5 * "One Ring to rule them all, One Ring to find them..."
8 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
9 * confused with the original package (see point 3 below). Thanks, Henry!
12 /* Additional note: this code is very heavily munged from Henry's version
13 * in places. In some spots I've traded clarity for efficiency, so don't
14 * blame Henry for some of the lack of readability.
17 /* The names of the functions have been changed from regcomp and
18 * regexec to pregcomp and pregexec in order to avoid conflicts
19 * with the POSIX routines of the same names.
22 #ifdef PERL_EXT_RE_BUILD
23 /* need to replace pregcomp et al, so enable that */
24 # ifndef PERL_IN_XSUB_RE
25 # define PERL_IN_XSUB_RE
27 /* need access to debugger hooks */
28 # if defined(PERL_EXT_RE_DEBUG) && !defined(DEBUGGING)
33 #ifdef PERL_IN_XSUB_RE
34 /* We *really* need to overwrite these symbols: */
35 # define Perl_regexec_flags my_regexec
36 # define Perl_regdump my_regdump
37 # define Perl_regprop my_regprop
38 # define Perl_re_intuit_start my_re_intuit_start
39 /* *These* symbols are masked to allow static link. */
40 # define Perl_pregexec my_pregexec
41 # define Perl_reginitcolors my_reginitcolors
43 # define PERL_NO_GET_CONTEXT
48 * pregcomp and pregexec -- regsub and regerror are not used in perl
50 * Copyright (c) 1986 by University of Toronto.
51 * Written by Henry Spencer. Not derived from licensed software.
53 * Permission is granted to anyone to use this software for any
54 * purpose on any computer system, and to redistribute it freely,
55 * subject to the following restrictions:
57 * 1. The author is not responsible for the consequences of use of
58 * this software, no matter how awful, even if they arise
61 * 2. The origin of this software must not be misrepresented, either
62 * by explicit claim or by omission.
64 * 3. Altered versions must be plainly marked as such, and must not
65 * be misrepresented as being the original software.
67 **** Alterations to Henry's code are...
69 **** Copyright (c) 1991-1999, Larry Wall
71 **** You may distribute under the terms of either the GNU General Public
72 **** License or the Artistic License, as specified in the README file.
74 * Beware that some of this code is subtly aware of the way operator
75 * precedence is structured in regular expressions. Serious changes in
76 * regular-expression syntax might require a total rethink.
79 #define PERL_IN_REGEXEC_C
82 #ifdef PERL_IN_XSUB_RE
83 # if defined(PERL_CAPI) || defined(PERL_OBJECT)
90 #define RF_tainted 1 /* tainted information used? */
91 #define RF_warned 2 /* warned about big count? */
92 #define RF_evaled 4 /* Did an EVAL with setting? */
93 #define RF_utf8 8 /* String contains multibyte chars? */
95 #define UTF (PL_reg_flags & RF_utf8)
97 #define RS_init 1 /* eval environment created */
98 #define RS_set 2 /* replsv value is set */
101 #define STATIC static
108 #define REGINCLASS(p,c) (ANYOF_FLAGS(p) ? reginclass(p,c) : ANYOF_BITMAP_TEST(p,c))
109 #define REGINCLASSUTF8(f,p) (ARG1(f) ? reginclassutf8(f,p) : swash_fetch((SV*)PL_regdata->data[ARG2(f)],p))
111 #define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
112 #define CHR_DIST(a,b) (UTF ? utf8_distance(a,b) : a - b)
114 #define reghop_c(pos,off) ((char*)reghop((U8*)pos, off))
115 #define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off))
116 #define HOP(pos,off) (UTF ? reghop((U8*)pos, off) : (U8*)(pos + off))
117 #define HOPMAYBE(pos,off) (UTF ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off))
118 #define HOPc(pos,off) ((char*)HOP(pos,off))
119 #define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off))
121 static void restore_pos(pTHXo_ void *arg);
125 S_regcppush(pTHX_ I32 parenfloor)
128 int retval = PL_savestack_ix;
129 int i = (PL_regsize - parenfloor) * 4;
133 for (p = PL_regsize; p > parenfloor; p--) {
134 SSPUSHINT(PL_regendp[p]);
135 SSPUSHINT(PL_regstartp[p]);
136 SSPUSHPTR(PL_reg_start_tmp[p]);
139 SSPUSHINT(PL_regsize);
140 SSPUSHINT(*PL_reglastparen);
141 SSPUSHPTR(PL_reginput);
143 SSPUSHINT(SAVEt_REGCONTEXT);
147 /* These are needed since we do not localize EVAL nodes: */
148 # define REGCP_SET DEBUG_r(PerlIO_printf(Perl_debug_log, \
149 " Setting an EVAL scope, savestack=%"IVdf"\n", \
150 (IV)PL_savestack_ix)); lastcp = PL_savestack_ix
152 # define REGCP_UNWIND DEBUG_r(lastcp != PL_savestack_ix ? \
153 PerlIO_printf(Perl_debug_log, \
154 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
155 (IV)lastcp, (IV)PL_savestack_ix) : 0); regcpblow(lastcp)
165 assert(i == SAVEt_REGCONTEXT);
167 input = (char *) SSPOPPTR;
168 *PL_reglastparen = SSPOPINT;
169 PL_regsize = SSPOPINT;
170 for (i -= 3; i > 0; i -= 4) {
171 paren = (U32)SSPOPINT;
172 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
173 PL_regstartp[paren] = SSPOPINT;
175 if (paren <= *PL_reglastparen)
176 PL_regendp[paren] = tmps;
178 PerlIO_printf(Perl_debug_log,
179 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
180 (UV)paren, (IV)PL_regstartp[paren],
181 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
182 (IV)PL_regendp[paren],
183 (paren > *PL_reglastparen ? "(no)" : ""));
187 if (*PL_reglastparen + 1 <= PL_regnpar) {
188 PerlIO_printf(Perl_debug_log,
189 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
190 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
193 for (paren = *PL_reglastparen + 1; paren <= PL_regnpar; paren++) {
194 if (paren > PL_regsize)
195 PL_regstartp[paren] = -1;
196 PL_regendp[paren] = -1;
202 S_regcp_set_to(pTHX_ I32 ss)
205 I32 tmp = PL_savestack_ix;
207 PL_savestack_ix = ss;
209 PL_savestack_ix = tmp;
213 typedef struct re_cc_state
217 struct re_cc_state *prev;
222 #define regcpblow(cp) LEAVE_SCOPE(cp)
225 * pregexec and friends
229 - pregexec - match a regexp against a string
232 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
233 char *strbeg, I32 minend, SV *screamer, U32 nosave)
234 /* strend: pointer to null at end of string */
235 /* strbeg: real beginning of string */
236 /* minend: end of match must be >=minend after stringarg. */
237 /* nosave: For optimizations. */
240 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
241 nosave ? 0 : REXEC_COPY_STR);
245 S_cache_re(pTHX_ regexp *prog)
248 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
250 PL_regprogram = prog->program;
252 PL_regnpar = prog->nparens;
253 PL_regdata = prog->data;
258 * Need to implement the following flags for reg_anch:
260 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
262 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
263 * INTUIT_AUTORITATIVE_ML
264 * INTUIT_ONCE_NOML - Intuit can match in one location only.
267 * Another flag for this function: SECOND_TIME (so that float substrs
268 * with giant delta may be not rechecked).
271 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
273 /* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
274 Otherwise, only SvCUR(sv) is used to get strbeg. */
276 /* XXXX We assume that strpos is strbeg unless sv. */
278 /* XXXX Some places assume that there is a fixed substring.
279 An update may be needed if optimizer marks as "INTUITable"
280 RExen without fixed substrings. Similarly, it is assumed that
281 lengths of all the strings are no more than minlen, thus they
282 cannot come from lookahead.
283 (Or minlen should take into account lookahead.) */
285 /* A failure to find a constant substring means that there is no need to make
286 an expensive call to REx engine, thus we celebrate a failure. Similarly,
287 finding a substring too deep into the string means that less calls to
288 regtry() should be needed.
290 REx compiler's optimizer found 4 possible hints:
291 a) Anchored substring;
293 c) Whether we are anchored (beginning-of-line or \G);
294 d) First node (of those at offset 0) which may distingush positions;
295 We use a)b)d) and multiline-part of c), and try to find a position in the
296 string which does not contradict any of them.
299 /* Most of decisions we do here should have been done at compile time.
