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 /* The ibcmp_utf8() uses to_uni_fold() which is more
971 * correct folding for Unicode than using lowercase.
972 * However, it doesn't work quite fully since the folding
973 * is a one-to-many mapping and the regex optimizer is
974 * unaware of this, so it may throw out good matches.
975 * Fortunately, not getting this right is allowed
976 * for Unicode Regular Expression Support level 1,
977 * only one-to-one matching is required. --jhi */
980 if ( utf8_to_uvchr((U8*)s, &len) == c1
982 !ibcmp_utf8(s, do_utf8, (I32)(strend - s),
984 && (norun || regtry(prog, s)) )
991 U8 tmpbuf [UTF8_MAXLEN+1];
992 U8 foldbuf[UTF8_MAXLEN_FOLD+1];
994 UV c = utf8_to_uvchr((U8*)s, &len);
997 uvchr_to_utf8(tmpbuf, c);
998 to_utf8_fold(tmpbuf, foldbuf, &foldlen);
999 f = utf8_to_uvchr(foldbuf, 0);
1001 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1002 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1003 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1004 if ( (c == c1 || c == c2 || f == c1 || f == c2)
1005 && (ln == len || ln == foldlen ||
1006 !ibcmp_utf8(s, do_utf8, (I32)(strend - s),
1008 && (norun || regtry(prog, s)) )
1018 && (ln == 1 || !(OP(c) == EXACTF
1020 : ibcmp_locale(s, m, ln)))
1021 && (norun || regtry(prog, s)) )
1027 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1028 && (ln == 1 || !(OP(c) == EXACTF
1030 : ibcmp_locale(s, m, ln)))
1031 && (norun || regtry(prog, s)) )
1038 PL_reg_flags |= RF_tainted;
1045 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1048 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1050 tmp = ((OP(c) == BOUND ?
1051 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1052 LOAD_UTF8_CHARCLASS(alnum,"a");
1053 while (s < strend) {
1054 if (tmp == !(OP(c) == BOUND ?
1055 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1056 isALNUM_LC_utf8((U8*)s)))
1059 if ((norun || regtry(prog, s)))
1066 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1067 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1068 while (s < strend) {
1070 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1072 if ((norun || regtry(prog, s)))
1078 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1082 PL_reg_flags |= RF_tainted;
1089 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1092 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1094 tmp = ((OP(c) == NBOUND ?
1095 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1096 LOAD_UTF8_CHARCLASS(alnum,"a");
1097 while (s < strend) {
1098 if (tmp == !(OP(c) == NBOUND ?
1099 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1100 isALNUM_LC_utf8((U8*)s)))
1102 else if ((norun || regtry(prog, s)))
1108 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1109 tmp = ((OP(c) == NBOUND ?
1110 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1111 while (s < strend) {
1113 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1115 else if ((norun || regtry(prog, s)))
1120 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1125 LOAD_UTF8_CHARCLASS(alnum,"a");
1126 while (s < strend) {
1127 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1128 if (tmp && (norun || regtry(prog, s)))
1139 while (s < strend) {
1141 if (tmp && (norun || regtry(prog, s)))
1153 PL_reg_flags |= RF_tainted;
1155 while (s < strend) {
1156 if (isALNUM_LC_utf8((U8*)s)) {
1157 if (tmp && (norun || regtry(prog, s)))
1168 while (s < strend) {
1169 if (isALNUM_LC(*s)) {
1170 if (tmp && (norun || regtry(prog, s)))
1183 LOAD_UTF8_CHARCLASS(alnum,"a");
1184 while (s < strend) {
1185 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1186 if (tmp && (norun || regtry(prog, s)))
1197 while (s < strend) {
1199 if (tmp && (norun || regtry(prog, s)))
1211 PL_reg_flags |= RF_tainted;
1213 while (s < strend) {
1214 if (!isALNUM_LC_utf8((U8*)s)) {
1215 if (tmp && (norun || regtry(prog, s)))
1226 while (s < strend) {
1227 if (!isALNUM_LC(*s)) {
1228 if (tmp && (norun || regtry(prog, s)))
1241 LOAD_UTF8_CHARCLASS(space," ");
1242 while (s < strend) {
1243 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1244 if (tmp && (norun || regtry(prog, s)))
1255 while (s < strend) {
1257 if (tmp && (norun || regtry(prog, s)))
1269 PL_reg_flags |= RF_tainted;
1271 while (s < strend) {
1272 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1273 if (tmp && (norun || regtry(prog, s)))
1284 while (s < strend) {
1285 if (isSPACE_LC(*s)) {
1286 if (tmp && (norun || regtry(prog, s)))
1299 LOAD_UTF8_CHARCLASS(space," ");
1300 while (s < strend) {
1301 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1302 if (tmp && (norun || regtry(prog, s)))
1313 while (s < strend) {
1315 if (tmp && (norun || regtry(prog, s)))
1327 PL_reg_flags |= RF_tainted;
1329 while (s < strend) {
1330 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1331 if (tmp && (norun || regtry(prog, s)))
1342 while (s < strend) {
1343 if (!isSPACE_LC(*s)) {
1344 if (tmp && (norun || regtry(prog, s)))
1357 LOAD_UTF8_CHARCLASS(digit,"0");
1358 while (s < strend) {
1359 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1360 if (tmp && (norun || regtry(prog, s)))
1371 while (s < strend) {
1373 if (tmp && (norun || regtry(prog, s)))
1385 PL_reg_flags |= RF_tainted;
1387 while (s < strend) {
1388 if (isDIGIT_LC_utf8((U8*)s)) {
1389 if (tmp && (norun || regtry(prog, s)))
1400 while (s < strend) {
1401 if (isDIGIT_LC(*s)) {
1402 if (tmp && (norun || regtry(prog, s)))
1415 LOAD_UTF8_CHARCLASS(digit,"0");
1416 while (s < strend) {
1417 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1418 if (tmp && (norun || regtry(prog, s)))
1429 while (s < strend) {
1431 if (tmp && (norun || regtry(prog, s)))
1443 PL_reg_flags |= RF_tainted;
1445 while (s < strend) {
1446 if (!isDIGIT_LC_utf8((U8*)s)) {
1447 if (tmp && (norun || regtry(prog, s)))
1458 while (s < strend) {
1459 if (!isDIGIT_LC(*s)) {
1460 if (tmp && (norun || regtry(prog, s)))
1472 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1481 - regexec_flags - match a regexp against a string
1484 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1485 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1486 /* strend: pointer to null at end of string */
1487 /* strbeg: real beginning of string */
1488 /* minend: end of match must be >=minend after stringarg. */
1489 /* data: May be used for some additional optimizations. */
1490 /* nosave: For optimizations. */
1493 register regnode *c;
1494 register char *startpos = stringarg;
1495 I32 minlen; /* must match at least this many chars */
1496 I32 dontbother = 0; /* how many characters not to try at end */
1497 /* I32 start_shift = 0; */ /* Offset of the start to find
1498 constant substr. */ /* CC */
1499 I32 end_shift = 0; /* Same for the end. */ /* CC */
1500 I32 scream_pos = -1; /* Internal iterator of scream. */
1502 SV* oreplsv = GvSV(PL_replgv);
1503 bool do_utf8 = DO_UTF8(sv);
1505 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
1512 PL_regnarrate = DEBUG_r_TEST;
1515 /* Be paranoid... */
1516 if (prog == NULL || startpos == NULL) {
1517 Perl_croak(aTHX_ "NULL regexp parameter");
1521 minlen = prog->minlen;
1522 if (strend - startpos < minlen) {
1523 DEBUG_r(PerlIO_printf(Perl_debug_log,
1524 "String too short [regexec_flags]...\n"));
1528 /* Check validity of program. */
1529 if (UCHARAT(prog->program) != REG_MAGIC) {
