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, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
71 **** 2000, 2001, 2002, 2003, by Larry Wall and others
73 **** You may distribute under the terms of either the GNU General Public
74 **** License or the Artistic License, as specified in the README file.
76 * Beware that some of this code is subtly aware of the way operator
77 * precedence is structured in regular expressions. Serious changes in
78 * regular-expression syntax might require a total rethink.
81 #define PERL_IN_REGEXEC_C
86 #define RF_tainted 1 /* tainted information used? */
87 #define RF_warned 2 /* warned about big count? */
88 #define RF_evaled 4 /* Did an EVAL with setting? */
89 #define RF_utf8 8 /* String contains multibyte chars? */
90 #define RF_false 16 /* odd number of nested negatives */
92 #define UTF ((PL_reg_flags & RF_utf8) != 0)
94 #define RS_init 1 /* eval environment created */
95 #define RS_set 2 /* replsv value is set */
101 #define REGINCLASS(p,c) (ANYOF_FLAGS(p) ? reginclass(p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c)))
107 #define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv))
108 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
110 #define reghop_c(pos,off) ((char*)reghop((U8*)pos, off))
111 #define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off))
112 #define HOP(pos,off) (PL_reg_match_utf8 ? reghop((U8*)pos, off) : (U8*)(pos + off))
113 #define HOPMAYBE(pos,off) (PL_reg_match_utf8 ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off))
114 #define HOPc(pos,off) ((char*)HOP(pos,off))
115 #define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off))
117 #define HOPBACK(pos, off) ( \
118 (PL_reg_match_utf8) \
119 ? reghopmaybe((U8*)pos, -off) \
120 : (pos - off >= PL_bostr) \
124 #define HOPBACKc(pos, off) (char*)HOPBACK(pos, off)
126 #define reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim))
127 #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
128 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
129 #define HOPMAYBE3(pos,off,lim) (PL_reg_match_utf8 ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
130 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
131 #define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim))
133 #define LOAD_UTF8_CHARCLASS(a,b) STMT_START { if (!CAT2(PL_utf8_,a)) { ENTER; save_re_context(); (void)CAT2(is_utf8_, a)((U8*)b); LEAVE; } } STMT_END
135 /* for use after a quantifier and before an EXACT-like node -- japhy */
136 #define JUMPABLE(rn) ( \
137 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
138 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
139 OP(rn) == PLUS || OP(rn) == MINMOD || \
140 (PL_regkind[(U8)OP(rn)] == CURLY && ARG1(rn) > 0) \
143 #define HAS_TEXT(rn) ( \
144 PL_regkind[(U8)OP(rn)] == EXACT || PL_regkind[(U8)OP(rn)] == REF \
148 Search for mandatory following text node; for lookahead, the text must
149 follow but for lookbehind (rn->flags != 0) we skip to the next step.
151 #define FIND_NEXT_IMPT(rn) STMT_START { \
152 while (JUMPABLE(rn)) \
153 if (OP(rn) == SUSPEND || PL_regkind[(U8)OP(rn)] == CURLY) \
154 rn = NEXTOPER(NEXTOPER(rn)); \
155 else if (OP(rn) == PLUS) \
157 else if (OP(rn) == IFMATCH) \
158 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
159 else rn += NEXT_OFF(rn); \
162 static void restore_pos(pTHX_ void *arg);
165 S_regcppush(pTHX_ I32 parenfloor)
167 int retval = PL_savestack_ix;
168 #define REGCP_PAREN_ELEMS 4
169 int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
172 if (paren_elems_to_push < 0)
173 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
175 #define REGCP_OTHER_ELEMS 6
176 SSGROW(paren_elems_to_push + REGCP_OTHER_ELEMS);
177 for (p = PL_regsize; p > parenfloor; p--) {
178 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
179 SSPUSHINT(PL_regendp[p]);
180 SSPUSHINT(PL_regstartp[p]);
181 SSPUSHPTR(PL_reg_start_tmp[p]);
184 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
185 SSPUSHINT(PL_regsize);
186 SSPUSHINT(*PL_reglastparen);
187 SSPUSHINT(*PL_reglastcloseparen);
188 SSPUSHPTR(PL_reginput);
189 #define REGCP_FRAME_ELEMS 2
190 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
191 * are needed for the regexp context stack bookkeeping. */
192 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
193 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
198 /* These are needed since we do not localize EVAL nodes: */
199 # define REGCP_SET(cp) DEBUG_r(PerlIO_printf(Perl_debug_log, \
200 " Setting an EVAL scope, savestack=%"IVdf"\n", \
201 (IV)PL_savestack_ix)); cp = PL_savestack_ix
203 # define REGCP_UNWIND(cp) DEBUG_r(cp != PL_savestack_ix ? \
204 PerlIO_printf(Perl_debug_log, \
205 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
206 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
216 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
218 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
219 i = SSPOPINT; /* Parentheses elements to pop. */
220 input = (char *) SSPOPPTR;
221 *PL_reglastcloseparen = SSPOPINT;
222 *PL_reglastparen = SSPOPINT;
223 PL_regsize = SSPOPINT;
225 /* Now restore the parentheses context. */
226 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
227 i > 0; i -= REGCP_PAREN_ELEMS) {
228 paren = (U32)SSPOPINT;
229 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
230 PL_regstartp[paren] = SSPOPINT;
232 if (paren <= *PL_reglastparen)
233 PL_regendp[paren] = tmps;
235 PerlIO_printf(Perl_debug_log,
236 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
237 (UV)paren, (IV)PL_regstartp[paren],
238 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
239 (IV)PL_regendp[paren],
240 (paren > *PL_reglastparen ? "(no)" : ""));
244 if ((I32)(*PL_reglastparen + 1) <= PL_regnpar) {
245 PerlIO_printf(Perl_debug_log,
246 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
247 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
251 /* It would seem that the similar code in regtry()
252 * already takes care of this, and in fact it is in
253 * a better location to since this code can #if 0-ed out
254 * but the code in regtry() is needed or otherwise tests
255 * requiring null fields (pat.t#187 and split.t#{13,14}
256 * (as of patchlevel 7877) will fail. Then again,
257 * this code seems to be necessary or otherwise
258 * building DynaLoader will fail:
259 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
261 for (paren = *PL_reglastparen + 1; (I32)paren <= PL_regnpar; paren++) {
262 if ((I32)paren > PL_regsize)
263 PL_regstartp[paren] = -1;
264 PL_regendp[paren] = -1;
271 S_regcp_set_to(pTHX_ I32 ss)
273 I32 tmp = PL_savestack_ix;
275 PL_savestack_ix = ss;
277 PL_savestack_ix = tmp;
281 typedef struct re_cc_state
285 struct re_cc_state *prev;
290 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
292 #define TRYPAREN(paren, n, input) { \
295 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
296 PL_regendp[paren] = input - PL_bostr; \
299 PL_regendp[paren] = -1; \
301 if (regmatch(next)) \
304 PL_regendp[paren] = -1; \
309 * pregexec and friends
313 - pregexec - match a regexp against a string
316 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
317 char *strbeg, I32 minend, SV *screamer, U32 nosave)
318 /* strend: pointer to null at end of string */
319 /* strbeg: real beginning of string */
320 /* minend: end of match must be >=minend after stringarg. */
321 /* nosave: For optimizations. */
324 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
325 nosave ? 0 : REXEC_COPY_STR);
329 S_cache_re(pTHX_ regexp *prog)
331 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
333 PL_regprogram = prog->program;
335 PL_regnpar = prog->nparens;
336 PL_regdata = prog->data;
341 * Need to implement the following flags for reg_anch:
343 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
345 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
346 * INTUIT_AUTORITATIVE_ML
347 * INTUIT_ONCE_NOML - Intuit can match in one location only.
350 * Another flag for this function: SECOND_TIME (so that float substrs
351 * with giant delta may be not rechecked).
354 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
356 /* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
357 Otherwise, only SvCUR(sv) is used to get strbeg. */
359 /* XXXX We assume that strpos is strbeg unless sv. */
361 /* XXXX Some places assume that there is a fixed substring.
362 An update may be needed if optimizer marks as "INTUITable"
363 RExen without fixed substrings. Similarly, it is assumed that
364 lengths of all the strings are no more than minlen, thus they
365 cannot come from lookahead.
366 (Or minlen should take into account lookahead.) */
368 /* A failure to find a constant substring means that there is no need to make
369 an expensive call to REx engine, thus we celebrate a failure. Similarly,
370 finding a substring too deep into the string means that less calls to
371 regtry() should be needed.
373 REx compiler's optimizer found 4 possible hints:
374 a) Anchored substring;
376 c) Whether we are anchored (beginning-of-line or \G);
377 d) First node (of those at offset 0) which may distingush positions;
378 We use a)b)d) and multiline-part of c), and try to find a position in the
379 string which does not contradict any of them.
382 /* Most of decisions we do here should have been done at compile time.
383 The nodes of the REx which we used for the search should have been
384 deleted from the finite automaton. */
387 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
388 char *strend, U32 flags, re_scream_pos_data *data)
390 register I32 start_shift = 0;
391 /* Should be nonnegative! */
392 register I32 end_shift = 0;
397 int do_utf8 = sv ? SvUTF8(sv) : 0; /* if no sv we have to assume bytes */
399 register char *other_last = Nullch; /* other substr checked before this */
400 char *check_at = Nullch; /* check substr found at this pos */
402 char *i_strpos = strpos;
403 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
405 RX_MATCH_UTF8_set(prog,do_utf8);
407 if (prog->reganch & ROPT_UTF8) {
408 DEBUG_r(PerlIO_printf(Perl_debug_log,
409 "UTF-8 regex...\n"));
410 PL_reg_flags |= RF_utf8;
414 char *s = PL_reg_match_utf8 ?
415 sv_uni_display(dsv, sv, 60, UNI_DISPLAY_REGEX) :
417 int len = PL_reg_match_utf8 ?
