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
1796 if (prog->nparens) {
1797 for (i = prog->nparens; i >= 1; i--) {
1803 if (regmatch(prog->program + 1)) {
1804 prog->endp[0] = PL_reginput - PL_bostr;
1807 REGCP_UNWIND(lastcp);
1811 #define RE_UNWIND_BRANCH 1
1812 #define RE_UNWIND_BRANCHJ 2
1816 typedef struct { /* XX: makes sense to enlarge it... */
1820 } re_unwind_generic_t;
1833 } re_unwind_branch_t;
1835 typedef union re_unwind_t {
1837 re_unwind_generic_t generic;
1838 re_unwind_branch_t branch;
1842 - regmatch - main matching routine
1844 * Conceptually the strategy is simple: check to see whether the current
1845 * node matches, call self recursively to see whether the rest matches,
1846 * and then act accordingly. In practice we make some effort to avoid
1847 * recursion, in particular by going through "ordinary" nodes (that don't
1848 * need to know whether the rest of the match failed) by a loop instead of
1851 /* [lwall] I've hoisted the register declarations to the outer block in order to
1852 * maybe save a little bit of pushing and popping on the stack. It also takes
1853 * advantage of machines that use a register save mask on subroutine entry.
1855 STATIC I32 /* 0 failure, 1 success */
1856 S_regmatch(pTHX_ regnode *prog)
1859 register regnode *scan; /* Current node. */
1860 regnode *next; /* Next node. */
1861 regnode *inner; /* Next node in internal branch. */
1862 register I32 nextchr; /* renamed nextchr - nextchar colides with
1863 function of same name */
1864 register I32 n; /* no or next */
1865 register I32 ln; /* len or last */
1866 register char *s; /* operand or save */
1867 register char *locinput = PL_reginput;
1868 register I32 c1, c2, paren; /* case fold search, parenth */
1869 int minmod = 0, sw = 0, logical = 0;
1871 I32 firstcp = PL_savestack_ix;
1877 /* Note that nextchr is a byte even in UTF */
1878 nextchr = UCHARAT(locinput);
1880 while (scan != NULL) {
1881 #define sayNO_L (logical ? (logical = 0, sw = 0, goto cont) : sayNO)
1883 # define sayYES goto yes
1884 # define sayNO goto no
1885 # define sayYES_FINAL goto yes_final
1886 # define sayYES_LOUD goto yes_loud
1887 # define sayNO_FINAL goto no_final
1888 # define sayNO_SILENT goto do_no
1889 # define saySAME(x) if (x) goto yes; else goto no
1890 # define REPORT_CODE_OFF 24
1892 # define sayYES return 1
1893 # define sayNO return 0
1894 # define sayYES_FINAL return 1
1895 # define sayYES_LOUD return 1
1896 # define sayNO_FINAL return 0
1897 # define sayNO_SILENT return 0
1898 # define saySAME(x) return x
1901 SV *prop = sv_newmortal();
1902 int docolor = *PL_colors[0];
1903 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
1904 int l = (PL_regeol - locinput > taill ? taill : PL_regeol - locinput);
1905 /* The part of the string before starttry has one color
1906 (pref0_len chars), between starttry and current
1907 position another one (pref_len - pref0_len chars),
1908 after the current position the third one.
1909 We assume that pref0_len <= pref_len, otherwise we
1910 decrease pref0_len. */
1911 int pref_len = (locinput - PL_bostr > (5 + taill) - l
1912 ? (5 + taill) - l : locinput - PL_bostr);
1913 int pref0_len = pref_len - (locinput - PL_reg_starttry);
1915 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
1916 l = ( PL_regeol - locinput > (5 + taill) - pref_len
1917 ? (5 + taill) - pref_len : PL_regeol - locinput);
1920 if (pref0_len > pref_len)
1921 pref0_len = pref_len;
1922 regprop(prop, scan);
1923 PerlIO_printf(Perl_debug_log,
1924 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
1925 (IV)(locinput - PL_bostr),
1926 PL_colors[4], pref0_len,
1927 locinput - pref_len, PL_colors[5],
1928 PL_colors[2], pref_len - pref0_len,
1929 locinput - pref_len + pref0_len, PL_colors[3],
1930 (docolor ? "" : "> <"),
1931 PL_colors[0], l, locinput, PL_colors[1],
1932 15 - l - pref_len + 1,
1934 (IV)(scan - PL_regprogram), PL_regindent*2, "",
1938 next = scan + NEXT_OFF(scan);
1944 if (locinput == PL_bostr
1945 ? PL_regprev == '\n'
1947 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
1949 /* regtill = regbol; */
1954 if (locinput == PL_bostr
1955 ? PL_regprev == '\n'
1956 : ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
1962 if (locinput == PL_bostr)
1966 if (locinput == PL_reg_ganch)
1976 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
1981 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
1983 if (PL_regeol - locinput > 1)
1987 if (PL_regeol != locinput)
1991 if (nextchr & 0x80) {
1992 locinput += PL_utf8skip[nextchr];
1993 if (locinput > PL_regeol)
1995 nextchr = UCHARAT(locinput);
1998 if (!nextchr && locinput >= PL_regeol)
2000 nextchr = UCHARAT(++locinput);
2003 if (!nextchr && locinput >= PL_regeol)
2005 nextchr = UCHARAT(++locinput);
2008 if (nextchr & 0x80) {
2009 locinput += PL_utf8skip[nextchr];
2010 if (locinput > PL_regeol)
2012 nextchr = UCHARAT(locinput);
2015 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2017 nextchr = UCHARAT(++locinput);
