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
60 * pregcomp and pregexec -- regsub and regerror are not used in perl
62 * Copyright (c) 1986 by University of Toronto.
63 * Written by Henry Spencer. Not derived from licensed software.
65 * Permission is granted to anyone to use this software for any
66 * purpose on any computer system, and to redistribute it freely,
67 * subject to the following restrictions:
69 * 1. The author is not responsible for the consequences of use of
70 * this software, no matter how awful, even if they arise
73 * 2. The origin of this software must not be misrepresented, either
74 * by explicit claim or by omission.
76 * 3. Altered versions must be plainly marked as such, and must not
77 * be misrepresented as being the original software.
79 **** Alterations to Henry's code are...
81 **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
82 **** 2000, 2001, 2002, 2003, 2004, 2005, by Larry Wall and others
84 **** You may distribute under the terms of either the GNU General Public
85 **** License or the Artistic License, as specified in the README file.
87 * Beware that some of this code is subtly aware of the way operator
88 * precedence is structured in regular expressions. Serious changes in
89 * regular-expression syntax might require a total rethink.
92 #define PERL_IN_REGEXEC_C
97 #define RF_tainted 1 /* tainted information used? */
98 #define RF_warned 2 /* warned about big count? */
99 #define RF_evaled 4 /* Did an EVAL with setting? */
100 #define RF_utf8 8 /* String contains multibyte chars? */
101 #define RF_false 16 /* odd number of nested negatives */
103 #define UTF ((PL_reg_flags & RF_utf8) != 0)
105 #define RS_init 1 /* eval environment created */
106 #define RS_set 2 /* replsv value is set */
109 #define STATIC static
112 #define REGINCLASS(p,c) (ANYOF_FLAGS(p) ? reginclass(p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c)))
118 #define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv))
119 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
121 #define reghop_c(pos,off) ((char*)reghop((U8*)pos, off))
122 #define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off))
123 #define HOP(pos,off) (PL_reg_match_utf8 ? reghop((U8*)pos, off) : (U8*)(pos + off))
124 #define HOPMAYBE(pos,off) (PL_reg_match_utf8 ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off))
125 #define HOPc(pos,off) ((char*)HOP(pos,off))
126 #define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off))
128 #define HOPBACK(pos, off) ( \
129 (PL_reg_match_utf8) \
130 ? reghopmaybe((U8*)pos, -off) \
131 : (pos - off >= PL_bostr) \
135 #define HOPBACKc(pos, off) (char*)HOPBACK(pos, off)
137 #define reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim))
138 #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
139 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
140 #define HOPMAYBE3(pos,off,lim) (PL_reg_match_utf8 ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
141 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
142 #define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim))
144 #define LOAD_UTF8_CHARCLASS(a,b) STMT_START { if (!CAT2(PL_utf8_,a)) { ENTER; save_re_context(); (void)CAT2(is_utf8_, a)((const U8*)b); LEAVE; } } STMT_END
146 /* for use after a quantifier and before an EXACT-like node -- japhy */
147 #define JUMPABLE(rn) ( \
148 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
149 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
150 OP(rn) == PLUS || OP(rn) == MINMOD || \
151 (PL_regkind[(U8)OP(rn)] == CURLY && ARG1(rn) > 0) \
154 #define HAS_TEXT(rn) ( \
155 PL_regkind[(U8)OP(rn)] == EXACT || PL_regkind[(U8)OP(rn)] == REF \
159 Search for mandatory following text node; for lookahead, the text must
160 follow but for lookbehind (rn->flags != 0) we skip to the next step.
162 #define FIND_NEXT_IMPT(rn) STMT_START { \
163 while (JUMPABLE(rn)) \
164 if (OP(rn) == SUSPEND || PL_regkind[(U8)OP(rn)] == CURLY) \
165 rn = NEXTOPER(NEXTOPER(rn)); \
166 else if (OP(rn) == PLUS) \
168 else if (OP(rn) == IFMATCH) \
169 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
170 else rn += NEXT_OFF(rn); \
173 static void restore_pos(pTHX_ void *arg);
176 S_regcppush(pTHX_ I32 parenfloor)
178 int retval = PL_savestack_ix;
179 #define REGCP_PAREN_ELEMS 4
180 int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
183 if (paren_elems_to_push < 0)
184 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
186 #define REGCP_OTHER_ELEMS 6
187 SSGROW(paren_elems_to_push + REGCP_OTHER_ELEMS);
188 for (p = PL_regsize; p > parenfloor; p--) {
189 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
190 SSPUSHINT(PL_regendp[p]);
191 SSPUSHINT(PL_regstartp[p]);
192 SSPUSHPTR(PL_reg_start_tmp[p]);
195 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
196 SSPUSHINT(PL_regsize);
197 SSPUSHINT(*PL_reglastparen);
198 SSPUSHINT(*PL_reglastcloseparen);
199 SSPUSHPTR(PL_reginput);
200 #define REGCP_FRAME_ELEMS 2
201 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
202 * are needed for the regexp context stack bookkeeping. */
203 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
204 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
209 /* These are needed since we do not localize EVAL nodes: */
210 # define REGCP_SET(cp) DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, \
211 " Setting an EVAL scope, savestack=%"IVdf"\n", \
212 (IV)PL_savestack_ix)); cp = PL_savestack_ix
214 # define REGCP_UNWIND(cp) DEBUG_EXECUTE_r(cp != PL_savestack_ix ? \
215 PerlIO_printf(Perl_debug_log, \
216 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
217 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
227 GET_RE_DEBUG_FLAGS_DECL;
229 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
231 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
232 i = SSPOPINT; /* Parentheses elements to pop. */
233 input = (char *) SSPOPPTR;
234 *PL_reglastcloseparen = SSPOPINT;
235 *PL_reglastparen = SSPOPINT;
236 PL_regsize = SSPOPINT;
238 /* Now restore the parentheses context. */
239 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
240 i > 0; i -= REGCP_PAREN_ELEMS) {
241 paren = (U32)SSPOPINT;
242 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
243 PL_regstartp[paren] = SSPOPINT;
245 if (paren <= *PL_reglastparen)
246 PL_regendp[paren] = tmps;
248 PerlIO_printf(Perl_debug_log,
249 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
250 (UV)paren, (IV)PL_regstartp[paren],
251 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
252 (IV)PL_regendp[paren],
253 (paren > *PL_reglastparen ? "(no)" : ""));
257 if ((I32)(*PL_reglastparen + 1) <= PL_regnpar) {
258 PerlIO_printf(Perl_debug_log,
259 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
260 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
264 /* It would seem that the similar code in regtry()
265 * already takes care of this, and in fact it is in
266 * a better location to since this code can #if 0-ed out
267 * but the code in regtry() is needed or otherwise tests
268 * requiring null fields (pat.t#187 and split.t#{13,14}
269 * (as of patchlevel 7877) will fail. Then again,
270 * this code seems to be necessary or otherwise
271 * building DynaLoader will fail:
272 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
274 for (paren = *PL_reglastparen + 1; (I32)paren <= PL_regnpar; paren++) {
275 if ((I32)paren > PL_regsize)
276 PL_regstartp[paren] = -1;
277 PL_regendp[paren] = -1;
284 S_regcp_set_to(pTHX_ I32 ss)
286 I32 tmp = PL_savestack_ix;
288 PL_savestack_ix = ss;
290 PL_savestack_ix = tmp;
294 typedef struct re_cc_state
298 struct re_cc_state *prev;
303 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
305 #define TRYPAREN(paren, n, input) { \
308 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
309 PL_regendp[paren] = input - PL_bostr; \
312 PL_regendp[paren] = -1; \
314 if (regmatch(next)) \
317 PL_regendp[paren] = -1; \
322 * pregexec and friends
326 - pregexec - match a regexp against a string
329 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
330 char *strbeg, I32 minend, SV *screamer, U32 nosave)
331 /* strend: pointer to null at end of string */
332 /* strbeg: real beginning of string */
333 /* minend: end of match must be >=minend after stringarg. */
334 /* nosave: For optimizations. */
337 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
338 nosave ? 0 : REXEC_COPY_STR);
342 S_cache_re(pTHX_ regexp *prog)
344 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
346 PL_regprogram = prog->program;
348 PL_regnpar = prog->nparens;
349 PL_regdata = prog->data;
354 * Need to implement the following flags for reg_anch:
356 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
358 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
359 * INTUIT_AUTORITATIVE_ML
360 * INTUIT_ONCE_NOML - Intuit can match in one location only.
363 * Another flag for this function: SECOND_TIME (so that float substrs
364 * with giant delta may be not rechecked).
367 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
369 /* If SCREAM, then SvPVX(sv) should be compatible with strpos and strend.
370 Otherwise, only SvCUR(sv) is used to get strbeg. */
372 /* XXXX We assume that strpos is strbeg unless sv. */
374 /* XXXX Some places assume that there is a fixed substring.
375 An update may be needed if optimizer marks as "INTUITable"
376 RExen without fixed substrings. Similarly, it is assumed that
377 lengths of all the strings are no more than minlen, thus they
378 cannot come from lookahead.
379 (Or minlen should take into account lookahead.) */
381 /* A failure to find a constant substring means that there is no need to make
382 an expensive call to REx engine, thus we celebrate a failure. Similarly,
383 finding a substring too deep into the string means that less calls to
384 regtry() should be needed.
386 REx compiler's optimizer found 4 possible hints:
387 a) Anchored substring;
389 c) Whether we are anchored (beginning-of-line or \G);
390 d) First node (of those at offset 0) which may distingush positions;
391 We use a)b)d) and multiline-part of c), and try to find a position in the
392 string which does not contradict any of them.
395 /* Most of decisions we do here should have been done at compile time.
396 The nodes of the REx which we used for the search should have been
397 deleted from the finite automaton. */
400 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
401 char *strend, U32 flags, re_scream_pos_data *data)
403 register I32 start_shift = 0;
404 /* Should be nonnegative! */
405 register I32 end_shift = 0;
410 int do_utf8 = sv ? SvUTF8(sv) : 0; /* if no sv we have to assume bytes */
412 register char *other_last = Nullch; /* other substr checked before this */
413 char *check_at = Nullch; /* check substr found at this pos */
414 I32 multiline = prog->reganch & PMf_MULTILINE;
416 char *i_strpos = strpos;
417 SV *dsv = PERL_DEBUG_PAD_ZERO(0);
420 GET_RE_DEBUG_FLAGS_DECL;
422 RX_MATCH_UTF8_set(prog,do_utf8);
424 if (prog->reganch & ROPT_UTF8) {
425 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
426 "UTF-8 regex...\n"));
427 PL_reg_flags |= RF_utf8;
431 char *s = PL_reg_match_utf8 ?
432 sv_uni_display(dsv, sv, 60, UNI_DISPLAY_REGEX) :
434 int len = PL_reg_match_utf8 ?