418 strlen(s) : strend - strpos;
421 if (PL_reg_match_utf8)
422 DEBUG_r(PerlIO_printf(Perl_debug_log,
423 "UTF-8 target...\n"));
424 PerlIO_printf(Perl_debug_log,
425 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
426 PL_colors[4],PL_colors[5],PL_colors[0],
429 (strlen(prog->precomp) > 60 ? "..." : ""),
431 (int)(len > 60 ? 60 : len),
433 (len > 60 ? "..." : "")
437 /* CHR_DIST() would be more correct here but it makes things slow. */
438 if (prog->minlen > strend - strpos) {
439 DEBUG_r(PerlIO_printf(Perl_debug_log,
440 "String too short... [re_intuit_start]\n"));
443 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
446 if (!prog->check_utf8 && prog->check_substr)
447 to_utf8_substr(prog);
448 check = prog->check_utf8;
450 if (!prog->check_substr && prog->check_utf8)
451 to_byte_substr(prog);
452 check = prog->check_substr;
454 if (check == &PL_sv_undef) {
455 DEBUG_r(PerlIO_printf(Perl_debug_log,
456 "Non-utf string cannot match utf check string\n"));
459 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
460 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
461 || ( (prog->reganch & ROPT_ANCH_BOL)
462 && !PL_multiline ) ); /* Check after \n? */
465 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
466 | ROPT_IMPLICIT)) /* not a real BOL */
467 /* SvCUR is not set on references: SvRV and SvPVX overlap */
469 && (strpos != strbeg)) {
470 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
473 if (prog->check_offset_min == prog->check_offset_max &&
474 !(prog->reganch & ROPT_CANY_SEEN)) {
475 /* Substring at constant offset from beg-of-str... */
478 s = HOP3c(strpos, prog->check_offset_min, strend);
480 slen = SvCUR(check); /* >= 1 */
482 if ( strend - s > slen || strend - s < slen - 1
483 || (strend - s == slen && strend[-1] != '\n')) {
484 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
487 /* Now should match s[0..slen-2] */
489 if (slen && (*SvPVX(check) != *s
491 && memNE(SvPVX(check), s, slen)))) {
493 DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
497 else if (*SvPVX(check) != *s
498 || ((slen = SvCUR(check)) > 1
499 && memNE(SvPVX(check), s, slen)))
501 goto success_at_start;
504 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
506 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
507 end_shift = prog->minlen - start_shift -
508 CHR_SVLEN(check) + (SvTAIL(check) != 0);
510 I32 end = prog->check_offset_max + CHR_SVLEN(check)
511 - (SvTAIL(check) != 0);
512 I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
514 if (end_shift < eshift)
518 else { /* Can match at random position */
521 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
522 /* Should be nonnegative! */
523 end_shift = prog->minlen - start_shift -
524 CHR_SVLEN(check) + (SvTAIL(check) != 0);
527 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
529 Perl_croak(aTHX_ "panic: end_shift");
533 /* Find a possible match in the region s..strend by looking for
534 the "check" substring in the region corrected by start/end_shift. */
535 if (flags & REXEC_SCREAM) {
536 I32 p = -1; /* Internal iterator of scream. */
537 I32 *pp = data ? data->scream_pos : &p;
539 if (PL_screamfirst[BmRARE(check)] >= 0
540 || ( BmRARE(check) == '\n'
541 && (BmPREVIOUS(check) == SvCUR(check) - 1)
543 s = screaminstr(sv, check,
544 start_shift + (s - strbeg), end_shift, pp, 0);
547 /* we may be pointing at the wrong string */
548 if (s && RX_MATCH_COPIED(prog))
549 s = strbeg + (s - SvPVX(sv));
551 *data->scream_olds = s;
553 else if (prog->reganch & ROPT_CANY_SEEN)
554 s = fbm_instr((U8*)(s + start_shift),
555 (U8*)(strend - end_shift),
556 check, PL_multiline ? FBMrf_MULTILINE : 0);
558 s = fbm_instr(HOP3(s, start_shift, strend),
559 HOP3(strend, -end_shift, strbeg),
560 check, PL_multiline ? FBMrf_MULTILINE : 0);
562 /* Update the count-of-usability, remove useless subpatterns,
565 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
566 (s ? "Found" : "Did not find"),
567 (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) ? "anchored" : "floating"),
569 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
571 PL_colors[1], (SvTAIL(check) ? "$" : ""),
572 (s ? " at offset " : "...\n") ) );
579 /* Finish the diagnostic message */
580 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
582 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
583 Start with the other substr.
584 XXXX no SCREAM optimization yet - and a very coarse implementation
585 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
586 *always* match. Probably should be marked during compile...
587 Probably it is right to do no SCREAM here...
590 if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8) : (prog->float_substr && prog->anchored_substr)) {
591 /* Take into account the "other" substring. */
592 /* XXXX May be hopelessly wrong for UTF... */
595 if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
598 char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
602 t = s - prog->check_offset_max;
603 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
605 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
610 t = HOP3c(t, prog->anchored_offset, strend);
611 if (t < other_last) /* These positions already checked */
613 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
616 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
617 /* On end-of-str: see comment below. */
618 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
619 if (must == &PL_sv_undef) {
621 DEBUG_r(must = prog->anchored_utf8); /* for debug */
626 HOP3(HOP3(last1, prog->anchored_offset, strend)
627 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
629 PL_multiline ? FBMrf_MULTILINE : 0
631 DEBUG_r(PerlIO_printf(Perl_debug_log,
632 "%s anchored substr `%s%.*s%s'%s",
633 (s ? "Found" : "Contradicts"),
636 - (SvTAIL(must)!=0)),
638 PL_colors[1], (SvTAIL(must) ? "$" : "")));
640 if (last1 >= last2) {
641 DEBUG_r(PerlIO_printf(Perl_debug_log,
642 ", giving up...\n"));
645 DEBUG_r(PerlIO_printf(Perl_debug_log,
646 ", trying floating at offset %ld...\n",
647 (long)(HOP3c(s1, 1, strend) - i_strpos)));
648 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
649 s = HOP3c(last, 1, strend);
653 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
654 (long)(s - i_strpos)));
655 t = HOP3c(s, -prog->anchored_offset, strbeg);
656 other_last = HOP3c(s, 1, strend);
664 else { /* Take into account the floating substring. */
669 t = HOP3c(s, -start_shift, strbeg);
671 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
672 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
673 last = HOP3c(t, prog->float_max_offset, strend);
674 s = HOP3c(t, prog->float_min_offset, strend);
677 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
678 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
679 /* fbm_instr() takes into account exact value of end-of-str
680 if the check is SvTAIL(ed). Since false positives are OK,
681 and end-of-str is not later than strend we are OK. */
682 if (must == &PL_sv_undef) {
684 DEBUG_r(must = prog->float_utf8); /* for debug message */
687 s = fbm_instr((unsigned char*)s,
688 (unsigned char*)last + SvCUR(must)
690 must, PL_multiline ? FBMrf_MULTILINE : 0);
691 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
692 (s ? "Found" : "Contradicts"),
694 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
696 PL_colors[1], (SvTAIL(must) ? "$" : "")));
699 DEBUG_r(PerlIO_printf(Perl_debug_log,
700 ", giving up...\n"));
703 DEBUG_r(PerlIO_printf(Perl_debug_log,
704 ", trying anchored starting at offset %ld...\n",
705 (long)(s1 + 1 - i_strpos)));
707 s = HOP3c(t, 1, strend);
711 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
712 (long)(s - i_strpos)));
713 other_last = s; /* Fix this later. --Hugo */
722 t = s - prog->check_offset_max;
723 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
725 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
727 /* Fixed substring is found far enough so that the match
728 cannot start at strpos. */
730 if (ml_anch && t[-1] != '\n') {
731 /* Eventually fbm_*() should handle this, but often
732 anchored_offset is not 0, so this check will not be wasted. */
733 /* XXXX In the code below we prefer to look for "^" even in
734 presence of anchored substrings. And we search even
735 beyond the found float position. These pessimizations
736 are historical artefacts only. */
738 while (t < strend - prog->minlen) {
740 if (t < check_at - prog->check_offset_min) {
741 if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
742 /* Since we moved from the found position,
743 we definitely contradict the found anchored
744 substr. Due to the above check we do not
745 contradict "check" substr.
746 Thus we can arrive here only if check substr
747 is float. Redo checking for "other"=="fixed".
750 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
751 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
752 goto do_other_anchored;
754 /* We don't contradict the found floating substring. */
755 /* XXXX Why not check for STCLASS? */
757 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
758 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
761 /* Position contradicts check-string */
762 /* XXXX probably better to look for check-string
763 than for "\n", so one should lower the limit for t? */
764 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
765 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
766 other_last = strpos = s = t + 1;
771 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
772 PL_colors[0],PL_colors[1]));
776 DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
777 PL_colors[0],PL_colors[1]));
781 ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
784 /* The found string does not prohibit matching at strpos,
785 - no optimization of calling REx engine can be performed,
786 unless it was an MBOL and we are not after MBOL,
787 or a future STCLASS check will fail this. */
789 /* Even in this situation we may use MBOL flag if strpos is offset
790 wrt the start of the string. */
791 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
792 && (strpos != strbeg) && strpos[-1] != '\n'
793 /* May be due to an implicit anchor of m{.*foo} */
794 && !(prog->reganch & ROPT_IMPLICIT))
799 DEBUG_r( if (ml_anch)
800 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
801 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
804 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
806 prog->check_utf8 /* Could be deleted already */
807 && --BmUSEFUL(prog->check_utf8) < 0
808 && (prog->check_utf8 == prog->float_utf8)
810 prog->check_substr /* Could be deleted already */
811 && --BmUSEFUL(prog->check_substr) < 0
812 && (prog->check_substr == prog->float_substr)
815 /* If flags & SOMETHING - do not do it many times on the same match */
816 DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
817 SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr);
818 if (do_utf8 ? prog->check_substr : prog->check_utf8)
819 SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8);
820 prog->check_substr = prog->check_utf8 = Nullsv; /* disable */
821 prog->float_substr = prog->float_utf8 = Nullsv; /* clear */
822 check = Nullsv; /* abort */
824 /* XXXX This is a remnant of the old implementation. It
825 looks wasteful, since now INTUIT can use many
827 prog->reganch &= ~RE_USE_INTUIT;
834 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
835 if (prog->regstclass) {
836 /* minlen == 0 is possible if regstclass is \b or \B,
837 and the fixed substr is ''$.
838 Since minlen is already taken into account, s+1 is before strend;
839 accidentally, minlen >= 1 guaranties no false positives at s + 1
840 even for \b or \B. But (minlen? 1 : 0) below assumes that
841 regstclass does not come from lookahead... */
842 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
843 This leaves EXACTF only, which is dealt with in find_byclass(). */
844 U8* str = (U8*)STRING(prog->regstclass);
845 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
846 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
848 char *endpos = (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
849 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
850 : (prog->float_substr || prog->float_utf8
851 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
854 char *startpos = strbeg;
858 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
863 if (endpos == strend) {
864 DEBUG_r( PerlIO_printf(Perl_debug_log,
865 "Could not match STCLASS...\n") );
868 DEBUG_r( PerlIO_printf(Perl_debug_log,
869 "This position contradicts STCLASS...\n") );
870 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
872 /* Contradict one of substrings */
873 if (prog->anchored_substr || prog->anchored_utf8) {
874 if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
875 DEBUG_r( what = "anchored" );
877 s = HOP3c(t, 1, strend);
878 if (s + start_shift + end_shift > strend) {
879 /* XXXX Should be taken into account earlier? */
880 DEBUG_r( PerlIO_printf(Perl_debug_log,
881 "Could not match STCLASS...\n") );
886 DEBUG_r( PerlIO_printf(Perl_debug_log,
887 "Looking for %s substr starting at offset %ld...\n",
888 what, (long)(s + start_shift - i_strpos)) );
891 /* Have both, check_string is floating */
892 if (t + start_shift >= check_at) /* Contradicts floating=check */
893 goto retry_floating_check;
894 /* Recheck anchored substring, but not floating... */
898 DEBUG_r( PerlIO_printf(Perl_debug_log,
899 "Looking for anchored substr starting at offset %ld...\n",
900 (long)(other_last - i_strpos)) );
901 goto do_other_anchored;
903 /* Another way we could have checked stclass at the
904 current position only: */
909 DEBUG_r( PerlIO_printf(Perl_debug_log,
910 "Looking for /%s^%s/m starting at offset %ld...\n",
911 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
914 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
916 /* Check is floating subtring. */
917 retry_floating_check:
918 t = check_at - start_shift;
919 DEBUG_r( what = "floating" );
920 goto hop_and_restart;
923 DEBUG_r(PerlIO_printf(Perl_debug_log,
924 "By STCLASS: moving %ld --> %ld\n",
925 (long)(t - i_strpos), (long)(s - i_strpos))
929 DEBUG_r(PerlIO_printf(Perl_debug_log,
930 "Does not contradict STCLASS...\n");
935 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
936 PL_colors[4], (check ? "Guessed" : "Giving up"),
937 PL_colors[5], (long)(s - i_strpos)) );
940 fail_finish: /* Substring not found */
941 if (prog->check_substr || prog->check_utf8) /* could be removed already */
942 BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
944 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
945 PL_colors[4],PL_colors[5]));
949 /* We know what class REx starts with. Try to find this position... */
951 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
953 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
960 register I32 tmp = 1; /* Scratch variable? */
961 register bool do_utf8 = PL_reg_match_utf8;
963 /* We know what class it must start with. */
968 if ((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
969 !UTF8_IS_INVARIANT((U8)s[0]) ?
