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() */
964 if ( utf8_to_uvchr((U8*)s, &len) == c1
966 ibcmp_utf8(s, do_utf8, m, UTF, ln)) )
972 UV c = utf8_to_uvchr((U8*)s, &len);
973 if ( (c == c1 || c == c2)
975 ibcmp_utf8(s, do_utf8, m, UTF, ln)) )
984 && (ln == 1 || !(OP(c) == EXACTF
986 : ibcmp_locale(s, m, ln)))
987 && (norun || regtry(prog, s)) )
993 if ( (*(U8*)s == c1 || *(U8*)s == c2)
994 && (ln == 1 || !(OP(c) == EXACTF
996 : ibcmp_locale(s, m, ln)))
997 && (norun || regtry(prog, s)) )
1004 PL_reg_flags |= RF_tainted;
1011 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1014 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1016 tmp = ((OP(c) == BOUND ?
1017 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1018 LOAD_UTF8_CHARCLASS(alnum,"a");
1019 while (s < strend) {
1020 if (tmp == !(OP(c) == BOUND ?
1021 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1022 isALNUM_LC_utf8((U8*)s)))
1025 if ((norun || regtry(prog, s)))
1032 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1033 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1034 while (s < strend) {
1036 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1038 if ((norun || regtry(prog, s)))
1044 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1048 PL_reg_flags |= RF_tainted;
1055 U8 *r = reghop3((U8*)s, -1, (U8*)startpos);
1058 tmp = (I32)utf8n_to_uvchr(r, s - (char*)r, 0, 0);
1060 tmp = ((OP(c) == NBOUND ?
1061 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1062 LOAD_UTF8_CHARCLASS(alnum,"a");
1063 while (s < strend) {
1064 if (tmp == !(OP(c) == NBOUND ?
1065 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1066 isALNUM_LC_utf8((U8*)s)))
1068 else if ((norun || regtry(prog, s)))
1074 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1075 tmp = ((OP(c) == NBOUND ?
1076 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1077 while (s < strend) {
1079 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1081 else if ((norun || regtry(prog, s)))
1086 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1091 LOAD_UTF8_CHARCLASS(alnum,"a");
1092 while (s < strend) {
1093 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1094 if (tmp && (norun || regtry(prog, s)))
1105 while (s < strend) {
1107 if (tmp && (norun || regtry(prog, s)))
1119 PL_reg_flags |= RF_tainted;
1121 while (s < strend) {
1122 if (isALNUM_LC_utf8((U8*)s)) {
1123 if (tmp && (norun || regtry(prog, s)))
1134 while (s < strend) {
1135 if (isALNUM_LC(*s)) {
1136 if (tmp && (norun || regtry(prog, s)))
1149 LOAD_UTF8_CHARCLASS(alnum,"a");
1150 while (s < strend) {
1151 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1152 if (tmp && (norun || regtry(prog, s)))
1163 while (s < strend) {
1165 if (tmp && (norun || regtry(prog, s)))
1177 PL_reg_flags |= RF_tainted;
1179 while (s < strend) {
1180 if (!isALNUM_LC_utf8((U8*)s)) {
1181 if (tmp && (norun || regtry(prog, s)))
1192 while (s < strend) {
1193 if (!isALNUM_LC(*s)) {
1194 if (tmp && (norun || regtry(prog, s)))
1207 LOAD_UTF8_CHARCLASS(space," ");
1208 while (s < strend) {
1209 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1210 if (tmp && (norun || regtry(prog, s)))
1221 while (s < strend) {
1223 if (tmp && (norun || regtry(prog, s)))
1235 PL_reg_flags |= RF_tainted;
1237 while (s < strend) {
1238 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1239 if (tmp && (norun || regtry(prog, s)))
1250 while (s < strend) {
1251 if (isSPACE_LC(*s)) {
1252 if (tmp && (norun || regtry(prog, s)))
1265 LOAD_UTF8_CHARCLASS(space," ");
1266 while (s < strend) {
1267 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1268 if (tmp && (norun || regtry(prog, s)))
1279 while (s < strend) {
1281 if (tmp && (norun || regtry(prog, s)))
1293 PL_reg_flags |= RF_tainted;
1295 while (s < strend) {
1296 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1297 if (tmp && (norun || regtry(prog, s)))
1308 while (s < strend) {
1309 if (!isSPACE_LC(*s)) {
1310 if (tmp && (norun || regtry(prog, s)))
1323 LOAD_UTF8_CHARCLASS(digit,"0");
1324 while (s < strend) {
1325 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1326 if (tmp && (norun || regtry(prog, s)))
1337 while (s < strend) {
1339 if (tmp && (norun || regtry(prog, s)))
1351 PL_reg_flags |= RF_tainted;
1353 while (s < strend) {
1354 if (isDIGIT_LC_utf8((U8*)s)) {
1355 if (tmp && (norun || regtry(prog, s)))
1366 while (s < strend) {
1367 if (isDIGIT_LC(*s)) {
1368 if (tmp && (norun || regtry(prog, s)))
1381 LOAD_UTF8_CHARCLASS(digit,"0");
1382 while (s < strend) {
1383 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1384 if (tmp && (norun || regtry(prog, s)))
1395 while (s < strend) {
1397 if (tmp && (norun || regtry(prog, s)))
1409 PL_reg_flags |= RF_tainted;
1411 while (s < strend) {
1412 if (!isDIGIT_LC_utf8((U8*)s)) {
1413 if (tmp && (norun || regtry(prog, s)))
1424 while (s < strend) {
1425 if (!isDIGIT_LC(*s)) {
1426 if (tmp && (norun || regtry(prog, s)))
1438 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1447 - regexec_flags - match a regexp against a string
1450 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1451 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1452 /* strend: pointer to null at end of string */
1453 /* strbeg: real beginning of string */
1454 /* minend: end of match must be >=minend after stringarg. */
1455 /* data: May be used for some additional optimizations. */
1456 /* nosave: For optimizations. */
1459 register regnode *c;
1460 register char *startpos = stringarg;
1461 I32 minlen; /* must match at least this many chars */
1462 I32 dontbother = 0; /* how many characters not to try at end */
1463 /* I32 start_shift = 0; */ /* Offset of the start to find
1464 constant substr. */ /* CC */
1465 I32 end_shift = 0; /* Same for the end. */ /* CC */
1466 I32 scream_pos = -1; /* Internal iterator of scream. */
1468 SV* oreplsv = GvSV(PL_replgv);
1469 bool do_utf8 = DO_UTF8(sv);
1471 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
1478 PL_regnarrate = DEBUG_r_TEST;
1481 /* Be paranoid... */
1482 if (prog == NULL || startpos == NULL) {
1483 Perl_croak(aTHX_ "NULL regexp parameter");
1487 minlen = prog->minlen;
1488 if (strend - startpos < minlen) {
1489 DEBUG_r(PerlIO_printf(Perl_debug_log,
1490 "String too short [regexec_flags]...\n"));
1494 /* Check validity of program. */
1495 if (UCHARAT(prog->program) != REG_MAGIC) {
1496 Perl_croak(aTHX_ "corrupted regexp program");
1500 PL_reg_eval_set = 0;
1503 if (prog->reganch & ROPT_UTF8)
1504 PL_reg_flags |= RF_utf8;
1506 /* Mark beginning of line for ^ and lookbehind. */
1507 PL_regbol = startpos;
1511 /* Mark end of line for $ (and such) */
1514 /* see how far we have to get to not match where we matched before */
1515 PL_regtill = startpos+minend;
1517 /* We start without call_cc context. */
1520 /* If there is a "must appear" string, look for it. */
1523 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1526 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1527 PL_reg_ganch = startpos;