435 strlen(s) : strend - strpos;
438 if (PL_reg_match_utf8)
439 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
440 "UTF-8 target...\n"));
441 PerlIO_printf(Perl_debug_log,
442 "%sGuessing start of match, REx%s `%s%.60s%s%s' against `%s%.*s%s%s'...\n",
443 PL_colors[4],PL_colors[5],PL_colors[0],
446 (strlen(prog->precomp) > 60 ? "..." : ""),
448 (int)(len > 60 ? 60 : len),
450 (len > 60 ? "..." : "")
454 /* CHR_DIST() would be more correct here but it makes things slow. */
455 if (prog->minlen > strend - strpos) {
456 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
457 "String too short... [re_intuit_start]\n"));
460 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
463 if (!prog->check_utf8 && prog->check_substr)
464 to_utf8_substr(prog);
465 check = prog->check_utf8;
467 if (!prog->check_substr && prog->check_utf8)
468 to_byte_substr(prog);
469 check = prog->check_substr;
471 if (check == &PL_sv_undef) {
472 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
473 "Non-utf string cannot match utf check string\n"));
476 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
477 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
478 || ( (prog->reganch & ROPT_ANCH_BOL)
479 && !multiline ) ); /* Check after \n? */
482 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
483 | ROPT_IMPLICIT)) /* not a real BOL */
484 /* SvCUR is not set on references: SvRV and SvPVX overlap */
486 && (strpos != strbeg)) {
487 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
490 if (prog->check_offset_min == prog->check_offset_max &&
491 !(prog->reganch & ROPT_CANY_SEEN)) {
492 /* Substring at constant offset from beg-of-str... */
495 s = HOP3c(strpos, prog->check_offset_min, strend);
497 slen = SvCUR(check); /* >= 1 */
499 if ( strend - s > slen || strend - s < slen - 1
500 || (strend - s == slen && strend[-1] != '\n')) {
501 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
504 /* Now should match s[0..slen-2] */
506 if (slen && (*SvPVX(check) != *s
508 && memNE(SvPVX(check), s, slen)))) {
510 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
514 else if (*SvPVX(check) != *s
515 || ((slen = SvCUR(check)) > 1
516 && memNE(SvPVX(check), s, slen)))
518 goto success_at_start;
521 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
523 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
524 end_shift = prog->minlen - start_shift -
525 CHR_SVLEN(check) + (SvTAIL(check) != 0);
527 I32 end = prog->check_offset_max + CHR_SVLEN(check)
528 - (SvTAIL(check) != 0);
529 I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
531 if (end_shift < eshift)
535 else { /* Can match at random position */
538 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
539 /* Should be nonnegative! */
540 end_shift = prog->minlen - start_shift -
541 CHR_SVLEN(check) + (SvTAIL(check) != 0);
544 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
546 Perl_croak(aTHX_ "panic: end_shift");
550 /* Find a possible match in the region s..strend by looking for
551 the "check" substring in the region corrected by start/end_shift. */
552 if (flags & REXEC_SCREAM) {
553 I32 p = -1; /* Internal iterator of scream. */
554 I32 *pp = data ? data->scream_pos : &p;
556 if (PL_screamfirst[BmRARE(check)] >= 0
557 || ( BmRARE(check) == '\n'
558 && (BmPREVIOUS(check) == SvCUR(check) - 1)
560 s = screaminstr(sv, check,
561 start_shift + (s - strbeg), end_shift, pp, 0);
564 /* we may be pointing at the wrong string */
565 if (s && RX_MATCH_COPIED(prog))
566 s = strbeg + (s - SvPVX(sv));
568 *data->scream_olds = s;
570 else if (prog->reganch & ROPT_CANY_SEEN)
571 s = fbm_instr((U8*)(s + start_shift),
572 (U8*)(strend - end_shift),
573 check, multiline ? FBMrf_MULTILINE : 0);
575 s = fbm_instr(HOP3(s, start_shift, strend),
576 HOP3(strend, -end_shift, strbeg),
577 check, multiline ? FBMrf_MULTILINE : 0);
579 /* Update the count-of-usability, remove useless subpatterns,
582 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s %s substr `%s%.*s%s'%s%s",
583 (s ? "Found" : "Did not find"),
584 (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) ? "anchored" : "floating"),
586 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
588 PL_colors[1], (SvTAIL(check) ? "$" : ""),
589 (s ? " at offset " : "...\n") ) );
596 /* Finish the diagnostic message */
597 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
599 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
600 Start with the other substr.
601 XXXX no SCREAM optimization yet - and a very coarse implementation
602 XXXX /ttx+/ results in anchored=`ttx', floating=`x'. floating will
603 *always* match. Probably should be marked during compile...
604 Probably it is right to do no SCREAM here...
607 if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8) : (prog->float_substr && prog->anchored_substr)) {
608 /* Take into account the "other" substring. */
609 /* XXXX May be hopelessly wrong for UTF... */
612 if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
615 char *last = HOP3c(s, -start_shift, strbeg), *last1, *last2;
619 t = s - prog->check_offset_max;
620 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
622 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
627 t = HOP3c(t, prog->anchored_offset, strend);
628 if (t < other_last) /* These positions already checked */
630 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
633 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
634 /* On end-of-str: see comment below. */
635 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
636 if (must == &PL_sv_undef) {
638 DEBUG_EXECUTE_r(must = prog->anchored_utf8); /* for debug */
643 HOP3(HOP3(last1, prog->anchored_offset, strend)
644 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
646 multiline ? FBMrf_MULTILINE : 0
648 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
649 "%s anchored substr `%s%.*s%s'%s",
650 (s ? "Found" : "Contradicts"),
653 - (SvTAIL(must)!=0)),
655 PL_colors[1], (SvTAIL(must) ? "$" : "")));
657 if (last1 >= last2) {
658 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
659 ", giving up...\n"));
662 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
663 ", trying floating at offset %ld...\n",
664 (long)(HOP3c(s1, 1, strend) - i_strpos)));
665 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
666 s = HOP3c(last, 1, strend);
670 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
671 (long)(s - i_strpos)));
672 t = HOP3c(s, -prog->anchored_offset, strbeg);
673 other_last = HOP3c(s, 1, strend);
681 else { /* Take into account the floating substring. */
686 t = HOP3c(s, -start_shift, strbeg);
688 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
689 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
690 last = HOP3c(t, prog->float_max_offset, strend);
691 s = HOP3c(t, prog->float_min_offset, strend);
694 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
695 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
696 /* fbm_instr() takes into account exact value of end-of-str
697 if the check is SvTAIL(ed). Since false positives are OK,
698 and end-of-str is not later than strend we are OK. */
699 if (must == &PL_sv_undef) {
701 DEBUG_EXECUTE_r(must = prog->float_utf8); /* for debug message */
704 s = fbm_instr((unsigned char*)s,
705 (unsigned char*)last + SvCUR(must)
707 must, multiline ? FBMrf_MULTILINE : 0);
708 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s floating substr `%s%.*s%s'%s",
709 (s ? "Found" : "Contradicts"),
711 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
713 PL_colors[1], (SvTAIL(must) ? "$" : "")));
716 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
717 ", giving up...\n"));
720 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
721 ", trying anchored starting at offset %ld...\n",
722 (long)(s1 + 1 - i_strpos)));
724 s = HOP3c(t, 1, strend);
728 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
729 (long)(s - i_strpos)));
730 other_last = s; /* Fix this later. --Hugo */
739 t = s - prog->check_offset_max;
740 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
742 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
744 /* Fixed substring is found far enough so that the match
745 cannot start at strpos. */
747 if (ml_anch && t[-1] != '\n') {
748 /* Eventually fbm_*() should handle this, but often
749 anchored_offset is not 0, so this check will not be wasted. */
750 /* XXXX In the code below we prefer to look for "^" even in
751 presence of anchored substrings. And we search even
752 beyond the found float position. These pessimizations
753 are historical artefacts only. */
755 while (t < strend - prog->minlen) {
757 if (t < check_at - prog->check_offset_min) {
758 if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
759 /* Since we moved from the found position,
760 we definitely contradict the found anchored
761 substr. Due to the above check we do not
762 contradict "check" substr.
763 Thus we can arrive here only if check substr
764 is float. Redo checking for "other"=="fixed".
767 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
768 PL_colors[0],PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
769 goto do_other_anchored;
771 /* We don't contradict the found floating substring. */
772 /* XXXX Why not check for STCLASS? */
774 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
775 PL_colors[0],PL_colors[1], (long)(s - i_strpos)));
778 /* Position contradicts check-string */
779 /* XXXX probably better to look for check-string
780 than for "\n", so one should lower the limit for t? */
781 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
782 PL_colors[0],PL_colors[1], (long)(t + 1 - i_strpos)));
783 other_last = strpos = s = t + 1;
788 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
789 PL_colors[0],PL_colors[1]));
793 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
794 PL_colors[0],PL_colors[1]));
798 ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
801 /* The found string does not prohibit matching at strpos,
802 - no optimization of calling REx engine can be performed,
803 unless it was an MBOL and we are not after MBOL,
804 or a future STCLASS check will fail this. */
806 /* Even in this situation we may use MBOL flag if strpos is offset
807 wrt the start of the string. */
808 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
809 && (strpos != strbeg) && strpos[-1] != '\n'
810 /* May be due to an implicit anchor of m{.*foo} */
811 && !(prog->reganch & ROPT_IMPLICIT))
816 DEBUG_EXECUTE_r( if (ml_anch)
817 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
818 (long)(strpos - i_strpos), PL_colors[0],PL_colors[1]);
821 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
823 prog->check_utf8 /* Could be deleted already */
824 && --BmUSEFUL(prog->check_utf8) < 0
825 && (prog->check_utf8 == prog->float_utf8)
827 prog->check_substr /* Could be deleted already */
828 && --BmUSEFUL(prog->check_substr) < 0
829 && (prog->check_substr == prog->float_substr)
832 /* If flags & SOMETHING - do not do it many times on the same match */
833 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
834 SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr);
835 if (do_utf8 ? prog->check_substr : prog->check_utf8)
836 SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8);
837 prog->check_substr = prog->check_utf8 = Nullsv; /* disable */
838 prog->float_substr = prog->float_utf8 = Nullsv; /* clear */
839 check = Nullsv; /* abort */
841 /* XXXX This is a remnant of the old implementation. It
842 looks wasteful, since now INTUIT can use many
844 prog->reganch &= ~RE_USE_INTUIT;
851 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
852 if (prog->regstclass) {
853 /* minlen == 0 is possible if regstclass is \b or \B,
854 and the fixed substr is ''$.
855 Since minlen is already taken into account, s+1 is before strend;
856 accidentally, minlen >= 1 guaranties no false positives at s + 1
857 even for \b or \B. But (minlen? 1 : 0) below assumes that
858 regstclass does not come from lookahead... */
859 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
860 This leaves EXACTF only, which is dealt with in find_byclass(). */
861 U8* str = (U8*)STRING(prog->regstclass);
862 int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
863 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
865 char *endpos = (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
866 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
867 : (prog->float_substr || prog->float_utf8
868 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
871 char *startpos = strbeg;
875 s = find_byclass(prog, prog->regstclass, s, endpos, startpos, 1);
878 const char *what = 0;
880 if (endpos == strend) {
881 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
882 "Could not match STCLASS...\n") );
885 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
886 "This position contradicts STCLASS...\n") );
887 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
889 /* Contradict one of substrings */
890 if (prog->anchored_substr || prog->anchored_utf8) {
891 if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
892 DEBUG_EXECUTE_r( what = "anchored" );
894 s = HOP3c(t, 1, strend);
895 if (s + start_shift + end_shift > strend) {
896 /* XXXX Should be taken into account earlier? */
897 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
898 "Could not match STCLASS...\n") );
903 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
904 "Looking for %s substr starting at offset %ld...\n",
905 what, (long)(s + start_shift - i_strpos)) );
908 /* Have both, check_string is floating */
909 if (t + start_shift >= check_at) /* Contradicts floating=check */
910 goto retry_floating_check;
911 /* Recheck anchored substring, but not floating... */
915 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
916 "Looking for anchored substr starting at offset %ld...\n",
917 (long)(other_last - i_strpos)) );
918 goto do_other_anchored;
920 /* Another way we could have checked stclass at the
921 current position only: */
926 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
927 "Looking for /%s^%s/m starting at offset %ld...\n",
928 PL_colors[0],PL_colors[1], (long)(t - i_strpos)) );
931 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
933 /* Check is floating subtring. */
934 retry_floating_check:
935 t = check_at - start_shift;
936 DEBUG_EXECUTE_r( what = "floating" );
937 goto hop_and_restart;
940 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
941 "By STCLASS: moving %ld --> %ld\n",
942 (long)(t - i_strpos), (long)(s - i_strpos))
946 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
947 "Does not contradict STCLASS...\n");
952 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
953 PL_colors[4], (check ? "Guessed" : "Giving up"),
954 PL_colors[5], (long)(s - i_strpos)) );
957 fail_finish: /* Substring not found */
958 if (prog->check_substr || prog->check_utf8) /* could be removed already */
959 BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
961 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
962 PL_colors[4],PL_colors[5]));
966 /* We know what class REx starts with. Try to find this position... */
968 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, char *strend, char *startpos, I32 norun)
970 I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
974 register STRLEN uskip;
978 register I32 tmp = 1; /* Scratch variable? */
979 register bool do_utf8 = PL_reg_match_utf8;
981 /* We know what class it must start with. */
985 while (s + (uskip = UTF8SKIP(s)) <= strend) {
986 if ((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
987 !UTF8_IS_INVARIANT((U8)s[0]) ?
988 reginclass(c, (U8*)s, 0, do_utf8) :
989 REGINCLASS(c, (U8*)s)) {
990 if (tmp && (norun || regtry(prog, s)))
1001 while (s < strend) {
1004 if (REGINCLASS(c, (U8*)s) ||
1005 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
1006 /* The assignment of 2 is intentional:
1007 * for the folded sharp s, the skip is 2. */
1008 (skip = SHARP_S_SKIP))) {
1009 if (tmp && (norun || regtry(prog, s)))
1021 while (s < strend) {
1022 if (tmp && (norun || regtry(prog, s)))
1031 ln = STR_LEN(c); /* length to match in octets/bytes */
1032 lnc = (I32) ln; /* length to match in characters */
1034 STRLEN ulen1, ulen2;
1036 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
1037 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
1039 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1040 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1042 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE,
1043 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1044 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE,
1045 0, ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
1047 while (sm < ((U8 *) m + ln)) {
1062 c2 = PL_fold_locale[c1];
1064 e = HOP3c(strend, -((I32)lnc), s);
1067 e = s; /* Due to minlen logic of intuit() */
1069 /* The idea in the EXACTF* cases is to first find the
1070 * first character of the EXACTF* node and then, if
1071 * necessary, case-insensitively compare the full
1072 * text of the node. The c1 and c2 are the first
1073 * characters (though in Unicode it gets a bit
1074 * more complicated because there are more cases
1075 * than just upper and lower: one needs to use
1076 * the so-called folding case for case-insensitive
1077 * matching (called "loose matching" in Unicode).
