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 - 1 : 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];
987 c = utf8_to_uvchr((U8*)s, &len);
991 !ibcmp_utf8(s, &se, 0, do_utf8,
993 && (norun || regtry(prog, s)) )
996 uvchr_to_utf8(tmpbuf, c);
997 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
999 && (f == c1 || f == c2)
1000 && (ln == foldlen ||
1001 !ibcmp_utf8((char *)foldbuf,
1002 0, foldlen, do_utf8,
1005 && (norun || regtry(prog, s)) )
1013 c = utf8_to_uvchr((U8*)s, &len);
1015 /* Handle some of the three Greek sigmas cases.
1016 * Note that not all the possible combinations
1017 * are handled here: some of them are handled
1018 * handled by the standard folding rules, and
1019 * some of them (the character class or ANYOF
1020 * cases) are handled during compiletime in
1021 * regexec.c:S_regclass(). */
1022 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1023 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1024 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1026 if ( (c == c1 || c == c2)
1029 !ibcmp_utf8(s, &se, 0, do_utf8,
1031 && (norun || regtry(prog, s)) )
1034 uvchr_to_utf8(tmpbuf, c);
1035 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1037 && (f == c1 || f == c2)
1038 && (ln == foldlen ||
1039 !ibcmp_utf8((char *)foldbuf,
1040 0, foldlen, do_utf8,
1043 && (norun || regtry(prog, s)) )
1054 && (ln == 1 || !(OP(c) == EXACTF
1056 : ibcmp_locale(s, m, ln)))
1057 && (norun || regtry(prog, s)) )
1063 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1064 && (ln == 1 || !(OP(c) == EXACTF
1066 : ibcmp_locale(s, m, ln)))
1067 && (norun || regtry(prog, s)) )
1074 PL_reg_flags |= RF_tainted;
1081 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1084 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1086 tmp = ((OP(c) == BOUND ?
1087 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1088 LOAD_UTF8_CHARCLASS(alnum,"a");
1089 while (s < strend) {
1090 if (tmp == !(OP(c) == BOUND ?
1091 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1092 isALNUM_LC_utf8((U8*)s)))
1095 if ((norun || regtry(prog, s)))
1102 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1103 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1104 while (s < strend) {
1106 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1108 if ((norun || regtry(prog, s)))
1114 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1118 PL_reg_flags |= RF_tainted;
1125 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1128 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1130 tmp = ((OP(c) == NBOUND ?
1131 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1132 LOAD_UTF8_CHARCLASS(alnum,"a");
1133 while (s < strend) {
1134 if (tmp == !(OP(c) == NBOUND ?
1135 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1136 isALNUM_LC_utf8((U8*)s)))
1138 else if ((norun || regtry(prog, s)))
1144 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1145 tmp = ((OP(c) == NBOUND ?
1146 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1147 while (s < strend) {
1149 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1151 else if ((norun || regtry(prog, s)))
1156 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1161 LOAD_UTF8_CHARCLASS(alnum,"a");
1162 while (s < strend) {
1163 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1164 if (tmp && (norun || regtry(prog, s)))
1175 while (s < strend) {
1177 if (tmp && (norun || regtry(prog, s)))
1189 PL_reg_flags |= RF_tainted;
1191 while (s < strend) {
1192 if (isALNUM_LC_utf8((U8*)s)) {
1193 if (tmp && (norun || regtry(prog, s)))
1204 while (s < strend) {
1205 if (isALNUM_LC(*s)) {
1206 if (tmp && (norun || regtry(prog, s)))
1219 LOAD_UTF8_CHARCLASS(alnum,"a");
1220 while (s < strend) {
1221 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1222 if (tmp && (norun || regtry(prog, s)))
1233 while (s < strend) {
1235 if (tmp && (norun || regtry(prog, s)))
1247 PL_reg_flags |= RF_tainted;
1249 while (s < strend) {
1250 if (!isALNUM_LC_utf8((U8*)s)) {
1251 if (tmp && (norun || regtry(prog, s)))
1262 while (s < strend) {
1263 if (!isALNUM_LC(*s)) {
1264 if (tmp && (norun || regtry(prog, s)))
1277 LOAD_UTF8_CHARCLASS(space," ");
1278 while (s < strend) {
1279 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1280 if (tmp && (norun || regtry(prog, s)))
1291 while (s < strend) {
1293 if (tmp && (norun || regtry(prog, s)))
1305 PL_reg_flags |= RF_tainted;
1307 while (s < strend) {
1308 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1309 if (tmp && (norun || regtry(prog, s)))
1320 while (s < strend) {
1321 if (isSPACE_LC(*s)) {
1322 if (tmp && (norun || regtry(prog, s)))
1335 LOAD_UTF8_CHARCLASS(space," ");
1336 while (s < strend) {
1337 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1338 if (tmp && (norun || regtry(prog, s)))
1349 while (s < strend) {
1351 if (tmp && (norun || regtry(prog, s)))
1363 PL_reg_flags |= RF_tainted;
1365 while (s < strend) {
1366 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1367 if (tmp && (norun || regtry(prog, s)))
1378 while (s < strend) {
1379 if (!isSPACE_LC(*s)) {
1380 if (tmp && (norun || regtry(prog, s)))
1393 LOAD_UTF8_CHARCLASS(digit,"0");
1394 while (s < strend) {
1395 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1396 if (tmp && (norun || regtry(prog, s)))
1407 while (s < strend) {
1409 if (tmp && (norun || regtry(prog, s)))
1421 PL_reg_flags |= RF_tainted;
1423 while (s < strend) {
1424 if (isDIGIT_LC_utf8((U8*)s)) {
1425 if (tmp && (norun || regtry(prog, s)))
1436 while (s < strend) {
1437 if (isDIGIT_LC(*s)) {
1438 if (tmp && (norun || regtry(prog, s)))
1451 LOAD_UTF8_CHARCLASS(digit,"0");
1452 while (s < strend) {
1453 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1454 if (tmp && (norun || regtry(prog, s)))
1465 while (s < strend) {
1467 if (tmp && (norun || regtry(prog, s)))
1479 PL_reg_flags |= RF_tainted;
1481 while (s < strend) {
1482 if (!isDIGIT_LC_utf8((U8*)s)) {
1483 if (tmp && (norun || regtry(prog, s)))
1494 while (s < strend) {
1495 if (!isDIGIT_LC(*s)) {
1496 if (tmp && (norun || regtry(prog, s)))
1508 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1517 - regexec_flags - match a regexp against a string
1520 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1521 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1522 /* strend: pointer to null at end of string */
1523 /* strbeg: real beginning of string */
1524 /* minend: end of match must be >=minend after stringarg. */
1525 /* data: May be used for some additional optimizations. */
1526 /* nosave: For optimizations. */
1529 register regnode *c;
1530 register char *startpos = stringarg;
1531 I32 minlen; /* must match at least this many chars */
1532 I32 dontbother = 0; /* how many characters not to try at end */
1533 /* I32 start_shift = 0; */ /* Offset of the start to find
1534 constant substr. */ /* CC */
1535 I32 end_shift = 0; /* Same for the end. */ /* CC */
1536 I32 scream_pos = -1; /* Internal iterator of scream. */
1538 SV* oreplsv = GvSV(PL_replgv);
1539 bool do_utf8 = DO_UTF8(sv);
1541 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
1548 PL_regnarrate = DEBUG_r_TEST;
1551 /* Be paranoid... */
1552 if (prog == NULL || startpos == NULL) {
1553 Perl_croak(aTHX_ "NULL regexp parameter");
1557 minlen = prog->minlen;
1558 if (strend - startpos < minlen) {
1559 DEBUG_r(PerlIO_printf(Perl_debug_log,
