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);
397 char*s = UTF ? sv_uni_display(dsv, sv, 60, 0) : strpos;
398 int len = UTF ? strlen(s) : strend - strpos;
401 PerlIO_printf(Perl_debug_log,
402 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
403 PL_colors[4],PL_colors[5],PL_colors[0],
406 (strlen(prog->precomp) > 60 ? "..." : ""),
408 (int)(len > 60 ? 60 : len),
410 (len > 60 ? "..." : "")
414 if (prog->reganch & ROPT_UTF8)
415 PL_reg_flags |= RF_utf8;
417 if (prog->minlen > CHR_DIST((U8*)strend, (U8*)strpos)) {
418 DEBUG_r(PerlIO_printf(Perl_debug_log,
419 "String too short... [re_intuit_start]\n"));
422 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
424 check = prog->check_substr;
425 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
426 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
427 || ( (prog->reganch & ROPT_ANCH_BOL)
428 && !PL_multiline ) ); /* Check after \n? */
431 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
432 | ROPT_IMPLICIT)) /* not a real BOL */
433 /* SvCUR is not set on references: SvRV and SvPVX overlap */
435 && (strpos != strbeg)) {
436 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
439 if (prog->check_offset_min == prog->check_offset_max &&
440 !(prog->reganch & ROPT_CANY_SEEN)) {
441 /* Substring at constant offset from beg-of-str... */
444 s = HOP3c(strpos, prog->check_offset_min, strend);
446 slen = SvCUR(check); /* >= 1 */
448 if ( strend - s > slen || strend - s < slen - 1
449 || (strend - s == slen && strend[-1] != '\n')) {
450 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
453 /* Now should match s[0..slen-2] */
455 if (slen && (*SvPVX(check) != *s
457 && memNE(SvPVX(check), s, slen)))) {
459 DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
463 else if (*SvPVX(check) != *s
464 || ((slen = SvCUR(check)) > 1
465 && memNE(SvPVX(check), s, slen)))
467 goto success_at_start;
470 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
472 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
473 end_shift = prog->minlen - start_shift -
474 CHR_SVLEN(check) + (SvTAIL(check) != 0);
476 I32 end = prog->check_offset_max + CHR_SVLEN(check)
477 - (SvTAIL(check) != 0);
478 I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
480 if (end_shift < eshift)
484 else { /* Can match at random position */
487 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
488 /* Should be nonnegative! */
489 end_shift = prog->minlen - start_shift -
490 CHR_SVLEN(check) + (SvTAIL(check) != 0);
493 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
495 Perl_croak(aTHX_ "panic: end_shift");
499 /* Find a possible match in the region s..strend by looking for
500 the "check" substring in the region corrected by start/end_shift. */
501 if (flags & REXEC_SCREAM) {
502 I32 p = -1; /* Internal iterator of scream. */
503 I32 *pp = data ? data->scream_pos : &p;
505 if (PL_screamfirst[BmRARE(check)] >= 0
506 || ( BmRARE(check) == '\n'
507 && (BmPREVIOUS(check) == SvCUR(check) - 1)
509 s = screaminstr(sv, check,
510 start_shift + (s - strbeg), end_shift, pp, 0);
514 *data->scream_olds = s;
516 else if (prog->reganch & ROPT_CANY_SEEN)
517 s = fbm_instr((U8*)(s + start_shift),
518 (U8*)(strend - end_shift),
519 check, PL_multiline ? FBMrf_MULTILINE : 0);
521 s = fbm_instr(HOP3(s, start_shift, strend),
522 HOP3(strend, -end_shift, strbeg),
523 check, PL_multiline ? FBMrf_MULTILINE : 0);
525 /* Update the count-of-usability, remove useless subpatterns,
528 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
529 (s ? "Found" : "Did not find"),
530 ((check == prog->anchored_substr) ? "anchored" : "floating"),
532 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
534 PL_colors[1], (SvTAIL(check) ? "$" : ""),
535 (s ? " at offset " : "...\n") ) );
542 /* Finish the diagnostic message */
543 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
545 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
546 Start with the other substr.
547 XXXX no SCREAM optimization yet - and a very coarse implementation
548 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
549 *always* match. Probably should be marked during compile...
550 Probably it is right to do no SCREAM here...
553 if (prog->float_substr && prog->anchored_substr) {
554 /* Take into account the "other" substring. */
555 /* XXXX May be hopelessly wrong for UTF... */
558 if (check == prog->float_substr) {
561 char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
564 t = s - prog->check_offset_max;
565 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
566 && (!(prog->reganch & ROPT_UTF8)
567 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
572 t = HOP3c(t, prog->anchored_offset, strend);
573 if (t < other_last) /* These positions already checked */
575 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
578 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
579 /* On end-of-str: see comment below. */
580 s = fbm_instr((unsigned char*)t,
581 HOP3(HOP3(last1, prog->anchored_offset, strend)
582 + SvCUR(prog->anchored_substr),
583 -(SvTAIL(prog->anchored_substr)!=0), strbeg),
584 prog->anchored_substr,
585 PL_multiline ? FBMrf_MULTILINE : 0);
586 DEBUG_r(PerlIO_printf(Perl_debug_log,
587 "%s anchored substr `%s%.*s%s'%s",
588 (s ? "Found" : "Contradicts"),
590 (int)(SvCUR(prog->anchored_substr)
591 - (SvTAIL(prog->anchored_substr)!=0)),
592 SvPVX(prog->anchored_substr),
593 PL_colors[1], (SvTAIL(prog->anchored_substr) ? "$" : "")));
595 if (last1 >= last2) {
596 DEBUG_r(PerlIO_printf(Perl_debug_log,
597 ", giving up...\n"));
600 DEBUG_r(PerlIO_printf(Perl_debug_log,
601 ", trying floating at offset %ld...\n",
602 (long)(HOP3c(s1, 1, strend) - i_strpos)));
603 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
604 s = HOP3c(last, 1, strend);
608 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
609 (long)(s - i_strpos)));
610 t = HOP3c(s, -prog->anchored_offset, strbeg);
611 other_last = HOP3c(s, 1, strend);
619 else { /* Take into account the floating substring. */
623 t = HOP3c(s, -start_shift, strbeg);
625 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
626 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
627 last = HOP3c(t, prog->float_max_offset, strend);
628 s = HOP3c(t, prog->float_min_offset, strend);
631 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
632 /* fbm_instr() takes into account exact value of end-of-str
633 if the check is SvTAIL(ed). Since false positives are OK,
634 and end-of-str is not later than strend we are OK. */
635 s = fbm_instr((unsigned char*)s,
636 (unsigned char*)last + SvCUR(prog->float_substr)
637 - (SvTAIL(prog->float_substr)!=0),
638 prog->float_substr, PL_multiline ? FBMrf_MULTILINE : 0);
639 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
640 (s ? "Found" : "Contradicts"),
642 (int)(SvCUR(prog->float_substr)
643 - (SvTAIL(prog->float_substr)!=0)),
644 SvPVX(prog->float_substr),
645 PL_colors[1], (SvTAIL(prog->float_substr) ? "$" : "")));
648 DEBUG_r(PerlIO_printf(Perl_debug_log,
649 ", giving up...\n"));
652 DEBUG_r(PerlIO_printf(Perl_debug_log,
653 ", trying anchored starting at offset %ld...\n",
654 (long)(s1 + 1 - i_strpos)));
656 s = HOP3c(t, 1, strend);
660 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
661 (long)(s - i_strpos)));
662 other_last = s; /* Fix this later. --Hugo */
671 t = s - prog->check_offset_max;
672 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
673 && (!(prog->reganch & ROPT_UTF8)
674 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
676 /* Fixed substring is found far enough so that the match
677 cannot start at strpos. */
679 if (ml_anch && t[-1] != '\n') {
680 /* Eventually fbm_*() should handle this, but often
681 anchored_offset is not 0, so this check will not be wasted. */
682 /* XXXX In the code below we prefer to look for "^" even in
683 presence of anchored substrings. And we search even
684 beyond the found float position. These pessimizations
685 are historical artefacts only. */
687 while (t < strend - prog->minlen) {
689 if (t < check_at - prog->check_offset_min) {
690 if (prog->anchored_substr) {
691 /* Since we moved from the found position,
692 we definitely contradict the found anchored
693 substr. Due to the above check we do not
694 contradict "check" substr.
695 Thus we can arrive here only if check substr
696 is float. Redo checking for "other"=="fixed".
699 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
700 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
701 goto do_other_anchored;
703 /* We don't contradict the found floating substring. */
704 /* XXXX Why not check for STCLASS? */
706 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
707 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
710 /* Position contradicts check-string */
711 /* XXXX probably better to look for check-string
712 than for "\n", so one should lower the limit for t? */
713 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
714 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
715 other_last = strpos = s = t + 1;
720 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
721 PL_colors[0],PL_colors[1]));
725 DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
726 PL_colors[0],PL_colors[1]));
730 ++BmUSEFUL(prog->check_substr); /* hooray/5 */
733 /* The found string does not prohibit matching at strpos,
734 - no optimization of calling REx engine can be performed,
735 unless it was an MBOL and we are not after MBOL,
736 or a future STCLASS check will fail this. */
738 /* Even in this situation we may use MBOL flag if strpos is offset
739 wrt the start of the string. */
740 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
741 && (strpos != strbeg) && strpos[-1] != '\n'
742 /* May be due to an implicit anchor of m{.*foo} */
743 && !(prog->reganch & ROPT_IMPLICIT))
748 DEBUG_r( if (ml_anch)
749 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
750 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
753 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
754 && prog->check_substr /* Could be deleted already */
755 && --BmUSEFUL(prog->check_substr) < 0
756 && prog->check_substr == prog->float_substr)
758 /* If flags & SOMETHING - do not do it many times on the same match */
759 DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
760 SvREFCNT_dec(prog->check_substr);
761 prog->check_substr = Nullsv; /* disable */
762 prog->float_substr = Nullsv; /* clear */
763 check = Nullsv; /* abort */
765 /* XXXX This is a remnant of the old implementation. It
766 looks wasteful, since now INTUIT can use many
768 prog->reganch &= ~RE_USE_INTUIT;
775 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
776 if (prog->regstclass) {
777 /* minlen == 0 is possible if regstclass is \b or \B,
778 and the fixed substr is ''$.
