2
* This code is derived from OpenBSD's libc/regex, original license follows:
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.
8
* This code is derived from software contributed to Berkeley by
11
* Redistribution and use in source and binary forms, with or without
12
* modification, are permitted provided that the following conditions
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.
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
35
* @(#)engine.c 8.5 (Berkeley) 3/20/94
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
46
#define matcher smatcher
49
#define dissect sdissect
50
#define backref sbackref
58
#define matcher lmatcher
61
#define dissect ldissect
62
#define backref lbackref
70
/* another structure passed up and down to avoid zillions of parameters */
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] */
81
states st; /* current states */
82
states fresh; /* states for a fresh start */
83
states tmp; /* temporary */
84
states empty; /* empty set of states */
87
static int matcher(struct re_guts *, const char *, size_t,
88
llvm_regmatch_t[], int);
89
static const char *dissect(struct match *, const char *, const char *, sopno,
91
static const char *backref(struct match *, const char *, const char *, sopno,
93
static const char *fast(struct match *, const char *, const char *, sopno, sopno);
94
static const char *slow(struct match *, const char *, const char *, sopno, sopno);
95
static states step(struct re_guts *, sopno, sopno, states, int, states);
96
#define MAX_RECURSION 100
99
#define BOLEOL (BOL+2)
100
#define NOTHING (BOL+3)
103
#define CODEMAX (BOL+5) /* highest code used */
104
#define NONCHAR(c) ((c) > CHAR_MAX)
105
#define NNONCHAR (CODEMAX-CHAR_MAX)
107
static void print(struct match *, char *, states, int, FILE *);
110
static void at(struct match *, char *, char *, char *, sopno, sopno);
113
static char *pchar(int);
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®_TRACE) (void)printf("=%s\n", (str)); }
122
#define SP(t, s, c) /* nothing */
123
#define AT(t, p1, p2, s1, s2) /* nothing */
124
#define NOTE(s) /* nothing */
128
- matcher - the actual matching engine
130
static int /* 0 success, REG_NOMATCH failure */
131
matcher(struct re_guts *g, const char *string, size_t nmatch,
132
llvm_regmatch_t pmatch[],
138
struct match *m = &mv;
140
const sopno gf = g->firststate+1; /* +1 for OEND */
141
const sopno gl = g->laststate;
145
/* simplify the situation where possible */
146
if (g->cflags®_NOSUB)
148
if (eflags®_STARTEND) {
149
start = string + pmatch[0].rm_so;
150
stop = string + pmatch[0].rm_eo;
153
stop = start + strlen(start);
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)
164
if (dp == stop) /* we didn't find g->must */
168
/* match struct setup */
183
/* this loop does only one repetition except for backrefs */
185
endp = fast(m, start, stop, gf, gl);
186
if (endp == NULL) { /* a miss */
188
free((void*)m->lastpos);
192
if (nmatch == 0 && !g->backrefs)
193
break; /* no further info needed */
196
assert(m->coldp != NULL);
198
NOTE("finding start");
199
endp = slow(m, m->coldp, stop, gf, gl);
202
assert(m->coldp < m->endp);
205
if (nmatch == 1 && !g->backrefs)
206
break; /* no further info needed */
208
/* oh my, he wants 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) {
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®_BACKR)) {
220
dp = dissect(m, m->coldp, endp, gf, gl);
222
if (g->nplus > 0 && m->lastpos == NULL)
223
m->lastpos = (const char **)malloc((g->nplus+1) *
225
if (g->nplus > 0 && m->lastpos == NULL) {
230
NOTE("backref dissect");
231
dp = backref(m, m->coldp, endp, gf, gl, (sopno)0, 0);
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);
240
if (dp != NULL || endp <= m->coldp)
