5 * "One Ring to rule them all, One Ring to find them..."
8 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
9 * confused with the original package (see point 3 below). Thanks, Henry!
12 /* Additional note: this code is very heavily munged from Henry's version
13 * in places. In some spots I've traded clarity for efficiency, so don't
14 * blame Henry for some of the lack of readability.
17 /* The names of the functions have been changed from regcomp and
18 * regexec to pregcomp and pregexec in order to avoid conflicts
19 * with the POSIX routines of the same names.
22 #ifdef PERL_EXT_RE_BUILD
23 /* need to replace pregcomp et al, so enable that */
24 # ifndef PERL_IN_XSUB_RE
25 # define PERL_IN_XSUB_RE
27 /* need access to debugger hooks */
28 # if defined(PERL_EXT_RE_DEBUG) && !defined(DEBUGGING)
33 #ifdef PERL_IN_XSUB_RE
34 /* We *really* need to overwrite these symbols: */
35 # define Perl_regexec_flags my_regexec
36 # define Perl_regdump my_regdump
37 # define Perl_regprop my_regprop
38 # define Perl_re_intuit_start my_re_intuit_start
39 /* *These* symbols are masked to allow static link. */
40 # define Perl_pregexec my_pregexec
41 # define Perl_reginitcolors my_reginitcolors
42 # define Perl_regclass_swash my_regclass_swash
44 # define PERL_NO_GET_CONTEXT
49 * pregcomp and pregexec -- regsub and regerror are not used in perl
51 * Copyright (c) 1986 by University of Toronto.
52 * Written by Henry Spencer. Not derived from licensed software.
54 * Permission is granted to anyone to use this software for any
55 * purpose on any computer system, and to redistribute it freely,
56 * subject to the following restrictions:
58 * 1. The author is not responsible for the consequences of use of
59 * this software, no matter how awful, even if they arise
62 * 2. The origin of this software must not be misrepresented, either
63 * by explicit claim or by omission.
65 * 3. Altered versions must be plainly marked as such, and must not
66 * be misrepresented as being the original software.
68 **** Alterations to Henry's code are...
70 **** Copyright (c) 1991-2001, Larry Wall
72 **** You may distribute under the terms of either the GNU General Public
73 **** License or the Artistic License, as specified in the README file.
75 * Beware that some of this code is subtly aware of the way operator
76 * precedence is structured in regular expressions. Serious changes in
77 * regular-expression syntax might require a total rethink.
80 #define PERL_IN_REGEXEC_C
85 #define RF_tainted 1 /* tainted information used? */
86 #define RF_warned 2 /* warned about big count? */
87 #define RF_evaled 4 /* Did an EVAL with setting? */
88 #define RF_utf8 8 /* String contains multibyte chars? */
90 #define UTF (PL_reg_flags & RF_utf8)
92 #define RS_init 1 /* eval environment created */
93 #define RS_set 2 /* replsv value is set */
103 #define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
104 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
106 #define reghop_c(pos,off) ((char*)reghop((U8*)pos, off))
107 #define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off))
108 #define HOP(pos,off) (PL_reg_match_utf8 ? reghop((U8*)pos, off) : (U8*)(pos + off))
109 #define HOPMAYBE(pos,off) (PL_reg_match_utf8 ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off))
110 #define HOPc(pos,off) ((char*)HOP(pos,off))
111 #define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off))
113 #define HOPBACK(pos, off) ( \
114 (UTF && PL_reg_match_utf8) \
115 ? reghopmaybe((U8*)pos, -off) \
116 : (pos - off >= PL_bostr) \
120 #define HOPBACKc(pos, off) (char*)HOPBACK(pos, off)
122 #define reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim))
123 #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
124 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
125 #define HOPMAYBE3(pos,off,lim) (PL_reg_match_utf8 ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
126 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
127 #define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim))
129 #define LOAD_UTF8_CHARCLASS(a,b) STMT_START { if (!CAT2(PL_utf8_,a)) (void)CAT2(is_utf8_, a)((U8*)b); } STMT_END
131 /* for use after a quantifier and before an EXACT-like node -- japhy */
132 #define JUMPABLE(rn) ( \
133 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
134 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
135 OP(rn) == PLUS || OP(rn) == MINMOD || \
136 (PL_regkind[(U8)OP(rn)] == CURLY && ARG1(rn) > 0) \
139 #define HAS_TEXT(rn) ( \
140 PL_regkind[(U8)OP(rn)] == EXACT || PL_regkind[(U8)OP(rn)] == REF \
143 #define FIND_NEXT_IMPT(rn) STMT_START { \
144 while (JUMPABLE(rn)) \
145 if (OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
146 PL_regkind[(U8)OP(rn)] == CURLY) \
147 rn = NEXTOPER(NEXTOPER(rn)); \
148 else if (OP(rn) == PLUS) \
150 else rn += NEXT_OFF(rn); \
153 static void restore_pos(pTHX_ void *arg);
156 S_regcppush(pTHX_ I32 parenfloor)
158 int retval = PL_savestack_ix;
159 #define REGCP_PAREN_ELEMS 4
160 int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
163 if (paren_elems_to_push < 0)
164 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
166 #define REGCP_OTHER_ELEMS 6
167 SSCHECK(paren_elems_to_push + REGCP_OTHER_ELEMS);
168 for (p = PL_regsize; p > parenfloor; p--) {
169 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
170 SSPUSHINT(PL_regendp[p]);
171 SSPUSHINT(PL_regstartp[p]);
172 SSPUSHPTR(PL_reg_start_tmp[p]);
175 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
176 SSPUSHINT(PL_regsize);
177 SSPUSHINT(*PL_reglastparen);
178 SSPUSHINT(*PL_reglastcloseparen);
179 SSPUSHPTR(PL_reginput);
180 #define REGCP_FRAME_ELEMS 2
181 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
182 * are needed for the regexp context stack bookkeeping. */
183 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
184 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
189 /* These are needed since we do not localize EVAL nodes: */
190 # define REGCP_SET(cp) DEBUG_r(PerlIO_printf(Perl_debug_log, \
191 " Setting an EVAL scope, savestack=%"IVdf"\n", \
192 (IV)PL_savestack_ix)); cp = PL_savestack_ix
194 # define REGCP_UNWIND(cp) DEBUG_r(cp != PL_savestack_ix ? \
195 PerlIO_printf(Perl_debug_log, \
196 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
197 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
207 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
209 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
210 i = SSPOPINT; /* Parentheses elements to pop. */
211 input = (char *) SSPOPPTR;
212 *PL_reglastcloseparen = SSPOPINT;
213 *PL_reglastparen = SSPOPINT;
214 PL_regsize = SSPOPINT;
216 /* Now restore the parentheses context. */
217 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
218 i > 0; i -= REGCP_PAREN_ELEMS) {
219 paren = (U32)SSPOPINT;
220 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
221 PL_regstartp[paren] = SSPOPINT;
223 if (paren <= *PL_reglastparen)
224 PL_regendp[paren] = tmps;
226 PerlIO_printf(Perl_debug_log,
227 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
228 (UV)paren, (IV)PL_regstartp[paren],
229 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
230 (IV)PL_regendp[paren],
231 (paren > *PL_reglastparen ? "(no)" : ""));
235 if (*PL_reglastparen + 1 <= PL_regnpar) {
236 PerlIO_printf(Perl_debug_log,
237 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
238 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
242 /* It would seem that the similar code in regtry()
243 * already takes care of this, and in fact it is in
244 * a better location to since this code can #if 0-ed out
245 * but the code in regtry() is needed or otherwise tests
246 * requiring null fields (pat.t#187 and split.t#{13,14}
247 * (as of patchlevel 7877) will fail. Then again,
248 * this code seems to be necessary or otherwise
249 * building DynaLoader will fail:
250 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
252 for (paren = *PL_reglastparen + 1; paren <= PL_regnpar; paren++) {
253 if (paren > PL_regsize)
254 PL_regstartp[paren] = -1;
255 PL_regendp[paren] = -1;
262 S_regcp_set_to(pTHX_ I32 ss)
264 I32 tmp = PL_savestack_ix;
266 PL_savestack_ix = ss;
268 PL_savestack_ix = tmp;
272 typedef struct re_cc_state
276 struct re_cc_state *prev;
281 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
283 #define TRYPAREN(paren, n, input) { \
286 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
287 PL_regendp[paren] = input - PL_bostr; \
290 PL_regendp[paren] = -1; \
292 if (regmatch(next)) \
295 PL_regendp[paren] = -1; \
300 * pregexec and friends
304 - pregexec - match a regexp against a string
307 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
308 char *strbeg, I32 minend, SV *screamer, U32 nosave)
309 /* strend: pointer to null at end of string */
310 /* strbeg: real beginning of string */
311 /* minend: end of match must be >=minend after stringarg. */
312 /* nosave: For optimizations. */
315 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
316 nosave ? 0 : REXEC_COPY_STR);
320 S_cache_re(pTHX_ regexp *prog)
322 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
324 PL_regprogram = prog->program;
326 PL_regnpar = prog->nparens;
327 PL_regdata = prog->data;
332 * Need to implement the following flags for reg_anch:
334 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
336 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
337 * INTUIT_AUTORITATIVE_ML
338 * INTUIT_ONCE_NOML - Intuit can match in one location only.
341 * Another flag for this function: SECOND_TIME (so that float substrs
342 * with giant delta may be not rechecked).
345 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
347 /* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
348 Otherwise, only SvCUR(sv) is used to get strbeg. */
350 /* XXXX We assume that strpos is strbeg unless sv. */
352 /* XXXX Some places assume that there is a fixed substring.
353 An update may be needed if optimizer marks as "INTUITable"
354 RExen without fixed substrings. Similarly, it is assumed that
355 lengths of all the strings are no more than minlen, thus they
356 cannot come from lookahead.
357 (Or minlen should take into account lookahead.) */
359 /* A failure to find a constant substring means that there is no need to make
360 an expensive call to REx engine, thus we celebrate a failure. Similarly,
361 finding a substring too deep into the string means that less calls to
362 regtry() should be needed.
364 REx compiler's optimizer found 4 possible hints:
365 a) Anchored substring;
367 c) Whether we are anchored (beginning-of-line or \G);
368 d) First node (of those at offset 0) which may distingush positions;
369 We use a)b)d) and multiline-part of c), and try to find a position in the
370 string which does not contradict any of them.
373 /* Most of decisions we do here should have been done at compile time.
374 The nodes of the REx which we used for the search should have been
375 deleted from the finite automaton. */
378 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
379 char *strend, U32 flags, re_scream_pos_data *data)
381 register I32 start_shift = 0;
382 /* Should be nonnegative! */
383 register I32 end_shift = 0;
389 register char *other_last = Nullch; /* other substr checked before this */
390 char *check_at = Nullch; /* check substr found at this pos */
392 char *i_strpos = strpos;
393 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
396 if (prog->reganch & ROPT_UTF8) {
397 DEBUG_r(PerlIO_printf(Perl_debug_log,
398 "UTF-8 regex...\n"));
399 PL_reg_flags |= RF_utf8;
403 char *s = PL_reg_match_utf8 ?
404 sv_uni_display(dsv, sv, 60, 0) : strpos;
405 int len = PL_reg_match_utf8 ?