779 Since minlen is already taken into account, s+1 is before strend;
780 accidentally, minlen >= 1 guaranties no false positives at s + 1
781 even for \b or \B. But (minlen? 1 : 0) below assumes that
782 regstclass does not come from lookahead... */
783 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
784 This leaves EXACTF only, which is dealt with in find_byclass(). */
785 U8* str = (U8*)STRING(prog->regstclass);
786 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
787 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
789 char *endpos = (prog->anchored_substr || ml_anch)
790 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
791 : (prog->float_substr
792 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
795 char *startpos = strbeg;
798 if (prog->reganch & ROPT_UTF8) {
799 PL_regdata = prog->data;
802 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
807 if (endpos == strend) {
808 DEBUG_r( PerlIO_printf(Perl_debug_log,
809 "Could not match STCLASS...\n") );
812 DEBUG_r( PerlIO_printf(Perl_debug_log,
813 "This position contradicts STCLASS...\n") );
814 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
816 /* Contradict one of substrings */
817 if (prog->anchored_substr) {
818 if (prog->anchored_substr == check) {
819 DEBUG_r( what = "anchored" );
821 s = HOP3c(t, 1, strend);
822 if (s + start_shift + end_shift > strend) {
823 /* XXXX Should be taken into account earlier? */
824 DEBUG_r( PerlIO_printf(Perl_debug_log,
825 "Could not match STCLASS...\n") );
830 DEBUG_r( PerlIO_printf(Perl_debug_log,
831 "Looking for %s substr starting at offset %ld...\n",
832 what, (long)(s + start_shift - i_strpos)) );
835 /* Have both, check_string is floating */
836 if (t + start_shift >= check_at) /* Contradicts floating=check */
837 goto retry_floating_check;
838 /* Recheck anchored substring, but not floating... */
842 DEBUG_r( PerlIO_printf(Perl_debug_log,
843 "Looking for anchored substr starting at offset %ld...\n",
844 (long)(other_last - i_strpos)) );
845 goto do_other_anchored;
847 /* Another way we could have checked stclass at the
848 current position only: */
853 DEBUG_r( PerlIO_printf(Perl_debug_log,
854 "Looking for /%s^%s/m starting at offset %ld...\n",
855 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
858 if (!prog->float_substr) /* Could have been deleted */
860 /* Check is floating subtring. */
861 retry_floating_check:
862 t = check_at - start_shift;
863 DEBUG_r( what = "floating" );
864 goto hop_and_restart;
867 DEBUG_r(PerlIO_printf(Perl_debug_log,
868 "By STCLASS: moving %ld --> %ld\n",
869 (long)(t - i_strpos), (long)(s - i_strpos))
873 DEBUG_r(PerlIO_printf(Perl_debug_log,
874 "Does not contradict STCLASS...\n");
879 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
880 PL_colors[4], (check ? "Guessed" : "Giving up"),
881 PL_colors[5], (long)(s - i_strpos)) );
884 fail_finish: /* Substring not found */
885 if (prog->check_substr) /* could be removed already */
886 BmUSEFUL(prog->check_substr) += 5; /* hooray */
888 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
889 PL_colors[4],PL_colors[5]));
893 /* We know what class REx starts with. Try to find this position... */
895 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
897 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
903 register I32 tmp = 1; /* Scratch variable? */
904 register bool do_utf8 = PL_reg_match_utf8;
906 /* We know what class it must start with. */
910 if (reginclass(c, (U8*)s, do_utf8)) {
911 if (tmp && (norun || regtry(prog, s)))
918 s += do_utf8 ? UTF8SKIP(s) : 1;
923 if (tmp && (norun || regtry(prog, s)))
935 U8 tmpbuf1[UTF8_MAXLEN*2+1];
936 U8 tmpbuf2[UTF8_MAXLEN*2+1];
938 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
939 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
941 c1 = utf8_to_uvuni(tmpbuf1, 0);
942 c2 = utf8_to_uvuni(tmpbuf2, 0);
953 c2 = PL_fold_locale[c1];
958 e = s; /* Due to minlen logic of intuit() */
962 /* The ibcmp_utf8() uses to_uni_fold() which is more
963 * correct folding for Unicode than using lowercase.
964 * However, it doesn't work quite fully since the folding
965 * is a one-to-many mapping and the regex optimizer is
966 * unaware of this, so it may throw out good matches.
967 * Fortunately, not getting this right is allowed
968 * for Unicode Regular Expression Support level 1,
969 * only one-to-one matching is required. --jhi */
972 if ( utf8_to_uvchr((U8*)s, &len) == c1
974 ibcmp_utf8(s, do_utf8, m, UTF, ln)) )
980 UV c = utf8_to_uvchr((U8*)s, &len);
981 if ( (c == c1 || c == c2)
983 ibcmp_utf8(s, do_utf8, m, UTF, ln)) )
992 && (ln == 1 || !(OP(c) == EXACTF
994 : ibcmp_locale(s, m, ln)))
995 && (norun || regtry(prog, s)) )
1001 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1002 && (ln == 1 || !(OP(c) == EXACTF
1004 : ibcmp_locale(s, m, ln)))
1005 && (norun || regtry(prog, s)) )
1012 PL_reg_flags |= RF_tainted;
1019 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1022 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1024 tmp = ((OP(c) == BOUND ?
1025 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1026 LOAD_UTF8_CHARCLASS(alnum,"a");
1027 while (s < strend) {
1028 if (tmp == !(OP(c) == BOUND ?
1029 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1030 isALNUM_LC_utf8((U8*)s)))
1033 if ((norun || regtry(prog, s)))
1040 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1041 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1042 while (s < strend) {
1044 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1046 if ((norun || regtry(prog, s)))
1052 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1056 PL_reg_flags |= RF_tainted;
1063 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1066 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1068 tmp = ((OP(c) == NBOUND ?
1069 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1070 LOAD_UTF8_CHARCLASS(alnum,"a");
1071 while (s < strend) {
1072 if (tmp == !(OP(c) == NBOUND ?
1073 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1074 isALNUM_LC_utf8((U8*)s)))
1076 else if ((norun || regtry(prog, s)))
1082 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1083 tmp = ((OP(c) == NBOUND ?
1084 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1085 while (s < strend) {
1087 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1089 else if ((norun || regtry(prog, s)))
1094 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1099 LOAD_UTF8_CHARCLASS(alnum,"a");
1100 while (s < strend) {
1101 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1102 if (tmp && (norun || regtry(prog, s)))
1113 while (s < strend) {
1115 if (tmp && (norun || regtry(prog, s)))
1127 PL_reg_flags |= RF_tainted;
1129 while (s < strend) {
1130 if (isALNUM_LC_utf8((U8*)s)) {
1131 if (tmp && (norun || regtry(prog, s)))
1142 while (s < strend) {
1143 if (isALNUM_LC(*s)) {
1144 if (tmp && (norun || regtry(prog, s)))
1157 LOAD_UTF8_CHARCLASS(alnum,"a");
1158 while (s < strend) {
1159 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1160 if (tmp && (norun || regtry(prog, s)))
1171 while (s < strend) {
1173 if (tmp && (norun || regtry(prog, s)))
1185 PL_reg_flags |= RF_tainted;
1187 while (s < strend) {
1188 if (!isALNUM_LC_utf8((U8*)s)) {
1189 if (tmp && (norun || regtry(prog, s)))
1200 while (s < strend) {
1201 if (!isALNUM_LC(*s)) {
1202 if (tmp && (norun || regtry(prog, s)))
1215 LOAD_UTF8_CHARCLASS(space," ");
1216 while (s < strend) {
1217 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1218 if (tmp && (norun || regtry(prog, s)))
1229 while (s < strend) {
1231 if (tmp && (norun || regtry(prog, s)))
1243 PL_reg_flags |= RF_tainted;
1245 while (s < strend) {
1246 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1247 if (tmp && (norun || regtry(prog, s)))
1258 while (s < strend) {
1259 if (isSPACE_LC(*s)) {
1260 if (tmp && (norun || regtry(prog, s)))
1273 LOAD_UTF8_CHARCLASS(space," ");
1274 while (s < strend) {
1275 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1276 if (tmp && (norun || regtry(prog, s)))
1287 while (s < strend) {
1289 if (tmp && (norun || regtry(prog, s)))
1301 PL_reg_flags |= RF_tainted;
1303 while (s < strend) {
1304 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1305 if (tmp && (norun || regtry(prog, s)))
1316 while (s < strend) {
1317 if (!isSPACE_LC(*s)) {
1318 if (tmp && (norun || regtry(prog, s)))
1331 LOAD_UTF8_CHARCLASS(digit,"0");
1332 while (s < strend) {
1333 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1334 if (tmp && (norun || regtry(prog, s)))
1345 while (s < strend) {
1347 if (tmp && (norun || regtry(prog, s)))
1359 PL_reg_flags |= RF_tainted;
1361 while (s < strend) {
1362 if (isDIGIT_LC_utf8((U8*)s)) {
1363 if (tmp && (norun || regtry(prog, s)))
1374 while (s < strend) {
1375 if (isDIGIT_LC(*s)) {
1376 if (tmp && (norun || regtry(prog, s)))
1389 LOAD_UTF8_CHARCLASS(digit,"0");
1390 while (s < strend) {
1391 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1392 if (tmp && (norun || regtry(prog, s)))
1403 while (s < strend) {
1405 if (tmp && (norun || regtry(prog, s)))
1417 PL_reg_flags |= RF_tainted;
1419 while (s < strend) {
1420 if (!isDIGIT_LC_utf8((U8*)s)) {
1421 if (tmp && (norun || regtry(prog, s)))
1432 while (s < strend) {
1433 if (!isDIGIT_LC(*s)) {
1434 if (tmp && (norun || regtry(prog, s)))
1446 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1455 - regexec_flags - match a regexp against a string
1458 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1459 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1460 /* strend: pointer to null at end of string */
1461 /* strbeg: real beginning of string */
1462 /* minend: end of match must be >=minend after stringarg. */
1463 /* data: May be used for some additional optimizations. */
1464 /* nosave: For optimizations. */
1467 register regnode *c;
1468 register char *startpos = stringarg;
1469 I32 minlen; /* must match at least this many chars */
1470 I32 dontbother = 0; /* how many characters not to try at end */
1471 /* I32 start_shift = 0; */ /* Offset of the start to find
1472 constant substr. */ /* CC */
1473 I32 end_shift = 0; /* Same for the end. */ /* CC */
1474 I32 scream_pos = -1; /* Internal iterator of scream. */
1476 SV* oreplsv = GvSV(PL_replgv);
1477 bool do_utf8 = DO_UTF8(sv);
1479 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
1486 PL_regnarrate = DEBUG_r_TEST;
1489 /* Be paranoid... */
1490 if (prog == NULL || startpos == NULL) {
1491 Perl_croak(aTHX_ "NULL regexp parameter");
1495 minlen = prog->minlen;
1496 if (strend - startpos < minlen) {
1497 DEBUG_r(PerlIO_printf(Perl_debug_log,
1498 "String too short [regexec_flags]...\n"));
1502 /* Check validity of program. */
1503 if (UCHARAT(prog->program) != REG_MAGIC) {
1504 Perl_croak(aTHX_ "corrupted regexp program");
1508 PL_reg_eval_set = 0;
1511 if (prog->reganch & ROPT_UTF8)
1512 PL_reg_flags |= RF_utf8;
1514 /* Mark beginning of line for ^ and lookbehind. */
1515 PL_regbol = startpos;
1519 /* Mark end of line for $ (and such) */
1522 /* see how far we have to get to not match where we matched before */
1523 PL_regtill = startpos+minend;
1525 /* We start without call_cc context. */
1528 /* If there is a "must appear" string, look for it. */
1531 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1534 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1535 PL_reg_ganch = startpos;
1536 else if (sv && SvTYPE(sv) >= SVt_PVMG
1538 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1539 && mg->mg_len >= 0) {
1540 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1541 if (prog->reganch & ROPT_ANCH_GPOS) {
1542 if (s > PL_reg_ganch)
1547 else /* pos() not defined */
1548 PL_reg_ganch = strbeg;
1551 if (do_utf8 == (UTF!=0) &&
1552 !(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
1553 re_scream_pos_data d;
1555 d.scream_olds = &scream_olds;
1556 d.scream_pos = &scream_pos;
1557 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1559 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1560 goto phooey; /* not present */
1565 char *s = do_utf8 ? sv_uni_display(dsv, sv, 60, 0) : startpos;
1566 int len = do_utf8 ? strlen(s) : strend - startpos;
1569 PerlIO_printf(Perl_debug_log,
1570 "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
1571 PL_colors[4],PL_colors[5],PL_colors[0],
1574 (strlen(prog->precomp) > 60 ? "..." : ""),
1576 (int)(len > 60 ? 60 : len),
1578 (len > 60 ? "..." : "")
1582 /* Simplest case: anchored match need be tried only once. */
1583 /* [unless only anchor is BOL and multiline is set] */
1584 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1585 if (s == startpos && regtry(prog, startpos))
1587 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1588 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1593 dontbother = minlen - 1;
1594 end = HOP3c(strend, -dontbother, strbeg) - 1;
1595 /* for multiline we only have to try after newlines */
1596 if (prog->check_substr) {
1600 if (regtry(prog, s))
1605 if (prog->reganch & RE_USE_INTUIT) {
1606 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1617 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1618 if (regtry(prog, s))
1625 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1626 if (regtry(prog, PL_reg_ganch))
1631 /* Messy cases: unanchored match. */
1632 if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
1633 /* we have /x+whatever/ */
1634 /* it must be a one character string (XXXX Except UTF?) */
1635 char ch = SvPVX(prog->anchored_substr)[0];
1641 while (s < strend) {
1643 DEBUG_r( did_match = 1 );
1644 if (regtry(prog, s)) goto got_it;
1646 while (s < strend && *s == ch)
1653 while (s < strend) {
1655 DEBUG_r( did_match = 1 );
1656 if (regtry(prog, s)) goto got_it;
1658 while (s < strend && *s == ch)
1664 DEBUG_r(if (!did_match)
1665 PerlIO_printf(Perl_debug_log,
1666 "Did not find anchored character...\n")
1670 else if (do_utf8 == (UTF!=0) &&
1671 (prog->anchored_substr != Nullsv
1672 || (prog->float_substr != Nullsv
1673 && prog->float_max_offset < strend - s))) {
1674 SV *must = prog->anchored_substr
1675 ? prog->anchored_substr : prog->float_substr;
1677 prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset;
1679 prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
1680 char *last = HOP3c(strend, /* Cannot start after this */
1681 -(I32)(CHR_SVLEN(must)
1682 - (SvTAIL(must) != 0) + back_min), strbeg);
1683 char *last1; /* Last position checked before */
1689 last1 = HOPc(s, -1);
1691 last1 = s - 1; /* bogus */
1693 /* XXXX check_substr already used to find `s', can optimize if
1694 check_substr==must. */
1696 dontbother = end_shift;
1697 strend = HOPc(strend, -dontbother);
1698 while ( (s <= last) &&
1699 ((flags & REXEC_SCREAM)
1700 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1701 end_shift, &scream_pos, 0))
1702 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1703 (unsigned char*)strend, must,
1704 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1705 DEBUG_r( did_match = 1 );
1706 if (HOPc(s, -back_max) > last1) {
1707 last1 = HOPc(s, -back_min);
1708 s = HOPc(s, -back_max);
1711 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1713 last1 = HOPc(s, -back_min);
1717 while (s <= last1) {
1718 if (regtry(prog, s))
1724 while (s <= last1) {
1725 if (regtry(prog, s))
1731 DEBUG_r(if (!did_match)
1732 PerlIO_printf(Perl_debug_log,
1733 "Did not find %s substr `%s%.*s%s'%s...\n",
1734 ((must == prog->anchored_substr)
1735 ? "anchored" : "floating"),
1737 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1739 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1743 else if ((c = prog->regstclass)) {
1744 if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT)
1745 /* don't bother with what can't match */
1746 strend = HOPc(strend, -(minlen - 1));
1748 SV *prop = sv_newmortal();
1750 PerlIO_printf(Perl_debug_log, "Matching stclass `%s' against `%s'\n", SvPVX(prop), UTF ? sv_uni_display(dsv, sv, 60, 0) : s);
1752 if (find_byclass(prog, c, s, strend, startpos, 0))
1754 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1758 if (prog->float_substr != Nullsv) { /* Trim the end. */
1761 if (flags & REXEC_SCREAM) {
1762 last = screaminstr(sv, prog->float_substr, s - strbeg,
1763 end_shift, &scream_pos, 1); /* last one */
1765 last = scream_olds; /* Only one occurrence. */
1769 char *little = SvPV(prog->float_substr, len);
1771 if (SvTAIL(prog->float_substr)) {
1772 if (memEQ(strend - len + 1, little, len - 1))
1773 last = strend - len + 1;
1774 else if (!PL_multiline)
1775 last = memEQ(strend - len, little, len)
1776 ? strend - len : Nullch;
1782 last = rninstr(s, strend, little, little + len);
1784 last = strend; /* matching `$' */
1788 DEBUG_r(PerlIO_printf(Perl_debug_log,
1789 "%sCan't trim the tail, match fails (should not happen)%s\n",
1790 PL_colors[4],PL_colors[5]));
1791 goto phooey; /* Should not happen! */
1793 dontbother = strend - last + prog->float_min_offset;
1795 if (minlen && (dontbother < minlen))
1796 dontbother = minlen - 1;
1797 strend -= dontbother; /* this one's always in bytes! */
1798 /* We don't know much -- general case. */
1801 if (regtry(prog, s))
1810 if (regtry(prog, s))
1812 } while (s++ < strend);
1820 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1822 if (PL_reg_eval_set) {
1823 /* Preserve the current value of $^R */
1824 if (oreplsv != GvSV(PL_replgv))
1825 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1826 restored, the value remains
1828 restore_pos(aTHX_ 0);
1831 /* make sure $`, $&, $', and $digit will work later */
1832 if ( !(flags & REXEC_NOT_FIRST) ) {
1833 if (RX_MATCH_COPIED(prog)) {
1834 Safefree(prog->subbeg);
1835 RX_MATCH_COPIED_off(prog);
1837 if (flags & REXEC_COPY_STR) {
1838 I32 i = PL_regeol - startpos + (stringarg - strbeg);
1840 s = savepvn(strbeg, i);
1843 RX_MATCH_COPIED_on(prog);
1846 prog->subbeg = strbeg;
