothing */
121 #define NOTE(s) /* nothing */
122 #endif
124 /*
125 - matcher - the actual matching engine
126 */
127 static int /* 0 success, REG_NOMATCH failure */
128 matcher(struct re_guts *g, char *string, size_t nmatch, llvm_regmatch_t pmatch[],
129 int eflags)
130 {
131 char *endp;
132 size_t i;
133 struct match mv;
134 struct match *m = &mv;
135 char *dp;
136 const sopno gf = g->firststate+1; /* +1 for OEND */
137 const sopno gl = g->laststate;
138 char *start;
139 char *stop;
141 /* simplify the situation where possible */
142 if (g->cflags®_NOSUB)
143 nmatch = 0;
144 if (eflags®_STARTEND) {
145 start = string + pmatch[0].rm_so;
146 stop = string + pmatch[0].rm_eo;
147 } else {
148 start = string;
149 stop = start + strlen(start);
150 }
151 if (stop < start)
152 return(REG_INVARG);
154 /* prescreening; this does wonders for this rather slow code */
155 if (g->must != NULL) {
156 for (dp = start; dp < stop; dp++)
157 if (*dp == g->must[0] && stop - dp >= g->mlen &&
158 memcmp(dp, g->must, (size_t)g->mlen) == 0)
159 break;
160 if (dp == stop) /* we didn't find g->must */
161 return(REG_NOMATCH);
162 }
164 /* match struct setup */
165 m->g = g;
166 m->eflags = eflags;
167 m->pmatch = NULL;
168 m->lastpos = NULL;
169 m->offp = string;
170 m->beginp = start;
171 m->endp = stop;
172 STATESETUP(m, 4);
173 SETUP(m->st);
174 SETUP(m->fresh);
175 SETUP(m->tmp);
176 SETUP(m->empty);
177 CLEAR(m->empty);
179 /* this loop does only one repetition except for backrefs */
180 for (;;) {
181 endp = fast(m, start, stop, gf, gl);
182 if (endp == NULL) { /* a miss */
183 free(m->pmatch);
184 free(m->lastpos);
185 STATETEARDOWN(m);
186 return(REG_NOMATCH);
187 }
188 if (nmatch == 0 && !g->backrefs)
189 break; /* no further info needed */
191 /* where? */
192 assert(m->coldp != NULL);
193 for (;;) {
194 NOTE("finding start");
195 endp = slow(m, m->coldp, stop, gf, gl);
196 if (endp != NULL)
197 break;
198 assert(m->coldp < m->endp);
199 m->coldp++;
200 }
201 if (nmatch == 1 && !g->backrefs)
202 break; /* no further info needed */
204 /* oh my, he wants the subexpressions... */
205 if (m->pmatch == NULL)
206 m->pmatch = (llvm_regmatch_t *)malloc((m->g->nsub + 1) *
207 sizeof(llvm_regmatch_t));
208 if (m->pmatch == NULL) {
209 STATETEARDOWN(m);
210 return(REG_ESPACE);
211 }
212 for (i = 1; i <= m->g->nsub; i++)
213 m->pmatch[i].rm_so = m->pmatch[i].rm_eo = -1;
214 if (!g->backrefs && !(m->eflags®_BACKR)) {
215 NOTE("dissecting");
216 dp = dissect(m, m->coldp, endp, gf, gl);
217 } else {
218 if (g->nplus > 0 && m->lastpos == NULL)
219 m->lastpos = (char **)malloc((g->nplus+1) *
220 sizeof(char *));
221 if (g->nplus > 0 && m->lastpos == NULL) {
222 free(m->pmatch);
223 STATETEARDOWN(m);
224 return(REG_ESPACE);
225 }
226 NOTE("backref dissect");
227 dp = backref(m, m->coldp, endp, gf, gl, (sopno)0, 0);
228 }
229 if (dp != NULL)
230 break;
232 /* uh-oh... we couldn't find a subexpression-level match */
