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];
963 e = do_utf8 ? s + ln : strend - ln;
966 e = s; /* Due to minlen logic of intuit() */
968 /* The idea in the EXACTF* cases is to first find the
969 * first character of the EXACTF* node and then, if
970 * necessary, case-insensitively compare the full
971 * text of the node. The c1 and c2 are the first
972 * characters (though in Unicode it gets a bit
973 * more complicated because there are more cases
974 * than just upper and lower: one is really supposed
975 * to use the so-called folding case for case-insensitive
976 * matching (called "loose matching" in Unicode). */
980 U8 tmpbuf [UTF8_MAXLEN+1];
981 U8 foldbuf[UTF8_MAXLEN_FOLD+1];
986 c = utf8_to_uvchr((U8*)s, &len);
989 ibcmp_utf8(s, (char **)0, 0, do_utf8,
990 m, (char **)0, ln, UTF))
991 && (norun || regtry(prog, s)) )
994 uvchr_to_utf8(tmpbuf, c);
995 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
997 && (f == c1 || f == c2)
999 !ibcmp_utf8((char *) foldbuf,
1000 (char **)0, foldlen, do_utf8,
1002 (char **)0, ln, UTF))
1003 && (norun || regtry(prog, s)) )
1011 c = utf8_to_uvchr((U8*)s, &len);
1013 /* Handle some of the three Greek sigmas cases.
1014 * Note that not all the possible combinations
1015 * are handled here: some of them are handled
1016 * handled by the standard folding rules, and
1017 * some of them (the character class or ANYOF
1018 * cases) are handled during compiletime in
1019 * regexec.c:S_regclass(). */
1020 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1021 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1022 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1024 if ( (c == c1 || c == c2)
1026 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1027 m, (char **)0, ln, UTF))
1028 && (norun || regtry(prog, s)) )
1031 uvchr_to_utf8(tmpbuf, c);
1032 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1034 && (f == c1 || f == c2)
1035 && (ln == foldlen ||
1036 !ibcmp_utf8((char *)foldbuf,
1037 (char **)0, foldlen, do_utf8,
1039 (char **)0, ln, UTF))
1040 && (norun || regtry(prog, s)) )
1051 && (ln == 1 || !(OP(c) == EXACTF
1053 : ibcmp_locale(s, m, ln)))
1054 && (norun || regtry(prog, s)) )
1060 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1061 && (ln == 1 || !(OP(c) == EXACTF
1063 : ibcmp_locale(s, m, ln)))
1064 && (norun || regtry(prog, s)) )
1071 PL_reg_flags |= RF_tainted;
1078 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1081 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1083 tmp = ((OP(c) == BOUND ?
1084 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1085 LOAD_UTF8_CHARCLASS(alnum,"a");
1086 while (s < strend) {
1087 if (tmp == !(OP(c) == BOUND ?
1088 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1089 isALNUM_LC_utf8((U8*)s)))
1092 if ((norun || regtry(prog, s)))
1099 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1100 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1101 while (s < strend) {
1103 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1105 if ((norun || regtry(prog, s)))
1111 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1115 PL_reg_flags |= RF_tainted;
1122 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1125 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1127 tmp = ((OP(c) == NBOUND ?
1128 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1129 LOAD_UTF8_CHARCLASS(alnum,"a");
1130 while (s < strend) {
1131 if (tmp == !(OP(c) == NBOUND ?
1132 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1133 isALNUM_LC_utf8((U8*)s)))
1135 else if ((norun || regtry(prog, s)))
1141 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1142 tmp = ((OP(c) == NBOUND ?
1143 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1144 while (s < strend) {
1146 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1148 else if ((norun || regtry(prog, s)))
1153 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1158 LOAD_UTF8_CHARCLASS(alnum,"a");
1159 while (s < strend) {
1160 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1161 if (tmp && (norun || regtry(prog, s)))
1172 while (s < strend) {
1174 if (tmp && (norun || regtry(prog, s)))
1186 PL_reg_flags |= RF_tainted;
1188 while (s < strend) {
1189 if (isALNUM_LC_utf8((U8*)s)) {
1190 if (tmp && (norun || regtry(prog, s)))
1201 while (s < strend) {
1202 if (isALNUM_LC(*s)) {
1203 if (tmp && (norun || regtry(prog, s)))
1216 LOAD_UTF8_CHARCLASS(alnum,"a");
1217 while (s < strend) {
1218 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1219 if (tmp && (norun || regtry(prog, s)))
1230 while (s < strend) {
1232 if (tmp && (norun || regtry(prog, s)))
1244 PL_reg_flags |= RF_tainted;
1246 while (s < strend) {
1247 if (!isALNUM_LC_utf8((U8*)s)) {
1248 if (tmp && (norun || regtry(prog, s)))
1259 while (s < strend) {
1260 if (!isALNUM_LC(*s)) {
1261 if (tmp && (norun || regtry(prog, s)))
1274 LOAD_UTF8_CHARCLASS(space," ");
1275 while (s < strend) {
1276 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1277 if (tmp && (norun || regtry(prog, s)))
1288 while (s < strend) {
1290 if (tmp && (norun || regtry(prog, s)))
1302 PL_reg_flags |= RF_tainted;
1304 while (s < strend) {
1305 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1306 if (tmp && (norun || regtry(prog, s)))
1317 while (s < strend) {
1318 if (isSPACE_LC(*s)) {
1319 if (tmp && (norun || regtry(prog, s)))
1332 LOAD_UTF8_CHARCLASS(space," ");
1333 while (s < strend) {
1334 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1335 if (tmp && (norun || regtry(prog, s)))
1346 while (s < strend) {
1348 if (tmp && (norun || regtry(prog, s)))
1360 PL_reg_flags |= RF_tainted;
1362 while (s < strend) {
1363 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1364 if (tmp && (norun || regtry(prog, s)))
1375 while (s < strend) {
1376 if (!isSPACE_LC(*s)) {
1377 if (tmp && (norun || regtry(prog, s)))
1390 LOAD_UTF8_CHARCLASS(digit,"0");
1391 while (s < strend) {
1392 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1393 if (tmp && (norun || regtry(prog, s)))
1404 while (s < strend) {
1406 if (tmp && (norun || regtry(prog, s)))
1418 PL_reg_flags |= RF_tainted;
1420 while (s < strend) {
1421 if (isDIGIT_LC_utf8((U8*)s)) {
1422 if (tmp && (norun || regtry(prog, s)))
1433 while (s < strend) {
1434 if (isDIGIT_LC(*s)) {
1435 if (tmp && (norun || regtry(prog, s)))
1448 LOAD_UTF8_CHARCLASS(digit,"0");
1449 while (s < strend) {
1450 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1451 if (tmp && (norun || regtry(prog, s)))
1462 while (s < strend) {
1464 if (tmp && (norun || regtry(prog, s)))
1476 PL_reg_flags |= RF_tainted;
1478 while (s < strend) {
1479 if (!isDIGIT_LC_utf8((U8*)s)) {
1480 if (tmp && (norun || regtry(prog, s)))
1491 while (s < strend) {
1492 if (!isDIGIT_LC(*s)) {
1493 if (tmp && (norun || regtry(prog, s)))
1505 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1514 - regexec_flags - match a regexp against a string
1517 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1518 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1519 /* strend: pointer to null at end of string */
1520 /* strbeg: real beginning of string */
1521 /* minend: end of match must be >=minend after stringarg. */
1522 /* data: May be used for some additional optimizations. */
1523 /* nosave: For optimizations. */
1526 register regnode *c;
1527 register char *startpos = stringarg;
1528 I32 minlen; /* must match at least this many chars */
1529 I32 dontbother = 0; /* how many characters not to try at end */
1530 /* I32 start_shift = 0; */ /* Offset of the start to find
1531 constant substr. */ /* CC */
1532 I32 end_shift = 0; /* Same for the end. */ /* CC */
1533 I32 scream_pos = -1; /* Internal iterator of scream. */
1535 SV* oreplsv = GvSV(PL_replgv);
1536 bool do_utf8 = DO_UTF8(sv);
1538 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
1545 PL_regnarrate = DEBUG_r_TEST;
1548 /* Be paranoid... */
1549 if (prog == NULL || startpos == NULL) {
1550 Perl_croak(aTHX_ "NULL regexp parameter");
1554 minlen = prog->minlen;
1555 if (strend - startpos < minlen) {
1556 DEBUG_r(PerlIO_printf(Perl_debug_log,
1557 "String too short [regexec_flags]...\n"));
1561 /* Check validity of program. */
1562 if (UCHARAT(prog->program) != REG_MAGIC) {
1563 Perl_croak(aTHX_ "corrupted regexp program");
1567 PL_reg_eval_set = 0;
1570 if (prog->reganch & ROPT_UTF8)
1571 PL_reg_flags |= RF_utf8;
1573 /* Mark beginning of line for ^ and lookbehind. */
1574 PL_regbol = startpos;
1578 /* Mark end of line for $ (and such) */
1581 /* see how far we have to get to not match where we matched before */
1582 PL_regtill = startpos+minend;
1584 /* We start without call_cc context. */
1587 /* If there is a "must appear" string, look for it. */
1590 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1593 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1594 PL_reg_ganch = startpos;
1595 else if (sv && SvTYPE(sv) >= SVt_PVMG
1597 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1598 && mg->mg_len >= 0) {
1599 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1600 if (prog->reganch & ROPT_ANCH_GPOS) {
1601 if (s > PL_reg_ganch)
1606 else /* pos() not defined */
1607 PL_reg_ganch = strbeg;
1610 if (do_utf8 == (UTF!=0) &&
1611 !(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
1612 re_scream_pos_data d;
1614 d.scream_olds = &scream_olds;
1615 d.scream_pos = &scream_pos;
1616 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1618 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1619 goto phooey; /* not present */
1624 char *s = do_utf8 ? sv_uni_display(dsv, sv, 60, 0) : startpos;
1625 int len = do_utf8 ? strlen(s) : strend - startpos;
1628 PerlIO_printf(Perl_debug_log,
1629 "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
1630 PL_colors[4],PL_colors[5],PL_colors[0],
1633 (strlen(prog->precomp) > 60 ? "..." : ""),
1635 (int)(len > 60 ? 60 : len),
1637 (len > 60 ? "..." : "")
1641 /* Simplest case: anchored match need be tried only once. */
1642 /* [unless only anchor is BOL and multiline is set] */
1643 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1644 if (s == startpos && regtry(prog, startpos))
1646 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1647 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1652 dontbother = minlen - 1;
1653 end = HOP3c(strend, -dontbother, strbeg) - 1;
1654 /* for multiline we only have to try after newlines */
1655 if (prog->check_substr) {
1659 if (regtry(prog, s))
1664 if (prog->reganch & RE_USE_INTUIT) {
1665 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1676 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1677 if (regtry(prog, s))
1684 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1685 if (regtry(prog, PL_reg_ganch))
1690 /* Messy cases: unanchored match. */
1691 if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
1692 /* we have /x+whatever/ */
1693 /* it must be a one character string (XXXX Except UTF?) */
1694 char ch = SvPVX(prog->anchored_substr)[0];
1700 while (s < strend) {
1702 DEBUG_r( did_match = 1 );
1703 if (regtry(prog, s)) goto got_it;
1705 while (s < strend && *s == ch)
1712 while (s < strend) {
1714 DEBUG_r( did_match = 1 );
1715 if (regtry(prog, s)) goto got_it;
1717 while (s < strend && *s == ch)
1723 DEBUG_r(if (!did_match)
1724 PerlIO_printf(Perl_debug_log,
1725 "Did not find anchored character...\n")
1729 else if (do_utf8 == (UTF!=0) &&
1730 (prog->anchored_substr != Nullsv
1731 || (prog->float_substr != Nullsv
1732 && prog->float_max_offset < strend - s))) {
1733 SV *must = prog->anchored_substr
1734 ? prog->anchored_substr : prog->float_substr;
1736 prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset;
1738 prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
1739 char *last = HOP3c(strend, /* Cannot start after this */
1740 -(I32)(CHR_SVLEN(must)
1741 - (SvTAIL(must) != 0) + back_min), strbeg);
1742 char *last1; /* Last position checked before */
1748 last1 = HOPc(s, -1);
1750 last1 = s - 1; /* bogus */
1752 /* XXXX check_substr already used to find `s', can optimize if
1753 check_substr==must. */
1755 dontbother = end_shift;
1756 strend = HOPc(strend, -dontbother);
1757 while ( (s <= last) &&
1758 ((flags & REXEC_SCREAM)
1759 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1760 end_shift, &scream_pos, 0))
1761 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1762 (unsigned char*)strend, must,
1763 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1764 DEBUG_r( did_match = 1 );
1765 if (HOPc(s, -back_max) > last1) {
1766 last1 = HOPc(s, -back_min);
1767 s = HOPc(s, -back_max);
1770 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1772 last1 = HOPc(s, -back_min);
1776 while (s <= last1) {
1777 if (regtry(prog, s))
1783 while (s <= last1) {
1784 if (regtry(prog, s))
1790 DEBUG_r(if (!did_match)
1791 PerlIO_printf(Perl_debug_log,
1792 "Did not find %s substr `%s%.*s%s'%s...\n",
1793 ((must == prog->anchored_substr)
1794 ? "anchored" : "floating"),
1796 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1798 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1802 else if ((c = prog->regstclass)) {
1803 if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT)
1804 /* don't bother with what can't match */
1805 strend = HOPc(strend, -(minlen - 1));
1807 SV *prop = sv_newmortal();
1809 PerlIO_printf(Perl_debug_log, "Matching stclass `%s' against `%s'\n", SvPVX(prop), UTF ? sv_uni_display(dsv, sv, 60, 0) : s);
1811 if (find_byclass(prog, c, s, strend, startpos, 0))
1813 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1817 if (prog->float_substr != Nullsv) { /* Trim the end. */
1820 if (flags & REXEC_SCREAM) {
1821 last = screaminstr(sv, prog->float_substr, s - strbeg,
1822 end_shift, &scream_pos, 1); /* last one */
1824 last = scream_olds; /* Only one occurrence. */
1828 char *little = SvPV(prog->float_substr, len);
1830 if (SvTAIL(prog->float_substr)) {
1831 if (memEQ(strend - len + 1, little, len - 1))
1832 last = strend - len + 1;
1833 else if (!PL_multiline)
1834 last = memEQ(strend - len, little, len)
1835 ? strend - len : Nullch;
1841 last = rninstr(s, strend, little, little + len);
1843 last = strend; /* matching `$' */
1847 DEBUG_r(PerlIO_printf(Perl_debug_log,
1848 "%sCan't trim the tail, match fails (should not happen)%s\n",
1849 PL_colors[4],PL_colors[5]));
1850 goto phooey; /* Should not happen! */
1852 dontbother = strend - last + prog->float_min_offset;
1854 if (minlen && (dontbother < minlen))
1855 dontbother = minlen - 1;
1856 strend -= dontbother; /* this one's always in bytes! */
1857 /* We don't know much -- general case. */
1860 if (regtry(prog, s))
1869 if (regtry(prog, s))
1871 } while (s++ < strend);
1879 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1881 if (PL_reg_eval_set) {
1882 /* Preserve the current value of $^R */
1883 if (oreplsv != GvSV(PL_replgv))
1884 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1885 restored, the value remains
1887 restore_pos(aTHX_ 0);
1890 /* make sure $`, $&, $', and $digit will work later */
1891 if ( !(flags & REXEC_NOT_FIRST) ) {
1892 if (RX_MATCH_COPIED(prog)) {
1893 Safefree(prog->subbeg);
1894 RX_MATCH_COPIED_off(prog);
1896 if (flags & REXEC_COPY_STR) {
1897 I32 i = PL_regeol - startpos + (stringarg - strbeg);
1899 s = savepvn(strbeg, i);
1902 RX_MATCH_COPIED_on(prog);
1905 prog->subbeg = strbeg;
1906 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
1913 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
1914 PL_colors[4],PL_colors[5]));
1915 if (PL_reg_eval_set)
1916 restore_pos(aTHX_ 0);
1921 - regtry - try match at specific point
1923 STATIC I32 /* 0 failure, 1 success */
1924 S_regtry(pTHX_ regexp *prog, char *startpos)
1932 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
1934 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
1937 PL_reg_eval_set = RS_init;
1939 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
1940 (IV)(PL_stack_sp - PL_stack_base));
1942 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
1943 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
1944 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
1946 /* Apparently this is not needed, judging by wantarray. */
1947 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
1948 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
1951 /* Make $_ available to executed code. */
1952 if (PL_reg_sv != DEFSV) {
1953 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
1958 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
1959 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
1960 /* prepare for quick setting of pos */
1961 sv_magic(PL_reg_sv, (SV*)0,
1962 PERL_MAGIC_regex_global, Nullch, 0);
1963 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
1967 PL_reg_oldpos = mg->mg_len;
1968 SAVEDESTRUCTOR_X(restore_pos, 0);
1970 if (!PL_reg_curpm) {
1971 Newz(22,PL_reg_curpm, 1, PMOP);
1974 SV* repointer = newSViv(0);
1975 /* so we know which PL_regex_padav element is PL_reg_curpm */
1976 SvFLAGS(repointer) |= SVf_BREAK;
1977 av_push(PL_regex_padav,repointer);
1978 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
1979 PL_regex_pad = AvARRAY(PL_regex_padav);
1983 PM_SETRE(PL_reg_curpm, prog);
1984 PL_reg_oldcurpm = PL_curpm;
1985 PL_curpm = PL_reg_curpm;
1986 if (RX_MATCH_COPIED(prog)) {
1987 /* Here is a serious problem: we cannot rewrite subbeg,
1988 since it may be needed if this match fails. Thus
1989 $` inside (?{}) could fail... */
1990 PL_reg_oldsaved = prog->subbeg;
1991 PL_reg_oldsavedlen = prog->sublen;
1992 RX_MATCH_COPIED_off(prog);
1995 PL_reg_oldsaved = Nullch;
1996 prog->subbeg = PL_bostr;
1997 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
1999 prog->startp[0] = startpos - PL_bostr;
2000 PL_reginput = startpos;
2001 PL_regstartp = prog->startp;
2002 PL_regendp = prog->endp;
2003 PL_reglastparen = &prog->lastparen;
2004 PL_reglastcloseparen = &prog->lastcloseparen;
2005 prog->lastparen = 0;
2007 DEBUG_r(PL_reg_starttry = startpos);
2008 if (PL_reg_start_tmpl <= prog->nparens) {
2009 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2010 if(PL_reg_start_tmp)
2011 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2013 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2017 sv_setpvn(PERL_DEBUG_PAD(0), "", 0);
2018 sv_setpvn(PERL_DEBUG_PAD(1), "", 0);
2019 sv_setpvn(PERL_DEBUG_PAD(2), "", 0);
2022 /* XXXX What this code is doing here?!!! There should be no need
2023 to do this again and again, PL_reglastparen should take care of
2026 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2027 * Actually, the code in regcppop() (which Ilya may be meaning by
2028 * PL_reglastparen), is not needed at all by the test suite
2029 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2030 * enough, for building DynaLoader, or otherwise this
2031 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2032 * will happen. Meanwhile, this code *is* needed for the
2033 * above-mentioned test suite tests to succeed. The common theme
2034 * on those tests seems to be returning null fields from matches.
2039 if (prog->nparens) {
2040 for (i = prog->nparens; i > *PL_reglastparen; i--) {
2047 if (regmatch(prog->program + 1)) {
2048 prog->endp[0] = PL_reginput - PL_bostr;
2051 REGCP_UNWIND(lastcp);
2055 #define RE_UNWIND_BRANCH 1
2056 #define RE_UNWIND_BRANCHJ 2
2060 typedef struct { /* XX: makes sense to enlarge it... */
2064 } re_unwind_generic_t;
2077 } re_unwind_branch_t;
2079 typedef union re_unwind_t {
2081 re_unwind_generic_t generic;
2082 re_unwind_branch_t branch;
2085 #define sayYES goto yes
2086 #define sayNO goto no
2087 #define sayYES_FINAL goto yes_final
2088 #define sayYES_LOUD goto yes_loud
2089 #define sayNO_FINAL goto no_final
2090 #define sayNO_SILENT goto do_no
2091 #define saySAME(x) if (x) goto yes; else goto no
2093 #define REPORT_CODE_OFF 24
2096 - regmatch - main matching routine
2098 * Conceptually the strategy is simple: check to see whether the current
2099 * node matches, call self recursively to see whether the rest matches,
2100 * and then act accordingly. In practice we make some effort to avoid
2101 * recursion, in particular by going through "ordinary" nodes (that don't
2102 * need to know whether the rest of the match failed) by a loop instead of
2105 /* [lwall] I've hoisted the register declarations to the outer block in order to
2106 * maybe save a little bit of pushing and popping on the stack. It also takes
2107 * advantage of machines that use a register save mask on subroutine entry.
2109 STATIC I32 /* 0 failure, 1 success */
2110 S_regmatch(pTHX_ regnode *prog)
2112 register regnode *scan; /* Current node. */
2113 regnode *next; /* Next node. */
2114 regnode *inner; /* Next node in internal branch. */
2115 register I32 nextchr; /* renamed nextchr - nextchar colides with
2116 function of same name */
2117 register I32 n; /* no or next */
2118 register I32 ln = 0; /* len or last */
2119 register char *s = Nullch; /* operand or save */
2120 register char *locinput = PL_reginput;
2121 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2122 int minmod = 0, sw = 0, logical = 0;
2125 I32 firstcp = PL_savestack_ix;
2127 register bool do_utf8 = PL_reg_match_utf8;
2129 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2130 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2131 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2138 /* Note that nextchr is a byte even in UTF */
2139 nextchr = UCHARAT(locinput);
2141 while (scan != NULL) {
2144 SV *prop = sv_newmortal();
2145 int docolor = *PL_colors[0];
2146 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2147 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2148 /* The part of the string before starttry has one color
2149 (pref0_len chars), between starttry and current
2150 position another one (pref_len - pref0_len chars),
2151 after the current position the third one.
2152 We assume that pref0_len <= pref_len, otherwise we
2153 decrease pref0_len. */
2154 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2155 ? (5 + taill) - l : locinput - PL_bostr;
2158 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2160 pref0_len = pref_len - (locinput - PL_reg_starttry);
2161 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2162 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2163 ? (5 + taill) - pref_len : PL_regeol - locinput);
2164 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2168 if (pref0_len > pref_len)
2169 pref0_len = pref_len;
2170 regprop(prop, scan);
2174 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2176 locinput - pref_len;
2177 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2178 char *s1 = do_utf8 ?
2179 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2180 pref_len - pref0_len, 60, 0) :
2181 locinput - pref_len + pref0_len;
2182 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2183 char *s2 = do_utf8 ?
2184 pv_uni_display(dsv2, (U8*)locinput,
2185 PL_regeol - locinput, 60, 0) :
2187 int len2 = do_utf8 ? strlen(s2) : l;
2188 PerlIO_printf(Perl_debug_log,
2189 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2190 (IV)(locinput - PL_bostr),
2197 (docolor ? "" : "> <"),
2201 15 - l - pref_len + 1,
2203 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2208 next = scan + NEXT_OFF(scan);
2214 if (locinput == PL_bostr || (PL_multiline &&
2215 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2217 /* regtill = regbol; */
2222 if (locinput == PL_bostr ||
2223 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2229 if (locinput == PL_bostr)
2233 if (locinput == PL_reg_ganch)
2243 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2248 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2250 if (PL_regeol - locinput > 1)
2254 if (PL_regeol != locinput)
2258 if (!nextchr && locinput >= PL_regeol)
2261 locinput += PL_utf8skip[nextchr];
2262 if (locinput > PL_regeol)
2264 nextchr = UCHARAT(locinput);
2267 nextchr = UCHARAT(++locinput);
2270 if (!nextchr && locinput >= PL_regeol)
2272 nextchr = UCHARAT(++locinput);
2275 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2278 locinput += PL_utf8skip[nextchr];
2279 if (locinput > PL_regeol)
2281 nextchr = UCHARAT(locinput);
2284 nextchr = UCHARAT(++locinput);
2289 if (do_utf8 != (UTF!=0)) {
2290 /* The target and the pattern have differing utf8ness. */
2296 /* The target is utf8, the pattern is not utf8. */
2300 if (NATIVE_TO_UNI(*(U8*)s) !=
2301 utf8_to_uvchr((U8*)l, &ulen))
2308 /* The target is not utf8, the pattern is utf8. */
2312 if (NATIVE_TO_UNI(*((U8*)l)) !=
2313 utf8_to_uvchr((U8*)s, &ulen))
2320 nextchr = UCHARAT(locinput);
2323 /* The target and the pattern have the same utf8ness. */
2324 /* Inline the first character, for speed. */
2325 if (UCHARAT(s) != nextchr)
2327 if (PL_regeol - locinput < ln)
2329 if (ln > 1 && memNE(s, locinput, ln))
2332 nextchr = UCHARAT(locinput);
2335 PL_reg_flags |= RF_tainted;
2341 if (do_utf8 || UTF) {
2342 /* Either target or the pattern are utf8. */
2344 char *e = PL_regeol;
2346 if (ibcmp_utf8(s, 0, ln, do_utf8,
2350 nextchr = UCHARAT(locinput);
2354 /* Neither the target and the pattern are utf8. */
2356 /* Inline the first character, for speed. */
2357 if (UCHARAT(s) != nextchr &&
2358 UCHARAT(s) != ((OP(scan) == EXACTF)
2359 ? PL_fold : PL_fold_locale)[nextchr])
2361 if (PL_regeol - locinput < ln)
2363 if (ln > 1 && (OP(scan) == EXACTF
2364 ? ibcmp(s, locinput, ln)
2365 : ibcmp_locale(s, locinput, ln)))
2368 nextchr = UCHARAT(locinput);
2372 STRLEN inclasslen = PL_regeol - locinput;
2374 if (!reginclasslen(scan, (U8*)locinput, &inclasslen, do_utf8))
2376 if (locinput >= PL_regeol)
2378 locinput += inclasslen;
2379 nextchr = UCHARAT(locinput);
2383 nextchr = UCHARAT(locinput);
2384 if (!reginclass(scan, (U8*)locinput, do_utf8))
2386 if (!nextchr && locinput >= PL_regeol)
2388 nextchr = UCHARAT(++locinput);
2392 PL_reg_flags |= RF_tainted;
2398 LOAD_UTF8_CHARCLASS(alnum,"a");
2399 if (!(OP(scan) == ALNUM
2400 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2401 : isALNUM_LC_utf8((U8*)locinput)))
2405 locinput += PL_utf8skip[nextchr];
2406 nextchr = UCHARAT(locinput);
2409 if (!(OP(scan) == ALNUM
2410 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2412 nextchr = UCHARAT(++locinput);
2415 PL_reg_flags |= RF_tainted;
2418 if (!nextchr && locinput >= PL_regeol)
2421 LOAD_UTF8_CHARCLASS(alnum,"a");
2422 if (OP(scan) == NALNUM
2423 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2424 : isALNUM_LC_utf8((U8*)locinput))
2428 locinput += PL_utf8skip[nextchr];
2429 nextchr = UCHARAT(locinput);
2432 if (OP(scan) == NALNUM
2433 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2435 nextchr = UCHARAT(++locinput);
2439 PL_reg_flags |= RF_tainted;
2443 /* was last char in word? */
2445 if (locinput == PL_bostr)
2448 U8 *r = reghop((U8*)locinput, -1);
2450 ln = utf8n_to_uvchr(r, s - (char*)r, 0, 0);
2452 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2453 ln = isALNUM_uni(ln);
2454 LOAD_UTF8_CHARCLASS(alnum,"a");
2455 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2458 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2459 n = isALNUM_LC_utf8((U8*)locinput);
2463 ln = (locinput != PL_bostr) ?
2464 UCHARAT(locinput - 1) : '\n';
2465 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2467 n = isALNUM(nextchr);
2470 ln = isALNUM_LC(ln);
2471 n = isALNUM_LC(nextchr);
2474 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2475 OP(scan) == BOUNDL))
2479 PL_reg_flags |= RF_tainted;
2485 if (UTF8_IS_CONTINUED(nextchr)) {
2486 LOAD_UTF8_CHARCLASS(space," ");
2487 if (!(OP(scan) == SPACE
2488 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2489 : isSPACE_LC_utf8((U8*)locinput)))
2493 locinput += PL_utf8skip[nextchr];
2494 nextchr = UCHARAT(locinput);
2497 if (!(OP(scan) == SPACE
2498 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2500 nextchr = UCHARAT(++locinput);
2503 if (!(OP(scan) == SPACE
2504 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2506 nextchr = UCHARAT(++locinput);
2510 PL_reg_flags |= RF_tainted;
2513 if (!nextchr && locinput >= PL_regeol)
2516 LOAD_UTF8_CHARCLASS(space," ");
2517 if (OP(scan) == NSPACE
2518 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2519 : isSPACE_LC_utf8((U8*)locinput))
2523 locinput += PL_utf8skip[nextchr];
2524 nextchr = UCHARAT(locinput);
2527 if (OP(scan) == NSPACE
2528 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2530 nextchr = UCHARAT(++locinput);
2533 PL_reg_flags |= RF_tainted;
2539 LOAD_UTF8_CHARCLASS(digit,"0");
2540 if (!(OP(scan) == DIGIT
2541 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2542 : isDIGIT_LC_utf8((U8*)locinput)))
2546 locinput += PL_utf8skip[nextchr];
2547 nextchr = UCHARAT(locinput);
2550 if (!(OP(scan) == DIGIT
2551 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2553 nextchr = UCHARAT(++locinput);
2556 PL_reg_flags |= RF_tainted;
2559 if (!nextchr && locinput >= PL_regeol)
2562 LOAD_UTF8_CHARCLASS(digit,"0");
2563 if (OP(scan) == NDIGIT
2564 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2565 : isDIGIT_LC_utf8((U8*)locinput))
2569 locinput += PL_utf8skip[nextchr];
2570 nextchr = UCHARAT(locinput);
2573 if (OP(scan) == NDIGIT
2574 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2576 nextchr = UCHARAT(++locinput);
2579 if (locinput >= PL_regeol)
2582 LOAD_UTF8_CHARCLASS(mark,"~");
2583 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2585 locinput += PL_utf8skip[nextchr];
2586 while (locinput < PL_regeol &&
2587 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2588 locinput += UTF8SKIP(locinput);
2589 if (locinput > PL_regeol)
2594 nextchr = UCHARAT(locinput);
2597 PL_reg_flags |= RF_tainted;
2601 n = ARG(scan); /* which paren pair */
2602 ln = PL_regstartp[n];
2603 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2604 if (*PL_reglastparen < n || ln == -1)
2605 sayNO; /* Do not match unless seen CLOSEn. */
2606 if (ln == PL_regendp[n])
2610 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2612 char *e = PL_bostr + PL_regendp[n];
2614 * Note that we can't do the "other character" lookup trick as
2615 * in the 8-bit case (no pun intended) because in Unicode we
2616 * have to map both upper and title case to lower case.
2618 if (OP(scan) == REFF) {
2619 STRLEN ulen1, ulen2;
2620 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2621 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2625 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2626 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2627 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2634 nextchr = UCHARAT(locinput);
2638 /* Inline the first character, for speed. */
2639 if (UCHARAT(s) != nextchr &&
2641 (UCHARAT(s) != ((OP(scan) == REFF
2642 ? PL_fold : PL_fold_locale)[nextchr]))))
2644 ln = PL_regendp[n] - ln;
2645 if (locinput + ln > PL_regeol)
2647 if (ln > 1 && (OP(scan) == REF
2648 ? memNE(s, locinput, ln)
2650 ? ibcmp(s, locinput, ln)
2651 : ibcmp_locale(s, locinput, ln))))
2654 nextchr = UCHARAT(locinput);
2665 OP_4tree *oop = PL_op;
2666 COP *ocurcop = PL_curcop;
2667 SV **ocurpad = PL_curpad;
2671 PL_op = (OP_4tree*)PL_regdata->data[n];
2672 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2673 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2674 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2678 CALLRUNOPS(aTHX); /* Scalar context. */
2681 ret = Nullsv; /* protect against empty (?{}) blocks. */
2689 PL_curpad = ocurpad;
2690 PL_curcop = ocurcop;
2692 if (logical == 2) { /* Postponed subexpression. */
2694 MAGIC *mg = Null(MAGIC*);
2696 CHECKPOINT cp, lastcp;
2698 if(SvROK(ret) || SvRMAGICAL(ret)) {
2699 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2702 mg = mg_find(sv, PERL_MAGIC_qr);
2705 re = (regexp *)mg->mg_obj;
2706 (void)ReREFCNT_inc(re);
2710 char *t = SvPV(ret, len);
2712 char *oprecomp = PL_regprecomp;
2713 I32 osize = PL_regsize;
2714 I32 onpar = PL_regnpar;
2717 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2719 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2720 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2722 PL_regprecomp = oprecomp;
2727 PerlIO_printf(Perl_debug_log,
2728 "Entering embedded `%s%.60s%s%s'\n",
2732 (strlen(re->precomp) > 60 ? "..." : ""))
2735 state.prev = PL_reg_call_cc;
2736 state.cc = PL_regcc;
2737 state.re = PL_reg_re;
2741 cp = regcppush(0); /* Save *all* the positions. */
2744 state.ss = PL_savestack_ix;
2745 *PL_reglastparen = 0;
2746 *PL_reglastcloseparen = 0;
2747 PL_reg_call_cc = &state;
2748 PL_reginput = locinput;
2750 /* XXXX This is too dramatic a measure... */
2753 if (regmatch(re->program + 1)) {
2754 /* Even though we succeeded, we need to restore
2755 global variables, since we may be wrapped inside
2756 SUSPEND, thus the match may be not finished yet. */
2758 /* XXXX Do this only if SUSPENDed? */
2759 PL_reg_call_cc = state.prev;
2760 PL_regcc = state.cc;
2761 PL_reg_re = state.re;
2762 cache_re(PL_reg_re);
2764 /* XXXX This is too dramatic a measure... */
2767 /* These are needed even if not SUSPEND. */
2773 REGCP_UNWIND(lastcp);
2775 PL_reg_call_cc = state.prev;
2776 PL_regcc = state.cc;
2777 PL_reg_re = state.re;
2778 cache_re(PL_reg_re);
2780 /* XXXX This is too dramatic a measure... */
2790 sv_setsv(save_scalar(PL_replgv), ret);
2794 n = ARG(scan); /* which paren pair */
2795 PL_reg_start_tmp[n] = locinput;
2800 n = ARG(scan); /* which paren pair */
2801 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2802 PL_regendp[n] = locinput - PL_bostr;
2803 if (n > *PL_reglastparen)
2804 *PL_reglastparen = n;
2805 *PL_reglastcloseparen = n;
2808 n = ARG(scan); /* which paren pair */
2809 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2812 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2814 next = NEXTOPER(NEXTOPER(scan));
2816 next = scan + ARG(scan);
2817 if (OP(next) == IFTHEN) /* Fake one. */
2818 next = NEXTOPER(NEXTOPER(next));
2822 logical = scan->flags;
2824 /*******************************************************************
2825 PL_regcc contains infoblock about the innermost (...)* loop, and
2826 a pointer to the next outer infoblock.
2828 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2830 1) After matching X, regnode for CURLYX is processed;
2832 2) This regnode creates infoblock on the stack, and calls
2833 regmatch() recursively with the starting point at WHILEM node;
2835 3) Each hit of WHILEM node tries to match A and Z (in the order
2836 depending on the current iteration, min/max of {min,max} and
2837 greediness). The information about where are nodes for "A"
2838 and "Z" is read from the infoblock, as is info on how many times "A"
2839 was already matched, and greediness.
2841 4) After A matches, the same WHILEM node is hit again.
2843 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2844 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2845 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2846 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2847 of the external loop.
2849 Currently present infoblocks form a tree with a stem formed by PL_curcc
2850 and whatever it mentions via ->next, and additional attached trees
2851 corresponding to temporarily unset infoblocks as in "5" above.
2853 In the following picture infoblocks for outer loop of
2854 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2855 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2856 infoblocks are drawn below the "reset" infoblock.
2858 In fact in the picture below we do not show failed matches for Z and T
2859 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2860 more obvious *why* one needs to *temporary* unset infoblocks.]
2862 Matched REx position InfoBlocks Comment
2866 Y A)*?Z)*?T x <- O <- I
2867 YA )*?Z)*?T x <- O <- I
2868 YA A)*?Z)*?T x <- O <- I
2869 YAA )*?Z)*?T x <- O <- I
2870 YAA Z)*?T x <- O # Temporary unset I
2873 YAAZ Y(A)*?Z)*?T x <- O
2876 YAAZY (A)*?Z)*?T x <- O
2879 YAAZY A)*?Z)*?T x <- O <- I
2882 YAAZYA )*?Z)*?T x <- O <- I
2885 YAAZYA Z)*?T x <- O # Temporary unset I
2891 YAAZYAZ T x # Temporary unset O
2898 *******************************************************************/
2901 CHECKPOINT cp = PL_savestack_ix;
2902 /* No need to save/restore up to this paren */
2903 I32 parenfloor = scan->flags;
2905 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2907 cc.oldcc = PL_regcc;
2909 /* XXXX Probably it is better to teach regpush to support
2910 parenfloor > PL_regsize... */
2911 if (parenfloor > *PL_reglastparen)
2912 parenfloor = *PL_reglastparen; /* Pessimization... */
2913 cc.parenfloor = parenfloor;
2915 cc.min = ARG1(scan);
2916 cc.max = ARG2(scan);
2917 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2921 PL_reginput = locinput;
2922 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2924 PL_regcc = cc.oldcc;
2930 * This is really hard to understand, because after we match
2931 * what we're trying to match, we must make sure the rest of
2932 * the REx is going to match for sure, and to do that we have
2933 * to go back UP the parse tree by recursing ever deeper. And
2934 * if it fails, we have to reset our parent's current state
2935 * that we can try again after backing off.
2938 CHECKPOINT cp, lastcp;
2939 CURCUR* cc = PL_regcc;
2940 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2942 n = cc->cur + 1; /* how many we know we matched */
2943 PL_reginput = locinput;
2946 PerlIO_printf(Perl_debug_log,
2947 "%*s %ld out of %ld..%ld cc=%lx\n",
2948 REPORT_CODE_OFF+PL_regindent*2, "",
2949 (long)n, (long)cc->min,
2950 (long)cc->max, (long)cc)
2953 /* If degenerate scan matches "", assume scan done. */
2955 if (locinput == cc->lastloc && n >= cc->min) {
2956 PL_regcc = cc->oldcc;
2960 PerlIO_printf(Perl_debug_log,
2961 "%*s empty match detected, try continuation...\n",
2962 REPORT_CODE_OFF+PL_regindent*2, "")
2964 if (regmatch(cc->next))
2972 /* First just match a string of min scans. */
2976 cc->lastloc = locinput;
2977 if (regmatch(cc->scan))
2980 cc->lastloc = lastloc;
2985 /* Check whether we already were at this position.
2986 Postpone detection until we know the match is not
2987 *that* much linear. */
2988 if (!PL_reg_maxiter) {
2989 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
2990 PL_reg_leftiter = PL_reg_maxiter;
2992 if (PL_reg_leftiter-- == 0) {
2993 I32 size = (PL_reg_maxiter + 7)/8;
2994 if (PL_reg_poscache) {
2995 if (PL_reg_poscache_size < size) {
2996 Renew(PL_reg_poscache, size, char);
2997 PL_reg_poscache_size = size;
2999 Zero(PL_reg_poscache, size, char);
3002 PL_reg_poscache_size = size;
3003 Newz(29, PL_reg_poscache, size, char);
3006 PerlIO_printf(Perl_debug_log,
3007 "%sDetected a super-linear match, switching on caching%s...\n",
3008 PL_colors[4], PL_colors[5])
3011 if (PL_reg_leftiter < 0) {
3012 I32 o = locinput - PL_bostr, b;
3014 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3017 if (PL_reg_poscache[o] & (1<<b)) {
3019 PerlIO_printf(Perl_debug_log,
3020 "%*s already tried at this position...\n",
3021 REPORT_CODE_OFF+PL_regindent*2, "")
3025 PL_reg_poscache[o] |= (1<<b);
3029 /* Prefer next over scan for minimal matching. */
3032 PL_regcc = cc->oldcc;
3035 cp = regcppush(cc->parenfloor);
3037 if (regmatch(cc->next)) {
3039 sayYES; /* All done. */
3041 REGCP_UNWIND(lastcp);
3047 if (n >= cc->max) { /* Maximum greed exceeded? */
3048 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3049 && !(PL_reg_flags & RF_warned)) {
3050 PL_reg_flags |= RF_warned;
3051 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3052 "Complex regular subexpression recursion",
3059 PerlIO_printf(Perl_debug_log,
3060 "%*s trying longer...\n",
3061 REPORT_CODE_OFF+PL_regindent*2, "")
3063 /* Try scanning more and see if it helps. */
3064 PL_reginput = locinput;
3066 cc->lastloc = locinput;
3067 cp = regcppush(cc->parenfloor);
3069 if (regmatch(cc->scan)) {
3073 REGCP_UNWIND(lastcp);
3076 cc->lastloc = lastloc;
3080 /* Prefer scan over next for maximal matching. */
3082 if (n < cc->max) { /* More greed allowed? */
3083 cp = regcppush(cc->parenfloor);
3085 cc->lastloc = locinput;
3087 if (regmatch(cc->scan)) {
3091 REGCP_UNWIND(lastcp);
3092 regcppop(); /* Restore some previous $<digit>s? */
3093 PL_reginput = locinput;
3095 PerlIO_printf(Perl_debug_log,
3096 "%*s failed, try continuation...\n",
3097 REPORT_CODE_OFF+PL_regindent*2, "")
3100 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3101 && !(PL_reg_flags & RF_warned)) {
3102 PL_reg_flags |= RF_warned;
3103 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3104 "Complex regular subexpression recursion",
3108 /* Failed deeper matches of scan, so see if this one works. */
3109 PL_regcc = cc->oldcc;
3112 if (regmatch(cc->next))
3118 cc->lastloc = lastloc;
3123 next = scan + ARG(scan);
3126 inner = NEXTOPER(NEXTOPER(scan));
3129 inner = NEXTOPER(scan);
3133 if (OP(next) != c1) /* No choice. */
3134 next = inner; /* Avoid recursion. */
3136 I32 lastparen = *PL_reglastparen;
3138 re_unwind_branch_t *uw;
3140 /* Put unwinding data on stack */
3141 unwind1 = SSNEWt(1,re_unwind_branch_t);
3142 uw = SSPTRt(unwind1,re_unwind_branch_t);
3145 uw->type = ((c1 == BRANCH)
3147 : RE_UNWIND_BRANCHJ);
3148 uw->lastparen = lastparen;
3150 uw->locinput = locinput;
3151 uw->nextchr = nextchr;
3153 uw->regindent = ++PL_regindent;
3156 REGCP_SET(uw->lastcp);
3158 /* Now go into the first branch */
3171 /* We suppose that the next guy does not need
3172 backtracking: in particular, it is of constant length,
3173 and has no parenths to influence future backrefs. */
3174 ln = ARG1(scan); /* min to match */
3175 n = ARG2(scan); /* max to match */
3176 paren = scan->flags;
3178 if (paren > PL_regsize)
3180 if (paren > *PL_reglastparen)
3181 *PL_reglastparen = paren;
3183 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3185 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3186 PL_reginput = locinput;
3189 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3191 /* if we matched something zero-length we don't need to
3192 backtrack - capturing parens are already defined, so
3193 the caveat in the maximal case doesn't apply
3195 XXXX if ln == 0, we can redo this check first time
3196 through the following loop
3199 n = ln; /* don't backtrack */
3200 locinput = PL_reginput;
3201 if (HAS_TEXT(next) || JUMPABLE(next)) {
3202 regnode *text_node = next;
3204 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3206 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3208 if (PL_regkind[(U8)OP(text_node)] == REF) {
3210 n = ARG(text_node); /* which paren pair */
3211 ln = PL_regstartp[n];
3212 /* assume yes if we haven't seen CLOSEn */
3214 *PL_reglastparen < n ||
3221 c1 = *(PL_bostr + ln);
3223 else { c1 = (U8)*STRING(text_node); }
3224 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3226 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3227 c2 = PL_fold_locale[c1];
3236 /* This may be improved if l == 0. */
3237 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3238 /* If it could work, try it. */
3240 UCHARAT(PL_reginput) == c1 ||
3241 UCHARAT(PL_reginput) == c2)
3245 PL_regstartp[paren] =
3246 HOPc(PL_reginput, -l) - PL_bostr;
3247 PL_regendp[paren] = PL_reginput - PL_bostr;
3250 PL_regendp[paren] = -1;
3254 REGCP_UNWIND(lastcp);
3256 /* Couldn't or didn't -- move forward. */
3257 PL_reginput = locinput;
3258 if (regrepeat_hard(scan, 1, &l)) {
3260 locinput = PL_reginput;
3267 n = regrepeat_hard(scan, n, &l);
3268 /* if we matched something zero-length we don't need to
3269 backtrack, unless the minimum count is zero and we
3270 are capturing the result - in that case the capture
3271 being defined or not may affect later execution
3273 if (n != 0 && l == 0 && !(paren && ln == 0))
3274 ln = n; /* don't backtrack */
3275 locinput = PL_reginput;
3277 PerlIO_printf(Perl_debug_log,
3278 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3279 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3283 if (HAS_TEXT(next) || JUMPABLE(next)) {
3284 regnode *text_node = next;
3286 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3288 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3290 if (PL_regkind[(U8)OP(text_node)] == REF) {
3292 n = ARG(text_node); /* which paren pair */
3293 ln = PL_regstartp[n];
3294 /* assume yes if we haven't seen CLOSEn */
3296 *PL_reglastparen < n ||
3303 c1 = *(PL_bostr + ln);
3305 else { c1 = (U8)*STRING(text_node); }
3307 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3309 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3310 c2 = PL_fold_locale[c1];
3321 /* If it could work, try it. */
3323 UCHARAT(PL_reginput) == c1 ||
3324 UCHARAT(PL_reginput) == c2)
3327 PerlIO_printf(Perl_debug_log,
3328 "%*s trying tail with n=%"IVdf"...\n",
3329 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3333 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3334 PL_regendp[paren] = PL_reginput - PL_bostr;
3337 PL_regendp[paren] = -1;
3341 REGCP_UNWIND(lastcp);
3343 /* Couldn't or didn't -- back up. */
3345 locinput = HOPc(locinput, -l);
3346 PL_reginput = locinput;
3353 paren = scan->flags; /* Which paren to set */
3354 if (paren > PL_regsize)
3356 if (paren > *PL_reglastparen)
3357 *PL_reglastparen = paren;
3358 ln = ARG1(scan); /* min to match */
3359 n = ARG2(scan); /* max to match */
3360 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3364 ln = ARG1(scan); /* min to match */
3365 n = ARG2(scan); /* max to match */
3366 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3371 scan = NEXTOPER(scan);
3377 scan = NEXTOPER(scan);
3381 * Lookahead to avoid useless match attempts
3382 * when we know what character comes next.
3386 * Used to only do .*x and .*?x, but now it allows
3387 * for )'s, ('s and (?{ ... })'s to be in the way
3388 * of the quantifier and the EXACT-like node. -- japhy
3391 if (HAS_TEXT(next) || JUMPABLE(next)) {
3393 regnode *text_node = next;
3395 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3397 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3399 if (PL_regkind[(U8)OP(text_node)] == REF) {
3401 n = ARG(text_node); /* which paren pair */
3402 ln = PL_regstartp[n];
3403 /* assume yes if we haven't seen CLOSEn */
3405 *PL_reglastparen < n ||
3410 goto assume_ok_easy;
3412 s = (U8*)PL_bostr + ln;
3414 else { s = (U8*)STRING(text_node); }
3418 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3420 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3421 c2 = PL_fold_locale[c1];
3424 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3425 STRLEN ulen1, ulen2;
3426 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3427 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3429 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3430 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3432 c1 = utf8_to_uvuni(tmpbuf1, 0);
3433 c2 = utf8_to_uvuni(tmpbuf2, 0);
3436 c2 = c1 = utf8_to_uvchr(s, NULL);
3444 PL_reginput = locinput;
3448 if (ln && regrepeat(scan, ln) < ln)
3450 locinput = PL_reginput;
3453 char *e; /* Should not check after this */
3454 char *old = locinput;
3456 if (n == REG_INFTY) {
3459 while (UTF8_IS_CONTINUATION(*(U8*)e))
3465 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3469 e = locinput + n - ln;
3475 /* Find place 'next' could work */
3478 while (locinput <= e &&
3479 UCHARAT(locinput) != c1)
3482 while (locinput <= e
3483 && UCHARAT(locinput) != c1
3484 && UCHARAT(locinput) != c2)
3487 count = locinput - old;
3494 utf8_to_uvchr((U8*)locinput, &len) != c1;
3499 for (count = 0; locinput <= e; count++) {
3500 UV c = utf8_to_uvchr((U8*)locinput, &len);
3501 if (c == c1 || c == c2)
3509 /* PL_reginput == old now */
3510 if (locinput != old) {
3511 ln = 1; /* Did some */
3512 if (regrepeat(scan, count) < count)
3515 /* PL_reginput == locinput now */
3516 TRYPAREN(paren, ln, locinput);
3517 PL_reginput = locinput; /* Could be reset... */
3518 REGCP_UNWIND(lastcp);
3519 /* Couldn't or didn't -- move forward. */
3522 locinput += UTF8SKIP(locinput);
3528 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3532 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3534 c = UCHARAT(PL_reginput);
3535 /* If it could work, try it. */
3536 if (c == c1 || c == c2)
3538 TRYPAREN(paren, n, PL_reginput);
3539 REGCP_UNWIND(lastcp);
3542 /* If it could work, try it. */
3543 else if (c1 == -1000)
3545 TRYPAREN(paren, n, PL_reginput);
3546 REGCP_UNWIND(lastcp);
3548 /* Couldn't or didn't -- move forward. */
3549 PL_reginput = locinput;
3550 if (regrepeat(scan, 1)) {
3552 locinput = PL_reginput;
3560 n = regrepeat(scan, n);
3561 locinput = PL_reginput;
3562 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3563 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3564 ln = n; /* why back off? */
3565 /* ...because $ and \Z can match before *and* after
3566 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3567 We should back off by one in this case. */
3568 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3577 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3579 c = UCHARAT(PL_reginput);
3581 /* If it could work, try it. */
3582 if (c1 == -1000 || c == c1 || c == c2)
3584 TRYPAREN(paren, n, PL_reginput);
3585 REGCP_UNWIND(lastcp);
3587 /* Couldn't or didn't -- back up. */
3589 PL_reginput = locinput = HOPc(locinput, -1);
3597 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3599 c = UCHARAT(PL_reginput);
3601 /* If it could work, try it. */
3602 if (c1 == -1000 || c == c1 || c == c2)
3604 TRYPAREN(paren, n, PL_reginput);
3605 REGCP_UNWIND(lastcp);
3607 /* Couldn't or didn't -- back up. */
3609 PL_reginput = locinput = HOPc(locinput, -1);
3616 if (PL_reg_call_cc) {
3617 re_cc_state *cur_call_cc = PL_reg_call_cc;
3618 CURCUR *cctmp = PL_regcc;
3619 regexp *re = PL_reg_re;
3620 CHECKPOINT cp, lastcp;
3622 cp = regcppush(0); /* Save *all* the positions. */
3624 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3626 PL_reginput = locinput; /* Make position available to
3628 cache_re(PL_reg_call_cc->re);
3629 PL_regcc = PL_reg_call_cc->cc;
3630 PL_reg_call_cc = PL_reg_call_cc->prev;
3631 if (regmatch(cur_call_cc->node)) {
3632 PL_reg_call_cc = cur_call_cc;
3636 REGCP_UNWIND(lastcp);
3638 PL_reg_call_cc = cur_call_cc;
3644 PerlIO_printf(Perl_debug_log,
3645 "%*s continuation failed...\n",
3646 REPORT_CODE_OFF+PL_regindent*2, "")
3650 if (locinput < PL_regtill) {
3651 DEBUG_r(PerlIO_printf(Perl_debug_log,
3652 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3654 (long)(locinput - PL_reg_starttry),
3655 (long)(PL_regtill - PL_reg_starttry),
3657 sayNO_FINAL; /* Cannot match: too short. */
3659 PL_reginput = locinput; /* put where regtry can find it */
3660 sayYES_FINAL; /* Success! */
3662 PL_reginput = locinput; /* put where regtry can find it */
3663 sayYES_LOUD; /* Success! */
3666 PL_reginput = locinput;
3671 s = HOPBACKc(locinput, scan->flags);
3677 PL_reginput = locinput;
3682 s = HOPBACKc(locinput, scan->flags);
3688 PL_reginput = locinput;
3691 inner = NEXTOPER(NEXTOPER(scan));
3692 if (regmatch(inner) != n) {
3707 if (OP(scan) == SUSPEND) {
3708 locinput = PL_reginput;
3709 nextchr = UCHARAT(locinput);
3714 next = scan + ARG(scan);
3719 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3720 PTR2UV(scan), OP(scan));
3721 Perl_croak(aTHX_ "regexp memory corruption");
3728 * We get here only if there's trouble -- normally "case END" is
3729 * the terminating point.
3731 Perl_croak(aTHX_ "corrupted regexp pointers");
3737 PerlIO_printf(Perl_debug_log,
3738 "%*s %scould match...%s\n",
3739 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3743 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3744 PL_colors[4],PL_colors[5]));
3750 #if 0 /* Breaks $^R */
3758 PerlIO_printf(Perl_debug_log,
3759 "%*s %sfailed...%s\n",
3760 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3766 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3769 case RE_UNWIND_BRANCH:
3770 case RE_UNWIND_BRANCHJ:
3772 re_unwind_branch_t *uwb = &(uw->branch);
3773 I32 lastparen = uwb->lastparen;
3775 REGCP_UNWIND(uwb->lastcp);
3776 for (n = *PL_reglastparen; n > lastparen; n--)
3778 *PL_reglastparen = n;
3779 scan = next = uwb->next;
3781 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3782 ? BRANCH : BRANCHJ) ) { /* Failure */
3789 /* Have more choice yet. Reuse the same uwb. */
3791 if ((n = (uwb->type == RE_UNWIND_BRANCH
3792 ? NEXT_OFF(next) : ARG(next))))
3795 next = NULL; /* XXXX Needn't unwinding in this case... */
3797 next = NEXTOPER(scan);
3798 if (uwb->type == RE_UNWIND_BRANCHJ)
3799 next = NEXTOPER(next);
3800 locinput = uwb->locinput;
3801 nextchr = uwb->nextchr;
3803 PL_regindent = uwb->regindent;
3810 Perl_croak(aTHX_ "regexp unwind memory corruption");
3821 - regrepeat - repeatedly match something simple, report how many
3824 * [This routine now assumes that it will only match on things of length 1.
3825 * That was true before, but now we assume scan - reginput is the count,
3826 * rather than incrementing count on every character. [Er, except utf8.]]
3829 S_regrepeat(pTHX_ regnode *p, I32 max)
3831 register char *scan;
3833 register char *loceol = PL_regeol;
3834 register I32 hardcount = 0;
3835 register bool do_utf8 = PL_reg_match_utf8;
3838 if (max != REG_INFTY && max < loceol - scan)
3839 loceol = scan + max;
3844 while (scan < loceol && hardcount < max && *scan != '\n') {
3845 scan += UTF8SKIP(scan);
3849 while (scan < loceol && *scan != '\n')
3859 case EXACT: /* length of string is 1 */
3861 while (scan < loceol && UCHARAT(scan) == c)
3864 case EXACTF: /* length of string is 1 */
3866 while (scan < loceol &&
3867 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3870 case EXACTFL: /* length of string is 1 */
3871 PL_reg_flags |= RF_tainted;
3873 while (scan < loceol &&
3874 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3880 while (hardcount < max && scan < loceol &&
3881 reginclass(p, (U8*)scan, do_utf8)) {
3882 scan += UTF8SKIP(scan);
3886 while (scan < loceol && reginclass(p, (U8*)scan, do_utf8))
3893 LOAD_UTF8_CHARCLASS(alnum,"a");
3894 while (hardcount < max && scan < loceol &&
3895 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3896 scan += UTF8SKIP(scan);
3900 while (scan < loceol && isALNUM(*scan))
3905 PL_reg_flags |= RF_tainted;
3908 while (hardcount < max && scan < loceol &&
3909 isALNUM_LC_utf8((U8*)scan)) {
3910 scan += UTF8SKIP(scan);
3914 while (scan < loceol && isALNUM_LC(*scan))
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(space," ");
3950 while (hardcount < max && scan < loceol &&
3952 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3953 scan += UTF8SKIP(scan);
3957 while (scan < loceol && isSPACE(*scan))
3962 PL_reg_flags |= RF_tainted;
3965 while (hardcount < max && scan < loceol &&
3966 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3967 scan += UTF8SKIP(scan);
3971 while (scan < loceol && isSPACE_LC(*scan))
3978 LOAD_UTF8_CHARCLASS(space," ");
3979 while (hardcount < max && scan < loceol &&
3981 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3982 scan += UTF8SKIP(scan);
3986 while (scan < loceol && !isSPACE(*scan))
3991 PL_reg_flags |= RF_tainted;
3994 while (hardcount < max && scan < loceol &&
3995 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3996 scan += UTF8SKIP(scan);
4000 while (scan < loceol && !isSPACE_LC(*scan))
4007 LOAD_UTF8_CHARCLASS(digit,"0");
4008 while (hardcount < max && scan < loceol &&
4009 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4010 scan += UTF8SKIP(scan);
4014 while (scan < loceol && isDIGIT(*scan))
4021 LOAD_UTF8_CHARCLASS(digit,"0");
4022 while (hardcount < max && scan < loceol &&
4023 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4024 scan += UTF8SKIP(scan);
4028 while (scan < loceol && !isDIGIT(*scan))
4032 default: /* Called on something of 0 width. */
4033 break; /* So match right here or not at all. */
4039 c = scan - PL_reginput;
4044 SV *prop = sv_newmortal();
4047 PerlIO_printf(Perl_debug_log,
4048 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4049 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4056 - regrepeat_hard - repeatedly match something, report total lenth and length
4058 * The repeater is supposed to have constant length.
4062 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4064 register char *scan = Nullch;
4065 register char *start;
4066 register char *loceol = PL_regeol;
4068 I32 count = 0, res = 1;
4073 start = PL_reginput;
4074 if (PL_reg_match_utf8) {
4075 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4078 while (start < PL_reginput) {
4080 start += UTF8SKIP(start);
4091 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4093 *lp = l = PL_reginput - start;
4094 if (max != REG_INFTY && l*max < loceol - scan)
4095 loceol = scan + l*max;
4108 - regclass_swash - prepare the utf8 swash
4112 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** listsvp, SV **altsvp)
4118 if (PL_regdata && PL_regdata->count) {
4121 if (PL_regdata->what[n] == 's') {
4122 SV *rv = (SV*)PL_regdata->data[n];
4123 AV *av = (AV*)SvRV((SV*)rv);
4126 si = *av_fetch(av, 0, FALSE);
4127 a = av_fetch(av, 1, FALSE);
4128 b = av_fetch(av, 2, FALSE);
4132 else if (si && doinit) {
4133 sw = swash_init("utf8", "", si, 1, 0);
4134 (void)av_store(av, 1, sw);
4150 - reginclass - determine if a character falls into a character class
4154 S_reginclasslen(pTHX_ register regnode *n, register U8* p, STRLEN* lenp, register bool do_utf8)
4156 char flags = ANYOF_FLAGS(n);
4162 c = do_utf8 ? utf8_to_uvchr(p, &len) : *p;
4164 plen = lenp ? *lenp : UNISKIP(c);
4165 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4168 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4169 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4172 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4176 SV *sw = regclass_swash(n, TRUE, 0, (SV**)&av);
4179 if (swash_fetch(sw, p, do_utf8))
4181 else if (flags & ANYOF_FOLD) {
4182 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
4185 if (!match && lenp && av) {
4188 for (i = 0; i <= av_len(av); i++) {
4189 SV* sv = *av_fetch(av, i, FALSE);
4191 char *s = SvPV(sv, len);
4193 if (len <= plen && memEQ(s, p, len)) {
4201 to_utf8_fold(p, tmpbuf, &tmplen);
4202 if (swash_fetch(sw, tmpbuf, do_utf8))
4206 to_utf8_upper(p, tmpbuf, &tmplen);
4207 if (swash_fetch(sw, tmpbuf, do_utf8))
4213 if (match && lenp && *lenp == 0)
4216 if (!match && c < 256) {
4217 if (ANYOF_BITMAP_TEST(n, c))
4219 else if (flags & ANYOF_FOLD) {
4222 if (flags & ANYOF_LOCALE) {
4223 PL_reg_flags |= RF_tainted;
4224 f = PL_fold_locale[c];
4228 if (f != c && ANYOF_BITMAP_TEST(n, f))
4232 if (!match && (flags & ANYOF_CLASS)) {
4233 PL_reg_flags |= RF_tainted;
4235 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4236 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4237 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4238 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4239 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4240 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4241 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4242 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4243 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4244 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4245 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4246 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4247 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4248 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4249 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4250 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4251 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4252 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4253 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4254 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4255 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4256 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4257 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4258 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4259 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4260 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4261 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4262 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4263 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4264 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4265 ) /* How's that for a conditional? */
4272 return (flags & ANYOF_INVERT) ? !match : match;
4276 S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8)
4278 return S_reginclasslen(aTHX_ n, p, 0, do_utf8);
4282 S_reghop(pTHX_ U8 *s, I32 off)
4284 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4288 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4291 while (off-- && s < lim) {
4292 /* XXX could check well-formedness here */
4300 if (UTF8_IS_CONTINUED(*s)) {
4301 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4304 /* XXX could check well-formedness here */
4312 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4314 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4318 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4321 while (off-- && s < lim) {
4322 /* XXX could check well-formedness here */
4332 if (UTF8_IS_CONTINUED(*s)) {
4333 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4336 /* XXX could check well-formedness here */
4348 restore_pos(pTHX_ void *arg)
4350 if (PL_reg_eval_set) {
4351 if (PL_reg_oldsaved) {
4352 PL_reg_re->subbeg = PL_reg_oldsaved;
4353 PL_reg_re->sublen = PL_reg_oldsavedlen;
4354 RX_MATCH_COPIED_on(PL_reg_re);
4356 PL_reg_magic->mg_len = PL_reg_oldpos;
4357 PL_reg_eval_set = 0;
4358 PL_curpm = PL_reg_oldcurpm;