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);
1618 RX_MATCH_UTF8_set(prog,do_utf8);
1622 PL_regnarrate = DEBUG_r_TEST;
1625 /* Be paranoid... */
1626 if (prog == NULL || startpos == NULL) {
1627 Perl_croak(aTHX_ "NULL regexp parameter");
1631 minlen = prog->minlen;
1632 if (strend - startpos < minlen) {
1633 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1634 "String too short [regexec_flags]...\n"));
1638 /* Check validity of program. */
1639 if (UCHARAT(prog->program) != REG_MAGIC) {
1640 Perl_croak(aTHX_ "corrupted regexp program");
1644 PL_reg_eval_set = 0;
1647 if (prog->reganch & ROPT_UTF8)
1648 PL_reg_flags |= RF_utf8;
1650 /* Mark beginning of line for ^ and lookbehind. */
1651 PL_regbol = startpos;
1655 /* Mark end of line for $ (and such) */
1658 /* see how far we have to get to not match where we matched before */
1659 PL_regtill = startpos+minend;
1661 /* We start without call_cc context. */
1664 /* If there is a "must appear" string, look for it. */
1667 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1670 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1671 PL_reg_ganch = startpos;
1672 else if (sv && SvTYPE(sv) >= SVt_PVMG
1674 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1675 && mg->mg_len >= 0) {
1676 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1677 if (prog->reganch & ROPT_ANCH_GPOS) {
1678 if (s > PL_reg_ganch)
1683 else /* pos() not defined */
1684 PL_reg_ganch = strbeg;
1687 if (!(flags & REXEC_CHECKED) && (prog->check_substr != NULL || prog->check_utf8 != NULL)) {
1688 re_scream_pos_data d;
1690 d.scream_olds = &scream_olds;
1691 d.scream_pos = &scream_pos;
1692 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1694 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1695 goto phooey; /* not present */
1700 const char * const s0 = UTF
1701 ? pv_uni_display(dsv0, (U8*)prog->precomp, prog->prelen, 60,
1704 const int len0 = UTF ? SvCUR(dsv0) : prog->prelen;
1705 const char * const s1 = do_utf8 ? sv_uni_display(dsv1, sv, 60,
1706 UNI_DISPLAY_REGEX) : startpos;
1707 const int len1 = do_utf8 ? SvCUR(dsv1) : strend - startpos;
1710 PerlIO_printf(Perl_debug_log,
1711 "%sMatching REx%s \"%s%*.*s%s%s\" against \"%s%.*s%s%s\"\n",
1712 PL_colors[4], PL_colors[5], PL_colors[0],
1715 len0 > 60 ? "..." : "",
1717 (int)(len1 > 60 ? 60 : len1),
1719 (len1 > 60 ? "..." : "")
1723 /* Simplest case: anchored match need be tried only once. */
1724 /* [unless only anchor is BOL and multiline is set] */
1725 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1726 if (s == startpos && regtry(prog, startpos))
1728 else if (multiline || (prog->reganch & ROPT_IMPLICIT)
1729 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1734 dontbother = minlen - 1;
1735 end = HOP3c(strend, -dontbother, strbeg) - 1;
1736 /* for multiline we only have to try after newlines */
1737 if (prog->check_substr || prog->check_utf8) {
1741 if (regtry(prog, s))
1746 if (prog->reganch & RE_USE_INTUIT) {
1747 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1758 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1759 if (regtry(prog, s))
1766 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1767 if (regtry(prog, PL_reg_ganch))
1772 /* Messy cases: unanchored match. */
1773 if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) {
1774 /* we have /x+whatever/ */
1775 /* it must be a one character string (XXXX Except UTF?) */
1780 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1781 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1782 ch = SvPVX_const(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
1785 while (s < strend) {
1787 DEBUG_EXECUTE_r( did_match = 1 );
1788 if (regtry(prog, s)) goto got_it;
1790 while (s < strend && *s == ch)
1797 while (s < strend) {
1799 DEBUG_EXECUTE_r( did_match = 1 );
1800 if (regtry(prog, s)) goto got_it;
1802 while (s < strend && *s == ch)
1808 DEBUG_EXECUTE_r(if (!did_match)
1809 PerlIO_printf(Perl_debug_log,
1810 "Did not find anchored character...\n")
1813 else if (prog->anchored_substr != NULL
1814 || prog->anchored_utf8 != NULL
1815 || ((prog->float_substr != NULL || prog->float_utf8 != NULL)
1816 && prog->float_max_offset < strend - s)) {
1821 char *last1; /* Last position checked before */
1825 if (prog->anchored_substr || prog->anchored_utf8) {
1826 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1827 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1828 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1829 back_max = back_min = prog->anchored_offset;
1831 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1832 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1833 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1834 back_max = prog->float_max_offset;
1835 back_min = prog->float_min_offset;
1837 if (must == &PL_sv_undef)
1838 /* could not downgrade utf8 check substring, so must fail */
1841 last = HOP3c(strend, /* Cannot start after this */
1842 -(I32)(CHR_SVLEN(must)
1843 - (SvTAIL(must) != 0) + back_min), strbeg);
1846 last1 = HOPc(s, -1);
1848 last1 = s - 1; /* bogus */
1850 /* XXXX check_substr already used to find "s", can optimize if
1851 check_substr==must. */
1853 dontbother = end_shift;
1854 strend = HOPc(strend, -dontbother);
1855 while ( (s <= last) &&
1856 ((flags & REXEC_SCREAM)
1857 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1858 end_shift, &scream_pos, 0))
1859 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1860 (unsigned char*)strend, must,
1861 multiline ? FBMrf_MULTILINE : 0))) ) {
1862 /* we may be pointing at the wrong string */
1863 if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog))
1864 s = strbeg + (s - SvPVX_const(sv));
1865 DEBUG_EXECUTE_r( did_match = 1 );
1866 if (HOPc(s, -back_max) > last1) {
1867 last1 = HOPc(s, -back_min);
1868 s = HOPc(s, -back_max);
1871 char * const t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1873 last1 = HOPc(s, -back_min);
1877 while (s <= last1) {
1878 if (regtry(prog, s))
1884 while (s <= last1) {
1885 if (regtry(prog, s))
1891 DEBUG_EXECUTE_r(if (!did_match)
1892 PerlIO_printf(Perl_debug_log,
1893 "Did not find %s substr \"%s%.*s%s\"%s...\n",
1894 ((must == prog->anchored_substr || must == prog->anchored_utf8)
1895 ? "anchored" : "floating"),
1897 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1899 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1903 else if ((c = prog->regstclass)) {
1905 I32 op = (U8)OP(prog->regstclass);
1906 /* don't bother with what can't match */
1907 if (PL_regkind[op] != EXACT && op != CANY)
1908 strend = HOPc(strend, -(minlen - 1));
1911 SV *prop = sv_newmortal();
1919 pv_uni_display(dsv0, (U8*)SvPVX_const(prop), SvCUR(prop), 60,
1920 UNI_DISPLAY_REGEX) :
1922 len0 = UTF ? SvCUR(dsv0) : SvCUR(prop);
1924 sv_uni_display(dsv1, sv, 60, UNI_DISPLAY_REGEX) : s;
1925 len1 = UTF ? SvCUR(dsv1) : strend - s;
1926 PerlIO_printf(Perl_debug_log,
1927 "Matching stclass \"%*.*s\" against \"%*.*s\"\n",
1931 if (find_byclass(prog, c, s, strend, 0))
1933 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1937 if (prog->float_substr != NULL || prog->float_utf8 != NULL) {
1942 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1943 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1944 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
1946 if (flags & REXEC_SCREAM) {
1947 last = screaminstr(sv, float_real, s - strbeg,
1948 end_shift, &scream_pos, 1); /* last one */
1950 last = scream_olds; /* Only one occurrence. */
1951 /* we may be pointing at the wrong string */
1952 else if (RX_MATCH_COPIED(prog))
1953 s = strbeg + (s - SvPVX_const(sv));
1957 const char * const little = SvPV_const(float_real, len);
1959 if (SvTAIL(float_real)) {
1960 if (memEQ(strend - len + 1, little, len - 1))
1961 last = strend - len + 1;
1962 else if (!multiline)
1963 last = memEQ(strend - len, little, len)
1964 ? strend - len : NULL;
1970 last = rninstr(s, strend, little, little + len);
1972 last = strend; /* matching "$" */
1976 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1977 "%sCan't trim the tail, match fails (should not happen)%s\n",
1978 PL_colors[4], PL_colors[5]));
1979 goto phooey; /* Should not happen! */
1981 dontbother = strend - last + prog->float_min_offset;
1983 if (minlen && (dontbother < minlen))
1984 dontbother = minlen - 1;
1985 strend -= dontbother; /* this one's always in bytes! */
1986 /* We don't know much -- general case. */
1989 if (regtry(prog, s))
1998 if (regtry(prog, s))
2000 } while (s++ < strend);
2008 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
2010 if (PL_reg_eval_set) {
2011 /* Preserve the current value of $^R */
2012 if (oreplsv != GvSV(PL_replgv))
2013 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
2014 restored, the value remains
2016 restore_pos(aTHX_ prog);
2019 /* make sure $`, $&, $', and $digit will work later */
2020 if ( !(flags & REXEC_NOT_FIRST) ) {
2021 RX_MATCH_COPY_FREE(prog);
2022 if (flags & REXEC_COPY_STR) {
2023 I32 i = PL_regeol - startpos + (stringarg - strbeg);
2024 #ifdef PERL_OLD_COPY_ON_WRITE
2026 || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) {
2028 PerlIO_printf(Perl_debug_log,
2029 "Copy on write: regexp capture, type %d\n",
2032 prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv);
2033 prog->subbeg = (char *)SvPVX_const(prog->saved_copy);
2034 assert (SvPOKp(prog->saved_copy));
2038 RX_MATCH_COPIED_on(prog);
2039 s = savepvn(strbeg, i);
2045 prog->subbeg = strbeg;
2046 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2053 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2054 PL_colors[4], PL_colors[5]));
2055 if (PL_reg_eval_set)
2056 restore_pos(aTHX_ prog);
2061 - regtry - try match at specific point
2063 STATIC I32 /* 0 failure, 1 success */
2064 S_regtry(pTHX_ regexp *prog, char *startpos)
2070 GET_RE_DEBUG_FLAGS_DECL;
2073 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
2075 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
2078 PL_reg_eval_set = RS_init;
2079 DEBUG_EXECUTE_r(DEBUG_s(
2080 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2081 (IV)(PL_stack_sp - PL_stack_base));
2083 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
2084 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2085 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2087 /* Apparently this is not needed, judging by wantarray. */
2088 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
2089 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2092 /* Make $_ available to executed code. */
2093 if (PL_reg_sv != DEFSV) {
2098 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
2099 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
2100 /* prepare for quick setting of pos */
2101 #ifdef PERL_OLD_COPY_ON_WRITE
2103 sv_force_normal_flags(sv, 0);
2105 mg = sv_magicext(PL_reg_sv, (SV*)0, PERL_MAGIC_regex_global,
2106 &PL_vtbl_mglob, NULL, 0);
2110 PL_reg_oldpos = mg->mg_len;
2111 SAVEDESTRUCTOR_X(restore_pos, prog);
2113 if (!PL_reg_curpm) {
2114 Newxz(PL_reg_curpm, 1, PMOP);
2117 SV* repointer = newSViv(0);
2118 /* so we know which PL_regex_padav element is PL_reg_curpm */
2119 SvFLAGS(repointer) |= SVf_BREAK;
2120 av_push(PL_regex_padav,repointer);
2121 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2122 PL_regex_pad = AvARRAY(PL_regex_padav);
2126 PM_SETRE(PL_reg_curpm, prog);
2127 PL_reg_oldcurpm = PL_curpm;
2128 PL_curpm = PL_reg_curpm;
2129 if (RX_MATCH_COPIED(prog)) {
2130 /* Here is a serious problem: we cannot rewrite subbeg,
2131 since it may be needed if this match fails. Thus
2132 $` inside (?{}) could fail... */
2133 PL_reg_oldsaved = prog->subbeg;
2134 PL_reg_oldsavedlen = prog->sublen;
2135 #ifdef PERL_OLD_COPY_ON_WRITE
2136 PL_nrs = prog->saved_copy;
2138 RX_MATCH_COPIED_off(prog);
2141 PL_reg_oldsaved = NULL;
2142 prog->subbeg = PL_bostr;
2143 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2145 prog->startp[0] = startpos - PL_bostr;
2146 PL_reginput = startpos;
2147 PL_regstartp = prog->startp;
2148 PL_regendp = prog->endp;
2149 PL_reglastparen = &prog->lastparen;
2150 PL_reglastcloseparen = &prog->lastcloseparen;
2151 prog->lastparen = 0;
2152 prog->lastcloseparen = 0;
2154 DEBUG_EXECUTE_r(PL_reg_starttry = startpos);
2155 if (PL_reg_start_tmpl <= prog->nparens) {
2156 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2157 if(PL_reg_start_tmp)
2158 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2160 Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2163 /* XXXX What this code is doing here?!!! There should be no need
2164 to do this again and again, PL_reglastparen should take care of
2167 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2168 * Actually, the code in regcppop() (which Ilya may be meaning by
2169 * PL_reglastparen), is not needed at all by the test suite
2170 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2171 * enough, for building DynaLoader, or otherwise this
2172 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2173 * will happen. Meanwhile, this code *is* needed for the
2174 * above-mentioned test suite tests to succeed. The common theme
2175 * on those tests seems to be returning null fields from matches.
