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;
920 tmp = (I32)(s != startpos) ? utf8_to_uv(reghop((U8*)s, -1),
923 tmp = ((OP(c) == BOUNDUTF8 ? isALNUM_uni(tmp) : isALNUM_LC_uni(tmp)) != 0);
925 if (tmp == !(OP(c) == BOUNDUTF8 ?
926 swash_fetch(PL_utf8_alnum, (U8*)s) :
927 isALNUM_LC_utf8((U8*)s)))
930 if ((norun || regtry(prog, s)))
935 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
939 PL_reg_flags |= RF_tainted;
942 tmp = (s != startpos) ? UCHARAT(s - 1) : '\n';
943 tmp = ((OP(c) == NBOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
945 if (tmp == !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
947 else if ((norun || regtry(prog, s)))
951 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
955 PL_reg_flags |= RF_tainted;
958 tmp = (I32)(s != startpos) ? utf8_to_uv(reghop((U8*)s, -1),
961 tmp = ((OP(c) == NBOUNDUTF8 ? isALNUM_uni(tmp) : isALNUM_LC_uni(tmp)) != 0);
963 if (tmp == !(OP(c) == NBOUNDUTF8 ?
964 swash_fetch(PL_utf8_alnum, (U8*)s) :
965 isALNUM_LC_utf8((U8*)s)))
967 else if ((norun || regtry(prog, s)))
971 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
977 if (tmp && (norun || regtry(prog, s)))
989 if (swash_fetch(PL_utf8_alnum, (U8*)s)) {
990 if (tmp && (norun || regtry(prog, s)))
1001 PL_reg_flags |= RF_tainted;
1002 while (s < strend) {
1003 if (isALNUM_LC(*s)) {
1004 if (tmp && (norun || regtry(prog, s)))
1015 PL_reg_flags |= RF_tainted;
1016 while (s < strend) {
1017 if (isALNUM_LC_utf8((U8*)s)) {
1018 if (tmp && (norun || regtry(prog, s)))
1029 while (s < strend) {
1031 if (tmp && (norun || regtry(prog, s)))
1042 while (s < strend) {
1043 if (!swash_fetch(PL_utf8_alnum, (U8*)s)) {
1044 if (tmp && (norun || regtry(prog, s)))
1055 PL_reg_flags |= RF_tainted;
1056 while (s < strend) {
1057 if (!isALNUM_LC(*s)) {
1058 if (tmp && (norun || regtry(prog, s)))
1069 PL_reg_flags |= RF_tainted;
1070 while (s < strend) {
1071 if (!isALNUM_LC_utf8((U8*)s)) {
1072 if (tmp && (norun || regtry(prog, s)))
1083 while (s < strend) {
1085 if (tmp && (norun || regtry(prog, s)))
1096 while (s < strend) {
1097 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s)) {
1098 if (tmp && (norun || regtry(prog, s)))
1109 PL_reg_flags |= RF_tainted;
1110 while (s < strend) {
1111 if (isSPACE_LC(*s)) {
1112 if (tmp && (norun || regtry(prog, s)))
1123 PL_reg_flags |= RF_tainted;
1124 while (s < strend) {
1125 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1126 if (tmp && (norun || regtry(prog, s)))
1137 while (s < strend) {
1139 if (tmp && (norun || regtry(prog, s)))
1150 while (s < strend) {
1151 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s))) {
1152 if (tmp && (norun || regtry(prog, s)))
1163 PL_reg_flags |= RF_tainted;
1164 while (s < strend) {
1165 if (!isSPACE_LC(*s)) {
1166 if (tmp && (norun || regtry(prog, s)))
1177 PL_reg_flags |= RF_tainted;
1178 while (s < strend) {
1179 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1180 if (tmp && (norun || regtry(prog, s)))
1191 while (s < strend) {
1193 if (tmp && (norun || regtry(prog, s)))
1204 while (s < strend) {
1205 if (swash_fetch(PL_utf8_digit,(U8*)s)) {
1206 if (tmp && (norun || regtry(prog, s)))
1217 PL_reg_flags |= RF_tainted;
1218 while (s < strend) {
1219 if (isDIGIT_LC(*s)) {
1220 if (tmp && (norun || regtry(prog, s)))
1231 PL_reg_flags |= RF_tainted;
1232 while (s < strend) {
1233 if (isDIGIT_LC_utf8((U8*)s)) {
1234 if (tmp && (norun || regtry(prog, s)))
1245 while (s < strend) {
1247 if (tmp && (norun || regtry(prog, s)))
1258 while (s < strend) {
1259 if (!swash_fetch(PL_utf8_digit,(U8*)s)) {
1260 if (tmp && (norun || regtry(prog, s)))
1271 PL_reg_flags |= RF_tainted;
1272 while (s < strend) {
1273 if (!isDIGIT_LC(*s)) {
1274 if (tmp && (norun || regtry(prog, s)))
1285 PL_reg_flags |= RF_tainted;
1286 while (s < strend) {
1287 if (!isDIGIT_LC_utf8((U8*)s)) {
1288 if (tmp && (norun || regtry(prog, s)))
1299 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1308 - regexec_flags - match a regexp against a string
1311 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1312 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1313 /* strend: pointer to null at end of string */
1314 /* strbeg: real beginning of string */
1315 /* minend: end of match must be >=minend after stringarg. */
1316 /* data: May be used for some additional optimizations. */
1317 /* nosave: For optimizations. */
1321 register regnode *c;
1322 register char *startpos = stringarg;
1323 I32 minlen; /* must match at least this many chars */
1324 I32 dontbother = 0; /* how many characters not to try at end */
1325 /* I32 start_shift = 0; */ /* Offset of the start to find
1326 constant substr. */ /* CC */
1327 I32 end_shift = 0; /* Same for the end. */ /* CC */
1328 I32 scream_pos = -1; /* Internal iterator of scream. */
1330 SV* oreplsv = GvSV(PL_replgv);
1336 PL_regnarrate = PL_debug & 512;
1339 /* Be paranoid... */
1340 if (prog == NULL || startpos == NULL) {
1341 Perl_croak(aTHX_ "NULL regexp parameter");
1345 minlen = prog->minlen;
1346 if (strend - startpos < minlen) goto phooey;
1348 if (startpos == strbeg) /* is ^ valid at stringarg? */
1351 PL_regprev = (U32)stringarg[-1];
1352 if (!PL_multiline && PL_regprev == '\n')
1353 PL_regprev = '\0'; /* force ^ to NOT match */
1356 /* Check validity of program. */
1357 if (UCHARAT(prog->program) != REG_MAGIC) {
1358 Perl_croak(aTHX_ "corrupted regexp program");
1362 PL_reg_eval_set = 0;
1365 if (prog->reganch & ROPT_UTF8)
1366 PL_reg_flags |= RF_utf8;
1368 /* Mark beginning of line for ^ and lookbehind. */
1369 PL_regbol = startpos;
1373 /* Mark end of line for $ (and such) */
1376 /* see how far we have to get to not match where we matched before */
1377 PL_regtill = startpos+minend;
1379 /* We start without call_cc context. */
1382 /* If there is a "must appear" string, look for it. */
1385 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1388 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1389 PL_reg_ganch = startpos;
1390 else if (sv && SvTYPE(sv) >= SVt_PVMG
1392 && (mg = mg_find(sv, 'g')) && mg->mg_len >= 0) {
1393 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1394 if (prog->reganch & ROPT_ANCH_GPOS) {
1395 if (s > PL_reg_ganch)
1400 else /* pos() not defined */
1401 PL_reg_ganch = strbeg;
1404 if (!(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
1405 re_scream_pos_data d;
1407 d.scream_olds = &scream_olds;
1408 d.scream_pos = &scream_pos;
1409 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1411 goto phooey; /* not present */
1414 DEBUG_r( if (!PL_colorset) reginitcolors() );
1415 DEBUG_r(PerlIO_printf(Perl_debug_log,
1416 "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
1417 PL_colors[4],PL_colors[5],PL_colors[0],
1420 (strlen(prog->precomp) > 60 ? "..." : ""),
1422 (int)(strend - startpos > 60 ? 60 : strend - startpos),
1423 startpos, PL_colors[1],
1424 (strend - startpos > 60 ? "..." : ""))
1427 /* Simplest case: anchored match need be tried only once. */
1428 /* [unless only anchor is BOL and multiline is set] */
1429 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1430 if (s == startpos && regtry(prog, startpos))
1432 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1433 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1438 dontbother = minlen - 1;
1439 end = HOPc(strend, -dontbother) - 1;
1440 /* for multiline we only have to try after newlines */
1441 if (prog->check_substr) {
1445 if (regtry(prog, s))
1450 if (prog->reganch & RE_USE_INTUIT) {
1451 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1462 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1463 if (regtry(prog, s))
1470 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1471 if (regtry(prog, PL_reg_ganch))
1476 /* Messy cases: unanchored match. */
1477 if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
1478 /* we have /x+whatever/ */
1479 /* it must be a one character string (XXXX Except UTF?) */