970 reginclass(c, (U8*)s, 0, do_utf8) :
971 REGINCLASS(c, (U8*)s)) {
972 if (tmp && (norun || regtry(prog, s)))
986 if (REGINCLASS(c, (U8*)s) ||
987 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
988 /* The assignment of 2 is intentional:
989 * for the folded sharp s, the skip is 2. */
990 (skip = SHARP_S_SKIP))) {
991 if (tmp && (norun || regtry(prog, s)))
1003 while (s < strend) {
1004 if (tmp && (norun || regtry(prog, s)))
1013 ln = STR_LEN(c); /* length to match in octets/bytes */
1014 lnc = (I32) ln; /* length to match in characters */
1016 STRLEN ulen1, ulen2;
1018 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
1019 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
1021 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1022 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1024 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN_UCLC,
1025 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1026 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN_UCLC,
1027 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1029 while (sm < ((U8 *) m + ln)) {
1044 c2 = PL_fold_locale[c1];
1046 e = HOP3c(strend, -((I32)lnc), s);
1049 e = s; /* Due to minlen logic of intuit() */
1051 /* The idea in the EXACTF* cases is to first find the
1052 * first character of the EXACTF* node and then, if
1053 * necessary, case-insensitively compare the full
1054 * text of the node. The c1 and c2 are the first
1055 * characters (though in Unicode it gets a bit
1056 * more complicated because there are more cases
1057 * than just upper and lower: one needs to use
1058 * the so-called folding case for case-insensitive
1059 * matching (called "loose matching" in Unicode).
1060 * ibcmp_utf8() will do just that. */
1064 U8 tmpbuf [UTF8_MAXLEN+1];
1065 U8 foldbuf[UTF8_MAXLEN_FOLD+1];
1066 STRLEN len, foldlen;
1069 /* Upper and lower of 1st char are equal -
1070 * probably not a "letter". */
1072 c = utf8n_to_uvchr((U8*)s, UTF8_MAXLEN, &len,
1074 0 : UTF8_ALLOW_ANY);
1077 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1078 m, (char **)0, ln, (bool)UTF))
1079 && (norun || regtry(prog, s)) )
1082 uvchr_to_utf8(tmpbuf, c);
1083 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1085 && (f == c1 || f == c2)
1086 && (ln == foldlen ||
1087 !ibcmp_utf8((char *) foldbuf,
1088 (char **)0, foldlen, do_utf8,
1090 (char **)0, ln, (bool)UTF))
1091 && (norun || regtry(prog, s)) )
1099 c = utf8n_to_uvchr((U8*)s, UTF8_MAXLEN, &len,
1101 0 : UTF8_ALLOW_ANY);
1103 /* Handle some of the three Greek sigmas cases.
1104 * Note that not all the possible combinations
1105 * are handled here: some of them are handled
1106 * by the standard folding rules, and some of
1107 * them (the character class or ANYOF cases)
1108 * are handled during compiletime in
1109 * regexec.c:S_regclass(). */
1110 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1111 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1112 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1114 if ( (c == c1 || c == c2)
1116 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1117 m, (char **)0, ln, (bool)UTF))
1118 && (norun || regtry(prog, s)) )
1121 uvchr_to_utf8(tmpbuf, c);
1122 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1124 && (f == c1 || f == c2)
1125 && (ln == foldlen ||
1126 !ibcmp_utf8((char *) foldbuf,
1127 (char **)0, foldlen, do_utf8,
1129 (char **)0, ln, (bool)UTF))
1130 && (norun || regtry(prog, s)) )
1141 && (ln == 1 || !(OP(c) == EXACTF
1143 : ibcmp_locale(s, m, ln)))
1144 && (norun || regtry(prog, s)) )
1150 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1151 && (ln == 1 || !(OP(c) == EXACTF
1153 : ibcmp_locale(s, m, ln)))
1154 && (norun || regtry(prog, s)) )
1161 PL_reg_flags |= RF_tainted;
1168 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1170 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1172 tmp = ((OP(c) == BOUND ?
1173 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1174 LOAD_UTF8_CHARCLASS(alnum,"a");
1175 while (s < strend) {
1176 if (tmp == !(OP(c) == BOUND ?
1177 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1178 isALNUM_LC_utf8((U8*)s)))
1181 if ((norun || regtry(prog, s)))
1188 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1189 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1190 while (s < strend) {
1192 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1194 if ((norun || regtry(prog, s)))
1200 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1204 PL_reg_flags |= RF_tainted;
1211 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1213 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1215 tmp = ((OP(c) == NBOUND ?
1216 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1217 LOAD_UTF8_CHARCLASS(alnum,"a");
1218 while (s < strend) {
1219 if (tmp == !(OP(c) == NBOUND ?
1220 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1221 isALNUM_LC_utf8((U8*)s)))
1223 else if ((norun || regtry(prog, s)))
1229 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1230 tmp = ((OP(c) == NBOUND ?
1231 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1232 while (s < strend) {
1234 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1236 else if ((norun || regtry(prog, s)))
1241 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1246 LOAD_UTF8_CHARCLASS(alnum,"a");
1247 while (s < strend) {
1248 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1249 if (tmp && (norun || regtry(prog, s)))
1260 while (s < strend) {
1262 if (tmp && (norun || regtry(prog, s)))
1274 PL_reg_flags |= RF_tainted;
1276 while (s < strend) {
1277 if (isALNUM_LC_utf8((U8*)s)) {
1278 if (tmp && (norun || regtry(prog, s)))
1289 while (s < strend) {
1290 if (isALNUM_LC(*s)) {
1291 if (tmp && (norun || regtry(prog, s)))
1304 LOAD_UTF8_CHARCLASS(alnum,"a");
1305 while (s < strend) {
1306 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1307 if (tmp && (norun || regtry(prog, s)))
1318 while (s < strend) {
1320 if (tmp && (norun || regtry(prog, s)))
1332 PL_reg_flags |= RF_tainted;
1334 while (s < strend) {
1335 if (!isALNUM_LC_utf8((U8*)s)) {
1336 if (tmp && (norun || regtry(prog, s)))
1347 while (s < strend) {
1348 if (!isALNUM_LC(*s)) {
1349 if (tmp && (norun || regtry(prog, s)))
1362 LOAD_UTF8_CHARCLASS(space," ");
1363 while (s < strend) {
1364 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1365 if (tmp && (norun || regtry(prog, s)))
1376 while (s < strend) {
1378 if (tmp && (norun || regtry(prog, s)))
1390 PL_reg_flags |= RF_tainted;
1392 while (s < strend) {
1393 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1394 if (tmp && (norun || regtry(prog, s)))
1405 while (s < strend) {
1406 if (isSPACE_LC(*s)) {
1407 if (tmp && (norun || regtry(prog, s)))
1420 LOAD_UTF8_CHARCLASS(space," ");
1421 while (s < strend) {
1422 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1423 if (tmp && (norun || regtry(prog, s)))
1434 while (s < strend) {
1436 if (tmp && (norun || regtry(prog, s)))
1448 PL_reg_flags |= RF_tainted;
1450 while (s < strend) {
1451 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1452 if (tmp && (norun || regtry(prog, s)))
1463 while (s < strend) {
1464 if (!isSPACE_LC(*s)) {
1465 if (tmp && (norun || regtry(prog, s)))
1478 LOAD_UTF8_CHARCLASS(digit,"0");
1479 while (s < strend) {
1480 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1481 if (tmp && (norun || regtry(prog, s)))
1492 while (s < strend) {
1494 if (tmp && (norun || regtry(prog, s)))
1506 PL_reg_flags |= RF_tainted;
1508 while (s < strend) {
1509 if (isDIGIT_LC_utf8((U8*)s)) {
1510 if (tmp && (norun || regtry(prog, s)))
1521 while (s < strend) {
1522 if (isDIGIT_LC(*s)) {
1523 if (tmp && (norun || regtry(prog, s)))
1536 LOAD_UTF8_CHARCLASS(digit,"0");
1537 while (s < strend) {
1538 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1539 if (tmp && (norun || regtry(prog, s)))
1550 while (s < strend) {
1552 if (tmp && (norun || regtry(prog, s)))
1564 PL_reg_flags |= RF_tainted;
1566 while (s < strend) {
1567 if (!isDIGIT_LC_utf8((U8*)s)) {
1568 if (tmp && (norun || regtry(prog, s)))
1579 while (s < strend) {
1580 if (!isDIGIT_LC(*s)) {
1581 if (tmp && (norun || regtry(prog, s)))
1593 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1602 - regexec_flags - match a regexp against a string
1605 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1606 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1607 /* strend: pointer to null at end of string */
1608 /* strbeg: real beginning of string */
1609 /* minend: end of match must be >=minend after stringarg. */
1610 /* data: May be used for some additional optimizations. */
1611 /* nosave: For optimizations. */
1614 register regnode *c;
1615 register char *startpos = stringarg;
1616 I32 minlen; /* must match at least this many chars */
1617 I32 dontbother = 0; /* how many characters not to try at end */
1618 /* I32 start_shift = 0; */ /* Offset of the start to find
1619 constant substr. */ /* CC */
1620 I32 end_shift = 0; /* Same for the end. */ /* CC */
1621 I32 scream_pos = -1; /* Internal iterator of scream. */
1623 SV* oreplsv = GvSV(PL_replgv);
1624 bool do_utf8 = DO_UTF8(sv);
1626 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
1627 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
1629 RX_MATCH_UTF8_set(prog,do_utf8);
1635 PL_regnarrate = DEBUG_r_TEST;
1638 /* Be paranoid... */
1639 if (prog == NULL || startpos == NULL) {
1640 Perl_croak(aTHX_ "NULL regexp parameter");
1644 minlen = prog->minlen;
1645 if (strend - startpos < minlen) {
1646 DEBUG_r(PerlIO_printf(Perl_debug_log,
1647 "String too short [regexec_flags]...\n"));
1651 /* Check validity of program. */
1652 if (UCHARAT(prog->program) != REG_MAGIC) {
1653 Perl_croak(aTHX_ "corrupted regexp program");
1657 PL_reg_eval_set = 0;
1660 if (prog->reganch & ROPT_UTF8)
1661 PL_reg_flags |= RF_utf8;
1663 /* Mark beginning of line for ^ and lookbehind. */
1664 PL_regbol = startpos;
1668 /* Mark end of line for $ (and such) */
1671 /* see how far we have to get to not match where we matched before */
1672 PL_regtill = startpos+minend;
1674 /* We start without call_cc context. */
1677 /* If there is a "must appear" string, look for it. */
1680 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1683 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1684 PL_reg_ganch = startpos;
1685 else if (sv && SvTYPE(sv) >= SVt_PVMG
1687 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1688 && mg->mg_len >= 0) {
1689 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1690 if (prog->reganch & ROPT_ANCH_GPOS) {
1691 if (s > PL_reg_ganch)
1696 else /* pos() not defined */
1697 PL_reg_ganch = strbeg;
1700 if (!(flags & REXEC_CHECKED) && (prog->check_substr != Nullsv || prog->check_utf8 != Nullsv)) {
1701 re_scream_pos_data d;
1703 d.scream_olds = &scream_olds;
1704 d.scream_pos = &scream_pos;
1705 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1707 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1708 goto phooey; /* not present */
1714 pv_uni_display(dsv0, (U8*)prog->precomp, prog->prelen, 60,
1715 UNI_DISPLAY_REGEX) :
1717 int len0 = UTF ? SvCUR(dsv0) : prog->prelen;
1718 char *s1 = do_utf8 ? sv_uni_display(dsv1, sv, 60,
1719 UNI_DISPLAY_REGEX) : startpos;
1720 int len1 = do_utf8 ? SvCUR(dsv1) : strend - startpos;
1723 PerlIO_printf(Perl_debug_log,
1724 "%sMatching REx%s `%s%*.*s%s%s' against `%s%.*s%s%s'\n",
1725 PL_colors[4],PL_colors[5],PL_colors[0],
1728 len0 > 60 ? "..." : "",
1730 (int)(len1 > 60 ? 60 : len1),
1732 (len1 > 60 ? "..." : "")
1736 /* Simplest case: anchored match need be tried only once. */
1737 /* [unless only anchor is BOL and multiline is set] */