1078 * ibcmp_utf8() will do just that. */
1082 U8 tmpbuf [UTF8_MAXBYTES+1];
1083 U8 foldbuf[UTF8_MAXBYTES_CASE+1];
1084 STRLEN len, foldlen;
1087 /* Upper and lower of 1st char are equal -
1088 * probably not a "letter". */
1090 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
1092 0 : UTF8_ALLOW_ANY);
1095 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1096 m, (char **)0, ln, (bool)UTF))
1097 && (norun || regtry(prog, s)) )
1100 uvchr_to_utf8(tmpbuf, c);
1101 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1103 && (f == c1 || f == c2)
1104 && (ln == foldlen ||
1105 !ibcmp_utf8((char *) foldbuf,
1106 (char **)0, foldlen, do_utf8,
1108 (char **)0, ln, (bool)UTF))
1109 && (norun || regtry(prog, s)) )
1117 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
1119 0 : UTF8_ALLOW_ANY);
1121 /* Handle some of the three Greek sigmas cases.
1122 * Note that not all the possible combinations
1123 * are handled here: some of them are handled
1124 * by the standard folding rules, and some of
1125 * them (the character class or ANYOF cases)
1126 * are handled during compiletime in
1127 * regexec.c:S_regclass(). */
1128 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1129 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1130 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1132 if ( (c == c1 || c == c2)
1134 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1135 m, (char **)0, ln, (bool)UTF))
1136 && (norun || regtry(prog, s)) )
1139 uvchr_to_utf8(tmpbuf, c);
1140 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1142 && (f == c1 || f == c2)
1143 && (ln == foldlen ||
1144 !ibcmp_utf8((char *) foldbuf,
1145 (char **)0, foldlen, do_utf8,
1147 (char **)0, ln, (bool)UTF))
1148 && (norun || regtry(prog, s)) )
1159 && (ln == 1 || !(OP(c) == EXACTF
1161 : ibcmp_locale(s, m, ln)))
1162 && (norun || regtry(prog, s)) )
1168 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1169 && (ln == 1 || !(OP(c) == EXACTF
1171 : ibcmp_locale(s, m, ln)))
1172 && (norun || regtry(prog, s)) )
1179 PL_reg_flags |= RF_tainted;
1186 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1188 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1190 tmp = ((OP(c) == BOUND ?
1191 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1192 LOAD_UTF8_CHARCLASS(alnum,"a");
1193 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1194 if (tmp == !(OP(c) == BOUND ?
1195 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1196 isALNUM_LC_utf8((U8*)s)))
1199 if ((norun || regtry(prog, s)))
1206 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1207 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1208 while (s < strend) {
1210 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1212 if ((norun || regtry(prog, s)))
1218 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1222 PL_reg_flags |= RF_tainted;
1229 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1231 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1233 tmp = ((OP(c) == NBOUND ?
1234 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1235 LOAD_UTF8_CHARCLASS(alnum,"a");
1236 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1237 if (tmp == !(OP(c) == NBOUND ?
1238 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1239 isALNUM_LC_utf8((U8*)s)))
1241 else if ((norun || regtry(prog, s)))
1247 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1248 tmp = ((OP(c) == NBOUND ?
1249 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1250 while (s < strend) {
1252 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1254 else if ((norun || regtry(prog, s)))
1259 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1264 LOAD_UTF8_CHARCLASS(alnum,"a");
1265 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1266 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1267 if (tmp && (norun || regtry(prog, s)))
1278 while (s < strend) {
1280 if (tmp && (norun || regtry(prog, s)))
1292 PL_reg_flags |= RF_tainted;
1294 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1295 if (isALNUM_LC_utf8((U8*)s)) {
1296 if (tmp && (norun || regtry(prog, s)))
1307 while (s < strend) {
1308 if (isALNUM_LC(*s)) {
1309 if (tmp && (norun || regtry(prog, s)))
1322 LOAD_UTF8_CHARCLASS(alnum,"a");
1323 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1324 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1325 if (tmp && (norun || regtry(prog, s)))
1336 while (s < strend) {
1338 if (tmp && (norun || regtry(prog, s)))
1350 PL_reg_flags |= RF_tainted;
1352 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1353 if (!isALNUM_LC_utf8((U8*)s)) {
1354 if (tmp && (norun || regtry(prog, s)))
1365 while (s < strend) {
1366 if (!isALNUM_LC(*s)) {
1367 if (tmp && (norun || regtry(prog, s)))
1380 LOAD_UTF8_CHARCLASS(space," ");
1381 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1382 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1383 if (tmp && (norun || regtry(prog, s)))
1394 while (s < strend) {
1396 if (tmp && (norun || regtry(prog, s)))
1408 PL_reg_flags |= RF_tainted;
1410 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1411 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1412 if (tmp && (norun || regtry(prog, s)))
1423 while (s < strend) {
1424 if (isSPACE_LC(*s)) {
1425 if (tmp && (norun || regtry(prog, s)))
1438 LOAD_UTF8_CHARCLASS(space," ");
1439 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1440 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1441 if (tmp && (norun || regtry(prog, s)))
1452 while (s < strend) {
1454 if (tmp && (norun || regtry(prog, s)))
1466 PL_reg_flags |= RF_tainted;
1468 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1469 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1470 if (tmp && (norun || regtry(prog, s)))
1481 while (s < strend) {
1482 if (!isSPACE_LC(*s)) {
1483 if (tmp && (norun || regtry(prog, s)))
1496 LOAD_UTF8_CHARCLASS(digit,"0");
1497 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1498 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1499 if (tmp && (norun || regtry(prog, s)))
1510 while (s < strend) {
1512 if (tmp && (norun || regtry(prog, s)))
1524 PL_reg_flags |= RF_tainted;
1526 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1527 if (isDIGIT_LC_utf8((U8*)s)) {
1528 if (tmp && (norun || regtry(prog, s)))
1539 while (s < strend) {
1540 if (isDIGIT_LC(*s)) {
1541 if (tmp && (norun || regtry(prog, s)))
1554 LOAD_UTF8_CHARCLASS(digit,"0");
1555 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1556 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1557 if (tmp && (norun || regtry(prog, s)))
1568 while (s < strend) {
1570 if (tmp && (norun || regtry(prog, s)))
1582 PL_reg_flags |= RF_tainted;
1584 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1585 if (!isDIGIT_LC_utf8((U8*)s)) {
1586 if (tmp && (norun || regtry(prog, s)))
1597 while (s < strend) {
1598 if (!isDIGIT_LC(*s)) {
1599 if (tmp && (norun || regtry(prog, s)))
1611 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1620 - regexec_flags - match a regexp against a string
1623 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1624 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1625 /* strend: pointer to null at end of string */
1626 /* strbeg: real beginning of string */
1627 /* minend: end of match must be >=minend after stringarg. */
1628 /* data: May be used for some additional optimizations. */
1629 /* nosave: For optimizations. */
1632 register regnode *c;
1633 register char *startpos = stringarg;
1634 I32 minlen; /* must match at least this many chars */
1635 I32 dontbother = 0; /* how many characters not to try at end */
1636 /* I32 start_shift = 0; */ /* Offset of the start to find
1637 constant substr. */ /* CC */
1638 I32 end_shift = 0; /* Same for the end. */ /* CC */
1639 I32 scream_pos = -1; /* Internal iterator of scream. */
1641 SV* oreplsv = GvSV(PL_replgv);
1642 bool do_utf8 = DO_UTF8(sv);
1643 I32 multiline = prog->reganch & PMf_MULTILINE;
1645 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
1646 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
1649 GET_RE_DEBUG_FLAGS_DECL;
1651 RX_MATCH_UTF8_set(prog,do_utf8);
1657 PL_regnarrate = DEBUG_r_TEST;
1660 /* Be paranoid... */
1661 if (prog == NULL || startpos == NULL) {
1662 Perl_croak(aTHX_ "NULL regexp parameter");
1666 minlen = prog->minlen;
1667 if (strend - startpos < minlen) {
1668 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1669 "String too short [regexec_flags]...\n"));
1673 /* Check validity of program. */
1674 if (UCHARAT(prog->program) != REG_MAGIC) {
1675 Perl_croak(aTHX_ "corrupted regexp program");
1679 PL_reg_eval_set = 0;
1682 if (prog->reganch & ROPT_UTF8)
1683 PL_reg_flags |= RF_utf8;
1685 /* Mark beginning of line for ^ and lookbehind. */
1686 PL_regbol = startpos;
1690 /* Mark end of line for $ (and such) */
1693 /* see how far we have to get to not match where we matched before */
1694 PL_regtill = startpos+minend;
1696 /* We start without call_cc context. */
1699 /* If there is a "must appear" string, look for it. */
1702 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1705 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1706 PL_reg_ganch = startpos;
1707 else if (sv && SvTYPE(sv) >= SVt_PVMG
1709 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1710 && mg->mg_len >= 0) {
1711 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1712 if (prog->reganch & ROPT_ANCH_GPOS) {
1713 if (s > PL_reg_ganch)
1718 else /* pos() not defined */
1719 PL_reg_ganch = strbeg;
1722 if (!(flags & REXEC_CHECKED) && (prog->check_substr != Nullsv || prog->check_utf8 != Nullsv)) {
1723 re_scream_pos_data d;
1725 d.scream_olds = &scream_olds;
1726 d.scream_pos = &scream_pos;
1727 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1729 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1730 goto phooey; /* not present */
1736 pv_uni_display(dsv0, (U8*)prog->precomp, prog->prelen, 60,
1737 UNI_DISPLAY_REGEX) :
1739 int len0 = UTF ? SvCUR(dsv0) : prog->prelen;
1740 char *s1 = do_utf8 ? sv_uni_display(dsv1, sv, 60,
1741 UNI_DISPLAY_REGEX) : startpos;
1742 int len1 = do_utf8 ? SvCUR(dsv1) : strend - startpos;
1745 PerlIO_printf(Perl_debug_log,
1746 "%sMatching REx%s `%s%*.*s%s%s' against `%s%.*s%s%s'\n",
1747 PL_colors[4],PL_colors[5],PL_colors[0],
1750 len0 > 60 ? "..." : "",
1752 (int)(len1 > 60 ? 60 : len1),
1754 (len1 > 60 ? "..." : "")
1758 /* Simplest case: anchored match need be tried only once. */
1759 /* [unless only anchor is BOL and multiline is set] */
1760 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1761 if (s == startpos && regtry(prog, startpos))
1763 else if (multiline || (prog->reganch & ROPT_IMPLICIT)
1764 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1769 dontbother = minlen - 1;
1770 end = HOP3c(strend, -dontbother, strbeg) - 1;
1771 /* for multiline we only have to try after newlines */
1772 if (prog->check_substr || prog->check_utf8) {
1776 if (regtry(prog, s))
1781 if (prog->reganch & RE_USE_INTUIT) {
1782 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1793 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1794 if (regtry(prog, s))
1801 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1802 if (regtry(prog, PL_reg_ganch))
1807 /* Messy cases: unanchored match. */
1808 if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) {
1809 /* we have /x+whatever/ */
1810 /* it must be a one character string (XXXX Except UTF?) */
1815 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1816 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1817 ch = SvPVX(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
1820 while (s < strend) {
1822 DEBUG_EXECUTE_r( did_match = 1 );
1823 if (regtry(prog, s)) goto got_it;
1825 while (s < strend && *s == ch)
1832 while (s < strend) {
1834 DEBUG_EXECUTE_r( did_match = 1 );
1835 if (regtry(prog, s)) goto got_it;
1837 while (s < strend && *s == ch)
1843 DEBUG_EXECUTE_r(if (!did_match)
1844 PerlIO_printf(Perl_debug_log,
1845 "Did not find anchored character...\n")
1849 else if (prog->anchored_substr != Nullsv
1850 || prog->anchored_utf8 != Nullsv
1851 || ((prog->float_substr != Nullsv || prog->float_utf8 != Nullsv)
1852 && prog->float_max_offset < strend - s)) {
1857 char *last1; /* Last position checked before */
1861 if (prog->anchored_substr || prog->anchored_utf8) {
1862 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1863 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1864 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1865 back_max = back_min = prog->anchored_offset;
1867 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1868 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1869 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1870 back_max = prog->float_max_offset;
1871 back_min = prog->float_min_offset;
1873 if (must == &PL_sv_undef)
1874 /* could not downgrade utf8 check substring, so must fail */
1877 last = HOP3c(strend, /* Cannot start after this */
1878 -(I32)(CHR_SVLEN(must)