1560 "String too short [regexec_flags]...\n"));
1564 /* Check validity of program. */
1565 if (UCHARAT(prog->program) != REG_MAGIC) {
1566 Perl_croak(aTHX_ "corrupted regexp program");
1570 PL_reg_eval_set = 0;
1573 if (prog->reganch & ROPT_UTF8)
1574 PL_reg_flags |= RF_utf8;
1576 /* Mark beginning of line for ^ and lookbehind. */
1577 PL_regbol = startpos;
1581 /* Mark end of line for $ (and such) */
1584 /* see how far we have to get to not match where we matched before */
1585 PL_regtill = startpos+minend;
1587 /* We start without call_cc context. */
1590 /* If there is a "must appear" string, look for it. */
1593 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1596 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1597 PL_reg_ganch = startpos;
1598 else if (sv && SvTYPE(sv) >= SVt_PVMG
1600 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1601 && mg->mg_len >= 0) {
1602 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1603 if (prog->reganch & ROPT_ANCH_GPOS) {
1604 if (s > PL_reg_ganch)
1609 else /* pos() not defined */
1610 PL_reg_ganch = strbeg;
1613 if (do_utf8 == (UTF!=0) &&
1614 !(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
1615 re_scream_pos_data d;
1617 d.scream_olds = &scream_olds;
1618 d.scream_pos = &scream_pos;
1619 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1621 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1622 goto phooey; /* not present */
1627 char *s = do_utf8 ? sv_uni_display(dsv, sv, 60, 0) : startpos;
1628 int len = do_utf8 ? strlen(s) : strend - startpos;
1631 PerlIO_printf(Perl_debug_log,
1632 "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
1633 PL_colors[4],PL_colors[5],PL_colors[0],
1636 (strlen(prog->precomp) > 60 ? "..." : ""),
1638 (int)(len > 60 ? 60 : len),
1640 (len > 60 ? "..." : "")
1644 /* Simplest case: anchored match need be tried only once. */
1645 /* [unless only anchor is BOL and multiline is set] */
1646 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1647 if (s == startpos && regtry(prog, startpos))
1649 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1650 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1655 dontbother = minlen - 1;
1656 end = HOP3c(strend, -dontbother, strbeg) - 1;
1657 /* for multiline we only have to try after newlines */
1658 if (prog->check_substr) {
1662 if (regtry(prog, s))
1667 if (prog->reganch & RE_USE_INTUIT) {
1668 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1679 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1680 if (regtry(prog, s))
1687 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1688 if (regtry(prog, PL_reg_ganch))
1693 /* Messy cases: unanchored match. */
1694 if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
1695 /* we have /x+whatever/ */
1696 /* it must be a one character string (XXXX Except UTF?) */
1697 char ch = SvPVX(prog->anchored_substr)[0];
1703 while (s < strend) {
1705 DEBUG_r( did_match = 1 );
1706 if (regtry(prog, s)) goto got_it;
1708 while (s < strend && *s == ch)
1715 while (s < strend) {
1717 DEBUG_r( did_match = 1 );
1718 if (regtry(prog, s)) goto got_it;
1720 while (s < strend && *s == ch)
1726 DEBUG_r(if (!did_match)
1727 PerlIO_printf(Perl_debug_log,
1728 "Did not find anchored character...\n")
1732 else if (do_utf8 == (UTF!=0) &&
1733 (prog->anchored_substr != Nullsv
1734 || (prog->float_substr != Nullsv
1735 && prog->float_max_offset < strend - s))) {
1736 SV *must = prog->anchored_substr
1737 ? prog->anchored_substr : prog->float_substr;
1739 prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset;
1741 prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
1742 char *last = HOP3c(strend, /* Cannot start after this */
1743 -(I32)(CHR_SVLEN(must)
1744 - (SvTAIL(must) != 0) + back_min), strbeg);
1745 char *last1; /* Last position checked before */
1751 last1 = HOPc(s, -1);
1753 last1 = s - 1; /* bogus */
1755 /* XXXX check_substr already used to find `s', can optimize if
1756 check_substr==must. */
1758 dontbother = end_shift;
1759 strend = HOPc(strend, -dontbother);
1760 while ( (s <= last) &&
1761 ((flags & REXEC_SCREAM)
1762 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1763 end_shift, &scream_pos, 0))
1764 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1765 (unsigned char*)strend, must,
1766 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1767 DEBUG_r( did_match = 1 );
1768 if (HOPc(s, -back_max) > last1) {
1769 last1 = HOPc(s, -back_min);
1770 s = HOPc(s, -back_max);
1773 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1775 last1 = HOPc(s, -back_min);
1779 while (s <= last1) {
1780 if (regtry(prog, s))
1786 while (s <= last1) {
1787 if (regtry(prog, s))
1793 DEBUG_r(if (!did_match)
1794 PerlIO_printf(Perl_debug_log,
1795 "Did not find %s substr `%s%.*s%s'%s...\n",
1796 ((must == prog->anchored_substr)
1797 ? "anchored" : "floating"),
1799 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1801 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1805 else if ((c = prog->regstclass)) {
1806 if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT)
1807 /* don't bother with what can't match */
1808 strend = HOPc(strend, -(minlen - 1));
1810 SV *prop = sv_newmortal();
1812 PerlIO_printf(Perl_debug_log, "Matching stclass `%s' against `%s'\n", SvPVX(prop), UTF ? sv_uni_display(dsv, sv, 60, 0) : s);
1814 if (find_byclass(prog, c, s, strend, startpos, 0))
1816 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1820 if (prog->float_substr != Nullsv) { /* Trim the end. */
1823 if (flags & REXEC_SCREAM) {
1824 last = screaminstr(sv, prog->float_substr, s - strbeg,
1825 end_shift, &scream_pos, 1); /* last one */
1827 last = scream_olds; /* Only one occurrence. */
1831 char *little = SvPV(prog->float_substr, len);
1833 if (SvTAIL(prog->float_substr)) {
1834 if (memEQ(strend - len + 1, little, len - 1))
1835 last = strend - len + 1;
1836 else if (!PL_multiline)
1837 last = memEQ(strend - len, little, len)
1838 ? strend - len : Nullch;
1844 last = rninstr(s, strend, little, little + len);
1846 last = strend; /* matching `$' */
1850 DEBUG_r(PerlIO_printf(Perl_debug_log,
1851 "%sCan't trim the tail, match fails (should not happen)%s\n",
1852 PL_colors[4],PL_colors[5]));
1853 goto phooey; /* Should not happen! */
1855 dontbother = strend - last + prog->float_min_offset;
1857 if (minlen && (dontbother < minlen))
1858 dontbother = minlen - 1;
1859 strend -= dontbother; /* this one's always in bytes! */
1860 /* We don't know much -- general case. */
1863 if (regtry(prog, s))
1872 if (regtry(prog, s))
1874 } while (s++ < strend);
1882 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1884 if (PL_reg_eval_set) {
1885 /* Preserve the current value of $^R */
1886 if (oreplsv != GvSV(PL_replgv))
1887 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1888 restored, the value remains
1890 restore_pos(aTHX_ 0);
1893 /* make sure $`, $&, $', and $digit will work later */
1894 if ( !(flags & REXEC_NOT_FIRST) ) {
1895 if (RX_MATCH_COPIED(prog)) {
1896 Safefree(prog->subbeg);
1897 RX_MATCH_COPIED_off(prog);
1899 if (flags & REXEC_COPY_STR) {
1900 I32 i = PL_regeol - startpos + (stringarg - strbeg);
1902 s = savepvn(strbeg, i);
1905 RX_MATCH_COPIED_on(prog);
1908 prog->subbeg = strbeg;
1909 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
1916 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
1917 PL_colors[4],PL_colors[5]));