233 assert(g->backrefs); /* must be back references doing it */
234 assert(g->nplus == 0 || m->lastpos != NULL);
235 for (;;) {
236 if (dp != NULL || endp <= m->coldp)
237 break; /* defeat */
238 NOTE("backoff");
239 endp = slow(m, m->coldp, endp-1, gf, gl);
240 if (endp == NULL)
241 break; /* defeat */
242 /* try it on a shorter possibility */
243 #ifndef NDEBUG
244 for (i = 1; i <= m->g->nsub; i++) {
245 assert(m->pmatch[i].rm_so == -1);
246 assert(m->pmatch[i].rm_eo == -1);
247 }
248 #endif
249 NOTE("backoff dissect");
250 dp = backref(m, m->coldp, endp, gf, gl, (sopno)0, 0);
251 }
252 assert(dp == NULL || dp == endp);
253 if (dp != NULL) /* found a shorter one */
254 break;
256 /* despite initial appearances, there is no match here */
257 NOTE("false alarm");
258 if (m->coldp == stop)
259 break;
260 start = m->coldp + 1; /* recycle starting later */
261 }
263 /* fill in the details if requested */
264 if (nmatch > 0) {
265 pmatch[0].rm_so = m->coldp - m->offp;
266 pmatch[0].rm_eo = endp - m->offp;
267 }
268 if (nmatch > 1) {
269 assert(m->pmatch != NULL);
270 for (i = 1; i < nmatch; i++)
271 if (i <= m->g->nsub)
272 pmatch[i] = m->pmatch[i];
273 else {
274 pmatch[i].rm_so = -1;
275 pmatch[i].rm_eo = -1;
276 }
277 }
279 if (m->pmatch != NULL)
280 free((char *)m->pmatch);
281 if (m->lastpos != NULL)
282 free((char *)m->lastpos);
283 STATETEARDOWN(m);
284 return(0);
285 }
287 /*
288 - dissect - figure out what matched what, no back references
289 */
290 static char * /* == stop (success) always */
291 dissect(struct match *m, char *start, char *stop, sopno startst, sopno stopst)
292 {
293 int i;
294 sopno ss; /* start sop of current subRE */
295 sopno es; /* end sop of current subRE */
296 char *sp; /* start of string matched by it */
297 char *stp; /* string matched by it cannot pass here */
298 char *rest; /* start of rest of string */
299 char *tail; /* string unmatched by rest of RE */
300 sopno ssub; /* start sop of subsubRE */
301 sopno esub; /* end sop of subsubRE */
302 char *ssp; /* start of string matched by subsubRE */
303 char *sep; /* end of string matched by subsubRE */
304 char *oldssp; /* previous ssp */
306 AT("diss", start, stop, startst, stopst);
307 sp = start;
308 for (ss = startst; ss < stopst; ss = es) {
309 /* identify end of subRE */
310 es = ss;
311 switch (OP(m->g->strip[es])) {
312 case OPLUS_:
313 case OQUEST_:
314 es += OPND(m->g->strip[es]);
315 break;
316 case OCH_:
317 while (OP(m->g->strip[es]) != O_CH)
318 es += OPND(m->g->strip[es]);
319 break;
320 }
321 es++;
323 /* figure out what it matched */
324 switch (OP(m->g->strip[ss])) {
325 case OEND:
326 assert(nope);
327 break;
328 case OCHAR:
329 sp++;
330 break;
331 case OBOL:
332 case OEOL:
333 case OBOW:
334 case OEOW:
335 break;
336 case OANY:
337 case OANYOF:
338 sp++;
339 break;
340 case OBACK_:
341 case O_BACK:
342 assert(nope);
343 break;
344 /* cases where length of match is hard to find */
345 case OQUEST_:
346 stp = stop;
347 for (;;) {
