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() */
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 to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1006 f = utf8_to_uvchr(foldbuf, 0);
1008 && (f == c1 || f == c2)
1009 && (ln == foldlen ||
1010 !ibcmp_utf8((char *)foldbuf,
1012 foldlen > ln ? ln : foldlen,
1014 && (norun || regtry(prog, s)) )
1022 c = utf8_to_uvchr((U8*)s, &len);
1024 /* Handle some of the three Greek sigmas cases.
1025 * Note that not all the possible combinations
1026 * are handled here: some of them are handled
1027 * handled by the standard folding rules, and
1028 * some of them (the character class or ANYOF
1029 * cases) are handled during compiletime in
1030 * regexec.c:S_regclass(). */
1031 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1032 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1033 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1035 if ( (c == c1 || c == c2)
1037 !ibcmp_utf8(s, do_utf8,
1038 strend - s > ln ? ln : strend - s,
1040 && (norun || regtry(prog, s)) )
1043 uvchr_to_utf8(tmpbuf, c);
1044 to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1045 f = utf8_to_uvchr(foldbuf, 0);
1047 && (f == c1 || f == c2)
1048 && (ln == foldlen ||
1049 !ibcmp_utf8((char *)foldbuf,
1051 foldlen > ln ? ln : foldlen,
1053 && (norun || regtry(prog, s)) )
1064 && (ln == 1 || !(OP(c) == EXACTF
1066 : ibcmp_locale(s, m, ln)))
1067 && (norun || regtry(prog, s)) )
1073 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1074 && (ln == 1 || !(OP(c) == EXACTF
1076 : ibcmp_locale(s, m, ln)))
1077 && (norun || regtry(prog, s)) )
1084 PL_reg_flags |= RF_tainted;
1091 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1094 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1096 tmp = ((OP(c) == BOUND ?
1097 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1098 LOAD_UTF8_CHARCLASS(alnum,"a");
1099 while (s < strend) {
1100 if (tmp == !(OP(c) == BOUND ?
1101 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1102 isALNUM_LC_utf8((U8*)s)))
1105 if ((norun || regtry(prog, s)))
1112 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1113 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1114 while (s < strend) {
1116 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1118 if ((norun || regtry(prog, s)))
1124 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1128 PL_reg_flags |= RF_tainted;
1135 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1138 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1140 tmp = ((OP(c) == NBOUND ?
1141 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1142 LOAD_UTF8_CHARCLASS(alnum,"a");
1143 while (s < strend) {
1144 if (tmp == !(OP(c) == NBOUND ?
1145 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1146 isALNUM_LC_utf8((U8*)s)))
1148 else if ((norun || regtry(prog, s)))
1154 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1155 tmp = ((OP(c) == NBOUND ?
1156 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1157 while (s < strend) {
1159 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1161 else if ((norun || regtry(prog, s)))
1166 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1171 LOAD_UTF8_CHARCLASS(alnum,"a");
1172 while (s < strend) {
1173 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1174 if (tmp && (norun || regtry(prog, s)))
1185 while (s < strend) {
1187 if (tmp && (norun || regtry(prog, s)))
1199 PL_reg_flags |= RF_tainted;
1201 while (s < strend) {
1202 if (isALNUM_LC_utf8((U8*)s)) {
1203 if (tmp && (norun || regtry(prog, s)))
1214 while (s < strend) {
1215 if (isALNUM_LC(*s)) {
1216 if (tmp && (norun || regtry(prog, s)))
1229 LOAD_UTF8_CHARCLASS(alnum,"a");
1230 while (s < strend) {
1231 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1232 if (tmp && (norun || regtry(prog, s)))
1243 while (s < strend) {
1245 if (tmp && (norun || regtry(prog, s)))
1257 PL_reg_flags |= RF_tainted;
1259 while (s < strend) {
1260 if (!isALNUM_LC_utf8((U8*)s)) {
1261 if (tmp && (norun || regtry(prog, s)))
1272 while (s < strend) {
1273 if (!isALNUM_LC(*s)) {
1274 if (tmp && (norun || regtry(prog, s)))
1287 LOAD_UTF8_CHARCLASS(space," ");
1288 while (s < strend) {
1289 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1290 if (tmp && (norun || regtry(prog, s)))
1301 while (s < strend) {
1303 if (tmp && (norun || regtry(prog, s)))
1315 PL_reg_flags |= RF_tainted;
1317 while (s < strend) {
1318 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1319 if (tmp && (norun || regtry(prog, s)))
1330 while (s < strend) {
1331 if (isSPACE_LC(*s)) {
1332 if (tmp && (norun || regtry(prog, s)))
1345 LOAD_UTF8_CHARCLASS(space," ");
1346 while (s < strend) {
1347 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1348 if (tmp && (norun || regtry(prog, s)))
1359 while (s < strend) {
1361 if (tmp && (norun || regtry(prog, s)))
1373 PL_reg_flags |= RF_tainted;
1375 while (s < strend) {
1376 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1377 if (tmp && (norun || regtry(prog, s)))
1388 while (s < strend) {
1389 if (!isSPACE_LC(*s)) {
1390 if (tmp && (norun || regtry(prog, s)))
1403 LOAD_UTF8_CHARCLASS(digit,"0");
1404 while (s < strend) {
1405 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1406 if (tmp && (norun || regtry(prog, s)))
1417 while (s < strend) {
1419 if (tmp && (norun || regtry(prog, s)))
1431 PL_reg_flags |= RF_tainted;
1433 while (s < strend) {
1434 if (isDIGIT_LC_utf8((U8*)s)) {
1435 if (tmp && (norun || regtry(prog, s)))
1446 while (s < strend) {
1447 if (isDIGIT_LC(*s)) {
1448 if (tmp && (norun || regtry(prog, s)))
1461 LOAD_UTF8_CHARCLASS(digit,"0");
1462 while (s < strend) {
1463 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1464 if (tmp && (norun || regtry(prog, s)))
1475 while (s < strend) {
1477 if (tmp && (norun || regtry(prog, s)))
1489 PL_reg_flags |= RF_tainted;
1491 while (s < strend) {
1492 if (!isDIGIT_LC_utf8((U8*)s)) {
1493 if (tmp && (norun || regtry(prog, s)))
1504 while (s < strend) {
1505 if (!isDIGIT_LC(*s)) {
1506 if (tmp && (norun || regtry(prog, s)))
1518 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1527 - regexec_flags - match a regexp against a string
1530 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1531 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1532 /* strend: pointer to null at end of string */
1533 /* strbeg: real beginning of string */
1534 /* minend: end of match must be >=minend after stringarg. */
1535 /* data: May be used for some additional optimizations. */
1536 /* nosave: For optimizations. */
1539 register regnode *c;
1540 register char *startpos = stringarg;
1541 I32 minlen; /* must match at least this many chars */
1542 I32 dontbother = 0; /* how many characters not to try at end */
1543 /* I32 start_shift = 0; */ /* Offset of the start to find
1544 constant substr. */ /* CC */
1545 I32 end_shift = 0; /* Same for the end. */ /* CC */
1546 I32 scream_pos = -1; /* Internal iterator of scream. */
1548 SV* oreplsv = GvSV(PL_replgv);
1549 bool do_utf8 = DO_UTF8(sv);
1551 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
1558 PL_regnarrate = DEBUG_r_TEST;
1561 /* Be paranoid... */
1562 if (prog == NULL || startpos == NULL) {
1563 Perl_croak(aTHX_ "NULL regexp parameter");
1567 minlen = prog->minlen;
1568 if (strend - startpos < minlen) {
1569 DEBUG_r(PerlIO_printf(Perl_debug_log,
1570 "String too short [regexec_flags]...\n"));
1574 /* Check validity of program. */
1575 if (UCHARAT(prog->program) != REG_MAGIC) {
1576 Perl_croak(aTHX_ "corrupted regexp program");
1580 PL_reg_eval_set = 0;
1583 if (prog->reganch & ROPT_UTF8)
1584 PL_reg_flags |= RF_utf8;
1586 /* Mark beginning of line for ^ and lookbehind. */
1587 PL_regbol = startpos;
1591 /* Mark end of line for $ (and such) */
1594 /* see how far we have to get to not match where we matched before */
1595 PL_regtill = startpos+minend;
1597 /* We start without call_cc context. */
1600 /* If there is a "must appear" string, look for it. */
1603 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1606 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1607 PL_reg_ganch = startpos;
1608 else if (sv && SvTYPE(sv) >= SVt_PVMG
1610 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1611 && mg->mg_len >= 0) {
1612 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1613 if (prog->reganch & ROPT_ANCH_GPOS) {
1614 if (s > PL_reg_ganch)
1619 else /* pos() not defined */
1620 PL_reg_ganch = strbeg;
1623 if (do_utf8 == (UTF!=0) &&
1624 !(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
1625 re_scream_pos_data d;
1627 d.scream_olds = &scream_olds;
1628 d.scream_pos = &scream_pos;
1629 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1631 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1632 goto phooey; /* not present */
1637 char *s = do_utf8 ? sv_uni_display(dsv, sv, 60, 0) : startpos;
1638 int len = do_utf8 ? strlen(s) : strend - startpos;
1641 PerlIO_printf(Perl_debug_log,
1642 "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
1643 PL_colors[4],PL_colors[5],PL_colors[0],
1646 (strlen(prog->precomp) > 60 ? "..." : ""),
1648 (int)(len > 60 ? 60 : len),
1650 (len > 60 ? "..." : "")
1654 /* Simplest case: anchored match need be tried only once. */
1655 /* [unless only anchor is BOL and multiline is set] */
1656 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1657 if (s == startpos && regtry(prog, startpos))
1659 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1660 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1665 dontbother = minlen - 1;
1666 end = HOP3c(strend, -dontbother, strbeg) - 1;
1667 /* for multiline we only have to try after newlines */
1668 if (prog->check_substr) {
1672 if (regtry(prog, s))
1677 if (prog->reganch & RE_USE_INTUIT) {
1678 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1689 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1690 if (regtry(prog, s))
1697 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1698 if (regtry(prog, PL_reg_ganch))
1703 /* Messy cases: unanchored match. */
1704 if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
1705 /* we have /x+whatever/ */
1706 /* it must be a one character string (XXXX Except UTF?) */
1707 char ch = SvPVX(prog->anchored_substr)[0];
1713 while (s < strend) {
1715 DEBUG_r( did_match = 1 );
1716 if (regtry(prog, s)) goto got_it;
1718 while (s < strend && *s == ch)
1725 while (s < strend) {
1727 DEBUG_r( did_match = 1 );
1728 if (regtry(prog, s)) goto got_it;
1730 while (s < strend && *s == ch)
1736 DEBUG_r(if (!did_match)
1737 PerlIO_printf(Perl_debug_log,
1738 "Did not find anchored character...\n")
1742 else if (do_utf8 == (UTF!=0) &&
1743 (prog->anchored_substr != Nullsv
1744 || (prog->float_substr != Nullsv
1745 && prog->float_max_offset < strend - s))) {
1746 SV *must = prog->anchored_substr
1747 ? prog->anchored_substr : prog->float_substr;
1749 prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset;
1751 prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
1752 char *last = HOP3c(strend, /* Cannot start after this */
1753 -(I32)(CHR_SVLEN(must)
1754 - (SvTAIL(must) != 0) + back_min), strbeg);
1755 char *last1; /* Last position checked before */
1761 last1 = HOPc(s, -1);
1763 last1 = s - 1; /* bogus */
1765 /* XXXX check_substr already used to find `s', can optimize if
1766 check_substr==must. */
1768 dontbother = end_shift;
1769 strend = HOPc(strend, -dontbother);
1770 while ( (s <= last) &&
1771 ((flags & REXEC_SCREAM)
1772 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1773 end_shift, &scream_pos, 0))
1774 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1775 (unsigned char*)strend, must,
1776 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1777 DEBUG_r( did_match = 1 );
1778 if (HOPc(s, -back_max) > last1) {
1779 last1 = HOPc(s, -back_min);
1780 s = HOPc(s, -back_max);
1783 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1785 last1 = HOPc(s, -back_min);
1789 while (s <= last1) {
1790 if (regtry(prog, s))
1796 while (s <= last1) {
1797 if (regtry(prog, s))
1803 DEBUG_r(if (!did_match)
1804 PerlIO_printf(Perl_debug_log,
1805 "Did not find %s substr `%s%.*s%s'%s...\n",
1806 ((must == prog->anchored_substr)
1807 ? "anchored" : "floating"),
1809 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1811 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1815 else if ((c = prog->regstclass)) {
1816 if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT)
1817 /* don't bother with what can't match */
1818 strend = HOPc(strend, -(minlen - 1));
1820 SV *prop = sv_newmortal();
1822 PerlIO_printf(Perl_debug_log, "Matching stclass `%s' against `%s'\n", SvPVX(prop), UTF ? sv_uni_display(dsv, sv, 60, 0) : s);
1824 if (find_byclass(prog, c, s, strend, startpos, 0))
1826 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1830 if (prog->float_substr != Nullsv) { /* Trim the end. */
1833 if (flags & REXEC_SCREAM) {
1834 last = screaminstr(sv, prog->float_substr, s - strbeg,
1835 end_shift, &scream_pos, 1); /* last one */
1837 last = scream_olds; /* Only one occurrence. */
1841 char *little = SvPV(prog->float_substr, len);
1843 if (SvTAIL(prog->float_substr)) {
1844 if (memEQ(strend - len + 1, little, len - 1))
1845 last = strend - len + 1;
1846 else if (!PL_multiline)
1847 last = memEQ(strend - len, little, len)
1848 ? strend - len : Nullch;
1854 last = rninstr(s, strend, little, little + len);
1856 last = strend; /* matching `$' */
1860 DEBUG_r(PerlIO_printf(Perl_debug_log,
1861 "%sCan't trim the tail, match fails (should not happen)%s\n",
1862 PL_colors[4],PL_colors[5]));
1863 goto phooey; /* Should not happen! */
1865 dontbother = strend - last + prog->float_min_offset;
1867 if (minlen && (dontbother < minlen))
1868 dontbother = minlen - 1;
1869 strend -= dontbother; /* this one's always in bytes! */
1870 /* We don't know much -- general case. */
1873 if (regtry(prog, s))
1882 if (regtry(prog, s))
1884 } while (s++ < strend);
1892 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1894 if (PL_reg_eval_set) {
1895 /* Preserve the current value of $^R */
1896 if (oreplsv != GvSV(PL_replgv))
1897 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1898 restored, the value remains
1900 restore_pos(aTHX_ 0);
1903 /* make sure $`, $&, $', and $digit will work later */
1904 if ( !(flags & REXEC_NOT_FIRST) ) {
1905 if (RX_MATCH_COPIED(prog)) {
1906 Safefree(prog->subbeg);
1907 RX_MATCH_COPIED_off(prog);
1909 if (flags & REXEC_COPY_STR) {
1910 I32 i = PL_regeol - startpos + (stringarg - strbeg);
1912 s = savepvn(strbeg, i);
1915 RX_MATCH_COPIED_on(prog);
1918 prog->subbeg = strbeg;
1919 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
1926 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
1927 PL_colors[4],PL_colors[5]));