1918 if (PL_reg_eval_set)
1919 restore_pos(aTHX_ 0);
1924 - regtry - try match at specific point
1926 STATIC I32 /* 0 failure, 1 success */
1927 S_regtry(pTHX_ regexp *prog, char *startpos)
1935 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
1937 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
1940 PL_reg_eval_set = RS_init;
1942 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
1943 (IV)(PL_stack_sp - PL_stack_base));
1945 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
1946 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
1947 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
1949 /* Apparently this is not needed, judging by wantarray. */
1950 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
1951 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
1954 /* Make $_ available to executed code. */
1955 if (PL_reg_sv != DEFSV) {
1956 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
1961 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
1962 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
1963 /* prepare for quick setting of pos */
1964 sv_magic(PL_reg_sv, (SV*)0,
1965 PERL_MAGIC_regex_global, Nullch, 0);
1966 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
1970 PL_reg_oldpos = mg->mg_len;
1971 SAVEDESTRUCTOR_X(restore_pos, 0);
1973 if (!PL_reg_curpm) {
1974 Newz(22,PL_reg_curpm, 1, PMOP);
1977 SV* repointer = newSViv(0);
1978 /* so we know which PL_regex_padav element is PL_reg_curpm */
1979 SvFLAGS(repointer) |= SVf_BREAK;
1980 av_push(PL_regex_padav,repointer);
1981 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
1982 PL_regex_pad = AvARRAY(PL_regex_padav);
1986 PM_SETRE(PL_reg_curpm, prog);
1987 PL_reg_oldcurpm = PL_curpm;
1988 PL_curpm = PL_reg_curpm;
1989 if (RX_MATCH_COPIED(prog)) {
1990 /* Here is a serious problem: we cannot rewrite subbeg,
1991 since it may be needed if this match fails. Thus
1992 $` inside (?{}) could fail... */
1993 PL_reg_oldsaved = prog->subbeg;
1994 PL_reg_oldsavedlen = prog->sublen;
1995 RX_MATCH_COPIED_off(prog);
1998 PL_reg_oldsaved = Nullch;
1999 prog->subbeg = PL_bostr;
2000 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2002 prog->startp[0] = startpos - PL_bostr;
2003 PL_reginput = startpos;
2004 PL_regstartp = prog->startp;
2005 PL_regendp = prog->endp;
2006 PL_reglastparen = &prog->lastparen;
2007 PL_reglastcloseparen = &prog->lastcloseparen;
2008 prog->lastparen = 0;
2010 DEBUG_r(PL_reg_starttry = startpos);
2011 if (PL_reg_start_tmpl <= prog->nparens) {
2012 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2013 if(PL_reg_start_tmp)
2014 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2016 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2020 sv_setpvn(PERL_DEBUG_PAD(0), "", 0);
2021 sv_setpvn(PERL_DEBUG_PAD(1), "", 0);
2022 sv_setpvn(PERL_DEBUG_PAD(2), "", 0);
2025 /* XXXX What this code is doing here?!!! There should be no need
2026 to do this again and again, PL_reglastparen should take care of
2029 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2030 * Actually, the code in regcppop() (which Ilya may be meaning by
2031 * PL_reglastparen), is not needed at all by the test suite
2032 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2033 * enough, for building DynaLoader, or otherwise this
2034 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2035 * will happen. Meanwhile, this code *is* needed for the
2036 * above-mentioned test suite tests to succeed. The common theme
2037 * on those tests seems to be returning null fields from matches.
2042 if (prog->nparens) {
2043 for (i = prog->nparens; i > *PL_reglastparen; i--) {
2050 if (regmatch(prog->program + 1)) {
2051 prog->endp[0] = PL_reginput - PL_bostr;
2054 REGCP_UNWIND(lastcp);
2058 #define RE_UNWIND_BRANCH 1
2059 #define RE_UNWIND_BRANCHJ 2
2063 typedef struct { /* XX: makes sense to enlarge it... */
2067 } re_unwind_generic_t;
2080 } re_unwind_branch_t;
2082 typedef union re_unwind_t {
2084 re_unwind_generic_t generic;
2085 re_unwind_branch_t branch;
2088 #define sayYES goto yes
2089 #define sayNO goto no
2090 #define sayYES_FINAL goto yes_final
2091 #define sayYES_LOUD goto yes_loud
2092 #define sayNO_FINAL goto no_final
2093 #define sayNO_SILENT goto do_no
2094 #define saySAME(x) if (x) goto yes; else goto no
2096 #define REPORT_CODE_OFF 24
2099 - regmatch - main matching routine
2101 * Conceptually the strategy is simple: check to see whether the current
2102 * node matches, call self recursively to see whether the rest matches,
2103 * and then act accordingly. In practice we make some effort to avoid
2104 * recursion, in particular by going through "ordinary" nodes (that don't
2105 * need to know whether the rest of the match failed) by a loop instead of
2108 /* [lwall] I've hoisted the register declarations to the outer block in order to
2109 * maybe save a little bit of pushing and popping on the stack. It also takes
2110 * advantage of machines that use a register save mask on subroutine entry.
2112 STATIC I32 /* 0 failure, 1 success */
2113 S_regmatch(pTHX_ regnode *prog)
2115 register regnode *scan; /* Current node. */
2116 regnode *next; /* Next node. */
2117 regnode *inner; /* Next node in internal branch. */
2118 register I32 nextchr; /* renamed nextchr - nextchar colides with
2119 function of same name */
2120 register I32 n; /* no or next */
2121 register I32 ln = 0; /* len or last */
2122 register char *s = Nullch; /* operand or save */
2123 register char *locinput = PL_reginput;
2124 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2125 int minmod = 0, sw = 0, logical = 0;
2128 I32 firstcp = PL_savestack_ix;
2130 register bool do_utf8 = PL_reg_match_utf8;
2132 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2133 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2134 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2141 /* Note that nextchr is a byte even in UTF */
2142 nextchr = UCHARAT(locinput);
2144 while (scan != NULL) {
2147 SV *prop = sv_newmortal();
2148 int docolor = *PL_colors[0];
2149 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2150 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2151 /* The part of the string before starttry has one color
2152 (pref0_len chars), between starttry and current
2153 position another one (pref_len - pref0_len chars),
2154 after the current position the third one.
2155 We assume that pref0_len <= pref_len, otherwise we
2156 decrease pref0_len. */
2157 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2158 ? (5 + taill) - l : locinput - PL_bostr;
2161 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2163 pref0_len = pref_len - (locinput - PL_reg_starttry);
2164 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2165 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2166 ? (5 + taill) - pref_len : PL_regeol - locinput);
2167 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2171 if (pref0_len > pref_len)
2172 pref0_len = pref_len;
2173 regprop(prop, scan);
2177 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2179 locinput - pref_len;
2180 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2181 char *s1 = do_utf8 ?
2182 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2183 pref_len - pref0_len, 60, 0) :
2184 locinput - pref_len + pref0_len;
2185 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2186 char *s2 = do_utf8 ?
