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];
966 e = s; /* Due to minlen logic of intuit() */
970 U8 tmpbuf [UTF8_MAXLEN+1];
971 U8 foldbuf[UTF8_MAXLEN_FOLD+1];
974 /* The ibcmp_utf8() uses to_uni_fold() which is more
975 * correct folding for Unicode than using lowercase.
976 * However, it doesn't work quite fully since the folding
977 * is a one-to-many mapping and the regex optimizer is
978 * unaware of this, so it may throw out good matches.
979 * Fortunately, not getting this right is allowed
980 * for Unicode Regular Expression Support level 1,
981 * only one-to-one matching is required. --jhi */
984 c = utf8_to_uvchr((U8*)s, &len);
985 uvchr_to_utf8(tmpbuf, c);
986 to_utf8_fold(tmpbuf, foldbuf, &foldlen);
987 f = utf8_to_uvchr(foldbuf, 0);
989 if ( (c == c1 && ln == len) ||
990 (f == c1 && ln == foldlen)
991 || !ibcmp_utf8(s, do_utf8, (I32)(strend - s),
993 && (norun || regtry(prog, s)) )
1000 c = utf8_to_uvchr((U8*)s, &len);
1001 uvchr_to_utf8(tmpbuf, c);
1002 to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1003 f = utf8_to_uvchr(foldbuf, 0);
1005 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1006 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1007 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1008 if ( ((c == c1 || c == c2) && ln == len) ||
1009 ((f == c1 || f == c2) && ln == foldlen)
1010 || !ibcmp_utf8(s, do_utf8, (I32)(strend - s),
1012 && (norun || regtry(prog, s)) )
1022 && (ln == 1 || !(OP(c) == EXACTF
1024 : ibcmp_locale(s, m, ln)))
1025 && (norun || regtry(prog, s)) )
1031 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1032 && (ln == 1 || !(OP(c) == EXACTF
1034 : ibcmp_locale(s, m, ln)))
1035 && (norun || regtry(prog, s)) )
1042 PL_reg_flags |= RF_tainted;
1049 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1052 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1054 tmp = ((OP(c) == BOUND ?
1055 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1056 LOAD_UTF8_CHARCLASS(alnum,"a");
1057 while (s < strend) {
1058 if (tmp == !(OP(c) == BOUND ?
1059 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1060 isALNUM_LC_utf8((U8*)s)))
1063 if ((norun || regtry(prog, s)))
1070 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1071 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1072 while (s < strend) {
1074 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1076 if ((norun || regtry(prog, s)))
1082 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1086 PL_reg_flags |= RF_tainted;
1093 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1096 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1098 tmp = ((OP(c) == NBOUND ?
1099 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1100 LOAD_UTF8_CHARCLASS(alnum,"a");
1101 while (s < strend) {
1102 if (tmp == !(OP(c) == NBOUND ?
1103 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1104 isALNUM_LC_utf8((U8*)s)))
1106 else if ((norun || regtry(prog, s)))
1112 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1113 tmp = ((OP(c) == NBOUND ?
1114 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1115 while (s < strend) {
1117 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1119 else if ((norun || regtry(prog, s)))
1124 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1129 LOAD_UTF8_CHARCLASS(alnum,"a");
1130 while (s < strend) {
1131 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1132 if (tmp && (norun || regtry(prog, s)))
1143 while (s < strend) {
1145 if (tmp && (norun || regtry(prog, s)))
1157 PL_reg_flags |= RF_tainted;
1159 while (s < strend) {
1160 if (isALNUM_LC_utf8((U8*)s)) {
1161 if (tmp && (norun || regtry(prog, s)))
1172 while (s < strend) {
1173 if (isALNUM_LC(*s)) {
1174 if (tmp && (norun || regtry(prog, s)))
1187 LOAD_UTF8_CHARCLASS(alnum,"a");
1188 while (s < strend) {
1189 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1190 if (tmp && (norun || regtry(prog, s)))
1201 while (s < strend) {
1203 if (tmp && (norun || regtry(prog, s)))
1215 PL_reg_flags |= RF_tainted;
1217 while (s < strend) {
1218 if (!isALNUM_LC_utf8((U8*)s)) {
1219 if (tmp && (norun || regtry(prog, s)))
1230 while (s < strend) {
1231 if (!isALNUM_LC(*s)) {
1232 if (tmp && (norun || regtry(prog, s)))
1245 LOAD_UTF8_CHARCLASS(space," ");
1246 while (s < strend) {
1247 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1248 if (tmp && (norun || regtry(prog, s)))
1259 while (s < strend) {
1261 if (tmp && (norun || regtry(prog, s)))
1273 PL_reg_flags |= RF_tainted;
1275 while (s < strend) {
1276 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1277 if (tmp && (norun || regtry(prog, s)))
1288 while (s < strend) {
1289 if (isSPACE_LC(*s)) {
1290 if (tmp && (norun || regtry(prog, s)))
1303 LOAD_UTF8_CHARCLASS(space," ");
1304 while (s < strend) {
1305 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1306 if (tmp && (norun || regtry(prog, s)))
1317 while (s < strend) {
1319 if (tmp && (norun || regtry(prog, s)))
1331 PL_reg_flags |= RF_tainted;
1333 while (s < strend) {
1334 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1335 if (tmp && (norun || regtry(prog, s)))
1346 while (s < strend) {
1347 if (!isSPACE_LC(*s)) {
1348 if (tmp && (norun || regtry(prog, s)))
1361 LOAD_UTF8_CHARCLASS(digit,"0");
1362 while (s < strend) {
1363 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1364 if (tmp && (norun || regtry(prog, s)))
1375 while (s < strend) {
1377 if (tmp && (norun || regtry(prog, s)))
1389 PL_reg_flags |= RF_tainted;
1391 while (s < strend) {
1392 if (isDIGIT_LC_utf8((U8*)s)) {
1393 if (tmp && (norun || regtry(prog, s)))
1404 while (s < strend) {
1405 if (isDIGIT_LC(*s)) {
1406 if (tmp && (norun || regtry(prog, s)))
1419 LOAD_UTF8_CHARCLASS(digit,"0");
1420 while (s < strend) {
1421 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1422 if (tmp && (norun || regtry(prog, s)))
1433 while (s < strend) {
1435 if (tmp && (norun || regtry(prog, s)))
1447 PL_reg_flags |= RF_tainted;
1449 while (s < strend) {
1450 if (!isDIGIT_LC_utf8((U8*)s)) {
1451 if (tmp && (norun || regtry(prog, s)))
1462 while (s < strend) {
1463 if (!isDIGIT_LC(*s)) {
1464 if (tmp && (norun || regtry(prog, s)))
1476 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1485 - regexec_flags - match a regexp against a string
1488 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1489 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1490 /* strend: pointer to null at end of string */
1491 /* strbeg: real beginning of string */
1492 /* minend: end of match must be >=minend after stringarg. */
1493 /* data: May be used for some additional optimizations. */
1494 /* nosave: For optimizations. */
1497 register regnode *c;
1498 register char *startpos = stringarg;
1499 I32 minlen; /* must match at least this many chars */
1500 I32 dontbother = 0; /* how many characters not to try at end */
1501 /* I32 start_shift = 0; */ /* Offset of the start to find
1502 constant substr. */ /* CC */
1503 I32 end_shift = 0; /* Same for the end. */ /* CC */
1504 I32 scream_pos = -1; /* Internal iterator of scream. */
1506 SV* oreplsv = GvSV(PL_replgv);
1507 bool do_utf8 = DO_UTF8(sv);
1509 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
1516 PL_regnarrate = DEBUG_r_TEST;
1519 /* Be paranoid... */
1520 if (prog == NULL || startpos == NULL) {
1521 Perl_croak(aTHX_ "NULL regexp parameter");
1525 minlen = prog->minlen;
1526 if (strend - startpos < minlen) {