348 /* how long could this one be? */
349 rest = slow(m, sp, stp, ss, es);
350 assert(rest != NULL); /* it did match */
351 /* could the rest match the rest? */
352 tail = slow(m, rest, stop, es, stopst);
353 if (tail == stop)
354 break; /* yes! */
355 /* no -- try a shorter match for this one */
356 stp = rest - 1;
357 assert(stp >= sp); /* it did work */
358 }
359 ssub = ss + 1;
360 esub = es - 1;
361 /* did innards match? */
362 if (slow(m, sp, rest, ssub, esub) != NULL) {
363 char *dp = dissect(m, sp, rest, ssub, esub);
364 (void)dp; /* avoid warning if assertions off */
365 assert(dp == rest);
366 } else /* no */
367 assert(sp == rest);
368 sp = rest;
369 break;
370 case OPLUS_:
371 stp = stop;
372 for (;;) {
373 /* how long could this one be? */
374 rest = slow(m, sp, stp, ss, es);
375 assert(rest != NULL); /* it did match */
376 /* could the rest match the rest? */
377 tail = slow(m, rest, stop, es, stopst);
378 if (tail == stop)
379 break; /* yes! */
380 /* no -- try a shorter match for this one */
381 stp = rest - 1;
382 assert(stp >= sp); /* it did work */
383 }
384 ssub = ss + 1;
385 esub = es - 1;
386 ssp = sp;
387 oldssp = ssp;
388 for (;;) { /* find last match of innards */
389 sep = slow(m, ssp, rest, ssub, esub);
390 if (sep == NULL || sep == ssp)
391 break; /* failed or matched null */
392 oldssp = ssp; /* on to next try */
393 ssp = sep;
394 }
395 if (sep == NULL) {
396 /* last successful match */
397 sep = ssp;
398 ssp = oldssp;
399 }
400 assert(sep == rest); /* must exhaust substring */
401 assert(slow(m, ssp, sep, ssub, esub) == rest);
402 {
403 char *dp = dissect(m, ssp, sep, ssub, esub);
404 (void)dp; /* avoid warning if assertions off */
405 assert(dp == sep);
406 }
407 sp = rest;
408 break;
409 case OCH_:
410 stp = stop;
411 for (;;) {
412 /* how long could this one be? */
413 rest = slow(m, sp, stp, ss, es);
414 assert(rest != NULL); /* it did match */
415 /* could the rest match the rest? */
416 tail = slow(m, rest, stop, es, stopst);
417 if (tail == stop)
418 break; /* yes! */
419 /* no -- try a shorter match for this one */
420 stp = rest - 1;
421 assert(stp >= sp); /* it did work */
422 }
423 ssub = ss + 1;
424 esub = ss + OPND(m->g->strip[ss]) - 1;
425 assert(OP(m->g->strip[esub]) == OOR1);
426 for (;;) { /* find first matching branch */
427 if (slow(m, sp, rest, ssub, esub) == rest)
428 break; /* it matched all of it */
429 /* that one missed, try next one */
430 assert(OP(m->g->strip[esub]) == OOR1);
431 esub++;
432 assert(OP(m->g->strip[esub]) == OOR2);
433 ssub = esub + 1;
434 esub += OPND(m->g->strip[esub]);
435 if (OP(m->g->strip[esub]) == OOR2)
436 esub--;
437 else
438 assert(OP(m->g->strip[esub]) == O_CH);
439 }
440 {
441 char *dp = dissect(m, sp, rest, ssub, esub);
442 (void)dp; /* avoid warning if assertions off */
443 assert(dp == rest);
444 }
445 sp = rest;
446 break;
447 case O_PLUS:
448 case O_QUEST:
449 case OOR1:
450 case OOR2:
451 case O_CH:
452 assert(nope);
453 break;
454 case OLPAREN:
455 i = OPND(m->g->strip[ss]);
456 assert(0 < i && i <= m->g->nsub);
457 m->pmatch[i].rm_so = sp - m->offp;
458 break;
459 case ORPAREN:
460 i = OPND(m->g->strip[ss]);
461 assert(0 < i && i <= m->g->nsub);
462 m->pmatch[i].rm_eo = sp - m->offp;
463 break;
464 default: /* uh oh */
465 assert(nope);
466 break;
467 }
468 }
470 assert(sp == stop);
471 return(sp);
472 }
474 /*
475 - backref - figure out what matched what, figuring in back references
476 */
477 static char * /* == stop (success) or NULL (failure) */
478 backref(struct match *m, char *start, char *stop, sopno startst, sopno stopst,
479 sopno lev, int rec) /* PLUS nesting level */
480 {
481 int i;
482 sopno ss; /* start sop of current subRE */
483 char *sp; /* start of string matched by it */
484 sopno ssub; /* start sop of subsubRE */
485 sopno esub; /* end sop of subsubRE */
486 char *ssp; /* start of string matched by subsubRE */
487 char *dp;
488 size_t len;
489 int hard;
490 sop s;
491 llvm_regoff_t offsave;
492 cset *cs;
494 AT("back", start, stop, startst, stopst);
495 sp = start;
497 /* get as far as we can with easy stuff */
498 hard = 0;
499 for (ss = startst; !hard && ss < stopst; ss++)
500 switch (OP(s = m->g->strip[ss])) {
501 case OCHAR:
502 if (sp == stop || *sp++ != (char)OPND(s))
503 return(NULL);
504 break;
505 case OANY:
506 if (sp == stop)
507 return(NULL);
508 sp++;
509 break;
510 case OANYOF:
511 cs = &m->g->sets[OPND(s)];
512 if (sp == stop || !CHIN(cs, *sp++))
513 return(NULL);
514 break;
515 case OBOL:
516 if ( (sp == m->beginp && !(m->eflags®_NOTBOL)) ||
517 (sp < m->endp && *(sp-1) == '\n' &&
518 (m->g->cflags®_NEWLINE)) )
519 { /* yes */ }
520 else
521 return(NULL);
522 break;
523 case OEOL:
524 if ( (sp == m->endp && !(m->eflags®_NOTEOL)) ||
525 (sp < m->endp && *sp == '\n' &&
526 (m->g->cflags®_NEWLINE)) )
527 { /* yes */ }
528 else
529 return(NULL);
530 break;
531 case OBOW:
532 if (( (sp == m->beginp && !(m->eflags®_NOTBOL)) ||
533 (sp < m->endp && *(sp-1) == '\n' &&
534 (m->g->cflags®_NEWLINE)) ||
535 (sp > m->beginp &&
536 !ISWORD(*(sp-1))) ) &&
537 (sp < m->endp && ISWORD(*sp)) )
538 { /* yes */ }
539 else
540 return(NULL);
541 break;
542 case OEOW:
543 if (( (sp == m->endp && !(m->eflags®_NOTEOL)) ||
544 (sp < m->endp && *sp == '\n' &&
545 (m->g->cflags®_NEWLINE)) ||
546 (sp < m->endp && !ISWORD(*sp)) ) &&
547 (sp > m->beginp && ISWORD(*(sp-1))) )
548 { /* yes */ }
549 else
550 return(NULL);
551 break;
552 case O_QUEST:
553 break;
554 case OOR1: /* matches null but needs to skip */
555 ss++;
556 s = m->g->strip[ss];
557 do {
558 assert(OP(s) == OOR2);
559 ss += OPND(s);
560 } while (OP(s = m->g->strip[ss]) != O_CH);
561 /* note that the ss++ gets us past the O_CH */
562 break;
563 default: /* have to make a choice */
564 hard = 1;
565 break;
566 }
567 if (!hard) { /* that was it! */
568 if (sp != stop)
569 return(NULL);
570 return(sp);
571 }
572 ss--; /* adjust for the for's final increment */