1847 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
1854 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
1855 PL_colors[4],PL_colors[5]));
1856 if (PL_reg_eval_set)
1857 restore_pos(aTHX_ 0);
1862 - regtry - try match at specific point
1864 STATIC I32 /* 0 failure, 1 success */
1865 S_regtry(pTHX_ regexp *prog, char *startpos)
1873 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
1875 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
1878 PL_reg_eval_set = RS_init;
1880 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
1881 (IV)(PL_stack_sp - PL_stack_base));
1883 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
1884 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
1885 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
1887 /* Apparently this is not needed, judging by wantarray. */
1888 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
1889 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
1892 /* Make $_ available to executed code. */
1893 if (PL_reg_sv != DEFSV) {
1894 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
1899 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
1900 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
1901 /* prepare for quick setting of pos */
1902 sv_magic(PL_reg_sv, (SV*)0,
1903 PERL_MAGIC_regex_global, Nullch, 0);
1904 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
1908 PL_reg_oldpos = mg->mg_len;
1909 SAVEDESTRUCTOR_X(restore_pos, 0);
1911 if (!PL_reg_curpm) {
1912 Newz(22,PL_reg_curpm, 1, PMOP);
1915 SV* repointer = newSViv(0);
1916 /* so we know which PL_regex_padav element is PL_reg_curpm */
1917 SvFLAGS(repointer) |= SVf_BREAK;
1918 av_push(PL_regex_padav,repointer);
1919 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
1920 PL_regex_pad = AvARRAY(PL_regex_padav);
1924 PM_SETRE(PL_reg_curpm, prog);
1925 PL_reg_oldcurpm = PL_curpm;
1926 PL_curpm = PL_reg_curpm;
1927 if (RX_MATCH_COPIED(prog)) {
1928 /* Here is a serious problem: we cannot rewrite subbeg,
1929 since it may be needed if this match fails. Thus
1930 $` inside (?{}) could fail... */
1931 PL_reg_oldsaved = prog->subbeg;
1932 PL_reg_oldsavedlen = prog->sublen;
1933 RX_MATCH_COPIED_off(prog);
1936 PL_reg_oldsaved = Nullch;
1937 prog->subbeg = PL_bostr;
1938 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
1940 prog->startp[0] = startpos - PL_bostr;
1941 PL_reginput = startpos;
1942 PL_regstartp = prog->startp;
1943 PL_regendp = prog->endp;
1944 PL_reglastparen = &prog->lastparen;
1945 PL_reglastcloseparen = &prog->lastcloseparen;
1946 prog->lastparen = 0;
1948 DEBUG_r(PL_reg_starttry = startpos);
1949 if (PL_reg_start_tmpl <= prog->nparens) {
1950 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
1951 if(PL_reg_start_tmp)
1952 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1954 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1958 sv_setpvn(PERL_DEBUG_PAD(0), "", 0);
1959 sv_setpvn(PERL_DEBUG_PAD(1), "", 0);
1960 sv_setpvn(PERL_DEBUG_PAD(2), "", 0);
1963 /* XXXX What this code is doing here?!!! There should be no need
1964 to do this again and again, PL_reglastparen should take care of
1967 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
1968 * Actually, the code in regcppop() (which Ilya may be meaning by
1969 * PL_reglastparen), is not needed at all by the test suite
1970 * (op/regexp, op/pat, op/split), but that code is needed, oddly
1971 * enough, for building DynaLoader, or otherwise this
1972 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
1973 * will happen. Meanwhile, this code *is* needed for the
1974 * above-mentioned test suite tests to succeed. The common theme
1975 * on those tests seems to be returning null fields from matches.
1980 if (prog->nparens) {
1981 for (i = prog->nparens; i > *PL_reglastparen; i--) {
1988 if (regmatch(prog->program + 1)) {
1989 prog->endp[0] = PL_reginput - PL_bostr;
1992 REGCP_UNWIND(lastcp);
1996 #define RE_UNWIND_BRANCH 1
1997 #define RE_UNWIND_BRANCHJ 2
2001 typedef struct { /* XX: makes sense to enlarge it... */
2005 } re_unwind_generic_t;
2018 } re_unwind_branch_t;
2020 typedef union re_unwind_t {
2022 re_unwind_generic_t generic;
2023 re_unwind_branch_t branch;
2026 #define sayYES goto yes
2027 #define sayNO goto no
2028 #define sayYES_FINAL goto yes_final
2029 #define sayYES_LOUD goto yes_loud
2030 #define sayNO_FINAL goto no_final
2031 #define sayNO_SILENT goto do_no
2032 #define saySAME(x) if (x) goto yes; else goto no
2034 #define REPORT_CODE_OFF 24
2037 - regmatch - main matching routine
2039 * Conceptually the strategy is simple: check to see whether the current
2040 * node matches, call self recursively to see whether the rest matches,
2041 * and then act accordingly. In practice we make some effort to avoid
2042 * recursion, in particular by going through "ordinary" nodes (that don't
2043 * need to know whether the rest of the match failed) by a loop instead of
2046 /* [lwall] I've hoisted the register declarations to the outer block in order to
2047 * maybe save a little bit of pushing and popping on the stack. It also takes
2048 * advantage of machines that use a register save mask on subroutine entry.
2050 STATIC I32 /* 0 failure, 1 success */
2051 S_regmatch(pTHX_ regnode *prog)
2053 register regnode *scan; /* Current node. */
2054 regnode *next; /* Next node. */
2055 regnode *inner; /* Next node in internal branch. */
2056 register I32 nextchr; /* renamed nextchr - nextchar colides with
2057 function of same name */
2058 register I32 n; /* no or next */
2059 register I32 ln = 0; /* len or last */
2060 register char *s = Nullch; /* operand or save */
2061 register char *locinput = PL_reginput;
2062 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2063 int minmod = 0, sw = 0, logical = 0;
2066 I32 firstcp = PL_savestack_ix;
2068 register bool do_utf8 = PL_reg_match_utf8;
2070 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2071 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2072 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2079 /* Note that nextchr is a byte even in UTF */
2080 nextchr = UCHARAT(locinput);
2082 while (scan != NULL) {
2085 SV *prop = sv_newmortal();
2086 int docolor = *PL_colors[0];
2087 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2088 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2089 /* The part of the string before starttry has one color
2090 (pref0_len chars), between starttry and current
2091 position another one (pref_len - pref0_len chars),
2092 after the current position the third one.
2093 We assume that pref0_len <= pref_len, otherwise we
2094 decrease pref0_len. */
2095 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2096 ? (5 + taill) - l : locinput - PL_bostr;
2099 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2101 pref0_len = pref_len - (locinput - PL_reg_starttry);
2102 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2103 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2104 ? (5 + taill) - pref_len : PL_regeol - locinput);
2105 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2109 if (pref0_len > pref_len)
2110 pref0_len = pref_len;
2111 regprop(prop, scan);
2115 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2117 locinput - pref_len;
2118 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2119 char *s1 = do_utf8 ?
2120 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2121 pref_len - pref0_len, 60, 0) :
2122 locinput - pref_len + pref0_len;
2123 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2124 char *s2 = do_utf8 ?
2125 pv_uni_display(dsv2, (U8*)locinput,
2126 PL_regeol - locinput, 60, 0) :
2128 int len2 = do_utf8 ? strlen(s2) : l;
2129 PerlIO_printf(Perl_debug_log,
2130 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2131 (IV)(locinput - PL_bostr),
2138 (docolor ? "" : "> <"),
2142 15 - l - pref_len + 1,
2144 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2149 next = scan + NEXT_OFF(scan);
2155 if (locinput == PL_bostr || (PL_multiline &&
2156 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2158 /* regtill = regbol; */
2163 if (locinput == PL_bostr ||
2164 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2170 if (locinput == PL_bostr)
2174 if (locinput == PL_reg_ganch)
2184 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2189 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2191 if (PL_regeol - locinput > 1)
2195 if (PL_regeol != locinput)
2199 if (!nextchr && locinput >= PL_regeol)
2202 locinput += PL_utf8skip[nextchr];
2203 if (locinput > PL_regeol)
2205 nextchr = UCHARAT(locinput);
2208 nextchr = UCHARAT(++locinput);
2211 if (!nextchr && locinput >= PL_regeol)
2213 nextchr = UCHARAT(++locinput);
2216 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2219 locinput += PL_utf8skip[nextchr];
2220 if (locinput > PL_regeol)
2222 nextchr = UCHARAT(locinput);
2225 nextchr = UCHARAT(++locinput);
2230 if (do_utf8 != (UTF!=0)) {
2231 /* The target and the pattern have differing "utf8ness". */
2237 /* The target is utf8, the pattern is not utf8. */
2241 if (NATIVE_TO_UNI(*(U8*)s) !=
2242 utf8_to_uvchr((U8*)l, &len))
2249 /* The target is not utf8, the pattern is utf8. */
2253 if (NATIVE_TO_UNI(*((U8*)l)) !=
2254 utf8_to_uvchr((U8*)s, &len))
2261 nextchr = UCHARAT(locinput);
2264 /* The target and the pattern have the same "utf8ness". */
2265 /* Inline the first character, for speed. */
2266 if (UCHARAT(s) != nextchr)
2268 if (PL_regeol - locinput < ln)
2270 if (ln > 1 && memNE(s, locinput, ln))
2273 nextchr = UCHARAT(locinput);
2276 PL_reg_flags |= RF_tainted;
2286 U8 tmpbuf[UTF8_MAXLEN*2+1];
2291 toLOWER_utf8((U8*)l, tmpbuf, &ulen);
2292 if (memNE(s, (char*)tmpbuf, ulen))
2298 nextchr = UCHARAT(locinput);
2302 /* Inline the first character, for speed. */
2303 if (UCHARAT(s) != nextchr &&
2304 UCHARAT(s) != ((OP(scan) == EXACTF)
2305 ? PL_fold : PL_fold_locale)[nextchr])
2307 if (PL_regeol - locinput < ln)
2309 if (ln > 1 && (OP(scan) == EXACTF
2310 ? ibcmp(s, locinput, ln)
2311 : ibcmp_locale(s, locinput, ln)))
2314 nextchr = UCHARAT(locinput);
2318 if (!reginclass(scan, (U8*)locinput, do_utf8))
2320 if (locinput >= PL_regeol)
2322 locinput += PL_utf8skip[nextchr];
2323 nextchr = UCHARAT(locinput);
2327 nextchr = UCHARAT(locinput);
2328 if (!reginclass(scan, (U8*)locinput, do_utf8))
2330 if (!nextchr && locinput >= PL_regeol)
2332 nextchr = UCHARAT(++locinput);
2336 PL_reg_flags |= RF_tainted;
2342 LOAD_UTF8_CHARCLASS(alnum,"a");
2343 if (!(OP(scan) == ALNUM
2344 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2345 : isALNUM_LC_utf8((U8*)locinput)))
2349 locinput += PL_utf8skip[nextchr];
2350 nextchr = UCHARAT(locinput);
2353 if (!(OP(scan) == ALNUM
2354 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2356 nextchr = UCHARAT(++locinput);
2359 PL_reg_flags |= RF_tainted;
2362 if (!nextchr && locinput >= PL_regeol)
2365 LOAD_UTF8_CHARCLASS(alnum,"a");
2366 if (OP(scan) == NALNUM
2367 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2368 : isALNUM_LC_utf8((U8*)locinput))
2372 locinput += PL_utf8skip[nextchr];
2373 nextchr = UCHARAT(locinput);
2376 if (OP(scan) == NALNUM
2377 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2379 nextchr = UCHARAT(++locinput);
2383 PL_reg_flags |= RF_tainted;
2387 /* was last char in word? */
2389 if (locinput == PL_bostr)
2392 U8 *r = reghop((U8*)locinput, -1);
2394 ln = utf8n_to_uvchr(r, s - (char*)r, 0, 0);
2396 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2397 ln = isALNUM_uni(ln);
2398 LOAD_UTF8_CHARCLASS(alnum,"a");
2399 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2402 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2403 n = isALNUM_LC_utf8((U8*)locinput);
2407 ln = (locinput != PL_bostr) ?
2408 UCHARAT(locinput - 1) : '\n';
2409 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2411 n = isALNUM(nextchr);
2414 ln = isALNUM_LC(ln);
2415 n = isALNUM_LC(nextchr);
2418 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2419 OP(scan) == BOUNDL))
2423 PL_reg_flags |= RF_tainted;
2429 if (UTF8_IS_CONTINUED(nextchr)) {
2430 LOAD_UTF8_CHARCLASS(space," ");
2431 if (!(OP(scan) == SPACE
2432 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2433 : isSPACE_LC_utf8((U8*)locinput)))
2437 locinput += PL_utf8skip[nextchr];
2438 nextchr = UCHARAT(locinput);
2441 if (!(OP(scan) == SPACE
2442 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2444 nextchr = UCHARAT(++locinput);
2447 if (!(OP(scan) == SPACE
2448 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2450 nextchr = UCHARAT(++locinput);
2454 PL_reg_flags |= RF_tainted;
2457 if (!nextchr && locinput >= PL_regeol)
2460 LOAD_UTF8_CHARCLASS(space," ");
2461 if (OP(scan) == NSPACE
2462 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2463 : isSPACE_LC_utf8((U8*)locinput))
2467 locinput += PL_utf8skip[nextchr];
2468 nextchr = UCHARAT(locinput);
2471 if (OP(scan) == NSPACE
2472 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2474 nextchr = UCHARAT(++locinput);
2477 PL_reg_flags |= RF_tainted;
2483 LOAD_UTF8_CHARCLASS(digit,"0");
2484 if (!(OP(scan) == DIGIT
2485 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2486 : isDIGIT_LC_utf8((U8*)locinput)))
2490 locinput += PL_utf8skip[nextchr];
2491 nextchr = UCHARAT(locinput);
2494 if (!(OP(scan) == DIGIT
2495 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2497 nextchr = UCHARAT(++locinput);
2500 PL_reg_flags |= RF_tainted;
2503 if (!nextchr && locinput >= PL_regeol)
2506 LOAD_UTF8_CHARCLASS(digit,"0");
2507 if (OP(scan) == NDIGIT
2508 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2509 : isDIGIT_LC_utf8((U8*)locinput))
2513 locinput += PL_utf8skip[nextchr];
2514 nextchr = UCHARAT(locinput);
2517 if (OP(scan) == NDIGIT
2518 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2520 nextchr = UCHARAT(++locinput);
2523 LOAD_UTF8_CHARCLASS(mark,"~");
2524 if (locinput >= PL_regeol ||
2525 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2527 locinput += PL_utf8skip[nextchr];
2528 while (locinput < PL_regeol &&
2529 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2530 locinput += UTF8SKIP(locinput);
2531 if (locinput > PL_regeol)
2533 nextchr = UCHARAT(locinput);
2536 PL_reg_flags |= RF_tainted;
2540 n = ARG(scan); /* which paren pair */
2541 ln = PL_regstartp[n];
2542 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2543 if (*PL_reglastparen < n || ln == -1)
2544 sayNO; /* Do not match unless seen CLOSEn. */
2545 if (ln == PL_regendp[n])
2549 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2551 char *e = PL_bostr + PL_regendp[n];
2553 * Note that we can't do the "other character" lookup trick as
2554 * in the 8-bit case (no pun intended) because in Unicode we
2555 * have to map both upper and title case to lower case.
2557 if (OP(scan) == REFF) {
2558 STRLEN ulen1, ulen2;
2559 U8 tmpbuf1[UTF8_MAXLEN*2+1];
2560 U8 tmpbuf2[UTF8_MAXLEN*2+1];
2564 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2565 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2566 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2573 nextchr = UCHARAT(locinput);
2577 /* Inline the first character, for speed. */
2578 if (UCHARAT(s) != nextchr &&
2580 (UCHARAT(s) != ((OP(scan) == REFF
2581 ? PL_fold : PL_fold_locale)[nextchr]))))
2583 ln = PL_regendp[n] - ln;
2584 if (locinput + ln > PL_regeol)
2586 if (ln > 1 && (OP(scan) == REF
2587 ? memNE(s, locinput, ln)
2589 ? ibcmp(s, locinput, ln)
2590 : ibcmp_locale(s, locinput, ln))))
2593 nextchr = UCHARAT(locinput);
2604 OP_4tree *oop = PL_op;
2605 COP *ocurcop = PL_curcop;
2606 SV **ocurpad = PL_curpad;
2610 PL_op = (OP_4tree*)PL_regdata->data[n];
2611 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2612 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2613 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2617 CALLRUNOPS(aTHX); /* Scalar context. */
2620 ret = Nullsv; /* protect against empty (?{}) blocks. */
2628 PL_curpad = ocurpad;
2629 PL_curcop = ocurcop;
2631 if (logical == 2) { /* Postponed subexpression. */
2633 MAGIC *mg = Null(MAGIC*);
2635 CHECKPOINT cp, lastcp;
2637 if(SvROK(ret) || SvRMAGICAL(ret)) {
2638 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2641 mg = mg_find(sv, PERL_MAGIC_qr);
2644 re = (regexp *)mg->mg_obj;
2645 (void)ReREFCNT_inc(re);
2649 char *t = SvPV(ret, len);
2651 char *oprecomp = PL_regprecomp;
2652 I32 osize = PL_regsize;
2653 I32 onpar = PL_regnpar;
2656 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2658 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2659 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2661 PL_regprecomp = oprecomp;
2666 PerlIO_printf(Perl_debug_log,
2667 "Entering embedded `%s%.60s%s%s'\n",
2671 (strlen(re->precomp) > 60 ? "..." : ""))
2674 state.prev = PL_reg_call_cc;
2675 state.cc = PL_regcc;
2676 state.re = PL_reg_re;
2680 cp = regcppush(0); /* Save *all* the positions. */
2683 state.ss = PL_savestack_ix;
2684 *PL_reglastparen = 0;
2685 *PL_reglastcloseparen = 0;
2686 PL_reg_call_cc = &state;
2687 PL_reginput = locinput;
2689 /* XXXX This is too dramatic a measure... */
2692 if (regmatch(re->program + 1)) {
2693 /* Even though we succeeded, we need to restore
2694 global variables, since we may be wrapped inside
2695 SUSPEND, thus the match may be not finished yet. */
2697 /* XXXX Do this only if SUSPENDed? */
2698 PL_reg_call_cc = state.prev;
2699 PL_regcc = state.cc;
2700 PL_reg_re = state.re;
2701 cache_re(PL_reg_re);
2703 /* XXXX This is too dramatic a measure... */
2706 /* These are needed even if not SUSPEND. */
2712 REGCP_UNWIND(lastcp);
2714 PL_reg_call_cc = state.prev;
2715 PL_regcc = state.cc;
2716 PL_reg_re = state.re;
2717 cache_re(PL_reg_re);
2719 /* XXXX This is too dramatic a measure... */
2729 sv_setsv(save_scalar(PL_replgv), ret);
2733 n = ARG(scan); /* which paren pair */
2734 PL_reg_start_tmp[n] = locinput;
2739 n = ARG(scan); /* which paren pair */
2740 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2741 PL_regendp[n] = locinput - PL_bostr;
2742 if (n > *PL_reglastparen)
2743 *PL_reglastparen = n;
2744 *PL_reglastcloseparen = n;
2747 n = ARG(scan); /* which paren pair */
2748 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2751 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2753 next = NEXTOPER(NEXTOPER(scan));
2755 next = scan + ARG(scan);
2756 if (OP(next) == IFTHEN) /* Fake one. */
2757 next = NEXTOPER(NEXTOPER(next));
2761 logical = scan->flags;
2763 /*******************************************************************
2764 PL_regcc contains infoblock about the innermost (...)* loop, and
2765 a pointer to the next outer infoblock.
2767 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2769 1) After matching X, regnode for CURLYX is processed;
2771 2) This regnode creates infoblock on the stack, and calls
2772 regmatch() recursively with the starting point at WHILEM node;
2774 3) Each hit of WHILEM node tries to match A and Z (in the order
2775 depending on the current iteration, min/max of {min,max} and
2776 greediness). The information about where are nodes for "A"
2777 and "Z" is read from the infoblock, as is info on how many times "A"
2778 was already matched, and greediness.
2780 4) After A matches, the same WHILEM node is hit again.
2782 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2783 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2784 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2785 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2786 of the external loop.
2788 Currently present infoblocks form a tree with a stem formed by PL_curcc
2789 and whatever it mentions via ->next, and additional attached trees
2790 corresponding to temporarily unset infoblocks as in "5" above.
2792 In the following picture infoblocks for outer loop of
2793 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2794 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2795 infoblocks are drawn below the "reset" infoblock.
2797 In fact in the picture below we do not show failed matches for Z and T
2798 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2799 more obvious *why* one needs to *temporary* unset infoblocks.]
2801 Matched REx position InfoBlocks Comment
2805 Y A)*?Z)*?T x <- O <- I
2806 YA )*?Z)*?T x <- O <- I
2807 YA A)*?Z)*?T x <- O <- I
2808 YAA )*?Z)*?T x <- O <- I
2809 YAA Z)*?T x <- O # Temporary unset I
2812 YAAZ Y(A)*?Z)*?T x <- O
2815 YAAZY (A)*?Z)*?T x <- O
2818 YAAZY A)*?Z)*?T x <- O <- I
2821 YAAZYA )*?Z)*?T x <- O <- I
2824 YAAZYA Z)*?T x <- O # Temporary unset I
2830 YAAZYAZ T x # Temporary unset O
2837 *******************************************************************/
2840 CHECKPOINT cp = PL_savestack_ix;
2841 /* No need to save/restore up to this paren */
2842 I32 parenfloor = scan->flags;
2844 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2846 cc.oldcc = PL_regcc;
2848 /* XXXX Probably it is better to teach regpush to support
2849 parenfloor > PL_regsize... */
2850 if (parenfloor > *PL_reglastparen)
2851 parenfloor = *PL_reglastparen; /* Pessimization... */
2852 cc.parenfloor = parenfloor;
2854 cc.min = ARG1(scan);
2855 cc.max = ARG2(scan);
2856 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2860 PL_reginput = locinput;
2861 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2863 PL_regcc = cc.oldcc;
2869 * This is really hard to understand, because after we match
2870 * what we're trying to match, we must make sure the rest of
2871 * the REx is going to match for sure, and to do that we have
2872 * to go back UP the parse tree by recursing ever deeper. And
2873 * if it fails, we have to reset our parent's current state
2874 * that we can try again after backing off.
2877 CHECKPOINT cp, lastcp;
2878 CURCUR* cc = PL_regcc;
2879 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2881 n = cc->cur + 1; /* how many we know we matched */
2882 PL_reginput = locinput;
2885 PerlIO_printf(Perl_debug_log,
2886 "%*s %ld out of %ld..%ld cc=%lx\n",
2887 REPORT_CODE_OFF+PL_regindent*2, "",
2888 (long)n, (long)cc->min,
2889 (long)cc->max, (long)cc)
2892 /* If degenerate scan matches "", assume scan done. */
2894 if (locinput == cc->lastloc && n >= cc->min) {
2895 PL_regcc = cc->oldcc;
2899 PerlIO_printf(Perl_debug_log,
2900 "%*s empty match detected, try continuation...\n",
2901 REPORT_CODE_OFF+PL_regindent*2, "")
2903 if (regmatch(cc->next))
2911 /* First just match a string of min scans. */
2915 cc->lastloc = locinput;
2916 if (regmatch(cc->scan))
2919 cc->lastloc = lastloc;
2924 /* Check whether we already were at this position.
2925 Postpone detection until we know the match is not
2926 *that* much linear. */
2927 if (!PL_reg_maxiter) {
2928 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
2929 PL_reg_leftiter = PL_reg_maxiter;
2931 if (PL_reg_leftiter-- == 0) {
2932 I32 size = (PL_reg_maxiter + 7)/8;
2933 if (PL_reg_poscache) {
2934 if (PL_reg_poscache_size < size) {
2935 Renew(PL_reg_poscache, size, char);
2936 PL_reg_poscache_size = size;
2938 Zero(PL_reg_poscache, size, char);
2941 PL_reg_poscache_size = size;
2942 Newz(29, PL_reg_poscache, size, char);
2945 PerlIO_printf(Perl_debug_log,
2946 "%sDetected a super-linear match, switching on caching%s...\n",
2947 PL_colors[4], PL_colors[5])
2950 if (PL_reg_leftiter < 0) {
2951 I32 o = locinput - PL_bostr, b;
2953 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
2956 if (PL_reg_poscache[o] & (1<<b)) {
2958 PerlIO_printf(Perl_debug_log,
2959 "%*s already tried at this position...\n",
2960 REPORT_CODE_OFF+PL_regindent*2, "")
2964 PL_reg_poscache[o] |= (1<<b);
2968 /* Prefer next over scan for minimal matching. */
2971 PL_regcc = cc->oldcc;
2974 cp = regcppush(cc->parenfloor);
2976 if (regmatch(cc->next)) {
2978 sayYES; /* All done. */
2980 REGCP_UNWIND(lastcp);
2986 if (n >= cc->max) { /* Maximum greed exceeded? */
2987 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2988 && !(PL_reg_flags & RF_warned)) {
2989 PL_reg_flags |= RF_warned;
2990 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2991 "Complex regular subexpression recursion",
2998 PerlIO_printf(Perl_debug_log,
2999 "%*s trying longer...\n",
3000 REPORT_CODE_OFF+PL_regindent*2, "")
3002 /* Try scanning more and see if it helps. */
3003 PL_reginput = locinput;
3005 cc->lastloc = locinput;
3006 cp = regcppush(cc->parenfloor);
3008 if (regmatch(cc->scan)) {
3012 REGCP_UNWIND(lastcp);
3015 cc->lastloc = lastloc;
3019 /* Prefer scan over next for maximal matching. */
3021 if (n < cc->max) { /* More greed allowed? */
3022 cp = regcppush(cc->parenfloor);
3024 cc->lastloc = locinput;
3026 if (regmatch(cc->scan)) {
3030 REGCP_UNWIND(lastcp);
3031 regcppop(); /* Restore some previous $<digit>s? */
3032 PL_reginput = locinput;
3034 PerlIO_printf(Perl_debug_log,
3035 "%*s failed, try continuation...\n",
3036 REPORT_CODE_OFF+PL_regindent*2, "")
3039 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3040 && !(PL_reg_flags & RF_warned)) {
3041 PL_reg_flags |= RF_warned;
3042 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3043 "Complex regular subexpression recursion",
3047 /* Failed deeper matches of scan, so see if this one works. */
3048 PL_regcc = cc->oldcc;
3051 if (regmatch(cc->next))
3057 cc->lastloc = lastloc;
3062 next = scan + ARG(scan);
3065 inner = NEXTOPER(NEXTOPER(scan));
3068 inner = NEXTOPER(scan);
3072 if (OP(next) != c1) /* No choice. */
3073 next = inner; /* Avoid recursion. */
3075 I32 lastparen = *PL_reglastparen;
3077 re_unwind_branch_t *uw;
3079 /* Put unwinding data on stack */
3080 unwind1 = SSNEWt(1,re_unwind_branch_t);
3081 uw = SSPTRt(unwind1,re_unwind_branch_t);
3084 uw->type = ((c1 == BRANCH)
3086 : RE_UNWIND_BRANCHJ);
3087 uw->lastparen = lastparen;
3089 uw->locinput = locinput;
3090 uw->nextchr = nextchr;
3092 uw->regindent = ++PL_regindent;
3095 REGCP_SET(uw->lastcp);
3097 /* Now go into the first branch */
3110 /* We suppose that the next guy does not need
3111 backtracking: in particular, it is of constant length,
3112 and has no parenths to influence future backrefs. */
3113 ln = ARG1(scan); /* min to match */
3114 n = ARG2(scan); /* max to match */
3115 paren = scan->flags;
3117 if (paren > PL_regsize)
3119 if (paren > *PL_reglastparen)
3120 *PL_reglastparen = paren;
3122 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3124 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3125 PL_reginput = locinput;
3128 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3130 /* if we matched something zero-length we don't need to
3131 backtrack - capturing parens are already defined, so
3132 the caveat in the maximal case doesn't apply
3134 XXXX if ln == 0, we can redo this check first time
3135 through the following loop
3138 n = ln; /* don't backtrack */
3139 locinput = PL_reginput;
3140 if (HAS_TEXT(next) || JUMPABLE(next)) {
3141 regnode *text_node = next;
3143 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3145 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3147 if (PL_regkind[(U8)OP(text_node)] == REF) {
3149 n = ARG(text_node); /* which paren pair */
3150 ln = PL_regstartp[n];
3151 /* assume yes if we haven't seen CLOSEn */
3153 *PL_reglastparen < n ||