1527 DEBUG_r(PerlIO_printf(Perl_debug_log,
1528 "String too short [regexec_flags]...\n"));
1532 /* Check validity of program. */
1533 if (UCHARAT(prog->program) != REG_MAGIC) {
1534 Perl_croak(aTHX_ "corrupted regexp program");
1538 PL_reg_eval_set = 0;
1541 if (prog->reganch & ROPT_UTF8)
1542 PL_reg_flags |= RF_utf8;
1544 /* Mark beginning of line for ^ and lookbehind. */
1545 PL_regbol = startpos;
1549 /* Mark end of line for $ (and such) */
1552 /* see how far we have to get to not match where we matched before */
1553 PL_regtill = startpos+minend;
1555 /* We start without call_cc context. */
1558 /* If there is a "must appear" string, look for it. */
1561 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1564 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1565 PL_reg_ganch = startpos;
1566 else if (sv && SvTYPE(sv) >= SVt_PVMG
1568 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1569 && mg->mg_len >= 0) {
1570 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1571 if (prog->reganch & ROPT_ANCH_GPOS) {
1572 if (s > PL_reg_ganch)
1577 else /* pos() not defined */
1578 PL_reg_ganch = strbeg;
1581 if (do_utf8 == (UTF!=0) &&
1582 !(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
1583 re_scream_pos_data d;
1585 d.scream_olds = &scream_olds;
1586 d.scream_pos = &scream_pos;
1587 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1589 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1590 goto phooey; /* not present */
1595 char *s = do_utf8 ? sv_uni_display(dsv, sv, 60, 0) : startpos;
1596 int len = do_utf8 ? strlen(s) : strend - startpos;
1599 PerlIO_printf(Perl_debug_log,
1600 "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
1601 PL_colors[4],PL_colors[5],PL_colors[0],
1604 (strlen(prog->precomp) > 60 ? "..." : ""),
1606 (int)(len > 60 ? 60 : len),
1608 (len > 60 ? "..." : "")
1612 /* Simplest case: anchored match need be tried only once. */
1613 /* [unless only anchor is BOL and multiline is set] */
1614 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1615 if (s == startpos && regtry(prog, startpos))
1617 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1618 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1623 dontbother = minlen - 1;
1624 end = HOP3c(strend, -dontbother, strbeg) - 1;
1625 /* for multiline we only have to try after newlines */
1626 if (prog->check_substr) {
1630 if (regtry(prog, s))
1635 if (prog->reganch & RE_USE_INTUIT) {
1636 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1647 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1648 if (regtry(prog, s))
1655 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1656 if (regtry(prog, PL_reg_ganch))
1661 /* Messy cases: unanchored match. */
1662 if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
1663 /* we have /x+whatever/ */
1664 /* it must be a one character string (XXXX Except UTF?) */
1665 char ch = SvPVX(prog->anchored_substr)[0];
1671 while (s < strend) {
1673 DEBUG_r( did_match = 1 );
1674 if (regtry(prog, s)) goto got_it;
1676 while (s < strend && *s == ch)
1683 while (s < strend) {
1685 DEBUG_r( did_match = 1 );
1686 if (regtry(prog, s)) goto got_it;
1688 while (s < strend && *s == ch)
1694 DEBUG_r(if (!did_match)
1695 PerlIO_printf(Perl_debug_log,
1696 "Did not find anchored character...\n")
1700 else if (do_utf8 == (UTF!=0) &&
1701 (prog->anchored_substr != Nullsv
1702 || (prog->float_substr != Nullsv
1703 && prog->float_max_offset < strend - s))) {
1704 SV *must = prog->anchored_substr
1705 ? prog->anchored_substr : prog->float_substr;
1707 prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset;
1709 prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
1710 char *last = HOP3c(strend, /* Cannot start after this */
1711 -(I32)(CHR_SVLEN(must)
1712 - (SvTAIL(must) != 0) + back_min), strbeg);
1713 char *last1; /* Last position checked before */
1719 last1 = HOPc(s, -1);
1721 last1 = s - 1; /* bogus */
1723 /* XXXX check_substr already used to find `s', can optimize if
1724 check_substr==must. */
1726 dontbother = end_shift;
1727 strend = HOPc(strend, -dontbother);
1728 while ( (s <= last) &&
1729 ((flags & REXEC_SCREAM)
1730 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1731 end_shift, &scream_pos, 0))
1732 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1733 (unsigned char*)strend, must,
1734 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1735 DEBUG_r( did_match = 1 );
1736 if (HOPc(s, -back_max) > last1) {
1737 last1 = HOPc(s, -back_min);
1738 s = HOPc(s, -back_max);
1741 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1743 last1 = HOPc(s, -back_min);
1747 while (s <= last1) {
1748 if (regtry(prog, s))
1754 while (s <= last1) {
1755 if (regtry(prog, s))
1761 DEBUG_r(if (!did_match)
1762 PerlIO_printf(Perl_debug_log,
1763 "Did not find %s substr `%s%.*s%s'%s...\n",
1764 ((must == prog->anchored_substr)
1765 ? "anchored" : "floating"),
1767 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1769 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1773 else if ((c = prog->regstclass)) {
1774 if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT)
1775 /* don't bother with what can't match */
1776 strend = HOPc(strend, -(minlen - 1));
1778 SV *prop = sv_newmortal();
1780 PerlIO_printf(Perl_debug_log, "Matching stclass `%s' against `%s'\n", SvPVX(prop), UTF ? sv_uni_display(dsv, sv, 60, 0) : s);
1782 if (find_byclass(prog, c, s, strend, startpos, 0))
1784 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1788 if (prog->float_substr != Nullsv) { /* Trim the end. */
1791 if (flags & REXEC_SCREAM) {
1792 last = screaminstr(sv, prog->float_substr, s - strbeg,
1793 end_shift, &scream_pos, 1); /* last one */
1795 last = scream_olds; /* Only one occurrence. */
1799 char *little = SvPV(prog->float_substr, len);
1801 if (SvTAIL(prog->float_substr)) {
1802 if (memEQ(strend - len + 1, little, len - 1))
1803 last = strend - len + 1;
1804 else if (!PL_multiline)
1805 last = memEQ(strend - len, little, len)
1806 ? strend - len : Nullch;
1812 last = rninstr(s, strend, little, little + len);
1814 last = strend; /* matching `$' */
1818 DEBUG_r(PerlIO_printf(Perl_debug_log,
1819 "%sCan't trim the tail, match fails (should not happen)%s\n",
1820 PL_colors[4],PL_colors[5]));
1821 goto phooey; /* Should not happen! */
1823 dontbother = strend - last + prog->float_min_offset;
1825 if (minlen && (dontbother < minlen))
1826 dontbother = minlen - 1;
1827 strend -= dontbother; /* this one's always in bytes! */
1828 /* We don't know much -- general case. */
1831 if (regtry(prog, s))
1840 if (regtry(prog, s))
1842 } while (s++ < strend);
1850 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1852 if (PL_reg_eval_set) {
1853 /* Preserve the current value of $^R */
1854 if (oreplsv != GvSV(PL_replgv))
1855 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1856 restored, the value remains
1858 restore_pos(aTHX_ 0);
1861 /* make sure $`, $&, $', and $digit will work later */
1862 if ( !(flags & REXEC_NOT_FIRST) ) {
1863 if (RX_MATCH_COPIED(prog)) {
1864 Safefree(prog->subbeg);
1865 RX_MATCH_COPIED_off(prog);
1867 if (flags & REXEC_COPY_STR) {
1868 I32 i = PL_regeol - startpos + (stringarg - strbeg);
1870 s = savepvn(strbeg, i);
1873 RX_MATCH_COPIED_on(prog);
1876 prog->subbeg = strbeg;
1877 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
1884 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