1480 char ch = SvPVX(prog->anchored_substr)[0];
1486 while (s < strend) {
1488 DEBUG_r( did_match = 1 );
1489 if (regtry(prog, s)) goto got_it;
1491 while (s < strend && *s == ch)
1498 while (s < strend) {
1500 DEBUG_r( did_match = 1 );
1501 if (regtry(prog, s)) goto got_it;
1503 while (s < strend && *s == ch)
1509 DEBUG_r(did_match ||
1510 PerlIO_printf(Perl_debug_log,
1511 "Did not find anchored character...\n"));
1514 else if (prog->anchored_substr != Nullsv
1515 || (prog->float_substr != Nullsv
1516 && prog->float_max_offset < strend - s)) {
1517 SV *must = prog->anchored_substr
1518 ? prog->anchored_substr : prog->float_substr;
1520 prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset;
1522 prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
1523 char *last = HOPc(strend, /* Cannot start after this */
1524 -(I32)(CHR_SVLEN(must)
1525 - (SvTAIL(must) != 0) + back_min));
1526 char *last1; /* Last position checked before */
1532 last1 = HOPc(s, -1);
1534 last1 = s - 1; /* bogus */
1536 /* XXXX check_substr already used to find `s', can optimize if
1537 check_substr==must. */
1539 dontbother = end_shift;
1540 strend = HOPc(strend, -dontbother);
1541 while ( (s <= last) &&
1542 ((flags & REXEC_SCREAM)
1543 ? (s = screaminstr(sv, must, HOPc(s, back_min) - strbeg,
1544 end_shift, &scream_pos, 0))
1545 : (s = fbm_instr((unsigned char*)HOP(s, back_min),
1546 (unsigned char*)strend, must,
1547 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1548 DEBUG_r( did_match = 1 );
1549 if (HOPc(s, -back_max) > last1) {
1550 last1 = HOPc(s, -back_min);
1551 s = HOPc(s, -back_max);
1554 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1556 last1 = HOPc(s, -back_min);
1560 while (s <= last1) {
1561 if (regtry(prog, s))
1567 while (s <= last1) {
1568 if (regtry(prog, s))
1574 DEBUG_r(did_match ||
1575 PerlIO_printf(Perl_debug_log, "Did not find %s substr `%s%.*s%s'%s...\n",
1576 ((must == prog->anchored_substr)
1577 ? "anchored" : "floating"),
1579 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1581 PL_colors[1], (SvTAIL(must) ? "$" : "")));
1584 else if ((c = prog->regstclass)) {
1585 if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT)
1586 /* don't bother with what can't match */
1587 strend = HOPc(strend, -(minlen - 1));
1588 if (find_byclass(prog, c, s, strend, startpos, 0))
1590 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1594 if (prog->float_substr != Nullsv) { /* Trim the end. */
1597 if (flags & REXEC_SCREAM) {
1598 last = screaminstr(sv, prog->float_substr, s - strbeg,
1599 end_shift, &scream_pos, 1); /* last one */
1601 last = scream_olds; /* Only one occurence. */
1605 char *little = SvPV(prog->float_substr, len);
1607 if (SvTAIL(prog->float_substr)) {
1608 if (memEQ(strend - len + 1, little, len - 1))
1609 last = strend - len + 1;
1610 else if (!PL_multiline)
1611 last = memEQ(strend - len, little, len)
1612 ? strend - len : Nullch;
1618 last = rninstr(s, strend, little, little + len);
1620 last = strend; /* matching `$' */
1624 DEBUG_r(PerlIO_printf(Perl_debug_log,
1625 "%sCan't trim the tail, match fails (should not happen)%s\n",
1626 PL_colors[4],PL_colors[5]));
1627 goto phooey; /* Should not happen! */
1629 dontbother = strend - last + prog->float_min_offset;
1631 if (minlen && (dontbother < minlen))
1632 dontbother = minlen - 1;
1633 strend -= dontbother; /* this one's always in bytes! */
1634 /* We don't know much -- general case. */
1637 if (regtry(prog, s))
1646 if (regtry(prog, s))
1648 } while (s++ < strend);
1656 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1658 if (PL_reg_eval_set) {
1659 /* Preserve the current value of $^R */
1660 if (oreplsv != GvSV(PL_replgv))
1661 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1662 restored, the value remains
1664 restore_pos(aTHXo_ 0);
1667 /* make sure $`, $&, $', and $digit will work later */
1668 if ( !(flags & REXEC_NOT_FIRST) ) {
1669 if (RX_MATCH_COPIED(prog)) {
1670 Safefree(prog->subbeg);
1671 RX_MATCH_COPIED_off(prog);
1673 if (flags & REXEC_COPY_STR) {
1674 I32 i = PL_regeol - startpos + (stringarg - strbeg);
1676 s = savepvn(strbeg, i);
1679 RX_MATCH_COPIED_on(prog);
1682 prog->subbeg = strbeg;
1683 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
1690 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
1691 PL_colors[4],PL_colors[5]));
1692 if (PL_reg_eval_set)
1693 restore_pos(aTHXo_ 0);
1698 - regtry - try match at specific point
1700 STATIC I32 /* 0 failure, 1 success */
1701 S_regtry(pTHX_ regexp *prog, char *startpos)
1710 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
1712 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
1715 PL_reg_eval_set = RS_init;
1717 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
1718 (IV)(PL_stack_sp - PL_stack_base));
1720 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
1721 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
1722 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
1724 /* Apparently this is not needed, judging by wantarray. */
1725 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
1726 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
1729 /* Make $_ available to executed code. */
1730 if (PL_reg_sv != DEFSV) {
1731 /* SAVE_DEFSV does *not* suffice here for USE_THREADS */
1736 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
1737 && (mg = mg_find(PL_reg_sv, 'g')))) {
1738 /* prepare for quick setting of pos */
1739 sv_magic(PL_reg_sv, (SV*)0, 'g', Nullch, 0);
1740 mg = mg_find(PL_reg_sv, 'g');
1744 PL_reg_oldpos = mg->mg_len;
1745 SAVEDESTRUCTOR_X(restore_pos, 0);
1748 Newz(22,PL_reg_curpm, 1, PMOP);
1749 PL_reg_curpm->op_pmregexp = prog;
1750 PL_reg_oldcurpm = PL_curpm;
1751 PL_curpm = PL_reg_curpm;
1752 if (RX_MATCH_COPIED(prog)) {
1753 /* Here is a serious problem: we cannot rewrite subbeg,
1754 since it may be needed if this match fails. Thus
1755 $` inside (?{}) could fail... */
1756 PL_reg_oldsaved = prog->subbeg;
1757 PL_reg_oldsavedlen = prog->sublen;
1758 RX_MATCH_COPIED_off(prog);
1761 PL_reg_oldsaved = Nullch;
1762 prog->subbeg = PL_bostr;
1763 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
1765 prog->startp[0] = startpos - PL_bostr;
1766 PL_reginput = startpos;
1767 PL_regstartp = prog->startp;
1768 PL_regendp = prog->endp;
1769 PL_reglastparen = &prog->lastparen;
1770 prog->lastparen = 0;
1772 DEBUG_r(PL_reg_starttry = startpos);
1773 if (PL_reg_start_tmpl <= prog->nparens) {
1774 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
1775 if(PL_reg_start_tmp)
1776 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1778 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1781 /* XXXX What this code is doing here?!!! There should be no need
1782 to do this again and again, PL_reglastparen should take care of
1786 if (prog->nparens) {
1787 for (i = prog->nparens; i >= 1; i--) {
1793 if (regmatch(prog->program + 1)) {
1794 prog->endp[0] = PL_reginput - PL_bostr;
1797 REGCP_UNWIND(lastcp);
1801 #define RE_UNWIND_BRANCH 1
1802 #define RE_UNWIND_BRANCHJ 2
1806 typedef struct { /* XX: makes sense to enlarge it... */
1810 } re_unwind_generic_t;
1823 } re_unwind_branch_t;
1825 typedef union re_unwind_t {
1827 re_unwind_generic_t generic;
1828 re_unwind_branch_t branch;
1832 - regmatch - main matching routine
1834 * Conceptually the strategy is simple: check to see whether the current
1835 * node matches, call self recursively to see whether the rest matches,
1836 * and then act accordingly. In practice we make some effort to avoid
1837 * recursion, in particular by going through "ordinary" nodes (that don't
1838 * need to know whether the rest of the match failed) by a loop instead of
1841 /* [lwall] I've hoisted the register declarations to the outer block in order to
1842 * maybe save a little bit of pushing and popping on the stack. It also takes
1843 * advantage of machines that use a register save mask on subroutine entry.
