xref: /freebsd/contrib/llvm-project/llvm/lib/Support/regengine.inc (revision c66ec88fed842fbaad62c30d510644ceb7bd2d71)
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 */
37
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 */
44
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
69
70/* another structure passed up and down to avoid zillions of parameters */
71struct match {
72	struct re_guts *g;
73	int eflags;
74	llvm_regmatch_t *pmatch;	/* [nsub+1] (0 element unused) */
75	const char *offp;		/* offsets work from here */
76	const char *beginp;		/* start of string -- virtual NUL precedes */
77	const char *endp;		/* end of string -- virtual NUL here */
78	const char *coldp;		/* can be no match starting before here */
79	const 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};
86
87static int matcher(struct re_guts *, const char *, size_t,
88                   llvm_regmatch_t[], int);
89static const char *dissect(struct match *, const char *, const char *, sopno,
90                           sopno);
91static const char *backref(struct match *, const char *, const char *, sopno,
92                           sopno, sopno, int);
93static const char *fast(struct match *, const char *, const char *, sopno, sopno);
94static const char *slow(struct match *, const char *, const char *, sopno, sopno);
95static states step(struct re_guts *, sopno, sopno, states, int, states);
96#define MAX_RECURSION	100
97#define	BOL	(OUT+1)
98#define	EOL	(BOL+1)
99#define	BOLEOL	(BOL+2)
100#define	NOTHING	(BOL+3)
101#define	BOW	(BOL+4)
102#define	EOW	(BOL+5)
103#define	CODEMAX	(BOL+5)		/* highest code used */
104#define	NONCHAR(c)	((c) > CHAR_MAX)
105#define	NNONCHAR	(CODEMAX-CHAR_MAX)
106#ifdef REDEBUG
107static void print(struct match *, char *, states, int, FILE *);
108#endif
109#ifdef REDEBUG
110static void at(struct match *, char *, char *, char *, sopno, sopno);
111#endif
112#ifdef REDEBUG
113static char *pchar(int);
114#endif
115
116#ifdef REDEBUG
117#define	SP(t, s, c)	print(m, t, s, c, stdout)
118#define	AT(t, p1, p2, s1, s2)	at(m, t, p1, p2, s1, s2)
119#define	NOTE(str)	{ if (m->eflags&REG_TRACE) (void)printf("=%s\n", (str)); }
120static int nope = 0;
121#else
122#define	SP(t, s, c)	/* nothing */
123#define	AT(t, p1, p2, s1, s2)	/* nothing */
124#define	NOTE(s)	/* nothing */
125#endif
126
127/*
128 - matcher - the actual matching engine
129 */
130static int			/* 0 success, REG_NOMATCH failure */
131matcher(struct re_guts *g, const char *string, size_t nmatch,
132        llvm_regmatch_t pmatch[],
133    int eflags)
134{
135	const char *endp;
136	size_t i;
137	struct match mv;
138	struct match *m = &mv;
139	const char *dp;
140	const sopno gf = g->firststate+1;	/* +1 for OEND */
141	const sopno gl = g->laststate;
142	const char *start;
143	const char *stop;
144
145	/* simplify the situation where possible */
146	if (g->cflags&REG_NOSUB)
147		nmatch = 0;
148	if (eflags&REG_STARTEND) {
149		start = string + pmatch[0].rm_so;
150		stop = string + pmatch[0].rm_eo;
151	} else {
152		start = string;
153		stop = start + strlen(start);
154	}
155	if (stop < start)
156		return(REG_INVARG);
157
158	/* prescreening; this does wonders for this rather slow code */
159	if (g->must != NULL) {
160		for (dp = start; dp < stop; dp++)
161			if (*dp == g->must[0] && stop - dp >= g->mlen &&
162				memcmp(dp, g->must, (size_t)g->mlen) == 0)
163				break;
164		if (dp == stop)		/* we didn't find g->must */
165			return(REG_NOMATCH);