1530 Perl_croak(aTHX_ "corrupted regexp program");
1534 PL_reg_eval_set = 0;
1537 if (prog->reganch & ROPT_UTF8)
1538 PL_reg_flags |= RF_utf8;
1540 /* Mark beginning of line for ^ and lookbehind. */
1541 PL_regbol = startpos;
1545 /* Mark end of line for $ (and such) */
1548 /* see how far we have to get to not match where we matched before */
1549 PL_regtill = startpos+minend;
1551 /* We start without call_cc context. */
1554 /* If there is a "must appear" string, look for it. */
1557 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1560 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1561 PL_reg_ganch = startpos;
1562 else if (sv && SvTYPE(sv) >= SVt_PVMG
1564 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1565 && mg->mg_len >= 0) {
1566 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1567 if (prog->reganch & ROPT_ANCH_GPOS) {
1568 if (s > PL_reg_ganch)
1573 else /* pos() not defined */
1574 PL_reg_ganch = strbeg;
1577 if (do_utf8 == (UTF!=0) &&
1578 !(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
1579 re_scream_pos_data d;
1581 d.scream_olds = &scream_olds;
1582 d.scream_pos = &scream_pos;
1583 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1585 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1586 goto phooey; /* not present */
1591 char *s = do_utf8 ? sv_uni_display(dsv, sv, 60, 0) : startpos;
1592 int len = do_utf8 ? strlen(s) : strend - startpos;
1595 PerlIO_printf(Perl_debug_log,
1596 "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
1597 PL_colors[4],PL_colors[5],PL_colors[0],
1600 (strlen(prog->precomp) > 60 ? "..." : ""),
1602 (int)(len > 60 ? 60 : len),
1604 (len > 60 ? "..." : "")
1608 /* Simplest case: anchored match need be tried only once. */
1609 /* [unless only anchor is BOL and multiline is set] */
1610 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1611 if (s == startpos && regtry(prog, startpos))
1613 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1614 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1619 dontbother = minlen - 1;
1620 end = HOP3c(strend, -dontbother, strbeg) - 1;
1621 /* for multiline we only have to try after newlines */
1622 if (prog->check_substr) {
1626 if (regtry(prog, s))
1631 if (prog->reganch & RE_USE_INTUIT) {
1632 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1643 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1644 if (regtry(prog, s))
1651 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1652 if (regtry(prog, PL_reg_ganch))
1657 /* Messy cases: unanchored match. */
1658 if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
1659 /* we have /x+whatever/ */
1660 /* it must be a one character string (XXXX Except UTF?) */
1661 char ch = SvPVX(prog->anchored_substr)[0];
1667 while (s < strend) {
1669 DEBUG_r( did_match = 1 );
1670 if (regtry(prog, s)) goto got_it;
1672 while (s < strend && *s == ch)
1679 while (s < strend) {
1681 DEBUG_r( did_match = 1 );
1682 if (regtry(prog, s)) goto got_it;
1684 while (s < strend && *s == ch)
1690 DEBUG_r(if (!did_match)
1691 PerlIO_printf(Perl_debug_log,
1692 "Did not find anchored character...\n")
1696 else if (do_utf8 == (UTF!=0) &&
1697 (prog->anchored_substr != Nullsv
1698 || (prog->float_substr != Nullsv
1699 && prog->float_max_offset < strend - s))) {
1700 SV *must = prog->anchored_substr
1701 ? prog->anchored_substr : prog->float_substr;
1703 prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset;
1705 prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
1706 char *last = HOP3c(strend, /* Cannot start after this */
1707 -(I32)(CHR_SVLEN(must)
1708 - (SvTAIL(must) != 0) + back_min), strbeg);
1709 char *last1; /* Last position checked before */
1715 last1 = HOPc(s, -1);
1717 last1 = s - 1; /* bogus */
1719 /* XXXX check_substr already used to find `s', can optimize if
1720 check_substr==must. */
1722 dontbother = end_shift;
1723 strend = HOPc(strend, -dontbother);
1724 while ( (s <= last) &&
1725 ((flags & REXEC_SCREAM)
1726 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1727 end_shift, &scream_pos, 0))
1728 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1729 (unsigned char*)strend, must,
1730 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1731 DEBUG_r( did_match = 1 );
1732 if (HOPc(s, -back_max) > last1) {
1733 last1 = HOPc(s, -back_min);
1734 s = HOPc(s, -back_max);
1737 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1739 last1 = HOPc(s, -back_min);
1743 while (s <= last1) {
1744 if (regtry(prog, s))
1750 while (s <= last1) {
1751 if (regtry(prog, s))
1757 DEBUG_r(if (!did_match)
1758 PerlIO_printf(Perl_debug_log,
1759 "Did not find %s substr `%s%.*s%s'%s...\n",
1760 ((must == prog->anchored_substr)
1761 ? "anchored" : "floating"),
1763 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1765 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1769 else if ((c = prog->regstclass)) {
1770 if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT)
1771 /* don't bother with what can't match */
1772 strend = HOPc(strend, -(minlen - 1));
1774 SV *prop = sv_newmortal();
1776 PerlIO_printf(Perl_debug_log, "Matching stclass `%s' against `%s'\n", SvPVX(prop), UTF ? sv_uni_display(dsv, sv, 60, 0) : s);
1778 if (find_byclass(prog, c, s, strend, startpos, 0))
1780 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1784 if (prog->float_substr != Nullsv) { /* Trim the end. */
1787 if (flags & REXEC_SCREAM) {
1788 last = screaminstr(sv, prog->float_substr, s - strbeg,
1789 end_shift, &scream_pos, 1); /* last one */
1791 last = scream_olds; /* Only one occurrence. */
1795 char *little = SvPV(prog->float_substr, len);
1797 if (SvTAIL(prog->float_substr)) {
1798 if (memEQ(strend - len + 1, little, len - 1))
1799 last = strend - len + 1;
1800 else if (!PL_multiline)
1801 last = memEQ(strend - len, little, len)
1802 ? strend - len : Nullch;
1808 last = rninstr(s, strend, little, little + len);
1810 last = strend; /* matching `$' */
1814 DEBUG_r(PerlIO_printf(Perl_debug_log,
1815 "%sCan't trim the tail, match fails (should not happen)%s\n",
1816 PL_colors[4],PL_colors[5]));
1817 goto phooey; /* Should not happen! */
1819 dontbother = strend - last + prog->float_min_offset;
1821 if (minlen && (dontbother < minlen))
1822 dontbother = minlen - 1;
1823 strend -= dontbother; /* this one's always in bytes! */
1824 /* We don't know much -- general case. */
1827 if (regtry(prog, s))
1836 if (regtry(prog, s))
1838 } while (s++ < strend);
1846 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1848 if (PL_reg_eval_set) {
1849 /* Preserve the current value of $^R */
1850 if (oreplsv != GvSV(PL_replgv))
1851 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1852 restored, the value remains
1854 restore_pos(aTHX_ 0);
1857 /* make sure $`, $&, $', and $digit will work later */
1858 if ( !(flags & REXEC_NOT_FIRST) ) {
1859 if (RX_MATCH_COPIED(prog)) {
1860 Safefree(prog->subbeg);
1861 RX_MATCH_COPIED_off(prog);
1863 if (flags & REXEC_COPY_STR) {
1864 I32 i = PL_regeol - startpos + (stringarg - strbeg);
1866 s = savepvn(strbeg, i);
1869 RX_MATCH_COPIED_on(prog);
1872 prog->subbeg = strbeg;
1873 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
1880 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
1881 PL_colors[4],PL_colors[5]));
1882 if (PL_reg_eval_set)
1883 restore_pos(aTHX_ 0);
1888 - regtry - try match at specific point
1890 STATIC I32 /* 0 failure, 1 success */
1891 S_regtry(pTHX_ regexp *prog, char *startpos)
1899 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
1901 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
1904 PL_reg_eval_set = RS_init;
1906 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
1907 (IV)(PL_stack_sp - PL_stack_base));
1909 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
1910 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
1911 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
1913 /* Apparently this is not needed, judging by wantarray. */
1914 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
1915 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
1918 /* Make $_ available to executed code. */
1919 if (PL_reg_sv != DEFSV) {
1920 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
1925 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
1926 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
1927 /* prepare for quick setting of pos */
1928 sv_magic(PL_reg_sv, (SV*)0,
1929 PERL_MAGIC_regex_global, Nullch, 0);
1930 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
1934 PL_reg_oldpos = mg->mg_len;
1935 SAVEDESTRUCTOR_X(restore_pos, 0);
1937 if (!PL_reg_curpm) {
1938 Newz(22,PL_reg_curpm, 1, PMOP);
1941 SV* repointer = newSViv(0);
1942 /* so we know which PL_regex_padav element is PL_reg_curpm */
1943 SvFLAGS(repointer) |= SVf_BREAK;
1944 av_push(PL_regex_padav,repointer);
1945 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
1946 PL_regex_pad = AvARRAY(PL_regex_padav);
1950 PM_SETRE(PL_reg_curpm, prog);
1951 PL_reg_oldcurpm = PL_curpm;
1952 PL_curpm = PL_reg_curpm;
1953 if (RX_MATCH_COPIED(prog)) {
1954 /* Here is a serious problem: we cannot rewrite subbeg,
1955 since it may be needed if this match fails. Thus
1956 $` inside (?{}) could fail... */
1957 PL_reg_oldsaved = prog->subbeg;
1958 PL_reg_oldsavedlen = prog->sublen;
1959 RX_MATCH_COPIED_off(prog);
1962 PL_reg_oldsaved = Nullch;
1963 prog->subbeg = PL_bostr;
1964 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
1966 prog->startp[0] = startpos - PL_bostr;
1967 PL_reginput = startpos;
1968 PL_regstartp = prog->startp;
1969 PL_regendp = prog->endp;
1970 PL_reglastparen = &prog->lastparen;
1971 PL_reglastcloseparen = &prog->lastcloseparen;
1972 prog->lastparen = 0;
1974 DEBUG_r(PL_reg_starttry = startpos);
1975 if (PL_reg_start_tmpl <= prog->nparens) {
1976 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
1977 if(PL_reg_start_tmp)
1978 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1980 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1984 sv_setpvn(PERL_DEBUG_PAD(0), "", 0);
1985 sv_setpvn(PERL_DEBUG_PAD(1), "", 0);
1986 sv_setpvn(PERL_DEBUG_PAD(2), "", 0);
1989 /* XXXX What this code is doing here?!!! There should be no need
1990 to do this again and again, PL_reglastparen should take care of
1993 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
1994 * Actually, the code in regcppop() (which Ilya may be meaning by
1995 * PL_reglastparen), is not needed at all by the test suite
1996 * (op/regexp, op/pat, op/split), but that code is needed, oddly
1997 * enough, for building DynaLoader, or otherwise this
1998 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
1999 * will happen. Meanwhile, this code *is* needed for the
2000 * above-mentioned test suite tests to succeed. The common theme
2001 * on those tests seems to be returning null fields from matches.
2006 if (prog->nparens) {
2007 for (i = prog->nparens; i > *PL_reglastparen; i--) {
2014 if (regmatch(prog->program + 1)) {
2015 prog->endp[0] = PL_reginput - PL_bostr;
2018 REGCP_UNWIND(lastcp);
2022 #define RE_UNWIND_BRANCH 1
2023 #define RE_UNWIND_BRANCHJ 2
2027 typedef struct { /* XX: makes sense to enlarge it... */
2031 } re_unwind_generic_t;
2044 } re_unwind_branch_t;
2046 typedef union re_unwind_t {
2048 re_unwind_generic_t generic;
2049 re_unwind_branch_t branch;
2052 #define sayYES goto yes
2053 #define sayNO goto no
2054 #define sayYES_FINAL goto yes_final
2055 #define sayYES_LOUD goto yes_loud
2056 #define sayNO_FINAL goto no_final
2057 #define sayNO_SILENT goto do_no
2058 #define saySAME(x) if (x) goto yes; else goto no
2060 #define REPORT_CODE_OFF 24
2063 - regmatch - main matching routine
2065 * Conceptually the strategy is simple: check to see whether the current
2066 * node matches, call self recursively to see whether the rest matches,
2067 * and then act accordingly. In practice we make some effort to avoid
2068 * recursion, in particular by going through "ordinary" nodes (that don't
2069 * need to know whether the rest of the match failed) by a loop instead of
2072 /* [lwall] I've hoisted the register declarations to the outer block in order to
2073 * maybe save a little bit of pushing and popping on the stack. It also takes
2074 * advantage of machines that use a register save mask on subroutine entry.
2076 STATIC I32 /* 0 failure, 1 success */
2077 S_regmatch(pTHX_ regnode *prog)
2079 register regnode *scan; /* Current node. */
2080 regnode *next; /* Next node. */
2081 regnode *inner; /* Next node in internal branch. */
2082 register I32 nextchr; /* renamed nextchr - nextchar colides with
2083 function of same name */
2084 register I32 n; /* no or next */
2085 register I32 ln = 0; /* len or last */
2086 register char *s = Nullch; /* operand or save */
2087 register char *locinput = PL_reginput;
2088 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2089 int minmod = 0, sw = 0, logical = 0;
2092 I32 firstcp = PL_savestack_ix;
2094 register bool do_utf8 = PL_reg_match_utf8;
2096 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2097 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2098 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2105 /* Note that nextchr is a byte even in UTF */
2106 nextchr = UCHARAT(locinput);
2108 while (scan != NULL) {
2111 SV *prop = sv_newmortal();
2112 int docolor = *PL_colors[0];
2113 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2114 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2115 /* The part of the string before starttry has one color
2116 (pref0_len chars), between starttry and current
2117 position another one (pref_len - pref0_len chars),
2118 after the current position the third one.
2119 We assume that pref0_len <= pref_len, otherwise we
2120 decrease pref0_len. */
2121 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2122 ? (5 + taill) - l : locinput - PL_bostr;
2125 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2127 pref0_len = pref_len - (locinput - PL_reg_starttry);
2128 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2129 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2130 ? (5 + taill) - pref_len : PL_regeol - locinput);
2131 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2135 if (pref0_len > pref_len)
2136 pref0_len = pref_len;
2137 regprop(prop, scan);
2141 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2143 locinput - pref_len;
2144 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2145 char *s1 = do_utf8 ?
2146 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2147 pref_len - pref0_len, 60, 0) :
2148 locinput - pref_len + pref0_len;
2149 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2150 char *s2 = do_utf8 ?
2151 pv_uni_display(dsv2, (U8*)locinput,
2152 PL_regeol - locinput, 60, 0) :
2154 int len2 = do_utf8 ? strlen(s2) : l;
2155 PerlIO_printf(Perl_debug_log,
2156 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2157 (IV)(locinput - PL_bostr),
2164 (docolor ? "" : "> <"),
2168 15 - l - pref_len + 1,
2170 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2175 next = scan + NEXT_OFF(scan);
2181 if (locinput == PL_bostr || (PL_multiline &&
2182 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2184 /* regtill = regbol; */
2189 if (locinput == PL_bostr ||
2190 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2196 if (locinput == PL_bostr)
2200 if (locinput == PL_reg_ganch)
2210 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2215 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2217 if (PL_regeol - locinput > 1)
2221 if (PL_regeol != locinput)
2225 if (!nextchr && locinput >= PL_regeol)
2228 locinput += PL_utf8skip[nextchr];
2229 if (locinput > PL_regeol)
2231 nextchr = UCHARAT(locinput);
2234 nextchr = UCHARAT(++locinput);
2237 if (!nextchr && locinput >= PL_regeol)
2239 nextchr = UCHARAT(++locinput);
2242 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2245 locinput += PL_utf8skip[nextchr];
2246 if (locinput > PL_regeol)
2248 nextchr = UCHARAT(locinput);
2251 nextchr = UCHARAT(++locinput);
2256 if (do_utf8 != (UTF!=0)) {
2257 /* The target and the pattern have differing utf8ness. */
2263 /* The target is utf8, the pattern is not utf8. */
2267 if (NATIVE_TO_UNI(*(U8*)s) !=
2268 utf8_to_uvchr((U8*)l, &ulen))
2275 /* The target is not utf8, the pattern is utf8. */
2279 if (NATIVE_TO_UNI(*((U8*)l)) !=
2280 utf8_to_uvchr((U8*)s, &ulen))
2287 nextchr = UCHARAT(locinput);
2290 /* The target and the pattern have the same utf8ness. */
2291 /* Inline the first character, for speed. */
2292 if (UCHARAT(s) != nextchr)
2294 if (PL_regeol - locinput < ln)
2296 if (ln > 1 && memNE(s, locinput, ln))
2299 nextchr = UCHARAT(locinput);
2302 PL_reg_flags |= RF_tainted;
2311 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
2313 if (do_utf8 != (UTF!=0)) {
2314 /* The target and the pattern have differing utf8ness. */
2315 STRLEN ulen1, ulen2;
2319 /* The target is utf8, the pattern is not utf8. */
2324 cs = to_uni_fold(NATIVE_TO_UNI(*(U8*)s),
2326 cl = utf8_to_uvchr((U8*)l, &ulen2);
2329 cl = to_uni_fold(cl, (U8*)l, &ulen2);
2330 if (ulen1 != ulen2 || cs != cl)
2338 /* The target is not utf8, the pattern is utf8. */
2343 cs = utf8_to_uvchr((U8*)s, &ulen1);
2345 cl = to_uni_fold(NATIVE_TO_UNI(*(U8*)l),
2349 cs = to_uni_fold(cs, (U8*)s, &ulen1);
2350 if (ulen1 != ulen2 || cs != cl)
2358 nextchr = UCHARAT(locinput);
2362 if (do_utf8 && UTF) {
2363 /* Both the target and the pattern are utf8. */
2369 if (UTF8SKIP(s) != UTF8SKIP(l) ||
2370 memNE(s, (char*)l, UTF8SKIP(s))) {
2371 to_utf8_fold((U8*)l, tmpbuf, &ulen);
2372 if (UTF8SKIP(s) != ulen ||
2373 memNE(s, (char*)tmpbuf, ulen))
2380 nextchr = UCHARAT(locinput);
2385 /* Neither the target and the pattern are utf8. */
2387 /* Inline the first character, for speed. */
2388 if (UCHARAT(s) != nextchr &&
2389 UCHARAT(s) != ((OP(scan) == EXACTF)
2390 ? PL_fold : PL_fold_locale)[nextchr])
2392 if (PL_regeol - locinput < ln)
2394 if (ln > 1 && (OP(scan) == EXACTF
2395 ? ibcmp(s, locinput, ln)
2396 : ibcmp_locale(s, locinput, ln)))
2399 nextchr = UCHARAT(locinput);
2403 if (!reginclass(scan, (U8*)locinput, do_utf8))
2405 if (locinput >= PL_regeol)
2407 locinput += PL_utf8skip[nextchr];
2408 nextchr = UCHARAT(locinput);
2412 nextchr = UCHARAT(locinput);
2413 if (!reginclass(scan, (U8*)locinput, do_utf8))
2415 if (!nextchr && locinput >= PL_regeol)
2417 nextchr = UCHARAT(++locinput);
2421 PL_reg_flags |= RF_tainted;
2427 LOAD_UTF8_CHARCLASS(alnum,"a");
2428 if (!(OP(scan) == ALNUM
2429 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2430 : isALNUM_LC_utf8((U8*)locinput)))
2434 locinput += PL_utf8skip[nextchr];
2435 nextchr = UCHARAT(locinput);
2438 if (!(OP(scan) == ALNUM
2439 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2441 nextchr = UCHARAT(++locinput);
2444 PL_reg_flags |= RF_tainted;
2447 if (!nextchr && locinput >= PL_regeol)
2450 LOAD_UTF8_CHARCLASS(alnum,"a");
2451 if (OP(scan) == NALNUM
2452 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2453 : isALNUM_LC_utf8((U8*)locinput))
2457 locinput += PL_utf8skip[nextchr];
2458 nextchr = UCHARAT(locinput);
2461 if (OP(scan) == NALNUM
2462 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2464 nextchr = UCHARAT(++locinput);
2468 PL_reg_flags |= RF_tainted;
2472 /* was last char in word? */
2474 if (locinput == PL_bostr)
2477 U8 *r = reghop((U8*)locinput, -1);
2479 ln = utf8n_to_uvchr(r, s - (char*)r, 0, 0);
2481 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2482 ln = isALNUM_uni(ln);
2483 LOAD_UTF8_CHARCLASS(alnum,"a");
2484 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2487 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2488 n = isALNUM_LC_utf8((U8*)locinput);
2492 ln = (locinput != PL_bostr) ?
2493 UCHARAT(locinput - 1) : '\n';
2494 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2496 n = isALNUM(nextchr);
2499 ln = isALNUM_LC(ln);
2500 n = isALNUM_LC(nextchr);
2503 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2504 OP(scan) == BOUNDL))
2508 PL_reg_flags |= RF_tainted;
2514 if (UTF8_IS_CONTINUED(nextchr)) {
2515 LOAD_UTF8_CHARCLASS(space," ");
2516 if (!(OP(scan) == SPACE
2517 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2518 : isSPACE_LC_utf8((U8*)locinput)))
2522 locinput += PL_utf8skip[nextchr];
2523 nextchr = UCHARAT(locinput);
2526 if (!(OP(scan) == SPACE
2527 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2529 nextchr = UCHARAT(++locinput);
2532 if (!(OP(scan) == SPACE
2533 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2535 nextchr = UCHARAT(++locinput);
2539 PL_reg_flags |= RF_tainted;
2542 if (!nextchr && locinput >= PL_regeol)
2545 LOAD_UTF8_CHARCLASS(space," ");
2546 if (OP(scan) == NSPACE
2547 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2548 : isSPACE_LC_utf8((U8*)locinput))
2552 locinput += PL_utf8skip[nextchr];
2553 nextchr = UCHARAT(locinput);
2556 if (OP(scan) == NSPACE
2557 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2559 nextchr = UCHARAT(++locinput);
2562 PL_reg_flags |= RF_tainted;
2568 LOAD_UTF8_CHARCLASS(digit,"0");
2569 if (!(OP(scan) == DIGIT
2570 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2571 : isDIGIT_LC_utf8((U8*)locinput)))
2575 locinput += PL_utf8skip[nextchr];
2576 nextchr = UCHARAT(locinput);
2579 if (!(OP(scan) == DIGIT
2580 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2582 nextchr = UCHARAT(++locinput);
2585 PL_reg_flags |= RF_tainted;
2588 if (!nextchr && locinput >= PL_regeol)
2591 LOAD_UTF8_CHARCLASS(digit,"0");
2592 if (OP(scan) == NDIGIT
2593 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2594 : isDIGIT_LC_utf8((U8*)locinput))
2598 locinput += PL_utf8skip[nextchr];
2599 nextchr = UCHARAT(locinput);
2602 if (OP(scan) == NDIGIT
2603 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2605 nextchr = UCHARAT(++locinput);
2608 if (locinput >= PL_regeol)
2611 LOAD_UTF8_CHARCLASS(mark,"~");
2612 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2614 locinput += PL_utf8skip[nextchr];
2615 while (locinput < PL_regeol &&
2616 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2617 locinput += UTF8SKIP(locinput);
2618 if (locinput > PL_regeol)
2623 nextchr = UCHARAT(locinput);
2626 PL_reg_flags |= RF_tainted;
2630 n = ARG(scan); /* which paren pair */
2631 ln = PL_regstartp[n];
2632 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2633 if (*PL_reglastparen < n || ln == -1)
2634 sayNO; /* Do not match unless seen CLOSEn. */
2635 if (ln == PL_regendp[n])
2639 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2641 char *e = PL_bostr + PL_regendp[n];
2643 * Note that we can't do the "other character" lookup trick as
2644 * in the 8-bit case (no pun intended) because in Unicode we
2645 * have to map both upper and title case to lower case.