1845 STATIC I32 /* 0 failure, 1 success */
1846 S_regmatch(pTHX_ regnode *prog)
1849 register regnode *scan; /* Current node. */
1850 regnode *next; /* Next node. */
1851 regnode *inner; /* Next node in internal branch. */
1852 register I32 nextchr; /* renamed nextchr - nextchar colides with
1853 function of same name */
1854 register I32 n; /* no or next */
1855 register I32 ln; /* len or last */
1856 register char *s; /* operand or save */
1857 register char *locinput = PL_reginput;
1858 register I32 c1, c2, paren; /* case fold search, parenth */
1859 int minmod = 0, sw = 0, logical = 0;
1861 I32 firstcp = PL_savestack_ix;
1867 /* Note that nextchr is a byte even in UTF */
1868 nextchr = UCHARAT(locinput);
1870 while (scan != NULL) {
1871 #define sayNO_L (logical ? (logical = 0, sw = 0, goto cont) : sayNO)
1873 # define sayYES goto yes
1874 # define sayNO goto no
1875 # define sayYES_FINAL goto yes_final
1876 # define sayYES_LOUD goto yes_loud
1877 # define sayNO_FINAL goto no_final
1878 # define sayNO_SILENT goto do_no
1879 # define saySAME(x) if (x) goto yes; else goto no
1880 # define REPORT_CODE_OFF 24
1882 # define sayYES return 1
1883 # define sayNO return 0
1884 # define sayYES_FINAL return 1
1885 # define sayYES_LOUD return 1
1886 # define sayNO_FINAL return 0
1887 # define sayNO_SILENT return 0
1888 # define saySAME(x) return x
1891 SV *prop = sv_newmortal();
1892 int docolor = *PL_colors[0];
1893 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
1894 int l = (PL_regeol - locinput > taill ? taill : PL_regeol - locinput);
1895 /* The part of the string before starttry has one color
1896 (pref0_len chars), between starttry and current
1897 position another one (pref_len - pref0_len chars),
1898 after the current position the third one.
1899 We assume that pref0_len <= pref_len, otherwise we
1900 decrease pref0_len. */
1901 int pref_len = (locinput - PL_bostr > (5 + taill) - l
1902 ? (5 + taill) - l : locinput - PL_bostr);
1903 int pref0_len = pref_len - (locinput - PL_reg_starttry);
1905 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
1906 l = ( PL_regeol - locinput > (5 + taill) - pref_len
1907 ? (5 + taill) - pref_len : PL_regeol - locinput);
1910 if (pref0_len > pref_len)
1911 pref0_len = pref_len;
1912 regprop(prop, scan);
1913 PerlIO_printf(Perl_debug_log,
1914 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
1915 (IV)(locinput - PL_bostr),
1916 PL_colors[4], pref0_len,
1917 locinput - pref_len, PL_colors[5],
1918 PL_colors[2], pref_len - pref0_len,
1919 locinput - pref_len + pref0_len, PL_colors[3],
1920 (docolor ? "" : "> <"),
1921 PL_colors[0], l, locinput, PL_colors[1],
1922 15 - l - pref_len + 1,
1924 (IV)(scan - PL_regprogram), PL_regindent*2, "",
1928 next = scan + NEXT_OFF(scan);
1934 if (locinput == PL_bostr
1935 ? PL_regprev == '\n'
1937 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
1939 /* regtill = regbol; */
1944 if (locinput == PL_bostr
1945 ? PL_regprev == '\n'
1946 : ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
1952 if (locinput == PL_bostr)
1956 if (locinput == PL_reg_ganch)
1966 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
1971 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
1973 if (PL_regeol - locinput > 1)
1977 if (PL_regeol != locinput)
1981 if (nextchr & 0x80) {
1982 locinput += PL_utf8skip[nextchr];
1983 if (locinput > PL_regeol)
1985 nextchr = UCHARAT(locinput);
1988 if (!nextchr && locinput >= PL_regeol)
1990 nextchr = UCHARAT(++locinput);
1993 if (!nextchr && locinput >= PL_regeol)
1995 nextchr = UCHARAT(++locinput);
1998 if (nextchr & 0x80) {
1999 locinput += PL_utf8skip[nextchr];
2000 if (locinput > PL_regeol)
2002 nextchr = UCHARAT(locinput);
2005 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2007 nextchr = UCHARAT(++locinput);
2010 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2012 nextchr = UCHARAT(++locinput);
2017 /* Inline the first character, for speed. */
2018 if (UCHARAT(s) != nextchr)
2020 if (PL_regeol - locinput < ln)
2022 if (ln > 1 && memNE(s, locinput, ln))
2025 nextchr = UCHARAT(locinput);
2028 PL_reg_flags |= RF_tainted;
2037 c1 = OP(scan) == EXACTF;
2041 if (utf8_to_uv((U8*)s, e - s, 0, 0) != (c1 ?