243
endp = slow(m, m->coldp, endp-1, gf, gl);
246
/* try it on a shorter possibility */
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);
253
NOTE("backoff dissect");
254
dp = backref(m, m->coldp, endp, gf, gl, (sopno)0, 0);
256
assert(dp == NULL || dp == endp);
257
if (dp != NULL) /* found a shorter one */
260
/* despite initial appearances, there is no match here */
262
if (m->coldp == stop)
264
start = m->coldp + 1; /* recycle starting later */
267
/* fill in the details if requested */
269
pmatch[0].rm_so = m->coldp - m->offp;
270
pmatch[0].rm_eo = endp - m->offp;
273
assert(m->pmatch != NULL);
274
for (i = 1; i < nmatch; i++)
276
pmatch[i] = m->pmatch[i];
278
pmatch[i].rm_so = -1;
279
pmatch[i].rm_eo = -1;
283
if (m->pmatch != NULL)
284
free((char *)m->pmatch);
285
if (m->lastpos != NULL)
286
free((char *)m->lastpos);
292
- dissect - figure out what matched what, no back references
294
static const char * /* == stop (success) always */
295
dissect(struct match *m, const char *start, const char *stop, sopno startst,
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 */
311
AT("diss", start, stop, startst, stopst);
313
for (ss = startst; ss < stopst; ss = es) {
314
/* identify end of subRE */
316
switch (OP(m->g->strip[es])) {
319
es += OPND(m->g->strip[es]);
322
while (OP(m->g->strip[es]) != O_CH)
323
es += OPND(m->g->strip[es]);
328
/* figure out what it matched */
329
switch (OP(m->g->strip[ss])) {
349
/* cases where length of match is hard to find */
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);
360
/* no -- try a shorter match for this one */
362
assert(stp >= sp); /* it did work */
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 */
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);
385
/* no -- try a shorter match for this one */
387
assert(stp >= sp); /* it did work */
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 */
401
/* last successful match */
405
assert(sep == rest); /* must exhaust substring */
406
assert(slow(m, ssp, sep, ssub, esub) == rest);
408
const char *dp = dissect(m, ssp, sep, ssub, esub);
409
(void)dp; /* avoid warning if assertions off */
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);
424
/* no -- try a shorter match for this one */
426
assert(stp >= sp); /* it did work */
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);
437
assert(OP(m->g->strip[esub]) == OOR2);
439
esub += OPND(m->g->strip[esub]);
440
if (OP(m->g->strip[esub]) == OOR2)
443
assert(OP(m->g->strip[esub]) == O_CH);
446
const char *dp = dissect(m, sp, rest, ssub, esub);
447
(void)dp; /* avoid warning if assertions off */
460
i = OPND(m->g->strip[ss]);
461
assert(0 < i && i <= m->g->nsub);
462
m->pmatch[i].rm_so = sp - m->offp;
465
i = OPND(m->g->strip[ss]);
466
assert(0 < i && i <= m->g->nsub);
467
m->pmatch[i].rm_eo = sp - m->offp;
480
- backref - figure out what matched what, figuring in back references
482
static const char * /* == stop (success) or NULL (failure) */
483
backref(struct match *m, const char *start, const char *stop, sopno startst,
484
sopno stopst, sopno lev, int rec) /* PLUS nesting level */
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 */
496
llvm_regoff_t offsave;
499
AT("back", start, stop, startst, stopst);
502
/* get as far as we can with easy stuff */
504
for (ss = startst; !hard && ss < stopst; ss++)
505
switch (OP(s = m->g->strip[ss])) {
507
if (sp == stop || *sp++ != (char)OPND(s))
516
cs = &m->g->sets[OPND(s)];
517
if (sp == stop || !CHIN(cs, *sp++))
521
if ( (sp == m->beginp && !(m->eflags®_NOTBOL)) ||
522
(sp < m->endp && *(sp-1) == '\n' &&
523
(m->g->cflags®_NEWLINE)) )
529
if ( (sp == m->endp && !(m->eflags®_NOTEOL)) ||
530
(sp < m->endp && *sp == '\n' &&
531
(m->g->cflags®_NEWLINE)) )