1528 else if (sv && SvTYPE(sv) >= SVt_PVMG
1530 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1531 && mg->mg_len >= 0) {
1532 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1533 if (prog->reganch & ROPT_ANCH_GPOS) {
1534 if (s > PL_reg_ganch)
1539 else /* pos() not defined */
1540 PL_reg_ganch = strbeg;
1543 if (do_utf8 == (UTF!=0) &&
1544 !(flags & REXEC_CHECKED) && prog->check_substr != Nullsv) {
1545 re_scream_pos_data d;
1547 d.scream_olds = &scream_olds;
1548 d.scream_pos = &scream_pos;
1549 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1551 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1552 goto phooey; /* not present */
1557 char *s = do_utf8 ? sv_uni_display(dsv, sv, 60, 0) : startpos;
1558 int len = do_utf8 ? strlen(s) : strend - startpos;
1561 PerlIO_printf(Perl_debug_log,
1562 "%sMatching REx%s `%s%.60s%s%s' against `%s%.*s%s%s'\n",
1563 PL_colors[4],PL_colors[5],PL_colors[0],
1566 (strlen(prog->precomp) > 60 ? "..." : ""),
1568 (int)(len > 60 ? 60 : len),
1570 (len > 60 ? "..." : "")
1574 /* Simplest case: anchored match need be tried only once. */
1575 /* [unless only anchor is BOL and multiline is set] */
1576 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1577 if (s == startpos && regtry(prog, startpos))
1579 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1580 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1585 dontbother = minlen - 1;
1586 end = HOP3c(strend, -dontbother, strbeg) - 1;
1587 /* for multiline we only have to try after newlines */
1588 if (prog->check_substr) {
1592 if (regtry(prog, s))
1597 if (prog->reganch & RE_USE_INTUIT) {
1598 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1609 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1610 if (regtry(prog, s))
1617 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1618 if (regtry(prog, PL_reg_ganch))
1623 /* Messy cases: unanchored match. */
1624 if (prog->anchored_substr && prog->reganch & ROPT_SKIP) {
1625 /* we have /x+whatever/ */
1626 /* it must be a one character string (XXXX Except UTF?) */
1627 char ch = SvPVX(prog->anchored_substr)[0];
1633 while (s < strend) {
1635 DEBUG_r( did_match = 1 );
1636 if (regtry(prog, s)) goto got_it;
1638 while (s < strend && *s == ch)
1645 while (s < strend) {
1647 DEBUG_r( did_match = 1 );
1648 if (regtry(prog, s)) goto got_it;
1650 while (s < strend && *s == ch)
1656 DEBUG_r(if (!did_match)
1657 PerlIO_printf(Perl_debug_log,
1658 "Did not find anchored character...\n")
1662 else if (do_utf8 == (UTF!=0) &&
1663 (prog->anchored_substr != Nullsv
1664 || (prog->float_substr != Nullsv
1665 && prog->float_max_offset < strend - s))) {
1666 SV *must = prog->anchored_substr
1667 ? prog->anchored_substr : prog->float_substr;
1669 prog->anchored_substr ? prog->anchored_offset : prog->float_max_offset;
1671 prog->anchored_substr ? prog->anchored_offset : prog->float_min_offset;
1672 char *last = HOP3c(strend, /* Cannot start after this */
1673 -(I32)(CHR_SVLEN(must)
1674 - (SvTAIL(must) != 0) + back_min), strbeg);
1675 char *last1; /* Last position checked before */
1681 last1 = HOPc(s, -1);
1683 last1 = s - 1; /* bogus */
1685 /* XXXX check_substr already used to find `s', can optimize if
1686 check_substr==must. */
1688 dontbother = end_shift;
1689 strend = HOPc(strend, -dontbother);
1690 while ( (s <= last) &&
1691 ((flags & REXEC_SCREAM)
1692 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1693 end_shift, &scream_pos, 0))
1694 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1695 (unsigned char*)strend, must,
1696 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1697 DEBUG_r( did_match = 1 );
1698 if (HOPc(s, -back_max) > last1) {
1699 last1 = HOPc(s, -back_min);
1700 s = HOPc(s, -back_max);
1703 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1705 last1 = HOPc(s, -back_min);
1709 while (s <= last1) {
1710 if (regtry(prog, s))
1716 while (s <= last1) {
1717 if (regtry(prog, s))
1723 DEBUG_r(if (!did_match)
1724 PerlIO_printf(Perl_debug_log,
1725 "Did not find %s substr `%s%.*s%s'%s...\n",
1726 ((must == prog->anchored_substr)
1727 ? "anchored" : "floating"),
1729 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1731 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1735 else if ((c = prog->regstclass)) {
1736 if (minlen && PL_regkind[(U8)OP(prog->regstclass)] != EXACT)
1737 /* don't bother with what can't match */
1738 strend = HOPc(strend, -(minlen - 1));
1740 SV *prop = sv_newmortal();
1742 PerlIO_printf(Perl_debug_log, "Matching stclass `%s' against `%s'\n", SvPVX(prop), UTF ? sv_uni_display(dsv, sv, 60, 0) : s);
1744 if (find_byclass(prog, c, s, strend, startpos, 0))
1746 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1750 if (prog->float_substr != Nullsv) { /* Trim the end. */
1753 if (flags & REXEC_SCREAM) {
1754 last = screaminstr(sv, prog->float_substr, s - strbeg,
1755 end_shift, &scream_pos, 1); /* last one */
1757 last = scream_olds; /* Only one occurrence. */
1761 char *little = SvPV(prog->float_substr, len);
1763 if (SvTAIL(prog->float_substr)) {
1764 if (memEQ(strend - len + 1, little, len - 1))
1765 last = strend - len + 1;
1766 else if (!PL_multiline)
1767 last = memEQ(strend - len, little, len)
1768 ? strend - len : Nullch;
1774 last = rninstr(s, strend, little, little + len);
1776 last = strend; /* matching `$' */
1780 DEBUG_r(PerlIO_printf(Perl_debug_log,
1781 "%sCan't trim the tail, match fails (should not happen)%s\n",
1782 PL_colors[4],PL_colors[5]));
1783 goto phooey; /* Should not happen! */
1785 dontbother = strend - last + prog->float_min_offset;
1787 if (minlen && (dontbother < minlen))
1788 dontbother = minlen - 1;
1789 strend -= dontbother; /* this one's always in bytes! */
1790 /* We don't know much -- general case. */
1793 if (regtry(prog, s))
1802 if (regtry(prog, s))
1804 } while (s++ < strend);
1812 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
1814 if (PL_reg_eval_set) {
1815 /* Preserve the current value of $^R */
1816 if (oreplsv != GvSV(PL_replgv))
1817 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
1818 restored, the value remains
1820 restore_pos(aTHX_ 0);
1823 /* make sure $`, $&, $', and $digit will work later */
1824 if ( !(flags & REXEC_NOT_FIRST) ) {
1825 if (RX_MATCH_COPIED(prog)) {
1826 Safefree(prog->subbeg);
1827 RX_MATCH_COPIED_off(prog);
1829 if (flags & REXEC_COPY_STR) {
1830 I32 i = PL_regeol - startpos + (stringarg - strbeg);
1832 s = savepvn(strbeg, i);
1835 RX_MATCH_COPIED_on(prog);
1838 prog->subbeg = strbeg;
1839 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
1846 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
1847 PL_colors[4],PL_colors[5]));
1848 if (PL_reg_eval_set)
1849 restore_pos(aTHX_ 0);
1854 - regtry - try match at specific point
1856 STATIC I32 /* 0 failure, 1 success */
1857 S_regtry(pTHX_ regexp *prog, char *startpos)
1865 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
1867 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
1870 PL_reg_eval_set = RS_init;
1872 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
1873 (IV)(PL_stack_sp - PL_stack_base));
1875 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
1876 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
1877 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
1879 /* Apparently this is not needed, judging by wantarray. */
1880 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
1881 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
1884 /* Make $_ available to executed code. */
1885 if (PL_reg_sv != DEFSV) {
1886 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
1891 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
1892 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
1893 /* prepare for quick setting of pos */
1894 sv_magic(PL_reg_sv, (SV*)0,
1895 PERL_MAGIC_regex_global, Nullch, 0);
1896 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
1900 PL_reg_oldpos = mg->mg_len;
1901 SAVEDESTRUCTOR_X(restore_pos, 0);
1903 if (!PL_reg_curpm) {
1904 Newz(22,PL_reg_curpm, 1, PMOP);
1907 SV* repointer = newSViv(0);
1908 /* so we know which PL_regex_padav element is PL_reg_curpm */
1909 SvFLAGS(repointer) |= SVf_BREAK;
1910 av_push(PL_regex_padav,repointer);
1911 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
1912 PL_regex_pad = AvARRAY(PL_regex_padav);
1916 PM_SETRE(PL_reg_curpm, prog);
1917 PL_reg_oldcurpm = PL_curpm;
1918 PL_curpm = PL_reg_curpm;
1919 if (RX_MATCH_COPIED(prog)) {
1920 /* Here is a serious problem: we cannot rewrite subbeg,
1921 since it may be needed if this match fails. Thus
1922 $` inside (?{}) could fail... */
1923 PL_reg_oldsaved = prog->subbeg;
1924 PL_reg_oldsavedlen = prog->sublen;
1925 RX_MATCH_COPIED_off(prog);
1928 PL_reg_oldsaved = Nullch;
1929 prog->subbeg = PL_bostr;
1930 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
1932 prog->startp[0] = startpos - PL_bostr;
1933 PL_reginput = startpos;
1934 PL_regstartp = prog->startp;
1935 PL_regendp = prog->endp;
1936 PL_reglastparen = &prog->lastparen;
1937 PL_reglastcloseparen = &prog->lastcloseparen;
1938 prog->lastparen = 0;
1940 DEBUG_r(PL_reg_starttry = startpos);
1941 if (PL_reg_start_tmpl <= prog->nparens) {
1942 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
1943 if(PL_reg_start_tmp)
1944 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1946 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
1950 sv_setpvn(PERL_DEBUG_PAD(0), "", 0);
1951 sv_setpvn(PERL_DEBUG_PAD(1), "", 0);
1952 sv_setpvn(PERL_DEBUG_PAD(2), "", 0);
1955 /* XXXX What this code is doing here?!!! There should be no need
1956 to do this again and again, PL_reglastparen should take care of
1959 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
1960 * Actually, the code in regcppop() (which Ilya may be meaning by
1961 * PL_reglastparen), is not needed at all by the test suite
1962 * (op/regexp, op/pat, op/split), but that code is needed, oddly
1963 * enough, for building DynaLoader, or otherwise this
1964 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
1965 * will happen. Meanwhile, this code *is* needed for the
1966 * above-mentioned test suite tests to succeed. The common theme
1967 * on those tests seems to be returning null fields from matches.
1972 if (prog->nparens) {
1973 for (i = prog->nparens; i > *PL_reglastparen; i--) {
1980 if (regmatch(prog->program + 1)) {
1981 prog->endp[0] = PL_reginput - PL_bostr;
1984 REGCP_UNWIND(lastcp);
1988 #define RE_UNWIND_BRANCH 1
1989 #define RE_UNWIND_BRANCHJ 2
1993 typedef struct { /* XX: makes sense to enlarge it... */
1997 } re_unwind_generic_t;
2010 } re_unwind_branch_t;
2012 typedef union re_unwind_t {
2014 re_unwind_generic_t generic;
2015 re_unwind_branch_t branch;
2018 #define sayYES goto yes
2019 #define sayNO goto no
2020 #define sayYES_FINAL goto yes_final
2021 #define sayYES_LOUD goto yes_loud
2022 #define sayNO_FINAL goto no_final
2023 #define sayNO_SILENT goto do_no
2024 #define saySAME(x) if (x) goto yes; else goto no
2026 #define REPORT_CODE_OFF 24
2029 - regmatch - main matching routine
2031 * Conceptually the strategy is simple: check to see whether the current
2032 * node matches, call self recursively to see whether the rest matches,
2033 * and then act accordingly. In practice we make some effort to avoid
2034 * recursion, in particular by going through "ordinary" nodes (that don't
2035 * need to know whether the rest of the match failed) by a loop instead of
2038 /* [lwall] I've hoisted the register declarations to the outer block in order to
2039 * maybe save a little bit of pushing and popping on the stack. It also takes
2040 * advantage of machines that use a register save mask on subroutine entry.
2042 STATIC I32 /* 0 failure, 1 success */
2043 S_regmatch(pTHX_ regnode *prog)
2045 register regnode *scan; /* Current node. */
2046 regnode *next; /* Next node. */
2047 regnode *inner; /* Next node in internal branch. */
2048 register I32 nextchr; /* renamed nextchr - nextchar colides with
2049 function of same name */
2050 register I32 n; /* no or next */
2051 register I32 ln = 0; /* len or last */
2052 register char *s = Nullch; /* operand or save */
2053 register char *locinput = PL_reginput;
2054 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2055 int minmod = 0, sw = 0, logical = 0;
2058 I32 firstcp = PL_savestack_ix;
2060 register bool do_utf8 = PL_reg_match_utf8;
2062 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2063 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2064 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2071 /* Note that nextchr is a byte even in UTF */
2072 nextchr = UCHARAT(locinput);
2074 while (scan != NULL) {
2077 SV *prop = sv_newmortal();
2078 int docolor = *PL_colors[0];
2079 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2080 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2081 /* The part of the string before starttry has one color
2082 (pref0_len chars), between starttry and current
2083 position another one (pref_len - pref0_len chars),
2084 after the current position the third one.
2085 We assume that pref0_len <= pref_len, otherwise we
2086 decrease pref0_len. */
2087 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2088 ? (5 + taill) - l : locinput - PL_bostr;
2091 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2093 pref0_len = pref_len - (locinput - PL_reg_starttry);
2094 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2095 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2096 ? (5 + taill) - pref_len : PL_regeol - locinput);
2097 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2101 if (pref0_len > pref_len)
2102 pref0_len = pref_len;
2103 regprop(prop, scan);
2107 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2109 locinput - pref_len;
2110 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2111 char *s1 = do_utf8 ?
2112 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2113 pref_len - pref0_len, 60, 0) :
2114 locinput - pref_len + pref0_len;
2115 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2116 char *s2 = do_utf8 ?