2042 toLOWER_utf8((U8*)l) :
2043 toLOWER_LC_utf8((U8*)l)))
2051 nextchr = UCHARAT(locinput);
2055 /* Inline the first character, for speed. */
2056 if (UCHARAT(s) != nextchr &&
2057 UCHARAT(s) != ((OP(scan) == EXACTF)
2058 ? PL_fold : PL_fold_locale)[nextchr])
2060 if (PL_regeol - locinput < ln)
2062 if (ln > 1 && (OP(scan) == EXACTF
2063 ? ibcmp(s, locinput, ln)
2064 : ibcmp_locale(s, locinput, ln)))
2067 nextchr = UCHARAT(locinput);
2070 if (!REGINCLASSUTF8(scan, (U8*)locinput))
2072 if (locinput >= PL_regeol)
2074 locinput += PL_utf8skip[nextchr];
2075 nextchr = UCHARAT(locinput);
2079 nextchr = UCHARAT(locinput);
2080 if (!REGINCLASS(scan, nextchr))
2082 if (!nextchr && locinput >= PL_regeol)
2084 nextchr = UCHARAT(++locinput);
2087 PL_reg_flags |= RF_tainted;
2092 if (!(OP(scan) == ALNUM
2093 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2095 nextchr = UCHARAT(++locinput);
2098 PL_reg_flags |= RF_tainted;
2103 if (nextchr & 0x80) {
2104 if (!(OP(scan) == ALNUMUTF8
2105 ? swash_fetch(PL_utf8_alnum, (U8*)locinput)
2106 : isALNUM_LC_utf8((U8*)locinput)))
2110 locinput += PL_utf8skip[nextchr];
2111 nextchr = UCHARAT(locinput);
2114 if (!(OP(scan) == ALNUMUTF8
2115 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2117 nextchr = UCHARAT(++locinput);
2120 PL_reg_flags |= RF_tainted;
2123 if (!nextchr && locinput >= PL_regeol)
2125 if (OP(scan) == NALNUM
2126 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2128 nextchr = UCHARAT(++locinput);
2131 PL_reg_flags |= RF_tainted;
2134 if (!nextchr && locinput >= PL_regeol)
2136 if (nextchr & 0x80) {
2137 if (OP(scan) == NALNUMUTF8
2138 ? swash_fetch(PL_utf8_alnum, (U8*)locinput)
2139 : isALNUM_LC_utf8((U8*)locinput))
2143 locinput += PL_utf8skip[nextchr];
2144 nextchr = UCHARAT(locinput);
2147 if (OP(scan) == NALNUMUTF8
2148 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2150 nextchr = UCHARAT(++locinput);
2154 PL_reg_flags |= RF_tainted;
2158 /* was last char in word? */
2159 ln = (locinput != PL_regbol) ? UCHARAT(locinput - 1) : PL_regprev;
2160 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2162 n = isALNUM(nextchr);
2165 ln = isALNUM_LC(ln);
2166 n = isALNUM_LC(nextchr);
2168 if (((!ln) == (!n)) == (OP(scan) == BOUND || OP(scan) == BOUNDL))
2173 PL_reg_flags |= RF_tainted;
2177 /* was last char in word? */
2178 ln = (locinput != PL_regbol)
2179 ? utf8_to_uv(reghop((U8*)locinput, -1),
2180 PL_regeol - locinput, 0, 0) : PL_regprev;
2181 if (OP(scan) == BOUNDUTF8 || OP(scan) == NBOUNDUTF8) {
2182 ln = isALNUM_uni(ln);
2183 n = swash_fetch(PL_utf8_alnum, (U8*)locinput);
2186 ln = isALNUM_LC_uni(ln);
2187 n = isALNUM_LC_utf8((U8*)locinput);
2189 if (((!ln) == (!n)) == (OP(scan) == BOUNDUTF8 || OP(scan) == BOUNDLUTF8))
2193 PL_reg_flags |= RF_tainted;
2198 if (!(OP(scan) == SPACE
2199 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2201 nextchr = UCHARAT(++locinput);
2204 PL_reg_flags |= RF_tainted;
2209 if (nextchr & 0x80) {
2210 if (!(OP(scan) == SPACEUTF8
2211 ? swash_fetch(PL_utf8_space, (U8*)locinput)
2212 : isSPACE_LC_utf8((U8*)locinput)))
2216 locinput += PL_utf8skip[nextchr];
2217 nextchr = UCHARAT(locinput);
2220 if (!(OP(scan) == SPACEUTF8
2221 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2223 nextchr = UCHARAT(++locinput);
2226 PL_reg_flags |= RF_tainted;
2229 if (!nextchr && locinput >= PL_regeol)
2231 if (OP(scan) == NSPACE
2232 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2234 nextchr = UCHARAT(++locinput);
2237 PL_reg_flags |= RF_tainted;
2240 if (!nextchr && locinput >= PL_regeol)
2242 if (nextchr & 0x80) {
2243 if (OP(scan) == NSPACEUTF8
2244 ? swash_fetch(PL_utf8_space, (U8*)locinput)
2245 : isSPACE_LC_utf8((U8*)locinput))
2249 locinput += PL_utf8skip[nextchr];
2250 nextchr = UCHARAT(locinput);
2253 if (OP(scan) == NSPACEUTF8
2254 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2256 nextchr = UCHARAT(++locinput);
2259 PL_reg_flags |= RF_tainted;
2264 if (!(OP(scan) == DIGIT
2265 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2267 nextchr = UCHARAT(++locinput);
2270 PL_reg_flags |= RF_tainted;
2275 if (nextchr & 0x80) {
2276 if (!(OP(scan) == DIGITUTF8
2277 ? swash_fetch(PL_utf8_digit, (U8*)locinput)
2278 : isDIGIT_LC_utf8((U8*)locinput)))
2282 locinput += PL_utf8skip[nextchr];
2283 nextchr = UCHARAT(locinput);
2286 if (!(OP(scan) == DIGITUTF8
2287 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2289 nextchr = UCHARAT(++locinput);
2292 PL_reg_flags |= RF_tainted;
2295 if (!nextchr && locinput >= PL_regeol)
2297 if (OP(scan) == NDIGIT
2298 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2300 nextchr = UCHARAT(++locinput);
2303 PL_reg_flags |= RF_tainted;
2306 if (!nextchr && locinput >= PL_regeol)
2308 if (nextchr & 0x80) {
2309 if (OP(scan) == NDIGITUTF8
2310 ? swash_fetch(PL_utf8_digit, (U8*)locinput)
2311 : isDIGIT_LC_utf8((U8*)locinput))
2315 locinput += PL_utf8skip[nextchr];
2316 nextchr = UCHARAT(locinput);
2319 if (OP(scan) == NDIGITUTF8
2320 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2322 nextchr = UCHARAT(++locinput);
2325 if (locinput >= PL_regeol || swash_fetch(PL_utf8_mark,(U8*)locinput))
2327 locinput += PL_utf8skip[nextchr];
2328 while (locinput < PL_regeol && swash_fetch(PL_utf8_mark,(U8*)locinput))
2329 locinput += UTF8SKIP(locinput);
2330 if (locinput > PL_regeol)
2332 nextchr = UCHARAT(locinput);
2335 PL_reg_flags |= RF_tainted;
2339 n = ARG(scan); /* which paren pair */
2340 ln = PL_regstartp[n];
2341 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2342 if (*PL_reglastparen < n || ln == -1)
2343 sayNO; /* Do not match unless seen CLOSEn. */
2344 if (ln == PL_regendp[n])
2348 if (UTF && OP(scan) != REF) { /* REF can do byte comparison */
2350 char *e = PL_bostr + PL_regendp[n];
2352 * Note that we can't do the "other character" lookup trick as
2353 * in the 8-bit case (no pun intended) because in Unicode we
2354 * have to map both upper and title case to lower case.