1738 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1739 if (s == startpos && regtry(prog, startpos))
1741 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1742 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1747 dontbother = minlen - 1;
1748 end = HOP3c(strend, -dontbother, strbeg) - 1;
1749 /* for multiline we only have to try after newlines */
1750 if (prog->check_substr || prog->check_utf8) {
1754 if (regtry(prog, s))
1759 if (prog->reganch & RE_USE_INTUIT) {
1760 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1771 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1772 if (regtry(prog, s))
1779 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1780 if (regtry(prog, PL_reg_ganch))
1785 /* Messy cases: unanchored match. */
1786 if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) {
1787 /* we have /x+whatever/ */
1788 /* it must be a one character string (XXXX Except UTF?) */
1793 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1794 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1795 ch = SvPVX(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
1798 while (s < strend) {
1800 DEBUG_r( did_match = 1 );
1801 if (regtry(prog, s)) goto got_it;
1803 while (s < strend && *s == ch)
1810 while (s < strend) {
1812 DEBUG_r( did_match = 1 );
1813 if (regtry(prog, s)) goto got_it;
1815 while (s < strend && *s == ch)
1821 DEBUG_r(if (!did_match)
1822 PerlIO_printf(Perl_debug_log,
1823 "Did not find anchored character...\n")
1827 else if (prog->anchored_substr != Nullsv
1828 || prog->anchored_utf8 != Nullsv
1829 || ((prog->float_substr != Nullsv || prog->float_utf8 != Nullsv)
1830 && prog->float_max_offset < strend - s)) {
1835 char *last1; /* Last position checked before */
1839 if (prog->anchored_substr || prog->anchored_utf8) {
1840 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1841 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1842 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1843 back_max = back_min = prog->anchored_offset;
1845 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1846 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1847 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1848 back_max = prog->float_max_offset;
1849 back_min = prog->float_min_offset;
1851 if (must == &PL_sv_undef)
1852 /* could not downgrade utf8 check substring, so must fail */
1855 last = HOP3c(strend, /* Cannot start after this */
1856 -(I32)(CHR_SVLEN(must)
1857 - (SvTAIL(must) != 0) + back_min), strbeg);
1860 last1 = HOPc(s, -1);
1862 last1 = s - 1; /* bogus */
1864 /* XXXX check_substr already used to find `s', can optimize if
1865 check_substr==must. */
1867 dontbother = end_shift;
1868 strend = HOPc(strend, -dontbother);
1869 while ( (s <= last) &&
1870 ((flags & REXEC_SCREAM)
1871 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1872 end_shift, &scream_pos, 0))
1873 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1874 (unsigned char*)strend, must,
1875 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1876 /* we may be pointing at the wrong string */
1877 if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog))
1878 s = strbeg + (s - SvPVX(sv));
1879 DEBUG_r( did_match = 1 );
1880 if (HOPc(s, -back_max) > last1) {
1881 last1 = HOPc(s, -back_min);
1882 s = HOPc(s, -back_max);
1885 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1887 last1 = HOPc(s, -back_min);
1891 while (s <= last1) {
1892 if (regtry(prog, s))
1898 while (s <= last1) {
1899 if (regtry(prog, s))
1905 DEBUG_r(if (!did_match)
1906 PerlIO_printf(Perl_debug_log,
1907 "Did not find %s substr `%s%.*s%s'%s...\n",
1908 ((must == prog->anchored_substr || must == prog->anchored_utf8)
1909 ? "anchored" : "floating"),
1911 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1913 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1917 else if ((c = prog->regstclass)) {
1919 I32 op = (U8)OP(prog->regstclass);
1920 /* don't bother with what can't match */
1921 if (PL_regkind[op] != EXACT && op != CANY)
1922 strend = HOPc(strend, -(minlen - 1));
1925 SV *prop = sv_newmortal();
1933 pv_uni_display(dsv0, (U8*)SvPVX(prop), SvCUR(prop), 60,
1934 UNI_DISPLAY_REGEX) :
1936 len0 = UTF ? SvCUR(dsv0) : SvCUR(prop);
1938 sv_uni_display(dsv1, sv, 60, UNI_DISPLAY_REGEX) : s;
1939 len1 = UTF ? SvCUR(dsv1) : strend - s;
1940 PerlIO_printf(Perl_debug_log,
1941 "Matching stclass `%*.*s' against `%*.*s'\n",
1945 if (find_byclass(prog, c, s, strend, startpos, 0))
1947 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1951 if (prog->float_substr != Nullsv || prog->float_utf8 != Nullsv) {
1956 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1957 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1958 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
1960 if (flags & REXEC_SCREAM) {
1961 last = screaminstr(sv, float_real, s - strbeg,
1962 end_shift, &scream_pos, 1); /* last one */
1964 last = scream_olds; /* Only one occurrence. */
1965 /* we may be pointing at the wrong string */
1966 else if (RX_MATCH_COPIED(prog))
1967 s = strbeg + (s - SvPVX(sv));
1971 char *little = SvPV(float_real, len);
1973 if (SvTAIL(float_real)) {
1974 if (memEQ(strend - len + 1, little, len - 1))
1975 last = strend - len + 1;
1976 else if (!PL_multiline)
1977 last = memEQ(strend - len, little, len)
1978 ? strend - len : Nullch;
1984 last = rninstr(s, strend, little, little + len);
1986 last = strend; /* matching `$' */
1990 DEBUG_r(PerlIO_printf(Perl_debug_log,
1991 "%sCan't trim the tail, match fails (should not happen)%s\n",
1992 PL_colors[4],PL_colors[5]));
1993 goto phooey; /* Should not happen! */
1995 dontbother = strend - last + prog->float_min_offset;
1997 if (minlen && (dontbother < minlen))
1998 dontbother = minlen - 1;
1999 strend -= dontbother; /* this one's always in bytes! */
2000 /* We don't know much -- general case. */
2003 if (regtry(prog, s))
2012 if (regtry(prog, s))
2014 } while (s++ < strend);
2022 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
2024 if (PL_reg_eval_set) {
2025 /* Preserve the current value of $^R */
2026 if (oreplsv != GvSV(PL_replgv))
2027 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
2028 restored, the value remains
2030 restore_pos(aTHX_ 0);
2033 /* make sure $`, $&, $', and $digit will work later */
2034 if ( !(flags & REXEC_NOT_FIRST) ) {
2035 if (RX_MATCH_COPIED(prog)) {
2036 Safefree(prog->subbeg);
2037 RX_MATCH_COPIED_off(prog);
2039 if (flags & REXEC_COPY_STR) {
2040 I32 i = PL_regeol - startpos + (stringarg - strbeg);
2042 s = savepvn(strbeg, i);
2045 RX_MATCH_COPIED_on(prog);
2048 prog->subbeg = strbeg;
2049 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2056 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2057 PL_colors[4],PL_colors[5]));
2058 if (PL_reg_eval_set)
2059 restore_pos(aTHX_ 0);
2064 - regtry - try match at specific point
2066 STATIC I32 /* 0 failure, 1 success */
2067 S_regtry(pTHX_ regexp *prog, char *startpos)
2075 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
2077 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
2080 PL_reg_eval_set = RS_init;
2082 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2083 (IV)(PL_stack_sp - PL_stack_base));
2085 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
2086 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2087 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2089 /* Apparently this is not needed, judging by wantarray. */
2090 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
2091 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2094 /* Make $_ available to executed code. */
2095 if (PL_reg_sv != DEFSV) {
2096 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
2101 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
2102 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
2103 /* prepare for quick setting of pos */
2104 sv_magic(PL_reg_sv, (SV*)0,
2105 PERL_MAGIC_regex_global, Nullch, 0);
2106 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
2110 PL_reg_oldpos = mg->mg_len;
2111 SAVEDESTRUCTOR_X(restore_pos, 0);
2113 if (!PL_reg_curpm) {
2114 Newz(22,PL_reg_curpm, 1, PMOP);
2117 SV* repointer = newSViv(0);
2118 /* so we know which PL_regex_padav element is PL_reg_curpm */
2119 SvFLAGS(repointer) |= SVf_BREAK;
2120 av_push(PL_regex_padav,repointer);
2121 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2122 PL_regex_pad = AvARRAY(PL_regex_padav);
2126 PM_SETRE(PL_reg_curpm, prog);
2127 PL_reg_oldcurpm = PL_curpm;
2128 PL_curpm = PL_reg_curpm;
2129 if (RX_MATCH_COPIED(prog)) {
2130 /* Here is a serious problem: we cannot rewrite subbeg,
2131 since it may be needed if this match fails. Thus
2132 $` inside (?{}) could fail... */
2133 PL_reg_oldsaved = prog->subbeg;
2134 PL_reg_oldsavedlen = prog->sublen;
2135 RX_MATCH_COPIED_off(prog);
2138 PL_reg_oldsaved = Nullch;
2139 prog->subbeg = PL_bostr;
2140 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2142 prog->startp[0] = startpos - PL_bostr;
2143 PL_reginput = startpos;
2144 PL_regstartp = prog->startp;
2145 PL_regendp = prog->endp;
2146 PL_reglastparen = &prog->lastparen;
2147 PL_reglastcloseparen = &prog->lastcloseparen;
2148 prog->lastparen = 0;
2149 prog->lastcloseparen = 0;
2151 DEBUG_r(PL_reg_starttry = startpos);
2152 if (PL_reg_start_tmpl <= prog->nparens) {
2153 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2154 if(PL_reg_start_tmp)
2155 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2157 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2160 /* XXXX What this code is doing here?!!! There should be no need
2161 to do this again and again, PL_reglastparen should take care of
2164 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2165 * Actually, the code in regcppop() (which Ilya may be meaning by
2166 * PL_reglastparen), is not needed at all by the test suite
2167 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2168 * enough, for building DynaLoader, or otherwise this
2169 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2170 * will happen. Meanwhile, this code *is* needed for the
2171 * above-mentioned test suite tests to succeed. The common theme
2172 * on those tests seems to be returning null fields from matches.
2177 if (prog->nparens) {
2178 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2185 if (regmatch(prog->program + 1)) {
2186 prog->endp[0] = PL_reginput - PL_bostr;
2189 REGCP_UNWIND(lastcp);
2193 #define RE_UNWIND_BRANCH 1
2194 #define RE_UNWIND_BRANCHJ 2
2198 typedef struct { /* XX: makes sense to enlarge it... */
2202 } re_unwind_generic_t;
2215 } re_unwind_branch_t;
2217 typedef union re_unwind_t {
2219 re_unwind_generic_t generic;
2220 re_unwind_branch_t branch;
2223 #define sayYES goto yes
2224 #define sayNO goto no
2225 #define sayNO_ANYOF goto no_anyof
2226 #define sayYES_FINAL goto yes_final
2227 #define sayYES_LOUD goto yes_loud
2228 #define sayNO_FINAL goto no_final
2229 #define sayNO_SILENT goto do_no
2230 #define saySAME(x) if (x) goto yes; else goto no
2232 #define REPORT_CODE_OFF 24
2235 - regmatch - main matching routine
2237 * Conceptually the strategy is simple: check to see whether the current
2238 * node matches, call self recursively to see whether the rest matches,
2239 * and then act accordingly. In practice we make some effort to avoid
2240 * recursion, in particular by going through "ordinary" nodes (that don't
2241 * need to know whether the rest of the match failed) by a loop instead of
2244 /* [lwall] I've hoisted the register declarations to the outer block in order to
2245 * maybe save a little bit of pushing and popping on the stack. It also takes
2246 * advantage of machines that use a register save mask on subroutine entry.
2248 STATIC I32 /* 0 failure, 1 success */
2249 S_regmatch(pTHX_ regnode *prog)
2251 register regnode *scan; /* Current node. */
2252 regnode *next; /* Next node. */
2253 regnode *inner; /* Next node in internal branch. */
2254 register I32 nextchr; /* renamed nextchr - nextchar colides with
2255 function of same name */
2256 register I32 n; /* no or next */
2257 register I32 ln = 0; /* len or last */
2258 register char *s = Nullch; /* operand or save */
2259 register char *locinput = PL_reginput;
2260 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2261 int minmod = 0, sw = 0, logical = 0;
2264 I32 firstcp = PL_savestack_ix;
2266 register bool do_utf8 = PL_reg_match_utf8;
2268 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2269 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2270 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2277 /* Note that nextchr is a byte even in UTF */
2278 nextchr = UCHARAT(locinput);
2280 while (scan != NULL) {
2283 SV *prop = sv_newmortal();
2284 int docolor = *PL_colors[0];
2285 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2286 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2287 /* The part of the string before starttry has one color
2288 (pref0_len chars), between starttry and current
2289 position another one (pref_len - pref0_len chars),
2290 after the current position the third one.
2291 We assume that pref0_len <= pref_len, otherwise we
2292 decrease pref0_len. */
2293 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2294 ? (5 + taill) - l : locinput - PL_bostr;
2297 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2299 pref0_len = pref_len - (locinput - PL_reg_starttry);
2300 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2301 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2302 ? (5 + taill) - pref_len : PL_regeol - locinput);
2303 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2307 if (pref0_len > pref_len)
2308 pref0_len = pref_len;
2309 regprop(prop, scan);
2312 do_utf8 && OP(scan) != CANY ?
2313 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2314 pref0_len, 60, UNI_DISPLAY_REGEX) :
2315 locinput - pref_len;
2316 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2317 char *s1 = do_utf8 && OP(scan) != CANY ?
2318 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2319 pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) :
2320 locinput - pref_len + pref0_len;
2321 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2322 char *s2 = do_utf8 && OP(scan) != CANY ?
2323 pv_uni_display(dsv2, (U8*)locinput,
2324 PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
2326 int len2 = do_utf8 ? strlen(s2) : l;
2327 PerlIO_printf(Perl_debug_log,
2328 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2329 (IV)(locinput - PL_bostr),
2336 (docolor ? "" : "> <"),
2340 15 - l - pref_len + 1,
2342 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2347 next = scan + NEXT_OFF(scan);
2353 if (locinput == PL_bostr || (PL_multiline &&
2354 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2356 /* regtill = regbol; */
2361 if (locinput == PL_bostr ||
2362 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2368 if (locinput == PL_bostr)
2372 if (locinput == PL_reg_ganch)
2382 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2387 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2389 if (PL_regeol - locinput > 1)
2393 if (PL_regeol != locinput)
2397 if (!nextchr && locinput >= PL_regeol)
2400 locinput += PL_utf8skip[nextchr];
2401 if (locinput > PL_regeol)
2403 nextchr = UCHARAT(locinput);
2406 nextchr = UCHARAT(++locinput);
2409 if (!nextchr && locinput >= PL_regeol)
2411 nextchr = UCHARAT(++locinput);
2414 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2417 locinput += PL_utf8skip[nextchr];
2418 if (locinput > PL_regeol)
2420 nextchr = UCHARAT(locinput);
2423 nextchr = UCHARAT(++locinput);
2428 if (do_utf8 != UTF) {
2429 /* The target and the pattern have differing utf8ness. */
2435 /* The target is utf8, the pattern is not utf8. */
2439 if (NATIVE_TO_UNI(*(U8*)s) !=
2440 utf8n_to_uvuni((U8*)l, UTF8_MAXLEN, &ulen,
2442 0 : UTF8_ALLOW_ANY))
2449 /* The target is not utf8, the pattern is utf8. */
2453 if (NATIVE_TO_UNI(*((U8*)l)) !=
2454 utf8n_to_uvuni((U8*)s, UTF8_MAXLEN, &ulen,
2456 0 : UTF8_ALLOW_ANY))
2463 nextchr = UCHARAT(locinput);
2466 /* The target and the pattern have the same utf8ness. */
2467 /* Inline the first character, for speed. */
2468 if (UCHARAT(s) != nextchr)
2470 if (PL_regeol - locinput < ln)
2472 if (ln > 1 && memNE(s, locinput, ln))
2475 nextchr = UCHARAT(locinput);
2478 PL_reg_flags |= RF_tainted;
2484 if (do_utf8 || UTF) {
2485 /* Either target or the pattern are utf8. */
2487 char *e = PL_regeol;
2489 if (ibcmp_utf8(s, 0, ln, (bool)UTF,
2490 l, &e, 0, do_utf8)) {
2491 /* One more case for the sharp s:
2492 * pack("U0U*", 0xDF) =~ /ss/i,
2493 * the 0xC3 0x9F are the UTF-8
2494 * byte sequence for the U+00DF. */
2496 toLOWER(s[0]) == 's' &&
2498 toLOWER(s[1]) == 's' &&
2505 nextchr = UCHARAT(locinput);
2509 /* Neither the target and the pattern are utf8. */
2511 /* Inline the first character, for speed. */
2512 if (UCHARAT(s) != nextchr &&
2513 UCHARAT(s) != ((OP(scan) == EXACTF)
2514 ? PL_fold : PL_fold_locale)[nextchr])
2516 if (PL_regeol - locinput < ln)
2518 if (ln > 1 && (OP(scan) == EXACTF
2519 ? ibcmp(s, locinput, ln)
2520 : ibcmp_locale(s, locinput, ln)))
2523 nextchr = UCHARAT(locinput);
2527 STRLEN inclasslen = PL_regeol - locinput;
2529 if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8))
2531 if (locinput >= PL_regeol)
2533 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
2534 nextchr = UCHARAT(locinput);
2539 nextchr = UCHARAT(locinput);
2540 if (!REGINCLASS(scan, (U8*)locinput))
2542 if (!nextchr && locinput >= PL_regeol)
2544 nextchr = UCHARAT(++locinput);
2548 /* If we might have the case of the German sharp s
2549 * in a casefolding Unicode character class. */
2551 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
2552 locinput += SHARP_S_SKIP;
2553 nextchr = UCHARAT(locinput);
2559 PL_reg_flags |= RF_tainted;
2565 LOAD_UTF8_CHARCLASS(alnum,"a");
2566 if (!(OP(scan) == ALNUM
2567 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2568 : isALNUM_LC_utf8((U8*)locinput)))
2572 locinput += PL_utf8skip[nextchr];
2573 nextchr = UCHARAT(locinput);
2576 if (!(OP(scan) == ALNUM
2577 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2579 nextchr = UCHARAT(++locinput);
2582 PL_reg_flags |= RF_tainted;
2585 if (!nextchr && locinput >= PL_regeol)
2588 LOAD_UTF8_CHARCLASS(alnum,"a");
2589 if (OP(scan) == NALNUM
2590 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2591 : isALNUM_LC_utf8((U8*)locinput))
2595 locinput += PL_utf8skip[nextchr];
2596 nextchr = UCHARAT(locinput);
2599 if (OP(scan) == NALNUM
2600 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2602 nextchr = UCHARAT(++locinput);
2606 PL_reg_flags |= RF_tainted;
2610 /* was last char in word? */
2612 if (locinput == PL_bostr)
2615 U8 *r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
2617 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
2619 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2620 ln = isALNUM_uni(ln);
2621 LOAD_UTF8_CHARCLASS(alnum,"a");
2622 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2625 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2626 n = isALNUM_LC_utf8((U8*)locinput);
2630 ln = (locinput != PL_bostr) ?
2631 UCHARAT(locinput - 1) : '\n';
2632 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2634 n = isALNUM(nextchr);
2637 ln = isALNUM_LC(ln);
2638 n = isALNUM_LC(nextchr);
2641 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2642 OP(scan) == BOUNDL))
2646 PL_reg_flags |= RF_tainted;
2652 if (UTF8_IS_CONTINUED(nextchr)) {
2653 LOAD_UTF8_CHARCLASS(space," ");
2654 if (!(OP(scan) == SPACE
2655 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2656 : isSPACE_LC_utf8((U8*)locinput)))
2660 locinput += PL_utf8skip[nextchr];
2661 nextchr = UCHARAT(locinput);
2664 if (!(OP(scan) == SPACE
2665 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2667 nextchr = UCHARAT(++locinput);
2670 if (!(OP(scan) == SPACE
2671 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2673 nextchr = UCHARAT(++locinput);
2677 PL_reg_flags |= RF_tainted;
2680 if (!nextchr && locinput >= PL_regeol)
2683 LOAD_UTF8_CHARCLASS(space," ");
2684 if (OP(scan) == NSPACE
2685 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2686 : isSPACE_LC_utf8((U8*)locinput))
2690 locinput += PL_utf8skip[nextchr];
2691 nextchr = UCHARAT(locinput);
2694 if (OP(scan) == NSPACE
2695 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2697 nextchr = UCHARAT(++locinput);
2700 PL_reg_flags |= RF_tainted;
2706 LOAD_UTF8_CHARCLASS(digit,"0");
2707 if (!(OP(scan) == DIGIT
2708 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2709 : isDIGIT_LC_utf8((U8*)locinput)))
2713 locinput += PL_utf8skip[nextchr];
2714 nextchr = UCHARAT(locinput);
2717 if (!(OP(scan) == DIGIT
2718 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2720 nextchr = UCHARAT(++locinput);
2723 PL_reg_flags |= RF_tainted;
2726 if (!nextchr && locinput >= PL_regeol)
2729 LOAD_UTF8_CHARCLASS(digit,"0");
2730 if (OP(scan) == NDIGIT
2731 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2732 : isDIGIT_LC_utf8((U8*)locinput))
2736 locinput += PL_utf8skip[nextchr];
2737 nextchr = UCHARAT(locinput);
2740 if (OP(scan) == NDIGIT
2741 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2743 nextchr = UCHARAT(++locinput);
2746 if (locinput >= PL_regeol)
2749 LOAD_UTF8_CHARCLASS(mark,"~");
2750 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2752 locinput += PL_utf8skip[nextchr];
2753 while (locinput < PL_regeol &&
2754 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2755 locinput += UTF8SKIP(locinput);
2756 if (locinput > PL_regeol)
2761 nextchr = UCHARAT(locinput);
2764 PL_reg_flags |= RF_tainted;
2768 n = ARG(scan); /* which paren pair */
2769 ln = PL_regstartp[n];
2770 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2771 if ((I32)*PL_reglastparen < n || ln == -1)
2772 sayNO; /* Do not match unless seen CLOSEn. */
2773 if (ln == PL_regendp[n])
2777 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2779 char *e = PL_bostr + PL_regendp[n];
2781 * Note that we can't do the "other character" lookup trick as
2782 * in the 8-bit case (no pun intended) because in Unicode we
2783 * have to map both upper and title case to lower case.