1928 if (PL_reg_eval_set)
1929 restore_pos(aTHX_ 0);
1934 - regtry - try match at specific point
1936 STATIC I32 /* 0 failure, 1 success */
1937 S_regtry(pTHX_ regexp *prog, char *startpos)
1945 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
1947 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
1950 PL_reg_eval_set = RS_init;
1952 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
1953 (IV)(PL_stack_sp - PL_stack_base));
1955 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
1956 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
1957 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
1959 /* Apparently this is not needed, judging by wantarray. */
1960 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
1961 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
1964 /* Make $_ available to executed code. */
1965 if (PL_reg_sv != DEFSV) {
1966 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
1971 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
1972 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
1973 /* prepare for quick setting of pos */
1974 sv_magic(PL_reg_sv, (SV*)0,
1975 PERL_MAGIC_regex_global, Nullch, 0);
1976 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
1980 PL_reg_oldpos = mg->mg_len;
1981 SAVEDESTRUCTOR_X(restore_pos, 0);
1983 if (!PL_reg_curpm) {
1984 Newz(22,PL_reg_curpm, 1, PMOP);
1987 SV* repointer = newSViv(0);
1988 /* so we know which PL_regex_padav element is PL_reg_curpm */
1989 SvFLAGS(repointer) |= SVf_BREAK;
1990 av_push(PL_regex_padav,repointer);
1991 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
1992 PL_regex_pad = AvARRAY(PL_regex_padav);
1996 PM_SETRE(PL_reg_curpm, prog);
1997 PL_reg_oldcurpm = PL_curpm;
1998 PL_curpm = PL_reg_curpm;
1999 if (RX_MATCH_COPIED(prog)) {
2000 /* Here is a serious problem: we cannot rewrite subbeg,
2001 since it may be needed if this match fails. Thus
2002 $` inside (?{}) could fail... */
2003 PL_reg_oldsaved = prog->subbeg;
2004 PL_reg_oldsavedlen = prog->sublen;
2005 RX_MATCH_COPIED_off(prog);
2008 PL_reg_oldsaved = Nullch;
2009 prog->subbeg = PL_bostr;
2010 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2012 prog->startp[0] = startpos - PL_bostr;
2013 PL_reginput = startpos;
2014 PL_regstartp = prog->startp;
2015 PL_regendp = prog->endp;
2016 PL_reglastparen = &prog->lastparen;
2017 PL_reglastcloseparen = &prog->lastcloseparen;
2018 prog->lastparen = 0;
2020 DEBUG_r(PL_reg_starttry = startpos);
2021 if (PL_reg_start_tmpl <= prog->nparens) {
2022 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2023 if(PL_reg_start_tmp)
2024 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2026 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2030 sv_setpvn(PERL_DEBUG_PAD(0), "", 0);
2031 sv_setpvn(PERL_DEBUG_PAD(1), "", 0);
2032 sv_setpvn(PERL_DEBUG_PAD(2), "", 0);
2035 /* XXXX What this code is doing here?!!! There should be no need
2036 to do this again and again, PL_reglastparen should take care of
2039 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2040 * Actually, the code in regcppop() (which Ilya may be meaning by
2041 * PL_reglastparen), is not needed at all by the test suite
2042 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2043 * enough, for building DynaLoader, or otherwise this
2044 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2045 * will happen. Meanwhile, this code *is* needed for the
2046 * above-mentioned test suite tests to succeed. The common theme
2047 * on those tests seems to be returning null fields from matches.
2052 if (prog->nparens) {
2053 for (i = prog->nparens; i > *PL_reglastparen; i--) {
2060 if (regmatch(prog->program + 1)) {
2061 prog->endp[0] = PL_reginput - PL_bostr;
2064 REGCP_UNWIND(lastcp);
2068 #define RE_UNWIND_BRANCH 1
2069 #define RE_UNWIND_BRANCHJ 2
2073 typedef struct { /* XX: makes sense to enlarge it... */
2077 } re_unwind_generic_t;
2090 } re_unwind_branch_t;
2092 typedef union re_unwind_t {
2094 re_unwind_generic_t generic;
2095 re_unwind_branch_t branch;
2098 #define sayYES goto yes
2099 #define sayNO goto no
2100 #define sayYES_FINAL goto yes_final
2101 #define sayYES_LOUD goto yes_loud
2102 #define sayNO_FINAL goto no_final
2103 #define sayNO_SILENT goto do_no
2104 #define saySAME(x) if (x) goto yes; else goto no
2106 #define REPORT_CODE_OFF 24
2109 - regmatch - main matching routine
2111 * Conceptually the strategy is simple: check to see whether the current
2112 * node matches, call self recursively to see whether the rest matches,
2113 * and then act accordingly. In practice we make some effort to avoid
2114 * recursion, in particular by going through "ordinary" nodes (that don't
2115 * need to know whether the rest of the match failed) by a loop instead of
2118 /* [lwall] I've hoisted the register declarations to the outer block in order to
2119 * maybe save a little bit of pushing and popping on the stack. It also takes
2120 * advantage of machines that use a register save mask on subroutine entry.
2122 STATIC I32 /* 0 failure, 1 success */
2123 S_regmatch(pTHX_ regnode *prog)
2125 register regnode *scan; /* Current node. */
2126 regnode *next; /* Next node. */
2127 regnode *inner; /* Next node in internal branch. */
2128 register I32 nextchr; /* renamed nextchr - nextchar colides with
2129 function of same name */
2130 register I32 n; /* no or next */
2131 register I32 ln = 0; /* len or last */
2132 register char *s = Nullch; /* operand or save */
2133 register char *locinput = PL_reginput;
2134 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2135 int minmod = 0, sw = 0, logical = 0;
2138 I32 firstcp = PL_savestack_ix;
2140 register bool do_utf8 = PL_reg_match_utf8;
2142 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2143 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2144 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2151 /* Note that nextchr is a byte even in UTF */
2152 nextchr = UCHARAT(locinput);
2154 while (scan != NULL) {
2157 SV *prop = sv_newmortal();
2158 int docolor = *PL_colors[0];
2159 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2160 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2161 /* The part of the string before starttry has one color
2162 (pref0_len chars), between starttry and current
2163 position another one (pref_len - pref0_len chars),
2164 after the current position the third one.
2165 We assume that pref0_len <= pref_len, otherwise we
2166 decrease pref0_len. */
2167 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2168 ? (5 + taill) - l : locinput - PL_bostr;
2171 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2173 pref0_len = pref_len - (locinput - PL_reg_starttry);
2174 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2175 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2176 ? (5 + taill) - pref_len : PL_regeol - locinput);
2177 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2181 if (pref0_len > pref_len)
2182 pref0_len = pref_len;
2183 regprop(prop, scan);
2187 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2189 locinput - pref_len;
2190 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2191 char *s1 = do_utf8 ?
2192 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2193 pref_len - pref0_len, 60, 0) :
2194 locinput - pref_len + pref0_len;
2195 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2196 char *s2 = do_utf8 ?
