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) 1991-2000, 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(cp) DEBUG_r(PerlIO_printf(Perl_debug_log, \
149 " Setting an EVAL scope, savestack=%"IVdf"\n", \
150 (IV)PL_savestack_ix)); cp = PL_savestack_ix
152 # define REGCP_UNWIND(cp) DEBUG_r(cp != PL_savestack_ix ? \
153 PerlIO_printf(Perl_debug_log, \
154 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
155 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
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) /* Ignores regcppush()ed data. */
224 #define TRYPAREN(paren, n, input) { \
227 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
228 PL_regendp[paren] = input - PL_bostr; \
231 PL_regendp[paren] = -1; \
233 if (regmatch(next)) \
236 PL_regendp[paren] = -1; \
241 * pregexec and friends
245 - pregexec - match a regexp against a string
248 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
249 char *strbeg, I32 minend, SV *screamer, U32 nosave)
250 /* strend: pointer to null at end of string */
251 /* strbeg: real beginning of string */
252 /* minend: end of match must be >=minend after stringarg. */
253 /* nosave: For optimizations. */
256 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
257 nosave ? 0 : REXEC_COPY_STR);
261 S_cache_re(pTHX_ regexp *prog)
264 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
266 PL_regprogram = prog->program;
268 PL_regnpar = prog->nparens;
269 PL_regdata = prog->data;
274 * Need to implement the following flags for reg_anch:
276 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
278 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
279 * INTUIT_AUTORITATIVE_ML
280 * INTUIT_ONCE_NOML - Intuit can match in one location only.
283 * Another flag for this function: SECOND_TIME (so that float substrs
284 * with giant delta may be not rechecked).
287 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
289 /* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
290 Otherwise, only SvCUR(sv) is used to get strbeg. */
292 /* XXXX We assume that strpos is strbeg unless sv. */
294 /* XXXX Some places assume that there is a fixed substring.
295 An update may be needed if optimizer marks as "INTUITable"
296 RExen without fixed substrings. Similarly, it is assumed that
297 lengths of all the strings are no more than minlen, thus they
298 cannot come from lookahead.
299 (Or minlen should take into account lookahead.) */
301 /* A failure to find a constant substring means that there is no need to make
302 an expensive call to REx engine, thus we celebrate a failure. Similarly,
303 finding a substring too deep into the string means that less calls to
304 regtry() should be needed.
306 REx compiler's optimizer found 4 possible hints:
307 a) Anchored substring;
309 c) Whether we are anchored (beginning-of-line or \G);
310 d) First node (of those at offset 0) which may distingush positions;
311 We use a)b)d) and multiline-part of c), and try to find a position in the
312 string which does not contradict any of them.
315 /* Most of decisions we do here should have been done at compile time.
316 The nodes of the REx which we used for the search should have been
317 deleted from the finite automaton. */
320 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
321 char *strend, U32 flags, re_scream_pos_data *data)
323 register I32 start_shift;
324 /* Should be nonnegative! */
325 register I32 end_shift;
332 register char *other_last = Nullch; /* other substr checked before this */
333 char *check_at; /* check substr found at this pos */
335 char *i_strpos = strpos;
338 DEBUG_r( if (!PL_colorset) reginitcolors() );
339 DEBUG_r(PerlIO_printf(Perl_debug_log,
340 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
341 PL_colors[4],PL_colors[5],PL_colors[0],
344 (strlen(prog->precomp) > 60 ? "..." : ""),
346 (int)(strend - strpos > 60 ? 60 : strend - strpos),
347 strpos, PL_colors[1],
348 (strend - strpos > 60 ? "..." : ""))
351 if (prog->minlen > strend - strpos) {
352 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too short...\n"));
355 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
356 check = prog->check_substr;
357 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
358 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
359 || ( (prog->reganch & ROPT_ANCH_BOL)
360 && !PL_multiline ) ); /* Check after \n? */
363 if ( !(prog->reganch & ROPT_ANCH_GPOS) /* Checked by the caller */
364 /* SvCUR is not set on references: SvRV and SvPVX overlap */
366 && (strpos != strbeg)) {
367 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
370 if (prog->check_offset_min == prog->check_offset_max) {
371 /* Substring at constant offset from beg-of-str... */
374 PL_regeol = strend; /* Used in HOP() */
375 s = HOPc(strpos, prog->check_offset_min);
377 slen = SvCUR(check); /* >= 1 */
379 if ( strend - s > slen || strend - s < slen - 1
380 || (strend - s == slen && strend[-1] != '\n')) {
381 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
384 /* Now should match s[0..slen-2] */
386 if (slen && (*SvPVX(check) != *s
388 && memNE(SvPVX(check), s, slen)))) {
390 DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
394 else if (*SvPVX(check) != *s
395 || ((slen = SvCUR(check)) > 1
396 && memNE(SvPVX(check), s, slen)))
398 goto success_at_start;
401 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
403 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
404 end_shift = prog->minlen - start_shift -
405 CHR_SVLEN(check) + (SvTAIL(check) != 0);
407 I32 end = prog->check_offset_max + CHR_SVLEN(check)
408 - (SvTAIL(check) != 0);
409 I32 eshift = strend - s - end;
411 if (end_shift < eshift)
415 else { /* Can match at random position */
418 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
419 /* Should be nonnegative! */
420 end_shift = prog->minlen - start_shift -
421 CHR_SVLEN(check) + (SvTAIL(check) != 0);
424 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
426 Perl_croak(aTHX_ "panic: end_shift");
430 /* Find a possible match in the region s..strend by looking for
431 the "check" substring in the region corrected by start/end_shift. */
432 if (flags & REXEC_SCREAM) {
433 I32 p = -1; /* Internal iterator of scream. */
434 I32 *pp = data ? data->scream_pos : &p;
436 if (PL_screamfirst[BmRARE(check)] >= 0
437 || ( BmRARE(check) == '\n'
438 && (BmPREVIOUS(check) == SvCUR(check) - 1)
440 s = screaminstr(sv, check,
441 start_shift + (s - strbeg), end_shift, pp, 0);
445 *data->scream_olds = s;
448 s = fbm_instr((unsigned char*)s + start_shift,
449 (unsigned char*)strend - end_shift,
450 check, PL_multiline ? FBMrf_MULTILINE : 0);
452 /* Update the count-of-usability, remove useless subpatterns,
455 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
456 (s ? "Found" : "Did not find"),
457 ((check == prog->anchored_substr) ? "anchored" : "floating"),
459 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
461 PL_colors[1], (SvTAIL(check) ? "$" : ""),
462 (s ? " at offset " : "...\n") ) );
469 /* Finish the diagnostic message */
470 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
472 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
473 Start with the other substr.
474 XXXX no SCREAM optimization yet - and a very coarse implementation
475 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
476 *always* match. Probably should be marked during compile...
477 Probably it is right to do no SCREAM here...
480 if (prog->float_substr && prog->anchored_substr) {
481 /* Take into account the "other" substring. */
482 /* XXXX May be hopelessly wrong for UTF... */
485 if (check == prog->float_substr) {
488 char *last = s - start_shift, *last1, *last2;
492 t = s - prog->check_offset_max;
493 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
494 && (!(prog->reganch & ROPT_UTF8)
495 || (PL_bostr = strpos, /* Used in regcopmaybe() */
496 (t = reghopmaybe_c(s, -(prog->check_offset_max)))
501 t += prog->anchored_offset;
502 if (t < other_last) /* These positions already checked */
505 last2 = last1 = strend - prog->minlen;
508 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
509 /* On end-of-str: see comment below. */
510 s = fbm_instr((unsigned char*)t,
511 (unsigned char*)last1 + prog->anchored_offset
512 + SvCUR(prog->anchored_substr)
513 - (SvTAIL(prog->anchored_substr)!=0),
514 prog->anchored_substr, PL_multiline ? FBMrf_MULTILINE : 0);
515 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s anchored substr `%s%.*s%s'%s",
516 (s ? "Found" : "Contradicts"),
518 (int)(SvCUR(prog->anchored_substr)
519 - (SvTAIL(prog->anchored_substr)!=0)),
520 SvPVX(prog->anchored_substr),
521 PL_colors[1], (SvTAIL(prog->anchored_substr) ? "$" : "")));
523 if (last1 >= last2) {
524 DEBUG_r(PerlIO_printf(Perl_debug_log,
525 ", giving up...\n"));
528 DEBUG_r(PerlIO_printf(Perl_debug_log,
529 ", trying floating at offset %ld...\n",
530 (long)(s1 + 1 - i_strpos)));
531 PL_regeol = strend; /* Used in HOP() */
532 other_last = last1 + prog->anchored_offset + 1;
537 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
538 (long)(s - i_strpos)));
539 t = s - prog->anchored_offset;
548 else { /* Take into account the floating substring. */
553 last1 = last = strend - prog->minlen + prog->float_min_offset;
554 if (last - t > prog->float_max_offset)
555 last = t + prog->float_max_offset;
556 s = t + prog->float_min_offset;
559 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
560 /* fbm_instr() takes into account exact value of end-of-str
561 if the check is SvTAIL(ed). Since false positives are OK,
562 and end-of-str is not later than strend we are OK. */
563 s = fbm_instr((unsigned char*)s,
564 (unsigned char*)last + SvCUR(prog->float_substr)
565 - (SvTAIL(prog->float_substr)!=0),
566 prog->float_substr, PL_multiline ? FBMrf_MULTILINE : 0);
567 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
568 (s ? "Found" : "Contradicts"),
570 (int)(SvCUR(prog->float_substr)
571 - (SvTAIL(prog->float_substr)!=0)),
572 SvPVX(prog->float_substr),
573 PL_colors[1], (SvTAIL(prog->float_substr) ? "$" : "")));
576 DEBUG_r(PerlIO_printf(Perl_debug_log,
577 ", giving up...\n"));
580 DEBUG_r(PerlIO_printf(Perl_debug_log,
581 ", trying anchored starting at offset %ld...\n",
582 (long)(s1 + 1 - i_strpos)));
583 other_last = last + 1;
584 PL_regeol = strend; /* Used in HOP() */
589 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
590 (long)(s - i_strpos)));
600 t = s - prog->check_offset_max;
602 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
603 && (!(prog->reganch & ROPT_UTF8)
604 || (PL_bostr = strpos, /* Used in regcopmaybe() */
605 ((t = reghopmaybe_c(s, -(prog->check_offset_max)))
608 /* Fixed substring is found far enough so that the match
609 cannot start at strpos. */
611 if (ml_anch && t[-1] != '\n') {
612 /* Eventually fbm_*() should handle this, but often
613 anchored_offset is not 0, so this check will not be wasted. */
614 /* XXXX In the code below we prefer to look for "^" even in
615 presence of anchored substrings. And we search even
616 beyond the found float position. These pessimizations
617 are historical artefacts only. */
619 while (t < strend - prog->minlen) {
621 if (t < check_at - prog->check_offset_min) {
622 if (prog->anchored_substr) {
623 /* Since we moved from the found position,
624 we definitely contradict the found anchored
625 substr. Due to the above check we do not
626 contradict "check" substr.
