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 && ibcmp_utf8(s, do_utf8, (I32)(strend - s),
1007 && (norun || regtry(prog, s)) )
1017 && (ln == 1 || !(OP(c) == EXACTF
1019 : ibcmp_locale(s, m, ln)))
1020 && (norun || regtry(prog, s)) )
1026 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1027 && (ln == 1 || !(OP(c) == EXACTF
1029 : ibcmp_locale(s, m, ln)))
1030 && (norun || regtry(prog, s)) )
1037 PL_reg_flags |= RF_tainted;
1044 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1047 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1049 tmp = ((OP(c) == BOUND ?
1050 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1051 LOAD_UTF8_CHARCLASS(alnum,"a");
1052 while (s < strend) {
1053 if (tmp == !(OP(c) == BOUND ?
1054 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1055 isALNUM_LC_utf8((U8*)s)))
1058 if ((norun || regtry(prog, s)))
1065 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1066 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1067 while (s < strend) {
1069 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1071 if ((norun || regtry(prog, s)))
1077 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1081 PL_reg_flags |= RF_tainted;
1088 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1091 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1093 tmp = ((OP(c) == NBOUND ?
1094 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1095 LOAD_UTF8_CHARCLASS(alnum,"a");
1096 while (s < strend) {
1097 if (tmp == !(OP(c) == NBOUND ?
1098 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1099 isALNUM_LC_utf8((U8*)s)))
1101 else if ((norun || regtry(prog, s)))
1107 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1108 tmp = ((OP(c) == NBOUND ?
1109 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1110 while (s < strend) {
1112 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1114 else if ((norun || regtry(prog, s)))
1119 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1124 LOAD_UTF8_CHARCLASS(alnum,"a");
1125 while (s < strend) {
1126 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1127 if (tmp && (norun || regtry(prog, s)))
1138 while (s < strend) {
1140 if (tmp && (norun || regtry(prog, s)))
1152 PL_reg_flags |= RF_tainted;
1154 while (s < strend) {
1155 if (isALNUM_LC_utf8((U8*)s)) {
1156 if (tmp && (norun || regtry(prog, s)))
1167 while (s < strend) {
1168 if (isALNUM_LC(*s)) {
1169 if (tmp && (norun || regtry(prog, s)))
1182 LOAD_UTF8_CHARCLASS(alnum,"a");
1183 while (s < strend) {
1184 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1185 if (tmp && (norun || regtry(prog, s)))
1196 while (s < strend) {
1198 if (tmp && (norun || regtry(prog, s)))
1210 PL_reg_flags |= RF_tainted;
1212 while (s < strend) {
1213 if (!isALNUM_LC_utf8((U8*)s)) {
1214 if (tmp && (norun || regtry(prog, s)))
1225 while (s < strend) {
1226 if (!isALNUM_LC(*s)) {
1227 if (tmp && (norun || regtry(prog, s)))
1240 LOAD_UTF8_CHARCLASS(space," ");
1241 while (s < strend) {
1242 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1243 if (tmp && (norun || regtry(prog, s)))
1254 while (s < strend) {
1256 if (tmp && (norun || regtry(prog, s)))
1268 PL_reg_flags |= RF_tainted;
1270 while (s < strend) {
1271 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1272 if (tmp && (norun || regtry(prog, s)))
1283 while (s < strend) {
1284 if (isSPACE_LC(*s)) {
1285 if (tmp && (norun || regtry(prog, s)))
1298 LOAD_UTF8_CHARCLASS(space," ");
1299 while (s < strend) {
1300 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1301 if (tmp && (norun || regtry(prog, s)))
1312 while (s < strend) {
1314 if (tmp && (norun || regtry(prog, s)))
1326 PL_reg_flags |= RF_tainted;
1328 while (s < strend) {
1329 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1330 if (tmp && (norun || regtry(prog, s)))
1341 while (s < strend) {
1342 if (!isSPACE_LC(*s)) {
1343 if (tmp && (norun || regtry(prog, s)))
1356 LOAD_UTF8_CHARCLASS(digit,"0");
1357 while (s < strend) {
1358 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1359 if (tmp && (norun || regtry(prog, s)))
1370 while (s < strend) {
1372 if (tmp && (norun || regtry(prog, s)))
1384 PL_reg_flags |= RF_tainted;
1386 while (s < strend) {
1387 if (isDIGIT_LC_utf8((U8*)s)) {
1388 if (tmp && (norun || regtry(prog, s)))
1399 while (s < strend) {
1400 if (isDIGIT_LC(*s)) {
1401 if (tmp && (norun || regtry(prog, s)))
1414 LOAD_UTF8_CHARCLASS(digit,"0");
1415 while (s < strend) {
1416 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1417 if (tmp && (norun || regtry(prog, s)))
1428 while (s < strend) {
1430 if (tmp && (norun || regtry(prog, s)))
1442 PL_reg_flags |= RF_tainted;
1444 while (s < strend) {
1445 if (!isDIGIT_LC_utf8((U8*)s)) {
1446 if (tmp && (norun || regtry(prog, s)))
1457 while (s < strend) {
1458 if (!isDIGIT_LC(*s)) {
1459 if (tmp && (norun || regtry(prog, s)))
1471 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1480 - regexec_flags - match a regexp against a string
1483 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1484 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1485 /* strend: pointer to null at end of string */
1486 /* strbeg: real beginning of string */
1487 /* minend: end of match must be >=minend after stringarg. */
1488 /* data: May be used for some additional optimizations. */
1489 /* nosave: For optimizations. */
1492 register regnode *c;
1493 register char *startpos = stringarg;
1494 I32 minlen; /* must match at least this many chars */
1495 I32 dontbother = 0; /* how many characters not to try at end */
1496 /* I32 start_shift = 0; */ /* Offset of the start to find
1497 constant substr. */ /* CC */
1498 I32 end_shift = 0; /* Same for the end. */ /* CC */
1499 I32 scream_pos = -1; /* Internal iterator of scream. */
1501 SV* oreplsv = GvSV(PL_replgv);
1502 bool do_utf8 = DO_UTF8(sv);
1504 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
1511 PL_regnarrate = DEBUG_r_TEST;
1514 /* Be paranoid... */
1515 if (prog == NULL || startpos == NULL) {
1516 Perl_croak(aTHX_ "NULL regexp parameter");
1520 minlen = prog->minlen;
1521 if (strend - startpos < minlen) {
1522 DEBUG_r(PerlIO_printf(Perl_debug_log,
1523 "String too short [regexec_flags]...\n"));
1527 /* Check validity of program. */
1528 if (UCHARAT(prog->program) != REG_MAGIC) {
1529 Perl_croak(aTHX_ "corrupted regexp program");
1533 PL_reg_eval_set = 0;
1536 if (prog->reganch & ROPT_UTF8)
1537 PL_reg_flags |= RF_utf8;
1539 /* Mark beginning of line for ^ and lookbehind. */
1540 PL_regbol = startpos;
1544 /* Mark end of line for $ (and such) */
1547 /* see how far we have to get to not match where we matched before */
1548 PL_regtill = startpos+minend;
1550 /* We start without call_cc context. */
1553 /* If there is a "must appear" string, look for it. */
1556 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1559 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1560 PL_reg_ganch = startpos;
1561 else if (sv && SvTYPE(sv) >= SVt_PVMG
1563 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1564 && mg->mg_len >= 0) {
1565 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1566 if (prog->reganch & ROPT_ANCH_GPOS) {
1567 if (s > PL_reg_ganch)
1572 else /* pos() not defined */
1573 PL_reg_ganch = strbeg;
1576 if (do_utf8 == (UTF!=0) &&
1577 !(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
1578 re_scream_pos_data d;
1580 d.scream_olds = &scream_olds;
1581 d.scream_pos = &scream_pos;
1582 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1584 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1585 goto phooey; /* not present */
1590 char *s = do_utf8 ? sv_uni_display(dsv, sv, 60, 0) : startpos;
1591 int len = do_utf8 ? strlen(s) : strend - startpos;
1594 PerlIO_printf(Perl_debug_log,
1595 "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
1596 PL_colors[4],PL_colors[5],PL_colors[0],
1599 (strlen(prog->precomp) > 60 ? "..." : ""),
1601 (int)(len > 60 ? 60 : len),
1603 (len > 60 ? "..." : "")
1607 /* Simplest case: anchored match need be tried only once. */
1608 /* [unless only anchor is BOL and multiline is set] */
1609 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1610 if (s == startpos && regtry(prog, startpos))
1612 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1613 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1618 dontbother = minlen - 1;
1619 end = HOP3c(strend, -dontbother, strbeg) - 1;
1620 /* for multiline we only have to try after newlines */
1621 if (prog->check_substr) {
1625 if (regtry(prog, s))
1630 if (prog->reganch & RE_USE_INTUIT) {