2356 if (OP(scan) == REFF) {
2360 if (toLOWER_utf8((U8*)s) != toLOWER_utf8((U8*)l))
2370 if (toLOWER_LC_utf8((U8*)s) != toLOWER_LC_utf8((U8*)l))
2377 nextchr = UCHARAT(locinput);
2381 /* Inline the first character, for speed. */
2382 if (UCHARAT(s) != nextchr &&
2384 (UCHARAT(s) != ((OP(scan) == REFF
2385 ? PL_fold : PL_fold_locale)[nextchr]))))
2387 ln = PL_regendp[n] - ln;
2388 if (locinput + ln > PL_regeol)
2390 if (ln > 1 && (OP(scan) == REF
2391 ? memNE(s, locinput, ln)
2393 ? ibcmp(s, locinput, ln)
2394 : ibcmp_locale(s, locinput, ln))))
2397 nextchr = UCHARAT(locinput);
2408 OP_4tree *oop = PL_op;
2409 COP *ocurcop = PL_curcop;
2410 SV **ocurpad = PL_curpad;
2414 PL_op = (OP_4tree*)PL_regdata->data[n];
2415 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2416 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2417 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2419 CALLRUNOPS(aTHX); /* Scalar context. */
2425 PL_curpad = ocurpad;
2426 PL_curcop = ocurcop;
2428 if (logical == 2) { /* Postponed subexpression. */
2430 MAGIC *mg = Null(MAGIC*);
2432 CHECKPOINT cp, lastcp;
2434 if(SvROK(ret) || SvRMAGICAL(ret)) {
2435 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2438 mg = mg_find(sv, 'r');
2441 re = (regexp *)mg->mg_obj;
2442 (void)ReREFCNT_inc(re);
2446 char *t = SvPV(ret, len);
2448 char *oprecomp = PL_regprecomp;
2449 I32 osize = PL_regsize;
2450 I32 onpar = PL_regnpar;
2453 pm.op_pmdynflags = (UTF||DO_UTF8(ret) ? PMdf_UTF8 : 0);
2454 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2456 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2457 sv_magic(ret,(SV*)ReREFCNT_inc(re),'r',0,0);
2458 PL_regprecomp = oprecomp;
2463 PerlIO_printf(Perl_debug_log,
2464 "Entering embedded `%s%.60s%s%s'\n",
2468 (strlen(re->precomp) > 60 ? "..." : ""))
2471 state.prev = PL_reg_call_cc;
2472 state.cc = PL_regcc;
2473 state.re = PL_reg_re;
2477 cp = regcppush(0); /* Save *all* the positions. */
2480 state.ss = PL_savestack_ix;
2481 *PL_reglastparen = 0;
2482 PL_reg_call_cc = &state;
2483 PL_reginput = locinput;
2485 /* XXXX This is too dramatic a measure... */
2488 if (regmatch(re->program + 1)) {
2489 /* Even though we succeeded, we need to restore
2490 global variables, since we may be wrapped inside
2491 SUSPEND, thus the match may be not finished yet. */
2493 /* XXXX Do this only if SUSPENDed? */
2494 PL_reg_call_cc = state.prev;
2495 PL_regcc = state.cc;
2496 PL_reg_re = state.re;
2497 cache_re(PL_reg_re);
2499 /* XXXX This is too dramatic a measure... */
2502 /* These are needed even if not SUSPEND. */
2508 REGCP_UNWIND(lastcp);
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... */
2524 sv_setsv(save_scalar(PL_replgv), ret);
2528 n = ARG(scan); /* which paren pair */
2529 PL_reg_start_tmp[n] = locinput;
2534 n = ARG(scan); /* which paren pair */
2535 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2536 PL_regendp[n] = locinput - PL_bostr;
2537 if (n > *PL_reglastparen)
2538 *PL_reglastparen = n;
2541 n = ARG(scan); /* which paren pair */
2542 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2545 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2547 next = NEXTOPER(NEXTOPER(scan));
2549 next = scan + ARG(scan);
2550 if (OP(next) == IFTHEN) /* Fake one. */
2551 next = NEXTOPER(NEXTOPER(next));
2555 logical = scan->flags;
2557 /*******************************************************************
2558 PL_regcc contains infoblock about the innermost (...)* loop, and
2559 a pointer to the next outer infoblock.
2561 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2563 1) After matching X, regnode for CURLYX is processed;
2565 2) This regnode creates infoblock on the stack, and calls
2566 regmatch() recursively with the starting point at WHILEM node;
2568 3) Each hit of WHILEM node tries to match A and Z (in the order
2569 depending on the current iteration, min/max of {min,max} and
2570 greediness). The information about where are nodes for "A"
2571 and "Z" is read from the infoblock, as is info on how many times "A"
2572 was already matched, and greediness.
2574 4) After A matches, the same WHILEM node is hit again.
2576 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2577 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2578 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2579 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2580 of the external loop.
2582 Currently present infoblocks form a tree with a stem formed by PL_curcc
2583 and whatever it mentions via ->next, and additional attached trees
2584 corresponding to temporarily unset infoblocks as in "5" above.
2586 In the following picture infoblocks for outer loop of
2587 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2588 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2589 infoblocks are drawn below the "reset" infoblock.
2591 In fact in the picture below we do not show failed matches for Z and T
2592 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2593 more obvious *why* one needs to *temporary* unset infoblocks.]
2595 Matched REx position InfoBlocks Comment
2599 Y A)*?Z)*?T x <- O <- I
2600 YA )*?Z)*?T x <- O <- I
2601 YA A)*?Z)*?T x <- O <- I
2602 YAA )*?Z)*?T x <- O <- I
2603 YAA Z)*?T x <- O # Temporary unset I
2606 YAAZ Y(A)*?Z)*?T x <- O
2609 YAAZY (A)*?Z)*?T x <- O
2612 YAAZY A)*?Z)*?T x <- O <- I
2615 YAAZYA )*?Z)*?T x <- O <- I
2618 YAAZYA Z)*?T x <- O # Temporary unset I
2624 YAAZYAZ T x # Temporary unset O
2631 *******************************************************************/
2634 CHECKPOINT cp = PL_savestack_ix;
2635 /* No need to save/restore up to this paren */
2636 I32 parenfloor = scan->flags;
2638 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2640 cc.oldcc = PL_regcc;
2642 /* XXXX Probably it is better to teach regpush to support
2643 parenfloor > PL_regsize... */
2644 if (parenfloor > *PL_reglastparen)
2645 parenfloor = *PL_reglastparen; /* Pessimization... */
2646 cc.parenfloor = parenfloor;
2648 cc.min = ARG1(scan);
2649 cc.max = ARG2(scan);
2650 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2654 PL_reginput = locinput;
2655 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2657 PL_regcc = cc.oldcc;
2663 * This is really hard to understand, because after we match
2664 * what we're trying to match, we must make sure the rest of
2665 * the REx is going to match for sure, and to do that we have
2666 * to go back UP the parse tree by recursing ever deeper. And
2667 * if it fails, we have to reset our parent's current state
2668 * that we can try again after backing off.
2671 CHECKPOINT cp, lastcp;
2672 CURCUR* cc = PL_regcc;
2673 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2675 n = cc->cur + 1; /* how many we know we matched */
2676 PL_reginput = locinput;
2679 PerlIO_printf(Perl_debug_log,
2680 "%*s %ld out of %ld..%ld cc=%lx\n",
2681 REPORT_CODE_OFF+PL_regindent*2, "",
2682 (long)n, (long)cc->min,
2683 (long)cc->max, (long)cc)
2686 /* If degenerate scan matches "", assume scan done. */
2688 if (locinput == cc->lastloc && n >= cc->min) {
2689 PL_regcc = cc->oldcc;
2693 PerlIO_printf(Perl_debug_log,
2694 "%*s empty match detected, try continuation...\n",
2695 REPORT_CODE_OFF+PL_regindent*2, "")
2697 if (regmatch(cc->next))
2705 /* First just match a string of min scans. */
2709 cc->lastloc = locinput;
2710 if (regmatch(cc->scan))
2713 cc->lastloc = lastloc;
2718 /* Check whether we already were at this position.