627 Thus we can arrive here only if check substr
628 is float. Redo checking for "other"=="fixed".
631 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
632 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
633 goto do_other_anchored;
635 /* We don't contradict the found floating substring. */
636 /* XXXX Why not check for STCLASS? */
638 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
639 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
642 /* Position contradicts check-string */
643 /* XXXX probably better to look for check-string
644 than for "\n", so one should lower the limit for t? */
645 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
646 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
647 other_last = strpos = s = t + 1;
652 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
653 PL_colors[0],PL_colors[1]));
657 DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
658 PL_colors[0],PL_colors[1]));
662 ++BmUSEFUL(prog->check_substr); /* hooray/5 */
666 /* The found string does not prohibit matching at strpos,
667 - no optimization of calling REx engine can be performed,
668 unless it was an MBOL and we are not after MBOL,
669 or a future STCLASS check will fail this. */
671 /* Even in this situation we may use MBOL flag if strpos is offset
672 wrt the start of the string. */
673 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
674 && (strpos != strbeg) && strpos[-1] != '\n'
675 /* May be due to an implicit anchor of m{.*foo} */
676 && !(prog->reganch & ROPT_IMPLICIT))
681 DEBUG_r( if (ml_anch)
682 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
683 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
686 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
687 && prog->check_substr /* Could be deleted already */
688 && --BmUSEFUL(prog->check_substr) < 0
689 && prog->check_substr == prog->float_substr)
691 /* If flags & SOMETHING - do not do it many times on the same match */
692 DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
693 SvREFCNT_dec(prog->check_substr);
694 prog->check_substr = Nullsv; /* disable */
695 prog->float_substr = Nullsv; /* clear */
696 check = Nullsv; /* abort */
698 /* XXXX This is a remnant of the old implementation. It
699 looks wasteful, since now INTUIT can use many
701 prog->reganch &= ~RE_USE_INTUIT;
708 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
709 if (prog->regstclass) {
710 /* minlen == 0 is possible if regstclass is \b or \B,
711 and the fixed substr is ''$.
712 Since minlen is already taken into account, s+1 is before strend;
713 accidentally, minlen >= 1 guaranties no false positives at s + 1
714 even for \b or \B. But (minlen? 1 : 0) below assumes that
715 regstclass does not come from lookahead... */
716 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
717 This leaves EXACTF only, which is dealt with in find_byclass(). */
718 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
719 ? STR_LEN(prog->regstclass)
721 char *endpos = (prog->anchored_substr || ml_anch)
722 ? s + (prog->minlen? cl_l : 0)
723 : (prog->float_substr ? check_at - start_shift + cl_l
725 char *startpos = strbeg;
728 if (prog->reganch & ROPT_UTF8) {
729 PL_regdata = prog->data; /* Used by REGINCLASS UTF logic */
732 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
737 if (endpos == strend) {
738 DEBUG_r( PerlIO_printf(Perl_debug_log,
739 "Could not match STCLASS...\n") );
742 DEBUG_r( PerlIO_printf(Perl_debug_log,
743 "This position contradicts STCLASS...\n") );
744 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
746 /* Contradict one of substrings */
747 if (prog->anchored_substr) {
748 if (prog->anchored_substr == check) {
749 DEBUG_r( what = "anchored" );
751 PL_regeol = strend; /* Used in HOP() */
753 if (s + start_shift + end_shift > strend) {
754 /* XXXX Should be taken into account earlier? */
755 DEBUG_r( PerlIO_printf(Perl_debug_log,
756 "Could not match STCLASS...\n") );
761 DEBUG_r( PerlIO_printf(Perl_debug_log,
762 "Looking for %s substr starting at offset %ld...\n",
763 what, (long)(s + start_shift - i_strpos)) );
766 /* Have both, check_string is floating */
767 if (t + start_shift >= check_at) /* Contradicts floating=check */
768 goto retry_floating_check;
769 /* Recheck anchored substring, but not floating... */
773 DEBUG_r( PerlIO_printf(Perl_debug_log,
774 "Looking for anchored substr starting at offset %ld...\n",
775 (long)(other_last - i_strpos)) );
776 goto do_other_anchored;
778 /* Another way we could have checked stclass at the
779 current position only: */
784 DEBUG_r( PerlIO_printf(Perl_debug_log,
785 "Looking for /%s^%s/m starting at offset %ld...\n",
786 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
789 if (!prog->float_substr) /* Could have been deleted */
791 /* Check is floating subtring. */
792 retry_floating_check:
793 t = check_at - start_shift;
794 DEBUG_r( what = "floating" );
795 goto hop_and_restart;
798 PerlIO_printf(Perl_debug_log,
799 "By STCLASS: moving %ld --> %ld\n",
800 (long)(t - i_strpos), (long)(s - i_strpos));
802 PerlIO_printf(Perl_debug_log,
803 "Does not contradict STCLASS...\n") );
806 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
807 PL_colors[4], (check ? "Guessed" : "Giving up"),
808 PL_colors[5], (long)(s - i_strpos)) );
811 fail_finish: /* Substring not found */
812 if (prog->check_substr) /* could be removed already */
813 BmUSEFUL(prog->check_substr) += 5; /* hooray */
815 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
816 PL_colors[4],PL_colors[5]));
820 /* We know what class REx starts with. Try to find this position... */
822 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
824 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
830 register I32 tmp = 1; /* Scratch variable? */
832 /* We know what class it must start with. */
836 if (REGINCLASSUTF8(c, (U8*)s)) {
837 if (tmp && (norun || regtry(prog, s)))
849 if (REGINCLASS(c, *(U8*)s)) {
850 if (tmp && (norun || regtry(prog, s)))
870 c2 = PL_fold_locale[c1];
875 e = s; /* Due to minlen logic of intuit() */
876 /* Here it is NOT UTF! */
880 && (ln == 1 || !(OP(c) == EXACTF
882 : ibcmp_locale(s, m, ln)))
883 && (norun || regtry(prog, s)) )
889 if ( (*(U8*)s == c1 || *(U8*)s == c2)
890 && (ln == 1 || !(OP(c) == EXACTF
892 : ibcmp_locale(s, m, ln)))
893 && (norun || regtry(prog, s)) )
900 PL_reg_flags |= RF_tainted;
903 tmp = (s != startpos) ? UCHARAT(s - 1) : '\n';
904 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
906 if (tmp == !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
908 if ((norun || regtry(prog, s)))
913 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
917 PL_reg_flags |= RF_tainted;
923 U8 *r = reghop((U8*)s, -1);
925 tmp = (I32)utf8_to_uv(r, s - (char*)r, 0, 0);
927 tmp = ((OP(c) == BOUNDUTF8 ?
928 isALNUM_uni(tmp) : isALNUM_LC_uni(tmp)) != 0);
930 if (tmp == !(OP(c) == BOUNDUTF8 ?
931 swash_fetch(PL_utf8_alnum, (U8*)s) :
932 isALNUM_LC_utf8((U8*)s)))
935 if ((norun || regtry(prog, s)))
940 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
944 PL_reg_flags |= RF_tainted;
947 tmp = (s != startpos) ? UCHARAT(s - 1) : '\n';
948 tmp = ((OP(c) == NBOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
950 if (tmp == !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
952 else if ((norun || regtry(prog, s)))
956 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
960 PL_reg_flags |= RF_tainted;
966 U8 *r = reghop((U8*)s, -1);
968 tmp = (I32)utf8_to_uv(r, s - (char*)r, 0, 0);
970 tmp = ((OP(c) == NBOUNDUTF8 ?
971 isALNUM_uni(tmp) : isALNUM_LC_uni(tmp)) != 0);
973 if (tmp == !(OP(c) == NBOUNDUTF8 ?