1631 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1642 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1643 if (regtry(prog, s))
1650 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1651 if (regtry(prog, PL_reg_ganch))
1656 /* Messy cases: unanchored match. */
1657 if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
1658 /* we have /x+whatever/ */
1659 /* it must be a one character string (XXXX Except UTF?) */
1660 char ch = SvPVX(prog->anchored_substr)[0];
1666 while (s < strend) {
1668 DEBUG_r( did_match = 1 );
1669 if (regtry(prog, s)) goto got_it;
1671 while (s < strend && *s == ch)
1678 while (s < strend) {
1680 DEBUG_r( did_match = 1 );
1681 if (regtry(prog, s)) goto got_it;
1683 while (s < strend && *s == ch)
1689 DEBUG_r(if (!did_match)
1690 PerlIO_printf(Perl_debug_log,
1691 "Did not find anchored character...\n")
1695 else if (do_utf8 == (UTF!=0) &&
1696 (prog->anchored_substr != Nullsv
1697 || (prog->float_substr != Nullsv
1698 && prog->float_max_offset < strend - s))) {
1699 SV *must = prog->anchored_substr
1700 ? prog->anchored_substr : prog->float_substr;
1702 prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset;
1704 prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
1705 char *last = HOP3c(strend, /* Cannot start after this */
1706 -(I32)(CHR_SVLEN(must)
1707 - (SvTAIL(must) != 0) + back_min), strbeg);
1708 char *last1; /* Last position checked before */
1714 last1 = HOPc(s, -1);
1716 last1 = s - 1; /* bogus */
1718 /* XXXX check_substr already used to find `s', can optimize if
1719 check_substr==must. */
1721 dontbother = end_shift;
1722 strend = HOPc(strend, -dontbother);
1723 while ( (s <= last) &&
1724 ((flags & REXEC_SCREAM)
1725 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1726 end_shift, &scream_pos, 0))
1727 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1728 (unsigned char*)strend, must,
1729 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1730 DEBUG_r( did_match = 1 );
1731 if (HOPc(s, -back_max) > last1) {
1732 last1 = HOPc(s, -back_min);
1733 s = HOPc(s, -back_max);
1736 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1738 last1 = HOPc(s, -back_min);
1742 while (s <= last1) {
1743 if (regtry(prog, s))
1749 while (s <= last1) {
1750 if (regtry(prog, s))
1756 DEBUG_r(if (!did_match)
1757 PerlIO_printf(Perl_debug_log,
1758 "Did not find %s substr `%s%.*s%s'%s...\n",
1759 ((must == prog->anchored_substr)
1760 ? "anchored" : "floating"),
1762 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1764 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1768 else if ((c = prog->regstclass)) {
1769 if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT)
1770 /* don't bother with what can't match */
1771 strend = HOPc(strend, -(minlen - 1));
1773 SV *prop = sv_newmortal();
1775 PerlIO_printf(Perl_debug_log, "Matching stclass `%s' against `%s'\n", SvPVX(prop), UTF ? sv_uni_display(dsv, sv, 60, 0) : s);
1777 if (find_byclass(prog, c, s, strend, startpos, 0))
1779 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1783 if (prog->float_substr != Nullsv) { /* Trim the end. */
1786 if (flags & REXEC_SCREAM) {
1787 last = screaminstr(sv, prog->float_substr, s - strbeg,
1788 end_shift, &scream_pos, 1); /* last one */
1790 last = scream_olds; /* Only one occurrence. */
1794 char *little = SvPV(prog->float_substr, len);
1796 if (SvTAIL(prog->float_substr)) {
1797 if (memEQ(strend - len + 1, little, len - 1))
1798 last = strend - len + 1;
1799 else if (!PL_multiline)
1800 last = memEQ(strend - len, little, len)
1801 ? strend - len : Nullch;
1807 last = rninstr(s, strend, little, little + len);
1809 last = strend; /* matching `$' */
1813 DEBUG_r(PerlIO_printf(Perl_debug_log,
1814 "%sCan't trim the tail, match fails (should not happen)%s\n",
1815 PL_colors[4],PL_colors[5]));
1816 goto phooey; /* Should not happen! */
1818 dontbother = strend - last + prog->float_min_offset;
1820 if (minlen && (dontbother < minlen))
1821 dontbother = minlen - 1;
1822 strend -= dontbother; /* this one's always in bytes! */
1823 /* We don't know much -- general case. */
1826 if (regtry(prog, s))
1835 if (regtry(prog, s))
1837 } while (s++ < strend);
1845 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1847 if (PL_reg_eval_set) {
1848 /* Preserve the current value of $^R */
1849 if (oreplsv != GvSV(PL_replgv))
1850 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1851 restored, the value remains
1853 restore_pos(aTHX_ 0);
1856 /* make sure $`, $&, $', and $digit will work later */
1857 if ( !(flags & REXEC_NOT_FIRST) ) {
1858 if (RX_MATCH_COPIED(prog)) {
1859 Safefree(prog->subbeg);
1860 RX_MATCH_COPIED_off(prog);
1862 if (flags & REXEC_COPY_STR) {
1863 I32 i = PL_regeol - startpos + (stringarg - strbeg);
1865 s = savepvn(strbeg, i);
1868 RX_MATCH_COPIED_on(prog);
1871 prog->subbeg = strbeg;
1872 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
1879 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
1880 PL_colors[4],PL_colors[5]));
1881 if (PL_reg_eval_set)
1882 restore_pos(aTHX_ 0);
1887 - regtry - try match at specific point
1889 STATIC I32 /* 0 failure, 1 success */
1890 S_regtry(pTHX_ regexp *prog, char *startpos)
1898 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
1900 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
1903 PL_reg_eval_set = RS_init;
1905 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
1906 (IV)(PL_stack_sp - PL_stack_base));
1908 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
1909 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
1910 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
1912 /* Apparently this is not needed, judging by wantarray. */
1913 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
1914 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
1917 /* Make $_ available to executed code. */
1918 if (PL_reg_sv != DEFSV) {
1919 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
1924 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
1925 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
1926 /* prepare for quick setting of pos */
1927 sv_magic(PL_reg_sv, (SV*)0,
1928 PERL_MAGIC_regex_global, Nullch, 0);
1929 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
1933 PL_reg_oldpos = mg->mg_len;
1934 SAVEDESTRUCTOR_X(restore_pos, 0);
1936 if (!PL_reg_curpm) {
1937 Newz(22,PL_reg_curpm, 1, PMOP);
1940 SV* repointer = newSViv(0);
1941 /* so we know which PL_regex_padav element is PL_reg_curpm */
1942 SvFLAGS(repointer) |= SVf_BREAK;
1943 av_push(PL_regex_padav,repointer);
1944 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
1945 PL_regex_pad = AvARRAY(PL_regex_padav);
1949 PM_SETRE(PL_reg_curpm, prog);
1950 PL_reg_oldcurpm = PL_curpm;
1951 PL_curpm = PL_reg_curpm;
1952 if (RX_MATCH_COPIED(prog)) {
1953 /* Here is a serious problem: we cannot rewrite subbeg,
1954 since it may be needed if this match fails. Thus
1955 $` inside (?{}) could fail... */
1956 PL_reg_oldsaved = prog->subbeg;
1957 PL_reg_oldsavedlen = prog->sublen;
1958 RX_MATCH_COPIED_off(prog);
1961 PL_reg_oldsaved = Nullch;
1962 prog->subbeg = PL_bostr;
1963 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
1965 prog->startp[0] = startpos - PL_bostr;
1966 PL_reginput = startpos;
1967 PL_regstartp = prog->startp;
1968 PL_regendp = prog->endp;
1969 PL_reglastparen = &prog->lastparen;
1970 PL_reglastcloseparen = &prog->lastcloseparen;
1971 prog->lastparen = 0;
1973 DEBUG_r(PL_reg_starttry = startpos);
1974 if (PL_reg_start_tmpl <= prog->nparens) {
1975 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
1976 if(PL_reg_start_tmp)
1977 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1979 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1983 sv_setpvn(PERL_DEBUG_PAD(0), "", 0);
1984 sv_setpvn(PERL_DEBUG_PAD(1), "", 0);
1985 sv_setpvn(PERL_DEBUG_PAD(2), "", 0);
1988 /* XXXX What this code is doing here?!!! There should be no need
1989 to do this again and again, PL_reglastparen should take care of
1992 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
1993 * Actually, the code in regcppop() (which Ilya may be meaning by
1994 * PL_reglastparen), is not needed at all by the test suite
1995 * (op/regexp, op/pat, op/split), but that code is needed, oddly
1996 * enough, for building DynaLoader, or otherwise this
1997 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
1998 * will happen. Meanwhile, this code *is* needed for the
1999 * above-mentioned test suite tests to succeed. The common theme
2000 * on those tests seems to be returning null fields from matches.