2197 pv_uni_display(dsv2, (U8*)locinput,
2198 PL_regeol - locinput, 60, 0) :
2200 int len2 = do_utf8 ? strlen(s2) : l;
2201 PerlIO_printf(Perl_debug_log,
2202 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2203 (IV)(locinput - PL_bostr),
2210 (docolor ? "" : "> <"),
2214 15 - l - pref_len + 1,
2216 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2221 next = scan + NEXT_OFF(scan);
2227 if (locinput == PL_bostr || (PL_multiline &&
2228 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2230 /* regtill = regbol; */
2235 if (locinput == PL_bostr ||
2236 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2242 if (locinput == PL_bostr)
2246 if (locinput == PL_reg_ganch)
2256 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2261 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2263 if (PL_regeol - locinput > 1)
2267 if (PL_regeol != locinput)
2271 if (!nextchr && locinput >= PL_regeol)
2274 locinput += PL_utf8skip[nextchr];
2275 if (locinput > PL_regeol)
2277 nextchr = UCHARAT(locinput);
2280 nextchr = UCHARAT(++locinput);
2283 if (!nextchr && locinput >= PL_regeol)
2285 nextchr = UCHARAT(++locinput);
2288 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2291 locinput += PL_utf8skip[nextchr];
2292 if (locinput > PL_regeol)
2294 nextchr = UCHARAT(locinput);
2297 nextchr = UCHARAT(++locinput);
2302 if (do_utf8 != (UTF!=0)) {
2303 /* The target and the pattern have differing utf8ness. */
2309 /* The target is utf8, the pattern is not utf8. */
2313 if (NATIVE_TO_UNI(*(U8*)s) !=
2314 utf8_to_uvchr((U8*)l, &ulen))
2321 /* The target is not utf8, the pattern is utf8. */
2325 if (NATIVE_TO_UNI(*((U8*)l)) !=
2326 utf8_to_uvchr((U8*)s, &ulen))
2333 nextchr = UCHARAT(locinput);
2336 /* The target and the pattern have the same utf8ness. */
2337 /* Inline the first character, for speed. */
2338 if (UCHARAT(s) != nextchr)
2340 if (PL_regeol - locinput < ln)
2342 if (ln > 1 && memNE(s, locinput, ln))
2345 nextchr = UCHARAT(locinput);
2348 PL_reg_flags |= RF_tainted;
2358 if (do_utf8 != (UTF!=0)) {
2359 /* The target and the pattern have differing utf8ness. */
2360 STRLEN ulen1, ulen2;
2364 /* The target is utf8, the pattern is not utf8. */
2369 cs = to_uni_fold(NATIVE_TO_UNI(*(U8*)s),
2371 cl = utf8_to_uvchr((U8*)l, &ulen2);
2374 cl = to_uni_fold(cl, (U8*)l, &ulen2);
2375 if (ulen1 != ulen2 || cs != cl)
2383 /* The target is not utf8, the pattern is utf8. */
2388 cs = utf8_to_uvchr((U8*)s, &ulen1);
2390 cl = to_uni_fold(NATIVE_TO_UNI(*(U8*)l),
2394 cs = to_uni_fold(cs, (U8*)s, &ulen1);
2395 if (ulen1 != ulen2 || cs != cl)
2403 nextchr = UCHARAT(locinput);
2407 if (do_utf8 && UTF) {
2408 /* Both the target and the pattern are utf8. */
2412 if (UTF8SKIP(s) != UTF8SKIP(l) ||
2413 memNE(s, (char*)l, UTF8SKIP(s))) {
2414 U8 lfoldbuf[UTF8_MAXLEN_FOLD+1];
2417 /* Try one of them folded. */
2419 to_utf8_fold((U8*)l, lfoldbuf, &lfoldlen);
2420 if (UTF8SKIP(s) != lfoldlen ||
2421 memNE(s, (char*)lfoldbuf, lfoldlen)) {
2422 U8 sfoldbuf[UTF8_MAXLEN_FOLD+1];
2425 /* Try both of them folded. */
2427 to_utf8_fold((U8*)s, sfoldbuf, &sfoldlen);
2428 if (sfoldlen != lfoldlen ||
2429 memNE((char*)sfoldbuf,
2430 (char*)lfoldbuf, lfoldlen))
2438 nextchr = UCHARAT(locinput);
2443 /* Neither the target and the pattern are utf8. */
2445 /* Inline the first character, for speed. */
2446 if (UCHARAT(s) != nextchr &&
2447 UCHARAT(s) != ((OP(scan) == EXACTF)
2448 ? PL_fold : PL_fold_locale)[nextchr])
2450 if (PL_regeol - locinput < ln)
2452 if (ln > 1 && (OP(scan) == EXACTF
2453 ? ibcmp(s, locinput, ln)
2454 : ibcmp_locale(s, locinput, ln)))
2457 nextchr = UCHARAT(locinput);
2461 if (!reginclass(scan, (U8*)locinput, do_utf8))
2463 if (locinput >= PL_regeol)
2465 locinput += PL_utf8skip[nextchr];
2466 nextchr = UCHARAT(locinput);
2470 nextchr = UCHARAT(locinput);
2471 if (!reginclass(scan, (U8*)locinput, do_utf8))
2473 if (!nextchr && locinput >= PL_regeol)
2475 nextchr = UCHARAT(++locinput);
2479 PL_reg_flags |= RF_tainted;
2485 LOAD_UTF8_CHARCLASS(alnum,"a");
2486 if (!(OP(scan) == ALNUM
2487 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2488 : isALNUM_LC_utf8((U8*)locinput)))
2492 locinput += PL_utf8skip[nextchr];
2493 nextchr = UCHARAT(locinput);
2496 if (!(OP(scan) == ALNUM
2497 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2499 nextchr = UCHARAT(++locinput);
2502 PL_reg_flags |= RF_tainted;
2505 if (!nextchr && locinput >= PL_regeol)
2508 LOAD_UTF8_CHARCLASS(alnum,"a");
2509 if (OP(scan) == NALNUM
2510 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2511 : isALNUM_LC_utf8((U8*)locinput))
2515 locinput += PL_utf8skip[nextchr];
2516 nextchr = UCHARAT(locinput);
2519 if (OP(scan) == NALNUM
2520 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2522 nextchr = UCHARAT(++locinput);
2526 PL_reg_flags |= RF_tainted;
2530 /* was last char in word? */
2532 if (locinput == PL_bostr)
2535 U8 *r = reghop((U8*)locinput, -1);
2537 ln = utf8n_to_uvchr(r, s - (char*)r, 0, 0);
2539 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2540 ln = isALNUM_uni(ln);
2541 LOAD_UTF8_CHARCLASS(alnum,"a");
2542 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2545 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2546 n = isALNUM_LC_utf8((U8*)locinput);
2550 ln = (locinput != PL_bostr) ?
2551 UCHARAT(locinput - 1) : '\n';
2552 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2554 n = isALNUM(nextchr);
2557 ln = isALNUM_LC(ln);
2558 n = isALNUM_LC(nextchr);
2561 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2562 OP(scan) == BOUNDL))
2566 PL_reg_flags |= RF_tainted;
2572 if (UTF8_IS_CONTINUED(nextchr)) {
2573 LOAD_UTF8_CHARCLASS(space," ");
2574 if (!(OP(scan) == SPACE
2575 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2576 : isSPACE_LC_utf8((U8*)locinput)))
2580 locinput += PL_utf8skip[nextchr];
2581 nextchr = UCHARAT(locinput);
2584 if (!(OP(scan) == SPACE
2585 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2587 nextchr = UCHARAT(++locinput);
2590 if (!(OP(scan) == SPACE
2591 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2593 nextchr = UCHARAT(++locinput);
2597 PL_reg_flags |= RF_tainted;
2600 if (!nextchr && locinput >= PL_regeol)
2603 LOAD_UTF8_CHARCLASS(space," ");
2604 if (OP(scan) == NSPACE
2605 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2606 : isSPACE_LC_utf8((U8*)locinput))
2610 locinput += PL_utf8skip[nextchr];
2611 nextchr = UCHARAT(locinput);
2614 if (OP(scan) == NSPACE
2615 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2617 nextchr = UCHARAT(++locinput);
2620 PL_reg_flags |= RF_tainted;
2626 LOAD_UTF8_CHARCLASS(digit,"0");
2627 if (!(OP(scan) == DIGIT
2628 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2629 : isDIGIT_LC_utf8((U8*)locinput)))
2633 locinput += PL_utf8skip[nextchr];
2634 nextchr = UCHARAT(locinput);
2637 if (!(OP(scan) == DIGIT
2638 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2640 nextchr = UCHARAT(++locinput);
2643 PL_reg_flags |= RF_tainted;
2646 if (!nextchr && locinput >= PL_regeol)
2649 LOAD_UTF8_CHARCLASS(digit,"0");
2650 if (OP(scan) == NDIGIT
2651 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2652 : isDIGIT_LC_utf8((U8*)locinput))
2656 locinput += PL_utf8skip[nextchr];
2657 nextchr = UCHARAT(locinput);
2660 if (OP(scan) == NDIGIT
2661 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2663 nextchr = UCHARAT(++locinput);
2666 if (locinput >= PL_regeol)
2669 LOAD_UTF8_CHARCLASS(mark,"~");
2670 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2672 locinput += PL_utf8skip[nextchr];
2673 while (locinput < PL_regeol &&
2674 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2675 locinput += UTF8SKIP(locinput);
2676 if (locinput > PL_regeol)
2681 nextchr = UCHARAT(locinput);
2684 PL_reg_flags |= RF_tainted;
2688 n = ARG(scan); /* which paren pair */
2689 ln = PL_regstartp[n];
2690 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2691 if (*PL_reglastparen < n || ln == -1)
2692 sayNO; /* Do not match unless seen CLOSEn. */
2693 if (ln == PL_regendp[n])
2697 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2699 char *e = PL_bostr + PL_regendp[n];
2701 * Note that we can't do the "other character" lookup trick as
2702 * in the 8-bit case (no pun intended) because in Unicode we
2703 * have to map both upper and title case to lower case.
2705 if (OP(scan) == REFF) {
2706 STRLEN ulen1, ulen2;
2707 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2708 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2712 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2713 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2714 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2721 nextchr = UCHARAT(locinput);
2725 /* Inline the first character, for speed. */
2726 if (UCHARAT(s) != nextchr &&
2728 (UCHARAT(s) != ((OP(scan) == REFF
2729 ? PL_fold : PL_fold_locale)[nextchr]))))
2731 ln = PL_regendp[n] - ln;
2732 if (locinput + ln > PL_regeol)
2734 if (ln > 1 && (OP(scan) == REF
2735 ? memNE(s, locinput, ln)
2737 ? ibcmp(s, locinput, ln)
2738 : ibcmp_locale(s, locinput, ln))))
2741 nextchr = UCHARAT(locinput);
2752 OP_4tree *oop = PL_op;
2753 COP *ocurcop = PL_curcop;
2754 SV **ocurpad = PL_curpad;
2758 PL_op = (OP_4tree*)PL_regdata->data[n];
2759 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2760 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2761 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2765 CALLRUNOPS(aTHX); /* Scalar context. */
2768 ret = Nullsv; /* protect against empty (?{}) blocks. */
2776 PL_curpad = ocurpad;
2777 PL_curcop = ocurcop;
2779 if (logical == 2) { /* Postponed subexpression. */
2781 MAGIC *mg = Null(MAGIC*);
2783 CHECKPOINT cp, lastcp;
2785 if(SvROK(ret) || SvRMAGICAL(ret)) {
2786 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2789 mg = mg_find(sv, PERL_MAGIC_qr);
2792 re = (regexp *)mg->mg_obj;
2793 (void)ReREFCNT_inc(re);
2797 char *t = SvPV(ret, len);
2799 char *oprecomp = PL_regprecomp;
2800 I32 osize = PL_regsize;
2801 I32 onpar = PL_regnpar;
2804 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2806 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2807 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2809 PL_regprecomp = oprecomp;
2814 PerlIO_printf(Perl_debug_log,
2815 "Entering embedded `%s%.60s%s%s'\n",
2819 (strlen(re->precomp) > 60 ? "..." : ""))
2822 state.prev = PL_reg_call_cc;
2823 state.cc = PL_regcc;
2824 state.re = PL_reg_re;
2828 cp = regcppush(0); /* Save *all* the positions. */
2831 state.ss = PL_savestack_ix;
2832 *PL_reglastparen = 0;
2833 *PL_reglastcloseparen = 0;
2834 PL_reg_call_cc = &state;
2835 PL_reginput = locinput;
2837 /* XXXX This is too dramatic a measure... */
2840 if (regmatch(re->program + 1)) {
2841 /* Even though we succeeded, we need to restore
2842 global variables, since we may be wrapped inside
2843 SUSPEND, thus the match may be not finished yet. */
2845 /* XXXX Do this only if SUSPENDed? */
2846 PL_reg_call_cc = state.prev;
2847 PL_regcc = state.cc;
2848 PL_reg_re = state.re;
2849 cache_re(PL_reg_re);
2851 /* XXXX This is too dramatic a measure... */
2854 /* These are needed even if not SUSPEND. */
2860 REGCP_UNWIND(lastcp);
2862 PL_reg_call_cc = state.prev;
2863 PL_regcc = state.cc;
2864 PL_reg_re = state.re;
2865 cache_re(PL_reg_re);
2867 /* XXXX This is too dramatic a measure... */
2877 sv_setsv(save_scalar(PL_replgv), ret);
2881 n = ARG(scan); /* which paren pair */
2882 PL_reg_start_tmp[n] = locinput;
2887 n = ARG(scan); /* which paren pair */
2888 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2889 PL_regendp[n] = locinput - PL_bostr;
2890 if (n > *PL_reglastparen)
2891 *PL_reglastparen = n;
2892 *PL_reglastcloseparen = n;
2895 n = ARG(scan); /* which paren pair */
2896 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2899 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2901 next = NEXTOPER(NEXTOPER(scan));
2903 next = scan + ARG(scan);
2904 if (OP(next) == IFTHEN) /* Fake one. */
2905 next = NEXTOPER(NEXTOPER(next));
2909 logical = scan->flags;
2911 /*******************************************************************
2912 PL_regcc contains infoblock about the innermost (...)* loop, and
2913 a pointer to the next outer infoblock.
2915 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2917 1) After matching X, regnode for CURLYX is processed;
2919 2) This regnode creates infoblock on the stack, and calls
2920 regmatch() recursively with the starting point at WHILEM node;
2922 3) Each hit of WHILEM node tries to match A and Z (in the order
2923 depending on the current iteration, min/max of {min,max} and
2924 greediness). The information about where are nodes for "A"
2925 and "Z" is read from the infoblock, as is info on how many times "A"
2926 was already matched, and greediness.
2928 4) After A matches, the same WHILEM node is hit again.
2930 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2931 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2932 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2933 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2934 of the external loop.
2936 Currently present infoblocks form a tree with a stem formed by PL_curcc
2937 and whatever it mentions via ->next, and additional attached trees
2938 corresponding to temporarily unset infoblocks as in "5" above.
2940 In the following picture infoblocks for outer loop of
2941 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2942 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2943 infoblocks are drawn below the "reset" infoblock.
2945 In fact in the picture below we do not show failed matches for Z and T
2946 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2947 more obvious *why* one needs to *temporary* unset infoblocks.]
2949 Matched REx position InfoBlocks Comment
2953 Y A)*?Z)*?T x <- O <- I
2954 YA )*?Z)*?T x <- O <- I
2955 YA A)*?Z)*?T x <- O <- I
2956 YAA )*?Z)*?T x <- O <- I
2957 YAA Z)*?T x <- O # Temporary unset I
2960 YAAZ Y(A)*?Z)*?T x <- O
2963 YAAZY (A)*?Z)*?T x <- O
2966 YAAZY A)*?Z)*?T x <- O <- I
2969 YAAZYA )*?Z)*?T x <- O <- I
2972 YAAZYA Z)*?T x <- O # Temporary unset I
2978 YAAZYAZ T x # Temporary unset O
2985 *******************************************************************/
2988 CHECKPOINT cp = PL_savestack_ix;
2989 /* No need to save/restore up to this paren */
2990 I32 parenfloor = scan->flags;
2992 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2994 cc.oldcc = PL_regcc;
2996 /* XXXX Probably it is better to teach regpush to support
2997 parenfloor > PL_regsize... */
2998 if (parenfloor > *PL_reglastparen)
2999 parenfloor = *PL_reglastparen; /* Pessimization... */
3000 cc.parenfloor = parenfloor;
3002 cc.min = ARG1(scan);
3003 cc.max = ARG2(scan);
3004 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3008 PL_reginput = locinput;
3009 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
3011 PL_regcc = cc.oldcc;
3017 * This is really hard to understand, because after we match
3018 * what we're trying to match, we must make sure the rest of
3019 * the REx is going to match for sure, and to do that we have
3020 * to go back UP the parse tree by recursing ever deeper. And
3021 * if it fails, we have to reset our parent's current state
3022 * that we can try again after backing off.
3025 CHECKPOINT cp, lastcp;
3026 CURCUR* cc = PL_regcc;
3027 char *lastloc = cc->lastloc; /* Detection of 0-len. */
3029 n = cc->cur + 1; /* how many we know we matched */
3030 PL_reginput = locinput;
3033 PerlIO_printf(Perl_debug_log,
3034 "%*s %ld out of %ld..%ld cc=%lx\n",
3035 REPORT_CODE_OFF+PL_regindent*2, "",
3036 (long)n, (long)cc->min,
3037 (long)cc->max, (long)cc)
3040 /* If degenerate scan matches "", assume scan done. */
3042 if (locinput == cc->lastloc && n >= cc->min) {
3043 PL_regcc = cc->oldcc;
3047 PerlIO_printf(Perl_debug_log,
3048 "%*s empty match detected, try continuation...\n",
3049 REPORT_CODE_OFF+PL_regindent*2, "")
3051 if (regmatch(cc->next))
3059 /* First just match a string of min scans. */
3063 cc->lastloc = locinput;
3064 if (regmatch(cc->scan))
3067 cc->lastloc = lastloc;
3072 /* Check whether we already were at this position.