1885 PL_colors[4],PL_colors[5]));
1886 if (PL_reg_eval_set)
1887 restore_pos(aTHX_ 0);
1892 - regtry - try match at specific point
1894 STATIC I32 /* 0 failure, 1 success */
1895 S_regtry(pTHX_ regexp *prog, char *startpos)
1903 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
1905 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
1908 PL_reg_eval_set = RS_init;
1910 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
1911 (IV)(PL_stack_sp - PL_stack_base));
1913 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
1914 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
1915 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
1917 /* Apparently this is not needed, judging by wantarray. */
1918 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
1919 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
1922 /* Make $_ available to executed code. */
1923 if (PL_reg_sv != DEFSV) {
1924 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
1929 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
1930 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
1931 /* prepare for quick setting of pos */
1932 sv_magic(PL_reg_sv, (SV*)0,
1933 PERL_MAGIC_regex_global, Nullch, 0);
1934 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
1938 PL_reg_oldpos = mg->mg_len;
1939 SAVEDESTRUCTOR_X(restore_pos, 0);
1941 if (!PL_reg_curpm) {
1942 Newz(22,PL_reg_curpm, 1, PMOP);
1945 SV* repointer = newSViv(0);
1946 /* so we know which PL_regex_padav element is PL_reg_curpm */
1947 SvFLAGS(repointer) |= SVf_BREAK;
1948 av_push(PL_regex_padav,repointer);
1949 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
1950 PL_regex_pad = AvARRAY(PL_regex_padav);
1954 PM_SETRE(PL_reg_curpm, prog);
1955 PL_reg_oldcurpm = PL_curpm;
1956 PL_curpm = PL_reg_curpm;
1957 if (RX_MATCH_COPIED(prog)) {
1958 /* Here is a serious problem: we cannot rewrite subbeg,
1959 since it may be needed if this match fails. Thus
1960 $` inside (?{}) could fail... */
1961 PL_reg_oldsaved = prog->subbeg;
1962 PL_reg_oldsavedlen = prog->sublen;
1963 RX_MATCH_COPIED_off(prog);
1966 PL_reg_oldsaved = Nullch;
1967 prog->subbeg = PL_bostr;
1968 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
1970 prog->startp[0] = startpos - PL_bostr;
1971 PL_reginput = startpos;
1972 PL_regstartp = prog->startp;
1973 PL_regendp = prog->endp;
1974 PL_reglastparen = &prog->lastparen;
1975 PL_reglastcloseparen = &prog->lastcloseparen;
1976 prog->lastparen = 0;
1978 DEBUG_r(PL_reg_starttry = startpos);
1979 if (PL_reg_start_tmpl <= prog->nparens) {
1980 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
1981 if(PL_reg_start_tmp)
1982 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1984 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1988 sv_setpvn(PERL_DEBUG_PAD(0), "", 0);
1989 sv_setpvn(PERL_DEBUG_PAD(1), "", 0);
1990 sv_setpvn(PERL_DEBUG_PAD(2), "", 0);
1993 /* XXXX What this code is doing here?!!! There should be no need
1994 to do this again and again, PL_reglastparen should take care of
1997 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
1998 * Actually, the code in regcppop() (which Ilya may be meaning by
1999 * PL_reglastparen), is not needed at all by the test suite
2000 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2001 * enough, for building DynaLoader, or otherwise this
2002 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2003 * will happen. Meanwhile, this code *is* needed for the
2004 * above-mentioned test suite tests to succeed. The common theme
2005 * on those tests seems to be returning null fields from matches.
2010 if (prog->nparens) {
2011 for (i = prog->nparens; i > *PL_reglastparen; i--) {
2018 if (regmatch(prog->program + 1)) {
2019 prog->endp[0] = PL_reginput - PL_bostr;
2022 REGCP_UNWIND(lastcp);
2026 #define RE_UNWIND_BRANCH 1
2027 #define RE_UNWIND_BRANCHJ 2
2031 typedef struct { /* XX: makes sense to enlarge it... */
2035 } re_unwind_generic_t;
2048 } re_unwind_branch_t;
2050 typedef union re_unwind_t {
2052 re_unwind_generic_t generic;
2053 re_unwind_branch_t branch;
2056 #define sayYES goto yes
2057 #define sayNO goto no
2058 #define sayYES_FINAL goto yes_final
2059 #define sayYES_LOUD goto yes_loud
2060 #define sayNO_FINAL goto no_final
2061 #define sayNO_SILENT goto do_no
2062 #define saySAME(x) if (x) goto yes; else goto no
2064 #define REPORT_CODE_OFF 24
2067 - regmatch - main matching routine
2069 * Conceptually the strategy is simple: check to see whether the current
2070 * node matches, call self recursively to see whether the rest matches,
2071 * and then act accordingly. In practice we make some effort to avoid
2072 * recursion, in particular by going through "ordinary" nodes (that don't
2073 * need to know whether the rest of the match failed) by a loop instead of
2076 /* [lwall] I've hoisted the register declarations to the outer block in order to
2077 * maybe save a little bit of pushing and popping on the stack. It also takes
2078 * advantage of machines that use a register save mask on subroutine entry.
2080 STATIC I32 /* 0 failure, 1 success */
2081 S_regmatch(pTHX_ regnode *prog)
2083 register regnode *scan; /* Current node. */
2084 regnode *next; /* Next node. */
2085 regnode *inner; /* Next node in internal branch. */
2086 register I32 nextchr; /* renamed nextchr - nextchar colides with
2087 function of same name */
2088 register I32 n; /* no or next */
2089 register I32 ln = 0; /* len or last */
2090 register char *s = Nullch; /* operand or save */
2091 register char *locinput = PL_reginput;
2092 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2093 int minmod = 0, sw = 0, logical = 0;
2096 I32 firstcp = PL_savestack_ix;
2098 register bool do_utf8 = PL_reg_match_utf8;
2100 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2101 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2102 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2109 /* Note that nextchr is a byte even in UTF */
2110 nextchr = UCHARAT(locinput);
2112 while (scan != NULL) {
2115 SV *prop = sv_newmortal();
2116 int docolor = *PL_colors[0];
2117 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2118 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2119 /* The part of the string before starttry has one color
2120 (pref0_len chars), between starttry and current
2121 position another one (pref_len - pref0_len chars),
2122 after the current position the third one.
2123 We assume that pref0_len <= pref_len, otherwise we
2124 decrease pref0_len. */
2125 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2126 ? (5 + taill) - l : locinput - PL_bostr;
2129 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2131 pref0_len = pref_len - (locinput - PL_reg_starttry);
2132 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2133 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2134 ? (5 + taill) - pref_len : PL_regeol - locinput);
2135 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2139 if (pref0_len > pref_len)
2140 pref0_len = pref_len;
2141 regprop(prop, scan);
2145 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2147 locinput - pref_len;
2148 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2149 char *s1 = do_utf8 ?
2150 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2151 pref_len - pref0_len, 60, 0) :
2152 locinput - pref_len + pref0_len;
2153 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2154 char *s2 = do_utf8 ?