1879 - (SvTAIL(must) != 0) + back_min), strbeg);
1882 last1 = HOPc(s, -1);
1884 last1 = s - 1; /* bogus */
1886 /* XXXX check_substr already used to find `s', can optimize if
1887 check_substr==must. */
1889 dontbother = end_shift;
1890 strend = HOPc(strend, -dontbother);
1891 while ( (s <= last) &&
1892 ((flags & REXEC_SCREAM)
1893 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1894 end_shift, &scream_pos, 0))
1895 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1896 (unsigned char*)strend, must,
1897 multiline ? FBMrf_MULTILINE : 0))) ) {
1898 /* we may be pointing at the wrong string */
1899 if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog))
1900 s = strbeg + (s - SvPVX(sv));
1901 DEBUG_EXECUTE_r( did_match = 1 );
1902 if (HOPc(s, -back_max) > last1) {
1903 last1 = HOPc(s, -back_min);
1904 s = HOPc(s, -back_max);
1907 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1909 last1 = HOPc(s, -back_min);
1913 while (s <= last1) {
1914 if (regtry(prog, s))
1920 while (s <= last1) {
1921 if (regtry(prog, s))
1927 DEBUG_EXECUTE_r(if (!did_match)
1928 PerlIO_printf(Perl_debug_log,
1929 "Did not find %s substr `%s%.*s%s'%s...\n",
1930 ((must == prog->anchored_substr || must == prog->anchored_utf8)
1931 ? "anchored" : "floating"),
1933 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1935 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1939 else if ((c = prog->regstclass)) {
1941 I32 op = (U8)OP(prog->regstclass);
1942 /* don't bother with what can't match */
1943 if (PL_regkind[op] != EXACT && op != CANY)
1944 strend = HOPc(strend, -(minlen - 1));
1947 SV *prop = sv_newmortal();
1955 pv_uni_display(dsv0, (U8*)SvPVX(prop), SvCUR(prop), 60,
1956 UNI_DISPLAY_REGEX) :
1958 len0 = UTF ? SvCUR(dsv0) : SvCUR(prop);
1960 sv_uni_display(dsv1, sv, 60, UNI_DISPLAY_REGEX) : s;
1961 len1 = UTF ? SvCUR(dsv1) : strend - s;
1962 PerlIO_printf(Perl_debug_log,
1963 "Matching stclass `%*.*s' against `%*.*s'\n",
1967 if (find_byclass(prog, c, s, strend, startpos, 0))
1969 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1973 if (prog->float_substr != Nullsv || prog->float_utf8 != Nullsv) {
1978 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1979 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1980 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
1982 if (flags & REXEC_SCREAM) {
1983 last = screaminstr(sv, float_real, s - strbeg,
1984 end_shift, &scream_pos, 1); /* last one */
1986 last = scream_olds; /* Only one occurrence. */
1987 /* we may be pointing at the wrong string */
1988 else if (RX_MATCH_COPIED(prog))
1989 s = strbeg + (s - SvPVX(sv));
1993 char *little = SvPV(float_real, len);
1995 if (SvTAIL(float_real)) {
1996 if (memEQ(strend - len + 1, little, len - 1))
1997 last = strend - len + 1;
1998 else if (!multiline)
1999 last = memEQ(strend - len, little, len)
2000 ? strend - len : Nullch;
2006 last = rninstr(s, strend, little, little + len);
2008 last = strend; /* matching `$' */
2012 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
2013 "%sCan't trim the tail, match fails (should not happen)%s\n",
2014 PL_colors[4],PL_colors[5]));
2015 goto phooey; /* Should not happen! */
2017 dontbother = strend - last + prog->float_min_offset;
2019 if (minlen && (dontbother < minlen))
2020 dontbother = minlen - 1;
2021 strend -= dontbother; /* this one's always in bytes! */
2022 /* We don't know much -- general case. */
2025 if (regtry(prog, s))
2034 if (regtry(prog, s))
2036 } while (s++ < strend);
2044 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
2046 if (PL_reg_eval_set) {
2047 /* Preserve the current value of $^R */
2048 if (oreplsv != GvSV(PL_replgv))
2049 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
2050 restored, the value remains
2052 restore_pos(aTHX_ 0);
2055 /* make sure $`, $&, $', and $digit will work later */
2056 if ( !(flags & REXEC_NOT_FIRST) ) {
2057 RX_MATCH_COPY_FREE(prog);
2058 if (flags & REXEC_COPY_STR) {
2059 I32 i = PL_regeol - startpos + (stringarg - strbeg);
2060 #ifdef PERL_COPY_ON_WRITE
2062 || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) {
2064 PerlIO_printf(Perl_debug_log,
2065 "Copy on write: regexp capture, type %d\n",
2068 prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv);
2069 prog->subbeg = SvPVX(prog->saved_copy);
2070 assert (SvPOKp(prog->saved_copy));
2074 RX_MATCH_COPIED_on(prog);
2075 s = savepvn(strbeg, i);
2081 prog->subbeg = strbeg;
2082 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2089 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2090 PL_colors[4],PL_colors[5]));
2091 if (PL_reg_eval_set)
2092 restore_pos(aTHX_ 0);
2097 - regtry - try match at specific point
2099 STATIC I32 /* 0 failure, 1 success */
2100 S_regtry(pTHX_ regexp *prog, char *startpos)
2106 GET_RE_DEBUG_FLAGS_DECL;
2109 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
2111 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
2114 PL_reg_eval_set = RS_init;
2115 DEBUG_EXECUTE_r(DEBUG_s(
2116 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2117 (IV)(PL_stack_sp - PL_stack_base));
2119 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
2120 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2121 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2123 /* Apparently this is not needed, judging by wantarray. */
2124 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
2125 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2128 /* Make $_ available to executed code. */
2129 if (PL_reg_sv != DEFSV) {
2134 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
2135 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
2136 /* prepare for quick setting of pos */
2137 sv_magic(PL_reg_sv, (SV*)0,
2138 PERL_MAGIC_regex_global, Nullch, 0);
2139 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
2143 PL_reg_oldpos = mg->mg_len;
2144 SAVEDESTRUCTOR_X(restore_pos, 0);
2146 if (!PL_reg_curpm) {
2147 Newz(22,PL_reg_curpm, 1, PMOP);
2150 SV* repointer = newSViv(0);
2151 /* so we know which PL_regex_padav element is PL_reg_curpm */
2152 SvFLAGS(repointer) |= SVf_BREAK;
2153 av_push(PL_regex_padav,repointer);
2154 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2155 PL_regex_pad = AvARRAY(PL_regex_padav);
2159 PM_SETRE(PL_reg_curpm, prog);
2160 PL_reg_oldcurpm = PL_curpm;
2161 PL_curpm = PL_reg_curpm;
2162 if (RX_MATCH_COPIED(prog)) {
2163 /* Here is a serious problem: we cannot rewrite subbeg,
2164 since it may be needed if this match fails. Thus
2165 $` inside (?{}) could fail... */
2166 PL_reg_oldsaved = prog->subbeg;
2167 PL_reg_oldsavedlen = prog->sublen;
2168 #ifdef PERL_COPY_ON_WRITE
2169 PL_nrs = prog->saved_copy;
2171 RX_MATCH_COPIED_off(prog);
2174 PL_reg_oldsaved = Nullch;
2175 prog->subbeg = PL_bostr;
2176 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2178 prog->startp[0] = startpos - PL_bostr;
2179 PL_reginput = startpos;
2180 PL_regstartp = prog->startp;
2181 PL_regendp = prog->endp;
2182 PL_reglastparen = &prog->lastparen;
2183 PL_reglastcloseparen = &prog->lastcloseparen;
2184 prog->lastparen = 0;
2185 prog->lastcloseparen = 0;
2187 DEBUG_EXECUTE_r(PL_reg_starttry = startpos);
2188 if (PL_reg_start_tmpl <= prog->nparens) {
2189 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2190 if(PL_reg_start_tmp)
2191 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2193 New(22,PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2196 /* XXXX What this code is doing here?!!! There should be no need
2197 to do this again and again, PL_reglastparen should take care of
2200 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2201 * Actually, the code in regcppop() (which Ilya may be meaning by
2202 * PL_reglastparen), is not needed at all by the test suite
2203 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2204 * enough, for building DynaLoader, or otherwise this
2205 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2206 * will happen. Meanwhile, this code *is* needed for the
2207 * above-mentioned test suite tests to succeed. The common theme
2208 * on those tests seems to be returning null fields from matches.
2213 if (prog->nparens) {
2214 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2221 if (regmatch(prog->program + 1)) {
2222 prog->endp[0] = PL_reginput - PL_bostr;
2225 REGCP_UNWIND(lastcp);
2229 #define RE_UNWIND_BRANCH 1
2230 #define RE_UNWIND_BRANCHJ 2
2234 typedef struct { /* XX: makes sense to enlarge it... */
2238 } re_unwind_generic_t;
2251 } re_unwind_branch_t;
2253 typedef union re_unwind_t {
2255 re_unwind_generic_t generic;
2256 re_unwind_branch_t branch;
2259 #define sayYES goto yes
2260 #define sayNO goto no
2261 #define sayNO_ANYOF goto no_anyof
2262 #define sayYES_FINAL goto yes_final
2263 #define sayYES_LOUD goto yes_loud
2264 #define sayNO_FINAL goto no_final
2265 #define sayNO_SILENT goto do_no
2266 #define saySAME(x) if (x) goto yes; else goto no
2268 /* this is used to determine how far from the left messages like
2269 'failed...' are printed. Currently 29 makes these messages line
2270 up with the opcode they refer to. Earlier perls used 25 which
2271 left these messages outdented making reviewing a debug output
2274 #define REPORT_CODE_OFF 29
2277 /* Make sure there is a test for this +1 options in re_tests */
2278 #define TRIE_INITAL_ACCEPT_BUFFLEN 4;
2280 #define TRIE_CHECK_STATE_IS_ACCEPTING STMT_START { \
2281 if ( trie->states[ state ].wordnum ) { \
2282 if ( !accepted ) { \
2285 bufflen = TRIE_INITAL_ACCEPT_BUFFLEN ; \
2286 sv_accept_buff=NEWSV( 1234, \
2287 bufflen * sizeof(reg_trie_accepted) - 1 ); \
2288 SvCUR_set( sv_accept_buff, sizeof(reg_trie_accepted) ); \
2289 SvPOK_on( sv_accept_buff ); \
2290 sv_2mortal( sv_accept_buff ); \
2291 accept_buff = (reg_trie_accepted*)SvPV_nolen( sv_accept_buff );\
2293 if ( accepted >= bufflen ) { \
2295 accept_buff =(reg_trie_accepted*)SvGROW( sv_accept_buff, \
2296 bufflen * sizeof(reg_trie_accepted) ); \
2298 SvCUR_set( sv_accept_buff,SvCUR( sv_accept_buff ) \
2299 + sizeof( reg_trie_accepted ) ); \
2301 accept_buff[ accepted ].wordnum = trie->states[ state ].wordnum; \
2302 accept_buff[ accepted ].endpos = uc; \
2306 #define TRIE_HANDLE_CHAR STMT_START { \
2307 if ( uvc < 256 ) { \
2308 charid = trie->charmap[ uvc ]; \
2311 if( trie->widecharmap ) { \
2312 SV** svpp = (SV**)NULL; \
2313 svpp = hv_fetch( trie->widecharmap, (char*)&uvc, \
2314 sizeof( UV ), 0 ); \
2316 charid = (U16)SvIV( *svpp ); \
2321 ( base + charid > trie->uniquecharcount ) && \
2322 ( base + charid - 1 - trie->uniquecharcount < trie->lasttrans) && \
2323 trie->trans[ base + charid - 1 - trie->uniquecharcount ].check == state ) \
2325 state = trie->trans[ base + charid - 1 - trie->uniquecharcount ].next; \
2333 - regmatch - main matching routine
2335 * Conceptually the strategy is simple: check to see whether the current
2336 * node matches, call self recursively to see whether the rest matches,
2337 * and then act accordingly. In practice we make some effort to avoid
2338 * recursion, in particular by going through "ordinary" nodes (that don't
2339 * need to know whether the rest of the match failed) by a loop instead of
2342 /* [lwall] I've hoisted the register declarations to the outer block in order to
2343 * maybe save a little bit of pushing and popping on the stack. It also takes
2344 * advantage of machines that use a register save mask on subroutine entry.
2346 STATIC I32 /* 0 failure, 1 success */
2347 S_regmatch(pTHX_ regnode *prog)
2349 register regnode *scan; /* Current node. */
2350 regnode *next; /* Next node. */
2351 regnode *inner; /* Next node in internal branch. */
2352 register I32 nextchr; /* renamed nextchr - nextchar colides with
2353 function of same name */
2354 register I32 n; /* no or next */
2355 register I32 ln = 0; /* len or last */
2356 register char *s = Nullch; /* operand or save */
2357 register char *locinput = PL_reginput;
2358 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2359 int minmod = 0, sw = 0, logical = 0;
2362 /* used by the trie code */
2363 SV *sv_accept_buff; /* accepting states we have traversed */
2364 reg_trie_accepted *accept_buff; /* "" */
2365 reg_trie_data *trie; /* what trie are we using right now */
2366 U32 accepted = 0; /* how many accepting states we have seen*/
2369 I32 firstcp = PL_savestack_ix;
2371 register bool do_utf8 = PL_reg_match_utf8;
2373 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2374 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2375 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2387 /* Note that nextchr is a byte even in UTF */
2388 nextchr = UCHARAT(locinput);
2390 while (scan != NULL) {
2393 SV *prop = sv_newmortal();
2394 int docolor = *PL_colors[0];
2395 int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2396 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2397 /* The part of the string before starttry has one color
2398 (pref0_len chars), between starttry and current
2399 position another one (pref_len - pref0_len chars),
2400 after the current position the third one.
2401 We assume that pref0_len <= pref_len, otherwise we
2402 decrease pref0_len. */
2403 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2404 ? (5 + taill) - l : locinput - PL_bostr;
2407 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2409 pref0_len = pref_len - (locinput - PL_reg_starttry);
2410 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2411 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2412 ? (5 + taill) - pref_len : PL_regeol - locinput);
2413 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2417 if (pref0_len > pref_len)
2418 pref0_len = pref_len;
2419 regprop(prop, scan);
2422 do_utf8 && OP(scan) != CANY ?
2423 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2424 pref0_len, 60, UNI_DISPLAY_REGEX) :
2425 locinput - pref_len;
2426 int len0 = do_utf8 ? strlen(s0) : pref0_len;
2427 char *s1 = do_utf8 && OP(scan) != CANY ?
2428 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2429 pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) :
2430 locinput - pref_len + pref0_len;
2431 int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2432 char *s2 = do_utf8 && OP(scan) != CANY ?
2433 pv_uni_display(dsv2, (U8*)locinput,
2434 PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
2436 int len2 = do_utf8 ? strlen(s2) : l;
2437 PerlIO_printf(Perl_debug_log,
2438 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2439 (IV)(locinput - PL_bostr),
2446 (docolor ? "" : "> <"),
2450 15 - l - pref_len + 1,
2452 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2457 next = scan + NEXT_OFF(scan);
2463 if (locinput == PL_bostr)
2465 /* regtill = regbol; */
2470 if (locinput == PL_bostr ||
2471 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2477 if (locinput == PL_bostr)
2481 if (locinput == PL_reg_ganch)
2487 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2492 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2494 if (PL_regeol - locinput > 1)
2498 if (PL_regeol != locinput)
2502 if (!nextchr && locinput >= PL_regeol)
2505 locinput += PL_utf8skip[nextchr];
2506 if (locinput > PL_regeol)
2508 nextchr = UCHARAT(locinput);
2511 nextchr = UCHARAT(++locinput);
2514 if (!nextchr && locinput >= PL_regeol)
2516 nextchr = UCHARAT(++locinput);
2519 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2522 locinput += PL_utf8skip[nextchr];
2523 if (locinput > PL_regeol)
2525 nextchr = UCHARAT(locinput);
2528 nextchr = UCHARAT(++locinput);
2534 traverse the TRIE keeping track of all accepting states
2535 we transition through until we get to a failing node.
2537 we use two slightly different pieces of code to handle
2538 the traversal depending on whether its case sensitive or
2539 not. we reuse the accept code however. (this should probably
2540 be turned into a macro.)
2547 U32 uniflags = ckWARN( WARN_UTF8 ) ? 0 : UTF8_ALLOW_ANY;
2548 U8 *uc = ( U8* )locinput;
2555 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
2556 U8 *uscan = (U8*)NULL;
2560 trie = (reg_trie_data*)PL_regdata->data[ ARG( scan ) ];
2562 while ( state && uc <= (U8*)PL_regeol ) {
2564 TRIE_CHECK_STATE_IS_ACCEPTING;
2566 base = trie->states[ state ].trans.base;
2568 DEBUG_TRIE_EXECUTE_r(
2569 PerlIO_printf( Perl_debug_log,
2570 "%*s %sState: %4x, Base: %4x Accepted: %4x ",
2571 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
2572 state, base, accepted );
2577 if ( do_utf8 || UTF ) {
2579 uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags );
2584 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags );
2585 uvc = to_uni_fold( uvc, foldbuf, &foldlen );
2586 foldlen -= UNISKIP( uvc );
2587 uscan = foldbuf + UNISKIP( uvc );
2599 DEBUG_TRIE_EXECUTE_r(
2600 PerlIO_printf( Perl_debug_log,
2601 "Charid:%3x CV:%4x After State: %4x%s\n",
2602 charid, uvc, state, PL_colors[5] );
2611 /* unreached codepoint: we jump into the middle of the next case
2612 from previous if blocks */
2615 U32 uniflags = ckWARN( WARN_UTF8 ) ? 0 : UTF8_ALLOW_ANY;
2616 U8 *uc = (U8*)locinput;
2625 trie = (reg_trie_data*)PL_regdata->data[ ARG( scan ) ];
2627 while ( state && uc <= (U8*)PL_regeol ) {
2629 TRIE_CHECK_STATE_IS_ACCEPTING;
2631 base = trie->states[ state ].trans.base;
2633 DEBUG_TRIE_EXECUTE_r(
2634 PerlIO_printf( Perl_debug_log,
2635 "%*s %sState: %4x, Base: %4x Accepted: %4x ",
2636 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
2637 state, base, accepted );
2642 if ( do_utf8 || UTF ) {
2643 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags );
2654 DEBUG_TRIE_EXECUTE_r(
2655 PerlIO_printf( Perl_debug_log,
2656 "Charid:%3x CV:%4x After State: %4x%s\n",
2657 charid, uvc, state, PL_colors[5] );
2667 There was at least one accepting state that we
2668 transitioned through. Presumably the number of accepting
2669 states is going to be low, typically one or two. So we
2670 simply scan through to find the one with lowest wordnum.
2671 Once we find it, we swap the last state into its place
2672 and decrement the size. We then try to match the rest of
2673 the pattern at the point where the word ends, if we
2674 succeed then we end the loop, otherwise the loop
2675 eventually terminates once all of the accepting states
2682 if ( accepted == 1 ) {
2684 SV **tmp = av_fetch( trie->words, accept_buff[ 0 ].wordnum-1, 0 );
2685 PerlIO_printf( Perl_debug_log,
2686 "%*s %sonly one match : #%d <%s>%s\n",
2687 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],
2688 accept_buff[ 0 ].wordnum,
2689 tmp ? SvPV_nolen( *tmp ) : "not compiled under -Dr",
2692 PL_reginput = (char *)accept_buff[ 0 ].endpos;
2693 /* in this case we free tmps/leave before we call regmatch
2694 as we wont be using accept_buff again. */
2697 gotit = regmatch( scan + NEXT_OFF( scan ) );
2700 PerlIO_printf( Perl_debug_log,"%*s %sgot %d possible matches%s\n",
2701 REPORT_CODE_OFF + PL_regindent * 2, "",PL_colors[4], accepted,
2704 while ( !gotit && accepted-- ) {
2707 for( cur = 1 ; cur <= accepted ; cur++ ) {
2708 DEBUG_TRIE_EXECUTE_r(
2709 PerlIO_printf( Perl_debug_log,
2710 "%*s %sgot %d (%d) as best, looking at %d (%d)%s\n",
2711 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
2712 best, accept_buff[ best ].wordnum, cur,
2713 accept_buff[ cur ].wordnum, PL_colors[5] );
2716 if ( accept_buff[ cur ].wordnum < accept_buff[ best ].wordnum )
2720 SV **tmp = av_fetch( trie->words, accept_buff[ best ].wordnum - 1, 0 );
2721 PerlIO_printf( Perl_debug_log, "%*s %strying alternation #%d <%s> at 0x%p%s\n",
2722 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],
2723 accept_buff[best].wordnum,
2724 tmp ? SvPV_nolen( *tmp ) : "not compiled under -Dr",scan,
2727 if ( best<accepted ) {
2728 reg_trie_accepted tmp = accept_buff[ best ];
2729 accept_buff[ best ] = accept_buff[ accepted ];
2730 accept_buff[ accepted ] = tmp;
2733 PL_reginput = (char *)accept_buff[ best ].endpos;
2736 as far as I can tell we only need the SAVETMPS/FREETMPS
2737 for re's with EVAL in them but I'm leaving them in for
2738 all until I can be sure.
2741 gotit = regmatch( scan + NEXT_OFF( scan ) ) ;
2754 /* unreached codepoint */
2758 if (do_utf8 != UTF) {
2759 /* The target and the pattern have differing utf8ness. */
2765 /* The target is utf8, the pattern is not utf8. */
2769 if (NATIVE_TO_UNI(*(U8*)s) !=
2770 utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen,
2772 0 : UTF8_ALLOW_ANY))
2779 /* The target is not utf8, the pattern is utf8. */
2783 if (NATIVE_TO_UNI(*((U8*)l)) !=
2784 utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen,
2786 0 : UTF8_ALLOW_ANY))
2793 nextchr = UCHARAT(locinput);
2796 /* The target and the pattern have the same utf8ness. */
2797 /* Inline the first character, for speed. */
2798 if (UCHARAT(s) != nextchr)
2800 if (PL_regeol - locinput < ln)
2802 if (ln > 1 && memNE(s, locinput, ln))
2805 nextchr = UCHARAT(locinput);
2808 PL_reg_flags |= RF_tainted;
2814 if (do_utf8 || UTF) {
2815 /* Either target or the pattern are utf8. */
2817 char *e = PL_regeol;
2819 if (ibcmp_utf8(s, 0, ln, (bool)UTF,
2820 l, &e, 0, do_utf8)) {
2821 /* One more case for the sharp s:
2822 * pack("U0U*", 0xDF) =~ /ss/i,
2823 * the 0xC3 0x9F are the UTF-8
2824 * byte sequence for the U+00DF. */
2826 toLOWER(s[0]) == 's' &&
2828 toLOWER(s[1]) == 's' &&
2835 nextchr = UCHARAT(locinput);
2839 /* Neither the target and the pattern are utf8. */
2841 /* Inline the first character, for speed. */
2842 if (UCHARAT(s) != nextchr &&
2843 UCHARAT(s) != ((OP(scan) == EXACTF)
2844 ? PL_fold : PL_fold_locale)[nextchr])
2846 if (PL_regeol - locinput < ln)
2848 if (ln > 1 && (OP(scan) == EXACTF
2849 ? ibcmp(s, locinput, ln)
2850 : ibcmp_locale(s, locinput, ln)))
2853 nextchr = UCHARAT(locinput);
2857 STRLEN inclasslen = PL_regeol - locinput;
2859 if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8))
2861 if (locinput >= PL_regeol)
2863 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
2864 nextchr = UCHARAT(locinput);
2869 nextchr = UCHARAT(locinput);
2870 if (!REGINCLASS(scan, (U8*)locinput))
2872 if (!nextchr && locinput >= PL_regeol)
2874 nextchr = UCHARAT(++locinput);
2878 /* If we might have the case of the German sharp s
2879 * in a casefolding Unicode character class. */
2881 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
2882 locinput += SHARP_S_SKIP;
2883 nextchr = UCHARAT(locinput);
2889 PL_reg_flags |= RF_tainted;
2895 LOAD_UTF8_CHARCLASS(alnum,"a");
2896 if (!(OP(scan) == ALNUM
2897 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2898 : isALNUM_LC_utf8((U8*)locinput)))
2902 locinput += PL_utf8skip[nextchr];
2903 nextchr = UCHARAT(locinput);
2906 if (!(OP(scan) == ALNUM
2907 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2909 nextchr = UCHARAT(++locinput);
2912 PL_reg_flags |= RF_tainted;
2915 if (!nextchr && locinput >= PL_regeol)
2918 LOAD_UTF8_CHARCLASS(alnum,"a");
2919 if (OP(scan) == NALNUM
2920 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2921 : isALNUM_LC_utf8((U8*)locinput))
2925 locinput += PL_utf8skip[nextchr];
2926 nextchr = UCHARAT(locinput);
2929 if (OP(scan) == NALNUM
2930 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2932 nextchr = UCHARAT(++locinput);
2936 PL_reg_flags |= RF_tainted;
2940 /* was last char in word? */
2942 if (locinput == PL_bostr)
2945 U8 *r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
2947 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
2949 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2950 ln = isALNUM_uni(ln);
2951 LOAD_UTF8_CHARCLASS(alnum,"a");
2952 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2955 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2956 n = isALNUM_LC_utf8((U8*)locinput);
2960 ln = (locinput != PL_bostr) ?
2961 UCHARAT(locinput - 1) : '\n';
2962 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2964 n = isALNUM(nextchr);
2967 ln = isALNUM_LC(ln);
2968 n = isALNUM_LC(nextchr);
2971 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
2972 OP(scan) == BOUNDL))
2976 PL_reg_flags |= RF_tainted;
2982 if (UTF8_IS_CONTINUED(nextchr)) {
2983 LOAD_UTF8_CHARCLASS(space," ");
2984 if (!(OP(scan) == SPACE
2985 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
2986 : isSPACE_LC_utf8((U8*)locinput)))
2990 locinput += PL_utf8skip[nextchr];
2991 nextchr = UCHARAT(locinput);
2994 if (!(OP(scan) == SPACE
2995 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
2997 nextchr = UCHARAT(++locinput);
3000 if (!(OP(scan) == SPACE
3001 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3003 nextchr = UCHARAT(++locinput);
3007 PL_reg_flags |= RF_tainted;
3010 if (!nextchr && locinput >= PL_regeol)
3013 LOAD_UTF8_CHARCLASS(space," ");
3014 if (OP(scan) == NSPACE
3015 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
3016 : isSPACE_LC_utf8((U8*)locinput))
3020 locinput += PL_utf8skip[nextchr];
3021 nextchr = UCHARAT(locinput);
3024 if (OP(scan) == NSPACE
3025 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
3027 nextchr = UCHARAT(++locinput);
3030 PL_reg_flags |= RF_tainted;
3036 LOAD_UTF8_CHARCLASS(digit,"0");
3037 if (!(OP(scan) == DIGIT
3038 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
3039 : isDIGIT_LC_utf8((U8*)locinput)))
3043 locinput += PL_utf8skip[nextchr];
3044 nextchr = UCHARAT(locinput);
3047 if (!(OP(scan) == DIGIT
3048 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
3050 nextchr = UCHARAT(++locinput);
3053 PL_reg_flags |= RF_tainted;
3056 if (!nextchr && locinput >= PL_regeol)
3059 LOAD_UTF8_CHARCLASS(digit,"0");
3060 if (OP(scan) == NDIGIT
3061 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
3062 : isDIGIT_LC_utf8((U8*)locinput))
3066 locinput += PL_utf8skip[nextchr];
3067 nextchr = UCHARAT(locinput);
3070 if (OP(scan) == NDIGIT
3071 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
3073 nextchr = UCHARAT(++locinput);
3076 if (locinput >= PL_regeol)
3079 LOAD_UTF8_CHARCLASS(mark,"~");
3080 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3082 locinput += PL_utf8skip[nextchr];
3083 while (locinput < PL_regeol &&
3084 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3085 locinput += UTF8SKIP(locinput);
3086 if (locinput > PL_regeol)
3091 nextchr = UCHARAT(locinput);
3094 PL_reg_flags |= RF_tainted;
3098 n = ARG(scan); /* which paren pair */
3099 ln = PL_regstartp[n];
3100 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3101 if ((I32)*PL_reglastparen < n || ln == -1)
3102 sayNO; /* Do not match unless seen CLOSEn. */
3103 if (ln == PL_regendp[n])
3107 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
3109 char *e = PL_bostr + PL_regendp[n];
3111 * Note that we can't do the "other character" lookup trick as
3112 * in the 8-bit case (no pun intended) because in Unicode we
3113 * have to map both upper and title case to lower case.
3115 if (OP(scan) == REFF) {
3116 STRLEN ulen1, ulen2;
3117 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
3118 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
3122 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
3123 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
3124 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
3131 nextchr = UCHARAT(locinput);
3135 /* Inline the first character, for speed. */
3136 if (UCHARAT(s) != nextchr &&
3138 (UCHARAT(s) != ((OP(scan) == REFF
3139 ? PL_fold : PL_fold_locale)[nextchr]))))
3141 ln = PL_regendp[n] - ln;
3142 if (locinput + ln > PL_regeol)
3144 if (ln > 1 && (OP(scan) == REF
3145 ? memNE(s, locinput, ln)
3147 ? ibcmp(s, locinput, ln)
3148 : ibcmp_locale(s, locinput, ln))))
3151 nextchr = UCHARAT(locinput);
3162 OP_4tree *oop = PL_op;
3163 COP *ocurcop = PL_curcop;
3166 struct regexp *oreg = PL_reg_re;
3169 PL_op = (OP_4tree*)PL_regdata->data[n];
3170 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
3171 PAD_SAVE_LOCAL(old_comppad, (PAD*)PL_regdata->data[n + 2]);
3172 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
3176 CALLRUNOPS(aTHX); /* Scalar context. */
3179 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
3187 PAD_RESTORE_LOCAL(old_comppad);
3188 PL_curcop = ocurcop;
3190 if (logical == 2) { /* Postponed subexpression. */
3192 MAGIC *mg = Null(MAGIC*);
3194 CHECKPOINT cp, lastcp;
3198 if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret)))
3199 mg = mg_find(sv, PERL_MAGIC_qr);
3200 else if (SvSMAGICAL(ret)) {
3201 if (SvGMAGICAL(ret))
3202 sv_unmagic(ret, PERL_MAGIC_qr);
3204 mg = mg_find(ret, PERL_MAGIC_qr);
3208 re = (regexp *)mg->mg_obj;
3209 (void)ReREFCNT_inc(re);
3213 char *t = SvPV(ret, len);
3215 char *oprecomp = PL_regprecomp;
3216 I32 osize = PL_regsize;
3217 I32 onpar = PL_regnpar;
3220 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
3221 re = CALLREGCOMP(aTHX_ t, t + len, &pm);
3223 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
3225 sv_magic(ret,(SV*)ReREFCNT_inc(re),
3227 PL_regprecomp = oprecomp;
3232 PerlIO_printf(Perl_debug_log,
3233 "Entering embedded `%s%.60s%s%s'\n",
3237 (strlen(re->precomp) > 60 ? "..." : ""))
3240 state.prev = PL_reg_call_cc;
3241 state.cc = PL_regcc;
3242 state.re = PL_reg_re;
3246 cp = regcppush(0); /* Save *all* the positions. */
3249 state.ss = PL_savestack_ix;
3250 *PL_reglastparen = 0;
3251 *PL_reglastcloseparen = 0;
3252 PL_reg_call_cc = &state;
3253 PL_reginput = locinput;
3254 toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
3255 ((re->reganch & ROPT_UTF8) != 0);
3256 if (toggleutf) PL_reg_flags ^= RF_utf8;
3258 /* XXXX This is too dramatic a measure... */
3261 if (regmatch(re->program + 1)) {
3262 /* Even though we succeeded, we need to restore
3263 global variables, since we may be wrapped inside
3264 SUSPEND, thus the match may be not finished yet. */
3266 /* XXXX Do this only if SUSPENDed? */
3267 PL_reg_call_cc = state.prev;
3268 PL_regcc = state.cc;
3269 PL_reg_re = state.re;
3270 cache_re(PL_reg_re);
3271 if (toggleutf) PL_reg_flags ^= RF_utf8;
3273 /* XXXX This is too dramatic a measure... */
3276 /* These are needed even if not SUSPEND. */
3282 REGCP_UNWIND(lastcp);
3284 PL_reg_call_cc = state.prev;
3285 PL_regcc = state.cc;
3286 PL_reg_re = state.re;
3287 cache_re(PL_reg_re);
3288 if (toggleutf) PL_reg_flags ^= RF_utf8;
3290 /* XXXX This is too dramatic a measure... */
3300 sv_setsv(save_scalar(PL_replgv), ret);
3306 n = ARG(scan); /* which paren pair */
3307 PL_reg_start_tmp[n] = locinput;
3312 n = ARG(scan); /* which paren pair */
3313 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
3314 PL_regendp[n] = locinput - PL_bostr;
3315 if (n > (I32)*PL_reglastparen)
3316 *PL_reglastparen = n;
3317 *PL_reglastcloseparen = n;
3320 n = ARG(scan); /* which paren pair */
3321 sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
3324 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3326 next = NEXTOPER(NEXTOPER(scan));
3328 next = scan + ARG(scan);
3329 if (OP(next) == IFTHEN) /* Fake one. */
3330 next = NEXTOPER(NEXTOPER(next));
3334 logical = scan->flags;
3336 /*******************************************************************
3337 PL_regcc contains infoblock about the innermost (...)* loop, and
3338 a pointer to the next outer infoblock.
3340 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
3342 1) After matching X, regnode for CURLYX is processed;
3344 2) This regnode creates infoblock on the stack, and calls
3345 regmatch() recursively with the starting point at WHILEM node;
3347 3) Each hit of WHILEM node tries to match A and Z (in the order
3348 depending on the current iteration, min/max of {min,max} and
3349 greediness). The information about where are nodes for "A"
3350 and "Z" is read from the infoblock, as is info on how many times "A"
3351 was already matched, and greediness.
3353 4) After A matches, the same WHILEM node is hit again.
3355 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
3356 of the same pair. Thus when WHILEM tries to match Z, it temporarily
3357 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
3358 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
3359 of the external loop.
3361 Currently present infoblocks form a tree with a stem formed by PL_curcc
3362 and whatever it mentions via ->next, and additional attached trees
3363 corresponding to temporarily unset infoblocks as in "5" above.
3365 In the following picture infoblocks for outer loop of
3366 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
3367 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
3368 infoblocks are drawn below the "reset" infoblock.
3370 In fact in the picture below we do not show failed matches for Z and T
3371 by WHILEM blocks. [We illustrate minimal matches, since for them it is
3372 more obvious *why* one needs to *temporary* unset infoblocks.]
3374 Matched REx position InfoBlocks Comment
3378 Y A)*?Z)*?T x <- O <- I
3379 YA )*?Z)*?T x <- O <- I
3380 YA A)*?Z)*?T x <- O <- I
3381 YAA )*?Z)*?T x <- O <- I
3382 YAA Z)*?T x <- O # Temporary unset I
3385 YAAZ Y(A)*?Z)*?T x <- O
3388 YAAZY (A)*?Z)*?T x <- O
3391 YAAZY A)*?Z)*?T x <- O <- I
3394 YAAZYA )*?Z)*?T x <- O <- I
3397 YAAZYA Z)*?T x <- O # Temporary unset I
3403 YAAZYAZ T x # Temporary unset O
3410 *******************************************************************/
3413 CHECKPOINT cp = PL_savestack_ix;
3414 /* No need to save/restore up to this paren */
3415 I32 parenfloor = scan->flags;
3417 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3419 cc.oldcc = PL_regcc;
3421 /* XXXX Probably it is better to teach regpush to support
3422 parenfloor > PL_regsize... */
3423 if (parenfloor > (I32)*PL_reglastparen)
3424 parenfloor = *PL_reglastparen; /* Pessimization... */
3425 cc.parenfloor = parenfloor;
3427 cc.min = ARG1(scan);
3428 cc.max = ARG2(scan);
3429 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3433 PL_reginput = locinput;
3434 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
3436 PL_regcc = cc.oldcc;
3442 * This is really hard to understand, because after we match
3443 * what we're trying to match, we must make sure the rest of
3444 * the REx is going to match for sure, and to do that we have
3445 * to go back UP the parse tree by recursing ever deeper. And
3446 * if it fails, we have to reset our parent's current state
3447 * that we can try again after backing off.
3450 CHECKPOINT cp, lastcp;
3451 CURCUR* cc = PL_regcc;
3452 char *lastloc = cc->lastloc; /* Detection of 0-len. */
3454 n = cc->cur + 1; /* how many we know we matched */
3455 PL_reginput = locinput;
3458 PerlIO_printf(Perl_debug_log,
3459 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3460 REPORT_CODE_OFF+PL_regindent*2, "",
3461 (long)n, (long)cc->min,
3462 (long)cc->max, PTR2UV(cc))
3465 /* If degenerate scan matches "", assume scan done. */
3467 if (locinput == cc->lastloc && n >= cc->min) {
3468 PL_regcc = cc->oldcc;
3472 PerlIO_printf(Perl_debug_log,
3473 "%*s empty match detected, try continuation...\n",
3474 REPORT_CODE_OFF+PL_regindent*2, "")
3476 if (regmatch(cc->next))
3484 /* First just match a string of min scans. */
3488 cc->lastloc = locinput;
3489 if (regmatch(cc->scan))
3492 cc->lastloc = lastloc;
3497 /* Check whether we already were at this position.
3498 Postpone detection until we know the match is not
3499 *that* much linear. */
3500 if (!PL_reg_maxiter) {
3501 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3502 PL_reg_leftiter = PL_reg_maxiter;
3504 if (PL_reg_leftiter-- == 0) {
3505 I32 size = (PL_reg_maxiter + 7)/8;
3506 if (PL_reg_poscache) {
3507 if ((I32)PL_reg_poscache_size < size) {
3508 Renew(PL_reg_poscache, size, char);
3509 PL_reg_poscache_size = size;
3511 Zero(PL_reg_poscache, size, char);
3514 PL_reg_poscache_size = size;
3515 Newz(29, PL_reg_poscache, size, char);
3518 PerlIO_printf(Perl_debug_log,
3519 "%sDetected a super-linear match, switching on caching%s...\n",
3520 PL_colors[4], PL_colors[5])
3523 if (PL_reg_leftiter < 0) {
3524 I32 o = locinput - PL_bostr, b;
3526 o = (scan->flags & 0xf) - 1 + o * (scan->flags>>4);
3529 if (PL_reg_poscache[o] & (1<<b)) {
3531 PerlIO_printf(Perl_debug_log,
3532 "%*s already tried at this position...\n",
3533 REPORT_CODE_OFF+PL_regindent*2, "")
3535 if (PL_reg_flags & RF_false)
3540 PL_reg_poscache[o] |= (1<<b);
3544 /* Prefer next over scan for minimal matching. */
3547 PL_regcc = cc->oldcc;
3550 cp = regcppush(cc->parenfloor);
3552 if (regmatch(cc->next)) {
3554 sayYES; /* All done. */
3556 REGCP_UNWIND(lastcp);
3562 if (n >= cc->max) { /* Maximum greed exceeded? */
3563 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3564 && !(PL_reg_flags & RF_warned)) {
3565 PL_reg_flags |= RF_warned;
3566 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3567 "Complex regular subexpression recursion",
3574 PerlIO_printf(Perl_debug_log,
3575 "%*s trying longer...\n",
3576 REPORT_CODE_OFF+PL_regindent*2, "")
3578 /* Try scanning more and see if it helps. */
3579 PL_reginput = locinput;
3581 cc->lastloc = locinput;
3582 cp = regcppush(cc->parenfloor);
3584 if (regmatch(cc->scan)) {
3588 REGCP_UNWIND(lastcp);
3591 cc->lastloc = lastloc;
3595 /* Prefer scan over next for maximal matching. */
3597 if (n < cc->max) { /* More greed allowed? */
3598 cp = regcppush(cc->parenfloor);
3600 cc->lastloc = locinput;
3602 if (regmatch(cc->scan)) {
3606 REGCP_UNWIND(lastcp);
3607 regcppop(); /* Restore some previous $<digit>s? */
3608 PL_reginput = locinput;
3610 PerlIO_printf(Perl_debug_log,
3611 "%*s failed, try continuation...\n",
3612 REPORT_CODE_OFF+PL_regindent*2, "")
3615 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3616 && !(PL_reg_flags & RF_warned)) {
3617 PL_reg_flags |= RF_warned;
3618 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3619 "Complex regular subexpression recursion",
3623 /* Failed deeper matches of scan, so see if this one works. */
3624 PL_regcc = cc->oldcc;
3627 if (regmatch(cc->next))
3633 cc->lastloc = lastloc;
3638 next = scan + ARG(scan);
3641 inner = NEXTOPER(NEXTOPER(scan));
3644 inner = NEXTOPER(scan);
3648 if (OP(next) != c1) /* No choice. */
3649 next = inner; /* Avoid recursion. */
3651 I32 lastparen = *PL_reglastparen;
3653 re_unwind_branch_t *uw;
3655 /* Put unwinding data on stack */
3656 unwind1 = SSNEWt(1,re_unwind_branch_t);
3657 uw = SSPTRt(unwind1,re_unwind_branch_t);
3660 uw->type = ((c1 == BRANCH)
3662 : RE_UNWIND_BRANCHJ);
3663 uw->lastparen = lastparen;
3665 uw->locinput = locinput;
3666 uw->nextchr = nextchr;
3668 uw->regindent = ++PL_regindent;
3671 REGCP_SET(uw->lastcp);
3673 /* Now go into the first branch */
3686 /* We suppose that the next guy does not need
3687 backtracking: in particular, it is of constant non-zero length,
3688 and has no parenths to influence future backrefs. */
3689 ln = ARG1(scan); /* min to match */
3690 n = ARG2(scan); /* max to match */
3691 paren = scan->flags;
3693 if (paren > PL_regsize)
3695 if (paren > (I32)*PL_reglastparen)
3696 *PL_reglastparen = paren;
3698 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3700 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3701 PL_reginput = locinput;
3704 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3706 locinput = PL_reginput;
3707 if (HAS_TEXT(next) || JUMPABLE(next)) {
3708 regnode *text_node = next;
3710 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3712 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3714 if (PL_regkind[(U8)OP(text_node)] == REF) {
3718 else { c1 = (U8)*STRING(text_node); }
3719 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3721 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3722 c2 = PL_fold_locale[c1];
3731 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3732 /* If it could work, try it. */
3734 UCHARAT(PL_reginput) == c1 ||
3735 UCHARAT(PL_reginput) == c2)
3739 PL_regstartp[paren] =
3740 HOPc(PL_reginput, -l) - PL_bostr;
3741 PL_regendp[paren] = PL_reginput - PL_bostr;
3744 PL_regendp[paren] = -1;
3748 REGCP_UNWIND(lastcp);
3750 /* Couldn't or didn't -- move forward. */
3751 PL_reginput = locinput;
3752 if (regrepeat_hard(scan, 1, &l)) {
3754 locinput = PL_reginput;
3761 n = regrepeat_hard(scan, n, &l);
3762 locinput = PL_reginput;
3764 PerlIO_printf(Perl_debug_log,
3765 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3766 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3770 if (HAS_TEXT(next) || JUMPABLE(next)) {
3771 regnode *text_node = next;
3773 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3775 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3777 if (PL_regkind[(U8)OP(text_node)] == REF) {
3781 else { c1 = (U8)*STRING(text_node); }
3783 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3785 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3786 c2 = PL_fold_locale[c1];
3797 /* If it could work, try it. */
3799 UCHARAT(PL_reginput) == c1 ||
3800 UCHARAT(PL_reginput) == c2)
3803 PerlIO_printf(Perl_debug_log,
3804 "%*s trying tail with n=%"IVdf"...\n",
3805 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3809 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3810 PL_regendp[paren] = PL_reginput - PL_bostr;
3813 PL_regendp[paren] = -1;
3817 REGCP_UNWIND(lastcp);
3819 /* Couldn't or didn't -- back up. */
3821 locinput = HOPc(locinput, -l);
3822 PL_reginput = locinput;
3829 paren = scan->flags; /* Which paren to set */
3830 if (paren > PL_regsize)
3832 if (paren > (I32)*PL_reglastparen)
3833 *PL_reglastparen = paren;
3834 ln = ARG1(scan); /* min to match */
3835 n = ARG2(scan); /* max to match */
3836 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3840 ln = ARG1(scan); /* min to match */
3841 n = ARG2(scan); /* max to match */
3842 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3847 scan = NEXTOPER(scan);
3853 scan = NEXTOPER(scan);
3857 * Lookahead to avoid useless match attempts
3858 * when we know what character comes next.
3862 * Used to only do .*x and .*?x, but now it allows
3863 * for )'s, ('s and (?{ ... })'s to be in the way
3864 * of the quantifier and the EXACT-like node. -- japhy
3867 if (HAS_TEXT(next) || JUMPABLE(next)) {
3869 regnode *text_node = next;
3871 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3873 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3875 if (PL_regkind[(U8)OP(text_node)] == REF) {
3877 goto assume_ok_easy;
3879 else { s = (U8*)STRING(text_node); }
3883 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3885 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3886 c2 = PL_fold_locale[c1];
3889 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3890 STRLEN ulen1, ulen2;
3891 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
3892 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
3894 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3895 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3897 c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0,
3899 0 : UTF8_ALLOW_ANY);
3900 c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0,
3902 0 : UTF8_ALLOW_ANY);
3905 c2 = c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0,
3907 0 : UTF8_ALLOW_ANY);
3915 PL_reginput = locinput;
3919 if (ln && regrepeat(scan, ln) < ln)
3921 locinput = PL_reginput;
3924 char *e; /* Should not check after this */
3925 char *old = locinput;
3928 if (n == REG_INFTY) {
3931 while (UTF8_IS_CONTINUATION(*(U8*)e))
3937 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3941 e = locinput + n - ln;
3946 /* Find place 'next' could work */
3949 while (locinput <= e &&
3950 UCHARAT(locinput) != c1)
3953 while (locinput <= e
3954 && UCHARAT(locinput) != c1
3955 && UCHARAT(locinput) != c2)
3958 count = locinput - old;
3963 /* count initialised to
3964 * utf8_distance(old, locinput) */
3965 while (locinput <= e &&
3966 utf8n_to_uvchr((U8*)locinput,
3967 UTF8_MAXBYTES, &len,
3969 0 : UTF8_ALLOW_ANY) != (UV)c1) {
3974 /* count initialised to
3975 * utf8_distance(old, locinput) */
3976 while (locinput <= e) {
3977 UV c = utf8n_to_uvchr((U8*)locinput,
3978 UTF8_MAXBYTES, &len,
3980 0 : UTF8_ALLOW_ANY);
3981 if (c == (UV)c1 || c == (UV)c2)
3990 /* PL_reginput == old now */
3991 if (locinput != old) {
3992 ln = 1; /* Did some */
3993 if (regrepeat(scan, count) < count)
3996 /* PL_reginput == locinput now */
3997 TRYPAREN(paren, ln, locinput);
3998 PL_reginput = locinput; /* Could be reset... */
3999 REGCP_UNWIND(lastcp);
4000 /* Couldn't or didn't -- move forward. */
4003 locinput += UTF8SKIP(locinput);
4010 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
4014 c = utf8n_to_uvchr((U8*)PL_reginput,
4017 0 : UTF8_ALLOW_ANY);
4019 c = UCHARAT(PL_reginput);
4020 /* If it could work, try it. */
4021 if (c == (UV)c1 || c == (UV)c2)
4023 TRYPAREN(paren, ln, PL_reginput);
4024 REGCP_UNWIND(lastcp);
4027 /* If it could work, try it. */
4028 else if (c1 == -1000)
4030 TRYPAREN(paren, ln, PL_reginput);
4031 REGCP_UNWIND(lastcp);
4033 /* Couldn't or didn't -- move forward. */
4034 PL_reginput = locinput;
4035 if (regrepeat(scan, 1)) {
4037 locinput = PL_reginput;
4045 n = regrepeat(scan, n);
4046 locinput = PL_reginput;
4047 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
4048 (OP(next) != MEOL ||
4049 OP(next) == SEOL || OP(next) == EOS))
4051 ln = n; /* why back off? */
4052 /* ...because $ and \Z can match before *and* after
4053 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
4054 We should back off by one in this case. */
4055 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
4064 c = utf8n_to_uvchr((U8*)PL_reginput,
4067 0 : UTF8_ALLOW_ANY);
4069 c = UCHARAT(PL_reginput);
4071 /* If it could work, try it. */
4072 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
4074 TRYPAREN(paren, n, PL_reginput);
4075 REGCP_UNWIND(lastcp);
4077 /* Couldn't or didn't -- back up. */
4079 PL_reginput = locinput = HOPc(locinput, -1);
4087 c = utf8n_to_uvchr((U8*)PL_reginput,
4090 0 : UTF8_ALLOW_ANY);
4092 c = UCHARAT(PL_reginput);
4094 /* If it could work, try it. */
4095 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
4097 TRYPAREN(paren, n, PL_reginput);
4098 REGCP_UNWIND(lastcp);
4100 /* Couldn't or didn't -- back up. */
4102 PL_reginput = locinput = HOPc(locinput, -1);
4109 if (PL_reg_call_cc) {
4110 re_cc_state *cur_call_cc = PL_reg_call_cc;
4111 CURCUR *cctmp = PL_regcc;
4112 regexp *re = PL_reg_re;
4113 CHECKPOINT cp, lastcp;
4115 cp = regcppush(0); /* Save *all* the positions. */
4117 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
4119 PL_reginput = locinput; /* Make position available to
4121 cache_re(PL_reg_call_cc->re);
4122 PL_regcc = PL_reg_call_cc->cc;
4123 PL_reg_call_cc = PL_reg_call_cc->prev;
4124 if (regmatch(cur_call_cc->node)) {
4125 PL_reg_call_cc = cur_call_cc;
4129 REGCP_UNWIND(lastcp);
4131 PL_reg_call_cc = cur_call_cc;
4137 PerlIO_printf(Perl_debug_log,
4138 "%*s continuation failed...\n",
4139 REPORT_CODE_OFF+PL_regindent*2, "")
4143 if (locinput < PL_regtill) {
4144 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
4145 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
4147 (long)(locinput - PL_reg_starttry),
4148 (long)(PL_regtill - PL_reg_starttry),
4150 sayNO_FINAL; /* Cannot match: too short. */
4152 PL_reginput = locinput; /* put where regtry can find it */
4153 sayYES_FINAL; /* Success! */
4155 PL_reginput = locinput; /* put where regtry can find it */
4156 sayYES_LOUD; /* Success! */
4159 PL_reginput = locinput;
4164 s = HOPBACKc(locinput, scan->flags);
4170 PL_reginput = locinput;
4171 PL_reg_flags ^= RF_false;
4176 s = HOPBACKc(locinput, scan->flags);
4182 PL_reginput = locinput;
4185 inner = NEXTOPER(NEXTOPER(scan));
4186 if (regmatch(inner) != n) {
4188 PL_reg_flags ^= RF_false;
4199 PL_reg_flags ^= RF_false;
4205 if (OP(scan) == SUSPEND) {
4206 locinput = PL_reginput;
4207 nextchr = UCHARAT(locinput);
4212 next = scan + ARG(scan);
4217 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
4218 PTR2UV(scan), OP(scan));
4219 Perl_croak(aTHX_ "regexp memory corruption");
4226 * We get here only if there's trouble -- normally "case END" is
4227 * the terminating point.
4229 Perl_croak(aTHX_ "corrupted regexp pointers");
4235 PerlIO_printf(Perl_debug_log,
4236 "%*s %scould match...%s\n",
4237 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],PL_colors[5])
4241 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
4242 PL_colors[4],PL_colors[5]));
4248 #if 0 /* Breaks $^R */
4256 PerlIO_printf(Perl_debug_log,
4257 "%*s %sfailed...%s\n",
4258 REPORT_CODE_OFF+PL_regindent*2, "",PL_colors[4],PL_colors[5])
4264 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
4267 case RE_UNWIND_BRANCH:
4268 case RE_UNWIND_BRANCHJ:
4270 re_unwind_branch_t *uwb = &(uw->branch);
4271 I32 lastparen = uwb->lastparen;
4273 REGCP_UNWIND(uwb->lastcp);
4274 for (n = *PL_reglastparen; n > lastparen; n--)
4276 *PL_reglastparen = n;
4277 scan = next = uwb->next;
4279 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
4280 ? BRANCH : BRANCHJ) ) { /* Failure */
4287 /* Have more choice yet. Reuse the same uwb. */
4289 if ((n = (uwb->type == RE_UNWIND_BRANCH
4290 ? NEXT_OFF(next) : ARG(next))))
4293 next = NULL; /* XXXX Needn't unwinding in this case... */
4295 next = NEXTOPER(scan);
4296 if (uwb->type == RE_UNWIND_BRANCHJ)
4297 next = NEXTOPER(next);
4298 locinput = uwb->locinput;
4299 nextchr = uwb->nextchr;
4301 PL_regindent = uwb->regindent;
4308 Perl_croak(aTHX_ "regexp unwind memory corruption");
4319 - regrepeat - repeatedly match something simple, report how many
4322 * [This routine now assumes that it will only match on things of length 1.
4323 * That was true before, but now we assume scan - reginput is the count,
4324 * rather than incrementing count on every character. [Er, except utf8.]]
4327 S_regrepeat(pTHX_ regnode *p, I32 max)
4329 register char *scan;
4331 register char *loceol = PL_regeol;
4332 register I32 hardcount = 0;
4333 register bool do_utf8 = PL_reg_match_utf8;
4336 if (max == REG_INFTY)
4338 else if (max < loceol - scan)
4339 loceol = scan + max;
4344 while (scan < loceol && hardcount < max && *scan != '\n') {
4345 scan += UTF8SKIP(scan);
4349 while (scan < loceol && *scan != '\n')
4356 while (scan < loceol && hardcount < max) {
4357 scan += UTF8SKIP(scan);
4367 case EXACT: /* length of string is 1 */
4369 while (scan < loceol && UCHARAT(scan) == c)
4372 case EXACTF: /* length of string is 1 */
4374 while (scan < loceol &&
4375 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
4378 case EXACTFL: /* length of string is 1 */
4379 PL_reg_flags |= RF_tainted;
4381 while (scan < loceol &&
4382 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
4388 while (hardcount < max && scan < loceol &&
4389 reginclass(p, (U8*)scan, 0, do_utf8)) {
4390 scan += UTF8SKIP(scan);
4394 while (scan < loceol && REGINCLASS(p, (U8*)scan))
4401 LOAD_UTF8_CHARCLASS(alnum,"a");
4402 while (hardcount < max && scan < loceol &&
4403 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4404 scan += UTF8SKIP(scan);
4408 while (scan < loceol && isALNUM(*scan))
4413 PL_reg_flags |= RF_tainted;
4416 while (hardcount < max && scan < loceol &&
4417 isALNUM_LC_utf8((U8*)scan)) {
4418 scan += UTF8SKIP(scan);
4422 while (scan < loceol && isALNUM_LC(*scan))
4429 LOAD_UTF8_CHARCLASS(alnum,"a");
4430 while (hardcount < max && scan < loceol &&
4431 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4432 scan += UTF8SKIP(scan);
4436 while (scan < loceol && !isALNUM(*scan))
4441 PL_reg_flags |= RF_tainted;
4444 while (hardcount < max && scan < loceol &&
4445 !isALNUM_LC_utf8((U8*)scan)) {
4446 scan += UTF8SKIP(scan);
4450 while (scan < loceol && !isALNUM_LC(*scan))
4457 LOAD_UTF8_CHARCLASS(space," ");
4458 while (hardcount < max && scan < loceol &&
4460 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4461 scan += UTF8SKIP(scan);
4465 while (scan < loceol && isSPACE(*scan))
4470 PL_reg_flags |= RF_tainted;
4473 while (hardcount < max && scan < loceol &&
4474 (*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4475 scan += UTF8SKIP(scan);
4479 while (scan < loceol && isSPACE_LC(*scan))
4486 LOAD_UTF8_CHARCLASS(space," ");
4487 while (hardcount < max && scan < loceol &&
4489 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4490 scan += UTF8SKIP(scan);
4494 while (scan < loceol && !isSPACE(*scan))
4499 PL_reg_flags |= RF_tainted;
4502 while (hardcount < max && scan < loceol &&
4503 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4504 scan += UTF8SKIP(scan);
4508 while (scan < loceol && !isSPACE_LC(*scan))
4515 LOAD_UTF8_CHARCLASS(digit,"0");
4516 while (hardcount < max && scan < loceol &&
4517 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4518 scan += UTF8SKIP(scan);
4522 while (scan < loceol && isDIGIT(*scan))
4529 LOAD_UTF8_CHARCLASS(digit,"0");
4530 while (hardcount < max && scan < loceol &&
4531 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4532 scan += UTF8SKIP(scan);
4536 while (scan < loceol && !isDIGIT(*scan))
4540 default: /* Called on something of 0 width. */
4541 break; /* So match right here or not at all. */
4547 c = scan - PL_reginput;
4552 SV *prop = sv_newmortal();
4556 PerlIO_printf(Perl_debug_log,
4557 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4558 REPORT_CODE_OFF+1, "", SvPVX(prop),(IV)c,(IV)max);
4566 - regrepeat_hard - repeatedly match something, report total lenth and length
4568 * The repeater is supposed to have constant non-zero length.
4572 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4574 register char *scan = Nullch;
4575 register char *start;
4576 register char *loceol = PL_regeol;
4578 I32 count = 0, res = 1;
4583 start = PL_reginput;
4584 if (PL_reg_match_utf8) {
4585 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4588 while (start < PL_reginput) {
4590 start += UTF8SKIP(start);
4601 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4603 *lp = l = PL_reginput - start;
4604 if (max != REG_INFTY && l*max < loceol - scan)
4605 loceol = scan + l*max;
4618 - regclass_swash - prepare the utf8 swash
4622 Perl_regclass_swash(pTHX_ register regnode* node, bool doinit, SV** listsvp, SV **altsvp)
4628 if (PL_regdata && PL_regdata->count) {
4631 if (PL_regdata->what[n] == 's') {
4632 SV *rv = (SV*)PL_regdata->data[n];
4633 AV *av = (AV*)SvRV((SV*)rv);
4634 SV **ary = AvARRAY(av);
4637 /* See the end of regcomp.c:S_reglass() for
4638 * documentation of these array elements. */
4641 a = SvTYPE(ary[1]) == SVt_RV ? &ary[1] : 0;
4642 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
4646 else if (si && doinit) {
4647 sw = swash_init("utf8", "", si, 1, 0);
4648 (void)av_store(av, 1, sw);
4664 - reginclass - determine if a character falls into a character class
4666 The n is the ANYOF regnode, the p is the target string, lenp
4667 is pointer to the maximum length of how far to go in the p
4668 (if the lenp is zero, UTF8SKIP(p) is used),
4669 do_utf8 tells whether the target string is in UTF-8.
4674 S_reginclass(pTHX_ register regnode *n, register U8* p, STRLEN* lenp, register bool do_utf8)
4676 char flags = ANYOF_FLAGS(n);
4682 if (do_utf8 && !UTF8_IS_INVARIANT(c))
4683 c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &len,
4684 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
4686 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
4687 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4690 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4691 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4694 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4698 SV *sw = regclass_swash(n, TRUE, 0, (SV**)&av);
4701 if (swash_fetch(sw, p, do_utf8))
4703 else if (flags & ANYOF_FOLD) {
4704 if (!match && lenp && av) {
4707 for (i = 0; i <= av_len(av); i++) {
4708 SV* sv = *av_fetch(av, i, FALSE);
4710 char *s = SvPV(sv, len);
4712 if (len <= plen && memEQ(s, (char*)p, len)) {
4720 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
4723 to_utf8_fold(p, tmpbuf, &tmplen);
4724 if (swash_fetch(sw, tmpbuf, do_utf8))
4730 if (match && lenp && *lenp == 0)
4731 *lenp = UNISKIP(NATIVE_TO_UNI(c));
4733 if (!match && c < 256) {
4734 if (ANYOF_BITMAP_TEST(n, c))
4736 else if (flags & ANYOF_FOLD) {
4739 if (flags & ANYOF_LOCALE) {
4740 PL_reg_flags |= RF_tainted;
4741 f = PL_fold_locale[c];
4745 if (f != c && ANYOF_BITMAP_TEST(n, f))
4749 if (!match && (flags & ANYOF_CLASS)) {
4750 PL_reg_flags |= RF_tainted;
4752 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4753 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4754 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4755 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4756 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4757 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4758 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4759 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4760 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4761 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4762 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4763 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4764 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4765 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4766 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4767 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4768 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4769 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4770 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4771 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4772 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4773 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4774 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4775 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4776 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4777 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4778 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4779 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4780 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4781 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4782 ) /* How's that for a conditional? */
4789 return (flags & ANYOF_INVERT) ? !match : match;
4793 S_reghop(pTHX_ U8 *s, I32 off)
4795 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4799 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4802 while (off-- && s < lim) {
4803 /* XXX could check well-formedness here */
4811 if (UTF8_IS_CONTINUED(*s)) {
4812 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4815 /* XXX could check well-formedness here */
4823 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4825 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4829 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4832 while (off-- && s < lim) {
4833 /* XXX could check well-formedness here */
4843 if (UTF8_IS_CONTINUED(*s)) {
4844 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4847 /* XXX could check well-formedness here */
4859 restore_pos(pTHX_ void *arg)
4861 if (PL_reg_eval_set) {
4862 if (PL_reg_oldsaved) {
4863 PL_reg_re->subbeg = PL_reg_oldsaved;
4864 PL_reg_re->sublen = PL_reg_oldsavedlen;
4865 #ifdef PERL_COPY_ON_WRITE
4866 PL_reg_re->saved_copy = PL_nrs;
4868 RX_MATCH_COPIED_on(PL_reg_re);
4870 PL_reg_magic->mg_len = PL_reg_oldpos;
4871 PL_reg_eval_set = 0;
4872 PL_curpm = PL_reg_oldcurpm;
4877 S_to_utf8_substr(pTHX_ register regexp *prog)
4880 if (prog->float_substr && !prog->float_utf8) {
4881 prog->float_utf8 = sv = newSVsv(prog->float_substr);
4882 sv_utf8_upgrade(sv);
4883 if (SvTAIL(prog->float_substr))
4885 if (prog->float_substr == prog->check_substr)
4886 prog->check_utf8 = sv;
4888 if (prog->anchored_substr && !prog->anchored_utf8) {
4889 prog->anchored_utf8 = sv = newSVsv(prog->anchored_substr);
4890 sv_utf8_upgrade(sv);
4891 if (SvTAIL(prog->anchored_substr))
4893 if (prog->anchored_substr == prog->check_substr)
4894 prog->check_utf8 = sv;
4899 S_to_byte_substr(pTHX_ register regexp *prog)
4902 if (prog->float_utf8 && !prog->float_substr) {
4903 prog->float_substr = sv = newSVsv(prog->float_utf8);
4904 if (sv_utf8_downgrade(sv, TRUE)) {
4905 if (SvTAIL(prog->float_utf8))
4909 prog->float_substr = sv = &PL_sv_undef;
4911 if (prog->float_utf8 == prog->check_utf8)
4912 prog->check_substr = sv;
4914 if (prog->anchored_utf8 && !prog->anchored_substr) {
4915 prog->anchored_substr = sv = newSVsv(prog->anchored_utf8);
4916 if (sv_utf8_downgrade(sv, TRUE)) {
4917 if (SvTAIL(prog->anchored_utf8))
4921 prog->anchored_substr = sv = &PL_sv_undef;
4923 if (prog->anchored_utf8 == prog->check_utf8)
4924 prog->check_substr = sv;