2180 if (prog->nparens) {
2182 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2189 if (regmatch(prog, prog->program + 1)) {
2190 prog->endp[0] = PL_reginput - PL_bostr;
2193 REGCP_UNWIND(lastcp);
2197 #define RE_UNWIND_BRANCH 1
2198 #define RE_UNWIND_BRANCHJ 2
2202 typedef struct { /* XX: makes sense to enlarge it... */
2206 } re_unwind_generic_t;
2219 } re_unwind_branch_t;
2221 typedef union re_unwind_t {
2223 re_unwind_generic_t generic;
2224 re_unwind_branch_t branch;
2227 #define sayYES goto yes
2228 #define sayNO goto no
2229 #define sayNO_ANYOF goto no_anyof
2230 #define sayYES_FINAL goto yes_final
2231 #define sayYES_LOUD goto yes_loud
2232 #define sayNO_FINAL goto no_final
2233 #define sayNO_SILENT goto do_no
2234 #define saySAME(x) if (x) goto yes; else goto no
2236 #define POSCACHE_SUCCESS 0 /* caching success rather than failure */
2237 #define POSCACHE_SEEN 1 /* we know what we're caching */
2238 #define POSCACHE_START 2 /* the real cache: this bit maps to pos 0 */
2239 #define CACHEsayYES STMT_START { \
2240 if (st->u.whilem.cache_offset | st->u.whilem.cache_bit) { \
2241 if (!(PL_reg_poscache[0] & (1<<POSCACHE_SEEN))) \
2242 PL_reg_poscache[0] |= (1<<POSCACHE_SUCCESS) || (1<<POSCACHE_SEEN); \
2243 else if (!(PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))) { \
2244 /* cache records failure, but this is success */ \
2246 PerlIO_printf(Perl_debug_log, \
2247 "%*s (remove success from failure cache)\n", \
2248 REPORT_CODE_OFF+PL_regindent*2, "") \
2250 PL_reg_poscache[st->u.whilem.cache_offset] &= ~(1<<st->u.whilem.cache_bit); \
2255 #define CACHEsayNO STMT_START { \
2256 if (st->u.whilem.cache_offset | st->u.whilem.cache_bit) { \
2257 if (!(PL_reg_poscache[0] & (1<<POSCACHE_SEEN))) \
2258 PL_reg_poscache[0] |= (1<<POSCACHE_SEEN); \
2259 else if ((PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))) { \
2260 /* cache records success, but this is failure */ \
2262 PerlIO_printf(Perl_debug_log, \
2263 "%*s (remove failure from success cache)\n", \
2264 REPORT_CODE_OFF+PL_regindent*2, "") \
2266 PL_reg_poscache[st->u.whilem.cache_offset] &= ~(1<<st->u.whilem.cache_bit); \
2272 /* this is used to determine how far from the left messages like
2273 'failed...' are printed. Currently 29 makes these messages line
2274 up with the opcode they refer to. Earlier perls used 25 which
2275 left these messages outdented making reviewing a debug output
2278 #define REPORT_CODE_OFF 29
2281 /* Make sure there is a test for this +1 options in re_tests */
2282 #define TRIE_INITAL_ACCEPT_BUFFLEN 4;
2284 /* grab a new slab and return the first slot in it */
2286 STATIC regmatch_state *
2289 regmatch_slab *s = PL_regmatch_slab->next;
2291 Newx(s, 1, regmatch_slab);
2292 s->prev = PL_regmatch_slab;
2294 PL_regmatch_slab->next = s;
2296 PL_regmatch_slab = s;
2297 return &s->states[0];
2300 /* simulate a recursive call to regmatch */
2302 #define REGMATCH(ns, where) \
2305 st->resume_state = resume_##where; \
2306 goto start_recurse; \
2307 resume_point_##where:
2310 - regmatch - main matching routine
2312 * Conceptually the strategy is simple: check to see whether the current
2313 * node matches, call self recursively to see whether the rest matches,
2314 * and then act accordingly. In practice we make some effort to avoid
2315 * recursion, in particular by going through "ordinary" nodes (that don't
2316 * need to know whether the rest of the match failed) by a loop instead of
2319 /* [lwall] I've hoisted the register declarations to the outer block in order to
2320 * maybe save a little bit of pushing and popping on the stack. It also takes
2321 * advantage of machines that use a register save mask on subroutine entry.
2323 * This function used to be heavily recursive, but since this had the
2324 * effect of blowing the CPU stack on complex regexes, it has been
2325 * restructured to be iterative, and to save state onto the heap rather
2326 * than the stack. Essentially whereever regmatch() used to be called, it
2327 * pushes the current state, notes where to return, then jumps back into
2330 * Originally the structure of this function used to look something like
2335 while (scan != NULL) {
2336 a++; // do stuff with a and b
2342 if (regmatch(...)) // recurse
2352 * Now it looks something like this:
2360 regmatch_state *st = new();
2362 st->a++; // do stuff with a and b
2364 while (scan != NULL) {
2372 st->resume_state = resume_FOO;
2373 goto start_recurse; // recurse
2382 st = new(); push a new state
2383 st->a = 1; st->b = 2;
2390 switch (resume_state) {
2392 goto resume_point_FOO;
2399 * WARNING: this means that any line in this function that contains a
2400 * REGMATCH() or TRYPAREN() is actually simulating a recursive call to
2401 * regmatch() using gotos instead. Thus the values of any local variables
2402 * not saved in the regmatch_state structure will have been lost when
2403 * execution resumes on the next line .
2405 * States (ie the st pointer) are allocated in slabs of about 4K in size.
2406 * PL_regmatch_state always points to the currently active state, and
2407 * PL_regmatch_slab points to the slab currently containing PL_regmatch_state.
2408 * The first time regmatch is called, the first slab is allocated, and is
2409 * never freed until interpreter desctruction. When the slab is full,
2410 * a new one is allocated chained to the end. At exit from regmatch, slabs
2411 * allocated since entry are freed.
2415 STATIC I32 /* 0 failure, 1 success */
2416 S_regmatch(pTHX_ regexp *rex, regnode *prog)
2419 register const bool do_utf8 = PL_reg_match_utf8;
2420 const U32 uniflags = UTF8_ALLOW_DEFAULT;
2422 regmatch_slab *orig_slab;
2423 regmatch_state *orig_state;
2425 /* the current state. This is a cached copy of PL_regmatch_state */
2426 register regmatch_state *st;
2428 /* cache heavy used fields of st in registers */
2429 register regnode *scan;
2430 register regnode *next;
2431 register I32 n = 0; /* initialize to shut up compiler warning */
2432 register char *locinput = PL_reginput;
2434 /* these variables are NOT saved during a recusive RFEGMATCH: */
2435 register I32 nextchr; /* is always set to UCHARAT(locinput) */
2436 bool result; /* return value of S_regmatch */
2437 regnode *inner; /* Next node in internal branch. */
2438 int depth = 0; /* depth of recursion */
2441 SV *re_debug_flags = NULL;
2446 /* on first ever call to regmatch, allocate first slab */
2447 if (!PL_regmatch_slab) {
2448 Newx(PL_regmatch_slab, 1, regmatch_slab);
2449 PL_regmatch_slab->prev = NULL;
2450 PL_regmatch_slab->next = NULL;
2451 PL_regmatch_state = &PL_regmatch_slab->states[0] - 1;
2454 /* remember current high-water mark for exit */
2455 /* XXX this should be done with SAVE* instead */
2456 orig_slab = PL_regmatch_slab;
2457 orig_state = PL_regmatch_state;
2459 /* grab next free state slot */
2460 st = ++PL_regmatch_state;
2461 if (st > &(PL_regmatch_slab->states[PERL_REGMATCH_SLAB_SLOTS-1]))
2462 st = PL_regmatch_state = S_push_slab(aTHX);
2469 /* Note that nextchr is a byte even in UTF */
2470 nextchr = UCHARAT(locinput);
2472 while (scan != NULL) {
2475 SV * const prop = sv_newmortal();
2476 const int docolor = *PL_colors[0];
2477 const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2478 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2479 /* The part of the string before starttry has one color
2480 (pref0_len chars), between starttry and current
2481 position another one (pref_len - pref0_len chars),
2482 after the current position the third one.