2647 if (OP(scan) == REFF) {
2648 STRLEN ulen1, ulen2;
2649 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2650 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2654 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2655 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2656 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2663 nextchr = UCHARAT(locinput);
2667 /* Inline the first character, for speed. */
2668 if (UCHARAT(s) != nextchr &&
2670 (UCHARAT(s) != ((OP(scan) == REFF
2671 ? PL_fold : PL_fold_locale)[nextchr]))))
2673 ln = PL_regendp[n] - ln;
2674 if (locinput + ln > PL_regeol)
2676 if (ln > 1 && (OP(scan) == REF
2677 ? memNE(s, locinput, ln)
2679 ? ibcmp(s, locinput, ln)
2680 : ibcmp_locale(s, locinput, ln))))
2683 nextchr = UCHARAT(locinput);
2694 OP_4tree *oop = PL_op;
2695 COP *ocurcop = PL_curcop;
2696 SV **ocurpad = PL_curpad;
2700 PL_op = (OP_4tree*)PL_regdata->data[n];
2701 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2702 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2703 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2707 CALLRUNOPS(aTHX); /* Scalar context. */
2710 ret = Nullsv; /* protect against empty (?{}) blocks. */
2718 PL_curpad = ocurpad;
2719 PL_curcop = ocurcop;
2721 if (logical == 2) { /* Postponed subexpression. */
2723 MAGIC *mg = Null(MAGIC*);
2725 CHECKPOINT cp, lastcp;
2727 if(SvROK(ret) || SvRMAGICAL(ret)) {
2728 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2731 mg = mg_find(sv, PERL_MAGIC_qr);
2734 re = (regexp *)mg->mg_obj;
2735 (void)ReREFCNT_inc(re);
2739 char *t = SvPV(ret, len);
2741 char *oprecomp = PL_regprecomp;
2742 I32 osize = PL_regsize;
2743 I32 onpar = PL_regnpar;
2746 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2748 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2749 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2751 PL_regprecomp = oprecomp;
2756 PerlIO_printf(Perl_debug_log,
2757 "Entering embedded `%s%.60s%s%s'\n",
2761 (strlen(re->precomp) > 60 ? "..." : ""))
2764 state.prev = PL_reg_call_cc;
2765 state.cc = PL_regcc;
2766 state.re = PL_reg_re;
2770 cp = regcppush(0); /* Save *all* the positions. */
2773 state.ss = PL_savestack_ix;
2774 *PL_reglastparen = 0;
2775 *PL_reglastcloseparen = 0;
2776 PL_reg_call_cc = &state;
2777 PL_reginput = locinput;
2779 /* XXXX This is too dramatic a measure... */
2782 if (regmatch(re->program + 1)) {
2783 /* Even though we succeeded, we need to restore
2784 global variables, since we may be wrapped inside
2785 SUSPEND, thus the match may be not finished yet. */
2787 /* XXXX Do this only if SUSPENDed? */
2788 PL_reg_call_cc = state.prev;
2789 PL_regcc = state.cc;
2790 PL_reg_re = state.re;
2791 cache_re(PL_reg_re);
2793 /* XXXX This is too dramatic a measure... */
2796 /* These are needed even if not SUSPEND. */
2802 REGCP_UNWIND(lastcp);
2804 PL_reg_call_cc = state.prev;
2805 PL_regcc = state.cc;
2806 PL_reg_re = state.re;
2807 cache_re(PL_reg_re);
2809 /* XXXX This is too dramatic a measure... */
2819 sv_setsv(save_scalar(PL_replgv), ret);
2823 n = ARG(scan); /* which paren pair */
2824 PL_reg_start_tmp[n] = locinput;
2829 n = ARG(scan); /* which paren pair */
2830 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2831 PL_regendp[n] = locinput - PL_bostr;
2832 if (n > *PL_reglastparen)
2833 *PL_reglastparen = n;
2834 *PL_reglastcloseparen = n;
2837 n = ARG(scan); /* which paren pair */
2838 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2841 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2843 next = NEXTOPER(NEXTOPER(scan));
2845 next = scan + ARG(scan);
2846 if (OP(next) == IFTHEN) /* Fake one. */
2847 next = NEXTOPER(NEXTOPER(next));
2851 logical = scan->flags;
2853 /*******************************************************************
2854 PL_regcc contains infoblock about the innermost (...)* loop, and
2855 a pointer to the next outer infoblock.
2857 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2859 1) After matching X, regnode for CURLYX is processed;
2861 2) This regnode creates infoblock on the stack, and calls
2862 regmatch() recursively with the starting point at WHILEM node;
2864 3) Each hit of WHILEM node tries to match A and Z (in the order
2865 depending on the current iteration, min/max of {min,max} and
2866 greediness). The information about where are nodes for "A"
2867 and "Z" is read from the infoblock, as is info on how many times "A"
2868 was already matched, and greediness.
2870 4) After A matches, the same WHILEM node is hit again.
2872 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2873 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2874 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2875 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2876 of the external loop.
2878 Currently present infoblocks form a tree with a stem formed by PL_curcc
2879 and whatever it mentions via ->next, and additional attached trees
2880 corresponding to temporarily unset infoblocks as in "5" above.
2882 In the following picture infoblocks for outer loop of
2883 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2884 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2885 infoblocks are drawn below the "reset" infoblock.
2887 In fact in the picture below we do not show failed matches for Z and T
2888 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2889 more obvious *why* one needs to *temporary* unset infoblocks.]
2891 Matched REx position InfoBlocks Comment
2895 Y A)*?Z)*?T x <- O <- I
2896 YA )*?Z)*?T x <- O <- I
2897 YA A)*?Z)*?T x <- O <- I
2898 YAA )*?Z)*?T x <- O <- I
2899 YAA Z)*?T x <- O # Temporary unset I
2902 YAAZ Y(A)*?Z)*?T x <- O
2905 YAAZY (A)*?Z)*?T x <- O
2908 YAAZY A)*?Z)*?T x <- O <- I
2911 YAAZYA )*?Z)*?T x <- O <- I
2914 YAAZYA Z)*?T x <- O # Temporary unset I
2920 YAAZYAZ T x # Temporary unset O
2927 *******************************************************************/
2930 CHECKPOINT cp = PL_savestack_ix;
2931 /* No need to save/restore up to this paren */
2932 I32 parenfloor = scan->flags;
2934 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2936 cc.oldcc = PL_regcc;
2938 /* XXXX Probably it is better to teach regpush to support
2939 parenfloor > PL_regsize... */
2940 if (parenfloor > *PL_reglastparen)
2941 parenfloor = *PL_reglastparen; /* Pessimization... */
2942 cc.parenfloor = parenfloor;
2944 cc.min = ARG1(scan);
2945 cc.max = ARG2(scan);
2946 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2950 PL_reginput = locinput;
2951 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2953 PL_regcc = cc.oldcc;
2959 * This is really hard to understand, because after we match
2960 * what we're trying to match, we must make sure the rest of
2961 * the REx is going to match for sure, and to do that we have
2962 * to go back UP the parse tree by recursing ever deeper. And
2963 * if it fails, we have to reset our parent's current state
2964 * that we can try again after backing off.
2967 CHECKPOINT cp, lastcp;
2968 CURCUR* cc = PL_regcc;
2969 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2971 n = cc->cur + 1; /* how many we know we matched */
2972 PL_reginput = locinput;
2975 PerlIO_printf(Perl_debug_log,
2976 "%*s %ld out of %ld..%ld cc=%lx\n",
2977 REPORT_CODE_OFF+PL_regindent*2, "",
2978 (long)n, (long)cc->min,
2979 (long)cc->max, (long)cc)
2982 /* If degenerate scan matches "", assume scan done. */
2984 if (locinput == cc->lastloc && n >= cc->min) {
2985 PL_regcc = cc->oldcc;
2989 PerlIO_printf(Perl_debug_log,
2990 "%*s empty match detected, try continuation...\n",
2991 REPORT_CODE_OFF+PL_regindent*2, "")
2993 if (regmatch(cc->next))
3001 /* First just match a string of min scans. */
3005 cc->lastloc = locinput;
3006 if (regmatch(cc->scan))
3009 cc->lastloc = lastloc;
3014 /* Check whether we already were at this position.
3015 Postpone detection until we know the match is not
3016 *that* much linear. */
3017 if (!PL_reg_maxiter) {
3018 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3019 PL_reg_leftiter = PL_reg_maxiter;
3021 if (PL_reg_leftiter-- == 0) {
3022 I32 size = (PL_reg_maxiter + 7)/8;
3023 if (PL_reg_poscache) {
3024 if (PL_reg_poscache_size < size) {
3025 Renew(PL_reg_poscache, size, char);
3026 PL_reg_poscache_size = size;
3028 Zero(PL_reg_poscache, size, char);
3031 PL_reg_poscache_size = size;
3032 Newz(29, PL_reg_poscache, size, char);
3035 PerlIO_printf(Perl_debug_log,
3036 "%sDetected a super-linear match, switching on caching%s...\n",
3037 PL_colors[4], PL_colors[5])
3040 if (PL_reg_leftiter < 0) {
3041 I32 o = locinput - PL_bostr, b;
3043 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3046 if (PL_reg_poscache[o] & (1<<b)) {
3048 PerlIO_printf(Perl_debug_log,
3049 "%*s already tried at this position...\n",
3050 REPORT_CODE_OFF+PL_regindent*2, "")
3054 PL_reg_poscache[o] |= (1<<b);
3058 /* Prefer next over scan for minimal matching. */
3061 PL_regcc = cc->oldcc;
3064 cp = regcppush(cc->parenfloor);
3066 if (regmatch(cc->next)) {
3068 sayYES; /* All done. */
3070 REGCP_UNWIND(lastcp);
3076 if (n >= cc->max) { /* Maximum greed exceeded? */
3077 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3078 && !(PL_reg_flags & RF_warned)) {
3079 PL_reg_flags |= RF_warned;
3080 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3081 "Complex regular subexpression recursion",
3088 PerlIO_printf(Perl_debug_log,
3089 "%*s trying longer...\n",
3090 REPORT_CODE_OFF+PL_regindent*2, "")
3092 /* Try scanning more and see if it helps. */
3093 PL_reginput = locinput;
3095 cc->lastloc = locinput;
3096 cp = regcppush(cc->parenfloor);
3098 if (regmatch(cc->scan)) {
3102 REGCP_UNWIND(lastcp);
3105 cc->lastloc = lastloc;
3109 /* Prefer scan over next for maximal matching. */
3111 if (n < cc->max) { /* More greed allowed? */
3112 cp = regcppush(cc->parenfloor);
3114 cc->lastloc = locinput;
3116 if (regmatch(cc->scan)) {
3120 REGCP_UNWIND(lastcp);
3121 regcppop(); /* Restore some previous $<digit>s? */
3122 PL_reginput = locinput;
3124 PerlIO_printf(Perl_debug_log,
3125 "%*s failed, try continuation...\n",
3126 REPORT_CODE_OFF+PL_regindent*2, "")
3129 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3130 && !(PL_reg_flags & RF_warned)) {
3131 PL_reg_flags |= RF_warned;
3132 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3133 "Complex regular subexpression recursion",
3137 /* Failed deeper matches of scan, so see if this one works. */
3138 PL_regcc = cc->oldcc;
3141 if (regmatch(cc->next))
3147 cc->lastloc = lastloc;
3152 next = scan + ARG(scan);
3155 inner = NEXTOPER(NEXTOPER(scan));
3158 inner = NEXTOPER(scan);
3162 if (OP(next) != c1) /* No choice. */
3163 next = inner; /* Avoid recursion. */
3165 I32 lastparen = *PL_reglastparen;
3167 re_unwind_branch_t *uw;
3169 /* Put unwinding data on stack */
3170 unwind1 = SSNEWt(1,re_unwind_branch_t);
3171 uw = SSPTRt(unwind1,re_unwind_branch_t);
3174 uw->type = ((c1 == BRANCH)
3176 : RE_UNWIND_BRANCHJ);
3177 uw->lastparen = lastparen;
3179 uw->locinput = locinput;
3180 uw->nextchr = nextchr;
3182 uw->regindent = ++PL_regindent;
3185 REGCP_SET(uw->lastcp);
3187 /* Now go into the first branch */
3200 /* We suppose that the next guy does not need
3201 backtracking: in particular, it is of constant length,
3202 and has no parenths to influence future backrefs. */
3203 ln = ARG1(scan); /* min to match */
3204 n = ARG2(scan); /* max to match */
3205 paren = scan->flags;
3207 if (paren > PL_regsize)
3209 if (paren > *PL_reglastparen)
3210 *PL_reglastparen = paren;
3212 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3214 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3215 PL_reginput = locinput;
3218 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3220 /* if we matched something zero-length we don't need to
3221 backtrack - capturing parens are already defined, so
3222 the caveat in the maximal case doesn't apply
3224 XXXX if ln == 0, we can redo this check first time
3225 through the following loop
3228 n = ln; /* don't backtrack */
3229 locinput = PL_reginput;
3230 if (HAS_TEXT(next) || JUMPABLE(next)) {
3231 regnode *text_node = next;
3233 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3235 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3237 if (PL_regkind[(U8)OP(text_node)] == REF) {
3239 n = ARG(text_node); /* which paren pair */
3240 ln = PL_regstartp[n];
3241 /* assume yes if we haven't seen CLOSEn */
3243 *PL_reglastparen < n ||
3250 c1 = *(PL_bostr + ln);
3252 else { c1 = (U8)*STRING(text_node); }
3253 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3255 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3256 c2 = PL_fold_locale[c1];
3265 /* This may be improved if l == 0. */
3266 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3267 /* If it could work, try it. */
3269 UCHARAT(PL_reginput) == c1 ||
3270 UCHARAT(PL_reginput) == c2)
3274 PL_regstartp[paren] =
3275 HOPc(PL_reginput, -l) - PL_bostr;
3276 PL_regendp[paren] = PL_reginput - PL_bostr;
3279 PL_regendp[paren] = -1;
3283 REGCP_UNWIND(lastcp);
3285 /* Couldn't or didn't -- move forward. */
3286 PL_reginput = locinput;
3287 if (regrepeat_hard(scan, 1, &l)) {
3289 locinput = PL_reginput;
3296 n = regrepeat_hard(scan, n, &l);
3297 /* if we matched something zero-length we don't need to
3298 backtrack, unless the minimum count is zero and we
3299 are capturing the result - in that case the capture
3300 being defined or not may affect later execution
3302 if (n != 0 && l == 0 && !(paren && ln == 0))
3303 ln = n; /* don't backtrack */
3304 locinput = PL_reginput;
3306 PerlIO_printf(Perl_debug_log,
3307 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3308 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3312 if (HAS_TEXT(next) || JUMPABLE(next)) {
3313 regnode *text_node = next;
3315 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3317 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3319 if (PL_regkind[(U8)OP(text_node)] == REF) {
3321 n = ARG(text_node); /* which paren pair */
3322 ln = PL_regstartp[n];
3323 /* assume yes if we haven't seen CLOSEn */
3325 *PL_reglastparen < n ||
3332 c1 = *(PL_bostr + ln);
3334 else { c1 = (U8)*STRING(text_node); }
3336 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3338 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3339 c2 = PL_fold_locale[c1];
3350 /* If it could work, try it. */
3352 UCHARAT(PL_reginput) == c1 ||
3353 UCHARAT(PL_reginput) == c2)
3356 PerlIO_printf(Perl_debug_log,
3357 "%*s trying tail with n=%"IVdf"...\n",
3358 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3362 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3363 PL_regendp[paren] = PL_reginput - PL_bostr;
3366 PL_regendp[paren] = -1;
3370 REGCP_UNWIND(lastcp);
3372 /* Couldn't or didn't -- back up. */
3374 locinput = HOPc(locinput, -l);
3375 PL_reginput = locinput;
3382 paren = scan->flags; /* Which paren to set */
3383 if (paren > PL_regsize)
3385 if (paren > *PL_reglastparen)
3386 *PL_reglastparen = paren;
3387 ln = ARG1(scan); /* min to match */
3388 n = ARG2(scan); /* max to match */
3389 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3393 ln = ARG1(scan); /* min to match */
3394 n = ARG2(scan); /* max to match */
3395 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3400 scan = NEXTOPER(scan);
3406 scan = NEXTOPER(scan);
3410 * Lookahead to avoid useless match attempts
3411 * when we know what character comes next.