574 /* the hard stuff */
575 AT("hard", sp, stop, ss, stopst);
576 s = m->g->strip[ss];
577 switch (OP(s)) {
578 case OBACK_: /* the vilest depths */
579 i = OPND(s);
580 assert(0 < i && i <= m->g->nsub);
581 if (m->pmatch[i].rm_eo == -1)
582 return(NULL);
583 assert(m->pmatch[i].rm_so != -1);
584 len = m->pmatch[i].rm_eo - m->pmatch[i].rm_so;
585 if (len == 0 && rec++ > MAX_RECURSION)
586 return(NULL);
587 assert(stop - m->beginp >= len);
588 if (sp > stop - len)
589 return(NULL); /* not enough left to match */
590 ssp = m->offp + m->pmatch[i].rm_so;
591 if (memcmp(sp, ssp, len) != 0)
592 return(NULL);
593 while (m->g->strip[ss] != SOP(O_BACK, i))
594 ss++;
595 return(backref(m, sp+len, stop, ss+1, stopst, lev, rec));
596 break;
597 case OQUEST_: /* to null or not */
598 dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
599 if (dp != NULL)
600 return(dp); /* not */
601 return(backref(m, sp, stop, ss+OPND(s)+1, stopst, lev, rec));
602 break;
603 case OPLUS_:
604 assert(m->lastpos != NULL);
605 assert(lev+1 <= m->g->nplus);
606 m->lastpos[lev+1] = sp;
607 return(backref(m, sp, stop, ss+1, stopst, lev+1, rec));
608 break;
609 case O_PLUS:
610 if (sp == m->lastpos[lev]) /* last pass matched null */
611 return(backref(m, sp, stop, ss+1, stopst, lev-1, rec));
612 /* try another pass */
613 m->lastpos[lev] = sp;
614 dp = backref(m, sp, stop, ss-OPND(s)+1, stopst, lev, rec);
615 if (dp == NULL)
616 return(backref(m, sp, stop, ss+1, stopst, lev-1, rec));
617 else
618 return(dp);
619 break;
620 case OCH_: /* find the right one, if any */
621 ssub = ss + 1;
622 esub = ss + OPND(s) - 1;
623 assert(OP(m->g->strip[esub]) == OOR1);
624 for (;;) { /* find first matching branch */
625 dp = backref(m, sp, stop, ssub, esub, lev, rec);
626 if (dp != NULL)
627 return(dp);
628 /* that one missed, try next one */
629 if (OP(m->g->strip[esub]) == O_CH)
630 return(NULL); /* there is none */
631 esub++;
632 assert(OP(m->g->strip[esub]) == OOR2);
633 ssub = esub + 1;
634 esub += OPND(m->g->strip[esub]);
635 if (OP(m->g->strip[esub]) == OOR2)
636 esub--;
637 else
638 assert(OP(m->g->strip[esub]) == O_CH);
639 }
640 break;
641 case OLPAREN: /* must undo assignment if rest fails */
642 i = OPND(s);
643 assert(0 < i && i <= m->g->nsub);
644 offsave = m->pmatch[i].rm_so;
645 m->pmatch[i].rm_so = sp - m->offp;
646 dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
647 if (dp != NULL)
648 return(dp);
649 m->pmatch[i].rm_so = offsave;
650 return(NULL);
651 break;
652 case ORPAREN: /* must undo assignment if rest fails */
653 i = OPND(s);
654 assert(0 < i && i <= m->g->nsub);
655 offsave = m->pmatch[i].rm_eo;
656 m->pmatch[i].rm_eo = sp - m->offp;
657 dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
658 if (dp != NULL)
659 return(dp);
660 m->pmatch[i].rm_eo = offsave;
661 return(NULL);
662 break;
663 default: /* uh oh */
664 assert(nope);
665 break;
666 }
668 /* "can't happen" */
669 assert(nope);
670 /* NOTREACHED */