2483 We assume that pref0_len <= pref_len, otherwise we
2484 decrease pref0_len. */
2485 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2486 ? (5 + taill) - l : locinput - PL_bostr;
2489 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2491 pref0_len = pref_len - (locinput - PL_reg_starttry);
2492 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2493 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2494 ? (5 + taill) - pref_len : PL_regeol - locinput);
2495 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2499 if (pref0_len > pref_len)
2500 pref0_len = pref_len;
2501 regprop(prop, scan);
2503 const char * const s0 =
2504 do_utf8 && OP(scan) != CANY ?
2505 pv_uni_display(PERL_DEBUG_PAD(0), (U8*)(locinput - pref_len),
2506 pref0_len, 60, UNI_DISPLAY_REGEX) :
2507 locinput - pref_len;
2508 const int len0 = do_utf8 ? strlen(s0) : pref0_len;
2509 const char * const s1 = do_utf8 && OP(scan) != CANY ?
2510 pv_uni_display(PERL_DEBUG_PAD(1),
2511 (U8*)(locinput - pref_len + pref0_len),
2512 pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) :
2513 locinput - pref_len + pref0_len;
2514 const int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2515 const char * const s2 = do_utf8 && OP(scan) != CANY ?
2516 pv_uni_display(PERL_DEBUG_PAD(2), (U8*)locinput,
2517 PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
2519 const int len2 = do_utf8 ? strlen(s2) : l;
2520 PerlIO_printf(Perl_debug_log,
2521 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2522 (IV)(locinput - PL_bostr),
2529 (docolor ? "" : "> <"),
2533 15 - l - pref_len + 1,
2535 (IV)(scan - rex->program), PL_regindent*2, "",
2540 next = scan + NEXT_OFF(scan);
2546 if (locinput == PL_bostr)
2548 /* regtill = regbol; */
2553 if (locinput == PL_bostr ||
2554 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2560 if (locinput == PL_bostr)
2564 if (locinput == PL_reg_ganch)
2570 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2575 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2577 if (PL_regeol - locinput > 1)
2581 if (PL_regeol != locinput)
2585 if (!nextchr && locinput >= PL_regeol)
2588 locinput += PL_utf8skip[nextchr];
2589 if (locinput > PL_regeol)
2591 nextchr = UCHARAT(locinput);
2594 nextchr = UCHARAT(++locinput);
2597 if (!nextchr && locinput >= PL_regeol)
2599 nextchr = UCHARAT(++locinput);
2602 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2605 locinput += PL_utf8skip[nextchr];
2606 if (locinput > PL_regeol)
2608 nextchr = UCHARAT(locinput);
2611 nextchr = UCHARAT(++locinput);
2617 traverse the TRIE keeping track of all accepting states
2618 we transition through until we get to a failing node.
2626 U8 *uc = ( U8* )locinput;
2633 U8 *uscan = (U8*)NULL;
2635 SV *sv_accept_buff = NULL;
2636 const enum { trie_plain, trie_utf8, trie_uft8_fold }
2637 trie_type = do_utf8 ?
2638 (OP(scan) == TRIE ? trie_utf8 : trie_uft8_fold)
2641 /* what trie are we using right now */
2643 = (reg_trie_data*)PL_reg_re->data->data[ ARG( scan ) ];
2644 st->u.trie.accepted = 0; /* how many accepting states we have seen */
2647 while ( state && uc <= (U8*)PL_regeol ) {
2649 if (trie->states[ state ].wordnum) {
2650 if (!st->u.trie.accepted ) {
2653 bufflen = TRIE_INITAL_ACCEPT_BUFFLEN;
2654 sv_accept_buff=newSV(bufflen *
2655 sizeof(reg_trie_accepted) - 1);
2656 SvCUR_set(sv_accept_buff,
2657 sizeof(reg_trie_accepted));
2658 SvPOK_on(sv_accept_buff);
2659 sv_2mortal(sv_accept_buff);
2660 st->u.trie.accept_buff =
2661 (reg_trie_accepted*)SvPV_nolen(sv_accept_buff );
2664 if (st->u.trie.accepted >= bufflen) {
2666 st->u.trie.accept_buff =(reg_trie_accepted*)
2667 SvGROW(sv_accept_buff,
2668 bufflen * sizeof(reg_trie_accepted));
2670 SvCUR_set(sv_accept_buff,SvCUR(sv_accept_buff)
2671 + sizeof(reg_trie_accepted));
2673 st->u.trie.accept_buff[st->u.trie.accepted].wordnum = trie->states[state].wordnum;
2674 st->u.trie.accept_buff[st->u.trie.accepted].endpos = uc;
2675 ++st->u.trie.accepted;
2678 base = trie->states[ state ].trans.base;
2680 DEBUG_TRIE_EXECUTE_r(
2681 PerlIO_printf( Perl_debug_log,
2682 "%*s %sState: %4"UVxf", Base: %4"UVxf", Accepted: %4"UVxf" ",
2683 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
2684 (UV)state, (UV)base, (UV)st->u.trie.accepted );
2688 switch (trie_type) {
2689 case trie_uft8_fold:
2691 uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags );
2696 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
2697 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags );
2698 uvc = to_uni_fold( uvc, foldbuf, &foldlen );
2699 foldlen -= UNISKIP( uvc );
2700 uscan = foldbuf + UNISKIP( uvc );
2704 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN,
2713 charid = trie->charmap[ uvc ];
2717 if (trie->widecharmap) {
2718 SV** svpp = (SV**)NULL;
2719 svpp = hv_fetch(trie->widecharmap,
2720 (char*)&uvc, sizeof(UV), 0);
2722 charid = (U16)SvIV(*svpp);
2727 (base + charid > trie->uniquecharcount )
2728 && (base + charid - 1 - trie->uniquecharcount
2730 && trie->trans[base + charid - 1 -
2731 trie->uniquecharcount].check == state)
2733 state = trie->trans[base + charid - 1 -
2734 trie->uniquecharcount ].next;
2745 DEBUG_TRIE_EXECUTE_r(
2746 PerlIO_printf( Perl_debug_log,
2747 "Charid:%3x CV:%4"UVxf" After State: %4"UVxf"%s\n",
2748 charid, uvc, (UV)state, PL_colors[5] );
2751 if (!st->u.trie.accepted )
2755 There was at least one accepting state that we
2756 transitioned through. Presumably the number of accepting
2757 states is going to be low, typically one or two. So we
2758 simply scan through to find the one with lowest wordnum.
2759 Once we find it, we swap the last state into its place
2760 and decrement the size. We then try to match the rest of
2761 the pattern at the point where the word ends, if we
2762 succeed then we end the loop, otherwise the loop
2763 eventually terminates once all of the accepting states
2767 if ( st->u.trie.accepted == 1 ) {
2769 SV ** const tmp = av_fetch( trie->words, st->u.trie.accept_buff[ 0 ].wordnum-1, 0 );
2770 PerlIO_printf( Perl_debug_log,
2771 "%*s %sonly one match : #%d <%s>%s\n",
2772 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],
2773 st->u.trie.accept_buff[ 0 ].wordnum,
2774 tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr",
2777 PL_reginput = (char *)st->u.trie.accept_buff[ 0 ].endpos;
2778 /* in this case we free tmps/leave before we call regmatch
2779 as we wont be using accept_buff again. */
2782 REGMATCH(scan + NEXT_OFF(scan), TRIE1);
2783 /*** all unsaved local vars undefined at this point */
2786 PerlIO_printf( Perl_debug_log,"%*s %sgot %"IVdf" possible matches%s\n",
2787 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], (IV)st->u.trie.accepted,
2790 while ( !result && st->u.trie.accepted-- ) {
2793 for( cur = 1 ; cur <= st->u.trie.accepted ; cur++ ) {
2794 DEBUG_TRIE_EXECUTE_r(
2795 PerlIO_printf( Perl_debug_log,
2796 "%*s %sgot %"IVdf" (%d) as best, looking at %"IVdf" (%d)%s\n",
2797 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
2798 (IV)best, st->u.trie.accept_buff[ best ].wordnum, (IV)cur,
2799 st->u.trie.accept_buff[ cur ].wordnum, PL_colors[5] );
2802 if (st->u.trie.accept_buff[cur].wordnum <
2803 st->u.trie.accept_buff[best].wordnum)
2807 reg_trie_data * const trie = (reg_trie_data*)
2808 PL_reg_re->data->data[ARG(scan)];
2809 SV ** const tmp = av_fetch( trie->words, st->u.trie.accept_buff[ best ].wordnum - 1, 0 );
2810 PerlIO_printf( Perl_debug_log, "%*s %strying alternation #%d <%s> at 0x%p%s\n",
2811 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],
2812 st->u.trie.accept_buff[best].wordnum,
2813 tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr",scan,
2816 if ( best<st->u.trie.accepted ) {
2817 reg_trie_accepted tmp = st->u.trie.accept_buff[ best ];
2818 st->u.trie.accept_buff[ best ] = st->u.trie.accept_buff[ st->u.trie.accepted ];
2819 st->u.trie.accept_buff[ st->u.trie.accepted ] = tmp;
2820 best = st->u.trie.accepted;
2822 PL_reginput = (char *)st->u.trie.accept_buff[ best ].endpos;
2825 as far as I can tell we only need the SAVETMPS/FREETMPS
2826 for re's with EVAL in them but I'm leaving them in for
2827 all until I can be sure.