2020 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2022 nextchr = UCHARAT(++locinput);
2027 /* Inline the first character, for speed. */
2028 if (UCHARAT(s) != nextchr)
2030 if (PL_regeol - locinput < ln)
2032 if (ln > 1 && memNE(s, locinput, ln))
2035 nextchr = UCHARAT(locinput);
2038 PL_reg_flags |= RF_tainted;
2047 c1 = OP(scan) == EXACTF;
2051 if (utf8_to_uv((U8*)s, e - s, 0, 0) !=
2053 toLOWER_utf8((U8*)l) :
2054 toLOWER_LC_utf8((U8*)l)))
2062 nextchr = UCHARAT(locinput);
2066 /* Inline the first character, for speed. */
2067 if (UCHARAT(s) != nextchr &&
2068 UCHARAT(s) != ((OP(scan) == EXACTF)
2069 ? PL_fold : PL_fold_locale)[nextchr])
2071 if (PL_regeol - locinput < ln)
2073 if (ln > 1 && (OP(scan) == EXACTF
2074 ? ibcmp(s, locinput, ln)
2075 : ibcmp_locale(s, locinput, ln)))
2078 nextchr = UCHARAT(locinput);
2081 if (!REGINCLASSUTF8(scan, (U8*)locinput))
2083 if (locinput >= PL_regeol)
2085 locinput += PL_utf8skip[nextchr];
2086 nextchr = UCHARAT(locinput);
2090 nextchr = UCHARAT(locinput);
2091 if (!REGINCLASS(scan, nextchr))
2093 if (!nextchr && locinput >= PL_regeol)
2095 nextchr = UCHARAT(++locinput);
2098 PL_reg_flags |= RF_tainted;
2103 if (!(OP(scan) == ALNUM
2104 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2106 nextchr = UCHARAT(++locinput);
2109 PL_reg_flags |= RF_tainted;
2114 if (nextchr & 0x80) {
2115 if (!(OP(scan) == ALNUMUTF8
2116 ? swash_fetch(PL_utf8_alnum, (U8*)locinput)
2117 : isALNUM_LC_utf8((U8*)locinput)))
2121 locinput += PL_utf8skip[nextchr];
2122 nextchr = UCHARAT(locinput);
2125 if (!(OP(scan) == ALNUMUTF8
2126 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2128 nextchr = UCHARAT(++locinput);
2131 PL_reg_flags |= RF_tainted;
2134 if (!nextchr && locinput >= PL_regeol)
2136 if (OP(scan) == NALNUM
2137 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2139 nextchr = UCHARAT(++locinput);
2142 PL_reg_flags |= RF_tainted;
2145 if (!nextchr && locinput >= PL_regeol)
2147 if (nextchr & 0x80) {
2148 if (OP(scan) == NALNUMUTF8
2149 ? swash_fetch(PL_utf8_alnum, (U8*)locinput)
2150 : isALNUM_LC_utf8((U8*)locinput))
2154 locinput += PL_utf8skip[nextchr];
2155 nextchr = UCHARAT(locinput);
2158 if (OP(scan) == NALNUMUTF8
2159 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2161 nextchr = UCHARAT(++locinput);
2165 PL_reg_flags |= RF_tainted;
2169 /* was last char in word? */
2170 ln = (locinput != PL_regbol) ? UCHARAT(locinput - 1) : PL_regprev;
2171 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2173 n = isALNUM(nextchr);
2176 ln = isALNUM_LC(ln);
2177 n = isALNUM_LC(nextchr);
2179 if (((!ln) == (!n)) == (OP(scan) == BOUND || OP(scan) == BOUNDL))
2184 PL_reg_flags |= RF_tainted;
2188 /* was last char in word? */
2189 if (locinput == PL_regbol)
2192 U8 *r = reghop((U8*)locinput, -1);
2194 ln = utf8_to_uv(r, s - (char*)r, 0, 0);
2196 if (OP(scan) == BOUNDUTF8 || OP(scan) == NBOUNDUTF8) {
2197 ln = isALNUM_uni(ln);
2198 n = swash_fetch(PL_utf8_alnum, (U8*)locinput);
2201 ln = isALNUM_LC_uni(ln);
2202 n = isALNUM_LC_utf8((U8*)locinput);
2204 if (((!ln) == (!n)) == (OP(scan) == BOUNDUTF8 || OP(scan) == BOUNDLUTF8))
2208 PL_reg_flags |= RF_tainted;
2213 if (!(OP(scan) == SPACE
2214 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2216 nextchr = UCHARAT(++locinput);
2219 PL_reg_flags |= RF_tainted;
2224 if (nextchr & 0x80) {
2225 if (!(OP(scan) == SPACEUTF8
2226 ? swash_fetch(PL_utf8_space, (U8*)locinput)
2227 : isSPACE_LC_utf8((U8*)locinput)))
2231 locinput += PL_utf8skip[nextchr];
2232 nextchr = UCHARAT(locinput);
2235 if (!(OP(scan) == SPACEUTF8
2236 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2238 nextchr = UCHARAT(++locinput);
2241 PL_reg_flags |= RF_tainted;
2244 if (!nextchr && locinput >= PL_regeol)
2246 if (OP(scan) == NSPACE
2247 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2249 nextchr = UCHARAT(++locinput);
2252 PL_reg_flags |= RF_tainted;
2255 if (!nextchr && locinput >= PL_regeol)
2257 if (nextchr & 0x80) {
2258 if (OP(scan) == NSPACEUTF8
2259 ? swash_fetch(PL_utf8_space, (U8*)locinput)
2260 : isSPACE_LC_utf8((U8*)locinput))
2264 locinput += PL_utf8skip[nextchr];
2265 nextchr = UCHARAT(locinput);
2268 if (OP(scan) == NSPACEUTF8
2269 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2271 nextchr = UCHARAT(++locinput);
2274 PL_reg_flags |= RF_tainted;
2279 if (!(OP(scan) == DIGIT
2280 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2282 nextchr = UCHARAT(++locinput);
2285 PL_reg_flags |= RF_tainted;
2290 if (nextchr & 0x80) {
2291 if (!(OP(scan) == DIGITUTF8
2292 ? swash_fetch(PL_utf8_digit, (U8*)locinput)
2293 : isDIGIT_LC_utf8((U8*)locinput)))
2297 locinput += PL_utf8skip[nextchr];
2298 nextchr = UCHARAT(locinput);
2301 if (!(OP(scan) == DIGITUTF8
2302 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2304 nextchr = UCHARAT(++locinput);
2307 PL_reg_flags |= RF_tainted;
2310 if (!nextchr && locinput >= PL_regeol)
2312 if (OP(scan) == NDIGIT