2155 pv_uni_display(dsv2, (U8*)locinput,
2156 PL_regeol - locinput, 60, 0) :
2158 int len2 = do_utf8 ? strlen(s2) : l;
2159 PerlIO_printf(Perl_debug_log,
2160 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2161 (IV)(locinput - PL_bostr),
2168 (docolor ? "" : "> <"),
2172 15 - l - pref_len + 1,
2174 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2179 next = scan + NEXT_OFF(scan);
2185 if (locinput == PL_bostr || (PL_multiline &&
2186 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2188 /* regtill = regbol; */
2193 if (locinput == PL_bostr ||
2194 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2200 if (locinput == PL_bostr)
2204 if (locinput == PL_reg_ganch)
2214 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2219 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2221 if (PL_regeol - locinput > 1)
2225 if (PL_regeol != locinput)
2229 if (!nextchr && locinput >= PL_regeol)
2232 locinput += PL_utf8skip[nextchr];
2233 if (locinput > PL_regeol)
2235 nextchr = UCHARAT(locinput);
2238 nextchr = UCHARAT(++locinput);
2241 if (!nextchr && locinput >= PL_regeol)
2243 nextchr = UCHARAT(++locinput);
2246 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2249 locinput += PL_utf8skip[nextchr];
2250 if (locinput > PL_regeol)
2252 nextchr = UCHARAT(locinput);
2255 nextchr = UCHARAT(++locinput);
2260 if (do_utf8 != (UTF!=0)) {
2261 /* The target and the pattern have differing utf8ness. */
2267 /* The target is utf8, the pattern is not utf8. */
2271 if (NATIVE_TO_UNI(*(U8*)s) !=
2272 utf8_to_uvchr((U8*)l, &ulen))
2279 /* The target is not utf8, the pattern is utf8. */
2283 if (NATIVE_TO_UNI(*((U8*)l)) !=
2284 utf8_to_uvchr((U8*)s, &ulen))
2291 nextchr = UCHARAT(locinput);
2294 /* The target and the pattern have the same utf8ness. */
2295 /* Inline the first character, for speed. */
2296 if (UCHARAT(s) != nextchr)
2298 if (PL_regeol - locinput < ln)
2300 if (ln > 1 && memNE(s, locinput, ln))
2303 nextchr = UCHARAT(locinput);
2306 PL_reg_flags |= RF_tainted;
2315 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
2317 if (do_utf8 != (UTF!=0)) {
2318 /* The target and the pattern have differing utf8ness. */
2319 STRLEN ulen1, ulen2;
2323 /* The target is utf8, the pattern is not utf8. */
2328 cs = to_uni_fold(NATIVE_TO_UNI(*(U8*)s),
2330 cl = utf8_to_uvchr((U8*)l, &ulen2);
2333 cl = to_uni_fold(cl, (U8*)l, &ulen2);
2334 if (ulen1 != ulen2 || cs != cl)
2342 /* The target is not utf8, the pattern is utf8. */
2347 cs = utf8_to_uvchr((U8*)s, &ulen1);
2349 cl = to_uni_fold(NATIVE_TO_UNI(*(U8*)l),
2353 cs = to_uni_fold(cs, (U8*)s, &ulen1);
2354 if (ulen1 != ulen2 || cs != cl)
2362 nextchr = UCHARAT(locinput);
2366 if (do_utf8 && UTF) {
2367 /* Both the target and the pattern are utf8. */
2373 if (UTF8SKIP(s) != UTF8SKIP(l) ||
2374 memNE(s, (char*)l, UTF8SKIP(s))) {
2375 to_utf8_fold((U8*)l, tmpbuf, &ulen);
2376 if (UTF8SKIP(s) != ulen ||
2377 memNE(s, (char*)tmpbuf, ulen))
2384 nextchr = UCHARAT(locinput);
2389 /* Neither the target and the pattern are utf8. */
2391 /* Inline the first character, for speed. */
2392 if (UCHARAT(s) != nextchr &&
2393 UCHARAT(s) != ((OP(scan) == EXACTF)
2394 ? PL_fold : PL_fold_locale)[nextchr])
2396 if (PL_regeol - locinput < ln)
2398 if (ln > 1 && (OP(scan) == EXACTF
2399 ? ibcmp(s, locinput, ln)
2400 : ibcmp_locale(s, locinput, ln)))
2403 nextchr = UCHARAT(locinput);
2407 if (!reginclass(scan, (U8*)locinput, do_utf8))
2409 if (locinput >= PL_regeol)
2411 locinput += PL_utf8skip[nextchr];
2412 nextchr = UCHARAT(locinput);
2416 nextchr = UCHARAT(locinput);
2417 if (!reginclass(scan, (U8*)locinput, do_utf8))
2419 if (!nextchr && locinput >= PL_regeol)
2421 nextchr = UCHARAT(++locinput);
2425 PL_reg_flags |= RF_tainted;
2431 LOAD_UTF8_CHARCLASS(alnum,"a");
2432 if (!(OP(scan) == ALNUM
2433 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2434 : isALNUM_LC_utf8((U8*)locinput)))
2438 locinput += PL_utf8skip[nextchr];
2439 nextchr = UCHARAT(locinput);
2442 if (!(OP(scan) == ALNUM
2443 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2445 nextchr = UCHARAT(++locinput);
2448 PL_reg_flags |= RF_tainted;
2451 if (!nextchr && locinput >= PL_regeol)
2454 LOAD_UTF8_CHARCLASS(alnum,"a");
2455 if (OP(scan) == NALNUM
2456 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2457 : isALNUM_LC_utf8((U8*)locinput))
2461 locinput += PL_utf8skip[nextchr];
2462 nextchr = UCHARAT(locinput);
2465 if (OP(scan) == NALNUM
2466 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2468 nextchr = UCHARAT(++locinput);
2472 PL_reg_flags |= RF_tainted;
2476 /* was last char in word? */
2478 if (locinput == PL_bostr)
2481 U8 *r = reghop((U8*)locinput, -1);
2483 ln = utf8n_to_uvchr(r, s - (char*)r, 0, 0);
2485 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2486 ln = isALNUM_uni(ln);
2487 LOAD_UTF8_CHARCLASS(alnum,"a");
2488 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2491 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2492 n = isALNUM_LC_utf8((U8*)locinput);
2496 ln = (locinput != PL_bostr) ?
2497 UCHARAT(locinput - 1) : '\n';
2498 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2500 n = isALNUM(nextchr);
2503 ln = isALNUM_LC(ln);
2504 n = isALNUM_LC(nextchr);
2507 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2508 OP(scan) == BOUNDL))
2512 PL_reg_flags |= RF_tainted;
2518 if (UTF8_IS_CONTINUED(nextchr)) {
2519 LOAD_UTF8_CHARCLASS(space," ");
2520 if (!(OP(scan) == SPACE
2521 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2522 : isSPACE_LC_utf8((U8*)locinput)))
2526 locinput += PL_utf8skip[nextchr];
2527 nextchr = UCHARAT(locinput);
2530 if (!(OP(scan) == SPACE
2531 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2533 nextchr = UCHARAT(++locinput);
2536 if (!(OP(scan) == SPACE
2537 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2539 nextchr = UCHARAT(++locinput);
2543 PL_reg_flags |= RF_tainted;
2546 if (!nextchr && locinput >= PL_regeol)
2549 LOAD_UTF8_CHARCLASS(space," ");
2550 if (OP(scan) == NSPACE
2551 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2552 : isSPACE_LC_utf8((U8*)locinput))
2556 locinput += PL_utf8skip[nextchr];
2557 nextchr = UCHARAT(locinput);
2560 if (OP(scan) == NSPACE
2561 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2563 nextchr = UCHARAT(++locinput);
2566 PL_reg_flags |= RF_tainted;
2572 LOAD_UTF8_CHARCLASS(digit,"0");
2573 if (!(OP(scan) == DIGIT
2574 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2575 : isDIGIT_LC_utf8((U8*)locinput)))
2579 locinput += PL_utf8skip[nextchr];
2580 nextchr = UCHARAT(locinput);
2583 if (!(OP(scan) == DIGIT
2584 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2586 nextchr = UCHARAT(++locinput);
2589 PL_reg_flags |= RF_tainted;
2592 if (!nextchr && locinput >= PL_regeol)
2595 LOAD_UTF8_CHARCLASS(digit,"0");
2596 if (OP(scan) == NDIGIT
2597 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2598 : isDIGIT_LC_utf8((U8*)locinput))
2602 locinput += PL_utf8skip[nextchr];
2603 nextchr = UCHARAT(locinput);
2606 if (OP(scan) == NDIGIT
2607 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2609 nextchr = UCHARAT(++locinput);
2612 if (locinput >= PL_regeol)
2615 LOAD_UTF8_CHARCLASS(mark,"~");
2616 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2618 locinput += PL_utf8skip[nextchr];
2619 while (locinput < PL_regeol &&
2620 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2621 locinput += UTF8SKIP(locinput);
2622 if (locinput > PL_regeol)
2627 nextchr = UCHARAT(locinput);
2630 PL_reg_flags |= RF_tainted;
2634 n = ARG(scan); /* which paren pair */
2635 ln = PL_regstartp[n];
2636 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2637 if (*PL_reglastparen < n || ln == -1)
2638 sayNO; /* Do not match unless seen CLOSEn. */
2639 if (ln == PL_regendp[n])
2643 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2645 char *e = PL_bostr + PL_regendp[n];
2647 * Note that we can't do the "other character" lookup trick as
2648 * in the 8-bit case (no pun intended) because in Unicode we
2649 * have to map both upper and title case to lower case.
2651 if (OP(scan) == REFF) {
2652 STRLEN ulen1, ulen2;
2653 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2654 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2658 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2659 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2660 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2667 nextchr = UCHARAT(locinput);
2671 /* Inline the first character, for speed. */
2672 if (UCHARAT(s) != nextchr &&
2674 (UCHARAT(s) != ((OP(scan) == REFF
2675 ? PL_fold : PL_fold_locale)[nextchr]))))
2677 ln = PL_regendp[n] - ln;
2678 if (locinput + ln > PL_regeol)
2680 if (ln > 1 && (OP(scan) == REF
2681 ? memNE(s, locinput, ln)
2683 ? ibcmp(s, locinput, ln)
2684 : ibcmp_locale(s, locinput, ln))))
2687 nextchr = UCHARAT(locinput);
2698 OP_4tree *oop = PL_op;
2699 COP *ocurcop = PL_curcop;
2700 SV **ocurpad = PL_curpad;
2704 PL_op = (OP_4tree*)PL_regdata->data[n];
2705 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2706 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2707 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2711 CALLRUNOPS(aTHX); /* Scalar context. */
2714 ret = Nullsv; /* protect against empty (?{}) blocks. */
2722 PL_curpad = ocurpad;
2723 PL_curcop = ocurcop;
2725 if (logical == 2) { /* Postponed subexpression. */
2727 MAGIC *mg = Null(MAGIC*);
2729 CHECKPOINT cp, lastcp;
2731 if(SvROK(ret) || SvRMAGICAL(ret)) {
2732 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2735 mg = mg_find(sv, PERL_MAGIC_qr);
2738 re = (regexp *)mg->mg_obj;
2739 (void)ReREFCNT_inc(re);
2743 char *t = SvPV(ret, len);
2745 char *oprecomp = PL_regprecomp;
2746 I32 osize = PL_regsize;
2747 I32 onpar = PL_regnpar;
2750 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2752 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2753 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2755 PL_regprecomp = oprecomp;
2760 PerlIO_printf(Perl_debug_log,
2761 "Entering embedded `%s%.60s%s%s'\n",
2765 (strlen(re->precomp) > 60 ? "..." : ""))
2768 state.prev = PL_reg_call_cc;
2769 state.cc = PL_regcc;
2770 state.re = PL_reg_re;
2774 cp = regcppush(0); /* Save *all* the positions. */
2777 state.ss = PL_savestack_ix;
2778 *PL_reglastparen = 0;
2779 *PL_reglastcloseparen = 0;
2780 PL_reg_call_cc = &state;
2781 PL_reginput = locinput;
2783 /* XXXX This is too dramatic a measure... */
2786 if (regmatch(re->program + 1)) {
2787 /* Even though we succeeded, we need to restore
2788 global variables, since we may be wrapped inside
2789 SUSPEND, thus the match may be not finished yet. */
2791 /* XXXX Do this only if SUSPENDed? */
2792 PL_reg_call_cc = state.prev;
2793 PL_regcc = state.cc;
2794 PL_reg_re = state.re;
2795 cache_re(PL_reg_re);
2797 /* XXXX This is too dramatic a measure... */
2800 /* These are needed even if not SUSPEND. */
2806 REGCP_UNWIND(lastcp);
2808 PL_reg_call_cc = state.prev;
2809 PL_regcc = state.cc;
2810 PL_reg_re = state.re;
2811 cache_re(PL_reg_re);
2813 /* XXXX This is too dramatic a measure... */
2823 sv_setsv(save_scalar(PL_replgv), ret);
2827 n = ARG(scan); /* which paren pair */
2828 PL_reg_start_tmp[n] = locinput;
2833 n = ARG(scan); /* which paren pair */
2834 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2835 PL_regendp[n] = locinput - PL_bostr;
2836 if (n > *PL_reglastparen)
2837 *PL_reglastparen = n;
2838 *PL_reglastcloseparen = n;
2841 n = ARG(scan); /* which paren pair */
2842 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2845 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2847 next = NEXTOPER(NEXTOPER(scan));
2849 next = scan + ARG(scan);
2850 if (OP(next) == IFTHEN) /* Fake one. */
2851 next = NEXTOPER(NEXTOPER(next));
2855 logical = scan->flags;
2857 /*******************************************************************
2858 PL_regcc contains infoblock about the innermost (...)* loop, and
2859 a pointer to the next outer infoblock.
2861 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2863 1) After matching X, regnode for CURLYX is processed;
2865 2) This regnode creates infoblock on the stack, and calls
2866 regmatch() recursively with the starting point at WHILEM node;
2868 3) Each hit of WHILEM node tries to match A and Z (in the order
2869 depending on the current iteration, min/max of {min,max} and
2870 greediness). The information about where are nodes for "A"
2871 and "Z" is read from the infoblock, as is info on how many times "A"
2872 was already matched, and greediness.
2874 4) After A matches, the same WHILEM node is hit again.
2876 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2877 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2878 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2879 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2880 of the external loop.
2882 Currently present infoblocks form a tree with a stem formed by PL_curcc
2883 and whatever it mentions via ->next, and additional attached trees
2884 corresponding to temporarily unset infoblocks as in "5" above.
2886 In the following picture infoblocks for outer loop of
2887 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2888 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2889 infoblocks are drawn below the "reset" infoblock.
2891 In fact in the picture below we do not show failed matches for Z and T
2892 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2893 more obvious *why* one needs to *temporary* unset infoblocks.]
2895 Matched REx position InfoBlocks Comment
2899 Y A)*?Z)*?T x <- O <- I
2900 YA )*?Z)*?T x <- O <- I
2901 YA A)*?Z)*?T x <- O <- I
2902 YAA )*?Z)*?T x <- O <- I
2903 YAA Z)*?T x <- O # Temporary unset I
2906 YAAZ Y(A)*?Z)*?T x <- O
2909 YAAZY (A)*?Z)*?T x <- O
2912 YAAZY A)*?Z)*?T x <- O <- I
2915 YAAZYA )*?Z)*?T x <- O <- I
2918 YAAZYA Z)*?T x <- O # Temporary unset I
2924 YAAZYAZ T x # Temporary unset O
2931 *******************************************************************/
2934 CHECKPOINT cp = PL_savestack_ix;
2935 /* No need to save/restore up to this paren */
2936 I32 parenfloor = scan->flags;
2938 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2940 cc.oldcc = PL_regcc;
2942 /* XXXX Probably it is better to teach regpush to support
2943 parenfloor > PL_regsize... */
2944 if (parenfloor > *PL_reglastparen)
2945 parenfloor = *PL_reglastparen; /* Pessimization... */
2946 cc.parenfloor = parenfloor;
2948 cc.min = ARG1(scan);
2949 cc.max = ARG2(scan);
2950 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2954 PL_reginput = locinput;
2955 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2957 PL_regcc = cc.oldcc;
2963 * This is really hard to understand, because after we match
2964 * what we're trying to match, we must make sure the rest of
2965 * the REx is going to match for sure, and to do that we have
2966 * to go back UP the parse tree by recursing ever deeper. And
2967 * if it fails, we have to reset our parent's current state
2968 * that we can try again after backing off.
2971 CHECKPOINT cp, lastcp;
2972 CURCUR* cc = PL_regcc;
2973 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2975 n = cc->cur + 1; /* how many we know we matched */
2976 PL_reginput = locinput;
2979 PerlIO_printf(Perl_debug_log,
2980 "%*s %ld out of %ld..%ld cc=%lx\n",
2981 REPORT_CODE_OFF+PL_regindent*2, "",
2982 (long)n, (long)cc->min,
2983 (long)cc->max, (long)cc)
2986 /* If degenerate scan matches "", assume scan done. */
2988 if (locinput == cc->lastloc && n >= cc->min) {
2989 PL_regcc = cc->oldcc;
2993 PerlIO_printf(Perl_debug_log,
2994 "%*s empty match detected, try continuation...\n",
2995 REPORT_CODE_OFF+PL_regindent*2, "")
2997 if (regmatch(cc->next))
3005 /* First just match a string of min scans. */
3009 cc->lastloc = locinput;
3010 if (regmatch(cc->scan))
3013 cc->lastloc = lastloc;
3018 /* Check whether we already were at this position.
3019 Postpone detection until we know the match is not
3020 *that* much linear. */
3021 if (!PL_reg_maxiter) {
3022 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3023 PL_reg_leftiter = PL_reg_maxiter;
3025 if (PL_reg_leftiter-- == 0) {
3026 I32 size = (PL_reg_maxiter + 7)/8;
3027 if (PL_reg_poscache) {
3028 if (PL_reg_poscache_size < size) {
3029 Renew(PL_reg_poscache, size, char);
3030 PL_reg_poscache_size = size;
3032 Zero(PL_reg_poscache, size, char);
3035 PL_reg_poscache_size = size;
3036 Newz(29, PL_reg_poscache, size, char);
3039 PerlIO_printf(Perl_debug_log,
3040 "%sDetected a super-linear match, switching on caching%s...\n",
3041 PL_colors[4], PL_colors[5])
3044 if (PL_reg_leftiter < 0) {
3045 I32 o = locinput - PL_bostr, b;
3047 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3050 if (PL_reg_poscache[o] & (1<<b)) {
3052 PerlIO_printf(Perl_debug_log,
3053 "%*s already tried at this position...\n",
3054 REPORT_CODE_OFF+PL_regindent*2, "")
3058 PL_reg_poscache[o] |= (1<<b);
3062 /* Prefer next over scan for minimal matching. */
3065 PL_regcc = cc->oldcc;
3068 cp = regcppush(cc->parenfloor);
3070 if (regmatch(cc->next)) {
3072 sayYES; /* All done. */
3074 REGCP_UNWIND(lastcp);
3080 if (n >= cc->max) { /* Maximum greed exceeded? */
3081 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3082 && !(PL_reg_flags & RF_warned)) {
3083 PL_reg_flags |= RF_warned;
3084 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3085 "Complex regular subexpression recursion",
3092 PerlIO_printf(Perl_debug_log,
3093 "%*s trying longer...\n",
3094 REPORT_CODE_OFF+PL_regindent*2, "")
3096 /* Try scanning more and see if it helps. */
3097 PL_reginput = locinput;
3099 cc->lastloc = locinput;
3100 cp = regcppush(cc->parenfloor);
3102 if (regmatch(cc->scan)) {
3106 REGCP_UNWIND(lastcp);
3109 cc->lastloc = lastloc;
3113 /* Prefer scan over next for maximal matching. */
3115 if (n < cc->max) { /* More greed allowed? */
3116 cp = regcppush(cc->parenfloor);
3118 cc->lastloc = locinput;
3120 if (regmatch(cc->scan)) {
3124 REGCP_UNWIND(lastcp);
3125 regcppop(); /* Restore some previous $<digit>s? */
3126 PL_reginput = locinput;
3128 PerlIO_printf(Perl_debug_log,
3129 "%*s failed, try continuation...\n",
3130 REPORT_CODE_OFF+PL_regindent*2, "")
3133 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3134 && !(PL_reg_flags & RF_warned)) {
3135 PL_reg_flags |= RF_warned;
3136 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3137 "Complex regular subexpression recursion",
3141 /* Failed deeper matches of scan, so see if this one works. */
3142 PL_regcc = cc->oldcc;
3145 if (regmatch(cc->next))
3151 cc->lastloc = lastloc;
3156 next = scan + ARG(scan);
3159 inner = NEXTOPER(NEXTOPER(scan));
3162 inner = NEXTOPER(scan);
3166 if (OP(next) != c1) /* No choice. */
3167 next = inner; /* Avoid recursion. */
3169 I32 lastparen = *PL_reglastparen;
3171 re_unwind_branch_t *uw;
3173 /* Put unwinding data on stack */
3174 unwind1 = SSNEWt(1,re_unwind_branch_t);
3175 uw = SSPTRt(unwind1,re_unwind_branch_t);
3178 uw->type = ((c1 == BRANCH)
3180 : RE_UNWIND_BRANCHJ);
3181 uw->lastparen = lastparen;
3183 uw->locinput = locinput;
3184 uw->nextchr = nextchr;
3186 uw->regindent = ++PL_regindent;
3189 REGCP_SET(uw->lastcp);
3191 /* Now go into the first branch */
3204 /* We suppose that the next guy does not need
3205 backtracking: in particular, it is of constant length,
3206 and has no parenths to influence future backrefs. */
3207 ln = ARG1(scan); /* min to match */
3208 n = ARG2(scan); /* max to match */
3209 paren = scan->flags;
3211 if (paren > PL_regsize)
3213 if (paren > *PL_reglastparen)
3214 *PL_reglastparen = paren;
3216 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3218 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3219 PL_reginput = locinput;
3222 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3224 /* if we matched something zero-length we don't need to
3225 backtrack - capturing parens are already defined, so
3226 the caveat in the maximal case doesn't apply
3228 XXXX if ln == 0, we can redo this check first time
3229 through the following loop
3232 n = ln; /* don't backtrack */
3233 locinput = PL_reginput;
3234 if (HAS_TEXT(next) || JUMPABLE(next)) {
3235 regnode *text_node = next;
3237 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3239 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3241 if (PL_regkind[(U8)OP(text_node)] == REF) {
3243 n = ARG(text_node); /* which paren pair */
3244 ln = PL_regstartp[n];
3245 /* assume yes if we haven't seen CLOSEn */
3247 *PL_reglastparen < n ||
3254 c1 = *(PL_bostr + ln);
3256 else { c1 = (U8)*STRING(text_node); }
3257 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3259 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3260 c2 = PL_fold_locale[c1];
3269 /* This may be improved if l == 0. */
3270 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3271 /* If it could work, try it. */
3273 UCHARAT(PL_reginput) == c1 ||
3274 UCHARAT(PL_reginput) == c2)
3278 PL_regstartp[paren] =
3279 HOPc(PL_reginput, -l) - PL_bostr;
3280 PL_regendp[paren] = PL_reginput - PL_bostr;
3283 PL_regendp[paren] = -1;
3287 REGCP_UNWIND(lastcp);
3289 /* Couldn't or didn't -- move forward. */
3290 PL_reginput = locinput;
3291 if (regrepeat_hard(scan, 1, &l)) {
3293 locinput = PL_reginput;
3300 n = regrepeat_hard(scan, n, &l);
3301 /* if we matched something zero-length we don't need to
3302 backtrack, unless the minimum count is zero and we
3303 are capturing the result - in that case the capture
3304 being defined or not may affect later execution
3306 if (n != 0 && l == 0 && !(paren && ln == 0))
3307 ln = n; /* don't backtrack */
3308 locinput = PL_reginput;
3310 PerlIO_printf(Perl_debug_log,
3311 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3312 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3316 if (HAS_TEXT(next) || JUMPABLE(next)) {
3317 regnode *text_node = next;
3319 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3321 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3323 if (PL_regkind[(U8)OP(text_node)] == REF) {
3325 n = ARG(text_node); /* which paren pair */
3326 ln = PL_regstartp[n];
3327 /* assume yes if we haven't seen CLOSEn */
3329 *PL_reglastparen < n ||
3336 c1 = *(PL_bostr + ln);
3338 else { c1 = (U8)*STRING(text_node); }
3340 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3342 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3343 c2 = PL_fold_locale[c1];
3354 /* If it could work, try it. */
3356 UCHARAT(PL_reginput) == c1 ||
3357 UCHARAT(PL_reginput) == c2)
3360 PerlIO_printf(Perl_debug_log,
3361 "%*s trying tail with n=%"IVdf"...\n",
3362 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3366 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3367 PL_regendp[paren] = PL_reginput - PL_bostr;
3370 PL_regendp[paren] = -1;
3374 REGCP_UNWIND(lastcp);
3376 /* Couldn't or didn't -- back up. */
3378 locinput = HOPc(locinput, -l);
3379 PL_reginput = locinput;
3386 paren = scan->flags; /* Which paren to set */
3387 if (paren > PL_regsize)
3389 if (paren > *PL_reglastparen)
3390 *PL_reglastparen = paren;
3391 ln = ARG1(scan); /* min to match */
3392 n = ARG2(scan); /* max to match */
3393 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3397 ln = ARG1(scan); /* min to match */
3398 n = ARG2(scan); /* max to match */
3399 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3404 scan = NEXTOPER(scan);
3410 scan = NEXTOPER(scan);
3414 * Lookahead to avoid useless match attempts
3415 * when we know what character comes next.
3419 * Used to only do .*x and .*?x, but now it allows
3420 * for )'s, ('s and (?{ ... })'s to be in the way
3421 * of the quantifier and the EXACT-like node. -- japhy
3424 if (HAS_TEXT(next) || JUMPABLE(next)) {
3426 regnode *text_node = next;
3428 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3430 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3432 if (PL_regkind[(U8)OP(text_node)] == REF) {
3434 n = ARG(text_node); /* which paren pair */
3435 ln = PL_regstartp[n];
3436 /* assume yes if we haven't seen CLOSEn */
3438 *PL_reglastparen < n ||
3443 goto assume_ok_easy;
3445 s = (U8*)PL_bostr + ln;
3447 else { s = (U8*)STRING(text_node); }
3451 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3453 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3454 c2 = PL_fold_locale[c1];
3457 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3458 STRLEN ulen1, ulen2;
3459 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3460 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3462 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3463 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3465 c1 = utf8_to_uvuni(tmpbuf1, 0);
3466 c2 = utf8_to_uvuni(tmpbuf2, 0);
3469 c2 = c1 = utf8_to_uvchr(s, NULL);
3477 PL_reginput = locinput;
3481 if (ln && regrepeat(scan, ln) < ln)
3483 locinput = PL_reginput;
3486 char *e; /* Should not check after this */
3487 char *old = locinput;
3489 if (n == REG_INFTY) {
3492 while (UTF8_IS_CONTINUATION(*(U8*)e))
3498 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3502 e = locinput + n - ln;
3508 /* Find place 'next' could work */
3511 while (locinput <= e &&
3512 UCHARAT(locinput) != c1)
3515 while (locinput <= e
3516 && UCHARAT(locinput) != c1
3517 && UCHARAT(locinput) != c2)
3520 count = locinput - old;
3527 utf8_to_uvchr((U8*)locinput, &len) != c1;
3532 for (count = 0; locinput <= e; count++) {
3533 UV c = utf8_to_uvchr((U8*)locinput, &len);
3534 if (c == c1 || c == c2)
3542 /* PL_reginput == old now */
3543 if (locinput != old) {
3544 ln = 1; /* Did some */
3545 if (regrepeat(scan, count) < count)
3548 /* PL_reginput == locinput now */
3549 TRYPAREN(paren, ln, locinput);
3550 PL_reginput = locinput; /* Could be reset... */
3551 REGCP_UNWIND(lastcp);
3552 /* Couldn't or didn't -- move forward. */
3555 locinput += UTF8SKIP(locinput);
3561 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3565 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3567 c = UCHARAT(PL_reginput);
3568 /* If it could work, try it. */
3569 if (c == c1 || c == c2)
3571 TRYPAREN(paren, n, PL_reginput);
3572 REGCP_UNWIND(lastcp);
3575 /* If it could work, try it. */
3576 else if (c1 == -1000)
3578 TRYPAREN(paren, n, PL_reginput);
3579 REGCP_UNWIND(lastcp);
3581 /* Couldn't or didn't -- move forward. */
3582 PL_reginput = locinput;
3583 if (regrepeat(scan, 1)) {
3585 locinput = PL_reginput;
3593 n = regrepeat(scan, n);
3594 locinput = PL_reginput;
3595 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3596 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3597 ln = n; /* why back off? */
3598 /* ...because $ and \Z can match before *and* after
3599 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3600 We should back off by one in this case. */
3601 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3610 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3612 c = UCHARAT(PL_reginput);
3614 /* If it could work, try it. */
3615 if (c1 == -1000 || c == c1 || c == c2)
3617 TRYPAREN(paren, n, PL_reginput);
3618 REGCP_UNWIND(lastcp);
3620 /* Couldn't or didn't -- back up. */
3622 PL_reginput = locinput = HOPc(locinput, -1);
3630 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3632 c = UCHARAT(PL_reginput);
3634 /* If it could work, try it. */
3635 if (c1 == -1000 || c == c1 || c == c2)
3637 TRYPAREN(paren, n, PL_reginput);
3638 REGCP_UNWIND(lastcp);
3640 /* Couldn't or didn't -- back up. */
3642 PL_reginput = locinput = HOPc(locinput, -1);
3649 if (PL_reg_call_cc) {
3650 re_cc_state *cur_call_cc = PL_reg_call_cc;
3651 CURCUR *cctmp = PL_regcc;
3652 regexp *re = PL_reg_re;
3653 CHECKPOINT cp, lastcp;
3655 cp = regcppush(0); /* Save *all* the positions. */
3657 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3659 PL_reginput = locinput; /* Make position available to
3661 cache_re(PL_reg_call_cc->re);
3662 PL_regcc = PL_reg_call_cc->cc;
3663 PL_reg_call_cc = PL_reg_call_cc->prev;
3664 if (regmatch(cur_call_cc->node)) {
3665 PL_reg_call_cc = cur_call_cc;
3669 REGCP_UNWIND(lastcp);
3671 PL_reg_call_cc = cur_call_cc;
3677 PerlIO_printf(Perl_debug_log,
3678 "%*s continuation failed...\n",
3679 REPORT_CODE_OFF+PL_regindent*2, "")
3683 if (locinput < PL_regtill) {
3684 DEBUG_r(PerlIO_printf(Perl_debug_log,
3685 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3687 (long)(locinput - PL_reg_starttry),
3688 (long)(PL_regtill - PL_reg_starttry),
3690 sayNO_FINAL; /* Cannot match: too short. */
3692 PL_reginput = locinput; /* put where regtry can find it */
3693 sayYES_FINAL; /* Success! */
3695 PL_reginput = locinput; /* put where regtry can find it */
3696 sayYES_LOUD; /* Success! */
3699 PL_reginput = locinput;
3704 s = HOPBACKc(locinput, scan->flags);
3710 PL_reginput = locinput;
3715 s = HOPBACKc(locinput, scan->flags);
3721 PL_reginput = locinput;
3724 inner = NEXTOPER(NEXTOPER(scan));
3725 if (regmatch(inner) != n) {
3740 if (OP(scan) == SUSPEND) {
3741 locinput = PL_reginput;
3742 nextchr = UCHARAT(locinput);
3747 next = scan + ARG(scan);
3752 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3753 PTR2UV(scan), OP(scan));
3754 Perl_croak(aTHX_ "regexp memory corruption");
3761 * We get here only if there's trouble -- normally "case END" is
3762 * the terminating point.
3764 Perl_croak(aTHX_ "corrupted regexp pointers");
3770 PerlIO_printf(Perl_debug_log,
3771 "%*s %scould match...%s\n",
3772 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3776 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3777 PL_colors[4],PL_colors[5]));
3783 #if 0 /* Breaks $^R */
3791 PerlIO_printf(Perl_debug_log,
3792 "%*s %sfailed...%s\n",
3793 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3799 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3802 case RE_UNWIND_BRANCH:
3803 case RE_UNWIND_BRANCHJ:
3805 re_unwind_branch_t *uwb = &(uw->branch);
3806 I32 lastparen = uwb->lastparen;
3808 REGCP_UNWIND(uwb->lastcp);
3809 for (n = *PL_reglastparen; n > lastparen; n--)
3811 *PL_reglastparen = n;
3812 scan = next = uwb->next;
3814 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3815 ? BRANCH : BRANCHJ) ) { /* Failure */
3822 /* Have more choice yet. Reuse the same uwb. */
3824 if ((n = (uwb->type == RE_UNWIND_BRANCH
3825 ? NEXT_OFF(next) : ARG(next))))
3828 next = NULL; /* XXXX Needn't unwinding in this case... */
3830 next = NEXTOPER(scan);
3831 if (uwb->type == RE_UNWIND_BRANCHJ)
3832 next = NEXTOPER(next);
3833 locinput = uwb->locinput;
3834 nextchr = uwb->nextchr;
3836 PL_regindent = uwb->regindent;
3843 Perl_croak(aTHX_ "regexp unwind memory corruption");
3854 - regrepeat - repeatedly match something simple, report how many
3857 * [This routine now assumes that it will only match on things of length 1.
3858 * That was true before, but now we assume scan - reginput is the count,
3859 * rather than incrementing count on every character. [Er, except utf8.]]
3862 S_regrepeat(pTHX_ regnode *p, I32 max)
3864 register char *scan;
3866 register char *loceol = PL_regeol;
3867 register I32 hardcount = 0;
3868 register bool do_utf8 = PL_reg_match_utf8;
3871 if (max != REG_INFTY && max < loceol - scan)
3872 loceol = scan + max;
3877 while (scan < loceol && hardcount < max && *scan != '\n') {
3878 scan += UTF8SKIP(scan);
3882 while (scan < loceol && *scan != '\n')
3892 case EXACT: /* length of string is 1 */
3894 while (scan < loceol && UCHARAT(scan) == c)
3897 case EXACTF: /* length of string is 1 */
3899 while (scan < loceol &&
3900 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3903 case EXACTFL: /* length of string is 1 */
3904 PL_reg_flags |= RF_tainted;
3906 while (scan < loceol &&
3907 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3913 while (hardcount < max && scan < loceol &&
3914 reginclass(p, (U8*)scan, do_utf8)) {
3915 scan += UTF8SKIP(scan);
3919 while (scan < loceol && reginclass(p, (U8*)scan, do_utf8))
3926 LOAD_UTF8_CHARCLASS(alnum,"a");
3927 while (hardcount < max && scan < loceol &&
3928 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3929 scan += UTF8SKIP(scan);
3933 while (scan < loceol && isALNUM(*scan))
3938 PL_reg_flags |= RF_tainted;
3941 while (hardcount < max && scan < loceol &&
3942 isALNUM_LC_utf8((U8*)scan)) {
3943 scan += UTF8SKIP(scan);
3947 while (scan < loceol && isALNUM_LC(*scan))
3954 LOAD_UTF8_CHARCLASS(alnum,"a");
3955 while (hardcount < max && scan < loceol &&
3956 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3957 scan += UTF8SKIP(scan);
3961 while (scan < loceol && !isALNUM(*scan))
3966 PL_reg_flags |= RF_tainted;
3969 while (hardcount < max && scan < loceol &&
3970 !isALNUM_LC_utf8((U8*)scan)) {
3971 scan += UTF8SKIP(scan);
3975 while (scan < loceol && !isALNUM_LC(*scan))
3982 LOAD_UTF8_CHARCLASS(space," ");
3983 while (hardcount < max && scan < loceol &&
3985 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3986 scan += UTF8SKIP(scan);
3990 while (scan < loceol && isSPACE(*scan))
3995 PL_reg_flags |= RF_tainted;
3998 while (hardcount < max && scan < loceol &&
3999 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4000 scan += UTF8SKIP(scan);
4004 while (scan < loceol && isSPACE_LC(*scan))
4011 LOAD_UTF8_CHARCLASS(space," ");
4012 while (hardcount < max && scan < loceol &&
4014 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4015 scan += UTF8SKIP(scan);
4019 while (scan < loceol && !isSPACE(*scan))
4024 PL_reg_flags |= RF_tainted;
4027 while (hardcount < max && scan < loceol &&
4028 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4029 scan += UTF8SKIP(scan);
4033 while (scan < loceol && !isSPACE_LC(*scan))
4040 LOAD_UTF8_CHARCLASS(digit,"0");
4041 while (hardcount < max && scan < loceol &&
4042 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4043 scan += UTF8SKIP(scan);
4047 while (scan < loceol && isDIGIT(*scan))
4054 LOAD_UTF8_CHARCLASS(digit,"0");
4055 while (hardcount < max && scan < loceol &&
4056 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4057 scan += UTF8SKIP(scan);
4061 while (scan < loceol && !isDIGIT(*scan))
4065 default: /* Called on something of 0 width. */
4066 break; /* So match right here or not at all. */
4072 c = scan - PL_reginput;
4077 SV *prop = sv_newmortal();
4080 PerlIO_printf(Perl_debug_log,
4081 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4082 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4089 - regrepeat_hard - repeatedly match something, report total lenth and length
4091 * The repeater is supposed to have constant length.
4095 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4097 register char *scan = Nullch;
4098 register char *start;
4099 register char *loceol = PL_regeol;
4101 I32 count = 0, res = 1;
4106 start = PL_reginput;
4107 if (PL_reg_match_utf8) {
4108 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4111 while (start < PL_reginput) {
4113 start += UTF8SKIP(start);
4124 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4126 *lp = l = PL_reginput - start;
4127 if (max != REG_INFTY && l*max < loceol - scan)
4128 loceol = scan + l*max;
4141 - regclass_swash - prepare the utf8 swash
4145 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** initsvp)
4150 if (PL_regdata && PL_regdata->count) {
4153 if (PL_regdata->what[n] == 's') {
4154 SV *rv = (SV*)PL_regdata->data[n];
4155 AV *av = (AV*)SvRV((SV*)rv);
4158 si = *av_fetch(av, 0, FALSE);
4159 a = av_fetch(av, 1, FALSE);
4163 else if (si && doinit) {
4164 sw = swash_init("utf8", "", si, 1, 0);
4165 (void)av_store(av, 1, sw);
4177 - reginclass - determine if a character falls into a character class
4181 S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8)
4183 char flags = ANYOF_FLAGS(n);
4188 c = do_utf8 ? utf8_to_uvchr(p, &len) : *p;
4190 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4191 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4192 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4195 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4198 SV *sw = regclass_swash(n, TRUE, 0);
4201 if (swash_fetch(sw, p, do_utf8))
4203 else if (flags & ANYOF_FOLD) {
4205 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
4207 to_utf8_fold(p, tmpbuf, &ulen);
4208 if (swash_fetch(sw, tmpbuf, do_utf8))
4210 to_utf8_upper(p, tmpbuf, &ulen);
4211 if (swash_fetch(sw, tmpbuf, do_utf8))
4217 if (!match && c < 256) {
4218 if (ANYOF_BITMAP_TEST(n, c))
4220 else if (flags & ANYOF_FOLD) {
4223 if (flags & ANYOF_LOCALE) {
4224 PL_reg_flags |= RF_tainted;
4225 f = PL_fold_locale[c];
4229 if (f != c && ANYOF_BITMAP_TEST(n, f))
4233 if (!match && (flags & ANYOF_CLASS)) {
4234 PL_reg_flags |= RF_tainted;
4236 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4237 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4238 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4239 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4240 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4241 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4242 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4243 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4244 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4245 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4246 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4247 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4248 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4249 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4250 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4251 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4252 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4253 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4254 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4255 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4256 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4257 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4258 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4259 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4260 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4261 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4262 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4263 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4264 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4265 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4266 ) /* How's that for a conditional? */
4273 return (flags & ANYOF_INVERT) ? !match : match;
4277 S_reghop(pTHX_ U8 *s, I32 off)
4279 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4283 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4286 while (off-- && s < lim) {
4287 /* XXX could check well-formedness here */
4295 if (UTF8_IS_CONTINUED(*s)) {
4296 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4299 /* XXX could check well-formedness here */
4307 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4309 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4313 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4316 while (off-- && s < lim) {
4317 /* XXX could check well-formedness here */
4327 if (UTF8_IS_CONTINUED(*s)) {
4328 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4331 /* XXX could check well-formedness here */
4343 restore_pos(pTHX_ void *arg)
4345 if (PL_reg_eval_set) {
4346 if (PL_reg_oldsaved) {
4347 PL_reg_re->subbeg = PL_reg_oldsaved;
4348 PL_reg_re->sublen = PL_reg_oldsavedlen;
4349 RX_MATCH_COPIED_on(PL_reg_re);
4351 PL_reg_magic->mg_len = PL_reg_oldpos;
4352 PL_reg_eval_set = 0;
4353 PL_curpm = PL_reg_oldcurpm;