2830 REGMATCH(scan + NEXT_OFF(scan), TRIE2);
2831 /*** all unsaved local vars undefined at this point */
2844 /* unreached codepoint */
2846 char *s = STRING(scan);
2847 st->ln = STR_LEN(scan);
2848 if (do_utf8 != UTF) {
2849 /* The target and the pattern have differing utf8ness. */
2851 const char *e = s + st->ln;
2854 /* The target is utf8, the pattern is not utf8. */
2859 if (NATIVE_TO_UNI(*(U8*)s) !=
2860 utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen,
2868 /* The target is not utf8, the pattern is utf8. */
2873 if (NATIVE_TO_UNI(*((U8*)l)) !=
2874 utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen,
2882 nextchr = UCHARAT(locinput);
2885 /* The target and the pattern have the same utf8ness. */
2886 /* Inline the first character, for speed. */
2887 if (UCHARAT(s) != nextchr)
2889 if (PL_regeol - locinput < st->ln)
2891 if (st->ln > 1 && memNE(s, locinput, st->ln))
2894 nextchr = UCHARAT(locinput);
2898 PL_reg_flags |= RF_tainted;
2901 char *s = STRING(scan);
2902 st->ln = STR_LEN(scan);
2904 if (do_utf8 || UTF) {
2905 /* Either target or the pattern are utf8. */
2907 char *e = PL_regeol;
2909 if (ibcmp_utf8(s, 0, st->ln, (bool)UTF,
2910 l, &e, 0, do_utf8)) {
2911 /* One more case for the sharp s:
2912 * pack("U0U*", 0xDF) =~ /ss/i,
2913 * the 0xC3 0x9F are the UTF-8
2914 * byte sequence for the U+00DF. */
2916 toLOWER(s[0]) == 's' &&
2918 toLOWER(s[1]) == 's' &&
2925 nextchr = UCHARAT(locinput);
2929 /* Neither the target and the pattern are utf8. */
2931 /* Inline the first character, for speed. */
2932 if (UCHARAT(s) != nextchr &&
2933 UCHARAT(s) != ((OP(scan) == EXACTF)
2934 ? PL_fold : PL_fold_locale)[nextchr])
2936 if (PL_regeol - locinput < st->ln)
2938 if (st->ln > 1 && (OP(scan) == EXACTF
2939 ? ibcmp(s, locinput, st->ln)
2940 : ibcmp_locale(s, locinput, st->ln)))
2943 nextchr = UCHARAT(locinput);
2948 STRLEN inclasslen = PL_regeol - locinput;
2950 if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8))
2952 if (locinput >= PL_regeol)
2954 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
2955 nextchr = UCHARAT(locinput);
2960 nextchr = UCHARAT(locinput);
2961 if (!REGINCLASS(scan, (U8*)locinput))
2963 if (!nextchr && locinput >= PL_regeol)
2965 nextchr = UCHARAT(++locinput);
2969 /* If we might have the case of the German sharp s
2970 * in a casefolding Unicode character class. */
2972 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
2973 locinput += SHARP_S_SKIP;
2974 nextchr = UCHARAT(locinput);
2980 PL_reg_flags |= RF_tainted;
2986 LOAD_UTF8_CHARCLASS_ALNUM();
2987 if (!(OP(scan) == ALNUM
2988 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2989 : isALNUM_LC_utf8((U8*)locinput)))
2993 locinput += PL_utf8skip[nextchr];
2994 nextchr = UCHARAT(locinput);
2997 if (!(OP(scan) == ALNUM
2998 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
3000 nextchr = UCHARAT(++locinput);
3003 PL_reg_flags |= RF_tainted;
3006 if (!nextchr && locinput >= PL_regeol)
3009 LOAD_UTF8_CHARCLASS_ALNUM();
3010 if (OP(scan) == NALNUM
3011 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
3012 : isALNUM_LC_utf8((U8*)locinput))
3016 locinput += PL_utf8skip[nextchr];
3017 nextchr = UCHARAT(locinput);
3020 if (OP(scan) == NALNUM
3021 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
3023 nextchr = UCHARAT(++locinput);
3027 PL_reg_flags |= RF_tainted;
3031 /* was last char in word? */
3033 if (locinput == PL_bostr)
3036 const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
3038 st->ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, uniflags);
3040 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
3041 st->ln = isALNUM_uni(st->ln);
3042 LOAD_UTF8_CHARCLASS_ALNUM();
3043 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
3046 st->ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(st->ln));
3047 n = isALNUM_LC_utf8((U8*)locinput);
3051 st->ln = (locinput != PL_bostr) ?
3052 UCHARAT(locinput - 1) : '\n';
3053 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
3054 st->ln = isALNUM(st->ln);
3055 n = isALNUM(nextchr);
3058 st->ln = isALNUM_LC(st->ln);
3059 n = isALNUM_LC(nextchr);
3062 if (((!st->ln) == (!n)) == (OP(scan) == BOUND ||
3063 OP(scan) == BOUNDL))
3067 PL_reg_flags |= RF_tainted;
3073 if (UTF8_IS_CONTINUED(nextchr)) {
3074 LOAD_UTF8_CHARCLASS_SPACE();
3075 if (!(OP(scan) == SPACE
3076 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
3077 : isSPACE_LC_utf8((U8*)locinput)))
3081 locinput += PL_utf8skip[nextchr];
3082 nextchr = UCHARAT(locinput);
3085 if (!(OP(scan) == SPACE
3086 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3088 nextchr = UCHARAT(++locinput);
3091 if (!(OP(scan) == SPACE
3092 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3094 nextchr = UCHARAT(++locinput);
3098 PL_reg_flags |= RF_tainted;
3101 if (!nextchr && locinput >= PL_regeol)
3104 LOAD_UTF8_CHARCLASS_SPACE();
3105 if (OP(scan) == NSPACE
3106 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
3107 : isSPACE_LC_utf8((U8*)locinput))
3111 locinput += PL_utf8skip[nextchr];
3112 nextchr = UCHARAT(locinput);
3115 if (OP(scan) == NSPACE
3116 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
3118 nextchr = UCHARAT(++locinput);
3121 PL_reg_flags |= RF_tainted;
3127 LOAD_UTF8_CHARCLASS_DIGIT();
3128 if (!(OP(scan) == DIGIT
3129 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
3130 : isDIGIT_LC_utf8((U8*)locinput)))
3134 locinput += PL_utf8skip[nextchr];
3135 nextchr = UCHARAT(locinput);
3138 if (!(OP(scan) == DIGIT
3139 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
3141 nextchr = UCHARAT(++locinput);
3144 PL_reg_flags |= RF_tainted;
3147 if (!nextchr && locinput >= PL_regeol)
3150 LOAD_UTF8_CHARCLASS_DIGIT();
3151 if (OP(scan) == NDIGIT
3152 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
3153 : isDIGIT_LC_utf8((U8*)locinput))
3157 locinput += PL_utf8skip[nextchr];
3158 nextchr = UCHARAT(locinput);
3161 if (OP(scan) == NDIGIT
3162 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
3164 nextchr = UCHARAT(++locinput);
3167 if (locinput >= PL_regeol)
3170 LOAD_UTF8_CHARCLASS_MARK();
3171 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3173 locinput += PL_utf8skip[nextchr];
3174 while (locinput < PL_regeol &&
3175 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3176 locinput += UTF8SKIP(locinput);
3177 if (locinput > PL_regeol)
3182 nextchr = UCHARAT(locinput);
3185 PL_reg_flags |= RF_tainted;
3190 n = ARG(scan); /* which paren pair */
3191 st->ln = PL_regstartp[n];
3192 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3193 if ((I32)*PL_reglastparen < n || st->ln == -1)
3194 sayNO; /* Do not match unless seen CLOSEn. */
3195 if (st->ln == PL_regendp[n])
3198 s = PL_bostr + st->ln;
3199 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
3201 const char *e = PL_bostr + PL_regendp[n];
3203 * Note that we can't do the "other character" lookup trick as
3204 * in the 8-bit case (no pun intended) because in Unicode we
3205 * have to map both upper and title case to lower case.
3207 if (OP(scan) == REFF) {
3209 STRLEN ulen1, ulen2;
3210 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
3211 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
3215 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
3216 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
3217 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
3224 nextchr = UCHARAT(locinput);
3228 /* Inline the first character, for speed. */
3229 if (UCHARAT(s) != nextchr &&
3231 (UCHARAT(s) != ((OP(scan) == REFF
3232 ? PL_fold : PL_fold_locale)[nextchr]))))
3234 st->ln = PL_regendp[n] - st->ln;
3235 if (locinput + st->ln > PL_regeol)
3237 if (st->ln > 1 && (OP(scan) == REF
3238 ? memNE(s, locinput, st->ln)
3240 ? ibcmp(s, locinput, st->ln)
3241 : ibcmp_locale(s, locinput, st->ln))))
3244 nextchr = UCHARAT(locinput);
3257 /* execute the code in the {...} */
3259 SV ** const before = SP;
3260 OP_4tree * const oop = PL_op;
3261 COP * const ocurcop = PL_curcop;
3263 struct regexp * const oreg = PL_reg_re;
3266 PL_op = (OP_4tree*)PL_reg_re->data->data[n];
3267 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
3268 PAD_SAVE_LOCAL(old_comppad, (PAD*)PL_reg_re->data->data[n + 2]);
3269 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
3271 CALLRUNOPS(aTHX); /* Scalar context. */
3274 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
3281 PAD_RESTORE_LOCAL(old_comppad);
3282 PL_curcop = ocurcop;
3286 sv_setsv(save_scalar(PL_replgv), ret);
3290 if (st->logical == 2) { /* Postponed subexpression: /(??{...})/ */
3296 /* extract RE object from returned value; compiling if
3301 if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret)))
3302 mg = mg_find(sv, PERL_MAGIC_qr);
3303 else if (SvSMAGICAL(ret)) {
3304 if (SvGMAGICAL(ret))
3305 sv_unmagic(ret, PERL_MAGIC_qr);
3307 mg = mg_find(ret, PERL_MAGIC_qr);
3311 re = (regexp *)mg->mg_obj;
3312 (void)ReREFCNT_inc(re);
3316 const char * const t = SvPV_const(ret, len);
3318 const I32 osize = PL_regsize;
3321 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
3322 re = CALLREGCOMP(aTHX_ (char*)t, (char*)t + len, &pm);
3324 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
3326 sv_magic(ret,(SV*)ReREFCNT_inc(re),
3332 PerlIO_printf(Perl_debug_log,
3333 "Entering embedded \"%s%.60s%s%s\"\n",
3337 (strlen(re->precomp) > 60 ? "..." : ""))
3340 state.prev = PL_reg_call_cc;
3342 state.re = PL_reg_re;
3346 st->u.eval.cp = regcppush(0); /* Save *all* the positions. */
3347 REGCP_SET(st->u.eval.lastcp);
3349 state.ss = PL_savestack_ix;
3350 *PL_reglastparen = 0;
3351 *PL_reglastcloseparen = 0;
3352 PL_reg_call_cc = &state;
3353 PL_reginput = locinput;
3354 toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
3355 ((re->reganch & ROPT_UTF8) != 0);
3356 if (toggleutf) PL_reg_flags ^= RF_utf8;
3358 /* XXXX This is too dramatic a measure... */
3361 /* XXX the only recursion left in regmatch() */
3362 if (regmatch(re, re->program + 1)) {
3363 /* Even though we succeeded, we need to restore
3364 global variables, since we may be wrapped inside
3365 SUSPEND, thus the match may be not finished yet. */
3367 /* XXXX Do this only if SUSPENDed? */
3368 PL_reg_call_cc = state.prev;
3370 PL_reg_re = state.re;
3372 if (toggleutf) PL_reg_flags ^= RF_utf8;
3374 /* XXXX This is too dramatic a measure... */
3377 /* These are needed even if not SUSPEND. */
3379 regcpblow(st->u.eval.cp);
3383 REGCP_UNWIND(st->u.eval.lastcp);
3385 PL_reg_call_cc = state.prev;
3387 PL_reg_re = state.re;
3388 if (toggleutf) PL_reg_flags ^= RF_utf8;
3390 /* XXXX This is too dramatic a measure... */
3397 /* /(?(?{...})X|Y)/ */
3398 st->sw = SvTRUE(ret);
3403 n = ARG(scan); /* which paren pair */
3404 PL_reg_start_tmp[n] = locinput;
3409 n = ARG(scan); /* which paren pair */
3410 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
3411 PL_regendp[n] = locinput - PL_bostr;
3412 if (n > (I32)*PL_reglastparen)
3413 *PL_reglastparen = n;
3414 *PL_reglastcloseparen = n;
3417 n = ARG(scan); /* which paren pair */
3418 st->sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
3421 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3423 next = NEXTOPER(NEXTOPER(scan));
3425 next = scan + ARG(scan);
3426 if (OP(next) == IFTHEN) /* Fake one. */
3427 next = NEXTOPER(NEXTOPER(next));
3431 st->logical = scan->flags;
3433 /*******************************************************************
3434 cc points to the regmatch_state associated with the most recent CURLYX.