974 swash_fetch(PL_utf8_alnum, (U8*)s) :
975 isALNUM_LC_utf8((U8*)s)))
977 else if ((norun || regtry(prog, s)))
981 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
987 if (tmp && (norun || regtry(prog, s)))
999 if (swash_fetch(PL_utf8_alnum, (U8*)s)) {
1000 if (tmp && (norun || regtry(prog, s)))
1011 PL_reg_flags |= RF_tainted;
1012 while (s < strend) {
1013 if (isALNUM_LC(*s)) {
1014 if (tmp && (norun || regtry(prog, s)))
1025 PL_reg_flags |= RF_tainted;
1026 while (s < strend) {
1027 if (isALNUM_LC_utf8((U8*)s)) {
1028 if (tmp && (norun || regtry(prog, s)))
1039 while (s < strend) {
1041 if (tmp && (norun || regtry(prog, s)))
1052 while (s < strend) {
1053 if (!swash_fetch(PL_utf8_alnum, (U8*)s)) {
1054 if (tmp && (norun || regtry(prog, s)))
1065 PL_reg_flags |= RF_tainted;
1066 while (s < strend) {
1067 if (!isALNUM_LC(*s)) {
1068 if (tmp && (norun || regtry(prog, s)))
1079 PL_reg_flags |= RF_tainted;
1080 while (s < strend) {
1081 if (!isALNUM_LC_utf8((U8*)s)) {
1082 if (tmp && (norun || regtry(prog, s)))
1093 while (s < strend) {
1095 if (tmp && (norun || regtry(prog, s)))
1106 while (s < strend) {
1107 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s)) {
1108 if (tmp && (norun || regtry(prog, s)))
1119 PL_reg_flags |= RF_tainted;
1120 while (s < strend) {
1121 if (isSPACE_LC(*s)) {
1122 if (tmp && (norun || regtry(prog, s)))
1133 PL_reg_flags |= RF_tainted;
1134 while (s < strend) {
1135 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1136 if (tmp && (norun || regtry(prog, s)))
1147 while (s < strend) {
1149 if (tmp && (norun || regtry(prog, s)))
1160 while (s < strend) {
1161 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s))) {
1162 if (tmp && (norun || regtry(prog, s)))
1173 PL_reg_flags |= RF_tainted;
1174 while (s < strend) {
1175 if (!isSPACE_LC(*s)) {
1176 if (tmp && (norun || regtry(prog, s)))
1187 PL_reg_flags |= RF_tainted;
1188 while (s < strend) {
1189 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1190 if (tmp && (norun || regtry(prog, s)))
1201 while (s < strend) {
1203 if (tmp && (norun || regtry(prog, s)))
1214 while (s < strend) {
1215 if (swash_fetch(PL_utf8_digit,(U8*)s)) {
1216 if (tmp && (norun || regtry(prog, s)))
1227 PL_reg_flags |= RF_tainted;
1228 while (s < strend) {
1229 if (isDIGIT_LC(*s)) {
1230 if (tmp && (norun || regtry(prog, s)))
1241 PL_reg_flags |= RF_tainted;
1242 while (s < strend) {
1243 if (isDIGIT_LC_utf8((U8*)s)) {
1244 if (tmp && (norun || regtry(prog, s)))
1255 while (s < strend) {
1257 if (tmp && (norun || regtry(prog, s)))
1268 while (s < strend) {
1269 if (!swash_fetch(PL_utf8_digit,(U8*)s)) {
1270 if (tmp && (norun || regtry(prog, s)))
1281 PL_reg_flags |= RF_tainted;
1282 while (s < strend) {
1283 if (!isDIGIT_LC(*s)) {
1284 if (tmp && (norun || regtry(prog, s)))
1295 PL_reg_flags |= RF_tainted;
1296 while (s < strend) {
1297 if (!isDIGIT_LC_utf8((U8*)s)) {
1298 if (tmp && (norun || regtry(prog, s)))
1309 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1318 - regexec_flags - match a regexp against a string
1321 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1322 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1323 /* strend: pointer to null at end of string */
1324 /* strbeg: real beginning of string */
1325 /* minend: end of match must be >=minend after stringarg. */
1326 /* data: May be used for some additional optimizations. */
1327 /* nosave: For optimizations. */
1331 register regnode *c;
1332 register char *startpos = stringarg;
1333 I32 minlen; /* must match at least this many chars */
1334 I32 dontbother = 0; /* how many characters not to try at end */
1335 /* I32 start_shift = 0; */ /* Offset of the start to find
1336 constant substr. */ /* CC */
1337 I32 end_shift = 0; /* Same for the end. */ /* CC */
1338 I32 scream_pos = -1; /* Internal iterator of scream. */
1340 SV* oreplsv = GvSV(PL_replgv);
1346 PL_regnarrate = PL_debug & 512;
1349 /* Be paranoid... */
1350 if (prog == NULL || startpos == NULL) {
1351 Perl_croak(aTHX_ "NULL regexp parameter");
1355 minlen = prog->minlen;
1356 if (strend - startpos < minlen) goto phooey;
1358 if (startpos == strbeg) /* is ^ valid at stringarg? */
1361 PL_regprev = (U32)stringarg[-1];
1362 if (!PL_multiline && PL_regprev == '\n')
1363 PL_regprev = '\0'; /* force ^ to NOT match */
1366 /* Check validity of program. */
1367 if (UCHARAT(prog->program) != REG_MAGIC) {
1368 Perl_croak(aTHX_ "corrupted regexp program");
1372 PL_reg_eval_set = 0;
1375 if (prog->reganch & ROPT_UTF8)
1376 PL_reg_flags |= RF_utf8;
1378 /* Mark beginning of line for ^ and lookbehind. */
1379 PL_regbol = startpos;
1383 /* Mark end of line for $ (and such) */
1386 /* see how far we have to get to not match where we matched before */
1387 PL_regtill = startpos+minend;
1389 /* We start without call_cc context. */
1392 /* If there is a "must appear" string, look for it. */
1395 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1398 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1399 PL_reg_ganch = startpos;
1400 else if (sv && SvTYPE(sv) >= SVt_PVMG
1402 && (mg = mg_find(sv, 'g')) && mg->mg_len >= 0) {
1403 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1404 if (prog->reganch & ROPT_ANCH_GPOS) {
1405 if (s > PL_reg_ganch)
1410 else /* pos() not defined */
1411 PL_reg_ganch = strbeg;
1414 if (!(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
1415 re_scream_pos_data d;
1417 d.scream_olds = &scream_olds;
1418 d.scream_pos = &scream_pos;
1419 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1421 goto phooey; /* not present */
1424 DEBUG_r( if (!PL_colorset) reginitcolors() );
1425 DEBUG_r(PerlIO_printf(Perl_debug_log,
1426 "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
1427 PL_colors[4],PL_colors[5],PL_colors[0],
1430 (strlen(prog->precomp) > 60 ? "..." : ""),
1432 (int)(strend - startpos > 60 ? 60 : strend - startpos),
1433 startpos, PL_colors[1],
1434 (strend - startpos > 60 ? "..." : ""))
1437 /* Simplest case: anchored match need be tried only once. */
1438 /* [unless only anchor is BOL and multiline is set] */
1439 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1440 if (s == startpos && regtry(prog, startpos))
1442 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1443 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1448 dontbother = minlen - 1;
1449 end = HOPc(strend, -dontbother) - 1;
1450 /* for multiline we only have to try after newlines */
1451 if (prog->check_substr) {
1455 if (regtry(prog, s))
1460 if (prog->reganch & RE_USE_INTUIT) {
1461 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1472 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1473 if (regtry(prog, s))
1480 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1481 if (regtry(prog, PL_reg_ganch))
1486 /* Messy cases: unanchored match. */
1487 if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
1488 /* we have /x+whatever/ */
1489 /* it must be a one character string (XXXX Except UTF?) */
1490 char ch = SvPVX(prog->anchored_substr)[0];
1496 while (s < strend) {
1498 DEBUG_r( did_match = 1 );
1499 if (regtry(prog, s)) goto got_it;
1501 while (s < strend && *s == ch)
1508 while (s < strend) {
1510 DEBUG_r( did_match = 1 );
1511 if (regtry(prog, s)) goto got_it;
1513 while (s < strend && *s == ch)
1519 DEBUG_r(did_match ||
1520 PerlIO_printf(Perl_debug_log,
1521 "Did not find anchored character...\n"));
1524 else if (prog->anchored_substr != Nullsv
1525 || (prog->float_substr != Nullsv
1526 && prog->float_max_offset < strend - s)) {
1527 SV *must = prog->anchored_substr
1528 ? prog->anchored_substr : prog->float_substr;
1530 prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset;
1532 prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
1533 char *last = HOPc(strend, /* Cannot start after this */
1534 -(I32)(CHR_SVLEN(must)
1535 - (SvTAIL(must) != 0) + back_min));
1536 char *last1; /* Last position checked before */
1542 last1 = HOPc(s, -1);
1544 last1 = s - 1; /* bogus */
1546 /* XXXX check_substr already used to find `s', can optimize if
1547 check_substr==must. */
1549 dontbother = end_shift;
1550 strend = HOPc(strend, -dontbother);
1551 while ( (s <= last) &&
1552 ((flags & REXEC_SCREAM)
1553 ? (s = screaminstr(sv, must, HOPc(s, back_min) - strbeg,
1554 end_shift, &scream_pos, 0))
1555 : (s = fbm_instr((unsigned char*)HOP(s, back_min),