671 return NULL;
672 }
674 /*
675 - fast - step through the string at top speed
676 */
677 static char * /* where tentative match ended, or NULL */
678 fast(struct match *m, char *start, char *stop, sopno startst, sopno stopst)
679 {
680 states st = m->st;
681 states fresh = m->fresh;
682 states tmp = m->tmp;
683 char *p = start;
684 int c = (start == m->beginp) ? OUT : *(start-1);
685 int lastc; /* previous c */
686 int flagch;
687 int i;
688 char *coldp; /* last p after which no match was underway */
690 CLEAR(st);
691 SET1(st, startst);
692 st = step(m->g, startst, stopst, st, NOTHING, st);
693 ASSIGN(fresh, st);
694 SP("start", st, *p);
695 coldp = NULL;
696 for (;;) {
697 /* next character */
698 lastc = c;
699 c = (p == m->endp) ? OUT : *p;
700 if (EQ(st, fresh))
701 coldp = p;
703 /* is there an EOL and/or BOL between lastc and c? */
704 flagch = '\0';
705 i = 0;
706 if ( (lastc == '\n' && m->g->cflags®_NEWLINE) ||
707 (lastc == OUT && !(m->eflags®_NOTBOL)) ) {
708 flagch = BOL;
709 i = m->g->nbol;
710 }
711 if ( (c == '\n' && m->g->cflags®_NEWLINE) ||
712 (c == OUT && !(m->eflags®_NOTEOL)) ) {
713 flagch = (flagch == BOL) ? BOLEOL : EOL;
714 i += m->g->neol;
715 }
716 if (i != 0) {
717 for (; i > 0; i--)
718 st = step(m->g, startst, stopst, st, flagch, st);
719 SP("boleol", st, c);
720 }
722 /* how about a word boundary? */
723 if ( (flagch == BOL || (lastc != OUT && !ISWORD(lastc))) &&
724 (c != OUT && ISWORD(c)) ) {
725 flagch = BOW;
726 }
727 if ( (lastc != OUT && ISWORD(lastc)) &&
728 (flagch == EOL || (c != OUT && !ISWORD(c))) ) {
729 flagch = EOW;
730 }
731 if (flagch == BOW || flagch == EOW) {
732 st = step(m->g, startst, stopst, st, flagch, st);
733 SP("boweow", st, c);
734 }
736 /* are we done? */
737 if (ISSET(st, stopst) || p == stop)
738 break; /* NOTE BREAK OUT */
740 /* no, we must deal with this character */
741 ASSIGN(tmp, st);
742 ASSIGN(st, fresh);
743 assert(c != OUT);
744 st = step(m->g, startst, stopst, tmp, c, st);
745 SP("aft", st, c);
746 assert(EQ(step(m->g, startst, stopst, st, NOTHING, st), st));
747 p++;
748 }
750 assert(coldp != NULL);
751 m->coldp = coldp;
752 if (ISSET(st, stopst))
753 return(p+1);
754 else
755 return(NULL);
756 }
758 /*
759 - slow - step through the string more deliberately
760 */
761 static char * /* where it ended */
762 slow(struct match *m, char *start, char *stop, sopno startst, sopno stopst)
763 {
764 states st = m->st;
765 states empty = m->empty;
766 states tmp = m->tmp;
767 char *p = start;
768 int c = (start == m->beginp) ? OUT : *(start-1);
769 int lastc; /* previous c */
770 int flagch;
771 int i;
772 char *matchp; /* last p at which a match ended */
774 AT("slow", start, stop, startst, stopst);
775 CLEAR(st);
776 SET1(st, startst);
777 SP("sstart", st, *p);
778 st = step(m->g, startst, stopst, st, NOTHING, st);
779 matchp = NULL;
780 for (;;) {
781 /* next character */
782 lastc = c;
783 c = (p == m->endp) ? OUT : *p;
785 /* is there an EOL and/or BOL between lastc and c? */
786 flagch = '\0';
787 i = 0;
788 if ( (lastc == '\n' && m->g->cflags®_NEWLINE) ||
789 (lastc == OUT && !(m->eflags®_NOTBOL)) ) {
790 flagch = BOL;
791 i = m->g->nbol;
792 }
793 if ( (c == '\n' && m->g->cflags®_NEWLINE) ||
794 (c == OUT && !(m->eflags®_NOTEOL)) ) {
795 flagch = (flagch == BOL) ? BOLEOL : EOL;
796 i += m->g->neol;
797 }
798 if (i != 0) {
799 for (; i > 0; i--)
800 st = step(m->g, startst, stopst, st, flagch, st);
801 SP("sboleol", st, c);
802 }
804 /* how about a word boundary? */
805 if ( (flagch == BOL || (lastc != OUT && !ISWORD(lastc))) &&
806 (c != OUT && ISWORD(c)) ) {
807 flagch = BOW;
808 }
809 if ( (lastc != OUT && ISWORD(lastc)) &&
810 (flagch == EOL || (c != OUT && !ISWORD(c))) ) {
811 flagch = EOW;
812 }
813 if (flagch == BOW || flagch == EOW) {
814 st = step(m->g, startst, stopst, st, flagch, st);
815 SP("sboweow", st, c);
816 }
818 /* are we done? */
819 if (ISSET(st, stopst))
820 matchp = p;
821 if (EQ(st, empty) || p == stop)
822 break; /* NOTE BREAK OUT */
824 /* no, we must deal with this character */
825 ASSIGN(tmp, st);
826 ASSIGN(st, empty);
827 assert(c != OUT);
828 st = step(m->g, startst, stopst, tmp, c, st);
829 SP("saft", st, c);
830 assert(EQ(step(m->g, startst, stopst, st, NOTHING, st), st));
831 p++;
832 }
834 return(matchp);
835 }
838 /*
839 - step - map set of states reachable before char to set reachable after
840 */
841 static states
842 step(struct re_guts *g,
843 sopno start, /* start state within strip */
844 sopno stop, /* state after stop state within strip */
845 states bef, /* states reachable before */
846 int ch, /* character or NONCHAR code */
847 states aft) /* states already known reachable after */
848 {
849 cset *cs;
850 sop s;
851 sopno pc;
852 onestate here; /* note, macros know this name */
853 sopno look;
854 int i;
856 for (pc = start, INIT(here, pc); pc != stop; pc++, INC(here)) {
857 s = g->strip[pc];
858 switch (OP(s)) {
859 case OEND:
860 assert(pc == stop-1);
861 break;
862 case OCHAR:
863 /* only characters can match */
864 assert(!NONCHAR(ch) || ch != (char)OPND(s));
865 if (ch == (char)OPND(s))
866 FWD(aft, bef, 1);
867 break;
868 case OBOL:
869 if (ch == BOL || ch == BOLEOL)
870 FWD(aft, bef, 1);
871 break;
872 case OEOL:
873 if (ch == EOL || ch == BOLEOL)
874 FWD(aft, bef, 1);
875 break;
876 case OBOW:
877 if (ch == BOW)
878 FWD(aft, bef, 1);
879 break;
880 case OEOW:
881 if (ch == EOW)
882 FWD(aft, bef, 1);
883 break;
884 case OANY:
885 if (!NONCHAR(ch))
886 FWD(aft, bef, 1);
887 break;
888 case OANYOF:
889 cs = &g->sets[OPND(s)];
890 if (!NONCHAR(ch) && CHIN(cs, ch))
891 FWD(aft, bef, 1);
892 break;
893 case OBACK_: /* ignored here */
894 case O_BACK:
895 FWD(aft, aft, 1);
896 break;
897 case OPLUS_: /* forward, this is just an empty */
898 FWD(aft, aft, 1);
899 break;
900 case O_PLUS: /* both forward and back */
901 FWD(aft, aft, 1);
902 i = ISSETBACK(aft, OPND(s));