3160 c1 = *(PL_bostr + ln);
3162 else { c1 = (U8)*STRING(text_node); }
3163 if (OP(next) == EXACTF)
3165 else if (OP(text_node) == EXACTFL)
3166 c2 = PL_fold_locale[c1];
3175 /* This may be improved if l == 0. */
3176 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3177 /* If it could work, try it. */
3179 UCHARAT(PL_reginput) == c1 ||
3180 UCHARAT(PL_reginput) == c2)
3184 PL_regstartp[paren] =
3185 HOPc(PL_reginput, -l) - PL_bostr;
3186 PL_regendp[paren] = PL_reginput - PL_bostr;
3189 PL_regendp[paren] = -1;
3193 REGCP_UNWIND(lastcp);
3195 /* Couldn't or didn't -- move forward. */
3196 PL_reginput = locinput;
3197 if (regrepeat_hard(scan, 1, &l)) {
3199 locinput = PL_reginput;
3206 n = regrepeat_hard(scan, n, &l);
3207 /* if we matched something zero-length we don't need to
3208 backtrack, unless the minimum count is zero and we
3209 are capturing the result - in that case the capture
3210 being defined or not may affect later execution
3212 if (n != 0 && l == 0 && !(paren && ln == 0))
3213 ln = n; /* don't backtrack */
3214 locinput = PL_reginput;
3216 PerlIO_printf(Perl_debug_log,
3217 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3218 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3222 if (HAS_TEXT(next) || JUMPABLE(next)) {
3223 regnode *text_node = next;
3225 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3227 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3229 if (PL_regkind[(U8)OP(text_node)] == REF) {
3231 n = ARG(text_node); /* which paren pair */
3232 ln = PL_regstartp[n];
3233 /* assume yes if we haven't seen CLOSEn */
3235 *PL_reglastparen < n ||
3242 c1 = *(PL_bostr + ln);
3244 else { c1 = (U8)*STRING(text_node); }
3246 if (OP(text_node) == EXACTF)
3248 else if (OP(text_node) == EXACTFL)
3249 c2 = PL_fold_locale[c1];
3260 /* If it could work, try it. */
3262 UCHARAT(PL_reginput) == c1 ||
3263 UCHARAT(PL_reginput) == c2)
3266 PerlIO_printf(Perl_debug_log,
3267 "%*s trying tail with n=%"IVdf"...\n",
3268 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3272 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3273 PL_regendp[paren] = PL_reginput - PL_bostr;
3276 PL_regendp[paren] = -1;
3280 REGCP_UNWIND(lastcp);
3282 /* Couldn't or didn't -- back up. */
3284 locinput = HOPc(locinput, -l);
3285 PL_reginput = locinput;
3292 paren = scan->flags; /* Which paren to set */
3293 if (paren > PL_regsize)
3295 if (paren > *PL_reglastparen)
3296 *PL_reglastparen = paren;
3297 ln = ARG1(scan); /* min to match */
3298 n = ARG2(scan); /* max to match */
3299 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3303 ln = ARG1(scan); /* min to match */
3304 n = ARG2(scan); /* max to match */
3305 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3310 scan = NEXTOPER(scan);
3316 scan = NEXTOPER(scan);
3320 * Lookahead to avoid useless match attempts
3321 * when we know what character comes next.
3325 * Used to only do .*x and .*?x, but now it allows
3326 * for )'s, ('s and (?{ ... })'s to be in the way
3327 * of the quantifier and the EXACT-like node. -- japhy
3330 if (HAS_TEXT(next) || JUMPABLE(next)) {
3332 regnode *text_node = next;
3334 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3336 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3338 if (PL_regkind[(U8)OP(text_node)] == REF) {
3340 n = ARG(text_node); /* which paren pair */
3341 ln = PL_regstartp[n];
3342 /* assume yes if we haven't seen CLOSEn */
3344 *PL_reglastparen < n ||
3349 goto assume_ok_easy;
3351 s = (U8*)PL_bostr + ln;
3353 else { s = (U8*)STRING(text_node); }
3357 if (OP(text_node) == EXACTF)
3359 else if (OP(text_node) == EXACTFL)
3360 c2 = PL_fold_locale[c1];
3363 if (OP(text_node) == EXACTF) {
3364 STRLEN ulen1, ulen2;
3365 U8 tmpbuf1[UTF8_MAXLEN*2+1];
3366 U8 tmpbuf2[UTF8_MAXLEN*2+1];
3368 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3369 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3371 c1 = utf8_to_uvuni(tmpbuf1, 0);
3372 c2 = utf8_to_uvuni(tmpbuf2, 0);
3375 c2 = c1 = utf8_to_uvchr(s, NULL);
3383 PL_reginput = locinput;
3387 if (ln && regrepeat(scan, ln) < ln)
3389 locinput = PL_reginput;
3392 char *e; /* Should not check after this */
3393 char *old = locinput;
3395 if (n == REG_INFTY) {
3398 while (UTF8_IS_CONTINUATION(*(U8*)e))
3404 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3408 e = locinput + n - ln;
3414 /* Find place 'next' could work */
3417 while (locinput <= e &&
3418 UCHARAT(locinput) != c1)
3421 while (locinput <= e
3422 && UCHARAT(locinput) != c1
3423 && UCHARAT(locinput) != c2)
3426 count = locinput - old;
3433 utf8_to_uvchr((U8*)locinput, &len) != c1;
3438 for (count = 0; locinput <= e; count++) {
3439 UV c = utf8_to_uvchr((U8*)locinput, &len);
3440 if (c == c1 || c == c2)
3448 /* PL_reginput == old now */
3449 if (locinput != old) {
3450 ln = 1; /* Did some */
3451 if (regrepeat(scan, count) < count)
3454 /* PL_reginput == locinput now */
3455 TRYPAREN(paren, ln, locinput);
3456 PL_reginput = locinput; /* Could be reset... */
3457 REGCP_UNWIND(lastcp);
3458 /* Couldn't or didn't -- move forward. */
3461 locinput += UTF8SKIP(locinput);
3467 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3471 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3473 c = UCHARAT(PL_reginput);
3474 /* If it could work, try it. */
3475 if (c == c1 || c == c2)
3477 TRYPAREN(paren, n, PL_reginput);
3478 REGCP_UNWIND(lastcp);
3481 /* If it could work, try it. */
3482 else if (c1 == -1000)
3484 TRYPAREN(paren, n, PL_reginput);
3485 REGCP_UNWIND(lastcp);
3487 /* Couldn't or didn't -- move forward. */
3488 PL_reginput = locinput;
3489 if (regrepeat(scan, 1)) {
3491 locinput = PL_reginput;
3499 n = regrepeat(scan, n);
3500 locinput = PL_reginput;
3501 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3502 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3503 ln = n; /* why back off? */
3504 /* ...because $ and \Z can match before *and* after
3505 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3506 We should back off by one in this case. */
3507 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3516 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3518 c = UCHARAT(PL_reginput);
3520 /* If it could work, try it. */
3521 if (c1 == -1000 || c == c1 || c == c2)
3523 TRYPAREN(paren, n, PL_reginput);
3524 REGCP_UNWIND(lastcp);
3526 /* Couldn't or didn't -- back up. */
3528 PL_reginput = locinput = HOPc(locinput, -1);
3536 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3538 c = UCHARAT(PL_reginput);
3540 /* If it could work, try it. */
3541 if (c1 == -1000 || c == c1 || c == c2)
3543 TRYPAREN(paren, n, PL_reginput);
3544 REGCP_UNWIND(lastcp);
3546 /* Couldn't or didn't -- back up. */
3548 PL_reginput = locinput = HOPc(locinput, -1);
3555 if (PL_reg_call_cc) {
3556 re_cc_state *cur_call_cc = PL_reg_call_cc;
3557 CURCUR *cctmp = PL_regcc;
3558 regexp *re = PL_reg_re;
3559 CHECKPOINT cp, lastcp;
3561 cp = regcppush(0); /* Save *all* the positions. */
3563 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3565 PL_reginput = locinput; /* Make position available to
3567 cache_re(PL_reg_call_cc->re);
3568 PL_regcc = PL_reg_call_cc->cc;
3569 PL_reg_call_cc = PL_reg_call_cc->prev;
3570 if (regmatch(cur_call_cc->node)) {
3571 PL_reg_call_cc = cur_call_cc;
3575 REGCP_UNWIND(lastcp);
3577 PL_reg_call_cc = cur_call_cc;
3583 PerlIO_printf(Perl_debug_log,
3584 "%*s continuation failed...\n",
3585 REPORT_CODE_OFF+PL_regindent*2, "")
3589 if (locinput < PL_regtill) {
3590 DEBUG_r(PerlIO_printf(Perl_debug_log,
3591 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3593 (long)(locinput - PL_reg_starttry),
3594 (long)(PL_regtill - PL_reg_starttry),
3596 sayNO_FINAL; /* Cannot match: too short. */
3598 PL_reginput = locinput; /* put where regtry can find it */
3599 sayYES_FINAL; /* Success! */
3601 PL_reginput = locinput; /* put where regtry can find it */
3602 sayYES_LOUD; /* Success! */
3605 PL_reginput = locinput;
3610 s = HOPBACKc(locinput, scan->flags);
3616 PL_reginput = locinput;
3621 s = HOPBACKc(locinput, scan->flags);
3627 PL_reginput = locinput;
3630 inner = NEXTOPER(NEXTOPER(scan));
3631 if (regmatch(inner) != n) {
3646 if (OP(scan) == SUSPEND) {
3647 locinput = PL_reginput;
3648 nextchr = UCHARAT(locinput);
3653 next = scan + ARG(scan);
3658 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3659 PTR2UV(scan), OP(scan));
3660 Perl_croak(aTHX_ "regexp memory corruption");
3667 * We get here only if there's trouble -- normally "case END" is
3668 * the terminating point.
3670 Perl_croak(aTHX_ "corrupted regexp pointers");
3676 PerlIO_printf(Perl_debug_log,
3677 "%*s %scould match...%s\n",
3678 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3682 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3683 PL_colors[4],PL_colors[5]));
3689 #if 0 /* Breaks $^R */
3697 PerlIO_printf(Perl_debug_log,
3698 "%*s %sfailed...%s\n",
3699 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3705 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3708 case RE_UNWIND_BRANCH:
3709 case RE_UNWIND_BRANCHJ:
3711 re_unwind_branch_t *uwb = &(uw->branch);
3712 I32 lastparen = uwb->lastparen;
3714 REGCP_UNWIND(uwb->lastcp);
3715 for (n = *PL_reglastparen; n > lastparen; n--)
3717 *PL_reglastparen = n;
3718 scan = next = uwb->next;
3720 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3721 ? BRANCH : BRANCHJ) ) { /* Failure */
3728 /* Have more choice yet. Reuse the same uwb. */
3730 if ((n = (uwb->type == RE_UNWIND_BRANCH
3731 ? NEXT_OFF(next) : ARG(next))))
3734 next = NULL; /* XXXX Needn't unwinding in this case... */
3736 next = NEXTOPER(scan);
3737 if (uwb->type == RE_UNWIND_BRANCHJ)
3738 next = NEXTOPER(next);
3739 locinput = uwb->locinput;
3740 nextchr = uwb->nextchr;
3742 PL_regindent = uwb->regindent;
3749 Perl_croak(aTHX_ "regexp unwind memory corruption");
3760 - regrepeat - repeatedly match something simple, report how many
3763 * [This routine now assumes that it will only match on things of length 1.
3764 * That was true before, but now we assume scan - reginput is the count,
3765 * rather than incrementing count on every character. [Er, except utf8.]]
3768 S_regrepeat(pTHX_ regnode *p, I32 max)
3770 register char *scan;
3772 register char *loceol = PL_regeol;
3773 register I32 hardcount = 0;
3774 register bool do_utf8 = PL_reg_match_utf8;
3777 if (max != REG_INFTY && max < loceol - scan)
3778 loceol = scan + max;
3783 while (scan < loceol && hardcount < max && *scan != '\n') {
3784 scan += UTF8SKIP(scan);
3788 while (scan < loceol && *scan != '\n')
3798 case EXACT: /* length of string is 1 */
3800 while (scan < loceol && UCHARAT(scan) == c)
3803 case EXACTF: /* length of string is 1 */
3805 while (scan < loceol &&
3806 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3809 case EXACTFL: /* length of string is 1 */
3810 PL_reg_flags |= RF_tainted;
3812 while (scan < loceol &&
3813 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3819 while (hardcount < max && scan < loceol &&
3820 reginclass(p, (U8*)scan, do_utf8)) {
3821 scan += UTF8SKIP(scan);
3825 while (scan < loceol && reginclass(p, (U8*)scan, do_utf8))
3832 LOAD_UTF8_CHARCLASS(alnum,"a");
3833 while (hardcount < max && scan < loceol &&
3834 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3835 scan += UTF8SKIP(scan);
3839 while (scan < loceol && isALNUM(*scan))
3844 PL_reg_flags |= RF_tainted;
3847 while (hardcount < max && scan < loceol &&
3848 isALNUM_LC_utf8((U8*)scan)) {
3849 scan += UTF8SKIP(scan);
3853 while (scan < loceol && isALNUM_LC(*scan))
3860 LOAD_UTF8_CHARCLASS(alnum,"a");
3861 while (hardcount < max && scan < loceol &&
3862 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3863 scan += UTF8SKIP(scan);
3867 while (scan < loceol && !isALNUM(*scan))
3872 PL_reg_flags |= RF_tainted;
3875 while (hardcount < max && scan < loceol &&
3876 !isALNUM_LC_utf8((U8*)scan)) {
3877 scan += UTF8SKIP(scan);
3881 while (scan < loceol && !isALNUM_LC(*scan))
3888 LOAD_UTF8_CHARCLASS(space," ");
3889 while (hardcount < max && scan < loceol &&
3891 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3892 scan += UTF8SKIP(scan);
3896 while (scan < loceol && isSPACE(*scan))
3901 PL_reg_flags |= RF_tainted;
3904 while (hardcount < max && scan < loceol &&
3905 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3906 scan += UTF8SKIP(scan);
3910 while (scan < loceol && isSPACE_LC(*scan))
3917 LOAD_UTF8_CHARCLASS(space," ");
3918 while (hardcount < max && scan < loceol &&
3920 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3921 scan += UTF8SKIP(scan);
3925 while (scan < loceol && !isSPACE(*scan))
3930 PL_reg_flags |= RF_tainted;
3933 while (hardcount < max && scan < loceol &&
3934 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3935 scan += UTF8SKIP(scan);
3939 while (scan < loceol && !isSPACE_LC(*scan))
3946 LOAD_UTF8_CHARCLASS(digit,"0");
3947 while (hardcount < max && scan < loceol &&
3948 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
3949 scan += UTF8SKIP(scan);
3953 while (scan < loceol && isDIGIT(*scan))
3960 LOAD_UTF8_CHARCLASS(digit,"0");
3961 while (hardcount < max && scan < loceol &&
3962 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
3963 scan += UTF8SKIP(scan);
3967 while (scan < loceol && !isDIGIT(*scan))
3971 default: /* Called on something of 0 width. */
3972 break; /* So match right here or not at all. */
3978 c = scan - PL_reginput;
3983 SV *prop = sv_newmortal();
3986 PerlIO_printf(Perl_debug_log,
3987 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
3988 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
3995 - regrepeat_hard - repeatedly match something, report total lenth and length
3997 * The repeater is supposed to have constant length.
4001 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4003 register char *scan = Nullch;
4004 register char *start;
4005 register char *loceol = PL_regeol;
4007 I32 count = 0, res = 1;
4012 start = PL_reginput;
4013 if (PL_reg_match_utf8) {
4014 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4017 while (start < PL_reginput) {
4019 start += UTF8SKIP(start);
4030 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4032 *lp = l = PL_reginput - start;
4033 if (max != REG_INFTY && l*max < loceol - scan)
4034 loceol = scan + l*max;
4047 - regclass_swash - prepare the utf8 swash
4051 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** initsvp)
4056 if (PL_regdata && PL_regdata->count) {
4059 if (PL_regdata->what[n] == 's') {
4060 SV *rv = (SV*)PL_regdata->data[n];
4061 AV *av = (AV*)SvRV((SV*)rv);
4064 si = *av_fetch(av, 0, FALSE);
4065 a = av_fetch(av, 1, FALSE);
4069 else if (si && doinit) {
4070 sw = swash_init("utf8", "", si, 1, 0);
4071 (void)av_store(av, 1, sw);
4083 - reginclass - determine if a character falls into a character class
4087 S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8)
4089 char flags = ANYOF_FLAGS(n);
4094 c = do_utf8 ? utf8_to_uvchr(p, &len) : *p;
4096 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4097 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4098 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4101 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4104 SV *sw = regclass_swash(n, TRUE, 0);
4107 if (swash_fetch(sw, p, do_utf8))
4109 else if (flags & ANYOF_FOLD) {
4111 U8 tmpbuf[UTF8_MAXLEN*2+1];
4113 toLOWER_utf8(p, tmpbuf, &ulen);
4114 if (swash_fetch(sw, tmpbuf, do_utf8))
4120 if (!match && c < 256) {
4121 if (ANYOF_BITMAP_TEST(n, c))
4123 else if (flags & ANYOF_FOLD) {
4126 if (flags & ANYOF_LOCALE) {
4127 PL_reg_flags |= RF_tainted;
4128 f = PL_fold_locale[c];
4132 if (f != c && ANYOF_BITMAP_TEST(n, f))
4136 if (!match && (flags & ANYOF_CLASS)) {
4137 PL_reg_flags |= RF_tainted;
4139 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4140 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4141 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4142 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4143 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4144 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4145 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4146 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4147 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4148 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4149 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4150 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4151 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4152 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4153 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4154 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4155 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4156 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4157 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4158 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4159 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4160 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4161 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4162 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4163 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4164 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4165 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4166 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4167 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4168 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4169 ) /* How's that for a conditional? */
4176 return (flags & ANYOF_INVERT) ? !match : match;
4180 S_reghop(pTHX_ U8 *s, I32 off)
4182 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4186 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4189 while (off-- && s < lim) {
4190 /* XXX could check well-formedness here */
4198 if (UTF8_IS_CONTINUED(*s)) {
4199 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4202 /* XXX could check well-formedness here */
4210 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4212 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4216 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4219 while (off-- && s < lim) {
4220 /* XXX could check well-formedness here */
4230 if (UTF8_IS_CONTINUED(*s)) {
4231 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4234 /* XXX could check well-formedness here */
4246 restore_pos(pTHX_ void *arg)
4248 if (PL_reg_eval_set) {
4249 if (PL_reg_oldsaved) {
4250 PL_reg_re->subbeg = PL_reg_oldsaved;
4251 PL_reg_re->sublen = PL_reg_oldsavedlen;
4252 RX_MATCH_COPIED_on(PL_reg_re);
4254 PL_reg_magic->mg_len = PL_reg_oldpos;
4255 PL_reg_eval_set = 0;
4256 PL_curpm = PL_reg_oldcurpm;