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) 1991-2003, Larry Wall
72 **** You may distribute under the terms of either the GNU General Public
73 **** License or the Artistic License, as specified in the README file.
75 * Beware that some of this code is subtly aware of the way operator
76 * precedence is structured in regular expressions. Serious changes in
77 * regular-expression syntax might require a total rethink.
80 #define PERL_IN_REGEXEC_C
85 #define RF_tainted 1 /* tainted information used? */
86 #define RF_warned 2 /* warned about big count? */
87 #define RF_evaled 4 /* Did an EVAL with setting? */
88 #define RF_utf8 8 /* String contains multibyte chars? */
90 #define UTF ((PL_reg_flags & RF_utf8) != 0)
92 #define RS_init 1 /* eval environment created */
93 #define RS_set 2 /* replsv value is set */
99 #define REGINCLASS(p,c) (ANYOF_FLAGS(p) ? reginclass(p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c)))
105 #define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv))
106 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
108 #define reghop_c(pos,off) ((char*)reghop((U8*)pos, off))
109 #define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off))
110 #define HOP(pos,off) (PL_reg_match_utf8 ? reghop((U8*)pos, off) : (U8*)(pos + off))
111 #define HOPMAYBE(pos,off) (PL_reg_match_utf8 ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off))
112 #define HOPc(pos,off) ((char*)HOP(pos,off))
113 #define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off))
115 #define HOPBACK(pos, off) ( \
116 (PL_reg_match_utf8) \
117 ? reghopmaybe((U8*)pos, -off) \
118 : (pos - off >= PL_bostr) \
122 #define HOPBACKc(pos, off) (char*)HOPBACK(pos, off)
124 #define reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim))
125 #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
126 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
127 #define HOPMAYBE3(pos,off,lim) (PL_reg_match_utf8 ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
128 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
129 #define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim))
131 #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
133 /* for use after a quantifier and before an EXACT-like node -- japhy */
134 #define JUMPABLE(rn) ( \
135 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
136 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
137 OP(rn) == PLUS || OP(rn) == MINMOD || \
138 (PL_regkind[(U8)OP(rn)] == CURLY && ARG1(rn) > 0) \
141 #define HAS_TEXT(rn) ( \
142 PL_regkind[(U8)OP(rn)] == EXACT || PL_regkind[(U8)OP(rn)] == REF \
146 Search for mandatory following text node; for lookahead, the text must
147 follow but for lookbehind (rn->flags != 0) we skip to the next step.
149 #define FIND_NEXT_IMPT(rn) STMT_START { \
150 while (JUMPABLE(rn)) \
151 if (OP(rn) == SUSPEND || PL_regkind[(U8)OP(rn)] == CURLY) \
152 rn = NEXTOPER(NEXTOPER(rn)); \
153 else if (OP(rn) == PLUS) \
155 else if (OP(rn) == IFMATCH) \
156 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
157 else rn += NEXT_OFF(rn); \
160 static void restore_pos(pTHX_ void *arg);
163 S_regcppush(pTHX_ I32 parenfloor)
165 int retval = PL_savestack_ix;
166 #define REGCP_PAREN_ELEMS 4
167 int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
170 if (paren_elems_to_push < 0)
171 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
173 #define REGCP_OTHER_ELEMS 6
174 SSCHECK(paren_elems_to_push + REGCP_OTHER_ELEMS);
175 for (p = PL_regsize; p > parenfloor; p--) {
176 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
177 SSPUSHINT(PL_regendp[p]);
178 SSPUSHINT(PL_regstartp[p]);
179 SSPUSHPTR(PL_reg_start_tmp[p]);
182 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
183 SSPUSHINT(PL_regsize);
184 SSPUSHINT(*PL_reglastparen);
185 SSPUSHINT(*PL_reglastcloseparen);
186 SSPUSHPTR(PL_reginput);
187 #define REGCP_FRAME_ELEMS 2
188 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
189 * are needed for the regexp context stack bookkeeping. */
190 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
191 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
196 /* These are needed since we do not localize EVAL nodes: */
197 # define REGCP_SET(cp) DEBUG_r(PerlIO_printf(Perl_debug_log, \
198 " Setting an EVAL scope, savestack=%"IVdf"\n", \
199 (IV)PL_savestack_ix)); cp = PL_savestack_ix
201 # define REGCP_UNWIND(cp) DEBUG_r(cp != PL_savestack_ix ? \
202 PerlIO_printf(Perl_debug_log, \
203 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
204 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
214 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
216 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
217 i = SSPOPINT; /* Parentheses elements to pop. */
218 input = (char *) SSPOPPTR;
219 *PL_reglastcloseparen = SSPOPINT;
220 *PL_reglastparen = SSPOPINT;
221 PL_regsize = SSPOPINT;
223 /* Now restore the parentheses context. */
224 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
225 i > 0; i -= REGCP_PAREN_ELEMS) {
226 paren = (U32)SSPOPINT;
227 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
228 PL_regstartp[paren] = SSPOPINT;
230 if (paren <= *PL_reglastparen)
231 PL_regendp[paren] = tmps;
233 PerlIO_printf(Perl_debug_log,
234 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
235 (UV)paren, (IV)PL_regstartp[paren],
236 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
237 (IV)PL_regendp[paren],
238 (paren > *PL_reglastparen ? "(no)" : ""));
242 if ((I32)(*PL_reglastparen + 1) <= PL_regnpar) {
243 PerlIO_printf(Perl_debug_log,
244 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
245 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
249 /* It would seem that the similar code in regtry()
250 * already takes care of this, and in fact it is in
251 * a better location to since this code can #if 0-ed out
252 * but the code in regtry() is needed or otherwise tests
253 * requiring null fields (pat.t#187 and split.t#{13,14}
254 * (as of patchlevel 7877) will fail. Then again,
255 * this code seems to be necessary or otherwise
256 * building DynaLoader will fail:
257 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
259 for (paren = *PL_reglastparen + 1; (I32)paren <= PL_regnpar; paren++) {
260 if ((I32)paren > PL_regsize)
261 PL_regstartp[paren] = -1;
262 PL_regendp[paren] = -1;
269 S_regcp_set_to(pTHX_ I32 ss)
271 I32 tmp = PL_savestack_ix;
273 PL_savestack_ix = ss;
275 PL_savestack_ix = tmp;
279 typedef struct re_cc_state
283 struct re_cc_state *prev;
288 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
290 #define TRYPAREN(paren, n, input) { \
293 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
294 PL_regendp[paren] = input - PL_bostr; \
297 PL_regendp[paren] = -1; \
299 if (regmatch(next)) \
302 PL_regendp[paren] = -1; \
307 * pregexec and friends
311 - pregexec - match a regexp against a string
314 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
315 char *strbeg, I32 minend, SV *screamer, U32 nosave)
316 /* strend: pointer to null at end of string */
317 /* strbeg: real beginning of string */
318 /* minend: end of match must be >=minend after stringarg. */
319 /* nosave: For optimizations. */
322 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
323 nosave ? 0 : REXEC_COPY_STR);
327 S_cache_re(pTHX_ regexp *prog)
329 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
331 PL_regprogram = prog->program;
333 PL_regnpar = prog->nparens;
334 PL_regdata = prog->data;
339 * Need to implement the following flags for reg_anch:
341 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
343 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
344 * INTUIT_AUTORITATIVE_ML
345 * INTUIT_ONCE_NOML - Intuit can match in one location only.
348 * Another flag for this function: SECOND_TIME (so that float substrs
349 * with giant delta may be not rechecked).
352 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
354 /* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
355 Otherwise, only SvCUR(sv) is used to get strbeg. */
357 /* XXXX We assume that strpos is strbeg unless sv. */
359 /* XXXX Some places assume that there is a fixed substring.
360 An update may be needed if optimizer marks as "INTUITable"
361 RExen without fixed substrings. Similarly, it is assumed that
362 lengths of all the strings are no more than minlen, thus they
363 cannot come from lookahead.
364 (Or minlen should take into account lookahead.) */
366 /* A failure to find a constant substring means that there is no need to make
367 an expensive call to REx engine, thus we celebrate a failure. Similarly,
368 finding a substring too deep into the string means that less calls to
369 regtry() should be needed.
371 REx compiler's optimizer found 4 possible hints:
372 a) Anchored substring;
374 c) Whether we are anchored (beginning-of-line or \G);
375 d) First node (of those at offset 0) which may distingush positions;
376 We use a)b)d) and multiline-part of c), and try to find a position in the
377 string which does not contradict any of them.
380 /* Most of decisions we do here should have been done at compile time.
381 The nodes of the REx which we used for the search should have been
382 deleted from the finite automaton. */
385 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
386 char *strend, U32 flags, re_scream_pos_data *data)
388 register I32 start_shift = 0;
389 /* Should be nonnegative! */
390 register I32 end_shift = 0;
395 int do_utf8 = sv ? SvUTF8(sv) : 0; /* if no sv we have to assume bytes */
397 register char *other_last = Nullch; /* other substr checked before this */
398 char *check_at = Nullch; /* check substr found at this pos */
400 char *i_strpos = strpos;
401 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
403 RX_MATCH_UTF8_set(prog,do_utf8);
405 if (prog->reganch & ROPT_UTF8) {
406 DEBUG_r(PerlIO_printf(Perl_debug_log,
407 "UTF-8 regex...\n"));
408 PL_reg_flags |= RF_utf8;
412 char *s = PL_reg_match_utf8 ?
413 sv_uni_display(dsv, sv, 60, UNI_DISPLAY_REGEX) :
415 int len = PL_reg_match_utf8 ?
416 strlen(s) : strend - strpos;
419 if (PL_reg_match_utf8)
420 DEBUG_r(PerlIO_printf(Perl_debug_log,
421 "UTF-8 target...\n"));
422 PerlIO_printf(Perl_debug_log,
423 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
424 PL_colors[4],PL_colors[5],PL_colors[0],
427 (strlen(prog->precomp) > 60 ? "..." : ""),
429 (int)(len > 60 ? 60 : len),
431 (len > 60 ? "..." : "")
435 /* CHR_DIST() would be more correct here but it makes things slow. */
436 if (prog->minlen > strend - strpos) {
437 DEBUG_r(PerlIO_printf(Perl_debug_log,
438 "String too short... [re_intuit_start]\n"));
441 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
444 if (!prog->check_utf8 && prog->check_substr)
445 to_utf8_substr(prog);
446 check = prog->check_utf8;
448 if (!prog->check_substr && prog->check_utf8)
449 to_byte_substr(prog);
450 check = prog->check_substr;
452 if (check == &PL_sv_undef) {
453 DEBUG_r(PerlIO_printf(Perl_debug_log,
454 "Non-utf string cannot match utf check string\n"));
457 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
458 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
459 || ( (prog->reganch & ROPT_ANCH_BOL)
460 && !PL_multiline ) ); /* Check after \n? */
463 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
464 | ROPT_IMPLICIT)) /* not a real BOL */
465 /* SvCUR is not set on references: SvRV and SvPVX overlap */
467 && (strpos != strbeg)) {
468 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
471 if (prog->check_offset_min == prog->check_offset_max &&
472 !(prog->reganch & ROPT_CANY_SEEN)) {
473 /* Substring at constant offset from beg-of-str... */
476 s = HOP3c(strpos, prog->check_offset_min, strend);
478 slen = SvCUR(check); /* >= 1 */
480 if ( strend - s > slen || strend - s < slen - 1
481 || (strend - s == slen && strend[-1] != '\n')) {
482 DEBUG_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
485 /* Now should match s[0..slen-2] */
487 if (slen && (*SvPVX(check) != *s
489 && memNE(SvPVX(check), s, slen)))) {
491 DEBUG_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
495 else if (*SvPVX(check) != *s
496 || ((slen = SvCUR(check)) > 1
497 && memNE(SvPVX(check), s, slen)))
499 goto success_at_start;
502 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
504 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
505 end_shift = prog->minlen - start_shift -
506 CHR_SVLEN(check) + (SvTAIL(check) != 0);
508 I32 end = prog->check_offset_max + CHR_SVLEN(check)
509 - (SvTAIL(check) != 0);
510 I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
512 if (end_shift < eshift)
516 else { /* Can match at random position */
519 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
520 /* Should be nonnegative! */
521 end_shift = prog->minlen - start_shift -
522 CHR_SVLEN(check) + (SvTAIL(check) != 0);
525 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
527 Perl_croak(aTHX_ "panic: end_shift");
531 /* Find a possible match in the region s..strend by looking for
532 the "check" substring in the region corrected by start/end_shift. */
533 if (flags & REXEC_SCREAM) {
534 I32 p = -1; /* Internal iterator of scream. */
535 I32 *pp = data ? data->scream_pos : &p;
537 if (PL_screamfirst[BmRARE(check)] >= 0
538 || ( BmRARE(check) == '\n'
539 && (BmPREVIOUS(check) == SvCUR(check) - 1)
541 s = screaminstr(sv, check,
542 start_shift + (s - strbeg), end_shift, pp, 0);
545 /* we may be pointing at the wrong string */
546 if (s && RX_MATCH_COPIED(prog))
547 s = strbeg + (s - SvPVX(sv));
549 *data->scream_olds = s;
551 else if (prog->reganch & ROPT_CANY_SEEN)
552 s = fbm_instr((U8*)(s + start_shift),
553 (U8*)(strend - end_shift),
554 check, PL_multiline ? FBMrf_MULTILINE : 0);
556 s = fbm_instr(HOP3(s, start_shift, strend),
557 HOP3(strend, -end_shift, strbeg),
558 check, PL_multiline ? FBMrf_MULTILINE : 0);
560 /* Update the count-of-usability, remove useless subpatterns,
563 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
564 (s ? "Found" : "Did not find"),
565 (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) ? "anchored" : "floating"),
567 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
569 PL_colors[1], (SvTAIL(check) ? "$" : ""),
570 (s ? " at offset " : "...\n") ) );
577 /* Finish the diagnostic message */
578 DEBUG_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
580 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
581 Start with the other substr.
582 XXXX no SCREAM optimization yet - and a very coarse implementation
583 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
584 *always* match. Probably should be marked during compile...
585 Probably it is right to do no SCREAM here...
588 if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8) : (prog->float_substr && prog->anchored_substr)) {
589 /* Take into account the "other" substring. */
590 /* XXXX May be hopelessly wrong for UTF... */
593 if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
596 char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
600 t = s - prog->check_offset_max;
601 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
603 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
608 t = HOP3c(t, prog->anchored_offset, strend);
609 if (t < other_last) /* These positions already checked */
611 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
614 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
615 /* On end-of-str: see comment below. */
616 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
617 if (must == &PL_sv_undef) {
619 DEBUG_r(must = prog->anchored_utf8); /* for debug */
624 HOP3(HOP3(last1, prog->anchored_offset, strend)
625 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
627 PL_multiline ? FBMrf_MULTILINE : 0
629 DEBUG_r(PerlIO_printf(Perl_debug_log,
630 "%s anchored substr `%s%.*s%s'%s",
631 (s ? "Found" : "Contradicts"),
634 - (SvTAIL(must)!=0)),
636 PL_colors[1], (SvTAIL(must) ? "$" : "")));
638 if (last1 >= last2) {
639 DEBUG_r(PerlIO_printf(Perl_debug_log,
640 ", giving up...\n"));
643 DEBUG_r(PerlIO_printf(Perl_debug_log,
644 ", trying floating at offset %ld...\n",
645 (long)(HOP3c(s1, 1, strend) - i_strpos)));
646 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
647 s = HOP3c(last, 1, strend);
651 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
652 (long)(s - i_strpos)));
653 t = HOP3c(s, -prog->anchored_offset, strbeg);
654 other_last = HOP3c(s, 1, strend);
662 else { /* Take into account the floating substring. */
667 t = HOP3c(s, -start_shift, strbeg);
669 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
670 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
671 last = HOP3c(t, prog->float_max_offset, strend);
672 s = HOP3c(t, prog->float_min_offset, strend);
675 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
676 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
677 /* fbm_instr() takes into account exact value of end-of-str
678 if the check is SvTAIL(ed). Since false positives are OK,
679 and end-of-str is not later than strend we are OK. */
680 if (must == &PL_sv_undef) {
682 DEBUG_r(must = prog->float_utf8); /* for debug message */
685 s = fbm_instr((unsigned char*)s,
686 (unsigned char*)last + SvCUR(must)
688 must, PL_multiline ? FBMrf_MULTILINE : 0);
689 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
690 (s ? "Found" : "Contradicts"),
692 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
694 PL_colors[1], (SvTAIL(must) ? "$" : "")));
697 DEBUG_r(PerlIO_printf(Perl_debug_log,
698 ", giving up...\n"));
701 DEBUG_r(PerlIO_printf(Perl_debug_log,
702 ", trying anchored starting at offset %ld...\n",
703 (long)(s1 + 1 - i_strpos)));
705 s = HOP3c(t, 1, strend);
709 DEBUG_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
710 (long)(s - i_strpos)));
711 other_last = s; /* Fix this later. --Hugo */
720 t = s - prog->check_offset_max;
721 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
723 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
725 /* Fixed substring is found far enough so that the match
726 cannot start at strpos. */
728 if (ml_anch && t[-1] != '\n') {
729 /* Eventually fbm_*() should handle this, but often
730 anchored_offset is not 0, so this check will not be wasted. */
731 /* XXXX In the code below we prefer to look for "^" even in
732 presence of anchored substrings. And we search even
733 beyond the found float position. These pessimizations
734 are historical artefacts only. */
736 while (t < strend - prog->minlen) {
738 if (t < check_at - prog->check_offset_min) {
739 if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
740 /* Since we moved from the found position,
741 we definitely contradict the found anchored
742 substr. Due to the above check we do not
743 contradict "check" substr.
744 Thus we can arrive here only if check substr
745 is float. Redo checking for "other"=="fixed".
748 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
749 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
750 goto do_other_anchored;
752 /* We don't contradict the found floating substring. */
753 /* XXXX Why not check for STCLASS? */
755 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
756 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
759 /* Position contradicts check-string */
760 /* XXXX probably better to look for check-string
761 than for "\n", so one should lower the limit for t? */
762 DEBUG_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
763 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
764 other_last = strpos = s = t + 1;
769 DEBUG_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
770 PL_colors[0],PL_colors[1]));
774 DEBUG_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
775 PL_colors[0],PL_colors[1]));
779 ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
782 /* The found string does not prohibit matching at strpos,
783 - no optimization of calling REx engine can be performed,
784 unless it was an MBOL and we are not after MBOL,
785 or a future STCLASS check will fail this. */
787 /* Even in this situation we may use MBOL flag if strpos is offset
788 wrt the start of the string. */
789 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
790 && (strpos != strbeg) && strpos[-1] != '\n'
791 /* May be due to an implicit anchor of m{.*foo} */
792 && !(prog->reganch & ROPT_IMPLICIT))
797 DEBUG_r( if (ml_anch)
798 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
799 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
802 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
804 prog->check_utf8 /* Could be deleted already */
805 && --BmUSEFUL(prog->check_utf8) < 0
806 && (prog->check_utf8 == prog->float_utf8)
808 prog->check_substr /* Could be deleted already */
809 && --BmUSEFUL(prog->check_substr) < 0
810 && (prog->check_substr == prog->float_substr)
813 /* If flags & SOMETHING - do not do it many times on the same match */
814 DEBUG_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
815 SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr);
816 if (do_utf8 ? prog->check_substr : prog->check_utf8)
817 SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8);
818 prog->check_substr = prog->check_utf8 = Nullsv; /* disable */
819 prog->float_substr = prog->float_utf8 = Nullsv; /* clear */
820 check = Nullsv; /* abort */
822 /* XXXX This is a remnant of the old implementation. It
823 looks wasteful, since now INTUIT can use many
825 prog->reganch &= ~RE_USE_INTUIT;
832 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
833 if (prog->regstclass) {
834 /* minlen == 0 is possible if regstclass is \b or \B,
835 and the fixed substr is ''$.
836 Since minlen is already taken into account, s+1 is before strend;
837 accidentally, minlen >= 1 guaranties no false positives at s + 1
838 even for \b or \B. But (minlen? 1 : 0) below assumes that
839 regstclass does not come from lookahead... */
840 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
841 This leaves EXACTF only, which is dealt with in find_byclass(). */
842 U8* str = (U8*)STRING(prog->regstclass);
843 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
844 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
846 char *endpos = (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
847 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
848 : (prog->float_substr || prog->float_utf8
849 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
852 char *startpos = strbeg;
856 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
861 if (endpos == strend) {
862 DEBUG_r( PerlIO_printf(Perl_debug_log,
863 "Could not match STCLASS...\n") );
866 DEBUG_r( PerlIO_printf(Perl_debug_log,
867 "This position contradicts STCLASS...\n") );
868 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
870 /* Contradict one of substrings */
871 if (prog->anchored_substr || prog->anchored_utf8) {
872 if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
873 DEBUG_r( what = "anchored" );
875 s = HOP3c(t, 1, strend);
876 if (s + start_shift + end_shift > strend) {
877 /* XXXX Should be taken into account earlier? */
878 DEBUG_r( PerlIO_printf(Perl_debug_log,
879 "Could not match STCLASS...\n") );
884 DEBUG_r( PerlIO_printf(Perl_debug_log,
885 "Looking for %s substr starting at offset %ld...\n",
886 what, (long)(s + start_shift - i_strpos)) );
889 /* Have both, check_string is floating */
890 if (t + start_shift >= check_at) /* Contradicts floating=check */
891 goto retry_floating_check;
892 /* Recheck anchored substring, but not floating... */
896 DEBUG_r( PerlIO_printf(Perl_debug_log,
897 "Looking for anchored substr starting at offset %ld...\n",
898 (long)(other_last - i_strpos)) );
899 goto do_other_anchored;
901 /* Another way we could have checked stclass at the
902 current position only: */
907 DEBUG_r( PerlIO_printf(Perl_debug_log,
908 "Looking for /%s^%s/m starting at offset %ld...\n",
909 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
912 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
914 /* Check is floating subtring. */
915 retry_floating_check:
916 t = check_at - start_shift;
917 DEBUG_r( what = "floating" );
918 goto hop_and_restart;
921 DEBUG_r(PerlIO_printf(Perl_debug_log,
922 "By STCLASS: moving %ld --> %ld\n",
923 (long)(t - i_strpos), (long)(s - i_strpos))
927 DEBUG_r(PerlIO_printf(Perl_debug_log,
928 "Does not contradict STCLASS...\n");
933 DEBUG_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
934 PL_colors[4], (check ? "Guessed" : "Giving up"),
935 PL_colors[5], (long)(s - i_strpos)) );
938 fail_finish: /* Substring not found */
939 if (prog->check_substr || prog->check_utf8) /* could be removed already */
940 BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
942 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
943 PL_colors[4],PL_colors[5]));
947 /* We know what class REx starts with. Try to find this position... */
949 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
951 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
957 register I32 tmp = 1; /* Scratch variable? */
958 register bool do_utf8 = PL_reg_match_utf8;
960 /* We know what class it must start with. */
965 if ((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
966 !UTF8_IS_INVARIANT((U8)s[0]) ?
967 reginclass(c, (U8*)s, 0, do_utf8) :
968 REGINCLASS(c, (U8*)s)) {
969 if (tmp && (norun || regtry(prog, s)))
983 if (REGINCLASS(c, (U8*)s) ||
984 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
985 /* The assignment of 2 is intentional:
986 * for the folded sharp s, the skip is 2. */
987 (skip = SHARP_S_SKIP))) {
988 if (tmp && (norun || regtry(prog, s)))
1000 while (s < strend) {
1001 if (tmp && (norun || regtry(prog, s)))
1012 STRLEN ulen1, ulen2;
1013 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
1014 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
1016 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1017 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1019 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXLEN_UCLC,
1020 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1021 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXLEN_UCLC,
1022 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1033 c2 = PL_fold_locale[c1];
1035 e = HOP3c(strend, -(I32)ln, s);
1038 e = s; /* Due to minlen logic of intuit() */
1040 /* The idea in the EXACTF* cases is to first find the
1041 * first character of the EXACTF* node and then, if
1042 * necessary, case-insensitively compare the full
1043 * text of the node. The c1 and c2 are the first
1044 * characters (though in Unicode it gets a bit
1045 * more complicated because there are more cases
1046 * than just upper and lower: one needs to use
1047 * the so-called folding case for case-insensitive
1048 * matching (called "loose matching" in Unicode).
1049 * ibcmp_utf8() will do just that. */
1053 U8 tmpbuf [UTF8_MAXLEN+1];
1054 U8 foldbuf[UTF8_MAXLEN_FOLD+1];
1055 STRLEN len, foldlen;
1059 c = utf8n_to_uvchr((U8*)s, UTF8_MAXLEN, &len,
1061 0 : UTF8_ALLOW_ANY);
1064 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1065 m, (char **)0, ln, (bool)UTF))
1066 && (norun || regtry(prog, s)) )
1069 uvchr_to_utf8(tmpbuf, c);
1070 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1072 && (f == c1 || f == c2)
1073 && (ln == foldlen ||
1074 !ibcmp_utf8((char *) foldbuf,
1075 (char **)0, foldlen, do_utf8,
1077 (char **)0, ln, (bool)UTF))
1078 && (norun || regtry(prog, s)) )
1086 c = utf8n_to_uvchr((U8*)s, UTF8_MAXLEN, &len,
1088 0 : UTF8_ALLOW_ANY);
1090 /* Handle some of the three Greek sigmas cases.
1091 * Note that not all the possible combinations
1092 * are handled here: some of them are handled
1093 * by the standard folding rules, and some of
1094 * them (the character class or ANYOF cases)
1095 * are handled during compiletime in
1096 * regexec.c:S_regclass(). */
1097 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1098 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1099 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1101 if ( (c == c1 || c == c2)
1103 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1104 m, (char **)0, ln, (bool)UTF))
1105 && (norun || regtry(prog, s)) )
1108 uvchr_to_utf8(tmpbuf, c);
1109 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1111 && (f == c1 || f == c2)
1112 && (ln == foldlen ||
1113 !ibcmp_utf8((char *) foldbuf,
1114 (char **)0, foldlen, do_utf8,
1116 (char **)0, ln, (bool)UTF))
1117 && (norun || regtry(prog, s)) )
1128 && (ln == 1 || !(OP(c) == EXACTF
1130 : ibcmp_locale(s, m, ln)))
1131 && (norun || regtry(prog, s)) )
1137 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1138 && (ln == 1 || !(OP(c) == EXACTF
1140 : ibcmp_locale(s, m, ln)))
1141 && (norun || regtry(prog, s)) )
1148 PL_reg_flags |= RF_tainted;
1155 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1157 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1159 tmp = ((OP(c) == BOUND ?
1160 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1161 LOAD_UTF8_CHARCLASS(alnum,"a");
1162 while (s < strend) {
1163 if (tmp == !(OP(c) == BOUND ?
1164 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1165 isALNUM_LC_utf8((U8*)s)))
1168 if ((norun || regtry(prog, s)))
1175 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1176 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1177 while (s < strend) {
1179 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1181 if ((norun || regtry(prog, s)))
1187 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1191 PL_reg_flags |= RF_tainted;
1198 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1200 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1202 tmp = ((OP(c) == NBOUND ?
1203 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1204 LOAD_UTF8_CHARCLASS(alnum,"a");
1205 while (s < strend) {
1206 if (tmp == !(OP(c) == NBOUND ?
1207 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1208 isALNUM_LC_utf8((U8*)s)))
1210 else if ((norun || regtry(prog, s)))
1216 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1217 tmp = ((OP(c) == NBOUND ?
1218 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1219 while (s < strend) {
1221 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1223 else if ((norun || regtry(prog, s)))
1228 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1233 LOAD_UTF8_CHARCLASS(alnum,"a");
1234 while (s < strend) {
1235 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1236 if (tmp && (norun || regtry(prog, s)))
1247 while (s < strend) {
1249 if (tmp && (norun || regtry(prog, s)))
1261 PL_reg_flags |= RF_tainted;
1263 while (s < strend) {
1264 if (isALNUM_LC_utf8((U8*)s)) {
1265 if (tmp && (norun || regtry(prog, s)))
1276 while (s < strend) {
1277 if (isALNUM_LC(*s)) {
1278 if (tmp && (norun || regtry(prog, s)))
1291 LOAD_UTF8_CHARCLASS(alnum,"a");
1292 while (s < strend) {
1293 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1294 if (tmp && (norun || regtry(prog, s)))
1305 while (s < strend) {
1307 if (tmp && (norun || regtry(prog, s)))
1319 PL_reg_flags |= RF_tainted;
1321 while (s < strend) {
1322 if (!isALNUM_LC_utf8((U8*)s)) {
1323 if (tmp && (norun || regtry(prog, s)))
1334 while (s < strend) {
1335 if (!isALNUM_LC(*s)) {
1336 if (tmp && (norun || regtry(prog, s)))
1349 LOAD_UTF8_CHARCLASS(space," ");
1350 while (s < strend) {
1351 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1352 if (tmp && (norun || regtry(prog, s)))
1363 while (s < strend) {
1365 if (tmp && (norun || regtry(prog, s)))
1377 PL_reg_flags |= RF_tainted;
1379 while (s < strend) {
1380 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1381 if (tmp && (norun || regtry(prog, s)))
1392 while (s < strend) {
1393 if (isSPACE_LC(*s)) {
1394 if (tmp && (norun || regtry(prog, s)))
1407 LOAD_UTF8_CHARCLASS(space," ");
1408 while (s < strend) {
1409 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1410 if (tmp && (norun || regtry(prog, s)))
1421 while (s < strend) {
1423 if (tmp && (norun || regtry(prog, s)))
1435 PL_reg_flags |= RF_tainted;
1437 while (s < strend) {
1438 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1439 if (tmp && (norun || regtry(prog, s)))
1450 while (s < strend) {
1451 if (!isSPACE_LC(*s)) {
1452 if (tmp && (norun || regtry(prog, s)))
1465 LOAD_UTF8_CHARCLASS(digit,"0");
1466 while (s < strend) {
1467 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1468 if (tmp && (norun || regtry(prog, s)))
1479 while (s < strend) {
1481 if (tmp && (norun || regtry(prog, s)))
1493 PL_reg_flags |= RF_tainted;
1495 while (s < strend) {
1496 if (isDIGIT_LC_utf8((U8*)s)) {
1497 if (tmp && (norun || regtry(prog, s)))
1508 while (s < strend) {
1509 if (isDIGIT_LC(*s)) {
1510 if (tmp && (norun || regtry(prog, s)))
1523 LOAD_UTF8_CHARCLASS(digit,"0");
1524 while (s < strend) {
1525 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1526 if (tmp && (norun || regtry(prog, s)))
1537 while (s < strend) {
1539 if (tmp && (norun || regtry(prog, s)))
1551 PL_reg_flags |= RF_tainted;
1553 while (s < strend) {
1554 if (!isDIGIT_LC_utf8((U8*)s)) {
1555 if (tmp && (norun || regtry(prog, s)))
1566 while (s < strend) {
1567 if (!isDIGIT_LC(*s)) {
1568 if (tmp && (norun || regtry(prog, s)))
1580 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1589 - regexec_flags - match a regexp against a string
1592 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1593 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1594 /* strend: pointer to null at end of string */
1595 /* strbeg: real beginning of string */
1596 /* minend: end of match must be >=minend after stringarg. */
1597 /* data: May be used for some additional optimizations. */
1598 /* nosave: For optimizations. */
1601 register regnode *c;
1602 register char *startpos = stringarg;
1603 I32 minlen; /* must match at least this many chars */
1604 I32 dontbother = 0; /* how many characters not to try at end */
1605 /* I32 start_shift = 0; */ /* Offset of the start to find
1606 constant substr. */ /* CC */
1607 I32 end_shift = 0; /* Same for the end. */ /* CC */
1608 I32 scream_pos = -1; /* Internal iterator of scream. */
1610 SV* oreplsv = GvSV(PL_replgv);
1611 bool do_utf8 = DO_UTF8(sv);
1613 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
1614 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
1616 RX_MATCH_UTF8_set(prog,do_utf8);
1622 PL_regnarrate = DEBUG_r_TEST;
1625 /* Be paranoid... */
1626 if (prog == NULL || startpos == NULL) {
1627 Perl_croak(aTHX_ "NULL regexp parameter");
1631 minlen = prog->minlen;
1632 if (strend - startpos < minlen) {
1633 DEBUG_r(PerlIO_printf(Perl_debug_log,
1634 "String too short [regexec_flags]...\n"));
1638 /* Check validity of program. */
1639 if (UCHARAT(prog->program) != REG_MAGIC) {
1640 Perl_croak(aTHX_ "corrupted regexp program");
1644 PL_reg_eval_set = 0;
1647 if (prog->reganch & ROPT_UTF8)
1648 PL_reg_flags |= RF_utf8;
1650 /* Mark beginning of line for ^ and lookbehind. */
1651 PL_regbol = startpos;
1655 /* Mark end of line for $ (and such) */
1658 /* see how far we have to get to not match where we matched before */
1659 PL_regtill = startpos+minend;
1661 /* We start without call_cc context. */
1664 /* If there is a "must appear" string, look for it. */
1667 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1670 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1671 PL_reg_ganch = startpos;
1672 else if (sv && SvTYPE(sv) >= SVt_PVMG
1674 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1675 && mg->mg_len >= 0) {
1676 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1677 if (prog->reganch & ROPT_ANCH_GPOS) {
1678 if (s > PL_reg_ganch)
1683 else /* pos() not defined */
1684 PL_reg_ganch = strbeg;
1687 if (!(flags & REXEC_CHECKED) && (prog->check_substr != Nullsv || prog->check_utf8 != Nullsv)) {
1688 re_scream_pos_data d;
1690 d.scream_olds = &scream_olds;
1691 d.scream_pos = &scream_pos;
1692 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1694 DEBUG_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1695 goto phooey; /* not present */
1701 pv_uni_display(dsv0, (U8*)prog->precomp, prog->prelen, 60,
1702 UNI_DISPLAY_REGEX) :
1704 int len0 = UTF ? SvCUR(dsv0) : prog->prelen;
1705 char *s1 = do_utf8 ? sv_uni_display(dsv1, sv, 60,
1706 UNI_DISPLAY_REGEX) : startpos;
1707 int len1 = do_utf8 ? SvCUR(dsv1) : strend - startpos;
1710 PerlIO_printf(Perl_debug_log,
1711 "%sMatching REx%s `%s%*.*s%s%s' against `%s%.*s%s%s'\n",
1712 PL_colors[4],PL_colors[5],PL_colors[0],
1715 len0 > 60 ? "..." : "",
1717 (int)(len1 > 60 ? 60 : len1),
1719 (len1 > 60 ? "..." : "")
1723 /* Simplest case: anchored match need be tried only once. */
1724 /* [unless only anchor is BOL and multiline is set] */
1725 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1726 if (s == startpos && regtry(prog, startpos))
1728 else if (PL_multiline || (prog->reganch & ROPT_IMPLICIT)
1729 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1734 dontbother = minlen - 1;
1735 end = HOP3c(strend, -dontbother, strbeg) - 1;
1736 /* for multiline we only have to try after newlines */
1737 if (prog->check_substr || prog->check_utf8) {
1741 if (regtry(prog, s))
1746 if (prog->reganch & RE_USE_INTUIT) {
1747 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1758 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1759 if (regtry(prog, s))
1766 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1767 if (regtry(prog, PL_reg_ganch))
1772 /* Messy cases: unanchored match. */
1773 if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) {
1774 /* we have /x+whatever/ */
1775 /* it must be a one character string (XXXX Except UTF?) */
1780 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1781 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1782 ch = SvPVX(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
1785 while (s < strend) {
1787 DEBUG_r( did_match = 1 );
1788 if (regtry(prog, s)) goto got_it;
1790 while (s < strend && *s == ch)
1797 while (s < strend) {
1799 DEBUG_r( did_match = 1 );
1800 if (regtry(prog, s)) goto got_it;
1802 while (s < strend && *s == ch)
1808 DEBUG_r(if (!did_match)
1809 PerlIO_printf(Perl_debug_log,
1810 "Did not find anchored character...\n")
1814 else if (prog->anchored_substr != Nullsv
1815 || prog->anchored_utf8 != Nullsv
1816 || ((prog->float_substr != Nullsv || prog->float_utf8 != Nullsv)
1817 && prog->float_max_offset < strend - s)) {
1822 char *last1; /* Last position checked before */
1826 if (prog->anchored_substr || prog->anchored_utf8) {
1827 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1828 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1829 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1830 back_max = back_min = prog->anchored_offset;
1832 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1833 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1834 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1835 back_max = prog->float_max_offset;
1836 back_min = prog->float_min_offset;
1838 if (must == &PL_sv_undef)
1839 /* could not downgrade utf8 check substring, so must fail */
1842 last = HOP3c(strend, /* Cannot start after this */
1843 -(I32)(CHR_SVLEN(must)
1844 - (SvTAIL(must) != 0) + back_min), strbeg);
1847 last1 = HOPc(s, -1);
1849 last1 = s - 1; /* bogus */
1851 /* XXXX check_substr already used to find `s', can optimize if
1852 check_substr==must. */
1854 dontbother = end_shift;
1855 strend = HOPc(strend, -dontbother);
1856 while ( (s <= last) &&
1857 ((flags & REXEC_SCREAM)
1858 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1859 end_shift, &scream_pos, 0))
1860 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1861 (unsigned char*)strend, must,
1862 PL_multiline ? FBMrf_MULTILINE : 0))) ) {
1863 /* we may be pointing at the wrong string */
1864 if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog))
1865 s = strbeg + (s - SvPVX(sv));
1866 DEBUG_r( did_match = 1 );
1867 if (HOPc(s, -back_max) > last1) {
1868 last1 = HOPc(s, -back_min);
1869 s = HOPc(s, -back_max);
1872 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1874 last1 = HOPc(s, -back_min);
1878 while (s <= last1) {
1879 if (regtry(prog, s))
1885 while (s <= last1) {
1886 if (regtry(prog, s))
1892 DEBUG_r(if (!did_match)
1893 PerlIO_printf(Perl_debug_log,
1894 "Did not find %s substr `%s%.*s%s'%s...\n",
1895 ((must == prog->anchored_substr || must == prog->anchored_utf8)
1896 ? "anchored" : "floating"),
1898 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1900 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1904 else if ((c = prog->regstclass)) {
1906 I32 op = (U8)OP(prog->regstclass);
1907 /* don't bother with what can't match */
1908 if (PL_regkind[op] != EXACT && op != CANY)
1909 strend = HOPc(strend, -(minlen - 1));
1912 SV *prop = sv_newmortal();
1920 pv_uni_display(dsv0, (U8*)SvPVX(prop), SvCUR(prop), 60,
1921 UNI_DISPLAY_REGEX) :
1923 len0 = UTF ? SvCUR(dsv0) : SvCUR(prop);
1925 sv_uni_display(dsv1, sv, 60, UNI_DISPLAY_REGEX) : s;
1926 len1 = UTF ? SvCUR(dsv1) : strend - s;
1927 PerlIO_printf(Perl_debug_log,
1928 "Matching stclass `%*.*s' against `%*.*s'\n",
1932 if (find_byclass(prog, c, s, strend, startpos, 0))
1934 DEBUG_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1938 if (prog->float_substr != Nullsv || prog->float_utf8 != Nullsv) {
1943 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1944 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1945 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
1947 if (flags & REXEC_SCREAM) {
1948 last = screaminstr(sv, float_real, s - strbeg,
1949 end_shift, &scream_pos, 1); /* last one */
1951 last = scream_olds; /* Only one occurrence. */
1952 /* we may be pointing at the wrong string */
1953 else if (RX_MATCH_COPIED(prog))
1954 s = strbeg + (s - SvPVX(sv));
1958 char *little = SvPV(float_real, len);
1960 if (SvTAIL(float_real)) {
1961 if (memEQ(strend - len + 1, little, len - 1))
1962 last = strend - len + 1;
1963 else if (!PL_multiline)
1964 last = memEQ(strend - len, little, len)
1965 ? strend - len : Nullch;
1971 last = rninstr(s, strend, little, little + len);
1973 last = strend; /* matching `$' */
1977 DEBUG_r(PerlIO_printf(Perl_debug_log,
1978 "%sCan't trim the tail, match fails (should not happen)%s\n",
1979 PL_colors[4],PL_colors[5]));
1980 goto phooey; /* Should not happen! */
1982 dontbother = strend - last + prog->float_min_offset;
1984 if (minlen && (dontbother < minlen))
1985 dontbother = minlen - 1;
1986 strend -= dontbother; /* this one's always in bytes! */
1987 /* We don't know much -- general case. */
1990 if (regtry(prog, s))
1999 if (regtry(prog, s))
2001 } while (s++ < strend);
2009 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
2011 if (PL_reg_eval_set) {
2012 /* Preserve the current value of $^R */
2013 if (oreplsv != GvSV(PL_replgv))
2014 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
2015 restored, the value remains
2017 restore_pos(aTHX_ 0);
2020 /* make sure $`, $&, $', and $digit will work later */
2021 if ( !(flags & REXEC_NOT_FIRST) ) {
2022 if (RX_MATCH_COPIED(prog)) {
2023 Safefree(prog->subbeg);
2024 RX_MATCH_COPIED_off(prog);
2026 if (flags & REXEC_COPY_STR) {
2027 I32 i = PL_regeol - startpos + (stringarg - strbeg);
2029 s = savepvn(strbeg, i);
2032 RX_MATCH_COPIED_on(prog);
2035 prog->subbeg = strbeg;
2036 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2043 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2044 PL_colors[4],PL_colors[5]));
2045 if (PL_reg_eval_set)
2046 restore_pos(aTHX_ 0);
2051 - regtry - try match at specific point
2053 STATIC I32 /* 0 failure, 1 success */
2054 S_regtry(pTHX_ regexp *prog, char *startpos)
2062 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
2064 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
2067 PL_reg_eval_set = RS_init;
2069 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2070 (IV)(PL_stack_sp - PL_stack_base));
2072 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
2073 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2074 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2076 /* Apparently this is not needed, judging by wantarray. */
2077 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
2078 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2081 /* Make $_ available to executed code. */
2082 if (PL_reg_sv != DEFSV) {
2083 /* SAVE_DEFSV does *not* suffice here for USE_5005THREADS */
2088 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
2089 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
2090 /* prepare for quick setting of pos */
2091 sv_magic(PL_reg_sv, (SV*)0,
2092 PERL_MAGIC_regex_global, Nullch, 0);
2093 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
2097 PL_reg_oldpos = mg->mg_len;
2098 SAVEDESTRUCTOR_X(restore_pos, 0);
2100 if (!PL_reg_curpm) {
2101 Newz(22,PL_reg_curpm, 1, PMOP);
2104 SV* repointer = newSViv(0);
2105 /* so we know which PL_regex_padav element is PL_reg_curpm */
2106 SvFLAGS(repointer) |= SVf_BREAK;
2107 av_push(PL_regex_padav,repointer);
2108 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2109 PL_regex_pad = AvARRAY(PL_regex_padav);
2113 PM_SETRE(PL_reg_curpm, prog);
2114 PL_reg_oldcurpm = PL_curpm;
2115 PL_curpm = PL_reg_curpm;
2116 if (RX_MATCH_COPIED(prog)) {
2117 /* Here is a serious problem: we cannot rewrite subbeg,
2118 since it may be needed if this match fails. Thus
2119 $` inside (?{}) could fail... */
2120 PL_reg_oldsaved = prog->subbeg;
2121 PL_reg_oldsavedlen = prog->sublen;
2122 RX_MATCH_COPIED_off(prog);
2125 PL_reg_oldsaved = Nullch;
2126 prog->subbeg = PL_bostr;
2127 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2129 prog->startp[0] = startpos - PL_bostr;
2130 PL_reginput = startpos;
2131 PL_regstartp = prog->startp;
2132 PL_regendp = prog->endp;
2133 PL_reglastparen = &prog->lastparen;
2134 PL_reglastcloseparen = &prog->lastcloseparen;
2135 prog->lastparen = 0;
2136 prog->lastcloseparen = 0;
2138 DEBUG_r(PL_reg_starttry = startpos);
2139 if (PL_reg_start_tmpl <= prog->nparens) {
2140 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2141 if(PL_reg_start_tmp)
2142 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2144 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2147 /* XXXX What this code is doing here?!!! There should be no need
2148 to do this again and again, PL_reglastparen should take care of
2151 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2152 * Actually, the code in regcppop() (which Ilya may be meaning by
2153 * PL_reglastparen), is not needed at all by the test suite
2154 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2155 * enough, for building DynaLoader, or otherwise this
2156 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2157 * will happen. Meanwhile, this code *is* needed for the
2158 * above-mentioned test suite tests to succeed. The common theme
2159 * on those tests seems to be returning null fields from matches.
2164 if (prog->nparens) {
2165 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2172 if (regmatch(prog->program + 1)) {
2173 prog->endp[0] = PL_reginput - PL_bostr;
2176 REGCP_UNWIND(lastcp);
2180 #define RE_UNWIND_BRANCH 1
2181 #define RE_UNWIND_BRANCHJ 2
2185 typedef struct { /* XX: makes sense to enlarge it... */
2189 } re_unwind_generic_t;
2202 } re_unwind_branch_t;
2204 typedef union re_unwind_t {
2206 re_unwind_generic_t generic;
2207 re_unwind_branch_t branch;
2210 #define sayYES goto yes
2211 #define sayNO goto no
2212 #define sayNO_ANYOF goto no_anyof
2213 #define sayYES_FINAL goto yes_final
2214 #define sayYES_LOUD goto yes_loud
2215 #define sayNO_FINAL goto no_final
2216 #define sayNO_SILENT goto do_no
2217 #define saySAME(x) if (x) goto yes; else goto no
2219 #define REPORT_CODE_OFF 24
2222 - regmatch - main matching routine
2224 * Conceptually the strategy is simple: check to see whether the current
2225 * node matches, call self recursively to see whether the rest matches,
2226 * and then act accordingly. In practice we make some effort to avoid
2227 * recursion, in particular by going through "ordinary" nodes (that don't
2228 * need to know whether the rest of the match failed) by a loop instead of
2231 /* [lwall] I've hoisted the register declarations to the outer block in order to
2232 * maybe save a little bit of pushing and popping on the stack. It also takes
2233 * advantage of machines that use a register save mask on subroutine entry.
2235 STATIC I32 /* 0 failure, 1 success */
2236 S_regmatch(pTHX_ regnode *prog)
2238 register regnode *scan; /* Current node. */
2239 regnode *next; /* Next node. */
2240 regnode *inner; /* Next node in internal branch. */
2241 register I32 nextchr; /* renamed nextchr - nextchar colides with
2242 function of same name */
2243 register I32 n; /* no or next */
2244 register I32 ln = 0; /* len or last */
2245 register char *s = Nullch; /* operand or save */
2246 register char *locinput = PL_reginput;
2247 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2248 int minmod = 0, sw = 0, logical = 0;
2251 I32 firstcp = PL_savestack_ix;
2253 register bool do_utf8 = PL_reg_match_utf8;
2255 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2256 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2257 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2264 /* Note that nextchr is a byte even in UTF */
2265 nextchr = UCHARAT(locinput);
2267 while (scan != NULL) {
2270 SV *prop = sv_newmortal();
2271 int docolor = *PL_colors[0];
2272 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2273 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2274 /* The part of the string before starttry has one color
2275 (pref0_len chars), between starttry and current
2276 position another one (pref_len - pref0_len chars),
2277 after the current position the third one.
2278 We assume that pref0_len <= pref_len, otherwise we
2279 decrease pref0_len. */
2280 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2281 ? (5 + taill) - l : locinput - PL_bostr;
2284 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2286 pref0_len = pref_len - (locinput - PL_reg_starttry);
2287 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2288 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2289 ? (5 + taill) - pref_len : PL_regeol - locinput);
2290 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2294 if (pref0_len > pref_len)
2295 pref0_len = pref_len;
2296 regprop(prop, scan);
2299 do_utf8 && OP(scan) != CANY ?
2300 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2301 pref0_len, 60, UNI_DISPLAY_REGEX) :
2302 locinput - pref_len;
2303 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2304 char *s1 = do_utf8 && OP(scan) != CANY ?
2305 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2306 pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) :
2307 locinput - pref_len + pref0_len;
2308 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2309 char *s2 = do_utf8 && OP(scan) != CANY ?
2310 pv_uni_display(dsv2, (U8*)locinput,
2311 PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
2313 int len2 = do_utf8 ? strlen(s2) : l;
2314 PerlIO_printf(Perl_debug_log,
2315 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2316 (IV)(locinput - PL_bostr),
2323 (docolor ? "" : "> <"),
2327 15 - l - pref_len + 1,
2329 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2334 next = scan + NEXT_OFF(scan);
2340 if (locinput == PL_bostr || (PL_multiline &&
2341 (nextchr || locinput < PL_regeol) && locinput[-1] == '\n') )
2343 /* regtill = regbol; */
2348 if (locinput == PL_bostr ||
2349 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2355 if (locinput == PL_bostr)
2359 if (locinput == PL_reg_ganch)
2369 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2374 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2376 if (PL_regeol - locinput > 1)
2380 if (PL_regeol != locinput)
2384 if (!nextchr && locinput >= PL_regeol)
2387 locinput += PL_utf8skip[nextchr];
2388 if (locinput > PL_regeol)
2390 nextchr = UCHARAT(locinput);
2393 nextchr = UCHARAT(++locinput);
2396 if (!nextchr && locinput >= PL_regeol)
2398 nextchr = UCHARAT(++locinput);
2401 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2404 locinput += PL_utf8skip[nextchr];
2405 if (locinput > PL_regeol)
2407 nextchr = UCHARAT(locinput);
2410 nextchr = UCHARAT(++locinput);
2415 if (do_utf8 != UTF) {
2416 /* The target and the pattern have differing utf8ness. */
2422 /* The target is utf8, the pattern is not utf8. */
2426 if (NATIVE_TO_UNI(*(U8*)s) !=
2427 utf8n_to_uvuni((U8*)l, UTF8_MAXLEN, &ulen,
2429 0 : UTF8_ALLOW_ANY))
2436 /* The target is not utf8, the pattern is utf8. */
2440 if (NATIVE_TO_UNI(*((U8*)l)) !=
2441 utf8n_to_uvuni((U8*)s, UTF8_MAXLEN, &ulen,
2443 0 : UTF8_ALLOW_ANY))
2450 nextchr = UCHARAT(locinput);
2453 /* The target and the pattern have the same utf8ness. */
2454 /* Inline the first character, for speed. */
2455 if (UCHARAT(s) != nextchr)
2457 if (PL_regeol - locinput < ln)
2459 if (ln > 1 && memNE(s, locinput, ln))
2462 nextchr = UCHARAT(locinput);
2465 PL_reg_flags |= RF_tainted;
2471 if (do_utf8 || UTF) {
2472 /* Either target or the pattern are utf8. */
2474 char *e = PL_regeol;
2476 if (ibcmp_utf8(s, 0, ln, (bool)UTF,
2477 l, &e, 0, do_utf8)) {
2478 /* One more case for the sharp s:
2479 * pack("U0U*", 0xDF) =~ /ss/i,
2480 * the 0xC3 0x9F are the UTF-8
2481 * byte sequence for the U+00DF. */
2483 toLOWER(s[0]) == 's' &&
2485 toLOWER(s[1]) == 's' &&
2492 nextchr = UCHARAT(locinput);
2496 /* Neither the target and the pattern are utf8. */
2498 /* Inline the first character, for speed. */
2499 if (UCHARAT(s) != nextchr &&
2500 UCHARAT(s) != ((OP(scan) == EXACTF)
2501 ? PL_fold : PL_fold_locale)[nextchr])
2503 if (PL_regeol - locinput < ln)
2505 if (ln > 1 && (OP(scan) == EXACTF
2506 ? ibcmp(s, locinput, ln)
2507 : ibcmp_locale(s, locinput, ln)))
2510 nextchr = UCHARAT(locinput);
2514 STRLEN inclasslen = PL_regeol - locinput;
2516 if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8))
2518 if (locinput >= PL_regeol)
2520 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
2521 nextchr = UCHARAT(locinput);
2526 nextchr = UCHARAT(locinput);
2527 if (!REGINCLASS(scan, (U8*)locinput))
2529 if (!nextchr && locinput >= PL_regeol)
2531 nextchr = UCHARAT(++locinput);
2535 /* If we might have the case of the German sharp s
2536 * in a casefolding Unicode character class. */
2538 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
2539 locinput += SHARP_S_SKIP;
2540 nextchr = UCHARAT(locinput);
2546 PL_reg_flags |= RF_tainted;
2552 LOAD_UTF8_CHARCLASS(alnum,"a");
2553 if (!(OP(scan) == ALNUM
2554 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2555 : isALNUM_LC_utf8((U8*)locinput)))
2559 locinput += PL_utf8skip[nextchr];
2560 nextchr = UCHARAT(locinput);
2563 if (!(OP(scan) == ALNUM
2564 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2566 nextchr = UCHARAT(++locinput);
2569 PL_reg_flags |= RF_tainted;
2572 if (!nextchr && locinput >= PL_regeol)
2575 LOAD_UTF8_CHARCLASS(alnum,"a");
2576 if (OP(scan) == NALNUM
2577 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2578 : isALNUM_LC_utf8((U8*)locinput))
2582 locinput += PL_utf8skip[nextchr];
2583 nextchr = UCHARAT(locinput);
2586 if (OP(scan) == NALNUM
2587 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2589 nextchr = UCHARAT(++locinput);
2593 PL_reg_flags |= RF_tainted;
2597 /* was last char in word? */
2599 if (locinput == PL_bostr)
2602 U8 *r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
2604 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
2606 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2607 ln = isALNUM_uni(ln);
2608 LOAD_UTF8_CHARCLASS(alnum,"a");
2609 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2612 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2613 n = isALNUM_LC_utf8((U8*)locinput);
2617 ln = (locinput != PL_bostr) ?
2618 UCHARAT(locinput - 1) : '\n';
2619 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2621 n = isALNUM(nextchr);
2624 ln = isALNUM_LC(ln);
2625 n = isALNUM_LC(nextchr);
2628 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2629 OP(scan) == BOUNDL))
2633 PL_reg_flags |= RF_tainted;
2639 if (UTF8_IS_CONTINUED(nextchr)) {
2640 LOAD_UTF8_CHARCLASS(space," ");
2641 if (!(OP(scan) == SPACE
2642 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2643 : isSPACE_LC_utf8((U8*)locinput)))
2647 locinput += PL_utf8skip[nextchr];
2648 nextchr = UCHARAT(locinput);
2651 if (!(OP(scan) == SPACE
2652 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2654 nextchr = UCHARAT(++locinput);
2657 if (!(OP(scan) == SPACE
2658 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2660 nextchr = UCHARAT(++locinput);
2664 PL_reg_flags |= RF_tainted;
2667 if (!nextchr && locinput >= PL_regeol)
2670 LOAD_UTF8_CHARCLASS(space," ");
2671 if (OP(scan) == NSPACE
2672 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2673 : isSPACE_LC_utf8((U8*)locinput))
2677 locinput += PL_utf8skip[nextchr];
2678 nextchr = UCHARAT(locinput);
2681 if (OP(scan) == NSPACE
2682 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
2684 nextchr = UCHARAT(++locinput);
2687 PL_reg_flags |= RF_tainted;
2693 LOAD_UTF8_CHARCLASS(digit,"0");
2694 if (!(OP(scan) == DIGIT
2695 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2696 : isDIGIT_LC_utf8((U8*)locinput)))
2700 locinput += PL_utf8skip[nextchr];
2701 nextchr = UCHARAT(locinput);
2704 if (!(OP(scan) == DIGIT
2705 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
2707 nextchr = UCHARAT(++locinput);
2710 PL_reg_flags |= RF_tainted;
2713 if (!nextchr && locinput >= PL_regeol)
2716 LOAD_UTF8_CHARCLASS(digit,"0");
2717 if (OP(scan) == NDIGIT
2718 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
2719 : isDIGIT_LC_utf8((U8*)locinput))
2723 locinput += PL_utf8skip[nextchr];
2724 nextchr = UCHARAT(locinput);
2727 if (OP(scan) == NDIGIT
2728 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
2730 nextchr = UCHARAT(++locinput);
2733 if (locinput >= PL_regeol)
2736 LOAD_UTF8_CHARCLASS(mark,"~");
2737 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2739 locinput += PL_utf8skip[nextchr];
2740 while (locinput < PL_regeol &&
2741 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
2742 locinput += UTF8SKIP(locinput);
2743 if (locinput > PL_regeol)
2748 nextchr = UCHARAT(locinput);
2751 PL_reg_flags |= RF_tainted;
2755 n = ARG(scan); /* which paren pair */
2756 ln = PL_regstartp[n];
2757 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2758 if ((I32)*PL_reglastparen < n || ln == -1)
2759 sayNO; /* Do not match unless seen CLOSEn. */
2760 if (ln == PL_regendp[n])
2764 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
2766 char *e = PL_bostr + PL_regendp[n];
2768 * Note that we can't do the "other character" lookup trick as
2769 * in the 8-bit case (no pun intended) because in Unicode we
2770 * have to map both upper and title case to lower case.
2772 if (OP(scan) == REFF) {
2773 STRLEN ulen1, ulen2;
2774 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
2775 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
2779 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
2780 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
2781 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
2788 nextchr = UCHARAT(locinput);
2792 /* Inline the first character, for speed. */
2793 if (UCHARAT(s) != nextchr &&
2795 (UCHARAT(s) != ((OP(scan) == REFF
2796 ? PL_fold : PL_fold_locale)[nextchr]))))
2798 ln = PL_regendp[n] - ln;
2799 if (locinput + ln > PL_regeol)
2801 if (ln > 1 && (OP(scan) == REF
2802 ? memNE(s, locinput, ln)
2804 ? ibcmp(s, locinput, ln)
2805 : ibcmp_locale(s, locinput, ln))))
2808 nextchr = UCHARAT(locinput);
2819 OP_4tree *oop = PL_op;
2820 COP *ocurcop = PL_curcop;
2825 PL_op = (OP_4tree*)PL_regdata->data[n];
2826 DEBUG_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
2827 PAD_SAVE_LOCAL(old_comppad, (PAD*)PL_regdata->data[n + 2]);
2828 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
2832 CALLRUNOPS(aTHX); /* Scalar context. */
2835 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
2843 PAD_RESTORE_LOCAL(old_comppad);
2844 PL_curcop = ocurcop;
2846 if (logical == 2) { /* Postponed subexpression. */
2848 MAGIC *mg = Null(MAGIC*);
2850 CHECKPOINT cp, lastcp;
2854 if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret)))
2855 mg = mg_find(sv, PERL_MAGIC_qr);
2856 else if (SvSMAGICAL(ret)) {
2857 if (SvGMAGICAL(ret))
2858 sv_unmagic(ret, PERL_MAGIC_qr);
2860 mg = mg_find(ret, PERL_MAGIC_qr);
2864 re = (regexp *)mg->mg_obj;
2865 (void)ReREFCNT_inc(re);
2869 char *t = SvPV(ret, len);
2871 char *oprecomp = PL_regprecomp;
2872 I32 osize = PL_regsize;
2873 I32 onpar = PL_regnpar;
2876 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
2877 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
2879 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
2881 sv_magic(ret,(SV*)ReREFCNT_inc(re),
2883 PL_regprecomp = oprecomp;
2888 PerlIO_printf(Perl_debug_log,
2889 "Entering embedded `%s%.60s%s%s'\n",
2893 (strlen(re->precomp) > 60 ? "..." : ""))
2896 state.prev = PL_reg_call_cc;
2897 state.cc = PL_regcc;
2898 state.re = PL_reg_re;
2902 cp = regcppush(0); /* Save *all* the positions. */
2905 state.ss = PL_savestack_ix;
2906 *PL_reglastparen = 0;
2907 *PL_reglastcloseparen = 0;
2908 PL_reg_call_cc = &state;
2909 PL_reginput = locinput;
2910 toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
2911 ((re->reganch & ROPT_UTF8) != 0);
2912 if (toggleutf) PL_reg_flags ^= RF_utf8;
2914 /* XXXX This is too dramatic a measure... */
2917 if (regmatch(re->program + 1)) {
2918 /* Even though we succeeded, we need to restore
2919 global variables, since we may be wrapped inside
2920 SUSPEND, thus the match may be not finished yet. */
2922 /* XXXX Do this only if SUSPENDed? */
2923 PL_reg_call_cc = state.prev;
2924 PL_regcc = state.cc;
2925 PL_reg_re = state.re;
2926 cache_re(PL_reg_re);
2927 if (toggleutf) PL_reg_flags ^= RF_utf8;
2929 /* XXXX This is too dramatic a measure... */
2932 /* These are needed even if not SUSPEND. */
2938 REGCP_UNWIND(lastcp);
2940 PL_reg_call_cc = state.prev;
2941 PL_regcc = state.cc;
2942 PL_reg_re = state.re;
2943 cache_re(PL_reg_re);
2944 if (toggleutf) PL_reg_flags ^= RF_utf8;
2946 /* XXXX This is too dramatic a measure... */
2956 sv_setsv(save_scalar(PL_replgv), ret);
2960 n = ARG(scan); /* which paren pair */
2961 PL_reg_start_tmp[n] = locinput;
2966 n = ARG(scan); /* which paren pair */
2967 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
2968 PL_regendp[n] = locinput - PL_bostr;
2969 if (n > (I32)*PL_reglastparen)
2970 *PL_reglastparen = n;
2971 *PL_reglastcloseparen = n;
2974 n = ARG(scan); /* which paren pair */
2975 sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
2978 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
2980 next = NEXTOPER(NEXTOPER(scan));
2982 next = scan + ARG(scan);
2983 if (OP(next) == IFTHEN) /* Fake one. */
2984 next = NEXTOPER(NEXTOPER(next));
2988 logical = scan->flags;
2990 /*******************************************************************
2991 PL_regcc contains infoblock about the innermost (...)* loop, and
2992 a pointer to the next outer infoblock.
2994 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
2996 1) After matching X, regnode for CURLYX is processed;
2998 2) This regnode creates infoblock on the stack, and calls
2999 regmatch() recursively with the starting point at WHILEM node;
3001 3) Each hit of WHILEM node tries to match A and Z (in the order
3002 depending on the current iteration, min/max of {min,max} and
3003 greediness). The information about where are nodes for "A"
3004 and "Z" is read from the infoblock, as is info on how many times "A"
3005 was already matched, and greediness.
3007 4) After A matches, the same WHILEM node is hit again.
3009 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
3010 of the same pair. Thus when WHILEM tries to match Z, it temporarily
3011 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
3012 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
3013 of the external loop.
3015 Currently present infoblocks form a tree with a stem formed by PL_curcc
3016 and whatever it mentions via ->next, and additional attached trees
3017 corresponding to temporarily unset infoblocks as in "5" above.
3019 In the following picture infoblocks for outer loop of
3020 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
3021 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
3022 infoblocks are drawn below the "reset" infoblock.
3024 In fact in the picture below we do not show failed matches for Z and T
3025 by WHILEM blocks. [We illustrate minimal matches, since for them it is
3026 more obvious *why* one needs to *temporary* unset infoblocks.]
3028 Matched REx position InfoBlocks Comment
3032 Y A)*?Z)*?T x <- O <- I
3033 YA )*?Z)*?T x <- O <- I
3034 YA A)*?Z)*?T x <- O <- I
3035 YAA )*?Z)*?T x <- O <- I
3036 YAA Z)*?T x <- O # Temporary unset I
3039 YAAZ Y(A)*?Z)*?T x <- O
3042 YAAZY (A)*?Z)*?T x <- O
3045 YAAZY A)*?Z)*?T x <- O <- I
3048 YAAZYA )*?Z)*?T x <- O <- I
3051 YAAZYA Z)*?T x <- O # Temporary unset I
3057 YAAZYAZ T x # Temporary unset O
3064 *******************************************************************/
3067 CHECKPOINT cp = PL_savestack_ix;
3068 /* No need to save/restore up to this paren */
3069 I32 parenfloor = scan->flags;
3071 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3073 cc.oldcc = PL_regcc;
3075 /* XXXX Probably it is better to teach regpush to support
3076 parenfloor > PL_regsize... */
3077 if (parenfloor > (I32)*PL_reglastparen)
3078 parenfloor = *PL_reglastparen; /* Pessimization... */
3079 cc.parenfloor = parenfloor;
3081 cc.min = ARG1(scan);
3082 cc.max = ARG2(scan);
3083 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3087 PL_reginput = locinput;
3088 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
3090 PL_regcc = cc.oldcc;
3096 * This is really hard to understand, because after we match
3097 * what we're trying to match, we must make sure the rest of
3098 * the REx is going to match for sure, and to do that we have
3099 * to go back UP the parse tree by recursing ever deeper. And
3100 * if it fails, we have to reset our parent's current state
3101 * that we can try again after backing off.
3104 CHECKPOINT cp, lastcp;
3105 CURCUR* cc = PL_regcc;
3106 char *lastloc = cc->lastloc; /* Detection of 0-len. */
3108 n = cc->cur + 1; /* how many we know we matched */
3109 PL_reginput = locinput;
3112 PerlIO_printf(Perl_debug_log,
3113 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3114 REPORT_CODE_OFF+PL_regindent*2, "",
3115 (long)n, (long)cc->min,
3116 (long)cc->max, PTR2UV(cc))
3119 /* If degenerate scan matches "", assume scan done. */
3121 if (locinput == cc->lastloc && n >= cc->min) {
3122 PL_regcc = cc->oldcc;
3126 PerlIO_printf(Perl_debug_log,
3127 "%*s empty match detected, try continuation...\n",
3128 REPORT_CODE_OFF+PL_regindent*2, "")
3130 if (regmatch(cc->next))
3138 /* First just match a string of min scans. */
3142 cc->lastloc = locinput;
3143 if (regmatch(cc->scan))
3146 cc->lastloc = lastloc;
3151 /* Check whether we already were at this position.
3152 Postpone detection until we know the match is not
3153 *that* much linear. */
3154 if (!PL_reg_maxiter) {
3155 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3156 PL_reg_leftiter = PL_reg_maxiter;
3158 if (PL_reg_leftiter-- == 0) {
3159 I32 size = (PL_reg_maxiter + 7)/8;
3160 if (PL_reg_poscache) {
3161 if ((I32)PL_reg_poscache_size < size) {
3162 Renew(PL_reg_poscache, size, char);
3163 PL_reg_poscache_size = size;
3165 Zero(PL_reg_poscache, size, char);
3168 PL_reg_poscache_size = size;
3169 Newz(29, PL_reg_poscache, size, char);
3172 PerlIO_printf(Perl_debug_log,
3173 "%sDetected a super-linear match, switching on caching%s...\n",
3174 PL_colors[4], PL_colors[5])
3177 if (PL_reg_leftiter < 0) {
3178 I32 o = locinput - PL_bostr, b;
3180 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3183 if (PL_reg_poscache[o] & (1<<b)) {
3185 PerlIO_printf(Perl_debug_log,
3186 "%*s already tried at this position...\n",
3187 REPORT_CODE_OFF+PL_regindent*2, "")
3191 PL_reg_poscache[o] |= (1<<b);
3195 /* Prefer next over scan for minimal matching. */
3198 PL_regcc = cc->oldcc;
3201 cp = regcppush(cc->parenfloor);
3203 if (regmatch(cc->next)) {
3205 sayYES; /* All done. */
3207 REGCP_UNWIND(lastcp);
3213 if (n >= cc->max) { /* Maximum greed exceeded? */
3214 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3215 && !(PL_reg_flags & RF_warned)) {
3216 PL_reg_flags |= RF_warned;
3217 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3218 "Complex regular subexpression recursion",
3225 PerlIO_printf(Perl_debug_log,
3226 "%*s trying longer...\n",
3227 REPORT_CODE_OFF+PL_regindent*2, "")
3229 /* Try scanning more and see if it helps. */
3230 PL_reginput = locinput;
3232 cc->lastloc = locinput;
3233 cp = regcppush(cc->parenfloor);
3235 if (regmatch(cc->scan)) {
3239 REGCP_UNWIND(lastcp);
3242 cc->lastloc = lastloc;
3246 /* Prefer scan over next for maximal matching. */
3248 if (n < cc->max) { /* More greed allowed? */
3249 cp = regcppush(cc->parenfloor);
3251 cc->lastloc = locinput;
3253 if (regmatch(cc->scan)) {
3257 REGCP_UNWIND(lastcp);
3258 regcppop(); /* Restore some previous $<digit>s? */
3259 PL_reginput = locinput;
3261 PerlIO_printf(Perl_debug_log,
3262 "%*s failed, try continuation...\n",
3263 REPORT_CODE_OFF+PL_regindent*2, "")
3266 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3267 && !(PL_reg_flags & RF_warned)) {
3268 PL_reg_flags |= RF_warned;
3269 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3270 "Complex regular subexpression recursion",
3274 /* Failed deeper matches of scan, so see if this one works. */
3275 PL_regcc = cc->oldcc;
3278 if (regmatch(cc->next))
3284 cc->lastloc = lastloc;
3289 next = scan + ARG(scan);
3292 inner = NEXTOPER(NEXTOPER(scan));
3295 inner = NEXTOPER(scan);
3299 if (OP(next) != c1) /* No choice. */
3300 next = inner; /* Avoid recursion. */
3302 I32 lastparen = *PL_reglastparen;
3304 re_unwind_branch_t *uw;
3306 /* Put unwinding data on stack */
3307 unwind1 = SSNEWt(1,re_unwind_branch_t);
3308 uw = SSPTRt(unwind1,re_unwind_branch_t);
3311 uw->type = ((c1 == BRANCH)
3313 : RE_UNWIND_BRANCHJ);
3314 uw->lastparen = lastparen;
3316 uw->locinput = locinput;
3317 uw->nextchr = nextchr;
3319 uw->regindent = ++PL_regindent;
3322 REGCP_SET(uw->lastcp);
3324 /* Now go into the first branch */
3337 /* We suppose that the next guy does not need
3338 backtracking: in particular, it is of constant length,
3339 and has no parenths to influence future backrefs. */
3340 ln = ARG1(scan); /* min to match */
3341 n = ARG2(scan); /* max to match */
3342 paren = scan->flags;
3344 if (paren > PL_regsize)
3346 if (paren > (I32)*PL_reglastparen)
3347 *PL_reglastparen = paren;
3349 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3351 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3352 PL_reginput = locinput;
3355 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3357 /* if we matched something zero-length we don't need to
3358 backtrack - capturing parens are already defined, so
3359 the caveat in the maximal case doesn't apply
3361 XXXX if ln == 0, we can redo this check first time
3362 through the following loop
3365 n = ln; /* don't backtrack */
3366 locinput = PL_reginput;
3367 if (HAS_TEXT(next) || JUMPABLE(next)) {
3368 regnode *text_node = next;
3370 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3372 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3374 if (PL_regkind[(U8)OP(text_node)] == REF) {
3376 n = ARG(text_node); /* which paren pair */
3377 ln = PL_regstartp[n];
3378 /* assume yes if we haven't seen CLOSEn */
3380 (I32)*PL_reglastparen < n ||
3387 c1 = *(PL_bostr + ln);
3389 else { c1 = (U8)*STRING(text_node); }
3390 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3392 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3393 c2 = PL_fold_locale[c1];
3402 /* This may be improved if l == 0. */
3403 while (n >= ln || (n == REG_INFTY && ln > 0 && l)) { /* ln overflow ? */
3404 /* If it could work, try it. */
3406 UCHARAT(PL_reginput) == c1 ||
3407 UCHARAT(PL_reginput) == c2)
3411 PL_regstartp[paren] =
3412 HOPc(PL_reginput, -l) - PL_bostr;
3413 PL_regendp[paren] = PL_reginput - PL_bostr;
3416 PL_regendp[paren] = -1;
3420 REGCP_UNWIND(lastcp);
3422 /* Couldn't or didn't -- move forward. */
3423 PL_reginput = locinput;
3424 if (regrepeat_hard(scan, 1, &l)) {
3426 locinput = PL_reginput;
3433 n = regrepeat_hard(scan, n, &l);
3434 /* if we matched something zero-length we don't need to
3435 backtrack, unless the minimum count is zero and we
3436 are capturing the result - in that case the capture
3437 being defined or not may affect later execution
3439 if (n != 0 && l == 0 && !(paren && ln == 0))
3440 ln = n; /* don't backtrack */
3441 locinput = PL_reginput;
3443 PerlIO_printf(Perl_debug_log,
3444 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3445 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3449 if (HAS_TEXT(next) || JUMPABLE(next)) {
3450 regnode *text_node = next;
3452 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3454 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3456 if (PL_regkind[(U8)OP(text_node)] == REF) {
3458 n = ARG(text_node); /* which paren pair */
3459 ln = PL_regstartp[n];
3460 /* assume yes if we haven't seen CLOSEn */
3462 (I32)*PL_reglastparen < n ||
3469 c1 = *(PL_bostr + ln);
3471 else { c1 = (U8)*STRING(text_node); }
3473 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3475 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3476 c2 = PL_fold_locale[c1];
3487 /* If it could work, try it. */
3489 UCHARAT(PL_reginput) == c1 ||
3490 UCHARAT(PL_reginput) == c2)
3493 PerlIO_printf(Perl_debug_log,
3494 "%*s trying tail with n=%"IVdf"...\n",
3495 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3499 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3500 PL_regendp[paren] = PL_reginput - PL_bostr;
3503 PL_regendp[paren] = -1;
3507 REGCP_UNWIND(lastcp);
3509 /* Couldn't or didn't -- back up. */
3511 locinput = HOPc(locinput, -l);
3512 PL_reginput = locinput;
3519 paren = scan->flags; /* Which paren to set */
3520 if (paren > PL_regsize)
3522 if (paren > (I32)*PL_reglastparen)
3523 *PL_reglastparen = paren;
3524 ln = ARG1(scan); /* min to match */
3525 n = ARG2(scan); /* max to match */
3526 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3530 ln = ARG1(scan); /* min to match */
3531 n = ARG2(scan); /* max to match */
3532 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3537 scan = NEXTOPER(scan);
3543 scan = NEXTOPER(scan);
3547 * Lookahead to avoid useless match attempts
3548 * when we know what character comes next.
3552 * Used to only do .*x and .*?x, but now it allows
3553 * for )'s, ('s and (?{ ... })'s to be in the way
3554 * of the quantifier and the EXACT-like node. -- japhy
3557 if (HAS_TEXT(next) || JUMPABLE(next)) {
3559 regnode *text_node = next;
3561 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3563 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3565 if (PL_regkind[(U8)OP(text_node)] == REF) {
3567 n = ARG(text_node); /* which paren pair */
3568 ln = PL_regstartp[n];
3569 /* assume yes if we haven't seen CLOSEn */
3571 (I32)*PL_reglastparen < n ||
3576 goto assume_ok_easy;
3578 s = (U8*)PL_bostr + ln;
3580 else { s = (U8*)STRING(text_node); }
3584 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3586 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3587 c2 = PL_fold_locale[c1];
3590 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3591 STRLEN ulen1, ulen2;
3592 U8 tmpbuf1[UTF8_MAXLEN_UCLC+1];
3593 U8 tmpbuf2[UTF8_MAXLEN_UCLC+1];
3595 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3596 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3598 c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXLEN, 0,
3600 0 : UTF8_ALLOW_ANY);
3601 c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXLEN, 0,
3603 0 : UTF8_ALLOW_ANY);
3606 c2 = c1 = utf8n_to_uvchr(s, UTF8_MAXLEN, 0,
3608 0 : UTF8_ALLOW_ANY);
3616 PL_reginput = locinput;
3620 if (ln && regrepeat(scan, ln) < ln)
3622 locinput = PL_reginput;
3625 char *e; /* Should not check after this */
3626 char *old = locinput;
3629 if (n == REG_INFTY) {
3632 while (UTF8_IS_CONTINUATION(*(U8*)e))
3638 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3642 e = locinput + n - ln;
3647 /* Find place 'next' could work */
3650 while (locinput <= e &&
3651 UCHARAT(locinput) != c1)
3654 while (locinput <= e
3655 && UCHARAT(locinput) != c1
3656 && UCHARAT(locinput) != c2)
3659 count = locinput - old;
3664 /* count initialised to
3665 * utf8_distance(old, locinput) */
3666 while (locinput <= e &&
3667 utf8n_to_uvchr((U8*)locinput,
3670 0 : UTF8_ALLOW_ANY) != (UV)c1) {
3675 /* count initialised to
3676 * utf8_distance(old, locinput) */
3677 while (locinput <= e) {
3678 UV c = utf8n_to_uvchr((U8*)locinput,
3681 0 : UTF8_ALLOW_ANY);
3682 if (c == (UV)c1 || c == (UV)c2)
3691 /* PL_reginput == old now */
3692 if (locinput != old) {
3693 ln = 1; /* Did some */
3694 if (regrepeat(scan, count) < count)
3697 /* PL_reginput == locinput now */
3698 TRYPAREN(paren, ln, locinput);
3699 PL_reginput = locinput; /* Could be reset... */
3700 REGCP_UNWIND(lastcp);
3701 /* Couldn't or didn't -- move forward. */
3704 locinput += UTF8SKIP(locinput);
3711 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3715 c = utf8n_to_uvchr((U8*)PL_reginput,
3718 0 : UTF8_ALLOW_ANY);
3720 c = UCHARAT(PL_reginput);
3721 /* If it could work, try it. */
3722 if (c == (UV)c1 || c == (UV)c2)
3724 TRYPAREN(paren, ln, PL_reginput);
3725 REGCP_UNWIND(lastcp);
3728 /* If it could work, try it. */
3729 else if (c1 == -1000)
3731 TRYPAREN(paren, ln, PL_reginput);
3732 REGCP_UNWIND(lastcp);
3734 /* Couldn't or didn't -- move forward. */
3735 PL_reginput = locinput;
3736 if (regrepeat(scan, 1)) {
3738 locinput = PL_reginput;
3746 n = regrepeat(scan, n);
3747 locinput = PL_reginput;
3748 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
3749 ((!PL_multiline && OP(next) != MEOL) ||
3750 OP(next) == SEOL || OP(next) == EOS))
3752 ln = n; /* why back off? */
3753 /* ...because $ and \Z can match before *and* after
3754 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
3755 We should back off by one in this case. */
3756 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
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);
3788 c = utf8n_to_uvchr((U8*)PL_reginput,
3791 0 : UTF8_ALLOW_ANY);
3793 c = UCHARAT(PL_reginput);
3795 /* If it could work, try it. */
3796 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
3798 TRYPAREN(paren, n, PL_reginput);
3799 REGCP_UNWIND(lastcp);
3801 /* Couldn't or didn't -- back up. */
3803 PL_reginput = locinput = HOPc(locinput, -1);
3810 if (PL_reg_call_cc) {
3811 re_cc_state *cur_call_cc = PL_reg_call_cc;
3812 CURCUR *cctmp = PL_regcc;
3813 regexp *re = PL_reg_re;
3814 CHECKPOINT cp, lastcp;
3816 cp = regcppush(0); /* Save *all* the positions. */
3818 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
3820 PL_reginput = locinput; /* Make position available to
3822 cache_re(PL_reg_call_cc->re);
3823 PL_regcc = PL_reg_call_cc->cc;
3824 PL_reg_call_cc = PL_reg_call_cc->prev;
3825 if (regmatch(cur_call_cc->node)) {
3826 PL_reg_call_cc = cur_call_cc;
3830 REGCP_UNWIND(lastcp);
3832 PL_reg_call_cc = cur_call_cc;
3838 PerlIO_printf(Perl_debug_log,
3839 "%*s continuation failed...\n",
3840 REPORT_CODE_OFF+PL_regindent*2, "")
3844 if (locinput < PL_regtill) {
3845 DEBUG_r(PerlIO_printf(Perl_debug_log,
3846 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
3848 (long)(locinput - PL_reg_starttry),
3849 (long)(PL_regtill - PL_reg_starttry),
3851 sayNO_FINAL; /* Cannot match: too short. */
3853 PL_reginput = locinput; /* put where regtry can find it */
3854 sayYES_FINAL; /* Success! */
3856 PL_reginput = locinput; /* put where regtry can find it */
3857 sayYES_LOUD; /* Success! */
3860 PL_reginput = locinput;
3865 s = HOPBACKc(locinput, scan->flags);
3871 PL_reginput = locinput;
3876 s = HOPBACKc(locinput, scan->flags);
3882 PL_reginput = locinput;
3885 inner = NEXTOPER(NEXTOPER(scan));
3886 if (regmatch(inner) != n) {
3901 if (OP(scan) == SUSPEND) {
3902 locinput = PL_reginput;
3903 nextchr = UCHARAT(locinput);
3908 next = scan + ARG(scan);
3913 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
3914 PTR2UV(scan), OP(scan));
3915 Perl_croak(aTHX_ "regexp memory corruption");
3922 * We get here only if there's trouble -- normally "case END" is
3923 * the terminating point.
3925 Perl_croak(aTHX_ "corrupted regexp pointers");
3931 PerlIO_printf(Perl_debug_log,
3932 "%*s %scould match...%s\n",
3933 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
3937 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
3938 PL_colors[4],PL_colors[5]));
3944 #if 0 /* Breaks $^R */
3952 PerlIO_printf(Perl_debug_log,
3953 "%*s %sfailed...%s\n",
3954 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
3960 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
3963 case RE_UNWIND_BRANCH:
3964 case RE_UNWIND_BRANCHJ:
3966 re_unwind_branch_t *uwb = &(uw->branch);
3967 I32 lastparen = uwb->lastparen;
3969 REGCP_UNWIND(uwb->lastcp);
3970 for (n = *PL_reglastparen; n > lastparen; n--)
3972 *PL_reglastparen = n;
3973 scan = next = uwb->next;
3975 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
3976 ? BRANCH : BRANCHJ) ) { /* Failure */
3983 /* Have more choice yet. Reuse the same uwb. */
3985 if ((n = (uwb->type == RE_UNWIND_BRANCH
3986 ? NEXT_OFF(next) : ARG(next))))
3989 next = NULL; /* XXXX Needn't unwinding in this case... */
3991 next = NEXTOPER(scan);
3992 if (uwb->type == RE_UNWIND_BRANCHJ)
3993 next = NEXTOPER(next);
3994 locinput = uwb->locinput;
3995 nextchr = uwb->nextchr;
3997 PL_regindent = uwb->regindent;
4004 Perl_croak(aTHX_ "regexp unwind memory corruption");
4015 - regrepeat - repeatedly match something simple, report how many
4018 * [This routine now assumes that it will only match on things of length 1.
4019 * That was true before, but now we assume scan - reginput is the count,
4020 * rather than incrementing count on every character. [Er, except utf8.]]
4023 S_regrepeat(pTHX_ regnode *p, I32 max)
4025 register char *scan;
4027 register char *loceol = PL_regeol;
4028 register I32 hardcount = 0;
4029 register bool do_utf8 = PL_reg_match_utf8;
4032 if (max == REG_INFTY)
4034 else if (max < loceol - scan)
4035 loceol = scan + max;
4040 while (scan < loceol && hardcount < max && *scan != '\n') {
4041 scan += UTF8SKIP(scan);
4045 while (scan < loceol && *scan != '\n')
4052 while (scan < loceol && hardcount < max) {
4053 scan += UTF8SKIP(scan);
4063 case EXACT: /* length of string is 1 */
4065 while (scan < loceol && UCHARAT(scan) == c)
4068 case EXACTF: /* length of string is 1 */
4070 while (scan < loceol &&
4071 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
4074 case EXACTFL: /* length of string is 1 */
4075 PL_reg_flags |= RF_tainted;
4077 while (scan < loceol &&
4078 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
4084 while (hardcount < max && scan < loceol &&
4085 reginclass(p, (U8*)scan, 0, do_utf8)) {
4086 scan += UTF8SKIP(scan);
4090 while (scan < loceol && REGINCLASS(p, (U8*)scan))
4097 LOAD_UTF8_CHARCLASS(alnum,"a");
4098 while (hardcount < max && scan < loceol &&
4099 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4100 scan += UTF8SKIP(scan);
4104 while (scan < loceol && isALNUM(*scan))
4109 PL_reg_flags |= RF_tainted;
4112 while (hardcount < max && scan < loceol &&
4113 isALNUM_LC_utf8((U8*)scan)) {
4114 scan += UTF8SKIP(scan);
4118 while (scan < loceol && isALNUM_LC(*scan))
4125 LOAD_UTF8_CHARCLASS(alnum,"a");
4126 while (hardcount < max && scan < loceol &&
4127 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4128 scan += UTF8SKIP(scan);
4132 while (scan < loceol && !isALNUM(*scan))
4137 PL_reg_flags |= RF_tainted;
4140 while (hardcount < max && scan < loceol &&
4141 !isALNUM_LC_utf8((U8*)scan)) {
4142 scan += UTF8SKIP(scan);
4146 while (scan < loceol && !isALNUM_LC(*scan))
4153 LOAD_UTF8_CHARCLASS(space," ");
4154 while (hardcount < max && scan < loceol &&
4156 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4157 scan += UTF8SKIP(scan);
4161 while (scan < loceol && isSPACE(*scan))
4166 PL_reg_flags |= RF_tainted;
4169 while (hardcount < max && scan < loceol &&
4170 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4171 scan += UTF8SKIP(scan);
4175 while (scan < loceol && isSPACE_LC(*scan))
4182 LOAD_UTF8_CHARCLASS(space," ");
4183 while (hardcount < max && scan < loceol &&
4185 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4186 scan += UTF8SKIP(scan);
4190 while (scan < loceol && !isSPACE(*scan))
4195 PL_reg_flags |= RF_tainted;
4198 while (hardcount < max && scan < loceol &&
4199 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4200 scan += UTF8SKIP(scan);
4204 while (scan < loceol && !isSPACE_LC(*scan))
4211 LOAD_UTF8_CHARCLASS(digit,"0");
4212 while (hardcount < max && scan < loceol &&
4213 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4214 scan += UTF8SKIP(scan);
4218 while (scan < loceol && isDIGIT(*scan))
4225 LOAD_UTF8_CHARCLASS(digit,"0");
4226 while (hardcount < max && scan < loceol &&
4227 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4228 scan += UTF8SKIP(scan);
4232 while (scan < loceol && !isDIGIT(*scan))
4236 default: /* Called on something of 0 width. */
4237 break; /* So match right here or not at all. */
4243 c = scan - PL_reginput;
4248 SV *prop = sv_newmortal();
4251 PerlIO_printf(Perl_debug_log,
4252 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4253 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4260 - regrepeat_hard - repeatedly match something, report total lenth and length
4262 * The repeater is supposed to have constant length.
4266 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4268 register char *scan = Nullch;
4269 register char *start;
4270 register char *loceol = PL_regeol;
4272 I32 count = 0, res = 1;
4277 start = PL_reginput;
4278 if (PL_reg_match_utf8) {
4279 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4282 while (start < PL_reginput) {
4284 start += UTF8SKIP(start);
4295 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4297 *lp = l = PL_reginput - start;
4298 if (max != REG_INFTY && l*max < loceol - scan)
4299 loceol = scan + l*max;
4312 - regclass_swash - prepare the utf8 swash
4316 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** listsvp, SV **altsvp)
4322 if (PL_regdata && PL_regdata->count) {
4325 if (PL_regdata->what[n] == 's') {
4326 SV *rv = (SV*)PL_regdata->data[n];
4327 AV *av = (AV*)SvRV((SV*)rv);
4328 SV **ary = AvARRAY(av);
4331 /* See the end of regcomp.c:S_reglass() for
4332 * documentation of these array elements. */
4335 a = SvTYPE(ary[1]) == SVt_RV ? &ary[1] : 0;
4336 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
4340 else if (si && doinit) {
4341 sw = swash_init("utf8", "", si, 1, 0);
4342 (void)av_store(av, 1, sw);
4358 - reginclass - determine if a character falls into a character class
4360 The n is the ANYOF regnode, the p is the target string, lenp
4361 is pointer to the maximum length of how far to go in the p
4362 (if the lenp is zero, UTF8SKIP(p) is used),
4363 do_utf8 tells whether the target string is in UTF-8.
4368 S_reginclass(pTHX_ register regnode *n, register U8* p, STRLEN* lenp, register bool do_utf8)
4370 char flags = ANYOF_FLAGS(n);
4376 if (do_utf8 && !UTF8_IS_INVARIANT(c))
4377 c = utf8n_to_uvchr(p, UTF8_MAXLEN, &len,
4378 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
4380 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
4381 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4384 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4385 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4388 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4392 SV *sw = regclass_swash(n, TRUE, 0, (SV**)&av);
4395 if (swash_fetch(sw, p, do_utf8))
4397 else if (flags & ANYOF_FOLD) {
4398 if (!match && lenp && av) {
4401 for (i = 0; i <= av_len(av); i++) {
4402 SV* sv = *av_fetch(av, i, FALSE);
4404 char *s = SvPV(sv, len);
4406 if (len <= plen && memEQ(s, (char*)p, len)) {
4414 U8 tmpbuf[UTF8_MAXLEN_FOLD+1];
4417 to_utf8_fold(p, tmpbuf, &tmplen);
4418 if (swash_fetch(sw, tmpbuf, do_utf8))
4424 if (match && lenp && *lenp == 0)
4425 *lenp = UNISKIP(NATIVE_TO_UNI(c));
4427 if (!match && c < 256) {
4428 if (ANYOF_BITMAP_TEST(n, c))
4430 else if (flags & ANYOF_FOLD) {
4433 if (flags & ANYOF_LOCALE) {
4434 PL_reg_flags |= RF_tainted;
4435 f = PL_fold_locale[c];
4439 if (f != c && ANYOF_BITMAP_TEST(n, f))
4443 if (!match && (flags & ANYOF_CLASS)) {
4444 PL_reg_flags |= RF_tainted;
4446 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4447 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4448 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4449 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4450 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4451 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4452 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4453 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4454 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4455 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4456 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4457 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4458 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4459 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4460 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4461 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4462 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4463 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4464 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4465 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4466 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4467 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4468 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4469 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4470 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4471 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4472 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4473 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4474 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4475 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4476 ) /* How's that for a conditional? */
4483 return (flags & ANYOF_INVERT) ? !match : match;
4487 S_reghop(pTHX_ U8 *s, I32 off)
4489 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4493 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4496 while (off-- && s < lim) {
4497 /* XXX could check well-formedness here */
4505 if (UTF8_IS_CONTINUED(*s)) {
4506 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4509 /* XXX could check well-formedness here */
4517 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4519 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4523 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4526 while (off-- && s < lim) {
4527 /* XXX could check well-formedness here */
4537 if (UTF8_IS_CONTINUED(*s)) {
4538 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4541 /* XXX could check well-formedness here */
4553 restore_pos(pTHX_ void *arg)
4555 if (PL_reg_eval_set) {
4556 if (PL_reg_oldsaved) {
4557 PL_reg_re->subbeg = PL_reg_oldsaved;
4558 PL_reg_re->sublen = PL_reg_oldsavedlen;
4559 RX_MATCH_COPIED_on(PL_reg_re);
4561 PL_reg_magic->mg_len = PL_reg_oldpos;
4562 PL_reg_eval_set = 0;
4563 PL_curpm = PL_reg_oldcurpm;
4568 S_to_utf8_substr(pTHX_ register regexp *prog)
4571 if (prog->float_substr && !prog->float_utf8) {
4572 prog->float_utf8 = sv = NEWSV(117, 0);
4573 SvSetSV(sv, prog->float_substr);
4574 sv_utf8_upgrade(sv);
4575 if (SvTAIL(prog->float_substr))
4577 if (prog->float_substr == prog->check_substr)
4578 prog->check_utf8 = sv;
4580 if (prog->anchored_substr && !prog->anchored_utf8) {
4581 prog->anchored_utf8 = sv = NEWSV(118, 0);
4582 SvSetSV(sv, prog->anchored_substr);
4583 sv_utf8_upgrade(sv);
4584 if (SvTAIL(prog->anchored_substr))
4586 if (prog->anchored_substr == prog->check_substr)
4587 prog->check_utf8 = sv;
4592 S_to_byte_substr(pTHX_ register regexp *prog)
4595 if (prog->float_utf8 && !prog->float_substr) {
4596 prog->float_substr = sv = NEWSV(117, 0);
4597 SvSetSV(sv, prog->float_utf8);
4598 if (sv_utf8_downgrade(sv, TRUE)) {
4599 if (SvTAIL(prog->float_utf8))
4603 prog->float_substr = sv = &PL_sv_undef;
4605 if (prog->float_utf8 == prog->check_utf8)
4606 prog->check_substr = sv;
4608 if (prog->anchored_utf8 && !prog->anchored_substr) {
4609 prog->anchored_substr = sv = NEWSV(118, 0);
4610 SvSetSV(sv, prog->anchored_utf8);
4611 if (sv_utf8_downgrade(sv, TRUE)) {
4612 if (SvTAIL(prog->anchored_utf8))
4616 prog->anchored_substr = sv = &PL_sv_undef;
4618 if (prog->anchored_utf8 == prog->check_utf8)
4619 prog->check_substr = sv;