2187 pv_uni_display(dsv2, (U8*)locinput,
2188 PL_regeol - locinput, 60, 0) :
2190 int len2 = do_utf8 ? strlen(s2) : l;
2191 PerlIO_printf(Perl_debug_log,
2192 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2193 (IV)(locinput - PL_bostr),
2200 (docolor ? "" : "> <"),
2204 15 - l - pref_len + 1,
2206 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2211 next = scan + NEXT_OFF(scan);
2217 if (locinput == PL_bostr || (PL_multiline &&
2218 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2220 /* regtill = regbol; */
2225 if (locinput == PL_bostr ||
2226 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2232 if (locinput == PL_bostr)
2236 if (locinput == PL_reg_ganch)
2246 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2251 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2253 if (PL_regeol - locinput > 1)
2257 if (PL_regeol != locinput)
2261 if (!nextchr && locinput >= PL_regeol)
2264 locinput += PL_utf8skip[nextchr];
2265 if (locinput > PL_regeol)
2267 nextchr = UCHARAT(locinput);
2270 nextchr = UCHARAT(++locinput);
2273 if (!nextchr && locinput >= PL_regeol)
2275 nextchr = UCHARAT(++locinput);
2278 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2281 locinput += PL_utf8skip[nextchr];
2282 if (locinput > PL_regeol)
2284 nextchr = UCHARAT(locinput);
2287 nextchr = UCHARAT(++locinput);
2292 if (do_utf8 != (UTF!=0)) {
2293 /* The target and the pattern have differing utf8ness. */
2299 /* The target is utf8, the pattern is not utf8. */
2303 if (NATIVE_TO_UNI(*(U8*)s) !=
2304 utf8_to_uvchr((U8*)l, &ulen))
2311 /* The target is not utf8, the pattern is utf8. */
2315 if (NATIVE_TO_UNI(*((U8*)l)) !=
2316 utf8_to_uvchr((U8*)s, &ulen))
2323 nextchr = UCHARAT(locinput);
2326 /* The target and the pattern have the same utf8ness. */
2327 /* Inline the first character, for speed. */
2328 if (UCHARAT(s) != nextchr)
2330 if (PL_regeol - locinput < ln)
2332 if (ln > 1 && memNE(s, locinput, ln))
2335 nextchr = UCHARAT(locinput);
2338 PL_reg_flags |= RF_tainted;
2344 if (do_utf8 || UTF) {
2345 /* Either target or the pattern are utf8. */
2347 char *e = PL_regeol;
2349 if (ibcmp_utf8(s, 0, ln, do_utf8,
2353 nextchr = UCHARAT(locinput);
2357 /* Neither the target and the pattern are utf8. */
2359 /* Inline the first character, for speed. */
2360 if (UCHARAT(s) != nextchr &&
2361 UCHARAT(s) != ((OP(scan) == EXACTF)
2362 ? PL_fold : PL_fold_locale)[nextchr])
2364 if (PL_regeol - locinput < ln)
2366 if (ln > 1 && (OP(scan) == EXACTF
2367 ? ibcmp(s, locinput, ln)
2368 : ibcmp_locale(s, locinput, ln)))
2371 nextchr = UCHARAT(locinput);
2375 if (!reginclass(scan, (U8*)locinput, do_utf8))
2377 if (locinput >= PL_regeol)
2379 locinput += PL_utf8skip[nextchr];
2380 nextchr = UCHARAT(locinput);
2384 nextchr = UCHARAT(locinput);
2385 if (!reginclass(scan, (U8*)locinput, do_utf8))
2387 if (!nextchr && locinput >= PL_regeol)
2389 nextchr = UCHARAT(++locinput);
2393 PL_reg_flags |= RF_tainted;
2399 LOAD_UTF8_CHARCLASS(alnum,"a");
2400 if (!(OP(scan) == ALNUM
2401 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2402 : isALNUM_LC_utf8((U8*)locinput)))
2406 locinput += PL_utf8skip[nextchr];
2407 nextchr = UCHARAT(locinput);
2410 if (!(OP(scan) == ALNUM
2411 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2413 nextchr = UCHARAT(++locinput);
2416 PL_reg_flags |= RF_tainted;
2419 if (!nextchr && locinput >= PL_regeol)
2422 LOAD_UTF8_CHARCLASS(alnum,"a");
2423 if (OP(scan) == NALNUM
2424 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2425 : isALNUM_LC_utf8((U8*)locinput))
2429 locinput += PL_utf8skip[nextchr];
2430 nextchr = UCHARAT(locinput);
2433 if (OP(scan) == NALNUM
2434 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2436 nextchr = UCHARAT(++locinput);
2440 PL_reg_flags |= RF_tainted;
2444 /* was last char in word? */
2446 if (locinput == PL_bostr)
2449 U8 *r = reghop((U8*)locinput, -1);
2451 ln = utf8n_to_uvchr(r, s - (char*)r, 0, 0);
2453 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2454 ln = isALNUM_uni(ln);
2455 LOAD_UTF8_CHARCLASS(alnum,"a");
2456 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2459 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2460 n = isALNUM_LC_utf8((U8*)locinput);
2464 ln = (locinput != PL_bostr) ?
2465 UCHARAT(locinput - 1) : '\n';
2466 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2468 n = isALNUM(nextchr);
2471 ln = isALNUM_LC(ln);
2472 n = isALNUM_LC(nextchr);
2475 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2476 OP(scan) == BOUNDL))
2480 PL_reg_flags |= RF_tainted;
2486 if (UTF8_IS_CONTINUED(nextchr)) {
2487 LOAD_UTF8_CHARCLASS(space," ");
2488 if (!(OP(scan) == SPACE
2489 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2490 : isSPACE_LC_utf8((U8*)locinput)))
2494 locinput += PL_utf8skip[nextchr];
2495 nextchr = UCHARAT(locinput);
2498 if (!(OP(scan) == SPACE
2499 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2501 nextchr = UCHARAT(++locinput);
2504 if (!(OP(scan) == SPACE
2505 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2507 nextchr = UCHARAT(++locinput);
2511 PL_reg_flags |= RF_tainted;
2514 if (!nextchr && locinput >= PL_regeol)
2517 LOAD_UTF8_CHARCLASS(space," ");
2518 if (OP(scan) == NSPACE
2519 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2520 : isSPACE_LC_utf8((U8*)locinput))
2524 locinput += PL_utf8skip[nextchr];
2525 nextchr = UCHARAT(locinput);
2528 if (OP(scan) == NSPACE
2529 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2531 nextchr = UCHARAT(++locinput);
2534 PL_reg_flags |= RF_tainted;
2540 LOAD_UTF8_CHARCLASS(digit,"0");
2541 if (!(OP(scan) == DIGIT
2542 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2543 : isDIGIT_LC_utf8((U8*)locinput)))
2547 locinput += PL_utf8skip[nextchr];
2548 nextchr = UCHARAT(locinput);
2551 if (!(OP(scan) == DIGIT
2552 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2554 nextchr = UCHARAT(++locinput);
2557 PL_reg_flags |= RF_tainted;
2560 if (!nextchr && locinput >= PL_regeol)
2563 LOAD_UTF8_CHARCLASS(digit,"0");
2564 if (OP(scan) == NDIGIT
2565 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2566 : isDIGIT_LC_utf8((U8*)locinput))
2570 locinput += PL_utf8skip[nextchr];
2571 nextchr = UCHARAT(locinput);
2574 if (OP(scan) == NDIGIT
2575 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2577 nextchr = UCHARAT(++locinput);
2580 if (locinput >= PL_regeol)
2583 LOAD_UTF8_CHARCLASS(mark,"~");
2584 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2586 locinput += PL_utf8skip[nextchr];
2587 while (locinput < PL_regeol &&
2588 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2589 locinput += UTF8SKIP(locinput);
2590 if (locinput > PL_regeol)
2595 nextchr = UCHARAT(locinput);
2598 PL_reg_flags |= RF_tainted;
2602 n = ARG(scan); /* which paren pair */
2603 ln = PL_regstartp[n];
2604 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2605 if (*PL_reglastparen < n || ln == -1)
2606 sayNO; /* Do not match unless seen CLOSEn. */
2607 if (ln == PL_regendp[n])
2611 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2613 char *e = PL_bostr + PL_regendp[n];
2615 * Note that we can't do the "other character" lookup trick as
2616 * in the 8-bit case (no pun intended) because in Unicode we
2617 * have to map both upper and title case to lower case.
