5 * "One Ring to rule them all, One Ring to find them..."
8 /* This file contains functions for executing a regular expression. See
9 * also regcomp.c which funnily enough, contains functions for compiling
10 * a regular expression.
12 * This file is also copied at build time to ext/re/re_exec.c, where
13 * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT.
14 * This causes the main functions to be compiled under new names and with
15 * debugging support added, which makes "use re 'debug'" work.
19 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
20 * confused with the original package (see point 3 below). Thanks, Henry!
23 /* Additional note: this code is very heavily munged from Henry's version
24 * in places. In some spots I've traded clarity for efficiency, so don't
25 * blame Henry for some of the lack of readability.
28 /* The names of the functions have been changed from regcomp and
29 * regexec to pregcomp and pregexec in order to avoid conflicts
30 * with the POSIX routines of the same names.
33 #ifdef PERL_EXT_RE_BUILD
34 /* need to replace pregcomp et al, so enable that */
35 # ifndef PERL_IN_XSUB_RE
36 # define PERL_IN_XSUB_RE
38 /* need access to debugger hooks */
39 # if defined(PERL_EXT_RE_DEBUG) && !defined(DEBUGGING)
44 #ifdef PERL_IN_XSUB_RE
45 /* We *really* need to overwrite these symbols: */
46 # define Perl_regexec_flags my_regexec
47 # define Perl_regdump my_regdump
48 # define Perl_regprop my_regprop
49 # define Perl_re_intuit_start my_re_intuit_start
50 /* *These* symbols are masked to allow static link. */
51 # define Perl_pregexec my_pregexec
52 # define Perl_reginitcolors my_reginitcolors
53 # define Perl_regclass_swash my_regclass_swash
55 # define PERL_NO_GET_CONTEXT
59 * pregcomp and pregexec -- regsub and regerror are not used in perl
61 * Copyright (c) 1986 by University of Toronto.
62 * Written by Henry Spencer. Not derived from licensed software.
64 * Permission is granted to anyone to use this software for any
65 * purpose on any computer system, and to redistribute it freely,
66 * subject to the following restrictions:
68 * 1. The author is not responsible for the consequences of use of
69 * this software, no matter how awful, even if they arise
72 * 2. The origin of this software must not be misrepresented, either
73 * by explicit claim or by omission.
75 * 3. Altered versions must be plainly marked as such, and must not
76 * be misrepresented as being the original software.
78 **** Alterations to Henry's code are...
80 **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
81 **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, by Larry Wall and others
83 **** You may distribute under the terms of either the GNU General Public
84 **** License or the Artistic License, as specified in the README file.
86 * Beware that some of this code is subtly aware of the way operator
87 * precedence is structured in regular expressions. Serious changes in
88 * regular-expression syntax might require a total rethink.
91 #define PERL_IN_REGEXEC_C
96 #define RF_tainted 1 /* tainted information used? */
97 #define RF_warned 2 /* warned about big count? */
98 #define RF_evaled 4 /* Did an EVAL with setting? */
99 #define RF_utf8 8 /* String contains multibyte chars? */
101 #define UTF ((PL_reg_flags & RF_utf8) != 0)
103 #define RS_init 1 /* eval environment created */
104 #define RS_set 2 /* replsv value is set */
107 #define STATIC static
110 #define REGINCLASS(p,c) (ANYOF_FLAGS(p) ? reginclass(p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c)))
116 #define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv))
117 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
119 #define HOPc(pos,off) ((char *)(PL_reg_match_utf8 \
120 ? reghop3((U8*)pos, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr)) \
122 #define HOPBACKc(pos, off) ((char*) \
123 ((PL_reg_match_utf8) \
124 ? reghopmaybe3((U8*)pos, -off, ((U8*)(off < 0 ? PL_regeol : PL_bostr))) \
125 : (pos - off >= PL_bostr) \
130 #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
131 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
132 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
134 #define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \
135 if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)str); assert(ok); LEAVE; } } STMT_END
136 #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a")
137 #define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0")
138 #define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ")
139 #define LOAD_UTF8_CHARCLASS_MARK() LOAD_UTF8_CHARCLASS(mark, "\xcd\x86")
141 /* for use after a quantifier and before an EXACT-like node -- japhy */
142 #define JUMPABLE(rn) ( \
143 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
144 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
145 OP(rn) == PLUS || OP(rn) == MINMOD || \
146 (PL_regkind[(U8)OP(rn)] == CURLY && ARG1(rn) > 0) \
149 #define HAS_TEXT(rn) ( \
150 PL_regkind[(U8)OP(rn)] == EXACT || PL_regkind[(U8)OP(rn)] == REF \
154 Search for mandatory following text node; for lookahead, the text must
155 follow but for lookbehind (rn->flags != 0) we skip to the next step.
157 #define FIND_NEXT_IMPT(rn) STMT_START { \
158 while (JUMPABLE(rn)) \
159 if (OP(rn) == SUSPEND || PL_regkind[(U8)OP(rn)] == CURLY) \
160 rn = NEXTOPER(NEXTOPER(rn)); \
161 else if (OP(rn) == PLUS) \
163 else if (OP(rn) == IFMATCH) \
164 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
165 else rn += NEXT_OFF(rn); \
168 static void restore_pos(pTHX_ void *arg);
171 S_regcppush(pTHX_ I32 parenfloor)
174 const int retval = PL_savestack_ix;
175 #define REGCP_PAREN_ELEMS 4
176 const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
179 if (paren_elems_to_push < 0)
180 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
182 #define REGCP_OTHER_ELEMS 6
183 SSGROW(paren_elems_to_push + REGCP_OTHER_ELEMS);
184 for (p = PL_regsize; p > parenfloor; p--) {
185 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
186 SSPUSHINT(PL_regendp[p]);
187 SSPUSHINT(PL_regstartp[p]);
188 SSPUSHPTR(PL_reg_start_tmp[p]);
191 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
192 SSPUSHINT(PL_regsize);
193 SSPUSHINT(*PL_reglastparen);
194 SSPUSHINT(*PL_reglastcloseparen);
195 SSPUSHPTR(PL_reginput);
196 #define REGCP_FRAME_ELEMS 2
197 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
198 * are needed for the regexp context stack bookkeeping. */
199 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
200 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
205 /* These are needed since we do not localize EVAL nodes: */
206 # define REGCP_SET(cp) DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, \
207 " Setting an EVAL scope, savestack=%"IVdf"\n", \
208 (IV)PL_savestack_ix)); cp = PL_savestack_ix
210 # define REGCP_UNWIND(cp) DEBUG_EXECUTE_r(cp != PL_savestack_ix ? \
211 PerlIO_printf(Perl_debug_log, \
212 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
213 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
216 S_regcppop(pTHX_ const regexp *rex)
222 GET_RE_DEBUG_FLAGS_DECL;
224 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
226 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
227 i = SSPOPINT; /* Parentheses elements to pop. */
228 input = (char *) SSPOPPTR;
229 *PL_reglastcloseparen = SSPOPINT;
230 *PL_reglastparen = SSPOPINT;
231 PL_regsize = SSPOPINT;
233 /* Now restore the parentheses context. */
234 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
235 i > 0; i -= REGCP_PAREN_ELEMS) {
237 U32 paren = (U32)SSPOPINT;
238 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
239 PL_regstartp[paren] = SSPOPINT;
241 if (paren <= *PL_reglastparen)
242 PL_regendp[paren] = tmps;
244 PerlIO_printf(Perl_debug_log,
245 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
246 (UV)paren, (IV)PL_regstartp[paren],
247 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
248 (IV)PL_regendp[paren],
249 (paren > *PL_reglastparen ? "(no)" : ""));
253 if ((I32)(*PL_reglastparen + 1) <= rex->nparens) {
254 PerlIO_printf(Perl_debug_log,
255 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
256 (IV)(*PL_reglastparen + 1), (IV)rex->nparens);
260 /* It would seem that the similar code in regtry()
261 * already takes care of this, and in fact it is in
262 * a better location to since this code can #if 0-ed out
263 * but the code in regtry() is needed or otherwise tests
264 * requiring null fields (pat.t#187 and split.t#{13,14}
265 * (as of patchlevel 7877) will fail. Then again,
266 * this code seems to be necessary or otherwise
267 * building DynaLoader will fail:
268 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
270 for (i = *PL_reglastparen + 1; i <= rex->nparens; i++) {
272 PL_regstartp[i] = -1;
279 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
281 #define TRYPAREN(paren, n, input, where) { \
284 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
285 PL_regendp[paren] = input - PL_bostr; \
288 PL_regendp[paren] = -1; \
290 REGMATCH(next, where); \
294 PL_regendp[paren] = -1; \
299 * pregexec and friends
303 - pregexec - match a regexp against a string
306 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
307 char *strbeg, I32 minend, SV *screamer, U32 nosave)
308 /* strend: pointer to null at end of string */
309 /* strbeg: real beginning of string */
310 /* minend: end of match must be >=minend after stringarg. */
311 /* nosave: For optimizations. */
314 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
315 nosave ? 0 : REXEC_COPY_STR);
320 * Need to implement the following flags for reg_anch:
322 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
324 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
325 * INTUIT_AUTORITATIVE_ML
326 * INTUIT_ONCE_NOML - Intuit can match in one location only.
329 * Another flag for this function: SECOND_TIME (so that float substrs
330 * with giant delta may be not rechecked).
333 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
335 /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend.
336 Otherwise, only SvCUR(sv) is used to get strbeg. */
338 /* XXXX We assume that strpos is strbeg unless sv. */
340 /* XXXX Some places assume that there is a fixed substring.
341 An update may be needed if optimizer marks as "INTUITable"
342 RExen without fixed substrings. Similarly, it is assumed that
343 lengths of all the strings are no more than minlen, thus they
344 cannot come from lookahead.
345 (Or minlen should take into account lookahead.) */
347 /* A failure to find a constant substring means that there is no need to make
348 an expensive call to REx engine, thus we celebrate a failure. Similarly,
349 finding a substring too deep into the string means that less calls to
350 regtry() should be needed.
352 REx compiler's optimizer found 4 possible hints:
353 a) Anchored substring;
355 c) Whether we are anchored (beginning-of-line or \G);
356 d) First node (of those at offset 0) which may distingush positions;
357 We use a)b)d) and multiline-part of c), and try to find a position in the
358 string which does not contradict any of them.
361 /* Most of decisions we do here should have been done at compile time.
362 The nodes of the REx which we used for the search should have been
363 deleted from the finite automaton. */
366 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
367 char *strend, U32 flags, re_scream_pos_data *data)
370 register I32 start_shift = 0;
371 /* Should be nonnegative! */
372 register I32 end_shift = 0;
377 const int do_utf8 = sv ? SvUTF8(sv) : 0; /* if no sv we have to assume bytes */
379 register char *other_last = NULL; /* other substr checked before this */
380 char *check_at = NULL; /* check substr found at this pos */
381 const I32 multiline = prog->reganch & PMf_MULTILINE;
383 const char * const i_strpos = strpos;
384 SV * const dsv = PERL_DEBUG_PAD_ZERO(0);
387 GET_RE_DEBUG_FLAGS_DECL;
389 RX_MATCH_UTF8_set(prog,do_utf8);
391 if (prog->reganch & ROPT_UTF8) {
392 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
393 "UTF-8 regex...\n"));
394 PL_reg_flags |= RF_utf8;
398 const char *s = PL_reg_match_utf8 ?
399 sv_uni_display(dsv, sv, 60, UNI_DISPLAY_REGEX) :
401 const int len = PL_reg_match_utf8 ?
402 strlen(s) : strend - strpos;
405 if (PL_reg_match_utf8)
406 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
407 "UTF-8 target...\n"));
408 PerlIO_printf(Perl_debug_log,
409 "%sGuessing start of match, REx%s \"%s%.60s%s%s\" against \"%s%.*s%s%s\"...\n",
410 PL_colors[4], PL_colors[5], PL_colors[0],
413 (strlen(prog->precomp) > 60 ? "..." : ""),
415 (int)(len > 60 ? 60 : len),
417 (len > 60 ? "..." : "")
421 /* CHR_DIST() would be more correct here but it makes things slow. */
422 if (prog->minlen > strend - strpos) {
423 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
424 "String too short... [re_intuit_start]\n"));
427 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
430 if (!prog->check_utf8 && prog->check_substr)
431 to_utf8_substr(prog);
432 check = prog->check_utf8;
434 if (!prog->check_substr && prog->check_utf8)
435 to_byte_substr(prog);
436 check = prog->check_substr;
438 if (check == &PL_sv_undef) {
439 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
440 "Non-utf string cannot match utf check string\n"));
443 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
444 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
445 || ( (prog->reganch & ROPT_ANCH_BOL)
446 && !multiline ) ); /* Check after \n? */
449 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
450 | ROPT_IMPLICIT)) /* not a real BOL */
451 /* SvCUR is not set on references: SvRV and SvPVX_const overlap */
453 && (strpos != strbeg)) {
454 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
457 if (prog->check_offset_min == prog->check_offset_max &&
458 !(prog->reganch & ROPT_CANY_SEEN)) {
459 /* Substring at constant offset from beg-of-str... */
462 s = HOP3c(strpos, prog->check_offset_min, strend);
464 slen = SvCUR(check); /* >= 1 */
466 if ( strend - s > slen || strend - s < slen - 1
467 || (strend - s == slen && strend[-1] != '\n')) {
468 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
471 /* Now should match s[0..slen-2] */
473 if (slen && (*SvPVX_const(check) != *s
475 && memNE(SvPVX_const(check), s, slen)))) {
477 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
481 else if (*SvPVX_const(check) != *s
482 || ((slen = SvCUR(check)) > 1
483 && memNE(SvPVX_const(check), s, slen)))
486 goto success_at_start;
489 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
491 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
492 end_shift = prog->minlen - start_shift -
493 CHR_SVLEN(check) + (SvTAIL(check) != 0);
495 const I32 end = prog->check_offset_max + CHR_SVLEN(check)
496 - (SvTAIL(check) != 0);
497 const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
499 if (end_shift < eshift)
503 else { /* Can match at random position */
506 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
507 /* Should be nonnegative! */
508 end_shift = prog->minlen - start_shift -
509 CHR_SVLEN(check) + (SvTAIL(check) != 0);
512 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
514 Perl_croak(aTHX_ "panic: end_shift");
518 /* Find a possible match in the region s..strend by looking for
519 the "check" substring in the region corrected by start/end_shift. */
520 if (flags & REXEC_SCREAM) {
521 I32 p = -1; /* Internal iterator of scream. */
522 I32 * const pp = data ? data->scream_pos : &p;
524 if (PL_screamfirst[BmRARE(check)] >= 0
525 || ( BmRARE(check) == '\n'
526 && (BmPREVIOUS(check) == SvCUR(check) - 1)
528 s = screaminstr(sv, check,
529 start_shift + (s - strbeg), end_shift, pp, 0);
532 /* we may be pointing at the wrong string */
533 if (s && RX_MATCH_COPIED(prog))
534 s = strbeg + (s - SvPVX_const(sv));
536 *data->scream_olds = s;
538 else if (prog->reganch & ROPT_CANY_SEEN)
539 s = fbm_instr((U8*)(s + start_shift),
540 (U8*)(strend - end_shift),
541 check, multiline ? FBMrf_MULTILINE : 0);
543 s = fbm_instr(HOP3(s, start_shift, strend),
544 HOP3(strend, -end_shift, strbeg),
545 check, multiline ? FBMrf_MULTILINE : 0);
547 /* Update the count-of-usability, remove useless subpatterns,
550 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s %s substr \"%s%.*s%s\"%s%s",
551 (s ? "Found" : "Did not find"),
552 (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) ? "anchored" : "floating"),
554 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
556 PL_colors[1], (SvTAIL(check) ? "$" : ""),
557 (s ? " at offset " : "...\n") ) );
564 /* Finish the diagnostic message */
565 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
567 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
568 Start with the other substr.
569 XXXX no SCREAM optimization yet - and a very coarse implementation
570 XXXX /ttx+/ results in anchored="ttx", floating="x". floating will
571 *always* match. Probably should be marked during compile...