2719 Postpone detection until we know the match is not
2720 *that* much linear. */
2721 if (!PL_reg_maxiter) {
2722 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
2723 PL_reg_leftiter = PL_reg_maxiter;
2725 if (PL_reg_leftiter-- == 0) {
2726 I32 size = (PL_reg_maxiter + 7)/8;
2727 if (PL_reg_poscache) {
2728 if (PL_reg_poscache_size < size) {
2729 Renew(PL_reg_poscache, size, char);
2730 PL_reg_poscache_size = size;
2732 Zero(PL_reg_poscache, size, char);
2735 PL_reg_poscache_size = size;
2736 Newz(29, PL_reg_poscache, size, char);
2739 PerlIO_printf(Perl_debug_log,
2740 "%sDetected a super-linear match, switching on caching%s...\n",
2741 PL_colors[4], PL_colors[5])
2744 if (PL_reg_leftiter < 0) {
2745 I32 o = locinput - PL_bostr, b;
2747 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
2750 if (PL_reg_poscache[o] & (1<<b)) {
2752 PerlIO_printf(Perl_debug_log,
2753 "%*s already tried at this position...\n",
2754 REPORT_CODE_OFF+PL_regindent*2, "")
2758 PL_reg_poscache[o] |= (1<<b);
2762 /* Prefer next over scan for minimal matching. */
2765 PL_regcc = cc->oldcc;
2768 cp = regcppush(cc->parenfloor);
2770 if (regmatch(cc->next)) {
2772 sayYES; /* All done. */
2774 REGCP_UNWIND(lastcp);
2780 if (n >= cc->max) { /* Maximum greed exceeded? */
2781 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2782 && !(PL_reg_flags & RF_warned)) {
2783 PL_reg_flags |= RF_warned;
2784 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2785 "Complex regular subexpression recursion",
2792 PerlIO_printf(Perl_debug_log,
2793 "%*s trying longer...\n",
2794 REPORT_CODE_OFF+PL_regindent*2, "")
2796 /* Try scanning more and see if it helps. */
2797 PL_reginput = locinput;
2799 cc->lastloc = locinput;
2800 cp = regcppush(cc->parenfloor);
2802 if (regmatch(cc->scan)) {
2806 REGCP_UNWIND(lastcp);
2809 cc->lastloc = lastloc;
2813 /* Prefer scan over next for maximal matching. */
2815 if (n < cc->max) { /* More greed allowed? */
2816 cp = regcppush(cc->parenfloor);
2818 cc->lastloc = locinput;
2820 if (regmatch(cc->scan)) {
2824 REGCP_UNWIND(lastcp);
2825 regcppop(); /* Restore some previous $<digit>s? */
2826 PL_reginput = locinput;
2828 PerlIO_printf(Perl_debug_log,
2829 "%*s failed, try continuation...\n",
2830 REPORT_CODE_OFF+PL_regindent*2, "")
2833 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2834 && !(PL_reg_flags & RF_warned)) {
2835 PL_reg_flags |= RF_warned;
2836 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2837 "Complex regular subexpression recursion",
2841 /* Failed deeper matches of scan, so see if this one works. */
2842 PL_regcc = cc->oldcc;
2845 if (regmatch(cc->next))
2851 cc->lastloc = lastloc;
2856 next = scan + ARG(scan);
2859 inner = NEXTOPER(NEXTOPER(scan));
2862 inner = NEXTOPER(scan);
2867 if (OP(next) != c1) /* No choice. */
2868 next = inner; /* Avoid recursion. */
2870 I32 lastparen = *PL_reglastparen;
2872 re_unwind_branch_t *uw;
2874 /* Put unwinding data on stack */
2875 unwind1 = SSNEWt(1,re_unwind_branch_t);
2876 uw = SSPTRt(unwind1,re_unwind_branch_t);
2879 uw->type = ((c1 == BRANCH)
2881 : RE_UNWIND_BRANCHJ);
2882 uw->lastparen = lastparen;
2884 uw->locinput = locinput;
2885 uw->nextchr = nextchr;
2887 uw->regindent = ++PL_regindent;
2890 REGCP_SET(uw->lastcp);
2892 /* Now go into the first branch */
2905 /* We suppose that the next guy does not need
2906 backtracking: in particular, it is of constant length,
2907 and has no parenths to influence future backrefs. */
2908 ln = ARG1(scan); /* min to match */
2909 n = ARG2(scan); /* max to match */
2910 paren = scan->flags;
2912 if (paren > PL_regsize)
2914 if (paren > *PL_reglastparen)
2915 *PL_reglastparen = paren;
2917 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
2919 scan += NEXT_OFF(scan); /* Skip former OPEN. */
2920 PL_reginput = locinput;
2923 if (ln && regrepeat_hard(scan, ln, &l) < ln)
2925 if (ln && l == 0 && n >= ln
2926 /* In fact, this is tricky. If paren, then the
2927 fact that we did/didnot match may influence
2928 future execution. */
2929 && !(paren && ln == 0))
2931 locinput = PL_reginput;
2932 if (PL_regkind[(U8)OP(next)] == EXACT) {
2933 c1 = (U8)*STRING(next);
2934 if (OP(next) == EXACTF)
2936 else if (OP(next) == EXACTFL)
2937 c2 = PL_fold_locale[c1];
2944 /* This may be improved if l == 0. */
2945 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
2946 /* If it could work, try it. */
2948 UCHARAT(PL_reginput) == c1 ||
2949 UCHARAT(PL_reginput) == c2)
2953 PL_regstartp[paren] =
2954 HOPc(PL_reginput, -l) - PL_bostr;
2955 PL_regendp[paren] = PL_reginput - PL_bostr;
2958 PL_regendp[paren] = -1;
2962 REGCP_UNWIND(lastcp);
2964 /* Couldn't or didn't -- move forward. */
2965 PL_reginput = locinput;
2966 if (regrepeat_hard(scan, 1, &l)) {
2968 locinput = PL_reginput;
2975 n = regrepeat_hard(scan, n, &l);
2976 if (n != 0 && l == 0
2977 /* In fact, this is tricky. If paren, then the
2978 fact that we did/didnot match may influence
2979 future execution. */
2980 && !(paren && ln == 0))
2982 locinput = PL_reginput;
2984 PerlIO_printf(Perl_debug_log,
2985 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
2986 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
2990 if (PL_regkind[(U8)OP(next)] == EXACT) {
2991 c1 = (U8)*STRING(next);
2992 if (OP(next) == EXACTF)
2994 else if (OP(next) == EXACTFL)
2995 c2 = PL_fold_locale[c1];
3004 /* If it could work, try it. */
3006 UCHARAT(PL_reginput) == c1 ||
3007 UCHARAT(PL_reginput) == c2)
3010 PerlIO_printf(Perl_debug_log,
3011 "%*s trying tail with n=%"IVdf"...\n",
3012 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3016 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3017 PL_regendp[paren] = PL_reginput - PL_bostr;
3020 PL_regendp[paren] = -1;
3024 REGCP_UNWIND(lastcp);
3026 /* Couldn't or didn't -- back up. */
3028 locinput = HOPc(locinput, -l);
3029 PL_reginput = locinput;
3036 paren = scan->flags; /* Which paren to set */
3037 if (paren > PL_regsize)
3039 if (paren > *PL_reglastparen)
3040 *PL_reglastparen = paren;
3041 ln = ARG1(scan); /* min to match */
3042 n = ARG2(scan); /* max to match */
3043 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3047 ln = ARG1(scan); /* min to match */
3048 n = ARG2(scan); /* max to match */
3049 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3054 scan = NEXTOPER(scan);
3060 scan = NEXTOPER(scan);
3064 * Lookahead to avoid useless match attempts
3065 * when we know what character comes next.