1556 (unsigned char*)strend, must,
1557 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1558 DEBUG_r( did_match = 1 );
1559 if (HOPc(s, -back_max) > last1) {
1560 last1 = HOPc(s, -back_min);
1561 s = HOPc(s, -back_max);
1564 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1566 last1 = HOPc(s, -back_min);
1570 while (s <= last1) {
1571 if (regtry(prog, s))
1577 while (s <= last1) {
1578 if (regtry(prog, s))
1584 DEBUG_r(did_match ||
1585 PerlIO_printf(Perl_debug_log, "Did not find %s substr `%s%.*s%s'%s...\n",
1586 ((must == prog->anchored_substr)
1587 ? "anchored" : "floating"),
1589 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1591 PL_colors[1], (SvTAIL(must) ? "$" : "")));
1594 else if ((c = prog->regstclass)) {
1595 if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT)
1596 /* don't bother with what can't match */
1597 strend = HOPc(strend, -(minlen - 1));
1598 if (find_byclass(prog, c, s, strend, startpos, 0))
1600 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1604 if (prog->float_substr != Nullsv) { /* Trim the end. */
1607 if (flags & REXEC_SCREAM) {
1608 last = screaminstr(sv, prog->float_substr, s - strbeg,
1609 end_shift, &scream_pos, 1); /* last one */
1611 last = scream_olds; /* Only one occurence. */
1615 char *little = SvPV(prog->float_substr, len);
1617 if (SvTAIL(prog->float_substr)) {
1618 if (memEQ(strend - len + 1, little, len - 1))
1619 last = strend - len + 1;
1620 else if (!PL_multiline)
1621 last = memEQ(strend - len, little, len)
1622 ? strend - len : Nullch;
1628 last = rninstr(s, strend, little, little + len);
1630 last = strend; /* matching `$' */
1634 DEBUG_r(PerlIO_printf(Perl_debug_log,
1635 "%sCan't trim the tail, match fails (should not happen)%s\n",
1636 PL_colors[4],PL_colors[5]));
1637 goto phooey; /* Should not happen! */
1639 dontbother = strend - last + prog->float_min_offset;
1641 if (minlen && (dontbother < minlen))
1642 dontbother = minlen - 1;
1643 strend -= dontbother; /* this one's always in bytes! */
1644 /* We don't know much -- general case. */
1647 if (regtry(prog, s))
1656 if (regtry(prog, s))
1658 } while (s++ < strend);
1666 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1668 if (PL_reg_eval_set) {
1669 /* Preserve the current value of $^R */
1670 if (oreplsv != GvSV(PL_replgv))
1671 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1672 restored, the value remains
1674 restore_pos(aTHXo_ 0);
1677 /* make sure $`, $&, $', and $digit will work later */
1678 if ( !(flags & REXEC_NOT_FIRST) ) {
1679 if (RX_MATCH_COPIED(prog)) {
1680 Safefree(prog->subbeg);
1681 RX_MATCH_COPIED_off(prog);
1683 if (flags & REXEC_COPY_STR) {
1684 I32 i = PL_regeol - startpos + (stringarg - strbeg);
1686 s = savepvn(strbeg, i);
1689 RX_MATCH_COPIED_on(prog);
1692 prog->subbeg = strbeg;
1693 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
1700 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
1701 PL_colors[4],PL_colors[5]));
1702 if (PL_reg_eval_set)
1703 restore_pos(aTHXo_ 0);
1708 - regtry - try match at specific point
1710 STATIC I32 /* 0 failure, 1 success */
1711 S_regtry(pTHX_ regexp *prog, char *startpos)
1720 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
1722 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
1725 PL_reg_eval_set = RS_init;
1727 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
1728 (IV)(PL_stack_sp - PL_stack_base));
1730 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
1731 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
1732 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
1734 /* Apparently this is not needed, judging by wantarray. */
1735 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
1736 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
1739 /* Make $_ available to executed code. */
1740 if (PL_reg_sv != DEFSV) {
1741 /* SAVE_DEFSV does *not* suffice here for USE_THREADS */
1746 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
1747 && (mg = mg_find(PL_reg_sv, 'g')))) {
1748 /* prepare for quick setting of pos */
1749 sv_magic(PL_reg_sv, (SV*)0, 'g', Nullch, 0);
1750 mg = mg_find(PL_reg_sv, 'g');
1754 PL_reg_oldpos = mg->mg_len;
1755 SAVEDESTRUCTOR_X(restore_pos, 0);
1758 Newz(22,PL_reg_curpm, 1, PMOP);
1759 PL_reg_curpm->op_pmregexp = prog;
1760 PL_reg_oldcurpm = PL_curpm;
1761 PL_curpm = PL_reg_curpm;
1762 if (RX_MATCH_COPIED(prog)) {
1763 /* Here is a serious problem: we cannot rewrite subbeg,
1764 since it may be needed if this match fails. Thus
1765 $` inside (?{}) could fail... */
1766 PL_reg_oldsaved = prog->subbeg;
1767 PL_reg_oldsavedlen = prog->sublen;
1768 RX_MATCH_COPIED_off(prog);
1771 PL_reg_oldsaved = Nullch;
1772 prog->subbeg = PL_bostr;
1773 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
1775 prog->startp[0] = startpos - PL_bostr;
1776 PL_reginput = startpos;
1777 PL_regstartp = prog->startp;
1778 PL_regendp = prog->endp;
1779 PL_reglastparen = &prog->lastparen;
1780 prog->lastparen = 0;
1782 DEBUG_r(PL_reg_starttry = startpos);
1783 if (PL_reg_start_tmpl <= prog->nparens) {
1784 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
1785 if(PL_reg_start_tmp)
1786 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1788 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1791 /* XXXX What this code is doing here?!!! There should be no need
1792 to do this again and again, PL_reglastparen should take care of
1794 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this. --jhi */
1797 if (prog->nparens) {
1798 for (i = prog->nparens; i >= 1; i--) {
1804 if (regmatch(prog->program + 1)) {
1805 prog->endp[0] = PL_reginput - PL_bostr;
1808 REGCP_UNWIND(lastcp);
1812 #define RE_UNWIND_BRANCH 1
1813 #define RE_UNWIND_BRANCHJ 2
1817 typedef struct { /* XX: makes sense to enlarge it... */
1821 } re_unwind_generic_t;
1834 } re_unwind_branch_t;
1836 typedef union re_unwind_t {
1838 re_unwind_generic_t generic;
1839 re_unwind_branch_t branch;
1843 - regmatch - main matching routine
1845 * Conceptually the strategy is simple: check to see whether the current
1846 * node matches, call self recursively to see whether the rest matches,
1847 * and then act accordingly. In practice we make some effort to avoid
1848 * recursion, in particular by going through "ordinary" nodes (that don't
1849 * need to know whether the rest of the match failed) by a loop instead of
1852 /* [lwall] I've hoisted the register declarations to the outer block in order to
1853 * maybe save a little bit of pushing and popping on the stack. It also takes
1854 * advantage of machines that use a register save mask on subroutine entry.
1856 STATIC I32 /* 0 failure, 1 success */
1857 S_regmatch(pTHX_ regnode *prog)
1860 register regnode *scan; /* Current node. */
1861 regnode *next; /* Next node. */
1862 regnode *inner; /* Next node in internal branch. */
1863 register I32 nextchr; /* renamed nextchr - nextchar colides with
1864 function of same name */
1865 register I32 n; /* no or next */
1866 register I32 ln; /* len or last */
1867 register char *s; /* operand or save */
1868 register char *locinput = PL_reginput;
1869 register I32 c1, c2, paren; /* case fold search, parenth */
1870 int minmod = 0, sw = 0, logical = 0;
1872 I32 firstcp = PL_savestack_ix;
1878 /* Note that nextchr is a byte even in UTF */
1879 nextchr = UCHARAT(locinput);
1881 while (scan != NULL) {
1882 #define sayNO_L (logical ? (logical = 0, sw = 0, goto cont) : sayNO)
1884 # define sayYES goto yes
1885 # define sayNO goto no
1886 # define sayYES_FINAL goto yes_final
1887 # define sayYES_LOUD goto yes_loud
1888 # define sayNO_FINAL goto no_final
1889 # define sayNO_SILENT goto do_no
1890 # define saySAME(x) if (x) goto yes; else goto no
1891 # define REPORT_CODE_OFF 24
1893 # define sayYES return 1
1894 # define sayNO return 0
1895 # define sayYES_FINAL return 1
1896 # define sayYES_LOUD return 1
1897 # define sayNO_FINAL return 0
1898 # define sayNO_SILENT return 0
1899 # define saySAME(x) return x
1902 SV *prop = sv_newmortal();
1903 int docolor = *PL_colors[0];
1904 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
1905 int l = (PL_regeol - locinput > taill ? taill : PL_regeol - locinput);
1906 /* The part of the string before starttry has one color
1907 (pref0_len chars), between starttry and current
1908 position another one (pref_len - pref0_len chars),
1909 after the current position the third one.