537
if (( (sp == m->beginp && !(m->eflags®_NOTBOL)) ||
538
(sp < m->endp && *(sp-1) == '\n' &&
539
(m->g->cflags®_NEWLINE)) ||
541
!ISWORD(*(sp-1))) ) &&
542
(sp < m->endp && ISWORD(*sp)) )
548
if (( (sp == m->endp && !(m->eflags®_NOTEOL)) ||
549
(sp < m->endp && *sp == '\n' &&
550
(m->g->cflags®_NEWLINE)) ||
551
(sp < m->endp && !ISWORD(*sp)) ) &&
552
(sp > m->beginp && ISWORD(*(sp-1))) )
559
case OOR1: /* matches null but needs to skip */
563
assert(OP(s) == OOR2);
565
} while (OP(s = m->g->strip[ss]) != O_CH);
566
/* note that the ss++ gets us past the O_CH */
568
default: /* have to make a choice */
572
if (!hard) { /* that was it! */
577
ss--; /* adjust for the for's final increment */
580
AT("hard", sp, stop, ss, stopst);
583
case OBACK_: /* the vilest depths */
585
assert(0 < i && i <= m->g->nsub);
586
if (m->pmatch[i].rm_eo == -1)
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)
592
assert(stop - m->beginp >= 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)
598
while (m->g->strip[ss] != SOP(O_BACK, i))
600
return(backref(m, sp+len, stop, ss+1, stopst, lev, rec));
602
case OQUEST_: /* to null or not */
603
dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
605
return(dp); /* not */
606
return(backref(m, sp, stop, ss+OPND(s)+1, stopst, lev, rec));
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));
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);
621
return(backref(m, sp, stop, ss+1, stopst, lev-1, rec));
625
case OCH_: /* find the right one, if any */
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);
633
/* that one missed, try next one */
634
if (OP(m->g->strip[esub]) == O_CH)
635
return(NULL); /* there is none */
637
assert(OP(m->g->strip[esub]) == OOR2);
639
esub += OPND(m->g->strip[esub]);
640
if (OP(m->g->strip[esub]) == OOR2)
643
assert(OP(m->g->strip[esub]) == O_CH);
646
case OLPAREN: /* must undo assignment if rest fails */
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);
654
m->pmatch[i].rm_so = offsave;
657
case ORPAREN: /* must undo assignment if rest fails */
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);
665
m->pmatch[i].rm_eo = offsave;
680
- fast - step through the string at top speed
682
static const char * /* where tentative match ended, or NULL */
683
fast(struct match *m, const char *start, const char *stop, sopno startst,
687
states fresh = m->fresh;
689
const char *p = start;
690
int c = (start == m->beginp) ? OUT : *(start-1);
691
int lastc; /* previous c */
694
const char *coldp; /* last p after which no match was underway */
698
st = step(m->g, startst, stopst, st, NOTHING, st);
705
c = (p == m->endp) ? OUT : *p;
709
/* is there an EOL and/or BOL between lastc and c? */
712
if ( (lastc == '\n' && m->g->cflags®_NEWLINE) ||
713
(lastc == OUT && !(m->eflags®_NOTBOL)) ) {
717
if ( (c == '\n' && m->g->cflags®_NEWLINE) ||
718
(c == OUT && !(m->eflags®_NOTEOL)) ) {
719
flagch = (flagch == BOL) ? BOLEOL : EOL;
724
st = step(m->g, startst, stopst, st, flagch, st);
728
/* how about a word boundary? */
729
if ( (flagch == BOL || (lastc != OUT && !ISWORD(lastc))) &&
730
(c != OUT && ISWORD(c)) ) {
733
if ( (lastc != OUT && ISWORD(lastc)) &&
734
(flagch == EOL || (c != OUT && !ISWORD(c))) ) {
737
if (flagch == BOW || flagch == EOW) {
738
st = step(m->g, startst, stopst, st, flagch, st);
743
if (ISSET(st, stopst) || p == stop)
744
break; /* NOTE BREAK OUT */
746
/* no, we must deal with this character */
750
st = step(m->g, startst, stopst, tmp, c, st);
752
assert(EQ(step(m->g, startst, stopst, st, NOTHING, st), st));
756
assert(coldp != NULL);
758
if (ISSET(st, stopst))
765
- slow - step through the string more deliberately
767
static const char * /* where it ended */
768