166	}
167
168	/* match struct setup */
169	m->g = g;
170	m->eflags = eflags;
171	m->pmatch = NULL;
172	m->lastpos = NULL;
173	m->offp = string;
174	m->beginp = start;
175	m->endp = stop;
176	STATESETUP(m, 4);
177	SETUP(m->st);
178	SETUP(m->fresh);
179	SETUP(m->tmp);
180	SETUP(m->empty);
181	CLEAR(m->empty);
182
183	/* this loop does only one repetition except for backrefs */
184	for (;;) {
185		endp = fast(m, start, stop, gf, gl);
186		if (endp == NULL) {		/* a miss */
187			free(m->pmatch);
188			free((void*)m->lastpos);
189			STATETEARDOWN(m);
190			return(REG_NOMATCH);
191		}
192		if (nmatch == 0 && !g->backrefs)
193			break;		/* no further info needed */
194
195		/* where? */
196		assert(m->coldp != NULL);
197		for (;;) {
198			NOTE("finding start");
199			endp = slow(m, m->coldp, stop, gf, gl);
200			if (endp != NULL)
201				break;
202			assert(m->coldp < m->endp);
203			m->coldp++;
204		}
205		if (nmatch == 1 && !g->backrefs)
206			break;		/* no further info needed */
207
208		/* oh my, they want the subexpressions... */
209		if (m->pmatch == NULL)
210			m->pmatch = (llvm_regmatch_t *)malloc((m->g->nsub + 1) *
211							sizeof(llvm_regmatch_t));
212		if (m->pmatch == NULL) {
213			STATETEARDOWN(m);
214			return(REG_ESPACE);
215		}
216		for (i = 1; i <= m->g->nsub; i++)
217			m->pmatch[i].rm_so = m->pmatch[i].rm_eo = -1;
218		if (!g->backrefs && !(m->eflags&REG_BACKR)) {
219			NOTE("dissecting");
220			dp = dissect(m, m->coldp, endp, gf, gl);
221		} else {
222			if (g->nplus > 0 && m->lastpos == NULL)
223				m->lastpos = (const char **)malloc((g->nplus+1) *
224							sizeof(char *));
225			if (g->nplus > 0 && m->lastpos == NULL) {
226				free(m->pmatch);
227				STATETEARDOWN(m);
228				return(REG_ESPACE);
229			}
230			NOTE("backref dissect");
231			dp = backref(m, m->coldp, endp, gf, gl, (sopno)0, 0);
232		}
233		if (dp != NULL)
234			break;
235
236		/* uh-oh... we couldn't find a subexpression-level match */
237		assert(g->backrefs);	/* must be back references doing it */
238		assert(g->nplus == 0 || m->lastpos != NULL);
239		for (;;) {
240			if (dp != NULL || endp <= m->coldp)
241				break;		/* defeat */
242			NOTE("backoff");
243			endp = slow(m, m->coldp, endp-1, gf, gl);
244			if (endp == NULL)
245				break;		/* defeat */
246			/* try it on a shorter possibility */
247#ifndef NDEBUG
248			for (i = 1; i <= m->g->nsub; i++) {
249				assert(m->pmatch[i].rm_so == -1);
250				assert(m->pmatch[i].rm_eo == -1);
251			}
252#endif
253			NOTE("backoff dissect");
254			dp = backref(m, m->coldp, endp, gf, gl, (sopno)0, 0);
255		}
256		assert(dp == NULL || dp == endp);
257		if (dp != NULL)		/* found a shorter one */
258			break;
259
260		/* despite initial appearances, there is no match here */
261		NOTE("false alarm");
262		if (m->coldp == stop)
263			break;
264		start = m->coldp + 1;	/* recycle starting later */
265	}
266
267	/* fill in the details if requested */
268	if (nmatch > 0) {
269		pmatch[0].rm_so = m->coldp - m->offp;
270		pmatch[0].rm_eo = endp - m->offp;
271	}
272	if (nmatch > 1) {
273		assert(m->pmatch != NULL);
274		for (i = 1; i < nmatch; i++)
275			if (i <= m->g->nsub)
276				pmatch[i] = m->pmatch[i];
277			else {
278				pmatch[i].rm_so = -1;
279				pmatch[i].rm_eo = -1;
280			}
281	}
282
283	if (m->pmatch != NULL)
284		free((char *)m->pmatch);
285	if (m->lastpos != NULL)
286		free((char *)m->lastpos);
287	STATETEARDOWN(m);
288	return(0);
289}
290
291/*