406 strlen(s) : strend - strpos;
409 if (PL_reg_match_utf8)
410 DEBUG_r(PerlIO_printf(Perl_debug_log,
411 "UTF-8 target...\n"));
412 PerlIO_printf(Perl_debug_log,
413 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
414 PL_colors[4],PL_colors[5],PL_colors[0],
417 (strlen(prog->precomp) > 60 ? "..." : ""),
419 (int)(len > 60 ? 60 : len),
421 (len > 60 ? "..." : "")
425 if (prog->minlen > CHR_DIST((U8*)strend, (U8*)strpos)) {
426 DEBUG_r(PerlIO_printf(Perl_debug_log,
427 "String too short... [re_intuit_start]\n"));
430 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
432 check = prog->check_substr;
433 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
434 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
435 || ( (prog->reganch & ROPT_ANCH_BOL)
436 && !PL_multiline ) ); /* Check after \n? */
439 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
440 | ROPT_IMPLICIT)) /* not a real BOL */
441 /* SvCUR is not set on references: SvRV and SvPVX overlap */
443 && (strpos != strbeg)) {
444 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
447 if (prog->check_offset_min == prog->check_offset_max &&
448 !(prog->reganch & ROPT_CANY_SEEN)) {
449 /* Substring at constant offset from beg-of-str... */
452 s = HOP3c(strpos, prog->check_offset_min, strend);
454 slen = SvCUR(check); /* >= 1 */
456 if ( strend - s > slen || strend - s < slen - 1
457 || (strend - s == slen && strend[-1] != '\n')) {
458 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
461 /* Now should match s[0..slen-2] */
463 if (slen && (*SvPVX(check) != *s
465 && memNE(SvPVX(check), s, slen)))) {
467 DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
471 else if (*SvPVX(check) != *s
472 || ((slen = SvCUR(check)) > 1
473 && memNE(SvPVX(check), s, slen)))
475 goto success_at_start;
478 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
480 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
481 end_shift = prog->minlen - start_shift -
482 CHR_SVLEN(check) + (SvTAIL(check) != 0);
484 I32 end = prog->check_offset_max + CHR_SVLEN(check)
485 - (SvTAIL(check) != 0);
486 I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
488 if (end_shift < eshift)
492 else { /* Can match at random position */
495 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
496 /* Should be nonnegative! */
497 end_shift = prog->minlen - start_shift -
498 CHR_SVLEN(check) + (SvTAIL(check) != 0);
501 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
503 Perl_croak(aTHX_ "panic: end_shift");
507 /* Find a possible match in the region s..strend by looking for
508 the "check" substring in the region corrected by start/end_shift. */
509 if (flags & REXEC_SCREAM) {
510 I32 p = -1; /* Internal iterator of scream. */
511 I32 *pp = data ? data->scream_pos : &p;
513 if (PL_screamfirst[BmRARE(check)] >= 0
514 || ( BmRARE(check) == '\n'
515 && (BmPREVIOUS(check) == SvCUR(check) - 1)
517 s = screaminstr(sv, check,
518 start_shift + (s - strbeg), end_shift, pp, 0);
522 *data->scream_olds = s;
524 else if (prog->reganch & ROPT_CANY_SEEN)
525 s = fbm_instr((U8*)(s + start_shift),
526 (U8*)(strend - end_shift),
527 check, PL_multiline ? FBMrf_MULTILINE : 0);
529 s = fbm_instr(HOP3(s, start_shift, strend),
530 HOP3(strend, -end_shift, strbeg),
531 check, PL_multiline ? FBMrf_MULTILINE : 0);
533 /* Update the count-of-usability, remove useless subpatterns,
536 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
537 (s ? "Found" : "Did not find"),
538 ((check == prog->anchored_substr) ? "anchored" : "floating"),
540 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
542 PL_colors[1], (SvTAIL(check) ? "$" : ""),
543 (s ? " at offset " : "...\n") ) );
550 /* Finish the diagnostic message */
551 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
553 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
554 Start with the other substr.
555 XXXX no SCREAM optimization yet - and a very coarse implementation
556 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
557 *always* match. Probably should be marked during compile...
558 Probably it is right to do no SCREAM here...
561 if (prog->float_substr && prog->anchored_substr) {
562 /* Take into account the "other" substring. */
563 /* XXXX May be hopelessly wrong for UTF... */
566 if (check == prog->float_substr) {
569 char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
572 t = s - prog->check_offset_max;
573 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
574 && (!(prog->reganch & ROPT_UTF8)
575 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
580 t = HOP3c(t, prog->anchored_offset, strend);
581 if (t < other_last) /* These positions already checked */
583 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
586 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
587 /* On end-of-str: see comment below. */
588 s = fbm_instr((unsigned char*)t,
589 HOP3(HOP3(last1, prog->anchored_offset, strend)
590 + SvCUR(prog->anchored_substr),
591 -(SvTAIL(prog->anchored_substr)!=0), strbeg),
592 prog->anchored_substr,
593 PL_multiline ? FBMrf_MULTILINE : 0);
594 DEBUG_r(PerlIO_printf(Perl_debug_log,
595 "%s anchored substr `%s%.*s%s'%s",
596 (s ? "Found" : "Contradicts"),
598 (int)(SvCUR(prog->anchored_substr)
599 - (SvTAIL(prog->anchored_substr)!=0)),
600 SvPVX(prog->anchored_substr),
601 PL_colors[1], (SvTAIL(prog->anchored_substr) ? "$" : "")));
603 if (last1 >= last2) {
604 DEBUG_r(PerlIO_printf(Perl_debug_log,
605 ", giving up...\n"));
608 DEBUG_r(PerlIO_printf(Perl_debug_log,
609 ", trying floating at offset %ld...\n",
610 (long)(HOP3c(s1, 1, strend) - i_strpos)));
611 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
612 s = HOP3c(last, 1, strend);
616 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
617 (long)(s - i_strpos)));
618 t = HOP3c(s, -prog->anchored_offset, strbeg);
619 other_last = HOP3c(s, 1, strend);
627 else { /* Take into account the floating substring. */
631 t = HOP3c(s, -start_shift, strbeg);
633 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
634 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
635 last = HOP3c(t, prog->float_max_offset, strend);
636 s = HOP3c(t, prog->float_min_offset, strend);
639 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
640 /* fbm_instr() takes into account exact value of end-of-str
641 if the check is SvTAIL(ed). Since false positives are OK,
642 and end-of-str is not later than strend we are OK. */
643 s = fbm_instr((unsigned char*)s,
644 (unsigned char*)last + SvCUR(prog->float_substr)
645 - (SvTAIL(prog->float_substr)!=0),
646 prog->float_substr, PL_multiline ? FBMrf_MULTILINE : 0);
647 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
648 (s ? "Found" : "Contradicts"),
650 (int)(SvCUR(prog->float_substr)
651 - (SvTAIL(prog->float_substr)!=0)),
652 SvPVX(prog->float_substr),
653 PL_colors[1], (SvTAIL(prog->float_substr) ? "$" : "")));
656 DEBUG_r(PerlIO_printf(Perl_debug_log,
657 ", giving up...\n"));
660 DEBUG_r(PerlIO_printf(Perl_debug_log,
661 ", trying anchored starting at offset %ld...\n",
662 (long)(s1 + 1 - i_strpos)));
664 s = HOP3c(t, 1, strend);
668 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
669 (long)(s - i_strpos)));
670 other_last = s; /* Fix this later. --Hugo */
679 t = s - prog->check_offset_max;
680 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
681 && (!(prog->reganch & ROPT_UTF8)
682 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
684 /* Fixed substring is found far enough so that the match
685 cannot start at strpos. */
687 if (ml_anch && t[-1] != '\n') {
688 /* Eventually fbm_*() should handle this, but often
689 anchored_offset is not 0, so this check will not be wasted. */
690 /* XXXX In the code below we prefer to look for "^" even in
691 presence of anchored substrings. And we search even
692 beyond the found float position. These pessimizations
693 are historical artefacts only. */
695 while (t < strend - prog->minlen) {
697 if (t < check_at - prog->check_offset_min) {
698 if (prog->anchored_substr) {
699 /* Since we moved from the found position,
700 we definitely contradict the found anchored
701 substr. Due to the above check we do not
702 contradict "check" substr.
703 Thus we can arrive here only if check substr
704 is float. Redo checking for "other"=="fixed".
707 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
708 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
709 goto do_other_anchored;
711 /* We don't contradict the found floating substring. */
712 /* XXXX Why not check for STCLASS? */
714 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
715 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
718 /* Position contradicts check-string */
719 /* XXXX probably better to look for check-string
720 than for "\n", so one should lower the limit for t? */
721 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
722 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
723 other_last = strpos = s = t + 1;
728 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
729 PL_colors[0],PL_colors[1]));
733 DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
734 PL_colors[0],PL_colors[1]));
738 ++BmUSEFUL(prog->check_substr); /* hooray/5 */
741 /* The found string does not prohibit matching at strpos,
742 - no optimization of calling REx engine can be performed,
743 unless it was an MBOL and we are not after MBOL,
744 or a future STCLASS check will fail this. */
746 /* Even in this situation we may use MBOL flag if strpos is offset
747 wrt the start of the string. */
748 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
749 && (strpos != strbeg) && strpos[-1] != '\n'
750 /* May be due to an implicit anchor of m{.*foo} */
751 && !(prog->reganch & ROPT_IMPLICIT))
756 DEBUG_r( if (ml_anch)
757 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
758 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
761 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
762 && prog->check_substr /* Could be deleted already */
763 && --BmUSEFUL(prog->check_substr) < 0
764 && prog->check_substr == prog->float_substr)
766 /* If flags & SOMETHING - do not do it many times on the same match */
767 DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
768 SvREFCNT_dec(prog->check_substr);
769 prog->check_substr = Nullsv; /* disable */
770 prog->float_substr = Nullsv; /* clear */
771 check = Nullsv; /* abort */
773 /* XXXX This is a remnant of the old implementation. It
774 looks wasteful, since now INTUIT can use many
776 prog->reganch &= ~RE_USE_INTUIT;
783 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
784 if (prog->regstclass) {
785 /* minlen == 0 is possible if regstclass is \b or \B,
786 and the fixed substr is ''$.
787 Since minlen is already taken into account, s+1 is before strend;
788 accidentally, minlen >= 1 guaranties no false positives at s + 1
789 even for \b or \B. But (minlen? 1 : 0) below assumes that
790 regstclass does not come from lookahead... */
791 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
792 This leaves EXACTF only, which is dealt with in find_byclass(). */
793 U8* str = (U8*)STRING(prog->regstclass);
794 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
795 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
797 char *endpos = (prog->anchored_substr || ml_anch)
798 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
799 : (prog->float_substr
800 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
803 char *startpos = strbeg;
806 if (prog->reganch & ROPT_UTF8) {
807 PL_regdata = prog->data;
810 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
815 if (endpos == strend) {
816 DEBUG_r( PerlIO_printf(Perl_debug_log,
817 "Could not match STCLASS...\n") );
820 DEBUG_r( PerlIO_printf(Perl_debug_log,
821 "This position contradicts STCLASS...\n") );
822 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
824 /* Contradict one of substrings */
825 if (prog->anchored_substr) {
826 if (prog->anchored_substr == check) {
827 DEBUG_r( what = "anchored" );
829 s = HOP3c(t, 1, strend);
830 if (s + start_shift + end_shift > strend) {
831 /* XXXX Should be taken into account earlier? */
832 DEBUG_r( PerlIO_printf(Perl_debug_log,
833 "Could not match STCLASS...\n") );
838 DEBUG_r( PerlIO_printf(Perl_debug_log,
839 "Looking for %s substr starting at offset %ld...\n",
840 what, (long)(s + start_shift - i_strpos)) );
843 /* Have both, check_string is floating */
844 if (t + start_shift >= check_at) /* Contradicts floating=check */
845 goto retry_floating_check;
846 /* Recheck anchored substring, but not floating... */
850 DEBUG_r( PerlIO_printf(Perl_debug_log,
851 "Looking for anchored substr starting at offset %ld...\n",
852 (long)(other_last - i_strpos)) );
853 goto do_other_anchored;
855 /* Another way we could have checked stclass at the
856 current position only: */
861 DEBUG_r( PerlIO_printf(Perl_debug_log,
862 "Looking for /%s^%s/m starting at offset %ld...\n",
863 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
866 if (!prog->float_substr) /* Could have been deleted */
868 /* Check is floating subtring. */
869 retry_floating_check:
870 t = check_at - start_shift;
871 DEBUG_r( what = "floating" );
872 goto hop_and_restart;
875 DEBUG_r(PerlIO_printf(Perl_debug_log,
876 "By STCLASS: moving %ld --> %ld\n",
877 (long)(t - i_strpos), (long)(s - i_strpos))
881 DEBUG_r(PerlIO_printf(Perl_debug_log,
882 "Does not contradict STCLASS...\n");
887 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
888 PL_colors[4], (check ? "Guessed" : "Giving up"),
889 PL_colors[5], (long)(s - i_strpos)) );
892 fail_finish: /* Substring not found */
893 if (prog->check_substr) /* could be removed already */
894 BmUSEFUL(prog->check_substr) += 5; /* hooray */
896 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
897 PL_colors[4],PL_colors[5]));
901 /* We know what class REx starts with. Try to find this position... */
903 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
905 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
911 register I32 tmp = 1; /* Scratch variable? */
912 register bool do_utf8 = PL_reg_match_utf8;
914 /* We know what class it must start with. */
918 if (reginclass(c, (U8*)s, do_utf8)) {
919 if (tmp && (norun || regtry(prog, s)))
926 s += do_utf8 ? UTF8SKIP(s) : 1;
931 if (tmp && (norun || regtry(prog, s)))
943 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
944 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
946 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
947 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
949 c1 = utf8_to_uvuni(tmpbuf1, 0);
950 c2 = utf8_to_uvuni(tmpbuf2, 0);
961 c2 = PL_fold_locale[c1];
966 e = s; /* Due to minlen logic of intuit() */
970 U8 tmpbuf [UTF8_MAXLEN+1];
971 U8 foldbuf[UTF8_MAXLEN_FOLD+1];
974 /* The ibcmp_utf8() uses to_uni_fold() which is more
975 * correct folding for Unicode than using lowercase.
976 * However, it doesn't work quite fully since the folding
977 * is a one-to-many mapping and the regex optimizer is
978 * unaware of this, so it may throw out good matches.
979 * Fortunately, not getting this right is allowed
980 * for Unicode Regular Expression Support level 1,
981 * only one-to-one matching is required. --jhi */
985 c = utf8_to_uvchr((U8*)s, &len);
988 !ibcmp_utf8(s, do_utf8, strend - s,
990 && (norun || regtry(prog, s)) )
993 uvchr_to_utf8(tmpbuf, c);
994 to_utf8_fold(tmpbuf, foldbuf, &foldlen);
995 f = utf8_to_uvchr(foldbuf, 0);
997 && (f == c1 || f == c2)
999 !ibcmp_utf8((char *)foldbuf,
1002 && (norun || regtry(prog, s)) )
1010 c = utf8_to_uvchr((U8*)s, &len);
1012 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1013 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1014 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1016 if ( (c == c1 || c == c2)
1018 !ibcmp_utf8(s, do_utf8, strend - s,
1020 && (norun || regtry(prog, s)) )
1023 uvchr_to_utf8(tmpbuf, c);
1024 to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1025 f = utf8_to_uvchr(foldbuf, 0);
1027 && (f == c1 || f == c2)
1028 && (ln == foldlen ||
1029 !ibcmp_utf8((char *)foldbuf,
1032 && (norun || regtry(prog, s)) )
1043 && (ln == 1 || !(OP(c) == EXACTF
1045 : ibcmp_locale(s, m, ln)))
1046 && (norun || regtry(prog, s)) )
1052 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1053 && (ln == 1 || !(OP(c) == EXACTF
1055 : ibcmp_locale(s, m, ln)))
1056 && (norun || regtry(prog, s)) )
1063 PL_reg_flags |= RF_tainted;
1070 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1073 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1075 tmp = ((OP(c) == BOUND ?
1076 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1077 LOAD_UTF8_CHARCLASS(alnum,"a");
1078 while (s < strend) {
1079 if (tmp == !(OP(c) == BOUND ?
1080 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1081 isALNUM_LC_utf8((U8*)s)))
1084 if ((norun || regtry(prog, s)))
1091 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1092 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1093 while (s < strend) {
1095 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1097 if ((norun || regtry(prog, s)))
1103 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1107 PL_reg_flags |= RF_tainted;
1114 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1117 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1119 tmp = ((OP(c) == NBOUND ?
1120 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1121 LOAD_UTF8_CHARCLASS(alnum,"a");
1122 while (s < strend) {
1123 if (tmp == !(OP(c) == NBOUND ?
1124 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1125 isALNUM_LC_utf8((U8*)s)))
1127 else if ((norun || regtry(prog, s)))
1133 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1134 tmp = ((OP(c) == NBOUND ?
1135 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1136 while (s < strend) {
1138 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1140 else if ((norun || regtry(prog, s)))
1145 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1150 LOAD_UTF8_CHARCLASS(alnum,"a");
1151 while (s < strend) {
1152 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1153 if (tmp && (norun || regtry(prog, s)))
1164 while (s < strend) {
1166 if (tmp && (norun || regtry(prog, s)))
1178 PL_reg_flags |= RF_tainted;
1180 while (s < strend) {
1181 if (isALNUM_LC_utf8((U8*)s)) {
1182 if (tmp && (norun || regtry(prog, s)))
1193 while (s < strend) {
1194 if (isALNUM_LC(*s)) {
1195 if (tmp && (norun || regtry(prog, s)))
1208 LOAD_UTF8_CHARCLASS(alnum,"a");
1209 while (s < strend) {
1210 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1211 if (tmp && (norun || regtry(prog, s)))
1222 while (s < strend) {
1224 if (tmp && (norun || regtry(prog, s)))
1236 PL_reg_flags |= RF_tainted;
1238 while (s < strend) {
1239 if (!isALNUM_LC_utf8((U8*)s)) {
1240 if (tmp && (norun || regtry(prog, s)))
1251 while (s < strend) {
1252 if (!isALNUM_LC(*s)) {
1253 if (tmp && (norun || regtry(prog, s)))
1266 LOAD_UTF8_CHARCLASS(space," ");
1267 while (s < strend) {
1268 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1269 if (tmp && (norun || regtry(prog, s)))
1280 while (s < strend) {
1282 if (tmp && (norun || regtry(prog, s)))
1294 PL_reg_flags |= RF_tainted;
1296 while (s < strend) {
1297 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1298 if (tmp && (norun || regtry(prog, s)))
1309 while (s < strend) {
1310 if (isSPACE_LC(*s)) {
1311 if (tmp && (norun || regtry(prog, s)))
1324 LOAD_UTF8_CHARCLASS(space," ");
1325 while (s < strend) {
1326 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1327 if (tmp && (norun || regtry(prog, s)))
1338 while (s < strend) {
1340 if (tmp && (norun || regtry(prog, s)))
1352 PL_reg_flags |= RF_tainted;
1354 while (s < strend) {
1355 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1356 if (tmp && (norun || regtry(prog, s)))
1367 while (s < strend) {
1368 if (!isSPACE_LC(*s)) {
1369 if (tmp && (norun || regtry(prog, s)))
1382 LOAD_UTF8_CHARCLASS(digit,"0");
1383 while (s < strend) {
1384 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1385 if (tmp && (norun || regtry(prog, s)))
1396 while (s < strend) {
1398 if (tmp && (norun || regtry(prog, s)))
1410 PL_reg_flags |= RF_tainted;
1412 while (s < strend) {
1413 if (isDIGIT_LC_utf8((U8*)s)) {
1414 if (tmp && (norun || regtry(prog, s)))
1425 while (s < strend) {
1426 if (isDIGIT_LC(*s)) {
1427 if (tmp && (norun || regtry(prog, s)))
1440 LOAD_UTF8_CHARCLASS(digit,"0");
1441 while (s < strend) {
1442 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1443 if (tmp && (norun || regtry(prog, s)))
1454 while (s < strend) {
1456 if (tmp && (norun || regtry(prog, s)))
1468 PL_reg_flags |= RF_tainted;
1470 while (s < strend) {
1471 if (!isDIGIT_LC_utf8((U8*)s)) {
1472 if (tmp && (norun || regtry(prog, s)))
1483 while (s < strend) {
1484 if (!isDIGIT_LC(*s)) {
1485 if (tmp && (norun || regtry(prog, s)))
1497 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1506 - regexec_flags - match a regexp against a string
1509 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1510 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1511 /* strend: pointer to null at end of string */
1512 /* strbeg: real beginning of string */
1513 /* minend: end of match must be >=minend after stringarg. */
1514 /* data: May be used for some additional optimizations. */
1515 /* nosave: For optimizations. */
1518 register regnode *c;
1519 register char *startpos = stringarg;
1520 I32 minlen; /* must match at least this many chars */
1521 I32 dontbother = 0; /* how many characters not to try at end */
1522 /* I32 start_shift = 0; */ /* Offset of the start to find
1523 constant substr. */ /* CC */
1524 I32 end_shift = 0; /* Same for the end. */ /* CC */
1525 I32 scream_pos = -1; /* Internal iterator of scream. */
1527 SV* oreplsv = GvSV(PL_replgv);
1528 bool do_utf8 = DO_UTF8(sv);
1530 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
1537 PL_regnarrate = DEBUG_r_TEST;