slow(struct match *m, const char *start, const char *stop, sopno startst,
772
states empty = m->empty;
774
const char *p = start;
775
int c = (start == m->beginp) ? OUT : *(start-1);
776
int lastc; /* previous c */
779
const char *matchp; /* last p at which a match ended */
781
AT("slow", start, stop, startst, stopst);
784
SP("sstart", st, *p);
785
st = step(m->g, startst, stopst, st, NOTHING, st);
790
c = (p == m->endp) ? OUT : *p;
792
/* is there an EOL and/or BOL between lastc and c? */
795
if ( (lastc == '\n' && m->g->cflags®_NEWLINE) ||
796
(lastc == OUT && !(m->eflags®_NOTBOL)) ) {
800
if ( (c == '\n' && m->g->cflags®_NEWLINE) ||
801
(c == OUT && !(m->eflags®_NOTEOL)) ) {
802
flagch = (flagch == BOL) ? BOLEOL : EOL;
807
st = step(m->g, startst, stopst, st, flagch, st);
808
SP("sboleol", st, c);
811
/* how about a word boundary? */
812
if ( (flagch == BOL || (lastc != OUT && !ISWORD(lastc))) &&
813
(c != OUT && ISWORD(c)) ) {
816
if ( (lastc != OUT && ISWORD(lastc)) &&
817
(flagch == EOL || (c != OUT && !ISWORD(c))) ) {
820
if (flagch == BOW || flagch == EOW) {
821
st = step(m->g, startst, stopst, st, flagch, st);
822
SP("sboweow", st, c);
826
if (ISSET(st, stopst))
828
if (EQ(st, empty) || p == stop)
829
break; /* NOTE BREAK OUT */
831
/* no, we must deal with this character */
835
st = step(m->g, startst, stopst, tmp, c, st);
837
assert(EQ(step(m->g, startst, stopst, st, NOTHING, st), st));
846
- step - map set of states reachable before char to set reachable after
849
step(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 */
859
onestate here; /* note, macros know this name */
863
for (pc = start, INIT(here, pc); pc != stop; pc++, INC(here)) {
867
assert(pc == stop-1);
870
/* only characters can match */
871
assert(!NONCHAR(ch) || ch != (char)OPND(s));
872
if (ch == (char)OPND(s))
876
if (ch == BOL || ch == BOLEOL)
880
if (ch == EOL || ch == BOLEOL)
896
cs = &g->sets[OPND(s)];
897
if (!NONCHAR(ch) && CHIN(cs, ch))
900
case OBACK_: /* ignored here */
904
case OPLUS_: /* forward, this is just an empty */
907
case O_PLUS: /* both forward and back */
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 */
917
case OQUEST_: /* two branches, both forward */
919
FWD(aft, aft, OPND(s));
921
case O_QUEST: /* just an empty */
924
case OLPAREN: /* not significant here */
928
case OCH_: /* mark the first two branches */
930
assert(OP(g->strip[pc+OPND(s)]) == OOR2);
931
FWD(aft, aft, OPND(s));
933
case OOR1: /* done a branch, find the O_CH */
934
if (ISSTATEIN(aft, here)) {
936
OP(s = g->strip[pc+look]) != O_CH;
938
assert(OP(s) == OOR2);
942
case OOR2: /* propagate OCH_'s marking */
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));
949
case O_CH: /* just empty */
952
default: /* ooooops... */
963
- print - print a set of states
966
print(struct match *m, char *caption, states st, int ch, FILE *d)
968
struct re_guts *g = m->g;
972
if (!(m->eflags®_TRACE))
975
(void)fprintf(d, "%s", caption);
977
(void)fprintf(d, " %s", pchar(ch));
978
for (i = 0; i < g->nstates; i++)
980
(void)fprintf(d, "%s%d", (first) ? "\t" : ", ", i);
983
(void)fprintf(d, "\n");
987
- at - print current situation
990
at(struct match *m, char *title, char *start, char *stop, sopno startst,
993
if (!(m->eflags®_TRACE))
996
(void)printf("%s %s-", title, pchar(*start));
997
(void)printf("%s ", pchar(*stop));
998
(void)printf("%ld-%ld\n", (long)startst, (long)stopst);
1002
#define PCHARDONE /* never again */
1004
- pchar - make a character printable
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.
1011
static char * /* -> representation */
1014
static char pbuf[10];
1016
if (isprint(ch) || ch == ' ')
1017
(void)snprintf(pbuf, sizeof pbuf, "%c", ch);
1019
(void)snprintf(pbuf, sizeof pbuf, "\\%o", ch);