3073 Postpone detection until we know the match is not
3074 *that* much linear. */
3075 if (!PL_reg_maxiter) {
3076 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3077 PL_reg_leftiter = PL_reg_maxiter;
3079 if (PL_reg_leftiter-- == 0) {
3080 I32 size = (PL_reg_maxiter + 7)/8;
3081 if (PL_reg_poscache) {
3082 if (PL_reg_poscache_size < size) {
3083 Renew(PL_reg_poscache, size, char);
3084 PL_reg_poscache_size = size;
3086 Zero(PL_reg_poscache, size, char);
3089 PL_reg_poscache_size = size;
3090 Newz(29, PL_reg_poscache, size, char);
3093 PerlIO_printf(Perl_debug_log,
3094 "%sDetected a super-linear match, switching on caching%s...\n",
3095 PL_colors[4], PL_colors[5])
3098 if (PL_reg_leftiter < 0) {
3099 I32 o = locinput - PL_bostr, b;
3101 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3104 if (PL_reg_poscache[o] & (1<<b)) {
3106 PerlIO_printf(Perl_debug_log,
3107 "%*s already tried at this position...\n",
3108 REPORT_CODE_OFF+PL_regindent*2, "")
3112 PL_reg_poscache[o] |= (1<<b);
3116 /* Prefer next over scan for minimal matching. */
3119 PL_regcc = cc->oldcc;
3122 cp = regcppush(cc->parenfloor);
3124 if (regmatch(cc->next)) {
3126 sayYES; /* All done. */
3128 REGCP_UNWIND(lastcp);
3134 if (n >= cc->max) { /* Maximum greed exceeded? */
3135 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3136 && !(PL_reg_flags & RF_warned)) {
3137 PL_reg_flags |= RF_warned;
3138 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3139 "Complex regular subexpression recursion",
3146 PerlIO_printf(Perl_debug_log,
3147 "%*s trying longer...\n",
3148 REPORT_CODE_OFF+PL_regindent*2, "")
3150 /* Try scanning more and see if it helps. */
3151 PL_reginput = locinput;
3153 cc->lastloc = locinput;
3154 cp = regcppush(cc->parenfloor);
3156 if (regmatch(cc->scan)) {
3160 REGCP_UNWIND(lastcp);
3163 cc->lastloc = lastloc;
3167 /* Prefer scan over next for maximal matching. */
3169 if (n < cc->max) { /* More greed allowed? */
3170 cp = regcppush(cc->parenfloor);
3172 cc->lastloc = locinput;
3174 if (regmatch(cc->scan)) {
3178 REGCP_UNWIND(lastcp);
3179 regcppop(); /* Restore some previous $<digit>s? */
3180 PL_reginput = locinput;
3182 PerlIO_printf(Perl_debug_log,
3183 "%*s failed, try continuation...\n",
3184 REPORT_CODE_OFF+PL_regindent*2, "")
3187 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3188 && !(PL_reg_flags & RF_warned)) {
3189 PL_reg_flags |= RF_warned;
3190 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3191 "Complex regular subexpression recursion",
3195 /* Failed deeper matches of scan, so see if this one works. */
3196 PL_regcc = cc->oldcc;
3199 if (regmatch(cc->next))
3205 cc->lastloc = lastloc;
3210 next = scan + ARG(scan);
3213 inner = NEXTOPER(NEXTOPER(scan));
3216 inner = NEXTOPER(scan);
3220 if (OP(next) != c1) /* No choice. */
3221 next = inner; /* Avoid recursion. */
3223 I32 lastparen = *PL_reglastparen;
3225 re_unwind_branch_t *uw;
3227 /* Put unwinding data on stack */
3228 unwind1 = SSNEWt(1,re_unwind_branch_t);
3229 uw = SSPTRt(unwind1,re_unwind_branch_t);
3232 uw->type = ((c1 == BRANCH)
3234 : RE_UNWIND_BRANCHJ);
3235 uw->lastparen = lastparen;
3237 uw->locinput = locinput;
3238 uw->nextchr = nextchr;
3240 uw->regindent = ++PL_regindent;
3243 REGCP_SET(uw->lastcp);
3245 /* Now go into the first branch */
3258 /* We suppose that the next guy does not need
3259 backtracking: in particular, it is of constant length,
3260 and has no parenths to influence future backrefs. */
3261 ln = ARG1(scan); /* min to match */
3262 n = ARG2(scan); /* max to match */
3263 paren = scan->flags;
3265 if (paren > PL_regsize)
3267 if (paren > *PL_reglastparen)
3268 *PL_reglastparen = paren;
3270 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3272 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3273 PL_reginput = locinput;
3276 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3278 /* if we matched something zero-length we don't need to
3279 backtrack - capturing parens are already defined, so
3280 the caveat in the maximal case doesn't apply
3282 XXXX if ln == 0, we can redo this check first time
3283 through the following loop
3286 n = ln; /* don't backtrack */
3287 locinput = PL_reginput;
3288 if (HAS_TEXT(next) || JUMPABLE(next)) {
3289 regnode *text_node = next;
3291 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3293 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3295 if (PL_regkind[(U8)OP(text_node)] == REF) {
3297 n = ARG(text_node); /* which paren pair */
3298 ln = PL_regstartp[n];
3299 /* assume yes if we haven't seen CLOSEn */
3301 *PL_reglastparen < n ||
3308 c1 = *(PL_bostr + ln);
3310 else { c1 = (U8)*STRING(text_node); }
3311 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3313 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3314 c2 = PL_fold_locale[c1];
3323 /* This may be improved if l == 0. */
3324 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3325 /* If it could work, try it. */
3327 UCHARAT(PL_reginput) == c1 ||
3328 UCHARAT(PL_reginput) == c2)
3332 PL_regstartp[paren] =
3333 HOPc(PL_reginput, -l) - PL_bostr;
3334 PL_regendp[paren] = PL_reginput - PL_bostr;
3337 PL_regendp[paren] = -1;
3341 REGCP_UNWIND(lastcp);
3343 /* Couldn't or didn't -- move forward. */
3344 PL_reginput = locinput;
3345 if (regrepeat_hard(scan, 1, &l)) {
3347 locinput = PL_reginput;
3354 n = regrepeat_hard(scan, n, &l);
3355 /* if we matched something zero-length we don't need to
3356 backtrack, unless the minimum count is zero and we
3357 are capturing the result - in that case the capture
3358 being defined or not may affect later execution
3360 if (n != 0 && l == 0 && !(paren && ln == 0))
3361 ln = n; /* don't backtrack */
3362 locinput = PL_reginput;
3364 PerlIO_printf(Perl_debug_log,
3365 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3366 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3370 if (HAS_TEXT(next) || JUMPABLE(next)) {
3371 regnode *text_node = next;
3373 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3375 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3377 if (PL_regkind[(U8)OP(text_node)] == REF) {
3379 n = ARG(text_node); /* which paren pair */
3380 ln = PL_regstartp[n];
3381 /* assume yes if we haven't seen CLOSEn */
3383 *PL_reglastparen < n ||
3390 c1 = *(PL_bostr + ln);
3392 else { c1 = (U8)*STRING(text_node); }
3394 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3396 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3397 c2 = PL_fold_locale[c1];
3408 /* If it could work, try it. */
3410 UCHARAT(PL_reginput) == c1 ||
3411 UCHARAT(PL_reginput) == c2)
3414 PerlIO_printf(Perl_debug_log,
3415 "%*s trying tail with n=%"IVdf"...\n",
3416 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3420 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3421 PL_regendp[paren] = PL_reginput - PL_bostr;
3424 PL_regendp[paren] = -1;
3428 REGCP_UNWIND(lastcp);
3430 /* Couldn't or didn't -- back up. */
3432 locinput = HOPc(locinput, -l);
3433 PL_reginput = locinput;
3440 paren = scan->flags; /* Which paren to set */
3441 if (paren > PL_regsize)
3443 if (paren > *PL_reglastparen)
3444 *PL_reglastparen = paren;
3445 ln = ARG1(scan); /* min to match */
3446 n = ARG2(scan); /* max to match */
3447 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3451 ln = ARG1(scan); /* min to match */
3452 n = ARG2(scan); /* max to match */
3453 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3458 scan = NEXTOPER(scan);
3464 scan = NEXTOPER(scan);
3468 * Lookahead to avoid useless match attempts
3469 * when we know what character comes next.
3473 * Used to only do .*x and .*?x, but now it allows
3474 * for )'s, ('s and (?{ ... })'s to be in the way
3475 * of the quantifier and the EXACT-like node. -- japhy
3478 if (HAS_TEXT(next) || JUMPABLE(next)) {
3480 regnode *text_node = next;
3482 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3484 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3486 if (PL_regkind[(U8)OP(text_node)] == REF) {
3488 n = ARG(text_node); /* which paren pair */
3489 ln = PL_regstartp[n];
3490 /* assume yes if we haven't seen CLOSEn */
3492 *PL_reglastparen < n ||
3497 goto assume_ok_easy;
3499 s = (U8*)PL_bostr + ln;
3501 else { s = (U8*)STRING(text_node); }
3505 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3507 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3508 c2 = PL_fold_locale[c1];
3511 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3512 STRLEN ulen1, ulen2;
3513 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3514 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3516 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3517 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3519 c1 = utf8_to_uvuni(tmpbuf1, 0);
3520 c2 = utf8_to_uvuni(tmpbuf2, 0);
3523 c2 = c1 = utf8_to_uvchr(s, NULL);
3531 PL_reginput = locinput;
3535 if (ln && regrepeat(scan, ln) < ln)
3537 locinput = PL_reginput;
3540 char *e; /* Should not check after this */
3541 char *old = locinput;
3543 if (n == REG_INFTY) {
3546 while (UTF8_IS_CONTINUATION(*(U8*)e))
3552 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3556 e = locinput + n - ln;
3562 /* Find place 'next' could work */
3565 while (locinput <= e &&
3566 UCHARAT(locinput) != c1)
3569 while (locinput <= e
3570 && UCHARAT(locinput) != c1
3571 && UCHARAT(locinput) != c2)
3574 count = locinput - old;
3581 utf8_to_uvchr((U8*)locinput, &len) != c1;
3586 for (count = 0; locinput <= e; count++) {
3587 UV c = utf8_to_uvchr((U8*)locinput, &len);
3588 if (c == c1 || c == c2)
3596 /* PL_reginput == old now */
3597 if (locinput != old) {
3598 ln = 1; /* Did some */
3599 if (regrepeat(scan, count) < count)
3602 /* PL_reginput == locinput now */
3603 TRYPAREN(paren, ln, locinput);
3604 PL_reginput = locinput; /* Could be reset... */
3605 REGCP_UNWIND(lastcp);
3606 /* Couldn't or didn't -- move forward. */
3609 locinput += UTF8SKIP(locinput);
3615 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3619 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3621 c = UCHARAT(PL_reginput);
3622 /* If it could work, try it. */
3623 if (c == c1 || c == c2)
3625 TRYPAREN(paren, n, PL_reginput);
3626 REGCP_UNWIND(lastcp);
3629 /* If it could work, try it. */
3630 else if (c1 == -1000)
3632 TRYPAREN(paren, n, PL_reginput);
3633 REGCP_UNWIND(lastcp);
3635 /* Couldn't or didn't -- move forward. */
3636 PL_reginput = locinput;
3637 if (regrepeat(scan, 1)) {
3639 locinput = PL_reginput;
3647 n = regrepeat(scan, n);
3648 locinput = PL_reginput;
3649 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3650 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3651 ln = n; /* why back off? */
3652 /* ...because $ and \Z can match before *and* after
3653 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3654 We should back off by one in this case. */
3655 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3664 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3666 c = UCHARAT(PL_reginput);
3668 /* If it could work, try it. */
3669 if (c1 == -1000 || c == c1 || c == c2)
3671 TRYPAREN(paren, n, PL_reginput);
3672 REGCP_UNWIND(lastcp);
3674 /* Couldn't or didn't -- back up. */
3676 PL_reginput = locinput = HOPc(locinput, -1);
3684 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3686 c = UCHARAT(PL_reginput);
3688 /* If it could work, try it. */
3689 if (c1 == -1000 || c == c1 || c == c2)
3691 TRYPAREN(paren, n, PL_reginput);
3692 REGCP_UNWIND(lastcp);
3694 /* Couldn't or didn't -- back up. */
3696 PL_reginput = locinput = HOPc(locinput, -1);
3703 if (PL_reg_call_cc) {
3704 re_cc_state *cur_call_cc = PL_reg_call_cc;
3705 CURCUR *cctmp = PL_regcc;
3706 regexp *re = PL_reg_re;
3707 CHECKPOINT cp, lastcp;
3709 cp = regcppush(0); /* Save *all* the positions. */
3711 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3713 PL_reginput = locinput; /* Make position available to
3715 cache_re(PL_reg_call_cc->re);
3716 PL_regcc = PL_reg_call_cc->cc;
3717 PL_reg_call_cc = PL_reg_call_cc->prev;
3718 if (regmatch(cur_call_cc->node)) {
3719 PL_reg_call_cc = cur_call_cc;
3723 REGCP_UNWIND(lastcp);
3725 PL_reg_call_cc = cur_call_cc;
3731 PerlIO_printf(Perl_debug_log,
3732 "%*s continuation failed...\n",
3733 REPORT_CODE_OFF+PL_regindent*2, "")
3737 if (locinput < PL_regtill) {
3738 DEBUG_r(PerlIO_printf(Perl_debug_log,
3739 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3741 (long)(locinput - PL_reg_starttry),
3742 (long)(PL_regtill - PL_reg_starttry),
3744 sayNO_FINAL; /* Cannot match: too short. */
3746 PL_reginput = locinput; /* put where regtry can find it */
3747 sayYES_FINAL; /* Success! */
3749 PL_reginput = locinput; /* put where regtry can find it */
3750 sayYES_LOUD; /* Success! */
3753 PL_reginput = locinput;
3758 s = HOPBACKc(locinput, scan->flags);
3764 PL_reginput = locinput;
3769 s = HOPBACKc(locinput, scan->flags);
3775 PL_reginput = locinput;
3778 inner = NEXTOPER(NEXTOPER(scan));
3779 if (regmatch(inner) != n) {
3794 if (OP(scan) == SUSPEND) {
3795 locinput = PL_reginput;
3796 nextchr = UCHARAT(locinput);
3801 next = scan + ARG(scan);
3806 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3807 PTR2UV(scan), OP(scan));
3808 Perl_croak(aTHX_ "regexp memory corruption");
3815 * We get here only if there's trouble -- normally "case END" is
3816 * the terminating point.
3818 Perl_croak(aTHX_ "corrupted regexp pointers");
3824 PerlIO_printf(Perl_debug_log,
3825 "%*s %scould match...%s\n",
3826 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3830 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3831 PL_colors[4],PL_colors[5]));
3837 #if 0 /* Breaks $^R */
3845 PerlIO_printf(Perl_debug_log,
3846 "%*s %sfailed...%s\n",
3847 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3853 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3856 case RE_UNWIND_BRANCH:
3857 case RE_UNWIND_BRANCHJ:
3859 re_unwind_branch_t *uwb = &(uw->branch);
3860 I32 lastparen = uwb->lastparen;
3862 REGCP_UNWIND(uwb->lastcp);
3863 for (n = *PL_reglastparen; n > lastparen; n--)
3865 *PL_reglastparen = n;
3866 scan = next = uwb->next;
3868 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3869 ? BRANCH : BRANCHJ) ) { /* Failure */
3876 /* Have more choice yet. Reuse the same uwb. */
3878 if ((n = (uwb->type == RE_UNWIND_BRANCH
3879 ? NEXT_OFF(next) : ARG(next))))
3882 next = NULL; /* XXXX Needn't unwinding in this case... */
3884 next = NEXTOPER(scan);
3885 if (uwb->type == RE_UNWIND_BRANCHJ)
3886 next = NEXTOPER(next);
3887 locinput = uwb->locinput;
3888 nextchr = uwb->nextchr;
3890 PL_regindent = uwb->regindent;
3897 Perl_croak(aTHX_ "regexp unwind memory corruption");
3908 - regrepeat - repeatedly match something simple, report how many
3911 * [This routine now assumes that it will only match on things of length 1.
3912 * That was true before, but now we assume scan - reginput is the count,
3913 * rather than incrementing count on every character. [Er, except utf8.]]
3916 S_regrepeat(pTHX_ regnode *p, I32 max)
3918 register char *scan;
3920 register char *loceol = PL_regeol;
3921 register I32 hardcount = 0;
3922 register bool do_utf8 = PL_reg_match_utf8;
3925 if (max != REG_INFTY && max < loceol - scan)
3926 loceol = scan + max;
3931 while (scan < loceol && hardcount < max && *scan != '\n') {
3932 scan += UTF8SKIP(scan);
3936 while (scan < loceol && *scan != '\n')
3946 case EXACT: /* length of string is 1 */
3948 while (scan < loceol && UCHARAT(scan) == c)
3951 case EXACTF: /* length of string is 1 */
3953 while (scan < loceol &&
3954 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3957 case EXACTFL: /* length of string is 1 */
3958 PL_reg_flags |= RF_tainted;
3960 while (scan < loceol &&
3961 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3967 while (hardcount < max && scan < loceol &&
3968 reginclass(p, (U8*)scan, do_utf8)) {
3969 scan += UTF8SKIP(scan);
3973 while (scan < loceol && reginclass(p, (U8*)scan, do_utf8))
3980 LOAD_UTF8_CHARCLASS(alnum,"a");
3981 while (hardcount < max && scan < loceol &&
3982 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3983 scan += UTF8SKIP(scan);
3987 while (scan < loceol && isALNUM(*scan))
3992 PL_reg_flags |= RF_tainted;
3995 while (hardcount < max && scan < loceol &&
3996 isALNUM_LC_utf8((U8*)scan)) {
3997 scan += UTF8SKIP(scan);
4001 while (scan < loceol && isALNUM_LC(*scan))
4008 LOAD_UTF8_CHARCLASS(alnum,"a");
4009 while (hardcount < max && scan < loceol &&
4010 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4011 scan += UTF8SKIP(scan);
4015 while (scan < loceol && !isALNUM(*scan))
4020 PL_reg_flags |= RF_tainted;
4023 while (hardcount < max && scan < loceol &&
4024 !isALNUM_LC_utf8((U8*)scan)) {
4025 scan += UTF8SKIP(scan);
4029 while (scan < loceol && !isALNUM_LC(*scan))
4036 LOAD_UTF8_CHARCLASS(space," ");
4037 while (hardcount < max && scan < loceol &&
4039 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4040 scan += UTF8SKIP(scan);
4044 while (scan < loceol && isSPACE(*scan))
4049 PL_reg_flags |= RF_tainted;
4052 while (hardcount < max && scan < loceol &&
4053 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4054 scan += UTF8SKIP(scan);
4058 while (scan < loceol && isSPACE_LC(*scan))
4065 LOAD_UTF8_CHARCLASS(space," ");
4066 while (hardcount < max && scan < loceol &&
4068 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4069 scan += UTF8SKIP(scan);
4073 while (scan < loceol && !isSPACE(*scan))
4078 PL_reg_flags |= RF_tainted;
4081 while (hardcount < max && scan < loceol &&
4082 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4083 scan += UTF8SKIP(scan);
4087 while (scan < loceol && !isSPACE_LC(*scan))
4094 LOAD_UTF8_CHARCLASS(digit,"0");
4095 while (hardcount < max && scan < loceol &&
4096 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4097 scan += UTF8SKIP(scan);
4101 while (scan < loceol && isDIGIT(*scan))
4108 LOAD_UTF8_CHARCLASS(digit,"0");
4109 while (hardcount < max && scan < loceol &&
4110 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4111 scan += UTF8SKIP(scan);
4115 while (scan < loceol && !isDIGIT(*scan))
4119 default: /* Called on something of 0 width. */
4120 break; /* So match right here or not at all. */
4126 c = scan - PL_reginput;
4131 SV *prop = sv_newmortal();
4134 PerlIO_printf(Perl_debug_log,
4135 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4136 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4143 - regrepeat_hard - repeatedly match something, report total lenth and length
4145 * The repeater is supposed to have constant length.
4149 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4151 register char *scan = Nullch;
4152 register char *start;
4153 register char *loceol = PL_regeol;
4155 I32 count = 0, res = 1;
4160 start = PL_reginput;
4161 if (PL_reg_match_utf8) {
4162 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4165 while (start < PL_reginput) {
4167 start += UTF8SKIP(start);
4178 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4180 *lp = l = PL_reginput - start;
4181 if (max != REG_INFTY && l*max < loceol - scan)
4182 loceol = scan + l*max;
4195 - regclass_swash - prepare the utf8 swash
4199 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** initsvp)
4204 if (PL_regdata && PL_regdata->count) {
4207 if (PL_regdata->what[n] == 's') {
4208 SV *rv = (SV*)PL_regdata->data[n];
4209 AV *av = (AV*)SvRV((SV*)rv);
4212 si = *av_fetch(av, 0, FALSE);
4213 a = av_fetch(av, 1, FALSE);
4217 else if (si && doinit) {
4218 sw = swash_init("utf8", "", si, 1, 0);
4219 (void)av_store(av, 1, sw);
4231 - reginclass - determine if a character falls into a character class
4235 S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8)
4237 char flags = ANYOF_FLAGS(n);
4242 c = do_utf8 ? utf8_to_uvchr(p, &len) : *p;
4244 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4245 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4246 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4249 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4252 SV *sw = regclass_swash(n, TRUE, 0);
4255 if (swash_fetch(sw, p, do_utf8))
4257 else if (flags & ANYOF_FOLD) {
4259 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
4261 to_utf8_fold(p, tmpbuf, &ulen);
4262 if (swash_fetch(sw, tmpbuf, do_utf8))
4264 to_utf8_upper(p, tmpbuf, &ulen);
4265 if (swash_fetch(sw, tmpbuf, do_utf8))
4271 if (!match && c < 256) {
4272 if (ANYOF_BITMAP_TEST(n, c))
4274 else if (flags & ANYOF_FOLD) {
4277 if (flags & ANYOF_LOCALE) {
4278 PL_reg_flags |= RF_tainted;
4279 f = PL_fold_locale[c];
4283 if (f != c && ANYOF_BITMAP_TEST(n, f))
4287 if (!match && (flags & ANYOF_CLASS)) {
4288 PL_reg_flags |= RF_tainted;
4290 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4291 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4292 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4293 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4294 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4295 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4296 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4297 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4298 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4299 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4300 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4301 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4302 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4303 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4304 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4305 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4306 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4307 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4308 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4309 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4310 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4311 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4312 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4313 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4314 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4315 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4316 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4317 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4318 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4319 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4320 ) /* How's that for a conditional? */
4327 return (flags & ANYOF_INVERT) ? !match : match;
4331 S_reghop(pTHX_ U8 *s, I32 off)
4333 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4337 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4340 while (off-- && s < lim) {
4341 /* XXX could check well-formedness here */
4349 if (UTF8_IS_CONTINUED(*s)) {
4350 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4353 /* XXX could check well-formedness here */
4361 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4363 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4367 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4370 while (off-- && s < lim) {
4371 /* XXX could check well-formedness here */
4381 if (UTF8_IS_CONTINUED(*s)) {
4382 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4385 /* XXX could check well-formedness here */
4397 restore_pos(pTHX_ void *arg)
4399 if (PL_reg_eval_set) {
4400 if (PL_reg_oldsaved) {
4401 PL_reg_re->subbeg = PL_reg_oldsaved;
4402 PL_reg_re->sublen = PL_reg_oldsavedlen;
4403 RX_MATCH_COPIED_on(PL_reg_re);
4405 PL_reg_magic->mg_len = PL_reg_oldpos;
4406 PL_reg_eval_set = 0;
4407 PL_curpm = PL_reg_oldcurpm;