5 * "One Ring to rule them all, One Ring to find them..."
8 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
9 * confused with the original package (see point 3 below). Thanks, Henry!
12 /* Additional note: this code is very heavily munged from Henry's version
13 * in places. In some spots I've traded clarity for efficiency, so don't
14 * blame Henry for some of the lack of readability.
17 /* The names of the functions have been changed from regcomp and
18 * regexec to pregcomp and pregexec in order to avoid conflicts
19 * with the POSIX routines of the same names.
22 #ifdef PERL_EXT_RE_BUILD
23 /* need to replace pregcomp et al, so enable that */
24 # ifndef PERL_IN_XSUB_RE
25 # define PERL_IN_XSUB_RE
27 /* need access to debugger hooks */
28 # if defined(PERL_EXT_RE_DEBUG) && !defined(DEBUGGING)
33 #ifdef PERL_IN_XSUB_RE
34 /* We *really* need to overwrite these symbols: */
35 # define Perl_regexec_flags my_regexec
36 # define Perl_regdump my_regdump
37 # define Perl_regprop my_regprop
38 # define Perl_re_intuit_start my_re_intuit_start
39 /* *These* symbols are masked to allow static link. */
40 # define Perl_pregexec my_pregexec
41 # define Perl_reginitcolors my_reginitcolors
42 # define Perl_regclass_swash my_regclass_swash
44 # define PERL_NO_GET_CONTEXT
49 * pregcomp and pregexec -- regsub and regerror are not used in perl
51 * Copyright (c) 1986 by University of Toronto.
52 * Written by Henry Spencer. Not derived from licensed software.
54 * Permission is granted to anyone to use this software for any
55 * purpose on any computer system, and to redistribute it freely,
56 * subject to the following restrictions:
58 * 1. The author is not responsible for the consequences of use of
59 * this software, no matter how awful, even if they arise
62 * 2. The origin of this software must not be misrepresented, either
63 * by explicit claim or by omission.
65 * 3. Altered versions must be plainly marked as such, and must not
66 * be misrepresented as being the original software.
68 **** Alterations to Henry's code are...
70 **** Copyright (c) 1991-2001, Larry Wall
72 **** You may distribute under the terms of either the GNU General Public
73 **** License or the Artistic License, as specified in the README file.
75 * Beware that some of this code is subtly aware of the way operator
76 * precedence is structured in regular expressions. Serious changes in
77 * regular-expression syntax might require a total rethink.
80 #define PERL_IN_REGEXEC_C
85 #define RF_tainted 1 /* tainted information used? */
86 #define RF_warned 2 /* warned about big count? */
87 #define RF_evaled 4 /* Did an EVAL with setting? */
88 #define RF_utf8 8 /* String contains multibyte chars? */
90 #define UTF (PL_reg_flags & RF_utf8)
92 #define RS_init 1 /* eval environment created */
93 #define RS_set 2 /* replsv value is set */
103 #define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
104 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
106 #define reghop_c(pos,off) ((char*)reghop((U8*)pos, off))
107 #define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off))
108 #define HOP(pos,off) (PL_reg_match_utf8 ? reghop((U8*)pos, off) : (U8*)(pos + off))
109 #define HOPMAYBE(pos,off) (PL_reg_match_utf8 ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off))
110 #define HOPc(pos,off) ((char*)HOP(pos,off))
111 #define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off))
113 #define HOPBACK(pos, off) ( \
114 (UTF && PL_reg_match_utf8) \
115 ? reghopmaybe((U8*)pos, -off) \
116 : (pos - off >= PL_bostr) \
120 #define HOPBACKc(pos, off) (char*)HOPBACK(pos, off)
122 #define reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim))
123 #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
124 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
125 #define HOPMAYBE3(pos,off,lim) (PL_reg_match_utf8 ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
126 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
127 #define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim))
129 #define LOAD_UTF8_CHARCLASS(a,b) STMT_START { if (!CAT2(PL_utf8_,a)) (void)CAT2(is_utf8_, a)((U8*)b); } STMT_END
131 /* for use after a quantifier and before an EXACT-like node -- japhy */
132 #define JUMPABLE(rn) ( \
133 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
134 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
135 OP(rn) == PLUS || OP(rn) == MINMOD || \
136 (PL_regkind[(U8)OP(rn)] == CURLY && ARG1(rn) > 0) \
139 #define HAS_TEXT(rn) ( \
140 PL_regkind[(U8)OP(rn)] == EXACT || PL_regkind[(U8)OP(rn)] == REF \
143 #define FIND_NEXT_IMPT(rn) STMT_START { \
144 while (JUMPABLE(rn)) \
145 if (OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
146 PL_regkind[(U8)OP(rn)] == CURLY) \
147 rn = NEXTOPER(NEXTOPER(rn)); \
148 else if (OP(rn) == PLUS) \
150 else rn += NEXT_OFF(rn); \
153 static void restore_pos(pTHX_ void *arg);
156 S_regcppush(pTHX_ I32 parenfloor)
158 int retval = PL_savestack_ix;
159 #define REGCP_PAREN_ELEMS 4
160 int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
163 if (paren_elems_to_push < 0)
164 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
166 #define REGCP_OTHER_ELEMS 6
167 SSCHECK(paren_elems_to_push + REGCP_OTHER_ELEMS);
168 for (p = PL_regsize; p > parenfloor; p--) {
169 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
170 SSPUSHINT(PL_regendp[p]);
171 SSPUSHINT(PL_regstartp[p]);
172 SSPUSHPTR(PL_reg_start_tmp[p]);
175 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
176 SSPUSHINT(PL_regsize);
177 SSPUSHINT(*PL_reglastparen);
178 SSPUSHINT(*PL_reglastcloseparen);
179 SSPUSHPTR(PL_reginput);
180 #define REGCP_FRAME_ELEMS 2
181 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
182 * are needed for the regexp context stack bookkeeping. */
183 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
184 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
189 /* These are needed since we do not localize EVAL nodes: */
190 # define REGCP_SET(cp) DEBUG_r(PerlIO_printf(Perl_debug_log, \
191 " Setting an EVAL scope, savestack=%"IVdf"\n", \
192 (IV)PL_savestack_ix)); cp = PL_savestack_ix
194 # define REGCP_UNWIND(cp) DEBUG_r(cp != PL_savestack_ix ? \
195 PerlIO_printf(Perl_debug_log, \
196 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
197 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
207 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
209 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
210 i = SSPOPINT; /* Parentheses elements to pop. */
211 input = (char *) SSPOPPTR;
212 *PL_reglastcloseparen = SSPOPINT;
213 *PL_reglastparen = SSPOPINT;
214 PL_regsize = SSPOPINT;
216 /* Now restore the parentheses context. */
217 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
218 i > 0; i -= REGCP_PAREN_ELEMS) {
219 paren = (U32)SSPOPINT;
220 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
221 PL_regstartp[paren] = SSPOPINT;
223 if (paren <= *PL_reglastparen)
224 PL_regendp[paren] = tmps;
226 PerlIO_printf(Perl_debug_log,
227 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
228 (UV)paren, (IV)PL_regstartp[paren],
229 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
230 (IV)PL_regendp[paren],
231 (paren > *PL_reglastparen ? "(no)" : ""));
235 if (*PL_reglastparen + 1 <= PL_regnpar) {
236 PerlIO_printf(Perl_debug_log,
237 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
238 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
242 /* It would seem that the similar code in regtry()
243 * already takes care of this, and in fact it is in
244 * a better location to since this code can #if 0-ed out
245 * but the code in regtry() is needed or otherwise tests
246 * requiring null fields (pat.t#187 and split.t#{13,14}
247 * (as of patchlevel 7877) will fail. Then again,
248 * this code seems to be necessary or otherwise
249 * building DynaLoader will fail:
250 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
252 for (paren = *PL_reglastparen + 1; paren <= PL_regnpar; paren++) {
253 if (paren > PL_regsize)
254 PL_regstartp[paren] = -1;
255 PL_regendp[paren] = -1;
262 S_regcp_set_to(pTHX_ I32 ss)
264 I32 tmp = PL_savestack_ix;
266 PL_savestack_ix = ss;
268 PL_savestack_ix = tmp;
272 typedef struct re_cc_state
276 struct re_cc_state *prev;
281 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
283 #define TRYPAREN(paren, n, input) { \
286 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
287 PL_regendp[paren] = input - PL_bostr; \
290 PL_regendp[paren] = -1; \
292 if (regmatch(next)) \
295 PL_regendp[paren] = -1; \
300 * pregexec and friends
304 - pregexec - match a regexp against a string
307 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
308 char *strbeg, I32 minend, SV *screamer, U32 nosave)
309 /* strend: pointer to null at end of string */
310 /* strbeg: real beginning of string */
311 /* minend: end of match must be >=minend after stringarg. */
312 /* nosave: For optimizations. */
315 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
316 nosave ? 0 : REXEC_COPY_STR);
320 S_cache_re(pTHX_ regexp *prog)
322 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
324 PL_regprogram = prog->program;
326 PL_regnpar = prog->nparens;
327 PL_regdata = prog->data;
332 * Need to implement the following flags for reg_anch:
334 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
336 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
337 * INTUIT_AUTORITATIVE_ML
338 * INTUIT_ONCE_NOML - Intuit can match in one location only.
341 * Another flag for this function: SECOND_TIME (so that float substrs
342 * with giant delta may be not rechecked).
345 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
347 /* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
348 Otherwise, only SvCUR(sv) is used to get strbeg. */
350 /* XXXX We assume that strpos is strbeg unless sv. */
352 /* XXXX Some places assume that there is a fixed substring.
353 An update may be needed if optimizer marks as "INTUITable"
354 RExen without fixed substrings. Similarly, it is assumed that
355 lengths of all the strings are no more than minlen, thus they
356 cannot come from lookahead.
357 (Or minlen should take into account lookahead.) */
359 /* A failure to find a constant substring means that there is no need to make
360 an expensive call to REx engine, thus we celebrate a failure. Similarly,
361 finding a substring too deep into the string means that less calls to
362 regtry() should be needed.
364 REx compiler's optimizer found 4 possible hints:
365 a) Anchored substring;
367 c) Whether we are anchored (beginning-of-line or \G);
368 d) First node (of those at offset 0) which may distingush positions;
369 We use a)b)d) and multiline-part of c), and try to find a position in the
370 string which does not contradict any of them.
373 /* Most of decisions we do here should have been done at compile time.
374 The nodes of the REx which we used for the search should have been
375 deleted from the finite automaton. */
378 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
379 char *strend, U32 flags, re_scream_pos_data *data)
381 register I32 start_shift = 0;
382 /* Should be nonnegative! */
383 register I32 end_shift = 0;
389 register char *other_last = Nullch; /* other substr checked before this */
390 char *check_at = Nullch; /* check substr found at this pos */
392 char *i_strpos = strpos;
393 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
396 if (prog->reganch & ROPT_UTF8) {
397 DEBUG_r(PerlIO_printf(Perl_debug_log,
398 "UTF-8 regex...\n"));
399 PL_reg_flags |= RF_utf8;
403 char *s = PL_reg_match_utf8 ?
404 sv_uni_display(dsv, sv, 60, 0) : strpos;
405 int len = PL_reg_match_utf8 ?
406 strlen(s) : strend - strpos;
409 if (PL_reg_match_utf8)
410 DEBUG_r(PerlIO_printf(Perl_debug_log,
411 "UTF-8 target...\n"));
412 PerlIO_printf(Perl_debug_log,
413 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
414 PL_colors[4],PL_colors[5],PL_colors[0],
417 (strlen(prog->precomp) > 60 ? "..." : ""),
419 (int)(len > 60 ? 60 : len),
421 (len > 60 ? "..." : "")
425 if (prog->minlen > CHR_DIST((U8*)strend, (U8*)strpos)) {
426 DEBUG_r(PerlIO_printf(Perl_debug_log,
427 "String too short... [re_intuit_start]\n"));
430 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
432 check = prog->check_substr;
433 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
434 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
435 || ( (prog->reganch & ROPT_ANCH_BOL)
436 && !PL_multiline ) ); /* Check after \n? */
439 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
440 | ROPT_IMPLICIT)) /* not a real BOL */
441 /* SvCUR is not set on references: SvRV and SvPVX overlap */
443 && (strpos != strbeg)) {
444 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
447 if (prog->check_offset_min == prog->check_offset_max &&
448 !(prog->reganch & ROPT_CANY_SEEN)) {
449 /* Substring at constant offset from beg-of-str... */
452 s = HOP3c(strpos, prog->check_offset_min, strend);
454 slen = SvCUR(check); /* >= 1 */
456 if ( strend - s > slen || strend - s < slen - 1
457 || (strend - s == slen && strend[-1] != '\n')) {
458 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
461 /* Now should match s[0..slen-2] */
463 if (slen && (*SvPVX(check) != *s
465 && memNE(SvPVX(check), s, slen)))) {
467 DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
471 else if (*SvPVX(check) != *s
472 || ((slen = SvCUR(check)) > 1
473 && memNE(SvPVX(check), s, slen)))
475 goto success_at_start;
478 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
480 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
481 end_shift = prog->minlen - start_shift -
482 CHR_SVLEN(check) + (SvTAIL(check) != 0);
484 I32 end = prog->check_offset_max + CHR_SVLEN(check)
485 - (SvTAIL(check) != 0);
486 I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
488 if (end_shift < eshift)
492 else { /* Can match at random position */
495 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
496 /* Should be nonnegative! */
497 end_shift = prog->minlen - start_shift -
498 CHR_SVLEN(check) + (SvTAIL(check) != 0);
501 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
503 Perl_croak(aTHX_ "panic: end_shift");
507 /* Find a possible match in the region s..strend by looking for
508 the "check" substring in the region corrected by start/end_shift. */
509 if (flags & REXEC_SCREAM) {
510 I32 p = -1; /* Internal iterator of scream. */
511 I32 *pp = data ? data->scream_pos : &p;
513 if (PL_screamfirst[BmRARE(check)] >= 0
514 || ( BmRARE(check) == '\n'
515 && (BmPREVIOUS(check) == SvCUR(check) - 1)
517 s = screaminstr(sv, check,
518 start_shift + (s - strbeg), end_shift, pp, 0);
522 *data->scream_olds = s;
524 else if (prog->reganch & ROPT_CANY_SEEN)
525 s = fbm_instr((U8*)(s + start_shift),
526 (U8*)(strend - end_shift),
527 check, PL_multiline ? FBMrf_MULTILINE : 0);
529 s = fbm_instr(HOP3(s, start_shift, strend),
530 HOP3(strend, -end_shift, strbeg),
531 check, PL_multiline ? FBMrf_MULTILINE : 0);
533 /* Update the count-of-usability, remove useless subpatterns,
536 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
537 (s ? "Found" : "Did not find"),
538 ((check == prog->anchored_substr) ? "anchored" : "floating"),
540 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
542 PL_colors[1], (SvTAIL(check) ? "$" : ""),
543 (s ? " at offset " : "...\n") ) );
550 /* Finish the diagnostic message */
551 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
553 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
554 Start with the other substr.
555 XXXX no SCREAM optimization yet - and a very coarse implementation
556 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
557 *always* match. Probably should be marked during compile...
558 Probably it is right to do no SCREAM here...
561 if (prog->float_substr && prog->anchored_substr) {
562 /* Take into account the "other" substring. */
563 /* XXXX May be hopelessly wrong for UTF... */
566 if (check == prog->float_substr) {
569 char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
572 t = s - prog->check_offset_max;
573 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
574 && (!(prog->reganch & ROPT_UTF8)
575 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
580 t = HOP3c(t, prog->anchored_offset, strend);
581 if (t < other_last) /* These positions already checked */
583 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
586 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
587 /* On end-of-str: see comment below. */
588 s = fbm_instr((unsigned char*)t,
589 HOP3(HOP3(last1, prog->anchored_offset, strend)
590 + SvCUR(prog->anchored_substr),
591 -(SvTAIL(prog->anchored_substr)!=0), strbeg),
592 prog->anchored_substr,
593 PL_multiline ? FBMrf_MULTILINE : 0);
594 DEBUG_r(PerlIO_printf(Perl_debug_log,
595 "%s anchored substr `%s%.*s%s'%s",
596 (s ? "Found" : "Contradicts"),
598 (int)(SvCUR(prog->anchored_substr)
599 - (SvTAIL(prog->anchored_substr)!=0)),
600 SvPVX(prog->anchored_substr),
601 PL_colors[1], (SvTAIL(prog->anchored_substr) ? "$" : "")));
603 if (last1 >= last2) {
604 DEBUG_r(PerlIO_printf(Perl_debug_log,
605 ", giving up...\n"));
608 DEBUG_r(PerlIO_printf(Perl_debug_log,
609 ", trying floating at offset %ld...\n",
610 (long)(HOP3c(s1, 1, strend) - i_strpos)));
611 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
612 s = HOP3c(last, 1, strend);
616 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
617 (long)(s - i_strpos)));
618 t = HOP3c(s, -prog->anchored_offset, strbeg);
619 other_last = HOP3c(s, 1, strend);
627 else { /* Take into account the floating substring. */
631 t = HOP3c(s, -start_shift, strbeg);
633 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
634 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
635 last = HOP3c(t, prog->float_max_offset, strend);
636 s = HOP3c(t, prog->float_min_offset, strend);
639 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
640 /* fbm_instr() takes into account exact value of end-of-str
641 if the check is SvTAIL(ed). Since false positives are OK,
642 and end-of-str is not later than strend we are OK. */
643 s = fbm_instr((unsigned char*)s,
644 (unsigned char*)last + SvCUR(prog->float_substr)
645 - (SvTAIL(prog->float_substr)!=0),
646 prog->float_substr, PL_multiline ? FBMrf_MULTILINE : 0);
647 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
648 (s ? "Found" : "Contradicts"),
650 (int)(SvCUR(prog->float_substr)
651 - (SvTAIL(prog->float_substr)!=0)),
652 SvPVX(prog->float_substr),
653 PL_colors[1], (SvTAIL(prog->float_substr) ? "$" : "")));
656 DEBUG_r(PerlIO_printf(Perl_debug_log,
657 ", giving up...\n"));
660 DEBUG_r(PerlIO_printf(Perl_debug_log,
661 ", trying anchored starting at offset %ld...\n",
662 (long)(s1 + 1 - i_strpos)));
664 s = HOP3c(t, 1, strend);
668 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
669 (long)(s - i_strpos)));
670 other_last = s; /* Fix this later. --Hugo */
679 t = s - prog->check_offset_max;
680 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
681 && (!(prog->reganch & ROPT_UTF8)
682 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
684 /* Fixed substring is found far enough so that the match
685 cannot start at strpos. */
687 if (ml_anch && t[-1] != '\n') {
688 /* Eventually fbm_*() should handle this, but often
689 anchored_offset is not 0, so this check will not be wasted. */
690 /* XXXX In the code below we prefer to look for "^" even in
691 presence of anchored substrings. And we search even
692 beyond the found float position. These pessimizations
693 are historical artefacts only. */
695 while (t < strend - prog->minlen) {
697 if (t < check_at - prog->check_offset_min) {
698 if (prog->anchored_substr) {
699 /* Since we moved from the found position,
700 we definitely contradict the found anchored
701 substr. Due to the above check we do not
702 contradict "check" substr.
703 Thus we can arrive here only if check substr
704 is float. Redo checking for "other"=="fixed".
707 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
708 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
709 goto do_other_anchored;
711 /* We don't contradict the found floating substring. */
712 /* XXXX Why not check for STCLASS? */
714 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
715 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
718 /* Position contradicts check-string */
719 /* XXXX probably better to look for check-string
720 than for "\n", so one should lower the limit for t? */
721 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
722 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
723 other_last = strpos = s = t + 1;
728 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
729 PL_colors[0],PL_colors[1]));
733 DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
734 PL_colors[0],PL_colors[1]));
738 ++BmUSEFUL(prog->check_substr); /* hooray/5 */
741 /* The found string does not prohibit matching at strpos,
742 - no optimization of calling REx engine can be performed,
743 unless it was an MBOL and we are not after MBOL,
744 or a future STCLASS check will fail this. */
746 /* Even in this situation we may use MBOL flag if strpos is offset
747 wrt the start of the string. */
748 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
749 && (strpos != strbeg) && strpos[-1] != '\n'
750 /* May be due to an implicit anchor of m{.*foo} */
751 && !(prog->reganch & ROPT_IMPLICIT))
756 DEBUG_r( if (ml_anch)
757 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
758 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
761 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
762 && prog->check_substr /* Could be deleted already */
763 && --BmUSEFUL(prog->check_substr) < 0
764 && prog->check_substr == prog->float_substr)
766 /* If flags & SOMETHING - do not do it many times on the same match */
767 DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
768 SvREFCNT_dec(prog->check_substr);
769 prog->check_substr = Nullsv; /* disable */
770 prog->float_substr = Nullsv; /* clear */
771 check = Nullsv; /* abort */
773 /* XXXX This is a remnant of the old implementation. It
774 looks wasteful, since now INTUIT can use many
776 prog->reganch &= ~RE_USE_INTUIT;
783 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
784 if (prog->regstclass) {
785 /* minlen == 0 is possible if regstclass is \b or \B,
786 and the fixed substr is ''$.
787 Since minlen is already taken into account, s+1 is before strend;
788 accidentally, minlen >= 1 guaranties no false positives at s + 1
789 even for \b or \B. But (minlen? 1 : 0) below assumes that
790 regstclass does not come from lookahead... */
791 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
792 This leaves EXACTF only, which is dealt with in find_byclass(). */
793 U8* str = (U8*)STRING(prog->regstclass);
794 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
795 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
797 char *endpos = (prog->anchored_substr || ml_anch)
798 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
799 : (prog->float_substr
800 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
803 char *startpos = strbeg;
806 if (prog->reganch & ROPT_UTF8) {
807 PL_regdata = prog->data;
810 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
815 if (endpos == strend) {
816 DEBUG_r( PerlIO_printf(Perl_debug_log,
817 "Could not match STCLASS...\n") );
820 DEBUG_r( PerlIO_printf(Perl_debug_log,
821 "This position contradicts STCLASS...\n") );
822 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
824 /* Contradict one of substrings */
825 if (prog->anchored_substr) {
826 if (prog->anchored_substr == check) {
827 DEBUG_r( what = "anchored" );
829 s = HOP3c(t, 1, strend);
830 if (s + start_shift + end_shift > strend) {
831 /* XXXX Should be taken into account earlier? */
832 DEBUG_r( PerlIO_printf(Perl_debug_log,
833 "Could not match STCLASS...\n") );
838 DEBUG_r( PerlIO_printf(Perl_debug_log,
839 "Looking for %s substr starting at offset %ld...\n",
840 what, (long)(s + start_shift - i_strpos)) );
843 /* Have both, check_string is floating */
844 if (t + start_shift >= check_at) /* Contradicts floating=check */
845 goto retry_floating_check;
846 /* Recheck anchored substring, but not floating... */
850 DEBUG_r( PerlIO_printf(Perl_debug_log,
851 "Looking for anchored substr starting at offset %ld...\n",
852 (long)(other_last - i_strpos)) );
853 goto do_other_anchored;
855 /* Another way we could have checked stclass at the
856 current position only: */
861 DEBUG_r( PerlIO_printf(Perl_debug_log,
862 "Looking for /%s^%s/m starting at offset %ld...\n",
863 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
866 if (!prog->float_substr) /* Could have been deleted */
868 /* Check is floating subtring. */
869 retry_floating_check:
870 t = check_at - start_shift;
871 DEBUG_r( what = "floating" );
872 goto hop_and_restart;
875 DEBUG_r(PerlIO_printf(Perl_debug_log,
876 "By STCLASS: moving %ld --> %ld\n",
877 (long)(t - i_strpos), (long)(s - i_strpos))
881 DEBUG_r(PerlIO_printf(Perl_debug_log,
882 "Does not contradict STCLASS...\n");
887 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
888 PL_colors[4], (check ? "Guessed" : "Giving up"),
889 PL_colors[5], (long)(s - i_strpos)) );
892 fail_finish: /* Substring not found */
893 if (prog->check_substr) /* could be removed already */
894 BmUSEFUL(prog->check_substr) += 5; /* hooray */
896 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
897 PL_colors[4],PL_colors[5]));
901 /* We know what class REx starts with. Try to find this position... */
903 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
905 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
911 register I32 tmp = 1; /* Scratch variable? */
912 register bool do_utf8 = PL_reg_match_utf8;
914 /* We know what class it must start with. */
918 if (reginclass(c, (U8*)s, do_utf8)) {
919 if (tmp && (norun || regtry(prog, s)))
926 s += do_utf8 ? UTF8SKIP(s) : 1;
931 if (tmp && (norun || regtry(prog, s)))
943 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
944 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
946 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
947 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
949 c1 = utf8_to_uvuni(tmpbuf1, 0);
950 c2 = utf8_to_uvuni(tmpbuf2, 0);
961 c2 = PL_fold_locale[c1];
963 e = do_utf8 ? s + ln - 1 : strend - ln;
966 e = s; /* Due to minlen logic of intuit() */
968 /* The idea in the EXACTF* cases is to first find the
969 * first character of the EXACTF* node and then, if
970 * necessary, case-insensitively compare the full
971 * text of the node. The c1 and c2 are the first
972 * characters (though in Unicode it gets a bit
973 * more complicated because there are more cases
974 * than just upper and lower: one is really supposed
975 * to use the so-called folding case for case-insensitive
976 * matching (called "loose matching" in Unicode). */
980 U8 tmpbuf [UTF8_MAXLEN+1];
981 U8 foldbuf[UTF8_MAXLEN_FOLD+1];
984 /* The ibcmp_utf8() uses to_uni_fold() which is more
985 * correct folding for Unicode than using lowercase.
986 * However, it doesn't work quite fully since the folding
987 * is a one-to-many mapping and the regex optimizer is
988 * unaware of this, so it may throw out good matches.
989 * Fortunately, not getting this right is allowed
990 * for Unicode Regular Expression Support level 1,
991 * only one-to-one matching is required. --jhi */
995 c = utf8_to_uvchr((U8*)s, &len);
998 !ibcmp_utf8(s, do_utf8,
999 strend - s > ln ? ln : strend - s,
1001 && (norun || regtry(prog, s)) )
1004 uvchr_to_utf8(tmpbuf, c);
1005 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1007 && (f == c1 || f == c2)
1008 && (ln == foldlen ||
1009 !ibcmp_utf8((char *)foldbuf,
1011 foldlen > ln ? ln : foldlen,
1013 && (norun || regtry(prog, s)) )
1021 c = utf8_to_uvchr((U8*)s, &len);
1023 /* Handle some of the three Greek sigmas cases.
1024 * Note that not all the possible combinations
1025 * are handled here: some of them are handled
1026 * handled by the standard folding rules, and
1027 * some of them (the character class or ANYOF
1028 * cases) are handled during compiletime in
1029 * regexec.c:S_regclass(). */
1030 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1031 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1032 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1034 if ( (c == c1 || c == c2)
1036 !ibcmp_utf8(s, do_utf8,
1037 strend - s > ln ? ln : strend - s,
1039 && (norun || regtry(prog, s)) )
1042 uvchr_to_utf8(tmpbuf, c);
1043 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1045 && (f == c1 || f == c2)
1046 && (ln == foldlen ||
1047 !ibcmp_utf8((char *)foldbuf,
1049 foldlen > ln ? ln : foldlen,
1051 && (norun || regtry(prog, s)) )
1062 && (ln == 1 || !(OP(c) == EXACTF
1064 : ibcmp_locale(s, m, ln)))
1065 && (norun || regtry(prog, s)) )
1071 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1072 && (ln == 1 || !(OP(c) == EXACTF
1074 : ibcmp_locale(s, m, ln)))
1075 && (norun || regtry(prog, s)) )
1082 PL_reg_flags |= RF_tainted;
1089 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1092 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1094 tmp = ((OP(c) == BOUND ?
1095 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1096 LOAD_UTF8_CHARCLASS(alnum,"a");
1097 while (s < strend) {
1098 if (tmp == !(OP(c) == BOUND ?
1099 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1100 isALNUM_LC_utf8((U8*)s)))
1103 if ((norun || regtry(prog, s)))
1110 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1111 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1112 while (s < strend) {
1114 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1116 if ((norun || regtry(prog, s)))
1122 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1126 PL_reg_flags |= RF_tainted;
1133 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1136 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1138 tmp = ((OP(c) == NBOUND ?
1139 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1140 LOAD_UTF8_CHARCLASS(alnum,"a");
1141 while (s < strend) {
1142 if (tmp == !(OP(c) == NBOUND ?
1143 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1144 isALNUM_LC_utf8((U8*)s)))
1146 else if ((norun || regtry(prog, s)))
1152 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1153 tmp = ((OP(c) == NBOUND ?
1154 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1155 while (s < strend) {
1157 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1159 else if ((norun || regtry(prog, s)))
1164 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1169 LOAD_UTF8_CHARCLASS(alnum,"a");
1170 while (s < strend) {
1171 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1172 if (tmp && (norun || regtry(prog, s)))
1183 while (s < strend) {
1185 if (tmp && (norun || regtry(prog, s)))
1197 PL_reg_flags |= RF_tainted;
1199 while (s < strend) {
1200 if (isALNUM_LC_utf8((U8*)s)) {
1201 if (tmp && (norun || regtry(prog, s)))
1212 while (s < strend) {
1213 if (isALNUM_LC(*s)) {
1214 if (tmp && (norun || regtry(prog, s)))
1227 LOAD_UTF8_CHARCLASS(alnum,"a");
1228 while (s < strend) {
1229 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1230 if (tmp && (norun || regtry(prog, s)))
1241 while (s < strend) {
1243 if (tmp && (norun || regtry(prog, s)))
1255 PL_reg_flags |= RF_tainted;
1257 while (s < strend) {
1258 if (!isALNUM_LC_utf8((U8*)s)) {
1259 if (tmp && (norun || regtry(prog, s)))
1270 while (s < strend) {
1271 if (!isALNUM_LC(*s)) {
1272 if (tmp && (norun || regtry(prog, s)))
1285 LOAD_UTF8_CHARCLASS(space," ");
1286 while (s < strend) {
1287 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1288 if (tmp && (norun || regtry(prog, s)))
1299 while (s < strend) {
1301 if (tmp && (norun || regtry(prog, s)))
1313 PL_reg_flags |= RF_tainted;
1315 while (s < strend) {
1316 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1317 if (tmp && (norun || regtry(prog, s)))
1328 while (s < strend) {
1329 if (isSPACE_LC(*s)) {
1330 if (tmp && (norun || regtry(prog, s)))
1343 LOAD_UTF8_CHARCLASS(space," ");
1344 while (s < strend) {
1345 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1346 if (tmp && (norun || regtry(prog, s)))
1357 while (s < strend) {
1359 if (tmp && (norun || regtry(prog, s)))
1371 PL_reg_flags |= RF_tainted;
1373 while (s < strend) {
1374 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1375 if (tmp && (norun || regtry(prog, s)))
1386 while (s < strend) {
1387 if (!isSPACE_LC(*s)) {
1388 if (tmp && (norun || regtry(prog, s)))
1401 LOAD_UTF8_CHARCLASS(digit,"0");
1402 while (s < strend) {
1403 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1404 if (tmp && (norun || regtry(prog, s)))
1415 while (s < strend) {
1417 if (tmp && (norun || regtry(prog, s)))
1429 PL_reg_flags |= RF_tainted;
1431 while (s < strend) {
1432 if (isDIGIT_LC_utf8((U8*)s)) {
1433 if (tmp && (norun || regtry(prog, s)))
1444 while (s < strend) {
1445 if (isDIGIT_LC(*s)) {
1446 if (tmp && (norun || regtry(prog, s)))
1459 LOAD_UTF8_CHARCLASS(digit,"0");
1460 while (s < strend) {
1461 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1462 if (tmp && (norun || regtry(prog, s)))
1473 while (s < strend) {
1475 if (tmp && (norun || regtry(prog, s)))
1487 PL_reg_flags |= RF_tainted;
1489 while (s < strend) {
1490 if (!isDIGIT_LC_utf8((U8*)s)) {
1491 if (tmp && (norun || regtry(prog, s)))
1502 while (s < strend) {
1503 if (!isDIGIT_LC(*s)) {
1504 if (tmp && (norun || regtry(prog, s)))
1516 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1525 - regexec_flags - match a regexp against a string
1528 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1529 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1530 /* strend: pointer to null at end of string */
1531 /* strbeg: real beginning of string */
1532 /* minend: end of match must be >=minend after stringarg. */
1533 /* data: May be used for some additional optimizations. */
1534 /* nosave: For optimizations. */
1537 register regnode *c;
1538 register char *startpos = stringarg;
1539 I32 minlen; /* must match at least this many chars */
1540 I32 dontbother = 0; /* how many characters not to try at end */
1541 /* I32 start_shift = 0; */ /* Offset of the start to find
1542 constant substr. */ /* CC */
1543 I32 end_shift = 0; /* Same for the end. */ /* CC */
1544 I32 scream_pos = -1; /* Internal iterator of scream. */
1546 SV* oreplsv = GvSV(PL_replgv);
1547 bool do_utf8 = DO_UTF8(sv);
1549 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
1556 PL_regnarrate = DEBUG_r_TEST;
1559 /* Be paranoid... */
1560 if (prog == NULL || startpos == NULL) {
1561 Perl_croak(aTHX_ "NULL regexp parameter");
1565 minlen = prog->minlen;
1566 if (strend - startpos < minlen) {
1567 DEBUG_r(PerlIO_printf(Perl_debug_log,
1568 "String too short [regexec_flags]...\n"));
1572 /* Check validity of program. */
1573 if (UCHARAT(prog->program) != REG_MAGIC) {
1574 Perl_croak(aTHX_ "corrupted regexp program");
1578 PL_reg_eval_set = 0;
1581 if (prog->reganch & ROPT_UTF8)
1582 PL_reg_flags |= RF_utf8;
1584 /* Mark beginning of line for ^ and lookbehind. */
1585 PL_regbol = startpos;
1589 /* Mark end of line for $ (and such) */
1592 /* see how far we have to get to not match where we matched before */
1593 PL_regtill = startpos+minend;
1595 /* We start without call_cc context. */
1598 /* If there is a "must appear" string, look for it. */
1601 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1604 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1605 PL_reg_ganch = startpos;
1606 else if (sv && SvTYPE(sv) >= SVt_PVMG
1608 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1609 && mg->mg_len >= 0) {
1610 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1611 if (prog->reganch & ROPT_ANCH_GPOS) {
1612 if (s > PL_reg_ganch)
1617 else /* pos() not defined */
1618 PL_reg_ganch = strbeg;
1621 if (do_utf8 == (UTF!=0) &&
1622 !(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
1623 re_scream_pos_data d;
1625 d.scream_olds = &scream_olds;
1626 d.scream_pos = &scream_pos;
1627 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1629 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1630 goto phooey; /* not present */
1635 char *s = do_utf8 ? sv_uni_display(dsv, sv, 60, 0) : startpos;
1636 int len = do_utf8 ? strlen(s) : strend - startpos;
1639 PerlIO_printf(Perl_debug_log,
1640 "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
1641 PL_colors[4],PL_colors[5],PL_colors[0],
1644 (strlen(prog->precomp) > 60 ? "..." : ""),
1646 (int)(len > 60 ? 60 : len),
1648 (len > 60 ? "..." : "")
1652 /* Simplest case: anchored match need be tried only once. */
1653 /* [unless only anchor is BOL and multiline is set] */
1654 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1655 if (s == startpos && regtry(prog, startpos))
1657 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1658 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1663 dontbother = minlen - 1;
1664 end = HOP3c(strend, -dontbother, strbeg) - 1;
1665 /* for multiline we only have to try after newlines */
1666 if (prog->check_substr) {
1670 if (regtry(prog, s))
1675 if (prog->reganch & RE_USE_INTUIT) {
1676 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1687 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1688 if (regtry(prog, s))
1695 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1696 if (regtry(prog, PL_reg_ganch))
1701 /* Messy cases: unanchored match. */
1702 if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
1703 /* we have /x+whatever/ */
1704 /* it must be a one character string (XXXX Except UTF?) */
1705 char ch = SvPVX(prog->anchored_substr)[0];
1711 while (s < strend) {
1713 DEBUG_r( did_match = 1 );
1714 if (regtry(prog, s)) goto got_it;
1716 while (s < strend && *s == ch)
1723 while (s < strend) {
1725 DEBUG_r( did_match = 1 );
1726 if (regtry(prog, s)) goto got_it;
1728 while (s < strend && *s == ch)
1734 DEBUG_r(if (!did_match)
1735 PerlIO_printf(Perl_debug_log,
1736 "Did not find anchored character...\n")
1740 else if (do_utf8 == (UTF!=0) &&
1741 (prog->anchored_substr != Nullsv
1742 || (prog->float_substr != Nullsv
1743 && prog->float_max_offset < strend - s))) {
1744 SV *must = prog->anchored_substr
1745 ? prog->anchored_substr : prog->float_substr;
1747 prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset;
1749 prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
1750 char *last = HOP3c(strend, /* Cannot start after this */
1751 -(I32)(CHR_SVLEN(must)
1752 - (SvTAIL(must) != 0) + back_min), strbeg);
1753 char *last1; /* Last position checked before */
1759 last1 = HOPc(s, -1);
1761 last1 = s - 1; /* bogus */
1763 /* XXXX check_substr already used to find `s', can optimize if
1764 check_substr==must. */
1766 dontbother = end_shift;
1767 strend = HOPc(strend, -dontbother);
1768 while ( (s <= last) &&
1769 ((flags & REXEC_SCREAM)
1770 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1771 end_shift, &scream_pos, 0))
1772 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1773 (unsigned char*)strend, must,
1774 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1775 DEBUG_r( did_match = 1 );
1776 if (HOPc(s, -back_max) > last1) {
1777 last1 = HOPc(s, -back_min);
1778 s = HOPc(s, -back_max);
1781 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1783 last1 = HOPc(s, -back_min);
1787 while (s <= last1) {
1788 if (regtry(prog, s))
1794 while (s <= last1) {
1795 if (regtry(prog, s))
1801 DEBUG_r(if (!did_match)
1802 PerlIO_printf(Perl_debug_log,
1803 "Did not find %s substr `%s%.*s%s'%s...\n",
1804 ((must == prog->anchored_substr)
1805 ? "anchored" : "floating"),
1807 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1809 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1813 else if ((c = prog->regstclass)) {
1814 if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT)
1815 /* don't bother with what can't match */
1816 strend = HOPc(strend, -(minlen - 1));
1818 SV *prop = sv_newmortal();
1820 PerlIO_printf(Perl_debug_log, "Matching stclass `%s' against `%s'\n", SvPVX(prop), UTF ? sv_uni_display(dsv, sv, 60, 0) : s);
1822 if (find_byclass(prog, c, s, strend, startpos, 0))
1824 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1828 if (prog->float_substr != Nullsv) { /* Trim the end. */
1831 if (flags & REXEC_SCREAM) {
1832 last = screaminstr(sv, prog->float_substr, s - strbeg,
1833 end_shift, &scream_pos, 1); /* last one */
1835 last = scream_olds; /* Only one occurrence. */
1839 char *little = SvPV(prog->float_substr, len);
1841 if (SvTAIL(prog->float_substr)) {
1842 if (memEQ(strend - len + 1, little, len - 1))
1843 last = strend - len + 1;
1844 else if (!PL_multiline)
1845 last = memEQ(strend - len, little, len)
1846 ? strend - len : Nullch;
1852 last = rninstr(s, strend, little, little + len);
1854 last = strend; /* matching `$' */
1858 DEBUG_r(PerlIO_printf(Perl_debug_log,
1859 "%sCan't trim the tail, match fails (should not happen)%s\n",
1860 PL_colors[4],PL_colors[5]));
1861 goto phooey; /* Should not happen! */
1863 dontbother = strend - last + prog->float_min_offset;
1865 if (minlen && (dontbother < minlen))
1866 dontbother = minlen - 1;
1867 strend -= dontbother; /* this one's always in bytes! */
1868 /* We don't know much -- general case. */
1871 if (regtry(prog, s))
1880 if (regtry(prog, s))
1882 } while (s++ < strend);
1890 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1892 if (PL_reg_eval_set) {
1893 /* Preserve the current value of $^R */
1894 if (oreplsv != GvSV(PL_replgv))
1895 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1896 restored, the value remains
1898 restore_pos(aTHX_ 0);
1901 /* make sure $`, $&, $', and $digit will work later */
1902 if ( !(flags & REXEC_NOT_FIRST) ) {
1903 if (RX_MATCH_COPIED(prog)) {
1904 Safefree(prog->subbeg);
1905 RX_MATCH_COPIED_off(prog);
1907 if (flags & REXEC_COPY_STR) {
1908 I32 i = PL_regeol - startpos + (stringarg - strbeg);
1910 s = savepvn(strbeg, i);
1913 RX_MATCH_COPIED_on(prog);
1916 prog->subbeg = strbeg;
1917 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
1924 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
1925 PL_colors[4],PL_colors[5]));
1926 if (PL_reg_eval_set)
1927 restore_pos(aTHX_ 0);
1932 - regtry - try match at specific point
1934 STATIC I32 /* 0 failure, 1 success */
1935 S_regtry(pTHX_ regexp *prog, char *startpos)
1943 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
1945 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
1948 PL_reg_eval_set = RS_init;
1950 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
1951 (IV)(PL_stack_sp - PL_stack_base));
1953 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
1954 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
1955 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
1957 /* Apparently this is not needed, judging by wantarray. */
1958 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
1959 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
1962 /* Make $_ available to executed code. */
1963 if (PL_reg_sv != DEFSV) {
1964 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
1969 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
1970 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
1971 /* prepare for quick setting of pos */
1972 sv_magic(PL_reg_sv, (SV*)0,
1973 PERL_MAGIC_regex_global, Nullch, 0);
1974 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
1978 PL_reg_oldpos = mg->mg_len;
1979 SAVEDESTRUCTOR_X(restore_pos, 0);
1981 if (!PL_reg_curpm) {
1982 Newz(22,PL_reg_curpm, 1, PMOP);
1985 SV* repointer = newSViv(0);
1986 /* so we know which PL_regex_padav element is PL_reg_curpm */
1987 SvFLAGS(repointer) |= SVf_BREAK;
1988 av_push(PL_regex_padav,repointer);
1989 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
1990 PL_regex_pad = AvARRAY(PL_regex_padav);
1994 PM_SETRE(PL_reg_curpm, prog);
1995 PL_reg_oldcurpm = PL_curpm;
1996 PL_curpm = PL_reg_curpm;
1997 if (RX_MATCH_COPIED(prog)) {
1998 /* Here is a serious problem: we cannot rewrite subbeg,
1999 since it may be needed if this match fails. Thus
2000 $` inside (?{}) could fail... */
2001 PL_reg_oldsaved = prog->subbeg;
2002 PL_reg_oldsavedlen = prog->sublen;
2003 RX_MATCH_COPIED_off(prog);
2006 PL_reg_oldsaved = Nullch;
2007 prog->subbeg = PL_bostr;
2008 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2010 prog->startp[0] = startpos - PL_bostr;
2011 PL_reginput = startpos;
2012 PL_regstartp = prog->startp;
2013 PL_regendp = prog->endp;
2014 PL_reglastparen = &prog->lastparen;
2015 PL_reglastcloseparen = &prog->lastcloseparen;
2016 prog->lastparen = 0;
2018 DEBUG_r(PL_reg_starttry = startpos);
2019 if (PL_reg_start_tmpl <= prog->nparens) {
2020 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2021 if(PL_reg_start_tmp)
2022 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2024 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2028 sv_setpvn(PERL_DEBUG_PAD(0), "", 0);
2029 sv_setpvn(PERL_DEBUG_PAD(1), "", 0);
2030 sv_setpvn(PERL_DEBUG_PAD(2), "", 0);
2033 /* XXXX What this code is doing here?!!! There should be no need
2034 to do this again and again, PL_reglastparen should take care of
2037 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2038 * Actually, the code in regcppop() (which Ilya may be meaning by
2039 * PL_reglastparen), is not needed at all by the test suite
2040 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2041 * enough, for building DynaLoader, or otherwise this
2042 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2043 * will happen. Meanwhile, this code *is* needed for the
2044 * above-mentioned test suite tests to succeed. The common theme
2045 * on those tests seems to be returning null fields from matches.
2050 if (prog->nparens) {
2051 for (i = prog->nparens; i > *PL_reglastparen; i--) {
2058 if (regmatch(prog->program + 1)) {
2059 prog->endp[0] = PL_reginput - PL_bostr;
2062 REGCP_UNWIND(lastcp);
2066 #define RE_UNWIND_BRANCH 1
2067 #define RE_UNWIND_BRANCHJ 2
2071 typedef struct { /* XX: makes sense to enlarge it... */
2075 } re_unwind_generic_t;
2088 } re_unwind_branch_t;
2090 typedef union re_unwind_t {
2092 re_unwind_generic_t generic;
2093 re_unwind_branch_t branch;
2096 #define sayYES goto yes
2097 #define sayNO goto no
2098 #define sayYES_FINAL goto yes_final
2099 #define sayYES_LOUD goto yes_loud
2100 #define sayNO_FINAL goto no_final
2101 #define sayNO_SILENT goto do_no
2102 #define saySAME(x) if (x) goto yes; else goto no
2104 #define REPORT_CODE_OFF 24
2107 - regmatch - main matching routine
2109 * Conceptually the strategy is simple: check to see whether the current
2110 * node matches, call self recursively to see whether the rest matches,
2111 * and then act accordingly. In practice we make some effort to avoid
2112 * recursion, in particular by going through "ordinary" nodes (that don't
2113 * need to know whether the rest of the match failed) by a loop instead of
2116 /* [lwall] I've hoisted the register declarations to the outer block in order to
2117 * maybe save a little bit of pushing and popping on the stack. It also takes
2118 * advantage of machines that use a register save mask on subroutine entry.
2120 STATIC I32 /* 0 failure, 1 success */
2121 S_regmatch(pTHX_ regnode *prog)
2123 register regnode *scan; /* Current node. */
2124 regnode *next; /* Next node. */
2125 regnode *inner; /* Next node in internal branch. */
2126 register I32 nextchr; /* renamed nextchr - nextchar colides with
2127 function of same name */
2128 register I32 n; /* no or next */
2129 register I32 ln = 0; /* len or last */
2130 register char *s = Nullch; /* operand or save */
2131 register char *locinput = PL_reginput;
2132 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2133 int minmod = 0, sw = 0, logical = 0;
2136 I32 firstcp = PL_savestack_ix;
2138 register bool do_utf8 = PL_reg_match_utf8;
2140 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2141 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2142 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2149 /* Note that nextchr is a byte even in UTF */
2150 nextchr = UCHARAT(locinput);
2152 while (scan != NULL) {
2155 SV *prop = sv_newmortal();
2156 int docolor = *PL_colors[0];
2157 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2158 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2159 /* The part of the string before starttry has one color
2160 (pref0_len chars), between starttry and current
2161 position another one (pref_len - pref0_len chars),
2162 after the current position the third one.
2163 We assume that pref0_len <= pref_len, otherwise we
2164 decrease pref0_len. */
2165 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2166 ? (5 + taill) - l : locinput - PL_bostr;
2169 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2171 pref0_len = pref_len - (locinput - PL_reg_starttry);
2172 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2173 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2174 ? (5 + taill) - pref_len : PL_regeol - locinput);
2175 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2179 if (pref0_len > pref_len)
2180 pref0_len = pref_len;
2181 regprop(prop, scan);
2185 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2187 locinput - pref_len;
2188 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2189 char *s1 = do_utf8 ?
2190 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2191 pref_len - pref0_len, 60, 0) :
2192 locinput - pref_len + pref0_len;
2193 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2194 char *s2 = do_utf8 ?
2195 pv_uni_display(dsv2, (U8*)locinput,
2196 PL_regeol - locinput, 60, 0) :
2198 int len2 = do_utf8 ? strlen(s2) : l;
2199 PerlIO_printf(Perl_debug_log,
2200 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2201 (IV)(locinput - PL_bostr),
2208 (docolor ? "" : "> <"),
2212 15 - l - pref_len + 1,
2214 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2219 next = scan + NEXT_OFF(scan);
2225 if (locinput == PL_bostr || (PL_multiline &&
2226 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2228 /* regtill = regbol; */
2233 if (locinput == PL_bostr ||
2234 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2240 if (locinput == PL_bostr)
2244 if (locinput == PL_reg_ganch)
2254 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2259 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2261 if (PL_regeol - locinput > 1)
2265 if (PL_regeol != locinput)
2269 if (!nextchr && locinput >= PL_regeol)
2272 locinput += PL_utf8skip[nextchr];
2273 if (locinput > PL_regeol)
2275 nextchr = UCHARAT(locinput);
2278 nextchr = UCHARAT(++locinput);
2281 if (!nextchr && locinput >= PL_regeol)
2283 nextchr = UCHARAT(++locinput);
2286 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2289 locinput += PL_utf8skip[nextchr];
2290 if (locinput > PL_regeol)
2292 nextchr = UCHARAT(locinput);
2295 nextchr = UCHARAT(++locinput);
2300 if (do_utf8 != (UTF!=0)) {
2301 /* The target and the pattern have differing utf8ness. */
2307 /* The target is utf8, the pattern is not utf8. */
2311 if (NATIVE_TO_UNI(*(U8*)s) !=
2312 utf8_to_uvchr((U8*)l, &ulen))
2319 /* The target is not utf8, the pattern is utf8. */
2323 if (NATIVE_TO_UNI(*((U8*)l)) !=
2324 utf8_to_uvchr((U8*)s, &ulen))
2331 nextchr = UCHARAT(locinput);
2334 /* The target and the pattern have the same utf8ness. */
2335 /* Inline the first character, for speed. */
2336 if (UCHARAT(s) != nextchr)
2338 if (PL_regeol - locinput < ln)
2340 if (ln > 1 && memNE(s, locinput, ln))
2343 nextchr = UCHARAT(locinput);
2346 PL_reg_flags |= RF_tainted;
2356 if (do_utf8 != (UTF!=0)) {
2357 /* The target and the pattern have differing utf8ness. */
2358 STRLEN ulen1, ulen2;
2362 /* The target is utf8, the pattern is not utf8. */
2367 cs = to_uni_fold(NATIVE_TO_UNI(*(U8*)s),
2369 cl = utf8_to_uvchr((U8*)l, &ulen2);
2372 cl = to_uni_fold(cl, (U8*)l, &ulen2);
2373 if (ulen1 != ulen2 || cs != cl)
2381 /* The target is not utf8, the pattern is utf8. */
2386 cs = utf8_to_uvchr((U8*)s, &ulen1);
2388 cl = to_uni_fold(NATIVE_TO_UNI(*(U8*)l),
2392 cs = to_uni_fold(cs, (U8*)s, &ulen1);
2393 if (ulen1 != ulen2 || cs != cl)
2401 nextchr = UCHARAT(locinput);
2405 if (do_utf8 && UTF) {
2406 /* Both the target and the pattern are utf8. */
2407 U8 lfoldbuf[UTF8_MAXLEN_FOLD+1], *lf;
2408 U8 sfoldbuf[UTF8_MAXLEN_FOLD+1], *sf;
2409 STRLEN lfoldlen, sfoldlen;
2414 /* Fold them and walk them characterwise. */
2417 to_utf8_fold((U8*)l, lfoldbuf, &lfoldlen);
2423 to_utf8_fold((U8*)s, sfoldbuf, &sfoldlen);
2428 while (llen && slen) {
2429 if (UTF8SKIP(lf) != UTF8SKIP(sf) ||
2430 memNE((char*)lf, (char*)sf, UTF8SKIP(lf)))
2432 llen -= UTF8SKIP(lf);
2434 slen -= UTF8SKIP(sf);
2442 nextchr = UCHARAT(locinput);
2447 /* Neither the target and the pattern are utf8. */
2449 /* Inline the first character, for speed. */
2450 if (UCHARAT(s) != nextchr &&
2451 UCHARAT(s) != ((OP(scan) == EXACTF)
2452 ? PL_fold : PL_fold_locale)[nextchr])
2454 if (PL_regeol - locinput < ln)
2456 if (ln > 1 && (OP(scan) == EXACTF
2457 ? ibcmp(s, locinput, ln)
2458 : ibcmp_locale(s, locinput, ln)))
2461 nextchr = UCHARAT(locinput);
2465 if (!reginclass(scan, (U8*)locinput, do_utf8))
2467 if (locinput >= PL_regeol)
2469 locinput += PL_utf8skip[nextchr];
2470 nextchr = UCHARAT(locinput);
2474 nextchr = UCHARAT(locinput);
2475 if (!reginclass(scan, (U8*)locinput, do_utf8))
2477 if (!nextchr && locinput >= PL_regeol)
2479 nextchr = UCHARAT(++locinput);
2483 PL_reg_flags |= RF_tainted;
2489 LOAD_UTF8_CHARCLASS(alnum,"a");
2490 if (!(OP(scan) == ALNUM
2491 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2492 : isALNUM_LC_utf8((U8*)locinput)))
2496 locinput += PL_utf8skip[nextchr];
2497 nextchr = UCHARAT(locinput);
2500 if (!(OP(scan) == ALNUM
2501 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2503 nextchr = UCHARAT(++locinput);
2506 PL_reg_flags |= RF_tainted;
2509 if (!nextchr && locinput >= PL_regeol)
2512 LOAD_UTF8_CHARCLASS(alnum,"a");
2513 if (OP(scan) == NALNUM
2514 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2515 : isALNUM_LC_utf8((U8*)locinput))
2519 locinput += PL_utf8skip[nextchr];
2520 nextchr = UCHARAT(locinput);
2523 if (OP(scan) == NALNUM
2524 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2526 nextchr = UCHARAT(++locinput);
2530 PL_reg_flags |= RF_tainted;
2534 /* was last char in word? */
2536 if (locinput == PL_bostr)
2539 U8 *r = reghop((U8*)locinput, -1);
2541 ln = utf8n_to_uvchr(r, s - (char*)r, 0, 0);
2543 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2544 ln = isALNUM_uni(ln);
2545 LOAD_UTF8_CHARCLASS(alnum,"a");
2546 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2549 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2550 n = isALNUM_LC_utf8((U8*)locinput);
2554 ln = (locinput != PL_bostr) ?
2555 UCHARAT(locinput - 1) : '\n';
2556 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2558 n = isALNUM(nextchr);
2561 ln = isALNUM_LC(ln);
2562 n = isALNUM_LC(nextchr);
2565 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2566 OP(scan) == BOUNDL))
2570 PL_reg_flags |= RF_tainted;
2576 if (UTF8_IS_CONTINUED(nextchr)) {
2577 LOAD_UTF8_CHARCLASS(space," ");
2578 if (!(OP(scan) == SPACE
2579 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2580 : isSPACE_LC_utf8((U8*)locinput)))
2584 locinput += PL_utf8skip[nextchr];
2585 nextchr = UCHARAT(locinput);
2588 if (!(OP(scan) == SPACE
2589 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2591 nextchr = UCHARAT(++locinput);
2594 if (!(OP(scan) == SPACE
2595 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2597 nextchr = UCHARAT(++locinput);
2601 PL_reg_flags |= RF_tainted;
2604 if (!nextchr && locinput >= PL_regeol)
2607 LOAD_UTF8_CHARCLASS(space," ");
2608 if (OP(scan) == NSPACE
2609 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2610 : isSPACE_LC_utf8((U8*)locinput))
2614 locinput += PL_utf8skip[nextchr];
2615 nextchr = UCHARAT(locinput);
2618 if (OP(scan) == NSPACE
2619 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2621 nextchr = UCHARAT(++locinput);
2624 PL_reg_flags |= RF_tainted;
2630 LOAD_UTF8_CHARCLASS(digit,"0");
2631 if (!(OP(scan) == DIGIT
2632 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2633 : isDIGIT_LC_utf8((U8*)locinput)))
2637 locinput += PL_utf8skip[nextchr];
2638 nextchr = UCHARAT(locinput);
2641 if (!(OP(scan) == DIGIT
2642 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2644 nextchr = UCHARAT(++locinput);
2647 PL_reg_flags |= RF_tainted;
2650 if (!nextchr && locinput >= PL_regeol)
2653 LOAD_UTF8_CHARCLASS(digit,"0");
2654 if (OP(scan) == NDIGIT
2655 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2656 : isDIGIT_LC_utf8((U8*)locinput))
2660 locinput += PL_utf8skip[nextchr];
2661 nextchr = UCHARAT(locinput);
2664 if (OP(scan) == NDIGIT
2665 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2667 nextchr = UCHARAT(++locinput);
2670 if (locinput >= PL_regeol)
2673 LOAD_UTF8_CHARCLASS(mark,"~");
2674 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2676 locinput += PL_utf8skip[nextchr];
2677 while (locinput < PL_regeol &&
2678 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2679 locinput += UTF8SKIP(locinput);
2680 if (locinput > PL_regeol)
2685 nextchr = UCHARAT(locinput);
2688 PL_reg_flags |= RF_tainted;
2692 n = ARG(scan); /* which paren pair */
2693 ln = PL_regstartp[n];
2694 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2695 if (*PL_reglastparen < n || ln == -1)
2696 sayNO; /* Do not match unless seen CLOSEn. */
2697 if (ln == PL_regendp[n])
2701 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2703 char *e = PL_bostr + PL_regendp[n];
2705 * Note that we can't do the "other character" lookup trick as
2706 * in the 8-bit case (no pun intended) because in Unicode we
2707 * have to map both upper and title case to lower case.
2709 if (OP(scan) == REFF) {
2710 STRLEN ulen1, ulen2;
2711 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2712 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2716 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2717 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2718 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2725 nextchr = UCHARAT(locinput);
2729 /* Inline the first character, for speed. */
2730 if (UCHARAT(s) != nextchr &&
2732 (UCHARAT(s) != ((OP(scan) == REFF
2733 ? PL_fold : PL_fold_locale)[nextchr]))))
2735 ln = PL_regendp[n] - ln;
2736 if (locinput + ln > PL_regeol)
2738 if (ln > 1 && (OP(scan) == REF
2739 ? memNE(s, locinput, ln)
2741 ? ibcmp(s, locinput, ln)
2742 : ibcmp_locale(s, locinput, ln))))
2745 nextchr = UCHARAT(locinput);
2756 OP_4tree *oop = PL_op;
2757 COP *ocurcop = PL_curcop;
2758 SV **ocurpad = PL_curpad;
2762 PL_op = (OP_4tree*)PL_regdata->data[n];
2763 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2764 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2765 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2769 CALLRUNOPS(aTHX); /* Scalar context. */
2772 ret = Nullsv; /* protect against empty (?{}) blocks. */
2780 PL_curpad = ocurpad;
2781 PL_curcop = ocurcop;
2783 if (logical == 2) { /* Postponed subexpression. */
2785 MAGIC *mg = Null(MAGIC*);
2787 CHECKPOINT cp, lastcp;
2789 if(SvROK(ret) || SvRMAGICAL(ret)) {
2790 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2793 mg = mg_find(sv, PERL_MAGIC_qr);
2796 re = (regexp *)mg->mg_obj;
2797 (void)ReREFCNT_inc(re);
2801 char *t = SvPV(ret, len);
2803 char *oprecomp = PL_regprecomp;
2804 I32 osize = PL_regsize;
2805 I32 onpar = PL_regnpar;
2808 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2810 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2811 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2813 PL_regprecomp = oprecomp;
2818 PerlIO_printf(Perl_debug_log,
2819 "Entering embedded `%s%.60s%s%s'\n",
2823 (strlen(re->precomp) > 60 ? "..." : ""))
2826 state.prev = PL_reg_call_cc;
2827 state.cc = PL_regcc;
2828 state.re = PL_reg_re;
2832 cp = regcppush(0); /* Save *all* the positions. */
2835 state.ss = PL_savestack_ix;
2836 *PL_reglastparen = 0;
2837 *PL_reglastcloseparen = 0;
2838 PL_reg_call_cc = &state;
2839 PL_reginput = locinput;
2841 /* XXXX This is too dramatic a measure... */
2844 if (regmatch(re->program + 1)) {
2845 /* Even though we succeeded, we need to restore
2846 global variables, since we may be wrapped inside
2847 SUSPEND, thus the match may be not finished yet. */
2849 /* XXXX Do this only if SUSPENDed? */
2850 PL_reg_call_cc = state.prev;
2851 PL_regcc = state.cc;
2852 PL_reg_re = state.re;
2853 cache_re(PL_reg_re);
2855 /* XXXX This is too dramatic a measure... */
2858 /* These are needed even if not SUSPEND. */
2864 REGCP_UNWIND(lastcp);
2866 PL_reg_call_cc = state.prev;
2867 PL_regcc = state.cc;
2868 PL_reg_re = state.re;
2869 cache_re(PL_reg_re);
2871 /* XXXX This is too dramatic a measure... */
2881 sv_setsv(save_scalar(PL_replgv), ret);
2885 n = ARG(scan); /* which paren pair */
2886 PL_reg_start_tmp[n] = locinput;
2891 n = ARG(scan); /* which paren pair */
2892 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2893 PL_regendp[n] = locinput - PL_bostr;
2894 if (n > *PL_reglastparen)
2895 *PL_reglastparen = n;
2896 *PL_reglastcloseparen = n;
2899 n = ARG(scan); /* which paren pair */
2900 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2903 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2905 next = NEXTOPER(NEXTOPER(scan));
2907 next = scan + ARG(scan);
2908 if (OP(next) == IFTHEN) /* Fake one. */
2909 next = NEXTOPER(NEXTOPER(next));
2913 logical = scan->flags;
2915 /*******************************************************************
2916 PL_regcc contains infoblock about the innermost (...)* loop, and
2917 a pointer to the next outer infoblock.
2919 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2921 1) After matching X, regnode for CURLYX is processed;
2923 2) This regnode creates infoblock on the stack, and calls
2924 regmatch() recursively with the starting point at WHILEM node;
2926 3) Each hit of WHILEM node tries to match A and Z (in the order
2927 depending on the current iteration, min/max of {min,max} and
2928 greediness). The information about where are nodes for "A"
2929 and "Z" is read from the infoblock, as is info on how many times "A"
2930 was already matched, and greediness.
2932 4) After A matches, the same WHILEM node is hit again.
2934 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2935 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2936 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2937 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2938 of the external loop.
2940 Currently present infoblocks form a tree with a stem formed by PL_curcc
2941 and whatever it mentions via ->next, and additional attached trees
2942 corresponding to temporarily unset infoblocks as in "5" above.
2944 In the following picture infoblocks for outer loop of
2945 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2946 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2947 infoblocks are drawn below the "reset" infoblock.
2949 In fact in the picture below we do not show failed matches for Z and T
2950 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2951 more obvious *why* one needs to *temporary* unset infoblocks.]
2953 Matched REx position InfoBlocks Comment
2957 Y A)*?Z)*?T x <- O <- I
2958 YA )*?Z)*?T x <- O <- I
2959 YA A)*?Z)*?T x <- O <- I
2960 YAA )*?Z)*?T x <- O <- I
2961 YAA Z)*?T x <- O # Temporary unset I
2964 YAAZ Y(A)*?Z)*?T x <- O
2967 YAAZY (A)*?Z)*?T x <- O
2970 YAAZY A)*?Z)*?T x <- O <- I
2973 YAAZYA )*?Z)*?T x <- O <- I
2976 YAAZYA Z)*?T x <- O # Temporary unset I
2982 YAAZYAZ T x # Temporary unset O
2989 *******************************************************************/
2992 CHECKPOINT cp = PL_savestack_ix;
2993 /* No need to save/restore up to this paren */
2994 I32 parenfloor = scan->flags;
2996 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2998 cc.oldcc = PL_regcc;
3000 /* XXXX Probably it is better to teach regpush to support
3001 parenfloor > PL_regsize... */
3002 if (parenfloor > *PL_reglastparen)
3003 parenfloor = *PL_reglastparen; /* Pessimization... */
3004 cc.parenfloor = parenfloor;
3006 cc.min = ARG1(scan);
3007 cc.max = ARG2(scan);
3008 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3012 PL_reginput = locinput;
3013 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
3015 PL_regcc = cc.oldcc;
3021 * This is really hard to understand, because after we match
3022 * what we're trying to match, we must make sure the rest of
3023 * the REx is going to match for sure, and to do that we have
3024 * to go back UP the parse tree by recursing ever deeper. And
3025 * if it fails, we have to reset our parent's current state
3026 * that we can try again after backing off.
3029 CHECKPOINT cp, lastcp;
3030 CURCUR* cc = PL_regcc;
3031 char *lastloc = cc->lastloc; /* Detection of 0-len. */
3033 n = cc->cur + 1; /* how many we know we matched */
3034 PL_reginput = locinput;
3037 PerlIO_printf(Perl_debug_log,
3038 "%*s %ld out of %ld..%ld cc=%lx\n",
3039 REPORT_CODE_OFF+PL_regindent*2, "",
3040 (long)n, (long)cc->min,
3041 (long)cc->max, (long)cc)
3044 /* If degenerate scan matches "", assume scan done. */
3046 if (locinput == cc->lastloc && n >= cc->min) {
3047 PL_regcc = cc->oldcc;
3051 PerlIO_printf(Perl_debug_log,
3052 "%*s empty match detected, try continuation...\n",
3053 REPORT_CODE_OFF+PL_regindent*2, "")
3055 if (regmatch(cc->next))
3063 /* First just match a string of min scans. */
3067 cc->lastloc = locinput;
3068 if (regmatch(cc->scan))
3071 cc->lastloc = lastloc;
3076 /* Check whether we already were at this position.
3077 Postpone detection until we know the match is not
3078 *that* much linear. */
3079 if (!PL_reg_maxiter) {
3080 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3081 PL_reg_leftiter = PL_reg_maxiter;
3083 if (PL_reg_leftiter-- == 0) {
3084 I32 size = (PL_reg_maxiter + 7)/8;
3085 if (PL_reg_poscache) {
3086 if (PL_reg_poscache_size < size) {
3087 Renew(PL_reg_poscache, size, char);
3088 PL_reg_poscache_size = size;
3090 Zero(PL_reg_poscache, size, char);
3093 PL_reg_poscache_size = size;
3094 Newz(29, PL_reg_poscache, size, char);
3097 PerlIO_printf(Perl_debug_log,
3098 "%sDetected a super-linear match, switching on caching%s...\n",
3099 PL_colors[4], PL_colors[5])
3102 if (PL_reg_leftiter < 0) {
3103 I32 o = locinput - PL_bostr, b;
3105 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3108 if (PL_reg_poscache[o] & (1<<b)) {
3110 PerlIO_printf(Perl_debug_log,
3111 "%*s already tried at this position...\n",
3112 REPORT_CODE_OFF+PL_regindent*2, "")
3116 PL_reg_poscache[o] |= (1<<b);
3120 /* Prefer next over scan for minimal matching. */
3123 PL_regcc = cc->oldcc;
3126 cp = regcppush(cc->parenfloor);
3128 if (regmatch(cc->next)) {
3130 sayYES; /* All done. */
3132 REGCP_UNWIND(lastcp);
3138 if (n >= cc->max) { /* Maximum greed exceeded? */
3139 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3140 && !(PL_reg_flags & RF_warned)) {
3141 PL_reg_flags |= RF_warned;
3142 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3143 "Complex regular subexpression recursion",
3150 PerlIO_printf(Perl_debug_log,
3151 "%*s trying longer...\n",
3152 REPORT_CODE_OFF+PL_regindent*2, "")
3154 /* Try scanning more and see if it helps. */
3155 PL_reginput = locinput;
3157 cc->lastloc = locinput;
3158 cp = regcppush(cc->parenfloor);
3160 if (regmatch(cc->scan)) {
3164 REGCP_UNWIND(lastcp);
3167 cc->lastloc = lastloc;
3171 /* Prefer scan over next for maximal matching. */
3173 if (n < cc->max) { /* More greed allowed? */
3174 cp = regcppush(cc->parenfloor);
3176 cc->lastloc = locinput;
3178 if (regmatch(cc->scan)) {
3182 REGCP_UNWIND(lastcp);
3183 regcppop(); /* Restore some previous $<digit>s? */
3184 PL_reginput = locinput;
3186 PerlIO_printf(Perl_debug_log,
3187 "%*s failed, try continuation...\n",
3188 REPORT_CODE_OFF+PL_regindent*2, "")
3191 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3192 && !(PL_reg_flags & RF_warned)) {
3193 PL_reg_flags |= RF_warned;
3194 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3195 "Complex regular subexpression recursion",
3199 /* Failed deeper matches of scan, so see if this one works. */
3200 PL_regcc = cc->oldcc;
3203 if (regmatch(cc->next))
3209 cc->lastloc = lastloc;
3214 next = scan + ARG(scan);
3217 inner = NEXTOPER(NEXTOPER(scan));
3220 inner = NEXTOPER(scan);
3224 if (OP(next) != c1) /* No choice. */
3225 next = inner; /* Avoid recursion. */
3227 I32 lastparen = *PL_reglastparen;
3229 re_unwind_branch_t *uw;
3231 /* Put unwinding data on stack */
3232 unwind1 = SSNEWt(1,re_unwind_branch_t);
3233 uw = SSPTRt(unwind1,re_unwind_branch_t);
3236 uw->type = ((c1 == BRANCH)
3238 : RE_UNWIND_BRANCHJ);
3239 uw->lastparen = lastparen;
3241 uw->locinput = locinput;
3242 uw->nextchr = nextchr;
3244 uw->regindent = ++PL_regindent;
3247 REGCP_SET(uw->lastcp);
3249 /* Now go into the first branch */
3262 /* We suppose that the next guy does not need
3263 backtracking: in particular, it is of constant length,
3264 and has no parenths to influence future backrefs. */
3265 ln = ARG1(scan); /* min to match */
3266 n = ARG2(scan); /* max to match */
3267 paren = scan->flags;
3269 if (paren > PL_regsize)
3271 if (paren > *PL_reglastparen)
3272 *PL_reglastparen = paren;
3274 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3276 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3277 PL_reginput = locinput;
3280 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3282 /* if we matched something zero-length we don't need to
3283 backtrack - capturing parens are already defined, so
3284 the caveat in the maximal case doesn't apply
3286 XXXX if ln == 0, we can redo this check first time
3287 through the following loop
3290 n = ln; /* don't backtrack */
3291 locinput = PL_reginput;
3292 if (HAS_TEXT(next) || JUMPABLE(next)) {
3293 regnode *text_node = next;
3295 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3297 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3299 if (PL_regkind[(U8)OP(text_node)] == REF) {
3301 n = ARG(text_node); /* which paren pair */
3302 ln = PL_regstartp[n];
3303 /* assume yes if we haven't seen CLOSEn */
3305 *PL_reglastparen < n ||
3312 c1 = *(PL_bostr + ln);
3314 else { c1 = (U8)*STRING(text_node); }
3315 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3317 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3318 c2 = PL_fold_locale[c1];
3327 /* This may be improved if l == 0. */
3328 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3329 /* If it could work, try it. */
3331 UCHARAT(PL_reginput) == c1 ||
3332 UCHARAT(PL_reginput) == c2)
3336 PL_regstartp[paren] =
3337 HOPc(PL_reginput, -l) - PL_bostr;
3338 PL_regendp[paren] = PL_reginput - PL_bostr;
3341 PL_regendp[paren] = -1;
3345 REGCP_UNWIND(lastcp);
3347 /* Couldn't or didn't -- move forward. */
3348 PL_reginput = locinput;
3349 if (regrepeat_hard(scan, 1, &l)) {
3351 locinput = PL_reginput;
3358 n = regrepeat_hard(scan, n, &l);
3359 /* if we matched something zero-length we don't need to
3360 backtrack, unless the minimum count is zero and we
3361 are capturing the result - in that case the capture
3362 being defined or not may affect later execution
3364 if (n != 0 && l == 0 && !(paren && ln == 0))
3365 ln = n; /* don't backtrack */
3366 locinput = PL_reginput;
3368 PerlIO_printf(Perl_debug_log,
3369 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3370 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3374 if (HAS_TEXT(next) || JUMPABLE(next)) {
3375 regnode *text_node = next;
3377 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3379 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3381 if (PL_regkind[(U8)OP(text_node)] == REF) {
3383 n = ARG(text_node); /* which paren pair */
3384 ln = PL_regstartp[n];
3385 /* assume yes if we haven't seen CLOSEn */
3387 *PL_reglastparen < n ||
3394 c1 = *(PL_bostr + ln);
3396 else { c1 = (U8)*STRING(text_node); }
3398 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3400 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3401 c2 = PL_fold_locale[c1];
3412 /* If it could work, try it. */
3414 UCHARAT(PL_reginput) == c1 ||
3415 UCHARAT(PL_reginput) == c2)
3418 PerlIO_printf(Perl_debug_log,
3419 "%*s trying tail with n=%"IVdf"...\n",
3420 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3424 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3425 PL_regendp[paren] = PL_reginput - PL_bostr;
3428 PL_regendp[paren] = -1;
3432 REGCP_UNWIND(lastcp);
3434 /* Couldn't or didn't -- back up. */
3436 locinput = HOPc(locinput, -l);
3437 PL_reginput = locinput;
3444 paren = scan->flags; /* Which paren to set */
3445 if (paren > PL_regsize)
3447 if (paren > *PL_reglastparen)
3448 *PL_reglastparen = paren;
3449 ln = ARG1(scan); /* min to match */
3450 n = ARG2(scan); /* max to match */
3451 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3455 ln = ARG1(scan); /* min to match */
3456 n = ARG2(scan); /* max to match */
3457 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3462 scan = NEXTOPER(scan);
3468 scan = NEXTOPER(scan);
3472 * Lookahead to avoid useless match attempts
3473 * when we know what character comes next.
3477 * Used to only do .*x and .*?x, but now it allows
3478 * for )'s, ('s and (?{ ... })'s to be in the way
3479 * of the quantifier and the EXACT-like node. -- japhy
3482 if (HAS_TEXT(next) || JUMPABLE(next)) {
3484 regnode *text_node = next;
3486 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3488 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3490 if (PL_regkind[(U8)OP(text_node)] == REF) {
3492 n = ARG(text_node); /* which paren pair */
3493 ln = PL_regstartp[n];
3494 /* assume yes if we haven't seen CLOSEn */
3496 *PL_reglastparen < n ||
3501 goto assume_ok_easy;
3503 s = (U8*)PL_bostr + ln;
3505 else { s = (U8*)STRING(text_node); }
3509 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3511 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3512 c2 = PL_fold_locale[c1];
3515 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3516 STRLEN ulen1, ulen2;
3517 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3518 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3520 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3521 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3523 c1 = utf8_to_uvuni(tmpbuf1, 0);
3524 c2 = utf8_to_uvuni(tmpbuf2, 0);
3527 c2 = c1 = utf8_to_uvchr(s, NULL);
3535 PL_reginput = locinput;
3539 if (ln && regrepeat(scan, ln) < ln)
3541 locinput = PL_reginput;
3544 char *e; /* Should not check after this */
3545 char *old = locinput;
3547 if (n == REG_INFTY) {
3550 while (UTF8_IS_CONTINUATION(*(U8*)e))
3556 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3560 e = locinput + n - ln;
3566 /* Find place 'next' could work */
3569 while (locinput <= e &&
3570 UCHARAT(locinput) != c1)
3573 while (locinput <= e
3574 && UCHARAT(locinput) != c1
3575 && UCHARAT(locinput) != c2)
3578 count = locinput - old;
3585 utf8_to_uvchr((U8*)locinput, &len) != c1;
3590 for (count = 0; locinput <= e; count++) {
3591 UV c = utf8_to_uvchr((U8*)locinput, &len);
3592 if (c == c1 || c == c2)
3600 /* PL_reginput == old now */
3601 if (locinput != old) {
3602 ln = 1; /* Did some */
3603 if (regrepeat(scan, count) < count)
3606 /* PL_reginput == locinput now */
3607 TRYPAREN(paren, ln, locinput);
3608 PL_reginput = locinput; /* Could be reset... */
3609 REGCP_UNWIND(lastcp);
3610 /* Couldn't or didn't -- move forward. */
3613 locinput += UTF8SKIP(locinput);
3619 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3623 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3625 c = UCHARAT(PL_reginput);
3626 /* If it could work, try it. */
3627 if (c == c1 || c == c2)
3629 TRYPAREN(paren, n, PL_reginput);
3630 REGCP_UNWIND(lastcp);
3633 /* If it could work, try it. */
3634 else if (c1 == -1000)
3636 TRYPAREN(paren, n, PL_reginput);
3637 REGCP_UNWIND(lastcp);
3639 /* Couldn't or didn't -- move forward. */
3640 PL_reginput = locinput;
3641 if (regrepeat(scan, 1)) {
3643 locinput = PL_reginput;
3651 n = regrepeat(scan, n);
3652 locinput = PL_reginput;
3653 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3654 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3655 ln = n; /* why back off? */
3656 /* ...because $ and \Z can match before *and* after
3657 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3658 We should back off by one in this case. */
3659 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3668 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3670 c = UCHARAT(PL_reginput);
3672 /* If it could work, try it. */
3673 if (c1 == -1000 || c == c1 || c == c2)
3675 TRYPAREN(paren, n, PL_reginput);
3676 REGCP_UNWIND(lastcp);
3678 /* Couldn't or didn't -- back up. */
3680 PL_reginput = locinput = HOPc(locinput, -1);
3688 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3690 c = UCHARAT(PL_reginput);
3692 /* If it could work, try it. */
3693 if (c1 == -1000 || c == c1 || c == c2)
3695 TRYPAREN(paren, n, PL_reginput);
3696 REGCP_UNWIND(lastcp);
3698 /* Couldn't or didn't -- back up. */
3700 PL_reginput = locinput = HOPc(locinput, -1);
3707 if (PL_reg_call_cc) {
3708 re_cc_state *cur_call_cc = PL_reg_call_cc;
3709 CURCUR *cctmp = PL_regcc;
3710 regexp *re = PL_reg_re;
3711 CHECKPOINT cp, lastcp;
3713 cp = regcppush(0); /* Save *all* the positions. */
3715 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3717 PL_reginput = locinput; /* Make position available to
3719 cache_re(PL_reg_call_cc->re);
3720 PL_regcc = PL_reg_call_cc->cc;
3721 PL_reg_call_cc = PL_reg_call_cc->prev;
3722 if (regmatch(cur_call_cc->node)) {
3723 PL_reg_call_cc = cur_call_cc;
3727 REGCP_UNWIND(lastcp);
3729 PL_reg_call_cc = cur_call_cc;
3735 PerlIO_printf(Perl_debug_log,
3736 "%*s continuation failed...\n",
3737 REPORT_CODE_OFF+PL_regindent*2, "")
3741 if (locinput < PL_regtill) {
3742 DEBUG_r(PerlIO_printf(Perl_debug_log,
3743 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3745 (long)(locinput - PL_reg_starttry),
3746 (long)(PL_regtill - PL_reg_starttry),
3748 sayNO_FINAL; /* Cannot match: too short. */
3750 PL_reginput = locinput; /* put where regtry can find it */
3751 sayYES_FINAL; /* Success! */
3753 PL_reginput = locinput; /* put where regtry can find it */
3754 sayYES_LOUD; /* Success! */
3757 PL_reginput = locinput;
3762 s = HOPBACKc(locinput, scan->flags);
3768 PL_reginput = locinput;
3773 s = HOPBACKc(locinput, scan->flags);
3779 PL_reginput = locinput;
3782 inner = NEXTOPER(NEXTOPER(scan));
3783 if (regmatch(inner) != n) {
3798 if (OP(scan) == SUSPEND) {
3799 locinput = PL_reginput;
3800 nextchr = UCHARAT(locinput);
3805 next = scan + ARG(scan);
3810 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3811 PTR2UV(scan), OP(scan));
3812 Perl_croak(aTHX_ "regexp memory corruption");
3819 * We get here only if there's trouble -- normally "case END" is
3820 * the terminating point.
3822 Perl_croak(aTHX_ "corrupted regexp pointers");
3828 PerlIO_printf(Perl_debug_log,
3829 "%*s %scould match...%s\n",
3830 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3834 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3835 PL_colors[4],PL_colors[5]));
3841 #if 0 /* Breaks $^R */
3849 PerlIO_printf(Perl_debug_log,
3850 "%*s %sfailed...%s\n",
3851 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3857 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3860 case RE_UNWIND_BRANCH:
3861 case RE_UNWIND_BRANCHJ:
3863 re_unwind_branch_t *uwb = &(uw->branch);
3864 I32 lastparen = uwb->lastparen;
3866 REGCP_UNWIND(uwb->lastcp);
3867 for (n = *PL_reglastparen; n > lastparen; n--)
3869 *PL_reglastparen = n;
3870 scan = next = uwb->next;
3872 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3873 ? BRANCH : BRANCHJ) ) { /* Failure */
3880 /* Have more choice yet. Reuse the same uwb. */
3882 if ((n = (uwb->type == RE_UNWIND_BRANCH
3883 ? NEXT_OFF(next) : ARG(next))))
3886 next = NULL; /* XXXX Needn't unwinding in this case... */
3888 next = NEXTOPER(scan);
3889 if (uwb->type == RE_UNWIND_BRANCHJ)
3890 next = NEXTOPER(next);
3891 locinput = uwb->locinput;
3892 nextchr = uwb->nextchr;
3894 PL_regindent = uwb->regindent;
3901 Perl_croak(aTHX_ "regexp unwind memory corruption");
3912 - regrepeat - repeatedly match something simple, report how many
3915 * [This routine now assumes that it will only match on things of length 1.
3916 * That was true before, but now we assume scan - reginput is the count,
3917 * rather than incrementing count on every character. [Er, except utf8.]]
3920 S_regrepeat(pTHX_ regnode *p, I32 max)
3922 register char *scan;
3924 register char *loceol = PL_regeol;
3925 register I32 hardcount = 0;
3926 register bool do_utf8 = PL_reg_match_utf8;
3929 if (max != REG_INFTY && max < loceol - scan)
3930 loceol = scan + max;
3935 while (scan < loceol && hardcount < max && *scan != '\n') {
3936 scan += UTF8SKIP(scan);
3940 while (scan < loceol && *scan != '\n')
3950 case EXACT: /* length of string is 1 */
3952 while (scan < loceol && UCHARAT(scan) == c)
3955 case EXACTF: /* length of string is 1 */
3957 while (scan < loceol &&
3958 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3961 case EXACTFL: /* length of string is 1 */
3962 PL_reg_flags |= RF_tainted;
3964 while (scan < loceol &&
3965 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3971 while (hardcount < max && scan < loceol &&
3972 reginclass(p, (U8*)scan, do_utf8)) {
3973 scan += UTF8SKIP(scan);
3977 while (scan < loceol && reginclass(p, (U8*)scan, do_utf8))
3984 LOAD_UTF8_CHARCLASS(alnum,"a");
3985 while (hardcount < max && scan < loceol &&
3986 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3987 scan += UTF8SKIP(scan);
3991 while (scan < loceol && isALNUM(*scan))
3996 PL_reg_flags |= RF_tainted;
3999 while (hardcount < max && scan < loceol &&
4000 isALNUM_LC_utf8((U8*)scan)) {
4001 scan += UTF8SKIP(scan);
4005 while (scan < loceol && isALNUM_LC(*scan))
4012 LOAD_UTF8_CHARCLASS(alnum,"a");
4013 while (hardcount < max && scan < loceol &&
4014 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4015 scan += UTF8SKIP(scan);
4019 while (scan < loceol && !isALNUM(*scan))
4024 PL_reg_flags |= RF_tainted;
4027 while (hardcount < max && scan < loceol &&
4028 !isALNUM_LC_utf8((U8*)scan)) {
4029 scan += UTF8SKIP(scan);
4033 while (scan < loceol && !isALNUM_LC(*scan))
4040 LOAD_UTF8_CHARCLASS(space," ");
4041 while (hardcount < max && scan < loceol &&
4043 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4044 scan += UTF8SKIP(scan);
4048 while (scan < loceol && isSPACE(*scan))
4053 PL_reg_flags |= RF_tainted;
4056 while (hardcount < max && scan < loceol &&
4057 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4058 scan += UTF8SKIP(scan);
4062 while (scan < loceol && isSPACE_LC(*scan))
4069 LOAD_UTF8_CHARCLASS(space," ");
4070 while (hardcount < max && scan < loceol &&
4072 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4073 scan += UTF8SKIP(scan);
4077 while (scan < loceol && !isSPACE(*scan))
4082 PL_reg_flags |= RF_tainted;
4085 while (hardcount < max && scan < loceol &&
4086 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4087 scan += UTF8SKIP(scan);
4091 while (scan < loceol && !isSPACE_LC(*scan))
4098 LOAD_UTF8_CHARCLASS(digit,"0");
4099 while (hardcount < max && scan < loceol &&
4100 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4101 scan += UTF8SKIP(scan);
4105 while (scan < loceol && isDIGIT(*scan))
4112 LOAD_UTF8_CHARCLASS(digit,"0");
4113 while (hardcount < max && scan < loceol &&
4114 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4115 scan += UTF8SKIP(scan);
4119 while (scan < loceol && !isDIGIT(*scan))
4123 default: /* Called on something of 0 width. */
4124 break; /* So match right here or not at all. */
4130 c = scan - PL_reginput;
4135 SV *prop = sv_newmortal();
4138 PerlIO_printf(Perl_debug_log,
4139 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4140 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4147 - regrepeat_hard - repeatedly match something, report total lenth and length
4149 * The repeater is supposed to have constant length.
4153 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4155 register char *scan = Nullch;
4156 register char *start;
4157 register char *loceol = PL_regeol;
4159 I32 count = 0, res = 1;
4164 start = PL_reginput;
4165 if (PL_reg_match_utf8) {
4166 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4169 while (start < PL_reginput) {
4171 start += UTF8SKIP(start);
4182 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4184 *lp = l = PL_reginput - start;
4185 if (max != REG_INFTY && l*max < loceol - scan)
4186 loceol = scan + l*max;
4199 - regclass_swash - prepare the utf8 swash
4203 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** initsvp)
4208 if (PL_regdata && PL_regdata->count) {
4211 if (PL_regdata->what[n] == 's') {
4212 SV *rv = (SV*)PL_regdata->data[n];
4213 AV *av = (AV*)SvRV((SV*)rv);
4216 si = *av_fetch(av, 0, FALSE);
4217 a = av_fetch(av, 1, FALSE);
4221 else if (si && doinit) {
4222 sw = swash_init("utf8", "", si, 1, 0);
4223 (void)av_store(av, 1, sw);
4235 - reginclass - determine if a character falls into a character class
4239 S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8)
4241 char flags = ANYOF_FLAGS(n);
4246 c = do_utf8 ? utf8_to_uvchr(p, &len) : *p;
4248 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4249 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4250 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4253 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4256 SV *sw = regclass_swash(n, TRUE, 0);
4259 if (swash_fetch(sw, p, do_utf8))
4261 else if (flags & ANYOF_FOLD) {
4263 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
4265 to_utf8_fold(p, tmpbuf, &ulen);
4266 if (swash_fetch(sw, tmpbuf, do_utf8))
4268 to_utf8_upper(p, tmpbuf, &ulen);
4269 if (swash_fetch(sw, tmpbuf, do_utf8))
4275 if (!match && c < 256) {
4276 if (ANYOF_BITMAP_TEST(n, c))
4278 else if (flags & ANYOF_FOLD) {
4281 if (flags & ANYOF_LOCALE) {
4282 PL_reg_flags |= RF_tainted;
4283 f = PL_fold_locale[c];
4287 if (f != c && ANYOF_BITMAP_TEST(n, f))
4291 if (!match && (flags & ANYOF_CLASS)) {
4292 PL_reg_flags |= RF_tainted;
4294 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4295 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4296 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4297 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4298 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4299 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4300 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4301 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4302 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4303 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4304 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4305 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4306 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4307 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4308 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4309 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4310 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4311 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4312 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4313 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4314 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4315 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4316 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4317 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4318 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4319 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4320 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4321 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4322 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4323 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4324 ) /* How's that for a conditional? */
4331 return (flags & ANYOF_INVERT) ? !match : match;
4335 S_reghop(pTHX_ U8 *s, I32 off)
4337 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4341 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4344 while (off-- && s < lim) {
4345 /* XXX could check well-formedness here */
4353 if (UTF8_IS_CONTINUED(*s)) {
4354 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4357 /* XXX could check well-formedness here */
4365 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4367 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4371 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4374 while (off-- && s < lim) {
4375 /* XXX could check well-formedness here */
4385 if (UTF8_IS_CONTINUED(*s)) {
4386 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4389 /* XXX could check well-formedness here */
4401 restore_pos(pTHX_ void *arg)
4403 if (PL_reg_eval_set) {
4404 if (PL_reg_oldsaved) {
4405 PL_reg_re->subbeg = PL_reg_oldsaved;
4406 PL_reg_re->sublen = PL_reg_oldsavedlen;
4407 RX_MATCH_COPIED_on(PL_reg_re);
4409 PL_reg_magic->mg_len = PL_reg_oldpos;
4410 PL_reg_eval_set = 0;
4411 PL_curpm = PL_reg_oldcurpm;