1910 We assume that pref0_len <= pref_len, otherwise we
1911 decrease pref0_len. */
1912 int pref_len = (locinput - PL_bostr > (5 + taill) - l
1913 ? (5 + taill) - l : locinput - PL_bostr);
1914 int pref0_len = pref_len - (locinput - PL_reg_starttry);
1916 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
1917 l = ( PL_regeol - locinput > (5 + taill) - pref_len
1918 ? (5 + taill) - pref_len : PL_regeol - locinput);
1921 if (pref0_len > pref_len)
1922 pref0_len = pref_len;
1923 regprop(prop, scan);
1924 PerlIO_printf(Perl_debug_log,
1925 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
1926 (IV)(locinput - PL_bostr),
1927 PL_colors[4], pref0_len,
1928 locinput - pref_len, PL_colors[5],
1929 PL_colors[2], pref_len - pref0_len,
1930 locinput - pref_len + pref0_len, PL_colors[3],
1931 (docolor ? "" : "> <"),
1932 PL_colors[0], l, locinput, PL_colors[1],
1933 15 - l - pref_len + 1,
1935 (IV)(scan - PL_regprogram), PL_regindent*2, "",
1939 next = scan + NEXT_OFF(scan);
1945 if (locinput == PL_bostr
1946 ? PL_regprev == '\n'
1948 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
1950 /* regtill = regbol; */
1955 if (locinput == PL_bostr
1956 ? PL_regprev == '\n'
1957 : ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
1963 if (locinput == PL_bostr)
1967 if (locinput == PL_reg_ganch)
1977 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
1982 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
1984 if (PL_regeol - locinput > 1)
1988 if (PL_regeol != locinput)
1992 if (nextchr & 0x80) {
1993 locinput += PL_utf8skip[nextchr];
1994 if (locinput > PL_regeol)
1996 nextchr = UCHARAT(locinput);
1999 if (!nextchr && locinput >= PL_regeol)
2001 nextchr = UCHARAT(++locinput);
2004 if (!nextchr && locinput >= PL_regeol)
2006 nextchr = UCHARAT(++locinput);
2009 if (nextchr & 0x80) {
2010 locinput += PL_utf8skip[nextchr];
2011 if (locinput > PL_regeol)
2013 nextchr = UCHARAT(locinput);
2016 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2018 nextchr = UCHARAT(++locinput);
2021 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2023 nextchr = UCHARAT(++locinput);
2028 /* Inline the first character, for speed. */
2029 if (UCHARAT(s) != nextchr)
2031 if (PL_regeol - locinput < ln)
2033 if (ln > 1 && memNE(s, locinput, ln))
2036 nextchr = UCHARAT(locinput);
2039 PL_reg_flags |= RF_tainted;
2048 c1 = OP(scan) == EXACTF;
2052 if (utf8_to_uv((U8*)s, e - s, 0, 0) !=
2054 toLOWER_utf8((U8*)l) :
2055 toLOWER_LC_utf8((U8*)l)))
2063 nextchr = UCHARAT(locinput);
2067 /* Inline the first character, for speed. */
2068 if (UCHARAT(s) != nextchr &&
2069 UCHARAT(s) != ((OP(scan) == EXACTF)
2070 ? PL_fold : PL_fold_locale)[nextchr])
2072 if (PL_regeol - locinput < ln)
2074 if (ln > 1 && (OP(scan) == EXACTF
2075 ? ibcmp(s, locinput, ln)
2076 : ibcmp_locale(s, locinput, ln)))
2079 nextchr = UCHARAT(locinput);
2082 if (!REGINCLASSUTF8(scan, (U8*)locinput))
2084 if (locinput >= PL_regeol)
2086 locinput += PL_utf8skip[nextchr];
2087 nextchr = UCHARAT(locinput);
2091 nextchr = UCHARAT(locinput);
2092 if (!REGINCLASS(scan, nextchr))
2094 if (!nextchr && locinput >= PL_regeol)
2096 nextchr = UCHARAT(++locinput);
2099 PL_reg_flags |= RF_tainted;
2104 if (!(OP(scan) == ALNUM
2105 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2107 nextchr = UCHARAT(++locinput);
2110 PL_reg_flags |= RF_tainted;
2115 if (nextchr & 0x80) {
2116 if (!(OP(scan) == ALNUMUTF8
2117 ? swash_fetch(PL_utf8_alnum, (U8*)locinput)
2118 : isALNUM_LC_utf8((U8*)locinput)))
2122 locinput += PL_utf8skip[nextchr];
2123 nextchr = UCHARAT(locinput);
2126 if (!(OP(scan) == ALNUMUTF8
2127 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2129 nextchr = UCHARAT(++locinput);
2132 PL_reg_flags |= RF_tainted;
2135 if (!nextchr && locinput >= PL_regeol)
2137 if (OP(scan) == NALNUM
2138 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2140 nextchr = UCHARAT(++locinput);
2143 PL_reg_flags |= RF_tainted;
2146 if (!nextchr && locinput >= PL_regeol)
2148 if (nextchr & 0x80) {
2149 if (OP(scan) == NALNUMUTF8
2150 ? swash_fetch(PL_utf8_alnum, (U8*)locinput)
2151 : isALNUM_LC_utf8((U8*)locinput))
2155 locinput += PL_utf8skip[nextchr];
2156 nextchr = UCHARAT(locinput);
2159 if (OP(scan) == NALNUMUTF8
2160 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2162 nextchr = UCHARAT(++locinput);
2166 PL_reg_flags |= RF_tainted;
2170 /* was last char in word? */
2171 ln = (locinput != PL_regbol) ? UCHARAT(locinput - 1) : PL_regprev;
2172 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2174 n = isALNUM(nextchr);
2177 ln = isALNUM_LC(ln);
2178 n = isALNUM_LC(nextchr);
2180 if (((!ln) == (!n)) == (OP(scan) == BOUND || OP(scan) == BOUNDL))
2185 PL_reg_flags |= RF_tainted;
2189 /* was last char in word? */
2190 if (locinput == PL_regbol)
2193 U8 *r = reghop((U8*)locinput, -1);
2195 ln = utf8_to_uv(r, s - (char*)r, 0, 0);
2197 if (OP(scan) == BOUNDUTF8 || OP(scan) == NBOUNDUTF8) {
2198 ln = isALNUM_uni(ln);
2199 n = swash_fetch(PL_utf8_alnum, (U8*)locinput);
2202 ln = isALNUM_LC_uni(ln);
2203 n = isALNUM_LC_utf8((U8*)locinput);
2205 if (((!ln) == (!n)) == (OP(scan) == BOUNDUTF8 || OP(scan) == BOUNDLUTF8))
2209 PL_reg_flags |= RF_tainted;
2214 if (!(OP(scan) == SPACE
2215 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2217 nextchr = UCHARAT(++locinput);
2220 PL_reg_flags |= RF_tainted;
2225 if (nextchr & 0x80) {
2226 if (!(OP(scan) == SPACEUTF8
2227 ? swash_fetch(PL_utf8_space, (U8*)locinput)
2228 : isSPACE_LC_utf8((U8*)locinput)))
2232 locinput += PL_utf8skip[nextchr];
2233 nextchr = UCHARAT(locinput);
2236 if (!(OP(scan) == SPACEUTF8
2237 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2239 nextchr = UCHARAT(++locinput);
2242 PL_reg_flags |= RF_tainted;
2245 if (!nextchr && locinput >= PL_regeol)
2247 if (OP(scan) == NSPACE
2248 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2250 nextchr = UCHARAT(++locinput);
2253 PL_reg_flags |= RF_tainted;
2256 if (!nextchr && locinput >= PL_regeol)
2258 if (nextchr & 0x80) {
2259 if (OP(scan) == NSPACEUTF8
2260 ? swash_fetch(PL_utf8_space, (U8*)locinput)
2261 : isSPACE_LC_utf8((U8*)locinput))
2265 locinput += PL_utf8skip[nextchr];
2266 nextchr = UCHARAT(locinput);
2269 if (OP(scan) == NSPACEUTF8
2270 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2272 nextchr = UCHARAT(++locinput);
2275 PL_reg_flags |= RF_tainted;
2280 if (!(OP(scan) == DIGIT
2281 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2283 nextchr = UCHARAT(++locinput);
2286 PL_reg_flags |= RF_tainted;
2291 if (nextchr & 0x80) {
2292 if (!(OP(scan) == DIGITUTF8
2293 ? swash_fetch(PL_utf8_digit, (U8*)locinput)
2294 : isDIGIT_LC_utf8((U8*)locinput)))
2298 locinput += PL_utf8skip[nextchr];
2299 nextchr = UCHARAT(locinput);
2302 if (!(OP(scan) == DIGITUTF8
2303 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2305 nextchr = UCHARAT(++locinput);
2308 PL_reg_flags |= RF_tainted;
2311 if (!nextchr && locinput >= PL_regeol)
2313 if (OP(scan) == NDIGIT
2314 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2316 nextchr = UCHARAT(++locinput);
2319 PL_reg_flags |= RF_tainted;
2322 if (!nextchr && locinput >= PL_regeol)
2324 if (nextchr & 0x80) {
2325 if (OP(scan) == NDIGITUTF8
2326 ? swash_fetch(PL_utf8_digit, (U8*)locinput)
2327 : isDIGIT_LC_utf8((U8*)locinput))
2331 locinput += PL_utf8skip[nextchr];
2332 nextchr = UCHARAT(locinput);
2335 if (OP(scan) == NDIGITUTF8
2336 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2338 nextchr = UCHARAT(++locinput);
2341 if (locinput >= PL_regeol || swash_fetch(PL_utf8_mark,(U8*)locinput))
2343 locinput += PL_utf8skip[nextchr];
2344 while (locinput < PL_regeol && swash_fetch(PL_utf8_mark,(U8*)locinput))
2345 locinput += UTF8SKIP(locinput);
2346 if (locinput > PL_regeol)
2348 nextchr = UCHARAT(locinput);
2351 PL_reg_flags |= RF_tainted;
2355 n = ARG(scan); /* which paren pair */
2356 ln = PL_regstartp[n];
2357 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2358 if (*PL_reglastparen < n || ln == -1)
2359 sayNO; /* Do not match unless seen CLOSEn. */
2360 if (ln == PL_regendp[n])
2364 if (UTF && OP(scan) != REF) { /* REF can do byte comparison */
2366 char *e = PL_bostr + PL_regendp[n];
2368 * Note that we can't do the "other character" lookup trick as
2369 * in the 8-bit case (no pun intended) because in Unicode we
2370 * have to map both upper and title case to lower case.
2372 if (OP(scan) == REFF) {
2376 if (toLOWER_utf8((U8*)s) != toLOWER_utf8((U8*)l))
2386 if (toLOWER_LC_utf8((U8*)s) != toLOWER_LC_utf8((U8*)l))
2393 nextchr = UCHARAT(locinput);
2397 /* Inline the first character, for speed. */
2398 if (UCHARAT(s) != nextchr &&
2400 (UCHARAT(s) != ((OP(scan) == REFF
2401 ? PL_fold : PL_fold_locale)[nextchr]))))
2403 ln = PL_regendp[n] - ln;
2404 if (locinput + ln > PL_regeol)
2406 if (ln > 1 && (OP(scan) == REF
2407 ? memNE(s, locinput, ln)
2409 ? ibcmp(s, locinput, ln)
2410 : ibcmp_locale(s, locinput, ln))))
2413 nextchr = UCHARAT(locinput);
2424 OP_4tree *oop = PL_op;
2425 COP *ocurcop = PL_curcop;
2426 SV **ocurpad = PL_curpad;
2430 PL_op = (OP_4tree*)PL_regdata->data[n];
2431 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2432 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2433 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2435 CALLRUNOPS(aTHX); /* Scalar context. */
2441 PL_curpad = ocurpad;
2442 PL_curcop = ocurcop;
2444 if (logical == 2) { /* Postponed subexpression. */
2446 MAGIC *mg = Null(MAGIC*);
2448 CHECKPOINT cp, lastcp;
2450 if(SvROK(ret) || SvRMAGICAL(ret)) {
2451 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2454 mg = mg_find(sv, 'r');
2457 re = (regexp *)mg->mg_obj;
2458 (void)ReREFCNT_inc(re);
2462 char *t = SvPV(ret, len);
2464 char *oprecomp = PL_regprecomp;
2465 I32 osize = PL_regsize;
2466 I32 onpar = PL_regnpar;
2469 pm.op_pmdynflags = (UTF||DO_UTF8(ret) ? PMdf_UTF8 : 0);
2470 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2472 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2473 sv_magic(ret,(SV*)ReREFCNT_inc(re),'r',0,0);
2474 PL_regprecomp = oprecomp;
2479 PerlIO_printf(Perl_debug_log,
2480 "Entering embedded `%s%.60s%s%s'\n",
2484 (strlen(re->precomp) > 60 ? "..." : ""))
2487 state.prev = PL_reg_call_cc;
2488 state.cc = PL_regcc;
2489 state.re = PL_reg_re;
2493 cp = regcppush(0); /* Save *all* the positions. */
2496 state.ss = PL_savestack_ix;
2497 *PL_reglastparen = 0;
2498 PL_reg_call_cc = &state;
2499 PL_reginput = locinput;
2501 /* XXXX This is too dramatic a measure... */
2504 if (regmatch(re->program + 1)) {
2505 /* Even though we succeeded, we need to restore
2506 global variables, since we may be wrapped inside
2507 SUSPEND, thus the match may be not finished yet. */
2509 /* XXXX Do this only if SUSPENDed? */
2510 PL_reg_call_cc = state.prev;
2511 PL_regcc = state.cc;
2512 PL_reg_re = state.re;
2513 cache_re(PL_reg_re);
2515 /* XXXX This is too dramatic a measure... */
2518 /* These are needed even if not SUSPEND. */
2524 REGCP_UNWIND(lastcp);
2526 PL_reg_call_cc = state.prev;
2527 PL_regcc = state.cc;
2528 PL_reg_re = state.re;
2529 cache_re(PL_reg_re);
2531 /* XXXX This is too dramatic a measure... */
2540 sv_setsv(save_scalar(PL_replgv), ret);
2544 n = ARG(scan); /* which paren pair */
2545 PL_reg_start_tmp[n] = locinput;
2550 n = ARG(scan); /* which paren pair */
2551 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2552 PL_regendp[n] = locinput - PL_bostr;
2553 if (n > *PL_reglastparen)
2554 *PL_reglastparen = n;
2557 n = ARG(scan); /* which paren pair */
2558 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2561 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2563 next = NEXTOPER(NEXTOPER(scan));
2565 next = scan + ARG(scan);
2566 if (OP(next) == IFTHEN) /* Fake one. */
2567 next = NEXTOPER(NEXTOPER(next));
2571 logical = scan->flags;
2573 /*******************************************************************
2574 PL_regcc contains infoblock about the innermost (...)* loop, and
2575 a pointer to the next outer infoblock.
2577 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2579 1) After matching X, regnode for CURLYX is processed;
2581 2) This regnode creates infoblock on the stack, and calls
2582 regmatch() recursively with the starting point at WHILEM node;
2584 3) Each hit of WHILEM node tries to match A and Z (in the order
2585 depending on the current iteration, min/max of {min,max} and
2586 greediness). The information about where are nodes for "A"
2587 and "Z" is read from the infoblock, as is info on how many times "A"
2588 was already matched, and greediness.
2590 4) After A matches, the same WHILEM node is hit again.
2592 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2593 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2594 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2595 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2596 of the external loop.
2598 Currently present infoblocks form a tree with a stem formed by PL_curcc
2599 and whatever it mentions via ->next, and additional attached trees
2600 corresponding to temporarily unset infoblocks as in "5" above.
2602 In the following picture infoblocks for outer loop of
2603 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2604 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2605 infoblocks are drawn below the "reset" infoblock.
2607 In fact in the picture below we do not show failed matches for Z and T
2608 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2609 more obvious *why* one needs to *temporary* unset infoblocks.]
2611 Matched REx position InfoBlocks Comment
2615 Y A)*?Z)*?T x <- O <- I
2616 YA )*?Z)*?T x <- O <- I
2617 YA A)*?Z)*?T x <- O <- I
2618 YAA )*?Z)*?T x <- O <- I
2619 YAA Z)*?T x <- O # Temporary unset I
2622 YAAZ Y(A)*?Z)*?T x <- O
2625 YAAZY (A)*?Z)*?T x <- O
2628 YAAZY A)*?Z)*?T x <- O <- I
2631 YAAZYA )*?Z)*?T x <- O <- I
2634 YAAZYA Z)*?T x <- O # Temporary unset I
2640 YAAZYAZ T x # Temporary unset O
2647 *******************************************************************/
2650 CHECKPOINT cp = PL_savestack_ix;
2651 /* No need to save/restore up to this paren */
2652 I32 parenfloor = scan->flags;
2654 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2656 cc.oldcc = PL_regcc;
2658 /* XXXX Probably it is better to teach regpush to support
2659 parenfloor > PL_regsize... */
2660 if (parenfloor > *PL_reglastparen)
2661 parenfloor = *PL_reglastparen; /* Pessimization... */
2662 cc.parenfloor = parenfloor;
2664 cc.min = ARG1(scan);
2665 cc.max = ARG2(scan);
2666 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2670 PL_reginput = locinput;
2671 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2673 PL_regcc = cc.oldcc;
2679 * This is really hard to understand, because after we match
2680 * what we're trying to match, we must make sure the rest of
2681 * the REx is going to match for sure, and to do that we have
2682 * to go back UP the parse tree by recursing ever deeper. And
2683 * if it fails, we have to reset our parent's current state
2684 * that we can try again after backing off.
2687 CHECKPOINT cp, lastcp;
2688 CURCUR* cc = PL_regcc;
2689 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2691 n = cc->cur + 1; /* how many we know we matched */
2692 PL_reginput = locinput;
2695 PerlIO_printf(Perl_debug_log,
2696 "%*s %ld out of %ld..%ld cc=%lx\n",
2697 REPORT_CODE_OFF+PL_regindent*2, "",
2698 (long)n, (long)cc->min,
2699 (long)cc->max, (long)cc)
2702 /* If degenerate scan matches "", assume scan done. */
2704 if (locinput == cc->lastloc && n >= cc->min) {
2705 PL_regcc = cc->oldcc;
2709 PerlIO_printf(Perl_debug_log,
2710 "%*s empty match detected, try continuation...\n",
2711 REPORT_CODE_OFF+PL_regindent*2, "")
2713 if (regmatch(cc->next))
2721 /* First just match a string of min scans. */
2725 cc->lastloc = locinput;
2726 if (regmatch(cc->scan))
2729 cc->lastloc = lastloc;
2734 /* Check whether we already were at this position.
2735 Postpone detection until we know the match is not
2736 *that* much linear. */
2737 if (!PL_reg_maxiter) {
2738 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
2739 PL_reg_leftiter = PL_reg_maxiter;
2741 if (PL_reg_leftiter-- == 0) {
2742 I32 size = (PL_reg_maxiter + 7)/8;
2743 if (PL_reg_poscache) {
2744 if (PL_reg_poscache_size < size) {
2745 Renew(PL_reg_poscache, size, char);
2746 PL_reg_poscache_size = size;
2748 Zero(PL_reg_poscache, size, char);
2751 PL_reg_poscache_size = size;
2752 Newz(29, PL_reg_poscache, size, char);
2755 PerlIO_printf(Perl_debug_log,
2756 "%sDetected a super-linear match, switching on caching%s...\n",
2757 PL_colors[4], PL_colors[5])
2760 if (PL_reg_leftiter < 0) {
2761 I32 o = locinput - PL_bostr, b;
2763 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
2766 if (PL_reg_poscache[o] & (1<<b)) {
2768 PerlIO_printf(Perl_debug_log,
2769 "%*s already tried at this position...\n",
2770 REPORT_CODE_OFF+PL_regindent*2, "")
2774 PL_reg_poscache[o] |= (1<<b);
2778 /* Prefer next over scan for minimal matching. */
2781 PL_regcc = cc->oldcc;
2784 cp = regcppush(cc->parenfloor);
2786 if (regmatch(cc->next)) {
2788 sayYES; /* All done. */
2790 REGCP_UNWIND(lastcp);
2796 if (n >= cc->max) { /* Maximum greed exceeded? */
2797 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2798 && !(PL_reg_flags & RF_warned)) {
2799 PL_reg_flags |= RF_warned;
2800 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2801 "Complex regular subexpression recursion",
2808 PerlIO_printf(Perl_debug_log,
2809 "%*s trying longer...\n",
2810 REPORT_CODE_OFF+PL_regindent*2, "")
2812 /* Try scanning more and see if it helps. */
2813 PL_reginput = locinput;
2815 cc->lastloc = locinput;
2816 cp = regcppush(cc->parenfloor);
2818 if (regmatch(cc->scan)) {
2822 REGCP_UNWIND(lastcp);
2825 cc->lastloc = lastloc;
2829 /* Prefer scan over next for maximal matching. */
2831 if (n < cc->max) { /* More greed allowed? */
2832 cp = regcppush(cc->parenfloor);
2834 cc->lastloc = locinput;
2836 if (regmatch(cc->scan)) {
2840 REGCP_UNWIND(lastcp);
2841 regcppop(); /* Restore some previous $<digit>s? */
2842 PL_reginput = locinput;
2844 PerlIO_printf(Perl_debug_log,
2845 "%*s failed, try continuation...\n",
2846 REPORT_CODE_OFF+PL_regindent*2, "")
2849 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2850 && !(PL_reg_flags & RF_warned)) {
2851 PL_reg_flags |= RF_warned;
2852 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2853 "Complex regular subexpression recursion",
2857 /* Failed deeper matches of scan, so see if this one works. */
2858 PL_regcc = cc->oldcc;
2861 if (regmatch(cc->next))
2867 cc->lastloc = lastloc;
2872 next = scan + ARG(scan);
2875 inner = NEXTOPER(NEXTOPER(scan));
2878 inner = NEXTOPER(scan);
2883 if (OP(next) != c1) /* No choice. */
2884 next = inner; /* Avoid recursion. */
2886 I32 lastparen = *PL_reglastparen;
2888 re_unwind_branch_t *uw;
2890 /* Put unwinding data on stack */
2891 unwind1 = SSNEWt(1,re_unwind_branch_t);
2892 uw = SSPTRt(unwind1,re_unwind_branch_t);
2895 uw->type = ((c1 == BRANCH)
2897 : RE_UNWIND_BRANCHJ);
2898 uw->lastparen = lastparen;
2900 uw->locinput = locinput;
2901 uw->nextchr = nextchr;
2903 uw->regindent = ++PL_regindent;
2906 REGCP_SET(uw->lastcp);
2908 /* Now go into the first branch */
2921 /* We suppose that the next guy does not need
2922 backtracking: in particular, it is of constant length,
2923 and has no parenths to influence future backrefs. */
2924 ln = ARG1(scan); /* min to match */
2925 n = ARG2(scan); /* max to match */
2926 paren = scan->flags;
2928 if (paren > PL_regsize)
2930 if (paren > *PL_reglastparen)
2931 *PL_reglastparen = paren;
2933 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
2935 scan += NEXT_OFF(scan); /* Skip former OPEN. */
2936 PL_reginput = locinput;
2939 if (ln && regrepeat_hard(scan, ln, &l) < ln)
2941 if (ln && l == 0 && n >= ln
2942 /* In fact, this is tricky. If paren, then the
2943 fact that we did/didnot match may influence
2944 future execution. */
2945 && !(paren && ln == 0))
2947 locinput = PL_reginput;
2948 if (PL_regkind[(U8)OP(next)] == EXACT) {
2949 c1 = (U8)*STRING(next);
2950 if (OP(next) == EXACTF)
2952 else if (OP(next) == EXACTFL)
2953 c2 = PL_fold_locale[c1];
2960 /* This may be improved if l == 0. */
2961 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
2962 /* If it could work, try it. */
2964 UCHARAT(PL_reginput) == c1 ||
2965 UCHARAT(PL_reginput) == c2)
2969 PL_regstartp[paren] =
2970 HOPc(PL_reginput, -l) - PL_bostr;
2971 PL_regendp[paren] = PL_reginput - PL_bostr;
2974 PL_regendp[paren] = -1;
2978 REGCP_UNWIND(lastcp);
2980 /* Couldn't or didn't -- move forward. */
2981 PL_reginput = locinput;
2982 if (regrepeat_hard(scan, 1, &l)) {
2984 locinput = PL_reginput;
2991 n = regrepeat_hard(scan, n, &l);
2992 if (n != 0 && l == 0
2993 /* In fact, this is tricky. If paren, then the
2994 fact that we did/didnot match may influence
2995 future execution. */
2996 && !(paren && ln == 0))
2998 locinput = PL_reginput;
3000 PerlIO_printf(Perl_debug_log,
3001 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3002 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3006 if (PL_regkind[(U8)OP(next)] == EXACT) {
3007 c1 = (U8)*STRING(next);
3008 if (OP(next) == EXACTF)
3010 else if (OP(next) == EXACTFL)
3011 c2 = PL_fold_locale[c1];
3020 /* If it could work, try it. */
3022 UCHARAT(PL_reginput) == c1 ||
3023 UCHARAT(PL_reginput) == c2)
3026 PerlIO_printf(Perl_debug_log,
3027 "%*s trying tail with n=%"IVdf"...\n",
3028 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3032 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3033 PL_regendp[paren] = PL_reginput - PL_bostr;
3036 PL_regendp[paren] = -1;
3040 REGCP_UNWIND(lastcp);
3042 /* Couldn't or didn't -- back up. */
3044 locinput = HOPc(locinput, -l);
3045 PL_reginput = locinput;
3052 paren = scan->flags; /* Which paren to set */
3053 if (paren > PL_regsize)
3055 if (paren > *PL_reglastparen)
3056 *PL_reglastparen = paren;
3057 ln = ARG1(scan); /* min to match */
3058 n = ARG2(scan); /* max to match */
3059 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3063 ln = ARG1(scan); /* min to match */
3064 n = ARG2(scan); /* max to match */
3065 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3070 scan = NEXTOPER(scan);
3076 scan = NEXTOPER(scan);
3080 * Lookahead to avoid useless match attempts
3081 * when we know what character comes next.
3083 if (PL_regkind[(U8)OP(next)] == EXACT) {
3084 c1 = (U8)*STRING(next);
3085 if (OP(next) == EXACTF)
3087 else if (OP(next) == EXACTFL)
3088 c2 = PL_fold_locale[c1];
3094 PL_reginput = locinput;
3098 if (ln && regrepeat(scan, ln) < ln)
3100 locinput = PL_reginput;
3103 char *e = locinput + n - ln; /* Should not check after this */
3104 char *old = locinput;
3106 if (e >= PL_regeol || (n == REG_INFTY))
3109 /* Find place 'next' could work */
3111 while (locinput <= e && *locinput != c1)
3114 while (locinput <= e
3121 /* PL_reginput == old now */
3122 if (locinput != old) {
3123 ln = 1; /* Did some */
3124 if (regrepeat(scan, locinput - old) <
3128 /* PL_reginput == locinput now */
3129 TRYPAREN(paren, ln, locinput);
3130 PL_reginput = locinput; /* Could be reset... */
3131 REGCP_UNWIND(lastcp);
3132 /* Couldn't or didn't -- move forward. */
3137 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3138 /* If it could work, try it. */
3140 UCHARAT(PL_reginput) == c1 ||
3141 UCHARAT(PL_reginput) == c2)
3143 TRYPAREN(paren, n, PL_reginput);
3144 REGCP_UNWIND(lastcp);
3146 /* Couldn't or didn't -- move forward. */
3147 PL_reginput = locinput;
3148 if (regrepeat(scan, 1)) {
3150 locinput = PL_reginput;
3158 n = regrepeat(scan, n);
3159 locinput = PL_reginput;
3160 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3161 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3162 ln = n; /* why back off? */
3163 /* ...because $ and \Z can match before *and* after
3164 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3165 We should back off by one in this case. */
3166 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3172 /* If it could work, try it. */
3174 UCHARAT(PL_reginput) == c1 ||
3175 UCHARAT(PL_reginput) == c2)
3177 TRYPAREN(paren, n, PL_reginput);
3178 REGCP_UNWIND(lastcp);
3180 /* Couldn't or didn't -- back up. */
3182 PL_reginput = locinput = HOPc(locinput, -1);
3187 /* If it could work, try it. */
3189 UCHARAT(PL_reginput) == c1 ||
3190 UCHARAT(PL_reginput) == c2)
3192 TRYPAREN(paren, n, PL_reginput);
3193 REGCP_UNWIND(lastcp);
3195 /* Couldn't or didn't -- back up. */
3197 PL_reginput = locinput = HOPc(locinput, -1);
3204 if (PL_reg_call_cc) {
3205 re_cc_state *cur_call_cc = PL_reg_call_cc;
3206 CURCUR *cctmp = PL_regcc;
3207 regexp *re = PL_reg_re;
3208 CHECKPOINT cp, lastcp;
3210 cp = regcppush(0); /* Save *all* the positions. */
3212 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3214 PL_reginput = locinput; /* Make position available to
3216 cache_re(PL_reg_call_cc->re);
3217 PL_regcc = PL_reg_call_cc->cc;
3218 PL_reg_call_cc = PL_reg_call_cc->prev;
3219 if (regmatch(cur_call_cc->node)) {
3220 PL_reg_call_cc = cur_call_cc;
3224 REGCP_UNWIND(lastcp);
3226 PL_reg_call_cc = cur_call_cc;
3232 PerlIO_printf(Perl_debug_log,
3233 "%*s continuation failed...\n",
3234 REPORT_CODE_OFF+PL_regindent*2, "")
3238 if (locinput < PL_regtill) {
3239 DEBUG_r(PerlIO_printf(Perl_debug_log,
3240 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3242 (long)(locinput - PL_reg_starttry),
3243 (long)(PL_regtill - PL_reg_starttry),
3245 sayNO_FINAL; /* Cannot match: too short. */
3247 PL_reginput = locinput; /* put where regtry can find it */
3248 sayYES_FINAL; /* Success! */
3250 PL_reginput = locinput; /* put where regtry can find it */
3251 sayYES_LOUD; /* Success! */
3254 PL_reginput = locinput;
3259 if (UTF) { /* XXXX This is absolutely
3260 broken, we read before
3262 s = HOPMAYBEc(locinput, -scan->flags);
3268 if (locinput < PL_bostr + scan->flags)
3270 PL_reginput = locinput - scan->flags;
3275 PL_reginput = locinput;
3280 if (UTF) { /* XXXX This is absolutely
3281 broken, we read before
3283 s = HOPMAYBEc(locinput, -scan->flags);
3284 if (!s || s < PL_bostr)
3289 if (locinput < PL_bostr + scan->flags)
3291 PL_reginput = locinput - scan->flags;
3296 PL_reginput = locinput;
3299 inner = NEXTOPER(NEXTOPER(scan));
3300 if (regmatch(inner) != n) {
3315 if (OP(scan) == SUSPEND) {
3316 locinput = PL_reginput;
3317 nextchr = UCHARAT(locinput);
3322 next = scan + ARG(scan);
3327 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3328 PTR2UV(scan), OP(scan));
3329 Perl_croak(aTHX_ "regexp memory corruption");
3336 * We get here only if there's trouble -- normally "case END" is
3337 * the terminating point.
3339 Perl_croak(aTHX_ "corrupted regexp pointers");
3345 PerlIO_printf(Perl_debug_log,
3346 "%*s %scould match...%s\n",
3347 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3351 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3352 PL_colors[4],PL_colors[5]));
3358 #if 0 /* Breaks $^R */
3366 PerlIO_printf(Perl_debug_log,
3367 "%*s %sfailed...%s\n",
3368 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3374 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3377 case RE_UNWIND_BRANCH:
3378 case RE_UNWIND_BRANCHJ:
3380 re_unwind_branch_t *uwb = &(uw->branch);
3381 I32 lastparen = uwb->lastparen;
3383 REGCP_UNWIND(uwb->lastcp);
3384 for (n = *PL_reglastparen; n > lastparen; n--)
3386 *PL_reglastparen = n;
3387 scan = next = uwb->next;
3389 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3390 ? BRANCH : BRANCHJ) ) { /* Failure */
3397 /* Have more choice yet. Reuse the same uwb. */
3399 if ((n = (uwb->type == RE_UNWIND_BRANCH
3400 ? NEXT_OFF(next) : ARG(next))))
3403 next = NULL; /* XXXX Needn't unwinding in this case... */
3405 next = NEXTOPER(scan);
3406 if (uwb->type == RE_UNWIND_BRANCHJ)
3407 next = NEXTOPER(next);
3408 locinput = uwb->locinput;
3409 nextchr = uwb->nextchr;
3411 PL_regindent = uwb->regindent;
3418 Perl_croak(aTHX_ "regexp unwind memory corruption");
3429 - regrepeat - repeatedly match something simple, report how many
3432 * [This routine now assumes that it will only match on things of length 1.