903 BACK(aft, aft, OPND(s));
904 if (!i && ISSETBACK(aft, OPND(s))) {
905 /* oho, must reconsider loop body */
906 pc -= OPND(s) + 1;
907 INIT(here, pc);
908 }
909 break;
910 case OQUEST_: /* two branches, both forward */
911 FWD(aft, aft, 1);
912 FWD(aft, aft, OPND(s));
913 break;
914 case O_QUEST: /* just an empty */
915 FWD(aft, aft, 1);
916 break;
917 case OLPAREN: /* not significant here */
918 case ORPAREN:
919 FWD(aft, aft, 1);
920 break;
921 case OCH_: /* mark the first two branches */
922 FWD(aft, aft, 1);
923 assert(OP(g->strip[pc+OPND(s)]) == OOR2);
924 FWD(aft, aft, OPND(s));
925 break;
926 case OOR1: /* done a branch, find the O_CH */
927 if (ISSTATEIN(aft, here)) {
928 for (look = 1;
929 OP(s = g->strip[pc+look]) != O_CH;
930 look += OPND(s))
931 assert(OP(s) == OOR2);
932 FWD(aft, aft, look);
933 }
934 break;
935 case OOR2: /* propagate OCH_'s marking */
936 FWD(aft, aft, 1);
937 if (OP(g->strip[pc+OPND(s)]) != O_CH) {
938 assert(OP(g->strip[pc+OPND(s)]) == OOR2);
939 FWD(aft, aft, OPND(s));
940 }
941 break;
942 case O_CH: /* just empty */
943 FWD(aft, aft, 1);
944 break;
945 default: /* ooooops... */
946 assert(nope);
947 break;
948 }
949 }
951 return(aft);
952 }
954 #ifdef REDEBUG
955 /*
956 - print - print a set of states
957 */
958 static void
959 print(struct match *m, char *caption, states st, int ch, FILE *d)
960 {
961 struct re_guts *g = m->g;
962 int i;
963 int first = 1;
965 if (!(m->eflags®_TRACE))
966 return;
968 (void)fprintf(d, "%s", caption);
969 if (ch != '\0')
970 (void)fprintf(d, " %s", pchar(ch));
971 for (i = 0; i < g->nstates; i++)
972 if (ISSET(st, i)) {
973 (void)fprintf(d, "%s%d", (first) ? "\t" : ", ", i);
974 first = 0;
975 }
976 (void)fprintf(d, "\n");
977 }
979 /*
980 - at - print current situation
981 */
982 static void
983 at(struct match *m, char *title, char *start, char *stop, sopno startst,
984 sopno stopst)
985 {
986 if (!(m->eflags®_TRACE))
987 return;
989 (void)printf("%s %s-", title, pchar(*start));
990 (void)printf("%s ", pchar(*stop));
991 (void)printf("%ld-%ld\n", (long)startst, (long)stopst);
992 }
994 #ifndef PCHARDONE
995 #define PCHARDONE /* never again */
996 /*
997 - pchar - make a character printable
998 *
999 * Is this identical to regchar() over in debug.c? Well, yes. But a
1000 * duplicate here avoids having a debugging-capable regexec.o tied to
1001 * a matching debug.o, and this is convenient. It all disappears in
1002 * the non-debug compilation anyway, so it doesn't matter much.
1003 */
1004 static char * /* -> representation */
1005 pchar(int ch)
1006 {
1007 static char pbuf[10];
1009 if (isprint(ch) || ch == ' ')
1010 (void)snprintf(pbuf, sizeof pbuf, "%c", ch);
1011 else
1012 (void)snprintf(pbuf, sizeof pbuf, "\\%o", ch);
1013 return(pbuf);
1014 }
1015 #endif
1016 #endif
1018 #undef matcher
1019 #undef fast
1020 #undef slow
1021 #undef dissect
1022 #undef backref
1023 #undef step
1024 #undef print
1025 #undef at
1026 #undef match
1027 #undef nope