292 - dissect - figure out what matched what, no back references
293 */
294static const char *			/* == stop (success) always */
295dissect(struct match *m, const char *start, const char *stop, sopno startst,
296        sopno stopst)
297{
298	int i;
299	sopno ss;	/* start sop of current subRE */
300	sopno es;	/* end sop of current subRE */
301	const char *sp;	/* start of string matched by it */
302	const char *stp;	/* string matched by it cannot pass here */
303	const char *rest;	/* start of rest of string */
304	const char *tail;	/* string unmatched by rest of RE */
305	sopno ssub;	/* start sop of subsubRE */
306	sopno esub;	/* end sop of subsubRE */
307	const char *ssp;	/* start of string matched by subsubRE */
308	const char *sep;	/* end of string matched by subsubRE */
309	const char *oldssp;	/* previous ssp */
310
311	AT("diss", start, stop, startst, stopst);
312	sp = start;
313	for (ss = startst; ss < stopst; ss = es) {
314		/* identify end of subRE */
315		es = ss;
316		switch (OP(m->g->strip[es])) {
317		case OPLUS_:
318		case OQUEST_:
319			es += OPND(m->g->strip[es]);
320			break;
321		case OCH_:
322			while (OP(m->g->strip[es]) != O_CH)
323				es += OPND(m->g->strip[es]);
324			break;
325		}
326		es++;
327
328		/* figure out what it matched */
329		switch (OP(m->g->strip[ss])) {
330		case OEND:
331			assert(nope);
332			break;
333		case OCHAR:
334			sp++;
335			break;
336		case OBOL:
337		case OEOL:
338		case OBOW:
339		case OEOW:
340			break;
341		case OANY:
342		case OANYOF:
343			sp++;
344			break;
345		case OBACK_:
346		case O_BACK:
347			assert(nope);
348			break;
349		/* cases where length of match is hard to find */
350		case OQUEST_:
351			stp = stop;
352			for (;;) {
353				/* how long could this one be? */
354				rest = slow(m, sp, stp, ss, es);
355				assert(rest != NULL);	/* it did match */
356				/* could the rest match the rest? */
357				tail = slow(m, rest, stop, es, stopst);
358				if (tail == stop)
359					break;		/* yes! */
360				/* no -- try a shorter match for this one */
361				stp = rest - 1;
362				assert(stp >= sp);	/* it did work */
363			}
364			ssub = ss + 1;
365			esub = es - 1;
366			/* did innards match? */
367			if (slow(m, sp, rest, ssub, esub) != NULL) {
368				const char *dp = dissect(m, sp, rest, ssub, esub);
369				(void)dp; /* avoid warning if assertions off */
370				assert(dp == rest);
371			} else		/* no */
372				assert(sp == rest);
373			sp = rest;
374			break;
375		case OPLUS_:
376			stp = stop;
377			for (;;) {
378				/* how long could this one be? */
379				rest = slow(m, sp, stp, ss, es);
380				assert(rest != NULL);	/* it did match */
381				/* could the rest match the rest? */
382				tail = slow(m, rest, stop, es, stopst);
383				if (tail == stop)
384					break;		/* yes! */
385				/* no -- try a shorter match for this one */
386				stp = rest - 1;
387				assert(stp >= sp);	/* it did work */
388			}
389			ssub = ss + 1;
390			esub = es - 1;
391			ssp = sp;
392			oldssp = ssp;
393			for (;;) {	/* find last match of innards */
394				sep = slow(m, ssp, rest, ssub, esub);
395				if (sep == NULL || sep == ssp)
396					break;	/* failed or matched null */
397				oldssp = ssp;	/* on to next try */
398				ssp = sep;
399			}
400			if (sep == NULL) {
401				/* last successful match */
402				sep = ssp;
403				ssp = oldssp;
404			}
405			assert(sep == rest);	/* must exhaust substring */
406			assert(slow(m, ssp, sep, ssub, esub) == rest);
407			{
408				const char *dp = dissect(m, ssp, sep, ssub, esub);
409				(void)dp; /* avoid warning if assertions off */
410				assert(dp == sep);
411			}
412			sp = rest;
413			break;
414		case OCH_:
415			stp = stop;
416			for (;;) {
417				/* how long could this one be? */
418				rest = slow(m, sp, stp, ss, es);
419				assert(rest != NULL);	/* it did match */
420				/* could the rest match the rest? */
421				tail = slow(m, rest, stop, es, stopst);
422				if (tail == stop)
423					break;		/* yes! */
424				/* no -- try a shorter match for this one */
425				stp = rest - 1;
426				assert(stp >= sp);	/* it did work */
427			}
428			ssub = ss + 1;
429			esub = ss + OPND(m->g->strip[ss]) - 1;
430			assert(OP(m->g->strip[esub]) == OOR1);
431			for (;;) {	/* find first matching branch */
432				if (slow(m, sp, rest, ssub, esub) == rest)
433					break;	/* it matched all of it */
434				/* that one missed, try next one */
435				assert(OP(m->g->strip[esub]) == OOR1);
436				esub++;
437				assert(OP(m->g->strip[esub]) == OOR2);
438				ssub = esub + 1;
439				esub += OPND(m->g->strip[esub]);
440				if (OP(m->g->strip[esub]) == OOR2)
441					esub--;
442				else
443					assert(OP(m->g->strip[esub]) == O_CH);
444			}
445			{
446				const char *dp = dissect(m, sp, rest, ssub, esub);
447				(void)dp; /* avoid warning if assertions off */
448				assert(dp == rest);
449			}
450			sp = rest;
451			break;
452		case O_PLUS:
453		case O_QUEST:
454		case OOR1:
455		case OOR2:
456		case O_CH:
457			assert(nope);
458			break;
459		case OLPAREN:
460			i = OPND(m->g->strip[ss]);
461			assert(0 < i && i <= m->g->nsub);
462			m->pmatch[i].rm_so = sp - m->offp;
463			break;
464		case ORPAREN:
465			i = OPND(m->g->strip[ss]);
466			assert(0 < i && i <= m->g->nsub);
467			m->pmatch[i].rm_eo = sp - m->offp;
468			break;
469		default:		/* uh oh */
470			assert(nope);
471			break;
472		}
473	}
474
475	assert(sp == stop);
476	return(sp);
477}
478
479/*
480 - backref - figure out what matched what, figuring in back references
481 */
482static const char *			/* == stop (success) or NULL (failure) */
483backref(struct match *m, const char *start, const char *stop, sopno startst,
484        sopno stopst, sopno lev, int rec)			/* PLUS nesting level */
485{
486	int i;
487	sopno ss;	/* start sop of current subRE */
488	const char *sp;	/* start of string matched by it */
489	sopno ssub;	/* start sop of subsubRE */
490	sopno esub;	/* end sop of subsubRE */
491	const char *ssp;	/* start of string matched by subsubRE */
492	const char *dp;
493	size_t len;
494	int hard;
495	sop s;
496	llvm_regoff_t offsave;
497	cset *cs;
498
499	AT("back", start, stop, startst, stopst);
500	sp = start;
501
502	/* get as far as we can with easy stuff */
503	hard = 0;
504	for (ss = startst; !hard && ss < stopst; ss++)
505		switch (OP(s = m->g->strip[ss])) {
506		case OCHAR:
507			if (sp == stop || *sp++ != (char)OPND(s))
508				return(NULL);
509			break;
510		case OANY:
511			if (sp == stop)
512				return(NULL);
513			sp++;
514			break;
515		case OANYOF:
516			cs = &m->g->sets[OPND(s)];
517			if (sp == stop || !CHIN(cs, *sp++))
518				return(NULL);
519			break;
520		case OBOL:
521			if ( (sp == m->beginp && !(m->eflags&REG_NOTBOL)) ||
522					(sp < m->endp && *(sp-1) == '\n' &&
523						(m->g->cflags&REG_NEWLINE)) )
524				{ /* yes */ }
525			else
526				return(NULL);
527			break;
528		case OEOL:
529			if ( (sp == m->endp && !(m->eflags&REG_NOTEOL)) ||