572 Probably it is right to do no SCREAM here...
575 if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8) : (prog->float_substr && prog->anchored_substr)) {
576 /* Take into account the "other" substring. */
577 /* XXXX May be hopelessly wrong for UTF... */
580 if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
583 char * const last = HOP3c(s, -start_shift, strbeg);
588 t = s - prog->check_offset_max;
589 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
591 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
596 t = HOP3c(t, prog->anchored_offset, strend);
597 if (t < other_last) /* These positions already checked */
599 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
602 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
603 /* On end-of-str: see comment below. */
604 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
605 if (must == &PL_sv_undef) {
607 DEBUG_EXECUTE_r(must = prog->anchored_utf8); /* for debug */
612 HOP3(HOP3(last1, prog->anchored_offset, strend)
613 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
615 multiline ? FBMrf_MULTILINE : 0
617 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
618 "%s anchored substr \"%s%.*s%s\"%s",
619 (s ? "Found" : "Contradicts"),
622 - (SvTAIL(must)!=0)),
624 PL_colors[1], (SvTAIL(must) ? "$" : "")));
626 if (last1 >= last2) {
627 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
628 ", giving up...\n"));
631 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
632 ", trying floating at offset %ld...\n",
633 (long)(HOP3c(s1, 1, strend) - i_strpos)));
634 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
635 s = HOP3c(last, 1, strend);
639 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
640 (long)(s - i_strpos)));
641 t = HOP3c(s, -prog->anchored_offset, strbeg);
642 other_last = HOP3c(s, 1, strend);
650 else { /* Take into account the floating substring. */
655 t = HOP3c(s, -start_shift, strbeg);
657 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
658 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
659 last = HOP3c(t, prog->float_max_offset, strend);
660 s = HOP3c(t, prog->float_min_offset, strend);
663 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
664 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
665 /* fbm_instr() takes into account exact value of end-of-str
666 if the check is SvTAIL(ed). Since false positives are OK,
667 and end-of-str is not later than strend we are OK. */
668 if (must == &PL_sv_undef) {
670 DEBUG_EXECUTE_r(must = prog->float_utf8); /* for debug message */
673 s = fbm_instr((unsigned char*)s,
674 (unsigned char*)last + SvCUR(must)
676 must, multiline ? FBMrf_MULTILINE : 0);
677 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s floating substr \"%s%.*s%s\"%s",
678 (s ? "Found" : "Contradicts"),
680 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
682 PL_colors[1], (SvTAIL(must) ? "$" : "")));
685 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
686 ", giving up...\n"));
689 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
690 ", trying anchored starting at offset %ld...\n",
691 (long)(s1 + 1 - i_strpos)));
693 s = HOP3c(t, 1, strend);
697 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
698 (long)(s - i_strpos)));
699 other_last = s; /* Fix this later. --Hugo */
708 t = s - prog->check_offset_max;
709 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
711 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
713 /* Fixed substring is found far enough so that the match
714 cannot start at strpos. */
716 if (ml_anch && t[-1] != '\n') {
717 /* Eventually fbm_*() should handle this, but often
718 anchored_offset is not 0, so this check will not be wasted. */
719 /* XXXX In the code below we prefer to look for "^" even in
720 presence of anchored substrings. And we search even
721 beyond the found float position. These pessimizations
722 are historical artefacts only. */
724 while (t < strend - prog->minlen) {
726 if (t < check_at - prog->check_offset_min) {
727 if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
728 /* Since we moved from the found position,
729 we definitely contradict the found anchored
730 substr. Due to the above check we do not
731 contradict "check" substr.
732 Thus we can arrive here only if check substr
733 is float. Redo checking for "other"=="fixed".
736 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
737 PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
738 goto do_other_anchored;
740 /* We don't contradict the found floating substring. */
741 /* XXXX Why not check for STCLASS? */
743 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
744 PL_colors[0], PL_colors[1], (long)(s - i_strpos)));
747 /* Position contradicts check-string */
748 /* XXXX probably better to look for check-string
749 than for "\n", so one should lower the limit for t? */
750 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
751 PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos)));
752 other_last = strpos = s = t + 1;
757 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
758 PL_colors[0], PL_colors[1]));
762 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
763 PL_colors[0], PL_colors[1]));
767 ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
770 /* The found string does not prohibit matching at strpos,
771 - no optimization of calling REx engine can be performed,
772 unless it was an MBOL and we are not after MBOL,
773 or a future STCLASS check will fail this. */
775 /* Even in this situation we may use MBOL flag if strpos is offset
776 wrt the start of the string. */
777 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
778 && (strpos != strbeg) && strpos[-1] != '\n'
779 /* May be due to an implicit anchor of m{.*foo} */
780 && !(prog->reganch & ROPT_IMPLICIT))
785 DEBUG_EXECUTE_r( if (ml_anch)
786 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
787 (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]);
790 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
792 prog->check_utf8 /* Could be deleted already */
793 && --BmUSEFUL(prog->check_utf8) < 0
794 && (prog->check_utf8 == prog->float_utf8)
796 prog->check_substr /* Could be deleted already */
797 && --BmUSEFUL(prog->check_substr) < 0
798 && (prog->check_substr == prog->float_substr)
801 /* If flags & SOMETHING - do not do it many times on the same match */
802 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
803 SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr);
804 if (do_utf8 ? prog->check_substr : prog->check_utf8)
805 SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8);
806 prog->check_substr = prog->check_utf8 = NULL; /* disable */
807 prog->float_substr = prog->float_utf8 = NULL; /* clear */
808 check = NULL; /* abort */
810 /* XXXX This is a remnant of the old implementation. It
811 looks wasteful, since now INTUIT can use many
813 prog->reganch &= ~RE_USE_INTUIT;
820 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
821 if (prog->regstclass) {
822 /* minlen == 0 is possible if regstclass is \b or \B,
823 and the fixed substr is ''$.
824 Since minlen is already taken into account, s+1 is before strend;
825 accidentally, minlen >= 1 guaranties no false positives at s + 1
826 even for \b or \B. But (minlen? 1 : 0) below assumes that
827 regstclass does not come from lookahead... */
828 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
829 This leaves EXACTF only, which is dealt with in find_byclass(). */
830 const U8* const str = (U8*)STRING(prog->regstclass);
831 const int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
832 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
834 const char * const endpos = (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
835 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
836 : (prog->float_substr || prog->float_utf8
837 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
843 s = find_byclass(prog, prog->regstclass, s, endpos, 1);
846 const char *what = NULL;
848 if (endpos == strend) {
849 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
850 "Could not match STCLASS...\n") );
853 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
854 "This position contradicts STCLASS...\n") );
855 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
857 /* Contradict one of substrings */
858 if (prog->anchored_substr || prog->anchored_utf8) {
859 if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
860 DEBUG_EXECUTE_r( what = "anchored" );
862 s = HOP3c(t, 1, strend);
863 if (s + start_shift + end_shift > strend) {
864 /* XXXX Should be taken into account earlier? */
865 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
866 "Could not match STCLASS...\n") );
871 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
872 "Looking for %s substr starting at offset %ld...\n",
873 what, (long)(s + start_shift - i_strpos)) );
876 /* Have both, check_string is floating */
877 if (t + start_shift >= check_at) /* Contradicts floating=check */
878 goto retry_floating_check;
879 /* Recheck anchored substring, but not floating... */
883 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
884 "Looking for anchored substr starting at offset %ld...\n",
885 (long)(other_last - i_strpos)) );
886 goto do_other_anchored;
888 /* Another way we could have checked stclass at the
889 current position only: */
894 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
895 "Looking for /%s^%s/m starting at offset %ld...\n",
896 PL_colors[0], PL_colors[1], (long)(t - i_strpos)) );
899 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
901 /* Check is floating subtring. */
902 retry_floating_check:
903 t = check_at - start_shift;
904 DEBUG_EXECUTE_r( what = "floating" );
905 goto hop_and_restart;
908 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
909 "By STCLASS: moving %ld --> %ld\n",
910 (long)(t - i_strpos), (long)(s - i_strpos))
914 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
915 "Does not contradict STCLASS...\n");
920 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
921 PL_colors[4], (check ? "Guessed" : "Giving up"),
922 PL_colors[5], (long)(s - i_strpos)) );
925 fail_finish: /* Substring not found */
926 if (prog->check_substr || prog->check_utf8) /* could be removed already */
927 BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
929 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
930 PL_colors[4], PL_colors[5]));
934 /* We know what class REx starts with. Try to find this position... */
936 S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, const char *strend, I32 norun)
939 const I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
943 register STRLEN uskip;
947 register I32 tmp = 1; /* Scratch variable? */
948 register const bool do_utf8 = PL_reg_match_utf8;
950 /* We know what class it must start with. */
954 while (s + (uskip = UTF8SKIP(s)) <= strend) {
955 if ((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
956 !UTF8_IS_INVARIANT((U8)s[0]) ?
957 reginclass(c, (U8*)s, 0, do_utf8) :
958 REGINCLASS(c, (U8*)s)) {
959 if (tmp && (norun || regtry(prog, s)))
973 if (REGINCLASS(c, (U8*)s) ||
974 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
975 /* The assignment of 2 is intentional:
976 * for the folded sharp s, the skip is 2. */
977 (skip = SHARP_S_SKIP))) {
978 if (tmp && (norun || regtry(prog, s)))
991 if (tmp && (norun || regtry(prog, s)))
1000 ln = STR_LEN(c); /* length to match in octets/bytes */
1001 lnc = (I32) ln; /* length to match in characters */
1003 STRLEN ulen1, ulen2;
1005 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
1006 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
1007 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1009 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1010 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1012 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE,
1014 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE,
1017 while (sm < ((U8 *) m + ln)) {
1032 c2 = PL_fold_locale[c1];
1034 e = HOP3c(strend, -((I32)lnc), s);
1037 e = s; /* Due to minlen logic of intuit() */
1039 /* The idea in the EXACTF* cases is to first find the
1040 * first character of the EXACTF* node and then, if
1041 * necessary, case-insensitively compare the full
1042 * text of the node. The c1 and c2 are the first
1043 * characters (though in Unicode it gets a bit
1044 * more complicated because there are more cases
1045 * than just upper and lower: one needs to use
1046 * the so-called folding case for case-insensitive
1047 * matching (called "loose matching" in Unicode).
1048 * ibcmp_utf8() will do just that. */
1052 U8 tmpbuf [UTF8_MAXBYTES+1];
1053 STRLEN len, foldlen;
1054 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1056 /* Upper and lower of 1st char are equal -
1057 * probably not a "letter". */
1059 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
1063 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1064 m, (char **)0, ln, (bool)UTF))
1065 && (norun || regtry(prog, s)) )
1068 U8 foldbuf[UTF8_MAXBYTES_CASE+1];
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_MAXBYTES, &len,
1089 /* Handle some of the three Greek sigmas cases.
1090 * Note that not all the possible combinations
1091 * are handled here: some of them are handled
1092 * by the standard folding rules, and some of
1093 * them (the character class or ANYOF cases)
1094 * are handled during compiletime in
1095 * regexec.c:S_regclass(). */
1096 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1097 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1098 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1100 if ( (c == c1 || c == c2)
1102 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1103 m, (char **)0, ln, (bool)UTF))
1104 && (norun || regtry(prog, s)) )
1107 U8 foldbuf[UTF8_MAXBYTES_CASE+1];
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 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1156 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT);
1158 tmp = ((OP(c) == BOUND ?
1159 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1160 LOAD_UTF8_CHARCLASS_ALNUM();
1161 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1162 if (tmp == !(OP(c) == BOUND ?
1163 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1164 isALNUM_LC_utf8((U8*)s)))
1167 if ((norun || regtry(prog, s)))
1174 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1175 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1176 while (s < strend) {
1178 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1180 if ((norun || regtry(prog, s)))
1186 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1190 PL_reg_flags |= RF_tainted;
1197 U8 * const r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1198 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, UTF8_ALLOW_DEFAULT);
1200 tmp = ((OP(c) == NBOUND ?
1201 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1202 LOAD_UTF8_CHARCLASS_ALNUM();
1203 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1204 if (tmp == !(OP(c) == NBOUND ?
1205 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1206 isALNUM_LC_utf8((U8*)s)))
1208 else if ((norun || regtry(prog, s)))
1214 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1215 tmp = ((OP(c) == NBOUND ?
1216 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1217 while (s < strend) {
1219 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1221 else if ((norun || regtry(prog, s)))
1226 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1231 LOAD_UTF8_CHARCLASS_ALNUM();
1232 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1233 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1234 if (tmp && (norun || regtry(prog, s)))
1245 while (s < strend) {
1247 if (tmp && (norun || regtry(prog, s)))
1259 PL_reg_flags |= RF_tainted;
1261 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1262 if (isALNUM_LC_utf8((U8*)s)) {
1263 if (tmp && (norun || regtry(prog, s)))
1274 while (s < strend) {
1275 if (isALNUM_LC(*s)) {
1276 if (tmp && (norun || regtry(prog, s)))
1289 LOAD_UTF8_CHARCLASS_ALNUM();
1290 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1291 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1292 if (tmp && (norun || regtry(prog, s)))
1303 while (s < strend) {
1305 if (tmp && (norun || regtry(prog, s)))
1317 PL_reg_flags |= RF_tainted;
1319 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1320 if (!isALNUM_LC_utf8((U8*)s)) {
1321 if (tmp && (norun || regtry(prog, s)))
1332 while (s < strend) {
1333 if (!isALNUM_LC(*s)) {
1334 if (tmp && (norun || regtry(prog, s)))
1347 LOAD_UTF8_CHARCLASS_SPACE();
1348 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1349 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1350 if (tmp && (norun || regtry(prog, s)))
1361 while (s < strend) {
1363 if (tmp && (norun || regtry(prog, s)))
1375 PL_reg_flags |= RF_tainted;
1377 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1378 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1379 if (tmp && (norun || regtry(prog, s)))
1390 while (s < strend) {
1391 if (isSPACE_LC(*s)) {
1392 if (tmp && (norun || regtry(prog, s)))
1405 LOAD_UTF8_CHARCLASS_SPACE();
1406 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1407 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1408 if (tmp && (norun || regtry(prog, s)))
1419 while (s < strend) {
1421 if (tmp && (norun || regtry(prog, s)))
1433 PL_reg_flags |= RF_tainted;
1435 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1436 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1437 if (tmp && (norun || regtry(prog, s)))
1448 while (s < strend) {
1449 if (!isSPACE_LC(*s)) {
1450 if (tmp && (norun || regtry(prog, s)))
1463 LOAD_UTF8_CHARCLASS_DIGIT();
1464 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1465 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1466 if (tmp && (norun || regtry(prog, s)))
1477 while (s < strend) {
1479 if (tmp && (norun || regtry(prog, s)))
1491 PL_reg_flags |= RF_tainted;
1493 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1494 if (isDIGIT_LC_utf8((U8*)s)) {
1495 if (tmp && (norun || regtry(prog, s)))
1506 while (s < strend) {
1507 if (isDIGIT_LC(*s)) {
1508 if (tmp && (norun || regtry(prog, s)))
1521 LOAD_UTF8_CHARCLASS_DIGIT();
1522 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1523 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1524 if (tmp && (norun || regtry(prog, s)))
1535 while (s < strend) {
1537 if (tmp && (norun || regtry(prog, s)))
1549 PL_reg_flags |= RF_tainted;
1551 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1552 if (!isDIGIT_LC_utf8((U8*)s)) {
1553 if (tmp && (norun || regtry(prog, s)))
1564 while (s < strend) {
1565 if (!isDIGIT_LC(*s)) {
1566 if (tmp && (norun || regtry(prog, s)))
1578 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1587 - regexec_flags - match a regexp against a string
1590 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1591 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1592 /* strend: pointer to null at end of string */
1593 /* strbeg: real beginning of string */
1594 /* minend: end of match must be >=minend after stringarg. */
1595 /* data: May be used for some additional optimizations. */
1596 /* nosave: For optimizations. */
1600 register regnode *c;
1601 register char *startpos = stringarg;
1602 I32 minlen; /* must match at least this many chars */
1603 I32 dontbother = 0; /* how many characters not to try at end */
1604 I32 end_shift = 0; /* Same for the end. */ /* CC */
1605 I32 scream_pos = -1; /* Internal iterator of scream. */
1606 char *scream_olds = NULL;
1607 SV* oreplsv = GvSV(PL_replgv);
1608 const bool do_utf8 = DO_UTF8(sv);
1609 const I32 multiline = prog->reganch & PMf_MULTILINE;
1611 SV * const dsv0 = PERL_DEBUG_PAD_ZERO(0);
1612 SV * const dsv1 = PERL_DEBUG_PAD_ZERO(1);
1615 GET_RE_DEBUG_FLAGS_DECL;
1617 PERL_UNUSED_ARG(data);
1619 /* Be paranoid... */
1620 if (prog == NULL || startpos == NULL) {
1621 Perl_croak(aTHX_ "NULL regexp parameter");
1627 PL_regnarrate = DEBUG_r_TEST;
1630 RX_MATCH_UTF8_set(prog, do_utf8);
1632 minlen = prog->minlen;
1633 if (strend - startpos < minlen) {
1634 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1635 "String too short [regexec_flags]...\n"));
1639 /* Check validity of program. */
1640 if (UCHARAT(prog->program) != REG_MAGIC) {
1641 Perl_croak(aTHX_ "corrupted regexp program");
1645 PL_reg_eval_set = 0;
1648 if (prog->reganch & ROPT_UTF8)
1649 PL_reg_flags |= RF_utf8;
1651 /* Mark beginning of line for ^ and lookbehind. */
1652 PL_regbol = startpos;
1656 /* Mark end of line for $ (and such) */
1659 /* see how far we have to get to not match where we matched before */
1660 PL_regtill = startpos+minend;
1662 /* If there is a "must appear" string, look for it. */
1665 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1668 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1669 PL_reg_ganch = startpos;
1670 else if (sv && SvTYPE(sv) >= SVt_PVMG
1672 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1673 && mg->mg_len >= 0) {
1674 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1675 if (prog->reganch & ROPT_ANCH_GPOS) {
1676 if (s > PL_reg_ganch)
1681 else /* pos() not defined */
1682 PL_reg_ganch = strbeg;
1685 if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) {
1686 re_scream_pos_data d;
1688 d.scream_olds = &scream_olds;
1689 d.scream_pos = &scream_pos;
1690 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1692 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1693 goto phooey; /* not present */
1698 const char * const s0 = UTF
1699 ? pv_uni_display(dsv0, (U8*)prog->precomp, prog->prelen, 60,
1702 const int len0 = UTF ? SvCUR(dsv0) : prog->prelen;
1703 const char * const s1 = do_utf8 ? sv_uni_display(dsv1, sv, 60,
1704 UNI_DISPLAY_REGEX) : startpos;
1705 const int len1 = do_utf8 ? SvCUR(dsv1) : strend - startpos;
1708 PerlIO_printf(Perl_debug_log,
1709 "%sMatching REx%s \"%s%*.*s%s%s\" against \"%s%.*s%s%s\"\n",
1710 PL_colors[4], PL_colors[5], PL_colors[0],
1713 len0 > 60 ? "..." : "",
1715 (int)(len1 > 60 ? 60 : len1),
1717 (len1 > 60 ? "..." : "")
1721 /* Simplest case: anchored match need be tried only once. */
1722 /* [unless only anchor is BOL and multiline is set] */
1723 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1724 if (s == startpos && regtry(prog, startpos))
1726 else if (multiline || (prog->reganch & ROPT_IMPLICIT)
1727 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1732 dontbother = minlen - 1;
1733 end = HOP3c(strend, -dontbother, strbeg) - 1;
1734 /* for multiline we only have to try after newlines */
1735 if (prog->check_substr || prog->check_utf8) {
1739 if (regtry(prog, s))
1744 if (prog->reganch & RE_USE_INTUIT) {
1745 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1756 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1757 if (regtry(prog, s))
1764 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1765 if (regtry(prog, PL_reg_ganch))
1770 /* Messy cases: unanchored match. */
1771 if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) {
1772 /* we have /x+whatever/ */
1773 /* it must be a one character string (XXXX Except UTF?) */
1778 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1779 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1780 ch = SvPVX_const(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
1783 while (s < strend) {
1785 DEBUG_EXECUTE_r( did_match = 1 );
1786 if (regtry(prog, s)) goto got_it;
1788 while (s < strend && *s == ch)
1795 while (s < strend) {
1797 DEBUG_EXECUTE_r( did_match = 1 );
1798 if (regtry(prog, s)) goto got_it;
1800 while (s < strend && *s == ch)
1806 DEBUG_EXECUTE_r(if (!did_match)
1807 PerlIO_printf(Perl_debug_log,
1808 "Did not find anchored character...\n")
1811 else if (prog->anchored_substr != NULL
1812 || prog->anchored_utf8 != NULL
1813 || ((prog->float_substr != NULL || prog->float_utf8 != NULL)
1814 && prog->float_max_offset < strend - s)) {
1819 char *last1; /* Last position checked before */
1823 if (prog->anchored_substr || prog->anchored_utf8) {
1824 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1825 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1826 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1827 back_max = back_min = prog->anchored_offset;
1829 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1830 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1831 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1832 back_max = prog->float_max_offset;
1833 back_min = prog->float_min_offset;
1835 if (must == &PL_sv_undef)
1836 /* could not downgrade utf8 check substring, so must fail */
1839 last = HOP3c(strend, /* Cannot start after this */
1840 -(I32)(CHR_SVLEN(must)
1841 - (SvTAIL(must) != 0) + back_min), strbeg);
1844 last1 = HOPc(s, -1);
1846 last1 = s - 1; /* bogus */
1848 /* XXXX check_substr already used to find "s", can optimize if
1849 check_substr==must. */
1851 dontbother = end_shift;
1852 strend = HOPc(strend, -dontbother);
1853 while ( (s <= last) &&
1854 ((flags & REXEC_SCREAM)
1855 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1856 end_shift, &scream_pos, 0))
1857 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1858 (unsigned char*)strend, must,
1859 multiline ? FBMrf_MULTILINE : 0))) ) {
1860 /* we may be pointing at the wrong string */
1861 if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog))
1862 s = strbeg + (s - SvPVX_const(sv));
1863 DEBUG_EXECUTE_r( did_match = 1 );
1864 if (HOPc(s, -back_max) > last1) {
1865 last1 = HOPc(s, -back_min);
1866 s = HOPc(s, -back_max);
1869 char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1871 last1 = HOPc(s, -back_min);
1875 while (s <= last1) {
1876 if (regtry(prog, s))
1882 while (s <= last1) {
1883 if (regtry(prog, s))
1889 DEBUG_EXECUTE_r(if (!did_match)
1890 PerlIO_printf(Perl_debug_log,
1891 "Did not find %s substr \"%s%.*s%s\"%s...\n",
1892 ((must == prog->anchored_substr || must == prog->anchored_utf8)
1893 ? "anchored" : "floating"),
1895 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1897 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1901 else if ((c = prog->regstclass)) {
1903 I32 op = (U8)OP(prog->regstclass);
1904 /* don't bother with what can't match */
1905 if (PL_regkind[op] != EXACT && op != CANY)
1906 strend = HOPc(strend, -(minlen - 1));
1909 SV *prop = sv_newmortal();
1917 pv_uni_display(dsv0, (U8*)SvPVX_const(prop), SvCUR(prop), 60,
1918 UNI_DISPLAY_REGEX) :
1920 len0 = UTF ? SvCUR(dsv0) : SvCUR(prop);
1922 sv_uni_display(dsv1, sv, 60, UNI_DISPLAY_REGEX) : s;
1923 len1 = UTF ? SvCUR(dsv1) : strend - s;
1924 PerlIO_printf(Perl_debug_log,
1925 "Matching stclass \"%*.*s\" against \"%*.*s\"\n",
1929 if (find_byclass(prog, c, s, strend, 0))
1931 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1935 if (prog->float_substr != NULL || prog->float_utf8 != NULL) {
1940 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1941 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1942 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
1944 if (flags & REXEC_SCREAM) {
1945 last = screaminstr(sv, float_real, s - strbeg,
1946 end_shift, &scream_pos, 1); /* last one */
1948 last = scream_olds; /* Only one occurrence. */
1949 /* we may be pointing at the wrong string */
1950 else if (RX_MATCH_COPIED(prog))
1951 s = strbeg + (s - SvPVX_const(sv));
1955 const char * const little = SvPV_const(float_real, len);
1957 if (SvTAIL(float_real)) {
1958 if (memEQ(strend - len + 1, little, len - 1))
1959 last = strend - len + 1;
1960 else if (!multiline)
1961 last = memEQ(strend - len, little, len)
1962 ? strend - len : NULL;
1968 last = rninstr(s, strend, little, little + len);
1970 last = strend; /* matching "$" */
1974 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1975 "%sCan't trim the tail, match fails (should not happen)%s\n",
1976 PL_colors[4], PL_colors[5]));
1977 goto phooey; /* Should not happen! */
1979 dontbother = strend - last + prog->float_min_offset;
1981 if (minlen && (dontbother < minlen))
1982 dontbother = minlen - 1;
1983 strend -= dontbother; /* this one's always in bytes! */
1984 /* We don't know much -- general case. */
1987 if (regtry(prog, s))
1996 if (regtry(prog, s))
1998 } while (s++ < strend);
2006 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
2008 if (PL_reg_eval_set) {
2009 /* Preserve the current value of $^R */
2010 if (oreplsv != GvSV(PL_replgv))
2011 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
2012 restored, the value remains
2014 restore_pos(aTHX_ prog);
2017 /* make sure $`, $&, $', and $digit will work later */
2018 if ( !(flags & REXEC_NOT_FIRST) ) {
2019 RX_MATCH_COPY_FREE(prog);
2020 if (flags & REXEC_COPY_STR) {
2021 I32 i = PL_regeol - startpos + (stringarg - strbeg);
2022 #ifdef PERL_OLD_COPY_ON_WRITE
2024 || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) {
2026 PerlIO_printf(Perl_debug_log,
2027 "Copy on write: regexp capture, type %d\n",
2030 prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv);
2031 prog->subbeg = (char *)SvPVX_const(prog->saved_copy);
2032 assert (SvPOKp(prog->saved_copy));
2036 RX_MATCH_COPIED_on(prog);
2037 s = savepvn(strbeg, i);
2043 prog->subbeg = strbeg;
2044 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2051 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2052 PL_colors[4], PL_colors[5]));
2053 if (PL_reg_eval_set)
2054 restore_pos(aTHX_ prog);
2059 - regtry - try match at specific point
2061 STATIC I32 /* 0 failure, 1 success */
2062 S_regtry(pTHX_ regexp *prog, char *startpos)
2068 GET_RE_DEBUG_FLAGS_DECL;
2071 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
2073 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
2076 PL_reg_eval_set = RS_init;
2077 DEBUG_EXECUTE_r(DEBUG_s(
2078 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2079 (IV)(PL_stack_sp - PL_stack_base));
2081 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
2082 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2083 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2085 /* Apparently this is not needed, judging by wantarray. */
2086 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
2087 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2090 /* Make $_ available to executed code. */
2091 if (PL_reg_sv != DEFSV) {
2096 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
2097 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
2098 /* prepare for quick setting of pos */
2099 #ifdef PERL_OLD_COPY_ON_WRITE
2101 sv_force_normal_flags(sv, 0);
2103 mg = sv_magicext(PL_reg_sv, (SV*)0, PERL_MAGIC_regex_global,
2104 &PL_vtbl_mglob, NULL, 0);
2108 PL_reg_oldpos = mg->mg_len;
2109 SAVEDESTRUCTOR_X(restore_pos, prog);
2111 if (!PL_reg_curpm) {
2112 Newxz(PL_reg_curpm, 1, PMOP);
2115 SV* repointer = newSViv(0);
2116 /* so we know which PL_regex_padav element is PL_reg_curpm */
2117 SvFLAGS(repointer) |= SVf_BREAK;
2118 av_push(PL_regex_padav,repointer);
2119 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2120 PL_regex_pad = AvARRAY(PL_regex_padav);
2124 PM_SETRE(PL_reg_curpm, prog);
2125 PL_reg_oldcurpm = PL_curpm;
2126 PL_curpm = PL_reg_curpm;
2127 if (RX_MATCH_COPIED(prog)) {
2128 /* Here is a serious problem: we cannot rewrite subbeg,
2129 since it may be needed if this match fails. Thus
2130 $` inside (?{}) could fail... */
2131 PL_reg_oldsaved = prog->subbeg;
2132 PL_reg_oldsavedlen = prog->sublen;
2133 #ifdef PERL_OLD_COPY_ON_WRITE
2134 PL_nrs = prog->saved_copy;
2136 RX_MATCH_COPIED_off(prog);
2139 PL_reg_oldsaved = NULL;
2140 prog->subbeg = PL_bostr;
2141 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2143 prog->startp[0] = startpos - PL_bostr;
2144 PL_reginput = startpos;
2145 PL_regstartp = prog->startp;
2146 PL_regendp = prog->endp;
2147 PL_reglastparen = &prog->lastparen;
2148 PL_reglastcloseparen = &prog->lastcloseparen;
2149 prog->lastparen = 0;
2150 prog->lastcloseparen = 0;
2152 DEBUG_EXECUTE_r(PL_reg_starttry = startpos);
2153 if (PL_reg_start_tmpl <= prog->nparens) {
2154 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2155 if(PL_reg_start_tmp)
2156 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2158 Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2161 /* XXXX What this code is doing here?!!! There should be no need
2162 to do this again and again, PL_reglastparen should take care of
2165 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2166 * Actually, the code in regcppop() (which Ilya may be meaning by
2167 * PL_reglastparen), is not needed at all by the test suite
2168 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2169 * enough, for building DynaLoader, or otherwise this
2170 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2171 * will happen. Meanwhile, this code *is* needed for the
2172 * above-mentioned test suite tests to succeed. The common theme
2173 * on those tests seems to be returning null fields from matches.
2178 if (prog->nparens) {
2180 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2187 if (regmatch(prog, prog->program + 1)) {
2188 prog->endp[0] = PL_reginput - PL_bostr;
2191 REGCP_UNWIND(lastcp);
2195 #define RE_UNWIND_BRANCH 1
2196 #define RE_UNWIND_BRANCHJ 2
2200 typedef struct { /* XX: makes sense to enlarge it... */
2204 } re_unwind_generic_t;
2217 } re_unwind_branch_t;
2219 typedef union re_unwind_t {
2221 re_unwind_generic_t generic;
2222 re_unwind_branch_t branch;
2225 #define sayYES goto yes
2226 #define sayNO goto no
2227 #define sayNO_ANYOF goto no_anyof
2228 #define sayYES_FINAL goto yes_final
2229 #define sayYES_LOUD goto yes_loud
2230 #define sayNO_FINAL goto no_final
2231 #define sayNO_SILENT goto do_no
2232 #define saySAME(x) if (x) goto yes; else goto no
2234 #define POSCACHE_SUCCESS 0 /* caching success rather than failure */
2235 #define POSCACHE_SEEN 1 /* we know what we're caching */
2236 #define POSCACHE_START 2 /* the real cache: this bit maps to pos 0 */
2237 #define CACHEsayYES STMT_START { \
2238 if (st->u.whilem.cache_offset | st->u.whilem.cache_bit) { \
2239 if (!(PL_reg_poscache[0] & (1<<POSCACHE_SEEN))) \
2240 PL_reg_poscache[0] |= (1<<POSCACHE_SUCCESS) || (1<<POSCACHE_SEEN); \
2241 else if (!(PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))) { \
2242 /* cache records failure, but this is success */ \
2244 PerlIO_printf(Perl_debug_log, \
2245 "%*s (remove success from failure cache)\n", \
2246 REPORT_CODE_OFF+PL_regindent*2, "") \
2248 PL_reg_poscache[st->u.whilem.cache_offset] &= ~(1<<st->u.whilem.cache_bit); \
2253 #define CACHEsayNO STMT_START { \
2254 if (st->u.whilem.cache_offset | st->u.whilem.cache_bit) { \
2255 if (!(PL_reg_poscache[0] & (1<<POSCACHE_SEEN))) \
2256 PL_reg_poscache[0] |= (1<<POSCACHE_SEEN); \
2257 else if ((PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))) { \
2258 /* cache records success, but this is failure */ \
2260 PerlIO_printf(Perl_debug_log, \
2261 "%*s (remove failure from success cache)\n", \
2262 REPORT_CODE_OFF+PL_regindent*2, "") \
2264 PL_reg_poscache[st->u.whilem.cache_offset] &= ~(1<<st->u.whilem.cache_bit); \
2270 /* this is used to determine how far from the left messages like
2271 'failed...' are printed. Currently 29 makes these messages line
2272 up with the opcode they refer to. Earlier perls used 25 which
2273 left these messages outdented making reviewing a debug output
2276 #define REPORT_CODE_OFF 29
2279 /* Make sure there is a test for this +1 options in re_tests */
2280 #define TRIE_INITAL_ACCEPT_BUFFLEN 4;
2282 /* grab a new slab and return the first slot in it */
2284 STATIC regmatch_state *
2287 regmatch_slab *s = PL_regmatch_slab->next;
2289 Newx(s, 1, regmatch_slab);
2290 s->prev = PL_regmatch_slab;
2292 PL_regmatch_slab->next = s;
2294 PL_regmatch_slab = s;
2295 return &s->states[0];
2298 /* simulate a recursive call to regmatch */
2300 #define REGMATCH(ns, where) \
2303 st->resume_state = resume_##where; \
2304 goto start_recurse; \
2305 resume_point_##where:
2308 /* push a new regex state. Set newst to point to it */
2310 #define PUSH_STATE(newst, resume) \
2312 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "PUSH STATE(%d)\n", depth)); \
2316 st->locinput = locinput; \
2317 st->resume_state = resume; \
2319 if (newst > &(PL_regmatch_slab->states[PERL_REGMATCH_SLAB_SLOTS-1])) \
2320 newst = S_push_slab(aTHX); \
2321 PL_regmatch_state = newst; \
2323 newst->minmod = 0; \
2325 newst->logical = 0; \
2326 newst->unwind = 0; \
2327 locinput = PL_reginput; \
2328 nextchr = UCHARAT(locinput);
2331 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "POP STATE(%d)\n", depth)); \
2334 if (st < &PL_regmatch_slab->states[0]) { \
2335 PL_regmatch_slab = PL_regmatch_slab->prev; \
2336 st = &PL_regmatch_slab->states[PERL_REGMATCH_SLAB_SLOTS-1]; \
2338 PL_regmatch_state = st; \
2342 locinput = st->locinput; \
2343 nextchr = UCHARAT(locinput);
2346 - regmatch - main matching routine
2348 * Conceptually the strategy is simple: check to see whether the current
2349 * node matches, call self recursively to see whether the rest matches,
2350 * and then act accordingly. In practice we make some effort to avoid
2351 * recursion, in particular by going through "ordinary" nodes (that don't
2352 * need to know whether the rest of the match failed) by a loop instead of
2355 /* [lwall] I've hoisted the register declarations to the outer block in order to
2356 * maybe save a little bit of pushing and popping on the stack. It also takes
2357 * advantage of machines that use a register save mask on subroutine entry.
2359 * This function used to be heavily recursive, but since this had the
2360 * effect of blowing the CPU stack on complex regexes, it has been
2361 * restructured to be iterative, and to save state onto the heap rather
2362 * than the stack. Essentially whereever regmatch() used to be called, it
2363 * pushes the current state, notes where to return, then jumps back into
2366 * Originally the structure of this function used to look something like
2371 while (scan != NULL) {
2372 a++; // do stuff with a and b
2378 if (regmatch(...)) // recurse
2388 * Now it looks something like this:
2396 regmatch_state *st = new();
2398 st->a++; // do stuff with a and b
2400 while (scan != NULL) {
2408 st->resume_state = resume_FOO;
2409 goto start_recurse; // recurse
2418 st = new(); push a new state
2419 st->a = 1; st->b = 2;
2426 switch (resume_state) {
2428 goto resume_point_FOO;
2435 * WARNING: this means that any line in this function that contains a
2436 * REGMATCH() or TRYPAREN() is actually simulating a recursive call to
2437 * regmatch() using gotos instead. Thus the values of any local variables
2438 * not saved in the regmatch_state structure will have been lost when
2439 * execution resumes on the next line .
2441 * States (ie the st pointer) are allocated in slabs of about 4K in size.
2442 * PL_regmatch_state always points to the currently active state, and
2443 * PL_regmatch_slab points to the slab currently containing PL_regmatch_state.
2444 * The first time regmatch is called, the first slab is allocated, and is
2445 * never freed until interpreter desctruction. When the slab is full,
2446 * a new one is allocated chained to the end. At exit from regmatch, slabs
2447 * allocated since entry are freed.
2451 STATIC I32 /* 0 failure, 1 success */
2452 S_regmatch(pTHX_ regexp *rex, regnode *prog)
2455 register const bool do_utf8 = PL_reg_match_utf8;
2456 const U32 uniflags = UTF8_ALLOW_DEFAULT;
2458 regmatch_slab *orig_slab;
2459 regmatch_state *orig_state;
2461 /* the current state. This is a cached copy of PL_regmatch_state */
2462 register regmatch_state *st;
2464 /* cache heavy used fields of st in registers */
2465 register regnode *scan;
2466 register regnode *next;
2467 register I32 n = 0; /* initialize to shut up compiler warning */
2468 register char *locinput = PL_reginput;
2470 /* these variables are NOT saved during a recusive RFEGMATCH: */
2471 register I32 nextchr; /* is always set to UCHARAT(locinput) */
2472 bool result; /* return value of S_regmatch */
2473 regnode *inner; /* Next node in internal branch. */
2474 int depth = 0; /* depth of recursion */
2475 regmatch_state *newst; /* when pushing a state, this is the new one */
2476 regmatch_state *cur_eval = NULL; /* most recent (??{}) state */
2479 SV *re_debug_flags = NULL;
2484 /* on first ever call to regmatch, allocate first slab */
2485 if (!PL_regmatch_slab) {
2486 Newx(PL_regmatch_slab, 1, regmatch_slab);
2487 PL_regmatch_slab->prev = NULL;
2488 PL_regmatch_slab->next = NULL;
2489 PL_regmatch_state = &PL_regmatch_slab->states[0] - 1;
2492 /* remember current high-water mark for exit */
2493 /* XXX this should be done with SAVE* instead */
2494 orig_slab = PL_regmatch_slab;
2495 orig_state = PL_regmatch_state;
2497 /* grab next free state slot */
2498 st = ++PL_regmatch_state;
2499 if (st > &(PL_regmatch_slab->states[PERL_REGMATCH_SLAB_SLOTS-1]))
2500 st = PL_regmatch_state = S_push_slab(aTHX);
2507 /* Note that nextchr is a byte even in UTF */
2508 nextchr = UCHARAT(locinput);
2510 while (scan != NULL) {
2513 SV * const prop = sv_newmortal();
2514 const int docolor = *PL_colors[0];
2515 const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2516 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2517 /* The part of the string before starttry has one color
2518 (pref0_len chars), between starttry and current
2519 position another one (pref_len - pref0_len chars),
2520 after the current position the third one.
2521 We assume that pref0_len <= pref_len, otherwise we
2522 decrease pref0_len. */
2523 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2524 ? (5 + taill) - l : locinput - PL_bostr;
2527 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2529 pref0_len = pref_len - (locinput - PL_reg_starttry);
2530 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2531 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2532 ? (5 + taill) - pref_len : PL_regeol - locinput);
2533 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2537 if (pref0_len > pref_len)
2538 pref0_len = pref_len;
2539 regprop(prop, scan);
2541 const char * const s0 =
2542 do_utf8 && OP(scan) != CANY ?
2543 pv_uni_display(PERL_DEBUG_PAD(0), (U8*)(locinput - pref_len),
2544 pref0_len, 60, UNI_DISPLAY_REGEX) :
2545 locinput - pref_len;
2546 const int len0 = do_utf8 ? strlen(s0) : pref0_len;
2547 const char * const s1 = do_utf8 && OP(scan) != CANY ?
2548 pv_uni_display(PERL_DEBUG_PAD(1),
2549 (U8*)(locinput - pref_len + pref0_len),
2550 pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) :
2551 locinput - pref_len + pref0_len;
2552 const int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2553 const char * const s2 = do_utf8 && OP(scan) != CANY ?
2554 pv_uni_display(PERL_DEBUG_PAD(2), (U8*)locinput,
2555 PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
2557 const int len2 = do_utf8 ? strlen(s2) : l;
2558 PerlIO_printf(Perl_debug_log,
2559 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2560 (IV)(locinput - PL_bostr),
2567 (docolor ? "" : "> <"),
2571 15 - l - pref_len + 1,
2573 (IV)(scan - rex->program), PL_regindent*2, "",
2578 next = scan + NEXT_OFF(scan);
2584 if (locinput == PL_bostr)
2586 /* regtill = regbol; */
2591 if (locinput == PL_bostr ||
2592 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2598 if (locinput == PL_bostr)
2602 if (locinput == PL_reg_ganch)
2608 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2613 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2615 if (PL_regeol - locinput > 1)
2619 if (PL_regeol != locinput)
2623 if (!nextchr && locinput >= PL_regeol)
2626 locinput += PL_utf8skip[nextchr];
2627 if (locinput > PL_regeol)
2629 nextchr = UCHARAT(locinput);
2632 nextchr = UCHARAT(++locinput);
2635 if (!nextchr && locinput >= PL_regeol)
2637 nextchr = UCHARAT(++locinput);
2640 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2643 locinput += PL_utf8skip[nextchr];
2644 if (locinput > PL_regeol)
2646 nextchr = UCHARAT(locinput);
2649 nextchr = UCHARAT(++locinput);
2655 traverse the TRIE keeping track of all accepting states
2656 we transition through until we get to a failing node.
2664 U8 *uc = ( U8* )locinput;
2671 U8 *uscan = (U8*)NULL;
2673 SV *sv_accept_buff = NULL;
2674 const enum { trie_plain, trie_utf8, trie_uft8_fold }
2675 trie_type = do_utf8 ?
2676 (OP(scan) == TRIE ? trie_utf8 : trie_uft8_fold)
2679 /* what trie are we using right now */
2681 = (reg_trie_data*)PL_reg_re->data->data[ ARG( scan ) ];
2682 st->u.trie.accepted = 0; /* how many accepting states we have seen */
2685 while ( state && uc <= (U8*)PL_regeol ) {
2687 if (trie->states[ state ].wordnum) {
2688 if (!st->u.trie.accepted ) {
2691 bufflen = TRIE_INITAL_ACCEPT_BUFFLEN;
2692 sv_accept_buff=newSV(bufflen *
2693 sizeof(reg_trie_accepted) - 1);
2694 SvCUR_set(sv_accept_buff,
2695 sizeof(reg_trie_accepted));
2696 SvPOK_on(sv_accept_buff);
2697 sv_2mortal(sv_accept_buff);
2698 st->u.trie.accept_buff =
2699 (reg_trie_accepted*)SvPV_nolen(sv_accept_buff );
2702 if (st->u.trie.accepted >= bufflen) {
2704 st->u.trie.accept_buff =(reg_trie_accepted*)
2705 SvGROW(sv_accept_buff,
2706 bufflen * sizeof(reg_trie_accepted));
2708 SvCUR_set(sv_accept_buff,SvCUR(sv_accept_buff)
2709 + sizeof(reg_trie_accepted));
2711 st->u.trie.accept_buff[st->u.trie.accepted].wordnum = trie->states[state].wordnum;
2712 st->u.trie.accept_buff[st->u.trie.accepted].endpos = uc;
2713 ++st->u.trie.accepted;
2716 base = trie->states[ state ].trans.base;
2718 DEBUG_TRIE_EXECUTE_r(
2719 PerlIO_printf( Perl_debug_log,
2720 "%*s %sState: %4"UVxf", Base: %4"UVxf", Accepted: %4"UVxf" ",
2721 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
2722 (UV)state, (UV)base, (UV)st->u.trie.accepted );
2726 switch (trie_type) {
2727 case trie_uft8_fold:
2729 uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags );
2734 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
2735 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags );
2736 uvc = to_uni_fold( uvc, foldbuf, &foldlen );
2737 foldlen -= UNISKIP( uvc );
2738 uscan = foldbuf + UNISKIP( uvc );
2742 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN,
2751 charid = trie->charmap[ uvc ];
2755 if (trie->widecharmap) {
2756 SV** svpp = (SV**)NULL;
2757 svpp = hv_fetch(trie->widecharmap,
2758 (char*)&uvc, sizeof(UV), 0);
2760 charid = (U16)SvIV(*svpp);
2765 (base + charid > trie->uniquecharcount )
2766 && (base + charid - 1 - trie->uniquecharcount
2768 && trie->trans[base + charid - 1 -
2769 trie->uniquecharcount].check == state)
2771 state = trie->trans[base + charid - 1 -
2772 trie->uniquecharcount ].next;
2783 DEBUG_TRIE_EXECUTE_r(
2784 PerlIO_printf( Perl_debug_log,
2785 "Charid:%3x CV:%4"UVxf" After State: %4"UVxf"%s\n",
2786 charid, uvc, (UV)state, PL_colors[5] );
2789 if (!st->u.trie.accepted )
2793 There was at least one accepting state that we
2794 transitioned through. Presumably the number of accepting
2795 states is going to be low, typically one or two. So we
2796 simply scan through to find the one with lowest wordnum.
2797 Once we find it, we swap the last state into its place
2798 and decrement the size. We then try to match the rest of
2799 the pattern at the point where the word ends, if we
2800 succeed then we end the loop, otherwise the loop
2801 eventually terminates once all of the accepting states
2805 if ( st->u.trie.accepted == 1 ) {
2807 SV ** const tmp = av_fetch( trie->words, st->u.trie.accept_buff[ 0 ].wordnum-1, 0 );
2808 PerlIO_printf( Perl_debug_log,
2809 "%*s %sonly one match : #%d <%s>%s\n",
2810 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],
2811 st->u.trie.accept_buff[ 0 ].wordnum,
2812 tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr",
2815 PL_reginput = (char *)st->u.trie.accept_buff[ 0 ].endpos;
2816 /* in this case we free tmps/leave before we call regmatch
2817 as we wont be using accept_buff again. */
2820 REGMATCH(scan + NEXT_OFF(scan), TRIE1);
2821 /*** all unsaved local vars undefined at this point */
2824 PerlIO_printf( Perl_debug_log,"%*s %sgot %"IVdf" possible matches%s\n",
2825 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], (IV)st->u.trie.accepted,
2828 while ( !result && st->u.trie.accepted-- ) {
2831 for( cur = 1 ; cur <= st->u.trie.accepted ; cur++ ) {
2832 DEBUG_TRIE_EXECUTE_r(
2833 PerlIO_printf( Perl_debug_log,
2834 "%*s %sgot %"IVdf" (%d) as best, looking at %"IVdf" (%d)%s\n",
2835 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
2836 (IV)best, st->u.trie.accept_buff[ best ].wordnum, (IV)cur,
2837 st->u.trie.accept_buff[ cur ].wordnum, PL_colors[5] );
2840 if (st->u.trie.accept_buff[cur].wordnum <
2841 st->u.trie.accept_buff[best].wordnum)
2845 reg_trie_data * const trie = (reg_trie_data*)
2846 PL_reg_re->data->data[ARG(scan)];
2847 SV ** const tmp = av_fetch( trie->words, st->u.trie.accept_buff[ best ].wordnum - 1, 0 );
2848 PerlIO_printf( Perl_debug_log, "%*s %strying alternation #%d <%s> at 0x%p%s\n",
2849 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],
2850 st->u.trie.accept_buff[best].wordnum,
2851 tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr",scan,
2854 if ( best<st->u.trie.accepted ) {
2855 reg_trie_accepted tmp = st->u.trie.accept_buff[ best ];
2856 st->u.trie.accept_buff[ best ] = st->u.trie.accept_buff[ st->u.trie.accepted ];
2857 st->u.trie.accept_buff[ st->u.trie.accepted ] = tmp;
2858 best = st->u.trie.accepted;
2860 PL_reginput = (char *)st->u.trie.accept_buff[ best ].endpos;
2863 as far as I can tell we only need the SAVETMPS/FREETMPS
2864 for re's with EVAL in them but I'm leaving them in for
2865 all until I can be sure.
2868 REGMATCH(scan + NEXT_OFF(scan), TRIE2);
2869 /*** all unsaved local vars undefined at this point */
2882 /* unreached codepoint */
2884 char *s = STRING(scan);
2885 st->ln = STR_LEN(scan);
2886 if (do_utf8 != UTF) {
2887 /* The target and the pattern have differing utf8ness. */
2889 const char *e = s + st->ln;
2892 /* The target is utf8, the pattern is not utf8. */
2897 if (NATIVE_TO_UNI(*(U8*)s) !=
2898 utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen,
2906 /* The target is not utf8, the pattern is utf8. */
2911 if (NATIVE_TO_UNI(*((U8*)l)) !=
2912 utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen,
2920 nextchr = UCHARAT(locinput);
2923 /* The target and the pattern have the same utf8ness. */
2924 /* Inline the first character, for speed. */
2925 if (UCHARAT(s) != nextchr)
2927 if (PL_regeol - locinput < st->ln)
2929 if (st->ln > 1 && memNE(s, locinput, st->ln))
2932 nextchr = UCHARAT(locinput);
2936 PL_reg_flags |= RF_tainted;
2939 char *s = STRING(scan);
2940 st->ln = STR_LEN(scan);
2942 if (do_utf8 || UTF) {
2943 /* Either target or the pattern are utf8. */
2945 char *e = PL_regeol;
2947 if (ibcmp_utf8(s, 0, st->ln, (bool)UTF,
2948 l, &e, 0, do_utf8)) {
2949 /* One more case for the sharp s:
2950 * pack("U0U*", 0xDF) =~ /ss/i,
2951 * the 0xC3 0x9F are the UTF-8
2952 * byte sequence for the U+00DF. */
2954 toLOWER(s[0]) == 's' &&
2956 toLOWER(s[1]) == 's' &&
2963 nextchr = UCHARAT(locinput);
2967 /* Neither the target and the pattern are utf8. */
2969 /* Inline the first character, for speed. */
2970 if (UCHARAT(s) != nextchr &&
2971 UCHARAT(s) != ((OP(scan) == EXACTF)
2972 ? PL_fold : PL_fold_locale)[nextchr])
2974 if (PL_regeol - locinput < st->ln)
2976 if (st->ln > 1 && (OP(scan) == EXACTF
2977 ? ibcmp(s, locinput, st->ln)
2978 : ibcmp_locale(s, locinput, st->ln)))
2981 nextchr = UCHARAT(locinput);
2986 STRLEN inclasslen = PL_regeol - locinput;
2988 if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8))
2990 if (locinput >= PL_regeol)
2992 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
2993 nextchr = UCHARAT(locinput);
2998 nextchr = UCHARAT(locinput);
2999 if (!REGINCLASS(scan, (U8*)locinput))
3001 if (!nextchr && locinput >= PL_regeol)
3003 nextchr = UCHARAT(++locinput);
3007 /* If we might have the case of the German sharp s
3008 * in a casefolding Unicode character class. */
3010 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
3011 locinput += SHARP_S_SKIP;
3012 nextchr = UCHARAT(locinput);
3018 PL_reg_flags |= RF_tainted;
3024 LOAD_UTF8_CHARCLASS_ALNUM();
3025 if (!(OP(scan) == ALNUM
3026 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
3027 : isALNUM_LC_utf8((U8*)locinput)))
3031 locinput += PL_utf8skip[nextchr];
3032 nextchr = UCHARAT(locinput);
3035 if (!(OP(scan) == ALNUM
3036 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
3038 nextchr = UCHARAT(++locinput);
3041 PL_reg_flags |= RF_tainted;
3044 if (!nextchr && locinput >= PL_regeol)
3047 LOAD_UTF8_CHARCLASS_ALNUM();
3048 if (OP(scan) == NALNUM
3049 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
3050 : isALNUM_LC_utf8((U8*)locinput))
3054 locinput += PL_utf8skip[nextchr];
3055 nextchr = UCHARAT(locinput);
3058 if (OP(scan) == NALNUM
3059 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
3061 nextchr = UCHARAT(++locinput);
3065 PL_reg_flags |= RF_tainted;
3069 /* was last char in word? */
3071 if (locinput == PL_bostr)
3074 const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
3076 st->ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags);
3078 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
3079 st->ln = isALNUM_uni(st->ln);
3080 LOAD_UTF8_CHARCLASS_ALNUM();
3081 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
3084 st->ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(st->ln));
3085 n = isALNUM_LC_utf8((U8*)locinput);
3089 st->ln = (locinput != PL_bostr) ?
3090 UCHARAT(locinput - 1) : '\n';
3091 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
3092 st->ln = isALNUM(st->ln);
3093 n = isALNUM(nextchr);
3096 st->ln = isALNUM_LC(st->ln);
3097 n = isALNUM_LC(nextchr);
3100 if (((!st->ln) == (!n)) == (OP(scan) == BOUND ||
3101 OP(scan) == BOUNDL))
3105 PL_reg_flags |= RF_tainted;
3111 if (UTF8_IS_CONTINUED(nextchr)) {
3112 LOAD_UTF8_CHARCLASS_SPACE();
3113 if (!(OP(scan) == SPACE
3114 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
3115 : isSPACE_LC_utf8((U8*)locinput)))
3119 locinput += PL_utf8skip[nextchr];
3120 nextchr = UCHARAT(locinput);
3123 if (!(OP(scan) == SPACE
3124 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3126 nextchr = UCHARAT(++locinput);
3129 if (!(OP(scan) == SPACE
3130 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3132 nextchr = UCHARAT(++locinput);
3136 PL_reg_flags |= RF_tainted;
3139 if (!nextchr && locinput >= PL_regeol)
3142 LOAD_UTF8_CHARCLASS_SPACE();
3143 if (OP(scan) == NSPACE
3144 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
3145 : isSPACE_LC_utf8((U8*)locinput))
3149 locinput += PL_utf8skip[nextchr];
3150 nextchr = UCHARAT(locinput);
3153 if (OP(scan) == NSPACE
3154 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
3156 nextchr = UCHARAT(++locinput);
3159 PL_reg_flags |= RF_tainted;
3165 LOAD_UTF8_CHARCLASS_DIGIT();
3166 if (!(OP(scan) == DIGIT
3167 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
3168 : isDIGIT_LC_utf8((U8*)locinput)))
3172 locinput += PL_utf8skip[nextchr];
3173 nextchr = UCHARAT(locinput);
3176 if (!(OP(scan) == DIGIT
3177 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
3179 nextchr = UCHARAT(++locinput);
3182 PL_reg_flags |= RF_tainted;
3185 if (!nextchr && locinput >= PL_regeol)
3188 LOAD_UTF8_CHARCLASS_DIGIT();
3189 if (OP(scan) == NDIGIT
3190 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
3191 : isDIGIT_LC_utf8((U8*)locinput))
3195 locinput += PL_utf8skip[nextchr];
3196 nextchr = UCHARAT(locinput);
3199 if (OP(scan) == NDIGIT
3200 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
3202 nextchr = UCHARAT(++locinput);
3205 if (locinput >= PL_regeol)
3208 LOAD_UTF8_CHARCLASS_MARK();
3209 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3211 locinput += PL_utf8skip[nextchr];
3212 while (locinput < PL_regeol &&
3213 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3214 locinput += UTF8SKIP(locinput);
3215 if (locinput > PL_regeol)
3220 nextchr = UCHARAT(locinput);
3223 PL_reg_flags |= RF_tainted;
3228 n = ARG(scan); /* which paren pair */
3229 st->ln = PL_regstartp[n];
3230 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3231 if ((I32)*PL_reglastparen < n || st->ln == -1)
3232 sayNO; /* Do not match unless seen CLOSEn. */
3233 if (st->ln == PL_regendp[n])
3236 s = PL_bostr + st->ln;
3237 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
3239 const char *e = PL_bostr + PL_regendp[n];
3241 * Note that we can't do the "other character" lookup trick as
3242 * in the 8-bit case (no pun intended) because in Unicode we
3243 * have to map both upper and title case to lower case.
3245 if (OP(scan) == REFF) {
3247 STRLEN ulen1, ulen2;
3248 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
3249 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
3253 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
3254 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
3255 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
3262 nextchr = UCHARAT(locinput);
3266 /* Inline the first character, for speed. */
3267 if (UCHARAT(s) != nextchr &&
3269 (UCHARAT(s) != ((OP(scan) == REFF
3270 ? PL_fold : PL_fold_locale)[nextchr]))))
3272 st->ln = PL_regendp[n] - st->ln;
3273 if (locinput + st->ln > PL_regeol)
3275 if (st->ln > 1 && (OP(scan) == REF
3276 ? memNE(s, locinput, st->ln)
3278 ? ibcmp(s, locinput, st->ln)
3279 : ibcmp_locale(s, locinput, st->ln))))
3282 nextchr = UCHARAT(locinput);
3295 /* execute the code in the {...} */
3297 SV ** const before = SP;
3298 OP_4tree * const oop = PL_op;
3299 COP * const ocurcop = PL_curcop;
3301 struct regexp * const oreg = PL_reg_re;
3304 PL_op = (OP_4tree*)PL_reg_re->data->data[n];
3305 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
3306 PAD_SAVE_LOCAL(old_comppad, (PAD*)PL_reg_re->data->data[n + 2]);
3307 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
3309 CALLRUNOPS(aTHX); /* Scalar context. */
3312 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
3319 PAD_RESTORE_LOCAL(old_comppad);
3320 PL_curcop = ocurcop;
3324 sv_setsv(save_scalar(PL_replgv), ret);
3328 if (st->logical == 2) { /* Postponed subexpression: /(??{...})/ */
3331 /* extract RE object from returned value; compiling if
3336 if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret)))
3337 mg = mg_find(sv, PERL_MAGIC_qr);
3338 else if (SvSMAGICAL(ret)) {
3339 if (SvGMAGICAL(ret))
3340 sv_unmagic(ret, PERL_MAGIC_qr);
3342 mg = mg_find(ret, PERL_MAGIC_qr);
3346 re = (regexp *)mg->mg_obj;
3347 (void)ReREFCNT_inc(re);
3351 const char * const t = SvPV_const(ret, len);
3353 const I32 osize = PL_regsize;
3356 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
3357 re = CALLREGCOMP(aTHX_ (char*)t, (char*)t + len, &pm);
3359 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
3361 sv_magic(ret,(SV*)ReREFCNT_inc(re),
3367 /* run the pattern returned from (??{...}) */
3370 PerlIO_printf(Perl_debug_log,
3371 "Entering embedded \"%s%.60s%s%s\"\n",
3375 (strlen(re->precomp) > 60 ? "..." : ""))
3378 st->u.eval.cp = regcppush(0); /* Save *all* the positions. */
3379 REGCP_SET(st->u.eval.lastcp);
3380 *PL_reglastparen = 0;
3381 *PL_reglastcloseparen = 0;
3382 PL_reginput = locinput;
3384 /* XXXX This is too dramatic a measure... */
3388 st->u.eval.toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
3389 ((re->reganch & ROPT_UTF8) != 0);
3390 if (st->u.eval.toggleutf) PL_reg_flags ^= RF_utf8;
3391 st->u.eval.prev_rex = rex;
3392 assert(rex == PL_reg_re); /* XXX */
3394 PL_reg_re = rex; /* XXX */
3396 st->u.eval.prev_eval = cur_eval;
3397 st->u.eval.prev_slab = PL_regmatch_slab;
3398 st->u.eval.depth = depth;
3400 PUSH_STATE(newst, resume_EVAL);
3403 /* now continue from first node in postoned RE */
3404 next = re->program + 1;
3408 /* /(?(?{...})X|Y)/ */
3409 st->sw = SvTRUE(ret);
3414 n = ARG(scan); /* which paren pair */
3415 PL_reg_start_tmp[n] = locinput;
3420 n = ARG(scan); /* which paren pair */
3421 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
3422 PL_regendp[n] = locinput - PL_bostr;
3423 if (n > (I32)*PL_reglastparen)
3424 *PL_reglastparen = n;
3425 *PL_reglastcloseparen = n;
3428 n = ARG(scan); /* which paren pair */
3429 st->sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
3432 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3434 next = NEXTOPER(NEXTOPER(scan));
3436 next = scan + ARG(scan);
3437 if (OP(next) == IFTHEN) /* Fake one. */
3438 next = NEXTOPER(NEXTOPER(next));
3442 st->logical = scan->flags;
3444 /*******************************************************************
3445 cc points to the regmatch_state associated with the most recent CURLYX.
3446 This struct contains info about the innermost (...)* loop (an
3447 "infoblock"), and a pointer to the next outer cc.
3449 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
3451 1) After matching Y, regnode for CURLYX is processed;
3453 2) This regnode populates cc, and calls regmatch() recursively
3454 with the starting point at WHILEM node;
3456 3) Each hit of WHILEM node tries to match A and Z (in the order
3457 depending on the current iteration, min/max of {min,max} and
3458 greediness). The information about where are nodes for "A"
3459 and "Z" is read from cc, as is info on how many times "A"
3460 was already matched, and greediness.
3462 4) After A matches, the same WHILEM node is hit again.
3464 5) Each time WHILEM is hit, cc is the infoblock created by CURLYX
3465 of the same pair. Thus when WHILEM tries to match Z, it temporarily
3466 resets cc, since this Y(A)*Z can be a part of some other loop:
3467 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
3468 of the external loop.
3470 Currently present infoblocks form a tree with a stem formed by st->cc
3471 and whatever it mentions via ->next, and additional attached trees
3472 corresponding to temporarily unset infoblocks as in "5" above.
3474 In the following picture, infoblocks for outer loop of
3475 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
3476 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
3477 infoblocks are drawn below the "reset" infoblock.
3479 In fact in the picture below we do not show failed matches for Z and T
3480 by WHILEM blocks. [We illustrate minimal matches, since for them it is
3481 more obvious *why* one needs to *temporary* unset infoblocks.]
3483 Matched REx position InfoBlocks Comment
3487 Y A)*?Z)*?T x <- O <- I
3488 YA )*?Z)*?T x <- O <- I
3489 YA A)*?Z)*?T x <- O <- I
3490 YAA )*?Z)*?T x <- O <- I
3491 YAA Z)*?T x <- O # Temporary unset I
3494 YAAZ Y(A)*?Z)*?T x <- O
3497 YAAZY (A)*?Z)*?T x <- O
3500 YAAZY A)*?Z)*?T x <- O <- I
3503 YAAZYA )*?Z)*?T x <- O <- I
3506 YAAZYA Z)*?T x <- O # Temporary unset I
3512 YAAZYAZ T x # Temporary unset O
3519 *******************************************************************/
3522 /* No need to save/restore up to this paren */
3523 I32 parenfloor = scan->flags;
3525 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3527 /* XXXX Probably it is better to teach regpush to support
3528 parenfloor > PL_regsize... */
3529 if (parenfloor > (I32)*PL_reglastparen)
3530 parenfloor = *PL_reglastparen; /* Pessimization... */
3532 st->u.curlyx.cp = PL_savestack_ix;
3533 st->u.curlyx.outercc = st->cc;
3535 /* these fields contain the state of the current curly.
3536 * they are accessed by subsequent WHILEMs;
3537 * cur and lastloc are also updated by WHILEM */
3538 st->u.curlyx.parenfloor = parenfloor;
3539 st->u.curlyx.cur = -1; /* this will be updated by WHILEM */
3540 st->u.curlyx.min = ARG1(scan);
3541 st->u.curlyx.max = ARG2(scan);
3542 st->u.curlyx.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3543 st->u.curlyx.lastloc = 0;
3544 /* st->next and st->minmod are also read by WHILEM */
3546 PL_reginput = locinput;
3547 REGMATCH(PREVOPER(next), CURLYX); /* start on the WHILEM */
3548 /*** all unsaved local vars undefined at this point */
3549 regcpblow(st->u.curlyx.cp);
3550 st->cc = st->u.curlyx.outercc;
3556 * This is really hard to understand, because after we match
3557 * what we're trying to match, we must make sure the rest of
3558 * the REx is going to match for sure, and to do that we have
3559 * to go back UP the parse tree by recursing ever deeper. And
3560 * if it fails, we have to reset our parent's current state
3561 * that we can try again after backing off.
3564 st->u.whilem.lastloc = st->cc->u.curlyx.lastloc; /* Detection of 0-len. */
3565 st->u.whilem.cache_offset = 0;
3566 st->u.whilem.cache_bit = 0;
3568 n = st->cc->u.curlyx.cur + 1; /* how many we know we matched */
3569 PL_reginput = locinput;
3572 PerlIO_printf(Perl_debug_log,
3573 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3574 REPORT_CODE_OFF+PL_regindent*2, "",
3575 (long)n, (long)st->cc->u.curlyx.min,
3576 (long)st->cc->u.curlyx.max, PTR2UV(st->cc))
3579 /* If degenerate scan matches "", assume scan done. */
3581 if (locinput == st->cc->u.curlyx.lastloc && n >= st->cc->u.curlyx.min) {
3582 st->u.whilem.savecc = st->cc;
3583 st->cc = st->cc->u.curlyx.outercc;
3585 st->ln = st->cc->u.curlyx.cur;
3587 PerlIO_printf(Perl_debug_log,
3588 "%*s empty match detected, try continuation...\n",
3589 REPORT_CODE_OFF+PL_regindent*2, "")
3591 REGMATCH(st->u.whilem.savecc->next, WHILEM1);
3592 /*** all unsaved local vars undefined at this point */
3593 st->cc = st->u.whilem.savecc;
3596 if (st->cc->u.curlyx.outercc)
3597 st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln;
3601 /* First just match a string of min scans. */
3603 if (n < st->cc->u.curlyx.min) {
3604 st->cc->u.curlyx.cur = n;
3605 st->cc->u.curlyx.lastloc = locinput;
3606 REGMATCH(st->cc->u.curlyx.scan, WHILEM2);
3607 /*** all unsaved local vars undefined at this point */
3610 st->cc->u.curlyx.cur = n - 1;
3611 st->cc->u.curlyx.lastloc = st->u.whilem.lastloc;
3616 /* Check whether we already were at this position.
3617 Postpone detection until we know the match is not
3618 *that* much linear. */
3619 if (!PL_reg_maxiter) {
3620 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3621 PL_reg_leftiter = PL_reg_maxiter;
3623 if (PL_reg_leftiter-- == 0) {
3624 const I32 size = (PL_reg_maxiter + 7 + POSCACHE_START)/8;
3625 if (PL_reg_poscache) {
3626 if ((I32)PL_reg_poscache_size < size) {
3627 Renew(PL_reg_poscache, size, char);
3628 PL_reg_poscache_size = size;
3630 Zero(PL_reg_poscache, size, char);
3633 PL_reg_poscache_size = size;
3634 Newxz(PL_reg_poscache, size, char);
3637 PerlIO_printf(Perl_debug_log,
3638 "%sDetected a super-linear match, switching on caching%s...\n",
3639 PL_colors[4], PL_colors[5])
3642 if (PL_reg_leftiter < 0) {
3643 st->u.whilem.cache_offset = locinput - PL_bostr;
3645 st->u.whilem.cache_offset = (scan->flags & 0xf) - 1 + POSCACHE_START
3646 + st->u.whilem.cache_offset * (scan->flags>>4);
3647 st->u.whilem.cache_bit = st->u.whilem.cache_offset % 8;
3648 st->u.whilem.cache_offset /= 8;
3649 if (PL_reg_poscache[st->u.whilem.cache_offset] & (1<<st->u.whilem.cache_bit)) {
3651 PerlIO_printf(Perl_debug_log,
3652 "%*s already tried at this position...\n",
3653 REPORT_CODE_OFF+PL_regindent*2, "")
3655 if (PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))
3656 /* cache records success */
3659 /* cache records failure */
3662 PL_reg_poscache[st->u.whilem.cache_offset] |= (1<<st->u.whilem.cache_bit);
3666 /* Prefer next over scan for minimal matching. */
3668 if (st->cc->minmod) {
3669 st->u.whilem.savecc = st->cc;
3670 st->cc = st->cc->u.curlyx.outercc;
3672 st->ln = st->cc->u.curlyx.cur;
3673 st->u.whilem.cp = regcppush(st->u.whilem.savecc->u.curlyx.parenfloor);
3674 REGCP_SET(st->u.whilem.lastcp);
3675 REGMATCH(st->u.whilem.savecc->next, WHILEM3);
3676 /*** all unsaved local vars undefined at this point */
3677 st->cc = st->u.whilem.savecc;
3679 regcpblow(st->u.whilem.cp);
3680 CACHEsayYES; /* All done. */
3682 REGCP_UNWIND(st->u.whilem.lastcp);
3684 if (st->cc->u.curlyx.outercc)
3685 st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln;
3687 if (n >= st->cc->u.curlyx.max) { /* Maximum greed exceeded? */
3688 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3689 && !(PL_reg_flags & RF_warned)) {
3690 PL_reg_flags |= RF_warned;
3691 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3692 "Complex regular subexpression recursion",
3699 PerlIO_printf(Perl_debug_log,
3700 "%*s trying longer...\n",
3701 REPORT_CODE_OFF+PL_regindent*2, "")
3703 /* Try scanning more and see if it helps. */
3704 PL_reginput = locinput;
3705 st->cc->u.curlyx.cur = n;
3706 st->cc->u.curlyx.lastloc = locinput;
3707 st->u.whilem.cp = regcppush(st->cc->u.curlyx.parenfloor);
3708 REGCP_SET(st->u.whilem.lastcp);
3709 REGMATCH(st->cc->u.curlyx.scan, WHILEM4);
3710 /*** all unsaved local vars undefined at this point */
3712 regcpblow(st->u.whilem.cp);
3715 REGCP_UNWIND(st->u.whilem.lastcp);
3717 st->cc->u.curlyx.cur = n - 1;
3718 st->cc->u.curlyx.lastloc = st->u.whilem.lastloc;
3722 /* Prefer scan over next for maximal matching. */
3724 if (n < st->cc->u.curlyx.max) { /* More greed allowed? */
3725 st->u.whilem.cp = regcppush(st->cc->u.curlyx.parenfloor);
3726 st->cc->u.curlyx.cur = n;
3727 st->cc->u.curlyx.lastloc = locinput;
3728 REGCP_SET(st->u.whilem.lastcp);
3729 REGMATCH(st->cc->u.curlyx.scan, WHILEM5);
3730 /*** all unsaved local vars undefined at this point */
3732 regcpblow(st->u.whilem.cp);
3735 REGCP_UNWIND(st->u.whilem.lastcp);
3736 regcppop(rex); /* Restore some previous $<digit>s? */
3737 PL_reginput = locinput;
3739 PerlIO_printf(Perl_debug_log,
3740 "%*s failed, try continuation...\n",
3741 REPORT_CODE_OFF+PL_regindent*2, "")
3744 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3745 && !(PL_reg_flags & RF_warned)) {
3746 PL_reg_flags |= RF_warned;
3747 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3748 "Complex regular subexpression recursion",
3752 /* Failed deeper matches of scan, so see if this one works. */
3753 st->u.whilem.savecc = st->cc;
3754 st->cc = st->cc->u.curlyx.outercc;
3756 st->ln = st->cc->u.curlyx.cur;
3757 REGMATCH(st->u.whilem.savecc->next, WHILEM6);
3758 /*** all unsaved local vars undefined at this point */
3759 st->cc = st->u.whilem.savecc;
3762 if (st->cc->u.curlyx.outercc)
3763 st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln;
3764 st->cc->u.curlyx.cur = n - 1;
3765 st->cc->u.curlyx.lastloc = st->u.whilem.lastloc;
3770 next = scan + ARG(scan);
3773 inner = NEXTOPER(NEXTOPER(scan));
3776 inner = NEXTOPER(scan);
3781 if (OP(next) != type) /* No choice. */
3782 next = inner; /* Avoid recursion. */
3784 const I32 lastparen = *PL_reglastparen;
3785 /* Put unwinding data on stack */
3786 const I32 unwind1 = SSNEWt(1,re_unwind_branch_t);
3787 re_unwind_branch_t * const uw = SSPTRt(unwind1,re_unwind_branch_t);
3789 uw->prev = st->unwind;
3790 st->unwind = unwind1;
3791 uw->type = ((type == BRANCH)
3793 : RE_UNWIND_BRANCHJ);
3794 uw->lastparen = lastparen;
3796 uw->locinput = locinput;
3797 uw->nextchr = nextchr;
3799 uw->regindent = ++PL_regindent;
3802 REGCP_SET(uw->lastcp);
3804 /* Now go into the first branch */
3814 st->u.curlym.l = st->u.curlym.matches = 0;
3816 /* We suppose that the next guy does not need
3817 backtracking: in particular, it is of constant non-zero length,
3818 and has no parenths to influence future backrefs. */
3819 st->ln = ARG1(scan); /* min to match */
3820 n = ARG2(scan); /* max to match */
3821 st->u.curlym.paren = scan->flags;
3822 if (st->u.curlym.paren) {
3823 if (st->u.curlym.paren > PL_regsize)
3824 PL_regsize = st->u.curlym.paren;
3825 if (st->u.curlym.paren > (I32)*PL_reglastparen)
3826 *PL_reglastparen = st->u.curlym.paren;
3828 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3829 if (st->u.curlym.paren)
3830 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3831 PL_reginput = locinput;
3832 st->u.curlym.maxwanted = st->minmod ? st->ln : n;
3833 if (st->u.curlym.maxwanted) {
3834 while (PL_reginput < PL_regeol && st->u.curlym.matches < st->u.curlym.maxwanted) {
3835 REGMATCH(scan, CURLYM1);
3836 /*** all unsaved local vars undefined at this point */
3839 /* on first match, determine length, u.curlym.l */
3840 if (!st->u.curlym.matches++) {
3841 if (PL_reg_match_utf8) {
3843 while (s < PL_reginput) {
3849 st->u.curlym.l = PL_reginput - locinput;
3851 if (st->u.curlym.l == 0) {
3852 st->u.curlym.matches = st->u.curlym.maxwanted;
3856 locinput = PL_reginput;
3860 PL_reginput = locinput;
3864 if (st->ln && st->u.curlym.matches < st->ln)
3866 if (HAS_TEXT(next) || JUMPABLE(next)) {
3867 regnode *text_node = next;
3869 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3871 if (! HAS_TEXT(text_node)) st->u.curlym.c1 = st->u.curlym.c2 = -1000;
3873 if (PL_regkind[(U8)OP(text_node)] == REF) {
3874 st->u.curlym.c1 = st->u.curlym.c2 = -1000;
3877 else { st->u.curlym.c1 = (U8)*STRING(text_node); }
3878 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3879 st->u.curlym.c2 = PL_fold[st->u.curlym.c1];
3880 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3881 st->u.curlym.c2 = PL_fold_locale[st->u.curlym.c1];
3883 st->u.curlym.c2 = st->u.curlym.c1;
3887 st->u.curlym.c1 = st->u.curlym.c2 = -1000;
3889 REGCP_SET(st->u.curlym.lastcp);
3890 while (n >= st->ln || (n == REG_INFTY && st->ln > 0)) { /* ln overflow ? */
3891 /* If it could work, try it. */
3892 if (st->u.curlym.c1 == -1000 ||
3893 UCHARAT(PL_reginput) == st->u.curlym.c1 ||
3894 UCHARAT(PL_reginput) == st->u.curlym.c2)
3896 if (st->u.curlym.paren) {
3898 PL_regstartp[st->u.curlym.paren] =
3899 HOPc(PL_reginput, -st->u.curlym.l) - PL_bostr;
3900 PL_regendp[st->u.curlym.paren] = PL_reginput - PL_bostr;
3903 PL_regendp[st->u.curlym.paren] = -1;
3905 REGMATCH(next, CURLYM2);
3906 /*** all unsaved local vars undefined at this point */
3909 REGCP_UNWIND(st->u.curlym.lastcp);
3911 /* Couldn't or didn't -- move forward. */
3912 PL_reginput = locinput;
3913 REGMATCH(scan, CURLYM3);
3914 /*** all unsaved local vars undefined at this point */
3917 locinput = PL_reginput;
3925 PerlIO_printf(Perl_debug_log,
3926 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3927 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3928 (IV) st->u.curlym.matches, (IV)st->u.curlym.l)
3930 if (st->u.curlym.matches >= st->ln) {
3931 if (HAS_TEXT(next) || JUMPABLE(next)) {
3932 regnode *text_node = next;
3934 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3936 if (! HAS_TEXT(text_node)) st->u.curlym.c1 = st->u.curlym.c2 = -1000;
3938 if (PL_regkind[(U8)OP(text_node)] == REF) {
3939 st->u.curlym.c1 = st->u.curlym.c2 = -1000;
3942 else { st->u.curlym.c1 = (U8)*STRING(text_node); }
3944 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3945 st->u.curlym.c2 = PL_fold[st->u.curlym.c1];
3946 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3947 st->u.curlym.c2 = PL_fold_locale[st->u.curlym.c1];
3949 st->u.curlym.c2 = st->u.curlym.c1;
3953 st->u.curlym.c1 = st->u.curlym.c2 = -1000;
3956 REGCP_SET(st->u.curlym.lastcp);
3957 while (st->u.curlym.matches >= st->ln) {
3958 /* If it could work, try it. */
3959 if (st->u.curlym.c1 == -1000 ||
3960 UCHARAT(PL_reginput) == st->u.curlym.c1 ||
3961 UCHARAT(PL_reginput) == st->u.curlym.c2)
3964 PerlIO_printf(Perl_debug_log,
3965 "%*s trying tail with matches=%"IVdf"...\n",
3966 (int)(REPORT_CODE_OFF+PL_regindent*2),
3967 "", (IV)st->u.curlym.matches)
3969 if (st->u.curlym.paren) {
3970 if (st->u.curlym.matches) {
3971 PL_regstartp[st->u.curlym.paren]
3972 = HOPc(PL_reginput, -st->u.curlym.l) - PL_bostr;
3973 PL_regendp[st->u.curlym.paren] = PL_reginput - PL_bostr;
3976 PL_regendp[st->u.curlym.paren] = -1;
3978 REGMATCH(next, CURLYM4);
3979 /*** all unsaved local vars undefined at this point */
3982 REGCP_UNWIND(st->u.curlym.lastcp);
3984 /* Couldn't or didn't -- back up. */
3985 st->u.curlym.matches--;
3986 locinput = HOPc(locinput, -st->u.curlym.l);
3987 PL_reginput = locinput;
3995 st->u.plus.paren = scan->flags; /* Which paren to set */
3996 if (st->u.plus.paren > PL_regsize)
3997 PL_regsize = st->u.plus.paren;
3998 if (st->u.plus.paren > (I32)*PL_reglastparen)
3999 *PL_reglastparen = st->u.plus.paren;
4000 st->ln = ARG1(scan); /* min to match */
4001 n = ARG2(scan); /* max to match */
4002 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
4005 st->u.plus.paren = 0;
4006 st->ln = ARG1(scan); /* min to match */
4007 n = ARG2(scan); /* max to match */
4008 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
4013 scan = NEXTOPER(scan);
4014 st->u.plus.paren = 0;
4019 scan = NEXTOPER(scan);
4020 st->u.plus.paren = 0;
4023 * Lookahead to avoid useless match attempts
4024 * when we know what character comes next.
4028 * Used to only do .*x and .*?x, but now it allows
4029 * for )'s, ('s and (?{ ... })'s to be in the way
4030 * of the quantifier and the EXACT-like node. -- japhy
4033 if (HAS_TEXT(next) || JUMPABLE(next)) {
4035 regnode *text_node = next;
4037 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
4039 if (! HAS_TEXT(text_node)) st->u.plus.c1 = st->u.plus.c2 = -1000;
4041 if (PL_regkind[(U8)OP(text_node)] == REF) {
4042 st->u.plus.c1 = st->u.plus.c2 = -1000;
4043 goto assume_ok_easy;
4045 else { s = (U8*)STRING(text_node); }
4048 st->u.plus.c2 = st->u.plus.c1 = *s;
4049 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
4050 st->u.plus.c2 = PL_fold[st->u.plus.c1];
4051 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
4052 st->u.plus.c2 = PL_fold_locale[st->u.plus.c1];
4055 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
4056 STRLEN ulen1, ulen2;
4057 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
4058 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
4060 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
4061 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
4063 st->u.plus.c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0,
4065 st->u.plus.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0,
4069 st->u.plus.c2 = st->u.plus.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0,
4076 st->u.plus.c1 = st->u.plus.c2 = -1000;
4078 PL_reginput = locinput;
4081 if (st->ln && regrepeat(scan, st->ln) < st->ln)
4083 locinput = PL_reginput;
4084 REGCP_SET(st->u.plus.lastcp);
4085 if (st->u.plus.c1 != -1000) {
4086 st->u.plus.old = locinput;
4087 st->u.plus.count = 0;
4089 if (n == REG_INFTY) {
4090 st->u.plus.e = PL_regeol - 1;
4092 while (UTF8_IS_CONTINUATION(*(U8*)st->u.plus.e))
4097 for (st->u.plus.e = locinput;
4098 m >0 && st->u.plus.e + UTF8SKIP(st->u.plus.e) <= PL_regeol; m--)
4099 st->u.plus.e += UTF8SKIP(st->u.plus.e);
4102 st->u.plus.e = locinput + n - st->ln;
4103 if (st->u.plus.e >= PL_regeol)
4104 st->u.plus.e = PL_regeol - 1;
4107 /* Find place 'next' could work */
4109 if (st->u.plus.c1 == st->u.plus.c2) {
4110 while (locinput <= st->u.plus.e &&
4111 UCHARAT(locinput) != st->u.plus.c1)
4114 while (locinput <= st->u.plus.e
4115 && UCHARAT(locinput) != st->u.plus.c1
4116 && UCHARAT(locinput) != st->u.plus.c2)
4119 st->u.plus.count = locinput - st->u.plus.old;
4122 if (st->u.plus.c1 == st->u.plus.c2) {
4124 /* count initialised to
4125 * utf8_distance(old, locinput) */
4126 while (locinput <= st->u.plus.e &&
4127 utf8n_to_uvchr((U8*)locinput,
4128 UTF8_MAXBYTES, &len,
4129 uniflags) != (UV)st->u.plus.c1) {
4135 /* count initialised to
4136 * utf8_distance(old, locinput) */
4137 while (locinput <= st->u.plus.e) {
4138 UV c = utf8n_to_uvchr((U8*)locinput,
4139 UTF8_MAXBYTES, &len,
4141 if (c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2)
4148 if (locinput > st->u.plus.e)
4150 /* PL_reginput == old now */
4151 if (locinput != st->u.plus.old) {
4152 st->ln = 1; /* Did some */
4153 if (regrepeat(scan, st->u.plus.count) < st->u.plus.count)
4156 /* PL_reginput == locinput now */
4157 TRYPAREN(st->u.plus.paren, st->ln, locinput, PLUS1);
4158 /*** all unsaved local vars undefined at this point */
4159 PL_reginput = locinput; /* Could be reset... */
4160 REGCP_UNWIND(st->u.plus.lastcp);
4161 /* Couldn't or didn't -- move forward. */
4162 st->u.plus.old = locinput;
4164 locinput += UTF8SKIP(locinput);
4167 st->u.plus.count = 1;
4171 while (n >= st->ln || (n == REG_INFTY && st->ln > 0)) { /* ln overflow ? */
4173 if (st->u.plus.c1 != -1000) {
4175 c = utf8n_to_uvchr((U8*)PL_reginput,
4179 c = UCHARAT(PL_reginput);
4180 /* If it could work, try it. */
4181 if (c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2)
4183 TRYPAREN(st->u.plus.paren, st->ln, PL_reginput, PLUS2);
4184 /*** all unsaved local vars undefined at this point */
4185 REGCP_UNWIND(st->u.plus.lastcp);
4188 /* If it could work, try it. */
4189 else if (st->u.plus.c1 == -1000)
4191 TRYPAREN(st->u.plus.paren, st->ln, PL_reginput, PLUS3);
4192 /*** all unsaved local vars undefined at this point */
4193 REGCP_UNWIND(st->u.plus.lastcp);
4195 /* Couldn't or didn't -- move forward. */
4196 PL_reginput = locinput;
4197 if (regrepeat(scan, 1)) {
4199 locinput = PL_reginput;
4206 n = regrepeat(scan, n);
4207 locinput = PL_reginput;
4208 if (st->ln < n && PL_regkind[(U8)OP(next)] == EOL &&
4209 (OP(next) != MEOL ||
4210 OP(next) == SEOL || OP(next) == EOS))
4212 st->ln = n; /* why back off? */
4213 /* ...because $ and \Z can match before *and* after
4214 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
4215 We should back off by one in this case. */
4216 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
4219 REGCP_SET(st->u.plus.lastcp);
4222 while (n >= st->ln) {
4223 if (st->u.plus.c1 != -1000) {
4225 c = utf8n_to_uvchr((U8*)PL_reginput,
4229 c = UCHARAT(PL_reginput);
4231 /* If it could work, try it. */
4232 if (st->u.plus.c1 == -1000 || c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2)
4234 TRYPAREN(st->u.plus.paren, n, PL_reginput, PLUS4);
4235 /*** all unsaved local vars undefined at this point */
4236 REGCP_UNWIND(st->u.plus.lastcp);
4238 /* Couldn't or didn't -- back up. */
4240 PL_reginput = locinput = HOPc(locinput, -1);
4248 /* we have successfully completed the execution of a
4249 * postponed re. Pop all states back to the last EVAL
4250 * then continue with the node following the (??{...})
4253 /* this simulates a POP_STATE, except that it pops several
4254 * levels, and doesn't restore locinput */
4257 PL_regmatch_slab = st->u.eval.prev_slab;
4258 cur_eval = st->u.eval.prev_eval;
4259 depth = st->u.eval.depth;
4261 PL_regmatch_state = st;
4266 if (st->u.eval.toggleutf)
4267 PL_reg_flags ^= RF_utf8;
4269 rex = st->u.eval.prev_rex;
4270 PL_reg_re = rex; /* XXX */
4271 /* XXXX This is too dramatic a measure... */
4274 /* Restore parens of the caller without popping the
4277 I32 tmp = PL_savestack_ix;
4278 PL_savestack_ix = st->u.eval.lastcp;
4280 PL_savestack_ix = tmp;
4284 PL_reginput = locinput;
4285 /* resume at node following the (??{...}) */
4290 if (locinput < PL_regtill) {
4291 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
4292 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
4294 (long)(locinput - PL_reg_starttry),
4295 (long)(PL_regtill - PL_reg_starttry),
4297 sayNO_FINAL; /* Cannot match: too short. */
4299 PL_reginput = locinput; /* put where regtry can find it */
4300 sayYES_FINAL; /* Success! */
4302 PL_reginput = locinput; /* put where regtry can find it */
4303 sayYES_LOUD; /* Success! */
4306 PL_reginput = locinput;
4311 char * const s = HOPBACKc(locinput, scan->flags);
4317 PL_reginput = locinput;
4322 char * const s = HOPBACKc(locinput, scan->flags);
4328 PL_reginput = locinput;
4331 REGMATCH(NEXTOPER(NEXTOPER(scan)), IFMATCH);
4332 /*** all unsaved local vars undefined at this point */
4348 if (OP(scan) == SUSPEND) {
4349 locinput = PL_reginput;
4350 nextchr = UCHARAT(locinput);
4355 next = scan + ARG(scan);
4360 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
4361 PTR2UV(scan), OP(scan));
4362 Perl_croak(aTHX_ "regexp memory corruption");
4370 /* simulate recursively calling regmatch(), but without actually
4371 * recursing - ie save the current state on the heap rather than on
4372 * the stack, then re-enter the loop. This avoids complex regexes
4373 * blowing the processor stack */
4377 /* push new state */
4378 regmatch_state *oldst = st;
4382 /* grab the next free state slot */
4384 if (st > &(PL_regmatch_slab->states[PERL_REGMATCH_SLAB_SLOTS-1]))
4385 st = S_push_slab(aTHX);
4386 PL_regmatch_state = st;
4390 oldst->locinput = locinput;
4393 locinput = PL_reginput;
4394 nextchr = UCHARAT(locinput);
4408 * We get here only if there's trouble -- normally "case END" is
4409 * the terminating point.
4411 Perl_croak(aTHX_ "corrupted regexp pointers");
4417 PerlIO_printf(Perl_debug_log,
4418 "%*s %scould match...%s\n",
4419 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5])
4423 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
4424 PL_colors[4], PL_colors[5]));
4431 /* XXX this is duplicate(ish) code to that in the do_no section.
4432 * eventually a yes should just pop the whole stack */
4434 /* restore previous state and re-enter */
4437 switch (st->resume_state) {
4439 goto resume_point_TRIE1;
4441 goto resume_point_TRIE2;
4445 goto resume_point_CURLYX;
4446 case resume_WHILEM1:
4447 goto resume_point_WHILEM1;
4448 case resume_WHILEM2:
4449 goto resume_point_WHILEM2;
4450 case resume_WHILEM3:
4451 goto resume_point_WHILEM3;
4452 case resume_WHILEM4:
4453 goto resume_point_WHILEM4;
4454 case resume_WHILEM5:
4455 goto resume_point_WHILEM5;
4456 case resume_WHILEM6:
4457 goto resume_point_WHILEM6;
4458 case resume_CURLYM1:
4459 goto resume_point_CURLYM1;
4460 case resume_CURLYM2:
4461 goto resume_point_CURLYM2;
4462 case resume_CURLYM3:
4463 goto resume_point_CURLYM3;
4464 case resume_CURLYM4:
4465 goto resume_point_CURLYM4;
4466 case resume_IFMATCH:
4467 goto resume_point_IFMATCH;
4469 goto resume_point_PLUS1;
4471 goto resume_point_PLUS2;
4473 goto resume_point_PLUS3;
4475 goto resume_point_PLUS4;
4477 Perl_croak(aTHX_ "regexp resume memory corruption");
4484 PerlIO_printf(Perl_debug_log,
4485 "%*s %sfailed...%s\n",
4486 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5])
4492 re_unwind_t * const uw = SSPTRt(st->unwind,re_unwind_t);
4495 case RE_UNWIND_BRANCH:
4496 case RE_UNWIND_BRANCHJ:
4498 re_unwind_branch_t * const uwb = &(uw->branch);
4499 const I32 lastparen = uwb->lastparen;
4501 REGCP_UNWIND(uwb->lastcp);
4502 for (n = *PL_reglastparen; n > lastparen; n--)
4504 *PL_reglastparen = n;
4505 scan = next = uwb->next;
4507 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
4508 ? BRANCH : BRANCHJ) ) { /* Failure */
4509 st->unwind = uwb->prev;
4515 /* Have more choice yet. Reuse the same uwb. */
4516 if ((n = (uwb->type == RE_UNWIND_BRANCH
4517 ? NEXT_OFF(next) : ARG(next))))
4520 next = NULL; /* XXXX Needn't unwinding in this case... */
4522 next = NEXTOPER(scan);
4523 if (uwb->type == RE_UNWIND_BRANCHJ)
4524 next = NEXTOPER(next);
4525 locinput = uwb->locinput;
4526 nextchr = uwb->nextchr;
4528 PL_regindent = uwb->regindent;
4535 Perl_croak(aTHX_ "regexp unwind memory corruption");
4546 /* there's a previous state to backtrack to */
4548 switch (st->resume_state) {
4550 goto resume_point_TRIE1;
4552 goto resume_point_TRIE2;
4554 /* we have failed an (??{...}). Restore state to the outer re
4555 * then re-throw the failure */
4556 if (st->u.eval.toggleutf)
4557 PL_reg_flags ^= RF_utf8;
4559 rex = st->u.eval.prev_rex;
4560 PL_reg_re = rex; /* XXX */
4561 cur_eval = st->u.eval.prev_eval;
4563 /* XXXX This is too dramatic a measure... */
4566 PL_reginput = locinput;
4567 REGCP_UNWIND(st->u.eval.lastcp);
4572 goto resume_point_CURLYX;
4573 case resume_WHILEM1:
4574 goto resume_point_WHILEM1;
4575 case resume_WHILEM2:
4576 goto resume_point_WHILEM2;
4577 case resume_WHILEM3:
4578 goto resume_point_WHILEM3;
4579 case resume_WHILEM4:
4580 goto resume_point_WHILEM4;
4581 case resume_WHILEM5:
4582 goto resume_point_WHILEM5;
4583 case resume_WHILEM6:
4584 goto resume_point_WHILEM6;
4585 case resume_CURLYM1:
4586 goto resume_point_CURLYM1;
4587 case resume_CURLYM2:
4588 goto resume_point_CURLYM2;
4589 case resume_CURLYM3:
4590 goto resume_point_CURLYM3;
4591 case resume_CURLYM4:
4592 goto resume_point_CURLYM4;
4593 case resume_IFMATCH:
4594 goto resume_point_IFMATCH;
4596 goto resume_point_PLUS1;
4598 goto resume_point_PLUS2;
4600 goto resume_point_PLUS3;
4602 goto resume_point_PLUS4;
4604 Perl_croak(aTHX_ "regexp resume memory corruption");
4610 /* restore original high-water mark */
4611 PL_regmatch_slab = orig_slab;
4612 PL_regmatch_state = orig_state;
4614 /* free all slabs above current one */
4615 if (orig_slab->next) {
4616 regmatch_slab *osl, *sl = orig_slab->next;
4617 orig_slab->next = NULL;
4630 - regrepeat - repeatedly match something simple, report how many
4633 * [This routine now assumes that it will only match on things of length 1.
4634 * That was true before, but now we assume scan - reginput is the count,
4635 * rather than incrementing count on every character. [Er, except utf8.]]
4638 S_regrepeat(pTHX_ const regnode *p, I32 max)
4641 register char *scan;
4643 register char *loceol = PL_regeol;
4644 register I32 hardcount = 0;
4645 register bool do_utf8 = PL_reg_match_utf8;
4648 if (max == REG_INFTY)
4650 else if (max < loceol - scan)
4651 loceol = scan + max;
4656 while (scan < loceol && hardcount < max && *scan != '\n') {
4657 scan += UTF8SKIP(scan);
4661 while (scan < loceol && *scan != '\n')
4668 while (scan < loceol && hardcount < max) {
4669 scan += UTF8SKIP(scan);
4679 case EXACT: /* length of string is 1 */
4681 while (scan < loceol && UCHARAT(scan) == c)
4684 case EXACTF: /* length of string is 1 */
4686 while (scan < loceol &&
4687 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
4690 case EXACTFL: /* length of string is 1 */
4691 PL_reg_flags |= RF_tainted;
4693 while (scan < loceol &&
4694 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
4700 while (hardcount < max && scan < loceol &&
4701 reginclass(p, (U8*)scan, 0, do_utf8)) {
4702 scan += UTF8SKIP(scan);
4706 while (scan < loceol && REGINCLASS(p, (U8*)scan))
4713 LOAD_UTF8_CHARCLASS_ALNUM();
4714 while (hardcount < max && scan < loceol &&
4715 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4716 scan += UTF8SKIP(scan);
4720 while (scan < loceol && isALNUM(*scan))
4725 PL_reg_flags |= RF_tainted;
4728 while (hardcount < max && scan < loceol &&
4729 isALNUM_LC_utf8((U8*)scan)) {
4730 scan += UTF8SKIP(scan);
4734 while (scan < loceol && isALNUM_LC(*scan))
4741 LOAD_UTF8_CHARCLASS_ALNUM();
4742 while (hardcount < max && scan < loceol &&
4743 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4744 scan += UTF8SKIP(scan);
4748 while (scan < loceol && !isALNUM(*scan))
4753 PL_reg_flags |= RF_tainted;
4756 while (hardcount < max && scan < loceol &&
4757 !isALNUM_LC_utf8((U8*)scan)) {
4758 scan += UTF8SKIP(scan);
4762 while (scan < loceol && !isALNUM_LC(*scan))
4769 LOAD_UTF8_CHARCLASS_SPACE();
4770 while (hardcount < max && scan < loceol &&
4772 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4773 scan += UTF8SKIP(scan);
4777 while (scan < loceol && isSPACE(*scan))
4782 PL_reg_flags |= RF_tainted;
4785 while (hardcount < max && scan < loceol &&
4786 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4787 scan += UTF8SKIP(scan);
4791 while (scan < loceol && isSPACE_LC(*scan))
4798 LOAD_UTF8_CHARCLASS_SPACE();
4799 while (hardcount < max && scan < loceol &&
4801 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4802 scan += UTF8SKIP(scan);
4806 while (scan < loceol && !isSPACE(*scan))
4811 PL_reg_flags |= RF_tainted;
4814 while (hardcount < max && scan < loceol &&
4815 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4816 scan += UTF8SKIP(scan);
4820 while (scan < loceol && !isSPACE_LC(*scan))
4827 LOAD_UTF8_CHARCLASS_DIGIT();
4828 while (hardcount < max && scan < loceol &&
4829 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4830 scan += UTF8SKIP(scan);
4834 while (scan < loceol && isDIGIT(*scan))
4841 LOAD_UTF8_CHARCLASS_DIGIT();
4842 while (hardcount < max && scan < loceol &&
4843 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4844 scan += UTF8SKIP(scan);
4848 while (scan < loceol && !isDIGIT(*scan))
4852 default: /* Called on something of 0 width. */
4853 break; /* So match right here or not at all. */
4859 c = scan - PL_reginput;
4863 SV *re_debug_flags = NULL;
4864 SV * const prop = sv_newmortal();
4868 PerlIO_printf(Perl_debug_log,
4869 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4870 REPORT_CODE_OFF+1, "", SvPVX_const(prop),(IV)c,(IV)max);
4879 - regclass_swash - prepare the utf8 swash
4883 Perl_regclass_swash(pTHX_ register const regnode* node, bool doinit, SV** listsvp, SV **altsvp)
4889 const struct reg_data *data = PL_reg_re ? PL_reg_re->data : NULL;
4891 if (data && data->count) {
4892 const U32 n = ARG(node);
4894 if (data->what[n] == 's') {
4895 SV * const rv = (SV*)data->data[n];
4896 AV * const av = (AV*)SvRV((SV*)rv);
4897 SV **const ary = AvARRAY(av);
4900 /* See the end of regcomp.c:S_regclass() for
4901 * documentation of these array elements. */
4904 a = SvROK(ary[1]) ? &ary[1] : 0;
4905 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
4909 else if (si && doinit) {
4910 sw = swash_init("utf8", "", si, 1, 0);
4911 (void)av_store(av, 1, sw);
4927 - reginclass - determine if a character falls into a character class
4929 The n is the ANYOF regnode, the p is the target string, lenp
4930 is pointer to the maximum length of how far to go in the p
4931 (if the lenp is zero, UTF8SKIP(p) is used),
4932 do_utf8 tells whether the target string is in UTF-8.
4937 S_reginclass(pTHX_ register const regnode *n, register const U8* p, STRLEN* lenp, register bool do_utf8)
4940 const char flags = ANYOF_FLAGS(n);
4946 if (do_utf8 && !UTF8_IS_INVARIANT(c)) {
4947 c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &len,
4948 (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) | UTF8_CHECK_ONLY);
4949 /* see [perl #37836] for UTF8_ALLOW_ANYUV */
4950 if (len == (STRLEN)-1)
4951 Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)");
4954 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
4955 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4958 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4959 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4962 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4966 SV * const sw = regclass_swash(n, TRUE, 0, (SV**)&av);
4969 if (swash_fetch(sw, p, do_utf8))
4971 else if (flags & ANYOF_FOLD) {
4972 if (!match && lenp && av) {
4974 for (i = 0; i <= av_len(av); i++) {
4975 SV* const sv = *av_fetch(av, i, FALSE);
4977 const char * const s = SvPV_const(sv, len);
4979 if (len <= plen && memEQ(s, (char*)p, len)) {
4987 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
4990 to_utf8_fold(p, tmpbuf, &tmplen);
4991 if (swash_fetch(sw, tmpbuf, do_utf8))
4997 if (match && lenp && *lenp == 0)
4998 *lenp = UNISKIP(NATIVE_TO_UNI(c));
5000 if (!match && c < 256) {
5001 if (ANYOF_BITMAP_TEST(n, c))
5003 else if (flags & ANYOF_FOLD) {
5006 if (flags & ANYOF_LOCALE) {
5007 PL_reg_flags |= RF_tainted;
5008 f = PL_fold_locale[c];
5012 if (f != c && ANYOF_BITMAP_TEST(n, f))
5016 if (!match && (flags & ANYOF_CLASS)) {
5017 PL_reg_flags |= RF_tainted;
5019 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
5020 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
5021 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
5022 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
5023 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
5024 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
5025 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
5026 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
5027 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
5028 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
5029 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
5030 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
5031 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
5032 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
5033 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
5034 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
5035 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
5036 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
5037 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
5038 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
5039 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
5040 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
5041 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
5042 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
5043 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
5044 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
5045 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
5046 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
5047 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
5048 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
5049 ) /* How's that for a conditional? */
5056 return (flags & ANYOF_INVERT) ? !match : match;
5060 S_reghop3(U8 *s, I32 off, U8* lim)
5064 while (off-- && s < lim) {
5065 /* XXX could check well-formedness here */
5073 if (UTF8_IS_CONTINUED(*s)) {
5074 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
5077 /* XXX could check well-formedness here */
5085 S_reghopmaybe3(U8* s, I32 off, U8* lim)
5089 while (off-- && s < lim) {
5090 /* XXX could check well-formedness here */
5100 if (UTF8_IS_CONTINUED(*s)) {
5101 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
5104 /* XXX could check well-formedness here */
5116 restore_pos(pTHX_ void *arg)
5119 regexp * const rex = (regexp *)arg;
5120 if (PL_reg_eval_set) {
5121 if (PL_reg_oldsaved) {
5122 rex->subbeg = PL_reg_oldsaved;
5123 rex->sublen = PL_reg_oldsavedlen;
5124 #ifdef PERL_OLD_COPY_ON_WRITE
5125 rex->saved_copy = PL_nrs;
5127 RX_MATCH_COPIED_on(rex);
5129 PL_reg_magic->mg_len = PL_reg_oldpos;
5130 PL_reg_eval_set = 0;
5131 PL_curpm = PL_reg_oldcurpm;
5136 S_to_utf8_substr(pTHX_ register regexp *prog)
5138 if (prog->float_substr && !prog->float_utf8) {
5139 SV* const sv = newSVsv(prog->float_substr);
5140 prog->float_utf8 = sv;
5141 sv_utf8_upgrade(sv);
5142 if (SvTAIL(prog->float_substr))
5144 if (prog->float_substr == prog->check_substr)
5145 prog->check_utf8 = sv;
5147 if (prog->anchored_substr && !prog->anchored_utf8) {
5148 SV* const sv = newSVsv(prog->anchored_substr);
5149 prog->anchored_utf8 = sv;
5150 sv_utf8_upgrade(sv);
5151 if (SvTAIL(prog->anchored_substr))
5153 if (prog->anchored_substr == prog->check_substr)
5154 prog->check_utf8 = sv;
5159 S_to_byte_substr(pTHX_ register regexp *prog)
5162 if (prog->float_utf8 && !prog->float_substr) {
5163 SV* sv = newSVsv(prog->float_utf8);
5164 prog->float_substr = sv;
5165 if (sv_utf8_downgrade(sv, TRUE)) {
5166 if (SvTAIL(prog->float_utf8))
5170 prog->float_substr = sv = &PL_sv_undef;
5172 if (prog->float_utf8 == prog->check_utf8)
5173 prog->check_substr = sv;
5175 if (prog->anchored_utf8 && !prog->anchored_substr) {
5176 SV* sv = newSVsv(prog->anchored_utf8);
5177 prog->anchored_substr = sv;
5178 if (sv_utf8_downgrade(sv, TRUE)) {
5179 if (SvTAIL(prog->anchored_utf8))
5183 prog->anchored_substr = sv = &PL_sv_undef;
5185 if (prog->anchored_utf8 == prog->check_utf8)
5186 prog->check_substr = sv;
5192 * c-indentation-style: bsd
5194 * indent-tabs-mode: t
5197 * ex: set ts=8 sts=4 sw=4 noet: