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;
959 register I32 tmp = 1; /* Scratch variable? */
960 register bool do_utf8 = PL_reg_match_utf8;
962 /* We know what class it must start with. */
967 if ((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
968 !UTF8_IS_INVARIANT((U8)s[0]) ?
969 reginclass(c, (U8*)s, 0, do_utf8) :
970 REGINCLASS(c, (U8*)s)) {
971 if (tmp && (norun || regtry(prog, s)))
985 if (REGINCLASS(c, (U8*)s) ||
986 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
987 /* The assignment of 2 is intentional:
988 * for the folded sharp s, the skip is 2. */
989 (skip = SHARP_S_SKIP))) {
990 if (tmp && (norun || regtry(prog, s)))
1002 while (s < strend) {
1003 if (tmp && (norun || regtry(prog, s)))
1014 STRLEN ulen1, ulen2;
1015 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
1016 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
1018 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1019 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1021 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN_UCLC,
1022 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1023 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN_UCLC,
1024 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1035 c2 = PL_fold_locale[c1];
1037 /* The last byte to try is ln-1 characters before strend
1038 * since the strend points one byte past the string. */
1039 e = HOP3c(strend, (I32)1 - (I32)ln, s);
1042 e = s; /* Due to minlen logic of intuit() */
1044 /* The idea in the EXACTF* cases is to first find the
1045 * first character of the EXACTF* node and then, if
1046 * necessary, case-insensitively compare the full
1047 * text of the node. The c1 and c2 are the first
1048 * characters (though in Unicode it gets a bit
1049 * more complicated because there are more cases
1050 * than just upper and lower: one needs to use
1051 * the so-called folding case for case-insensitive
1052 * matching (called "loose matching" in Unicode).
1053 * ibcmp_utf8() will do just that. */
1057 U8 tmpbuf [UTF8_MAXLEN+1];
1058 U8 foldbuf[UTF8_MAXLEN_FOLD+1];
1059 STRLEN len, foldlen;
1063 c = utf8n_to_uvchr((U8*)s, UTF8_MAXLEN, &len,
1065 0 : UTF8_ALLOW_ANY);
1068 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1069 m, (char **)0, ln, (bool)UTF))
1070 && (norun || regtry(prog, s)) )
1073 uvchr_to_utf8(tmpbuf, c);
1074 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1076 && (f == c1 || f == c2)
1077 && (ln == foldlen ||
1078 !ibcmp_utf8((char *) foldbuf,
1079 (char **)0, foldlen, do_utf8,
1081 (char **)0, ln, (bool)UTF))
1082 && (norun || regtry(prog, s)) )
1090 c = utf8n_to_uvchr((U8*)s, UTF8_MAXLEN, &len,
1092 0 : UTF8_ALLOW_ANY);
1094 /* Handle some of the three Greek sigmas cases.
1095 * Note that not all the possible combinations
1096 * are handled here: some of them are handled
1097 * by the standard folding rules, and some of
1098 * them (the character class or ANYOF cases)
1099 * are handled during compiletime in
1100 * regexec.c:S_regclass(). */
1101 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1102 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1103 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1105 if ( (c == c1 || c == c2)
1107 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1108 m, (char **)0, ln, (bool)UTF))
1109 && (norun || regtry(prog, s)) )
1112 uvchr_to_utf8(tmpbuf, c);
1113 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1115 && (f == c1 || f == c2)
1116 && (ln == foldlen ||
1117 !ibcmp_utf8((char *) foldbuf,
1118 (char **)0, foldlen, do_utf8,
1120 (char **)0, ln, (bool)UTF))
1121 && (norun || regtry(prog, s)) )
1132 && (ln == 1 || !(OP(c) == EXACTF
1134 : ibcmp_locale(s, m, ln)))
1135 && (norun || regtry(prog, s)) )
1141 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1142 && (ln == 1 || !(OP(c) == EXACTF
1144 : ibcmp_locale(s, m, ln)))
1145 && (norun || regtry(prog, s)) )
1152 PL_reg_flags |= RF_tainted;
1159 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1161 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1163 tmp = ((OP(c) == BOUND ?
1164 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1165 LOAD_UTF8_CHARCLASS(alnum,"a");
1166 while (s < strend) {
1167 if (tmp == !(OP(c) == BOUND ?
1168 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1169 isALNUM_LC_utf8((U8*)s)))
1172 if ((norun || regtry(prog, s)))
1179 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1180 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1181 while (s < strend) {
1183 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1185 if ((norun || regtry(prog, s)))
1191 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1195 PL_reg_flags |= RF_tainted;
1202 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1204 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1206 tmp = ((OP(c) == NBOUND ?
1207 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1208 LOAD_UTF8_CHARCLASS(alnum,"a");
1209 while (s < strend) {
1210 if (tmp == !(OP(c) == NBOUND ?
1211 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1212 isALNUM_LC_utf8((U8*)s)))
1214 else if ((norun || regtry(prog, s)))
1220 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1221 tmp = ((OP(c) == NBOUND ?
1222 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1223 while (s < strend) {
1225 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1227 else if ((norun || regtry(prog, s)))
1232 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1237 LOAD_UTF8_CHARCLASS(alnum,"a");
1238 while (s < strend) {
1239 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1240 if (tmp && (norun || regtry(prog, s)))
1251 while (s < strend) {
1253 if (tmp && (norun || regtry(prog, s)))
1265 PL_reg_flags |= RF_tainted;
1267 while (s < strend) {
1268 if (isALNUM_LC_utf8((U8*)s)) {
1269 if (tmp && (norun || regtry(prog, s)))
1280 while (s < strend) {
1281 if (isALNUM_LC(*s)) {
1282 if (tmp && (norun || regtry(prog, s)))
1295 LOAD_UTF8_CHARCLASS(alnum,"a");
1296 while (s < strend) {
1297 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1298 if (tmp && (norun || regtry(prog, s)))
1309 while (s < strend) {
1311 if (tmp && (norun || regtry(prog, s)))
1323 PL_reg_flags |= RF_tainted;
1325 while (s < strend) {
1326 if (!isALNUM_LC_utf8((U8*)s)) {
1327 if (tmp && (norun || regtry(prog, s)))
1338 while (s < strend) {
1339 if (!isALNUM_LC(*s)) {
1340 if (tmp && (norun || regtry(prog, s)))
1353 LOAD_UTF8_CHARCLASS(space," ");
1354 while (s < strend) {
1355 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1356 if (tmp && (norun || regtry(prog, s)))
1367 while (s < strend) {
1369 if (tmp && (norun || regtry(prog, s)))
1381 PL_reg_flags |= RF_tainted;
1383 while (s < strend) {
1384 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1385 if (tmp && (norun || regtry(prog, s)))
1396 while (s < strend) {
1397 if (isSPACE_LC(*s)) {
1398 if (tmp && (norun || regtry(prog, s)))
1411 LOAD_UTF8_CHARCLASS(space," ");
1412 while (s < strend) {
1413 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1414 if (tmp && (norun || regtry(prog, s)))
1425 while (s < strend) {
1427 if (tmp && (norun || regtry(prog, s)))
1439 PL_reg_flags |= RF_tainted;
1441 while (s < strend) {
1442 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1443 if (tmp && (norun || regtry(prog, s)))
1454 while (s < strend) {
1455 if (!isSPACE_LC(*s)) {
1456 if (tmp && (norun || regtry(prog, s)))
1469 LOAD_UTF8_CHARCLASS(digit,"0");
1470 while (s < strend) {
1471 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1472 if (tmp && (norun || regtry(prog, s)))
1483 while (s < strend) {
1485 if (tmp && (norun || regtry(prog, s)))
1497 PL_reg_flags |= RF_tainted;
1499 while (s < strend) {
1500 if (isDIGIT_LC_utf8((U8*)s)) {
1501 if (tmp && (norun || regtry(prog, s)))
1512 while (s < strend) {
1513 if (isDIGIT_LC(*s)) {
1514 if (tmp && (norun || regtry(prog, s)))
1527 LOAD_UTF8_CHARCLASS(digit,"0");
1528 while (s < strend) {
1529 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1530 if (tmp && (norun || regtry(prog, s)))
1541 while (s < strend) {
1543 if (tmp && (norun || regtry(prog, s)))
1555 PL_reg_flags |= RF_tainted;
1557 while (s < strend) {
1558 if (!isDIGIT_LC_utf8((U8*)s)) {
1559 if (tmp && (norun || regtry(prog, s)))
1570 while (s < strend) {
1571 if (!isDIGIT_LC(*s)) {
1572 if (tmp && (norun || regtry(prog, s)))
1584 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1593 - regexec_flags - match a regexp against a string
1596 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1597 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1598 /* strend: pointer to null at end of string */
1599 /* strbeg: real beginning of string */
1600 /* minend: end of match must be >=minend after stringarg. */
1601 /* data: May be used for some additional optimizations. */
1602 /* nosave: For optimizations. */
1605 register regnode *c;
1606 register char *startpos = stringarg;
1607 I32 minlen; /* must match at least this many chars */
1608 I32 dontbother = 0; /* how many characters not to try at end */
1609 /* I32 start_shift = 0; */ /* Offset of the start to find
1610 constant substr. */ /* CC */
1611 I32 end_shift = 0; /* Same for the end. */ /* CC */
1612 I32 scream_pos = -1; /* Internal iterator of scream. */
1614 SV* oreplsv = GvSV(PL_replgv);
1615 bool do_utf8 = DO_UTF8(sv);
1617 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
1618 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
1620 RX_MATCH_UTF8_set(prog,do_utf8);
1626 PL_regnarrate = DEBUG_r_TEST;
1629 /* Be paranoid... */
1630 if (prog == NULL || startpos == NULL) {
1631 Perl_croak(aTHX_ "NULL regexp parameter");
1635 minlen = prog->minlen;
1636 if (strend - startpos < minlen) {
1637 DEBUG_r(PerlIO_printf(Perl_debug_log,
1638 "String too short [regexec_flags]...\n"));
1642 /* Check validity of program. */
1643 if (UCHARAT(prog->program) != REG_MAGIC) {
1644 Perl_croak(aTHX_ "corrupted regexp program");
1648 PL_reg_eval_set = 0;
1651 if (prog->reganch & ROPT_UTF8)
1652 PL_reg_flags |= RF_utf8;
1654 /* Mark beginning of line for ^ and lookbehind. */
1655 PL_regbol = startpos;
1659 /* Mark end of line for $ (and such) */
1662 /* see how far we have to get to not match where we matched before */
1663 PL_regtill = startpos+minend;
1665 /* We start without call_cc context. */
1668 /* If there is a "must appear" string, look for it. */
1671 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1674 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1675 PL_reg_ganch = startpos;
1676 else if (sv && SvTYPE(sv) >= SVt_PVMG
1678 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1679 && mg->mg_len >= 0) {
1680 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1681 if (prog->reganch & ROPT_ANCH_GPOS) {
1682 if (s > PL_reg_ganch)
1687 else /* pos() not defined */
1688 PL_reg_ganch = strbeg;
1691 if (!(flags & REXEC_CHECKED) && (prog->check_substr != Nullsv || prog->check_utf8 != Nullsv)) {
1692 re_scream_pos_data d;
1694 d.scream_olds = &scream_olds;
1695 d.scream_pos = &scream_pos;
1696 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1698 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1699 goto phooey; /* not present */
1705 pv_uni_display(dsv0, (U8*)prog->precomp, prog->prelen, 60,
1706 UNI_DISPLAY_REGEX) :
1708 int len0 = UTF ? SvCUR(dsv0) : prog->prelen;
1709 char *s1 = do_utf8 ? sv_uni_display(dsv1, sv, 60,
1710 UNI_DISPLAY_REGEX) : startpos;
1711 int len1 = do_utf8 ? SvCUR(dsv1) : strend - startpos;
1714 PerlIO_printf(Perl_debug_log,
1715 "%sMatching REx%s `%s%*.*s%s%s' against `%s%.*s%s%s'\n",
1716 PL_colors[4],PL_colors[5],PL_colors[0],
1719 len0 > 60 ? "..." : "",
1721 (int)(len1 > 60 ? 60 : len1),
1723 (len1 > 60 ? "..." : "")
1727 /* Simplest case: anchored match need be tried only once. */
1728 /* [unless only anchor is BOL and multiline is set] */
1729 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1730 if (s == startpos && regtry(prog, startpos))
1732 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1733 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1738 dontbother = minlen - 1;
1739 end = HOP3c(strend, -dontbother, strbeg) - 1;
1740 /* for multiline we only have to try after newlines */
1741 if (prog->check_substr || prog->check_utf8) {
1745 if (regtry(prog, s))
1750 if (prog->reganch & RE_USE_INTUIT) {
1751 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1762 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1763 if (regtry(prog, s))
1770 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1771 if (regtry(prog, PL_reg_ganch))
1776 /* Messy cases: unanchored match. */
1777 if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) {
1778 /* we have /x+whatever/ */
1779 /* it must be a one character string (XXXX Except UTF?) */
1784 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1785 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1786 ch = SvPVX(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
1789 while (s < strend) {
1791 DEBUG_r( did_match = 1 );
1792 if (regtry(prog, s)) goto got_it;
1794 while (s < strend && *s == ch)
1801 while (s < strend) {
1803 DEBUG_r( did_match = 1 );
1804 if (regtry(prog, s)) goto got_it;
1806 while (s < strend && *s == ch)
1812 DEBUG_r(if (!did_match)
1813 PerlIO_printf(Perl_debug_log,
1814 "Did not find anchored character...\n")
1818 else if (prog->anchored_substr != Nullsv
1819 || prog->anchored_utf8 != Nullsv
1820 || ((prog->float_substr != Nullsv || prog->float_utf8 != Nullsv)
1821 && prog->float_max_offset < strend - s)) {
1826 char *last1; /* Last position checked before */
1830 if (prog->anchored_substr || prog->anchored_utf8) {
1831 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1832 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1833 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1834 back_max = back_min = prog->anchored_offset;
1836 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1837 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1838 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1839 back_max = prog->float_max_offset;
1840 back_min = prog->float_min_offset;
1842 if (must == &PL_sv_undef)
1843 /* could not downgrade utf8 check substring, so must fail */
1846 last = HOP3c(strend, /* Cannot start after this */
1847 -(I32)(CHR_SVLEN(must)
1848 - (SvTAIL(must) != 0) + back_min), strbeg);
1851 last1 = HOPc(s, -1);
1853 last1 = s - 1; /* bogus */
1855 /* XXXX check_substr already used to find `s', can optimize if
1856 check_substr==must. */
1858 dontbother = end_shift;
1859 strend = HOPc(strend, -dontbother);
1860 while ( (s <= last) &&
1861 ((flags & REXEC_SCREAM)
1862 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1863 end_shift, &scream_pos, 0))
1864 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1865 (unsigned char*)strend, must,
1866 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1867 /* we may be pointing at the wrong string */
1868 if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog))
1869 s = strbeg + (s - SvPVX(sv));
1870 DEBUG_r( did_match = 1 );
1871 if (HOPc(s, -back_max) > last1) {
1872 last1 = HOPc(s, -back_min);
1873 s = HOPc(s, -back_max);
1876 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1878 last1 = HOPc(s, -back_min);
1882 while (s <= last1) {
1883 if (regtry(prog, s))
1889 while (s <= last1) {
1890 if (regtry(prog, s))
1896 DEBUG_r(if (!did_match)
1897 PerlIO_printf(Perl_debug_log,
1898 "Did not find %s substr `%s%.*s%s'%s...\n",
1899 ((must == prog->anchored_substr || must == prog->anchored_utf8)
1900 ? "anchored" : "floating"),
1902 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1904 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1908 else if ((c = prog->regstclass)) {
1910 I32 op = (U8)OP(prog->regstclass);
1911 /* don't bother with what can't match */
1912 if (PL_regkind[op] != EXACT && op != CANY)
1913 strend = HOPc(strend, -(minlen - 1));
1916 SV *prop = sv_newmortal();
1924 pv_uni_display(dsv0, (U8*)SvPVX(prop), SvCUR(prop), 60,
1925 UNI_DISPLAY_REGEX) :
1927 len0 = UTF ? SvCUR(dsv0) : SvCUR(prop);
1929 sv_uni_display(dsv1, sv, 60, UNI_DISPLAY_REGEX) : s;
1930 len1 = UTF ? SvCUR(dsv1) : strend - s;
1931 PerlIO_printf(Perl_debug_log,
1932 "Matching stclass `%*.*s' against `%*.*s'\n",
1936 if (find_byclass(prog, c, s, strend, startpos, 0))
1938 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1942 if (prog->float_substr != Nullsv || prog->float_utf8 != Nullsv) {
1947 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1948 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1949 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
1951 if (flags & REXEC_SCREAM) {
1952 last = screaminstr(sv, float_real, s - strbeg,
1953 end_shift, &scream_pos, 1); /* last one */
1955 last = scream_olds; /* Only one occurrence. */
1956 /* we may be pointing at the wrong string */
1957 else if (RX_MATCH_COPIED(prog))
1958 s = strbeg + (s - SvPVX(sv));
1962 char *little = SvPV(float_real, len);
1964 if (SvTAIL(float_real)) {
1965 if (memEQ(strend - len + 1, little, len - 1))
1966 last = strend - len + 1;
1967 else if (!PL_multiline)
1968 last = memEQ(strend - len, little, len)
1969 ? strend - len : Nullch;
1975 last = rninstr(s, strend, little, little + len);
1977 last = strend; /* matching `$' */
1981 DEBUG_r(PerlIO_printf(Perl_debug_log,
1982 "%sCan't trim the tail, match fails (should not happen)%s\n",
1983 PL_colors[4],PL_colors[5]));
1984 goto phooey; /* Should not happen! */
1986 dontbother = strend - last + prog->float_min_offset;
1988 if (minlen && (dontbother < minlen))
1989 dontbother = minlen - 1;
1990 strend -= dontbother; /* this one's always in bytes! */
1991 /* We don't know much -- general case. */
1994 if (regtry(prog, s))
2003 if (regtry(prog, s))
2005 } while (s++ < strend);
2013 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
2015 if (PL_reg_eval_set) {
2016 /* Preserve the current value of $^R */
2017 if (oreplsv != GvSV(PL_replgv))
2018 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
2019 restored, the value remains
2021 restore_pos(aTHX_ 0);
2024 /* make sure $`, $&, $', and $digit will work later */
2025 if ( !(flags & REXEC_NOT_FIRST) ) {
2026 if (RX_MATCH_COPIED(prog)) {
2027 Safefree(prog->subbeg);
2028 RX_MATCH_COPIED_off(prog);
2030 if (flags & REXEC_COPY_STR) {
2031 I32 i = PL_regeol - startpos + (stringarg - strbeg);
2033 s = savepvn(strbeg, i);
2036 RX_MATCH_COPIED_on(prog);
2039 prog->subbeg = strbeg;
2040 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2047 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2048 PL_colors[4],PL_colors[5]));
2049 if (PL_reg_eval_set)
2050 restore_pos(aTHX_ 0);
2055 - regtry - try match at specific point
2057 STATIC I32 /* 0 failure, 1 success */
2058 S_regtry(pTHX_ regexp *prog, char *startpos)
2066 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
2068 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
2071 PL_reg_eval_set = RS_init;
2073 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2074 (IV)(PL_stack_sp - PL_stack_base));
2076 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
2077 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2078 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2080 /* Apparently this is not needed, judging by wantarray. */
2081 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
2082 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2085 /* Make $_ available to executed code. */
2086 if (PL_reg_sv != DEFSV) {
2087 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
2092 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
2093 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
2094 /* prepare for quick setting of pos */
2095 sv_magic(PL_reg_sv, (SV*)0,
2096 PERL_MAGIC_regex_global, Nullch, 0);
2097 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
2101 PL_reg_oldpos = mg->mg_len;
2102 SAVEDESTRUCTOR_X(restore_pos, 0);
2104 if (!PL_reg_curpm) {
2105 Newz(22,PL_reg_curpm, 1, PMOP);
2108 SV* repointer = newSViv(0);
2109 /* so we know which PL_regex_padav element is PL_reg_curpm */
2110 SvFLAGS(repointer) |= SVf_BREAK;
2111 av_push(PL_regex_padav,repointer);
2112 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2113 PL_regex_pad = AvARRAY(PL_regex_padav);
2117 PM_SETRE(PL_reg_curpm, prog);
2118 PL_reg_oldcurpm = PL_curpm;
2119 PL_curpm = PL_reg_curpm;
2120 if (RX_MATCH_COPIED(prog)) {
2121 /* Here is a serious problem: we cannot rewrite subbeg,
2122 since it may be needed if this match fails. Thus
2123 $` inside (?{}) could fail... */
2124 PL_reg_oldsaved = prog->subbeg;
2125 PL_reg_oldsavedlen = prog->sublen;
2126 RX_MATCH_COPIED_off(prog);
2129 PL_reg_oldsaved = Nullch;
2130 prog->subbeg = PL_bostr;
2131 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2133 prog->startp[0] = startpos - PL_bostr;
2134 PL_reginput = startpos;
2135 PL_regstartp = prog->startp;
2136 PL_regendp = prog->endp;
2137 PL_reglastparen = &prog->lastparen;
2138 PL_reglastcloseparen = &prog->lastcloseparen;
2139 prog->lastparen = 0;
2140 prog->lastcloseparen = 0;
2142 DEBUG_r(PL_reg_starttry = startpos);
2143 if (PL_reg_start_tmpl <= prog->nparens) {
2144 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2145 if(PL_reg_start_tmp)
2146 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2148 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2151 /* XXXX What this code is doing here?!!! There should be no need
2152 to do this again and again, PL_reglastparen should take care of
2155 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2156 * Actually, the code in regcppop() (which Ilya may be meaning by
2157 * PL_reglastparen), is not needed at all by the test suite
2158 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2159 * enough, for building DynaLoader, or otherwise this
2160 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2161 * will happen. Meanwhile, this code *is* needed for the
2162 * above-mentioned test suite tests to succeed. The common theme
2163 * on those tests seems to be returning null fields from matches.
2168 if (prog->nparens) {
2169 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2176 if (regmatch(prog->program + 1)) {
2177 prog->endp[0] = PL_reginput - PL_bostr;
2180 REGCP_UNWIND(lastcp);
2184 #define RE_UNWIND_BRANCH 1
2185 #define RE_UNWIND_BRANCHJ 2
2189 typedef struct { /* XX: makes sense to enlarge it... */
2193 } re_unwind_generic_t;
2206 } re_unwind_branch_t;
2208 typedef union re_unwind_t {
2210 re_unwind_generic_t generic;
2211 re_unwind_branch_t branch;
2214 #define sayYES goto yes
2215 #define sayNO goto no
2216 #define sayNO_ANYOF goto no_anyof
2217 #define sayYES_FINAL goto yes_final
2218 #define sayYES_LOUD goto yes_loud
2219 #define sayNO_FINAL goto no_final
2220 #define sayNO_SILENT goto do_no
2221 #define saySAME(x) if (x) goto yes; else goto no
2223 #define REPORT_CODE_OFF 24
2226 - regmatch - main matching routine
2228 * Conceptually the strategy is simple: check to see whether the current
2229 * node matches, call self recursively to see whether the rest matches,
2230 * and then act accordingly. In practice we make some effort to avoid
2231 * recursion, in particular by going through "ordinary" nodes (that don't
2232 * need to know whether the rest of the match failed) by a loop instead of
2235 /* [lwall] I've hoisted the register declarations to the outer block in order to
2236 * maybe save a little bit of pushing and popping on the stack. It also takes
2237 * advantage of machines that use a register save mask on subroutine entry.
2239 STATIC I32 /* 0 failure, 1 success */
2240 S_regmatch(pTHX_ regnode *prog)
2242 register regnode *scan; /* Current node. */
2243 regnode *next; /* Next node. */
2244 regnode *inner; /* Next node in internal branch. */
2245 register I32 nextchr; /* renamed nextchr - nextchar colides with
2246 function of same name */
2247 register I32 n; /* no or next */
2248 register I32 ln = 0; /* len or last */
2249 register char *s = Nullch; /* operand or save */
2250 register char *locinput = PL_reginput;
2251 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2252 int minmod = 0, sw = 0, logical = 0;
2255 I32 firstcp = PL_savestack_ix;
2257 register bool do_utf8 = PL_reg_match_utf8;
2259 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2260 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2261 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2268 /* Note that nextchr is a byte even in UTF */
2269 nextchr = UCHARAT(locinput);
2271 while (scan != NULL) {
2274 SV *prop = sv_newmortal();
2275 int docolor = *PL_colors[0];
2276 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2277 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2278 /* The part of the string before starttry has one color
2279 (pref0_len chars), between starttry and current
2280 position another one (pref_len - pref0_len chars),
2281 after the current position the third one.
2282 We assume that pref0_len <= pref_len, otherwise we
2283 decrease pref0_len. */
2284 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2285 ? (5 + taill) - l : locinput - PL_bostr;
2288 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2290 pref0_len = pref_len - (locinput - PL_reg_starttry);
2291 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2292 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2293 ? (5 + taill) - pref_len : PL_regeol - locinput);
2294 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2298 if (pref0_len > pref_len)
2299 pref0_len = pref_len;
2300 regprop(prop, scan);
2303 do_utf8 && OP(scan) != CANY ?
2304 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2305 pref0_len, 60, UNI_DISPLAY_REGEX) :
2306 locinput - pref_len;
2307 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2308 char *s1 = do_utf8 && OP(scan) != CANY ?
2309 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2310 pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) :
2311 locinput - pref_len + pref0_len;
2312 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2313 char *s2 = do_utf8 && OP(scan) != CANY ?
2314 pv_uni_display(dsv2, (U8*)locinput,
2315 PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
2317 int len2 = do_utf8 ? strlen(s2) : l;
2318 PerlIO_printf(Perl_debug_log,
2319 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2320 (IV)(locinput - PL_bostr),
2327 (docolor ? "" : "> <"),
2331 15 - l - pref_len + 1,
2333 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2338 next = scan + NEXT_OFF(scan);
2344 if (locinput == PL_bostr || (PL_multiline &&
2345 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2347 /* regtill = regbol; */
2352 if (locinput == PL_bostr ||
2353 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2359 if (locinput == PL_bostr)
2363 if (locinput == PL_reg_ganch)
2373 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2378 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2380 if (PL_regeol - locinput > 1)
2384 if (PL_regeol != locinput)
2388 if (!nextchr && locinput >= PL_regeol)
2391 locinput += PL_utf8skip[nextchr];
2392 if (locinput > PL_regeol)
2394 nextchr = UCHARAT(locinput);
2397 nextchr = UCHARAT(++locinput);
2400 if (!nextchr && locinput >= PL_regeol)
2402 nextchr = UCHARAT(++locinput);
2405 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2408 locinput += PL_utf8skip[nextchr];
2409 if (locinput > PL_regeol)
2411 nextchr = UCHARAT(locinput);
2414 nextchr = UCHARAT(++locinput);
2419 if (do_utf8 != UTF) {
2420 /* The target and the pattern have differing utf8ness. */
2426 /* The target is utf8, the pattern is not utf8. */
2430 if (NATIVE_TO_UNI(*(U8*)s) !=
2431 utf8n_to_uvuni((U8*)l, UTF8_MAXLEN, &ulen,
2433 0 : UTF8_ALLOW_ANY))
2440 /* The target is not utf8, the pattern is utf8. */
2444 if (NATIVE_TO_UNI(*((U8*)l)) !=
2445 utf8n_to_uvuni((U8*)s, UTF8_MAXLEN, &ulen,
2447 0 : UTF8_ALLOW_ANY))
2454 nextchr = UCHARAT(locinput);
2457 /* The target and the pattern have the same utf8ness. */
2458 /* Inline the first character, for speed. */
2459 if (UCHARAT(s) != nextchr)
2461 if (PL_regeol - locinput < ln)
2463 if (ln > 1 && memNE(s, locinput, ln))
2466 nextchr = UCHARAT(locinput);
2469 PL_reg_flags |= RF_tainted;
2475 if (do_utf8 || UTF) {
2476 /* Either target or the pattern are utf8. */
2478 char *e = PL_regeol;
2480 if (ibcmp_utf8(s, 0, ln, (bool)UTF,
2481 l, &e, 0, do_utf8)) {
2482 /* One more case for the sharp s:
2483 * pack("U0U*", 0xDF) =~ /ss/i,
2484 * the 0xC3 0x9F are the UTF-8
2485 * byte sequence for the U+00DF. */
2487 toLOWER(s[0]) == 's' &&
2489 toLOWER(s[1]) == 's' &&
2496 nextchr = UCHARAT(locinput);
2500 /* Neither the target and the pattern are utf8. */
2502 /* Inline the first character, for speed. */
2503 if (UCHARAT(s) != nextchr &&
2504 UCHARAT(s) != ((OP(scan) == EXACTF)
2505 ? PL_fold : PL_fold_locale)[nextchr])
2507 if (PL_regeol - locinput < ln)
2509 if (ln > 1 && (OP(scan) == EXACTF
2510 ? ibcmp(s, locinput, ln)
2511 : ibcmp_locale(s, locinput, ln)))
2514 nextchr = UCHARAT(locinput);
2518 STRLEN inclasslen = PL_regeol - locinput;
2520 if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8))
2522 if (locinput >= PL_regeol)
2524 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
2525 nextchr = UCHARAT(locinput);
2530 nextchr = UCHARAT(locinput);
2531 if (!REGINCLASS(scan, (U8*)locinput))
2533 if (!nextchr && locinput >= PL_regeol)
2535 nextchr = UCHARAT(++locinput);
2539 /* If we might have the case of the German sharp s
2540 * in a casefolding Unicode character class. */
2542 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
2543 locinput += SHARP_S_SKIP;
2544 nextchr = UCHARAT(locinput);
2550 PL_reg_flags |= RF_tainted;
2556 LOAD_UTF8_CHARCLASS(alnum,"a");
2557 if (!(OP(scan) == ALNUM
2558 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2559 : isALNUM_LC_utf8((U8*)locinput)))
2563 locinput += PL_utf8skip[nextchr];
2564 nextchr = UCHARAT(locinput);
2567 if (!(OP(scan) == ALNUM
2568 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2570 nextchr = UCHARAT(++locinput);
2573 PL_reg_flags |= RF_tainted;
2576 if (!nextchr && locinput >= PL_regeol)
2579 LOAD_UTF8_CHARCLASS(alnum,"a");
2580 if (OP(scan) == NALNUM
2581 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2582 : isALNUM_LC_utf8((U8*)locinput))
2586 locinput += PL_utf8skip[nextchr];
2587 nextchr = UCHARAT(locinput);
2590 if (OP(scan) == NALNUM
2591 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2593 nextchr = UCHARAT(++locinput);
2597 PL_reg_flags |= RF_tainted;
2601 /* was last char in word? */
2603 if (locinput == PL_bostr)
2606 U8 *r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
2608 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
2610 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2611 ln = isALNUM_uni(ln);
2612 LOAD_UTF8_CHARCLASS(alnum,"a");
2613 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2616 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2617 n = isALNUM_LC_utf8((U8*)locinput);
2621 ln = (locinput != PL_bostr) ?
2622 UCHARAT(locinput - 1) : '\n';
2623 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2625 n = isALNUM(nextchr);
2628 ln = isALNUM_LC(ln);
2629 n = isALNUM_LC(nextchr);
2632 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2633 OP(scan) == BOUNDL))
2637 PL_reg_flags |= RF_tainted;
2643 if (UTF8_IS_CONTINUED(nextchr)) {
2644 LOAD_UTF8_CHARCLASS(space," ");
2645 if (!(OP(scan) == SPACE
2646 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2647 : isSPACE_LC_utf8((U8*)locinput)))
2651 locinput += PL_utf8skip[nextchr];
2652 nextchr = UCHARAT(locinput);
2655 if (!(OP(scan) == SPACE
2656 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2658 nextchr = UCHARAT(++locinput);
2661 if (!(OP(scan) == SPACE
2662 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2664 nextchr = UCHARAT(++locinput);
2668 PL_reg_flags |= RF_tainted;
2671 if (!nextchr && locinput >= PL_regeol)
2674 LOAD_UTF8_CHARCLASS(space," ");
2675 if (OP(scan) == NSPACE
2676 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2677 : isSPACE_LC_utf8((U8*)locinput))
2681 locinput += PL_utf8skip[nextchr];
2682 nextchr = UCHARAT(locinput);
2685 if (OP(scan) == NSPACE
2686 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2688 nextchr = UCHARAT(++locinput);
2691 PL_reg_flags |= RF_tainted;
2697 LOAD_UTF8_CHARCLASS(digit,"0");
2698 if (!(OP(scan) == DIGIT
2699 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2700 : isDIGIT_LC_utf8((U8*)locinput)))
2704 locinput += PL_utf8skip[nextchr];
2705 nextchr = UCHARAT(locinput);
2708 if (!(OP(scan) == DIGIT
2709 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2711 nextchr = UCHARAT(++locinput);
2714 PL_reg_flags |= RF_tainted;
2717 if (!nextchr && locinput >= PL_regeol)
2720 LOAD_UTF8_CHARCLASS(digit,"0");
2721 if (OP(scan) == NDIGIT
2722 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2723 : isDIGIT_LC_utf8((U8*)locinput))
2727 locinput += PL_utf8skip[nextchr];
2728 nextchr = UCHARAT(locinput);
2731 if (OP(scan) == NDIGIT
2732 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2734 nextchr = UCHARAT(++locinput);
2737 if (locinput >= PL_regeol)
2740 LOAD_UTF8_CHARCLASS(mark,"~");
2741 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2743 locinput += PL_utf8skip[nextchr];
2744 while (locinput < PL_regeol &&
2745 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2746 locinput += UTF8SKIP(locinput);
2747 if (locinput > PL_regeol)
2752 nextchr = UCHARAT(locinput);
2755 PL_reg_flags |= RF_tainted;
2759 n = ARG(scan); /* which paren pair */
2760 ln = PL_regstartp[n];
2761 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2762 if ((I32)*PL_reglastparen < n || ln == -1)
2763 sayNO; /* Do not match unless seen CLOSEn. */
2764 if (ln == PL_regendp[n])
2768 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2770 char *e = PL_bostr + PL_regendp[n];
2772 * Note that we can't do the "other character" lookup trick as
2773 * in the 8-bit case (no pun intended) because in Unicode we
2774 * have to map both upper and title case to lower case.
2776 if (OP(scan) == REFF) {
2777 STRLEN ulen1, ulen2;
2778 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2779 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2783 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2784 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2785 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2792 nextchr = UCHARAT(locinput);
2796 /* Inline the first character, for speed. */
2797 if (UCHARAT(s) != nextchr &&
2799 (UCHARAT(s) != ((OP(scan) == REFF
2800 ? PL_fold : PL_fold_locale)[nextchr]))))
2802 ln = PL_regendp[n] - ln;
2803 if (locinput + ln > PL_regeol)
2805 if (ln > 1 && (OP(scan) == REF
2806 ? memNE(s, locinput, ln)
2808 ? ibcmp(s, locinput, ln)
2809 : ibcmp_locale(s, locinput, ln))))
2812 nextchr = UCHARAT(locinput);
2823 OP_4tree *oop = PL_op;
2824 COP *ocurcop = PL_curcop;
2827 struct regexp *oreg = PL_reg_re;
2830 PL_op = (OP_4tree*)PL_regdata->data[n];
2831 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2832 PAD_SAVE_LOCAL(old_comppad, (PAD*)PL_regdata->data[n + 2]);
2833 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2837 CALLRUNOPS(aTHX); /* Scalar context. */
2840 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
2848 PAD_RESTORE_LOCAL(old_comppad);
2849 PL_curcop = ocurcop;
2851 if (logical == 2) { /* Postponed subexpression. */
2853 MAGIC *mg = Null(MAGIC*);
2855 CHECKPOINT cp, lastcp;
2859 if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret)))
2860 mg = mg_find(sv, PERL_MAGIC_qr);
2861 else if (SvSMAGICAL(ret)) {
2862 if (SvGMAGICAL(ret))
2863 sv_unmagic(ret, PERL_MAGIC_qr);
2865 mg = mg_find(ret, PERL_MAGIC_qr);
2869 re = (regexp *)mg->mg_obj;
2870 (void)ReREFCNT_inc(re);
2874 char *t = SvPV(ret, len);
2876 char *oprecomp = PL_regprecomp;
2877 I32 osize = PL_regsize;
2878 I32 onpar = PL_regnpar;
2881 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
2882 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2884 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
2886 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2888 PL_regprecomp = oprecomp;
2893 PerlIO_printf(Perl_debug_log,
2894 "Entering embedded `%s%.60s%s%s'\n",
2898 (strlen(re->precomp) > 60 ? "..." : ""))
2901 state.prev = PL_reg_call_cc;
2902 state.cc = PL_regcc;
2903 state.re = PL_reg_re;
2907 cp = regcppush(0); /* Save *all* the positions. */
2910 state.ss = PL_savestack_ix;
2911 *PL_reglastparen = 0;
2912 *PL_reglastcloseparen = 0;
2913 PL_reg_call_cc = &state;
2914 PL_reginput = locinput;
2915 toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
2916 ((re->reganch & ROPT_UTF8) != 0);
2917 if (toggleutf) PL_reg_flags ^= RF_utf8;
2919 /* XXXX This is too dramatic a measure... */
2922 if (regmatch(re->program + 1)) {
2923 /* Even though we succeeded, we need to restore
2924 global variables, since we may be wrapped inside
2925 SUSPEND, thus the match may be not finished yet. */
2927 /* XXXX Do this only if SUSPENDed? */
2928 PL_reg_call_cc = state.prev;
2929 PL_regcc = state.cc;
2930 PL_reg_re = state.re;
2931 cache_re(PL_reg_re);
2932 if (toggleutf) PL_reg_flags ^= RF_utf8;
2934 /* XXXX This is too dramatic a measure... */
2937 /* These are needed even if not SUSPEND. */
2943 REGCP_UNWIND(lastcp);
2945 PL_reg_call_cc = state.prev;
2946 PL_regcc = state.cc;
2947 PL_reg_re = state.re;
2948 cache_re(PL_reg_re);
2949 if (toggleutf) PL_reg_flags ^= RF_utf8;
2951 /* XXXX This is too dramatic a measure... */
2961 sv_setsv(save_scalar(PL_replgv), ret);
2967 n = ARG(scan); /* which paren pair */
2968 PL_reg_start_tmp[n] = locinput;
2973 n = ARG(scan); /* which paren pair */
2974 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2975 PL_regendp[n] = locinput - PL_bostr;
2976 if (n > (I32)*PL_reglastparen)
2977 *PL_reglastparen = n;
2978 *PL_reglastcloseparen = n;
2981 n = ARG(scan); /* which paren pair */
2982 sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
2985 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2987 next = NEXTOPER(NEXTOPER(scan));
2989 next = scan + ARG(scan);
2990 if (OP(next) == IFTHEN) /* Fake one. */
2991 next = NEXTOPER(NEXTOPER(next));
2995 logical = scan->flags;
2997 /*******************************************************************
2998 PL_regcc contains infoblock about the innermost (...)* loop, and
2999 a pointer to the next outer infoblock.
3001 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
3003 1) After matching X, regnode for CURLYX is processed;
3005 2) This regnode creates infoblock on the stack, and calls
3006 regmatch() recursively with the starting point at WHILEM node;
3008 3) Each hit of WHILEM node tries to match A and Z (in the order
3009 depending on the current iteration, min/max of {min,max} and
3010 greediness). The information about where are nodes for "A"
3011 and "Z" is read from the infoblock, as is info on how many times "A"
3012 was already matched, and greediness.
3014 4) After A matches, the same WHILEM node is hit again.
3016 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
3017 of the same pair. Thus when WHILEM tries to match Z, it temporarily
3018 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
3019 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
3020 of the external loop.
3022 Currently present infoblocks form a tree with a stem formed by PL_curcc
3023 and whatever it mentions via ->next, and additional attached trees
3024 corresponding to temporarily unset infoblocks as in "5" above.
3026 In the following picture infoblocks for outer loop of
3027 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
3028 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
3029 infoblocks are drawn below the "reset" infoblock.
3031 In fact in the picture below we do not show failed matches for Z and T
3032 by WHILEM blocks. [We illustrate minimal matches, since for them it is
3033 more obvious *why* one needs to *temporary* unset infoblocks.]
3035 Matched REx position InfoBlocks Comment
3039 Y A)*?Z)*?T x <- O <- I
3040 YA )*?Z)*?T x <- O <- I
3041 YA A)*?Z)*?T x <- O <- I
3042 YAA )*?Z)*?T x <- O <- I
3043 YAA Z)*?T x <- O # Temporary unset I
3046 YAAZ Y(A)*?Z)*?T x <- O
3049 YAAZY (A)*?Z)*?T x <- O
3052 YAAZY A)*?Z)*?T x <- O <- I
3055 YAAZYA )*?Z)*?T x <- O <- I
3058 YAAZYA Z)*?T x <- O # Temporary unset I
3064 YAAZYAZ T x # Temporary unset O
3071 *******************************************************************/
3074 CHECKPOINT cp = PL_savestack_ix;
3075 /* No need to save/restore up to this paren */
3076 I32 parenfloor = scan->flags;
3078 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3080 cc.oldcc = PL_regcc;
3082 /* XXXX Probably it is better to teach regpush to support
3083 parenfloor > PL_regsize... */
3084 if (parenfloor > (I32)*PL_reglastparen)
3085 parenfloor = *PL_reglastparen; /* Pessimization... */
3086 cc.parenfloor = parenfloor;
3088 cc.min = ARG1(scan);
3089 cc.max = ARG2(scan);
3090 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3094 PL_reginput = locinput;
3095 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
3097 PL_regcc = cc.oldcc;
3103 * This is really hard to understand, because after we match
3104 * what we're trying to match, we must make sure the rest of
3105 * the REx is going to match for sure, and to do that we have
3106 * to go back UP the parse tree by recursing ever deeper. And
3107 * if it fails, we have to reset our parent's current state
3108 * that we can try again after backing off.
3111 CHECKPOINT cp, lastcp;
3112 CURCUR* cc = PL_regcc;
3113 char *lastloc = cc->lastloc; /* Detection of 0-len. */
3115 n = cc->cur + 1; /* how many we know we matched */
3116 PL_reginput = locinput;
3119 PerlIO_printf(Perl_debug_log,
3120 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3121 REPORT_CODE_OFF+PL_regindent*2, "",
3122 (long)n, (long)cc->min,
3123 (long)cc->max, PTR2UV(cc))
3126 /* If degenerate scan matches "", assume scan done. */
3128 if (locinput == cc->lastloc && n >= cc->min) {
3129 PL_regcc = cc->oldcc;
3133 PerlIO_printf(Perl_debug_log,
3134 "%*s empty match detected, try continuation...\n",
3135 REPORT_CODE_OFF+PL_regindent*2, "")
3137 if (regmatch(cc->next))
3145 /* First just match a string of min scans. */
3149 cc->lastloc = locinput;
3150 if (regmatch(cc->scan))
3153 cc->lastloc = lastloc;
3158 /* Check whether we already were at this position.
3159 Postpone detection until we know the match is not
3160 *that* much linear. */
3161 if (!PL_reg_maxiter) {
3162 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3163 PL_reg_leftiter = PL_reg_maxiter;
3165 if (PL_reg_leftiter-- == 0) {
3166 I32 size = (PL_reg_maxiter + 7)/8;
3167 if (PL_reg_poscache) {
3168 if ((I32)PL_reg_poscache_size < size) {
3169 Renew(PL_reg_poscache, size, char);
3170 PL_reg_poscache_size = size;
3172 Zero(PL_reg_poscache, size, char);
3175 PL_reg_poscache_size = size;
3176 Newz(29, PL_reg_poscache, size, char);
3179 PerlIO_printf(Perl_debug_log,
3180 "%sDetected a super-linear match, switching on caching%s...\n",
3181 PL_colors[4], PL_colors[5])
3184 if (PL_reg_leftiter < 0) {
3185 I32 o = locinput - PL_bostr, b;
3187 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3190 if (PL_reg_poscache[o] & (1<<b)) {
3192 PerlIO_printf(Perl_debug_log,
3193 "%*s already tried at this position...\n",
3194 REPORT_CODE_OFF+PL_regindent*2, "")
3196 if (PL_reg_flags & RF_false)
3201 PL_reg_poscache[o] |= (1<<b);
3205 /* Prefer next over scan for minimal matching. */
3208 PL_regcc = cc->oldcc;
3211 cp = regcppush(cc->parenfloor);
3213 if (regmatch(cc->next)) {
3215 sayYES; /* All done. */
3217 REGCP_UNWIND(lastcp);
3223 if (n >= cc->max) { /* Maximum greed exceeded? */
3224 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3225 && !(PL_reg_flags & RF_warned)) {
3226 PL_reg_flags |= RF_warned;
3227 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3228 "Complex regular subexpression recursion",
3235 PerlIO_printf(Perl_debug_log,
3236 "%*s trying longer...\n",
3237 REPORT_CODE_OFF+PL_regindent*2, "")
3239 /* Try scanning more and see if it helps. */
3240 PL_reginput = locinput;
3242 cc->lastloc = locinput;
3243 cp = regcppush(cc->parenfloor);
3245 if (regmatch(cc->scan)) {
3249 REGCP_UNWIND(lastcp);
3252 cc->lastloc = lastloc;
3256 /* Prefer scan over next for maximal matching. */
3258 if (n < cc->max) { /* More greed allowed? */
3259 cp = regcppush(cc->parenfloor);
3261 cc->lastloc = locinput;
3263 if (regmatch(cc->scan)) {
3267 REGCP_UNWIND(lastcp);
3268 regcppop(); /* Restore some previous $<digit>s? */
3269 PL_reginput = locinput;
3271 PerlIO_printf(Perl_debug_log,
3272 "%*s failed, try continuation...\n",
3273 REPORT_CODE_OFF+PL_regindent*2, "")
3276 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3277 && !(PL_reg_flags & RF_warned)) {
3278 PL_reg_flags |= RF_warned;
3279 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3280 "Complex regular subexpression recursion",
3284 /* Failed deeper matches of scan, so see if this one works. */
3285 PL_regcc = cc->oldcc;
3288 if (regmatch(cc->next))
3294 cc->lastloc = lastloc;
3299 next = scan + ARG(scan);
3302 inner = NEXTOPER(NEXTOPER(scan));
3305 inner = NEXTOPER(scan);
3309 if (OP(next) != c1) /* No choice. */
3310 next = inner; /* Avoid recursion. */
3312 I32 lastparen = *PL_reglastparen;
3314 re_unwind_branch_t *uw;
3316 /* Put unwinding data on stack */
3317 unwind1 = SSNEWt(1,re_unwind_branch_t);
3318 uw = SSPTRt(unwind1,re_unwind_branch_t);
3321 uw->type = ((c1 == BRANCH)
3323 : RE_UNWIND_BRANCHJ);
3324 uw->lastparen = lastparen;
3326 uw->locinput = locinput;
3327 uw->nextchr = nextchr;
3329 uw->regindent = ++PL_regindent;
3332 REGCP_SET(uw->lastcp);
3334 /* Now go into the first branch */
3347 /* We suppose that the next guy does not need
3348 backtracking: in particular, it is of constant length,
3349 and has no parenths to influence future backrefs. */
3350 ln = ARG1(scan); /* min to match */
3351 n = ARG2(scan); /* max to match */
3352 paren = scan->flags;
3354 if (paren > PL_regsize)
3356 if (paren > (I32)*PL_reglastparen)
3357 *PL_reglastparen = paren;
3359 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3361 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3362 PL_reginput = locinput;
3365 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3367 /* if we matched something zero-length we don't need to
3368 backtrack - capturing parens are already defined, so
3369 the caveat in the maximal case doesn't apply
3371 XXXX if ln == 0, we can redo this check first time
3372 through the following loop
3375 n = ln; /* don't backtrack */
3376 locinput = PL_reginput;
3377 if (HAS_TEXT(next) || JUMPABLE(next)) {
3378 regnode *text_node = next;
3380 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3382 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3384 if (PL_regkind[(U8)OP(text_node)] == REF) {
3388 else { c1 = (U8)*STRING(text_node); }
3389 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3391 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3392 c2 = PL_fold_locale[c1];
3401 /* This may be improved if l == 0. */
3402 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3403 /* If it could work, try it. */
3405 UCHARAT(PL_reginput) == c1 ||
3406 UCHARAT(PL_reginput) == c2)
3410 PL_regstartp[paren] =
3411 HOPc(PL_reginput, -l) - PL_bostr;
3412 PL_regendp[paren] = PL_reginput - PL_bostr;
3415 PL_regendp[paren] = -1;
3419 REGCP_UNWIND(lastcp);
3421 /* Couldn't or didn't -- move forward. */
3422 PL_reginput = locinput;
3423 if (regrepeat_hard(scan, 1, &l)) {
3425 locinput = PL_reginput;
3432 n = regrepeat_hard(scan, n, &l);
3433 /* if we matched something zero-length we don't need to
3434 backtrack, unless the minimum count is zero and we
3435 are capturing the result - in that case the capture
3436 being defined or not may affect later execution
3438 if (n != 0 && l == 0 && !(paren && ln == 0))
3439 ln = n; /* don't backtrack */
3440 locinput = PL_reginput;
3442 PerlIO_printf(Perl_debug_log,
3443 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3444 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3448 if (HAS_TEXT(next) || JUMPABLE(next)) {
3449 regnode *text_node = next;
3451 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3453 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3455 if (PL_regkind[(U8)OP(text_node)] == REF) {
3459 else { c1 = (U8)*STRING(text_node); }
3461 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3463 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3464 c2 = PL_fold_locale[c1];
3475 /* If it could work, try it. */
3477 UCHARAT(PL_reginput) == c1 ||
3478 UCHARAT(PL_reginput) == c2)
3481 PerlIO_printf(Perl_debug_log,
3482 "%*s trying tail with n=%"IVdf"...\n",
3483 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3487 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3488 PL_regendp[paren] = PL_reginput - PL_bostr;
3491 PL_regendp[paren] = -1;
3495 REGCP_UNWIND(lastcp);
3497 /* Couldn't or didn't -- back up. */
3499 locinput = HOPc(locinput, -l);
3500 PL_reginput = locinput;
3507 paren = scan->flags; /* Which paren to set */
3508 if (paren > PL_regsize)
3510 if (paren > (I32)*PL_reglastparen)
3511 *PL_reglastparen = paren;
3512 ln = ARG1(scan); /* min to match */
3513 n = ARG2(scan); /* max to match */
3514 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3518 ln = ARG1(scan); /* min to match */
3519 n = ARG2(scan); /* max to match */
3520 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3525 scan = NEXTOPER(scan);
3531 scan = NEXTOPER(scan);
3535 * Lookahead to avoid useless match attempts
3536 * when we know what character comes next.
3540 * Used to only do .*x and .*?x, but now it allows
3541 * for )'s, ('s and (?{ ... })'s to be in the way
3542 * of the quantifier and the EXACT-like node. -- japhy
3545 if (HAS_TEXT(next) || JUMPABLE(next)) {
3547 regnode *text_node = next;
3549 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3551 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3553 if (PL_regkind[(U8)OP(text_node)] == REF) {
3555 goto assume_ok_easy;
3557 else { s = (U8*)STRING(text_node); }
3561 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3563 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3564 c2 = PL_fold_locale[c1];
3567 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3568 STRLEN ulen1, ulen2;
3569 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3570 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3572 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3573 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3575 c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXLEN, 0,
3577 0 : UTF8_ALLOW_ANY);
3578 c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXLEN, 0,
3580 0 : UTF8_ALLOW_ANY);
3583 c2 = c1 = utf8n_to_uvchr(s, UTF8_MAXLEN, 0,
3585 0 : UTF8_ALLOW_ANY);
3593 PL_reginput = locinput;
3597 if (ln && regrepeat(scan, ln) < ln)
3599 locinput = PL_reginput;
3602 char *e; /* Should not check after this */
3603 char *old = locinput;
3606 if (n == REG_INFTY) {
3609 while (UTF8_IS_CONTINUATION(*(U8*)e))
3615 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3619 e = locinput + n - ln;
3624 /* Find place 'next' could work */
3627 while (locinput <= e &&
3628 UCHARAT(locinput) != c1)
3631 while (locinput <= e
3632 && UCHARAT(locinput) != c1
3633 && UCHARAT(locinput) != c2)
3636 count = locinput - old;
3641 /* count initialised to
3642 * utf8_distance(old, locinput) */
3643 while (locinput <= e &&
3644 utf8n_to_uvchr((U8*)locinput,
3647 0 : UTF8_ALLOW_ANY) != (UV)c1) {
3652 /* count initialised to
3653 * utf8_distance(old, locinput) */
3654 while (locinput <= e) {
3655 UV c = utf8n_to_uvchr((U8*)locinput,
3658 0 : UTF8_ALLOW_ANY);
3659 if (c == (UV)c1 || c == (UV)c2)
3668 /* PL_reginput == old now */
3669 if (locinput != old) {
3670 ln = 1; /* Did some */
3671 if (regrepeat(scan, count) < count)
3674 /* PL_reginput == locinput now */
3675 TRYPAREN(paren, ln, locinput);
3676 PL_reginput = locinput; /* Could be reset... */
3677 REGCP_UNWIND(lastcp);
3678 /* Couldn't or didn't -- move forward. */
3681 locinput += UTF8SKIP(locinput);
3688 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3692 c = utf8n_to_uvchr((U8*)PL_reginput,
3695 0 : UTF8_ALLOW_ANY);
3697 c = UCHARAT(PL_reginput);
3698 /* If it could work, try it. */
3699 if (c == (UV)c1 || c == (UV)c2)
3701 TRYPAREN(paren, ln, PL_reginput);
3702 REGCP_UNWIND(lastcp);
3705 /* If it could work, try it. */
3706 else if (c1 == -1000)
3708 TRYPAREN(paren, ln, PL_reginput);
3709 REGCP_UNWIND(lastcp);
3711 /* Couldn't or didn't -- move forward. */
3712 PL_reginput = locinput;
3713 if (regrepeat(scan, 1)) {
3715 locinput = PL_reginput;
3723 n = regrepeat(scan, n);
3724 locinput = PL_reginput;
3725 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3726 ((!PL_multiline && OP(next) != MEOL) ||
3727 OP(next) == SEOL || OP(next) == EOS))
3729 ln = n; /* why back off? */
3730 /* ...because $ and \Z can match before *and* after
3731 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3732 We should back off by one in this case. */
3733 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
3742 c = utf8n_to_uvchr((U8*)PL_reginput,
3745 0 : UTF8_ALLOW_ANY);
3747 c = UCHARAT(PL_reginput);
3749 /* If it could work, try it. */
3750 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3752 TRYPAREN(paren, n, PL_reginput);
3753 REGCP_UNWIND(lastcp);
3755 /* Couldn't or didn't -- back up. */
3757 PL_reginput = locinput = HOPc(locinput, -1);
3765 c = utf8n_to_uvchr((U8*)PL_reginput,
3768 0 : UTF8_ALLOW_ANY);
3770 c = UCHARAT(PL_reginput);
3772 /* If it could work, try it. */
3773 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3775 TRYPAREN(paren, n, PL_reginput);
3776 REGCP_UNWIND(lastcp);
3778 /* Couldn't or didn't -- back up. */
3780 PL_reginput = locinput = HOPc(locinput, -1);
3787 if (PL_reg_call_cc) {
3788 re_cc_state *cur_call_cc = PL_reg_call_cc;
3789 CURCUR *cctmp = PL_regcc;
3790 regexp *re = PL_reg_re;
3791 CHECKPOINT cp, lastcp;
3793 cp = regcppush(0); /* Save *all* the positions. */
3795 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3797 PL_reginput = locinput; /* Make position available to
3799 cache_re(PL_reg_call_cc->re);
3800 PL_regcc = PL_reg_call_cc->cc;
3801 PL_reg_call_cc = PL_reg_call_cc->prev;
3802 if (regmatch(cur_call_cc->node)) {
3803 PL_reg_call_cc = cur_call_cc;
3807 REGCP_UNWIND(lastcp);
3809 PL_reg_call_cc = cur_call_cc;
3815 PerlIO_printf(Perl_debug_log,
3816 "%*s continuation failed...\n",
3817 REPORT_CODE_OFF+PL_regindent*2, "")
3821 if (locinput < PL_regtill) {
3822 DEBUG_r(PerlIO_printf(Perl_debug_log,
3823 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3825 (long)(locinput - PL_reg_starttry),
3826 (long)(PL_regtill - PL_reg_starttry),
3828 sayNO_FINAL; /* Cannot match: too short. */
3830 PL_reginput = locinput; /* put where regtry can find it */
3831 sayYES_FINAL; /* Success! */
3833 PL_reginput = locinput; /* put where regtry can find it */
3834 sayYES_LOUD; /* Success! */
3837 PL_reginput = locinput;
3842 s = HOPBACKc(locinput, scan->flags);
3848 PL_reginput = locinput;
3849 PL_reg_flags ^= RF_false;
3854 s = HOPBACKc(locinput, scan->flags);
3860 PL_reginput = locinput;
3863 inner = NEXTOPER(NEXTOPER(scan));
3864 if (regmatch(inner) != n) {
3866 PL_reg_flags ^= RF_false;
3877 PL_reg_flags ^= RF_false;
3883 if (OP(scan) == SUSPEND) {
3884 locinput = PL_reginput;
3885 nextchr = UCHARAT(locinput);
3890 next = scan + ARG(scan);
3895 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3896 PTR2UV(scan), OP(scan));
3897 Perl_croak(aTHX_ "regexp memory corruption");
3904 * We get here only if there's trouble -- normally "case END" is
3905 * the terminating point.
3907 Perl_croak(aTHX_ "corrupted regexp pointers");
3913 PerlIO_printf(Perl_debug_log,
3914 "%*s %scould match...%s\n",
3915 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3919 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3920 PL_colors[4],PL_colors[5]));
3926 #if 0 /* Breaks $^R */
3934 PerlIO_printf(Perl_debug_log,
3935 "%*s %sfailed...%s\n",
3936 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3942 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3945 case RE_UNWIND_BRANCH:
3946 case RE_UNWIND_BRANCHJ:
3948 re_unwind_branch_t *uwb = &(uw->branch);
3949 I32 lastparen = uwb->lastparen;
3951 REGCP_UNWIND(uwb->lastcp);
3952 for (n = *PL_reglastparen; n > lastparen; n--)
3954 *PL_reglastparen = n;
3955 scan = next = uwb->next;
3957 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3958 ? BRANCH : BRANCHJ) ) { /* Failure */
3965 /* Have more choice yet. Reuse the same uwb. */
3967 if ((n = (uwb->type == RE_UNWIND_BRANCH
3968 ? NEXT_OFF(next) : ARG(next))))
3971 next = NULL; /* XXXX Needn't unwinding in this case... */
3973 next = NEXTOPER(scan);
3974 if (uwb->type == RE_UNWIND_BRANCHJ)
3975 next = NEXTOPER(next);
3976 locinput = uwb->locinput;
3977 nextchr = uwb->nextchr;
3979 PL_regindent = uwb->regindent;
3986 Perl_croak(aTHX_ "regexp unwind memory corruption");
3997 - regrepeat - repeatedly match something simple, report how many
4000 * [This routine now assumes that it will only match on things of length 1.
4001 * That was true before, but now we assume scan - reginput is the count,
4002 * rather than incrementing count on every character. [Er, except utf8.]]
4005 S_regrepeat(pTHX_ regnode *p, I32 max)
4007 register char *scan;
4009 register char *loceol = PL_regeol;
4010 register I32 hardcount = 0;
4011 register bool do_utf8 = PL_reg_match_utf8;
4014 if (max == REG_INFTY)
4016 else if (max < loceol - scan)
4017 loceol = scan + max;
4022 while (scan < loceol && hardcount < max && *scan != '\n') {
4023 scan += UTF8SKIP(scan);
4027 while (scan < loceol && *scan != '\n')
4034 while (scan < loceol && hardcount < max) {
4035 scan += UTF8SKIP(scan);
4045 case EXACT: /* length of string is 1 */
4047 while (scan < loceol && UCHARAT(scan) == c)
4050 case EXACTF: /* length of string is 1 */
4052 while (scan < loceol &&
4053 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
4056 case EXACTFL: /* length of string is 1 */
4057 PL_reg_flags |= RF_tainted;
4059 while (scan < loceol &&
4060 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
4066 while (hardcount < max && scan < loceol &&
4067 reginclass(p, (U8*)scan, 0, do_utf8)) {
4068 scan += UTF8SKIP(scan);
4072 while (scan < loceol && REGINCLASS(p, (U8*)scan))
4079 LOAD_UTF8_CHARCLASS(alnum,"a");
4080 while (hardcount < max && scan < loceol &&
4081 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4082 scan += UTF8SKIP(scan);
4086 while (scan < loceol && isALNUM(*scan))
4091 PL_reg_flags |= RF_tainted;
4094 while (hardcount < max && scan < loceol &&
4095 isALNUM_LC_utf8((U8*)scan)) {
4096 scan += UTF8SKIP(scan);
4100 while (scan < loceol && isALNUM_LC(*scan))
4107 LOAD_UTF8_CHARCLASS(alnum,"a");
4108 while (hardcount < max && scan < loceol &&
4109 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4110 scan += UTF8SKIP(scan);
4114 while (scan < loceol && !isALNUM(*scan))
4119 PL_reg_flags |= RF_tainted;
4122 while (hardcount < max && scan < loceol &&
4123 !isALNUM_LC_utf8((U8*)scan)) {
4124 scan += UTF8SKIP(scan);
4128 while (scan < loceol && !isALNUM_LC(*scan))
4135 LOAD_UTF8_CHARCLASS(space," ");
4136 while (hardcount < max && scan < loceol &&
4138 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4139 scan += UTF8SKIP(scan);
4143 while (scan < loceol && isSPACE(*scan))
4148 PL_reg_flags |= RF_tainted;
4151 while (hardcount < max && scan < loceol &&
4152 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4153 scan += UTF8SKIP(scan);
4157 while (scan < loceol && isSPACE_LC(*scan))
4164 LOAD_UTF8_CHARCLASS(space," ");
4165 while (hardcount < max && scan < loceol &&
4167 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4168 scan += UTF8SKIP(scan);
4172 while (scan < loceol && !isSPACE(*scan))
4177 PL_reg_flags |= RF_tainted;
4180 while (hardcount < max && scan < loceol &&
4181 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4182 scan += UTF8SKIP(scan);
4186 while (scan < loceol && !isSPACE_LC(*scan))
4193 LOAD_UTF8_CHARCLASS(digit,"0");
4194 while (hardcount < max && scan < loceol &&
4195 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4196 scan += UTF8SKIP(scan);
4200 while (scan < loceol && isDIGIT(*scan))
4207 LOAD_UTF8_CHARCLASS(digit,"0");
4208 while (hardcount < max && scan < loceol &&
4209 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4210 scan += UTF8SKIP(scan);
4214 while (scan < loceol && !isDIGIT(*scan))
4218 default: /* Called on something of 0 width. */
4219 break; /* So match right here or not at all. */
4225 c = scan - PL_reginput;
4230 SV *prop = sv_newmortal();
4233 PerlIO_printf(Perl_debug_log,
4234 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4235 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4242 - regrepeat_hard - repeatedly match something, report total lenth and length
4244 * The repeater is supposed to have constant length.
4248 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4250 register char *scan = Nullch;
4251 register char *start;
4252 register char *loceol = PL_regeol;
4254 I32 count = 0, res = 1;
4259 start = PL_reginput;
4260 if (PL_reg_match_utf8) {
4261 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4264 while (start < PL_reginput) {
4266 start += UTF8SKIP(start);
4277 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4279 *lp = l = PL_reginput - start;
4280 if (max != REG_INFTY && l*max < loceol - scan)
4281 loceol = scan + l*max;
4294 - regclass_swash - prepare the utf8 swash
4298 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** listsvp, SV **altsvp)
4304 if (PL_regdata && PL_regdata->count) {
4307 if (PL_regdata->what[n] == 's') {
4308 SV *rv = (SV*)PL_regdata->data[n];
4309 AV *av = (AV*)SvRV((SV*)rv);
4310 SV **ary = AvARRAY(av);
4313 /* See the end of regcomp.c:S_reglass() for
4314 * documentation of these array elements. */
4317 a = SvTYPE(ary[1]) == SVt_RV ? &ary[1] : 0;
4318 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
4322 else if (si && doinit) {
4323 sw = swash_init("utf8", "", si, 1, 0);
4324 (void)av_store(av, 1, sw);
4340 - reginclass - determine if a character falls into a character class
4342 The n is the ANYOF regnode, the p is the target string, lenp
4343 is pointer to the maximum length of how far to go in the p
4344 (if the lenp is zero, UTF8SKIP(p) is used),
4345 do_utf8 tells whether the target string is in UTF-8.
4350 S_reginclass(pTHX_ register regnode *n, register U8* p, STRLEN* lenp, register bool do_utf8)
4352 char flags = ANYOF_FLAGS(n);
4358 if (do_utf8 && !UTF8_IS_INVARIANT(c))
4359 c = utf8n_to_uvchr(p, UTF8_MAXLEN, &len,
4360 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
4362 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
4363 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4366 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4367 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4370 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4374 SV *sw = regclass_swash(n, TRUE, 0, (SV**)&av);
4377 if (swash_fetch(sw, p, do_utf8))
4379 else if (flags & ANYOF_FOLD) {
4380 if (!match && lenp && av) {
4383 for (i = 0; i <= av_len(av); i++) {
4384 SV* sv = *av_fetch(av, i, FALSE);
4386 char *s = SvPV(sv, len);
4388 if (len <= plen && memEQ(s, (char*)p, len)) {
4396 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
4399 to_utf8_fold(p, tmpbuf, &tmplen);
4400 if (swash_fetch(sw, tmpbuf, do_utf8))
4406 if (match && lenp && *lenp == 0)
4407 *lenp = UNISKIP(NATIVE_TO_UNI(c));
4409 if (!match && c < 256) {
4410 if (ANYOF_BITMAP_TEST(n, c))
4412 else if (flags & ANYOF_FOLD) {
4415 if (flags & ANYOF_LOCALE) {
4416 PL_reg_flags |= RF_tainted;
4417 f = PL_fold_locale[c];
4421 if (f != c && ANYOF_BITMAP_TEST(n, f))
4425 if (!match && (flags & ANYOF_CLASS)) {
4426 PL_reg_flags |= RF_tainted;
4428 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4429 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4430 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4431 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4432 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4433 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4434 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4435 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4436 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4437 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4438 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4439 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4440 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4441 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4442 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4443 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4444 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4445 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4446 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4447 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4448 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4449 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4450 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4451 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4452 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4453 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4454 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4455 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4456 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4457 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4458 ) /* How's that for a conditional? */
4465 return (flags & ANYOF_INVERT) ? !match : match;
4469 S_reghop(pTHX_ U8 *s, I32 off)
4471 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4475 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4478 while (off-- && s < lim) {
4479 /* XXX could check well-formedness here */
4487 if (UTF8_IS_CONTINUED(*s)) {
4488 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4491 /* XXX could check well-formedness here */
4499 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4501 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4505 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4508 while (off-- && s < lim) {
4509 /* XXX could check well-formedness here */
4519 if (UTF8_IS_CONTINUED(*s)) {
4520 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4523 /* XXX could check well-formedness here */
4535 restore_pos(pTHX_ void *arg)
4537 if (PL_reg_eval_set) {
4538 if (PL_reg_oldsaved) {
4539 PL_reg_re->subbeg = PL_reg_oldsaved;
4540 PL_reg_re->sublen = PL_reg_oldsavedlen;
4541 RX_MATCH_COPIED_on(PL_reg_re);
4543 PL_reg_magic->mg_len = PL_reg_oldpos;
4544 PL_reg_eval_set = 0;
4545 PL_curpm = PL_reg_oldcurpm;
4550 S_to_utf8_substr(pTHX_ register regexp *prog)
4553 if (prog->float_substr && !prog->float_utf8) {
4554 prog->float_utf8 = sv = NEWSV(117, 0);
4555 SvSetSV(sv, prog->float_substr);
4556 sv_utf8_upgrade(sv);
4557 if (SvTAIL(prog->float_substr))
4559 if (prog->float_substr == prog->check_substr)
4560 prog->check_utf8 = sv;
4562 if (prog->anchored_substr && !prog->anchored_utf8) {
4563 prog->anchored_utf8 = sv = NEWSV(118, 0);
4564 SvSetSV(sv, prog->anchored_substr);
4565 sv_utf8_upgrade(sv);
4566 if (SvTAIL(prog->anchored_substr))
4568 if (prog->anchored_substr == prog->check_substr)
4569 prog->check_utf8 = sv;
4574 S_to_byte_substr(pTHX_ register regexp *prog)
4577 if (prog->float_utf8 && !prog->float_substr) {
4578 prog->float_substr = sv = NEWSV(117, 0);
4579 SvSetSV(sv, prog->float_utf8);
4580 if (sv_utf8_downgrade(sv, TRUE)) {
4581 if (SvTAIL(prog->float_utf8))
4585 prog->float_substr = sv = &PL_sv_undef;
4587 if (prog->float_utf8 == prog->check_utf8)
4588 prog->check_substr = sv;
4590 if (prog->anchored_utf8 && !prog->anchored_substr) {
4591 prog->anchored_substr = sv = NEWSV(118, 0);
4592 SvSetSV(sv, prog->anchored_utf8);
4593 if (sv_utf8_downgrade(sv, TRUE)) {
4594 if (SvTAIL(prog->anchored_utf8))
4598 prog->anchored_substr = sv = &PL_sv_undef;
4600 if (prog->anchored_utf8 == prog->check_utf8)
4601 prog->check_substr = sv;