2117 pv_uni_display(dsv2, (U8*)locinput,
2118 PL_regeol - locinput, 60, 0) :
2120 int len2 = do_utf8 ? strlen(s2) : l;
2121 PerlIO_printf(Perl_debug_log,
2122 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2123 (IV)(locinput - PL_bostr),
2130 (docolor ? "" : "> <"),
2134 15 - l - pref_len + 1,
2136 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2141 next = scan + NEXT_OFF(scan);
2147 if (locinput == PL_bostr || (PL_multiline &&
2148 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2150 /* regtill = regbol; */
2155 if (locinput == PL_bostr ||
2156 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2162 if (locinput == PL_bostr)
2166 if (locinput == PL_reg_ganch)
2176 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2181 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2183 if (PL_regeol - locinput > 1)
2187 if (PL_regeol != locinput)
2191 if (!nextchr && locinput >= PL_regeol)
2194 locinput += PL_utf8skip[nextchr];
2195 if (locinput > PL_regeol)
2197 nextchr = UCHARAT(locinput);
2200 nextchr = UCHARAT(++locinput);
2203 if (!nextchr && locinput >= PL_regeol)
2205 nextchr = UCHARAT(++locinput);
2208 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2211 locinput += PL_utf8skip[nextchr];
2212 if (locinput > PL_regeol)
2214 nextchr = UCHARAT(locinput);
2217 nextchr = UCHARAT(++locinput);
2222 if (do_utf8 != (UTF!=0)) {
2223 /* The target and the pattern have differing "utf8ness". */
2229 /* The target is utf8, the pattern is not utf8. */
2233 if (NATIVE_TO_UNI(*(U8*)s) !=
2234 utf8_to_uvchr((U8*)l, &len))
2241 /* The target is not utf8, the pattern is utf8. */
2245 if (NATIVE_TO_UNI(*((U8*)l)) !=
2246 utf8_to_uvchr((U8*)s, &len))
2253 nextchr = UCHARAT(locinput);
2256 /* The target and the pattern have the same "utf8ness". */
2257 /* Inline the first character, for speed. */
2258 if (UCHARAT(s) != nextchr)
2260 if (PL_regeol - locinput < ln)
2262 if (ln > 1 && memNE(s, locinput, ln))
2265 nextchr = UCHARAT(locinput);
2268 PL_reg_flags |= RF_tainted;
2278 U8 tmpbuf[UTF8_MAXLEN*2+1];
2283 toLOWER_utf8((U8*)l, tmpbuf, &ulen);
2284 if (memNE(s, (char*)tmpbuf, ulen))
2290 nextchr = UCHARAT(locinput);
2294 /* Inline the first character, for speed. */
2295 if (UCHARAT(s) != nextchr &&
2296 UCHARAT(s) != ((OP(scan) == EXACTF)
2297 ? PL_fold : PL_fold_locale)[nextchr])
2299 if (PL_regeol - locinput < ln)
2301 if (ln > 1 && (OP(scan) == EXACTF
2302 ? ibcmp(s, locinput, ln)
2303 : ibcmp_locale(s, locinput, ln)))
2306 nextchr = UCHARAT(locinput);
2310 if (!reginclass(scan, (U8*)locinput, do_utf8))
2312 if (locinput >= PL_regeol)
2314 locinput += PL_utf8skip[nextchr];
2315 nextchr = UCHARAT(locinput);
2319 nextchr = UCHARAT(locinput);
2320 if (!reginclass(scan, (U8*)locinput, do_utf8))
2322 if (!nextchr && locinput >= PL_regeol)
2324 nextchr = UCHARAT(++locinput);
2328 PL_reg_flags |= RF_tainted;
2334 LOAD_UTF8_CHARCLASS(alnum,"a");
2335 if (!(OP(scan) == ALNUM
2336 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2337 : isALNUM_LC_utf8((U8*)locinput)))
2341 locinput += PL_utf8skip[nextchr];
2342 nextchr = UCHARAT(locinput);
2345 if (!(OP(scan) == ALNUM
2346 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2348 nextchr = UCHARAT(++locinput);
2351 PL_reg_flags |= RF_tainted;
2354 if (!nextchr && locinput >= PL_regeol)
2357 LOAD_UTF8_CHARCLASS(alnum,"a");
2358 if (OP(scan) == NALNUM
2359 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2360 : isALNUM_LC_utf8((U8*)locinput))
2364 locinput += PL_utf8skip[nextchr];
2365 nextchr = UCHARAT(locinput);
2368 if (OP(scan) == NALNUM
2369 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2371 nextchr = UCHARAT(++locinput);
2375 PL_reg_flags |= RF_tainted;
2379 /* was last char in word? */
2381 if (locinput == PL_bostr)
2384 U8 *r = reghop((U8*)locinput, -1);
2386 ln = utf8n_to_uvchr(r, s - (char*)r, 0, 0);
2388 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2389 ln = isALNUM_uni(ln);
2390 LOAD_UTF8_CHARCLASS(alnum,"a");
2391 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2394 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2395 n = isALNUM_LC_utf8((U8*)locinput);
2399 ln = (locinput != PL_bostr) ?
2400 UCHARAT(locinput - 1) : '\n';
2401 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2403 n = isALNUM(nextchr);
2406 ln = isALNUM_LC(ln);
2407 n = isALNUM_LC(nextchr);
2410 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2411 OP(scan) == BOUNDL))
2415 PL_reg_flags |= RF_tainted;
2421 if (UTF8_IS_CONTINUED(nextchr)) {
2422 LOAD_UTF8_CHARCLASS(space," ");
2423 if (!(OP(scan) == SPACE
2424 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2425 : isSPACE_LC_utf8((U8*)locinput)))
2429 locinput += PL_utf8skip[nextchr];
2430 nextchr = UCHARAT(locinput);
2433 if (!(OP(scan) == SPACE
2434 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2436 nextchr = UCHARAT(++locinput);
2439 if (!(OP(scan) == SPACE
2440 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2442 nextchr = UCHARAT(++locinput);
2446 PL_reg_flags |= RF_tainted;
2449 if (!nextchr && locinput >= PL_regeol)
2452 LOAD_UTF8_CHARCLASS(space," ");
2453 if (OP(scan) == NSPACE
2454 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2455 : isSPACE_LC_utf8((U8*)locinput))
2459 locinput += PL_utf8skip[nextchr];
2460 nextchr = UCHARAT(locinput);
2463 if (OP(scan) == NSPACE
2464 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2466 nextchr = UCHARAT(++locinput);
2469 PL_reg_flags |= RF_tainted;
2475 LOAD_UTF8_CHARCLASS(digit,"0");
2476 if (!(OP(scan) == DIGIT
2477 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2478 : isDIGIT_LC_utf8((U8*)locinput)))
2482 locinput += PL_utf8skip[nextchr];
2483 nextchr = UCHARAT(locinput);
2486 if (!(OP(scan) == DIGIT
2487 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2489 nextchr = UCHARAT(++locinput);
2492 PL_reg_flags |= RF_tainted;
2495 if (!nextchr && locinput >= PL_regeol)
2498 LOAD_UTF8_CHARCLASS(digit,"0");
2499 if (OP(scan) == NDIGIT
2500 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2501 : isDIGIT_LC_utf8((U8*)locinput))
2505 locinput += PL_utf8skip[nextchr];
2506 nextchr = UCHARAT(locinput);
2509 if (OP(scan) == NDIGIT
2510 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2512 nextchr = UCHARAT(++locinput);
2515 LOAD_UTF8_CHARCLASS(mark,"~");
2516 if (locinput >= PL_regeol ||
2517 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2519 locinput += PL_utf8skip[nextchr];
2520 while (locinput < PL_regeol &&
2521 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2522 locinput += UTF8SKIP(locinput);
2523 if (locinput > PL_regeol)
2525 nextchr = UCHARAT(locinput);
2528 PL_reg_flags |= RF_tainted;
2532 n = ARG(scan); /* which paren pair */
2533 ln = PL_regstartp[n];
2534 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2535 if (*PL_reglastparen < n || ln == -1)
2536 sayNO; /* Do not match unless seen CLOSEn. */
2537 if (ln == PL_regendp[n])
2541 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2543 char *e = PL_bostr + PL_regendp[n];
2545 * Note that we can't do the "other character" lookup trick as
2546 * in the 8-bit case (no pun intended) because in Unicode we
2547 * have to map both upper and title case to lower case.
2549 if (OP(scan) == REFF) {
2550 STRLEN ulen1, ulen2;
2551 U8 tmpbuf1[UTF8_MAXLEN*2+1];
2552 U8 tmpbuf2[UTF8_MAXLEN*2+1];
2556 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2557 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2558 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2565 nextchr = UCHARAT(locinput);
2569 /* Inline the first character, for speed. */
2570 if (UCHARAT(s) != nextchr &&
2572 (UCHARAT(s) != ((OP(scan) == REFF
2573 ? PL_fold : PL_fold_locale)[nextchr]))))
2575 ln = PL_regendp[n] - ln;
2576 if (locinput + ln > PL_regeol)
2578 if (ln > 1 && (OP(scan) == REF
2579 ? memNE(s, locinput, ln)
2581 ? ibcmp(s, locinput, ln)
2582 : ibcmp_locale(s, locinput, ln))))
2585 nextchr = UCHARAT(locinput);
2596 OP_4tree *oop = PL_op;
2597 COP *ocurcop = PL_curcop;
2598 SV **ocurpad = PL_curpad;
2602 PL_op = (OP_4tree*)PL_regdata->data[n];
2603 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2604 PL_curpad = AvARRAY((AV*)PL_regdata->data[n + 2]);
2605 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2609 CALLRUNOPS(aTHX); /* Scalar context. */
2612 ret = Nullsv; /* protect against empty (?{}) blocks. */
2620 PL_curpad = ocurpad;
2621 PL_curcop = ocurcop;
2623 if (logical == 2) { /* Postponed subexpression. */
2625 MAGIC *mg = Null(MAGIC*);
2627 CHECKPOINT cp, lastcp;
2629 if(SvROK(ret) || SvRMAGICAL(ret)) {
2630 SV *sv = SvROK(ret) ? SvRV(ret) : ret;
2633 mg = mg_find(sv, PERL_MAGIC_qr);
2636 re = (regexp *)mg->mg_obj;
2637 (void)ReREFCNT_inc(re);
2641 char *t = SvPV(ret, len);
2643 char *oprecomp = PL_regprecomp;
2644 I32 osize = PL_regsize;
2645 I32 onpar = PL_regnpar;
2648 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2650 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY)))
2651 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2653 PL_regprecomp = oprecomp;
2658 PerlIO_printf(Perl_debug_log,
2659 "Entering embedded `%s%.60s%s%s'\n",
2663 (strlen(re->precomp) > 60 ? "..." : ""))
2666 state.prev = PL_reg_call_cc;
2667 state.cc = PL_regcc;
2668 state.re = PL_reg_re;
2672 cp = regcppush(0); /* Save *all* the positions. */
2675 state.ss = PL_savestack_ix;
2676 *PL_reglastparen = 0;
2677 *PL_reglastcloseparen = 0;
2678 PL_reg_call_cc = &state;
2679 PL_reginput = locinput;
2681 /* XXXX This is too dramatic a measure... */
2684 if (regmatch(re->program + 1)) {
2685 /* Even though we succeeded, we need to restore
2686 global variables, since we may be wrapped inside
2687 SUSPEND, thus the match may be not finished yet. */
2689 /* XXXX Do this only if SUSPENDed? */
2690 PL_reg_call_cc = state.prev;
2691 PL_regcc = state.cc;
2692 PL_reg_re = state.re;
2693 cache_re(PL_reg_re);
2695 /* XXXX This is too dramatic a measure... */
2698 /* These are needed even if not SUSPEND. */
2704 REGCP_UNWIND(lastcp);
2706 PL_reg_call_cc = state.prev;
2707 PL_regcc = state.cc;
2708 PL_reg_re = state.re;
2709 cache_re(PL_reg_re);
2711 /* XXXX This is too dramatic a measure... */
2721 sv_setsv(save_scalar(PL_replgv), ret);
2725 n = ARG(scan); /* which paren pair */
2726 PL_reg_start_tmp[n] = locinput;
2731 n = ARG(scan); /* which paren pair */
2732 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2733 PL_regendp[n] = locinput - PL_bostr;
2734 if (n > *PL_reglastparen)
2735 *PL_reglastparen = n;
2736 *PL_reglastcloseparen = n;
2739 n = ARG(scan); /* which paren pair */
2740 sw = (*PL_reglastparen >= n && PL_regendp[n] != -1);
2743 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2745 next = NEXTOPER(NEXTOPER(scan));
2747 next = scan + ARG(scan);
2748 if (OP(next) == IFTHEN) /* Fake one. */
2749 next = NEXTOPER(NEXTOPER(next));
2753 logical = scan->flags;
2755 /*******************************************************************
2756 PL_regcc contains infoblock about the innermost (...)* loop, and
2757 a pointer to the next outer infoblock.
2759 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2761 1) After matching X, regnode for CURLYX is processed;
2763 2) This regnode creates infoblock on the stack, and calls
2764 regmatch() recursively with the starting point at WHILEM node;
2766 3) Each hit of WHILEM node tries to match A and Z (in the order
2767 depending on the current iteration, min/max of {min,max} and
2768 greediness). The information about where are nodes for "A"
2769 and "Z" is read from the infoblock, as is info on how many times "A"
2770 was already matched, and greediness.
2772 4) After A matches, the same WHILEM node is hit again.
2774 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
2775 of the same pair. Thus when WHILEM tries to match Z, it temporarily
2776 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
2777 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
2778 of the external loop.
2780 Currently present infoblocks form a tree with a stem formed by PL_curcc
2781 and whatever it mentions via ->next, and additional attached trees
2782 corresponding to temporarily unset infoblocks as in "5" above.
2784 In the following picture infoblocks for outer loop of
2785 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
2786 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
2787 infoblocks are drawn below the "reset" infoblock.
2789 In fact in the picture below we do not show failed matches for Z and T
2790 by WHILEM blocks. [We illustrate minimal matches, since for them it is
2791 more obvious *why* one needs to *temporary* unset infoblocks.]
2793 Matched REx position InfoBlocks Comment
2797 Y A)*?Z)*?T x <- O <- I
2798 YA )*?Z)*?T x <- O <- I
2799 YA A)*?Z)*?T x <- O <- I
2800 YAA )*?Z)*?T x <- O <- I
2801 YAA Z)*?T x <- O # Temporary unset I
2804 YAAZ Y(A)*?Z)*?T x <- O
2807 YAAZY (A)*?Z)*?T x <- O
2810 YAAZY A)*?Z)*?T x <- O <- I
2813 YAAZYA )*?Z)*?T x <- O <- I
2816 YAAZYA Z)*?T x <- O # Temporary unset I
2822 YAAZYAZ T x # Temporary unset O
2829 *******************************************************************/
2832 CHECKPOINT cp = PL_savestack_ix;
2833 /* No need to save/restore up to this paren */
2834 I32 parenfloor = scan->flags;
2836 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
2838 cc.oldcc = PL_regcc;
2840 /* XXXX Probably it is better to teach regpush to support
2841 parenfloor > PL_regsize... */
2842 if (parenfloor > *PL_reglastparen)
2843 parenfloor = *PL_reglastparen; /* Pessimization... */
2844 cc.parenfloor = parenfloor;
2846 cc.min = ARG1(scan);
2847 cc.max = ARG2(scan);
2848 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2852 PL_reginput = locinput;
2853 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
2855 PL_regcc = cc.oldcc;
2861 * This is really hard to understand, because after we match
2862 * what we're trying to match, we must make sure the rest of
2863 * the REx is going to match for sure, and to do that we have
2864 * to go back UP the parse tree by recursing ever deeper. And
2865 * if it fails, we have to reset our parent's current state
2866 * that we can try again after backing off.
2869 CHECKPOINT cp, lastcp;
2870 CURCUR* cc = PL_regcc;
2871 char *lastloc = cc->lastloc; /* Detection of 0-len. */
2873 n = cc->cur + 1; /* how many we know we matched */
2874 PL_reginput = locinput;
2877 PerlIO_printf(Perl_debug_log,
2878 "%*s %ld out of %ld..%ld cc=%lx\n",
2879 REPORT_CODE_OFF+PL_regindent*2, "",
2880 (long)n, (long)cc->min,
2881 (long)cc->max, (long)cc)
2884 /* If degenerate scan matches "", assume scan done. */
2886 if (locinput == cc->lastloc && n >= cc->min) {
2887 PL_regcc = cc->oldcc;
2891 PerlIO_printf(Perl_debug_log,
2892 "%*s empty match detected, try continuation...\n",
2893 REPORT_CODE_OFF+PL_regindent*2, "")
2895 if (regmatch(cc->next))
2903 /* First just match a string of min scans. */
2907 cc->lastloc = locinput;
2908 if (regmatch(cc->scan))
2911 cc->lastloc = lastloc;
2916 /* Check whether we already were at this position.