2005 if (prog->nparens) {
2006 for (i = prog->nparens; i > *PL_reglastparen; i--) {
2013 if (regmatch(prog->program + 1)) {
2014 prog->endp[0] = PL_reginput - PL_bostr;
2017 REGCP_UNWIND(lastcp);
2021 #define RE_UNWIND_BRANCH 1
2022 #define RE_UNWIND_BRANCHJ 2
2026 typedef struct { /* XX: makes sense to enlarge it... */
2030 } re_unwind_generic_t;
2043 } re_unwind_branch_t;
2045 typedef union re_unwind_t {
2047 re_unwind_generic_t generic;
2048 re_unwind_branch_t branch;
2051 #define sayYES goto yes
2052 #define sayNO goto no
2053 #define sayYES_FINAL goto yes_final
2054 #define sayYES_LOUD goto yes_loud
2055 #define sayNO_FINAL goto no_final
2056 #define sayNO_SILENT goto do_no
2057 #define saySAME(x) if (x) goto yes; else goto no
2059 #define REPORT_CODE_OFF 24
2062 - regmatch - main matching routine
2064 * Conceptually the strategy is simple: check to see whether the current
2065 * node matches, call self recursively to see whether the rest matches,
2066 * and then act accordingly. In practice we make some effort to avoid
2067 * recursion, in particular by going through "ordinary" nodes (that don't
2068 * need to know whether the rest of the match failed) by a loop instead of
2071 /* [lwall] I've hoisted the register declarations to the outer block in order to
2072 * maybe save a little bit of pushing and popping on the stack. It also takes
2073 * advantage of machines that use a register save mask on subroutine entry.
2075 STATIC I32 /* 0 failure, 1 success */
2076 S_regmatch(pTHX_ regnode *prog)
2078 register regnode *scan; /* Current node. */
2079 regnode *next; /* Next node. */
2080 regnode *inner; /* Next node in internal branch. */
2081 register I32 nextchr; /* renamed nextchr - nextchar colides with
2082 function of same name */
2083 register I32 n; /* no or next */
2084 register I32 ln = 0; /* len or last */
2085 register char *s = Nullch; /* operand or save */
2086 register char *locinput = PL_reginput;
2087 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2088 int minmod = 0, sw = 0, logical = 0;
2091 I32 firstcp = PL_savestack_ix;
2093 register bool do_utf8 = PL_reg_match_utf8;
2095 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2096 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2097 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2104 /* Note that nextchr is a byte even in UTF */
2105 nextchr = UCHARAT(locinput);
2107 while (scan != NULL) {
2110 SV *prop = sv_newmortal();
2111 int docolor = *PL_colors[0];
2112 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2113 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2114 /* The part of the string before starttry has one color
2115 (pref0_len chars), between starttry and current
2116 position another one (pref_len - pref0_len chars),
2117 after the current position the third one.
2118 We assume that pref0_len <= pref_len, otherwise we
2119 decrease pref0_len. */
2120 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2121 ? (5 + taill) - l : locinput - PL_bostr;
2124 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2126 pref0_len = pref_len - (locinput - PL_reg_starttry);
2127 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2128 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2129 ? (5 + taill) - pref_len : PL_regeol - locinput);
2130 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2134 if (pref0_len > pref_len)
2135 pref0_len = pref_len;
2136 regprop(prop, scan);
2140 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2142 locinput - pref_len;
2143 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2144 char *s1 = do_utf8 ?
2145 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2146 pref_len - pref0_len, 60, 0) :
2147 locinput - pref_len + pref0_len;
2148 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2149 char *s2 = do_utf8 ?
2150 pv_uni_display(dsv2, (U8*)locinput,
2151 PL_regeol - locinput, 60, 0) :
2153 int len2 = do_utf8 ? strlen(s2) : l;
2154 PerlIO_printf(Perl_debug_log,
2155 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2156 (IV)(locinput - PL_bostr),
2163 (docolor ? "" : "> <"),
2167 15 - l - pref_len + 1,
2169 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2174 next = scan + NEXT_OFF(scan);
2180 if (locinput == PL_bostr || (PL_multiline &&
2181 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2183 /* regtill = regbol; */
2188 if (locinput == PL_bostr ||
2189 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2195 if (locinput == PL_bostr)
2199 if (locinput == PL_reg_ganch)
2209 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2214 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2216 if (PL_regeol - locinput > 1)
2220 if (PL_regeol != locinput)
2224 if (!nextchr && locinput >= PL_regeol)
2227 locinput += PL_utf8skip[nextchr];
2228 if (locinput > PL_regeol)
2230 nextchr = UCHARAT(locinput);
2233 nextchr = UCHARAT(++locinput);
2236 if (!nextchr && locinput >= PL_regeol)
2238 nextchr = UCHARAT(++locinput);
2241 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2244 locinput += PL_utf8skip[nextchr];
2245 if (locinput > PL_regeol)
2247 nextchr = UCHARAT(locinput);
2250 nextchr = UCHARAT(++locinput);
2255 if (do_utf8 != (UTF!=0)) {
2256 /* The target and the pattern have differing utf8ness. */
2262 /* The target is utf8, the pattern is not utf8. */
2266 if (NATIVE_TO_UNI(*(U8*)s) !=
2267 utf8_to_uvchr((U8*)l, &ulen))
2274 /* The target is not utf8, the pattern is utf8. */
2278 if (NATIVE_TO_UNI(*((U8*)l)) !=
2279 utf8_to_uvchr((U8*)s, &ulen))
2286 nextchr = UCHARAT(locinput);
2289 /* The target and the pattern have the same utf8ness. */
2290 /* Inline the first character, for speed. */
2291 if (UCHARAT(s) != nextchr)
2293 if (PL_regeol - locinput < ln)
2295 if (ln > 1 && memNE(s, locinput, ln))
2298 nextchr = UCHARAT(locinput);
2301 PL_reg_flags |= RF_tainted;
2310 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
2312 if (do_utf8 != (UTF!=0)) {
2313 /* The target and the pattern have differing utf8ness. */
2314 STRLEN ulen1, ulen2;
2318 /* The target is utf8, the pattern is not utf8. */
2323 cs = to_uni_fold(NATIVE_TO_UNI(*(U8*)s),
2325 cl = utf8_to_uvchr((U8*)l, &ulen2);
2328 cl = to_uni_fold(cl, (U8*)l, &ulen2);
2329 if (ulen1 != ulen2 || cs != cl)
2337 /* The target is not utf8, the pattern is utf8. */
2342 cs = utf8_to_uvchr((U8*)s, &ulen1);
2344 cl = to_uni_fold(NATIVE_TO_UNI(*(U8*)l),
2348 cs = to_uni_fold(cs, (U8*)s, &ulen1);
2349 if (ulen1 != ulen2 || cs != cl)
2357 nextchr = UCHARAT(locinput);
2361 if (do_utf8 && UTF) {
2362 /* Both the target and the pattern are utf8. */
2368 if (UTF8SKIP(s) != UTF8SKIP(l) ||
2369 memNE(s, (char*)l, UTF8SKIP(s))) {
2370 to_utf8_fold((U8*)l, tmpbuf, &ulen);
2371 if (UTF8SKIP(s) != ulen ||
2372 memNE(s, (char*)tmpbuf, ulen))
2379 nextchr = UCHARAT(locinput);
2384 /* Neither the target and the pattern are utf8. */
2386 /* Inline the first character, for speed. */
2387 if (UCHARAT(s) != nextchr &&
2388 UCHARAT(s) != ((OP(scan) == EXACTF)
2389 ? PL_fold : PL_fold_locale)[nextchr])
2391 if (PL_regeol - locinput < ln)
2393 if (ln > 1 && (OP(scan) == EXACTF
2394 ? ibcmp(s, locinput, ln)
2395 : ibcmp_locale(s, locinput, ln)))
2398 nextchr = UCHARAT(locinput);
2402 if (!reginclass(scan, (U8*)locinput, do_utf8))
2404 if (locinput >= PL_regeol)
2406 locinput += PL_utf8skip[nextchr];
2407 nextchr = UCHARAT(locinput);
2411 nextchr = UCHARAT(locinput);
2412 if (!reginclass(scan, (U8*)locinput, do_utf8))
2414 if (!nextchr && locinput >= PL_regeol)
2416 nextchr = UCHARAT(++locinput);
2420 PL_reg_flags |= RF_tainted;
2426 LOAD_UTF8_CHARCLASS(alnum,"a");
2427 if (!(OP(scan) == ALNUM
2428 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2429 : isALNUM_LC_utf8((U8*)locinput)))
2433 locinput += PL_utf8skip[nextchr];
2434 nextchr = UCHARAT(locinput);
2437 if (!(OP(scan) == ALNUM
2438 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2440 nextchr = UCHARAT(++locinput);
2443 PL_reg_flags |= RF_tainted;
2446 if (!nextchr && locinput >= PL_regeol)
2449 LOAD_UTF8_CHARCLASS(alnum,"a");
2450 if (OP(scan) == NALNUM
2451 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2452 : isALNUM_LC_utf8((U8*)locinput))
2456 locinput += PL_utf8skip[nextchr];
2457 nextchr = UCHARAT(locinput);
2460 if (OP(scan) == NALNUM
2461 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2463 nextchr = UCHARAT(++locinput);
2467 PL_reg_flags |= RF_tainted;
2471 /* was last char in word? */
2473 if (locinput == PL_bostr)
2476 U8 *r = reghop((U8*)locinput, -1);
2478 ln = utf8n_to_uvchr(r, s - (char*)r, 0, 0);
2480 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2481 ln = isALNUM_uni(ln);
2482 LOAD_UTF8_CHARCLASS(alnum,"a");
2483 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2486 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2487 n = isALNUM_LC_utf8((U8*)locinput);
2491 ln = (locinput != PL_bostr) ?
2492 UCHARAT(locinput - 1) : '\n';
2493 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2495 n = isALNUM(nextchr);
2498 ln = isALNUM_LC(ln);
2499 n = isALNUM_LC(nextchr);
2502 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2503 OP(scan) == BOUNDL))
2507 PL_reg_flags |= RF_tainted;
2513 if (UTF8_IS_CONTINUED(nextchr)) {
2514 LOAD_UTF8_CHARCLASS(space," ");
2515 if (!(OP(scan) == SPACE
2516 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2517 : isSPACE_LC_utf8((U8*)locinput)))
2521 locinput += PL_utf8skip[nextchr];
2522 nextchr = UCHARAT(locinput);
2525 if (!(OP(scan) == SPACE
2526 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2528 nextchr = UCHARAT(++locinput);
2531 if (!(OP(scan) == SPACE
2532 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2534 nextchr = UCHARAT(++locinput);
2538 PL_reg_flags |= RF_tainted;
2541 if (!nextchr && locinput >= PL_regeol)
2544 LOAD_UTF8_CHARCLASS(space," ");
2545 if (OP(scan) == NSPACE
2546 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2547 : isSPACE_LC_utf8((U8*)locinput))
2551 locinput += PL_utf8skip[nextchr];
2552 nextchr = UCHARAT(locinput);
2555 if (OP(scan) == NSPACE
2556 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2558 nextchr = UCHARAT(++locinput);
2561 PL_reg_flags |= RF_tainted;
2567 LOAD_UTF8_CHARCLASS(digit,"0");
2568 if (!(OP(scan) == DIGIT
2569 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2570 : isDIGIT_LC_utf8((U8*)locinput)))
2574 locinput += PL_utf8skip[nextchr];
2575 nextchr = UCHARAT(locinput);
2578 if (!(OP(scan) == DIGIT
2579 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2581 nextchr = UCHARAT(++locinput);
2584 PL_reg_flags |= RF_tainted;
2587 if (!nextchr && locinput >= PL_regeol)
2590 LOAD_UTF8_CHARCLASS(digit,"0");
2591 if (OP(scan) == NDIGIT
2592 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2593 : isDIGIT_LC_utf8((U8*)locinput))
2597 locinput += PL_utf8skip[nextchr];
2598 nextchr = UCHARAT(locinput);
2601 if (OP(scan) == NDIGIT
2602 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2604 nextchr = UCHARAT(++locinput);
2607 if (locinput >= PL_regeol)
2610 LOAD_UTF8_CHARCLASS(mark,"~");
2611 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2613 locinput += PL_utf8skip[nextchr];
2614 while (locinput < PL_regeol &&
2615 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2616 locinput += UTF8SKIP(locinput);
2617 if (locinput > PL_regeol)
2622 nextchr = UCHARAT(locinput);
2625 PL_reg_flags |= RF_tainted;
2629 n = ARG(scan); /* which paren pair */
2630 ln = PL_regstartp[n];
2631 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2632 if (*PL_reglastparen < n || ln == -1)
2633 sayNO; /* Do not match unless seen CLOSEn. */
2634 if (ln == PL_regendp[n])
2638 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2640 char *e = PL_bostr + PL_regendp[n];
2642 * Note that we can't do the "other character" lookup trick as
2643 * in the 8-bit case (no pun intended) because in Unicode we
2644 * have to map both upper and title case to lower case.
2646 if (OP(scan) == REFF) {
2647 STRLEN ulen1, ulen2;
2648 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2649 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2653 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2654 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2655 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2662 nextchr = UCHARAT(locinput);
2666 /* Inline the first character, for speed. */
2667 if (UCHARAT(s) != nextchr &&
2669 (UCHARAT(s) != ((OP(scan) == REFF
2670 ? PL_fold : PL_fold_locale)[nextchr]))))
2672 ln = PL_regendp[n] - ln;
2673 if (locinput + ln > PL_regeol)
2675 if (ln > 1 && (OP(scan) == REF
2676 ? memNE(s, locinput, ln)
2678 ? ibcmp(s, locinput, ln)
2679 : ibcmp_locale(s, locinput, ln))))
2682 nextchr = UCHARAT(locinput);
2693 OP_4tree *oop = PL_op;
2694 COP *ocurcop = PL_curcop;
2695 SV **ocurpad = PL_curpad;
2699 PL_op = (OP_4tree*)PL_regdata->data[n];
2700 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2701 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2702 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2706 CALLRUNOPS(aTHX); /* Scalar context. */
2709 ret = Nullsv; /* protect against empty (?{}) blocks. */
2717 PL_curpad = ocurpad;
2718 PL_curcop = ocurcop;
2720 if (logical == 2) { /* Postponed subexpression. */
2722 MAGIC *mg = Null(MAGIC*);
2724 CHECKPOINT cp, lastcp;
2726 if(SvROK(ret) || SvRMAGICAL(ret)) {
2727 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2730 mg = mg_find(sv, PERL_MAGIC_qr);
2733 re = (regexp *)mg->mg_obj;
2734 (void)ReREFCNT_inc(re);
2738 char *t = SvPV(ret, len);
2740 char *oprecomp = PL_regprecomp;
2741 I32 osize = PL_regsize;
2742 I32 onpar = PL_regnpar;
2745 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2747 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2748 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2750 PL_regprecomp = oprecomp;
2755 PerlIO_printf(Perl_debug_log,
2756 "Entering embedded `%s%.60s%s%s'\n",
2760 (strlen(re->precomp) > 60 ? "..." : ""))
2763 state.prev = PL_reg_call_cc;
2764 state.cc = PL_regcc;
2765 state.re = PL_reg_re;
2769 cp = regcppush(0); /* Save *all* the positions. */
2772 state.ss = PL_savestack_ix;
2773 *PL_reglastparen = 0;
2774 *PL_reglastcloseparen = 0;
2775 PL_reg_call_cc = &state;
2776 PL_reginput = locinput;
2778 /* XXXX This is too dramatic a measure... */
2781 if (regmatch(re->program + 1)) {
2782 /* Even though we succeeded, we need to restore
2783 global variables, since we may be wrapped inside
2784 SUSPEND, thus the match may be not finished yet. */
2786 /* XXXX Do this only if SUSPENDed? */
2787 PL_reg_call_cc = state.prev;
2788 PL_regcc = state.cc;
2789 PL_reg_re = state.re;
2790 cache_re(PL_reg_re);
2792 /* XXXX This is too dramatic a measure... */
2795 /* These are needed even if not SUSPEND. */
2801 REGCP_UNWIND(lastcp);
2803 PL_reg_call_cc = state.prev;
2804 PL_regcc = state.cc;
2805 PL_reg_re = state.re;
2806 cache_re(PL_reg_re);
2808 /* XXXX This is too dramatic a measure... */
2818 sv_setsv(save_scalar(PL_replgv), ret);
2822 n = ARG(scan); /* which paren pair */
2823 PL_reg_start_tmp[n] = locinput;
2828 n = ARG(scan); /* which paren pair */
2829 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2830 PL_regendp[n] = locinput - PL_bostr;
2831 if (n > *PL_reglastparen)
2832 *PL_reglastparen = n;
2833 *PL_reglastcloseparen = n;
2836 n = ARG(scan); /* which paren pair */
2837 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2840 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2842 next = NEXTOPER(NEXTOPER(scan));
2844 next = scan + ARG(scan);
2845 if (OP(next) == IFTHEN) /* Fake one. */
2846 next = NEXTOPER(NEXTOPER(next));
2850 logical = scan->flags;
2852 /*******************************************************************
2853 PL_regcc contains infoblock about the innermost (...)* loop, and
2854 a pointer to the next outer infoblock.
2856 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2858 1) After matching X, regnode for CURLYX is processed;
2860 2) This regnode creates infoblock on the stack, and calls
2861 regmatch() recursively with the starting point at WHILEM node;
2863 3) Each hit of WHILEM node tries to match A and Z (in the order
2864 depending on the current iteration, min/max of {min,max} and
2865 greediness). The information about where are nodes for "A"
2866 and "Z" is read from the infoblock, as is info on how many times "A"
2867 was already matched, and greediness.
2869 4) After A matches, the same WHILEM node is hit again.
2871 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2872 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2873 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2874 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2875 of the external loop.
2877 Currently present infoblocks form a tree with a stem formed by PL_curcc
2878 and whatever it mentions via ->next, and additional attached trees
2879 corresponding to temporarily unset infoblocks as in "5" above.
2881 In the following picture infoblocks for outer loop of
2882 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2883 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2884 infoblocks are drawn below the "reset" infoblock.
2886 In fact in the picture below we do not show failed matches for Z and T
2887 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2888 more obvious *why* one needs to *temporary* unset infoblocks.]
2890 Matched REx position InfoBlocks Comment
2894 Y A)*?Z)*?T x <- O <- I
2895 YA )*?Z)*?T x <- O <- I
2896 YA A)*?Z)*?T x <- O <- I
2897 YAA )*?Z)*?T x <- O <- I
2898 YAA Z)*?T x <- O # Temporary unset I
2901 YAAZ Y(A)*?Z)*?T x <- O
2904 YAAZY (A)*?Z)*?T x <- O
2907 YAAZY A)*?Z)*?T x <- O <- I
2910 YAAZYA )*?Z)*?T x <- O <- I
2913 YAAZYA Z)*?T x <- O # Temporary unset I
2919 YAAZYAZ T x # Temporary unset O
2926 *******************************************************************/
2929 CHECKPOINT cp = PL_savestack_ix;
2930 /* No need to save/restore up to this paren */
2931 I32 parenfloor = scan->flags;
2933 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2935 cc.oldcc = PL_regcc;
2937 /* XXXX Probably it is better to teach regpush to support
2938 parenfloor > PL_regsize... */
2939 if (parenfloor > *PL_reglastparen)
2940 parenfloor = *PL_reglastparen; /* Pessimization... */
2941 cc.parenfloor = parenfloor;
2943 cc.min = ARG1(scan);
2944 cc.max = ARG2(scan);
2945 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2949 PL_reginput = locinput;
2950 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2952 PL_regcc = cc.oldcc;
2958 * This is really hard to understand, because after we match
2959 * what we're trying to match, we must make sure the rest of
2960 * the REx is going to match for sure, and to do that we have
2961 * to go back UP the parse tree by recursing ever deeper. And
2962 * if it fails, we have to reset our parent's current state
2963 * that we can try again after backing off.
2966 CHECKPOINT cp, lastcp;
2967 CURCUR* cc = PL_regcc;
2968 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2970 n = cc->cur + 1; /* how many we know we matched */
2971 PL_reginput = locinput;
2974 PerlIO_printf(Perl_debug_log,
2975 "%*s %ld out of %ld..%ld cc=%lx\n",
2976 REPORT_CODE_OFF+PL_regindent*2, "",
2977 (long)n, (long)cc->min,
2978 (long)cc->max, (long)cc)
2981 /* If degenerate scan matches "", assume scan done. */
2983 if (locinput == cc->lastloc && n >= cc->min) {
2984 PL_regcc = cc->oldcc;
2988 PerlIO_printf(Perl_debug_log,
2989 "%*s empty match detected, try continuation...\n",
2990 REPORT_CODE_OFF+PL_regindent*2, "")
2992 if (regmatch(cc->next))
3000 /* First just match a string of min scans. */
3004 cc->lastloc = locinput;
3005 if (regmatch(cc->scan))
3008 cc->lastloc = lastloc;
3013 /* Check whether we already were at this position.
3014 Postpone detection until we know the match is not
3015 *that* much linear. */
3016 if (!PL_reg_maxiter) {
3017 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3018 PL_reg_leftiter = PL_reg_maxiter;
3020 if (PL_reg_leftiter-- == 0) {
3021 I32 size = (PL_reg_maxiter + 7)/8;
3022 if (PL_reg_poscache) {
3023 if (PL_reg_poscache_size < size) {
3024 Renew(PL_reg_poscache, size, char);
3025 PL_reg_poscache_size = size;
3027 Zero(PL_reg_poscache, size, char);
3030 PL_reg_poscache_size = size;
3031 Newz(29, PL_reg_poscache, size, char);
3034 PerlIO_printf(Perl_debug_log,
3035 "%sDetected a super-linear match, switching on caching%s...\n",
3036 PL_colors[4], PL_colors[5])
3039 if (PL_reg_leftiter < 0) {
3040 I32 o = locinput - PL_bostr, b;
3042 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3045 if (PL_reg_poscache[o] & (1<<b)) {
3047 PerlIO_printf(Perl_debug_log,
3048 "%*s already tried at this position...\n",
3049 REPORT_CODE_OFF+PL_regindent*2, "")
3053 PL_reg_poscache[o] |= (1<<b);
3057 /* Prefer next over scan for minimal matching. */
3060 PL_regcc = cc->oldcc;
3063 cp = regcppush(cc->parenfloor);
3065 if (regmatch(cc->next)) {
3067 sayYES; /* All done. */
3069 REGCP_UNWIND(lastcp);
3075 if (n >= cc->max) { /* Maximum greed exceeded? */
3076 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3077 && !(PL_reg_flags & RF_warned)) {
3078 PL_reg_flags |= RF_warned;
3079 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3080 "Complex regular subexpression recursion",
3087 PerlIO_printf(Perl_debug_log,
3088 "%*s trying longer...\n",
3089 REPORT_CODE_OFF+PL_regindent*2, "")
3091 /* Try scanning more and see if it helps. */
3092 PL_reginput = locinput;
3094 cc->lastloc = locinput;
3095 cp = regcppush(cc->parenfloor);
3097 if (regmatch(cc->scan)) {
3101 REGCP_UNWIND(lastcp);
3104 cc->lastloc = lastloc;
3108 /* Prefer scan over next for maximal matching. */
3110 if (n < cc->max) { /* More greed allowed? */
3111 cp = regcppush(cc->parenfloor);
3113 cc->lastloc = locinput;
3115 if (regmatch(cc->scan)) {
3119 REGCP_UNWIND(lastcp);
3120 regcppop(); /* Restore some previous $<digit>s? */
3121 PL_reginput = locinput;
3123 PerlIO_printf(Perl_debug_log,
3124 "%*s failed, try continuation...\n",
3125 REPORT_CODE_OFF+PL_regindent*2, "")
3128 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3129 && !(PL_reg_flags & RF_warned)) {
3130 PL_reg_flags |= RF_warned;
3131 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3132 "Complex regular subexpression recursion",
3136 /* Failed deeper matches of scan, so see if this one works. */
3137 PL_regcc = cc->oldcc;
3140 if (regmatch(cc->next))
3146 cc->lastloc = lastloc;
3151 next = scan + ARG(scan);
3154 inner = NEXTOPER(NEXTOPER(scan));
3157 inner = NEXTOPER(scan);
3161 if (OP(next) != c1) /* No choice. */
3162 next = inner; /* Avoid recursion. */
3164 I32 lastparen = *PL_reglastparen;
3166 re_unwind_branch_t *uw;
3168 /* Put unwinding data on stack */
3169 unwind1 = SSNEWt(1,re_unwind_branch_t);
3170 uw = SSPTRt(unwind1,re_unwind_branch_t);
3173 uw->type = ((c1 == BRANCH)
3175 : RE_UNWIND_BRANCHJ);
3176 uw->lastparen = lastparen;
3178 uw->locinput = locinput;
3179 uw->nextchr = nextchr;
3181 uw->regindent = ++PL_regindent;
3184 REGCP_SET(uw->lastcp);
3186 /* Now go into the first branch */
3199 /* We suppose that the next guy does not need
3200 backtracking: in particular, it is of constant length,
3201 and has no parenths to influence future backrefs. */
3202 ln = ARG1(scan); /* min to match */
3203 n = ARG2(scan); /* max to match */
3204 paren = scan->flags;
3206 if (paren > PL_regsize)
3208 if (paren > *PL_reglastparen)
3209 *PL_reglastparen = paren;
3211 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3213 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3214 PL_reginput = locinput;
3217 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3219 /* if we matched something zero-length we don't need to
3220 backtrack - capturing parens are already defined, so
3221 the caveat in the maximal case doesn't apply
3223 XXXX if ln == 0, we can redo this check first time
3224 through the following loop
3227 n = ln; /* don't backtrack */
3228 locinput = PL_reginput;
3229 if (HAS_TEXT(next) || JUMPABLE(next)) {
3230 regnode *text_node = next;
3232 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3234 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3236 if (PL_regkind[(U8)OP(text_node)] == REF) {
3238 n = ARG(text_node); /* which paren pair */
3239 ln = PL_regstartp[n];
3240 /* assume yes if we haven't seen CLOSEn */
3242 *PL_reglastparen < n ||
3249 c1 = *(PL_bostr + ln);
3251 else { c1 = (U8)*STRING(text_node); }
3252 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3254 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3255 c2 = PL_fold_locale[c1];
3264 /* This may be improved if l == 0. */
3265 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3266 /* If it could work, try it. */
3268 UCHARAT(PL_reginput) == c1 ||
3269 UCHARAT(PL_reginput) == c2)
3273 PL_regstartp[paren] =
3274 HOPc(PL_reginput, -l) - PL_bostr;
3275 PL_regendp[paren] = PL_reginput - PL_bostr;
3278 PL_regendp[paren] = -1;
3282 REGCP_UNWIND(lastcp);
3284 /* Couldn't or didn't -- move forward. */
3285 PL_reginput = locinput;
3286 if (regrepeat_hard(scan, 1, &l)) {
3288 locinput = PL_reginput;
3295 n = regrepeat_hard(scan, n, &l);
3296 /* if we matched something zero-length we don't need to
3297 backtrack, unless the minimum count is zero and we
3298 are capturing the result - in that case the capture
3299 being defined or not may affect later execution
3301 if (n != 0 && l == 0 && !(paren && ln == 0))
3302 ln = n; /* don't backtrack */
3303 locinput = PL_reginput;
3305 PerlIO_printf(Perl_debug_log,
3306 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3307 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3311 if (HAS_TEXT(next) || JUMPABLE(next)) {
3312 regnode *text_node = next;
3314 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3316 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3318 if (PL_regkind[(U8)OP(text_node)] == REF) {
3320 n = ARG(text_node); /* which paren pair */
3321 ln = PL_regstartp[n];
3322 /* assume yes if we haven't seen CLOSEn */
3324 *PL_reglastparen < n ||
3331 c1 = *(PL_bostr + ln);
3333 else { c1 = (U8)*STRING(text_node); }
3335 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3337 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3338 c2 = PL_fold_locale[c1];
3349 /* If it could work, try it. */
3351 UCHARAT(PL_reginput) == c1 ||
3352 UCHARAT(PL_reginput) == c2)
3355 PerlIO_printf(Perl_debug_log,
3356 "%*s trying tail with n=%"IVdf"...\n",
3357 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3361 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3362 PL_regendp[paren] = PL_reginput - PL_bostr;
3365 PL_regendp[paren] = -1;
3369 REGCP_UNWIND(lastcp);
3371 /* Couldn't or didn't -- back up. */
3373 locinput = HOPc(locinput, -l);
3374 PL_reginput = locinput;
3381 paren = scan->flags; /* Which paren to set */
3382 if (paren > PL_regsize)
3384 if (paren > *PL_reglastparen)
3385 *PL_reglastparen = paren;
3386 ln = ARG1(scan); /* min to match */
3387 n = ARG2(scan); /* max to match */
3388 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3392 ln = ARG1(scan); /* min to match */
3393 n = ARG2(scan); /* max to match */
3394 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3399 scan = NEXTOPER(scan);
3405 scan = NEXTOPER(scan);
3409 * Lookahead to avoid useless match attempts
3410 * when we know what character comes next.
3414 * Used to only do .*x and .*?x, but now it allows
3415 * for )'s, ('s and (?{ ... })'s to be in the way
3416 * of the quantifier and the EXACT-like node. -- japhy
3419 if (HAS_TEXT(next) || JUMPABLE(next)) {
3421 regnode *text_node = next;
3423 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3425 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3427 if (PL_regkind[(U8)OP(text_node)] == REF) {
3429 n = ARG(text_node); /* which paren pair */
3430 ln = PL_regstartp[n];
3431 /* assume yes if we haven't seen CLOSEn */
3433 *PL_reglastparen < n ||
3438 goto assume_ok_easy;
3440 s = (U8*)PL_bostr + ln;
3442 else { s = (U8*)STRING(text_node); }
3446 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3448 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3449 c2 = PL_fold_locale[c1];
3452 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3453 STRLEN ulen1, ulen2;
3454 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3455 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3457 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3458 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3460 c1 = utf8_to_uvuni(tmpbuf1, 0);
3461 c2 = utf8_to_uvuni(tmpbuf2, 0);
3464 c2 = c1 = utf8_to_uvchr(s, NULL);
3472 PL_reginput = locinput;
3476 if (ln && regrepeat(scan, ln) < ln)
3478 locinput = PL_reginput;
3481 char *e; /* Should not check after this */
3482 char *old = locinput;
3484 if (n == REG_INFTY) {
3487 while (UTF8_IS_CONTINUATION(*(U8*)e))
3493 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3497 e = locinput + n - ln;
3503 /* Find place 'next' could work */
3506 while (locinput <= e &&
3507 UCHARAT(locinput) != c1)
3510 while (locinput <= e
3511 && UCHARAT(locinput) != c1
3512 && UCHARAT(locinput) != c2)
3515 count = locinput - old;
3522 utf8_to_uvchr((U8*)locinput, &len) != c1;
3527 for (count = 0; locinput <= e; count++) {
3528 UV c = utf8_to_uvchr((U8*)locinput, &len);
3529 if (c == c1 || c == c2)
3537 /* PL_reginput == old now */
3538 if (locinput != old) {
3539 ln = 1; /* Did some */
3540 if (regrepeat(scan, count) < count)
3543 /* PL_reginput == locinput now */
3544 TRYPAREN(paren, ln, locinput);
3545 PL_reginput = locinput; /* Could be reset... */
3546 REGCP_UNWIND(lastcp);
3547 /* Couldn't or didn't -- move forward. */
3550 locinput += UTF8SKIP(locinput);
3556 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3560 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3562 c = UCHARAT(PL_reginput);
3563 /* If it could work, try it. */
3564 if (c == c1 || c == c2)
3566 TRYPAREN(paren, n, PL_reginput);
3567 REGCP_UNWIND(lastcp);
3570 /* If it could work, try it. */
3571 else if (c1 == -1000)
3573 TRYPAREN(paren, n, PL_reginput);
3574 REGCP_UNWIND(lastcp);
3576 /* Couldn't or didn't -- move forward. */
3577 PL_reginput = locinput;
3578 if (regrepeat(scan, 1)) {
3580 locinput = PL_reginput;
3588 n = regrepeat(scan, n);
3589 locinput = PL_reginput;
3590 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3591 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3592 ln = n; /* why back off? */
3593 /* ...because $ and \Z can match before *and* after
3594 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3595 We should back off by one in this case. */
3596 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3605 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3607 c = UCHARAT(PL_reginput);
3609 /* If it could work, try it. */
3610 if (c1 == -1000 || c == c1 || c == c2)
3612 TRYPAREN(paren, n, PL_reginput);
3613 REGCP_UNWIND(lastcp);
3615 /* Couldn't or didn't -- back up. */
3617 PL_reginput = locinput = HOPc(locinput, -1);
3625 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3627 c = UCHARAT(PL_reginput);
3629 /* If it could work, try it. */
3630 if (c1 == -1000 || c == c1 || c == c2)
3632 TRYPAREN(paren, n, PL_reginput);
3633 REGCP_UNWIND(lastcp);
3635 /* Couldn't or didn't -- back up. */
3637 PL_reginput = locinput = HOPc(locinput, -1);
3644 if (PL_reg_call_cc) {
3645 re_cc_state *cur_call_cc = PL_reg_call_cc;
3646 CURCUR *cctmp = PL_regcc;
3647 regexp *re = PL_reg_re;
3648 CHECKPOINT cp, lastcp;
3650 cp = regcppush(0); /* Save *all* the positions. */
3652 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3654 PL_reginput = locinput; /* Make position available to
3656 cache_re(PL_reg_call_cc->re);
3657 PL_regcc = PL_reg_call_cc->cc;
3658 PL_reg_call_cc = PL_reg_call_cc->prev;
3659 if (regmatch(cur_call_cc->node)) {
3660 PL_reg_call_cc = cur_call_cc;
3664 REGCP_UNWIND(lastcp);
3666 PL_reg_call_cc = cur_call_cc;
3672 PerlIO_printf(Perl_debug_log,
3673 "%*s continuation failed...\n",
3674 REPORT_CODE_OFF+PL_regindent*2, "")
3678 if (locinput < PL_regtill) {
3679 DEBUG_r(PerlIO_printf(Perl_debug_log,
3680 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3682 (long)(locinput - PL_reg_starttry),
3683 (long)(PL_regtill - PL_reg_starttry),
3685 sayNO_FINAL; /* Cannot match: too short. */
3687 PL_reginput = locinput; /* put where regtry can find it */
3688 sayYES_FINAL; /* Success! */
3690 PL_reginput = locinput; /* put where regtry can find it */
3691 sayYES_LOUD; /* Success! */
3694 PL_reginput = locinput;
3699 s = HOPBACKc(locinput, scan->flags);
3705 PL_reginput = locinput;
3710 s = HOPBACKc(locinput, scan->flags);
3716 PL_reginput = locinput;
3719 inner = NEXTOPER(NEXTOPER(scan));
3720 if (regmatch(inner) != n) {
3735 if (OP(scan) == SUSPEND) {
3736 locinput = PL_reginput;
3737 nextchr = UCHARAT(locinput);
3742 next = scan + ARG(scan);
3747 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3748 PTR2UV(scan), OP(scan));
3749 Perl_croak(aTHX_ "regexp memory corruption");
3756 * We get here only if there's trouble -- normally "case END" is
3757 * the terminating point.
3759 Perl_croak(aTHX_ "corrupted regexp pointers");
3765 PerlIO_printf(Perl_debug_log,
3766 "%*s %scould match...%s\n",
3767 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3771 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3772 PL_colors[4],PL_colors[5]));
3778 #if 0 /* Breaks $^R */
3786 PerlIO_printf(Perl_debug_log,
3787 "%*s %sfailed...%s\n",
3788 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3794 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3797 case RE_UNWIND_BRANCH:
3798 case RE_UNWIND_BRANCHJ:
3800 re_unwind_branch_t *uwb = &(uw->branch);
3801 I32 lastparen = uwb->lastparen;
3803 REGCP_UNWIND(uwb->lastcp);
3804 for (n = *PL_reglastparen; n > lastparen; n--)
3806 *PL_reglastparen = n;
3807 scan = next = uwb->next;
3809 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3810 ? BRANCH : BRANCHJ) ) { /* Failure */
3817 /* Have more choice yet. Reuse the same uwb. */
3819 if ((n = (uwb->type == RE_UNWIND_BRANCH
3820 ? NEXT_OFF(next) : ARG(next))))
3823 next = NULL; /* XXXX Needn't unwinding in this case... */
3825 next = NEXTOPER(scan);
3826 if (uwb->type == RE_UNWIND_BRANCHJ)
3827 next = NEXTOPER(next);
3828 locinput = uwb->locinput;
3829 nextchr = uwb->nextchr;
3831 PL_regindent = uwb->regindent;
3838 Perl_croak(aTHX_ "regexp unwind memory corruption");
3849 - regrepeat - repeatedly match something simple, report how many
3852 * [This routine now assumes that it will only match on things of length 1.
3853 * That was true before, but now we assume scan - reginput is the count,
3854 * rather than incrementing count on every character. [Er, except utf8.]]
3857 S_regrepeat(pTHX_ regnode *p, I32 max)
3859 register char *scan;
3861 register char *loceol = PL_regeol;
3862 register I32 hardcount = 0;
3863 register bool do_utf8 = PL_reg_match_utf8;
3866 if (max != REG_INFTY && max < loceol - scan)
3867 loceol = scan + max;
3872 while (scan < loceol && hardcount < max && *scan != '\n') {
3873 scan += UTF8SKIP(scan);
3877 while (scan < loceol && *scan != '\n')
3887 case EXACT: /* length of string is 1 */
3889 while (scan < loceol && UCHARAT(scan) == c)
3892 case EXACTF: /* length of string is 1 */
3894 while (scan < loceol &&
3895 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3898 case EXACTFL: /* length of string is 1 */
3899 PL_reg_flags |= RF_tainted;
3901 while (scan < loceol &&
3902 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3908 while (hardcount < max && scan < loceol &&
3909 reginclass(p, (U8*)scan, do_utf8)) {
3910 scan += UTF8SKIP(scan);
3914 while (scan < loceol && reginclass(p, (U8*)scan, do_utf8))
3921 LOAD_UTF8_CHARCLASS(alnum,"a");
3922 while (hardcount < max && scan < loceol &&
3923 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3924 scan += UTF8SKIP(scan);
3928 while (scan < loceol && isALNUM(*scan))
3933 PL_reg_flags |= RF_tainted;
3936 while (hardcount < max && scan < loceol &&
3937 isALNUM_LC_utf8((U8*)scan)) {
3938 scan += UTF8SKIP(scan);
3942 while (scan < loceol && isALNUM_LC(*scan))
3949 LOAD_UTF8_CHARCLASS(alnum,"a");
3950 while (hardcount < max && scan < loceol &&
3951 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3952 scan += UTF8SKIP(scan);
3956 while (scan < loceol && !isALNUM(*scan))
3961 PL_reg_flags |= RF_tainted;
3964 while (hardcount < max && scan < loceol &&
3965 !isALNUM_LC_utf8((U8*)scan)) {
3966 scan += UTF8SKIP(scan);
3970 while (scan < loceol && !isALNUM_LC(*scan))
3977 LOAD_UTF8_CHARCLASS(space," ");
3978 while (hardcount < max && scan < loceol &&
3980 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3981 scan += UTF8SKIP(scan);
3985 while (scan < loceol && isSPACE(*scan))
3990 PL_reg_flags |= RF_tainted;
3993 while (hardcount < max && scan < loceol &&
3994 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3995 scan += UTF8SKIP(scan);
3999 while (scan < loceol && isSPACE_LC(*scan))
4006 LOAD_UTF8_CHARCLASS(space," ");
4007 while (hardcount < max && scan < loceol &&
4009 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4010 scan += UTF8SKIP(scan);
4014 while (scan < loceol && !isSPACE(*scan))
4019 PL_reg_flags |= RF_tainted;
4022 while (hardcount < max && scan < loceol &&
4023 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4024 scan += UTF8SKIP(scan);
4028 while (scan < loceol && !isSPACE_LC(*scan))
4035 LOAD_UTF8_CHARCLASS(digit,"0");
4036 while (hardcount < max && scan < loceol &&
4037 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4038 scan += UTF8SKIP(scan);
4042 while (scan < loceol && isDIGIT(*scan))
4049 LOAD_UTF8_CHARCLASS(digit,"0");
4050 while (hardcount < max && scan < loceol &&
4051 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4052 scan += UTF8SKIP(scan);
4056 while (scan < loceol && !isDIGIT(*scan))
4060 default: /* Called on something of 0 width. */
4061 break; /* So match right here or not at all. */
4067 c = scan - PL_reginput;
4072 SV *prop = sv_newmortal();
4075 PerlIO_printf(Perl_debug_log,
4076 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4077 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4084 - regrepeat_hard - repeatedly match something, report total lenth and length
4086 * The repeater is supposed to have constant length.
4090 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4092 register char *scan = Nullch;
4093 register char *start;
4094 register char *loceol = PL_regeol;
4096 I32 count = 0, res = 1;
4101 start = PL_reginput;
4102 if (PL_reg_match_utf8) {
4103 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4106 while (start < PL_reginput) {
4108 start += UTF8SKIP(start);
4119 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4121 *lp = l = PL_reginput - start;
4122 if (max != REG_INFTY && l*max < loceol - scan)
4123 loceol = scan + l*max;
4136 - regclass_swash - prepare the utf8 swash
4140 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** initsvp)
4145 if (PL_regdata && PL_regdata->count) {
4148 if (PL_regdata->what[n] == 's') {
4149 SV *rv = (SV*)PL_regdata->data[n];
4150 AV *av = (AV*)SvRV((SV*)rv);
4153 si = *av_fetch(av, 0, FALSE);
4154 a = av_fetch(av, 1, FALSE);
4158 else if (si && doinit) {
4159 sw = swash_init("utf8", "", si, 1, 0);
4160 (void)av_store(av, 1, sw);
4172 - reginclass - determine if a character falls into a character class
4176 S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8)
4178 char flags = ANYOF_FLAGS(n);
4183 c = do_utf8 ? utf8_to_uvchr(p, &len) : *p;
4185 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4186 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4187 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4190 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4193 SV *sw = regclass_swash(n, TRUE, 0);
4196 if (swash_fetch(sw, p, do_utf8))
4198 else if (flags & ANYOF_FOLD) {
4200 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
4202 to_utf8_fold(p, tmpbuf, &ulen);
4203 if (swash_fetch(sw, tmpbuf, do_utf8))
4205 to_utf8_upper(p, tmpbuf, &ulen);
4206 if (swash_fetch(sw, tmpbuf, do_utf8))
4212 if (!match && c < 256) {
4213 if (ANYOF_BITMAP_TEST(n, c))
4215 else if (flags & ANYOF_FOLD) {
4218 if (flags & ANYOF_LOCALE) {
4219 PL_reg_flags |= RF_tainted;
4220 f = PL_fold_locale[c];
4224 if (f != c && ANYOF_BITMAP_TEST(n, f))
4228 if (!match && (flags & ANYOF_CLASS)) {
4229 PL_reg_flags |= RF_tainted;
4231 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4232 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4233 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4234 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4235 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4236 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4237 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4238 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4239 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4240 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4241 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4242 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4243 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4244 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4245 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4246 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4247 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4248 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4249 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4250 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4251 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4252 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4253 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4254 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4255 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4256 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4257 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4258 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4259 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4260 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4261 ) /* How's that for a conditional? */
4268 return (flags & ANYOF_INVERT) ? !match : match;
4272 S_reghop(pTHX_ U8 *s, I32 off)
4274 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4278 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4281 while (off-- && s < lim) {
4282 /* XXX could check well-formedness here */
4290 if (UTF8_IS_CONTINUED(*s)) {
4291 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4294 /* XXX could check well-formedness here */
4302 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4304 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4308 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4311 while (off-- && s < lim) {
4312 /* XXX could check well-formedness here */
4322 if (UTF8_IS_CONTINUED(*s)) {
4323 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4326 /* XXX could check well-formedness here */
4338 restore_pos(pTHX_ void *arg)
4340 if (PL_reg_eval_set) {
4341 if (PL_reg_oldsaved) {
4342 PL_reg_re->subbeg = PL_reg_oldsaved;
4343 PL_reg_re->sublen = PL_reg_oldsavedlen;
4344 RX_MATCH_COPIED_on(PL_reg_re);
4346 PL_reg_magic->mg_len = PL_reg_oldpos;
4347 PL_reg_eval_set = 0;
4348 PL_curpm = PL_reg_oldcurpm;