2313 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2315 nextchr = UCHARAT(++locinput);
2318 PL_reg_flags |= RF_tainted;
2321 if (!nextchr && locinput >= PL_regeol)
2323 if (nextchr & 0x80) {
2324 if (OP(scan) == NDIGITUTF8
2325 ? swash_fetch(PL_utf8_digit, (U8*)locinput)
2326 : isDIGIT_LC_utf8((U8*)locinput))
2330 locinput += PL_utf8skip[nextchr];
2331 nextchr = UCHARAT(locinput);
2334 if (OP(scan) == NDIGITUTF8
2335 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2337 nextchr = UCHARAT(++locinput);
2340 if (locinput >= PL_regeol || swash_fetch(PL_utf8_mark,(U8*)locinput))
2342 locinput += PL_utf8skip[nextchr];
2343 while (locinput < PL_regeol && swash_fetch(PL_utf8_mark,(U8*)locinput))
2344 locinput += UTF8SKIP(locinput);
2345 if (locinput > PL_regeol)
2347 nextchr = UCHARAT(locinput);
2350 PL_reg_flags |= RF_tainted;
2354 n = ARG(scan); /* which paren pair */
2355 ln = PL_regstartp[n];
2356 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2357 if (*PL_reglastparen < n || ln == -1)
2358 sayNO; /* Do not match unless seen CLOSEn. */
2359 if (ln == PL_regendp[n])
2363 if (UTF && OP(scan) != REF) { /* REF can do byte comparison */
2365 char *e = PL_bostr + PL_regendp[n];
2367 * Note that we can't do the "other character" lookup trick as
2368 * in the 8-bit case (no pun intended) because in Unicode we
2369 * have to map both upper and title case to lower case.
2371 if (OP(scan) == REFF) {
2375 if (toLOWER_utf8((U8*)s) != toLOWER_utf8((U8*)l))
2385 if (toLOWER_LC_utf8((U8*)s) != toLOWER_LC_utf8((U8*)l))
2392 nextchr = UCHARAT(locinput);
2396 /* Inline the first character, for speed. */
2397 if (UCHARAT(s) != nextchr &&
2399 (UCHARAT(s) != ((OP(scan) == REFF
2400 ? PL_fold : PL_fold_locale)[nextchr]))))
2402 ln = PL_regendp[n] - ln;
2403 if (locinput + ln > PL_regeol)
2405 if (ln > 1 && (OP(scan) == REF
2406 ? memNE(s, locinput, ln)
2408 ? ibcmp(s, locinput, ln)
2409 : ibcmp_locale(s, locinput, ln))))
2412 nextchr = UCHARAT(locinput);
2423 OP_4tree *oop = PL_op;
2424 COP *ocurcop = PL_curcop;
2425 SV **ocurpad = PL_curpad;
2429 PL_op = (OP_4tree*)PL_regdata->data[n];
2430 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2431 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2432 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2434 CALLRUNOPS(aTHX); /* Scalar context. */
2440 PL_curpad = ocurpad;
2441 PL_curcop = ocurcop;
2443 if (logical == 2) { /* Postponed subexpression. */
2445 MAGIC *mg = Null(MAGIC*);
2447 CHECKPOINT cp, lastcp;
2449 if(SvROK(ret) || SvRMAGICAL(ret)) {
2450 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2453 mg = mg_find(sv, 'r');
2456 re = (regexp *)mg->mg_obj;
2457 (void)ReREFCNT_inc(re);
2461 char *t = SvPV(ret, len);
2463 char *oprecomp = PL_regprecomp;
2464 I32 osize = PL_regsize;
2465 I32 onpar = PL_regnpar;
2468 pm.op_pmdynflags = (UTF||DO_UTF8(ret) ? PMdf_UTF8 : 0);
2469 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2471 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2472 sv_magic(ret,(SV*)ReREFCNT_inc(re),'r',0,0);
2473 PL_regprecomp = oprecomp;
2478 PerlIO_printf(Perl_debug_log,
2479 "Entering embedded `%s%.60s%s%s'\n",
2483 (strlen(re->precomp) > 60 ? "..." : ""))
2486 state.prev = PL_reg_call_cc;
2487 state.cc = PL_regcc;
2488 state.re = PL_reg_re;
2492 cp = regcppush(0); /* Save *all* the positions. */
2495 state.ss = PL_savestack_ix;
2496 *PL_reglastparen = 0;
2497 PL_reg_call_cc = &state;
2498 PL_reginput = locinput;
2500 /* XXXX This is too dramatic a measure... */
2503 if (regmatch(re->program + 1)) {
2504 /* Even though we succeeded, we need to restore
2505 global variables, since we may be wrapped inside
2506 SUSPEND, thus the match may be not finished yet. */
2508 /* XXXX Do this only if SUSPENDed? */
2509 PL_reg_call_cc = state.prev;
2510 PL_regcc = state.cc;
2511 PL_reg_re = state.re;
2512 cache_re(PL_reg_re);
2514 /* XXXX This is too dramatic a measure... */
2517 /* These are needed even if not SUSPEND. */
2523 REGCP_UNWIND(lastcp);
2525 PL_reg_call_cc = state.prev;
2526 PL_regcc = state.cc;
2527 PL_reg_re = state.re;
2528 cache_re(PL_reg_re);
2530 /* XXXX This is too dramatic a measure... */
2539 sv_setsv(save_scalar(PL_replgv), ret);
2543 n = ARG(scan); /* which paren pair */
2544 PL_reg_start_tmp[n] = locinput;
2549 n = ARG(scan); /* which paren pair */
2550 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2551 PL_regendp[n] = locinput - PL_bostr;
2552 if (n > *PL_reglastparen)
2553 *PL_reglastparen = n;
2556 n = ARG(scan); /* which paren pair */
2557 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2560 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2562 next = NEXTOPER(NEXTOPER(scan));
2564 next = scan + ARG(scan);
2565 if (OP(next) == IFTHEN) /* Fake one. */
2566 next = NEXTOPER(NEXTOPER(next));
2570 logical = scan->flags;
2572 /*******************************************************************
2573 PL_regcc contains infoblock about the innermost (...)* loop, and
2574 a pointer to the next outer infoblock.