530					(sp < m->endp && *sp == '\n' &&
531						(m->g->cflags&REG_NEWLINE)) )
532				{ /* yes */ }
533			else
534				return(NULL);
535			break;
536		case OBOW:
537			if (( (sp == m->beginp && !(m->eflags&REG_NOTBOL)) ||
538					(sp < m->endp && *(sp-1) == '\n' &&
539						(m->g->cflags&REG_NEWLINE)) ||
540					(sp > m->beginp &&
541							!ISWORD(*(sp-1))) ) &&
542					(sp < m->endp && ISWORD(*sp)) )
543				{ /* yes */ }
544			else
545				return(NULL);
546			break;
547		case OEOW:
548			if (( (sp == m->endp && !(m->eflags&REG_NOTEOL)) ||
549					(sp < m->endp && *sp == '\n' &&
550						(m->g->cflags&REG_NEWLINE)) ||
551					(sp < m->endp && !ISWORD(*sp)) ) &&
552					(sp > m->beginp && ISWORD(*(sp-1))) )
553				{ /* yes */ }
554			else
555				return(NULL);
556			break;
557		case O_QUEST:
558			break;
559		case OOR1:	/* matches null but needs to skip */
560			ss++;
561			s = m->g->strip[ss];
562			do {
563				assert(OP(s) == OOR2);
564				ss += OPND(s);
565			} while (OP(s = m->g->strip[ss]) != O_CH);
566			/* note that the ss++ gets us past the O_CH */
567			break;
568		default:	/* have to make a choice */
569			hard = 1;
570			break;
571		}
572	if (!hard) {		/* that was it! */
573		if (sp != stop)
574			return(NULL);
575		return(sp);
576	}
577	ss--;			/* adjust for the for's final increment */
578
579	/* the hard stuff */
580	AT("hard", sp, stop, ss, stopst);
581	s = m->g->strip[ss];
582	switch (OP(s)) {
583	case OBACK_:		/* the vilest depths */
584		i = OPND(s);
585		assert(0 < i && i <= m->g->nsub);
586		if (m->pmatch[i].rm_eo == -1)
587			return(NULL);
588		assert(m->pmatch[i].rm_so != -1);
589		len = m->pmatch[i].rm_eo - m->pmatch[i].rm_so;
590		if (len == 0 && rec++ > MAX_RECURSION)
591			return(NULL);
592		assert(stop - m->beginp >= len);
593		if (sp > stop - len)
594			return(NULL);	/* not enough left to match */
595		ssp = m->offp + m->pmatch[i].rm_so;
596		if (memcmp(sp, ssp, len) != 0)
597			return(NULL);
598		while (m->g->strip[ss] != SOP(O_BACK, i))
599			ss++;
600		return(backref(m, sp+len, stop, ss+1, stopst, lev, rec));
601		break;
602	case OQUEST_:		/* to null or not */
603		dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
604		if (dp != NULL)
605			return(dp);	/* not */
606		return(backref(m, sp, stop, ss+OPND(s)+1, stopst, lev, rec));
607		break;
608	case OPLUS_:
609		assert(m->lastpos != NULL);
610		assert(lev+1 <= m->g->nplus);
611		m->lastpos[lev+1] = sp;
612		return(backref(m, sp, stop, ss+1, stopst, lev+1, rec));
613		break;
614	case O_PLUS:
615		if (sp == m->lastpos[lev])	/* last pass matched null */
616			return(backref(m, sp, stop, ss+1, stopst, lev-1, rec));
617		/* try another pass */
618		m->lastpos[lev] = sp;
619		dp = backref(m, sp, stop, ss-OPND(s)+1, stopst, lev, rec);
620		if (dp == NULL)
621			return(backref(m, sp, stop, ss+1, stopst, lev-1, rec));
622		else
623			return(dp);
624		break;
625	case OCH_:		/* find the right one, if any */
626		ssub = ss + 1;
627		esub = ss + OPND(s) - 1;
628		assert(OP(m->g->strip[esub]) == OOR1);
629		for (;;) {	/* find first matching branch */
630			dp = backref(m, sp, stop, ssub, esub, lev, rec);
631			if (dp != NULL)
632				return(dp);
633			/* that one missed, try next one */
634			if (OP(m->g->strip[esub]) == O_CH)
635				return(NULL);	/* there is none */
636			esub++;
637			assert(OP(m->g->strip[esub]) == OOR2);
638			ssub = esub + 1;
639			esub += OPND(m->g->strip[esub]);
640			if (OP(m->g->strip[esub]) == OOR2)
641				esub--;