3433 * That was true before, but now we assume scan - reginput is the count,
3434 * rather than incrementing count on every character. [Er, except utf8.]]
3437 S_regrepeat(pTHX_ regnode *p, I32 max)
3440 register char *scan;
3442 register char *loceol = PL_regeol;
3443 register I32 hardcount = 0;
3446 if (max != REG_INFTY && max < loceol - scan)
3447 loceol = scan + max;
3450 while (scan < loceol && *scan != '\n')
3458 while (scan < loceol && *scan != '\n') {
3459 scan += UTF8SKIP(scan);
3465 while (scan < loceol) {
3466 scan += UTF8SKIP(scan);
3470 case EXACT: /* length of string is 1 */
3472 while (scan < loceol && UCHARAT(scan) == c)
3475 case EXACTF: /* length of string is 1 */
3477 while (scan < loceol &&
3478 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3481 case EXACTFL: /* length of string is 1 */
3482 PL_reg_flags |= RF_tainted;
3484 while (scan < loceol &&
3485 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3490 while (scan < loceol && REGINCLASSUTF8(p, (U8*)scan)) {
3491 scan += UTF8SKIP(scan);
3496 while (scan < loceol && REGINCLASS(p, *scan))
3500 while (scan < loceol && isALNUM(*scan))
3505 while (scan < loceol && swash_fetch(PL_utf8_alnum, (U8*)scan)) {
3506 scan += UTF8SKIP(scan);
3511 PL_reg_flags |= RF_tainted;
3512 while (scan < loceol && isALNUM_LC(*scan))
3516 PL_reg_flags |= RF_tainted;
3518 while (scan < loceol && isALNUM_LC_utf8((U8*)scan)) {
3519 scan += UTF8SKIP(scan);
3525 while (scan < loceol && !isALNUM(*scan))
3530 while (scan < loceol && !swash_fetch(PL_utf8_alnum, (U8*)scan)) {
3531 scan += UTF8SKIP(scan);
3536 PL_reg_flags |= RF_tainted;
3537 while (scan < loceol && !isALNUM_LC(*scan))
3541 PL_reg_flags |= RF_tainted;
3543 while (scan < loceol && !isALNUM_LC_utf8((U8*)scan)) {
3544 scan += UTF8SKIP(scan);
3549 while (scan < loceol && isSPACE(*scan))
3554 while (scan < loceol && (*scan == ' ' || swash_fetch(PL_utf8_space,(U8*)scan))) {
3555 scan += UTF8SKIP(scan);
3560 PL_reg_flags |= RF_tainted;
3561 while (scan < loceol && isSPACE_LC(*scan))
3565 PL_reg_flags |= RF_tainted;
3567 while (scan < loceol && (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3568 scan += UTF8SKIP(scan);
3573 while (scan < loceol && !isSPACE(*scan))
3578 while (scan < loceol && !(*scan == ' ' || swash_fetch(PL_utf8_space,(U8*)scan))) {
3579 scan += UTF8SKIP(scan);
3584 PL_reg_flags |= RF_tainted;
3585 while (scan < loceol && !isSPACE_LC(*scan))
3589 PL_reg_flags |= RF_tainted;
3591 while (scan < loceol && !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3592 scan += UTF8SKIP(scan);
3597 while (scan < loceol && isDIGIT(*scan))
3602 while (scan < loceol && swash_fetch(PL_utf8_digit,(U8*)scan)) {
3603 scan += UTF8SKIP(scan);
3609 while (scan < loceol && !isDIGIT(*scan))
3614 while (scan < loceol && !swash_fetch(PL_utf8_digit,(U8*)scan)) {
3615 scan += UTF8SKIP(scan);
3619 default: /* Called on something of 0 width. */
3620 break; /* So match right here or not at all. */
3626 c = scan - PL_reginput;
3631 SV *prop = sv_newmortal();
3634 PerlIO_printf(Perl_debug_log,
3635 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
3636 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
3643 - regrepeat_hard - repeatedly match something, report total lenth and length
3645 * The repeater is supposed to have constant length.
3649 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
3652 register char *scan;
3653 register char *start;
3654 register char *loceol = PL_regeol;
3656 I32 count = 0, res = 1;
3661 start = PL_reginput;
3663 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3666 while (start < PL_reginput) {
3668 start += UTF8SKIP(start);
3679 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3681 *lp = l = PL_reginput - start;
3682 if (max != REG_INFTY && l*max < loceol - scan)
3683 loceol = scan + l*max;
3696 - reginclass - determine if a character falls into a character class
3700 S_reginclass(pTHX_ register regnode *p, register I32 c)
3703 char flags = ANYOF_FLAGS(p);
3707 if (ANYOF_BITMAP_TEST(p, c))
3709 else if (flags & ANYOF_FOLD) {
3711 if (flags & ANYOF_LOCALE) {
3712 PL_reg_flags |= RF_tainted;
3713 cf = PL_fold_locale[c];
3717 if (ANYOF_BITMAP_TEST(p, cf))
3721 if (!match && (flags & ANYOF_CLASS)) {
3722 PL_reg_flags |= RF_tainted;
3724 (ANYOF_CLASS_TEST(p, ANYOF_ALNUM) && isALNUM_LC(c)) ||
3725 (ANYOF_CLASS_TEST(p, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
3726 (ANYOF_CLASS_TEST(p, ANYOF_SPACE) && isSPACE_LC(c)) ||
3727 (ANYOF_CLASS_TEST(p, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
3728 (ANYOF_CLASS_TEST(p, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
3729 (ANYOF_CLASS_TEST(p, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
3730 (ANYOF_CLASS_TEST(p, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
3731 (ANYOF_CLASS_TEST(p, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
3732 (ANYOF_CLASS_TEST(p, ANYOF_ALPHA) && isALPHA_LC(c)) ||
3733 (ANYOF_CLASS_TEST(p, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
3734 (ANYOF_CLASS_TEST(p, ANYOF_ASCII) && isASCII(c)) ||
3735 (ANYOF_CLASS_TEST(p, ANYOF_NASCII) && !isASCII(c)) ||
3736 (ANYOF_CLASS_TEST(p, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
3737 (ANYOF_CLASS_TEST(p, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
3738 (ANYOF_CLASS_TEST(p, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
3739 (ANYOF_CLASS_TEST(p, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
3740 (ANYOF_CLASS_TEST(p, ANYOF_LOWER) && isLOWER_LC(c)) ||
3741 (ANYOF_CLASS_TEST(p, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
3742 (ANYOF_CLASS_TEST(p, ANYOF_PRINT) && isPRINT_LC(c)) ||
3743 (ANYOF_CLASS_TEST(p, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
3744 (ANYOF_CLASS_TEST(p, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
3745 (ANYOF_CLASS_TEST(p, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
3746 (ANYOF_CLASS_TEST(p, ANYOF_UPPER) && isUPPER_LC(c)) ||
3747 (ANYOF_CLASS_TEST(p, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
3748 (ANYOF_CLASS_TEST(p, ANYOF_XDIGIT) && isXDIGIT(c)) ||
3749 (ANYOF_CLASS_TEST(p, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
3750 (ANYOF_CLASS_TEST(p, ANYOF_PSXSPC) && isPSXSPC(c)) ||
3751 (ANYOF_CLASS_TEST(p, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
3752 (ANYOF_CLASS_TEST(p, ANYOF_BLANK) && isBLANK(c)) ||
3753 (ANYOF_CLASS_TEST(p, ANYOF_NBLANK) && !isBLANK(c))
3754 ) /* How's that for a conditional? */
3760 return (flags & ANYOF_INVERT) ? !match : match;
3764 S_reginclassutf8(pTHX_ regnode *f, U8 *p)
3767 char flags = ARG1(f);
3769 SV *sv = (SV*)PL_regdata->data[ARG2(f)];
3771 if (swash_fetch(sv, p))
3773 else if (flags & ANYOF_FOLD) {
3774 U8 tmpbuf[UTF8_MAXLEN];
3775 if (flags & ANYOF_LOCALE) {
3776 PL_reg_flags |= RF_tainted;
3777 uv_to_utf8(tmpbuf, toLOWER_LC_utf8(p));
3780 uv_to_utf8(tmpbuf, toLOWER_utf8(p));
3781 if (swash_fetch(sv, tmpbuf))
3785 /* UTF8 combined with ANYOF_CLASS is ill-defined. */
3787 return (flags & ANYOF_INVERT) ? !match : match;
3791 S_reghop(pTHX_ U8 *s, I32 off)
3795 while (off-- && s < (U8*)PL_regeol)
3800 if (s > (U8*)PL_bostr) {
3803 while (s > (U8*)PL_bostr && (*s & 0xc0) == 0x80)
3805 } /* XXX could check well-formedness here */
3813 S_reghopmaybe(pTHX_ U8* s, I32 off)
3817 while (off-- && s < (U8*)PL_regeol)
3824 if (s > (U8*)PL_bostr) {
3827 while (s > (U8*)PL_bostr && (*s & 0xc0) == 0x80)
3829 } /* XXX could check well-formedness here */
3845 restore_pos(pTHXo_ void *arg)
3848 if (PL_reg_eval_set) {
3849 if (PL_reg_oldsaved) {
3850 PL_reg_re->subbeg = PL_reg_oldsaved;
3851 PL_reg_re->sublen = PL_reg_oldsavedlen;
3852 RX_MATCH_COPIED_on(PL_reg_re);
3854 PL_reg_magic->mg_len = PL_reg_oldpos;
3855 PL_reg_eval_set = 0;
3856 PL_curpm = PL_reg_oldcurpm;