300 The nodes of the REx which we used for the search should have been
301 deleted from the finite automaton. */
304 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
305 char *strend, U32 flags, re_scream_pos_data *data)
307 register I32 start_shift;
308 /* Should be nonnegative! */
309 register I32 end_shift;
315 register char *other_last = Nullch; /* other substr checked before this */
316 char *check_at; /* check substr found at this pos */
318 char *i_strpos = strpos;
321 DEBUG_r( if (!PL_colorset) reginitcolors() );
322 DEBUG_r(PerlIO_printf(Perl_debug_log,
323 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
324 PL_colors[4],PL_colors[5],PL_colors[0],
327 (strlen(prog->precomp) > 60 ? "..." : ""),
329 (int)(strend - strpos > 60 ? 60 : strend - strpos),
330 strpos, PL_colors[1],
331 (strend - strpos > 60 ? "..." : ""))
334 if (prog->minlen > strend - strpos) {
335 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too short...\n"));
338 check = prog->check_substr;
339 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
340 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
341 || ( (prog->reganch & ROPT_ANCH_BOL)
342 && !PL_multiline ) ); /* Check after \n? */
344 if ((prog->check_offset_min == prog->check_offset_max) && !ml_anch) {
345 /* Substring at constant offset from beg-of-str... */
348 if ( !(prog->reganch & ROPT_ANCH_GPOS) /* Checked by the caller */
349 && (sv && (strpos + SvCUR(sv) != strend)) ) {
350 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
353 PL_regeol = strend; /* Used in HOP() */
354 s = HOPc(strpos, prog->check_offset_min);
356 slen = SvCUR(check); /* >= 1 */
358 if ( strend - s > slen || strend - s < slen - 1
359 || (strend - s == slen && strend[-1] != '\n')) {
360 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
363 /* Now should match s[0..slen-2] */
365 if (slen && (*SvPVX(check) != *s
367 && memNE(SvPVX(check), s, slen)))) {
369 DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
373 else if (*SvPVX(check) != *s
374 || ((slen = SvCUR(check)) > 1
375 && memNE(SvPVX(check), s, slen)))
377 goto success_at_start;
379 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
381 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
382 end_shift = prog->minlen - start_shift -
383 CHR_SVLEN(check) + (SvTAIL(check) != 0);
385 I32 end = prog->check_offset_max + CHR_SVLEN(check)
386 - (SvTAIL(check) != 0);
387 I32 eshift = strend - s - end;
389 if (end_shift < eshift)
393 else { /* Can match at random position */
396 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
397 /* Should be nonnegative! */
398 end_shift = prog->minlen - start_shift -
399 CHR_SVLEN(check) + (SvTAIL(check) != 0);
402 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
404 Perl_croak(aTHX_ "panic: end_shift");
408 /* Find a possible match in the region s..strend by looking for
409 the "check" substring in the region corrected by start/end_shift. */
410 if (flags & REXEC_SCREAM) {
411 char *strbeg = SvPVX(sv); /* XXXX Assume PV_force() on SCREAM! */
412 I32 p = -1; /* Internal iterator of scream. */
413 I32 *pp = data ? data->scream_pos : &p;
415 if (PL_screamfirst[BmRARE(check)] >= 0
416 || ( BmRARE(check) == '\n'
417 && (BmPREVIOUS(check) == SvCUR(check) - 1)
419 s = screaminstr(sv, check,
420 start_shift + (s - strbeg), end_shift, pp, 0);
424 *data->scream_olds = s;
427 s = fbm_instr((unsigned char*)s + start_shift,
428 (unsigned char*)strend - end_shift,
429 check, PL_multiline ? FBMrf_MULTILINE : 0);
431 /* Update the count-of-usability, remove useless subpatterns,
434 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
435 (s ? "Found" : "Did not find"),
436 ((check == prog->anchored_substr) ? "anchored" : "floating"),
438 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
440 PL_colors[1], (SvTAIL(check) ? "$" : ""),
441 (s ? " at offset " : "...\n") ) );
448 /* Finish the diagnostic message */
449 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
451 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
452 Start with the other substr.
453 XXXX no SCREAM optimization yet - and a very coarse implementation
454 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
455 *always* match. Probably should be marked during compile...
456 Probably it is right to do no SCREAM here...
459 if (prog->float_substr && prog->anchored_substr) {
460 /* Take into account the "other" substring. */
461 /* XXXX May be hopelessly wrong for UTF... */
464 if (check == prog->float_substr) {
467 char *last = s - start_shift, *last1, *last2;
471 t = s - prog->check_offset_max;
472 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
473 && (!(prog->reganch & ROPT_UTF8)
474 || (PL_bostr = strpos, /* Used in regcopmaybe() */
475 (t = reghopmaybe_c(s, -(prog->check_offset_max)))
480 t += prog->anchored_offset;
481 if (t < other_last) /* These positions already checked */
484 last2 = last1 = strend - prog->minlen;
487 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
488 /* On end-of-str: see comment below. */
489 s = fbm_instr((unsigned char*)t,
490 (unsigned char*)last1 + prog->anchored_offset
491 + SvCUR(prog->anchored_substr)
492 - (SvTAIL(prog->anchored_substr)!=0),
493 prog->anchored_substr, PL_multiline ? FBMrf_MULTILINE : 0);
494 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s anchored substr `%s%.*s%s'%s",
495 (s ? "Found" : "Contradicts"),
497 (int)(SvCUR(prog->anchored_substr)
498 - (SvTAIL(prog->anchored_substr)!=0)),
499 SvPVX(prog->anchored_substr),
500 PL_colors[1], (SvTAIL(prog->anchored_substr) ? "$" : "")));
502 if (last1 >= last2) {
503 DEBUG_r(PerlIO_printf(Perl_debug_log,
504 ", giving up...\n"));
507 DEBUG_r(PerlIO_printf(Perl_debug_log,
508 ", trying floating at offset %ld...\n",
509 (long)(s1 + 1 - i_strpos)));
510 PL_regeol = strend; /* Used in HOP() */
511 other_last = last1 + prog->anchored_offset + 1;
516 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
517 (long)(s - i_strpos)));
518 t = s - prog->anchored_offset;
527 else { /* Take into account the floating substring. */
532 last1 = last = strend - prog->minlen + prog->float_min_offset;
533 if (last - t > prog->float_max_offset)
534 last = t + prog->float_max_offset;
535 s = t + prog->float_min_offset;
538 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
539 /* fbm_instr() takes into account exact value of end-of-str
540 if the check is SvTAIL(ed). Since false positives are OK,
541 and end-of-str is not later than strend we are OK. */
542 s = fbm_instr((unsigned char*)s,
543 (unsigned char*)last + SvCUR(prog->float_substr)
544 - (SvTAIL(prog->float_substr)!=0),
545 prog->float_substr, PL_multiline ? FBMrf_MULTILINE : 0);
546 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
547 (s ? "Found" : "Contradicts"),
549 (int)(SvCUR(prog->float_substr)
550 - (SvTAIL(prog->float_substr)!=0)),
551 SvPVX(prog->float_substr),
552 PL_colors[1], (SvTAIL(prog->float_substr) ? "$" : "")));
555 DEBUG_r(PerlIO_printf(Perl_debug_log,
556 ", giving up...\n"));
559 DEBUG_r(PerlIO_printf(Perl_debug_log,
560 ", trying anchored starting at offset %ld...\n",
561 (long)(s1 + 1 - i_strpos)));
562 other_last = last + 1;
563 PL_regeol = strend; /* Used in HOP() */
568 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
569 (long)(s - i_strpos)));
579 t = s - prog->check_offset_max;
581 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
582 && (!(prog->reganch & ROPT_UTF8)
583 || (PL_bostr = strpos, /* Used in regcopmaybe() */
584 ((t = reghopmaybe_c(s, -(prog->check_offset_max)))
587 /* Fixed substring is found far enough so that the match
588 cannot start at strpos. */
590 if (ml_anch && t[-1] != '\n') {
591 /* Eventually fbm_*() should handle this, but often
592 anchored_offset is not 0, so this check will not be wasted. */
593 /* XXXX In the code below we prefer to look for "^" even in
594 presence of anchored substrings. And we search even
595 beyond the found float position. These pessimizations
596 are historical artefacts only. */
598 while (t < strend - prog->minlen) {
600 if (t < s - prog->check_offset_min) {
601 if (prog->anchored_substr) {
602 /* We definitely contradict the found anchored
603 substr. Due to the above check we do not
604 contradict "check" substr.
605 Thus we can arrive here only if check substr
606 is float. Redo checking for "other"=="fixed".
609 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
610 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
611 goto do_other_anchored;
614 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
615 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
618 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting at offset %ld...\n",
619 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
625 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
626 PL_colors[0],PL_colors[1]));
631 ++BmUSEFUL(prog->check_substr); /* hooray/5 */
635 /* The found string does not prohibit matching at beg-of-str
636 - no optimization of calling REx engine can be performed,
637 unless it was an MBOL and we are not after MBOL. */
639 /* Even in this situation we may use MBOL flag if strpos is offset
640 wrt the start of the string. */
642 && (strpos + SvCUR(sv) != strend) && strpos[-1] != '\n'
643 /* May be due to an implicit anchor of m{.*foo} */
644 && !(prog->reganch & ROPT_IMPLICIT))
649 DEBUG_r( if (ml_anch)
650 PerlIO_printf(Perl_debug_log, "Does not contradict /%s^%s/m...\n",
651 PL_colors[0],PL_colors[1]);
654 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
655 && prog->check_substr /* Could be deleted already */
656 && --BmUSEFUL(prog->check_substr) < 0
657 && prog->check_substr == prog->float_substr)
659 /* If flags & SOMETHING - do not do it many times on the same match */
660 SvREFCNT_dec(prog->check_substr);
661 prog->check_substr = Nullsv; /* disable */
662 prog->float_substr = Nullsv; /* clear */
664 /* XXXX This is a remnant of the old implementation. It
665 looks wasteful, since now INTUIT can use many
667 prog->reganch &= ~RE_USE_INTUIT;
674 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
675 if (prog->regstclass) {
676 /* minlen == 0 is possible if regstclass is \b or \B,
677 and the fixed substr is ''$.