2576 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2578 1) After matching X, regnode for CURLYX is processed;
2580 2) This regnode creates infoblock on the stack, and calls
2581 regmatch() recursively with the starting point at WHILEM node;
2583 3) Each hit of WHILEM node tries to match A and Z (in the order
2584 depending on the current iteration, min/max of {min,max} and
2585 greediness). The information about where are nodes for "A"
2586 and "Z" is read from the infoblock, as is info on how many times "A"
2587 was already matched, and greediness.
2589 4) After A matches, the same WHILEM node is hit again.
2591 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2592 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2593 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2594 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2595 of the external loop.
2597 Currently present infoblocks form a tree with a stem formed by PL_curcc
2598 and whatever it mentions via ->next, and additional attached trees
2599 corresponding to temporarily unset infoblocks as in "5" above.
2601 In the following picture infoblocks for outer loop of
2602 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2603 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2604 infoblocks are drawn below the "reset" infoblock.
2606 In fact in the picture below we do not show failed matches for Z and T
2607 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2608 more obvious *why* one needs to *temporary* unset infoblocks.]
2610 Matched REx position InfoBlocks Comment
2614 Y A)*?Z)*?T x <- O <- I
2615 YA )*?Z)*?T x <- O <- I
2616 YA A)*?Z)*?T x <- O <- I
2617 YAA )*?Z)*?T x <- O <- I
2618 YAA Z)*?T x <- O # Temporary unset I
2621 YAAZ Y(A)*?Z)*?T x <- O
2624 YAAZY (A)*?Z)*?T x <- O
2627 YAAZY A)*?Z)*?T x <- O <- I
2630 YAAZYA )*?Z)*?T x <- O <- I
2633 YAAZYA Z)*?T x <- O # Temporary unset I
2639 YAAZYAZ T x # Temporary unset O
2646 *******************************************************************/
2649 CHECKPOINT cp = PL_savestack_ix;
2650 /* No need to save/restore up to this paren */
2651 I32 parenfloor = scan->flags;
2653 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2655 cc.oldcc = PL_regcc;
2657 /* XXXX Probably it is better to teach regpush to support
2658 parenfloor > PL_regsize... */
2659 if (parenfloor > *PL_reglastparen)
2660 parenfloor = *PL_reglastparen; /* Pessimization... */
2661 cc.parenfloor = parenfloor;
2663 cc.min = ARG1(scan);
2664 cc.max = ARG2(scan);
2665 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2669 PL_reginput = locinput;
2670 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2672 PL_regcc = cc.oldcc;
2678 * This is really hard to understand, because after we match
2679 * what we're trying to match, we must make sure the rest of
2680 * the REx is going to match for sure, and to do that we have
2681 * to go back UP the parse tree by recursing ever deeper. And
2682 * if it fails, we have to reset our parent's current state
2683 * that we can try again after backing off.
2686 CHECKPOINT cp, lastcp;
2687 CURCUR* cc = PL_regcc;
2688 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2690 n = cc->cur + 1; /* how many we know we matched */
2691 PL_reginput = locinput;
2694 PerlIO_printf(Perl_debug_log,
2695 "%*s %ld out of %ld..%ld cc=%lx\n",
2696 REPORT_CODE_OFF+PL_regindent*2, "",
2697 (long)n, (long)cc->min,
2698 (long)cc->max, (long)cc)
2701 /* If degenerate scan matches "", assume scan done. */
2703 if (locinput == cc->lastloc && n >= cc->min) {
2704 PL_regcc = cc->oldcc;
2708 PerlIO_printf(Perl_debug_log,
2709 "%*s empty match detected, try continuation...\n",
2710 REPORT_CODE_OFF+PL_regindent*2, "")
2712 if (regmatch(cc->next))
2720 /* First just match a string of min scans. */
2724 cc->lastloc = locinput;
2725 if (regmatch(cc->scan))
2728 cc->lastloc = lastloc;
2733 /* Check whether we already were at this position.