3067 if (PL_regkind[(U8)OP(next)] == EXACT) {
3068 c1 = (U8)*STRING(next);
3069 if (OP(next) == EXACTF)
3071 else if (OP(next) == EXACTFL)
3072 c2 = PL_fold_locale[c1];
3078 PL_reginput = locinput;
3082 if (ln && regrepeat(scan, ln) < ln)
3084 locinput = PL_reginput;
3087 char *e = locinput + n - ln; /* Should not check after this */
3088 char *old = locinput;
3090 if (e >= PL_regeol || (n == REG_INFTY))
3093 /* Find place 'next' could work */
3095 while (locinput <= e && *locinput != c1)
3098 while (locinput <= e
3105 /* PL_reginput == old now */
3106 if (locinput != old) {
3107 ln = 1; /* Did some */
3108 if (regrepeat(scan, locinput - old) <
3112 /* PL_reginput == locinput now */
3113 TRYPAREN(paren, ln, locinput);
3114 PL_reginput = locinput; /* Could be reset... */
3115 REGCP_UNWIND(lastcp);
3116 /* Couldn't or didn't -- move forward. */
3121 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3122 /* If it could work, try it. */
3124 UCHARAT(PL_reginput) == c1 ||
3125 UCHARAT(PL_reginput) == c2)
3127 TRYPAREN(paren, n, PL_reginput);
3128 REGCP_UNWIND(lastcp);
3130 /* Couldn't or didn't -- move forward. */
3131 PL_reginput = locinput;
3132 if (regrepeat(scan, 1)) {
3134 locinput = PL_reginput;
3142 n = regrepeat(scan, n);
3143 locinput = PL_reginput;
3144 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3145 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3146 ln = n; /* why back off? */
3147 /* ...because $ and \Z can match before *and* after
3148 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3149 We should back off by one in this case. */
3150 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3156 /* If it could work, try it. */
3158 UCHARAT(PL_reginput) == c1 ||
3159 UCHARAT(PL_reginput) == c2)
3161 TRYPAREN(paren, n, PL_reginput);
3162 REGCP_UNWIND(lastcp);
3164 /* Couldn't or didn't -- back up. */
3166 PL_reginput = locinput = HOPc(locinput, -1);
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);
3188 if (PL_reg_call_cc) {
3189 re_cc_state *cur_call_cc = PL_reg_call_cc;
3190 CURCUR *cctmp = PL_regcc;
3191 regexp *re = PL_reg_re;
3192 CHECKPOINT cp, lastcp;
3194 cp = regcppush(0); /* Save *all* the positions. */
3196 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3198 PL_reginput = locinput; /* Make position available to
3200 cache_re(PL_reg_call_cc->re);
3201 PL_regcc = PL_reg_call_cc->cc;
3202 PL_reg_call_cc = PL_reg_call_cc->prev;
3203 if (regmatch(cur_call_cc->node)) {
3204 PL_reg_call_cc = cur_call_cc;
3208 REGCP_UNWIND(lastcp);
3210 PL_reg_call_cc = cur_call_cc;
3216 PerlIO_printf(Perl_debug_log,
3217 "%*s continuation failed...\n",
3218 REPORT_CODE_OFF+PL_regindent*2, "")
3222 if (locinput < PL_regtill) {
3223 DEBUG_r(PerlIO_printf(Perl_debug_log,
3224 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3226 (long)(locinput - PL_reg_starttry),
3227 (long)(PL_regtill - PL_reg_starttry),
3229 sayNO_FINAL; /* Cannot match: too short. */
3231 PL_reginput = locinput; /* put where regtry can find it */
3232 sayYES_FINAL; /* Success! */
3234 PL_reginput = locinput; /* put where regtry can find it */
3235 sayYES_LOUD; /* Success! */
3238 PL_reginput = locinput;
3243 if (UTF) { /* XXXX This is absolutely
3244 broken, we read before
3246 s = HOPMAYBEc(locinput, -scan->flags);
3252 if (locinput < PL_bostr + scan->flags)
3254 PL_reginput = locinput - scan->flags;
3259 PL_reginput = locinput;
3264 if (UTF) { /* XXXX This is absolutely
3265 broken, we read before
3267 s = HOPMAYBEc(locinput, -scan->flags);
3268 if (!s || s < PL_bostr)
3273 if (locinput < PL_bostr + scan->flags)
3275 PL_reginput = locinput - scan->flags;
3280 PL_reginput = locinput;
3283 inner = NEXTOPER(NEXTOPER(scan));
3284 if (regmatch(inner) != n) {
3299 if (OP(scan) == SUSPEND) {
3300 locinput = PL_reginput;
3301 nextchr = UCHARAT(locinput);
3306 next = scan + ARG(scan);
3311 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3312 PTR2UV(scan), OP(scan));
3313 Perl_croak(aTHX_ "regexp memory corruption");
3320 * We get here only if there's trouble -- normally "case END" is
3321 * the terminating point.
3323 Perl_croak(aTHX_ "corrupted regexp pointers");
3329 PerlIO_printf(Perl_debug_log,
3330 "%*s %scould match...%s\n",
3331 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3335 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3336 PL_colors[4],PL_colors[5]));
3342 #if 0 /* Breaks $^R */
3350 PerlIO_printf(Perl_debug_log,
3351 "%*s %sfailed...%s\n",
3352 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3358 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3361 case RE_UNWIND_BRANCH:
3362 case RE_UNWIND_BRANCHJ:
3364 re_unwind_branch_t *uwb = &(uw->branch);
3365 I32 lastparen = uwb->lastparen;
3367 REGCP_UNWIND(uwb->lastcp);
3368 for (n = *PL_reglastparen; n > lastparen; n--)
3370 *PL_reglastparen = n;
3371 scan = next = uwb->next;
3373 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3374 ? BRANCH : BRANCHJ) ) { /* Failure */
3381 /* Have more choice yet. Reuse the same uwb. */
3383 if ((n = (uwb->type == RE_UNWIND_BRANCH
3384 ? NEXT_OFF(next) : ARG(next))))
3387 next = NULL; /* XXXX Needn't unwinding in this case... */
3389 next = NEXTOPER(scan);
3390 if (uwb->type == RE_UNWIND_BRANCHJ)
3391 next = NEXTOPER(next);
3392 locinput = uwb->locinput;
3393 nextchr = uwb->nextchr;
3395 PL_regindent = uwb->regindent;
3402 Perl_croak(aTHX_ "regexp unwind memory corruption");
3413 - regrepeat - repeatedly match something simple, report how many
3416 * [This routine now assumes that it will only match on things of length 1.
3417 * That was true before, but now we assume scan - reginput is the count,
3418 * rather than incrementing count on every character. [Er, except utf8.]]
