5 * "One Ring to rule them all, One Ring to find them..."
8 /* This file contains functions for executing a regular expression. See
9 * also regcomp.c which funnily enough, contains functions for compiling
10 * a regular expression.
12 * This file is also copied at build time to ext/re/re_exec.c, where
13 * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT.
14 * This causes the main functions to be compiled under new names and with
15 * debugging support added, which makes "use re 'debug'" work.
18 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
19 * confused with the original package (see point 3 below). Thanks, Henry!
22 /* Additional note: this code is very heavily munged from Henry's version
23 * in places. In some spots I've traded clarity for efficiency, so don't
24 * blame Henry for some of the lack of readability.
27 /* The names of the functions have been changed from regcomp and
28 * regexec to pregcomp and pregexec in order to avoid conflicts
29 * with the POSIX routines of the same names.
32 #ifdef PERL_EXT_RE_BUILD
37 * pregcomp and pregexec -- regsub and regerror are not used in perl
39 * Copyright (c) 1986 by University of Toronto.
40 * Written by Henry Spencer. Not derived from licensed software.
42 * Permission is granted to anyone to use this software for any
43 * purpose on any computer system, and to redistribute it freely,
44 * subject to the following restrictions:
46 * 1. The author is not responsible for the consequences of use of
47 * this software, no matter how awful, even if they arise
50 * 2. The origin of this software must not be misrepresented, either
51 * by explicit claim or by omission.
53 * 3. Altered versions must be plainly marked as such, and must not
54 * be misrepresented as being the original software.
56 **** Alterations to Henry's code are...
58 **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
59 **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, by Larry Wall and others
61 **** You may distribute under the terms of either the GNU General Public
62 **** License or the Artistic License, as specified in the README file.
64 * Beware that some of this code is subtly aware of the way operator
65 * precedence is structured in regular expressions. Serious changes in
66 * regular-expression syntax might require a total rethink.
69 #define PERL_IN_REGEXEC_C
72 #ifdef PERL_IN_XSUB_RE
78 #define RF_tainted 1 /* tainted information used? */
79 #define RF_warned 2 /* warned about big count? */
80 #define RF_evaled 4 /* Did an EVAL with setting? */
81 #define RF_utf8 8 /* Pattern contains multibyte chars? */
83 #define UTF ((PL_reg_flags & RF_utf8) != 0)
85 #define RS_init 1 /* eval environment created */
86 #define RS_set 2 /* replsv value is set */
92 #define REGINCLASS(prog,p,c) (ANYOF_FLAGS(p) ? reginclass(prog,p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c)))
98 #define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv))
99 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
101 #define HOPc(pos,off) \
102 (char *)(PL_reg_match_utf8 \
103 ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \
105 #define HOPBACKc(pos, off) \
106 (char*)(PL_reg_match_utf8\
107 ? reghopmaybe3((U8*)pos, -off, (U8*)PL_bostr) \
108 : (pos - off >= PL_bostr) \
112 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)(pos), off, (U8*)(lim)) : (U8*)(pos + off))
113 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
115 #define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \
116 if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)str); assert(ok); LEAVE; } } STMT_END
117 #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a")
118 #define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0")
119 #define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ")
120 #define LOAD_UTF8_CHARCLASS_MARK() LOAD_UTF8_CHARCLASS(mark, "\xcd\x86")
122 /* TODO: Combine JUMPABLE and HAS_TEXT to cache OP(rn) */
124 /* for use after a quantifier and before an EXACT-like node -- japhy */
125 #define JUMPABLE(rn) ( \
126 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
127 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
128 OP(rn) == PLUS || OP(rn) == MINMOD || \
129 (PL_regkind[OP(rn)] == CURLY && ARG1(rn) > 0) \
132 #define HAS_TEXT(rn) ( \
133 PL_regkind[OP(rn)] == EXACT || PL_regkind[OP(rn)] == REF \
137 Search for mandatory following text node; for lookahead, the text must
138 follow but for lookbehind (rn->flags != 0) we skip to the next step.
140 #define FIND_NEXT_IMPT(rn) STMT_START { \
141 while (JUMPABLE(rn)) { \
142 const OPCODE type = OP(rn); \
143 if (type == SUSPEND || PL_regkind[type] == CURLY) \
144 rn = NEXTOPER(NEXTOPER(rn)); \
145 else if (type == PLUS) \
147 else if (type == IFMATCH) \
148 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
149 else rn += NEXT_OFF(rn); \
153 static void restore_pos(pTHX_ void *arg);
156 S_regcppush(pTHX_ I32 parenfloor)
159 const int retval = PL_savestack_ix;
160 #define REGCP_PAREN_ELEMS 4
161 const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
163 GET_RE_DEBUG_FLAGS_DECL;
165 if (paren_elems_to_push < 0)
166 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
168 #define REGCP_OTHER_ELEMS 6
169 SSGROW(paren_elems_to_push + REGCP_OTHER_ELEMS);
170 for (p = PL_regsize; p > parenfloor; p--) {
171 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
172 SSPUSHINT(PL_regendp[p]);
173 SSPUSHINT(PL_regstartp[p]);
174 SSPUSHPTR(PL_reg_start_tmp[p]);
176 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
177 " saving \\%"UVuf" %"IVdf"(%"IVdf")..%"IVdf"\n",
178 (UV)p, (IV)PL_regstartp[p],
179 (IV)(PL_reg_start_tmp[p] - PL_bostr),
183 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
184 SSPUSHINT(PL_regsize);
185 SSPUSHINT(*PL_reglastparen);
186 SSPUSHINT(*PL_reglastcloseparen);
187 SSPUSHPTR(PL_reginput);
188 #define REGCP_FRAME_ELEMS 2
189 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
190 * are needed for the regexp context stack bookkeeping. */
191 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
192 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
197 /* These are needed since we do not localize EVAL nodes: */
198 #define REGCP_SET(cp) \
200 PerlIO_printf(Perl_debug_log, \
201 " Setting an EVAL scope, savestack=%"IVdf"\n", \
202 (IV)PL_savestack_ix)); \
205 #define REGCP_UNWIND(cp) \
207 if (cp != PL_savestack_ix) \
208 PerlIO_printf(Perl_debug_log, \
209 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
210 (IV)(cp), (IV)PL_savestack_ix)); \
214 S_regcppop(pTHX_ const regexp *rex)
220 GET_RE_DEBUG_FLAGS_DECL;
222 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
224 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
225 i = SSPOPINT; /* Parentheses elements to pop. */
226 input = (char *) SSPOPPTR;
227 *PL_reglastcloseparen = SSPOPINT;
228 *PL_reglastparen = SSPOPINT;
229 PL_regsize = SSPOPINT;
231 /* Now restore the parentheses context. */
232 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
233 i > 0; i -= REGCP_PAREN_ELEMS) {
235 U32 paren = (U32)SSPOPINT;
236 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
237 PL_regstartp[paren] = SSPOPINT;
239 if (paren <= *PL_reglastparen)
240 PL_regendp[paren] = tmps;
242 PerlIO_printf(Perl_debug_log,
243 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
244 (UV)paren, (IV)PL_regstartp[paren],
245 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
246 (IV)PL_regendp[paren],
247 (paren > *PL_reglastparen ? "(no)" : ""));
251 if (*PL_reglastparen + 1 <= rex->nparens) {
252 PerlIO_printf(Perl_debug_log,
253 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
254 (IV)(*PL_reglastparen + 1), (IV)rex->nparens);
258 /* It would seem that the similar code in regtry()
259 * already takes care of this, and in fact it is in
260 * a better location to since this code can #if 0-ed out
261 * but the code in regtry() is needed or otherwise tests
262 * requiring null fields (pat.t#187 and split.t#{13,14}
263 * (as of patchlevel 7877) will fail. Then again,
264 * this code seems to be necessary or otherwise
265 * building DynaLoader will fail:
266 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
268 for (i = *PL_reglastparen + 1; (U32)i <= rex->nparens; i++) {
270 PL_regstartp[i] = -1;
277 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
280 * pregexec and friends
283 #ifndef PERL_IN_XSUB_RE
285 - pregexec - match a regexp against a string
288 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
289 char *strbeg, I32 minend, SV *screamer, U32 nosave)
290 /* strend: pointer to null at end of string */
291 /* strbeg: real beginning of string */
292 /* minend: end of match must be >=minend after stringarg. */
293 /* nosave: For optimizations. */
296 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
297 nosave ? 0 : REXEC_COPY_STR);
302 * Need to implement the following flags for reg_anch:
304 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
306 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
307 * INTUIT_AUTORITATIVE_ML
308 * INTUIT_ONCE_NOML - Intuit can match in one location only.
311 * Another flag for this function: SECOND_TIME (so that float substrs
312 * with giant delta may be not rechecked).
315 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
317 /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend.
318 Otherwise, only SvCUR(sv) is used to get strbeg. */
320 /* XXXX We assume that strpos is strbeg unless sv. */
322 /* XXXX Some places assume that there is a fixed substring.
323 An update may be needed if optimizer marks as "INTUITable"
324 RExen without fixed substrings. Similarly, it is assumed that
325 lengths of all the strings are no more than minlen, thus they
326 cannot come from lookahead.
327 (Or minlen should take into account lookahead.) */
329 /* A failure to find a constant substring means that there is no need to make
330 an expensive call to REx engine, thus we celebrate a failure. Similarly,
331 finding a substring too deep into the string means that less calls to
332 regtry() should be needed.
334 REx compiler's optimizer found 4 possible hints:
335 a) Anchored substring;
337 c) Whether we are anchored (beginning-of-line or \G);
338 d) First node (of those at offset 0) which may distingush positions;
339 We use a)b)d) and multiline-part of c), and try to find a position in the
340 string which does not contradict any of them.
343 /* Most of decisions we do here should have been done at compile time.
344 The nodes of the REx which we used for the search should have been
345 deleted from the finite automaton. */
348 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
349 char *strend, U32 flags, re_scream_pos_data *data)
352 register I32 start_shift = 0;
353 /* Should be nonnegative! */
354 register I32 end_shift = 0;
359 const bool do_utf8 = (sv && SvUTF8(sv)) ? 1 : 0; /* if no sv we have to assume bytes */
361 register char *other_last = NULL; /* other substr checked before this */
362 char *check_at = NULL; /* check substr found at this pos */
363 const I32 multiline = prog->reganch & PMf_MULTILINE;
365 const char * const i_strpos = strpos;
368 GET_RE_DEBUG_FLAGS_DECL;
370 RX_MATCH_UTF8_set(prog,do_utf8);
372 if (prog->reganch & ROPT_UTF8) {
373 PL_reg_flags |= RF_utf8;
376 debug_start_match(prog, do_utf8, strpos, strend,
377 sv ? "Guessing start of match in sv for"
378 : "Guessing start of match in string for");
381 /* CHR_DIST() would be more correct here but it makes things slow. */
382 if (prog->minlen > strend - strpos) {
383 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
384 "String too short... [re_intuit_start]\n"));
388 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
391 if (!prog->check_utf8 && prog->check_substr)
392 to_utf8_substr(prog);
393 check = prog->check_utf8;
395 if (!prog->check_substr && prog->check_utf8)
396 to_byte_substr(prog);
397 check = prog->check_substr;
399 if (check == &PL_sv_undef) {
400 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
401 "Non-utf8 string cannot match utf8 check string\n"));
404 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
405 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
406 || ( (prog->reganch & ROPT_ANCH_BOL)
407 && !multiline ) ); /* Check after \n? */
410 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
411 | ROPT_IMPLICIT)) /* not a real BOL */
412 /* SvCUR is not set on references: SvRV and SvPVX_const overlap */
414 && (strpos != strbeg)) {
415 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
418 if (prog->check_offset_min == prog->check_offset_max &&
419 !(prog->reganch & ROPT_CANY_SEEN)) {
420 /* Substring at constant offset from beg-of-str... */
423 s = HOP3c(strpos, prog->check_offset_min, strend);
426 slen = SvCUR(check); /* >= 1 */
428 if ( strend - s > slen || strend - s < slen - 1
429 || (strend - s == slen && strend[-1] != '\n')) {
430 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
433 /* Now should match s[0..slen-2] */
435 if (slen && (*SvPVX_const(check) != *s
437 && memNE(SvPVX_const(check), s, slen)))) {
439 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
443 else if (*SvPVX_const(check) != *s
444 || ((slen = SvCUR(check)) > 1
445 && memNE(SvPVX_const(check), s, slen)))
448 goto success_at_start;
451 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
453 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
454 end_shift = prog->check_end_shift;
457 const I32 end = prog->check_offset_max + CHR_SVLEN(check)
458 - (SvTAIL(check) != 0);
459 const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
461 if (end_shift < eshift)
465 else { /* Can match at random position */
468 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
469 end_shift = prog->check_end_shift;
471 /* end shift should be non negative here */
474 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
476 Perl_croak(aTHX_ "panic: end_shift: %"IVdf" pattern:\n%s\n ",
477 end_shift,prog->precomp);
481 /* Find a possible match in the region s..strend by looking for
482 the "check" substring in the region corrected by start/end_shift. */
485 I32 srch_start_shift = start_shift;
486 I32 srch_end_shift = end_shift;
487 if (srch_start_shift < 0 && strbeg - s > srch_start_shift) {
488 srch_end_shift -= ((strbeg - s) - srch_start_shift);
489 srch_start_shift = strbeg - s;
492 PerlIO_printf(Perl_debug_log, "Check offset min: %"IVdf" Start shift: %"IVdf" End shift %"IVdf" Real End Shift: %"IVdf"\n",
493 (IV)prog->check_offset_min,
494 (IV)srch_start_shift,
496 (IV)prog->check_end_shift);
499 if (flags & REXEC_SCREAM) {
500 I32 p = -1; /* Internal iterator of scream. */
501 I32 * const pp = data ? data->scream_pos : &p;
503 if (PL_screamfirst[BmRARE(check)] >= 0
504 || ( BmRARE(check) == '\n'
505 && (BmPREVIOUS(check) == SvCUR(check) - 1)
507 s = screaminstr(sv, check,
508 srch_start_shift + (s - strbeg), srch_end_shift, pp, 0);
511 /* we may be pointing at the wrong string */
512 if (s && RX_MATCH_COPIED(prog))
513 s = strbeg + (s - SvPVX_const(sv));
515 *data->scream_olds = s;
520 if (prog->reganch & ROPT_CANY_SEEN) {
521 start_point= (U8*)(s + srch_start_shift);
522 end_point= (U8*)(strend - srch_end_shift);
524 start_point= HOP3(s, srch_start_shift, srch_start_shift < 0 ? strbeg : strend);
525 end_point= HOP3(strend, -srch_end_shift, strbeg);
528 PerlIO_printf(Perl_debug_log, "fbm_instr len=%d str=<%.*s>\n",
529 (int)(end_point - start_point),
530 (int)(end_point - start_point),
534 s = fbm_instr( start_point, end_point,
535 check, multiline ? FBMrf_MULTILINE : 0);
538 /* Update the count-of-usability, remove useless subpatterns,
542 RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
543 SvPVX_const(check), RE_SV_DUMPLEN(check), 30);
544 PerlIO_printf(Perl_debug_log, "%s %s substr %s%s%s",
545 (s ? "Found" : "Did not find"),
546 (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)
547 ? "anchored" : "floating"),
550 (s ? " at offset " : "...\n") );
555 /* Finish the diagnostic message */
556 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
558 /* XXX dmq: first branch is for positive lookbehind...
559 Our check string is offset from the beginning of the pattern.
560 So we need to do any stclass tests offset forward from that
569 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
570 Start with the other substr.
571 XXXX no SCREAM optimization yet - and a very coarse implementation
572 XXXX /ttx+/ results in anchored="ttx", floating="x". floating will
573 *always* match. Probably should be marked during compile...
574 Probably it is right to do no SCREAM here...
577 if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8)
578 : (prog->float_substr && prog->anchored_substr))
580 /* Take into account the "other" substring. */
581 /* XXXX May be hopelessly wrong for UTF... */
584 if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
587 char * const last = HOP3c(s, -start_shift, strbeg);
589 char * const saved_s = s;
592 t = s - prog->check_offset_max;
593 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
595 || ((t = (char*)reghopmaybe3((U8*)s, -(prog->check_offset_max), (U8*)strpos))
600 t = HOP3c(t, prog->anchored_offset, strend);
601 if (t < other_last) /* These positions already checked */
603 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
606 /* XXXX It is not documented what units *_offsets are in.
607 We assume bytes, but this is clearly wrong.
608 Meaning this code needs to be carefully reviewed for errors.
612 /* On end-of-str: see comment below. */
613 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
614 if (must == &PL_sv_undef) {
616 DEBUG_r(must = prog->anchored_utf8); /* for debug */
621 HOP3(HOP3(last1, prog->anchored_offset, strend)
622 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
624 multiline ? FBMrf_MULTILINE : 0
627 RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
628 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
629 PerlIO_printf(Perl_debug_log, "%s anchored substr %s%s",
630 (s ? "Found" : "Contradicts"),
631 quoted, RE_SV_TAIL(must));
636 if (last1 >= last2) {
637 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
638 ", giving up...\n"));
641 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
642 ", trying floating at offset %ld...\n",
643 (long)(HOP3c(saved_s, 1, strend) - i_strpos)));
644 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
645 s = HOP3c(last, 1, strend);
649 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
650 (long)(s - i_strpos)));
651 t = HOP3c(s, -prog->anchored_offset, strbeg);
652 other_last = HOP3c(s, 1, strend);
660 else { /* Take into account the floating substring. */
662 char * const saved_s = s;
665 t = HOP3c(s, -start_shift, strbeg);
667 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
668 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
669 last = HOP3c(t, prog->float_max_offset, strend);
670 s = HOP3c(t, prog->float_min_offset, strend);
673 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
674 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
675 /* fbm_instr() takes into account exact value of end-of-str
676 if the check is SvTAIL(ed). Since false positives are OK,
677 and end-of-str is not later than strend we are OK. */
678 if (must == &PL_sv_undef) {
680 DEBUG_r(must = prog->float_utf8); /* for debug message */
683 s = fbm_instr((unsigned char*)s,
684 (unsigned char*)last + SvCUR(must)
686 must, multiline ? FBMrf_MULTILINE : 0);
688 RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
689 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
690 PerlIO_printf(Perl_debug_log, "%s floating substr %s%s",
691 (s ? "Found" : "Contradicts"),
692 quoted, RE_SV_TAIL(must));
696 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
697 ", giving up...\n"));
700 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
701 ", trying anchored starting at offset %ld...\n",
702 (long)(saved_s + 1 - i_strpos)));
704 s = HOP3c(t, 1, strend);
708 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
709 (long)(s - i_strpos)));
710 other_last = s; /* Fix this later. --Hugo */
720 t= HOP3( s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos);
723 PerlIO_printf(Perl_debug_log,
724 "Check offset min:%"IVdf" max:%"IVdf" S:%"IVdf" t:%"IVdf" D:%"IVdf" end:%"IVdf"\n",
725 (IV)prog->check_offset_min,
726 (IV)prog->check_offset_max,
734 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
736 || ((t = (char*)reghopmaybe3((U8*)s, -prog->check_offset_max, (prog->check_offset_max<0) ? strend : strpos))
739 /* Fixed substring is found far enough so that the match
740 cannot start at strpos. */
742 if (ml_anch && t[-1] != '\n') {
743 /* Eventually fbm_*() should handle this, but often
744 anchored_offset is not 0, so this check will not be wasted. */
745 /* XXXX In the code below we prefer to look for "^" even in
746 presence of anchored substrings. And we search even
747 beyond the found float position. These pessimizations
748 are historical artefacts only. */
750 while (t < strend - prog->minlen) {
752 if (t < check_at - prog->check_offset_min) {
753 if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
754 /* Since we moved from the found position,
755 we definitely contradict the found anchored
756 substr. Due to the above check we do not
757 contradict "check" substr.
758 Thus we can arrive here only if check substr
759 is float. Redo checking for "other"=="fixed".
762 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
763 PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
764 goto do_other_anchored;
766 /* We don't contradict the found floating substring. */
767 /* XXXX Why not check for STCLASS? */
769 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
770 PL_colors[0], PL_colors[1], (long)(s - i_strpos)));
773 /* Position contradicts check-string */
774 /* XXXX probably better to look for check-string
775 than for "\n", so one should lower the limit for t? */
776 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
777 PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos)));
778 other_last = strpos = s = t + 1;
783 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
784 PL_colors[0], PL_colors[1]));
788 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
789 PL_colors[0], PL_colors[1]));
793 ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
796 /* The found string does not prohibit matching at strpos,
797 - no optimization of calling REx engine can be performed,
798 unless it was an MBOL and we are not after MBOL,
799 or a future STCLASS check will fail this. */
801 /* Even in this situation we may use MBOL flag if strpos is offset
802 wrt the start of the string. */
803 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
804 && (strpos != strbeg) && strpos[-1] != '\n'
805 /* May be due to an implicit anchor of m{.*foo} */
806 && !(prog->reganch & ROPT_IMPLICIT))
811 DEBUG_EXECUTE_r( if (ml_anch)
812 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
813 (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]);
816 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
818 prog->check_utf8 /* Could be deleted already */
819 && --BmUSEFUL(prog->check_utf8) < 0
820 && (prog->check_utf8 == prog->float_utf8)
822 prog->check_substr /* Could be deleted already */
823 && --BmUSEFUL(prog->check_substr) < 0
824 && (prog->check_substr == prog->float_substr)
827 /* If flags & SOMETHING - do not do it many times on the same match */
828 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
829 SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr);
830 if (do_utf8 ? prog->check_substr : prog->check_utf8)
831 SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8);
832 prog->check_substr = prog->check_utf8 = NULL; /* disable */
833 prog->float_substr = prog->float_utf8 = NULL; /* clear */
834 check = NULL; /* abort */
836 /* XXXX This is a remnant of the old implementation. It
837 looks wasteful, since now INTUIT can use many
839 prog->reganch &= ~RE_USE_INTUIT;
846 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
847 /* trie stclasses are too expensive to use here, we are better off to
848 leave it to regmatch itself */
849 if (prog->regstclass && PL_regkind[OP(prog->regstclass)]!=TRIE) {
850 /* minlen == 0 is possible if regstclass is \b or \B,
851 and the fixed substr is ''$.
852 Since minlen is already taken into account, s+1 is before strend;
853 accidentally, minlen >= 1 guaranties no false positives at s + 1
854 even for \b or \B. But (minlen? 1 : 0) below assumes that
855 regstclass does not come from lookahead... */
856 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
857 This leaves EXACTF only, which is dealt with in find_byclass(). */
858 const U8* const str = (U8*)STRING(prog->regstclass);
859 const int cl_l = (PL_regkind[OP(prog->regstclass)] == EXACT
860 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
863 if (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
864 endpos= HOP3c(s, (prog->minlen ? cl_l : 0), strend);
865 else if (prog->float_substr || prog->float_utf8)
866 endpos= HOP3c(HOP3c(check_at, -start_shift, strbeg), cl_l, strend);
870 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "start_shift: %"IVdf" check_at: %d s: %d endpos: %d\n",
871 start_shift,check_at-strbeg,s-strbeg,endpos-strbeg));
874 s = find_byclass(prog, prog->regstclass, s, endpos, NULL);
877 const char *what = NULL;
879 if (endpos == strend) {
880 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
881 "Could not match STCLASS...\n") );
884 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
885 "This position contradicts STCLASS...\n") );
886 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
888 /* Contradict one of substrings */
889 if (prog->anchored_substr || prog->anchored_utf8) {
890 if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
891 DEBUG_EXECUTE_r( what = "anchored" );
893 s = HOP3c(t, 1, strend);
894 if (s + start_shift + end_shift > strend) {
895 /* XXXX Should be taken into account earlier? */
896 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
897 "Could not match STCLASS...\n") );
902 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
903 "Looking for %s substr starting at offset %ld...\n",
904 what, (long)(s + start_shift - i_strpos)) );
907 /* Have both, check_string is floating */
908 if (t + start_shift >= check_at) /* Contradicts floating=check */
909 goto retry_floating_check;
910 /* Recheck anchored substring, but not floating... */
914 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
915 "Looking for anchored substr starting at offset %ld...\n",
916 (long)(other_last - i_strpos)) );
917 goto do_other_anchored;
919 /* Another way we could have checked stclass at the
920 current position only: */
925 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
926 "Looking for /%s^%s/m starting at offset %ld...\n",
927 PL_colors[0], PL_colors[1], (long)(t - i_strpos)) );
930 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
932 /* Check is floating subtring. */
933 retry_floating_check:
934 t = check_at - start_shift;
935 DEBUG_EXECUTE_r( what = "floating" );
936 goto hop_and_restart;
939 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
940 "By STCLASS: moving %ld --> %ld\n",
941 (long)(t - i_strpos), (long)(s - i_strpos))
945 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
946 "Does not contradict STCLASS...\n");
951 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
952 PL_colors[4], (check ? "Guessed" : "Giving up"),
953 PL_colors[5], (long)(s - i_strpos)) );
956 fail_finish: /* Substring not found */
957 if (prog->check_substr || prog->check_utf8) /* could be removed already */
958 BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
960 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
961 PL_colors[4], PL_colors[5]));
967 #define REXEC_TRIE_READ_CHAR(trie_type, trie, uc, uscan, len, uvc, charid, \
968 foldlen, foldbuf, uniflags) STMT_START { \
969 switch (trie_type) { \
970 case trie_utf8_fold: \
972 uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags ); \
977 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \
978 uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \
979 foldlen -= UNISKIP( uvc ); \
980 uscan = foldbuf + UNISKIP( uvc ); \
984 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags ); \
992 charid = trie->charmap[ uvc ]; \
996 if (trie->widecharmap) { \
997 SV** const svpp = hv_fetch(trie->widecharmap, \
998 (char*)&uvc, sizeof(UV), 0); \
1000 charid = (U16)SvIV(*svpp); \
1005 #define REXEC_FBC_EXACTISH_CHECK(CoNd) \
1008 ibcmp_utf8(s, NULL, 0, do_utf8, \
1009 m, NULL, ln, (bool)UTF)) \
1010 && (!reginfo || regtry(reginfo, s)) ) \
1013 U8 foldbuf[UTF8_MAXBYTES_CASE+1]; \
1014 uvchr_to_utf8(tmpbuf, c); \
1015 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen); \
1017 && (f == c1 || f == c2) \
1018 && (ln == foldlen || \
1019 !ibcmp_utf8((char *) foldbuf, \
1020 NULL, foldlen, do_utf8, \
1022 NULL, ln, (bool)UTF)) \
1023 && (!reginfo || regtry(reginfo, s)) ) \
1028 #define REXEC_FBC_EXACTISH_SCAN(CoNd) \
1032 && (ln == 1 || !(OP(c) == EXACTF \
1034 : ibcmp_locale(s, m, ln))) \
1035 && (!reginfo || regtry(reginfo, s)) ) \
1041 #define REXEC_FBC_UTF8_SCAN(CoDe) \
1043 while (s + (uskip = UTF8SKIP(s)) <= strend) { \
1049 #define REXEC_FBC_SCAN(CoDe) \
1051 while (s < strend) { \
1057 #define REXEC_FBC_UTF8_CLASS_SCAN(CoNd) \
1058 REXEC_FBC_UTF8_SCAN( \
1060 if (tmp && (!reginfo || regtry(reginfo, s))) \
1069 #define REXEC_FBC_CLASS_SCAN(CoNd) \
1072 if (tmp && (!reginfo || regtry(reginfo, s))) \
1081 #define REXEC_FBC_TRYIT \
1082 if ((!reginfo || regtry(reginfo, s))) \
1085 #define REXEC_FBC_CSCAN_PRELOAD(UtFpReLoAd,CoNdUtF8,CoNd) \
1088 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1091 REXEC_FBC_CLASS_SCAN(CoNd); \
1095 #define REXEC_FBC_CSCAN_TAINT(CoNdUtF8,CoNd) \
1096 PL_reg_flags |= RF_tainted; \
1098 REXEC_FBC_UTF8_CLASS_SCAN(CoNdUtF8); \
1101 REXEC_FBC_CLASS_SCAN(CoNd); \
1105 #define DUMP_EXEC_POS(li,s,doutf8) \
1106 dump_exec_pos(li,s,(PL_regeol),(PL_bostr),(PL_reg_starttry),doutf8)
1108 /* We know what class REx starts with. Try to find this position... */
1109 /* if reginfo is NULL, its a dryrun */
1110 /* annoyingly all the vars in this routine have different names from their counterparts
1111 in regmatch. /grrr */
1114 S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s,
1115 const char *strend, const regmatch_info *reginfo)
1118 const I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
1122 register STRLEN uskip;
1126 register I32 tmp = 1; /* Scratch variable? */
1127 register const bool do_utf8 = PL_reg_match_utf8;
1129 /* We know what class it must start with. */
1133 REXEC_FBC_UTF8_CLASS_SCAN((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
1134 !UTF8_IS_INVARIANT((U8)s[0]) ?
1135 reginclass(prog, c, (U8*)s, 0, do_utf8) :
1136 REGINCLASS(prog, c, (U8*)s));
1139 while (s < strend) {
1142 if (REGINCLASS(prog, c, (U8*)s) ||
1143 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
1144 /* The assignment of 2 is intentional:
1145 * for the folded sharp s, the skip is 2. */
1146 (skip = SHARP_S_SKIP))) {
1147 if (tmp && (!reginfo || regtry(reginfo, s)))
1160 if (tmp && (!reginfo || regtry(reginfo, s)))
1168 ln = STR_LEN(c); /* length to match in octets/bytes */
1169 lnc = (I32) ln; /* length to match in characters */
1171 STRLEN ulen1, ulen2;
1173 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
1174 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
1175 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1177 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1178 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1180 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE,
1182 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE,
1185 while (sm < ((U8 *) m + ln)) {
1200 c2 = PL_fold_locale[c1];
1202 e = HOP3c(strend, -((I32)lnc), s);
1204 if (!reginfo && e < s)
1205 e = s; /* Due to minlen logic of intuit() */
1207 /* The idea in the EXACTF* cases is to first find the
1208 * first character of the EXACTF* node and then, if
1209 * necessary, case-insensitively compare the full
1210 * text of the node. The c1 and c2 are the first
1211 * characters (though in Unicode it gets a bit
1212 * more complicated because there are more cases
1213 * than just upper and lower: one needs to use
1214 * the so-called folding case for case-insensitive
1215 * matching (called "loose matching" in Unicode).
1216 * ibcmp_utf8() will do just that. */
1220 U8 tmpbuf [UTF8_MAXBYTES+1];
1221 STRLEN len, foldlen;
1222 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1224 /* Upper and lower of 1st char are equal -
1225 * probably not a "letter". */
1227 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
1229 REXEC_FBC_EXACTISH_CHECK(c == c1);
1234 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
1237 /* Handle some of the three Greek sigmas cases.
1238 * Note that not all the possible combinations
1239 * are handled here: some of them are handled
1240 * by the standard folding rules, and some of
1241 * them (the character class or ANYOF cases)
1242 * are handled during compiletime in
1243 * regexec.c:S_regclass(). */
1244 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1245 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1246 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1248 REXEC_FBC_EXACTISH_CHECK(c == c1 || c == c2);
1254 REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1);
1256 REXEC_FBC_EXACTISH_SCAN(*(U8*)s == c1 || *(U8*)s == c2);
1260 PL_reg_flags |= RF_tainted;
1267 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1268 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT);
1270 tmp = ((OP(c) == BOUND ?
1271 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1272 LOAD_UTF8_CHARCLASS_ALNUM();
1273 REXEC_FBC_UTF8_SCAN(
1274 if (tmp == !(OP(c) == BOUND ?
1275 (bool)swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1276 isALNUM_LC_utf8((U8*)s)))
1284 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1285 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1288 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1294 if ((!prog->minlen && tmp) && (!reginfo || regtry(reginfo, s)))
1298 PL_reg_flags |= RF_tainted;
1305 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1306 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT);
1308 tmp = ((OP(c) == NBOUND ?
1309 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1310 LOAD_UTF8_CHARCLASS_ALNUM();
1311 REXEC_FBC_UTF8_SCAN(
1312 if (tmp == !(OP(c) == NBOUND ?
1313 (bool)swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1314 isALNUM_LC_utf8((U8*)s)))
1316 else REXEC_FBC_TRYIT;
1320 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1321 tmp = ((OP(c) == NBOUND ?
1322 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1325 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1327 else REXEC_FBC_TRYIT;
1330 if ((!prog->minlen && !tmp) && (!reginfo || regtry(reginfo, s)))
1334 REXEC_FBC_CSCAN_PRELOAD(
1335 LOAD_UTF8_CHARCLASS_ALNUM(),
1336 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8),
1340 REXEC_FBC_CSCAN_TAINT(
1341 isALNUM_LC_utf8((U8*)s),
1345 REXEC_FBC_CSCAN_PRELOAD(
1346 LOAD_UTF8_CHARCLASS_ALNUM(),
1347 !swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8),
1351 REXEC_FBC_CSCAN_TAINT(
1352 !isALNUM_LC_utf8((U8*)s),
1356 REXEC_FBC_CSCAN_PRELOAD(
1357 LOAD_UTF8_CHARCLASS_SPACE(),
1358 *s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8),
1362 REXEC_FBC_CSCAN_TAINT(
1363 *s == ' ' || isSPACE_LC_utf8((U8*)s),
1367 REXEC_FBC_CSCAN_PRELOAD(
1368 LOAD_UTF8_CHARCLASS_SPACE(),
1369 !(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)),
1373 REXEC_FBC_CSCAN_TAINT(
1374 !(*s == ' ' || isSPACE_LC_utf8((U8*)s)),
1378 REXEC_FBC_CSCAN_PRELOAD(
1379 LOAD_UTF8_CHARCLASS_DIGIT(),
1380 swash_fetch(PL_utf8_digit,(U8*)s, do_utf8),
1384 REXEC_FBC_CSCAN_TAINT(
1385 isDIGIT_LC_utf8((U8*)s),
1389 REXEC_FBC_CSCAN_PRELOAD(
1390 LOAD_UTF8_CHARCLASS_DIGIT(),
1391 !swash_fetch(PL_utf8_digit,(U8*)s, do_utf8),
1395 REXEC_FBC_CSCAN_TAINT(
1396 !isDIGIT_LC_utf8((U8*)s),
1402 const enum { trie_plain, trie_utf8, trie_utf8_fold }
1403 trie_type = do_utf8 ?
1404 (c->flags == EXACT ? trie_utf8 : trie_utf8_fold)
1406 /* what trie are we using right now */
1408 = (reg_ac_data*)prog->data->data[ ARG( c ) ];
1409 reg_trie_data *trie=aho->trie;
1411 const char *last_start = strend - trie->minlen;
1413 const char *real_start = s;
1415 STRLEN maxlen = trie->maxlen;
1417 U8 **points; /* map of where we were in the input string
1418 when reading a given char. For ASCII this
1419 is unnecessary overhead as the relationship
1420 is always 1:1, but for unicode, especially
1421 case folded unicode this is not true. */
1422 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1426 GET_RE_DEBUG_FLAGS_DECL;
1428 /* We can't just allocate points here. We need to wrap it in
1429 * an SV so it gets freed properly if there is a croak while
1430 * running the match */
1433 sv_points=newSV(maxlen * sizeof(U8 *));
1434 SvCUR_set(sv_points,
1435 maxlen * sizeof(U8 *));
1436 SvPOK_on(sv_points);
1437 sv_2mortal(sv_points);
1438 points=(U8**)SvPV_nolen(sv_points );
1439 if ( trie_type != trie_utf8_fold
1440 && (trie->bitmap || OP(c)==AHOCORASICKC) )
1443 bitmap=(U8*)trie->bitmap;
1445 bitmap=(U8*)ANYOF_BITMAP(c);
1447 /* this is the Aho-Corasick algorithm modified a touch
1448 to include special handling for long "unknown char"
1449 sequences. The basic idea being that we use AC as long
1450 as we are dealing with a possible matching char, when
1451 we encounter an unknown char (and we have not encountered
1452 an accepting state) we scan forward until we find a legal
1454 AC matching is basically that of trie matching, except
1455 that when we encounter a failing transition, we fall back
1456 to the current states "fail state", and try the current char
1457 again, a process we repeat until we reach the root state,
1458 state 1, or a legal transition. If we fail on the root state
1459 then we can either terminate if we have reached an accepting
1460 state previously, or restart the entire process from the beginning
1464 while (s <= last_start) {
1465 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1473 U8 *uscan = (U8*)NULL;
1474 U8 *leftmost = NULL;
1476 U32 accepted_word= 0;
1480 while ( state && uc <= (U8*)strend ) {
1482 U32 word = aho->states[ state ].wordnum;
1486 DEBUG_TRIE_EXECUTE_r(
1487 if ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
1488 dump_exec_pos( (char *)uc, c, strend, real_start,
1489 (char *)uc, do_utf8 );
1490 PerlIO_printf( Perl_debug_log,
1491 " Scanning for legal start char...\n");
1494 while ( uc <= (U8*)last_start && !BITMAP_TEST(bitmap,*uc) ) {
1499 if (uc >(U8*)last_start) break;
1503 U8 *lpos= points[ (pointpos - trie->wordlen[word-1] ) % maxlen ];
1504 if (!leftmost || lpos < leftmost) {
1505 DEBUG_r(accepted_word=word);
1511 points[pointpos++ % maxlen]= uc;
1512 REXEC_TRIE_READ_CHAR(trie_type, trie, uc, uscan, len,
1513 uvc, charid, foldlen, foldbuf, uniflags);
1514 DEBUG_TRIE_EXECUTE_r({
1515 dump_exec_pos( (char *)uc, c, strend, real_start,
1517 PerlIO_printf(Perl_debug_log,
1518 " Charid:%3u CP:%4"UVxf" ",
1524 word = aho->states[ state ].wordnum;
1526 base = aho->states[ state ].trans.base;
1528 DEBUG_TRIE_EXECUTE_r({
1530 dump_exec_pos( (char *)uc, c, strend, real_start,
1532 PerlIO_printf( Perl_debug_log,
1533 "%sState: %4"UVxf", word=%"UVxf,
1534 failed ? " Fail transition to " : "",
1535 (UV)state, (UV)word);
1540 (base + charid > trie->uniquecharcount )
1541 && (base + charid - 1 - trie->uniquecharcount
1543 && trie->trans[base + charid - 1 -
1544 trie->uniquecharcount].check == state
1545 && (tmp=trie->trans[base + charid - 1 -
1546 trie->uniquecharcount ].next))
1548 DEBUG_TRIE_EXECUTE_r(
1549 PerlIO_printf( Perl_debug_log," - legal\n"));
1554 DEBUG_TRIE_EXECUTE_r(
1555 PerlIO_printf( Perl_debug_log," - fail\n"));
1557 state = aho->fail[state];
1561 /* we must be accepting here */
1562 DEBUG_TRIE_EXECUTE_r(
1563 PerlIO_printf( Perl_debug_log," - accepting\n"));
1572 if (!state) state = 1;
1575 if ( aho->states[ state ].wordnum ) {
1576 U8 *lpos = points[ (pointpos - trie->wordlen[aho->states[ state ].wordnum-1]) % maxlen ];
1577 if (!leftmost || lpos < leftmost) {
1578 DEBUG_r(accepted_word=aho->states[ state ].wordnum);
1583 s = (char*)leftmost;
1584 DEBUG_TRIE_EXECUTE_r({
1586 Perl_debug_log,"Matches word #%"UVxf" at position %d. Trying full pattern...\n",
1587 (UV)accepted_word, s - real_start
1590 if (!reginfo || regtry(reginfo, s)) {
1596 DEBUG_TRIE_EXECUTE_r({
1597 PerlIO_printf( Perl_debug_log,"Pattern failed. Looking for new start point...\n");
1600 DEBUG_TRIE_EXECUTE_r(
1601 PerlIO_printf( Perl_debug_log,"No match.\n"));
1610 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1619 - regexec_flags - match a regexp against a string
1622 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1623 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1624 /* strend: pointer to null at end of string */
1625 /* strbeg: real beginning of string */
1626 /* minend: end of match must be >=minend after stringarg. */
1627 /* data: May be used for some additional optimizations. */
1628 /* nosave: For optimizations. */
1632 register regnode *c;
1633 register char *startpos = stringarg;
1634 I32 minlen; /* must match at least this many chars */
1635 I32 dontbother = 0; /* how many characters not to try at end */
1636 I32 end_shift = 0; /* Same for the end. */ /* CC */
1637 I32 scream_pos = -1; /* Internal iterator of scream. */
1638 char *scream_olds = NULL;
1639 SV* const oreplsv = GvSV(PL_replgv);
1640 const bool do_utf8 = DO_UTF8(sv);
1643 regmatch_info reginfo; /* create some info to pass to regtry etc */
1645 GET_RE_DEBUG_FLAGS_DECL;
1647 PERL_UNUSED_ARG(data);
1649 /* Be paranoid... */
1650 if (prog == NULL || startpos == NULL) {
1651 Perl_croak(aTHX_ "NULL regexp parameter");
1655 multiline = prog->reganch & PMf_MULTILINE;
1656 reginfo.prog = prog;
1658 RX_MATCH_UTF8_set(prog, do_utf8);
1660 debug_start_match(prog, do_utf8, startpos, strend,
1664 minlen = prog->minlen;
1666 if (strend - startpos < (minlen+(prog->check_offset_min<0?prog->check_offset_min:0))) {
1667 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1668 "String too short [regexec_flags]...\n"));
1673 /* Check validity of program. */
1674 if (UCHARAT(prog->program) != REG_MAGIC) {
1675 Perl_croak(aTHX_ "corrupted regexp program");
1679 PL_reg_eval_set = 0;
1682 if (prog->reganch & ROPT_UTF8)
1683 PL_reg_flags |= RF_utf8;
1685 /* Mark beginning of line for ^ and lookbehind. */
1686 reginfo.bol = startpos; /* XXX not used ??? */
1690 /* Mark end of line for $ (and such) */
1693 /* see how far we have to get to not match where we matched before */
1694 reginfo.till = startpos+minend;
1696 /* If there is a "must appear" string, look for it. */
1699 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to set reginfo->ganch */
1702 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1703 reginfo.ganch = startpos;
1704 else if (sv && SvTYPE(sv) >= SVt_PVMG
1706 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1707 && mg->mg_len >= 0) {
1708 reginfo.ganch = strbeg + mg->mg_len; /* Defined pos() */
1709 if (prog->reganch & ROPT_ANCH_GPOS) {
1710 if (s > reginfo.ganch)
1715 else /* pos() not defined */
1716 reginfo.ganch = strbeg;
1719 if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) {
1720 re_scream_pos_data d;
1722 d.scream_olds = &scream_olds;
1723 d.scream_pos = &scream_pos;
1724 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1726 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1727 goto phooey; /* not present */
1733 /* Simplest case: anchored match need be tried only once. */
1734 /* [unless only anchor is BOL and multiline is set] */
1735 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1736 if (s == startpos && regtry(®info, startpos))
1738 else if (multiline || (prog->reganch & ROPT_IMPLICIT)
1739 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1744 dontbother = minlen - 1;
1745 end = HOP3c(strend, -dontbother, strbeg) - 1;
1746 /* for multiline we only have to try after newlines */
1747 if (prog->check_substr || prog->check_utf8) {
1751 if (regtry(®info, s))
1756 if (prog->reganch & RE_USE_INTUIT) {
1757 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1768 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1769 if (regtry(®info, s))
1776 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1777 if (regtry(®info, reginfo.ganch))
1782 /* Messy cases: unanchored match. */
1783 if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) {
1784 /* we have /x+whatever/ */
1785 /* it must be a one character string (XXXX Except UTF?) */
1790 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1791 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1792 ch = SvPVX_const(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
1797 DEBUG_EXECUTE_r( did_match = 1 );
1798 if (regtry(®info, s)) goto got_it;
1800 while (s < strend && *s == ch)
1808 DEBUG_EXECUTE_r( did_match = 1 );
1809 if (regtry(®info, s)) goto got_it;
1811 while (s < strend && *s == ch)
1816 DEBUG_EXECUTE_r(if (!did_match)
1817 PerlIO_printf(Perl_debug_log,
1818 "Did not find anchored character...\n")
1821 else if (prog->anchored_substr != NULL
1822 || prog->anchored_utf8 != NULL
1823 || ((prog->float_substr != NULL || prog->float_utf8 != NULL)
1824 && prog->float_max_offset < strend - s)) {
1829 char *last1; /* Last position checked before */
1833 if (prog->anchored_substr || prog->anchored_utf8) {
1834 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1835 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1836 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1837 back_max = back_min = prog->anchored_offset;
1839 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1840 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1841 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1842 back_max = prog->float_max_offset;
1843 back_min = prog->float_min_offset;
1847 if (must == &PL_sv_undef)
1848 /* could not downgrade utf8 check substring, so must fail */
1854 last = HOP3c(strend, /* Cannot start after this */
1855 -(I32)(CHR_SVLEN(must)
1856 - (SvTAIL(must) != 0) + back_min), strbeg);
1859 last1 = HOPc(s, -1);
1861 last1 = s - 1; /* bogus */
1863 /* XXXX check_substr already used to find "s", can optimize if
1864 check_substr==must. */
1866 dontbother = end_shift;
1867 strend = HOPc(strend, -dontbother);
1868 while ( (s <= last) &&
1869 ((flags & REXEC_SCREAM)
1870 ? (s = screaminstr(sv, must, HOP3c(s, back_min, (back_min<0 ? strbeg : strend)) - strbeg,
1871 end_shift, &scream_pos, 0))
1872 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, (back_min<0 ? strbeg : strend)),
1873 (unsigned char*)strend, must,
1874 multiline ? FBMrf_MULTILINE : 0))) ) {
1875 /* we may be pointing at the wrong string */
1876 if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog))
1877 s = strbeg + (s - SvPVX_const(sv));
1878 DEBUG_EXECUTE_r( did_match = 1 );
1879 if (HOPc(s, -back_max) > last1) {
1880 last1 = HOPc(s, -back_min);
1881 s = HOPc(s, -back_max);
1884 char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1886 last1 = HOPc(s, -back_min);
1890 while (s <= last1) {
1891 if (regtry(®info, s))
1897 while (s <= last1) {
1898 if (regtry(®info, s))
1904 DEBUG_EXECUTE_r(if (!did_match) {
1905 RE_PV_QUOTED_DECL(quoted, do_utf8, PERL_DEBUG_PAD_ZERO(0),
1906 SvPVX_const(must), RE_SV_DUMPLEN(must), 30);
1907 PerlIO_printf(Perl_debug_log, "Did not find %s substr %s%s...\n",
1908 ((must == prog->anchored_substr || must == prog->anchored_utf8)
1909 ? "anchored" : "floating"),
1910 quoted, RE_SV_TAIL(must));
1914 else if ( (c = prog->regstclass) ) {
1916 const OPCODE op = OP(prog->regstclass);
1917 /* don't bother with what can't match */
1918 if (PL_regkind[op] != EXACT && op != CANY && PL_regkind[op] != TRIE)
1919 strend = HOPc(strend, -(minlen - 1));
1922 SV * const prop = sv_newmortal();
1923 regprop(prog, prop, c);
1925 RE_PV_QUOTED_DECL(quoted,UTF,PERL_DEBUG_PAD_ZERO(1),
1927 PerlIO_printf(Perl_debug_log,
1928 "Matching stclass %.*s against %s (%d chars)\n",
1929 (int)SvCUR(prop), SvPVX_const(prop),
1930 quoted, (int)(strend - s));
1933 if (find_byclass(prog, c, s, strend, ®info))
1935 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass... [regexec_flags]\n"));
1939 if (prog->float_substr != NULL || prog->float_utf8 != NULL) {
1944 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1945 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1946 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
1948 if (flags & REXEC_SCREAM) {
1949 last = screaminstr(sv, float_real, s - strbeg,
1950 end_shift, &scream_pos, 1); /* last one */
1952 last = scream_olds; /* Only one occurrence. */
1953 /* we may be pointing at the wrong string */
1954 else if (RX_MATCH_COPIED(prog))
1955 s = strbeg + (s - SvPVX_const(sv));
1959 const char * const little = SvPV_const(float_real, len);
1961 if (SvTAIL(float_real)) {
1962 if (memEQ(strend - len + 1, little, len - 1))
1963 last = strend - len + 1;
1964 else if (!multiline)
1965 last = memEQ(strend - len, little, len)
1966 ? strend - len : NULL;
1972 last = rninstr(s, strend, little, little + len);
1974 last = strend; /* matching "$" */
1978 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1979 "%sCan't trim the tail, match fails (should not happen)%s\n",
1980 PL_colors[4], PL_colors[5]));
1981 goto phooey; /* Should not happen! */
1983 dontbother = strend - last + prog->float_min_offset;
1985 if (minlen && (dontbother < minlen))
1986 dontbother = minlen - 1;
1987 strend -= dontbother; /* this one's always in bytes! */
1988 /* We don't know much -- general case. */
1991 if (regtry(®info, s))
2000 if (regtry(®info, s))
2002 } while (s++ < strend);
2010 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
2012 if (PL_reg_eval_set) {
2013 /* Preserve the current value of $^R */
2014 if (oreplsv != GvSV(PL_replgv))
2015 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
2016 restored, the value remains
2018 restore_pos(aTHX_ prog);
2021 /* make sure $`, $&, $', and $digit will work later */
2022 if ( !(flags & REXEC_NOT_FIRST) ) {
2023 RX_MATCH_COPY_FREE(prog);
2024 if (flags & REXEC_COPY_STR) {
2025 const I32 i = PL_regeol - startpos + (stringarg - strbeg);
2026 #ifdef PERL_OLD_COPY_ON_WRITE
2028 || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) {
2030 PerlIO_printf(Perl_debug_log,
2031 "Copy on write: regexp capture, type %d\n",
2034 prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv);
2035 prog->subbeg = (char *)SvPVX_const(prog->saved_copy);
2036 assert (SvPOKp(prog->saved_copy));
2040 RX_MATCH_COPIED_on(prog);
2041 s = savepvn(strbeg, i);
2047 prog->subbeg = strbeg;
2048 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2055 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2056 PL_colors[4], PL_colors[5]));
2057 if (PL_reg_eval_set)
2058 restore_pos(aTHX_ prog);
2063 - regtry - try match at specific point
2065 STATIC I32 /* 0 failure, 1 success */
2066 S_regtry(pTHX_ const regmatch_info *reginfo, char *startpos)
2072 regexp *prog = reginfo->prog;
2073 GET_RE_DEBUG_FLAGS_DECL;
2076 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
2078 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
2081 PL_reg_eval_set = RS_init;
2082 DEBUG_EXECUTE_r(DEBUG_s(
2083 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2084 (IV)(PL_stack_sp - PL_stack_base));
2086 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
2087 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2088 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2090 /* Apparently this is not needed, judging by wantarray. */
2091 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
2092 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2095 /* Make $_ available to executed code. */
2096 if (reginfo->sv != DEFSV) {
2098 DEFSV = reginfo->sv;
2101 if (!(SvTYPE(reginfo->sv) >= SVt_PVMG && SvMAGIC(reginfo->sv)
2102 && (mg = mg_find(reginfo->sv, PERL_MAGIC_regex_global)))) {
2103 /* prepare for quick setting of pos */
2104 #ifdef PERL_OLD_COPY_ON_WRITE
2106 sv_force_normal_flags(sv, 0);
2108 mg = sv_magicext(reginfo->sv, NULL, PERL_MAGIC_regex_global,
2109 &PL_vtbl_mglob, NULL, 0);
2113 PL_reg_oldpos = mg->mg_len;
2114 SAVEDESTRUCTOR_X(restore_pos, prog);
2116 if (!PL_reg_curpm) {
2117 Newxz(PL_reg_curpm, 1, PMOP);
2120 SV* const repointer = newSViv(0);
2121 /* so we know which PL_regex_padav element is PL_reg_curpm */
2122 SvFLAGS(repointer) |= SVf_BREAK;
2123 av_push(PL_regex_padav,repointer);
2124 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2125 PL_regex_pad = AvARRAY(PL_regex_padav);
2129 PM_SETRE(PL_reg_curpm, prog);
2130 PL_reg_oldcurpm = PL_curpm;
2131 PL_curpm = PL_reg_curpm;
2132 if (RX_MATCH_COPIED(prog)) {
2133 /* Here is a serious problem: we cannot rewrite subbeg,
2134 since it may be needed if this match fails. Thus
2135 $` inside (?{}) could fail... */
2136 PL_reg_oldsaved = prog->subbeg;
2137 PL_reg_oldsavedlen = prog->sublen;
2138 #ifdef PERL_OLD_COPY_ON_WRITE
2139 PL_nrs = prog->saved_copy;
2141 RX_MATCH_COPIED_off(prog);
2144 PL_reg_oldsaved = NULL;
2145 prog->subbeg = PL_bostr;
2146 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2148 prog->startp[0] = startpos - PL_bostr;
2149 PL_reginput = startpos;
2150 PL_regstartp = prog->startp;
2151 PL_regendp = prog->endp;
2152 PL_reglastparen = &prog->lastparen;
2153 PL_reglastcloseparen = &prog->lastcloseparen;
2154 prog->lastparen = 0;
2155 prog->lastcloseparen = 0;
2157 DEBUG_EXECUTE_r(PL_reg_starttry = startpos);
2158 if (PL_reg_start_tmpl <= prog->nparens) {
2159 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2160 if(PL_reg_start_tmp)
2161 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2163 Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2166 /* XXXX What this code is doing here?!!! There should be no need
2167 to do this again and again, PL_reglastparen should take care of
2170 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2171 * Actually, the code in regcppop() (which Ilya may be meaning by
2172 * PL_reglastparen), is not needed at all by the test suite
2173 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2174 * enough, for building DynaLoader, or otherwise this
2175 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2176 * will happen. Meanwhile, this code *is* needed for the
2177 * above-mentioned test suite tests to succeed. The common theme
2178 * on those tests seems to be returning null fields from matches.
2183 if (prog->nparens) {
2185 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2192 if (regmatch(reginfo, prog->program + 1)) {
2193 prog->endp[0] = PL_reginput - PL_bostr;
2196 REGCP_UNWIND(lastcp);
2201 #define sayYES goto yes
2202 #define sayNO goto no
2203 #define sayNO_ANYOF goto no_anyof
2204 #define sayYES_FINAL goto yes_final
2205 #define sayNO_FINAL goto no_final
2206 #define sayNO_SILENT goto do_no
2207 #define saySAME(x) if (x) goto yes; else goto no
2209 #define CACHEsayNO STMT_START { \
2210 if (st->u.whilem.cache_offset | st->u.whilem.cache_bit) \
2211 PL_reg_poscache[st->u.whilem.cache_offset] |= \
2212 (1<<st->u.whilem.cache_bit); \
2217 /* this is used to determine how far from the left messages like
2218 'failed...' are printed. Currently 29 makes these messages line
2219 up with the opcode they refer to. Earlier perls used 25 which
2220 left these messages outdented making reviewing a debug output
2223 #define REPORT_CODE_OFF 29
2226 /* Make sure there is a test for this +1 options in re_tests */
2227 #define TRIE_INITAL_ACCEPT_BUFFLEN 4;
2229 #define CHRTEST_UNINIT -1001 /* c1/c2 haven't been calculated yet */
2230 #define CHRTEST_VOID -1000 /* the c1/c2 "next char" test should be skipped */
2232 #define SLAB_FIRST(s) (&(s)->states[0])
2233 #define SLAB_LAST(s) (&(s)->states[PERL_REGMATCH_SLAB_SLOTS-1])
2235 /* grab a new slab and return the first slot in it */
2237 STATIC regmatch_state *
2240 #if PERL_VERSION < 9
2243 regmatch_slab *s = PL_regmatch_slab->next;
2245 Newx(s, 1, regmatch_slab);
2246 s->prev = PL_regmatch_slab;
2248 PL_regmatch_slab->next = s;
2250 PL_regmatch_slab = s;
2251 return SLAB_FIRST(s);
2254 /* simulate a recursive call to regmatch */
2256 #define REGMATCH(ns, where) \
2259 st->resume_state = resume_##where; \
2260 goto start_recurse; \
2261 resume_point_##where:
2263 /* push a new state then goto it */
2265 #define PUSH_STATE_GOTO(state, node) \
2267 st->resume_state = state; \
2270 /* push a new state with success backtracking, then goto it */
2272 #define PUSH_YES_STATE_GOTO(state, node) \
2274 st->resume_state = state; \
2275 goto push_yes_state;
2280 - regmatch - main matching routine
2282 * Conceptually the strategy is simple: check to see whether the current
2283 * node matches, call self recursively to see whether the rest matches,
2284 * and then act accordingly. In practice we make some effort to avoid
2285 * recursion, in particular by going through "ordinary" nodes (that don't
2286 * need to know whether the rest of the match failed) by a loop instead of
2289 /* [lwall] I've hoisted the register declarations to the outer block in order to
2290 * maybe save a little bit of pushing and popping on the stack. It also takes
2291 * advantage of machines that use a register save mask on subroutine entry.
2293 * This function used to be heavily recursive, but since this had the
2294 * effect of blowing the CPU stack on complex regexes, it has been
2295 * restructured to be iterative, and to save state onto the heap rather
2296 * than the stack. Essentially whereever regmatch() used to be called, it
2297 * pushes the current state, notes where to return, then jumps back into
2300 * Originally the structure of this function used to look something like
2305 while (scan != NULL) {
2306 a++; // do stuff with a and b
2312 if (regmatch(...)) // recurse
2322 * Now it looks something like this:
2330 regmatch_state *st = new();
2332 st->a++; // do stuff with a and b
2334 while (scan != NULL) {
2342 st->resume_state = resume_FOO;
2343 goto start_recurse; // recurse
2352 st = new(); push a new state
2353 st->a = 1; st->b = 2;
2360 switch (resume_state) {
2362 goto resume_point_FOO;
2369 * WARNING: this means that any line in this function that contains a
2370 * REGMATCH() or TRYPAREN() is actually simulating a recursive call to
2371 * regmatch() using gotos instead. Thus the values of any local variables
2372 * not saved in the regmatch_state structure will have been lost when
2373 * execution resumes on the next line .
2375 * States (ie the st pointer) are allocated in slabs of about 4K in size.
2376 * PL_regmatch_state always points to the currently active state, and
2377 * PL_regmatch_slab points to the slab currently containing PL_regmatch_state.
2378 * The first time regmatch is called, the first slab is allocated, and is
2379 * never freed until interpreter desctruction. When the slab is full,
2380 * a new one is allocated chained to the end. At exit from regmatch, slabs
2381 * allocated since entry are freed.
2384 /* *** every FOO_fail should = FOO+1 */
2385 #define TRIE_next (REGNODE_MAX+1)
2386 #define TRIE_next_fail (REGNODE_MAX+2)
2387 #define EVAL_A (REGNODE_MAX+3)
2388 #define EVAL_A_fail (REGNODE_MAX+4)
2389 #define resume_CURLYX (REGNODE_MAX+5)
2390 #define resume_WHILEM1 (REGNODE_MAX+6)
2391 #define resume_WHILEM2 (REGNODE_MAX+7)
2392 #define resume_WHILEM3 (REGNODE_MAX+8)
2393 #define resume_WHILEM4 (REGNODE_MAX+9)
2394 #define resume_WHILEM5 (REGNODE_MAX+10)
2395 #define resume_WHILEM6 (REGNODE_MAX+11)
2396 #define BRANCH_next (REGNODE_MAX+12)
2397 #define BRANCH_next_fail (REGNODE_MAX+13)
2398 #define CURLYM_A (REGNODE_MAX+14)
2399 #define CURLYM_A_fail (REGNODE_MAX+15)
2400 #define CURLYM_B (REGNODE_MAX+16)
2401 #define CURLYM_B_fail (REGNODE_MAX+17)
2402 #define IFMATCH_A (REGNODE_MAX+18)
2403 #define IFMATCH_A_fail (REGNODE_MAX+19)
2404 #define CURLY_B_min_known (REGNODE_MAX+20)
2405 #define CURLY_B_min_known_fail (REGNODE_MAX+21)
2406 #define CURLY_B_min (REGNODE_MAX+22)
2407 #define CURLY_B_min_fail (REGNODE_MAX+23)
2408 #define CURLY_B_max (REGNODE_MAX+24)
2409 #define CURLY_B_max_fail (REGNODE_MAX+25)
2412 #define REG_NODE_NUM(x) ((x) ? (int)((x)-prog) : -1)
2416 S_debug_start_match(pTHX_ const regexp *prog, const bool do_utf8,
2417 const char *start, const char *end, const char *blurb)
2419 const bool utf8_pat= prog->reganch & ROPT_UTF8 ? 1 : 0;
2423 RE_PV_QUOTED_DECL(s0, utf8_pat, PERL_DEBUG_PAD_ZERO(0),
2424 prog->precomp, prog->prelen, 60);
2426 RE_PV_QUOTED_DECL(s1, do_utf8, PERL_DEBUG_PAD_ZERO(1),
2427 start, end - start, 60);
2429 PerlIO_printf(Perl_debug_log,
2430 "%s%s REx%s %s against %s\n",
2431 PL_colors[4], blurb, PL_colors[5], s0, s1);
2433 if (do_utf8||utf8_pat)
2434 PerlIO_printf(Perl_debug_log, "UTF-8 %s%s%s...\n",
2435 utf8_pat ? "pattern" : "",
2436 utf8_pat && do_utf8 ? " and " : "",
2437 do_utf8 ? "string" : ""
2443 S_dump_exec_pos(pTHX_ const char *locinput,
2444 const regnode *scan,
2445 const char *loc_regeol,
2446 const char *loc_bostr,
2447 const char *loc_reg_starttry,
2450 const int docolor = *PL_colors[0] || *PL_colors[2] || *PL_colors[4];
2451 const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2452 int l = (loc_regeol - locinput) > taill ? taill : (loc_regeol - locinput);
2453 /* The part of the string before starttry has one color
2454 (pref0_len chars), between starttry and current
2455 position another one (pref_len - pref0_len chars),
2456 after the current position the third one.
2457 We assume that pref0_len <= pref_len, otherwise we
2458 decrease pref0_len. */
2459 int pref_len = (locinput - loc_bostr) > (5 + taill) - l
2460 ? (5 + taill) - l : locinput - loc_bostr;
2463 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2465 pref0_len = pref_len - (locinput - loc_reg_starttry);
2466 if (l + pref_len < (5 + taill) && l < loc_regeol - locinput)
2467 l = ( loc_regeol - locinput > (5 + taill) - pref_len
2468 ? (5 + taill) - pref_len : loc_regeol - locinput);
2469 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2473 if (pref0_len > pref_len)
2474 pref0_len = pref_len;
2476 const int is_uni = (do_utf8 && OP(scan) != CANY) ? 1 : 0;
2478 RE_PV_COLOR_DECL(s0,len0,is_uni,PERL_DEBUG_PAD(0),
2479 (locinput - pref_len),pref0_len, 60, 4, 5);
2481 RE_PV_COLOR_DECL(s1,len1,is_uni,PERL_DEBUG_PAD(1),
2482 (locinput - pref_len + pref0_len),
2483 pref_len - pref0_len, 60, 2, 3);
2485 RE_PV_COLOR_DECL(s2,len2,is_uni,PERL_DEBUG_PAD(2),
2486 locinput, loc_regeol - locinput, 10, 0, 1);
2488 const STRLEN tlen=len0+len1+len2;
2489 PerlIO_printf(Perl_debug_log,
2490 "%4"IVdf" <%.*s%.*s%s%.*s>%*s|",
2491 (IV)(locinput - loc_bostr),
2494 (docolor ? "" : "> <"),
2496 tlen > 19 ? 0 : 19 - tlen,
2503 STATIC I32 /* 0 failure, 1 success */
2504 S_regmatch(pTHX_ const regmatch_info *reginfo, regnode *prog)
2506 #if PERL_VERSION < 9
2510 register const bool do_utf8 = PL_reg_match_utf8;
2511 const U32 uniflags = UTF8_ALLOW_DEFAULT;
2513 regexp *rex = reginfo->prog;
2515 regmatch_slab *orig_slab;
2516 regmatch_state *orig_state;
2518 /* the current state. This is a cached copy of PL_regmatch_state */
2519 register regmatch_state *st;
2521 /* cache heavy used fields of st in registers */
2522 register regnode *scan;
2523 register regnode *next;
2524 register I32 n = 0; /* initialize to shut up compiler warning */
2525 register char *locinput = PL_reginput;
2527 /* these variables are NOT saved during a recusive RFEGMATCH: */
2528 register I32 nextchr; /* is always set to UCHARAT(locinput) */
2529 bool result = 0; /* return value of S_regmatch */
2530 int depth = 0; /* depth of recursion */
2531 regmatch_state *yes_state = NULL; /* state to pop to on success of
2538 GET_RE_DEBUG_FLAGS_DECL;
2542 /* on first ever call to regmatch, allocate first slab */
2543 if (!PL_regmatch_slab) {
2544 Newx(PL_regmatch_slab, 1, regmatch_slab);
2545 PL_regmatch_slab->prev = NULL;
2546 PL_regmatch_slab->next = NULL;
2547 PL_regmatch_state = SLAB_FIRST(PL_regmatch_slab);
2550 /* remember current high-water mark for exit */
2551 /* XXX this should be done with SAVE* instead */
2552 orig_slab = PL_regmatch_slab;
2553 orig_state = PL_regmatch_state;
2555 /* grab next free state slot */
2556 st = ++PL_regmatch_state;
2557 if (st > SLAB_LAST(PL_regmatch_slab))
2558 st = PL_regmatch_state = S_push_slab(aTHX);
2565 /* Note that nextchr is a byte even in UTF */
2566 nextchr = UCHARAT(locinput);
2568 while (scan != NULL) {
2571 SV * const prop = sv_newmortal();
2572 regnode *rnext=regnext(scan);
2573 DUMP_EXEC_POS( locinput, scan, do_utf8 );
2574 regprop(rex, prop, scan);
2576 PerlIO_printf(Perl_debug_log,
2577 "%3"IVdf":%*s%s(%"IVdf")\n",
2578 (IV)(scan - rex->program), PL_regindent*2, "",
2580 (PL_regkind[OP(scan)] == END || !rnext) ?
2581 0 : (IV)(rnext - rex->program));
2584 next = scan + NEXT_OFF(scan);
2587 state_num = OP(scan);
2590 switch (state_num) {
2592 if (locinput == PL_bostr)
2594 /* reginfo->till = reginfo->bol; */
2599 if (locinput == PL_bostr ||
2600 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2606 if (locinput == PL_bostr)
2610 if (locinput == reginfo->ganch)
2616 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2621 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2623 if (PL_regeol - locinput > 1)
2627 if (PL_regeol != locinput)
2631 if (!nextchr && locinput >= PL_regeol)
2634 locinput += PL_utf8skip[nextchr];
2635 if (locinput > PL_regeol)
2637 nextchr = UCHARAT(locinput);
2640 nextchr = UCHARAT(++locinput);
2643 if (!nextchr && locinput >= PL_regeol)
2645 nextchr = UCHARAT(++locinput);
2648 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2651 locinput += PL_utf8skip[nextchr];
2652 if (locinput > PL_regeol)
2654 nextchr = UCHARAT(locinput);
2657 nextchr = UCHARAT(++locinput);
2661 #define ST st->u.trie
2663 /* In this case the charclass data is available inline so
2664 we can fail fast without a lot of extra overhead.
2666 if (scan->flags == EXACT || !do_utf8) {
2667 if(!ANYOF_BITMAP_TEST(scan, *locinput)) {
2669 PerlIO_printf(Perl_debug_log,
2670 "%*s %sfailed to match trie start class...%s\n",
2671 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5])
2680 /* what type of TRIE am I? (utf8 makes this contextual) */
2681 const enum { trie_plain, trie_utf8, trie_utf8_fold }
2682 trie_type = do_utf8 ?
2683 (scan->flags == EXACT ? trie_utf8 : trie_utf8_fold)
2686 /* what trie are we using right now */
2687 reg_trie_data * const trie
2688 = (reg_trie_data*)rex->data->data[ ARG( scan ) ];
2689 U32 state = trie->startstate;
2691 if (trie->bitmap && trie_type != trie_utf8_fold &&
2692 !TRIE_BITMAP_TEST(trie,*locinput)
2694 if (trie->states[ state ].wordnum) {
2696 PerlIO_printf(Perl_debug_log,
2697 "%*s %smatched empty string...%s\n",
2698 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5])
2703 PerlIO_printf(Perl_debug_log,
2704 "%*s %sfailed to match trie start class...%s\n",
2705 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5])
2712 U8 *uc = ( U8* )locinput;
2716 U8 *uscan = (U8*)NULL;
2718 SV *sv_accept_buff = NULL;
2719 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
2721 ST.accepted = 0; /* how many accepting states we have seen */
2723 ST.jump = trie->jump;
2732 traverse the TRIE keeping track of all accepting states
2733 we transition through until we get to a failing node.
2736 while ( state && uc <= (U8*)PL_regeol ) {
2737 U32 base = trie->states[ state ].trans.base;
2740 /* We use charid to hold the wordnum as we don't use it
2741 for charid until after we have done the wordnum logic.
2742 We define an alias just so that the wordnum logic reads
2745 #define got_wordnum charid
2746 got_wordnum = trie->states[ state ].wordnum;
2748 if ( got_wordnum ) {
2749 if ( ! ST.accepted ) {
2752 bufflen = TRIE_INITAL_ACCEPT_BUFFLEN;
2753 sv_accept_buff=newSV(bufflen *
2754 sizeof(reg_trie_accepted) - 1);
2755 SvCUR_set(sv_accept_buff, 0);
2756 SvPOK_on(sv_accept_buff);
2757 sv_2mortal(sv_accept_buff);
2760 (reg_trie_accepted*)SvPV_nolen(sv_accept_buff );
2763 if (ST.accepted >= bufflen) {
2765 ST.accept_buff =(reg_trie_accepted*)
2766 SvGROW(sv_accept_buff,
2767 bufflen * sizeof(reg_trie_accepted));
2769 SvCUR_set(sv_accept_buff,SvCUR(sv_accept_buff)
2770 + sizeof(reg_trie_accepted));
2773 ST.accept_buff[ST.accepted].wordnum = got_wordnum;
2774 ST.accept_buff[ST.accepted].endpos = uc;
2776 } while (trie->nextword && (got_wordnum= trie->nextword[got_wordnum]));
2780 DEBUG_TRIE_EXECUTE_r({
2781 DUMP_EXEC_POS( (char *)uc, scan, do_utf8 );
2782 PerlIO_printf( Perl_debug_log,
2783 "%*s %sState: %4"UVxf" Accepted: %4"UVxf" ",
2784 2+PL_regindent * 2, "", PL_colors[4],
2785 (UV)state, (UV)ST.accepted );
2789 REXEC_TRIE_READ_CHAR(trie_type, trie, uc, uscan, len,
2790 uvc, charid, foldlen, foldbuf, uniflags);
2793 (base + charid > trie->uniquecharcount )
2794 && (base + charid - 1 - trie->uniquecharcount
2796 && trie->trans[base + charid - 1 -
2797 trie->uniquecharcount].check == state)
2799 state = trie->trans[base + charid - 1 -
2800 trie->uniquecharcount ].next;
2811 DEBUG_TRIE_EXECUTE_r(
2812 PerlIO_printf( Perl_debug_log,
2813 "Charid:%3x CP:%4"UVxf" After State: %4"UVxf"%s\n",
2814 charid, uvc, (UV)state, PL_colors[5] );
2821 PerlIO_printf( Perl_debug_log,
2822 "%*s %sgot %"IVdf" possible matches%s\n",
2823 REPORT_CODE_OFF + PL_regindent * 2, "",
2824 PL_colors[4], (IV)ST.accepted, PL_colors[5] );
2830 case TRIE_next_fail: /* we failed - try next alterative */
2832 if ( ST.accepted == 1 ) {
2833 /* only one choice left - just continue */
2835 reg_trie_data * const trie
2836 = (reg_trie_data*)rex->data->data[ ARG(ST.me) ];
2837 SV ** const tmp = RX_DEBUG(reginfo->prog)
2838 ? av_fetch( trie->words, ST.accept_buff[ 0 ].wordnum-1, 0 )
2840 PerlIO_printf( Perl_debug_log,
2841 "%*s %sonly one match left: #%d <%s>%s\n",
2842 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],
2843 ST.accept_buff[ 0 ].wordnum,
2844 tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr",
2847 PL_reginput = (char *)ST.accept_buff[ 0 ].endpos;
2848 /* in this case we free tmps/leave before we call regmatch
2849 as we wont be using accept_buff again. */
2852 locinput = PL_reginput;
2853 nextchr = UCHARAT(locinput);
2858 scan = ST.B - ST.jump[ST.accept_buff[0].wordnum];
2860 continue; /* execute rest of RE */
2863 if (!ST.accepted-- ) {
2870 There are at least two accepting states left. Presumably
2871 the number of accepting states is going to be low,
2872 typically two. So we simply scan through to find the one
2873 with lowest wordnum. Once we find it, we swap the last
2874 state into its place and decrement the size. We then try to
2875 match the rest of the pattern at the point where the word
2876 ends. If we succeed, control just continues along the
2877 regex; if we fail we return here to try the next accepting
2884 for( cur = 1 ; cur <= ST.accepted ; cur++ ) {
2885 DEBUG_TRIE_EXECUTE_r(
2886 PerlIO_printf( Perl_debug_log,
2887 "%*s %sgot %"IVdf" (%d) as best, looking at %"IVdf" (%d)%s\n",
2888 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
2889 (IV)best, ST.accept_buff[ best ].wordnum, (IV)cur,
2890 ST.accept_buff[ cur ].wordnum, PL_colors[5] );
2893 if (ST.accept_buff[cur].wordnum <
2894 ST.accept_buff[best].wordnum)
2899 reg_trie_data * const trie
2900 = (reg_trie_data*)rex->data->data[ ARG(ST.me) ];
2901 SV ** const tmp = RX_DEBUG(reginfo->prog)
2902 ? av_fetch( trie->words, ST.accept_buff[ best ].wordnum - 1, 0 )
2904 PerlIO_printf( Perl_debug_log, "%*s %strying alternation #%d <%s> at node #%d %s\n",
2905 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],
2906 ST.accept_buff[best].wordnum,
2907 tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr", REG_NODE_NUM(scan),
2911 if ( best<ST.accepted ) {
2912 reg_trie_accepted tmp = ST.accept_buff[ best ];
2913 ST.accept_buff[ best ] = ST.accept_buff[ ST.accepted ];
2914 ST.accept_buff[ ST.accepted ] = tmp;
2917 PL_reginput = (char *)ST.accept_buff[ best ].endpos;
2919 PUSH_STATE_GOTO(TRIE_next, ST.B);
2922 PUSH_STATE_GOTO(TRIE_next, ST.B - ST.jump[ST.accept_buff[best].wordnum]);
2932 char *s = STRING(scan);
2933 st->ln = STR_LEN(scan);
2934 if (do_utf8 != UTF) {
2935 /* The target and the pattern have differing utf8ness. */
2937 const char * const e = s + st->ln;
2940 /* The target is utf8, the pattern is not utf8. */
2945 if (NATIVE_TO_UNI(*(U8*)s) !=
2946 utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen,
2954 /* The target is not utf8, the pattern is utf8. */
2959 if (NATIVE_TO_UNI(*((U8*)l)) !=
2960 utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen,
2968 nextchr = UCHARAT(locinput);
2971 /* The target and the pattern have the same utf8ness. */
2972 /* Inline the first character, for speed. */
2973 if (UCHARAT(s) != nextchr)
2975 if (PL_regeol - locinput < st->ln)
2977 if (st->ln > 1 && memNE(s, locinput, st->ln))
2980 nextchr = UCHARAT(locinput);
2984 PL_reg_flags |= RF_tainted;
2987 char * const s = STRING(scan);
2988 st->ln = STR_LEN(scan);
2990 if (do_utf8 || UTF) {
2991 /* Either target or the pattern are utf8. */
2992 const char * const l = locinput;
2993 char *e = PL_regeol;
2995 if (ibcmp_utf8(s, 0, st->ln, (bool)UTF,
2996 l, &e, 0, do_utf8)) {
2997 /* One more case for the sharp s:
2998 * pack("U0U*", 0xDF) =~ /ss/i,
2999 * the 0xC3 0x9F are the UTF-8
3000 * byte sequence for the U+00DF. */
3002 toLOWER(s[0]) == 's' &&
3004 toLOWER(s[1]) == 's' &&
3011 nextchr = UCHARAT(locinput);
3015 /* Neither the target and the pattern are utf8. */
3017 /* Inline the first character, for speed. */
3018 if (UCHARAT(s) != nextchr &&
3019 UCHARAT(s) != ((OP(scan) == EXACTF)
3020 ? PL_fold : PL_fold_locale)[nextchr])
3022 if (PL_regeol - locinput < st->ln)
3024 if (st->ln > 1 && (OP(scan) == EXACTF
3025 ? ibcmp(s, locinput, st->ln)
3026 : ibcmp_locale(s, locinput, st->ln)))
3029 nextchr = UCHARAT(locinput);
3034 STRLEN inclasslen = PL_regeol - locinput;
3036 if (!reginclass(rex, scan, (U8*)locinput, &inclasslen, do_utf8))
3038 if (locinput >= PL_regeol)
3040 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
3041 nextchr = UCHARAT(locinput);
3046 nextchr = UCHARAT(locinput);
3047 if (!REGINCLASS(rex, scan, (U8*)locinput))
3049 if (!nextchr && locinput >= PL_regeol)
3051 nextchr = UCHARAT(++locinput);
3055 /* If we might have the case of the German sharp s
3056 * in a casefolding Unicode character class. */
3058 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
3059 locinput += SHARP_S_SKIP;
3060 nextchr = UCHARAT(locinput);
3066 PL_reg_flags |= RF_tainted;
3072 LOAD_UTF8_CHARCLASS_ALNUM();
3073 if (!(OP(scan) == ALNUM
3074 ? (bool)swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
3075 : isALNUM_LC_utf8((U8*)locinput)))
3079 locinput += PL_utf8skip[nextchr];
3080 nextchr = UCHARAT(locinput);
3083 if (!(OP(scan) == ALNUM
3084 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
3086 nextchr = UCHARAT(++locinput);
3089 PL_reg_flags |= RF_tainted;
3092 if (!nextchr && locinput >= PL_regeol)
3095 LOAD_UTF8_CHARCLASS_ALNUM();
3096 if (OP(scan) == NALNUM
3097 ? (bool)swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
3098 : isALNUM_LC_utf8((U8*)locinput))
3102 locinput += PL_utf8skip[nextchr];
3103 nextchr = UCHARAT(locinput);
3106 if (OP(scan) == NALNUM
3107 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
3109 nextchr = UCHARAT(++locinput);
3113 PL_reg_flags |= RF_tainted;
3117 /* was last char in word? */
3119 if (locinput == PL_bostr)
3122 const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
3124 st->ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags);
3126 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
3127 st->ln = isALNUM_uni(st->ln);
3128 LOAD_UTF8_CHARCLASS_ALNUM();
3129 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
3132 st->ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(st->ln));
3133 n = isALNUM_LC_utf8((U8*)locinput);
3137 st->ln = (locinput != PL_bostr) ?
3138 UCHARAT(locinput - 1) : '\n';
3139 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
3140 st->ln = isALNUM(st->ln);
3141 n = isALNUM(nextchr);
3144 st->ln = isALNUM_LC(st->ln);
3145 n = isALNUM_LC(nextchr);
3148 if (((!st->ln) == (!n)) == (OP(scan) == BOUND ||
3149 OP(scan) == BOUNDL))
3153 PL_reg_flags |= RF_tainted;
3159 if (UTF8_IS_CONTINUED(nextchr)) {
3160 LOAD_UTF8_CHARCLASS_SPACE();
3161 if (!(OP(scan) == SPACE
3162 ? (bool)swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
3163 : isSPACE_LC_utf8((U8*)locinput)))
3167 locinput += PL_utf8skip[nextchr];
3168 nextchr = UCHARAT(locinput);
3171 if (!(OP(scan) == SPACE
3172 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3174 nextchr = UCHARAT(++locinput);
3177 if (!(OP(scan) == SPACE
3178 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3180 nextchr = UCHARAT(++locinput);
3184 PL_reg_flags |= RF_tainted;
3187 if (!nextchr && locinput >= PL_regeol)
3190 LOAD_UTF8_CHARCLASS_SPACE();
3191 if (OP(scan) == NSPACE
3192 ? (bool)swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
3193 : isSPACE_LC_utf8((U8*)locinput))
3197 locinput += PL_utf8skip[nextchr];
3198 nextchr = UCHARAT(locinput);
3201 if (OP(scan) == NSPACE
3202 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
3204 nextchr = UCHARAT(++locinput);
3207 PL_reg_flags |= RF_tainted;
3213 LOAD_UTF8_CHARCLASS_DIGIT();
3214 if (!(OP(scan) == DIGIT
3215 ? (bool)swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
3216 : isDIGIT_LC_utf8((U8*)locinput)))
3220 locinput += PL_utf8skip[nextchr];
3221 nextchr = UCHARAT(locinput);
3224 if (!(OP(scan) == DIGIT
3225 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
3227 nextchr = UCHARAT(++locinput);
3230 PL_reg_flags |= RF_tainted;
3233 if (!nextchr && locinput >= PL_regeol)
3236 LOAD_UTF8_CHARCLASS_DIGIT();
3237 if (OP(scan) == NDIGIT
3238 ? (bool)swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
3239 : isDIGIT_LC_utf8((U8*)locinput))
3243 locinput += PL_utf8skip[nextchr];
3244 nextchr = UCHARAT(locinput);
3247 if (OP(scan) == NDIGIT
3248 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
3250 nextchr = UCHARAT(++locinput);
3253 if (locinput >= PL_regeol)
3256 LOAD_UTF8_CHARCLASS_MARK();
3257 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3259 locinput += PL_utf8skip[nextchr];
3260 while (locinput < PL_regeol &&
3261 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3262 locinput += UTF8SKIP(locinput);
3263 if (locinput > PL_regeol)
3268 nextchr = UCHARAT(locinput);
3271 PL_reg_flags |= RF_tainted;
3276 n = ARG(scan); /* which paren pair */
3277 st->ln = PL_regstartp[n];
3278 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3279 if ((I32)*PL_reglastparen < n || st->ln == -1)
3280 sayNO; /* Do not match unless seen CLOSEn. */
3281 if (st->ln == PL_regendp[n])
3284 s = PL_bostr + st->ln;
3285 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
3287 const char *e = PL_bostr + PL_regendp[n];
3289 * Note that we can't do the "other character" lookup trick as
3290 * in the 8-bit case (no pun intended) because in Unicode we
3291 * have to map both upper and title case to lower case.
3293 if (OP(scan) == REFF) {
3295 STRLEN ulen1, ulen2;
3296 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
3297 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
3301 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
3302 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
3303 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
3310 nextchr = UCHARAT(locinput);
3314 /* Inline the first character, for speed. */
3315 if (UCHARAT(s) != nextchr &&
3317 (UCHARAT(s) != ((OP(scan) == REFF
3318 ? PL_fold : PL_fold_locale)[nextchr]))))
3320 st->ln = PL_regendp[n] - st->ln;
3321 if (locinput + st->ln > PL_regeol)
3323 if (st->ln > 1 && (OP(scan) == REF
3324 ? memNE(s, locinput, st->ln)
3326 ? ibcmp(s, locinput, st->ln)
3327 : ibcmp_locale(s, locinput, st->ln))))
3330 nextchr = UCHARAT(locinput);
3341 #define ST st->u.eval
3343 case EVAL: /* /(?{A})B/ /(??{A})B/ and /(?(?{A})X|Y)B/ */
3347 /* execute the code in the {...} */
3349 SV ** const before = SP;
3350 OP_4tree * const oop = PL_op;
3351 COP * const ocurcop = PL_curcop;
3355 PL_op = (OP_4tree*)rex->data->data[n];
3356 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
3357 PAD_SAVE_LOCAL(old_comppad, (PAD*)rex->data->data[n + 2]);
3358 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
3360 CALLRUNOPS(aTHX); /* Scalar context. */
3363 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
3370 PAD_RESTORE_LOCAL(old_comppad);
3371 PL_curcop = ocurcop;
3374 sv_setsv(save_scalar(PL_replgv), ret);
3378 if (st->logical == 2) { /* Postponed subexpression: /(??{...})/ */
3381 /* extract RE object from returned value; compiling if
3386 if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret)))
3387 mg = mg_find(sv, PERL_MAGIC_qr);
3388 else if (SvSMAGICAL(ret)) {
3389 if (SvGMAGICAL(ret))
3390 sv_unmagic(ret, PERL_MAGIC_qr);
3392 mg = mg_find(ret, PERL_MAGIC_qr);
3396 re = (regexp *)mg->mg_obj;
3397 (void)ReREFCNT_inc(re);
3401 const char * const t = SvPV_const(ret, len);
3403 const I32 osize = PL_regsize;
3406 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
3407 re = CALLREGCOMP(aTHX_ (char*)t, (char*)t + len, &pm);
3409 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
3411 sv_magic(ret,(SV*)ReREFCNT_inc(re),
3417 /* run the pattern returned from (??{...}) */
3418 ST.cp = regcppush(0); /* Save *all* the positions. */
3419 REGCP_SET(ST.lastcp);
3420 *PL_reglastparen = 0;
3421 *PL_reglastcloseparen = 0;
3422 PL_reginput = locinput;
3424 /* XXXX This is too dramatic a measure... */
3428 ST.toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
3429 ((re->reganch & ROPT_UTF8) != 0);
3430 if (ST.toggleutf) PL_reg_flags ^= RF_utf8;
3436 debug_start_match(re, do_utf8, locinput, PL_regeol,
3437 "Matching embedded");
3439 /* now continue from first node in postoned RE */
3440 PUSH_YES_STATE_GOTO(EVAL_A, re->program + 1);
3443 /* /(?(?{...})X|Y)/ */
3444 st->sw = SvTRUE(ret);
3449 case EVAL_A: /* successfully ran inner rex (??{rex}) */
3451 PL_reg_flags ^= RF_utf8;
3454 /* XXXX This is too dramatic a measure... */
3456 /* Restore parens of the caller without popping the
3459 const I32 tmp = PL_savestack_ix;
3460 PL_savestack_ix = ST.lastcp;
3462 PL_savestack_ix = tmp;
3464 PL_reginput = locinput;
3465 /* continue at the node following the (??{...}) */
3468 case EVAL_A_fail: /* unsuccessfully ran inner rex (??{rex}) */
3469 /* Restore state to the outer re then re-throw the failure */
3471 PL_reg_flags ^= RF_utf8;
3475 /* XXXX This is too dramatic a measure... */
3478 PL_reginput = locinput;
3479 REGCP_UNWIND(ST.lastcp);
3486 n = ARG(scan); /* which paren pair */
3487 PL_reg_start_tmp[n] = locinput;
3492 n = ARG(scan); /* which paren pair */
3493 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
3494 PL_regendp[n] = locinput - PL_bostr;
3495 if (n > (I32)*PL_reglastparen)
3496 *PL_reglastparen = n;
3497 *PL_reglastcloseparen = n;
3500 n = ARG(scan); /* which paren pair */
3501 st->sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
3504 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3506 next = NEXTOPER(NEXTOPER(scan));
3508 next = scan + ARG(scan);
3509 if (OP(next) == IFTHEN) /* Fake one. */
3510 next = NEXTOPER(NEXTOPER(next));
3514 st->logical = scan->flags;
3516 /*******************************************************************
3517 cc points to the regmatch_state associated with the most recent CURLYX.
3518 This struct contains info about the innermost (...)* loop (an
3519 "infoblock"), and a pointer to the next outer cc.
3521 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
3523 1) After matching Y, regnode for CURLYX is processed;
3525 2) This regnode populates cc, and calls regmatch() recursively
3526 with the starting point at WHILEM node;
3528 3) Each hit of WHILEM node tries to match A and Z (in the order
3529 depending on the current iteration, min/max of {min,max} and
3530 greediness). The information about where are nodes for "A"
3531 and "Z" is read from cc, as is info on how many times "A"
3532 was already matched, and greediness.
3534 4) After A matches, the same WHILEM node is hit again.
3536 5) Each time WHILEM is hit, cc is the infoblock created by CURLYX
3537 of the same pair. Thus when WHILEM tries to match Z, it temporarily
3538 resets cc, since this Y(A)*Z can be a part of some other loop:
3539 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
3540 of the external loop.
3542 Currently present infoblocks form a tree with a stem formed by st->cc
3543 and whatever it mentions via ->next, and additional attached trees
3544 corresponding to temporarily unset infoblocks as in "5" above.
3546 In the following picture, infoblocks for outer loop of
3547 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
3548 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
3549 infoblocks are drawn below the "reset" infoblock.
3551 In fact in the picture below we do not show failed matches for Z and T
3552 by WHILEM blocks. [We illustrate minimal matches, since for them it is
3553 more obvious *why* one needs to *temporary* unset infoblocks.]
3555 Matched REx position InfoBlocks Comment
3559 Y A)*?Z)*?T x <- O <- I
3560 YA )*?Z)*?T x <- O <- I
3561 YA A)*?Z)*?T x <- O <- I
3562 YAA )*?Z)*?T x <- O <- I
3563 YAA Z)*?T x <- O # Temporary unset I
3566 YAAZ Y(A)*?Z)*?T x <- O
3569 YAAZY (A)*?Z)*?T x <- O
3572 YAAZY A)*?Z)*?T x <- O <- I
3575 YAAZYA )*?Z)*?T x <- O <- I
3578 YAAZYA Z)*?T x <- O # Temporary unset I
3584 YAAZYAZ T x # Temporary unset O
3591 *******************************************************************/
3594 /* No need to save/restore up to this paren */
3595 parenfloor = scan->flags;
3599 CURLYX and WHILEM are always paired: they're the moral
3600 equivalent of pp_enteriter anbd pp_iter.
3602 The only time next could be null is if the node tree is
3603 corrupt. This was mentioned on p5p a few days ago.
3605 See http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2006-04/msg00556.html
3606 So we'll assert that this is true:
3609 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3611 /* XXXX Probably it is better to teach regpush to support
3612 parenfloor > PL_regsize... */
3613 if (parenfloor > (I32)*PL_reglastparen)
3614 parenfloor = *PL_reglastparen; /* Pessimization... */
3616 st->u.curlyx.cp = PL_savestack_ix;
3617 st->u.curlyx.outercc = st->cc;
3619 /* these fields contain the state of the current curly.
3620 * they are accessed by subsequent WHILEMs;
3621 * cur and lastloc are also updated by WHILEM */
3622 st->u.curlyx.parenfloor = parenfloor;
3623 st->u.curlyx.cur = -1; /* this will be updated by WHILEM */
3624 st->u.curlyx.min = ARG1(scan);
3625 st->u.curlyx.max = ARG2(scan);
3626 st->u.curlyx.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3627 st->u.curlyx.lastloc = 0;
3628 /* st->next and st->minmod are also read by WHILEM */
3630 PL_reginput = locinput;
3631 REGMATCH(PREVOPER(next), CURLYX); /* start on the WHILEM */
3632 /*** all unsaved local vars undefined at this point */
3633 regcpblow(st->u.curlyx.cp);
3634 st->cc = st->u.curlyx.outercc;
3640 * This is really hard to understand, because after we match
3641 * what we're trying to match, we must make sure the rest of
3642 * the REx is going to match for sure, and to do that we have
3643 * to go back UP the parse tree by recursing ever deeper. And
3644 * if it fails, we have to reset our parent's current state
3645 * that we can try again after backing off.
3650 st->cc gets initialised by CURLYX ready for use by WHILEM.
3651 So again, unless somethings been corrupted, st->cc cannot
3652 be null at that point in WHILEM.
3654 See http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2006-04/msg00556.html
3655 So we'll assert that this is true:
3658 st->u.whilem.lastloc = st->cc->u.curlyx.lastloc; /* Detection of 0-len. */
3659 st->u.whilem.cache_offset = 0;
3660 st->u.whilem.cache_bit = 0;
3662 n = st->cc->u.curlyx.cur + 1; /* how many we know we matched */
3663 PL_reginput = locinput;
3666 PerlIO_printf(Perl_debug_log,
3667 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3668 REPORT_CODE_OFF+PL_regindent*2, "",
3669 (long)n, (long)st->cc->u.curlyx.min,
3670 (long)st->cc->u.curlyx.max, PTR2UV(st->cc))
3673 /* If degenerate scan matches "", assume scan done. */
3675 if (locinput == st->cc->u.curlyx.lastloc && n >= st->cc->u.curlyx.min) {
3676 st->u.whilem.savecc = st->cc;
3677 st->cc = st->cc->u.curlyx.outercc;
3679 st->ln = st->cc->u.curlyx.cur;
3681 PerlIO_printf(Perl_debug_log,
3682 "%*s empty match detected, try continuation...\n",
3683 REPORT_CODE_OFF+PL_regindent*2, "")
3685 REGMATCH(st->u.whilem.savecc->next, WHILEM1);
3686 /*** all unsaved local vars undefined at this point */
3687 st->cc = st->u.whilem.savecc;
3690 if (st->cc->u.curlyx.outercc)
3691 st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln;
3695 /* First just match a string of min scans. */
3697 if (n < st->cc->u.curlyx.min) {
3698 st->cc->u.curlyx.cur = n;
3699 st->cc->u.curlyx.lastloc = locinput;
3700 REGMATCH(st->cc->u.curlyx.scan, WHILEM2);
3701 /*** all unsaved local vars undefined at this point */
3704 st->cc->u.curlyx.cur = n - 1;
3705 st->cc->u.curlyx.lastloc = st->u.whilem.lastloc;
3710 /* Check whether we already were at this position.
3711 Postpone detection until we know the match is not
3712 *that* much linear. */
3713 if (!PL_reg_maxiter) {
3714 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3715 /* possible overflow for long strings and many CURLYX's */
3716 if (PL_reg_maxiter < 0)
3717 PL_reg_maxiter = I32_MAX;
3718 PL_reg_leftiter = PL_reg_maxiter;
3720 if (PL_reg_leftiter-- == 0) {
3721 const I32 size = (PL_reg_maxiter + 7)/8;
3722 if (PL_reg_poscache) {
3723 if ((I32)PL_reg_poscache_size < size) {
3724 Renew(PL_reg_poscache, size, char);
3725 PL_reg_poscache_size = size;
3727 Zero(PL_reg_poscache, size, char);
3730 PL_reg_poscache_size = size;
3731 Newxz(PL_reg_poscache, size, char);
3734 PerlIO_printf(Perl_debug_log,
3735 "%sDetected a super-linear match, switching on caching%s...\n",
3736 PL_colors[4], PL_colors[5])
3739 if (PL_reg_leftiter < 0) {
3740 st->u.whilem.cache_offset = locinput - PL_bostr;
3742 st->u.whilem.cache_offset = (scan->flags & 0xf) - 1
3743 + st->u.whilem.cache_offset * (scan->flags>>4);
3744 st->u.whilem.cache_bit = st->u.whilem.cache_offset % 8;
3745 st->u.whilem.cache_offset /= 8;
3746 if (PL_reg_poscache[st->u.whilem.cache_offset] & (1<<st->u.whilem.cache_bit)) {
3748 PerlIO_printf(Perl_debug_log,
3749 "%*s already tried at this position...\n",
3750 REPORT_CODE_OFF+PL_regindent*2, "")
3752 sayNO; /* cache records failure */
3757 /* Prefer next over scan for minimal matching. */
3759 if (st->cc->minmod) {
3760 st->u.whilem.savecc = st->cc;
3761 st->cc = st->cc->u.curlyx.outercc;
3763 st->ln = st->cc->u.curlyx.cur;
3764 st->u.whilem.cp = regcppush(st->u.whilem.savecc->u.curlyx.parenfloor);
3765 REGCP_SET(st->u.whilem.lastcp);
3766 REGMATCH(st->u.whilem.savecc->next, WHILEM3);
3767 /*** all unsaved local vars undefined at this point */
3768 st->cc = st->u.whilem.savecc;
3770 regcpblow(st->u.whilem.cp);
3771 sayYES; /* All done. */
3773 REGCP_UNWIND(st->u.whilem.lastcp);
3775 if (st->cc->u.curlyx.outercc)
3776 st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln;
3778 if (n >= st->cc->u.curlyx.max) { /* Maximum greed exceeded? */
3779 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3780 && !(PL_reg_flags & RF_warned)) {
3781 PL_reg_flags |= RF_warned;
3782 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3783 "Complex regular subexpression recursion",
3790 PerlIO_printf(Perl_debug_log,
3791 "%*s trying longer...\n",
3792 REPORT_CODE_OFF+PL_regindent*2, "")
3794 /* Try scanning more and see if it helps. */
3795 PL_reginput = locinput;
3796 st->cc->u.curlyx.cur = n;
3797 st->cc->u.curlyx.lastloc = locinput;
3798 st->u.whilem.cp = regcppush(st->cc->u.curlyx.parenfloor);
3799 REGCP_SET(st->u.whilem.lastcp);
3800 REGMATCH(st->cc->u.curlyx.scan, WHILEM4);
3801 /*** all unsaved local vars undefined at this point */
3803 regcpblow(st->u.whilem.cp);
3806 REGCP_UNWIND(st->u.whilem.lastcp);
3808 st->cc->u.curlyx.cur = n - 1;
3809 st->cc->u.curlyx.lastloc = st->u.whilem.lastloc;
3813 /* Prefer scan over next for maximal matching. */
3815 if (n < st->cc->u.curlyx.max) { /* More greed allowed? */
3816 st->u.whilem.cp = regcppush(st->cc->u.curlyx.parenfloor);
3817 st->cc->u.curlyx.cur = n;
3818 st->cc->u.curlyx.lastloc = locinput;
3819 REGCP_SET(st->u.whilem.lastcp);
3820 REGMATCH(st->cc->u.curlyx.scan, WHILEM5);
3821 /*** all unsaved local vars undefined at this point */
3823 regcpblow(st->u.whilem.cp);
3826 REGCP_UNWIND(st->u.whilem.lastcp);
3827 regcppop(rex); /* Restore some previous $<digit>s? */
3828 PL_reginput = locinput;
3830 PerlIO_printf(Perl_debug_log,
3831 "%*s failed, try continuation...\n",
3832 REPORT_CODE_OFF+PL_regindent*2, "")
3835 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3836 && !(PL_reg_flags & RF_warned)) {
3837 PL_reg_flags |= RF_warned;
3838 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3839 "Complex regular subexpression recursion",
3843 /* Failed deeper matches of scan, so see if this one works. */
3844 st->u.whilem.savecc = st->cc;
3845 st->cc = st->cc->u.curlyx.outercc;
3847 st->ln = st->cc->u.curlyx.cur;
3848 REGMATCH(st->u.whilem.savecc->next, WHILEM6);
3849 /*** all unsaved local vars undefined at this point */
3850 st->cc = st->u.whilem.savecc;
3853 if (st->cc->u.curlyx.outercc)
3854 st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln;
3855 st->cc->u.curlyx.cur = n - 1;
3856 st->cc->u.curlyx.lastloc = st->u.whilem.lastloc;
3862 #define ST st->u.branch
3864 case BRANCHJ: /* /(...|A|...)/ with long next pointer */
3865 next = scan + ARG(scan);
3868 scan = NEXTOPER(scan);
3871 case BRANCH: /* /(...|A|...)/ */
3872 scan = NEXTOPER(scan); /* scan now points to inner node */
3873 if (!next || (OP(next) != BRANCH && OP(next) != BRANCHJ))
3874 /* last branch; skip state push and jump direct to node */
3876 ST.lastparen = *PL_reglastparen;
3877 ST.next_branch = next;
3879 PL_reginput = locinput;
3881 /* Now go into the branch */
3882 PUSH_STATE_GOTO(BRANCH_next, scan);
3885 case BRANCH_next_fail: /* that branch failed; try the next, if any */
3886 REGCP_UNWIND(ST.cp);
3887 for (n = *PL_reglastparen; n > ST.lastparen; n--)
3889 *PL_reglastparen = n;
3890 scan = ST.next_branch;
3891 /* no more branches? */
3892 if (!scan || (OP(scan) != BRANCH && OP(scan) != BRANCHJ))
3894 continue; /* execute next BRANCH[J] op */
3902 #define ST st->u.curlym
3904 case CURLYM: /* /A{m,n}B/ where A is fixed-length */
3906 /* This is an optimisation of CURLYX that enables us to push
3907 * only a single backtracking state, no matter now many matches
3908 * there are in {m,n}. It relies on the pattern being constant
3909 * length, with no parens to influence future backrefs
3913 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;