1 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
2 * confused with the original package (see point 3 below). Thanks, Henry!
5 /* Additional note: this code is very heavily munged from Henry's version
6 * in places. In some spots I've traded clarity for efficiency, so don't
7 * blame Henry for some of the lack of readability.
10 /* $RCSfile: regexec.c,v $$Revision: 4.0.1.3 $$Date: 91/11/05 18:23:55 $
13 * Revision 4.0.1.3 91/11/05 18:23:55 lwall
14 * patch11: prepared for ctype implementations that don't define isascii()
15 * patch11: initial .* in pattern had dependency on value of $*
17 * Revision 4.0.1.2 91/06/07 11:50:33 lwall
18 * patch4: new copyright notice
19 * patch4: // wouldn't use previous pattern if it started with a null character
21 * Revision 4.0.1.1 91/04/12 09:07:39 lwall
22 * patch1: regexec only allocated space for 9 subexpresssions
24 * Revision 4.0 91/03/20 01:39:16 lwall
30 * regcomp and regexec -- regsub and regerror are not used in perl
32 * Copyright (c) 1986 by University of Toronto.
33 * Written by Henry Spencer. Not derived from licensed software.
35 * Permission is granted to anyone to use this software for any
36 * purpose on any computer system, and to redistribute it freely,
37 * subject to the following restrictions:
39 * 1. The author is not responsible for the consequences of use of
40 * this software, no matter how awful, even if they arise
43 * 2. The origin of this software must not be misrepresented, either
44 * by explicit claim or by omission.
46 * 3. Altered versions must be plainly marked as such, and must not
47 * be misrepresented as being the original software.
49 **** Alterations to Henry's code are...
51 **** Copyright (c) 1991, Larry Wall
53 **** You may distribute under the terms of either the GNU General Public
54 **** License or the Artistic License, as specified in the README file.
56 * Beware that some of this code is subtly aware of the way operator
57 * precedence is structured in regular expressions. Serious changes in
58 * regular-expression syntax might require a total rethink.
77 * Global work variables for regexec().
79 static char *regprecomp;
80 static char *reginput; /* String-input pointer. */
81 static char regprev; /* char before regbol, \n if none */
82 static char *regbol; /* Beginning of input, for ^ check. */
83 static char *regeol; /* End of input, for $ check. */
84 static char **regstartp; /* Pointer to startp array. */
85 static char **regendp; /* Ditto for endp. */
86 static char *reglastparen; /* Similarly for lastparen. */
89 static int regmyp_size = 0;
90 static char **regmystartp = Null(char**);
91 static char **regmyendp = Null(char**);
97 STATIC int regmatch();
98 STATIC int regrepeat();
100 extern int multiline;
103 - regexec - match a regexp against a string
106 regexec(prog, stringarg, strend, strbeg, minend, screamer, safebase)
107 register regexp *prog;
109 register char *strend; /* pointer to null at end of string */
110 char *strbeg; /* real beginning of string */
111 int minend; /* end of match must be at least minend after stringarg */
113 int safebase; /* no need to remember string in subbase */
118 register char *string = stringarg;
120 int minlen = 0; /* must match at least this many chars */
121 int dontbother = 0; /* how many characters not to try at end */
124 if (prog == NULL || string == NULL) {
125 fatal("NULL regexp parameter");
129 if (string == strbeg) /* is ^ valid at stringarg? */
132 regprev = stringarg[-1];
133 if (!multiline && regprev == '\n')
134 regprev = '\0'; /* force ^ to NOT match */
136 regprecomp = prog->precomp;
137 /* Check validity of program. */
138 if (UCHARAT(prog->program) != MAGIC) {
139 FAIL("corrupted regexp program");
142 if (prog->do_folding) {
145 New(1101,c,i+1,char);
146 (void)bcopy(string, c, i+1);
149 for (s = string; s < strend; s++)
154 /* If there is a "must appear" string, look for it. */
156 if (prog->regmust != Nullstr &&
157 (!(prog->reganch & ROPT_ANCH)
158 || (multiline && prog->regback >= 0)) ) {
159 if (stringarg == strbeg && screamer) {
160 if (screamfirst[prog->regmust->str_rare] >= 0)
161 s = screaminstr(screamer,prog->regmust);
167 s = fbminstr((unsigned char*)s, (unsigned char*)strend,
171 ++prog->regmust->str_u.str_useful; /* hooray */
172 goto phooey; /* not present */
174 else if (prog->regback >= 0) {
178 minlen = prog->regback + prog->regmust->str_cur;
180 else if (--prog->regmust->str_u.str_useful < 0) { /* boo */
181 str_free(prog->regmust);
182 prog->regmust = Nullstr; /* disable regmust */
187 minlen = prog->regmust->str_cur;
191 /* Mark beginning of line for ^ . */
194 /* Mark end of line for $ (and such) */
197 /* see how far we have to get to not match where we matched before */
198 regtill = string+minend;
200 /* Allocate our backreference arrays */
201 if ( regmyp_size < prog->nparens + 1 ) {
202 /* Allocate or enlarge the arrays */
203 regmyp_size = prog->nparens + 1;
204 if ( regmyp_size < 10 ) regmyp_size = 10; /* minimum */
206 /* reallocate larger */
207 Renew(regmystartp,regmyp_size,char*);
208 Renew(regmyendp, regmyp_size,char*);
211 /* Initial allocation */
212 New(1102,regmystartp,regmyp_size,char*);
213 New(1102,regmyendp, regmyp_size,char*);
218 /* Simplest case: anchored match need be tried only once. */
219 /* [unless multiline is set] */
220 if (prog->reganch & ROPT_ANCH) {
221 if (regtry(prog, string))
223 else if (multiline || (prog->reganch & ROPT_IMPLICIT)) {
225 dontbother = minlen - 1;
226 strend -= dontbother;
227 /* for multiline we only have to try after newlines */
232 if (s < strend && regtry(prog, s))
240 /* Messy cases: unanchored match. */
241 if (prog->regstart) {
242 if (prog->reganch & ROPT_SKIP) { /* we have /x+whatever/ */
243 /* it must be a one character string */
244 i = prog->regstart->str_ptr[0];
250 while (s < strend && *s == i)
256 else if (prog->regstart->str_pok == 3) {
257 /* We know what string it must start with. */
259 while ((s = fbminstr((unsigned char*)s,
260 (unsigned char*)strend, prog->regstart)) != NULL)
271 c = prog->regstart->str_ptr;
272 while ((s = ninstr(s, strend,
273 c, c + prog->regstart->str_cur )) != NULL) {
282 if (c = prog->regstclass) {
283 int doevery = (prog->reganch & ROPT_SKIP) == 0;
286 dontbother = minlen - 1;
287 strend -= dontbother; /* don't bother with what can't match */
289 /* We know what class it must start with. */
295 if (!(c[i >> 3] & (1 << (i&7)))) {
296 if (tmp && regtry(prog, s))
308 dontbother++,strend--;
314 tmp = isALNUM(regprev); /* assume not alphanumeric */
317 if (tmp != isALNUM(i)) {
324 if ((minlen || tmp) && regtry(prog,s))
329 dontbother++,strend--;
335 tmp = isALNUM(regprev); /* assume not alphanumeric */
338 if (tmp != isALNUM(i))
340 else if (regtry(prog, s))
344 if ((minlen || !tmp) && regtry(prog,s))
351 if (tmp && regtry(prog, s))
365 if (tmp && regtry(prog, s))
378 if (tmp && regtry(prog, s))
391 if (tmp && regtry(prog, s))
404 if (tmp && regtry(prog, s))
417 if (tmp && regtry(prog, s))
431 dontbother = minlen - 1;
432 strend -= dontbother;
433 /* We don't know much -- general case. */
437 } while (s++ < strend);
444 if ((!safebase && (prog->nparens || sawampersand)) || prog->do_folding){
445 strend += dontbother; /* uncheat */
446 if (safebase) /* no need for $digit later */
448 else if (strbeg != prog->subbase) {
449 i = strend - string + (stringarg - strbeg);
450 s = nsavestr(strbeg,i); /* so $digit will work later */
452 Safefree(prog->subbase);
453 prog->subbeg = prog->subbase = s;
458 s += (stringarg - strbeg);
459 for (i = 0; i <= prog->nparens; i++) {
461 prog->startp[i] = s + (prog->startp[i] - string);
462 prog->endp[i] = s + (prog->endp[i] - string);
465 if (prog->do_folding)
471 if (prog->do_folding)
477 - regtry - try match at specific point
479 static int /* 0 failure, 1 success */
489 regstartp = prog->startp;
490 regendp = prog->endp;
491 reglastparen = &prog->lastparen;
497 for (i = prog->nparens; i >= 0; i--) {
502 if (regmatch(prog->program + 1) && reginput >= regtill) {
503 prog->startp[0] = string;
504 prog->endp[0] = reginput;
511 - regmatch - main matching routine
513 * Conceptually the strategy is simple: check to see whether the current
514 * node matches, call self recursively to see whether the rest matches,
515 * and then act accordingly. In practice we make some effort to avoid
516 * recursion, in particular by going through "ordinary" nodes (that don't
517 * need to know whether the rest of the match failed) by a loop instead of
520 /* [lwall] I've hoisted the register declarations to the outer block in order to
521 * maybe save a little bit of pushing and popping on the stack. It also takes
522 * advantage of machines that use a register save mask on subroutine entry.
524 static int /* 0 failure, 1 success */
528 register char *scan; /* Current node. */
529 char *next; /* Next node. */
530 register int nextchar;
531 register int n; /* no or next */
532 register int ln; /* len or last */
533 register char *s; /* operand or save */
534 register char *locinput = reginput;
536 nextchar = *locinput;
539 if (scan != NULL && regnarrate)
540 fprintf(stderr, "%s(\n", regprop(scan));
542 while (scan != NULL) {
545 fprintf(stderr, "%s...\n", regprop(scan));
549 next = scan + NEXT(scan);
553 next = regnext(scan);
558 if (locinput == regbol ? regprev == '\n' :
559 ((nextchar || locinput < regeol) &&
560 locinput[-1] == '\n') )
562 /* regtill = regbol; */
567 if ((nextchar || locinput < regeol) && nextchar != '\n')
569 if (!multiline && regeol - locinput > 1)
571 /* regtill = regbol; */
574 if ((nextchar == '\0' && locinput >= regeol) ||
577 nextchar = *++locinput;
582 /* Inline the first character, for speed. */
585 if (regeol - locinput < ln)
587 if (ln > 1 && bcmp(s, locinput, ln) != 0)
590 nextchar = *locinput;
595 nextchar = UCHARAT(locinput);
596 if (s[nextchar >> 3] & (1 << (nextchar&7)))
598 if (!nextchar && locinput >= regeol)
600 nextchar = *++locinput;
605 if (!isALNUM(nextchar))
607 nextchar = *++locinput;
610 if (!nextchar && locinput >= regeol)
612 if (isALNUM(nextchar))
614 nextchar = *++locinput;
618 if (locinput == regbol) /* was last char in word? */
619 ln = isALNUM(regprev);
621 ln = isALNUM(locinput[-1]);
622 n = isALNUM(nextchar); /* is next char in word? */
623 if ((ln == n) == (OP(scan) == BOUND))
627 if (!nextchar && locinput >= regeol)
629 if (!isSPACE(nextchar))
631 nextchar = *++locinput;
636 if (isSPACE(nextchar))
638 nextchar = *++locinput;
641 if (!isDIGIT(nextchar))
643 nextchar = *++locinput;
646 if (!nextchar && locinput >= regeol)
648 if (isDIGIT(nextchar))
650 nextchar = *++locinput;
653 n = ARG1(scan); /* which paren pair */
659 if (s == regmyendp[n])
661 /* Inline the first character, for speed. */
664 ln = regmyendp[n] - s;
665 if (locinput + ln > regeol)
667 if (ln > 1 && bcmp(s, locinput, ln) != 0)
670 nextchar = *locinput;
678 n = ARG1(scan); /* which paren pair */
681 regmystartp[n] = locinput; /* for REF */
682 if (regmatch(next)) {
684 * Don't set startp if some later
685 * invocation of the same parentheses
688 if (regstartp[n] == NULL)
689 regstartp[n] = locinput;
695 n = ARG1(scan); /* which paren pair */
698 regmyendp[n] = locinput; /* for REF */
699 if (regmatch(next)) {
701 * Don't set endp if some later
702 * invocation of the same parentheses
705 if (regendp[n] == NULL) {
706 regendp[n] = locinput;
707 if (n > *reglastparen)
716 if (OP(next) != BRANCH) /* No choice. */
717 next = NEXTOPER(scan); /* Avoid recursion. */
721 if (regmatch(NEXTOPER(scan)))
730 scan = regnext(scan);
732 } while (scan != NULL && OP(scan) == BRANCH);
739 ln = ARG1(scan); /* min to match */
740 n = ARG2(scan); /* max to match */
741 scan = NEXTOPER(scan) + 4;
746 scan = NEXTOPER(scan);
750 * Lookahead to avoid useless match attempts
751 * when we know what character comes next.
755 scan = NEXTOPER(scan);
757 if (OP(next) == EXACTLY)
758 nextchar = *(OPERAND(next)+1);
762 n = regrepeat(scan, n);
763 if (!multiline && OP(next) == EOL && ln < n)
764 ln = n; /* why back off? */
766 /* If it could work, try it. */
767 if (nextchar == -1000 || *reginput == nextchar)
770 /* Couldn't or didn't -- back up. */
772 reginput = locinput + n;
776 reginput = locinput; /* put where regtry can find it */
777 return(1); /* Success! */
779 printf("%x %d\n",scan,scan[1]);
780 FAIL("regexp memory corruption");
787 * We get here only if there's trouble -- normally "case END" is
788 * the terminating point.
790 FAIL("corrupted regexp pointers");
798 - regrepeat - repeatedly match something simple, report how many
801 * [This routine now assumes that it will only match on things of length 1.
802 * That was true before, but now we assume scan - reginput is the count,
803 * rather than incrementing count on every character.]
813 register char *loceol = regeol;
816 if (max && max < loceol - scan)
821 while (scan < loceol && *scan != '\n')
824 case EXACTLY: /* length of string is 1 */
826 while (scan < loceol && *opnd == *scan)
831 while (scan < loceol && !(opnd[c >> 3] & (1 << (c & 7)))) {
837 while (scan < loceol && isALNUM(*scan))
841 while (scan < loceol && !isALNUM(*scan))
845 while (scan < loceol && isSPACE(*scan))
849 while (scan < loceol && !isSPACE(*scan))
853 while (scan < loceol && isDIGIT(*scan))
857 while (scan < loceol && !isDIGIT(*scan))
860 default: /* Oh dear. Called inappropriately. */
861 FAIL("internal regexp foulup");
872 - regnext - dig the "next" pointer out of a node
874 * [Note, when REGALIGN is defined there are two places in regmatch()
875 * that bypass this code for speed.]