3415 * Used to only do .*x and .*?x, but now it allows
3416 * for )'s, ('s and (?{ ... })'s to be in the way
3417 * of the quantifier and the EXACT-like node. -- japhy
3420 if (HAS_TEXT(next) || JUMPABLE(next)) {
3422 regnode *text_node = next;
3424 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3426 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3428 if (PL_regkind[(U8)OP(text_node)] == REF) {
3430 n = ARG(text_node); /* which paren pair */
3431 ln = PL_regstartp[n];
3432 /* assume yes if we haven't seen CLOSEn */
3434 *PL_reglastparen < n ||
3439 goto assume_ok_easy;
3441 s = (U8*)PL_bostr + ln;
3443 else { s = (U8*)STRING(text_node); }
3447 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3449 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3450 c2 = PL_fold_locale[c1];
3453 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3454 STRLEN ulen1, ulen2;
3455 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3456 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3458 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3459 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3461 c1 = utf8_to_uvuni(tmpbuf1, 0);
3462 c2 = utf8_to_uvuni(tmpbuf2, 0);
3465 c2 = c1 = utf8_to_uvchr(s, NULL);
3473 PL_reginput = locinput;
3477 if (ln && regrepeat(scan, ln) < ln)
3479 locinput = PL_reginput;
3482 char *e; /* Should not check after this */
3483 char *old = locinput;
3485 if (n == REG_INFTY) {
3488 while (UTF8_IS_CONTINUATION(*(U8*)e))
3494 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3498 e = locinput + n - ln;
3504 /* Find place 'next' could work */
3507 while (locinput <= e &&
3508 UCHARAT(locinput) != c1)
3511 while (locinput <= e
3512 && UCHARAT(locinput) != c1
3513 && UCHARAT(locinput) != c2)
3516 count = locinput - old;
3523 utf8_to_uvchr((U8*)locinput, &len) != c1;
3528 for (count = 0; locinput <= e; count++) {
3529 UV c = utf8_to_uvchr((U8*)locinput, &len);
3530 if (c == c1 || c == c2)
3538 /* PL_reginput == old now */
3539 if (locinput != old) {
3540 ln = 1; /* Did some */
3541 if (regrepeat(scan, count) < count)
3544 /* PL_reginput == locinput now */
3545 TRYPAREN(paren, ln, locinput);
3546 PL_reginput = locinput; /* Could be reset... */
3547 REGCP_UNWIND(lastcp);
3548 /* Couldn't or didn't -- move forward. */
3551 locinput += UTF8SKIP(locinput);
3557 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3561 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3563 c = UCHARAT(PL_reginput);
3564 /* If it could work, try it. */
3565 if (c == c1 || c == c2)
3567 TRYPAREN(paren, n, PL_reginput);
3568 REGCP_UNWIND(lastcp);
3571 /* If it could work, try it. */
3572 else if (c1 == -1000)
3574 TRYPAREN(paren, n, PL_reginput);
3575 REGCP_UNWIND(lastcp);
3577 /* Couldn't or didn't -- move forward. */
3578 PL_reginput = locinput;
3579 if (regrepeat(scan, 1)) {
3581 locinput = PL_reginput;
3589 n = regrepeat(scan, n);
3590 locinput = PL_reginput;
3591 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3592 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3593 ln = n; /* why back off? */
3594 /* ...because $ and \Z can match before *and* after
3595 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3596 We should back off by one in this case. */
3597 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3606 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3608 c = UCHARAT(PL_reginput);
3610 /* If it could work, try it. */
3611 if (c1 == -1000 || c == c1 || c == c2)
3613 TRYPAREN(paren, n, PL_reginput);
3614 REGCP_UNWIND(lastcp);
3616 /* Couldn't or didn't -- back up. */
3618 PL_reginput = locinput = HOPc(locinput, -1);
3626 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3628 c = UCHARAT(PL_reginput);
3630 /* If it could work, try it. */
3631 if (c1 == -1000 || c == c1 || c == c2)
3633 TRYPAREN(paren, n, PL_reginput);
3634 REGCP_UNWIND(lastcp);
3636 /* Couldn't or didn't -- back up. */
3638 PL_reginput = locinput = HOPc(locinput, -1);
3645 if (PL_reg_call_cc) {
3646 re_cc_state *cur_call_cc = PL_reg_call_cc;
3647 CURCUR *cctmp = PL_regcc;
3648 regexp *re = PL_reg_re;
3649 CHECKPOINT cp, lastcp;
3651 cp = regcppush(0); /* Save *all* the positions. */
3653 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3655 PL_reginput = locinput; /* Make position available to
3657 cache_re(PL_reg_call_cc->re);
3658 PL_regcc = PL_reg_call_cc->cc;
3659 PL_reg_call_cc = PL_reg_call_cc->prev;
3660 if (regmatch(cur_call_cc->node)) {
3661 PL_reg_call_cc = cur_call_cc;
3665 REGCP_UNWIND(lastcp);
3667 PL_reg_call_cc = cur_call_cc;
3673 PerlIO_printf(Perl_debug_log,
3674 "%*s continuation failed...\n",
3675 REPORT_CODE_OFF+PL_regindent*2, "")
3679 if (locinput < PL_regtill) {
3680 DEBUG_r(PerlIO_printf(Perl_debug_log,
3681 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3683 (long)(locinput - PL_reg_starttry),
3684 (long)(PL_regtill - PL_reg_starttry),
3686 sayNO_FINAL; /* Cannot match: too short. */
3688 PL_reginput = locinput; /* put where regtry can find it */
3689 sayYES_FINAL; /* Success! */
3691 PL_reginput = locinput; /* put where regtry can find it */
3692 sayYES_LOUD; /* Success! */
3695 PL_reginput = locinput;
3700 s = HOPBACKc(locinput, scan->flags);
3706 PL_reginput = locinput;
3711 s = HOPBACKc(locinput, scan->flags);
3717 PL_reginput = locinput;
3720 inner = NEXTOPER(NEXTOPER(scan));
3721 if (regmatch(inner) != n) {
3736 if (OP(scan) == SUSPEND) {
3737 locinput = PL_reginput;
3738 nextchr = UCHARAT(locinput);
3743 next = scan + ARG(scan);
3748 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3749 PTR2UV(scan), OP(scan));
3750 Perl_croak(aTHX_ "regexp memory corruption");
3757 * We get here only if there's trouble -- normally "case END" is
3758 * the terminating point.
3760 Perl_croak(aTHX_ "corrupted regexp pointers");
3766 PerlIO_printf(Perl_debug_log,
3767 "%*s %scould match...%s\n",
3768 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3772 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3773 PL_colors[4],PL_colors[5]));
3779 #if 0 /* Breaks $^R */
3787 PerlIO_printf(Perl_debug_log,
3788 "%*s %sfailed...%s\n",
3789 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3795 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3798 case RE_UNWIND_BRANCH:
3799 case RE_UNWIND_BRANCHJ:
3801 re_unwind_branch_t *uwb = &(uw->branch);
3802 I32 lastparen = uwb->lastparen;
3804 REGCP_UNWIND(uwb->lastcp);
3805 for (n = *PL_reglastparen; n > lastparen; n--)
3807 *PL_reglastparen = n;
3808 scan = next = uwb->next;
3810 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3811 ? BRANCH : BRANCHJ) ) { /* Failure */
3818 /* Have more choice yet. Reuse the same uwb. */
3820 if ((n = (uwb->type == RE_UNWIND_BRANCH
3821 ? NEXT_OFF(next) : ARG(next))))
3824 next = NULL; /* XXXX Needn't unwinding in this case... */
3826 next = NEXTOPER(scan);
3827 if (uwb->type == RE_UNWIND_BRANCHJ)
3828 next = NEXTOPER(next);
3829 locinput = uwb->locinput;
3830 nextchr = uwb->nextchr;
3832 PL_regindent = uwb->regindent;
3839 Perl_croak(aTHX_ "regexp unwind memory corruption");
3850 - regrepeat - repeatedly match something simple, report how many
3853 * [This routine now assumes that it will only match on things of length 1.