2619 if (OP(scan) == REFF) {
2620 STRLEN ulen1, ulen2;
2621 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2622 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2626 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2627 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2628 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2635 nextchr = UCHARAT(locinput);
2639 /* Inline the first character, for speed. */
2640 if (UCHARAT(s) != nextchr &&
2642 (UCHARAT(s) != ((OP(scan) == REFF
2643 ? PL_fold : PL_fold_locale)[nextchr]))))
2645 ln = PL_regendp[n] - ln;
2646 if (locinput + ln > PL_regeol)
2648 if (ln > 1 && (OP(scan) == REF
2649 ? memNE(s, locinput, ln)
2651 ? ibcmp(s, locinput, ln)
2652 : ibcmp_locale(s, locinput, ln))))
2655 nextchr = UCHARAT(locinput);
2666 OP_4tree *oop = PL_op;
2667 COP *ocurcop = PL_curcop;
2668 SV **ocurpad = PL_curpad;
2672 PL_op = (OP_4tree*)PL_regdata->data[n];
2673 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2674 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2675 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2679 CALLRUNOPS(aTHX); /* Scalar context. */
2682 ret = Nullsv; /* protect against empty (?{}) blocks. */
2690 PL_curpad = ocurpad;
2691 PL_curcop = ocurcop;
2693 if (logical == 2) { /* Postponed subexpression. */
2695 MAGIC *mg = Null(MAGIC*);
2697 CHECKPOINT cp, lastcp;
2699 if(SvROK(ret) || SvRMAGICAL(ret)) {
2700 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2703 mg = mg_find(sv, PERL_MAGIC_qr);
2706 re = (regexp *)mg->mg_obj;
2707 (void)ReREFCNT_inc(re);
2711 char *t = SvPV(ret, len);
2713 char *oprecomp = PL_regprecomp;
2714 I32 osize = PL_regsize;
2715 I32 onpar = PL_regnpar;
2718 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2720 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2721 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2723 PL_regprecomp = oprecomp;
2728 PerlIO_printf(Perl_debug_log,
2729 "Entering embedded `%s%.60s%s%s'\n",
2733 (strlen(re->precomp) > 60 ? "..." : ""))
2736 state.prev = PL_reg_call_cc;
2737 state.cc = PL_regcc;
2738 state.re = PL_reg_re;
2742 cp = regcppush(0); /* Save *all* the positions. */
2745 state.ss = PL_savestack_ix;
2746 *PL_reglastparen = 0;
2747 *PL_reglastcloseparen = 0;
2748 PL_reg_call_cc = &state;
2749 PL_reginput = locinput;
2751 /* XXXX This is too dramatic a measure... */
2754 if (regmatch(re->program + 1)) {
2755 /* Even though we succeeded, we need to restore
2756 global variables, since we may be wrapped inside
2757 SUSPEND, thus the match may be not finished yet. */
2759 /* XXXX Do this only if SUSPENDed? */
2760 PL_reg_call_cc = state.prev;
2761 PL_regcc = state.cc;
2762 PL_reg_re = state.re;
2763 cache_re(PL_reg_re);
2765 /* XXXX This is too dramatic a measure... */
2768 /* These are needed even if not SUSPEND. */
2774 REGCP_UNWIND(lastcp);
2776 PL_reg_call_cc = state.prev;
2777 PL_regcc = state.cc;
2778 PL_reg_re = state.re;
2779 cache_re(PL_reg_re);
2781 /* XXXX This is too dramatic a measure... */
2791 sv_setsv(save_scalar(PL_replgv), ret);
2795 n = ARG(scan); /* which paren pair */
2796 PL_reg_start_tmp[n] = locinput;
2801 n = ARG(scan); /* which paren pair */
2802 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2803 PL_regendp[n] = locinput - PL_bostr;
2804 if (n > *PL_reglastparen)
2805 *PL_reglastparen = n;
2806 *PL_reglastcloseparen = n;
2809 n = ARG(scan); /* which paren pair */
2810 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2813 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2815 next = NEXTOPER(NEXTOPER(scan));
2817 next = scan + ARG(scan);
2818 if (OP(next) == IFTHEN) /* Fake one. */
2819 next = NEXTOPER(NEXTOPER(next));
2823 logical = scan->flags;
2825 /*******************************************************************
2826 PL_regcc contains infoblock about the innermost (...)* loop, and
2827 a pointer to the next outer infoblock.
2829 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2831 1) After matching X, regnode for CURLYX is processed;
2833 2) This regnode creates infoblock on the stack, and calls
2834 regmatch() recursively with the starting point at WHILEM node;
2836 3) Each hit of WHILEM node tries to match A and Z (in the order
2837 depending on the current iteration, min/max of {min,max} and
2838 greediness). The information about where are nodes for "A"
2839 and "Z" is read from the infoblock, as is info on how many times "A"
2840 was already matched, and greediness.
2842 4) After A matches, the same WHILEM node is hit again.
2844 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2845 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2846 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2847 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2848 of the external loop.
2850 Currently present infoblocks form a tree with a stem formed by PL_curcc
2851 and whatever it mentions via ->next, and additional attached trees
2852 corresponding to temporarily unset infoblocks as in "5" above.
2854 In the following picture infoblocks for outer loop of
2855 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2856 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2857 infoblocks are drawn below the "reset" infoblock.
2859 In fact in the picture below we do not show failed matches for Z and T
2860 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2861 more obvious *why* one needs to *temporary* unset infoblocks.]
2863 Matched REx position InfoBlocks Comment
2867 Y A)*?Z)*?T x <- O <- I
2868 YA )*?Z)*?T x <- O <- I
2869 YA A)*?Z)*?T x <- O <- I
2870 YAA )*?Z)*?T x <- O <- I
2871 YAA Z)*?T x <- O # Temporary unset I
2874 YAAZ Y(A)*?Z)*?T x <- O
2877 YAAZY (A)*?Z)*?T x <- O
2880 YAAZY A)*?Z)*?T x <- O <- I
2883 YAAZYA )*?Z)*?T x <- O <- I
2886 YAAZYA Z)*?T x <- O # Temporary unset I
2892 YAAZYAZ T x # Temporary unset O
2899 *******************************************************************/
2902 CHECKPOINT cp = PL_savestack_ix;
2903 /* No need to save/restore up to this paren */
2904 I32 parenfloor = scan->flags;
2906 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2908 cc.oldcc = PL_regcc;
2910 /* XXXX Probably it is better to teach regpush to support
2911 parenfloor > PL_regsize... */
2912 if (parenfloor > *PL_reglastparen)
2913 parenfloor = *PL_reglastparen; /* Pessimization... */
2914 cc.parenfloor = parenfloor;
2916 cc.min = ARG1(scan);
2917 cc.max = ARG2(scan);
2918 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2922 PL_reginput = locinput;
2923 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2925 PL_regcc = cc.oldcc;
2931 * This is really hard to understand, because after we match
2932 * what we're trying to match, we must make sure the rest of
2933 * the REx is going to match for sure, and to do that we have
2934 * to go back UP the parse tree by recursing ever deeper. And
2935 * if it fails, we have to reset our parent's current state
2936 * that we can try again after backing off.
2939 CHECKPOINT cp, lastcp;
2940 CURCUR* cc = PL_regcc;
2941 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2943 n = cc->cur + 1; /* how many we know we matched */
2944 PL_reginput = locinput;
2947 PerlIO_printf(Perl_debug_log,
2948 "%*s %ld out of %ld..%ld cc=%lx\n",
2949 REPORT_CODE_OFF+PL_regindent*2, "",
2950 (long)n, (long)cc->min,
2951 (long)cc->max, (long)cc)
2954 /* If degenerate scan matches "", assume scan done. */
2956 if (locinput == cc->lastloc && n >= cc->min) {
2957 PL_regcc = cc->oldcc;
2961 PerlIO_printf(Perl_debug_log,
2962 "%*s empty match detected, try continuation...\n",
2963 REPORT_CODE_OFF+PL_regindent*2, "")
2965 if (regmatch(cc->next))
2973 /* First just match a string of min scans. */
2977 cc->lastloc = locinput;
2978 if (regmatch(cc->scan))
2981 cc->lastloc = lastloc;
2986 /* Check whether we already were at this position.
2987 Postpone detection until we know the match is not
2988 *that* much linear. */
2989 if (!PL_reg_maxiter) {
2990 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
2991 PL_reg_leftiter = PL_reg_maxiter;
2993 if (PL_reg_leftiter-- == 0) {
2994 I32 size = (PL_reg_maxiter + 7)/8;
2995 if (PL_reg_poscache) {
2996 if (PL_reg_poscache_size < size) {
2997 Renew(PL_reg_poscache, size, char);
2998 PL_reg_poscache_size = size;
3000 Zero(PL_reg_poscache, size, char);
3003 PL_reg_poscache_size = size;
3004 Newz(29, PL_reg_poscache, size, char);
3007 PerlIO_printf(Perl_debug_log,
3008 "%sDetected a super-linear match, switching on caching%s...\n",
3009 PL_colors[4], PL_colors[5])
3012 if (PL_reg_leftiter < 0) {
3013 I32 o = locinput - PL_bostr, b;
3015 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3018 if (PL_reg_poscache[o] & (1<<b)) {
3020 PerlIO_printf(Perl_debug_log,
3021 "%*s already tried at this position...\n",
3022 REPORT_CODE_OFF+PL_regindent*2, "")
3026 PL_reg_poscache[o] |= (1<<b);
3030 /* Prefer next over scan for minimal matching. */
3033 PL_regcc = cc->oldcc;
3036 cp = regcppush(cc->parenfloor);
3038 if (regmatch(cc->next)) {
3040 sayYES; /* All done. */
3042 REGCP_UNWIND(lastcp);
3048 if (n >= cc->max) { /* Maximum greed exceeded? */
3049 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3050 && !(PL_reg_flags & RF_warned)) {
3051 PL_reg_flags |= RF_warned;
3052 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3053 "Complex regular subexpression recursion",
3060 PerlIO_printf(Perl_debug_log,
3061 "%*s trying longer...\n",
3062 REPORT_CODE_OFF+PL_regindent*2, "")
3064 /* Try scanning more and see if it helps. */
3065 PL_reginput = locinput;
3067 cc->lastloc = locinput;
3068 cp = regcppush(cc->parenfloor);
3070 if (regmatch(cc->scan)) {
3074 REGCP_UNWIND(lastcp);
3077 cc->lastloc = lastloc;
3081 /* Prefer scan over next for maximal matching. */
3083 if (n < cc->max) { /* More greed allowed? */
3084 cp = regcppush(cc->parenfloor);
3086 cc->lastloc = locinput;
3088 if (regmatch(cc->scan)) {
3092 REGCP_UNWIND(lastcp);
3093 regcppop(); /* Restore some previous $<digit>s? */
3094 PL_reginput = locinput;
3096 PerlIO_printf(Perl_debug_log,
3097 "%*s failed, try continuation...\n",
3098 REPORT_CODE_OFF+PL_regindent*2, "")
3101 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3102 && !(PL_reg_flags & RF_warned)) {
3103 PL_reg_flags |= RF_warned;
3104 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3105 "Complex regular subexpression recursion",
3109 /* Failed deeper matches of scan, so see if this one works. */
3110 PL_regcc = cc->oldcc;
3113 if (regmatch(cc->next))
3119 cc->lastloc = lastloc;
3124 next = scan + ARG(scan);
3127 inner = NEXTOPER(NEXTOPER(scan));
3130 inner = NEXTOPER(scan);
3134 if (OP(next) != c1) /* No choice. */
3135 next = inner; /* Avoid recursion. */
3137 I32 lastparen = *PL_reglastparen;
3139 re_unwind_branch_t *uw;
3141 /* Put unwinding data on stack */
3142 unwind1 = SSNEWt(1,re_unwind_branch_t);
3143 uw = SSPTRt(unwind1,re_unwind_branch_t);
3146 uw->type = ((c1 == BRANCH)
3148 : RE_UNWIND_BRANCHJ);
3149 uw->lastparen = lastparen;
3151 uw->locinput = locinput;
3152 uw->nextchr = nextchr;
3154 uw->regindent = ++PL_regindent;
3157 REGCP_SET(uw->lastcp);
3159 /* Now go into the first branch */
3172 /* We suppose that the next guy does not need
3173 backtracking: in particular, it is of constant length,
3174 and has no parenths to influence future backrefs. */
3175 ln = ARG1(scan); /* min to match */
3176 n = ARG2(scan); /* max to match */
3177 paren = scan->flags;
3179 if (paren > PL_regsize)
3181 if (paren > *PL_reglastparen)
3182 *PL_reglastparen = paren;
3184 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3186 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3187 PL_reginput = locinput;
3190 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3192 /* if we matched something zero-length we don't need to
3193 backtrack - capturing parens are already defined, so
3194 the caveat in the maximal case doesn't apply
3196 XXXX if ln == 0, we can redo this check first time
3197 through the following loop
3200 n = ln; /* don't backtrack */
3201 locinput = PL_reginput;
3202 if (HAS_TEXT(next) || JUMPABLE(next)) {
3203 regnode *text_node = next;
3205 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3207 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3209 if (PL_regkind[(U8)OP(text_node)] == REF) {
3211 n = ARG(text_node); /* which paren pair */
3212 ln = PL_regstartp[n];
3213 /* assume yes if we haven't seen CLOSEn */
3215 *PL_reglastparen < n ||
3222 c1 = *(PL_bostr + ln);
3224 else { c1 = (U8)*STRING(text_node); }
3225 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3227 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3228 c2 = PL_fold_locale[c1];
3237 /* This may be improved if l == 0. */
3238 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3239 /* If it could work, try it. */
3241 UCHARAT(PL_reginput) == c1 ||
3242 UCHARAT(PL_reginput) == c2)
3246 PL_regstartp[paren] =
3247 HOPc(PL_reginput, -l) - PL_bostr;
3248 PL_regendp[paren] = PL_reginput - PL_bostr;
3251 PL_regendp[paren] = -1;
3255 REGCP_UNWIND(lastcp);
3257 /* Couldn't or didn't -- move forward. */
3258 PL_reginput = locinput;
3259 if (regrepeat_hard(scan, 1, &l)) {
3261 locinput = PL_reginput;
3268 n = regrepeat_hard(scan, n, &l);
3269 /* if we matched something zero-length we don't need to
3270 backtrack, unless the minimum count is zero and we
3271 are capturing the result - in that case the capture
3272 being defined or not may affect later execution
3274 if (n != 0 && l == 0 && !(paren && ln == 0))
3275 ln = n; /* don't backtrack */
3276 locinput = PL_reginput;
3278 PerlIO_printf(Perl_debug_log,
3279 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3280 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3284 if (HAS_TEXT(next) || JUMPABLE(next)) {
3285 regnode *text_node = next;
3287 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3289 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3291 if (PL_regkind[(U8)OP(text_node)] == REF) {
3293 n = ARG(text_node); /* which paren pair */
3294 ln = PL_regstartp[n];
3295 /* assume yes if we haven't seen CLOSEn */
3297 *PL_reglastparen < n ||
3304 c1 = *(PL_bostr + ln);
3306 else { c1 = (U8)*STRING(text_node); }
3308 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3310 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3311 c2 = PL_fold_locale[c1];
3322 /* If it could work, try it. */
3324 UCHARAT(PL_reginput) == c1 ||
3325 UCHARAT(PL_reginput) == c2)
3328 PerlIO_printf(Perl_debug_log,
3329 "%*s trying tail with n=%"IVdf"...\n",
3330 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3334 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3335 PL_regendp[paren] = PL_reginput - PL_bostr;
3338 PL_regendp[paren] = -1;
3342 REGCP_UNWIND(lastcp);
3344 /* Couldn't or didn't -- back up. */
3346 locinput = HOPc(locinput, -l);
3347 PL_reginput = locinput;
3354 paren = scan->flags; /* Which paren to set */
3355 if (paren > PL_regsize)
3357 if (paren > *PL_reglastparen)
3358 *PL_reglastparen = paren;
3359 ln = ARG1(scan); /* min to match */
3360 n = ARG2(scan); /* max to match */
3361 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3365 ln = ARG1(scan); /* min to match */
3366 n = ARG2(scan); /* max to match */
3367 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3372 scan = NEXTOPER(scan);
3378 scan = NEXTOPER(scan);
3382 * Lookahead to avoid useless match attempts
3383 * when we know what character comes next.