2734 Postpone detection until we know the match is not
2735 *that* much linear. */
2736 if (!PL_reg_maxiter) {
2737 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
2738 PL_reg_leftiter = PL_reg_maxiter;
2740 if (PL_reg_leftiter-- == 0) {
2741 I32 size = (PL_reg_maxiter + 7)/8;
2742 if (PL_reg_poscache) {
2743 if (PL_reg_poscache_size < size) {
2744 Renew(PL_reg_poscache, size, char);
2745 PL_reg_poscache_size = size;
2747 Zero(PL_reg_poscache, size, char);
2750 PL_reg_poscache_size = size;
2751 Newz(29, PL_reg_poscache, size, char);
2754 PerlIO_printf(Perl_debug_log,
2755 "%sDetected a super-linear match, switching on caching%s...\n",
2756 PL_colors[4], PL_colors[5])
2759 if (PL_reg_leftiter < 0) {
2760 I32 o = locinput - PL_bostr, b;
2762 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
2765 if (PL_reg_poscache[o] & (1<<b)) {
2767 PerlIO_printf(Perl_debug_log,
2768 "%*s already tried at this position...\n",
2769 REPORT_CODE_OFF+PL_regindent*2, "")
2773 PL_reg_poscache[o] |= (1<<b);
2777 /* Prefer next over scan for minimal matching. */
2780 PL_regcc = cc->oldcc;
2783 cp = regcppush(cc->parenfloor);
2785 if (regmatch(cc->next)) {
2787 sayYES; /* All done. */
2789 REGCP_UNWIND(lastcp);
2795 if (n >= cc->max) { /* Maximum greed exceeded? */
2796 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2797 && !(PL_reg_flags & RF_warned)) {
2798 PL_reg_flags |= RF_warned;
2799 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2800 "Complex regular subexpression recursion",
2807 PerlIO_printf(Perl_debug_log,
2808 "%*s trying longer...\n",
2809 REPORT_CODE_OFF+PL_regindent*2, "")
2811 /* Try scanning more and see if it helps. */
2812 PL_reginput = locinput;
2814 cc->lastloc = locinput;
2815 cp = regcppush(cc->parenfloor);
2817 if (regmatch(cc->scan)) {
2821 REGCP_UNWIND(lastcp);
2824 cc->lastloc = lastloc;
2828 /* Prefer scan over next for maximal matching. */
2830 if (n < cc->max) { /* More greed allowed? */
2831 cp = regcppush(cc->parenfloor);
2833 cc->lastloc = locinput;
2835 if (regmatch(cc->scan)) {
2839 REGCP_UNWIND(lastcp);
2840 regcppop(); /* Restore some previous $<digit>s? */
2841 PL_reginput = locinput;
2843 PerlIO_printf(Perl_debug_log,
2844 "%*s failed, try continuation...\n",
2845 REPORT_CODE_OFF+PL_regindent*2, "")
2848 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2849 && !(PL_reg_flags & RF_warned)) {
2850 PL_reg_flags |= RF_warned;
2851 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2852 "Complex regular subexpression recursion",
2856 /* Failed deeper matches of scan, so see if this one works. */
2857 PL_regcc = cc->oldcc;
2860 if (regmatch(cc->next))
2866 cc->lastloc = lastloc;
2871 next = scan + ARG(scan);
2874 inner = NEXTOPER(NEXTOPER(scan));
2877 inner = NEXTOPER(scan);
2882 if (OP(next) != c1) /* No choice. */
2883 next = inner; /* Avoid recursion. */
2885 I32 lastparen = *PL_reglastparen;
2887 re_unwind_branch_t *uw;
2889 /* Put unwinding data on stack */
2890 unwind1 = SSNEWt(1,re_unwind_branch_t);
2891 uw = SSPTRt(unwind1,re_unwind_branch_t);
2894 uw->type = ((c1 == BRANCH)
2896 : RE_UNWIND_BRANCHJ);
2897 uw->lastparen = lastparen;
2899 uw->locinput = locinput;
2900 uw->nextchr = nextchr;
2902 uw->regindent = ++PL_regindent;
2905 REGCP_SET(uw->lastcp);
2907 /* Now go into the first branch */
2920 /* We suppose that the next guy does not need
2921 backtracking: in particular, it is of constant length,
2922 and has no parenths to influence future backrefs. */
2923 ln = ARG1(scan); /* min to match */
2924 n = ARG2(scan); /* max to match */
2925 paren = scan->flags;
2927 if (paren > PL_regsize)
2929 if (paren > *PL_reglastparen)
2930 *PL_reglastparen = paren;
2932 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
2934 scan += NEXT_OFF(scan); /* Skip former OPEN. */
2935 PL_reginput = locinput;
2938 if (ln && regrepeat_hard(scan, ln, &l) < ln)
2940 if (ln && l == 0 && n >= ln
2941 /* In fact, this is tricky. If paren, then the
2942 fact that we did/didnot match may influence
2943 future execution. */
2944 && !(paren && ln == 0))
2946 locinput = PL_reginput;
2947 if (PL_regkind[(U8)OP(next)] == EXACT) {
2948 c1 = (U8)*STRING(next);
2949 if (OP(next) == EXACTF)
2951 else if (OP(next) == EXACTFL)
2952 c2 = PL_fold_locale[c1];
2959 /* This may be improved if l == 0. */
2960 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
2961 /* If it could work, try it. */
2963 UCHARAT(PL_reginput) == c1 ||
2964 UCHARAT(PL_reginput) == c2)
2968 PL_regstartp[paren] =
2969 HOPc(PL_reginput, -l) - PL_bostr;
2970 PL_regendp[paren] = PL_reginput - PL_bostr;
2973 PL_regendp[paren] = -1;
2977 REGCP_UNWIND(lastcp);
2979 /* Couldn't or didn't -- move forward. */
2980 PL_reginput = locinput;
2981 if (regrepeat_hard(scan, 1, &l)) {
2983 locinput = PL_reginput;
2990 n = regrepeat_hard(scan, n, &l);
2991 if (n != 0 && l == 0
2992 /* In fact, this is tricky. If paren, then the
2993 fact that we did/didnot match may influence
2994 future execution. */
2995 && !(paren && ln == 0))
2997 locinput = PL_reginput;
2999 PerlIO_printf(Perl_debug_log,
3000 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3001 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3005 if (PL_regkind[(U8)OP(next)] == EXACT) {
3006 c1 = (U8)*STRING(next);
3007 if (OP(next) == EXACTF)
3009 else if (OP(next) == EXACTFL)
3010 c2 = PL_fold_locale[c1];
3019 /* If it could work, try it. */
3021 UCHARAT(PL_reginput) == c1 ||
3022 UCHARAT(PL_reginput) == c2)
3025 PerlIO_printf(Perl_debug_log,
3026 "%*s trying tail with n=%"IVdf"...\n",
3027 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3031 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3032 PL_regendp[paren] = PL_reginput - PL_bostr;
3035 PL_regendp[paren] = -1;
3039 REGCP_UNWIND(lastcp);
3041 /* Couldn't or didn't -- back up. */
3043 locinput = HOPc(locinput, -l);
3044 PL_reginput = locinput;
3051 paren = scan->flags; /* Which paren to set */
3052 if (paren > PL_regsize)
3054 if (paren > *PL_reglastparen)
3055 *PL_reglastparen = paren;
3056 ln = ARG1(scan); /* min to match */
3057 n = ARG2(scan); /* max to match */
3058 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3062 ln = ARG1(scan); /* min to match */
3063 n = ARG2(scan); /* max to match */
3064 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3069 scan = NEXTOPER(scan);
3075 scan = NEXTOPER(scan);
3079 * Lookahead to avoid useless match attempts
3080 * when we know what character comes next.