2917 Postpone detection until we know the match is not
2918 *that* much linear. */
2919 if (!PL_reg_maxiter) {
2920 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
2921 PL_reg_leftiter = PL_reg_maxiter;
2923 if (PL_reg_leftiter-- == 0) {
2924 I32 size = (PL_reg_maxiter + 7)/8;
2925 if (PL_reg_poscache) {
2926 if (PL_reg_poscache_size < size) {
2927 Renew(PL_reg_poscache, size, char);
2928 PL_reg_poscache_size = size;
2930 Zero(PL_reg_poscache, size, char);
2933 PL_reg_poscache_size = size;
2934 Newz(29, PL_reg_poscache, size, char);
2937 PerlIO_printf(Perl_debug_log,
2938 "%sDetected a super-linear match, switching on caching%s...\n",
2939 PL_colors[4], PL_colors[5])
2942 if (PL_reg_leftiter < 0) {
2943 I32 o = locinput - PL_bostr, b;
2945 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
2948 if (PL_reg_poscache[o] & (1<<b)) {
2950 PerlIO_printf(Perl_debug_log,
2951 "%*s already tried at this position...\n",
2952 REPORT_CODE_OFF+PL_regindent*2, "")
2956 PL_reg_poscache[o] |= (1<<b);
2960 /* Prefer next over scan for minimal matching. */
2963 PL_regcc = cc->oldcc;
2966 cp = regcppush(cc->parenfloor);
2968 if (regmatch(cc->next)) {
2970 sayYES; /* All done. */
2972 REGCP_UNWIND(lastcp);
2978 if (n >= cc->max) { /* Maximum greed exceeded? */
2979 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
2980 && !(PL_reg_flags & RF_warned)) {
2981 PL_reg_flags |= RF_warned;
2982 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
2983 "Complex regular subexpression recursion",
2990 PerlIO_printf(Perl_debug_log,
2991 "%*s trying longer...\n",
2992 REPORT_CODE_OFF+PL_regindent*2, "")
2994 /* Try scanning more and see if it helps. */
2995 PL_reginput = locinput;
2997 cc->lastloc = locinput;
2998 cp = regcppush(cc->parenfloor);
3000 if (regmatch(cc->scan)) {
3004 REGCP_UNWIND(lastcp);
3007 cc->lastloc = lastloc;
3011 /* Prefer scan over next for maximal matching. */
3013 if (n < cc->max) { /* More greed allowed? */
3014 cp = regcppush(cc->parenfloor);
3016 cc->lastloc = locinput;
3018 if (regmatch(cc->scan)) {
3022 REGCP_UNWIND(lastcp);
3023 regcppop(); /* Restore some previous $<digit>s? */
3024 PL_reginput = locinput;
3026 PerlIO_printf(Perl_debug_log,
3027 "%*s failed, try continuation...\n",
3028 REPORT_CODE_OFF+PL_regindent*2, "")
3031 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3032 && !(PL_reg_flags & RF_warned)) {
3033 PL_reg_flags |= RF_warned;
3034 Perl_warner(aTHX_ WARN_REGEXP, "%s limit (%d) exceeded",
3035 "Complex regular subexpression recursion",
3039 /* Failed deeper matches of scan, so see if this one works. */
3040 PL_regcc = cc->oldcc;
3043 if (regmatch(cc->next))
3049 cc->lastloc = lastloc;
3054 next = scan + ARG(scan);
3057 inner = NEXTOPER(NEXTOPER(scan));
3060 inner = NEXTOPER(scan);
3064 if (OP(next) != c1) /* No choice. */
3065 next = inner; /* Avoid recursion. */
3067 I32 lastparen = *PL_reglastparen;
3069 re_unwind_branch_t *uw;
3071 /* Put unwinding data on stack */
3072 unwind1 = SSNEWt(1,re_unwind_branch_t);
3073 uw = SSPTRt(unwind1,re_unwind_branch_t);
3076 uw->type = ((c1 == BRANCH)
3078 : RE_UNWIND_BRANCHJ);
3079 uw->lastparen = lastparen;
3081 uw->locinput = locinput;
3082 uw->nextchr = nextchr;
3084 uw->regindent = ++PL_regindent;
3087 REGCP_SET(uw->lastcp);
3089 /* Now go into the first branch */
3102 /* We suppose that the next guy does not need
3103 backtracking: in particular, it is of constant length,
3104 and has no parenths to influence future backrefs. */
3105 ln = ARG1(scan); /* min to match */
3106 n = ARG2(scan); /* max to match */
3107 paren = scan->flags;
3109 if (paren > PL_regsize)
3111 if (paren > *PL_reglastparen)
3112 *PL_reglastparen = paren;
3114 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3116 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3117 PL_reginput = locinput;
3120 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3122 /* if we matched something zero-length we don't need to
3123 backtrack - capturing parens are already defined, so
3124 the caveat in the maximal case doesn't apply
3126 XXXX if ln == 0, we can redo this check first time
3127 through the following loop
3130 n = ln; /* don't backtrack */
3131 locinput = PL_reginput;
3132 if (HAS_TEXT(next) || JUMPABLE(next)) {
3133 regnode *text_node = next;
3135 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3137 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3139 if (PL_regkind[(U8)OP(text_node)] == REF) {
3141 n = ARG(text_node); /* which paren pair */
3142 ln = PL_regstartp[n];
3143 /* assume yes if we haven't seen CLOSEn */
3145 *PL_reglastparen < n ||
3152 c1 = *(PL_bostr + ln);
3154 else { c1 = (U8)*STRING(text_node); }
3155 if (OP(next) == EXACTF)
3157 else if (OP(text_node) == EXACTFL)
3158 c2 = PL_fold_locale[c1];
3167 /* This may be improved if l == 0. */
3168 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3169 /* If it could work, try it. */
3171 UCHARAT(PL_reginput) == c1 ||
3172 UCHARAT(PL_reginput) == c2)
3176 PL_regstartp[paren] =
3177 HOPc(PL_reginput, -l) - PL_bostr;
3178 PL_regendp[paren] = PL_reginput - PL_bostr;
3181 PL_regendp[paren] = -1;
3185 REGCP_UNWIND(lastcp);
3187 /* Couldn't or didn't -- move forward. */
3188 PL_reginput = locinput;
3189 if (regrepeat_hard(scan, 1, &l)) {
3191 locinput = PL_reginput;
3198 n = regrepeat_hard(scan, n, &l);
3199 /* if we matched something zero-length we don't need to
3200 backtrack, unless the minimum count is zero and we
3201 are capturing the result - in that case the capture
3202 being defined or not may affect later execution
3204 if (n != 0 && l == 0 && !(paren && ln == 0))
3205 ln = n; /* don't backtrack */
3206 locinput = PL_reginput;
3208 PerlIO_printf(Perl_debug_log,
3209 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3210 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3214 if (HAS_TEXT(next) || JUMPABLE(next)) {
3215 regnode *text_node = next;
3217 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3219 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3221 if (PL_regkind[(U8)OP(text_node)] == REF) {
3223 n = ARG(text_node); /* which paren pair */
3224 ln = PL_regstartp[n];
3225 /* assume yes if we haven't seen CLOSEn */
3227 *PL_reglastparen < n ||
3234 c1 = *(PL_bostr + ln);
3236 else { c1 = (U8)*STRING(text_node); }
3238 if (OP(text_node) == EXACTF)
3240 else if (OP(text_node) == EXACTFL)
3241 c2 = PL_fold_locale[c1];
3252 /* If it could work, try it. */
3254 UCHARAT(PL_reginput) == c1 ||
3255 UCHARAT(PL_reginput) == c2)
3258 PerlIO_printf(Perl_debug_log,
3259 "%*s trying tail with n=%"IVdf"...\n",
3260 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3264 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3265 PL_regendp[paren] = PL_reginput - PL_bostr;
3268 PL_regendp[paren] = -1;
3272 REGCP_UNWIND(lastcp);
3274 /* Couldn't or didn't -- back up. */
3276 locinput = HOPc(locinput, -l);
3277 PL_reginput = locinput;
3284 paren = scan->flags; /* Which paren to set */
3285 if (paren > PL_regsize)
3287 if (paren > *PL_reglastparen)
3288 *PL_reglastparen = paren;
3289 ln = ARG1(scan); /* min to match */
3290 n = ARG2(scan); /* max to match */
3291 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3295 ln = ARG1(scan); /* min to match */
3296 n = ARG2(scan); /* max to match */
3297 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3302 scan = NEXTOPER(scan);
3308 scan = NEXTOPER(scan);
3312 * Lookahead to avoid useless match attempts
3313 * when we know what character comes next.
3317 * Used to only do .*x and .*?x, but now it allows
3318 * for )'s, ('s and (?{ ... })'s to be in the way
3319 * of the quantifier and the EXACT-like node. -- japhy
3322 if (HAS_TEXT(next) || JUMPABLE(next)) {
3324 regnode *text_node = next;
3326 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3328 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3330 if (PL_regkind[(U8)OP(text_node)] == REF) {
3332 n = ARG(text_node); /* which paren pair */
3333 ln = PL_regstartp[n];
3334 /* assume yes if we haven't seen CLOSEn */
3336 *PL_reglastparen < n ||
3341 goto assume_ok_easy;
3343 s = (U8*)PL_bostr + ln;
3345 else { s = (U8*)STRING(text_node); }
3349 if (OP(text_node) == EXACTF)
3351 else if (OP(text_node) == EXACTFL)
3352 c2 = PL_fold_locale[c1];
3355 if (OP(text_node) == EXACTF) {
3356 STRLEN ulen1, ulen2;
3357 U8 tmpbuf1[UTF8_MAXLEN*2+1];
3358 U8 tmpbuf2[UTF8_MAXLEN*2+1];
3360 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3361 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3363 c1 = utf8_to_uvuni(tmpbuf1, 0);
3364 c2 = utf8_to_uvuni(tmpbuf2, 0);
3367 c2 = c1 = utf8_to_uvchr(s, NULL);
3375 PL_reginput = locinput;
3379 if (ln && regrepeat(scan, ln) < ln)
3381 locinput = PL_reginput;
3384 char *e; /* Should not check after this */
3385 char *old = locinput;
3387 if (n == REG_INFTY) {
3390 while (UTF8_IS_CONTINUATION(*(U8*)e))
3396 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3400 e = locinput + n - ln;
3406 /* Find place 'next' could work */
3409 while (locinput <= e &&
3410 UCHARAT(locinput) != c1)
3413 while (locinput <= e
3414 && UCHARAT(locinput) != c1
3415 && UCHARAT(locinput) != c2)
3418 count = locinput - old;
3425 utf8_to_uvchr((U8*)locinput, &len) != c1;
3430 for (count = 0; locinput <= e; count++) {
3431 UV c = utf8_to_uvchr((U8*)locinput, &len);
3432 if (c == c1 || c == c2)
3440 /* PL_reginput == old now */
3441 if (locinput != old) {
3442 ln = 1; /* Did some */
3443 if (regrepeat(scan, count) < count)
3446 /* PL_reginput == locinput now */
3447 TRYPAREN(paren, ln, locinput);
3448 PL_reginput = locinput; /* Could be reset... */
3449 REGCP_UNWIND(lastcp);
3450 /* Couldn't or didn't -- move forward. */
3453 locinput += UTF8SKIP(locinput);
3459 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3463 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3465 c = UCHARAT(PL_reginput);
3466 /* If it could work, try it. */
3467 if (c == c1 || c == c2)
3469 TRYPAREN(paren, n, PL_reginput);
3470 REGCP_UNWIND(lastcp);
3473 /* If it could work, try it. */
3474 else if (c1 == -1000)
3476 TRYPAREN(paren, n, PL_reginput);
3477 REGCP_UNWIND(lastcp);
3479 /* Couldn't or didn't -- move forward. */
3480 PL_reginput = locinput;
3481 if (regrepeat(scan, 1)) {
3483 locinput = PL_reginput;
3491 n = regrepeat(scan, n);
3492 locinput = PL_reginput;
3493 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3494 (!PL_multiline || OP(next) == SEOL || OP(next) == EOS)) {
3495 ln = n; /* why back off? */
3496 /* ...because $ and \Z can match before *and* after
3497 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3498 We should back off by one in this case. */
3499 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3508 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3510 c = UCHARAT(PL_reginput);
3512 /* If it could work, try it. */
3513 if (c1 == -1000 || c == c1 || c == c2)
3515 TRYPAREN(paren, n, PL_reginput);
3516 REGCP_UNWIND(lastcp);
3518 /* Couldn't or didn't -- back up. */
3520 PL_reginput = locinput = HOPc(locinput, -1);
3528 c = utf8_to_uvchr((U8*)PL_reginput, NULL);
3530 c = UCHARAT(PL_reginput);
3532 /* If it could work, try it. */
3533 if (c1 == -1000 || c == c1 || c == c2)
3535 TRYPAREN(paren, n, PL_reginput);
3536 REGCP_UNWIND(lastcp);
3538 /* Couldn't or didn't -- back up. */
3540 PL_reginput = locinput = HOPc(locinput, -1);
3547 if (PL_reg_call_cc) {
3548 re_cc_state *cur_call_cc = PL_reg_call_cc;
3549 CURCUR *cctmp = PL_regcc;
3550 regexp *re = PL_reg_re;
3551 CHECKPOINT cp, lastcp;
3553 cp = regcppush(0); /* Save *all* the positions. */
3555 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3557 PL_reginput = locinput; /* Make position available to
3559 cache_re(PL_reg_call_cc->re);
3560 PL_regcc = PL_reg_call_cc->cc;
3561 PL_reg_call_cc = PL_reg_call_cc->prev;
3562 if (regmatch(cur_call_cc->node)) {
3563 PL_reg_call_cc = cur_call_cc;
3567 REGCP_UNWIND(lastcp);
3569 PL_reg_call_cc = cur_call_cc;
3575 PerlIO_printf(Perl_debug_log,
3576 "%*s continuation failed...\n",
3577 REPORT_CODE_OFF+PL_regindent*2, "")
3581 if (locinput < PL_regtill) {
3582 DEBUG_r(PerlIO_printf(Perl_debug_log,
3583 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3585 (long)(locinput - PL_reg_starttry),
3586 (long)(PL_regtill - PL_reg_starttry),
3588 sayNO_FINAL; /* Cannot match: too short. */
3590 PL_reginput = locinput; /* put where regtry can find it */
3591 sayYES_FINAL; /* Success! */
3593 PL_reginput = locinput; /* put where regtry can find it */
3594 sayYES_LOUD; /* Success! */
3597 PL_reginput = locinput;
3602 s = HOPBACKc(locinput, scan->flags);
3608 PL_reginput = locinput;
3613 s = HOPBACKc(locinput, scan->flags);
3619 PL_reginput = locinput;
3622 inner = NEXTOPER(NEXTOPER(scan));
3623 if (regmatch(inner) != n) {
3638 if (OP(scan) == SUSPEND) {
3639 locinput = PL_reginput;
3640 nextchr = UCHARAT(locinput);
3645 next = scan + ARG(scan);
3650 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3651 PTR2UV(scan), OP(scan));
3652 Perl_croak(aTHX_ "regexp memory corruption");
3659 * We get here only if there's trouble -- normally "case END" is
3660 * the terminating point.
3662 Perl_croak(aTHX_ "corrupted regexp pointers");
3668 PerlIO_printf(Perl_debug_log,
3669 "%*s %scould match...%s\n",
3670 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3674 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3675 PL_colors[4],PL_colors[5]));
3681 #if 0 /* Breaks $^R */
3689 PerlIO_printf(Perl_debug_log,
3690 "%*s %sfailed...%s\n",
3691 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3697 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3700 case RE_UNWIND_BRANCH:
3701 case RE_UNWIND_BRANCHJ:
3703 re_unwind_branch_t *uwb = &(uw->branch);
3704 I32 lastparen = uwb->lastparen;
3706 REGCP_UNWIND(uwb->lastcp);
3707 for (n = *PL_reglastparen; n > lastparen; n--)
3709 *PL_reglastparen = n;
3710 scan = next = uwb->next;
3712 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3713 ? BRANCH : BRANCHJ) ) { /* Failure */
3720 /* Have more choice yet. Reuse the same uwb. */
3722 if ((n = (uwb->type == RE_UNWIND_BRANCH
3723 ? NEXT_OFF(next) : ARG(next))))
3726 next = NULL; /* XXXX Needn't unwinding in this case... */
3728 next = NEXTOPER(scan);
3729 if (uwb->type == RE_UNWIND_BRANCHJ)
3730 next = NEXTOPER(next);
3731 locinput = uwb->locinput;
3732 nextchr = uwb->nextchr;
3734 PL_regindent = uwb->regindent;
3741 Perl_croak(aTHX_ "regexp unwind memory corruption");
3752 - regrepeat - repeatedly match something simple, report how many
3755 * [This routine now assumes that it will only match on things of length 1.
3756 * That was true before, but now we assume scan - reginput is the count,
3757 * rather than incrementing count on every character. [Er, except utf8.]]
3760 S_regrepeat(pTHX_ regnode *p, I32 max)
3762 register char *scan;
3764 register char *loceol = PL_regeol;
3765 register I32 hardcount = 0;
3766 register bool do_utf8 = PL_reg_match_utf8;
3769 if (max != REG_INFTY && max < loceol - scan)
3770 loceol = scan + max;
3775 while (scan < loceol && hardcount < max && *scan != '\n') {
3776 scan += UTF8SKIP(scan);
3780 while (scan < loceol && *scan != '\n')
3790 case EXACT: /* length of string is 1 */
3792 while (scan < loceol && UCHARAT(scan) == c)
3795 case EXACTF: /* length of string is 1 */
3797 while (scan < loceol &&
3798 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
3801 case EXACTFL: /* length of string is 1 */
3802 PL_reg_flags |= RF_tainted;
3804 while (scan < loceol &&
3805 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
3811 while (hardcount < max && scan < loceol &&
3812 reginclass(p, (U8*)scan, do_utf8)) {
3813 scan += UTF8SKIP(scan);
3817 while (scan < loceol && reginclass(p, (U8*)scan, do_utf8))
3824 LOAD_UTF8_CHARCLASS(alnum,"a");
3825 while (hardcount < max && scan < loceol &&
3826 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3827 scan += UTF8SKIP(scan);
3831 while (scan < loceol && isALNUM(*scan))
3836 PL_reg_flags |= RF_tainted;
3839 while (hardcount < max && scan < loceol &&
3840 isALNUM_LC_utf8((U8*)scan)) {
3841 scan += UTF8SKIP(scan);
3845 while (scan < loceol && isALNUM_LC(*scan))
3852 LOAD_UTF8_CHARCLASS(alnum,"a");
3853 while (hardcount < max && scan < loceol &&
3854 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
3855 scan += UTF8SKIP(scan);
3859 while (scan < loceol && !isALNUM(*scan))
3864 PL_reg_flags |= RF_tainted;
3867 while (hardcount < max && scan < loceol &&
3868 !isALNUM_LC_utf8((U8*)scan)) {
3869 scan += UTF8SKIP(scan);
3873 while (scan < loceol && !isALNUM_LC(*scan))
3880 LOAD_UTF8_CHARCLASS(space," ");
3881 while (hardcount < max && scan < loceol &&
3883 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3884 scan += UTF8SKIP(scan);
3888 while (scan < loceol && isSPACE(*scan))
3893 PL_reg_flags |= RF_tainted;
3896 while (hardcount < max && scan < loceol &&
3897 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3898 scan += UTF8SKIP(scan);
3902 while (scan < loceol && isSPACE_LC(*scan))
3909 LOAD_UTF8_CHARCLASS(space," ");
3910 while (hardcount < max && scan < loceol &&
3912 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
3913 scan += UTF8SKIP(scan);
3917 while (scan < loceol && !isSPACE(*scan))
3922 PL_reg_flags |= RF_tainted;
3925 while (hardcount < max && scan < loceol &&
3926 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
3927 scan += UTF8SKIP(scan);
3931 while (scan < loceol && !isSPACE_LC(*scan))
3938 LOAD_UTF8_CHARCLASS(digit,"0");
3939 while (hardcount < max && scan < loceol &&
3940 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
3941 scan += UTF8SKIP(scan);
3945 while (scan < loceol && isDIGIT(*scan))
3952 LOAD_UTF8_CHARCLASS(digit,"0");
3953 while (hardcount < max && scan < loceol &&
3954 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
3955 scan += UTF8SKIP(scan);
3959 while (scan < loceol && !isDIGIT(*scan))
3963 default: /* Called on something of 0 width. */
3964 break; /* So match right here or not at all. */
3970 c = scan - PL_reginput;
3975 SV *prop = sv_newmortal();
3978 PerlIO_printf(Perl_debug_log,
3979 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
3980 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
3987 - regrepeat_hard - repeatedly match something, report total lenth and length
3989 * The repeater is supposed to have constant length.
3993 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
3995 register char *scan = Nullch;
3996 register char *start;
3997 register char *loceol = PL_regeol;
3999 I32 count = 0, res = 1;
4004 start = PL_reginput;
4005 if (PL_reg_match_utf8) {
4006 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4009 while (start < PL_reginput) {
4011 start += UTF8SKIP(start);
4022 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4024 *lp = l = PL_reginput - start;
4025 if (max != REG_INFTY && l*max < loceol - scan)
4026 loceol = scan + l*max;
4039 - regclass_swash - prepare the utf8 swash
4043 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** initsvp)
4048 if (PL_regdata && PL_regdata->count) {
4051 if (PL_regdata->what[n] == 's') {
4052 SV *rv = (SV*)PL_regdata->data[n];
4053 AV *av = (AV*)SvRV((SV*)rv);
4056 si = *av_fetch(av, 0, FALSE);
4057 a = av_fetch(av, 1, FALSE);
4061 else if (si && doinit) {
4062 sw = swash_init("utf8", "", si, 1, 0);
4063 (void)av_store(av, 1, sw);
4075 - reginclass - determine if a character falls into a character class
4079 S_reginclass(pTHX_ register regnode *n, register U8* p, register bool do_utf8)
4081 char flags = ANYOF_FLAGS(n);
4086 c = do_utf8 ? utf8_to_uvchr(p, &len) : *p;
4088 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4089 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4090 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4093 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4096 SV *sw = regclass_swash(n, TRUE, 0);
4099 if (swash_fetch(sw, p, do_utf8))
4101 else if (flags & ANYOF_FOLD) {
4103 U8 tmpbuf[UTF8_MAXLEN*2+1];
4105 toLOWER_utf8(p, tmpbuf, &ulen);
4106 if (swash_fetch(sw, tmpbuf, do_utf8))
4112 if (!match && c < 256) {
4113 if (ANYOF_BITMAP_TEST(n, c))
4115 else if (flags & ANYOF_FOLD) {
4118 if (flags & ANYOF_LOCALE) {
4119 PL_reg_flags |= RF_tainted;
4120 f = PL_fold_locale[c];
4124 if (f != c && ANYOF_BITMAP_TEST(n, f))
4128 if (!match && (flags & ANYOF_CLASS)) {
4129 PL_reg_flags |= RF_tainted;
4131 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4132 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4133 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4134 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4135 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4136 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4137 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4138 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4139 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4140 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4141 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4142 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4143 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4144 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4145 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4146 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4147 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4148 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4149 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4150 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4151 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4152 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4153 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4154 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4155 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4156 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4157 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4158 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4159 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4160 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4161 ) /* How's that for a conditional? */
4168 return (flags & ANYOF_INVERT) ? !match : match;
4172 S_reghop(pTHX_ U8 *s, I32 off)
4174 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4178 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4181 while (off-- && s < lim) {
4182 /* XXX could check well-formedness here */
4190 if (UTF8_IS_CONTINUED(*s)) {
4191 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4194 /* XXX could check well-formedness here */
4202 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4204 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4208 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4211 while (off-- && s < lim) {
4212 /* XXX could check well-formedness here */
4222 if (UTF8_IS_CONTINUED(*s)) {
4223 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4226 /* XXX could check well-formedness here */
4238 restore_pos(pTHX_ void *arg)
4240 if (PL_reg_eval_set) {
4241 if (PL_reg_oldsaved) {
4242 PL_reg_re->subbeg = PL_reg_oldsaved;
4243 PL_reg_re->sublen = PL_reg_oldsavedlen;
4244 RX_MATCH_COPIED_on(PL_reg_re);
4246 PL_reg_magic->mg_len = PL_reg_oldpos;
4247 PL_reg_eval_set = 0;
4248 PL_curpm = PL_reg_oldcurpm;