3387 * Used to only do .*x and .*?x, but now it allows
3388 * for )'s, ('s and (?{ ... })'s to be in the way
3389 * of the quantifier and the EXACT-like node. -- japhy
3392 if (HAS_TEXT(next) || JUMPABLE(next)) {
3394 regnode *text_node = next;
3396 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3398 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3400 if (PL_regkind[(U8)OP(text_node)] == REF) {
3402 n = ARG(text_node); /* which paren pair */
3403 ln = PL_regstartp[n];
3404 /* assume yes if we haven't seen CLOSEn */
3406 *PL_reglastparen < n ||
3411 goto assume_ok_easy;
3413 s = (U8*)PL_bostr + ln;
3415 else { s = (U8*)STRING(text_node); }
3419 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3421 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3422 c2 = PL_fold_locale[c1];
3425 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3426 STRLEN ulen1, ulen2;
3427 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3428 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3430 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3431 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3433 c1 = utf8_to_uvuni(tmpbuf1, 0);
3434 c2 = utf8_to_uvuni(tmpbuf2, 0);
3437 c2 = c1 = utf8_to_uvchr(s, NULL);
3445 PL_reginput = locinput;
3449 if (ln && regrepeat(scan, ln) < ln)
3451 locinput = PL_reginput;
3454 char *e; /* Should not check after this */
3455 char *old = locinput;
3457 if (n == REG_INFTY) {
3460 while (UTF8_IS_CONTINUATION(*(U8*)e))
3466 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3470 e = locinput + n - ln;
3476 /* Find place 'next' could work */
3479 while (locinput <= e &&
3480 UCHARAT(locinput) != c1)
3483 while (locinput <= e
3484 && UCHARAT(locinput) != c1
3485 && UCHARAT(locinput) != c2)
3488 count = locinput - old;
3495 utf8_to_uvchr((U8*)locinput, &len) != c1;
3500 for (count = 0; locinput <= e; count++) {
3501 UV c = utf8_to_uvchr((U8*)locinput, &len);
3502 if (c == c1 || c == c2)
3510 /* PL_reginput == old now */
3511 if (locinput != old) {
3512 ln = 1; /* Did some */
3513 if (regrepeat(scan, count) < count)
3516 /* PL_reginput == locinput now */
3517 TRYPAREN(paren, ln, locinput);
3518 PL_reginput = locinput; /* Could be reset... */
3519 REGCP_UNWIND(lastcp);
3520 /* Couldn't or didn't -- move forward. */
3523 locinput += UTF8SKIP(locinput);
3529 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3533 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3535 c = UCHARAT(PL_reginput);
3536 /* If it could work, try it. */
3537 if (c == c1 || c == c2)
3539 TRYPAREN(paren, n, PL_reginput);
3540 REGCP_UNWIND(lastcp);
3543 /* If it could work, try it. */
3544 else if (c1 == -1000)
3546 TRYPAREN(paren, n, PL_reginput);
3547 REGCP_UNWIND(lastcp);
3549 /* Couldn't or didn't -- move forward. */
3550 PL_reginput = locinput;
3551 if (regrepeat(scan, 1)) {
3553 locinput = PL_reginput;
3561 n = regrepeat(scan, n);
3562 locinput = PL_reginput;
3563 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3564 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3565 ln = n; /* why back off? */
3566 /* ...because $ and \Z can match before *and* after
3567 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3568 We should back off by one in this case. */
3569 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3578 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3580 c = UCHARAT(PL_reginput);
3582 /* If it could work, try it. */
3583 if (c1 == -1000 || c == c1 || c == c2)
3585 TRYPAREN(paren, n, PL_reginput);
3586 REGCP_UNWIND(lastcp);
3588 /* Couldn't or didn't -- back up. */
3590 PL_reginput = locinput = HOPc(locinput, -1);
3598 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3600 c = UCHARAT(PL_reginput);
3602 /* If it could work, try it. */
3603 if (c1 == -1000 || c == c1 || c == c2)
3605 TRYPAREN(paren, n, PL_reginput);
3606 REGCP_UNWIND(lastcp);
3608 /* Couldn't or didn't -- back up. */
3610 PL_reginput = locinput = HOPc(locinput, -1);
3617 if (PL_reg_call_cc) {
3618 re_cc_state *cur_call_cc = PL_reg_call_cc;
3619 CURCUR *cctmp = PL_regcc;
3620 regexp *re = PL_reg_re;
3621 CHECKPOINT cp, lastcp;
3623 cp = regcppush(0); /* Save *all* the positions. */
3625 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3627 PL_reginput = locinput; /* Make position available to
3629 cache_re(PL_reg_call_cc->re);
3630 PL_regcc = PL_reg_call_cc->cc;
3631 PL_reg_call_cc = PL_reg_call_cc->prev;
3632 if (regmatch(cur_call_cc->node)) {
3633 PL_reg_call_cc = cur_call_cc;
3637 REGCP_UNWIND(lastcp);
3639 PL_reg_call_cc = cur_call_cc;
3645 PerlIO_printf(Perl_debug_log,
3646 "%*s continuation failed...\n",
3647 REPORT_CODE_OFF+PL_regindent*2, "")
3651 if (locinput < PL_regtill) {
3652 DEBUG_r(PerlIO_printf(Perl_debug_log,
3653 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3655 (long)(locinput - PL_reg_starttry),
3656 (long)(PL_regtill - PL_reg_starttry),
3658 sayNO_FINAL; /* Cannot match: too short. */
3660 PL_reginput = locinput; /* put where regtry can find it */
3661 sayYES_FINAL; /* Success! */
3663 PL_reginput = locinput; /* put where regtry can find it */
3664 sayYES_LOUD; /* Success! */
3667 PL_reginput = locinput;
3672 s = HOPBACKc(locinput, scan->flags);
3678 PL_reginput = locinput;
3683 s = HOPBACKc(locinput, scan->flags);
3689 PL_reginput = locinput;
3692 inner = NEXTOPER(NEXTOPER(scan));
3693 if (regmatch(inner) != n) {
3708 if (OP(scan) == SUSPEND) {
3709 locinput = PL_reginput;
3710 nextchr = UCHARAT(locinput);
3715 next = scan + ARG(scan);
3720 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3721 PTR2UV(scan), OP(scan));
3722 Perl_croak(aTHX_ "regexp memory corruption");
3729 * We get here only if there's trouble -- normally "case END" is
3730 * the terminating point.
3732 Perl_croak(aTHX_ "corrupted regexp pointers");
3738 PerlIO_printf(Perl_debug_log,
3739 "%*s %scould match...%s\n",
3740 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3744 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3745 PL_colors[4],PL_colors[5]));
3751 #if 0 /* Breaks $^R */
3759 PerlIO_printf(Perl_debug_log,
3760 "%*s %sfailed...%s\n",
3761 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3767 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3770 case RE_UNWIND_BRANCH:
3771 case RE_UNWIND_BRANCHJ:
3773 re_unwind_branch_t *uwb = &(uw->branch);
3774 I32 lastparen = uwb->lastparen;
3776 REGCP_UNWIND(uwb->lastcp);
3777 for (n = *PL_reglastparen; n > lastparen; n--)
3779 *PL_reglastparen = n;
3780 scan = next = uwb->next;
3782 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3783 ? BRANCH : BRANCHJ) ) { /* Failure */
3790 /* Have more choice yet. Reuse the same uwb. */
3792 if ((n = (uwb->type == RE_UNWIND_BRANCH
3793 ? NEXT_OFF(next) : ARG(next))))
3796 next = NULL; /* XXXX Needn't unwinding in this case... */
3798 next = NEXTOPER(scan);
3799 if (uwb->type == RE_UNWIND_BRANCHJ)
3800 next = NEXTOPER(next);
3801 locinput = uwb->locinput;
3802 nextchr = uwb->nextchr;
3804 PL_regindent = uwb->regindent;
3811 Perl_croak(aTHX_ "regexp unwind memory corruption");
3822 - regrepeat - repeatedly match something simple, report how many
3825 * [This routine now assumes that it will only match on things of length 1.