642			else
643				assert(OP(m->g->strip[esub]) == O_CH);
644		}
645		break;
646	case OLPAREN:		/* must undo assignment if rest fails */
647		i = OPND(s);
648		assert(0 < i && i <= m->g->nsub);
649		offsave = m->pmatch[i].rm_so;
650		m->pmatch[i].rm_so = sp - m->offp;
651		dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
652		if (dp != NULL)
653			return(dp);
654		m->pmatch[i].rm_so = offsave;
655		return(NULL);
656		break;
657	case ORPAREN:		/* must undo assignment if rest fails */
658		i = OPND(s);
659		assert(0 < i && i <= m->g->nsub);
660		offsave = m->pmatch[i].rm_eo;
661		m->pmatch[i].rm_eo = sp - m->offp;
662		dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
663		if (dp != NULL)
664			return(dp);
665		m->pmatch[i].rm_eo = offsave;
666		return(NULL);
667		break;
668	default:		/* uh oh */
669		assert(nope);
670		break;
671	}
672
673	/* "can't happen" */
674	assert(nope);
675	/* NOTREACHED */
676        return NULL;
677}
678
679/*
680 - fast - step through the string at top speed
681 */
682static const char *			/* where tentative match ended, or NULL */
683fast(struct match *m, const char *start, const char *stop, sopno startst,
684     sopno stopst)
685{
686	states st = m->st;
687	states fresh = m->fresh;
688	states tmp = m->tmp;
689	const char *p = start;
690	int c = (start == m->beginp) ? OUT : *(start-1);
691	int lastc;	/* previous c */
692	int flagch;
693	int i;
694	const char *coldp;	/* last p after which no match was underway */
695
696	CLEAR(st);
697	SET1(st, startst);
698	st = step(m->g, startst, stopst, st, NOTHING, st);
699	ASSIGN(fresh, st);
700	SP("start", st, *p);
701	coldp = NULL;
702	for (;;) {
703		/* next character */
704		lastc = c;
705		c = (p == m->endp) ? OUT : *p;
706		if (EQ(st, fresh))
707			coldp = p;
708
709		/* is there an EOL and/or BOL between lastc and c? */
710		flagch = '\0';
711		i = 0;
712		if ( (lastc == '\n' && m->g->cflags&REG_NEWLINE) ||
713				(lastc == OUT && !(m->eflags&REG_NOTBOL)) ) {
714			flagch = BOL;
715			i = m->g->nbol;
716		}
717		if ( (c == '\n' && m->g->cflags&REG_NEWLINE) ||
718				(c == OUT && !(m->eflags&REG_NOTEOL)) ) {
719			flagch = (flagch == BOL) ? BOLEOL : EOL;
720			i += m->g->neol;
721		}
722		if (i != 0) {
723			for (; i > 0; i--)
724				st = step(m->g, startst, stopst, st, flagch, st);
725			SP("boleol", st, c);
726		}
727
728		/* how about a word boundary? */
729		if ( (flagch == BOL || (lastc != OUT && !ISWORD(lastc))) &&
730					(c != OUT && ISWORD(c)) ) {
731			flagch = BOW;
732		}
733		if ( (lastc != OUT && ISWORD(lastc)) &&
734				(flagch == EOL || (c != OUT && !ISWORD(c))) ) {
735			flagch = EOW;
736		}
737		if (flagch == BOW || flagch == EOW) {
738			st = step(m->g, startst, stopst, st, flagch, st);
739			SP("boweow", st, c);
740		}
741
742		/* are we done? */
743		if (ISSET(st, stopst) || p == stop)
744			break;		/* NOTE BREAK OUT */
745
746		/* no, we must deal with this character */
747		ASSIGN(tmp, st);
748		ASSIGN(st, fresh);
749		assert(c != OUT);
750		st = step(m->g, startst, stopst, tmp, c, st);
751		SP("aft", st, c);
752		assert(EQ(step(m->g, startst, stopst, st, NOTHING, st), st));
753		p++;
754	}
755
756	assert(coldp != NULL);
757	m->coldp = coldp;
758	if (ISSET(st, stopst))
759		return(p+1);
760	else
761		return(NULL);
762}
763
764/*
765 - slow - step through the string more deliberately
766 */
767static const char *			/* where it ended */
768slow(struct match *m, const char *start, const char *stop, sopno startst,
769     sopno stopst)
770{
771	states st = m->st;
772	states empty = m->empty;
773	states tmp = m->tmp;
774	const char *p = start;
775	int c = (start == m->beginp) ? OUT : *(start-1);
776	int lastc;	/* previous c */
777	int flagch;
778	int i;
779	const char *matchp;	/* last p at which a match ended */
780
781	AT("slow", start, stop, startst, stopst);
782	CLEAR(st);
783	SET1(st, startst);
784	SP("sstart", st, *p);
785	st = step(m->g, startst, stopst, st, NOTHING, st);
786	matchp = NULL;
787	for (;;) {
788		/* next character */
789		lastc = c;
790		c = (p == m->endp) ? OUT : *p;
791
792		/* is there an EOL and/or BOL between lastc and c? */
793		flagch = '\0';
794		i = 0;
795		if ( (lastc == '\n' && m->g->cflags&REG_NEWLINE) ||
796				(lastc == OUT && !(m->eflags&REG_NOTBOL)) ) {
797			flagch = BOL;
798			i = m->g->nbol;
799		}
800		if ( (c == '\n' && m->g->cflags&REG_NEWLINE) ||
801				(c == OUT && !(m->eflags&REG_NOTEOL)) ) {
802			flagch = (flagch == BOL) ? BOLEOL : EOL;
803			i += m->g->neol;
804		}
805		if (i != 0) {
806			for (; i > 0; i--)
807				st = step(m->g, startst, stopst, st, flagch, st);
808			SP("sboleol", st, c);
809		}
810
811		/* how about a word boundary? */
812		if ( (flagch == BOL || (lastc != OUT && !ISWORD(lastc))) &&
813					(c != OUT && ISWORD(c)) ) {
814			flagch = BOW;
815		}
816		if ( (lastc != OUT && ISWORD(lastc)) &&
817				(flagch == EOL || (c != OUT && !ISWORD(c))) ) {
818			flagch = EOW;
819		}
820		if (flagch == BOW || flagch == EOW) {
821			st = step(m->g, startst, stopst, st, flagch, st);
822			SP("sboweow", st, c);
823		}
824
825		/* are we done? */
826		if (ISSET(st, stopst))
827			matchp = p;
828		if (EQ(st, empty) || p == stop)
829			break;		/* NOTE BREAK OUT */
830
831		/* no, we must deal with this character */
832		ASSIGN(tmp, st);
833		ASSIGN(st, empty);
834		assert(c != OUT);
835		st = step(m->g, startst, stopst, tmp, c, st);
836		SP("saft", st, c);
837		assert(EQ(step(m->g, startst, stopst, st, NOTHING, st), st));
838		p++;
839	}
840
841	return(matchp);
842}
843
844
845/*
846 - step - map set of states reachable before char to set reachable after
847 */
848static states
849step(struct re_guts *g,
850    sopno start,		/* start state within strip */
851    sopno stop,			/* state after stop state within strip */
852    states bef,			/* states reachable before */
853    int ch,			/* character or NONCHAR code */
854    states aft)			/* states already known reachable after */
855{
856	cset *cs;
857	sop s;
858	sopno pc;
859	onestate here;		/* note, macros know this name */
860	sopno look;
861	int i;
862
863	for (pc = start, INIT(here, pc); pc != stop; pc++, INC(here)) {
864		s = g->strip[pc];
865		switch (OP(s)) {
866		case OEND:
867			assert(pc == stop-1);
868			break;
869		case OCHAR:
870			/* only characters can match */
871			assert(!NONCHAR(ch) || ch != (char)OPND(s));
872			if (ch == (char)OPND(s))
873				FWD(aft, bef, 1);
874			break;
875		case OBOL:
876			if (ch == BOL || ch == BOLEOL)
877				FWD(aft, bef, 1);
878			break;
879		case OEOL:
880			if (ch == EOL || ch == BOLEOL)
881				FWD(aft, bef, 1);
882			break;
883		case OBOW:
884			if (ch == BOW)
885				FWD(aft, bef, 1);
886			break;
887		case OEOW:
888			if (ch == EOW)
889				FWD(aft, bef, 1);
890			break;
891		case OANY:
892			if (!NONCHAR(ch))
893				FWD(aft, bef, 1);
894			break;
895		case OANYOF:
896			cs = &g->sets[OPND(s)];
897			if (!NONCHAR(ch) && CHIN(cs, ch))
898				FWD(aft, bef, 1);
899			break;
900		case OBACK_:		/* ignored here */
901		case O_BACK:
902			FWD(aft, aft, 1);
903			break;
904		case OPLUS_:		/* forward, this is just an empty */
905			FWD(aft, aft, 1);
906			break;
907		case O_PLUS:		/* both forward and back */
908			FWD(aft, aft, 1);
909			i = ISSETBACK(aft, OPND(s));
910			BACK(aft, aft, OPND(s));
911			if (!i && ISSETBACK(aft, OPND(s))) {
912				/* oho, must reconsider loop body */
913				pc -= OPND(s) + 1;
914				INIT(here, pc);
915			}
916			break;
917		case OQUEST_:		/* two branches, both forward */
918			FWD(aft, aft, 1);
919			FWD(aft, aft, OPND(s));
920			break;
921		case O_QUEST:		/* just an empty */
922			FWD(aft, aft, 1);
923			break;
924		case OLPAREN:		/* not significant here */
925		case ORPAREN:
926			FWD(aft, aft, 1);
927			break;
928		case OCH_:		/* mark the first two branches */
929			FWD(aft, aft, 1);
930			assert(OP(g->strip[pc+OPND(s)]) == OOR2);
931			FWD(aft, aft, OPND(s));
932			break;
933		case OOR1:		/* done a branch, find the O_CH */
934			if (ISSTATEIN(aft, here)) {
935				for (look = 1;
936						OP(s = g->strip[pc+look]) != O_CH;
937						look += OPND(s))
938					assert(OP(s) == OOR2);
939				FWD(aft, aft, look);
940			}
941			break;
942		case OOR2:		/* propagate OCH_'s marking */
943			FWD(aft, aft, 1);
944			if (OP(g->strip[pc+OPND(s)]) != O_CH) {
945				assert(OP(g->strip[pc+OPND(s)]) == OOR2);
946				FWD(aft, aft, OPND(s));
947			}
948			break;
949		case O_CH:		/* just empty */
950			FWD(aft, aft, 1);
951			break;
952		default:		/* ooooops... */
953			assert(nope);
954			break;
955		}
956	}
957
958	return(aft);
959}
960
961#ifdef REDEBUG
962/*
963 - print - print a set of states
964 */
965static void
966print(struct match *m, char *caption, states st, int ch, FILE *d)
967{
968	struct re_guts *g = m->g;
969	int i;
970	int first = 1;
971
972	if (!(m->eflags&REG_TRACE))
973		return;
974
975	(void)fprintf(d, "%s", caption);
976	if (ch != '\0')
977		(void)fprintf(d, " %s", pchar(ch));
978	for (i = 0; i < g->nstates; i++)
979		if (ISSET(st, i)) {
980			(void)fprintf(d, "%s%d", (first) ? "\t" : ", ", i);
981			first = 0;
982		}
983	(void)fprintf(d, "\n");
984}
985
986/*
987 - at - print current situation
988 */
989static void
990at(struct match *m, char *title, char *start, char *stop, sopno startst,
991    sopno stopst)
992{
993	if (!(m->eflags&REG_TRACE))
994		return;
995
996	(void)printf("%s %s-", title, pchar(*start));
997	(void)printf("%s ", pchar(*stop));
998	(void)printf("%ld-%ld\n", (long)startst, (long)stopst);
999}
1000
1001#ifndef PCHARDONE
1002#define	PCHARDONE	/* never again */
1003/*
1004 - pchar - make a character printable
1005 *
1006 * Is this identical to regchar() over in debug.c?  Well, yes.  But a
1007 * duplicate here avoids having a debugging-capable regexec.o tied to
1008 * a matching debug.o, and this is convenient.  It all disappears in
1009 * the non-debug compilation anyway, so it doesn't matter much.
1010 */
1011static char *			/* -> representation */
1012pchar(int ch)
1013{
1014	static char pbuf[10];
1015
1016	if (isPrint(ch) || ch == ' ')
1017		(void)snprintf(pbuf, sizeof pbuf, "%c", ch);
1018	else
1019		(void)snprintf(pbuf, sizeof pbuf, "\\%o", ch);
1020	return(pbuf);
1021}
1022#endif
1023#endif
1024
1025#undef	matcher
1026#undef	fast
1027#undef	slow
1028#undef	dissect
1029#undef	backref
1030#undef	step
1031#undef	print
1032#undef	at
1033#undef	match
1034#undef	nope
1035