3854 * That was true before, but now we assume scan - reginput is the count,
3855 * rather than incrementing count on every character. [Er, except utf8.]]
3858 S_regrepeat(pTHX_ regnode *p, I32 max)
3860 register char *scan;
3862 register char *loceol = PL_regeol;
3863 register I32 hardcount = 0;
3864 register bool do_utf8 = PL_reg_match_utf8;
3867 if (max != REG_INFTY && max < loceol - scan)
3868 loceol = scan + max;
3873 while (scan < loceol && hardcount < max && *scan != '\n') {
3874 scan += UTF8SKIP(scan);
3878 while (scan < loceol && *scan != '\n')
3888 case EXACT: /* length of string is 1 */
3890 while (scan < loceol && UCHARAT(scan) == c)
3893 case EXACTF: /* length of string is 1 */
3895 while (scan < loceol &&
3896 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3899 case EXACTFL: /* length of string is 1 */
3900 PL_reg_flags |= RF_tainted;
3902 while (scan < loceol &&
3903 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3909 while (hardcount < max && scan < loceol &&
3910 reginclass(p, (U8*)scan, do_utf8)) {
3911 scan += UTF8SKIP(scan);
3915 while (scan < loceol && reginclass(p, (U8*)scan, do_utf8))
3922 LOAD_UTF8_CHARCLASS(alnum,"a");
3923 while (hardcount < max && scan < loceol &&
3924 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3925 scan += UTF8SKIP(scan);
3929 while (scan < loceol && isALNUM(*scan))
3934 PL_reg_flags |= RF_tainted;
3937 while (hardcount < max && scan < loceol &&
3938 isALNUM_LC_utf8((U8*)scan)) {
3939 scan += UTF8SKIP(scan);
3943 while (scan < loceol && isALNUM_LC(*scan))
3950 LOAD_UTF8_CHARCLASS(alnum,"a");
3951 while (hardcount < max && scan < loceol &&
3952 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3953 scan += UTF8SKIP(scan);
3957 while (scan < loceol && !isALNUM(*scan))
3962 PL_reg_flags |= RF_tainted;
3965 while (hardcount < max && scan < loceol &&
3966 !isALNUM_LC_utf8((U8*)scan)) {
3967 scan += UTF8SKIP(scan);
3971 while (scan < loceol && !isALNUM_LC(*scan))
3978 LOAD_UTF8_CHARCLASS(space," ");
3979 while (hardcount < max && scan < loceol &&
3981 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3982 scan += UTF8SKIP(scan);
3986 while (scan < loceol && isSPACE(*scan))
3991 PL_reg_flags |= RF_tainted;
3994 while (hardcount < max && scan < loceol &&
3995 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3996 scan += UTF8SKIP(scan);
4000 while (scan < loceol && isSPACE_LC(*scan))
4007 LOAD_UTF8_CHARCLASS(space," ");
4008 while (hardcount < max && scan < loceol &&
4010 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4011 scan += UTF8SKIP(scan);
4015 while (scan < loceol && !isSPACE(*scan))
4020 PL_reg_flags |= RF_tainted;
4023 while (hardcount < max && scan < loceol &&
4024 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4025 scan += UTF8SKIP(scan);
4029 while (scan < loceol && !isSPACE_LC(*scan))
4036 LOAD_UTF8_CHARCLASS(digit,"0");
4037 while (hardcount < max && scan < loceol &&
4038 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4039 scan += UTF8SKIP(scan);
4043 while (scan < loceol && isDIGIT(*scan))
4050 LOAD_UTF8_CHARCLASS(digit,"0");
4051 while (hardcount < max && scan < loceol &&
4052 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4053 scan += UTF8SKIP(scan);
4057 while (scan < loceol && !isDIGIT(*scan))
4061 default: /* Called on something of 0 width. */
4062 break; /* So match right here or not at all. */
4068 c = scan - PL_reginput;
4073 SV *prop = sv_newmortal();
4076 PerlIO_printf(Perl_debug_log,
4077 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4078 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4085 - regrepeat_hard - repeatedly match something, report total lenth and length
4087 * The repeater is supposed to have constant length.
4091 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4093 register char *scan = Nullch;
4094 register char *start;
4095 register char *loceol = PL_regeol;
4097 I32 count = 0, res = 1;
4102 start = PL_reginput;
4103 if (PL_reg_match_utf8) {
4104 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4107 while (start < PL_reginput) {
4109 start += UTF8SKIP(start);
4120 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4122 *lp = l = PL_reginput - start;
4123 if (max != REG_INFTY && l*max < loceol - scan)
4124 loceol = scan + l*max;
4137 - regclass_swash - prepare the utf8 swash
4141 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** initsvp)
4146 if (PL_regdata && PL_regdata->count) {
4149 if (PL_regdata->what[n] == 's') {
4150 SV *rv = (SV*)PL_regdata->data[n];
4151 AV *av = (AV*)SvRV((SV*)rv);
4154 si = *av_fetch(av, 0, FALSE);
4155 a = av_fetch(av, 1, FALSE);
4159 else if (si && doinit) {
4160 sw = swash_init("utf8", "", si, 1, 0);
4161 (void)av_store(av, 1, sw);
4173 - reginclass - determine if a character falls into a character class
4177 S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8)
4179 char flags = ANYOF_FLAGS(n);
4184 c = do_utf8 ? utf8_to_uvchr(p, &len) : *p;
4186 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4187 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4188 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4191 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4194 SV *sw = regclass_swash(n, TRUE, 0);
4197 if (swash_fetch(sw, p, do_utf8))
4199 else if (flags & ANYOF_FOLD) {
4201 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
4203 to_utf8_fold(p, tmpbuf, &ulen);
4204 if (swash_fetch(sw, tmpbuf, do_utf8))
4206 to_utf8_upper(p, tmpbuf, &ulen);
4207 if (swash_fetch(sw, tmpbuf, do_utf8))
4213 if (!match && c < 256) {
4214 if (ANYOF_BITMAP_TEST(n, c))
4216 else if (flags & ANYOF_FOLD) {
4219 if (flags & ANYOF_LOCALE) {
4220 PL_reg_flags |= RF_tainted;
4221 f = PL_fold_locale[c];
4225 if (f != c && ANYOF_BITMAP_TEST(n, f))
4229 if (!match && (flags & ANYOF_CLASS)) {
4230 PL_reg_flags |= RF_tainted;
4232 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4233 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4234 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4235 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4236 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4237 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4238 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4239 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4240 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4241 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4242 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4243 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4244 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4245 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4246 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4247 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4248 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4249 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4250 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4251 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4252 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4253 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4254 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4255 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4256 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4257 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4258 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4259 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4260 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4261 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4262 ) /* How's that for a conditional? */
4269 return (flags & ANYOF_INVERT) ? !match : match;
4273 S_reghop(pTHX_ U8 *s, I32 off)
4275 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4279 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4282 while (off-- && s < lim) {
4283 /* XXX could check well-formedness here */
4291 if (UTF8_IS_CONTINUED(*s)) {
4292 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4295 /* XXX could check well-formedness here */
4303 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4305 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4309 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4312 while (off-- && s < lim) {
4313 /* XXX could check well-formedness here */
4323 if (UTF8_IS_CONTINUED(*s)) {
4324 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4327 /* XXX could check well-formedness here */
4339 restore_pos(pTHX_ void *arg)
4341 if (PL_reg_eval_set) {
4342 if (PL_reg_oldsaved) {
4343 PL_reg_re->subbeg = PL_reg_oldsaved;
4344 PL_reg_re->sublen = PL_reg_oldsavedlen;
4345 RX_MATCH_COPIED_on(PL_reg_re);
4347 PL_reg_magic->mg_len = PL_reg_oldpos;
4348 PL_reg_eval_set = 0;
4349 PL_curpm = PL_reg_oldcurpm;