3421 S_regrepeat(pTHX_ regnode *p, I32 max)
3424 register char *scan;
3426 register char *loceol = PL_regeol;
3427 register I32 hardcount = 0;
3430 if (max != REG_INFTY && max < loceol - scan)
3431 loceol = scan + max;
3434 while (scan < loceol && *scan != '\n')
3442 while (scan < loceol && *scan != '\n') {
3443 scan += UTF8SKIP(scan);
3449 while (scan < loceol) {
3450 scan += UTF8SKIP(scan);
3454 case EXACT: /* length of string is 1 */
3456 while (scan < loceol && UCHARAT(scan) == c)
3459 case EXACTF: /* length of string is 1 */
3461 while (scan < loceol &&
3462 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3465 case EXACTFL: /* length of string is 1 */
3466 PL_reg_flags |= RF_tainted;
3468 while (scan < loceol &&
3469 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3474 while (scan < loceol && REGINCLASSUTF8(p, (U8*)scan)) {
3475 scan += UTF8SKIP(scan);
3480 while (scan < loceol && REGINCLASS(p, *scan))
3484 while (scan < loceol && isALNUM(*scan))
3489 while (scan < loceol && swash_fetch(PL_utf8_alnum, (U8*)scan)) {
3490 scan += UTF8SKIP(scan);
3495 PL_reg_flags |= RF_tainted;
3496 while (scan < loceol && isALNUM_LC(*scan))
3500 PL_reg_flags |= RF_tainted;
3502 while (scan < loceol && isALNUM_LC_utf8((U8*)scan)) {
3503 scan += UTF8SKIP(scan);
3509 while (scan < loceol && !isALNUM(*scan))
3514 while (scan < loceol && !swash_fetch(PL_utf8_alnum, (U8*)scan)) {
3515 scan += UTF8SKIP(scan);
3520 PL_reg_flags |= RF_tainted;
3521 while (scan < loceol && !isALNUM_LC(*scan))
3525 PL_reg_flags |= RF_tainted;
3527 while (scan < loceol && !isALNUM_LC_utf8((U8*)scan)) {
3528 scan += UTF8SKIP(scan);
3533 while (scan < loceol && isSPACE(*scan))
3538 while (scan < loceol && (*scan == ' ' || swash_fetch(PL_utf8_space,(U8*)scan))) {
3539 scan += UTF8SKIP(scan);
3544 PL_reg_flags |= RF_tainted;
3545 while (scan < loceol && isSPACE_LC(*scan))
3549 PL_reg_flags |= RF_tainted;
3551 while (scan < loceol && (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3552 scan += UTF8SKIP(scan);
3557 while (scan < loceol && !isSPACE(*scan))
3562 while (scan < loceol && !(*scan == ' ' || swash_fetch(PL_utf8_space,(U8*)scan))) {
3563 scan += UTF8SKIP(scan);
3568 PL_reg_flags |= RF_tainted;
3569 while (scan < loceol && !isSPACE_LC(*scan))
3573 PL_reg_flags |= RF_tainted;
3575 while (scan < loceol && !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3576 scan += UTF8SKIP(scan);
3581 while (scan < loceol && isDIGIT(*scan))
3586 while (scan < loceol && swash_fetch(PL_utf8_digit,(U8*)scan)) {
3587 scan += UTF8SKIP(scan);
3593 while (scan < loceol && !isDIGIT(*scan))
3598 while (scan < loceol && !swash_fetch(PL_utf8_digit,(U8*)scan)) {
3599 scan += UTF8SKIP(scan);
3603 default: /* Called on something of 0 width. */
3604 break; /* So match right here or not at all. */
3610 c = scan - PL_reginput;
3615 SV *prop = sv_newmortal();
3618 PerlIO_printf(Perl_debug_log,
3619 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
3620 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
3627 - regrepeat_hard - repeatedly match something, report total lenth and length
3629 * The repeater is supposed to have constant length.
3633 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
3636 register char *scan;
3637 register char *start;
3638 register char *loceol = PL_regeol;
3640 I32 count = 0, res = 1;
3645 start = PL_reginput;
3647 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3650 while (start < PL_reginput) {
3652 start += UTF8SKIP(start);
3663 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
3665 *lp = l = PL_reginput - start;
3666 if (max != REG_INFTY && l*max < loceol - scan)
3667 loceol = scan + l*max;
3680 - reginclass - determine if a character falls into a character class
3684 S_reginclass(pTHX_ register regnode *p, register I32 c)
3687 char flags = ANYOF_FLAGS(p);
3691 if (ANYOF_BITMAP_TEST(p, c))
3693 else if (flags & ANYOF_FOLD) {
3695 if (flags & ANYOF_LOCALE) {
3696 PL_reg_flags |= RF_tainted;
3697 cf = PL_fold_locale[c];
3701 if (ANYOF_BITMAP_TEST(p, cf))
3705 if (!match && (flags & ANYOF_CLASS)) {
3706 PL_reg_flags |= RF_tainted;
3708 (ANYOF_CLASS_TEST(p, ANYOF_ALNUM) && isALNUM_LC(c)) ||
3709 (ANYOF_CLASS_TEST(p, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
3710 (ANYOF_CLASS_TEST(p, ANYOF_SPACE) && isSPACE_LC(c)) ||
3711 (ANYOF_CLASS_TEST(p, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
3712 (ANYOF_CLASS_TEST(p, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
3713 (ANYOF_CLASS_TEST(p, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
3714 (ANYOF_CLASS_TEST(p, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
3715 (ANYOF_CLASS_TEST(p, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
3716 (ANYOF_CLASS_TEST(p, ANYOF_ALPHA) && isALPHA_LC(c)) ||
3717 (ANYOF_CLASS_TEST(p, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
3718 (ANYOF_CLASS_TEST(p, ANYOF_ASCII) && isASCII(c)) ||
3719 (ANYOF_CLASS_TEST(p, ANYOF_NASCII) && !isASCII(c)) ||
3720 (ANYOF_CLASS_TEST(p, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
3721 (ANYOF_CLASS_TEST(p, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
3722 (ANYOF_CLASS_TEST(p, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
3723 (ANYOF_CLASS_TEST(p, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
3724 (ANYOF_CLASS_TEST(p, ANYOF_LOWER) && isLOWER_LC(c)) ||
3725 (ANYOF_CLASS_TEST(p, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
3726 (ANYOF_CLASS_TEST(p, ANYOF_PRINT) && isPRINT_LC(c)) ||
3727 (ANYOF_CLASS_TEST(p, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
3728 (ANYOF_CLASS_TEST(p, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
3729 (ANYOF_CLASS_TEST(p, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
3730 (ANYOF_CLASS_TEST(p, ANYOF_UPPER) && isUPPER_LC(c)) ||
3731 (ANYOF_CLASS_TEST(p, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
3732 (ANYOF_CLASS_TEST(p, ANYOF_XDIGIT) && isXDIGIT(c)) ||
3733 (ANYOF_CLASS_TEST(p, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
3734 (ANYOF_CLASS_TEST(p, ANYOF_PSXSPC) && isPSXSPC(c)) ||
3735 (ANYOF_CLASS_TEST(p, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
3736 (ANYOF_CLASS_TEST(p, ANYOF_BLANK) && isBLANK(c)) ||
3737 (ANYOF_CLASS_TEST(p, ANYOF_NBLANK) && !isBLANK(c))
3738 ) /* How's that for a conditional? */
3744 return (flags & ANYOF_INVERT) ? !match : match;
3748 S_reginclassutf8(pTHX_ regnode *f, U8 *p)
3751 char flags = ARG1(f);
3753 SV *sv = (SV*)PL_regdata->data[ARG2(f)];
3755 if (swash_fetch(sv, p))
3757 else if (flags & ANYOF_FOLD) {
3758 U8 tmpbuf[UTF8_MAXLEN];
3759 if (flags & ANYOF_LOCALE) {
3760 PL_reg_flags |= RF_tainted;
3761 uv_to_utf8(tmpbuf, toLOWER_LC_utf8(p));
3764 uv_to_utf8(tmpbuf, toLOWER_utf8(p));
3765 if (swash_fetch(sv, tmpbuf))
3769 /* UTF8 combined with ANYOF_CLASS is ill-defined. */
3771 return (flags & ANYOF_INVERT) ? !match : match;
3775 S_reghop(pTHX_ U8 *s, I32 off)
3779 while (off-- && s < (U8*)PL_regeol)
3784 if (s > (U8*)PL_bostr) {
3787 while (s > (U8*)PL_bostr && (*s & 0xc0) == 0x80)
3789 } /* XXX could check well-formedness here */
3797 S_reghopmaybe(pTHX_ U8* s, I32 off)
3801 while (off-- && s < (U8*)PL_regeol)
3808 if (s > (U8*)PL_bostr) {
3811 while (s > (U8*)PL_bostr && (*s & 0xc0) == 0x80)
3813 } /* XXX could check well-formedness here */
3829 restore_pos(pTHXo_ void *arg)
3832 if (PL_reg_eval_set) {
3833 if (PL_reg_oldsaved) {
3834 PL_reg_re->subbeg = PL_reg_oldsaved;
3835 PL_reg_re->sublen = PL_reg_oldsavedlen;
3836 RX_MATCH_COPIED_on(PL_reg_re);
3838 PL_reg_magic->mg_len = PL_reg_oldpos;
3839 PL_reg_eval_set = 0;
3840 PL_curpm = PL_reg_oldcurpm;