678 Since minlen is already taken into account, s+1 is before strend;
679 accidentally, minlen >= 1 guaranties no false positives at s + 1
680 even for \b or \B. But (minlen? 1 : 0) below assumes that
681 regstclass does not come from lookahead... */
682 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
683 This leaves EXACTF only, which is dealt with in find_byclass(). */
684 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
685 ? STR_LEN(prog->regstclass)
687 char *endpos = (prog->anchored_substr || ml_anch)
688 ? s + (prog->minlen? cl_l : 0)
689 : (prog->float_substr ? check_at - start_shift + cl_l
691 char *startpos = sv ? strend - SvCUR(sv) : s;
694 if (prog->reganch & ROPT_UTF8) {
695 PL_regdata = prog->data; /* Used by REGINCLASS UTF logic */
698 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
703 if (endpos == strend) {
704 DEBUG_r( PerlIO_printf(Perl_debug_log,
705 "Could not match STCLASS...\n") );
708 DEBUG_r( PerlIO_printf(Perl_debug_log,
709 "This position contradicts STCLASS...\n") );
710 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
712 /* Contradict one of substrings */
713 if (prog->anchored_substr) {
714 if (prog->anchored_substr == check) {
715 DEBUG_r( what = "anchored" );
717 PL_regeol = strend; /* Used in HOP() */
719 if (s + start_shift + end_shift > strend) {
720 /* XXXX Should be taken into account earlier? */
721 DEBUG_r( PerlIO_printf(Perl_debug_log,
722 "Could not match STCLASS...\n") );
725 DEBUG_r( PerlIO_printf(Perl_debug_log,
726 "Trying %s substr starting at offset %ld...\n",
727 what, (long)(s + start_shift - i_strpos)) );
730 /* Have both, check_string is floating */
731 if (t + start_shift >= check_at) /* Contradicts floating=check */
732 goto retry_floating_check;
733 /* Recheck anchored substring, but not floating... */
735 DEBUG_r( PerlIO_printf(Perl_debug_log,
736 "Trying anchored substr starting at offset %ld...\n",
737 (long)(other_last - i_strpos)) );
738 goto do_other_anchored;
740 /* Another way we could have checked stclass at the
741 current position only: */
744 DEBUG_r( PerlIO_printf(Perl_debug_log,
745 "Trying /^/m starting at offset %ld...\n",
746 (long)(t - i_strpos)) );
749 if (!prog->float_substr) /* Could have been deleted */
751 /* Check is floating subtring. */
752 retry_floating_check:
753 t = check_at - start_shift;
754 DEBUG_r( what = "floating" );
755 goto hop_and_restart;
758 PerlIO_printf(Perl_debug_log,
759 "By STCLASS: moving %ld --> %ld\n",
760 (long)(t - i_strpos), (long)(s - i_strpos));
762 PerlIO_printf(Perl_debug_log,
763 "Does not contradict STCLASS...\n") );
765 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sGuessed:%s match at offset %ld\n",
766 PL_colors[4], PL_colors[5], (long)(s - i_strpos)) );
769 fail_finish: /* Substring not found */
770 if (prog->check_substr) /* could be removed already */
771 BmUSEFUL(prog->check_substr) += 5; /* hooray */
773 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
774 PL_colors[4],PL_colors[5]));
778 /* We know what class REx starts with. Try to find this position... */
780 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
782 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
788 register I32 tmp = 1; /* Scratch variable? */
790 /* We know what class it must start with. */
794 if (REGINCLASSUTF8(c, (U8*)s)) {
795 if (tmp && (norun || regtry(prog, s)))
807 if (REGINCLASS(c, *s)) {
808 if (tmp && (norun || regtry(prog, s)))
828 c2 = PL_fold_locale[c1];
833 e = s; /* Due to minlen logic of intuit() */
834 /* Here it is NOT UTF! */
838 && (ln == 1 || !(OP(c) == EXACTF
840 : ibcmp_locale(s, m, ln)))
841 && (norun || regtry(prog, s)) )
847 if ( (*s == c1 || *s == c2)
848 && (ln == 1 || !(OP(c) == EXACTF
850 : ibcmp_locale(s, m, ln)))
851 && (norun || regtry(prog, s)) )
858 PL_reg_flags |= RF_tainted;
861 tmp = (s != startpos) ? UCHARAT(s - 1) : '\n';
862 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
864 if (tmp == !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
866 if ((norun || regtry(prog, s)))
871 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
875 PL_reg_flags |= RF_tainted;
878 tmp = (I32)(s != startpos) ? utf8_to_uv(reghop((U8*)s, -1), 0) : '\n';
879 tmp = ((OP(c) == BOUNDUTF8 ? isALNUM_uni(tmp) : isALNUM_LC_uni(tmp)) != 0);
881 if (tmp == !(OP(c) == BOUNDUTF8 ?
882 swash_fetch(PL_utf8_alnum, (U8*)s) :
883 isALNUM_LC_utf8((U8*)s)))
886 if ((norun || regtry(prog, s)))
891 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
895 PL_reg_flags |= RF_tainted;
898 tmp = (s != startpos) ? UCHARAT(s - 1) : '\n';
899 tmp = ((OP(c) == NBOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
901 if (tmp == !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
903 else if ((norun || regtry(prog, s)))
907 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
911 PL_reg_flags |= RF_tainted;
914 tmp = (I32)(s != startpos) ? utf8_to_uv(reghop((U8*)s, -1), 0) : '\n';
915 tmp = ((OP(c) == NBOUNDUTF8 ? isALNUM_uni(tmp) : isALNUM_LC_uni(tmp)) != 0);
917 if (tmp == !(OP(c) == NBOUNDUTF8 ?
918 swash_fetch(PL_utf8_alnum, (U8*)s) :
919 isALNUM_LC_utf8((U8*)s)))
921 else if ((norun || regtry(prog, s)))
925 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
931 if (tmp && (norun || regtry(prog, s)))
943 if (swash_fetch(PL_utf8_alnum, (U8*)s)) {
944 if (tmp && (norun || regtry(prog, s)))
955 PL_reg_flags |= RF_tainted;
957 if (isALNUM_LC(*s)) {
958 if (tmp && (norun || regtry(prog, s)))
969 PL_reg_flags |= RF_tainted;
971 if (isALNUM_LC_utf8((U8*)s)) {
972 if (tmp && (norun || regtry(prog, s)))
985 if (tmp && (norun || regtry(prog, s)))
997 if (!swash_fetch(PL_utf8_alnum, (U8*)s)) {
998 if (tmp && (norun || regtry(prog, s)))
1009 PL_reg_flags |= RF_tainted;
1010 while (s < strend) {
1011 if (!isALNUM_LC(*s)) {
1012 if (tmp && (norun || regtry(prog, s)))
1023 PL_reg_flags |= RF_tainted;
1024 while (s < strend) {
1025 if (!isALNUM_LC_utf8((U8*)s)) {
1026 if (tmp && (norun || regtry(prog, s)))
1037 while (s < strend) {
1039 if (tmp && (norun || regtry(prog, s)))
1050 while (s < strend) {
1051 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s)) {
1052 if (tmp && (norun || regtry(prog, s)))
1063 PL_reg_flags |= RF_tainted;
1064 while (s < strend) {
1065 if (isSPACE_LC(*s)) {
1066 if (tmp && (norun || regtry(prog, s)))
1077 PL_reg_flags |= RF_tainted;
1078 while (s < strend) {
1079 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1080 if (tmp && (norun || regtry(prog, s)))
1091 while (s < strend) {
1093 if (tmp && (norun || regtry(prog, s)))
1104 while (s < strend) {
1105 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s))) {
1106 if (tmp && (norun || regtry(prog, s)))
1117 PL_reg_flags |= RF_tainted;
1118 while (s < strend) {
1119 if (!isSPACE_LC(*s)) {
1120 if (tmp && (norun || regtry(prog, s)))
1131 PL_reg_flags |= RF_tainted;
1132 while (s < strend) {
1133 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1134 if (tmp && (norun || regtry(prog, s)))
1145 while (s < strend) {
1147 if (tmp && (norun || regtry(prog, s)))
1158 while (s < strend) {
1159 if (swash_fetch(PL_utf8_digit,(U8*)s)) {
1160 if (tmp && (norun || regtry(prog, s)))
1171 PL_reg_flags |= RF_tainted;
1172 while (s < strend) {
1173 if (isDIGIT_LC(*s)) {
1174 if (tmp && (norun || regtry(prog, s)))
1185 PL_reg_flags |= RF_tainted;
1186 while (s < strend) {
1187 if (isDIGIT_LC_utf8((U8*)s)) {
1188 if (tmp && (norun || regtry(prog, s)))
1199 while (s < strend) {
1201 if (tmp && (norun || regtry(prog, s)))
1212 while (s < strend) {
1213 if (!swash_fetch(PL_utf8_digit,(U8*)s)) {
1214 if (tmp && (norun || regtry(prog, s)))
1225 PL_reg_flags |= RF_tainted;
1226 while (s < strend) {
1227 if (!isDIGIT_LC(*s)) {
1228 if (tmp && (norun || regtry(prog, s)))
1239 PL_reg_flags |= RF_tainted;
1240 while (s < strend) {
1241 if (!isDIGIT_LC_utf8((U8*)s)) {
1242 if (tmp && (norun || regtry(prog, s)))
1253 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1262 - regexec_flags - match a regexp against a string
1265 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1266 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1267 /* strend: pointer to null at end of string */
1268 /* strbeg: real beginning of string */
1269 /* minend: end of match must be >=minend after stringarg. */
1270 /* data: May be used for some additional optimizations. */
1271 /* nosave: For optimizations. */
1275 register regnode *c;
1276 register char *startpos = stringarg;
1278 I32 minlen; /* must match at least this many chars */
1279 I32 dontbother = 0; /* how many characters not to try at end */
1280 I32 start_shift = 0; /* Offset of the start to find
1281 constant substr. */ /* CC */
1282 I32 end_shift = 0; /* Same for the end. */ /* CC */
1283 I32 scream_pos = -1; /* Internal iterator of scream. */
1285 SV* oreplsv = GvSV(PL_replgv);
1291 PL_regnarrate = PL_debug & 512;
1294 /* Be paranoid... */
1295 if (prog == NULL || startpos == NULL) {
1296 Perl_croak(aTHX_ "NULL regexp parameter");
1300 minlen = prog->minlen;
1301 if (strend - startpos < minlen) goto phooey;
1303 if (startpos == strbeg) /* is ^ valid at stringarg? */
1306 PL_regprev = (U32)stringarg[-1];
1307 if (!PL_multiline && PL_regprev == '\n')
1308 PL_regprev = '\0'; /* force ^ to NOT match */
1311 /* Check validity of program. */
1312 if (UCHARAT(prog->program) != REG_MAGIC) {
1313 Perl_croak(aTHX_ "corrupted regexp program");
1317 PL_reg_eval_set = 0;
1320 if (prog->reganch & ROPT_UTF8)
1321 PL_reg_flags |= RF_utf8;
1323 /* Mark beginning of line for ^ and lookbehind. */
1324 PL_regbol = startpos;
1328 /* Mark end of line for $ (and such) */
1331 /* see how far we have to get to not match where we matched before */
1332 PL_regtill = startpos+minend;
1334 /* We start without call_cc context. */
1337 /* If there is a "must appear" string, look for it. */
1340 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1343 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1344 PL_reg_ganch = startpos;
1345 else if (sv && SvTYPE(sv) >= SVt_PVMG
1347 && (mg = mg_find(sv, 'g')) && mg->mg_len >= 0) {
1348 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1349 if (prog->reganch & ROPT_ANCH_GPOS) {
1350 if (s > PL_reg_ganch)
1355 else /* pos() not defined */
1356 PL_reg_ganch = strbeg;
1359 if (!(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
1360 re_scream_pos_data d;
1362 d.scream_olds = &scream_olds;
1363 d.scream_pos = &scream_pos;
1364 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1366 goto phooey; /* not present */
1369 DEBUG_r( if (!PL_colorset) reginitcolors() );
1370 DEBUG_r(PerlIO_printf(Perl_debug_log,
1371 "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
1372 PL_colors[4],PL_colors[5],PL_colors[0],
1375 (strlen(prog->precomp) > 60 ? "..." : ""),
1377 (int)(strend - startpos > 60 ? 60 : strend - startpos),
1378 startpos, PL_colors[1],
1379 (strend - startpos > 60 ? "..." : ""))
1382 /* Simplest case: anchored match need be tried only once. */
1383 /* [unless only anchor is BOL and multiline is set] */
1384 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1385 if (s == startpos && regtry(prog, startpos))
1387 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1388 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1393 dontbother = minlen - 1;
1394 end = HOPc(strend, -dontbother) - 1;
1395 /* for multiline we only have to try after newlines */
1396 if (prog->check_substr) {
1400 if (regtry(prog, s))
1405 if (prog->reganch & RE_USE_INTUIT) {
1406 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1417 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1418 if (regtry(prog, s))
1425 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1426 if (regtry(prog, PL_reg_ganch))
1431 /* Messy cases: unanchored match. */
1432 if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
1433 /* we have /x+whatever/ */
1434 /* it must be a one character string (XXXX Except UTF?) */
1435 char ch = SvPVX(prog->anchored_substr)[0];
1437 while (s < strend) {
1439 if (regtry(prog, s)) goto got_it;
1441 while (s < strend && *s == ch)
1448 while (s < strend) {
1450 if (regtry(prog, s)) goto got_it;
1452 while (s < strend && *s == ch)
1460 else if (prog->anchored_substr != Nullsv
1461 || (prog->float_substr != Nullsv
1462 && prog->float_max_offset < strend - s)) {
1463 SV *must = prog->anchored_substr
1464 ? prog->anchored_substr : prog->float_substr;
1466 prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset;
1468 prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
1469 I32 delta = back_max - back_min;
1470 char *last = HOPc(strend, /* Cannot start after this */
1471 -(I32)(CHR_SVLEN(must)
1472 - (SvTAIL(must) != 0) + back_min));
1473 char *last1; /* Last position checked before */
1476 last1 = HOPc(s, -1);
1478 last1 = s - 1; /* bogus */
1480 /* XXXX check_substr already used to find `s', can optimize if
1481 check_substr==must. */
1483 dontbother = end_shift;
1484 strend = HOPc(strend, -dontbother);
1485 while ( (s <= last) &&
1486 ((flags & REXEC_SCREAM)
1487 ? (s = screaminstr(sv, must, HOPc(s, back_min) - strbeg,
1488 end_shift, &scream_pos, 0))
1489 : (s = fbm_instr((unsigned char*)HOP(s, back_min),
1490 (unsigned char*)strend, must,
1491 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1492 if (HOPc(s, -back_max) > last1) {
1493 last1 = HOPc(s, -back_min);
1494 s = HOPc(s, -back_max);
1497 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1499 last1 = HOPc(s, -back_min);
1503 while (s <= last1) {
1504 if (regtry(prog, s))
1510 while (s <= last1) {
1511 if (regtry(prog, s))
1519 else if (c = prog->regstclass) {
1520 if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT)
1521 /* don't bother with what can't match */
1522 strend = HOPc(strend, -(minlen - 1));
1523 if (find_byclass(prog, c, s, strend, startpos, 0))
1528 if (prog->float_substr != Nullsv) { /* Trim the end. */
1530 I32 oldpos = scream_pos;
1532 if (flags & REXEC_SCREAM) {
1533 last = screaminstr(sv, prog->float_substr, s - strbeg,
1534 end_shift, &scream_pos, 1); /* last one */
1536 last = scream_olds; /* Only one occurence. */
1540 char *little = SvPV(prog->float_substr, len);
1542 if (SvTAIL(prog->float_substr)) {
1543 if (memEQ(strend - len + 1, little, len - 1))
1544 last = strend - len + 1;
1545 else if (!PL_multiline)
1546 last = memEQ(strend - len, little, len)
1547 ? strend - len : Nullch;
1553 last = rninstr(s, strend, little, little + len);
1555 last = strend; /* matching `$' */
1558 if (last == NULL) goto phooey; /* Should not happen! */
1559 dontbother = strend - last + prog->float_min_offset;
1561 if (minlen && (dontbother < minlen))
1562 dontbother = minlen - 1;
1563 strend -= dontbother; /* this one's always in bytes! */
1564 /* We don't know much -- general case. */
1567 if (regtry(prog, s))
1576 if (regtry(prog, s))
1578 } while (s++ < strend);
1586 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1588 if (PL_reg_eval_set) {
1589 /* Preserve the current value of $^R */
1590 if (oreplsv != GvSV(PL_replgv))
1591 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1592 restored, the value remains
1594 restore_pos(aTHXo_ 0);
1597 /* make sure $`, $&, $', and $digit will work later */
1598 if ( !(flags & REXEC_NOT_FIRST) ) {
1599 if (RX_MATCH_COPIED(prog)) {
1600 Safefree(prog->subbeg);
1601 RX_MATCH_COPIED_off(prog);
1603 if (flags & REXEC_COPY_STR) {
1604 I32 i = PL_regeol - startpos + (stringarg - strbeg);
1606 s = savepvn(strbeg, i);
1609 RX_MATCH_COPIED_on(prog);
1612 prog->subbeg = strbeg;
1613 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
1620 if (PL_reg_eval_set)
1621 restore_pos(aTHXo_ 0);
1626 - regtry - try match at specific point
1628 STATIC I32 /* 0 failure, 1 success */
1629 S_regtry(pTHX_ regexp *prog, char *startpos)
1637 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
1640 PL_reg_eval_set = RS_init;
1642 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
1643 (IV)(PL_stack_sp - PL_stack_base));
1645 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
1646 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
1647 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
1649 /* Apparently this is not needed, judging by wantarray. */
1650 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
1651 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
1654 /* Make $_ available to executed code. */
1655 if (PL_reg_sv != DEFSV) {
1656 /* SAVE_DEFSV does *not* suffice here for USE_THREADS */
1661 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
1662 && (mg = mg_find(PL_reg_sv, 'g')))) {
1663 /* prepare for quick setting of pos */
1664 sv_magic(PL_reg_sv, (SV*)0, 'g', Nullch, 0);
1665 mg = mg_find(PL_reg_sv, 'g');
1669 PL_reg_oldpos = mg->mg_len;
1670 SAVEDESTRUCTOR_X(restore_pos, 0);
1673 New(22,PL_reg_curpm, 1, PMOP);
1674 PL_reg_curpm->op_pmregexp = prog;
1675 PL_reg_oldcurpm = PL_curpm;
1676 PL_curpm = PL_reg_curpm;
1677 if (RX_MATCH_COPIED(prog)) {
1678 /* Here is a serious problem: we cannot rewrite subbeg,
1679 since it may be needed if this match fails. Thus
1680 $` inside (?{}) could fail... */
1681 PL_reg_oldsaved = prog->subbeg;
1682 PL_reg_oldsavedlen = prog->sublen;
1683 RX_MATCH_COPIED_off(prog);
1686 PL_reg_oldsaved = Nullch;
1687 prog->subbeg = PL_bostr;
1688 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
1690 prog->startp[0] = startpos - PL_bostr;
1691 PL_reginput = startpos;
1692 PL_regstartp = prog->startp;
1693 PL_regendp = prog->endp;
1694 PL_reglastparen = &prog->lastparen;
1695 prog->lastparen = 0;
1697 DEBUG_r(PL_reg_starttry = startpos);
1698 if (PL_reg_start_tmpl <= prog->nparens) {
1699 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
1700 if(PL_reg_start_tmp)
1701 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1703 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1706 /* XXXX What this code is doing here?!!! There should be no need
1707 to do this again and again, PL_reglastparen should take care of
1711 if (prog->nparens) {
1712 for (i = prog->nparens; i >= 1; i--) {
1718 if (regmatch(prog->program + 1)) {
1719 prog->endp[0] = PL_reginput - PL_bostr;
1727 - regmatch - main matching routine
1729 * Conceptually the strategy is simple: check to see whether the current
1730 * node matches, call self recursively to see whether the rest matches,
1731 * and then act accordingly. In practice we make some effort to avoid
1732 * recursion, in particular by going through "ordinary" nodes (that don't
1733 * need to know whether the rest of the match failed) by a loop instead of
1736 /* [lwall] I've hoisted the register declarations to the outer block in order to
1737 * maybe save a little bit of pushing and popping on the stack. It also takes
1738 * advantage of machines that use a register save mask on subroutine entry.
1740 STATIC I32 /* 0 failure, 1 success */
1741 S_regmatch(pTHX_ regnode *prog)
1744 register regnode *scan; /* Current node. */
1745 regnode *next; /* Next node. */
1746 regnode *inner; /* Next node in internal branch. */
1747 register I32 nextchr; /* renamed nextchr - nextchar colides with
1748 function of same name */
1749 register I32 n; /* no or next */
1750 register I32 ln; /* len or last */
1751 register char *s; /* operand or save */
1752 register char *locinput = PL_reginput;
1753 register I32 c1, c2, paren; /* case fold search, parenth */
1754 int minmod = 0, sw = 0, logical = 0;
1759 /* Note that nextchr is a byte even in UTF */
1760 nextchr = UCHARAT(locinput);
1762 while (scan != NULL) {
1763 #define sayNO_L (logical ? (logical = 0, sw = 0, goto cont) : sayNO)
1765 # define sayYES goto yes
1766 # define sayNO goto no
1767 # define sayYES_FINAL goto yes_final
1768 # define sayYES_LOUD goto yes_loud
1769 # define sayNO_FINAL goto no_final
1770 # define sayNO_SILENT goto do_no
1771 # define saySAME(x) if (x) goto yes; else goto no
1772 # define REPORT_CODE_OFF 24
1774 # define sayYES return 1
1775 # define sayNO return 0
1776 # define sayYES_FINAL return 1
1777 # define sayYES_LOUD return 1
1778 # define sayNO_FINAL return 0
1779 # define sayNO_SILENT return 0
1780 # define saySAME(x) return x
1783 SV *prop = sv_newmortal();
1784 int docolor = *PL_colors[0];
1785 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
1786 int l = (PL_regeol - locinput > taill ? taill : PL_regeol - locinput);
1787 /* The part of the string before starttry has one color
1788 (pref0_len chars), between starttry and current
1789 position another one (pref_len - pref0_len chars),
1790 after the current position the third one.
1791 We assume that pref0_len <= pref_len, otherwise we
1792 decrease pref0_len. */
1793 int pref_len = (locinput - PL_bostr > (5 + taill) - l
1794 ? (5 + taill) - l : locinput - PL_bostr);
1795 int pref0_len = pref_len - (locinput - PL_reg_starttry);
1797 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
1798 l = ( PL_regeol - locinput > (5 + taill) - pref_len
1799 ? (5 + taill) - pref_len : PL_regeol - locinput);
1802 if (pref0_len > pref_len)
1803 pref0_len = pref_len;
1804 regprop(prop, scan);
1805 PerlIO_printf(Perl_debug_log,
1806 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
1807 (IV)(locinput - PL_bostr),
1808 PL_colors[4], pref0_len,
1809 locinput - pref_len, PL_colors[5],
1810 PL_colors[2], pref_len - pref0_len,
1811 locinput - pref_len + pref0_len, PL_colors[3],
1812 (docolor ? "" : "> <"),
1813 PL_colors[0], l, locinput, PL_colors[1],
1814 15 - l - pref_len + 1,
1816 (IV)(scan - PL_regprogram), PL_regindent*2, "",
1820 next = scan + NEXT_OFF(scan);
1826 if (locinput == PL_bostr
1827 ? PL_regprev == '\n'
1829 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
1831 /* regtill = regbol; */
1836 if (locinput == PL_bostr
1837 ? PL_regprev == '\n'
1838 : ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
1844 if (locinput == PL_regbol && PL_regprev == '\n')
1848 if (locinput == PL_reg_ganch)
1858 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
1863 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
1865 if (PL_regeol - locinput > 1)
1869 if (PL_regeol != locinput)
1873 if (nextchr & 0x80) {
1874 locinput += PL_utf8skip[nextchr];
1875 if (locinput > PL_regeol)
1877 nextchr = UCHARAT(locinput);
1880 if (!nextchr && locinput >= PL_regeol)
1882 nextchr = UCHARAT(++locinput);
1885 if (!nextchr && locinput >= PL_regeol)
1887 nextchr = UCHARAT(++locinput);
1890 if (nextchr & 0x80) {
1891 locinput += PL_utf8skip[nextchr];
1892 if (locinput > PL_regeol)
1894 nextchr = UCHARAT(locinput);
1897 if (!nextchr && locinput >= PL_regeol || nextchr == '\n')
1899 nextchr = UCHARAT(++locinput);
1902 if (!nextchr && locinput >= PL_regeol || nextchr == '\n')
1904 nextchr = UCHARAT(++locinput);
1909 /* Inline the first character, for speed. */
1910 if (UCHARAT(s) != nextchr)
1912 if (PL_regeol - locinput < ln)
1914 if (ln > 1 && memNE(s, locinput, ln))
1917 nextchr = UCHARAT(locinput);
1920 PL_reg_flags |= RF_tainted;
1929 c1 = OP(scan) == EXACTF;
1933 if (utf8_to_uv((U8*)s, 0) != (c1 ?
1934 toLOWER_utf8((U8*)l) :
1935 toLOWER_LC_utf8((U8*)l)))
1943 nextchr = UCHARAT(locinput);
1947 /* Inline the first character, for speed. */
1948 if (UCHARAT(s) != nextchr &&
1949 UCHARAT(s) != ((OP(scan) == EXACTF)
1950 ? PL_fold : PL_fold_locale)[nextchr])
1952 if (PL_regeol - locinput < ln)
1954 if (ln > 1 && (OP(scan) == EXACTF
1955 ? ibcmp(s, locinput, ln)
1956 : ibcmp_locale(s, locinput, ln)))
1959 nextchr = UCHARAT(locinput);
1962 if (!REGINCLASSUTF8(scan, (U8*)locinput))
1964 if (locinput >= PL_regeol)
1966 locinput += PL_utf8skip[nextchr];
1967 nextchr = UCHARAT(locinput);
1971 nextchr = UCHARAT(locinput);
1972 if (!REGINCLASS(scan, nextchr))
1974 if (!nextchr && locinput >= PL_regeol)
1976 nextchr = UCHARAT(++locinput);
1979 PL_reg_flags |= RF_tainted;
1984 if (!(OP(scan) == ALNUM
1985 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
1987 nextchr = UCHARAT(++locinput);
1990 PL_reg_flags |= RF_tainted;
1995 if (nextchr & 0x80) {
1996 if (!(OP(scan) == ALNUMUTF8
1997 ? swash_fetch(PL_utf8_alnum, (U8*)locinput)
1998 : isALNUM_LC_utf8((U8*)locinput)))
2002 locinput += PL_utf8skip[nextchr];
2003 nextchr = UCHARAT(locinput);
2006 if (!(OP(scan) == ALNUMUTF8
2007 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2009 nextchr = UCHARAT(++locinput);
2012 PL_reg_flags |= RF_tainted;
2015 if (!nextchr && locinput >= PL_regeol)
2017 if (OP(scan) == NALNUM
2018 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2020 nextchr = UCHARAT(++locinput);
2023 PL_reg_flags |= RF_tainted;
2026 if (!nextchr && locinput >= PL_regeol)
2028 if (nextchr & 0x80) {
2029 if (OP(scan) == NALNUMUTF8
2030 ? swash_fetch(PL_utf8_alnum, (U8*)locinput)
2031 : isALNUM_LC_utf8((U8*)locinput))
2035 locinput += PL_utf8skip[nextchr];
2036 nextchr = UCHARAT(locinput);
2039 if (OP(scan) == NALNUMUTF8
2040 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2042 nextchr = UCHARAT(++locinput);
2046 PL_reg_flags |= RF_tainted;
2050 /* was last char in word? */
2051 ln = (locinput != PL_regbol) ? UCHARAT(locinput - 1) : PL_regprev;
2052 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2054 n = isALNUM(nextchr);
2057 ln = isALNUM_LC(ln);
2058 n = isALNUM_LC(nextchr);
2060 if (((!ln) == (!n)) == (OP(scan) == BOUND || OP(scan) == BOUNDL))
2065 PL_reg_flags |= RF_tainted;
2069 /* was last char in word? */
2070 ln = (locinput != PL_regbol)
2071 ? utf8_to_uv(reghop((U8*)locinput, -1), 0) : PL_regprev;
2072 if (OP(scan) == BOUNDUTF8 || OP(scan) == NBOUNDUTF8) {
2073 ln = isALNUM_uni(ln);
2074 n = swash_fetch(PL_utf8_alnum, (U8*)locinput);
2077 ln = isALNUM_LC_uni(ln);
2078 n = isALNUM_LC_utf8((U8*)locinput);
2080 if (((!ln) == (!n)) == (OP(scan) == BOUNDUTF8 || OP(scan) == BOUNDLUTF8))
2084 PL_reg_flags |= RF_tainted;
2087 if (!nextchr && locinput >= PL_regeol)
2089 if (!(OP(scan) == SPACE
2090 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2092 nextchr = UCHARAT(++locinput);
2095 PL_reg_flags |= RF_tainted;
2098 if (!nextchr && locinput >= PL_regeol)
2100 if (nextchr & 0x80) {
2101 if (!(OP(scan) == SPACEUTF8
2102 ? swash_fetch(PL_utf8_space,(U8*)locinput)
2103 : isSPACE_LC_utf8((U8*)locinput)))
2107 locinput += PL_utf8skip[nextchr];
2108 nextchr = UCHARAT(locinput);
2111 if (!(OP(scan) == SPACEUTF8
2112 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2114 nextchr = UCHARAT(++locinput);
2117 PL_reg_flags |= RF_tainted;
2122 if (OP(scan) == SPACE
2123 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2125 nextchr = UCHARAT(++locinput);
2128 PL_reg_flags |= RF_tainted;
2133 if (nextchr & 0x80) {
2134 if (OP(scan) == NSPACEUTF8
2135 ? swash_fetch(PL_utf8_space,(U8*)locinput)
2136 : isSPACE_LC_utf8((U8*)locinput))
2140 locinput += PL_utf8skip[nextchr];
2141 nextchr = UCHARAT(locinput);
2144 if (OP(scan) == NSPACEUTF8
2145 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2147 nextchr = UCHARAT(++locinput);
2150 PL_reg_flags |= RF_tainted;
2153 if (!nextchr && locinput >= PL_regeol)
2155 if (!(OP(scan) == DIGIT
2156 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2158 nextchr = UCHARAT(++locinput);
2161 PL_reg_flags |= RF_tainted;
2166 if (nextchr & 0x80) {
2167 if (OP(scan) == NDIGITUTF8
2168 ? swash_fetch(PL_utf8_digit,(U8*)locinput)
2169 : isDIGIT_LC_utf8((U8*)locinput))
2173 locinput += PL_utf8skip[nextchr];
2174 nextchr = UCHARAT(locinput);
2177 if (!isDIGIT(nextchr))
2179 nextchr = UCHARAT(++locinput);
2182 PL_reg_flags |= RF_tainted;
2187 if (OP(scan) == DIGIT
2188 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2190 nextchr = UCHARAT(++locinput);
2193 PL_reg_flags |= RF_tainted;
2196 if (!nextchr && locinput >= PL_regeol)
2198 if (nextchr & 0x80) {
2199 if (swash_fetch(PL_utf8_digit,(U8*)locinput))
2201 locinput += PL_utf8skip[nextchr];
2202 nextchr = UCHARAT(locinput);
2205 if (isDIGIT(nextchr))
2207 nextchr = UCHARAT(++locinput);
2210 if (locinput >= PL_regeol || swash_fetch(PL_utf8_mark,(U8*)locinput))
2212 locinput += PL_utf8skip[nextchr];
2213 while (locinput < PL_regeol && swash_fetch(PL_utf8_mark,(U8*)locinput))
2214 locinput += UTF8SKIP(locinput);
2215 if (locinput > PL_regeol)
2217 nextchr = UCHARAT(locinput);
2220 PL_reg_flags |= RF_tainted;
2224 n = ARG(scan); /* which paren pair */
2225 ln = PL_regstartp[n];
2226 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2227 if (*PL_reglastparen < n || ln == -1)
2228 sayNO; /* Do not match unless seen CLOSEn. */
2229 if (ln == PL_regendp[n])
2233 if (UTF && OP(scan) != REF) { /* REF can do byte comparison */
2235 char *e = PL_bostr + PL_regendp[n];
2237 * Note that we can't do the "other character" lookup trick as
2238 * in the 8-bit case (no pun intended) because in Unicode we
2239 * have to map both upper and title case to lower case.
2241 if (OP(scan) == REFF) {
2245 if (toLOWER_utf8((U8*)s) != toLOWER_utf8((U8*)l))
2255 if (toLOWER_LC_utf8((U8*)s) != toLOWER_LC_utf8((U8*)l))
2262 nextchr = UCHARAT(locinput);
2266 /* Inline the first character, for speed. */
2267 if (UCHARAT(s) != nextchr &&
2269 (UCHARAT(s) != ((OP(scan) == REFF
2270 ? PL_fold : PL_fold_locale)[nextchr]))))
2272 ln = PL_regendp[n] - ln;
2273 if (locinput + ln > PL_regeol)
2275 if (ln > 1 && (OP(scan) == REF
2276 ? memNE(s, locinput, ln)
2278 ? ibcmp(s, locinput, ln)
2279 : ibcmp_locale(s, locinput, ln))))
2282 nextchr = UCHARAT(locinput);
2293 OP_4tree *oop = PL_op;
2294 COP *ocurcop = PL_curcop;
2295 SV **ocurpad = PL_curpad;
2299 PL_op = (OP_4tree*)PL_regdata->data[n];
2300 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2301 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2302 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2304 CALLRUNOPS(aTHX); /* Scalar context. */
2310 PL_curpad = ocurpad;
2311 PL_curcop = ocurcop;
2313 if (logical == 2) { /* Postponed subexpression. */
2315 MAGIC *mg = Null(MAGIC*);
2317 CHECKPOINT cp, lastcp;
2319 if(SvROK(ret) || SvRMAGICAL(ret)) {
2320 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2323 mg = mg_find(sv, 'r');
2326 re = (regexp *)mg->mg_obj;
2327 (void)ReREFCNT_inc(re);
2331 char *t = SvPV(ret, len);
2333 char *oprecomp = PL_regprecomp;
2334 I32 osize = PL_regsize;
2335 I32 onpar = PL_regnpar;
2338 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2340 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2341 sv_magic(ret,(SV*)ReREFCNT_inc(re),'r',0,0);
2342 PL_regprecomp = oprecomp;
2347 PerlIO_printf(Perl_debug_log,
2348 "Entering embedded `%s%.60s%s%s'\n",
2352 (strlen(re->precomp) > 60 ? "..." : ""))
2355 state.prev = PL_reg_call_cc;
2356 state.cc = PL_regcc;
2357 state.re = PL_reg_re;
2361 cp = regcppush(0); /* Save *all* the positions. */
2364 state.ss = PL_savestack_ix;
2365 *PL_reglastparen = 0;
2366 PL_reg_call_cc = &state;
2367 PL_reginput = locinput;
2369 /* XXXX This is too dramatic a measure... */
2372 if (regmatch(re->program + 1)) {
2373 /* Even though we succeeded, we need to restore
2374 global variables, since we may be wrapped inside
2375 SUSPEND, thus the match may be not finished yet. */
2377 /* XXXX Do this only if SUSPENDed? */
2378 PL_reg_call_cc = state.prev;
2379 PL_regcc = state.cc;
2380 PL_reg_re = state.re;
2381 cache_re(PL_reg_re);
2383 /* XXXX This is too dramatic a measure... */
2386 /* These are needed even if not SUSPEND. */
2394 PL_reg_call_cc = state.prev;
2395 PL_regcc = state.cc;
2396 PL_reg_re = state.re;
2397 cache_re(PL_reg_re);
2399 /* XXXX This is too dramatic a measure... */
2408 sv_setsv(save_scalar(PL_replgv), ret);
2412 n = ARG(scan); /* which paren pair */
2413 PL_reg_start_tmp[n] = locinput;
2418 n = ARG(scan); /* which paren pair */
2419 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2420 PL_regendp[n] = locinput - PL_bostr;
2421 if (n > *PL_reglastparen)
2422 *PL_reglastparen = n;
2425 n = ARG(scan); /* which paren pair */
2426 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2429 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2431 next = NEXTOPER(NEXTOPER(scan));
2433 next = scan + ARG(scan);
2434 if (OP(next) == IFTHEN) /* Fake one. */
2435 next = NEXTOPER(NEXTOPER(next));
2439 logical = scan->flags;
2441 /*******************************************************************
2442 PL_regcc contains infoblock about the innermost (...)* loop, and
2443 a pointer to the next outer infoblock.
2445 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2447 1) After matching X, regnode for CURLYX is processed;
2449 2) This regnode creates infoblock on the stack, and calls
2450 regmatch() recursively with the starting point at WHILEM node;
2452 3) Each hit of WHILEM node tries to match A and Z (in the order
2453 depending on the current iteration, min/max of {min,max} and
2454 greediness). The information about where are nodes for "A"
2455 and "Z" is read from the infoblock, as is info on how many times "A"
2456 was already matched, and greediness.
2458 4) After A matches, the same WHILEM node is hit again.
2460 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2461 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2462 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2463 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2464 of the external loop.
2466 Currently present infoblocks form a tree with a stem formed by PL_curcc
2467 and whatever it mentions via ->next, and additional attached trees
2468 corresponding to temporarily unset infoblocks as in "5" above.
2470 In the following picture infoblocks for outer loop of
2471 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2472 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2473 infoblocks are drawn below the "reset" infoblock.
2475 In fact in the picture below we do not show failed matches for Z and T
2476 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2477 more obvious *why* one needs to *temporary* unset infoblocks.]
2479 Matched REx position InfoBlocks Comment
2483 Y A)*?Z)*?T x <- O <- I
2484 YA )*?Z)*?T x <- O <- I
2485 YA A)*?Z)*?T x <- O <- I
2486 YAA )*?Z)*?T x <- O <- I
2487 YAA Z)*?T x <- O # Temporary unset I
2490 YAAZ Y(A)*?Z)*?T x <- O
2493 YAAZY (A)*?Z)*?T x <- O
2496 YAAZY A)*?Z)*?T x <- O <- I
2499 YAAZYA )*?Z)*?T x <- O <- I
2502 YAAZYA Z)*?T x <- O # Temporary unset I
2508 YAAZYAZ T x # Temporary unset O
2515 *******************************************************************/
2518 CHECKPOINT cp = PL_savestack_ix;
2520 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2522 cc.oldcc = PL_regcc;
2524 cc.parenfloor = *PL_reglastparen;
2526 cc.min = ARG1(scan);
2527 cc.max = ARG2(scan);
2528 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2532 PL_reginput = locinput;
2533 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2535 PL_regcc = cc.oldcc;
2541 * This is really hard to understand, because after we match
2542 * what we're trying to match, we must make sure the rest of
2543 * the REx is going to match for sure, and to do that we have
2544 * to go back UP the parse tree by recursing ever deeper. And
2545 * if it fails, we have to reset our parent's current state
2546 * that we can try again after backing off.
2549 CHECKPOINT cp, lastcp;
2550 CURCUR* cc = PL_regcc;
2551 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2553 n = cc->cur + 1; /* how many we know we matched */
2554 PL_reginput = locinput;
2557 PerlIO_printf(Perl_debug_log,
2558 "%*s %ld out of %ld..%ld cc=%lx\n",
2559 REPORT_CODE_OFF+PL_regindent*2, "",
2560 (long)n, (long)cc->min,
2561 (long)cc->max, (long)cc)
2564 /* If degenerate scan matches "", assume scan done. */
2566 if (locinput == cc->lastloc && n >= cc->min) {
2567 PL_regcc = cc->oldcc;
2571 PerlIO_printf(Perl_debug_log,
2572 "%*s empty match detected, try continuation...\n",
2573 REPORT_CODE_OFF+PL_regindent*2, "")
2575 if (regmatch(cc->next))
2583 /* First just match a string of min scans. */
2587 cc->lastloc = locinput;
2588 if (regmatch(cc->scan))
2591 cc->lastloc = lastloc;
2596 /* Check whether we already were at this position.
2597 Postpone detection until we know the match is not
2598 *that* much linear. */
2599 if (!PL_reg_maxiter) {
2600 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
2601 PL_reg_leftiter = PL_reg_maxiter;
2603 if (PL_reg_leftiter-- == 0) {
2604 I32 size = (PL_reg_maxiter + 7)/8;
2605 if (PL_reg_poscache) {
2606 if (PL_reg_poscache_size < size) {
2607 Renew(PL_reg_poscache, size, char);
2608 PL_reg_poscache_size = size;
2610 Zero(PL_reg_poscache, size, char);
2613 PL_reg_poscache_size = size;
2614 Newz(29, PL_reg_poscache, size, char);
2617 PerlIO_printf(Perl_debug_log,
2618 "%sDetected a super-linear match, switching on caching%s...\n",
2619 PL_colors[4], PL_colors[5])
2622 if (PL_reg_leftiter < 0) {
2623 I32 o = locinput - PL_bostr, b;
2625 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
2628 if (PL_reg_poscache[o] & (1<<b)) {
2630 PerlIO_printf(Perl_debug_log,
2631 "%*s already tried at this position...\n",
2632 REPORT_CODE_OFF+PL_regindent*2, "")
2636 PL_reg_poscache[o] |= (1<<b);
2640 /* Prefer next over scan for minimal matching. */
2643 PL_regcc = cc->oldcc;
2646 cp = regcppush(cc->parenfloor);
2648 if (regmatch(cc->next)) {
2650 sayYES; /* All done. */
2658 if (n >= cc->max) { /* Maximum greed exceeded? */
2659 if (ckWARN(WARN_UNSAFE) && n >= REG_INFTY
2660 && !(PL_reg_flags & RF_warned)) {
2661 PL_reg_flags |= RF_warned;
2662 Perl_warner(aTHX_ WARN_UNSAFE, "%s limit (%d) exceeded",
2663 "Complex regular subexpression recursion",
2670 PerlIO_printf(Perl_debug_log,
2671 "%*s trying longer...\n",
2672 REPORT_CODE_OFF+PL_regindent*2, "")
2674 /* Try scanning more and see if it helps. */
2675 PL_reginput = locinput;
2677 cc->lastloc = locinput;
2678 cp = regcppush(cc->parenfloor);
2680 if (regmatch(cc->scan)) {
2687 cc->lastloc = lastloc;
2691 /* Prefer scan over next for maximal matching. */
2693 if (n < cc->max) { /* More greed allowed? */
2694 cp = regcppush(cc->parenfloor);
2696 cc->lastloc = locinput;
2698 if (regmatch(cc->scan)) {
2703 regcppop(); /* Restore some previous $<digit>s? */
2704 PL_reginput = locinput;
2706 PerlIO_printf(Perl_debug_log,
2707 "%*s failed, try continuation...\n",
2708 REPORT_CODE_OFF+PL_regindent*2, "")
2711 if (ckWARN(WARN_UNSAFE) && n >= REG_INFTY
2712 && !(PL_reg_flags & RF_warned)) {
2713 PL_reg_flags |= RF_warned;
2714 Perl_warner(aTHX_ WARN_UNSAFE, "%s limit (%d) exceeded",
2715 "Complex regular subexpression recursion",
2719 /* Failed deeper matches of scan, so see if this one works. */
2720 PL_regcc = cc->oldcc;
2723 if (regmatch(cc->next))
2729 cc->lastloc = lastloc;
2734 next = scan + ARG(scan);
2737 inner = NEXTOPER(NEXTOPER(scan));
2740 inner = NEXTOPER(scan);
2745 if (OP(next) != c1) /* No choice. */
2746 next = inner; /* Avoid recursion. */
2748 int lastparen = *PL_reglastparen;
2752 PL_reginput = locinput;
2753 if (regmatch(inner))
2756 for (n = *PL_reglastparen; n > lastparen; n--)
2758 *PL_reglastparen = n;
2761 if (n = (c1 == BRANCH ? NEXT_OFF(next) : ARG(next)))
2765 inner = NEXTOPER(scan);
2766 if (c1 == BRANCHJ) {
2767 inner = NEXTOPER(inner);
2769 } while (scan != NULL && OP(scan) == c1);
2783 /* We suppose that the next guy does not need
2784 backtracking: in particular, it is of constant length,
2785 and has no parenths to influence future backrefs. */
2786 ln = ARG1(scan); /* min to match */
2787 n = ARG2(scan); /* max to match */
2788 paren = scan->flags;
2790 if (paren > PL_regsize)
2792 if (paren > *PL_reglastparen)
2793 *PL_reglastparen = paren;
2795 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
2797 scan += NEXT_OFF(scan); /* Skip former OPEN. */
2798 PL_reginput = locinput;
2801 if (ln && regrepeat_hard(scan, ln, &l) < ln)
2803 if (ln && l == 0 && n >= ln
2804 /* In fact, this is tricky. If paren, then the
2805 fact that we did/didnot match may influence
2806 future execution. */
2807 && !(paren && ln == 0))
2809 locinput = PL_reginput;
2810 if (PL_regkind[(U8)OP(next)] == EXACT) {
2811 c1 = (U8)*STRING(next);
2812 if (OP(next) == EXACTF)
2814 else if (OP(next) == EXACTFL)
2815 c2 = PL_fold_locale[c1];
2822 /* This may be improved if l == 0. */
2823 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
2824 /* If it could work, try it. */
2826 UCHARAT(PL_reginput) == c1 ||
2827 UCHARAT(PL_reginput) == c2)
2831 PL_regstartp[paren] =
2832 HOPc(PL_reginput, -l) - PL_bostr;
2833 PL_regendp[paren] = PL_reginput - PL_bostr;
2836 PL_regendp[paren] = -1;
2842 /* Couldn't or didn't -- move forward. */
2843 PL_reginput = locinput;
2844 if (regrepeat_hard(scan, 1, &l)) {
2846 locinput = PL_reginput;
2853 n = regrepeat_hard(scan, n, &l);
2854 if (n != 0 && l == 0
2855 /* In fact, this is tricky. If paren, then the
2856 fact that we did/didnot match may influence
2857 future execution. */
2858 && !(paren && ln == 0))
2860 locinput = PL_reginput;
2862 PerlIO_printf(Perl_debug_log,
2863 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
2864 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
2868 if (PL_regkind[(U8)OP(next)] == EXACT) {
2869 c1 = (U8)*STRING(next);
2870 if (OP(next) == EXACTF)
2872 else if (OP(next) == EXACTFL)
2873 c2 = PL_fold_locale[c1];
2882 /* If it could work, try it. */
2884 UCHARAT(PL_reginput) == c1 ||
2885 UCHARAT(PL_reginput) == c2)
2888 PerlIO_printf(Perl_debug_log,
2889 "%*s trying tail with n=%"IVdf"...\n",
2890 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
2894 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
2895 PL_regendp[paren] = PL_reginput - PL_bostr;
2898 PL_regendp[paren] = -1;
2904 /* Couldn't or didn't -- back up. */
2906 locinput = HOPc(locinput, -l);
2907 PL_reginput = locinput;
2914 paren = scan->flags; /* Which paren to set */
2915 if (paren > PL_regsize)
2917 if (paren > *PL_reglastparen)
2918 *PL_reglastparen = paren;
2919 ln = ARG1(scan); /* min to match */
2920 n = ARG2(scan); /* max to match */
2921 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
2925 ln = ARG1(scan); /* min to match */
2926 n = ARG2(scan); /* max to match */
2927 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
2932 scan = NEXTOPER(scan);
2938 scan = NEXTOPER(scan);
2942 * Lookahead to avoid useless match attempts
2943 * when we know what character comes next.
2945 if (PL_regkind[(U8)OP(next)] == EXACT) {
2946 c1 = (U8)*STRING(next);
2947 if (OP(next) == EXACTF)
2949 else if (OP(next) == EXACTFL)
2950 c2 = PL_fold_locale[c1];
2956 PL_reginput = locinput;
2960 if (ln && regrepeat(scan, ln) < ln)
2962 locinput = PL_reginput;
2965 char *e = locinput + n - ln; /* Should not check after this */
2966 char *old = locinput;
2968 if (e >= PL_regeol || (n == REG_INFTY))
2971 /* Find place 'next' could work */
2973 while (locinput <= e && *locinput != c1)
2976 while (locinput <= e
2983 /* PL_reginput == old now */
2984 if (locinput != old) {
2985 ln = 1; /* Did some */
2986 if (regrepeat(scan, locinput - old) <
2990 /* PL_reginput == locinput now */
2993 PL_regstartp[paren] = HOPc(locinput, -1) - PL_bostr;
2994 PL_regendp[paren] = locinput - PL_bostr;
2997 PL_regendp[paren] = -1;
3001 PL_reginput = locinput; /* Could be reset... */
3003 /* Couldn't or didn't -- move forward. */
3008 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3009 /* If it could work, try it. */
3011 UCHARAT(PL_reginput) == c1 ||
3012 UCHARAT(PL_reginput) == c2)
3016 PL_regstartp[paren] = HOPc(PL_reginput, -1) - PL_bostr;
3017 PL_regendp[paren] = PL_reginput - PL_bostr;
3020 PL_regendp[paren] = -1;
3026 /* Couldn't or didn't -- move forward. */
3027 PL_reginput = locinput;
3028 if (regrepeat(scan, 1)) {
3030 locinput = PL_reginput;
3038 n = regrepeat(scan, n);
3039 locinput = PL_reginput;
3040 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3041 (!PL_multiline || OP(next) == SEOL))
3042 ln = n; /* why back off? */
3046 /* If it could work, try it. */
3048 UCHARAT(PL_reginput) == c1 ||
3049 UCHARAT(PL_reginput) == c2)
3053 PL_regstartp[paren] = HOPc(PL_reginput, -1) - PL_bostr;
3054 PL_regendp[paren] = PL_reginput - PL_bostr;
3057 PL_regendp[paren] = -1;
3063 /* Couldn't or didn't -- back up. */
3065 PL_reginput = locinput = HOPc(locinput, -1);
3070 /* If it could work, try it. */
3072 UCHARAT(PL_reginput) == c1 ||
3073 UCHARAT(PL_reginput) == c2)
3079 /* Couldn't or didn't -- back up. */
3081 PL_reginput = locinput = HOPc(locinput, -1);
3088 if (PL_reg_call_cc) {
3089 re_cc_state *cur_call_cc = PL_reg_call_cc;
3090 CURCUR *cctmp = PL_regcc;
3091 regexp *re = PL_reg_re;
3092 CHECKPOINT cp, lastcp;
3094 cp = regcppush(0); /* Save *all* the positions. */
3096 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3098 PL_reginput = locinput; /* Make position available to
3100 cache_re(PL_reg_call_cc->re);
3101 PL_regcc = PL_reg_call_cc->cc;
3102 PL_reg_call_cc = PL_reg_call_cc->prev;
3103 if (regmatch(cur_call_cc->node)) {
3104 PL_reg_call_cc = cur_call_cc;
3110 PL_reg_call_cc = cur_call_cc;
3116 PerlIO_printf(Perl_debug_log,
3117 "%*s continuation failed...\n",
3118 REPORT_CODE_OFF+PL_regindent*2, "")
3122 if (locinput < PL_regtill) {
3123 DEBUG_r(PerlIO_printf(Perl_debug_log,
3124 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3126 (long)(locinput - PL_reg_starttry),
3127 (long)(PL_regtill - PL_reg_starttry),
3129 sayNO_FINAL; /* Cannot match: too short. */
3131 PL_reginput = locinput; /* put where regtry can find it */
3132 sayYES_FINAL; /* Success! */
3134 PL_reginput = locinput; /* put where regtry can find it */
3135 sayYES_LOUD; /* Success! */
3138 PL_reginput = locinput;
3143 if (UTF) { /* XXXX This is absolutely
3144 broken, we read before
3146 s = HOPMAYBEc(locinput, -scan->flags);
3152 if (locinput < PL_bostr + scan->flags)
3154 PL_reginput = locinput - scan->flags;
3159 PL_reginput = locinput;
3164 if (UTF) { /* XXXX This is absolutely
3165 broken, we read before
3167 s = HOPMAYBEc(locinput, -scan->flags);
3168 if (!s || s < PL_bostr)
3173 if (locinput < PL_bostr + scan->flags)
3175 PL_reginput = locinput - scan->flags;
3180 PL_reginput = locinput;
3183 inner = NEXTOPER(NEXTOPER(scan));
3184 if (regmatch(inner) != n) {
3199 if (OP(scan) == SUSPEND) {
3200 locinput = PL_reginput;
3201 nextchr = UCHARAT(locinput);
3206 next = scan + ARG(scan);
3211 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3212 PTR2UV(scan), OP(scan));
3213 Perl_croak(aTHX_ "regexp memory corruption");
3219 * We get here only if there's trouble -- normally "case END" is
3220 * the terminating point.
3222 Perl_croak(aTHX_ "corrupted regexp pointers");
3228 PerlIO_printf(Perl_debug_log,
3229 "%*s %scould match...%s\n",
3230 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3234 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3235 PL_colors[4],PL_colors[5]));
3244 PerlIO_printf(Perl_debug_log,
3245 "%*s %sfailed...%s\n",
3246 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3258 - regrepeat - repeatedly match something simple, report how many
3261 * [This routine now assumes that it will only match on things of length 1.
3262 * That was true before, but now we assume scan - reginput is the count,
3263 * rather than incrementing count on every character. [Er, except utf8.]]
3266 S_regrepeat(pTHX_ regnode *p, I32 max)
3269 register char *scan;
3271 register char *loceol = PL_regeol;
3272 register I32 hardcount = 0;
3275 if (max != REG_INFTY && max < loceol - scan)
3276 loceol = scan + max;
3279 while (scan < loceol && *scan != '\n')
3287 while (scan < loceol && *scan != '\n') {
3288 scan += UTF8SKIP(scan);
3294 while (scan < loceol) {
3295 scan += UTF8SKIP(scan);
3299 case EXACT: /* length of string is 1 */
3301 while (scan < loceol && UCHARAT(scan) == c)
3304 case EXACTF: /* length of string is 1 */
3306 while (scan < loceol &&
3307 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3310 case EXACTFL: /* length of string is 1 */
3311 PL_reg_flags |= RF_tainted;
3313 while (scan < loceol &&
3314 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3319 while (scan < loceol && REGINCLASSUTF8(p, (U8*)scan)) {
3320 scan += UTF8SKIP(scan);
3325 while (scan < loceol && REGINCLASS(p, *scan))
3329 while (scan < loceol && isALNUM(*scan))
3334 while (scan < loceol && swash_fetch(PL_utf8_alnum, (U8*)scan)) {
3335 scan += UTF8SKIP(scan);
3340 PL_reg_flags |= RF_tainted;
3341 while (scan < loceol && isALNUM_LC(*scan))
3345 PL_reg_flags |= RF_tainted;
3347 while (scan < loceol && isALNUM_LC_utf8((U8*)scan)) {
3348 scan += UTF8SKIP(scan);
3354 while (scan < loceol && !isALNUM(*scan))
3359 while (scan < loceol && !swash_fetch(PL_utf8_alnum, (U8*)scan)) {
3360 scan += UTF8SKIP(scan);
3365 PL_reg_flags |= RF_tainted;
3366 while (scan < loceol && !isALNUM_LC(*scan))
3370 PL_reg_flags |= RF_tainted;
3372 while (scan < loceol && !isALNUM_LC_utf8((U8*)scan)) {
3373 scan += UTF8SKIP(scan);
3378 while (scan < loceol && isSPACE(*scan))
3383 while (scan < loceol && (*scan == ' ' || swash_fetch(PL_utf8_space,(U8*)scan))) {
3384 scan += UTF8SKIP(scan);
3389 PL_reg_flags |= RF_tainted;
3390 while (scan < loceol && isSPACE_LC(*scan))
3394 PL_reg_flags |= RF_tainted;
3396 while (scan < loceol && (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3397 scan += UTF8SKIP(scan);
3402 while (scan < loceol && !isSPACE(*scan))
3407 while (scan < loceol && !(*scan == ' ' || swash_fetch(PL_utf8_space,(U8*)scan))) {
3408 scan += UTF8SKIP(scan);
3413 PL_reg_flags |= RF_tainted;
3414 while (scan < loceol && !isSPACE_LC(*scan))
3418 PL_reg_flags |= RF_tainted;
3420 while (scan < loceol && !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3421 scan += UTF8SKIP(scan);
3426 while (scan < loceol && isDIGIT(*scan))
3431 while (scan < loceol && swash_fetch(PL_utf8_digit,(U8*)scan)) {
3432 scan += UTF8SKIP(scan);
3438 while (scan < loceol && !isDIGIT(*scan))
3443 while (scan < loceol && !swash_fetch(PL_utf8_digit,(U8*)scan)) {
3444 scan += UTF8SKIP(scan);
3448 default: /* Called on something of 0 width. */
3449 break; /* So match right here or not at all. */
3455 c = scan - PL_reginput;
3460 SV *prop = sv_newmortal();
3463 PerlIO_printf(Perl_debug_log,
3464 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
3465 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
3472 - regrepeat_hard - repeatedly match something, report total lenth and length
3474 * The repeater is supposed to have constant length.
3478 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
3481 register char *scan;
3482 register char *start;
3483 register char *loceol = PL_regeol;
3485 I32 count = 0, res = 1;
3490 start = PL_reginput;
3492 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3495 while (start < PL_reginput) {
3497 start += UTF8SKIP(start);
3508 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3510 *lp = l = PL_reginput - start;
3511 if (max != REG_INFTY && l*max < loceol - scan)
3512 loceol = scan + l*max;
3525 - reginclass - determine if a character falls into a character class
3529 S_reginclass(pTHX_ register regnode *p, register I32 c)
3532 char flags = ANYOF_FLAGS(p);
3536 if (ANYOF_BITMAP_TEST(p, c))
3538 else if (flags & ANYOF_FOLD) {
3540 if (flags & ANYOF_LOCALE) {
3541 PL_reg_flags |= RF_tainted;
3542 cf = PL_fold_locale[c];
3546 if (ANYOF_BITMAP_TEST(p, cf))
3550 if (!match && (flags & ANYOF_CLASS)) {
3551 PL_reg_flags |= RF_tainted;
3553 (ANYOF_CLASS_TEST(p, ANYOF_ALNUM) && isALNUM_LC(c)) ||
3554 (ANYOF_CLASS_TEST(p, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
3555 (ANYOF_CLASS_TEST(p, ANYOF_SPACE) && isSPACE_LC(c)) ||
3556 (ANYOF_CLASS_TEST(p, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
3557 (ANYOF_CLASS_TEST(p, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
3558 (ANYOF_CLASS_TEST(p, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
3559 (ANYOF_CLASS_TEST(p, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
3560 (ANYOF_CLASS_TEST(p, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
3561 (ANYOF_CLASS_TEST(p, ANYOF_ALPHA) && isALPHA_LC(c)) ||
3562 (ANYOF_CLASS_TEST(p, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
3563 (ANYOF_CLASS_TEST(p, ANYOF_ASCII) && isASCII(c)) ||
3564 (ANYOF_CLASS_TEST(p, ANYOF_NASCII) && !isASCII(c)) ||
3565 (ANYOF_CLASS_TEST(p, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
3566 (ANYOF_CLASS_TEST(p, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
3567 (ANYOF_CLASS_TEST(p, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
3568 (ANYOF_CLASS_TEST(p, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
3569 (ANYOF_CLASS_TEST(p, ANYOF_LOWER) && isLOWER_LC(c)) ||
3570 (ANYOF_CLASS_TEST(p, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
3571 (ANYOF_CLASS_TEST(p, ANYOF_PRINT) && isPRINT_LC(c)) ||
3572 (ANYOF_CLASS_TEST(p, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
3573 (ANYOF_CLASS_TEST(p, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
3574 (ANYOF_CLASS_TEST(p, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
3575 (ANYOF_CLASS_TEST(p, ANYOF_UPPER) && isUPPER_LC(c)) ||
3576 (ANYOF_CLASS_TEST(p, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
3577 (ANYOF_CLASS_TEST(p, ANYOF_XDIGIT) && isXDIGIT(c)) ||
3578 (ANYOF_CLASS_TEST(p, ANYOF_NXDIGIT) && !isXDIGIT(c))
3579 ) /* How's that for a conditional? */
3585 return (flags & ANYOF_INVERT) ? !match : match;
3589 S_reginclassutf8(pTHX_ regnode *f, U8 *p)
3592 char flags = ARG1(f);
3594 SV *sv = (SV*)PL_regdata->data[ARG2(f)];
3596 if (swash_fetch(sv, p))
3598 else if (flags & ANYOF_FOLD) {
3601 if (flags & ANYOF_LOCALE) {
3602 PL_reg_flags |= RF_tainted;
3603 uv_to_utf8(tmpbuf, toLOWER_LC_utf8(p));
3606 uv_to_utf8(tmpbuf, toLOWER_utf8(p));
3607 if (swash_fetch(sv, tmpbuf))
3611 /* UTF8 combined with ANYOF_CLASS is ill-defined. */
3613 return (flags & ANYOF_INVERT) ? !match : match;
3617 S_reghop(pTHX_ U8 *s, I32 off)
3621 while (off-- && s < (U8*)PL_regeol)
3626 if (s > (U8*)PL_bostr) {
3629 while (s > (U8*)PL_bostr && (*s & 0xc0) == 0x80)
3631 } /* XXX could check well-formedness here */
3639 S_reghopmaybe(pTHX_ U8* s, I32 off)
3643 while (off-- && s < (U8*)PL_regeol)
3650 if (s > (U8*)PL_bostr) {
3653 while (s > (U8*)PL_bostr && (*s & 0xc0) == 0x80)
3655 } /* XXX could check well-formedness here */
3671 restore_pos(pTHXo_ void *arg)
3674 if (PL_reg_eval_set) {
3675 if (PL_reg_oldsaved) {
3676 PL_reg_re->subbeg = PL_reg_oldsaved;
3677 PL_reg_re->sublen = PL_reg_oldsavedlen;
3678 RX_MATCH_COPIED_on(PL_reg_re);
3680 PL_reg_magic->mg_len = PL_reg_oldpos;
3681 PL_reg_eval_set = 0;
3682 PL_curpm = PL_reg_oldcurpm;