1540 /* Be paranoid... */
1541 if (prog == NULL || startpos == NULL) {
1542 Perl_croak(aTHX_ "NULL regexp parameter");
1546 minlen = prog->minlen;
1547 if (strend - startpos < minlen) {
1548 DEBUG_r(PerlIO_printf(Perl_debug_log,
1549 "String too short [regexec_flags]...\n"));
1553 /* Check validity of program. */
1554 if (UCHARAT(prog->program) != REG_MAGIC) {
1555 Perl_croak(aTHX_ "corrupted regexp program");
1559 PL_reg_eval_set = 0;
1562 if (prog->reganch & ROPT_UTF8)
1563 PL_reg_flags |= RF_utf8;
1565 /* Mark beginning of line for ^ and lookbehind. */
1566 PL_regbol = startpos;
1570 /* Mark end of line for $ (and such) */
1573 /* see how far we have to get to not match where we matched before */
1574 PL_regtill = startpos+minend;
1576 /* We start without call_cc context. */
1579 /* If there is a "must appear" string, look for it. */
1582 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1585 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1586 PL_reg_ganch = startpos;
1587 else if (sv && SvTYPE(sv) >= SVt_PVMG
1589 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1590 && mg->mg_len >= 0) {
1591 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1592 if (prog->reganch & ROPT_ANCH_GPOS) {
1593 if (s > PL_reg_ganch)
1598 else /* pos() not defined */
1599 PL_reg_ganch = strbeg;
1602 if (do_utf8 == (UTF!=0) &&
1603 !(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
1604 re_scream_pos_data d;
1606 d.scream_olds = &scream_olds;
1607 d.scream_pos = &scream_pos;
1608 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1610 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1611 goto phooey; /* not present */
1616 char *s = do_utf8 ? sv_uni_display(dsv, sv, 60, 0) : startpos;
1617 int len = do_utf8 ? strlen(s) : strend - startpos;
1620 PerlIO_printf(Perl_debug_log,
1621 "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
1622 PL_colors[4],PL_colors[5],PL_colors[0],
1625 (strlen(prog->precomp) > 60 ? "..." : ""),
1627 (int)(len > 60 ? 60 : len),
1629 (len > 60 ? "..." : "")
1633 /* Simplest case: anchored match need be tried only once. */
1634 /* [unless only anchor is BOL and multiline is set] */
1635 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1636 if (s == startpos && regtry(prog, startpos))
1638 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1639 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1644 dontbother = minlen - 1;
1645 end = HOP3c(strend, -dontbother, strbeg) - 1;
1646 /* for multiline we only have to try after newlines */
1647 if (prog->check_substr) {
1651 if (regtry(prog, s))
1656 if (prog->reganch & RE_USE_INTUIT) {
1657 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1668 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1669 if (regtry(prog, s))
1676 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1677 if (regtry(prog, PL_reg_ganch))
1682 /* Messy cases: unanchored match. */
1683 if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
1684 /* we have /x+whatever/ */
1685 /* it must be a one character string (XXXX Except UTF?) */
1686 char ch = SvPVX(prog->anchored_substr)[0];
1692 while (s < strend) {
1694 DEBUG_r( did_match = 1 );
1695 if (regtry(prog, s)) goto got_it;
1697 while (s < strend && *s == ch)
1704 while (s < strend) {
1706 DEBUG_r( did_match = 1 );
1707 if (regtry(prog, s)) goto got_it;
1709 while (s < strend && *s == ch)
1715 DEBUG_r(if (!did_match)
1716 PerlIO_printf(Perl_debug_log,
1717 "Did not find anchored character...\n")
1721 else if (do_utf8 == (UTF!=0) &&
1722 (prog->anchored_substr != Nullsv
1723 || (prog->float_substr != Nullsv
1724 && prog->float_max_offset < strend - s))) {
1725 SV *must = prog->anchored_substr
1726 ? prog->anchored_substr : prog->float_substr;
1728 prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset;
1730 prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
1731 char *last = HOP3c(strend, /* Cannot start after this */
1732 -(I32)(CHR_SVLEN(must)
1733 - (SvTAIL(must) != 0) + back_min), strbeg);
1734 char *last1; /* Last position checked before */
1740 last1 = HOPc(s, -1);
1742 last1 = s - 1; /* bogus */
1744 /* XXXX check_substr already used to find `s', can optimize if
1745 check_substr==must. */
1747 dontbother = end_shift;
1748 strend = HOPc(strend, -dontbother);
1749 while ( (s <= last) &&
1750 ((flags & REXEC_SCREAM)
1751 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1752 end_shift, &scream_pos, 0))
1753 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1754 (unsigned char*)strend, must,
1755 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1756 DEBUG_r( did_match = 1 );
1757 if (HOPc(s, -back_max) > last1) {
1758 last1 = HOPc(s, -back_min);
1759 s = HOPc(s, -back_max);
1762 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1764 last1 = HOPc(s, -back_min);
1768 while (s <= last1) {
1769 if (regtry(prog, s))
1775 while (s <= last1) {
1776 if (regtry(prog, s))
1782 DEBUG_r(if (!did_match)
1783 PerlIO_printf(Perl_debug_log,
1784 "Did not find %s substr `%s%.*s%s'%s...\n",
1785 ((must == prog->anchored_substr)
1786 ? "anchored" : "floating"),
1788 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1790 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1794 else if ((c = prog->regstclass)) {
1795 if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT)
1796 /* don't bother with what can't match */
1797 strend = HOPc(strend, -(minlen - 1));
1799 SV *prop = sv_newmortal();
1801 PerlIO_printf(Perl_debug_log, "Matching stclass `%s' against `%s'\n", SvPVX(prop), UTF ? sv_uni_display(dsv, sv, 60, 0) : s);
1803 if (find_byclass(prog, c, s, strend, startpos, 0))
1805 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1809 if (prog->float_substr != Nullsv) { /* Trim the end. */
1812 if (flags & REXEC_SCREAM) {
1813 last = screaminstr(sv, prog->float_substr, s - strbeg,
1814 end_shift, &scream_pos, 1); /* last one */
1816 last = scream_olds; /* Only one occurrence. */
1820 char *little = SvPV(prog->float_substr, len);
1822 if (SvTAIL(prog->float_substr)) {
1823 if (memEQ(strend - len + 1, little, len - 1))
1824 last = strend - len + 1;
1825 else if (!PL_multiline)
1826 last = memEQ(strend - len, little, len)
1827 ? strend - len : Nullch;
1833 last = rninstr(s, strend, little, little + len);
1835 last = strend; /* matching `$' */
1839 DEBUG_r(PerlIO_printf(Perl_debug_log,
1840 "%sCan't trim the tail, match fails (should not happen)%s\n",
1841 PL_colors[4],PL_colors[5]));
1842 goto phooey; /* Should not happen! */
1844 dontbother = strend - last + prog->float_min_offset;
1846 if (minlen && (dontbother < minlen))
1847 dontbother = minlen - 1;
1848 strend -= dontbother; /* this one's always in bytes! */
1849 /* We don't know much -- general case. */
1852 if (regtry(prog, s))
1861 if (regtry(prog, s))
1863 } while (s++ < strend);
1871 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1873 if (PL_reg_eval_set) {
1874 /* Preserve the current value of $^R */
1875 if (oreplsv != GvSV(PL_replgv))
1876 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1877 restored, the value remains
1879 restore_pos(aTHX_ 0);
1882 /* make sure $`, $&, $', and $digit will work later */
1883 if ( !(flags & REXEC_NOT_FIRST) ) {
1884 if (RX_MATCH_COPIED(prog)) {
1885 Safefree(prog->subbeg);
1886 RX_MATCH_COPIED_off(prog);
1888 if (flags & REXEC_COPY_STR) {
1889 I32 i = PL_regeol - startpos + (stringarg - strbeg);
1891 s = savepvn(strbeg, i);
1894 RX_MATCH_COPIED_on(prog);
1897 prog->subbeg = strbeg;
1898 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
1905 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
1906 PL_colors[4],PL_colors[5]));
1907 if (PL_reg_eval_set)
1908 restore_pos(aTHX_ 0);
1913 - regtry - try match at specific point
1915 STATIC I32 /* 0 failure, 1 success */
1916 S_regtry(pTHX_ regexp *prog, char *startpos)
1924 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
1926 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
1929 PL_reg_eval_set = RS_init;
1931 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
1932 (IV)(PL_stack_sp - PL_stack_base));
1934 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
1935 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
1936 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
1938 /* Apparently this is not needed, judging by wantarray. */
1939 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
1940 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
1943 /* Make $_ available to executed code. */
1944 if (PL_reg_sv != DEFSV) {
1945 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
1950 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
1951 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
1952 /* prepare for quick setting of pos */
1953 sv_magic(PL_reg_sv, (SV*)0,
1954 PERL_MAGIC_regex_global, Nullch, 0);
1955 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
1959 PL_reg_oldpos = mg->mg_len;
1960 SAVEDESTRUCTOR_X(restore_pos, 0);
1962 if (!PL_reg_curpm) {
1963 Newz(22,PL_reg_curpm, 1, PMOP);
1966 SV* repointer = newSViv(0);
1967 /* so we know which PL_regex_padav element is PL_reg_curpm */
1968 SvFLAGS(repointer) |= SVf_BREAK;
1969 av_push(PL_regex_padav,repointer);
1970 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
1971 PL_regex_pad = AvARRAY(PL_regex_padav);
1975 PM_SETRE(PL_reg_curpm, prog);
1976 PL_reg_oldcurpm = PL_curpm;
1977 PL_curpm = PL_reg_curpm;
1978 if (RX_MATCH_COPIED(prog)) {
1979 /* Here is a serious problem: we cannot rewrite subbeg,
1980 since it may be needed if this match fails. Thus
1981 $` inside (?{}) could fail... */
1982 PL_reg_oldsaved = prog->subbeg;
1983 PL_reg_oldsavedlen = prog->sublen;
1984 RX_MATCH_COPIED_off(prog);
1987 PL_reg_oldsaved = Nullch;
1988 prog->subbeg = PL_bostr;
1989 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
1991 prog->startp[0] = startpos - PL_bostr;
1992 PL_reginput = startpos;
1993 PL_regstartp = prog->startp;
1994 PL_regendp = prog->endp;
1995 PL_reglastparen = &prog->lastparen;
1996 PL_reglastcloseparen = &prog->lastcloseparen;
1997 prog->lastparen = 0;
1999 DEBUG_r(PL_reg_starttry = startpos);
2000 if (PL_reg_start_tmpl <= prog->nparens) {
2001 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2002 if(PL_reg_start_tmp)
2003 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2005 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2009 sv_setpvn(PERL_DEBUG_PAD(0), "", 0);
2010 sv_setpvn(PERL_DEBUG_PAD(1), "", 0);
2011 sv_setpvn(PERL_DEBUG_PAD(2), "", 0);
2014 /* XXXX What this code is doing here?!!! There should be no need
2015 to do this again and again, PL_reglastparen should take care of
2018 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2019 * Actually, the code in regcppop() (which Ilya may be meaning by
2020 * PL_reglastparen), is not needed at all by the test suite
2021 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2022 * enough, for building DynaLoader, or otherwise this
2023 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2024 * will happen. Meanwhile, this code *is* needed for the
2025 * above-mentioned test suite tests to succeed. The common theme
2026 * on those tests seems to be returning null fields from matches.
2031 if (prog->nparens) {
2032 for (i = prog->nparens; i > *PL_reglastparen; i--) {
2039 if (regmatch(prog->program + 1)) {
2040 prog->endp[0] = PL_reginput - PL_bostr;
2043 REGCP_UNWIND(lastcp);
2047 #define RE_UNWIND_BRANCH 1
2048 #define RE_UNWIND_BRANCHJ 2
2052 typedef struct { /* XX: makes sense to enlarge it... */
2056 } re_unwind_generic_t;
2069 } re_unwind_branch_t;
2071 typedef union re_unwind_t {
2073 re_unwind_generic_t generic;
2074 re_unwind_branch_t branch;
2077 #define sayYES goto yes
2078 #define sayNO goto no
2079 #define sayYES_FINAL goto yes_final
2080 #define sayYES_LOUD goto yes_loud
2081 #define sayNO_FINAL goto no_final
2082 #define sayNO_SILENT goto do_no
2083 #define saySAME(x) if (x) goto yes; else goto no
2085 #define REPORT_CODE_OFF 24
2088 - regmatch - main matching routine
2090 * Conceptually the strategy is simple: check to see whether the current
2091 * node matches, call self recursively to see whether the rest matches,
2092 * and then act accordingly. In practice we make some effort to avoid
2093 * recursion, in particular by going through "ordinary" nodes (that don't
2094 * need to know whether the rest of the match failed) by a loop instead of
2097 /* [lwall] I've hoisted the register declarations to the outer block in order to
2098 * maybe save a little bit of pushing and popping on the stack. It also takes
2099 * advantage of machines that use a register save mask on subroutine entry.
2101 STATIC I32 /* 0 failure, 1 success */
2102 S_regmatch(pTHX_ regnode *prog)
2104 register regnode *scan; /* Current node. */
2105 regnode *next; /* Next node. */
2106 regnode *inner; /* Next node in internal branch. */
2107 register I32 nextchr; /* renamed nextchr - nextchar colides with
2108 function of same name */
2109 register I32 n; /* no or next */
2110 register I32 ln = 0; /* len or last */
2111 register char *s = Nullch; /* operand or save */
2112 register char *locinput = PL_reginput;
2113 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2114 int minmod = 0, sw = 0, logical = 0;
2117 I32 firstcp = PL_savestack_ix;
2119 register bool do_utf8 = PL_reg_match_utf8;
2121 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2122 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2123 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2130 /* Note that nextchr is a byte even in UTF */
2131 nextchr = UCHARAT(locinput);
2133 while (scan != NULL) {
2136 SV *prop = sv_newmortal();
2137 int docolor = *PL_colors[0];
2138 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2139 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2140 /* The part of the string before starttry has one color
2141 (pref0_len chars), between starttry and current
2142 position another one (pref_len - pref0_len chars),
2143 after the current position the third one.
2144 We assume that pref0_len <= pref_len, otherwise we
2145 decrease pref0_len. */
2146 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2147 ? (5 + taill) - l : locinput - PL_bostr;
2150 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2152 pref0_len = pref_len - (locinput - PL_reg_starttry);
2153 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2154 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2155 ? (5 + taill) - pref_len : PL_regeol - locinput);
2156 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2160 if (pref0_len > pref_len)
2161 pref0_len = pref_len;
2162 regprop(prop, scan);
2166 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2168 locinput - pref_len;
2169 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2170 char *s1 = do_utf8 ?
2171 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2172 pref_len - pref0_len, 60, 0) :
2173 locinput - pref_len + pref0_len;
2174 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2175 char *s2 = do_utf8 ?
2176 pv_uni_display(dsv2, (U8*)locinput,
2177 PL_regeol - locinput, 60, 0) :
2179 int len2 = do_utf8 ? strlen(s2) : l;
2180 PerlIO_printf(Perl_debug_log,
2181 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2182 (IV)(locinput - PL_bostr),
2189 (docolor ? "" : "> <"),
2193 15 - l - pref_len + 1,
2195 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2200 next = scan + NEXT_OFF(scan);
2206 if (locinput == PL_bostr || (PL_multiline &&
2207 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2209 /* regtill = regbol; */
2214 if (locinput == PL_bostr ||
2215 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2221 if (locinput == PL_bostr)
2225 if (locinput == PL_reg_ganch)
2235 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2240 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2242 if (PL_regeol - locinput > 1)
2246 if (PL_regeol != locinput)
2250 if (!nextchr && locinput >= PL_regeol)
2253 locinput += PL_utf8skip[nextchr];
2254 if (locinput > PL_regeol)
2256 nextchr = UCHARAT(locinput);
2259 nextchr = UCHARAT(++locinput);
2262 if (!nextchr && locinput >= PL_regeol)
2264 nextchr = UCHARAT(++locinput);
2267 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2270 locinput += PL_utf8skip[nextchr];
2271 if (locinput > PL_regeol)
2273 nextchr = UCHARAT(locinput);
2276 nextchr = UCHARAT(++locinput);
2281 if (do_utf8 != (UTF!=0)) {
2282 /* The target and the pattern have differing utf8ness. */
2288 /* The target is utf8, the pattern is not utf8. */
2292 if (NATIVE_TO_UNI(*(U8*)s) !=
2293 utf8_to_uvchr((U8*)l, &ulen))
2300 /* The target is not utf8, the pattern is utf8. */
2304 if (NATIVE_TO_UNI(*((U8*)l)) !=
2305 utf8_to_uvchr((U8*)s, &ulen))
2312 nextchr = UCHARAT(locinput);
2315 /* The target and the pattern have the same utf8ness. */
2316 /* Inline the first character, for speed. */
2317 if (UCHARAT(s) != nextchr)
2319 if (PL_regeol - locinput < ln)
2321 if (ln > 1 && memNE(s, locinput, ln))
2324 nextchr = UCHARAT(locinput);
2327 PL_reg_flags |= RF_tainted;
2336 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
2338 if (do_utf8 != (UTF!=0)) {
2339 /* The target and the pattern have differing utf8ness. */
2340 STRLEN ulen1, ulen2;
2344 /* The target is utf8, the pattern is not utf8. */
2349 cs = to_uni_fold(NATIVE_TO_UNI(*(U8*)s),
2351 cl = utf8_to_uvchr((U8*)l, &ulen2);
2354 cl = to_uni_fold(cl, (U8*)l, &ulen2);
2355 if (ulen1 != ulen2 || cs != cl)
2363 /* The target is not utf8, the pattern is utf8. */
2368 cs = utf8_to_uvchr((U8*)s, &ulen1);
2370 cl = to_uni_fold(NATIVE_TO_UNI(*(U8*)l),
2374 cs = to_uni_fold(cs, (U8*)s, &ulen1);
2375 if (ulen1 != ulen2 || cs != cl)
2383 nextchr = UCHARAT(locinput);
2387 if (do_utf8 && UTF) {
2388 /* Both the target and the pattern are utf8. */
2394 if (UTF8SKIP(s) != UTF8SKIP(l) ||
2395 memNE(s, (char*)l, UTF8SKIP(s))) {
2396 U8 lfoldbuf[UTF8_MAXLEN_FOLD+1];
2399 to_utf8_fold((U8*)l, lfoldbuf, &lfoldlen);
2400 if (UTF8SKIP(s) != lfoldlen ||
2401 memNE(s, (char*)lfoldbuf, lfoldlen)) {
2402 U8 sfoldbuf[UTF8_MAXLEN_FOLD+1];
2405 to_utf8_fold((U8*)s, sfoldbuf, &sfoldlen);
2406 if (sfoldlen != lfoldlen ||
2407 memNE((char*)sfoldbuf,
2408 (char*)lfoldbuf, lfoldlen))
2416 nextchr = UCHARAT(locinput);
2421 /* Neither the target and the pattern are utf8. */
2423 /* Inline the first character, for speed. */
2424 if (UCHARAT(s) != nextchr &&
2425 UCHARAT(s) != ((OP(scan) == EXACTF)
2426 ? PL_fold : PL_fold_locale)[nextchr])
2428 if (PL_regeol - locinput < ln)
2430 if (ln > 1 && (OP(scan) == EXACTF
2431 ? ibcmp(s, locinput, ln)
2432 : ibcmp_locale(s, locinput, ln)))
2435 nextchr = UCHARAT(locinput);
2439 if (!reginclass(scan, (U8*)locinput, do_utf8))
2441 if (locinput >= PL_regeol)
2443 locinput += PL_utf8skip[nextchr];
2444 nextchr = UCHARAT(locinput);
2448 nextchr = UCHARAT(locinput);
2449 if (!reginclass(scan, (U8*)locinput, do_utf8))
2451 if (!nextchr && locinput >= PL_regeol)
2453 nextchr = UCHARAT(++locinput);
2457 PL_reg_flags |= RF_tainted;
2463 LOAD_UTF8_CHARCLASS(alnum,"a");
2464 if (!(OP(scan) == ALNUM
2465 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2466 : isALNUM_LC_utf8((U8*)locinput)))
2470 locinput += PL_utf8skip[nextchr];
2471 nextchr = UCHARAT(locinput);
2474 if (!(OP(scan) == ALNUM
2475 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2477 nextchr = UCHARAT(++locinput);
2480 PL_reg_flags |= RF_tainted;
2483 if (!nextchr && locinput >= PL_regeol)
2486 LOAD_UTF8_CHARCLASS(alnum,"a");
2487 if (OP(scan) == NALNUM
2488 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2489 : isALNUM_LC_utf8((U8*)locinput))
2493 locinput += PL_utf8skip[nextchr];
2494 nextchr = UCHARAT(locinput);
2497 if (OP(scan) == NALNUM
2498 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2500 nextchr = UCHARAT(++locinput);
2504 PL_reg_flags |= RF_tainted;
2508 /* was last char in word? */
2510 if (locinput == PL_bostr)
2513 U8 *r = reghop((U8*)locinput, -1);
2515 ln = utf8n_to_uvchr(r, s - (char*)r, 0, 0);
2517 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2518 ln = isALNUM_uni(ln);
2519 LOAD_UTF8_CHARCLASS(alnum,"a");
2520 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2523 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2524 n = isALNUM_LC_utf8((U8*)locinput);
2528 ln = (locinput != PL_bostr) ?
2529 UCHARAT(locinput - 1) : '\n';
2530 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2532 n = isALNUM(nextchr);
2535 ln = isALNUM_LC(ln);
2536 n = isALNUM_LC(nextchr);
2539 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2540 OP(scan) == BOUNDL))
2544 PL_reg_flags |= RF_tainted;
2550 if (UTF8_IS_CONTINUED(nextchr)) {
2551 LOAD_UTF8_CHARCLASS(space," ");
2552 if (!(OP(scan) == SPACE
2553 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2554 : isSPACE_LC_utf8((U8*)locinput)))
2558 locinput += PL_utf8skip[nextchr];
2559 nextchr = UCHARAT(locinput);
2562 if (!(OP(scan) == SPACE
2563 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2565 nextchr = UCHARAT(++locinput);
2568 if (!(OP(scan) == SPACE
2569 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2571 nextchr = UCHARAT(++locinput);
2575 PL_reg_flags |= RF_tainted;
2578 if (!nextchr && locinput >= PL_regeol)
2581 LOAD_UTF8_CHARCLASS(space," ");
2582 if (OP(scan) == NSPACE
2583 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2584 : isSPACE_LC_utf8((U8*)locinput))
2588 locinput += PL_utf8skip[nextchr];
2589 nextchr = UCHARAT(locinput);
2592 if (OP(scan) == NSPACE
2593 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2595 nextchr = UCHARAT(++locinput);
2598 PL_reg_flags |= RF_tainted;
2604 LOAD_UTF8_CHARCLASS(digit,"0");
2605 if (!(OP(scan) == DIGIT
2606 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2607 : isDIGIT_LC_utf8((U8*)locinput)))
2611 locinput += PL_utf8skip[nextchr];
2612 nextchr = UCHARAT(locinput);
2615 if (!(OP(scan) == DIGIT
2616 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2618 nextchr = UCHARAT(++locinput);
2621 PL_reg_flags |= RF_tainted;
2624 if (!nextchr && locinput >= PL_regeol)
2627 LOAD_UTF8_CHARCLASS(digit,"0");
2628 if (OP(scan) == NDIGIT
2629 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2630 : isDIGIT_LC_utf8((U8*)locinput))
2634 locinput += PL_utf8skip[nextchr];
2635 nextchr = UCHARAT(locinput);
2638 if (OP(scan) == NDIGIT
2639 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2641 nextchr = UCHARAT(++locinput);
2644 if (locinput >= PL_regeol)
2647 LOAD_UTF8_CHARCLASS(mark,"~");
2648 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2650 locinput += PL_utf8skip[nextchr];
2651 while (locinput < PL_regeol &&
2652 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2653 locinput += UTF8SKIP(locinput);
2654 if (locinput > PL_regeol)
2659 nextchr = UCHARAT(locinput);
2662 PL_reg_flags |= RF_tainted;
2666 n = ARG(scan); /* which paren pair */
2667 ln = PL_regstartp[n];
2668 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2669 if (*PL_reglastparen < n || ln == -1)
2670 sayNO; /* Do not match unless seen CLOSEn. */
2671 if (ln == PL_regendp[n])
2675 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2677 char *e = PL_bostr + PL_regendp[n];
2679 * Note that we can't do the "other character" lookup trick as
2680 * in the 8-bit case (no pun intended) because in Unicode we
2681 * have to map both upper and title case to lower case.
2683 if (OP(scan) == REFF) {
2684 STRLEN ulen1, ulen2;
2685 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2686 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2690 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2691 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2692 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2699 nextchr = UCHARAT(locinput);
2703 /* Inline the first character, for speed. */
2704 if (UCHARAT(s) != nextchr &&
2706 (UCHARAT(s) != ((OP(scan) == REFF
2707 ? PL_fold : PL_fold_locale)[nextchr]))))
2709 ln = PL_regendp[n] - ln;
2710 if (locinput + ln > PL_regeol)
2712 if (ln > 1 && (OP(scan) == REF
2713 ? memNE(s, locinput, ln)
2715 ? ibcmp(s, locinput, ln)
2716 : ibcmp_locale(s, locinput, ln))))
2719 nextchr = UCHARAT(locinput);
2730 OP_4tree *oop = PL_op;
2731 COP *ocurcop = PL_curcop;
2732 SV **ocurpad = PL_curpad;
2736 PL_op = (OP_4tree*)PL_regdata->data[n];
2737 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2738 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2739 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2743 CALLRUNOPS(aTHX); /* Scalar context. */
2746 ret = Nullsv; /* protect against empty (?{}) blocks. */
2754 PL_curpad = ocurpad;
2755 PL_curcop = ocurcop;
2757 if (logical == 2) { /* Postponed subexpression. */
2759 MAGIC *mg = Null(MAGIC*);
2761 CHECKPOINT cp, lastcp;
2763 if(SvROK(ret) || SvRMAGICAL(ret)) {
2764 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2767 mg = mg_find(sv, PERL_MAGIC_qr);
2770 re = (regexp *)mg->mg_obj;
2771 (void)ReREFCNT_inc(re);
2775 char *t = SvPV(ret, len);
2777 char *oprecomp = PL_regprecomp;
2778 I32 osize = PL_regsize;
2779 I32 onpar = PL_regnpar;
2782 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2784 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2785 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2787 PL_regprecomp = oprecomp;
2792 PerlIO_printf(Perl_debug_log,
2793 "Entering embedded `%s%.60s%s%s'\n",
2797 (strlen(re->precomp) > 60 ? "..." : ""))
2800 state.prev = PL_reg_call_cc;
2801 state.cc = PL_regcc;
2802 state.re = PL_reg_re;
2806 cp = regcppush(0); /* Save *all* the positions. */
2809 state.ss = PL_savestack_ix;
2810 *PL_reglastparen = 0;
2811 *PL_reglastcloseparen = 0;
2812 PL_reg_call_cc = &state;
2813 PL_reginput = locinput;
2815 /* XXXX This is too dramatic a measure... */
2818 if (regmatch(re->program + 1)) {
2819 /* Even though we succeeded, we need to restore
2820 global variables, since we may be wrapped inside
2821 SUSPEND, thus the match may be not finished yet. */
2823 /* XXXX Do this only if SUSPENDed? */
2824 PL_reg_call_cc = state.prev;
2825 PL_regcc = state.cc;
2826 PL_reg_re = state.re;
2827 cache_re(PL_reg_re);
2829 /* XXXX This is too dramatic a measure... */
2832 /* These are needed even if not SUSPEND. */
2838 REGCP_UNWIND(lastcp);
2840 PL_reg_call_cc = state.prev;
2841 PL_regcc = state.cc;
2842 PL_reg_re = state.re;
2843 cache_re(PL_reg_re);
2845 /* XXXX This is too dramatic a measure... */
2855 sv_setsv(save_scalar(PL_replgv), ret);
2859 n = ARG(scan); /* which paren pair */
2860 PL_reg_start_tmp[n] = locinput;
2865 n = ARG(scan); /* which paren pair */
2866 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2867 PL_regendp[n] = locinput - PL_bostr;
2868 if (n > *PL_reglastparen)
2869 *PL_reglastparen = n;
2870 *PL_reglastcloseparen = n;
2873 n = ARG(scan); /* which paren pair */
2874 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2877 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2879 next = NEXTOPER(NEXTOPER(scan));
2881 next = scan + ARG(scan);
2882 if (OP(next) == IFTHEN) /* Fake one. */
2883 next = NEXTOPER(NEXTOPER(next));
2887 logical = scan->flags;
2889 /*******************************************************************
2890 PL_regcc contains infoblock about the innermost (...)* loop, and
2891 a pointer to the next outer infoblock.
2893 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2895 1) After matching X, regnode for CURLYX is processed;
2897 2) This regnode creates infoblock on the stack, and calls
2898 regmatch() recursively with the starting point at WHILEM node;
2900 3) Each hit of WHILEM node tries to match A and Z (in the order
2901 depending on the current iteration, min/max of {min,max} and
2902 greediness). The information about where are nodes for "A"
2903 and "Z" is read from the infoblock, as is info on how many times "A"
2904 was already matched, and greediness.
2906 4) After A matches, the same WHILEM node is hit again.
2908 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2909 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2910 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2911 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2912 of the external loop.
2914 Currently present infoblocks form a tree with a stem formed by PL_curcc
2915 and whatever it mentions via ->next, and additional attached trees
2916 corresponding to temporarily unset infoblocks as in "5" above.
2918 In the following picture infoblocks for outer loop of
2919 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2920 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2921 infoblocks are drawn below the "reset" infoblock.
2923 In fact in the picture below we do not show failed matches for Z and T
2924 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2925 more obvious *why* one needs to *temporary* unset infoblocks.]
2927 Matched REx position InfoBlocks Comment
2931 Y A)*?Z)*?T x <- O <- I
2932 YA )*?Z)*?T x <- O <- I
2933 YA A)*?Z)*?T x <- O <- I
2934 YAA )*?Z)*?T x <- O <- I
2935 YAA Z)*?T x <- O # Temporary unset I
2938 YAAZ Y(A)*?Z)*?T x <- O
2941 YAAZY (A)*?Z)*?T x <- O
2944 YAAZY A)*?Z)*?T x <- O <- I
2947 YAAZYA )*?Z)*?T x <- O <- I
2950 YAAZYA Z)*?T x <- O # Temporary unset I
2956 YAAZYAZ T x # Temporary unset O
2963 *******************************************************************/
2966 CHECKPOINT cp = PL_savestack_ix;
2967 /* No need to save/restore up to this paren */
2968 I32 parenfloor = scan->flags;
2970 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2972 cc.oldcc = PL_regcc;
2974 /* XXXX Probably it is better to teach regpush to support
2975 parenfloor > PL_regsize... */
2976 if (parenfloor > *PL_reglastparen)
2977 parenfloor = *PL_reglastparen; /* Pessimization... */
2978 cc.parenfloor = parenfloor;
2980 cc.min = ARG1(scan);
2981 cc.max = ARG2(scan);
2982 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2986 PL_reginput = locinput;
2987 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2989 PL_regcc = cc.oldcc;
2995 * This is really hard to understand, because after we match
2996 * what we're trying to match, we must make sure the rest of
2997 * the REx is going to match for sure, and to do that we have
2998 * to go back UP the parse tree by recursing ever deeper. And
2999 * if it fails, we have to reset our parent's current state
3000 * that we can try again after backing off.
3003 CHECKPOINT cp, lastcp;
3004 CURCUR* cc = PL_regcc;
3005 char *lastloc = cc->lastloc; /* Detection of 0-len. */
3007 n = cc->cur + 1; /* how many we know we matched */
3008 PL_reginput = locinput;
3011 PerlIO_printf(Perl_debug_log,
3012 "%*s %ld out of %ld..%ld cc=%lx\n",
3013 REPORT_CODE_OFF+PL_regindent*2, "",
3014 (long)n, (long)cc->min,
3015 (long)cc->max, (long)cc)
3018 /* If degenerate scan matches "", assume scan done. */
3020 if (locinput == cc->lastloc && n >= cc->min) {
3021 PL_regcc = cc->oldcc;
3025 PerlIO_printf(Perl_debug_log,
3026 "%*s empty match detected, try continuation...\n",
3027 REPORT_CODE_OFF+PL_regindent*2, "")
3029 if (regmatch(cc->next))
3037 /* First just match a string of min scans. */
3041 cc->lastloc = locinput;
3042 if (regmatch(cc->scan))
3045 cc->lastloc = lastloc;
3050 /* Check whether we already were at this position.
3051 Postpone detection until we know the match is not
3052 *that* much linear. */
3053 if (!PL_reg_maxiter) {
3054 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3055 PL_reg_leftiter = PL_reg_maxiter;
3057 if (PL_reg_leftiter-- == 0) {
3058 I32 size = (PL_reg_maxiter + 7)/8;
3059 if (PL_reg_poscache) {
3060 if (PL_reg_poscache_size < size) {
3061 Renew(PL_reg_poscache, size, char);
3062 PL_reg_poscache_size = size;
3064 Zero(PL_reg_poscache, size, char);
3067 PL_reg_poscache_size = size;
3068 Newz(29, PL_reg_poscache, size, char);
3071 PerlIO_printf(Perl_debug_log,
3072 "%sDetected a super-linear match, switching on caching%s...\n",
3073 PL_colors[4], PL_colors[5])
3076 if (PL_reg_leftiter < 0) {
3077 I32 o = locinput - PL_bostr, b;
3079 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3082 if (PL_reg_poscache[o] & (1<<b)) {
3084 PerlIO_printf(Perl_debug_log,
3085 "%*s already tried at this position...\n",
3086 REPORT_CODE_OFF+PL_regindent*2, "")
3090 PL_reg_poscache[o] |= (1<<b);
3094 /* Prefer next over scan for minimal matching. */
3097 PL_regcc = cc->oldcc;
3100 cp = regcppush(cc->parenfloor);
3102 if (regmatch(cc->next)) {
3104 sayYES; /* All done. */
3106 REGCP_UNWIND(lastcp);
3112 if (n >= cc->max) { /* Maximum greed exceeded? */
3113 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3114 && !(PL_reg_flags & RF_warned)) {
3115 PL_reg_flags |= RF_warned;
3116 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3117 "Complex regular subexpression recursion",
3124 PerlIO_printf(Perl_debug_log,
3125 "%*s trying longer...\n",
3126 REPORT_CODE_OFF+PL_regindent*2, "")
3128 /* Try scanning more and see if it helps. */
3129 PL_reginput = locinput;
3131 cc->lastloc = locinput;
3132 cp = regcppush(cc->parenfloor);
3134 if (regmatch(cc->scan)) {
3138 REGCP_UNWIND(lastcp);
3141 cc->lastloc = lastloc;
3145 /* Prefer scan over next for maximal matching. */
3147 if (n < cc->max) { /* More greed allowed? */
3148 cp = regcppush(cc->parenfloor);
3150 cc->lastloc = locinput;
3152 if (regmatch(cc->scan)) {
3156 REGCP_UNWIND(lastcp);
3157 regcppop(); /* Restore some previous $<digit>s? */
3158 PL_reginput = locinput;
3160 PerlIO_printf(Perl_debug_log,
3161 "%*s failed, try continuation...\n",
3162 REPORT_CODE_OFF+PL_regindent*2, "")
3165 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3166 && !(PL_reg_flags & RF_warned)) {
3167 PL_reg_flags |= RF_warned;
3168 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3169 "Complex regular subexpression recursion",
3173 /* Failed deeper matches of scan, so see if this one works. */
3174 PL_regcc = cc->oldcc;
3177 if (regmatch(cc->next))
3183 cc->lastloc = lastloc;
3188 next = scan + ARG(scan);
3191 inner = NEXTOPER(NEXTOPER(scan));
3194 inner = NEXTOPER(scan);
3198 if (OP(next) != c1) /* No choice. */
3199 next = inner; /* Avoid recursion. */
3201 I32 lastparen = *PL_reglastparen;
3203 re_unwind_branch_t *uw;
3205 /* Put unwinding data on stack */
3206 unwind1 = SSNEWt(1,re_unwind_branch_t);
3207 uw = SSPTRt(unwind1,re_unwind_branch_t);
3210 uw->type = ((c1 == BRANCH)
3212 : RE_UNWIND_BRANCHJ);
3213 uw->lastparen = lastparen;
3215 uw->locinput = locinput;
3216 uw->nextchr = nextchr;
3218 uw->regindent = ++PL_regindent;
3221 REGCP_SET(uw->lastcp);
3223 /* Now go into the first branch */
3236 /* We suppose that the next guy does not need
3237 backtracking: in particular, it is of constant length,
3238 and has no parenths to influence future backrefs. */
3239 ln = ARG1(scan); /* min to match */
3240 n = ARG2(scan); /* max to match */
3241 paren = scan->flags;
3243 if (paren > PL_regsize)
3245 if (paren > *PL_reglastparen)
3246 *PL_reglastparen = paren;
3248 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3250 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3251 PL_reginput = locinput;
3254 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3256 /* if we matched something zero-length we don't need to
3257 backtrack - capturing parens are already defined, so
3258 the caveat in the maximal case doesn't apply
3260 XXXX if ln == 0, we can redo this check first time
3261 through the following loop
3264 n = ln; /* don't backtrack */
3265 locinput = PL_reginput;
3266 if (HAS_TEXT(next) || JUMPABLE(next)) {
3267 regnode *text_node = next;
3269 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3271 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3273 if (PL_regkind[(U8)OP(text_node)] == REF) {
3275 n = ARG(text_node); /* which paren pair */
3276 ln = PL_regstartp[n];
3277 /* assume yes if we haven't seen CLOSEn */
3279 *PL_reglastparen < n ||
3286 c1 = *(PL_bostr + ln);
3288 else { c1 = (U8)*STRING(text_node); }
3289 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3291 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3292 c2 = PL_fold_locale[c1];
3301 /* This may be improved if l == 0. */
3302 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3303 /* If it could work, try it. */
3305 UCHARAT(PL_reginput) == c1 ||
3306 UCHARAT(PL_reginput) == c2)
3310 PL_regstartp[paren] =
3311 HOPc(PL_reginput, -l) - PL_bostr;
3312 PL_regendp[paren] = PL_reginput - PL_bostr;
3315 PL_regendp[paren] = -1;
3319 REGCP_UNWIND(lastcp);
3321 /* Couldn't or didn't -- move forward. */
3322 PL_reginput = locinput;
3323 if (regrepeat_hard(scan, 1, &l)) {
3325 locinput = PL_reginput;
3332 n = regrepeat_hard(scan, n, &l);
3333 /* if we matched something zero-length we don't need to
3334 backtrack, unless the minimum count is zero and we
3335 are capturing the result - in that case the capture
3336 being defined or not may affect later execution
3338 if (n != 0 && l == 0 && !(paren && ln == 0))
3339 ln = n; /* don't backtrack */
3340 locinput = PL_reginput;
3342 PerlIO_printf(Perl_debug_log,
3343 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3344 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3348 if (HAS_TEXT(next) || JUMPABLE(next)) {
3349 regnode *text_node = next;
3351 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3353 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3355 if (PL_regkind[(U8)OP(text_node)] == REF) {
3357 n = ARG(text_node); /* which paren pair */
3358 ln = PL_regstartp[n];
3359 /* assume yes if we haven't seen CLOSEn */
3361 *PL_reglastparen < n ||
3368 c1 = *(PL_bostr + ln);
3370 else { c1 = (U8)*STRING(text_node); }
3372 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3374 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3375 c2 = PL_fold_locale[c1];
3386 /* If it could work, try it. */
3388 UCHARAT(PL_reginput) == c1 ||
3389 UCHARAT(PL_reginput) == c2)
3392 PerlIO_printf(Perl_debug_log,
3393 "%*s trying tail with n=%"IVdf"...\n",
3394 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3398 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3399 PL_regendp[paren] = PL_reginput - PL_bostr;
3402 PL_regendp[paren] = -1;
3406 REGCP_UNWIND(lastcp);
3408 /* Couldn't or didn't -- back up. */
3410 locinput = HOPc(locinput, -l);
3411 PL_reginput = locinput;
3418 paren = scan->flags; /* Which paren to set */
3419 if (paren > PL_regsize)
3421 if (paren > *PL_reglastparen)
3422 *PL_reglastparen = paren;
3423 ln = ARG1(scan); /* min to match */
3424 n = ARG2(scan); /* max to match */
3425 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3429 ln = ARG1(scan); /* min to match */
3430 n = ARG2(scan); /* max to match */
3431 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3436 scan = NEXTOPER(scan);
3442 scan = NEXTOPER(scan);
3446 * Lookahead to avoid useless match attempts
3447 * when we know what character comes next.
3451 * Used to only do .*x and .*?x, but now it allows
3452 * for )'s, ('s and (?{ ... })'s to be in the way
3453 * of the quantifier and the EXACT-like node. -- japhy
3456 if (HAS_TEXT(next) || JUMPABLE(next)) {
3458 regnode *text_node = next;
3460 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3462 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3464 if (PL_regkind[(U8)OP(text_node)] == REF) {
3466 n = ARG(text_node); /* which paren pair */
3467 ln = PL_regstartp[n];
3468 /* assume yes if we haven't seen CLOSEn */
3470 *PL_reglastparen < n ||
3475 goto assume_ok_easy;
3477 s = (U8*)PL_bostr + ln;
3479 else { s = (U8*)STRING(text_node); }
3483 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3485 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3486 c2 = PL_fold_locale[c1];
3489 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3490 STRLEN ulen1, ulen2;
3491 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3492 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3494 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3495 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3497 c1 = utf8_to_uvuni(tmpbuf1, 0);
3498 c2 = utf8_to_uvuni(tmpbuf2, 0);
3501 c2 = c1 = utf8_to_uvchr(s, NULL);
3509 PL_reginput = locinput;
3513 if (ln && regrepeat(scan, ln) < ln)
3515 locinput = PL_reginput;
3518 char *e; /* Should not check after this */
3519 char *old = locinput;
3521 if (n == REG_INFTY) {
3524 while (UTF8_IS_CONTINUATION(*(U8*)e))
3530 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3534 e = locinput + n - ln;
3540 /* Find place 'next' could work */
3543 while (locinput <= e &&
3544 UCHARAT(locinput) != c1)
3547 while (locinput <= e
3548 && UCHARAT(locinput) != c1
3549 && UCHARAT(locinput) != c2)
3552 count = locinput - old;
3559 utf8_to_uvchr((U8*)locinput, &len) != c1;
3564 for (count = 0; locinput <= e; count++) {
3565 UV c = utf8_to_uvchr((U8*)locinput, &len);
3566 if (c == c1 || c == c2)
3574 /* PL_reginput == old now */
3575 if (locinput != old) {
3576 ln = 1; /* Did some */
3577 if (regrepeat(scan, count) < count)
3580 /* PL_reginput == locinput now */
3581 TRYPAREN(paren, ln, locinput);
3582 PL_reginput = locinput; /* Could be reset... */
3583 REGCP_UNWIND(lastcp);
3584 /* Couldn't or didn't -- move forward. */
3587 locinput += UTF8SKIP(locinput);
3593 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3597 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3599 c = UCHARAT(PL_reginput);
3600 /* If it could work, try it. */
3601 if (c == c1 || c == c2)
3603 TRYPAREN(paren, n, PL_reginput);
3604 REGCP_UNWIND(lastcp);
3607 /* If it could work, try it. */
3608 else if (c1 == -1000)
3610 TRYPAREN(paren, n, PL_reginput);
3611 REGCP_UNWIND(lastcp);
3613 /* Couldn't or didn't -- move forward. */
3614 PL_reginput = locinput;
3615 if (regrepeat(scan, 1)) {
3617 locinput = PL_reginput;
3625 n = regrepeat(scan, n);
3626 locinput = PL_reginput;
3627 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3628 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3629 ln = n; /* why back off? */
3630 /* ...because $ and \Z can match before *and* after
3631 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3632 We should back off by one in this case. */
3633 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3642 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3644 c = UCHARAT(PL_reginput);
3646 /* If it could work, try it. */
3647 if (c1 == -1000 || c == c1 || c == c2)
3649 TRYPAREN(paren, n, PL_reginput);
3650 REGCP_UNWIND(lastcp);
3652 /* Couldn't or didn't -- back up. */
3654 PL_reginput = locinput = HOPc(locinput, -1);
3662 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3664 c = UCHARAT(PL_reginput);
3666 /* If it could work, try it. */
3667 if (c1 == -1000 || c == c1 || c == c2)
3669 TRYPAREN(paren, n, PL_reginput);
3670 REGCP_UNWIND(lastcp);
3672 /* Couldn't or didn't -- back up. */
3674 PL_reginput = locinput = HOPc(locinput, -1);
3681 if (PL_reg_call_cc) {
3682 re_cc_state *cur_call_cc = PL_reg_call_cc;
3683 CURCUR *cctmp = PL_regcc;
3684 regexp *re = PL_reg_re;
3685 CHECKPOINT cp, lastcp;
3687 cp = regcppush(0); /* Save *all* the positions. */
3689 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3691 PL_reginput = locinput; /* Make position available to
3693 cache_re(PL_reg_call_cc->re);
3694 PL_regcc = PL_reg_call_cc->cc;
3695 PL_reg_call_cc = PL_reg_call_cc->prev;
3696 if (regmatch(cur_call_cc->node)) {
3697 PL_reg_call_cc = cur_call_cc;
3701 REGCP_UNWIND(lastcp);
3703 PL_reg_call_cc = cur_call_cc;
3709 PerlIO_printf(Perl_debug_log,
3710 "%*s continuation failed...\n",
3711 REPORT_CODE_OFF+PL_regindent*2, "")
3715 if (locinput < PL_regtill) {
3716 DEBUG_r(PerlIO_printf(Perl_debug_log,
3717 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3719 (long)(locinput - PL_reg_starttry),
3720 (long)(PL_regtill - PL_reg_starttry),
3722 sayNO_FINAL; /* Cannot match: too short. */
3724 PL_reginput = locinput; /* put where regtry can find it */
3725 sayYES_FINAL; /* Success! */
3727 PL_reginput = locinput; /* put where regtry can find it */
3728 sayYES_LOUD; /* Success! */
3731 PL_reginput = locinput;
3736 s = HOPBACKc(locinput, scan->flags);
3742 PL_reginput = locinput;
3747 s = HOPBACKc(locinput, scan->flags);
3753 PL_reginput = locinput;
3756 inner = NEXTOPER(NEXTOPER(scan));
3757 if (regmatch(inner) != n) {
3772 if (OP(scan) == SUSPEND) {
3773 locinput = PL_reginput;
3774 nextchr = UCHARAT(locinput);
3779 next = scan + ARG(scan);
3784 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3785 PTR2UV(scan), OP(scan));
3786 Perl_croak(aTHX_ "regexp memory corruption");
3793 * We get here only if there's trouble -- normally "case END" is
3794 * the terminating point.
3796 Perl_croak(aTHX_ "corrupted regexp pointers");
3802 PerlIO_printf(Perl_debug_log,
3803 "%*s %scould match...%s\n",
3804 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3808 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3809 PL_colors[4],PL_colors[5]));
3815 #if 0 /* Breaks $^R */
3823 PerlIO_printf(Perl_debug_log,
3824 "%*s %sfailed...%s\n",
3825 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3831 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3834 case RE_UNWIND_BRANCH:
3835 case RE_UNWIND_BRANCHJ:
3837 re_unwind_branch_t *uwb = &(uw->branch);
3838 I32 lastparen = uwb->lastparen;
3840 REGCP_UNWIND(uwb->lastcp);
3841 for (n = *PL_reglastparen; n > lastparen; n--)
3843 *PL_reglastparen = n;
3844 scan = next = uwb->next;
3846 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3847 ? BRANCH : BRANCHJ) ) { /* Failure */
3854 /* Have more choice yet. Reuse the same uwb. */
3856 if ((n = (uwb->type == RE_UNWIND_BRANCH
3857 ? NEXT_OFF(next) : ARG(next))))
3860 next = NULL; /* XXXX Needn't unwinding in this case... */
3862 next = NEXTOPER(scan);
3863 if (uwb->type == RE_UNWIND_BRANCHJ)
3864 next = NEXTOPER(next);
3865 locinput = uwb->locinput;
3866 nextchr = uwb->nextchr;
3868 PL_regindent = uwb->regindent;
3875 Perl_croak(aTHX_ "regexp unwind memory corruption");
3886 - regrepeat - repeatedly match something simple, report how many
3889 * [This routine now assumes that it will only match on things of length 1.
3890 * That was true before, but now we assume scan - reginput is the count,
3891 * rather than incrementing count on every character. [Er, except utf8.]]
3894 S_regrepeat(pTHX_ regnode *p, I32 max)
3896 register char *scan;
3898 register char *loceol = PL_regeol;
3899 register I32 hardcount = 0;
3900 register bool do_utf8 = PL_reg_match_utf8;
3903 if (max != REG_INFTY && max < loceol - scan)
3904 loceol = scan + max;
3909 while (scan < loceol && hardcount < max && *scan != '\n') {
3910 scan += UTF8SKIP(scan);
3914 while (scan < loceol && *scan != '\n')
3924 case EXACT: /* length of string is 1 */
3926 while (scan < loceol && UCHARAT(scan) == c)
3929 case EXACTF: /* length of string is 1 */
3931 while (scan < loceol &&
3932 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3935 case EXACTFL: /* length of string is 1 */
3936 PL_reg_flags |= RF_tainted;
3938 while (scan < loceol &&
3939 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3945 while (hardcount < max && scan < loceol &&
3946 reginclass(p, (U8*)scan, do_utf8)) {
3947 scan += UTF8SKIP(scan);
3951 while (scan < loceol && reginclass(p, (U8*)scan, do_utf8))
3958 LOAD_UTF8_CHARCLASS(alnum,"a");
3959 while (hardcount < max && scan < loceol &&
3960 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3961 scan += UTF8SKIP(scan);
3965 while (scan < loceol && isALNUM(*scan))
3970 PL_reg_flags |= RF_tainted;
3973 while (hardcount < max && scan < loceol &&
3974 isALNUM_LC_utf8((U8*)scan)) {
3975 scan += UTF8SKIP(scan);
3979 while (scan < loceol && isALNUM_LC(*scan))
3986 LOAD_UTF8_CHARCLASS(alnum,"a");
3987 while (hardcount < max && scan < loceol &&
3988 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3989 scan += UTF8SKIP(scan);
3993 while (scan < loceol && !isALNUM(*scan))
3998 PL_reg_flags |= RF_tainted;
4001 while (hardcount < max && scan < loceol &&
4002 !isALNUM_LC_utf8((U8*)scan)) {
4003 scan += UTF8SKIP(scan);
4007 while (scan < loceol && !isALNUM_LC(*scan))
4014 LOAD_UTF8_CHARCLASS(space," ");
4015 while (hardcount < max && scan < loceol &&
4017 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4018 scan += UTF8SKIP(scan);
4022 while (scan < loceol && isSPACE(*scan))
4027 PL_reg_flags |= RF_tainted;
4030 while (hardcount < max && scan < loceol &&
4031 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4032 scan += UTF8SKIP(scan);
4036 while (scan < loceol && isSPACE_LC(*scan))
4043 LOAD_UTF8_CHARCLASS(space," ");
4044 while (hardcount < max && scan < loceol &&
4046 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4047 scan += UTF8SKIP(scan);
4051 while (scan < loceol && !isSPACE(*scan))
4056 PL_reg_flags |= RF_tainted;
4059 while (hardcount < max && scan < loceol &&
4060 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4061 scan += UTF8SKIP(scan);
4065 while (scan < loceol && !isSPACE_LC(*scan))
4072 LOAD_UTF8_CHARCLASS(digit,"0");
4073 while (hardcount < max && scan < loceol &&
4074 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4075 scan += UTF8SKIP(scan);
4079 while (scan < loceol && isDIGIT(*scan))
4086 LOAD_UTF8_CHARCLASS(digit,"0");
4087 while (hardcount < max && scan < loceol &&
4088 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4089 scan += UTF8SKIP(scan);
4093 while (scan < loceol && !isDIGIT(*scan))
4097 default: /* Called on something of 0 width. */
4098 break; /* So match right here or not at all. */
4104 c = scan - PL_reginput;
4109 SV *prop = sv_newmortal();
4112 PerlIO_printf(Perl_debug_log,
4113 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4114 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4121 - regrepeat_hard - repeatedly match something, report total lenth and length
4123 * The repeater is supposed to have constant length.
4127 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4129 register char *scan = Nullch;
4130 register char *start;
4131 register char *loceol = PL_regeol;
4133 I32 count = 0, res = 1;
4138 start = PL_reginput;
4139 if (PL_reg_match_utf8) {
4140 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4143 while (start < PL_reginput) {
4145 start += UTF8SKIP(start);
4156 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4158 *lp = l = PL_reginput - start;
4159 if (max != REG_INFTY && l*max < loceol - scan)
4160 loceol = scan + l*max;
4173 - regclass_swash - prepare the utf8 swash
4177 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** initsvp)
4182 if (PL_regdata && PL_regdata->count) {
4185 if (PL_regdata->what[n] == 's') {
4186 SV *rv = (SV*)PL_regdata->data[n];
4187 AV *av = (AV*)SvRV((SV*)rv);
4190 si = *av_fetch(av, 0, FALSE);
4191 a = av_fetch(av, 1, FALSE);
4195 else if (si && doinit) {
4196 sw = swash_init("utf8", "", si, 1, 0);
4197 (void)av_store(av, 1, sw);
4209 - reginclass - determine if a character falls into a character class
4213 S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8)
4215 char flags = ANYOF_FLAGS(n);
4220 c = do_utf8 ? utf8_to_uvchr(p, &len) : *p;
4222 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4223 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4224 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4227 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4230 SV *sw = regclass_swash(n, TRUE, 0);
4233 if (swash_fetch(sw, p, do_utf8))
4235 else if (flags & ANYOF_FOLD) {
4237 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
4239 to_utf8_fold(p, tmpbuf, &ulen);
4240 if (swash_fetch(sw, tmpbuf, do_utf8))
4242 to_utf8_upper(p, tmpbuf, &ulen);
4243 if (swash_fetch(sw, tmpbuf, do_utf8))
4249 if (!match && c < 256) {
4250 if (ANYOF_BITMAP_TEST(n, c))
4252 else if (flags & ANYOF_FOLD) {
4255 if (flags & ANYOF_LOCALE) {
4256 PL_reg_flags |= RF_tainted;
4257 f = PL_fold_locale[c];
4261 if (f != c && ANYOF_BITMAP_TEST(n, f))
4265 if (!match && (flags & ANYOF_CLASS)) {
4266 PL_reg_flags |= RF_tainted;
4268 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4269 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4270 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4271 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4272 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4273 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4274 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4275 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4276 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4277 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4278 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4279 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4280 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4281 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4282 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4283 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4284 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4285 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4286 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4287 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4288 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4289 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4290 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4291 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4292 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4293 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4294 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4295 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4296 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4297 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4298 ) /* How's that for a conditional? */
4305 return (flags & ANYOF_INVERT) ? !match : match;
4309 S_reghop(pTHX_ U8 *s, I32 off)
4311 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4315 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4318 while (off-- && s < lim) {
4319 /* XXX could check well-formedness here */
4327 if (UTF8_IS_CONTINUED(*s)) {
4328 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4331 /* XXX could check well-formedness here */
4339 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4341 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4345 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4348 while (off-- && s < lim) {
4349 /* XXX could check well-formedness here */
4359 if (UTF8_IS_CONTINUED(*s)) {
4360 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4363 /* XXX could check well-formedness here */
4375 restore_pos(pTHX_ void *arg)
4377 if (PL_reg_eval_set) {
4378 if (PL_reg_oldsaved) {
4379 PL_reg_re->subbeg = PL_reg_oldsaved;
4380 PL_reg_re->sublen = PL_reg_oldsavedlen;
4381 RX_MATCH_COPIED_on(PL_reg_re);
4383 PL_reg_magic->mg_len = PL_reg_oldpos;
4384 PL_reg_eval_set = 0;
4385 PL_curpm = PL_reg_oldcurpm;