3826 * That was true before, but now we assume scan - reginput is the count,
3827 * rather than incrementing count on every character. [Er, except utf8.]]
3830 S_regrepeat(pTHX_ regnode *p, I32 max)
3832 register char *scan;
3834 register char *loceol = PL_regeol;
3835 register I32 hardcount = 0;
3836 register bool do_utf8 = PL_reg_match_utf8;
3839 if (max != REG_INFTY && max < loceol - scan)
3840 loceol = scan + max;
3845 while (scan < loceol && hardcount < max && *scan != '\n') {
3846 scan += UTF8SKIP(scan);
3850 while (scan < loceol && *scan != '\n')
3860 case EXACT: /* length of string is 1 */
3862 while (scan < loceol && UCHARAT(scan) == c)
3865 case EXACTF: /* length of string is 1 */
3867 while (scan < loceol &&
3868 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3871 case EXACTFL: /* length of string is 1 */
3872 PL_reg_flags |= RF_tainted;
3874 while (scan < loceol &&
3875 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3881 while (hardcount < max && scan < loceol &&
3882 reginclass(p, (U8*)scan, do_utf8)) {
3883 scan += UTF8SKIP(scan);
3887 while (scan < loceol && reginclass(p, (U8*)scan, do_utf8))
3894 LOAD_UTF8_CHARCLASS(alnum,"a");
3895 while (hardcount < max && scan < loceol &&
3896 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3897 scan += UTF8SKIP(scan);
3901 while (scan < loceol && isALNUM(*scan))
3906 PL_reg_flags |= RF_tainted;
3909 while (hardcount < max && scan < loceol &&
3910 isALNUM_LC_utf8((U8*)scan)) {
3911 scan += UTF8SKIP(scan);
3915 while (scan < loceol && isALNUM_LC(*scan))
3922 LOAD_UTF8_CHARCLASS(alnum,"a");
3923 while (hardcount < max && scan < loceol &&
3924 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3925 scan += UTF8SKIP(scan);
3929 while (scan < loceol && !isALNUM(*scan))
3934 PL_reg_flags |= RF_tainted;
3937 while (hardcount < max && scan < loceol &&
3938 !isALNUM_LC_utf8((U8*)scan)) {
3939 scan += UTF8SKIP(scan);
3943 while (scan < loceol && !isALNUM_LC(*scan))
3950 LOAD_UTF8_CHARCLASS(space," ");
3951 while (hardcount < max && scan < loceol &&
3953 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3954 scan += UTF8SKIP(scan);
3958 while (scan < loceol && isSPACE(*scan))
3963 PL_reg_flags |= RF_tainted;
3966 while (hardcount < max && scan < loceol &&
3967 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3968 scan += UTF8SKIP(scan);
3972 while (scan < loceol && isSPACE_LC(*scan))
3979 LOAD_UTF8_CHARCLASS(space," ");
3980 while (hardcount < max && scan < loceol &&
3982 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3983 scan += UTF8SKIP(scan);
3987 while (scan < loceol && !isSPACE(*scan))
3992 PL_reg_flags |= RF_tainted;
3995 while (hardcount < max && scan < loceol &&
3996 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3997 scan += UTF8SKIP(scan);
4001 while (scan < loceol && !isSPACE_LC(*scan))
4008 LOAD_UTF8_CHARCLASS(digit,"0");
4009 while (hardcount < max && scan < loceol &&
4010 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4011 scan += UTF8SKIP(scan);
4015 while (scan < loceol && isDIGIT(*scan))
4022 LOAD_UTF8_CHARCLASS(digit,"0");
4023 while (hardcount < max && scan < loceol &&
4024 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4025 scan += UTF8SKIP(scan);
4029 while (scan < loceol && !isDIGIT(*scan))
4033 default: /* Called on something of 0 width. */
4034 break; /* So match right here or not at all. */
4040 c = scan - PL_reginput;
4045 SV *prop = sv_newmortal();
4048 PerlIO_printf(Perl_debug_log,
4049 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4050 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4057 - regrepeat_hard - repeatedly match something, report total lenth and length
4059 * The repeater is supposed to have constant length.
4063 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4065 register char *scan = Nullch;
4066 register char *start;
4067 register char *loceol = PL_regeol;
4069 I32 count = 0, res = 1;
4074 start = PL_reginput;
4075 if (PL_reg_match_utf8) {
4076 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4079 while (start < PL_reginput) {
4081 start += UTF8SKIP(start);
4092 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4094 *lp = l = PL_reginput - start;
4095 if (max != REG_INFTY && l*max < loceol - scan)
4096 loceol = scan + l*max;
4109 - regclass_swash - prepare the utf8 swash
4113 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** initsvp)
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);
4131 else if (si && doinit) {
4132 sw = swash_init("utf8", "", si, 1, 0);
4133 (void)av_store(av, 1, sw);
4145 - reginclass - determine if a character falls into a character class
4149 S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8)
4151 char flags = ANYOF_FLAGS(n);
4156 c = do_utf8 ? utf8_to_uvchr(p, &len) : *p;
4158 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4159 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4160 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4163 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4166 SV *sw = regclass_swash(n, TRUE, 0);
4169 if (swash_fetch(sw, p, do_utf8))
4171 else if (flags & ANYOF_FOLD) {
4172 U8 foldbuf[UTF8_MAXLEN_FOLD+1];
4175 to_utf8_fold(p, foldbuf, &foldlen);
4176 if (swash_fetch(sw, foldbuf, do_utf8))
4178 to_utf8_upper(p, foldbuf, &foldlen);
4179 if (swash_fetch(sw, foldbuf, do_utf8))
4185 if (!match && c < 256) {
4186 if (ANYOF_BITMAP_TEST(n, c))
4188 else if (flags & ANYOF_FOLD) {
4191 if (flags & ANYOF_LOCALE) {
4192 PL_reg_flags |= RF_tainted;
4193 f = PL_fold_locale[c];
4197 if (f != c && ANYOF_BITMAP_TEST(n, f))
4201 if (!match && (flags & ANYOF_CLASS)) {
4202 PL_reg_flags |= RF_tainted;
4204 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4205 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4206 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4207 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4208 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4209 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4210 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4211 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4212 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4213 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4214 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4215 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4216 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4217 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4218 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4219 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4220 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4221 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4222 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4223 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4224 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4225 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4226 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4227 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4228 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4229 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4230 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4231 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4232 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4233 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4234 ) /* How's that for a conditional? */
4241 return (flags & ANYOF_INVERT) ? !match : match;
4245 S_reghop(pTHX_ U8 *s, I32 off)
4247 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4251 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4254 while (off-- && s < lim) {
4255 /* XXX could check well-formedness here */
4263 if (UTF8_IS_CONTINUED(*s)) {
4264 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4267 /* XXX could check well-formedness here */
4275 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4277 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4281 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4284 while (off-- && s < lim) {
4285 /* XXX could check well-formedness here */
4295 if (UTF8_IS_CONTINUED(*s)) {
4296 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4299 /* XXX could check well-formedness here */
4311 restore_pos(pTHX_ void *arg)
4313 if (PL_reg_eval_set) {
4314 if (PL_reg_oldsaved) {
4315 PL_reg_re->subbeg = PL_reg_oldsaved;
4316 PL_reg_re->sublen = PL_reg_oldsavedlen;
4317 RX_MATCH_COPIED_on(PL_reg_re);
4319 PL_reg_magic->mg_len = PL_reg_oldpos;
4320 PL_reg_eval_set = 0;
4321 PL_curpm = PL_reg_oldcurpm;