3082 if (PL_regkind[(U8)OP(next)] == EXACT) {
3083 c1 = (U8)*STRING(next);
3084 if (OP(next) == EXACTF)
3086 else if (OP(next) == EXACTFL)
3087 c2 = PL_fold_locale[c1];
3093 PL_reginput = locinput;
3097 if (ln && regrepeat(scan, ln) < ln)
3099 locinput = PL_reginput;
3102 char *e = locinput + n - ln; /* Should not check after this */
3103 char *old = locinput;
3105 if (e >= PL_regeol || (n == REG_INFTY))
3108 /* Find place 'next' could work */
3110 while (locinput <= e && *locinput != c1)
3113 while (locinput <= e
3120 /* PL_reginput == old now */
3121 if (locinput != old) {
3122 ln = 1; /* Did some */
3123 if (regrepeat(scan, locinput - old) <
3127 /* PL_reginput == locinput now */
3128 TRYPAREN(paren, ln, locinput);
3129 PL_reginput = locinput; /* Could be reset... */
3130 REGCP_UNWIND(lastcp);
3131 /* Couldn't or didn't -- move forward. */
3136 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3137 /* If it could work, try it. */
3139 UCHARAT(PL_reginput) == c1 ||
3140 UCHARAT(PL_reginput) == c2)
3142 TRYPAREN(paren, n, PL_reginput);
3143 REGCP_UNWIND(lastcp);
3145 /* Couldn't or didn't -- move forward. */
3146 PL_reginput = locinput;
3147 if (regrepeat(scan, 1)) {
3149 locinput = PL_reginput;
3157 n = regrepeat(scan, n);
3158 locinput = PL_reginput;
3159 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3160 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3161 ln = n; /* why back off? */
3162 /* ...because $ and \Z can match before *and* after
3163 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3164 We should back off by one in this case. */
3165 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3171 /* If it could work, try it. */
3173 UCHARAT(PL_reginput) == c1 ||
3174 UCHARAT(PL_reginput) == c2)
3176 TRYPAREN(paren, n, PL_reginput);
3177 REGCP_UNWIND(lastcp);
3179 /* Couldn't or didn't -- back up. */
3181 PL_reginput = locinput = HOPc(locinput, -1);
3186 /* If it could work, try it. */
3188 UCHARAT(PL_reginput) == c1 ||
3189 UCHARAT(PL_reginput) == c2)
3191 TRYPAREN(paren, n, PL_reginput);
3192 REGCP_UNWIND(lastcp);
3194 /* Couldn't or didn't -- back up. */
3196 PL_reginput = locinput = HOPc(locinput, -1);
3203 if (PL_reg_call_cc) {
3204 re_cc_state *cur_call_cc = PL_reg_call_cc;
3205 CURCUR *cctmp = PL_regcc;
3206 regexp *re = PL_reg_re;
3207 CHECKPOINT cp, lastcp;
3209 cp = regcppush(0); /* Save *all* the positions. */
3211 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3213 PL_reginput = locinput; /* Make position available to
3215 cache_re(PL_reg_call_cc->re);
3216 PL_regcc = PL_reg_call_cc->cc;
3217 PL_reg_call_cc = PL_reg_call_cc->prev;
3218 if (regmatch(cur_call_cc->node)) {
3219 PL_reg_call_cc = cur_call_cc;
3223 REGCP_UNWIND(lastcp);
3225 PL_reg_call_cc = cur_call_cc;
3231 PerlIO_printf(Perl_debug_log,
3232 "%*s continuation failed...\n",
3233 REPORT_CODE_OFF+PL_regindent*2, "")
3237 if (locinput < PL_regtill) {
3238 DEBUG_r(PerlIO_printf(Perl_debug_log,
3239 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3241 (long)(locinput - PL_reg_starttry),
3242 (long)(PL_regtill - PL_reg_starttry),
3244 sayNO_FINAL; /* Cannot match: too short. */
3246 PL_reginput = locinput; /* put where regtry can find it */
3247 sayYES_FINAL; /* Success! */
3249 PL_reginput = locinput; /* put where regtry can find it */
3250 sayYES_LOUD; /* Success! */
3253 PL_reginput = locinput;
3258 if (UTF) { /* XXXX This is absolutely
3259 broken, we read before
3261 s = HOPMAYBEc(locinput, -scan->flags);
3267 if (locinput < PL_bostr + scan->flags)
3269 PL_reginput = locinput - scan->flags;
3274 PL_reginput = locinput;
3279 if (UTF) { /* XXXX This is absolutely
3280 broken, we read before
3282 s = HOPMAYBEc(locinput, -scan->flags);
3283 if (!s || s < PL_bostr)
3288 if (locinput < PL_bostr + scan->flags)
3290 PL_reginput = locinput - scan->flags;
3295 PL_reginput = locinput;
3298 inner = NEXTOPER(NEXTOPER(scan));
3299 if (regmatch(inner) != n) {
3314 if (OP(scan) == SUSPEND) {
3315 locinput = PL_reginput;
3316 nextchr = UCHARAT(locinput);
3321 next = scan + ARG(scan);
3326 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3327 PTR2UV(scan), OP(scan));
3328 Perl_croak(aTHX_ "regexp memory corruption");
3335 * We get here only if there's trouble -- normally "case END" is
3336 * the terminating point.
3338 Perl_croak(aTHX_ "corrupted regexp pointers");
3344 PerlIO_printf(Perl_debug_log,
3345 "%*s %scould match...%s\n",
3346 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3350 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3351 PL_colors[4],PL_colors[5]));
3357 #if 0 /* Breaks $^R */
3365 PerlIO_printf(Perl_debug_log,
3366 "%*s %sfailed...%s\n",
3367 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3373 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3376 case RE_UNWIND_BRANCH:
3377 case RE_UNWIND_BRANCHJ:
3379 re_unwind_branch_t *uwb = &(uw->branch);
3380 I32 lastparen = uwb->lastparen;
3382 REGCP_UNWIND(uwb->lastcp);
3383 for (n = *PL_reglastparen; n > lastparen; n--)
3385 *PL_reglastparen = n;
3386 scan = next = uwb->next;
3388 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3389 ? BRANCH : BRANCHJ) ) { /* Failure */
3396 /* Have more choice yet. Reuse the same uwb. */
3398 if ((n = (uwb->type == RE_UNWIND_BRANCH
3399 ? NEXT_OFF(next) : ARG(next))))
3402 next = NULL; /* XXXX Needn't unwinding in this case... */
3404 next = NEXTOPER(scan);
3405 if (uwb->type == RE_UNWIND_BRANCHJ)
3406 next = NEXTOPER(next);
3407 locinput = uwb->locinput;
3408 nextchr = uwb->nextchr;
3410 PL_regindent = uwb->regindent;
3417 Perl_croak(aTHX_ "regexp unwind memory corruption");
3428 - regrepeat - repeatedly match something simple, report how many
3431 * [This routine now assumes that it will only match on things of length 1.
3432 * That was true before, but now we assume scan - reginput is the count,
3433 * rather than incrementing count on every character. [Er, except utf8.]]
3436 S_regrepeat(pTHX_ regnode *p, I32 max)
3439 register char *scan;
3441 register char *loceol = PL_regeol;
3442 register I32 hardcount = 0;
3445 if (max != REG_INFTY && max < loceol - scan)
3446 loceol = scan + max;
3449 while (scan < loceol && *scan != '\n')
3457 while (scan < loceol && *scan != '\n') {
3458 scan += UTF8SKIP(scan);
3464 while (scan < loceol) {
3465 scan += UTF8SKIP(scan);
3469 case EXACT: /* length of string is 1 */
3471 while (scan < loceol && UCHARAT(scan) == c)
3474 case EXACTF: /* length of string is 1 */
3476 while (scan < loceol &&
3477 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3480 case EXACTFL: /* length of string is 1 */
3481 PL_reg_flags |= RF_tainted;
3483 while (scan < loceol &&
3484 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3489 while (scan < loceol && REGINCLASSUTF8(p, (U8*)scan)) {
3490 scan += UTF8SKIP(scan);
3495 while (scan < loceol && REGINCLASS(p, *scan))
3499 while (scan < loceol && isALNUM(*scan))
3504 while (scan < loceol && swash_fetch(PL_utf8_alnum, (U8*)scan)) {
3505 scan += UTF8SKIP(scan);
3510 PL_reg_flags |= RF_tainted;
3511 while (scan < loceol && isALNUM_LC(*scan))
3515 PL_reg_flags |= RF_tainted;
3517 while (scan < loceol && isALNUM_LC_utf8((U8*)scan)) {
3518 scan += UTF8SKIP(scan);
3524 while (scan < loceol && !isALNUM(*scan))
3529 while (scan < loceol && !swash_fetch(PL_utf8_alnum, (U8*)scan)) {
3530 scan += UTF8SKIP(scan);
3535 PL_reg_flags |= RF_tainted;
3536 while (scan < loceol && !isALNUM_LC(*scan))
3540 PL_reg_flags |= RF_tainted;
3542 while (scan < loceol && !isALNUM_LC_utf8((U8*)scan)) {
3543 scan += UTF8SKIP(scan);
3548 while (scan < loceol && isSPACE(*scan))
3553 while (scan < loceol && (*scan == ' ' || swash_fetch(PL_utf8_space,(U8*)scan))) {
3554 scan += UTF8SKIP(scan);
3559 PL_reg_flags |= RF_tainted;
3560 while (scan < loceol && isSPACE_LC(*scan))
3564 PL_reg_flags |= RF_tainted;
3566 while (scan < loceol && (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3567 scan += UTF8SKIP(scan);
3572 while (scan < loceol && !isSPACE(*scan))
3577 while (scan < loceol && !(*scan == ' ' || swash_fetch(PL_utf8_space,(U8*)scan))) {
3578 scan += UTF8SKIP(scan);
3583 PL_reg_flags |= RF_tainted;
3584 while (scan < loceol && !isSPACE_LC(*scan))
3588 PL_reg_flags |= RF_tainted;
3590 while (scan < loceol && !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3591 scan += UTF8SKIP(scan);
3596 while (scan < loceol && isDIGIT(*scan))
3601 while (scan < loceol && swash_fetch(PL_utf8_digit,(U8*)scan)) {
3602 scan += UTF8SKIP(scan);
3608 while (scan < loceol && !isDIGIT(*scan))
3613 while (scan < loceol && !swash_fetch(PL_utf8_digit,(U8*)scan)) {
3614 scan += UTF8SKIP(scan);
3618 default: /* Called on something of 0 width. */
3619 break; /* So match right here or not at all. */
3625 c = scan - PL_reginput;
3630 SV *prop = sv_newmortal();
3633 PerlIO_printf(Perl_debug_log,
3634 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
3635 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
3642 - regrepeat_hard - repeatedly match something, report total lenth and length
3644 * The repeater is supposed to have constant length.
3648 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
3651 register char *scan;
3652 register char *start;
3653 register char *loceol = PL_regeol;
3655 I32 count = 0, res = 1;
3660 start = PL_reginput;
3662 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3665 while (start < PL_reginput) {
3667 start += UTF8SKIP(start);
3678 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3680 *lp = l = PL_reginput - start;
3681 if (max != REG_INFTY && l*max < loceol - scan)
3682 loceol = scan + l*max;
3695 - reginclass - determine if a character falls into a character class
3699 S_reginclass(pTHX_ register regnode *p, register I32 c)
3702 char flags = ANYOF_FLAGS(p);
3706 if (ANYOF_BITMAP_TEST(p, c))
3708 else if (flags & ANYOF_FOLD) {
3710 if (flags & ANYOF_LOCALE) {
3711 PL_reg_flags |= RF_tainted;
3712 cf = PL_fold_locale[c];
3716 if (ANYOF_BITMAP_TEST(p, cf))
3720 if (!match && (flags & ANYOF_CLASS)) {
3721 PL_reg_flags |= RF_tainted;
3723 (ANYOF_CLASS_TEST(p, ANYOF_ALNUM) && isALNUM_LC(c)) ||
3724 (ANYOF_CLASS_TEST(p, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
3725 (ANYOF_CLASS_TEST(p, ANYOF_SPACE) && isSPACE_LC(c)) ||
3726 (ANYOF_CLASS_TEST(p, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
3727 (ANYOF_CLASS_TEST(p, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
3728 (ANYOF_CLASS_TEST(p, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
3729 (ANYOF_CLASS_TEST(p, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
3730 (ANYOF_CLASS_TEST(p, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
3731 (ANYOF_CLASS_TEST(p, ANYOF_ALPHA) && isALPHA_LC(c)) ||
3732 (ANYOF_CLASS_TEST(p, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
3733 (ANYOF_CLASS_TEST(p, ANYOF_ASCII) && isASCII(c)) ||
3734 (ANYOF_CLASS_TEST(p, ANYOF_NASCII) && !isASCII(c)) ||
3735 (ANYOF_CLASS_TEST(p, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
3736 (ANYOF_CLASS_TEST(p, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
3737 (ANYOF_CLASS_TEST(p, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
3738 (ANYOF_CLASS_TEST(p, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
3739 (ANYOF_CLASS_TEST(p, ANYOF_LOWER) && isLOWER_LC(c)) ||
3740 (ANYOF_CLASS_TEST(p, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
3741 (ANYOF_CLASS_TEST(p, ANYOF_PRINT) && isPRINT_LC(c)) ||
3742 (ANYOF_CLASS_TEST(p, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
3743 (ANYOF_CLASS_TEST(p, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
3744 (ANYOF_CLASS_TEST(p, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
3745 (ANYOF_CLASS_TEST(p, ANYOF_UPPER) && isUPPER_LC(c)) ||
3746 (ANYOF_CLASS_TEST(p, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
3747 (ANYOF_CLASS_TEST(p, ANYOF_XDIGIT) && isXDIGIT(c)) ||
3748 (ANYOF_CLASS_TEST(p, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
3749 (ANYOF_CLASS_TEST(p, ANYOF_PSXSPC) && isPSXSPC(c)) ||
3750 (ANYOF_CLASS_TEST(p, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
3751 (ANYOF_CLASS_TEST(p, ANYOF_BLANK) && isBLANK(c)) ||
3752 (ANYOF_CLASS_TEST(p, ANYOF_NBLANK) && !isBLANK(c))
3753 ) /* How's that for a conditional? */
3759 return (flags & ANYOF_INVERT) ? !match : match;
3763 S_reginclassutf8(pTHX_ regnode *f, U8 *p)
3766 char flags = ARG1(f);
3768 SV *sv = (SV*)PL_regdata->data[ARG2(f)];
3770 if (swash_fetch(sv, p))
3772 else if (flags & ANYOF_FOLD) {
3773 U8 tmpbuf[UTF8_MAXLEN];
3774 if (flags & ANYOF_LOCALE) {
3775 PL_reg_flags |= RF_tainted;
3776 uv_to_utf8(tmpbuf, toLOWER_LC_utf8(p));
3779 uv_to_utf8(tmpbuf, toLOWER_utf8(p));
3780 if (swash_fetch(sv, tmpbuf))
3784 /* UTF8 combined with ANYOF_CLASS is ill-defined. */
3786 return (flags & ANYOF_INVERT) ? !match : match;
3790 S_reghop(pTHX_ U8 *s, I32 off)
3794 while (off-- && s < (U8*)PL_regeol)
3799 if (s > (U8*)PL_bostr) {
3802 while (s > (U8*)PL_bostr && (*s & 0xc0) == 0x80)
3804 } /* XXX could check well-formedness here */
3812 S_reghopmaybe(pTHX_ U8* s, I32 off)
3816 while (off-- && s < (U8*)PL_regeol)
3823 if (s > (U8*)PL_bostr) {
3826 while (s > (U8*)PL_bostr && (*s & 0xc0) == 0x80)
3828 } /* XXX could check well-formedness here */
3844 restore_pos(pTHXo_ void *arg)
3847 if (PL_reg_eval_set) {
3848 if (PL_reg_oldsaved) {
3849 PL_reg_re->subbeg = PL_reg_oldsaved;
3850 PL_reg_re->sublen = PL_reg_oldsavedlen;
3851 RX_MATCH_COPIED_on(PL_reg_re);
3853 PL_reg_magic->mg_len = PL_reg_oldpos;
3854 PL_reg_eval_set = 0;
3855 PL_curpm = PL_reg_oldcurpm;