2785 if (OP(scan) == REFF) {
2786 STRLEN ulen1, ulen2;
2787 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2788 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2792 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2793 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2794 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2801 nextchr = UCHARAT(locinput);
2805 /* Inline the first character, for speed. */
2806 if (UCHARAT(s) != nextchr &&
2808 (UCHARAT(s) != ((OP(scan) == REFF
2809 ? PL_fold : PL_fold_locale)[nextchr]))))
2811 ln = PL_regendp[n] - ln;
2812 if (locinput + ln > PL_regeol)
2814 if (ln > 1 && (OP(scan) == REF
2815 ? memNE(s, locinput, ln)
2817 ? ibcmp(s, locinput, ln)
2818 : ibcmp_locale(s, locinput, ln))))
2821 nextchr = UCHARAT(locinput);
2832 OP_4tree *oop = PL_op;
2833 COP *ocurcop = PL_curcop;
2836 struct regexp *oreg = PL_reg_re;
2839 PL_op = (OP_4tree*)PL_regdata->data[n];
2840 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2841 PAD_SAVE_LOCAL(old_comppad, (PAD*)PL_regdata->data[n + 2]);
2842 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2846 CALLRUNOPS(aTHX); /* Scalar context. */
2849 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
2857 PAD_RESTORE_LOCAL(old_comppad);
2858 PL_curcop = ocurcop;
2860 if (logical == 2) { /* Postponed subexpression. */
2862 MAGIC *mg = Null(MAGIC*);
2864 CHECKPOINT cp, lastcp;
2868 if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret)))
2869 mg = mg_find(sv, PERL_MAGIC_qr);
2870 else if (SvSMAGICAL(ret)) {
2871 if (SvGMAGICAL(ret))
2872 sv_unmagic(ret, PERL_MAGIC_qr);
2874 mg = mg_find(ret, PERL_MAGIC_qr);
2878 re = (regexp *)mg->mg_obj;
2879 (void)ReREFCNT_inc(re);
2883 char *t = SvPV(ret, len);
2885 char *oprecomp = PL_regprecomp;
2886 I32 osize = PL_regsize;
2887 I32 onpar = PL_regnpar;
2890 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
2891 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2893 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
2895 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2897 PL_regprecomp = oprecomp;
2902 PerlIO_printf(Perl_debug_log,
2903 "Entering embedded `%s%.60s%s%s'\n",
2907 (strlen(re->precomp) > 60 ? "..." : ""))
2910 state.prev = PL_reg_call_cc;
2911 state.cc = PL_regcc;
2912 state.re = PL_reg_re;
2916 cp = regcppush(0); /* Save *all* the positions. */
2919 state.ss = PL_savestack_ix;
2920 *PL_reglastparen = 0;
2921 *PL_reglastcloseparen = 0;
2922 PL_reg_call_cc = &state;
2923 PL_reginput = locinput;
2924 toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
2925 ((re->reganch & ROPT_UTF8) != 0);
2926 if (toggleutf) PL_reg_flags ^= RF_utf8;
2928 /* XXXX This is too dramatic a measure... */
2931 if (regmatch(re->program + 1)) {
2932 /* Even though we succeeded, we need to restore
2933 global variables, since we may be wrapped inside
2934 SUSPEND, thus the match may be not finished yet. */
2936 /* XXXX Do this only if SUSPENDed? */
2937 PL_reg_call_cc = state.prev;
2938 PL_regcc = state.cc;
2939 PL_reg_re = state.re;
2940 cache_re(PL_reg_re);
2941 if (toggleutf) PL_reg_flags ^= RF_utf8;
2943 /* XXXX This is too dramatic a measure... */
2946 /* These are needed even if not SUSPEND. */
2952 REGCP_UNWIND(lastcp);
2954 PL_reg_call_cc = state.prev;
2955 PL_regcc = state.cc;
2956 PL_reg_re = state.re;
2957 cache_re(PL_reg_re);
2958 if (toggleutf) PL_reg_flags ^= RF_utf8;
2960 /* XXXX This is too dramatic a measure... */
2970 sv_setsv(save_scalar(PL_replgv), ret);
2976 n = ARG(scan); /* which paren pair */
2977 PL_reg_start_tmp[n] = locinput;
2982 n = ARG(scan); /* which paren pair */
2983 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2984 PL_regendp[n] = locinput - PL_bostr;
2985 if (n > (I32)*PL_reglastparen)
2986 *PL_reglastparen = n;
2987 *PL_reglastcloseparen = n;
2990 n = ARG(scan); /* which paren pair */
2991 sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
2994 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2996 next = NEXTOPER(NEXTOPER(scan));
2998 next = scan + ARG(scan);
2999 if (OP(next) == IFTHEN) /* Fake one. */
3000 next = NEXTOPER(NEXTOPER(next));
3004 logical = scan->flags;
3006 /*******************************************************************
3007 PL_regcc contains infoblock about the innermost (...)* loop, and
3008 a pointer to the next outer infoblock.
3010 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
3012 1) After matching X, regnode for CURLYX is processed;
3014 2) This regnode creates infoblock on the stack, and calls
3015 regmatch() recursively with the starting point at WHILEM node;
3017 3) Each hit of WHILEM node tries to match A and Z (in the order
3018 depending on the current iteration, min/max of {min,max} and
3019 greediness). The information about where are nodes for "A"
3020 and "Z" is read from the infoblock, as is info on how many times "A"
3021 was already matched, and greediness.
3023 4) After A matches, the same WHILEM node is hit again.
3025 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
3026 of the same pair. Thus when WHILEM tries to match Z, it temporarily
3027 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
3028 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
3029 of the external loop.
3031 Currently present infoblocks form a tree with a stem formed by PL_curcc
3032 and whatever it mentions via ->next, and additional attached trees
3033 corresponding to temporarily unset infoblocks as in "5" above.
3035 In the following picture infoblocks for outer loop of
3036 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
3037 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
3038 infoblocks are drawn below the "reset" infoblock.
3040 In fact in the picture below we do not show failed matches for Z and T
3041 by WHILEM blocks. [We illustrate minimal matches, since for them it is
3042 more obvious *why* one needs to *temporary* unset infoblocks.]
3044 Matched REx position InfoBlocks Comment
3048 Y A)*?Z)*?T x <- O <- I
3049 YA )*?Z)*?T x <- O <- I
3050 YA A)*?Z)*?T x <- O <- I
3051 YAA )*?Z)*?T x <- O <- I
3052 YAA Z)*?T x <- O # Temporary unset I
3055 YAAZ Y(A)*?Z)*?T x <- O
3058 YAAZY (A)*?Z)*?T x <- O
3061 YAAZY A)*?Z)*?T x <- O <- I
3064 YAAZYA )*?Z)*?T x <- O <- I
3067 YAAZYA Z)*?T x <- O # Temporary unset I
3073 YAAZYAZ T x # Temporary unset O
3080 *******************************************************************/
3083 CHECKPOINT cp = PL_savestack_ix;
3084 /* No need to save/restore up to this paren */
3085 I32 parenfloor = scan->flags;
3087 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3089 cc.oldcc = PL_regcc;
3091 /* XXXX Probably it is better to teach regpush to support
3092 parenfloor > PL_regsize... */
3093 if (parenfloor > (I32)*PL_reglastparen)
3094 parenfloor = *PL_reglastparen; /* Pessimization... */
3095 cc.parenfloor = parenfloor;
3097 cc.min = ARG1(scan);
3098 cc.max = ARG2(scan);
3099 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3103 PL_reginput = locinput;
3104 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
3106 PL_regcc = cc.oldcc;
3112 * This is really hard to understand, because after we match
3113 * what we're trying to match, we must make sure the rest of
3114 * the REx is going to match for sure, and to do that we have
3115 * to go back UP the parse tree by recursing ever deeper. And
3116 * if it fails, we have to reset our parent's current state
3117 * that we can try again after backing off.
3120 CHECKPOINT cp, lastcp;
3121 CURCUR* cc = PL_regcc;
3122 char *lastloc = cc->lastloc; /* Detection of 0-len. */
3124 n = cc->cur + 1; /* how many we know we matched */
3125 PL_reginput = locinput;
3128 PerlIO_printf(Perl_debug_log,
3129 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3130 REPORT_CODE_OFF+PL_regindent*2, "",
3131 (long)n, (long)cc->min,
3132 (long)cc->max, PTR2UV(cc))
3135 /* If degenerate scan matches "", assume scan done. */
3137 if (locinput == cc->lastloc && n >= cc->min) {
3138 PL_regcc = cc->oldcc;
3142 PerlIO_printf(Perl_debug_log,
3143 "%*s empty match detected, try continuation...\n",
3144 REPORT_CODE_OFF+PL_regindent*2, "")
3146 if (regmatch(cc->next))
3154 /* First just match a string of min scans. */
3158 cc->lastloc = locinput;
3159 if (regmatch(cc->scan))
3162 cc->lastloc = lastloc;
3167 /* Check whether we already were at this position.
3168 Postpone detection until we know the match is not
3169 *that* much linear. */
3170 if (!PL_reg_maxiter) {
3171 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3172 PL_reg_leftiter = PL_reg_maxiter;
3174 if (PL_reg_leftiter-- == 0) {
3175 I32 size = (PL_reg_maxiter + 7)/8;
3176 if (PL_reg_poscache) {
3177 if ((I32)PL_reg_poscache_size < size) {
3178 Renew(PL_reg_poscache, size, char);
3179 PL_reg_poscache_size = size;
3181 Zero(PL_reg_poscache, size, char);
3184 PL_reg_poscache_size = size;
3185 Newz(29, PL_reg_poscache, size, char);
3188 PerlIO_printf(Perl_debug_log,
3189 "%sDetected a super-linear match, switching on caching%s...\n",
3190 PL_colors[4], PL_colors[5])
3193 if (PL_reg_leftiter < 0) {
3194 I32 o = locinput - PL_bostr, b;
3196 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3199 if (PL_reg_poscache[o] & (1<<b)) {
3201 PerlIO_printf(Perl_debug_log,
3202 "%*s already tried at this position...\n",
3203 REPORT_CODE_OFF+PL_regindent*2, "")
3205 if (PL_reg_flags & RF_false)
3210 PL_reg_poscache[o] |= (1<<b);
3214 /* Prefer next over scan for minimal matching. */
3217 PL_regcc = cc->oldcc;
3220 cp = regcppush(cc->parenfloor);
3222 if (regmatch(cc->next)) {
3224 sayYES; /* All done. */
3226 REGCP_UNWIND(lastcp);
3232 if (n >= cc->max) { /* Maximum greed exceeded? */
3233 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3234 && !(PL_reg_flags & RF_warned)) {
3235 PL_reg_flags |= RF_warned;
3236 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3237 "Complex regular subexpression recursion",
3244 PerlIO_printf(Perl_debug_log,
3245 "%*s trying longer...\n",
3246 REPORT_CODE_OFF+PL_regindent*2, "")
3248 /* Try scanning more and see if it helps. */
3249 PL_reginput = locinput;
3251 cc->lastloc = locinput;
3252 cp = regcppush(cc->parenfloor);
3254 if (regmatch(cc->scan)) {
3258 REGCP_UNWIND(lastcp);
3261 cc->lastloc = lastloc;
3265 /* Prefer scan over next for maximal matching. */
3267 if (n < cc->max) { /* More greed allowed? */
3268 cp = regcppush(cc->parenfloor);
3270 cc->lastloc = locinput;
3272 if (regmatch(cc->scan)) {
3276 REGCP_UNWIND(lastcp);
3277 regcppop(); /* Restore some previous $<digit>s? */
3278 PL_reginput = locinput;
3280 PerlIO_printf(Perl_debug_log,
3281 "%*s failed, try continuation...\n",
3282 REPORT_CODE_OFF+PL_regindent*2, "")
3285 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3286 && !(PL_reg_flags & RF_warned)) {
3287 PL_reg_flags |= RF_warned;
3288 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3289 "Complex regular subexpression recursion",
3293 /* Failed deeper matches of scan, so see if this one works. */
3294 PL_regcc = cc->oldcc;
3297 if (regmatch(cc->next))
3303 cc->lastloc = lastloc;
3308 next = scan + ARG(scan);
3311 inner = NEXTOPER(NEXTOPER(scan));
3314 inner = NEXTOPER(scan);
3318 if (OP(next) != c1) /* No choice. */
3319 next = inner; /* Avoid recursion. */
3321 I32 lastparen = *PL_reglastparen;
3323 re_unwind_branch_t *uw;
3325 /* Put unwinding data on stack */
3326 unwind1 = SSNEWt(1,re_unwind_branch_t);
3327 uw = SSPTRt(unwind1,re_unwind_branch_t);
3330 uw->type = ((c1 == BRANCH)
3332 : RE_UNWIND_BRANCHJ);
3333 uw->lastparen = lastparen;
3335 uw->locinput = locinput;
3336 uw->nextchr = nextchr;
3338 uw->regindent = ++PL_regindent;
3341 REGCP_SET(uw->lastcp);
3343 /* Now go into the first branch */
3356 /* We suppose that the next guy does not need
3357 backtracking: in particular, it is of constant length,
3358 and has no parenths to influence future backrefs. */
3359 ln = ARG1(scan); /* min to match */
3360 n = ARG2(scan); /* max to match */
3361 paren = scan->flags;
3363 if (paren > PL_regsize)
3365 if (paren > (I32)*PL_reglastparen)
3366 *PL_reglastparen = paren;
3368 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3370 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3371 PL_reginput = locinput;
3374 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3376 /* if we matched something zero-length we don't need to
3377 backtrack - capturing parens are already defined, so
3378 the caveat in the maximal case doesn't apply
3380 XXXX if ln == 0, we can redo this check first time
3381 through the following loop
3384 n = ln; /* don't backtrack */
3385 locinput = PL_reginput;
3386 if (HAS_TEXT(next) || JUMPABLE(next)) {
3387 regnode *text_node = next;
3389 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3391 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3393 if (PL_regkind[(U8)OP(text_node)] == REF) {
3397 else { c1 = (U8)*STRING(text_node); }
3398 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3400 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3401 c2 = PL_fold_locale[c1];
3410 /* This may be improved if l == 0. */
3411 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3412 /* If it could work, try it. */
3414 UCHARAT(PL_reginput) == c1 ||
3415 UCHARAT(PL_reginput) == c2)
3419 PL_regstartp[paren] =
3420 HOPc(PL_reginput, -l) - PL_bostr;
3421 PL_regendp[paren] = PL_reginput - PL_bostr;
3424 PL_regendp[paren] = -1;
3428 REGCP_UNWIND(lastcp);
3430 /* Couldn't or didn't -- move forward. */
3431 PL_reginput = locinput;
3432 if (regrepeat_hard(scan, 1, &l)) {
3434 locinput = PL_reginput;
3441 n = regrepeat_hard(scan, n, &l);
3442 /* if we matched something zero-length we don't need to
3443 backtrack, unless the minimum count is zero and we
3444 are capturing the result - in that case the capture
3445 being defined or not may affect later execution
3447 if (n != 0 && l == 0 && !(paren && ln == 0))
3448 ln = n; /* don't backtrack */
3449 locinput = PL_reginput;
3451 PerlIO_printf(Perl_debug_log,
3452 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3453 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3457 if (HAS_TEXT(next) || JUMPABLE(next)) {
3458 regnode *text_node = next;
3460 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3462 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3464 if (PL_regkind[(U8)OP(text_node)] == REF) {
3468 else { c1 = (U8)*STRING(text_node); }
3470 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3472 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3473 c2 = PL_fold_locale[c1];
3484 /* If it could work, try it. */
3486 UCHARAT(PL_reginput) == c1 ||
3487 UCHARAT(PL_reginput) == c2)
3490 PerlIO_printf(Perl_debug_log,
3491 "%*s trying tail with n=%"IVdf"...\n",
3492 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3496 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3497 PL_regendp[paren] = PL_reginput - PL_bostr;
3500 PL_regendp[paren] = -1;
3504 REGCP_UNWIND(lastcp);
3506 /* Couldn't or didn't -- back up. */
3508 locinput = HOPc(locinput, -l);
3509 PL_reginput = locinput;
3516 paren = scan->flags; /* Which paren to set */
3517 if (paren > PL_regsize)
3519 if (paren > (I32)*PL_reglastparen)
3520 *PL_reglastparen = paren;
3521 ln = ARG1(scan); /* min to match */
3522 n = ARG2(scan); /* max to match */
3523 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3527 ln = ARG1(scan); /* min to match */
3528 n = ARG2(scan); /* max to match */
3529 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3534 scan = NEXTOPER(scan);
3540 scan = NEXTOPER(scan);
3544 * Lookahead to avoid useless match attempts
3545 * when we know what character comes next.
3549 * Used to only do .*x and .*?x, but now it allows
3550 * for )'s, ('s and (?{ ... })'s to be in the way
3551 * of the quantifier and the EXACT-like node. -- japhy
3554 if (HAS_TEXT(next) || JUMPABLE(next)) {
3556 regnode *text_node = next;
3558 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3560 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3562 if (PL_regkind[(U8)OP(text_node)] == REF) {
3564 goto assume_ok_easy;
3566 else { s = (U8*)STRING(text_node); }
3570 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3572 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3573 c2 = PL_fold_locale[c1];
3576 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3577 STRLEN ulen1, ulen2;
3578 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3579 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3581 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3582 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3584 c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXLEN, 0,
3586 0 : UTF8_ALLOW_ANY);
3587 c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXLEN, 0,
3589 0 : UTF8_ALLOW_ANY);
3592 c2 = c1 = utf8n_to_uvchr(s, UTF8_MAXLEN, 0,
3594 0 : UTF8_ALLOW_ANY);
3602 PL_reginput = locinput;
3606 if (ln && regrepeat(scan, ln) < ln)
3608 locinput = PL_reginput;
3611 char *e; /* Should not check after this */
3612 char *old = locinput;
3615 if (n == REG_INFTY) {
3618 while (UTF8_IS_CONTINUATION(*(U8*)e))
3624 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3628 e = locinput + n - ln;
3633 /* Find place 'next' could work */
3636 while (locinput <= e &&
3637 UCHARAT(locinput) != c1)
3640 while (locinput <= e
3641 && UCHARAT(locinput) != c1
3642 && UCHARAT(locinput) != c2)
3645 count = locinput - old;
3650 /* count initialised to
3651 * utf8_distance(old, locinput) */
3652 while (locinput <= e &&
3653 utf8n_to_uvchr((U8*)locinput,
3656 0 : UTF8_ALLOW_ANY) != (UV)c1) {
3661 /* count initialised to
3662 * utf8_distance(old, locinput) */
3663 while (locinput <= e) {
3664 UV c = utf8n_to_uvchr((U8*)locinput,
3667 0 : UTF8_ALLOW_ANY);
3668 if (c == (UV)c1 || c == (UV)c2)
3677 /* PL_reginput == old now */
3678 if (locinput != old) {
3679 ln = 1; /* Did some */
3680 if (regrepeat(scan, count) < count)
3683 /* PL_reginput == locinput now */
3684 TRYPAREN(paren, ln, locinput);
3685 PL_reginput = locinput; /* Could be reset... */
3686 REGCP_UNWIND(lastcp);
3687 /* Couldn't or didn't -- move forward. */
3690 locinput += UTF8SKIP(locinput);
3697 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3701 c = utf8n_to_uvchr((U8*)PL_reginput,
3704 0 : UTF8_ALLOW_ANY);
3706 c = UCHARAT(PL_reginput);
3707 /* If it could work, try it. */
3708 if (c == (UV)c1 || c == (UV)c2)
3710 TRYPAREN(paren, ln, PL_reginput);
3711 REGCP_UNWIND(lastcp);
3714 /* If it could work, try it. */
3715 else if (c1 == -1000)
3717 TRYPAREN(paren, ln, PL_reginput);
3718 REGCP_UNWIND(lastcp);
3720 /* Couldn't or didn't -- move forward. */
3721 PL_reginput = locinput;
3722 if (regrepeat(scan, 1)) {
3724 locinput = PL_reginput;
3732 n = regrepeat(scan, n);
3733 locinput = PL_reginput;
3734 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3735 ((!PL_multiline && OP(next) != MEOL) ||
3736 OP(next) == SEOL || OP(next) == EOS))
3738 ln = n; /* why back off? */
3739 /* ...because $ and \Z can match before *and* after
3740 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3741 We should back off by one in this case. */
3742 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3751 c = utf8n_to_uvchr((U8*)PL_reginput,
3754 0 : UTF8_ALLOW_ANY);
3756 c = UCHARAT(PL_reginput);
3758 /* If it could work, try it. */
3759 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3761 TRYPAREN(paren, n, PL_reginput);
3762 REGCP_UNWIND(lastcp);
3764 /* Couldn't or didn't -- back up. */
3766 PL_reginput = locinput = HOPc(locinput, -1);
3774 c = utf8n_to_uvchr((U8*)PL_reginput,
3777 0 : UTF8_ALLOW_ANY);
3779 c = UCHARAT(PL_reginput);
3781 /* If it could work, try it. */
3782 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3784 TRYPAREN(paren, n, PL_reginput);
3785 REGCP_UNWIND(lastcp);
3787 /* Couldn't or didn't -- back up. */
3789 PL_reginput = locinput = HOPc(locinput, -1);
3796 if (PL_reg_call_cc) {
3797 re_cc_state *cur_call_cc = PL_reg_call_cc;
3798 CURCUR *cctmp = PL_regcc;
3799 regexp *re = PL_reg_re;
3800 CHECKPOINT cp, lastcp;
3802 cp = regcppush(0); /* Save *all* the positions. */
3804 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3806 PL_reginput = locinput; /* Make position available to
3808 cache_re(PL_reg_call_cc->re);
3809 PL_regcc = PL_reg_call_cc->cc;
3810 PL_reg_call_cc = PL_reg_call_cc->prev;
3811 if (regmatch(cur_call_cc->node)) {
3812 PL_reg_call_cc = cur_call_cc;
3816 REGCP_UNWIND(lastcp);
3818 PL_reg_call_cc = cur_call_cc;
3824 PerlIO_printf(Perl_debug_log,
3825 "%*s continuation failed...\n",
3826 REPORT_CODE_OFF+PL_regindent*2, "")
3830 if (locinput < PL_regtill) {
3831 DEBUG_r(PerlIO_printf(Perl_debug_log,
3832 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3834 (long)(locinput - PL_reg_starttry),
3835 (long)(PL_regtill - PL_reg_starttry),
3837 sayNO_FINAL; /* Cannot match: too short. */
3839 PL_reginput = locinput; /* put where regtry can find it */
3840 sayYES_FINAL; /* Success! */
3842 PL_reginput = locinput; /* put where regtry can find it */
3843 sayYES_LOUD; /* Success! */
3846 PL_reginput = locinput;
3851 s = HOPBACKc(locinput, scan->flags);
3857 PL_reginput = locinput;
3858 PL_reg_flags ^= RF_false;
3863 s = HOPBACKc(locinput, scan->flags);
3869 PL_reginput = locinput;
3872 inner = NEXTOPER(NEXTOPER(scan));
3873 if (regmatch(inner) != n) {
3875 PL_reg_flags ^= RF_false;
3886 PL_reg_flags ^= RF_false;
3892 if (OP(scan) == SUSPEND) {
3893 locinput = PL_reginput;
3894 nextchr = UCHARAT(locinput);
3899 next = scan + ARG(scan);
3904 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3905 PTR2UV(scan), OP(scan));
3906 Perl_croak(aTHX_ "regexp memory corruption");
3913 * We get here only if there's trouble -- normally "case END" is
3914 * the terminating point.
3916 Perl_croak(aTHX_ "corrupted regexp pointers");
3922 PerlIO_printf(Perl_debug_log,
3923 "%*s %scould match...%s\n",
3924 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3928 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3929 PL_colors[4],PL_colors[5]));
3935 #if 0 /* Breaks $^R */
3943 PerlIO_printf(Perl_debug_log,
3944 "%*s %sfailed...%s\n",
3945 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3951 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3954 case RE_UNWIND_BRANCH:
3955 case RE_UNWIND_BRANCHJ:
3957 re_unwind_branch_t *uwb = &(uw->branch);
3958 I32 lastparen = uwb->lastparen;
3960 REGCP_UNWIND(uwb->lastcp);
3961 for (n = *PL_reglastparen; n > lastparen; n--)
3963 *PL_reglastparen = n;
3964 scan = next = uwb->next;
3966 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3967 ? BRANCH : BRANCHJ) ) { /* Failure */
3974 /* Have more choice yet. Reuse the same uwb. */
3976 if ((n = (uwb->type == RE_UNWIND_BRANCH
3977 ? NEXT_OFF(next) : ARG(next))))
3980 next = NULL; /* XXXX Needn't unwinding in this case... */
3982 next = NEXTOPER(scan);
3983 if (uwb->type == RE_UNWIND_BRANCHJ)
3984 next = NEXTOPER(next);
3985 locinput = uwb->locinput;
3986 nextchr = uwb->nextchr;
3988 PL_regindent = uwb->regindent;
3995 Perl_croak(aTHX_ "regexp unwind memory corruption");
4006 - regrepeat - repeatedly match something simple, report how many
4009 * [This routine now assumes that it will only match on things of length 1.
4010 * That was true before, but now we assume scan - reginput is the count,
4011 * rather than incrementing count on every character. [Er, except utf8.]]
4014 S_regrepeat(pTHX_ regnode *p, I32 max)
4016 register char *scan;
4018 register char *loceol = PL_regeol;
4019 register I32 hardcount = 0;
4020 register bool do_utf8 = PL_reg_match_utf8;
4023 if (max == REG_INFTY)
4025 else if (max < loceol - scan)
4026 loceol = scan + max;
4031 while (scan < loceol && hardcount < max && *scan != '\n') {
4032 scan += UTF8SKIP(scan);
4036 while (scan < loceol && *scan != '\n')
4043 while (scan < loceol && hardcount < max) {
4044 scan += UTF8SKIP(scan);
4054 case EXACT: /* length of string is 1 */
4056 while (scan < loceol && UCHARAT(scan) == c)
4059 case EXACTF: /* length of string is 1 */
4061 while (scan < loceol &&
4062 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
4065 case EXACTFL: /* length of string is 1 */
4066 PL_reg_flags |= RF_tainted;
4068 while (scan < loceol &&
4069 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
4075 while (hardcount < max && scan < loceol &&
4076 reginclass(p, (U8*)scan, 0, do_utf8)) {
4077 scan += UTF8SKIP(scan);
4081 while (scan < loceol && REGINCLASS(p, (U8*)scan))
4088 LOAD_UTF8_CHARCLASS(alnum,"a");
4089 while (hardcount < max && scan < loceol &&
4090 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4091 scan += UTF8SKIP(scan);
4095 while (scan < loceol && isALNUM(*scan))
4100 PL_reg_flags |= RF_tainted;
4103 while (hardcount < max && scan < loceol &&
4104 isALNUM_LC_utf8((U8*)scan)) {
4105 scan += UTF8SKIP(scan);
4109 while (scan < loceol && isALNUM_LC(*scan))
4116 LOAD_UTF8_CHARCLASS(alnum,"a");
4117 while (hardcount < max && scan < loceol &&
4118 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4119 scan += UTF8SKIP(scan);
4123 while (scan < loceol && !isALNUM(*scan))
4128 PL_reg_flags |= RF_tainted;
4131 while (hardcount < max && scan < loceol &&
4132 !isALNUM_LC_utf8((U8*)scan)) {
4133 scan += UTF8SKIP(scan);
4137 while (scan < loceol && !isALNUM_LC(*scan))
4144 LOAD_UTF8_CHARCLASS(space," ");
4145 while (hardcount < max && scan < loceol &&
4147 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4148 scan += UTF8SKIP(scan);
4152 while (scan < loceol && isSPACE(*scan))
4157 PL_reg_flags |= RF_tainted;
4160 while (hardcount < max && scan < loceol &&
4161 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4162 scan += UTF8SKIP(scan);
4166 while (scan < loceol && isSPACE_LC(*scan))
4173 LOAD_UTF8_CHARCLASS(space," ");
4174 while (hardcount < max && scan < loceol &&
4176 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4177 scan += UTF8SKIP(scan);
4181 while (scan < loceol && !isSPACE(*scan))
4186 PL_reg_flags |= RF_tainted;
4189 while (hardcount < max && scan < loceol &&
4190 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4191 scan += UTF8SKIP(scan);
4195 while (scan < loceol && !isSPACE_LC(*scan))
4202 LOAD_UTF8_CHARCLASS(digit,"0");
4203 while (hardcount < max && scan < loceol &&
4204 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4205 scan += UTF8SKIP(scan);
4209 while (scan < loceol && isDIGIT(*scan))
4216 LOAD_UTF8_CHARCLASS(digit,"0");
4217 while (hardcount < max && scan < loceol &&
4218 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4219 scan += UTF8SKIP(scan);
4223 while (scan < loceol && !isDIGIT(*scan))
4227 default: /* Called on something of 0 width. */
4228 break; /* So match right here or not at all. */
4234 c = scan - PL_reginput;
4239 SV *prop = sv_newmortal();
4242 PerlIO_printf(Perl_debug_log,
4243 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4244 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4251 - regrepeat_hard - repeatedly match something, report total lenth and length
4253 * The repeater is supposed to have constant length.
4257 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4259 register char *scan = Nullch;
4260 register char *start;
4261 register char *loceol = PL_regeol;
4263 I32 count = 0, res = 1;
4268 start = PL_reginput;
4269 if (PL_reg_match_utf8) {
4270 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4273 while (start < PL_reginput) {
4275 start += UTF8SKIP(start);
4286 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4288 *lp = l = PL_reginput - start;
4289 if (max != REG_INFTY && l*max < loceol - scan)
4290 loceol = scan + l*max;
4303 - regclass_swash - prepare the utf8 swash
4307 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** listsvp, SV **altsvp)
4313 if (PL_regdata && PL_regdata->count) {
4316 if (PL_regdata->what[n] == 's') {
4317 SV *rv = (SV*)PL_regdata->data[n];
4318 AV *av = (AV*)SvRV((SV*)rv);
4319 SV **ary = AvARRAY(av);
4322 /* See the end of regcomp.c:S_reglass() for
4323 * documentation of these array elements. */
4326 a = SvTYPE(ary[1]) == SVt_RV ? &ary[1] : 0;
4327 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
4331 else if (si && doinit) {
4332 sw = swash_init("utf8", "", si, 1, 0);
4333 (void)av_store(av, 1, sw);
4349 - reginclass - determine if a character falls into a character class
4351 The n is the ANYOF regnode, the p is the target string, lenp
4352 is pointer to the maximum length of how far to go in the p
4353 (if the lenp is zero, UTF8SKIP(p) is used),
4354 do_utf8 tells whether the target string is in UTF-8.
4359 S_reginclass(pTHX_ register regnode *n, register U8* p, STRLEN* lenp, register bool do_utf8)
4361 char flags = ANYOF_FLAGS(n);
4367 if (do_utf8 && !UTF8_IS_INVARIANT(c))
4368 c = utf8n_to_uvchr(p, UTF8_MAXLEN, &len,
4369 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
4371 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
4372 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4375 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4376 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4379 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4383 SV *sw = regclass_swash(n, TRUE, 0, (SV**)&av);
4386 if (swash_fetch(sw, p, do_utf8))
4388 else if (flags & ANYOF_FOLD) {
4389 if (!match && lenp && av) {
4392 for (i = 0; i <= av_len(av); i++) {
4393 SV* sv = *av_fetch(av, i, FALSE);
4395 char *s = SvPV(sv, len);
4397 if (len <= plen && memEQ(s, (char*)p, len)) {
4405 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
4408 to_utf8_fold(p, tmpbuf, &tmplen);
4409 if (swash_fetch(sw, tmpbuf, do_utf8))
4415 if (match && lenp && *lenp == 0)
4416 *lenp = UNISKIP(NATIVE_TO_UNI(c));
4418 if (!match && c < 256) {
4419 if (ANYOF_BITMAP_TEST(n, c))
4421 else if (flags & ANYOF_FOLD) {
4424 if (flags & ANYOF_LOCALE) {
4425 PL_reg_flags |= RF_tainted;
4426 f = PL_fold_locale[c];
4430 if (f != c && ANYOF_BITMAP_TEST(n, f))
4434 if (!match && (flags & ANYOF_CLASS)) {
4435 PL_reg_flags |= RF_tainted;
4437 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4438 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4439 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4440 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4441 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4442 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4443 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4444 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4445 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4446 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4447 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4448 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4449 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4450 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4451 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4452 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4453 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4454 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4455 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4456 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4457 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4458 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4459 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4460 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4461 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4462 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4463 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4464 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4465 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4466 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4467 ) /* How's that for a conditional? */
4474 return (flags & ANYOF_INVERT) ? !match : match;
4478 S_reghop(pTHX_ U8 *s, I32 off)
4480 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4484 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4487 while (off-- && s < lim) {
4488 /* XXX could check well-formedness here */
4496 if (UTF8_IS_CONTINUED(*s)) {
4497 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4500 /* XXX could check well-formedness here */
4508 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4510 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4514 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4517 while (off-- && s < lim) {
4518 /* XXX could check well-formedness here */
4528 if (UTF8_IS_CONTINUED(*s)) {
4529 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4532 /* XXX could check well-formedness here */
4544 restore_pos(pTHX_ void *arg)
4546 if (PL_reg_eval_set) {
4547 if (PL_reg_oldsaved) {
4548 PL_reg_re->subbeg = PL_reg_oldsaved;
4549 PL_reg_re->sublen = PL_reg_oldsavedlen;
4550 RX_MATCH_COPIED_on(PL_reg_re);
4552 PL_reg_magic->mg_len = PL_reg_oldpos;
4553 PL_reg_eval_set = 0;
4554 PL_curpm = PL_reg_oldcurpm;
4559 S_to_utf8_substr(pTHX_ register regexp *prog)
4562 if (prog->float_substr && !prog->float_utf8) {
4563 prog->float_utf8 = sv = NEWSV(117, 0);
4564 SvSetSV(sv, prog->float_substr);
4565 sv_utf8_upgrade(sv);
4566 if (SvTAIL(prog->float_substr))
4568 if (prog->float_substr == prog->check_substr)
4569 prog->check_utf8 = sv;
4571 if (prog->anchored_substr && !prog->anchored_utf8) {
4572 prog->anchored_utf8 = sv = NEWSV(118, 0);
4573 SvSetSV(sv, prog->anchored_substr);
4574 sv_utf8_upgrade(sv);
4575 if (SvTAIL(prog->anchored_substr))
4577 if (prog->anchored_substr == prog->check_substr)
4578 prog->check_utf8 = sv;
4583 S_to_byte_substr(pTHX_ register regexp *prog)
4586 if (prog->float_utf8 && !prog->float_substr) {
4587 prog->float_substr = sv = NEWSV(117, 0);
4588 SvSetSV(sv, prog->float_utf8);
4589 if (sv_utf8_downgrade(sv, TRUE)) {
4590 if (SvTAIL(prog->float_utf8))
4594 prog->float_substr = sv = &PL_sv_undef;
4596 if (prog->float_utf8 == prog->check_utf8)
4597 prog->check_substr = sv;
4599 if (prog->anchored_utf8 && !prog->anchored_substr) {
4600 prog->anchored_substr = sv = NEWSV(118, 0);
4601 SvSetSV(sv, prog->anchored_utf8);
4602 if (sv_utf8_downgrade(sv, TRUE)) {
4603 if (SvTAIL(prog->anchored_utf8))
4607 prog->anchored_substr = sv = &PL_sv_undef;
4609 if (prog->anchored_utf8 == prog->check_utf8)
4610 prog->check_substr = sv;