3435 This struct contains info about the innermost (...)* loop (an
3436 "infoblock"), and a pointer to the next outer cc.
3438 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
3440 1) After matching Y, regnode for CURLYX is processed;
3442 2) This regnode populates cc, and calls regmatch() recursively
3443 with the starting point at WHILEM node;
3445 3) Each hit of WHILEM node tries to match A and Z (in the order
3446 depending on the current iteration, min/max of {min,max} and
3447 greediness). The information about where are nodes for "A"
3448 and "Z" is read from cc, as is info on how many times "A"
3449 was already matched, and greediness.
3451 4) After A matches, the same WHILEM node is hit again.
3453 5) Each time WHILEM is hit, cc is the infoblock created by CURLYX
3454 of the same pair. Thus when WHILEM tries to match Z, it temporarily
3455 resets cc, since this Y(A)*Z can be a part of some other loop:
3456 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
3457 of the external loop.
3459 Currently present infoblocks form a tree with a stem formed by st->cc
3460 and whatever it mentions via ->next, and additional attached trees
3461 corresponding to temporarily unset infoblocks as in "5" above.
3463 In the following picture, infoblocks for outer loop of
3464 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
3465 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
3466 infoblocks are drawn below the "reset" infoblock.
3468 In fact in the picture below we do not show failed matches for Z and T
3469 by WHILEM blocks. [We illustrate minimal matches, since for them it is
3470 more obvious *why* one needs to *temporary* unset infoblocks.]
3472 Matched REx position InfoBlocks Comment
3476 Y A)*?Z)*?T x <- O <- I
3477 YA )*?Z)*?T x <- O <- I
3478 YA A)*?Z)*?T x <- O <- I
3479 YAA )*?Z)*?T x <- O <- I
3480 YAA Z)*?T x <- O # Temporary unset I
3483 YAAZ Y(A)*?Z)*?T x <- O
3486 YAAZY (A)*?Z)*?T x <- O
3489 YAAZY A)*?Z)*?T x <- O <- I
3492 YAAZYA )*?Z)*?T x <- O <- I
3495 YAAZYA Z)*?T x <- O # Temporary unset I
3501 YAAZYAZ T x # Temporary unset O
3508 *******************************************************************/
3511 /* No need to save/restore up to this paren */
3512 I32 parenfloor = scan->flags;
3514 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3516 /* XXXX Probably it is better to teach regpush to support
3517 parenfloor > PL_regsize... */
3518 if (parenfloor > (I32)*PL_reglastparen)
3519 parenfloor = *PL_reglastparen; /* Pessimization... */
3521 st->u.curlyx.cp = PL_savestack_ix;
3522 st->u.curlyx.outercc = st->cc;
3524 /* these fields contain the state of the current curly.
3525 * they are accessed by subsequent WHILEMs;
3526 * cur and lastloc are also updated by WHILEM */
3527 st->u.curlyx.parenfloor = parenfloor;
3528 st->u.curlyx.cur = -1; /* this will be updated by WHILEM */
3529 st->u.curlyx.min = ARG1(scan);
3530 st->u.curlyx.max = ARG2(scan);
3531 st->u.curlyx.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3532 st->u.curlyx.lastloc = 0;
3533 /* st->next and st->minmod are also read by WHILEM */
3535 PL_reginput = locinput;
3536 REGMATCH(PREVOPER(next), CURLYX); /* start on the WHILEM */
3537 /*** all unsaved local vars undefined at this point */
3538 regcpblow(st->u.curlyx.cp);
3539 st->cc = st->u.curlyx.outercc;
3545 * This is really hard to understand, because after we match
3546 * what we're trying to match, we must make sure the rest of
3547 * the REx is going to match for sure, and to do that we have
3548 * to go back UP the parse tree by recursing ever deeper. And
3549 * if it fails, we have to reset our parent's current state
3550 * that we can try again after backing off.
3553 st->u.whilem.lastloc = st->cc->u.curlyx.lastloc; /* Detection of 0-len. */
3554 st->u.whilem.cache_offset = 0;
3555 st->u.whilem.cache_bit = 0;
3557 n = st->cc->u.curlyx.cur + 1; /* how many we know we matched */
3558 PL_reginput = locinput;
3561 PerlIO_printf(Perl_debug_log,
3562 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3563 REPORT_CODE_OFF+PL_regindent*2, "",
3564 (long)n, (long)st->cc->u.curlyx.min,
3565 (long)st->cc->u.curlyx.max, PTR2UV(st->cc))
3568 /* If degenerate scan matches "", assume scan done. */
3570 if (locinput == st->cc->u.curlyx.lastloc && n >= st->cc->u.curlyx.min) {
3571 st->u.whilem.savecc = st->cc;
3572 st->cc = st->cc->u.curlyx.outercc;
3574 st->ln = st->cc->u.curlyx.cur;
3576 PerlIO_printf(Perl_debug_log,
3577 "%*s empty match detected, try continuation...\n",
3578 REPORT_CODE_OFF+PL_regindent*2, "")
3580 REGMATCH(st->u.whilem.savecc->next, WHILEM1);
3581 /*** all unsaved local vars undefined at this point */
3582 st->cc = st->u.whilem.savecc;
3585 if (st->cc->u.curlyx.outercc)
3586 st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln;
3590 /* First just match a string of min scans. */
3592 if (n < st->cc->u.curlyx.min) {
3593 st->cc->u.curlyx.cur = n;
3594 st->cc->u.curlyx.lastloc = locinput;
3595 REGMATCH(st->cc->u.curlyx.scan, WHILEM2);
3596 /*** all unsaved local vars undefined at this point */
3599 st->cc->u.curlyx.cur = n - 1;
3600 st->cc->u.curlyx.lastloc = st->u.whilem.lastloc;
3605 /* Check whether we already were at this position.
3606 Postpone detection until we know the match is not
3607 *that* much linear. */
3608 if (!PL_reg_maxiter) {
3609 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3610 PL_reg_leftiter = PL_reg_maxiter;
3612 if (PL_reg_leftiter-- == 0) {
3613 const I32 size = (PL_reg_maxiter + 7 + POSCACHE_START)/8;
3614 if (PL_reg_poscache) {
3615 if ((I32)PL_reg_poscache_size < size) {
3616 Renew(PL_reg_poscache, size, char);
3617 PL_reg_poscache_size = size;
3619 Zero(PL_reg_poscache, size, char);
3622 PL_reg_poscache_size = size;
3623 Newxz(PL_reg_poscache, size, char);
3626 PerlIO_printf(Perl_debug_log,
3627 "%sDetected a super-linear match, switching on caching%s...\n",
3628 PL_colors[4], PL_colors[5])
3631 if (PL_reg_leftiter < 0) {
3632 st->u.whilem.cache_offset = locinput - PL_bostr;
3634 st->u.whilem.cache_offset = (scan->flags & 0xf) - 1 + POSCACHE_START
3635 + st->u.whilem.cache_offset * (scan->flags>>4);
3636 st->u.whilem.cache_bit = st->u.whilem.cache_offset % 8;
3637 st->u.whilem.cache_offset /= 8;
3638 if (PL_reg_poscache[st->u.whilem.cache_offset] & (1<<st->u.whilem.cache_bit)) {
3640 PerlIO_printf(Perl_debug_log,
3641 "%*s already tried at this position...\n",
3642 REPORT_CODE_OFF+PL_regindent*2, "")
3644 if (PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))
3645 /* cache records success */
3648 /* cache records failure */
3651 PL_reg_poscache[st->u.whilem.cache_offset] |= (1<<st->u.whilem.cache_bit);
3655 /* Prefer next over scan for minimal matching. */
3657 if (st->cc->minmod) {
3658 st->u.whilem.savecc = st->cc;
3659 st->cc = st->cc->u.curlyx.outercc;
3661 st->ln = st->cc->u.curlyx.cur;
3662 st->u.whilem.cp = regcppush(st->u.whilem.savecc->u.curlyx.parenfloor);
3663 REGCP_SET(st->u.whilem.lastcp);
3664 REGMATCH(st->u.whilem.savecc->next, WHILEM3);
3665 /*** all unsaved local vars undefined at this point */
3666 st->cc = st->u.whilem.savecc;
3668 regcpblow(st->u.whilem.cp);
3669 CACHEsayYES; /* All done. */
3671 REGCP_UNWIND(st->u.whilem.lastcp);
3673 if (st->cc->u.curlyx.outercc)
3674 st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln;
3676 if (n >= st->cc->u.curlyx.max) { /* Maximum greed exceeded? */
3677 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3678 && !(PL_reg_flags & RF_warned)) {
3679 PL_reg_flags |= RF_warned;
3680 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3681 "Complex regular subexpression recursion",
3688 PerlIO_printf(Perl_debug_log,
3689 "%*s trying longer...\n",
3690 REPORT_CODE_OFF+PL_regindent*2, "")
3692 /* Try scanning more and see if it helps. */
3693 PL_reginput = locinput;
3694 st->cc->u.curlyx.cur = n;
3695 st->cc->u.curlyx.lastloc = locinput;
3696 st->u.whilem.cp = regcppush(st->cc->u.curlyx.parenfloor);
3697 REGCP_SET(st->u.whilem.lastcp);
3698 REGMATCH(st->cc->u.curlyx.scan, WHILEM4);
3699 /*** all unsaved local vars undefined at this point */
3701 regcpblow(st->u.whilem.cp);
3704 REGCP_UNWIND(st->u.whilem.lastcp);
3706 st->cc->u.curlyx.cur = n - 1;
3707 st->cc->u.curlyx.lastloc = st->u.whilem.lastloc;
3711 /* Prefer scan over next for maximal matching. */
3713 if (n < st->cc->u.curlyx.max) { /* More greed allowed? */
3714 st->u.whilem.cp = regcppush(st->cc->u.curlyx.parenfloor);
3715 st->cc->u.curlyx.cur = n;
3716 st->cc->u.curlyx.lastloc = locinput;
3717 REGCP_SET(st->u.whilem.lastcp);
3718 REGMATCH(st->cc->u.curlyx.scan, WHILEM5);
3719 /*** all unsaved local vars undefined at this point */
3721 regcpblow(st->u.whilem.cp);
3724 REGCP_UNWIND(st->u.whilem.lastcp);
3725 regcppop(rex); /* Restore some previous $<digit>s? */
3726 PL_reginput = locinput;
3728 PerlIO_printf(Perl_debug_log,
3729 "%*s failed, try continuation...\n",
3730 REPORT_CODE_OFF+PL_regindent*2, "")
3733 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3734 && !(PL_reg_flags & RF_warned)) {
3735 PL_reg_flags |= RF_warned;
3736 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3737 "Complex regular subexpression recursion",
3741 /* Failed deeper matches of scan, so see if this one works. */
3742 st->u.whilem.savecc = st->cc;
3743 st->cc = st->cc->u.curlyx.outercc;
3745 st->ln = st->cc->u.curlyx.cur;
3746 REGMATCH(st->u.whilem.savecc->next, WHILEM6);
3747 /*** all unsaved local vars undefined at this point */
3748 st->cc = st->u.whilem.savecc;
3751 if (st->cc->u.curlyx.outercc)
3752 st->cc->u.curlyx.outercc->u.curlyx.cur = st->ln;
3753 st->cc->u.curlyx.cur = n - 1;
3754 st->cc->u.curlyx.lastloc = st->u.whilem.lastloc;
3759 next = scan + ARG(scan);
3762 inner = NEXTOPER(NEXTOPER(scan));
3765 inner = NEXTOPER(scan);
3770 if (OP(next) != type) /* No choice. */
3771 next = inner; /* Avoid recursion. */
3773 const I32 lastparen = *PL_reglastparen;
3774 /* Put unwinding data on stack */
3775 const I32 unwind1 = SSNEWt(1,re_unwind_branch_t);
3776 re_unwind_branch_t * const uw = SSPTRt(unwind1,re_unwind_branch_t);
3778 uw->prev = st->unwind;
3779 st->unwind = unwind1;
3780 uw->type = ((type == BRANCH)
3782 : RE_UNWIND_BRANCHJ);
3783 uw->lastparen = lastparen;
3785 uw->locinput = locinput;
3786 uw->nextchr = nextchr;
3788 uw->regindent = ++PL_regindent;
3791 REGCP_SET(uw->lastcp);
3793 /* Now go into the first branch */
3803 st->u.curlym.l = st->u.curlym.matches = 0;
3805 /* We suppose that the next guy does not need
3806 backtracking: in particular, it is of constant non-zero length,
3807 and has no parenths to influence future backrefs. */
3808 st->ln = ARG1(scan); /* min to match */
3809 n = ARG2(scan); /* max to match */
3810 st->u.curlym.paren = scan->flags;
3811 if (st->u.curlym.paren) {
3812 if (st->u.curlym.paren > PL_regsize)
3813 PL_regsize = st->u.curlym.paren;
3814 if (st->u.curlym.paren > (I32)*PL_reglastparen)
3815 *PL_reglastparen = st->u.curlym.paren;
3817 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3818 if (st->u.curlym.paren)
3819 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3820 PL_reginput = locinput;
3821 st->u.curlym.maxwanted = st->minmod ? st->ln : n;
3822 if (st->u.curlym.maxwanted) {
3823 while (PL_reginput < PL_regeol && st->u.curlym.matches < st->u.curlym.maxwanted) {
3824 REGMATCH(scan, CURLYM1);
3825 /*** all unsaved local vars undefined at this point */
3828 /* on first match, determine length, u.curlym.l */
3829 if (!st->u.curlym.matches++) {
3830 if (PL_reg_match_utf8) {
3832 while (s < PL_reginput) {
3838 st->u.curlym.l = PL_reginput - locinput;
3840 if (st->u.curlym.l == 0) {
3841 st->u.curlym.matches = st->u.curlym.maxwanted;
3845 locinput = PL_reginput;
3849 PL_reginput = locinput;
3853 if (st->ln && st->u.curlym.matches < st->ln)
3855 if (HAS_TEXT(next) || JUMPABLE(next)) {
3856 regnode *text_node = next;
3858 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3860 if (! HAS_TEXT(text_node)) st->u.curlym.c1 = st->u.curlym.c2 = -1000;
3862 if (PL_regkind[(U8)OP(text_node)] == REF) {
3863 st->u.curlym.c1 = st->u.curlym.c2 = -1000;
3866 else { st->u.curlym.c1 = (U8)*STRING(text_node); }
3867 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3868 st->u.curlym.c2 = PL_fold[st->u.curlym.c1];
3869 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3870 st->u.curlym.c2 = PL_fold_locale[st->u.curlym.c1];
3872 st->u.curlym.c2 = st->u.curlym.c1;
3876 st->u.curlym.c1 = st->u.curlym.c2 = -1000;
3878 REGCP_SET(st->u.curlym.lastcp);
3879 while (n >= st->ln || (n == REG_INFTY && st->ln > 0)) { /* ln overflow ? */
3880 /* If it could work, try it. */
3881 if (st->u.curlym.c1 == -1000 ||
3882 UCHARAT(PL_reginput) == st->u.curlym.c1 ||
3883 UCHARAT(PL_reginput) == st->u.curlym.c2)
3885 if (st->u.curlym.paren) {
3887 PL_regstartp[st->u.curlym.paren] =
3888 HOPc(PL_reginput, -st->u.curlym.l) - PL_bostr;
3889 PL_regendp[st->u.curlym.paren] = PL_reginput - PL_bostr;
3892 PL_regendp[st->u.curlym.paren] = -1;
3894 REGMATCH(next, CURLYM2);
3895 /*** all unsaved local vars undefined at this point */
3898 REGCP_UNWIND(st->u.curlym.lastcp);
3900 /* Couldn't or didn't -- move forward. */
3901 PL_reginput = locinput;
3902 REGMATCH(scan, CURLYM3);
3903 /*** all unsaved local vars undefined at this point */
3906 locinput = PL_reginput;
3914 PerlIO_printf(Perl_debug_log,
3915 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3916 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3917 (IV) st->u.curlym.matches, (IV)st->u.curlym.l)
3919 if (st->u.curlym.matches >= st->ln) {
3920 if (HAS_TEXT(next) || JUMPABLE(next)) {
3921 regnode *text_node = next;
3923 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3925 if (! HAS_TEXT(text_node)) st->u.curlym.c1 = st->u.curlym.c2 = -1000;
3927 if (PL_regkind[(U8)OP(text_node)] == REF) {
3928 st->u.curlym.c1 = st->u.curlym.c2 = -1000;
3931 else { st->u.curlym.c1 = (U8)*STRING(text_node); }
3933 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3934 st->u.curlym.c2 = PL_fold[st->u.curlym.c1];
3935 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3936 st->u.curlym.c2 = PL_fold_locale[st->u.curlym.c1];
3938 st->u.curlym.c2 = st->u.curlym.c1;
3942 st->u.curlym.c1 = st->u.curlym.c2 = -1000;
3945 REGCP_SET(st->u.curlym.lastcp);
3946 while (st->u.curlym.matches >= st->ln) {
3947 /* If it could work, try it. */
3948 if (st->u.curlym.c1 == -1000 ||
3949 UCHARAT(PL_reginput) == st->u.curlym.c1 ||
3950 UCHARAT(PL_reginput) == st->u.curlym.c2)
3953 PerlIO_printf(Perl_debug_log,
3954 "%*s trying tail with matches=%"IVdf"...\n",
3955 (int)(REPORT_CODE_OFF+PL_regindent*2),
3956 "", (IV)st->u.curlym.matches)
3958 if (st->u.curlym.paren) {
3959 if (st->u.curlym.matches) {
3960 PL_regstartp[st->u.curlym.paren]
3961 = HOPc(PL_reginput, -st->u.curlym.l) - PL_bostr;
3962 PL_regendp[st->u.curlym.paren] = PL_reginput - PL_bostr;
3965 PL_regendp[st->u.curlym.paren] = -1;
3967 REGMATCH(next, CURLYM4);
3968 /*** all unsaved local vars undefined at this point */
3971 REGCP_UNWIND(st->u.curlym.lastcp);
3973 /* Couldn't or didn't -- back up. */
3974 st->u.curlym.matches--;
3975 locinput = HOPc(locinput, -st->u.curlym.l);
3976 PL_reginput = locinput;
3984 st->u.plus.paren = scan->flags; /* Which paren to set */
3985 if (st->u.plus.paren > PL_regsize)
3986 PL_regsize = st->u.plus.paren;
3987 if (st->u.plus.paren > (I32)*PL_reglastparen)
3988 *PL_reglastparen = st->u.plus.paren;
3989 st->ln = ARG1(scan); /* min to match */
3990 n = ARG2(scan); /* max to match */
3991 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3994 st->u.plus.paren = 0;
3995 st->ln = ARG1(scan); /* min to match */
3996 n = ARG2(scan); /* max to match */
3997 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
4002 scan = NEXTOPER(scan);
4003 st->u.plus.paren = 0;
4008 scan = NEXTOPER(scan);
4009 st->u.plus.paren = 0;
4012 * Lookahead to avoid useless match attempts
4013 * when we know what character comes next.
4017 * Used to only do .*x and .*?x, but now it allows
4018 * for )'s, ('s and (?{ ... })'s to be in the way
4019 * of the quantifier and the EXACT-like node. -- japhy
4022 if (HAS_TEXT(next) || JUMPABLE(next)) {
4024 regnode *text_node = next;
4026 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
4028 if (! HAS_TEXT(text_node)) st->u.plus.c1 = st->u.plus.c2 = -1000;
4030 if (PL_regkind[(U8)OP(text_node)] == REF) {
4031 st->u.plus.c1 = st->u.plus.c2 = -1000;
4032 goto assume_ok_easy;
4034 else { s = (U8*)STRING(text_node); }
4037 st->u.plus.c2 = st->u.plus.c1 = *s;
4038 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
4039 st->u.plus.c2 = PL_fold[st->u.plus.c1];
4040 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
4041 st->u.plus.c2 = PL_fold_locale[st->u.plus.c1];
4044 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
4045 STRLEN ulen1, ulen2;
4046 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
4047 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
4049 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
4050 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
4052 st->u.plus.c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0,
4054 st->u.plus.c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0,
4058 st->u.plus.c2 = st->u.plus.c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0,
4065 st->u.plus.c1 = st->u.plus.c2 = -1000;
4067 PL_reginput = locinput;
4070 if (st->ln && regrepeat(scan, st->ln) < st->ln)
4072 locinput = PL_reginput;
4073 REGCP_SET(st->u.plus.lastcp);
4074 if (st->u.plus.c1 != -1000) {
4075 st->u.plus.old = locinput;
4076 st->u.plus.count = 0;
4078 if (n == REG_INFTY) {
4079 st->u.plus.e = PL_regeol - 1;
4081 while (UTF8_IS_CONTINUATION(*(U8*)st->u.plus.e))
4086 for (st->u.plus.e = locinput;
4087 m >0 && st->u.plus.e + UTF8SKIP(st->u.plus.e) <= PL_regeol; m--)
4088 st->u.plus.e += UTF8SKIP(st->u.plus.e);
4091 st->u.plus.e = locinput + n - st->ln;
4092 if (st->u.plus.e >= PL_regeol)
4093 st->u.plus.e = PL_regeol - 1;
4096 /* Find place 'next' could work */
4098 if (st->u.plus.c1 == st->u.plus.c2) {
4099 while (locinput <= st->u.plus.e &&
4100 UCHARAT(locinput) != st->u.plus.c1)
4103 while (locinput <= st->u.plus.e
4104 && UCHARAT(locinput) != st->u.plus.c1
4105 && UCHARAT(locinput) != st->u.plus.c2)
4108 st->u.plus.count = locinput - st->u.plus.old;
4111 if (st->u.plus.c1 == st->u.plus.c2) {
4113 /* count initialised to
4114 * utf8_distance(old, locinput) */
4115 while (locinput <= st->u.plus.e &&
4116 utf8n_to_uvchr((U8*)locinput,
4117 UTF8_MAXBYTES, &len,
4118 uniflags) != (UV)st->u.plus.c1) {
4124 /* count initialised to
4125 * utf8_distance(old, locinput) */
4126 while (locinput <= st->u.plus.e) {
4127 UV c = utf8n_to_uvchr((U8*)locinput,
4128 UTF8_MAXBYTES, &len,
4130 if (c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2)
4137 if (locinput > st->u.plus.e)
4139 /* PL_reginput == old now */
4140 if (locinput != st->u.plus.old) {
4141 st->ln = 1; /* Did some */
4142 if (regrepeat(scan, st->u.plus.count) < st->u.plus.count)
4145 /* PL_reginput == locinput now */
4146 TRYPAREN(st->u.plus.paren, st->ln, locinput, PLUS1);
4147 /*** all unsaved local vars undefined at this point */
4148 PL_reginput = locinput; /* Could be reset... */
4149 REGCP_UNWIND(st->u.plus.lastcp);
4150 /* Couldn't or didn't -- move forward. */
4151 st->u.plus.old = locinput;
4153 locinput += UTF8SKIP(locinput);
4156 st->u.plus.count = 1;
4160 while (n >= st->ln || (n == REG_INFTY && st->ln > 0)) { /* ln overflow ? */
4162 if (st->u.plus.c1 != -1000) {
4164 c = utf8n_to_uvchr((U8*)PL_reginput,
4168 c = UCHARAT(PL_reginput);
4169 /* If it could work, try it. */
4170 if (c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2)
4172 TRYPAREN(st->u.plus.paren, st->ln, PL_reginput, PLUS2);
4173 /*** all unsaved local vars undefined at this point */
4174 REGCP_UNWIND(st->u.plus.lastcp);
4177 /* If it could work, try it. */
4178 else if (st->u.plus.c1 == -1000)
4180 TRYPAREN(st->u.plus.paren, st->ln, PL_reginput, PLUS3);
4181 /*** all unsaved local vars undefined at this point */
4182 REGCP_UNWIND(st->u.plus.lastcp);
4184 /* Couldn't or didn't -- move forward. */
4185 PL_reginput = locinput;
4186 if (regrepeat(scan, 1)) {
4188 locinput = PL_reginput;
4195 n = regrepeat(scan, n);
4196 locinput = PL_reginput;
4197 if (st->ln < n && PL_regkind[(U8)OP(next)] == EOL &&
4198 (OP(next) != MEOL ||
4199 OP(next) == SEOL || OP(next) == EOS))
4201 st->ln = n; /* why back off? */
4202 /* ...because $ and \Z can match before *and* after
4203 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
4204 We should back off by one in this case. */
4205 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
4208 REGCP_SET(st->u.plus.lastcp);
4211 while (n >= st->ln) {
4212 if (st->u.plus.c1 != -1000) {
4214 c = utf8n_to_uvchr((U8*)PL_reginput,
4218 c = UCHARAT(PL_reginput);
4220 /* If it could work, try it. */
4221 if (st->u.plus.c1 == -1000 || c == (UV)st->u.plus.c1 || c == (UV)st->u.plus.c2)
4223 TRYPAREN(st->u.plus.paren, n, PL_reginput, PLUS4);
4224 /*** all unsaved local vars undefined at this point */
4225 REGCP_UNWIND(st->u.plus.lastcp);
4227 /* Couldn't or didn't -- back up. */
4229 PL_reginput = locinput = HOPc(locinput, -1);
4236 if (PL_reg_call_cc) {
4237 st->u.end.cur_call_cc = PL_reg_call_cc;
4238 st->u.end.end_re = PL_reg_re;
4240 /* Save *all* the positions. */
4241 st->u.end.cp = regcppush(0);
4242 REGCP_SET(st->u.end.lastcp);
4244 /* Restore parens of the caller. */
4246 I32 tmp = PL_savestack_ix;
4247 PL_savestack_ix = PL_reg_call_cc->ss;
4249 PL_savestack_ix = tmp;
4252 /* Make position available to the callcc. */
4253 PL_reginput = locinput;
4255 PL_reg_re = PL_reg_call_cc->re;
4256 st->u.end.savecc = st->cc;
4257 st->cc = PL_reg_call_cc->cc;
4258 PL_reg_call_cc = PL_reg_call_cc->prev;
4259 REGMATCH(st->u.end.cur_call_cc->node, END);
4260 /*** all unsaved local vars undefined at this point */
4262 PL_reg_call_cc = st->u.end.cur_call_cc;
4263 regcpblow(st->u.end.cp);
4266 REGCP_UNWIND(st->u.end.lastcp);
4268 PL_reg_call_cc = st->u.end.cur_call_cc;
4269 st->cc = st->u.end.savecc;
4270 PL_reg_re = st->u.end.end_re;
4273 PerlIO_printf(Perl_debug_log,
4274 "%*s continuation failed...\n",
4275 REPORT_CODE_OFF+PL_regindent*2, "")
4279 if (locinput < PL_regtill) {
4280 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
4281 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
4283 (long)(locinput - PL_reg_starttry),
4284 (long)(PL_regtill - PL_reg_starttry),
4286 sayNO_FINAL; /* Cannot match: too short. */
4288 PL_reginput = locinput; /* put where regtry can find it */
4289 sayYES_FINAL; /* Success! */
4291 PL_reginput = locinput; /* put where regtry can find it */
4292 sayYES_LOUD; /* Success! */
4295 PL_reginput = locinput;
4300 char * const s = HOPBACKc(locinput, scan->flags);
4306 PL_reginput = locinput;
4311 char * const s = HOPBACKc(locinput, scan->flags);
4317 PL_reginput = locinput;
4320 REGMATCH(NEXTOPER(NEXTOPER(scan)), IFMATCH);
4321 /*** all unsaved local vars undefined at this point */
4337 if (OP(scan) == SUSPEND) {
4338 locinput = PL_reginput;
4339 nextchr = UCHARAT(locinput);
4344 next = scan + ARG(scan);
4349 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
4350 PTR2UV(scan), OP(scan));
4351 Perl_croak(aTHX_ "regexp memory corruption");
4359 /* simulate recursively calling regmatch(), but without actually
4360 * recursing - ie save the current state on the heap rather than on
4361 * the stack, then re-enter the loop. This avoids complex regexes
4362 * blowing the processor stack */
4366 /* push new state */
4367 regmatch_state *oldst = st;
4371 /* grab the next free state slot */
4373 if (st > &(PL_regmatch_slab->states[PERL_REGMATCH_SLAB_SLOTS-1]))
4374 st = S_push_slab(aTHX);
4375 PL_regmatch_state = st;
4379 oldst->locinput = locinput;
4380 oldst->reg_call_cc = PL_reg_call_cc;
4383 locinput = PL_reginput;
4384 nextchr = UCHARAT(locinput);
4396 * We get here only if there's trouble -- normally "case END" is
4397 * the terminating point.
4399 Perl_croak(aTHX_ "corrupted regexp pointers");
4405 PerlIO_printf(Perl_debug_log,
4406 "%*s %scould match...%s\n",
4407 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5])
4411 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
4412 PL_colors[4], PL_colors[5]));
4423 PerlIO_printf(Perl_debug_log,
4424 "%*s %sfailed...%s\n",
4425 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5])
4431 re_unwind_t * const uw = SSPTRt(st->unwind,re_unwind_t);
4434 case RE_UNWIND_BRANCH:
4435 case RE_UNWIND_BRANCHJ:
4437 re_unwind_branch_t * const uwb = &(uw->branch);
4438 const I32 lastparen = uwb->lastparen;
4440 REGCP_UNWIND(uwb->lastcp);
4441 for (n = *PL_reglastparen; n > lastparen; n--)
4443 *PL_reglastparen = n;
4444 scan = next = uwb->next;
4446 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
4447 ? BRANCH : BRANCHJ) ) { /* Failure */
4448 st->unwind = uwb->prev;
4454 /* Have more choice yet. Reuse the same uwb. */
4455 if ((n = (uwb->type == RE_UNWIND_BRANCH
4456 ? NEXT_OFF(next) : ARG(next))))
4459 next = NULL; /* XXXX Needn't unwinding in this case... */
4461 next = NEXTOPER(scan);
4462 if (uwb->type == RE_UNWIND_BRANCHJ)
4463 next = NEXTOPER(next);
4464 locinput = uwb->locinput;
4465 nextchr = uwb->nextchr;
4467 PL_regindent = uwb->regindent;
4474 Perl_croak(aTHX_ "regexp unwind memory corruption");
4485 /* restore previous state and re-enter */
4487 if (st < &PL_regmatch_slab->states[0]) {
4488 PL_regmatch_slab = PL_regmatch_slab->prev;
4489 st = &PL_regmatch_slab->states[PERL_REGMATCH_SLAB_SLOTS-1];
4491 PL_regmatch_state = st;
4493 PL_reg_call_cc = st->reg_call_cc;
4497 locinput = st->locinput;
4498 nextchr = UCHARAT(locinput);
4500 switch (st->resume_state) {
4502 goto resume_point_TRIE1;
4504 goto resume_point_TRIE2;
4506 goto resume_point_CURLYX;
4507 case resume_WHILEM1:
4508 goto resume_point_WHILEM1;
4509 case resume_WHILEM2:
4510 goto resume_point_WHILEM2;
4511 case resume_WHILEM3:
4512 goto resume_point_WHILEM3;
4513 case resume_WHILEM4:
4514 goto resume_point_WHILEM4;
4515 case resume_WHILEM5:
4516 goto resume_point_WHILEM5;
4517 case resume_WHILEM6:
4518 goto resume_point_WHILEM6;
4519 case resume_CURLYM1:
4520 goto resume_point_CURLYM1;
4521 case resume_CURLYM2:
4522 goto resume_point_CURLYM2;
4523 case resume_CURLYM3:
4524 goto resume_point_CURLYM3;
4525 case resume_CURLYM4:
4526 goto resume_point_CURLYM4;
4527 case resume_IFMATCH:
4528 goto resume_point_IFMATCH;
4530 goto resume_point_PLUS1;
4532 goto resume_point_PLUS2;
4534 goto resume_point_PLUS3;
4536 goto resume_point_PLUS4;
4538 goto resume_point_END;
4540 Perl_croak(aTHX_ "regexp resume memory corruption");
4544 /* restore original high-water mark */
4545 PL_regmatch_slab = orig_slab;
4546 PL_regmatch_state = orig_state;
4548 /* free all slabs above current one */
4549 if (orig_slab->next) {
4550 regmatch_slab *osl, *sl = orig_slab->next;
4551 orig_slab->next = NULL;
4564 - regrepeat - repeatedly match something simple, report how many
4567 * [This routine now assumes that it will only match on things of length 1.
4568 * That was true before, but now we assume scan - reginput is the count,
4569 * rather than incrementing count on every character. [Er, except utf8.]]
4572 S_regrepeat(pTHX_ const regnode *p, I32 max)
4575 register char *scan;
4577 register char *loceol = PL_regeol;
4578 register I32 hardcount = 0;
4579 register bool do_utf8 = PL_reg_match_utf8;
4582 if (max == REG_INFTY)
4584 else if (max < loceol - scan)
4585 loceol = scan + max;
4590 while (scan < loceol && hardcount < max && *scan != '\n') {
4591 scan += UTF8SKIP(scan);
4595 while (scan < loceol && *scan != '\n')
4602 while (scan < loceol && hardcount < max) {
4603 scan += UTF8SKIP(scan);
4613 case EXACT: /* length of string is 1 */
4615 while (scan < loceol && UCHARAT(scan) == c)
4618 case EXACTF: /* length of string is 1 */
4620 while (scan < loceol &&
4621 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
4624 case EXACTFL: /* length of string is 1 */
4625 PL_reg_flags |= RF_tainted;
4627 while (scan < loceol &&
4628 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
4634 while (hardcount < max && scan < loceol &&
4635 reginclass(p, (U8*)scan, 0, do_utf8)) {
4636 scan += UTF8SKIP(scan);
4640 while (scan < loceol && REGINCLASS(p, (U8*)scan))
4647 LOAD_UTF8_CHARCLASS_ALNUM();
4648 while (hardcount < max && scan < loceol &&
4649 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4650 scan += UTF8SKIP(scan);
4654 while (scan < loceol && isALNUM(*scan))
4659 PL_reg_flags |= RF_tainted;
4662 while (hardcount < max && scan < loceol &&
4663 isALNUM_LC_utf8((U8*)scan)) {
4664 scan += UTF8SKIP(scan);
4668 while (scan < loceol && isALNUM_LC(*scan))
4675 LOAD_UTF8_CHARCLASS_ALNUM();
4676 while (hardcount < max && scan < loceol &&
4677 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4678 scan += UTF8SKIP(scan);
4682 while (scan < loceol && !isALNUM(*scan))
4687 PL_reg_flags |= RF_tainted;
4690 while (hardcount < max && scan < loceol &&
4691 !isALNUM_LC_utf8((U8*)scan)) {
4692 scan += UTF8SKIP(scan);
4696 while (scan < loceol && !isALNUM_LC(*scan))
4703 LOAD_UTF8_CHARCLASS_SPACE();
4704 while (hardcount < max && scan < loceol &&
4706 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4707 scan += UTF8SKIP(scan);
4711 while (scan < loceol && isSPACE(*scan))
4716 PL_reg_flags |= RF_tainted;
4719 while (hardcount < max && scan < loceol &&
4720 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4721 scan += UTF8SKIP(scan);
4725 while (scan < loceol && isSPACE_LC(*scan))
4732 LOAD_UTF8_CHARCLASS_SPACE();
4733 while (hardcount < max && scan < loceol &&
4735 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4736 scan += UTF8SKIP(scan);
4740 while (scan < loceol && !isSPACE(*scan))
4745 PL_reg_flags |= RF_tainted;
4748 while (hardcount < max && scan < loceol &&
4749 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4750 scan += UTF8SKIP(scan);
4754 while (scan < loceol && !isSPACE_LC(*scan))
4761 LOAD_UTF8_CHARCLASS_DIGIT();
4762 while (hardcount < max && scan < loceol &&
4763 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4764 scan += UTF8SKIP(scan);
4768 while (scan < loceol && isDIGIT(*scan))
4775 LOAD_UTF8_CHARCLASS_DIGIT();
4776 while (hardcount < max && scan < loceol &&
4777 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4778 scan += UTF8SKIP(scan);
4782 while (scan < loceol && !isDIGIT(*scan))
4786 default: /* Called on something of 0 width. */
4787 break; /* So match right here or not at all. */
4793 c = scan - PL_reginput;
4797 SV *re_debug_flags = NULL;
4798 SV * const prop = sv_newmortal();
4802 PerlIO_printf(Perl_debug_log,
4803 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4804 REPORT_CODE_OFF+1, "", SvPVX_const(prop),(IV)c,(IV)max);
4813 - regclass_swash - prepare the utf8 swash
4817 Perl_regclass_swash(pTHX_ register const regnode* node, bool doinit, SV** listsvp, SV **altsvp)
4823 const struct reg_data *data = PL_reg_re ? PL_reg_re->data : NULL;
4825 if (data && data->count) {
4826 const U32 n = ARG(node);
4828 if (data->what[n] == 's') {
4829 SV * const rv = (SV*)data->data[n];
4830 AV * const av = (AV*)SvRV((SV*)rv);
4831 SV **const ary = AvARRAY(av);
4834 /* See the end of regcomp.c:S_regclass() for
4835 * documentation of these array elements. */
4838 a = SvROK(ary[1]) ? &ary[1] : 0;
4839 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
4843 else if (si && doinit) {
4844 sw = swash_init("utf8", "", si, 1, 0);
4845 (void)av_store(av, 1, sw);
4861 - reginclass - determine if a character falls into a character class
4863 The n is the ANYOF regnode, the p is the target string, lenp
4864 is pointer to the maximum length of how far to go in the p
4865 (if the lenp is zero, UTF8SKIP(p) is used),
4866 do_utf8 tells whether the target string is in UTF-8.
4871 S_reginclass(pTHX_ register const regnode *n, register const U8* p, STRLEN* lenp, register bool do_utf8)
4874 const char flags = ANYOF_FLAGS(n);
4880 if (do_utf8 && !UTF8_IS_INVARIANT(c)) {
4881 c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &len,
4882 (UTF8_ALLOW_DEFAULT & UTF8_ALLOW_ANYUV) | UTF8_CHECK_ONLY);
4883 /* see [perl #37836] for UTF8_ALLOW_ANYUV */
4884 if (len == (STRLEN)-1)
4885 Perl_croak(aTHX_ "Malformed UTF-8 character (fatal)");
4888 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
4889 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4892 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4893 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4896 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4900 SV * const sw = regclass_swash(n, TRUE, 0, (SV**)&av);
4903 if (swash_fetch(sw, p, do_utf8))
4905 else if (flags & ANYOF_FOLD) {
4906 if (!match && lenp && av) {
4908 for (i = 0; i <= av_len(av); i++) {
4909 SV* const sv = *av_fetch(av, i, FALSE);
4911 const char * const s = SvPV_const(sv, len);
4913 if (len <= plen && memEQ(s, (char*)p, len)) {
4921 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
4924 to_utf8_fold(p, tmpbuf, &tmplen);
4925 if (swash_fetch(sw, tmpbuf, do_utf8))
4931 if (match && lenp && *lenp == 0)
4932 *lenp = UNISKIP(NATIVE_TO_UNI(c));
4934 if (!match && c < 256) {
4935 if (ANYOF_BITMAP_TEST(n, c))
4937 else if (flags & ANYOF_FOLD) {
4940 if (flags & ANYOF_LOCALE) {
4941 PL_reg_flags |= RF_tainted;
4942 f = PL_fold_locale[c];
4946 if (f != c && ANYOF_BITMAP_TEST(n, f))
4950 if (!match && (flags & ANYOF_CLASS)) {
4951 PL_reg_flags |= RF_tainted;
4953 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4954 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4955 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4956 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4957 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4958 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4959 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4960 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4961 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4962 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4963 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4964 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4965 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4966 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4967 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4968 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4969 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4970 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4971 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4972 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4973 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4974 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4975 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4976 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4977 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4978 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4979 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4980 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4981 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4982 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4983 ) /* How's that for a conditional? */
4990 return (flags & ANYOF_INVERT) ? !match : match;
4994 S_reghop3(U8 *s, I32 off, U8* lim)
4998 while (off-- && s < lim) {
4999 /* XXX could check well-formedness here */
5007 if (UTF8_IS_CONTINUED(*s)) {
5008 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
5011 /* XXX could check well-formedness here */
5019 S_reghopmaybe3(U8* s, I32 off, U8* lim)
5023 while (off-- && s < lim) {
5024 /* XXX could check well-formedness here */
5034 if (UTF8_IS_CONTINUED(*s)) {
5035 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
5038 /* XXX could check well-formedness here */
5050 restore_pos(pTHX_ void *arg)
5053 regexp * const rex = (regexp *)arg;
5054 if (PL_reg_eval_set) {
5055 if (PL_reg_oldsaved) {
5056 rex->subbeg = PL_reg_oldsaved;
5057 rex->sublen = PL_reg_oldsavedlen;
5058 #ifdef PERL_OLD_COPY_ON_WRITE
5059 rex->saved_copy = PL_nrs;
5061 RX_MATCH_COPIED_on(rex);
5063 PL_reg_magic->mg_len = PL_reg_oldpos;
5064 PL_reg_eval_set = 0;
5065 PL_curpm = PL_reg_oldcurpm;
5070 S_to_utf8_substr(pTHX_ register regexp *prog)
5072 if (prog->float_substr && !prog->float_utf8) {
5073 SV* const sv = newSVsv(prog->float_substr);
5074 prog->float_utf8 = sv;
5075 sv_utf8_upgrade(sv);
5076 if (SvTAIL(prog->float_substr))
5078 if (prog->float_substr == prog->check_substr)
5079 prog->check_utf8 = sv;
5081 if (prog->anchored_substr && !prog->anchored_utf8) {
5082 SV* const sv = newSVsv(prog->anchored_substr);
5083 prog->anchored_utf8 = sv;
5084 sv_utf8_upgrade(sv);
5085 if (SvTAIL(prog->anchored_substr))
5087 if (prog->anchored_substr == prog->check_substr)
5088 prog->check_utf8 = sv;
5093 S_to_byte_substr(pTHX_ register regexp *prog)
5096 if (prog->float_utf8 && !prog->float_substr) {
5097 SV* sv = newSVsv(prog->float_utf8);
5098 prog->float_substr = sv;
5099 if (sv_utf8_downgrade(sv, TRUE)) {
5100 if (SvTAIL(prog->float_utf8))
5104 prog->float_substr = sv = &PL_sv_undef;
5106 if (prog->float_utf8 == prog->check_utf8)
5107 prog->check_substr = sv;
5109 if (prog->anchored_utf8 && !prog->anchored_substr) {
5110 SV* sv = newSVsv(prog->anchored_utf8);
5111 prog->anchored_substr = sv;
5112 if (sv_utf8_downgrade(sv, TRUE)) {
5113 if (SvTAIL(prog->anchored_utf8))
5117 prog->anchored_substr = sv = &PL_sv_undef;
5119 if (prog->anchored_utf8 == prog->check_utf8)
5120 prog->check_substr = sv;
5126 * c-indentation-style: bsd
5128 * indent-tabs-mode: t
5131 * ex: set ts=8 sts=4 sw=4 noet: