5 * "One Ring to rule them all, One Ring to find them..."
8 /* This file contains functions for executing a regular expression. See
9 * also regcomp.c which funnily enough, contains functions for compiling
10 * a regular expression.
12 * This file is also copied at build time to ext/re/re_exec.c, where
13 * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT.
14 * This causes the main functions to be compiled under new names and with
15 * debugging support added, which makes "use re 'debug'" work.
19 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
20 * confused with the original package (see point 3 below). Thanks, Henry!
23 /* Additional note: this code is very heavily munged from Henry's version
24 * in places. In some spots I've traded clarity for efficiency, so don't
25 * blame Henry for some of the lack of readability.
28 /* The names of the functions have been changed from regcomp and
29 * regexec to pregcomp and pregexec in order to avoid conflicts
30 * with the POSIX routines of the same names.
33 #ifdef PERL_EXT_RE_BUILD
34 /* need to replace pregcomp et al, so enable that */
35 # ifndef PERL_IN_XSUB_RE
36 # define PERL_IN_XSUB_RE
38 /* need access to debugger hooks */
39 # if defined(PERL_EXT_RE_DEBUG) && !defined(DEBUGGING)
44 #ifdef PERL_IN_XSUB_RE
45 /* We *really* need to overwrite these symbols: */
46 # define Perl_regexec_flags my_regexec
47 # define Perl_regdump my_regdump
48 # define Perl_regprop my_regprop
49 # define Perl_re_intuit_start my_re_intuit_start
50 /* *These* symbols are masked to allow static link. */
51 # define Perl_pregexec my_pregexec
52 # define Perl_reginitcolors my_reginitcolors
53 # define Perl_regclass_swash my_regclass_swash
55 # define PERL_NO_GET_CONTEXT
59 * pregcomp and pregexec -- regsub and regerror are not used in perl
61 * Copyright (c) 1986 by University of Toronto.
62 * Written by Henry Spencer. Not derived from licensed software.
64 * Permission is granted to anyone to use this software for any
65 * purpose on any computer system, and to redistribute it freely,
66 * subject to the following restrictions:
68 * 1. The author is not responsible for the consequences of use of
69 * this software, no matter how awful, even if they arise
72 * 2. The origin of this software must not be misrepresented, either
73 * by explicit claim or by omission.
75 * 3. Altered versions must be plainly marked as such, and must not
76 * be misrepresented as being the original software.
78 **** Alterations to Henry's code are...
80 **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
81 **** 2000, 2001, 2002, 2003, 2004, 2005, by Larry Wall and others
83 **** You may distribute under the terms of either the GNU General Public
84 **** License or the Artistic License, as specified in the README file.
86 * Beware that some of this code is subtly aware of the way operator
87 * precedence is structured in regular expressions. Serious changes in
88 * regular-expression syntax might require a total rethink.
91 #define PERL_IN_REGEXEC_C
96 #define RF_tainted 1 /* tainted information used? */
97 #define RF_warned 2 /* warned about big count? */
98 #define RF_evaled 4 /* Did an EVAL with setting? */
99 #define RF_utf8 8 /* String contains multibyte chars? */
101 #define UTF ((PL_reg_flags & RF_utf8) != 0)
103 #define RS_init 1 /* eval environment created */
104 #define RS_set 2 /* replsv value is set */
107 #define STATIC static
110 #define REGINCLASS(p,c) (ANYOF_FLAGS(p) ? reginclass(p,c,0,0) : ANYOF_BITMAP_TEST(p,*(c)))
116 #define CHR_SVLEN(sv) (do_utf8 ? sv_len_utf8(sv) : SvCUR(sv))
117 #define CHR_DIST(a,b) (PL_reg_match_utf8 ? utf8_distance(a,b) : a - b)
119 #define reghop_c(pos,off) ((char*)reghop((U8*)pos, off))
120 #define reghopmaybe_c(pos,off) ((char*)reghopmaybe((U8*)pos, off))
121 #define HOP(pos,off) (PL_reg_match_utf8 ? reghop((U8*)pos, off) : (U8*)(pos + off))
122 #define HOPMAYBE(pos,off) (PL_reg_match_utf8 ? reghopmaybe((U8*)pos, off) : (U8*)(pos + off))
123 #define HOPc(pos,off) ((char*)HOP(pos,off))
124 #define HOPMAYBEc(pos,off) ((char*)HOPMAYBE(pos,off))
126 #define HOPBACK(pos, off) ( \
127 (PL_reg_match_utf8) \
128 ? reghopmaybe((U8*)pos, -off) \
129 : (pos - off >= PL_bostr) \
133 #define HOPBACKc(pos, off) (char*)HOPBACK(pos, off)
135 #define reghop3_c(pos,off,lim) ((char*)reghop3((U8*)pos, off, (U8*)lim))
136 #define reghopmaybe3_c(pos,off,lim) ((char*)reghopmaybe3((U8*)pos, off, (U8*)lim))
137 #define HOP3(pos,off,lim) (PL_reg_match_utf8 ? reghop3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
138 #define HOPMAYBE3(pos,off,lim) (PL_reg_match_utf8 ? reghopmaybe3((U8*)pos, off, (U8*)lim) : (U8*)(pos + off))
139 #define HOP3c(pos,off,lim) ((char*)HOP3(pos,off,lim))
140 #define HOPMAYBE3c(pos,off,lim) ((char*)HOPMAYBE3(pos,off,lim))
142 #define LOAD_UTF8_CHARCLASS(class,str) STMT_START { \
143 if (!CAT2(PL_utf8_,class)) { bool ok; ENTER; save_re_context(); ok=CAT2(is_utf8_,class)((const U8*)str); assert(ok); LEAVE; } } STMT_END
144 #define LOAD_UTF8_CHARCLASS_ALNUM() LOAD_UTF8_CHARCLASS(alnum,"a")
145 #define LOAD_UTF8_CHARCLASS_DIGIT() LOAD_UTF8_CHARCLASS(digit,"0")
146 #define LOAD_UTF8_CHARCLASS_SPACE() LOAD_UTF8_CHARCLASS(space," ")
147 #define LOAD_UTF8_CHARCLASS_MARK() LOAD_UTF8_CHARCLASS(mark, "\xcd\x86")
149 /* for use after a quantifier and before an EXACT-like node -- japhy */
150 #define JUMPABLE(rn) ( \
151 OP(rn) == OPEN || OP(rn) == CLOSE || OP(rn) == EVAL || \
152 OP(rn) == SUSPEND || OP(rn) == IFMATCH || \
153 OP(rn) == PLUS || OP(rn) == MINMOD || \
154 (PL_regkind[(U8)OP(rn)] == CURLY && ARG1(rn) > 0) \
157 #define HAS_TEXT(rn) ( \
158 PL_regkind[(U8)OP(rn)] == EXACT || PL_regkind[(U8)OP(rn)] == REF \
162 Search for mandatory following text node; for lookahead, the text must
163 follow but for lookbehind (rn->flags != 0) we skip to the next step.
165 #define FIND_NEXT_IMPT(rn) STMT_START { \
166 while (JUMPABLE(rn)) \
167 if (OP(rn) == SUSPEND || PL_regkind[(U8)OP(rn)] == CURLY) \
168 rn = NEXTOPER(NEXTOPER(rn)); \
169 else if (OP(rn) == PLUS) \
171 else if (OP(rn) == IFMATCH) \
172 rn = (rn->flags == 0) ? NEXTOPER(NEXTOPER(rn)) : rn + ARG(rn); \
173 else rn += NEXT_OFF(rn); \
176 static void restore_pos(pTHX_ void *arg);
179 S_regcppush(pTHX_ I32 parenfloor)
181 const int retval = PL_savestack_ix;
182 #define REGCP_PAREN_ELEMS 4
183 const int paren_elems_to_push = (PL_regsize - parenfloor) * REGCP_PAREN_ELEMS;
186 if (paren_elems_to_push < 0)
187 Perl_croak(aTHX_ "panic: paren_elems_to_push < 0");
189 #define REGCP_OTHER_ELEMS 6
190 SSGROW(paren_elems_to_push + REGCP_OTHER_ELEMS);
191 for (p = PL_regsize; p > parenfloor; p--) {
192 /* REGCP_PARENS_ELEMS are pushed per pairs of parentheses. */
193 SSPUSHINT(PL_regendp[p]);
194 SSPUSHINT(PL_regstartp[p]);
195 SSPUSHPTR(PL_reg_start_tmp[p]);
198 /* REGCP_OTHER_ELEMS are pushed in any case, parentheses or no. */
199 SSPUSHINT(PL_regsize);
200 SSPUSHINT(*PL_reglastparen);
201 SSPUSHINT(*PL_reglastcloseparen);
202 SSPUSHPTR(PL_reginput);
203 #define REGCP_FRAME_ELEMS 2
204 /* REGCP_FRAME_ELEMS are part of the REGCP_OTHER_ELEMS and
205 * are needed for the regexp context stack bookkeeping. */
206 SSPUSHINT(paren_elems_to_push + REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
207 SSPUSHINT(SAVEt_REGCONTEXT); /* Magic cookie. */
212 /* These are needed since we do not localize EVAL nodes: */
213 # define REGCP_SET(cp) DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, \
214 " Setting an EVAL scope, savestack=%"IVdf"\n", \
215 (IV)PL_savestack_ix)); cp = PL_savestack_ix
217 # define REGCP_UNWIND(cp) DEBUG_EXECUTE_r(cp != PL_savestack_ix ? \
218 PerlIO_printf(Perl_debug_log, \
219 " Clearing an EVAL scope, savestack=%"IVdf"..%"IVdf"\n", \
220 (IV)(cp), (IV)PL_savestack_ix) : 0); regcpblow(cp)
229 GET_RE_DEBUG_FLAGS_DECL;
231 /* Pop REGCP_OTHER_ELEMS before the parentheses loop starts. */
233 assert(i == SAVEt_REGCONTEXT); /* Check that the magic cookie is there. */
234 i = SSPOPINT; /* Parentheses elements to pop. */
235 input = (char *) SSPOPPTR;
236 *PL_reglastcloseparen = SSPOPINT;
237 *PL_reglastparen = SSPOPINT;
238 PL_regsize = SSPOPINT;
240 /* Now restore the parentheses context. */
241 for (i -= (REGCP_OTHER_ELEMS - REGCP_FRAME_ELEMS);
242 i > 0; i -= REGCP_PAREN_ELEMS) {
244 paren = (U32)SSPOPINT;
245 PL_reg_start_tmp[paren] = (char *) SSPOPPTR;
246 PL_regstartp[paren] = SSPOPINT;
248 if (paren <= *PL_reglastparen)
249 PL_regendp[paren] = tmps;
251 PerlIO_printf(Perl_debug_log,
252 " restoring \\%"UVuf" to %"IVdf"(%"IVdf")..%"IVdf"%s\n",
253 (UV)paren, (IV)PL_regstartp[paren],
254 (IV)(PL_reg_start_tmp[paren] - PL_bostr),
255 (IV)PL_regendp[paren],
256 (paren > *PL_reglastparen ? "(no)" : ""));
260 if ((I32)(*PL_reglastparen + 1) <= PL_regnpar) {
261 PerlIO_printf(Perl_debug_log,
262 " restoring \\%"IVdf"..\\%"IVdf" to undef\n",
263 (IV)(*PL_reglastparen + 1), (IV)PL_regnpar);
267 /* It would seem that the similar code in regtry()
268 * already takes care of this, and in fact it is in
269 * a better location to since this code can #if 0-ed out
270 * but the code in regtry() is needed or otherwise tests
271 * requiring null fields (pat.t#187 and split.t#{13,14}
272 * (as of patchlevel 7877) will fail. Then again,
273 * this code seems to be necessary or otherwise
274 * building DynaLoader will fail:
275 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
277 for (paren = *PL_reglastparen + 1; (I32)paren <= PL_regnpar; paren++) {
278 if ((I32)paren > PL_regsize)
279 PL_regstartp[paren] = -1;
280 PL_regendp[paren] = -1;
287 S_regcp_set_to(pTHX_ I32 ss)
289 const I32 tmp = PL_savestack_ix;
291 PL_savestack_ix = ss;
293 PL_savestack_ix = tmp;
297 typedef struct re_cc_state
301 struct re_cc_state *prev;
306 #define regcpblow(cp) LEAVE_SCOPE(cp) /* Ignores regcppush()ed data. */
308 #define TRYPAREN(paren, n, input) { \
311 PL_regstartp[paren] = HOPc(input, -1) - PL_bostr; \
312 PL_regendp[paren] = input - PL_bostr; \
315 PL_regendp[paren] = -1; \
317 if (regmatch(next)) \
320 PL_regendp[paren] = -1; \
325 * pregexec and friends
329 - pregexec - match a regexp against a string
332 Perl_pregexec(pTHX_ register regexp *prog, char *stringarg, register char *strend,
333 char *strbeg, I32 minend, SV *screamer, U32 nosave)
334 /* strend: pointer to null at end of string */
335 /* strbeg: real beginning of string */
336 /* minend: end of match must be >=minend after stringarg. */
337 /* nosave: For optimizations. */
340 regexec_flags(prog, stringarg, strend, strbeg, minend, screamer, NULL,
341 nosave ? 0 : REXEC_COPY_STR);
345 S_cache_re(pTHX_ regexp *prog)
347 PL_regprecomp = prog->precomp; /* Needed for FAIL. */
349 PL_regprogram = prog->program;
351 PL_regnpar = prog->nparens;
352 PL_regdata = prog->data;
357 * Need to implement the following flags for reg_anch:
359 * USE_INTUIT_NOML - Useful to call re_intuit_start() first
361 * INTUIT_AUTORITATIVE_NOML - Can trust a positive answer
362 * INTUIT_AUTORITATIVE_ML
363 * INTUIT_ONCE_NOML - Intuit can match in one location only.
366 * Another flag for this function: SECOND_TIME (so that float substrs
367 * with giant delta may be not rechecked).
370 /* Assumptions: if ANCH_GPOS, then strpos is anchored. XXXX Check GPOS logic */
372 /* If SCREAM, then SvPVX_const(sv) should be compatible with strpos and strend.
373 Otherwise, only SvCUR(sv) is used to get strbeg. */
375 /* XXXX We assume that strpos is strbeg unless sv. */
377 /* XXXX Some places assume that there is a fixed substring.
378 An update may be needed if optimizer marks as "INTUITable"
379 RExen without fixed substrings. Similarly, it is assumed that
380 lengths of all the strings are no more than minlen, thus they
381 cannot come from lookahead.
382 (Or minlen should take into account lookahead.) */
384 /* A failure to find a constant substring means that there is no need to make
385 an expensive call to REx engine, thus we celebrate a failure. Similarly,
386 finding a substring too deep into the string means that less calls to
387 regtry() should be needed.
389 REx compiler's optimizer found 4 possible hints:
390 a) Anchored substring;
392 c) Whether we are anchored (beginning-of-line or \G);
393 d) First node (of those at offset 0) which may distingush positions;
394 We use a)b)d) and multiline-part of c), and try to find a position in the
395 string which does not contradict any of them.
398 /* Most of decisions we do here should have been done at compile time.
399 The nodes of the REx which we used for the search should have been
400 deleted from the finite automaton. */
403 Perl_re_intuit_start(pTHX_ regexp *prog, SV *sv, char *strpos,
404 char *strend, U32 flags, re_scream_pos_data *data)
406 register I32 start_shift = 0;
407 /* Should be nonnegative! */
408 register I32 end_shift = 0;
413 const int do_utf8 = sv ? SvUTF8(sv) : 0; /* if no sv we have to assume bytes */
415 register char *other_last = Nullch; /* other substr checked before this */
416 char *check_at = Nullch; /* check substr found at this pos */
417 const I32 multiline = prog->reganch & PMf_MULTILINE;
419 const char * const i_strpos = strpos;
420 SV * const dsv = PERL_DEBUG_PAD_ZERO(0);
423 GET_RE_DEBUG_FLAGS_DECL;
425 RX_MATCH_UTF8_set(prog,do_utf8);
427 if (prog->reganch & ROPT_UTF8) {
428 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
429 "UTF-8 regex...\n"));
430 PL_reg_flags |= RF_utf8;
434 const char *s = PL_reg_match_utf8 ?
435 sv_uni_display(dsv, sv, 60, UNI_DISPLAY_REGEX) :
437 const int len = PL_reg_match_utf8 ?
438 strlen(s) : strend - strpos;
441 if (PL_reg_match_utf8)
442 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
443 "UTF-8 target...\n"));
444 PerlIO_printf(Perl_debug_log,
445 "%sGuessing start of match, REx%s \"%s%.60s%s%s\" against \"%s%.*s%s%s\"...\n",
446 PL_colors[4], PL_colors[5], PL_colors[0],
449 (strlen(prog->precomp) > 60 ? "..." : ""),
451 (int)(len > 60 ? 60 : len),
453 (len > 60 ? "..." : "")
457 /* CHR_DIST() would be more correct here but it makes things slow. */
458 if (prog->minlen > strend - strpos) {
459 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
460 "String too short... [re_intuit_start]\n"));
463 strbeg = (sv && SvPOK(sv)) ? strend - SvCUR(sv) : strpos;
466 if (!prog->check_utf8 && prog->check_substr)
467 to_utf8_substr(prog);
468 check = prog->check_utf8;
470 if (!prog->check_substr && prog->check_utf8)
471 to_byte_substr(prog);
472 check = prog->check_substr;
474 if (check == &PL_sv_undef) {
475 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
476 "Non-utf string cannot match utf check string\n"));
479 if (prog->reganch & ROPT_ANCH) { /* Match at beg-of-str or after \n */
480 ml_anch = !( (prog->reganch & ROPT_ANCH_SINGLE)
481 || ( (prog->reganch & ROPT_ANCH_BOL)
482 && !multiline ) ); /* Check after \n? */
485 if ( !(prog->reganch & (ROPT_ANCH_GPOS /* Checked by the caller */
486 | ROPT_IMPLICIT)) /* not a real BOL */
487 /* SvCUR is not set on references: SvRV and SvPVX_const overlap */
489 && (strpos != strbeg)) {
490 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not at start...\n"));
493 if (prog->check_offset_min == prog->check_offset_max &&
494 !(prog->reganch & ROPT_CANY_SEEN)) {
495 /* Substring at constant offset from beg-of-str... */
498 s = HOP3c(strpos, prog->check_offset_min, strend);
500 slen = SvCUR(check); /* >= 1 */
502 if ( strend - s > slen || strend - s < slen - 1
503 || (strend - s == slen && strend[-1] != '\n')) {
504 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String too long...\n"));
507 /* Now should match s[0..slen-2] */
509 if (slen && (*SvPVX_const(check) != *s
511 && memNE(SvPVX_const(check), s, slen)))) {
513 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "String not equal...\n"));
517 else if (*SvPVX_const(check) != *s
518 || ((slen = SvCUR(check)) > 1
519 && memNE(SvPVX_const(check), s, slen)))
521 goto success_at_start;
524 /* Match is anchored, but substr is not anchored wrt beg-of-str. */
526 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
527 end_shift = prog->minlen - start_shift -
528 CHR_SVLEN(check) + (SvTAIL(check) != 0);
530 const I32 end = prog->check_offset_max + CHR_SVLEN(check)
531 - (SvTAIL(check) != 0);
532 const I32 eshift = CHR_DIST((U8*)strend, (U8*)s) - end;
534 if (end_shift < eshift)
538 else { /* Can match at random position */
541 start_shift = prog->check_offset_min; /* okay to underestimate on CC */
542 /* Should be nonnegative! */
543 end_shift = prog->minlen - start_shift -
544 CHR_SVLEN(check) + (SvTAIL(check) != 0);
547 #ifdef DEBUGGING /* 7/99: reports of failure (with the older version) */
549 Perl_croak(aTHX_ "panic: end_shift");
553 /* Find a possible match in the region s..strend by looking for
554 the "check" substring in the region corrected by start/end_shift. */
555 if (flags & REXEC_SCREAM) {
556 I32 p = -1; /* Internal iterator of scream. */
557 I32 * const pp = data ? data->scream_pos : &p;
559 if (PL_screamfirst[BmRARE(check)] >= 0
560 || ( BmRARE(check) == '\n'
561 && (BmPREVIOUS(check) == SvCUR(check) - 1)
563 s = screaminstr(sv, check,
564 start_shift + (s - strbeg), end_shift, pp, 0);
567 /* we may be pointing at the wrong string */
568 if (s && RX_MATCH_COPIED(prog))
569 s = strbeg + (s - SvPVX_const(sv));
571 *data->scream_olds = s;
573 else if (prog->reganch & ROPT_CANY_SEEN)
574 s = fbm_instr((U8*)(s + start_shift),
575 (U8*)(strend - end_shift),
576 check, multiline ? FBMrf_MULTILINE : 0);
578 s = fbm_instr(HOP3(s, start_shift, strend),
579 HOP3(strend, -end_shift, strbeg),
580 check, multiline ? FBMrf_MULTILINE : 0);
582 /* Update the count-of-usability, remove useless subpatterns,
585 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s %s substr \"%s%.*s%s\"%s%s",
586 (s ? "Found" : "Did not find"),
587 (check == (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) ? "anchored" : "floating"),
589 (int)(SvCUR(check) - (SvTAIL(check)!=0)),
591 PL_colors[1], (SvTAIL(check) ? "$" : ""),
592 (s ? " at offset " : "...\n") ) );
599 /* Finish the diagnostic message */
600 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%ld...\n", (long)(s - i_strpos)) );
602 /* Got a candidate. Check MBOL anchoring, and the *other* substr.
603 Start with the other substr.
604 XXXX no SCREAM optimization yet - and a very coarse implementation
605 XXXX /ttx+/ results in anchored="ttx", floating="x". floating will
606 *always* match. Probably should be marked during compile...
607 Probably it is right to do no SCREAM here...
610 if (do_utf8 ? (prog->float_utf8 && prog->anchored_utf8) : (prog->float_substr && prog->anchored_substr)) {
611 /* Take into account the "other" substring. */
612 /* XXXX May be hopelessly wrong for UTF... */
615 if (check == (do_utf8 ? prog->float_utf8 : prog->float_substr)) {
618 char * const last = HOP3c(s, -start_shift, strbeg);
623 t = s - prog->check_offset_max;
624 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
626 || ((t = reghopmaybe3_c(s, -(prog->check_offset_max), strpos))
631 t = HOP3c(t, prog->anchored_offset, strend);
632 if (t < other_last) /* These positions already checked */
634 last2 = last1 = HOP3c(strend, -prog->minlen, strbeg);
637 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
638 /* On end-of-str: see comment below. */
639 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
640 if (must == &PL_sv_undef) {
642 DEBUG_EXECUTE_r(must = prog->anchored_utf8); /* for debug */
647 HOP3(HOP3(last1, prog->anchored_offset, strend)
648 + SvCUR(must), -(SvTAIL(must)!=0), strbeg),
650 multiline ? FBMrf_MULTILINE : 0
652 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
653 "%s anchored substr \"%s%.*s%s\"%s",
654 (s ? "Found" : "Contradicts"),
657 - (SvTAIL(must)!=0)),
659 PL_colors[1], (SvTAIL(must) ? "$" : "")));
661 if (last1 >= last2) {
662 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
663 ", giving up...\n"));
666 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
667 ", trying floating at offset %ld...\n",
668 (long)(HOP3c(s1, 1, strend) - i_strpos)));
669 other_last = HOP3c(last1, prog->anchored_offset+1, strend);
670 s = HOP3c(last, 1, strend);
674 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
675 (long)(s - i_strpos)));
676 t = HOP3c(s, -prog->anchored_offset, strbeg);
677 other_last = HOP3c(s, 1, strend);
685 else { /* Take into account the floating substring. */
690 t = HOP3c(s, -start_shift, strbeg);
692 HOP3c(strend, -prog->minlen + prog->float_min_offset, strbeg);
693 if (CHR_DIST((U8*)last, (U8*)t) > prog->float_max_offset)
694 last = HOP3c(t, prog->float_max_offset, strend);
695 s = HOP3c(t, prog->float_min_offset, strend);
698 /* XXXX It is not documented what units *_offsets are in. Assume bytes. */
699 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
700 /* fbm_instr() takes into account exact value of end-of-str
701 if the check is SvTAIL(ed). Since false positives are OK,
702 and end-of-str is not later than strend we are OK. */
703 if (must == &PL_sv_undef) {
705 DEBUG_EXECUTE_r(must = prog->float_utf8); /* for debug message */
708 s = fbm_instr((unsigned char*)s,
709 (unsigned char*)last + SvCUR(must)
711 must, multiline ? FBMrf_MULTILINE : 0);
712 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s floating substr \"%s%.*s%s\"%s",
713 (s ? "Found" : "Contradicts"),
715 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
717 PL_colors[1], (SvTAIL(must) ? "$" : "")));
720 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
721 ", giving up...\n"));
724 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
725 ", trying anchored starting at offset %ld...\n",
726 (long)(s1 + 1 - i_strpos)));
728 s = HOP3c(t, 1, strend);
732 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, " at offset %ld...\n",
733 (long)(s - i_strpos)));
734 other_last = s; /* Fix this later. --Hugo */
743 t = s - prog->check_offset_max;
744 if (s - strpos > prog->check_offset_max /* signed-corrected t > strpos */
746 || ((t = reghopmaybe3_c(s, -prog->check_offset_max, strpos))
748 /* Fixed substring is found far enough so that the match
749 cannot start at strpos. */
751 if (ml_anch && t[-1] != '\n') {
752 /* Eventually fbm_*() should handle this, but often
753 anchored_offset is not 0, so this check will not be wasted. */
754 /* XXXX In the code below we prefer to look for "^" even in
755 presence of anchored substrings. And we search even
756 beyond the found float position. These pessimizations
757 are historical artefacts only. */
759 while (t < strend - prog->minlen) {
761 if (t < check_at - prog->check_offset_min) {
762 if (do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) {
763 /* Since we moved from the found position,
764 we definitely contradict the found anchored
765 substr. Due to the above check we do not
766 contradict "check" substr.
767 Thus we can arrive here only if check substr
768 is float. Redo checking for "other"=="fixed".
771 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld, rescanning for anchored from offset %ld...\n",
772 PL_colors[0], PL_colors[1], (long)(strpos - i_strpos), (long)(strpos - i_strpos + prog->anchored_offset)));
773 goto do_other_anchored;
775 /* We don't contradict the found floating substring. */
776 /* XXXX Why not check for STCLASS? */
778 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m at offset %ld...\n",
779 PL_colors[0], PL_colors[1], (long)(s - i_strpos)));
782 /* Position contradicts check-string */
783 /* XXXX probably better to look for check-string
784 than for "\n", so one should lower the limit for t? */
785 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Found /%s^%s/m, restarting lookup for check-string at offset %ld...\n",
786 PL_colors[0], PL_colors[1], (long)(t + 1 - i_strpos)));
787 other_last = strpos = s = t + 1;
792 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Did not find /%s^%s/m...\n",
793 PL_colors[0], PL_colors[1]));
797 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Starting position does not contradict /%s^%s/m...\n",
798 PL_colors[0], PL_colors[1]));
802 ++BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr); /* hooray/5 */
805 /* The found string does not prohibit matching at strpos,
806 - no optimization of calling REx engine can be performed,
807 unless it was an MBOL and we are not after MBOL,
808 or a future STCLASS check will fail this. */
810 /* Even in this situation we may use MBOL flag if strpos is offset
811 wrt the start of the string. */
812 if (ml_anch && sv && !SvROK(sv) /* See prev comment on SvROK */
813 && (strpos != strbeg) && strpos[-1] != '\n'
814 /* May be due to an implicit anchor of m{.*foo} */
815 && !(prog->reganch & ROPT_IMPLICIT))
820 DEBUG_EXECUTE_r( if (ml_anch)
821 PerlIO_printf(Perl_debug_log, "Position at offset %ld does not contradict /%s^%s/m...\n",
822 (long)(strpos - i_strpos), PL_colors[0], PL_colors[1]);
825 if (!(prog->reganch & ROPT_NAUGHTY) /* XXXX If strpos moved? */
827 prog->check_utf8 /* Could be deleted already */
828 && --BmUSEFUL(prog->check_utf8) < 0
829 && (prog->check_utf8 == prog->float_utf8)
831 prog->check_substr /* Could be deleted already */
832 && --BmUSEFUL(prog->check_substr) < 0
833 && (prog->check_substr == prog->float_substr)
836 /* If flags & SOMETHING - do not do it many times on the same match */
837 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "... Disabling check substring...\n"));
838 SvREFCNT_dec(do_utf8 ? prog->check_utf8 : prog->check_substr);
839 if (do_utf8 ? prog->check_substr : prog->check_utf8)
840 SvREFCNT_dec(do_utf8 ? prog->check_substr : prog->check_utf8);
841 prog->check_substr = prog->check_utf8 = Nullsv; /* disable */
842 prog->float_substr = prog->float_utf8 = Nullsv; /* clear */
843 check = Nullsv; /* abort */
845 /* XXXX This is a remnant of the old implementation. It
846 looks wasteful, since now INTUIT can use many
848 prog->reganch &= ~RE_USE_INTUIT;
855 /* XXXX BmUSEFUL already changed, maybe multiple change is meaningful... */
856 if (prog->regstclass) {
857 /* minlen == 0 is possible if regstclass is \b or \B,
858 and the fixed substr is ''$.
859 Since minlen is already taken into account, s+1 is before strend;
860 accidentally, minlen >= 1 guaranties no false positives at s + 1
861 even for \b or \B. But (minlen? 1 : 0) below assumes that
862 regstclass does not come from lookahead... */
863 /* If regstclass takes bytelength more than 1: If charlength==1, OK.
864 This leaves EXACTF only, which is dealt with in find_byclass(). */
865 const U8* const str = (U8*)STRING(prog->regstclass);
866 const int cl_l = (PL_regkind[(U8)OP(prog->regstclass)] == EXACT
867 ? CHR_DIST(str+STR_LEN(prog->regstclass), str)
869 const char * const endpos = (prog->anchored_substr || prog->anchored_utf8 || ml_anch)
870 ? HOP3c(s, (prog->minlen ? cl_l : 0), strend)
871 : (prog->float_substr || prog->float_utf8
872 ? HOP3c(HOP3c(check_at, -start_shift, strbeg),
878 s = find_byclass(prog, prog->regstclass, s, endpos, 1);
881 const char *what = 0;
883 if (endpos == strend) {
884 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
885 "Could not match STCLASS...\n") );
888 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
889 "This position contradicts STCLASS...\n") );
890 if ((prog->reganch & ROPT_ANCH) && !ml_anch)
892 /* Contradict one of substrings */
893 if (prog->anchored_substr || prog->anchored_utf8) {
894 if ((do_utf8 ? prog->anchored_utf8 : prog->anchored_substr) == check) {
895 DEBUG_EXECUTE_r( what = "anchored" );
897 s = HOP3c(t, 1, strend);
898 if (s + start_shift + end_shift > strend) {
899 /* XXXX Should be taken into account earlier? */
900 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
901 "Could not match STCLASS...\n") );
906 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
907 "Looking for %s substr starting at offset %ld...\n",
908 what, (long)(s + start_shift - i_strpos)) );
911 /* Have both, check_string is floating */
912 if (t + start_shift >= check_at) /* Contradicts floating=check */
913 goto retry_floating_check;
914 /* Recheck anchored substring, but not floating... */
918 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
919 "Looking for anchored substr starting at offset %ld...\n",
920 (long)(other_last - i_strpos)) );
921 goto do_other_anchored;
923 /* Another way we could have checked stclass at the
924 current position only: */
929 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log,
930 "Looking for /%s^%s/m starting at offset %ld...\n",
931 PL_colors[0], PL_colors[1], (long)(t - i_strpos)) );
934 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr)) /* Could have been deleted */
936 /* Check is floating subtring. */
937 retry_floating_check:
938 t = check_at - start_shift;
939 DEBUG_EXECUTE_r( what = "floating" );
940 goto hop_and_restart;
943 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
944 "By STCLASS: moving %ld --> %ld\n",
945 (long)(t - i_strpos), (long)(s - i_strpos))
949 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
950 "Does not contradict STCLASS...\n");
955 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%s%s:%s match at offset %ld\n",
956 PL_colors[4], (check ? "Guessed" : "Giving up"),
957 PL_colors[5], (long)(s - i_strpos)) );
960 fail_finish: /* Substring not found */
961 if (prog->check_substr || prog->check_utf8) /* could be removed already */
962 BmUSEFUL(do_utf8 ? prog->check_utf8 : prog->check_substr) += 5; /* hooray */
964 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch rejected by optimizer%s\n",
965 PL_colors[4], PL_colors[5]));
969 /* We know what class REx starts with. Try to find this position... */
971 S_find_byclass(pTHX_ regexp * prog, regnode *c, char *s, const char *strend, I32 norun)
974 const I32 doevery = (prog->reganch & ROPT_SKIP) == 0;
978 register STRLEN uskip;
982 register I32 tmp = 1; /* Scratch variable? */
983 register const bool do_utf8 = PL_reg_match_utf8;
985 /* We know what class it must start with. */
989 while (s + (uskip = UTF8SKIP(s)) <= strend) {
990 if ((ANYOF_FLAGS(c) & ANYOF_UNICODE) ||
991 !UTF8_IS_INVARIANT((U8)s[0]) ?
992 reginclass(c, (U8*)s, 0, do_utf8) :
993 REGINCLASS(c, (U8*)s)) {
994 if (tmp && (norun || regtry(prog, s)))
1005 while (s < strend) {
1008 if (REGINCLASS(c, (U8*)s) ||
1009 (ANYOF_FOLD_SHARP_S(c, s, strend) &&
1010 /* The assignment of 2 is intentional:
1011 * for the folded sharp s, the skip is 2. */
1012 (skip = SHARP_S_SKIP))) {
1013 if (tmp && (norun || regtry(prog, s)))
1025 while (s < strend) {
1026 if (tmp && (norun || regtry(prog, s)))
1035 ln = STR_LEN(c); /* length to match in octets/bytes */
1036 lnc = (I32) ln; /* length to match in characters */
1038 STRLEN ulen1, ulen2;
1040 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
1041 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
1042 const U32 uniflags = ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY;
1044 to_utf8_lower((U8*)m, tmpbuf1, &ulen1);
1045 to_utf8_upper((U8*)m, tmpbuf2, &ulen2);
1047 c1 = utf8n_to_uvchr(tmpbuf1, UTF8_MAXBYTES_CASE,
1049 c2 = utf8n_to_uvchr(tmpbuf2, UTF8_MAXBYTES_CASE,
1052 while (sm < ((U8 *) m + ln)) {
1067 c2 = PL_fold_locale[c1];
1069 e = HOP3c(strend, -((I32)lnc), s);
1072 e = s; /* Due to minlen logic of intuit() */
1074 /* The idea in the EXACTF* cases is to first find the
1075 * first character of the EXACTF* node and then, if
1076 * necessary, case-insensitively compare the full
1077 * text of the node. The c1 and c2 are the first
1078 * characters (though in Unicode it gets a bit
1079 * more complicated because there are more cases
1080 * than just upper and lower: one needs to use
1081 * the so-called folding case for case-insensitive
1082 * matching (called "loose matching" in Unicode).
1083 * ibcmp_utf8() will do just that. */
1087 U8 tmpbuf [UTF8_MAXBYTES+1];
1088 STRLEN len, foldlen;
1089 const U32 uniflags = ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY;
1091 /* Upper and lower of 1st char are equal -
1092 * probably not a "letter". */
1094 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
1098 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1099 m, (char **)0, ln, (bool)UTF))
1100 && (norun || regtry(prog, s)) )
1103 U8 foldbuf[UTF8_MAXBYTES_CASE+1];
1104 uvchr_to_utf8(tmpbuf, c);
1105 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1107 && (f == c1 || f == c2)
1108 && (ln == foldlen ||
1109 !ibcmp_utf8((char *) foldbuf,
1110 (char **)0, foldlen, do_utf8,
1112 (char **)0, ln, (bool)UTF))
1113 && (norun || regtry(prog, s)) )
1121 c = utf8n_to_uvchr((U8*)s, UTF8_MAXBYTES, &len,
1124 /* Handle some of the three Greek sigmas cases.
1125 * Note that not all the possible combinations
1126 * are handled here: some of them are handled
1127 * by the standard folding rules, and some of
1128 * them (the character class or ANYOF cases)
1129 * are handled during compiletime in
1130 * regexec.c:S_regclass(). */
1131 if (c == (UV)UNICODE_GREEK_CAPITAL_LETTER_SIGMA ||
1132 c == (UV)UNICODE_GREEK_SMALL_LETTER_FINAL_SIGMA)
1133 c = (UV)UNICODE_GREEK_SMALL_LETTER_SIGMA;
1135 if ( (c == c1 || c == c2)
1137 ibcmp_utf8(s, (char **)0, 0, do_utf8,
1138 m, (char **)0, ln, (bool)UTF))
1139 && (norun || regtry(prog, s)) )
1142 U8 foldbuf[UTF8_MAXBYTES_CASE+1];
1143 uvchr_to_utf8(tmpbuf, c);
1144 f = to_utf8_fold(tmpbuf, foldbuf, &foldlen);
1146 && (f == c1 || f == c2)
1147 && (ln == foldlen ||
1148 !ibcmp_utf8((char *) foldbuf,
1149 (char **)0, foldlen, do_utf8,
1151 (char **)0, ln, (bool)UTF))
1152 && (norun || regtry(prog, s)) )
1163 && (ln == 1 || !(OP(c) == EXACTF
1165 : ibcmp_locale(s, m, ln)))
1166 && (norun || regtry(prog, s)) )
1172 if ( (*(U8*)s == c1 || *(U8*)s == c2)
1173 && (ln == 1 || !(OP(c) == EXACTF
1175 : ibcmp_locale(s, m, ln)))
1176 && (norun || regtry(prog, s)) )
1183 PL_reg_flags |= RF_tainted;
1190 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1192 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1194 tmp = ((OP(c) == BOUND ?
1195 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1196 LOAD_UTF8_CHARCLASS_ALNUM();
1197 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1198 if (tmp == !(OP(c) == BOUND ?
1199 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1200 isALNUM_LC_utf8((U8*)s)))
1203 if ((norun || regtry(prog, s)))
1210 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1211 tmp = ((OP(c) == BOUND ? isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1212 while (s < strend) {
1214 !(OP(c) == BOUND ? isALNUM(*s) : isALNUM_LC(*s))) {
1216 if ((norun || regtry(prog, s)))
1222 if ((!prog->minlen && tmp) && (norun || regtry(prog, s)))
1226 PL_reg_flags |= RF_tainted;
1233 U8 *r = reghop3((U8*)s, -1, (U8*)PL_bostr);
1235 tmp = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
1237 tmp = ((OP(c) == NBOUND ?
1238 isALNUM_uni(tmp) : isALNUM_LC_uvchr(UNI_TO_NATIVE(tmp))) != 0);
1239 LOAD_UTF8_CHARCLASS_ALNUM();
1240 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1241 if (tmp == !(OP(c) == NBOUND ?
1242 swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8) :
1243 isALNUM_LC_utf8((U8*)s)))
1245 else if ((norun || regtry(prog, s)))
1251 tmp = (s != PL_bostr) ? UCHARAT(s - 1) : '\n';
1252 tmp = ((OP(c) == NBOUND ?
1253 isALNUM(tmp) : isALNUM_LC(tmp)) != 0);
1254 while (s < strend) {
1256 !(OP(c) == NBOUND ? isALNUM(*s) : isALNUM_LC(*s)))
1258 else if ((norun || regtry(prog, s)))
1263 if ((!prog->minlen && !tmp) && (norun || regtry(prog, s)))
1268 LOAD_UTF8_CHARCLASS_ALNUM();
1269 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1270 if (swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1271 if (tmp && (norun || regtry(prog, s)))
1282 while (s < strend) {
1284 if (tmp && (norun || regtry(prog, s)))
1296 PL_reg_flags |= RF_tainted;
1298 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1299 if (isALNUM_LC_utf8((U8*)s)) {
1300 if (tmp && (norun || regtry(prog, s)))
1311 while (s < strend) {
1312 if (isALNUM_LC(*s)) {
1313 if (tmp && (norun || regtry(prog, s)))
1326 LOAD_UTF8_CHARCLASS_ALNUM();
1327 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1328 if (!swash_fetch(PL_utf8_alnum, (U8*)s, do_utf8)) {
1329 if (tmp && (norun || regtry(prog, s)))
1340 while (s < strend) {
1342 if (tmp && (norun || regtry(prog, s)))
1354 PL_reg_flags |= RF_tainted;
1356 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1357 if (!isALNUM_LC_utf8((U8*)s)) {
1358 if (tmp && (norun || regtry(prog, s)))
1369 while (s < strend) {
1370 if (!isALNUM_LC(*s)) {
1371 if (tmp && (norun || regtry(prog, s)))
1384 LOAD_UTF8_CHARCLASS_SPACE();
1385 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1386 if (*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8)) {
1387 if (tmp && (norun || regtry(prog, s)))
1398 while (s < strend) {
1400 if (tmp && (norun || regtry(prog, s)))
1412 PL_reg_flags |= RF_tainted;
1414 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1415 if (*s == ' ' || isSPACE_LC_utf8((U8*)s)) {
1416 if (tmp && (norun || regtry(prog, s)))
1427 while (s < strend) {
1428 if (isSPACE_LC(*s)) {
1429 if (tmp && (norun || regtry(prog, s)))
1442 LOAD_UTF8_CHARCLASS_SPACE();
1443 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1444 if (!(*s == ' ' || swash_fetch(PL_utf8_space,(U8*)s, do_utf8))) {
1445 if (tmp && (norun || regtry(prog, s)))
1456 while (s < strend) {
1458 if (tmp && (norun || regtry(prog, s)))
1470 PL_reg_flags |= RF_tainted;
1472 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1473 if (!(*s == ' ' || isSPACE_LC_utf8((U8*)s))) {
1474 if (tmp && (norun || regtry(prog, s)))
1485 while (s < strend) {
1486 if (!isSPACE_LC(*s)) {
1487 if (tmp && (norun || regtry(prog, s)))
1500 LOAD_UTF8_CHARCLASS_DIGIT();
1501 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1502 if (swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1503 if (tmp && (norun || regtry(prog, s)))
1514 while (s < strend) {
1516 if (tmp && (norun || regtry(prog, s)))
1528 PL_reg_flags |= RF_tainted;
1530 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1531 if (isDIGIT_LC_utf8((U8*)s)) {
1532 if (tmp && (norun || regtry(prog, s)))
1543 while (s < strend) {
1544 if (isDIGIT_LC(*s)) {
1545 if (tmp && (norun || regtry(prog, s)))
1558 LOAD_UTF8_CHARCLASS_DIGIT();
1559 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1560 if (!swash_fetch(PL_utf8_digit,(U8*)s, do_utf8)) {
1561 if (tmp && (norun || regtry(prog, s)))
1572 while (s < strend) {
1574 if (tmp && (norun || regtry(prog, s)))
1586 PL_reg_flags |= RF_tainted;
1588 while (s + (uskip = UTF8SKIP(s)) <= strend) {
1589 if (!isDIGIT_LC_utf8((U8*)s)) {
1590 if (tmp && (norun || regtry(prog, s)))
1601 while (s < strend) {
1602 if (!isDIGIT_LC(*s)) {
1603 if (tmp && (norun || regtry(prog, s)))
1615 Perl_croak(aTHX_ "panic: unknown regstclass %d", (int)OP(c));
1624 - regexec_flags - match a regexp against a string
1627 Perl_regexec_flags(pTHX_ register regexp *prog, char *stringarg, register char *strend,
1628 char *strbeg, I32 minend, SV *sv, void *data, U32 flags)
1629 /* strend: pointer to null at end of string */
1630 /* strbeg: real beginning of string */
1631 /* minend: end of match must be >=minend after stringarg. */
1632 /* data: May be used for some additional optimizations. */
1633 /* nosave: For optimizations. */
1636 register regnode *c;
1637 register char *startpos = stringarg;
1638 I32 minlen; /* must match at least this many chars */
1639 I32 dontbother = 0; /* how many characters not to try at end */
1640 I32 end_shift = 0; /* Same for the end. */ /* CC */
1641 I32 scream_pos = -1; /* Internal iterator of scream. */
1643 SV* oreplsv = GvSV(PL_replgv);
1644 const bool do_utf8 = DO_UTF8(sv);
1645 const I32 multiline = prog->reganch & PMf_MULTILINE;
1647 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
1648 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
1651 GET_RE_DEBUG_FLAGS_DECL;
1653 PERL_UNUSED_ARG(data);
1654 RX_MATCH_UTF8_set(prog,do_utf8);
1660 PL_regnarrate = DEBUG_r_TEST;
1663 /* Be paranoid... */
1664 if (prog == NULL || startpos == NULL) {
1665 Perl_croak(aTHX_ "NULL regexp parameter");
1669 minlen = prog->minlen;
1670 if (strend - startpos < minlen) {
1671 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
1672 "String too short [regexec_flags]...\n"));
1676 /* Check validity of program. */
1677 if (UCHARAT(prog->program) != REG_MAGIC) {
1678 Perl_croak(aTHX_ "corrupted regexp program");
1682 PL_reg_eval_set = 0;
1685 if (prog->reganch & ROPT_UTF8)
1686 PL_reg_flags |= RF_utf8;
1688 /* Mark beginning of line for ^ and lookbehind. */
1689 PL_regbol = startpos;
1693 /* Mark end of line for $ (and such) */
1696 /* see how far we have to get to not match where we matched before */
1697 PL_regtill = startpos+minend;
1699 /* We start without call_cc context. */
1702 /* If there is a "must appear" string, look for it. */
1705 if (prog->reganch & ROPT_GPOS_SEEN) { /* Need to have PL_reg_ganch */
1708 if (flags & REXEC_IGNOREPOS) /* Means: check only at start */
1709 PL_reg_ganch = startpos;
1710 else if (sv && SvTYPE(sv) >= SVt_PVMG
1712 && (mg = mg_find(sv, PERL_MAGIC_regex_global))
1713 && mg->mg_len >= 0) {
1714 PL_reg_ganch = strbeg + mg->mg_len; /* Defined pos() */
1715 if (prog->reganch & ROPT_ANCH_GPOS) {
1716 if (s > PL_reg_ganch)
1721 else /* pos() not defined */
1722 PL_reg_ganch = strbeg;
1725 if (!(flags & REXEC_CHECKED) && (prog->check_substr != Nullsv || prog->check_utf8 != Nullsv)) {
1726 re_scream_pos_data d;
1728 d.scream_olds = &scream_olds;
1729 d.scream_pos = &scream_pos;
1730 s = re_intuit_start(prog, sv, s, strend, flags, &d);
1732 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Not present...\n"));
1733 goto phooey; /* not present */
1738 const char * const s0 = UTF
1739 ? pv_uni_display(dsv0, (U8*)prog->precomp, prog->prelen, 60,
1742 const int len0 = UTF ? SvCUR(dsv0) : prog->prelen;
1743 const char * const s1 = do_utf8 ? sv_uni_display(dsv1, sv, 60,
1744 UNI_DISPLAY_REGEX) : startpos;
1745 const int len1 = do_utf8 ? SvCUR(dsv1) : strend - startpos;
1748 PerlIO_printf(Perl_debug_log,
1749 "%sMatching REx%s \"%s%*.*s%s%s\" against \"%s%.*s%s%s\"\n",
1750 PL_colors[4], PL_colors[5], PL_colors[0],
1753 len0 > 60 ? "..." : "",
1755 (int)(len1 > 60 ? 60 : len1),
1757 (len1 > 60 ? "..." : "")
1761 /* Simplest case: anchored match need be tried only once. */
1762 /* [unless only anchor is BOL and multiline is set] */
1763 if (prog->reganch & (ROPT_ANCH & ~ROPT_ANCH_GPOS)) {
1764 if (s == startpos && regtry(prog, startpos))
1766 else if (multiline || (prog->reganch & ROPT_IMPLICIT)
1767 || (prog->reganch & ROPT_ANCH_MBOL)) /* XXXX SBOL? */
1772 dontbother = minlen - 1;
1773 end = HOP3c(strend, -dontbother, strbeg) - 1;
1774 /* for multiline we only have to try after newlines */
1775 if (prog->check_substr || prog->check_utf8) {
1779 if (regtry(prog, s))
1784 if (prog->reganch & RE_USE_INTUIT) {
1785 s = re_intuit_start(prog, sv, s + 1, strend, flags, NULL);
1796 if (*s++ == '\n') { /* don't need PL_utf8skip here */
1797 if (regtry(prog, s))
1804 } else if (prog->reganch & ROPT_ANCH_GPOS) {
1805 if (regtry(prog, PL_reg_ganch))
1810 /* Messy cases: unanchored match. */
1811 if ((prog->anchored_substr || prog->anchored_utf8) && prog->reganch & ROPT_SKIP) {
1812 /* we have /x+whatever/ */
1813 /* it must be a one character string (XXXX Except UTF?) */
1818 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1819 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1820 ch = SvPVX_const(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr)[0];
1823 while (s < strend) {
1825 DEBUG_EXECUTE_r( did_match = 1 );
1826 if (regtry(prog, s)) goto got_it;
1828 while (s < strend && *s == ch)
1835 while (s < strend) {
1837 DEBUG_EXECUTE_r( did_match = 1 );
1838 if (regtry(prog, s)) goto got_it;
1840 while (s < strend && *s == ch)
1846 DEBUG_EXECUTE_r(if (!did_match)
1847 PerlIO_printf(Perl_debug_log,
1848 "Did not find anchored character...\n")
1851 else if (prog->anchored_substr != Nullsv
1852 || prog->anchored_utf8 != Nullsv
1853 || ((prog->float_substr != Nullsv || prog->float_utf8 != Nullsv)
1854 && prog->float_max_offset < strend - s)) {
1859 char *last1; /* Last position checked before */
1863 if (prog->anchored_substr || prog->anchored_utf8) {
1864 if (!(do_utf8 ? prog->anchored_utf8 : prog->anchored_substr))
1865 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1866 must = do_utf8 ? prog->anchored_utf8 : prog->anchored_substr;
1867 back_max = back_min = prog->anchored_offset;
1869 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1870 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1871 must = do_utf8 ? prog->float_utf8 : prog->float_substr;
1872 back_max = prog->float_max_offset;
1873 back_min = prog->float_min_offset;
1875 if (must == &PL_sv_undef)
1876 /* could not downgrade utf8 check substring, so must fail */
1879 last = HOP3c(strend, /* Cannot start after this */
1880 -(I32)(CHR_SVLEN(must)
1881 - (SvTAIL(must) != 0) + back_min), strbeg);
1884 last1 = HOPc(s, -1);
1886 last1 = s - 1; /* bogus */
1888 /* XXXX check_substr already used to find "s", can optimize if
1889 check_substr==must. */
1891 dontbother = end_shift;
1892 strend = HOPc(strend, -dontbother);
1893 while ( (s <= last) &&
1894 ((flags & REXEC_SCREAM)
1895 ? (s = screaminstr(sv, must, HOP3c(s, back_min, strend) - strbeg,
1896 end_shift, &scream_pos, 0))
1897 : (s = fbm_instr((unsigned char*)HOP3(s, back_min, strend),
1898 (unsigned char*)strend, must,
1899 multiline ? FBMrf_MULTILINE : 0))) ) {
1900 /* we may be pointing at the wrong string */
1901 if ((flags & REXEC_SCREAM) && RX_MATCH_COPIED(prog))
1902 s = strbeg + (s - SvPVX_const(sv));
1903 DEBUG_EXECUTE_r( did_match = 1 );
1904 if (HOPc(s, -back_max) > last1) {
1905 last1 = HOPc(s, -back_min);
1906 s = HOPc(s, -back_max);
1909 char *t = (last1 >= PL_bostr) ? HOPc(last1, 1) : last1 + 1;
1911 last1 = HOPc(s, -back_min);
1915 while (s <= last1) {
1916 if (regtry(prog, s))
1922 while (s <= last1) {
1923 if (regtry(prog, s))
1929 DEBUG_EXECUTE_r(if (!did_match)
1930 PerlIO_printf(Perl_debug_log,
1931 "Did not find %s substr \"%s%.*s%s\"%s...\n",
1932 ((must == prog->anchored_substr || must == prog->anchored_utf8)
1933 ? "anchored" : "floating"),
1935 (int)(SvCUR(must) - (SvTAIL(must)!=0)),
1937 PL_colors[1], (SvTAIL(must) ? "$" : ""))
1941 else if ((c = prog->regstclass)) {
1943 I32 op = (U8)OP(prog->regstclass);
1944 /* don't bother with what can't match */
1945 if (PL_regkind[op] != EXACT && op != CANY)
1946 strend = HOPc(strend, -(minlen - 1));
1949 SV *prop = sv_newmortal();
1957 pv_uni_display(dsv0, (U8*)SvPVX_const(prop), SvCUR(prop), 60,
1958 UNI_DISPLAY_REGEX) :
1960 len0 = UTF ? SvCUR(dsv0) : SvCUR(prop);
1962 sv_uni_display(dsv1, sv, 60, UNI_DISPLAY_REGEX) : s;
1963 len1 = UTF ? SvCUR(dsv1) : strend - s;
1964 PerlIO_printf(Perl_debug_log,
1965 "Matching stclass \"%*.*s\" against \"%*.*s\"\n",
1969 if (find_byclass(prog, c, s, strend, 0))
1971 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "Contradicts stclass...\n"));
1975 if (prog->float_substr != Nullsv || prog->float_utf8 != Nullsv) {
1980 if (!(do_utf8 ? prog->float_utf8 : prog->float_substr))
1981 do_utf8 ? to_utf8_substr(prog) : to_byte_substr(prog);
1982 float_real = do_utf8 ? prog->float_utf8 : prog->float_substr;
1984 if (flags & REXEC_SCREAM) {
1985 last = screaminstr(sv, float_real, s - strbeg,
1986 end_shift, &scream_pos, 1); /* last one */
1988 last = scream_olds; /* Only one occurrence. */
1989 /* we may be pointing at the wrong string */
1990 else if (RX_MATCH_COPIED(prog))
1991 s = strbeg + (s - SvPVX_const(sv));
1995 const char * const little = SvPV_const(float_real, len);
1997 if (SvTAIL(float_real)) {
1998 if (memEQ(strend - len + 1, little, len - 1))
1999 last = strend - len + 1;
2000 else if (!multiline)
2001 last = memEQ(strend - len, little, len)
2002 ? strend - len : Nullch;
2008 last = rninstr(s, strend, little, little + len);
2010 last = strend; /* matching "$" */
2014 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
2015 "%sCan't trim the tail, match fails (should not happen)%s\n",
2016 PL_colors[4], PL_colors[5]));
2017 goto phooey; /* Should not happen! */
2019 dontbother = strend - last + prog->float_min_offset;
2021 if (minlen && (dontbother < minlen))
2022 dontbother = minlen - 1;
2023 strend -= dontbother; /* this one's always in bytes! */
2024 /* We don't know much -- general case. */
2027 if (regtry(prog, s))
2036 if (regtry(prog, s))
2038 } while (s++ < strend);
2046 RX_MATCH_TAINTED_set(prog, PL_reg_flags & RF_tainted);
2048 if (PL_reg_eval_set) {
2049 /* Preserve the current value of $^R */
2050 if (oreplsv != GvSV(PL_replgv))
2051 sv_setsv(oreplsv, GvSV(PL_replgv));/* So that when GvSV(replgv) is
2052 restored, the value remains
2054 restore_pos(aTHX_ 0);
2057 /* make sure $`, $&, $', and $digit will work later */
2058 if ( !(flags & REXEC_NOT_FIRST) ) {
2059 RX_MATCH_COPY_FREE(prog);
2060 if (flags & REXEC_COPY_STR) {
2061 I32 i = PL_regeol - startpos + (stringarg - strbeg);
2062 #ifdef PERL_OLD_COPY_ON_WRITE
2064 || (SvFLAGS(sv) & CAN_COW_MASK) == CAN_COW_FLAGS)) {
2066 PerlIO_printf(Perl_debug_log,
2067 "Copy on write: regexp capture, type %d\n",
2070 prog->saved_copy = sv_setsv_cow(prog->saved_copy, sv);
2071 prog->subbeg = (char *)SvPVX_const(prog->saved_copy);
2072 assert (SvPOKp(prog->saved_copy));
2076 RX_MATCH_COPIED_on(prog);
2077 s = savepvn(strbeg, i);
2083 prog->subbeg = strbeg;
2084 prog->sublen = PL_regeol - strbeg; /* strend may have been modified */
2091 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch failed%s\n",
2092 PL_colors[4], PL_colors[5]));
2093 if (PL_reg_eval_set)
2094 restore_pos(aTHX_ 0);
2099 - regtry - try match at specific point
2101 STATIC I32 /* 0 failure, 1 success */
2102 S_regtry(pTHX_ regexp *prog, char *startpos)
2108 GET_RE_DEBUG_FLAGS_DECL;
2111 PL_regindent = 0; /* XXXX Not good when matches are reenterable... */
2113 if ((prog->reganch & ROPT_EVAL_SEEN) && !PL_reg_eval_set) {
2116 PL_reg_eval_set = RS_init;
2117 DEBUG_EXECUTE_r(DEBUG_s(
2118 PerlIO_printf(Perl_debug_log, " setting stack tmpbase at %"IVdf"\n",
2119 (IV)(PL_stack_sp - PL_stack_base));
2121 SAVEI32(cxstack[cxstack_ix].blk_oldsp);
2122 cxstack[cxstack_ix].blk_oldsp = PL_stack_sp - PL_stack_base;
2123 /* Otherwise OP_NEXTSTATE will free whatever on stack now. */
2125 /* Apparently this is not needed, judging by wantarray. */
2126 /* SAVEI8(cxstack[cxstack_ix].blk_gimme);
2127 cxstack[cxstack_ix].blk_gimme = G_SCALAR; */
2130 /* Make $_ available to executed code. */
2131 if (PL_reg_sv != DEFSV) {
2136 if (!(SvTYPE(PL_reg_sv) >= SVt_PVMG && SvMAGIC(PL_reg_sv)
2137 && (mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global)))) {
2138 /* prepare for quick setting of pos */
2139 sv_magic(PL_reg_sv, (SV*)0,
2140 PERL_MAGIC_regex_global, Nullch, 0);
2141 mg = mg_find(PL_reg_sv, PERL_MAGIC_regex_global);
2145 PL_reg_oldpos = mg->mg_len;
2146 SAVEDESTRUCTOR_X(restore_pos, 0);
2148 if (!PL_reg_curpm) {
2149 Newxz(PL_reg_curpm, 1, PMOP);
2152 SV* repointer = newSViv(0);
2153 /* so we know which PL_regex_padav element is PL_reg_curpm */
2154 SvFLAGS(repointer) |= SVf_BREAK;
2155 av_push(PL_regex_padav,repointer);
2156 PL_reg_curpm->op_pmoffset = av_len(PL_regex_padav);
2157 PL_regex_pad = AvARRAY(PL_regex_padav);
2161 PM_SETRE(PL_reg_curpm, prog);
2162 PL_reg_oldcurpm = PL_curpm;
2163 PL_curpm = PL_reg_curpm;
2164 if (RX_MATCH_COPIED(prog)) {
2165 /* Here is a serious problem: we cannot rewrite subbeg,
2166 since it may be needed if this match fails. Thus
2167 $` inside (?{}) could fail... */
2168 PL_reg_oldsaved = prog->subbeg;
2169 PL_reg_oldsavedlen = prog->sublen;
2170 #ifdef PERL_OLD_COPY_ON_WRITE
2171 PL_nrs = prog->saved_copy;
2173 RX_MATCH_COPIED_off(prog);
2176 PL_reg_oldsaved = Nullch;
2177 prog->subbeg = PL_bostr;
2178 prog->sublen = PL_regeol - PL_bostr; /* strend may have been modified */
2180 prog->startp[0] = startpos - PL_bostr;
2181 PL_reginput = startpos;
2182 PL_regstartp = prog->startp;
2183 PL_regendp = prog->endp;
2184 PL_reglastparen = &prog->lastparen;
2185 PL_reglastcloseparen = &prog->lastcloseparen;
2186 prog->lastparen = 0;
2187 prog->lastcloseparen = 0;
2189 DEBUG_EXECUTE_r(PL_reg_starttry = startpos);
2190 if (PL_reg_start_tmpl <= prog->nparens) {
2191 PL_reg_start_tmpl = prog->nparens*3/2 + 3;
2192 if(PL_reg_start_tmp)
2193 Renew(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2195 Newx(PL_reg_start_tmp, PL_reg_start_tmpl, char*);
2198 /* XXXX What this code is doing here?!!! There should be no need
2199 to do this again and again, PL_reglastparen should take care of
2202 /* Tests pat.t#187 and split.t#{13,14} seem to depend on this code.
2203 * Actually, the code in regcppop() (which Ilya may be meaning by
2204 * PL_reglastparen), is not needed at all by the test suite
2205 * (op/regexp, op/pat, op/split), but that code is needed, oddly
2206 * enough, for building DynaLoader, or otherwise this
2207 * "Error: '*' not in typemap in DynaLoader.xs, line 164"
2208 * will happen. Meanwhile, this code *is* needed for the
2209 * above-mentioned test suite tests to succeed. The common theme
2210 * on those tests seems to be returning null fields from matches.
2215 if (prog->nparens) {
2216 for (i = prog->nparens; i > (I32)*PL_reglastparen; i--) {
2223 if (regmatch(prog->program + 1)) {
2224 prog->endp[0] = PL_reginput - PL_bostr;
2227 REGCP_UNWIND(lastcp);
2231 #define RE_UNWIND_BRANCH 1
2232 #define RE_UNWIND_BRANCHJ 2
2236 typedef struct { /* XX: makes sense to enlarge it... */
2240 } re_unwind_generic_t;
2253 } re_unwind_branch_t;
2255 typedef union re_unwind_t {
2257 re_unwind_generic_t generic;
2258 re_unwind_branch_t branch;
2261 #define sayYES goto yes
2262 #define sayNO goto no
2263 #define sayNO_ANYOF goto no_anyof
2264 #define sayYES_FINAL goto yes_final
2265 #define sayYES_LOUD goto yes_loud
2266 #define sayNO_FINAL goto no_final
2267 #define sayNO_SILENT goto do_no
2268 #define saySAME(x) if (x) goto yes; else goto no
2270 #define POSCACHE_SUCCESS 0 /* caching success rather than failure */
2271 #define POSCACHE_SEEN 1 /* we know what we're caching */
2272 #define POSCACHE_START 2 /* the real cache: this bit maps to pos 0 */
2273 #define CACHEsayYES STMT_START { \
2274 if (cache_offset | cache_bit) { \
2275 if (!(PL_reg_poscache[0] & (1<<POSCACHE_SEEN))) \
2276 PL_reg_poscache[0] |= (1<<POSCACHE_SUCCESS) || (1<<POSCACHE_SEEN); \
2277 else if (!(PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))) { \
2278 /* cache records failure, but this is success */ \
2280 PerlIO_printf(Perl_debug_log, \
2281 "%*s (remove success from failure cache)\n", \
2282 REPORT_CODE_OFF+PL_regindent*2, "") \
2284 PL_reg_poscache[cache_offset] &= ~(1<<cache_bit); \
2289 #define CACHEsayNO STMT_START { \
2290 if (cache_offset | cache_bit) { \
2291 if (!(PL_reg_poscache[0] & (1<<POSCACHE_SEEN))) \
2292 PL_reg_poscache[0] |= (1<<POSCACHE_SEEN); \
2293 else if ((PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))) { \
2294 /* cache records success, but this is failure */ \
2296 PerlIO_printf(Perl_debug_log, \
2297 "%*s (remove failure from success cache)\n", \
2298 REPORT_CODE_OFF+PL_regindent*2, "") \
2300 PL_reg_poscache[cache_offset] &= ~(1<<cache_bit); \
2306 /* this is used to determine how far from the left messages like
2307 'failed...' are printed. Currently 29 makes these messages line
2308 up with the opcode they refer to. Earlier perls used 25 which
2309 left these messages outdented making reviewing a debug output
2312 #define REPORT_CODE_OFF 29
2315 /* Make sure there is a test for this +1 options in re_tests */
2316 #define TRIE_INITAL_ACCEPT_BUFFLEN 4;
2318 #define TRIE_CHECK_STATE_IS_ACCEPTING STMT_START { \
2319 if ( trie->states[ state ].wordnum ) { \
2320 if ( !accepted ) { \
2323 bufflen = TRIE_INITAL_ACCEPT_BUFFLEN ; \
2324 sv_accept_buff=NEWSV( 1234, \
2325 bufflen * sizeof(reg_trie_accepted) - 1 ); \
2326 SvCUR_set( sv_accept_buff, sizeof(reg_trie_accepted) ); \
2327 SvPOK_on( sv_accept_buff ); \
2328 sv_2mortal( sv_accept_buff ); \
2329 accept_buff = (reg_trie_accepted*)SvPV_nolen( sv_accept_buff );\
2331 if ( accepted >= bufflen ) { \
2333 accept_buff =(reg_trie_accepted*)SvGROW( sv_accept_buff, \
2334 bufflen * sizeof(reg_trie_accepted) ); \
2336 SvCUR_set( sv_accept_buff,SvCUR( sv_accept_buff ) \
2337 + sizeof( reg_trie_accepted ) ); \
2339 accept_buff[ accepted ].wordnum = trie->states[ state ].wordnum; \
2340 accept_buff[ accepted ].endpos = uc; \
2344 #define TRIE_HANDLE_CHAR STMT_START { \
2345 if ( uvc < 256 ) { \
2346 charid = trie->charmap[ uvc ]; \
2349 if( trie->widecharmap ) { \
2350 SV** svpp = (SV**)NULL; \
2351 svpp = hv_fetch( trie->widecharmap, (char*)&uvc, \
2352 sizeof( UV ), 0 ); \
2354 charid = (U16)SvIV( *svpp ); \
2359 ( base + charid > trie->uniquecharcount ) && \
2360 ( base + charid - 1 - trie->uniquecharcount < trie->lasttrans) && \
2361 trie->trans[ base + charid - 1 - trie->uniquecharcount ].check == state ) \
2363 state = trie->trans[ base + charid - 1 - trie->uniquecharcount ].next; \
2371 - regmatch - main matching routine
2373 * Conceptually the strategy is simple: check to see whether the current
2374 * node matches, call self recursively to see whether the rest matches,
2375 * and then act accordingly. In practice we make some effort to avoid
2376 * recursion, in particular by going through "ordinary" nodes (that don't
2377 * need to know whether the rest of the match failed) by a loop instead of
2380 /* [lwall] I've hoisted the register declarations to the outer block in order to
2381 * maybe save a little bit of pushing and popping on the stack. It also takes
2382 * advantage of machines that use a register save mask on subroutine entry.
2384 STATIC I32 /* 0 failure, 1 success */
2385 S_regmatch(pTHX_ regnode *prog)
2388 register regnode *scan; /* Current node. */
2389 regnode *next; /* Next node. */
2390 regnode *inner; /* Next node in internal branch. */
2391 register I32 nextchr; /* renamed nextchr - nextchar colides with
2392 function of same name */
2393 register I32 n; /* no or next */
2394 register I32 ln = 0; /* len or last */
2395 register char *s = Nullch; /* operand or save */
2396 register char *locinput = PL_reginput;
2397 register I32 c1 = 0, c2 = 0, paren; /* case fold search, parenth */
2398 int minmod = 0, sw = 0, logical = 0;
2401 /* used by the trie code */
2402 SV *sv_accept_buff = 0; /* accepting states we have traversed */
2403 reg_trie_accepted *accept_buff = 0; /* "" */
2404 reg_trie_data *trie; /* what trie are we using right now */
2405 U32 accepted = 0; /* how many accepting states we have seen*/
2408 I32 firstcp = PL_savestack_ix;
2410 register const bool do_utf8 = PL_reg_match_utf8;
2412 SV *dsv0 = PERL_DEBUG_PAD_ZERO(0);
2413 SV *dsv1 = PERL_DEBUG_PAD_ZERO(1);
2414 SV *dsv2 = PERL_DEBUG_PAD_ZERO(2);
2416 SV *re_debug_flags = NULL;
2418 U32 uniflags = ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY;
2427 /* Note that nextchr is a byte even in UTF */
2428 nextchr = UCHARAT(locinput);
2430 while (scan != NULL) {
2433 SV *prop = sv_newmortal();
2434 const int docolor = *PL_colors[0];
2435 const int taill = (docolor ? 10 : 7); /* 3 chars for "> <" */
2436 int l = (PL_regeol - locinput) > taill ? taill : (PL_regeol - locinput);
2437 /* The part of the string before starttry has one color
2438 (pref0_len chars), between starttry and current
2439 position another one (pref_len - pref0_len chars),
2440 after the current position the third one.
2441 We assume that pref0_len <= pref_len, otherwise we
2442 decrease pref0_len. */
2443 int pref_len = (locinput - PL_bostr) > (5 + taill) - l
2444 ? (5 + taill) - l : locinput - PL_bostr;
2447 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput - pref_len)))
2449 pref0_len = pref_len - (locinput - PL_reg_starttry);
2450 if (l + pref_len < (5 + taill) && l < PL_regeol - locinput)
2451 l = ( PL_regeol - locinput > (5 + taill) - pref_len
2452 ? (5 + taill) - pref_len : PL_regeol - locinput);
2453 while (do_utf8 && UTF8_IS_CONTINUATION(*(U8*)(locinput + l)))
2457 if (pref0_len > pref_len)
2458 pref0_len = pref_len;
2459 regprop(prop, scan);
2461 const char * const s0 =
2462 do_utf8 && OP(scan) != CANY ?
2463 pv_uni_display(dsv0, (U8*)(locinput - pref_len),
2464 pref0_len, 60, UNI_DISPLAY_REGEX) :
2465 locinput - pref_len;
2466 const int len0 = do_utf8 ? strlen(s0) : pref0_len;
2467 const char * const s1 = do_utf8 && OP(scan) != CANY ?
2468 pv_uni_display(dsv1, (U8*)(locinput - pref_len + pref0_len),
2469 pref_len - pref0_len, 60, UNI_DISPLAY_REGEX) :
2470 locinput - pref_len + pref0_len;
2471 const int len1 = do_utf8 ? strlen(s1) : pref_len - pref0_len;
2472 const char * const s2 = do_utf8 && OP(scan) != CANY ?
2473 pv_uni_display(dsv2, (U8*)locinput,
2474 PL_regeol - locinput, 60, UNI_DISPLAY_REGEX) :
2476 const int len2 = do_utf8 ? strlen(s2) : l;
2477 PerlIO_printf(Perl_debug_log,
2478 "%4"IVdf" <%s%.*s%s%s%.*s%s%s%s%.*s%s>%*s|%3"IVdf":%*s%s\n",
2479 (IV)(locinput - PL_bostr),
2486 (docolor ? "" : "> <"),
2490 15 - l - pref_len + 1,
2492 (IV)(scan - PL_regprogram), PL_regindent*2, "",
2497 next = scan + NEXT_OFF(scan);
2503 if (locinput == PL_bostr)
2505 /* regtill = regbol; */
2510 if (locinput == PL_bostr ||
2511 ((nextchr || locinput < PL_regeol) && locinput[-1] == '\n'))
2517 if (locinput == PL_bostr)
2521 if (locinput == PL_reg_ganch)
2527 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2532 if ((nextchr || locinput < PL_regeol) && nextchr != '\n')
2534 if (PL_regeol - locinput > 1)
2538 if (PL_regeol != locinput)
2542 if (!nextchr && locinput >= PL_regeol)
2545 locinput += PL_utf8skip[nextchr];
2546 if (locinput > PL_regeol)
2548 nextchr = UCHARAT(locinput);
2551 nextchr = UCHARAT(++locinput);
2554 if (!nextchr && locinput >= PL_regeol)
2556 nextchr = UCHARAT(++locinput);
2559 if ((!nextchr && locinput >= PL_regeol) || nextchr == '\n')
2562 locinput += PL_utf8skip[nextchr];
2563 if (locinput > PL_regeol)
2565 nextchr = UCHARAT(locinput);
2568 nextchr = UCHARAT(++locinput);
2574 traverse the TRIE keeping track of all accepting states
2575 we transition through until we get to a failing node.
2577 we use two slightly different pieces of code to handle
2578 the traversal depending on whether its case sensitive or
2579 not. we reuse the accept code however. (this should probably
2580 be turned into a macro.)
2586 U8 *uc = ( U8* )locinput;
2593 U8 *uscan = (U8*)NULL;
2597 trie = (reg_trie_data*)PL_regdata->data[ ARG( scan ) ];
2599 while ( state && uc <= (U8*)PL_regeol ) {
2601 TRIE_CHECK_STATE_IS_ACCEPTING;
2603 base = trie->states[ state ].trans.base;
2605 DEBUG_TRIE_EXECUTE_r(
2606 PerlIO_printf( Perl_debug_log,
2607 "%*s %sState: %4"UVxf", Base: %4"UVxf", Accepted: %4"UVxf" ",
2608 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
2609 (UV)state, (UV)base, (UV)accepted );
2616 uvc = utf8n_to_uvuni( uscan, UTF8_MAXLEN, &len, uniflags );
2621 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
2622 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags );
2623 uvc = to_uni_fold( uvc, foldbuf, &foldlen );
2624 foldlen -= UNISKIP( uvc );
2625 uscan = foldbuf + UNISKIP( uvc );
2637 DEBUG_TRIE_EXECUTE_r(
2638 PerlIO_printf( Perl_debug_log,
2639 "Charid:%3x CV:%4"UVxf" After State: %4"UVxf"%s\n",
2640 charid, uvc, (UV)state, PL_colors[5] );
2649 /* unreached codepoint: we jump into the middle of the next case
2650 from previous if blocks */
2653 U8 *uc = (U8*)locinput;
2662 trie = (reg_trie_data*)PL_regdata->data[ ARG( scan ) ];
2664 while ( state && uc <= (U8*)PL_regeol ) {
2666 TRIE_CHECK_STATE_IS_ACCEPTING;
2668 base = trie->states[ state ].trans.base;
2670 DEBUG_TRIE_EXECUTE_r(
2671 PerlIO_printf( Perl_debug_log,
2672 "%*s %sState: %4"UVxf", Base: %4"UVxf", Accepted: %4"UVxf" ",
2673 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
2674 (UV)state, (UV)base, (UV)accepted );
2680 uvc = utf8n_to_uvuni( (U8*)uc, UTF8_MAXLEN, &len, uniflags );
2691 DEBUG_TRIE_EXECUTE_r(
2692 PerlIO_printf( Perl_debug_log,
2693 "Charid:%3x CV:%4"UVxf" After State: %4"UVxf"%s\n",
2694 charid, uvc, (UV)state, PL_colors[5] );
2704 There was at least one accepting state that we
2705 transitioned through. Presumably the number of accepting
2706 states is going to be low, typically one or two. So we
2707 simply scan through to find the one with lowest wordnum.
2708 Once we find it, we swap the last state into its place
2709 and decrement the size. We then try to match the rest of
2710 the pattern at the point where the word ends, if we
2711 succeed then we end the loop, otherwise the loop
2712 eventually terminates once all of the accepting states
2719 if ( accepted == 1 ) {
2721 SV **tmp = av_fetch( trie->words, accept_buff[ 0 ].wordnum-1, 0 );
2722 PerlIO_printf( Perl_debug_log,
2723 "%*s %sonly one match : #%d <%s>%s\n",
2724 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],
2725 accept_buff[ 0 ].wordnum,
2726 tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr",
2729 PL_reginput = (char *)accept_buff[ 0 ].endpos;
2730 /* in this case we free tmps/leave before we call regmatch
2731 as we wont be using accept_buff again. */
2734 gotit = regmatch( scan + NEXT_OFF( scan ) );
2737 PerlIO_printf( Perl_debug_log,"%*s %sgot %"IVdf" possible matches%s\n",
2738 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4], (IV)accepted,
2741 while ( !gotit && accepted-- ) {
2744 for( cur = 1 ; cur <= accepted ; cur++ ) {
2745 DEBUG_TRIE_EXECUTE_r(
2746 PerlIO_printf( Perl_debug_log,
2747 "%*s %sgot %"IVdf" (%d) as best, looking at %"IVdf" (%d)%s\n",
2748 REPORT_CODE_OFF + PL_regindent * 2, "", PL_colors[4],
2749 (IV)best, accept_buff[ best ].wordnum, (IV)cur,
2750 accept_buff[ cur ].wordnum, PL_colors[5] );
2753 if ( accept_buff[ cur ].wordnum < accept_buff[ best ].wordnum )
2757 SV **tmp = av_fetch( trie->words, accept_buff[ best ].wordnum - 1, 0 );
2758 PerlIO_printf( Perl_debug_log, "%*s %strying alternation #%d <%s> at 0x%p%s\n",
2759 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4],
2760 accept_buff[best].wordnum,
2761 tmp ? SvPV_nolen_const( *tmp ) : "not compiled under -Dr",scan,
2764 if ( best<accepted ) {
2765 reg_trie_accepted tmp = accept_buff[ best ];
2766 accept_buff[ best ] = accept_buff[ accepted ];
2767 accept_buff[ accepted ] = tmp;
2770 PL_reginput = (char *)accept_buff[ best ].endpos;
2773 as far as I can tell we only need the SAVETMPS/FREETMPS
2774 for re's with EVAL in them but I'm leaving them in for
2775 all until I can be sure.
2778 gotit = regmatch( scan + NEXT_OFF( scan ) ) ;
2791 /* unreached codepoint */
2795 if (do_utf8 != UTF) {
2796 /* The target and the pattern have differing utf8ness. */
2798 const char *e = s + ln;
2801 /* The target is utf8, the pattern is not utf8. */
2806 if (NATIVE_TO_UNI(*(U8*)s) !=
2807 utf8n_to_uvuni((U8*)l, UTF8_MAXBYTES, &ulen,
2815 /* The target is not utf8, the pattern is utf8. */
2820 if (NATIVE_TO_UNI(*((U8*)l)) !=
2821 utf8n_to_uvuni((U8*)s, UTF8_MAXBYTES, &ulen,
2829 nextchr = UCHARAT(locinput);
2832 /* The target and the pattern have the same utf8ness. */
2833 /* Inline the first character, for speed. */
2834 if (UCHARAT(s) != nextchr)
2836 if (PL_regeol - locinput < ln)
2838 if (ln > 1 && memNE(s, locinput, ln))
2841 nextchr = UCHARAT(locinput);
2844 PL_reg_flags |= RF_tainted;
2850 if (do_utf8 || UTF) {
2851 /* Either target or the pattern are utf8. */
2853 char *e = PL_regeol;
2855 if (ibcmp_utf8(s, 0, ln, (bool)UTF,
2856 l, &e, 0, do_utf8)) {
2857 /* One more case for the sharp s:
2858 * pack("U0U*", 0xDF) =~ /ss/i,
2859 * the 0xC3 0x9F are the UTF-8
2860 * byte sequence for the U+00DF. */
2862 toLOWER(s[0]) == 's' &&
2864 toLOWER(s[1]) == 's' &&
2871 nextchr = UCHARAT(locinput);
2875 /* Neither the target and the pattern are utf8. */
2877 /* Inline the first character, for speed. */
2878 if (UCHARAT(s) != nextchr &&
2879 UCHARAT(s) != ((OP(scan) == EXACTF)
2880 ? PL_fold : PL_fold_locale)[nextchr])
2882 if (PL_regeol - locinput < ln)
2884 if (ln > 1 && (OP(scan) == EXACTF
2885 ? ibcmp(s, locinput, ln)
2886 : ibcmp_locale(s, locinput, ln)))
2889 nextchr = UCHARAT(locinput);
2893 STRLEN inclasslen = PL_regeol - locinput;
2895 if (!reginclass(scan, (U8*)locinput, &inclasslen, do_utf8))
2897 if (locinput >= PL_regeol)
2899 locinput += inclasslen ? inclasslen : UTF8SKIP(locinput);
2900 nextchr = UCHARAT(locinput);
2905 nextchr = UCHARAT(locinput);
2906 if (!REGINCLASS(scan, (U8*)locinput))
2908 if (!nextchr && locinput >= PL_regeol)
2910 nextchr = UCHARAT(++locinput);
2914 /* If we might have the case of the German sharp s
2915 * in a casefolding Unicode character class. */
2917 if (ANYOF_FOLD_SHARP_S(scan, locinput, PL_regeol)) {
2918 locinput += SHARP_S_SKIP;
2919 nextchr = UCHARAT(locinput);
2925 PL_reg_flags |= RF_tainted;
2931 LOAD_UTF8_CHARCLASS_ALNUM();
2932 if (!(OP(scan) == ALNUM
2933 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2934 : isALNUM_LC_utf8((U8*)locinput)))
2938 locinput += PL_utf8skip[nextchr];
2939 nextchr = UCHARAT(locinput);
2942 if (!(OP(scan) == ALNUM
2943 ? isALNUM(nextchr) : isALNUM_LC(nextchr)))
2945 nextchr = UCHARAT(++locinput);
2948 PL_reg_flags |= RF_tainted;
2951 if (!nextchr && locinput >= PL_regeol)
2954 LOAD_UTF8_CHARCLASS_ALNUM();
2955 if (OP(scan) == NALNUM
2956 ? swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8)
2957 : isALNUM_LC_utf8((U8*)locinput))
2961 locinput += PL_utf8skip[nextchr];
2962 nextchr = UCHARAT(locinput);
2965 if (OP(scan) == NALNUM
2966 ? isALNUM(nextchr) : isALNUM_LC(nextchr))
2968 nextchr = UCHARAT(++locinput);
2972 PL_reg_flags |= RF_tainted;
2976 /* was last char in word? */
2978 if (locinput == PL_bostr)
2981 const U8 * const r = reghop3((U8*)locinput, -1, (U8*)PL_bostr);
2983 ln = utf8n_to_uvchr(r, UTF8SKIP(r), 0, 0);
2985 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
2986 ln = isALNUM_uni(ln);
2987 LOAD_UTF8_CHARCLASS_ALNUM();
2988 n = swash_fetch(PL_utf8_alnum, (U8*)locinput, do_utf8);
2991 ln = isALNUM_LC_uvchr(UNI_TO_NATIVE(ln));
2992 n = isALNUM_LC_utf8((U8*)locinput);
2996 ln = (locinput != PL_bostr) ?
2997 UCHARAT(locinput - 1) : '\n';
2998 if (OP(scan) == BOUND || OP(scan) == NBOUND) {
3000 n = isALNUM(nextchr);
3003 ln = isALNUM_LC(ln);
3004 n = isALNUM_LC(nextchr);
3007 if (((!ln) == (!n)) == (OP(scan) == BOUND ||
3008 OP(scan) == BOUNDL))
3012 PL_reg_flags |= RF_tainted;
3018 if (UTF8_IS_CONTINUED(nextchr)) {
3019 LOAD_UTF8_CHARCLASS_SPACE();
3020 if (!(OP(scan) == SPACE
3021 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
3022 : isSPACE_LC_utf8((U8*)locinput)))
3026 locinput += PL_utf8skip[nextchr];
3027 nextchr = UCHARAT(locinput);
3030 if (!(OP(scan) == SPACE
3031 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3033 nextchr = UCHARAT(++locinput);
3036 if (!(OP(scan) == SPACE
3037 ? isSPACE(nextchr) : isSPACE_LC(nextchr)))
3039 nextchr = UCHARAT(++locinput);
3043 PL_reg_flags |= RF_tainted;
3046 if (!nextchr && locinput >= PL_regeol)
3049 LOAD_UTF8_CHARCLASS_SPACE();
3050 if (OP(scan) == NSPACE
3051 ? swash_fetch(PL_utf8_space, (U8*)locinput, do_utf8)
3052 : isSPACE_LC_utf8((U8*)locinput))
3056 locinput += PL_utf8skip[nextchr];
3057 nextchr = UCHARAT(locinput);
3060 if (OP(scan) == NSPACE
3061 ? isSPACE(nextchr) : isSPACE_LC(nextchr))
3063 nextchr = UCHARAT(++locinput);
3066 PL_reg_flags |= RF_tainted;
3072 LOAD_UTF8_CHARCLASS_DIGIT();
3073 if (!(OP(scan) == DIGIT
3074 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
3075 : isDIGIT_LC_utf8((U8*)locinput)))
3079 locinput += PL_utf8skip[nextchr];
3080 nextchr = UCHARAT(locinput);
3083 if (!(OP(scan) == DIGIT
3084 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr)))
3086 nextchr = UCHARAT(++locinput);
3089 PL_reg_flags |= RF_tainted;
3092 if (!nextchr && locinput >= PL_regeol)
3095 LOAD_UTF8_CHARCLASS_DIGIT();
3096 if (OP(scan) == NDIGIT
3097 ? swash_fetch(PL_utf8_digit, (U8*)locinput, do_utf8)
3098 : isDIGIT_LC_utf8((U8*)locinput))
3102 locinput += PL_utf8skip[nextchr];
3103 nextchr = UCHARAT(locinput);
3106 if (OP(scan) == NDIGIT
3107 ? isDIGIT(nextchr) : isDIGIT_LC(nextchr))
3109 nextchr = UCHARAT(++locinput);
3112 if (locinput >= PL_regeol)
3115 LOAD_UTF8_CHARCLASS_MARK();
3116 if (swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3118 locinput += PL_utf8skip[nextchr];
3119 while (locinput < PL_regeol &&
3120 swash_fetch(PL_utf8_mark,(U8*)locinput, do_utf8))
3121 locinput += UTF8SKIP(locinput);
3122 if (locinput > PL_regeol)
3127 nextchr = UCHARAT(locinput);
3130 PL_reg_flags |= RF_tainted;
3134 n = ARG(scan); /* which paren pair */
3135 ln = PL_regstartp[n];
3136 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3137 if ((I32)*PL_reglastparen < n || ln == -1)
3138 sayNO; /* Do not match unless seen CLOSEn. */
3139 if (ln == PL_regendp[n])
3143 if (do_utf8 && OP(scan) != REF) { /* REF can do byte comparison */
3145 const char *e = PL_bostr + PL_regendp[n];
3147 * Note that we can't do the "other character" lookup trick as
3148 * in the 8-bit case (no pun intended) because in Unicode we
3149 * have to map both upper and title case to lower case.
3151 if (OP(scan) == REFF) {
3153 STRLEN ulen1, ulen2;
3154 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
3155 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
3159 toLOWER_utf8((U8*)s, tmpbuf1, &ulen1);
3160 toLOWER_utf8((U8*)l, tmpbuf2, &ulen2);
3161 if (ulen1 != ulen2 || memNE((char *)tmpbuf1, (char *)tmpbuf2, ulen1))
3168 nextchr = UCHARAT(locinput);
3172 /* Inline the first character, for speed. */
3173 if (UCHARAT(s) != nextchr &&
3175 (UCHARAT(s) != ((OP(scan) == REFF
3176 ? PL_fold : PL_fold_locale)[nextchr]))))
3178 ln = PL_regendp[n] - ln;
3179 if (locinput + ln > PL_regeol)
3181 if (ln > 1 && (OP(scan) == REF
3182 ? memNE(s, locinput, ln)
3184 ? ibcmp(s, locinput, ln)
3185 : ibcmp_locale(s, locinput, ln))))
3188 nextchr = UCHARAT(locinput);
3199 OP_4tree *oop = PL_op;
3200 COP *ocurcop = PL_curcop;
3203 struct regexp *oreg = PL_reg_re;
3206 PL_op = (OP_4tree*)PL_regdata->data[n];
3207 DEBUG_EXECUTE_r( PerlIO_printf(Perl_debug_log, " re_eval 0x%"UVxf"\n", PTR2UV(PL_op)) );
3208 PAD_SAVE_LOCAL(old_comppad, (PAD*)PL_regdata->data[n + 2]);
3209 PL_regendp[0] = PL_reg_magic->mg_len = locinput - PL_bostr;
3213 CALLRUNOPS(aTHX); /* Scalar context. */
3216 ret = &PL_sv_undef; /* protect against empty (?{}) blocks. */
3224 PAD_RESTORE_LOCAL(old_comppad);
3225 PL_curcop = ocurcop;
3227 if (logical == 2) { /* Postponed subexpression. */
3229 MAGIC *mg = Null(MAGIC*);
3231 CHECKPOINT cp, lastcp;
3235 if(SvROK(ret) && SvSMAGICAL(sv = SvRV(ret)))
3236 mg = mg_find(sv, PERL_MAGIC_qr);
3237 else if (SvSMAGICAL(ret)) {
3238 if (SvGMAGICAL(ret))
3239 sv_unmagic(ret, PERL_MAGIC_qr);
3241 mg = mg_find(ret, PERL_MAGIC_qr);
3245 re = (regexp *)mg->mg_obj;
3246 (void)ReREFCNT_inc(re);
3250 const char *t = SvPV_const(ret, len);
3252 char * const oprecomp = PL_regprecomp;
3253 const I32 osize = PL_regsize;
3254 const I32 onpar = PL_regnpar;
3257 if (DO_UTF8(ret)) pm.op_pmdynflags |= PMdf_DYN_UTF8;
3258 re = CALLREGCOMP(aTHX_ (char*)t, (char*)t + len, &pm);
3260 & (SVs_TEMP | SVs_PADTMP | SVf_READONLY
3262 sv_magic(ret,(SV*)ReREFCNT_inc(re),
3264 PL_regprecomp = oprecomp;
3269 PerlIO_printf(Perl_debug_log,
3270 "Entering embedded \"%s%.60s%s%s\"\n",
3274 (strlen(re->precomp) > 60 ? "..." : ""))
3277 state.prev = PL_reg_call_cc;
3278 state.cc = PL_regcc;
3279 state.re = PL_reg_re;
3283 cp = regcppush(0); /* Save *all* the positions. */
3286 state.ss = PL_savestack_ix;
3287 *PL_reglastparen = 0;
3288 *PL_reglastcloseparen = 0;
3289 PL_reg_call_cc = &state;
3290 PL_reginput = locinput;
3291 toggleutf = ((PL_reg_flags & RF_utf8) != 0) ^
3292 ((re->reganch & ROPT_UTF8) != 0);
3293 if (toggleutf) PL_reg_flags ^= RF_utf8;
3295 /* XXXX This is too dramatic a measure... */
3298 if (regmatch(re->program + 1)) {
3299 /* Even though we succeeded, we need to restore
3300 global variables, since we may be wrapped inside
3301 SUSPEND, thus the match may be not finished yet. */
3303 /* XXXX Do this only if SUSPENDed? */
3304 PL_reg_call_cc = state.prev;
3305 PL_regcc = state.cc;
3306 PL_reg_re = state.re;
3307 cache_re(PL_reg_re);
3308 if (toggleutf) PL_reg_flags ^= RF_utf8;
3310 /* XXXX This is too dramatic a measure... */
3313 /* These are needed even if not SUSPEND. */
3319 REGCP_UNWIND(lastcp);
3321 PL_reg_call_cc = state.prev;
3322 PL_regcc = state.cc;
3323 PL_reg_re = state.re;
3324 cache_re(PL_reg_re);
3325 if (toggleutf) PL_reg_flags ^= RF_utf8;
3327 /* XXXX This is too dramatic a measure... */
3337 sv_setsv(save_scalar(PL_replgv), ret);
3343 n = ARG(scan); /* which paren pair */
3344 PL_reg_start_tmp[n] = locinput;
3349 n = ARG(scan); /* which paren pair */
3350 PL_regstartp[n] = PL_reg_start_tmp[n] - PL_bostr;
3351 PL_regendp[n] = locinput - PL_bostr;
3352 if (n > (I32)*PL_reglastparen)
3353 *PL_reglastparen = n;
3354 *PL_reglastcloseparen = n;
3357 n = ARG(scan); /* which paren pair */
3358 sw = ((I32)*PL_reglastparen >= n && PL_regendp[n] != -1);
3361 PL_reg_leftiter = PL_reg_maxiter; /* Void cache */
3363 next = NEXTOPER(NEXTOPER(scan));
3365 next = scan + ARG(scan);
3366 if (OP(next) == IFTHEN) /* Fake one. */
3367 next = NEXTOPER(NEXTOPER(next));
3371 logical = scan->flags;
3373 /*******************************************************************
3374 PL_regcc contains infoblock about the innermost (...)* loop, and
3375 a pointer to the next outer infoblock.
3377 Here is how Y(A)*Z is processed (if it is compiled into CURLYX/WHILEM):
3379 1) After matching X, regnode for CURLYX is processed;
3381 2) This regnode creates infoblock on the stack, and calls
3382 regmatch() recursively with the starting point at WHILEM node;
3384 3) Each hit of WHILEM node tries to match A and Z (in the order
3385 depending on the current iteration, min/max of {min,max} and
3386 greediness). The information about where are nodes for "A"
3387 and "Z" is read from the infoblock, as is info on how many times "A"
3388 was already matched, and greediness.
3390 4) After A matches, the same WHILEM node is hit again.
3392 5) Each time WHILEM is hit, PL_regcc is the infoblock created by CURLYX
3393 of the same pair. Thus when WHILEM tries to match Z, it temporarily
3394 resets PL_regcc, since this Y(A)*Z can be a part of some other loop:
3395 as in (Y(A)*Z)*. If Z matches, the automaton will hit the WHILEM node
3396 of the external loop.
3398 Currently present infoblocks form a tree with a stem formed by PL_curcc
3399 and whatever it mentions via ->next, and additional attached trees
3400 corresponding to temporarily unset infoblocks as in "5" above.
3402 In the following picture infoblocks for outer loop of
3403 (Y(A)*?Z)*?T are denoted O, for inner I. NULL starting block
3404 is denoted by x. The matched string is YAAZYAZT. Temporarily postponed
3405 infoblocks are drawn below the "reset" infoblock.
3407 In fact in the picture below we do not show failed matches for Z and T
3408 by WHILEM blocks. [We illustrate minimal matches, since for them it is
3409 more obvious *why* one needs to *temporary* unset infoblocks.]
3411 Matched REx position InfoBlocks Comment
3415 Y A)*?Z)*?T x <- O <- I
3416 YA )*?Z)*?T x <- O <- I
3417 YA A)*?Z)*?T x <- O <- I
3418 YAA )*?Z)*?T x <- O <- I
3419 YAA Z)*?T x <- O # Temporary unset I
3422 YAAZ Y(A)*?Z)*?T x <- O
3425 YAAZY (A)*?Z)*?T x <- O
3428 YAAZY A)*?Z)*?T x <- O <- I
3431 YAAZYA )*?Z)*?T x <- O <- I
3434 YAAZYA Z)*?T x <- O # Temporary unset I
3440 YAAZYAZ T x # Temporary unset O
3447 *******************************************************************/
3450 CHECKPOINT cp = PL_savestack_ix;
3451 /* No need to save/restore up to this paren */
3452 I32 parenfloor = scan->flags;
3454 if (OP(PREVOPER(next)) == NOTHING) /* LONGJMP */
3456 cc.oldcc = PL_regcc;
3458 /* XXXX Probably it is better to teach regpush to support
3459 parenfloor > PL_regsize... */
3460 if (parenfloor > (I32)*PL_reglastparen)
3461 parenfloor = *PL_reglastparen; /* Pessimization... */
3462 cc.parenfloor = parenfloor;
3464 cc.min = ARG1(scan);
3465 cc.max = ARG2(scan);
3466 cc.scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3470 PL_reginput = locinput;
3471 n = regmatch(PREVOPER(next)); /* start on the WHILEM */
3473 PL_regcc = cc.oldcc;
3479 * This is really hard to understand, because after we match
3480 * what we're trying to match, we must make sure the rest of
3481 * the REx is going to match for sure, and to do that we have
3482 * to go back UP the parse tree by recursing ever deeper. And
3483 * if it fails, we have to reset our parent's current state
3484 * that we can try again after backing off.
3487 CHECKPOINT cp, lastcp;
3488 CURCUR* cc = PL_regcc;
3489 char *lastloc = cc->lastloc; /* Detection of 0-len. */
3490 I32 cache_offset = 0, cache_bit = 0;
3492 n = cc->cur + 1; /* how many we know we matched */
3493 PL_reginput = locinput;
3496 PerlIO_printf(Perl_debug_log,
3497 "%*s %ld out of %ld..%ld cc=%"UVxf"\n",
3498 REPORT_CODE_OFF+PL_regindent*2, "",
3499 (long)n, (long)cc->min,
3500 (long)cc->max, PTR2UV(cc))
3503 /* If degenerate scan matches "", assume scan done. */
3505 if (locinput == cc->lastloc && n >= cc->min) {
3506 PL_regcc = cc->oldcc;
3510 PerlIO_printf(Perl_debug_log,
3511 "%*s empty match detected, try continuation...\n",
3512 REPORT_CODE_OFF+PL_regindent*2, "")
3514 if (regmatch(cc->next))
3522 /* First just match a string of min scans. */
3526 cc->lastloc = locinput;
3527 if (regmatch(cc->scan))
3530 cc->lastloc = lastloc;
3535 /* Check whether we already were at this position.
3536 Postpone detection until we know the match is not
3537 *that* much linear. */
3538 if (!PL_reg_maxiter) {
3539 PL_reg_maxiter = (PL_regeol - PL_bostr + 1) * (scan->flags>>4);
3540 PL_reg_leftiter = PL_reg_maxiter;
3542 if (PL_reg_leftiter-- == 0) {
3543 const I32 size = (PL_reg_maxiter + 7 + POSCACHE_START)/8;
3544 if (PL_reg_poscache) {
3545 if ((I32)PL_reg_poscache_size < size) {
3546 Renew(PL_reg_poscache, size, char);
3547 PL_reg_poscache_size = size;
3549 Zero(PL_reg_poscache, size, char);
3552 PL_reg_poscache_size = size;
3553 Newxz(PL_reg_poscache, size, char);
3556 PerlIO_printf(Perl_debug_log,
3557 "%sDetected a super-linear match, switching on caching%s...\n",
3558 PL_colors[4], PL_colors[5])
3561 if (PL_reg_leftiter < 0) {
3562 cache_offset = locinput - PL_bostr;
3564 cache_offset = (scan->flags & 0xf) - 1 + POSCACHE_START
3565 + cache_offset * (scan->flags>>4);
3566 cache_bit = cache_offset % 8;
3568 if (PL_reg_poscache[cache_offset] & (1<<cache_bit)) {
3570 PerlIO_printf(Perl_debug_log,
3571 "%*s already tried at this position...\n",
3572 REPORT_CODE_OFF+PL_regindent*2, "")
3574 if (PL_reg_poscache[0] & (1<<POSCACHE_SUCCESS))
3575 /* cache records success */
3578 /* cache records failure */
3581 PL_reg_poscache[cache_offset] |= (1<<cache_bit);
3585 /* Prefer next over scan for minimal matching. */
3588 PL_regcc = cc->oldcc;
3591 cp = regcppush(cc->parenfloor);
3593 if (regmatch(cc->next)) {
3595 CACHEsayYES; /* All done. */
3597 REGCP_UNWIND(lastcp);
3603 if (n >= cc->max) { /* Maximum greed exceeded? */
3604 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3605 && !(PL_reg_flags & RF_warned)) {
3606 PL_reg_flags |= RF_warned;
3607 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3608 "Complex regular subexpression recursion",
3615 PerlIO_printf(Perl_debug_log,
3616 "%*s trying longer...\n",
3617 REPORT_CODE_OFF+PL_regindent*2, "")
3619 /* Try scanning more and see if it helps. */
3620 PL_reginput = locinput;
3622 cc->lastloc = locinput;
3623 cp = regcppush(cc->parenfloor);
3625 if (regmatch(cc->scan)) {
3629 REGCP_UNWIND(lastcp);
3632 cc->lastloc = lastloc;
3636 /* Prefer scan over next for maximal matching. */
3638 if (n < cc->max) { /* More greed allowed? */
3639 cp = regcppush(cc->parenfloor);
3641 cc->lastloc = locinput;
3643 if (regmatch(cc->scan)) {
3647 REGCP_UNWIND(lastcp);
3648 regcppop(); /* Restore some previous $<digit>s? */
3649 PL_reginput = locinput;
3651 PerlIO_printf(Perl_debug_log,
3652 "%*s failed, try continuation...\n",
3653 REPORT_CODE_OFF+PL_regindent*2, "")
3656 if (ckWARN(WARN_REGEXP) && n >= REG_INFTY
3657 && !(PL_reg_flags & RF_warned)) {
3658 PL_reg_flags |= RF_warned;
3659 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s limit (%d) exceeded",
3660 "Complex regular subexpression recursion",
3664 /* Failed deeper matches of scan, so see if this one works. */
3665 PL_regcc = cc->oldcc;
3668 if (regmatch(cc->next))
3674 cc->lastloc = lastloc;
3679 next = scan + ARG(scan);
3682 inner = NEXTOPER(NEXTOPER(scan));
3685 inner = NEXTOPER(scan);
3689 if (OP(next) != c1) /* No choice. */
3690 next = inner; /* Avoid recursion. */
3692 const I32 lastparen = *PL_reglastparen;
3694 re_unwind_branch_t *uw;
3696 /* Put unwinding data on stack */
3697 unwind1 = SSNEWt(1,re_unwind_branch_t);
3698 uw = SSPTRt(unwind1,re_unwind_branch_t);
3701 uw->type = ((c1 == BRANCH)
3703 : RE_UNWIND_BRANCHJ);
3704 uw->lastparen = lastparen;
3706 uw->locinput = locinput;
3707 uw->nextchr = nextchr;
3709 uw->regindent = ++PL_regindent;
3712 REGCP_SET(uw->lastcp);
3714 /* Now go into the first branch */
3727 /* We suppose that the next guy does not need
3728 backtracking: in particular, it is of constant non-zero length,
3729 and has no parenths to influence future backrefs. */
3730 ln = ARG1(scan); /* min to match */
3731 n = ARG2(scan); /* max to match */
3732 paren = scan->flags;
3734 if (paren > PL_regsize)
3736 if (paren > (I32)*PL_reglastparen)
3737 *PL_reglastparen = paren;
3739 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3741 scan += NEXT_OFF(scan); /* Skip former OPEN. */
3742 PL_reginput = locinput;
3745 if (ln && regrepeat_hard(scan, ln, &l) < ln)
3747 locinput = PL_reginput;
3748 if (HAS_TEXT(next) || JUMPABLE(next)) {
3749 regnode *text_node = next;
3751 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3753 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3755 if (PL_regkind[(U8)OP(text_node)] == REF) {
3759 else { c1 = (U8)*STRING(text_node); }
3760 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3762 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3763 c2 = PL_fold_locale[c1];
3772 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
3773 /* If it could work, try it. */
3775 UCHARAT(PL_reginput) == c1 ||
3776 UCHARAT(PL_reginput) == c2)
3780 PL_regstartp[paren] =
3781 HOPc(PL_reginput, -l) - PL_bostr;
3782 PL_regendp[paren] = PL_reginput - PL_bostr;
3785 PL_regendp[paren] = -1;
3789 REGCP_UNWIND(lastcp);
3791 /* Couldn't or didn't -- move forward. */
3792 PL_reginput = locinput;
3793 if (regrepeat_hard(scan, 1, &l)) {
3795 locinput = PL_reginput;
3802 n = regrepeat_hard(scan, n, &l);
3803 locinput = PL_reginput;
3805 PerlIO_printf(Perl_debug_log,
3806 "%*s matched %"IVdf" times, len=%"IVdf"...\n",
3807 (int)(REPORT_CODE_OFF+PL_regindent*2), "",
3811 if (HAS_TEXT(next) || JUMPABLE(next)) {
3812 regnode *text_node = next;
3814 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3816 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3818 if (PL_regkind[(U8)OP(text_node)] == REF) {
3822 else { c1 = (U8)*STRING(text_node); }
3824 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3826 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3827 c2 = PL_fold_locale[c1];
3838 /* If it could work, try it. */
3840 UCHARAT(PL_reginput) == c1 ||
3841 UCHARAT(PL_reginput) == c2)
3844 PerlIO_printf(Perl_debug_log,
3845 "%*s trying tail with n=%"IVdf"...\n",
3846 (int)(REPORT_CODE_OFF+PL_regindent*2), "", (IV)n)
3850 PL_regstartp[paren] = HOPc(PL_reginput, -l) - PL_bostr;
3851 PL_regendp[paren] = PL_reginput - PL_bostr;
3854 PL_regendp[paren] = -1;
3858 REGCP_UNWIND(lastcp);
3860 /* Couldn't or didn't -- back up. */
3862 locinput = HOPc(locinput, -l);
3863 PL_reginput = locinput;
3870 paren = scan->flags; /* Which paren to set */
3871 if (paren > PL_regsize)
3873 if (paren > (I32)*PL_reglastparen)
3874 *PL_reglastparen = paren;
3875 ln = ARG1(scan); /* min to match */
3876 n = ARG2(scan); /* max to match */
3877 scan = regnext(NEXTOPER(scan) + NODE_STEP_REGNODE);
3881 ln = ARG1(scan); /* min to match */
3882 n = ARG2(scan); /* max to match */
3883 scan = NEXTOPER(scan) + NODE_STEP_REGNODE;
3888 scan = NEXTOPER(scan);
3894 scan = NEXTOPER(scan);
3898 * Lookahead to avoid useless match attempts
3899 * when we know what character comes next.
3903 * Used to only do .*x and .*?x, but now it allows
3904 * for )'s, ('s and (?{ ... })'s to be in the way
3905 * of the quantifier and the EXACT-like node. -- japhy
3908 if (HAS_TEXT(next) || JUMPABLE(next)) {
3910 regnode *text_node = next;
3912 if (! HAS_TEXT(text_node)) FIND_NEXT_IMPT(text_node);
3914 if (! HAS_TEXT(text_node)) c1 = c2 = -1000;
3916 if (PL_regkind[(U8)OP(text_node)] == REF) {
3918 goto assume_ok_easy;
3920 else { s = (U8*)STRING(text_node); }
3924 if (OP(text_node) == EXACTF || OP(text_node) == REFF)
3926 else if (OP(text_node) == EXACTFL || OP(text_node) == REFFL)
3927 c2 = PL_fold_locale[c1];
3930 if (OP(text_node) == EXACTF || OP(text_node) == REFF) {
3931 STRLEN ulen1, ulen2;
3932 U8 tmpbuf1[UTF8_MAXBYTES_CASE+1];
3933 U8 tmpbuf2[UTF8_MAXBYTES_CASE+1];
3935 to_utf8_lower((U8*)s, tmpbuf1, &ulen1);
3936 to_utf8_upper((U8*)s, tmpbuf2, &ulen2);
3938 c1 = utf8n_to_uvuni(tmpbuf1, UTF8_MAXBYTES, 0,
3940 c2 = utf8n_to_uvuni(tmpbuf2, UTF8_MAXBYTES, 0,
3944 c2 = c1 = utf8n_to_uvchr(s, UTF8_MAXBYTES, 0,
3953 PL_reginput = locinput;
3957 if (ln && regrepeat(scan, ln) < ln)
3959 locinput = PL_reginput;
3962 char *e; /* Should not check after this */
3963 char *old = locinput;
3966 if (n == REG_INFTY) {
3969 while (UTF8_IS_CONTINUATION(*(U8*)e))
3975 m >0 && e + UTF8SKIP(e) <= PL_regeol; m--)
3979 e = locinput + n - ln;
3984 /* Find place 'next' could work */
3987 while (locinput <= e &&
3988 UCHARAT(locinput) != c1)
3991 while (locinput <= e
3992 && UCHARAT(locinput) != c1
3993 && UCHARAT(locinput) != c2)
3996 count = locinput - old;
4001 /* count initialised to
4002 * utf8_distance(old, locinput) */
4003 while (locinput <= e &&
4004 utf8n_to_uvchr((U8*)locinput,
4005 UTF8_MAXBYTES, &len,
4006 uniflags) != (UV)c1) {
4012 /* count initialised to
4013 * utf8_distance(old, locinput) */
4014 while (locinput <= e) {
4015 UV c = utf8n_to_uvchr((U8*)locinput,
4016 UTF8_MAXBYTES, &len,
4018 if (c == (UV)c1 || c == (UV)c2)
4027 /* PL_reginput == old now */
4028 if (locinput != old) {
4029 ln = 1; /* Did some */
4030 if (regrepeat(scan, count) < count)
4033 /* PL_reginput == locinput now */
4034 TRYPAREN(paren, ln, locinput);
4035 PL_reginput = locinput; /* Could be reset... */
4036 REGCP_UNWIND(lastcp);
4037 /* Couldn't or didn't -- move forward. */
4040 locinput += UTF8SKIP(locinput);
4047 while (n >= ln || (n == REG_INFTY && ln > 0)) { /* ln overflow ? */
4051 c = utf8n_to_uvchr((U8*)PL_reginput,
4055 c = UCHARAT(PL_reginput);
4056 /* If it could work, try it. */
4057 if (c == (UV)c1 || c == (UV)c2)
4059 TRYPAREN(paren, ln, PL_reginput);
4060 REGCP_UNWIND(lastcp);
4063 /* If it could work, try it. */
4064 else if (c1 == -1000)
4066 TRYPAREN(paren, ln, PL_reginput);
4067 REGCP_UNWIND(lastcp);
4069 /* Couldn't or didn't -- move forward. */
4070 PL_reginput = locinput;
4071 if (regrepeat(scan, 1)) {
4073 locinput = PL_reginput;
4081 n = regrepeat(scan, n);
4082 locinput = PL_reginput;
4083 if (ln < n && PL_regkind[(U8)OP(next)] == EOL &&
4084 (OP(next) != MEOL ||
4085 OP(next) == SEOL || OP(next) == EOS))
4087 ln = n; /* why back off? */
4088 /* ...because $ and \Z can match before *and* after
4089 newline at the end. Consider "\n\n" =~ /\n+\Z\n/.
4090 We should back off by one in this case. */
4091 if (UCHARAT(PL_reginput - 1) == '\n' && OP(next) != EOS)
4100 c = utf8n_to_uvchr((U8*)PL_reginput,
4104 c = UCHARAT(PL_reginput);
4106 /* If it could work, try it. */
4107 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
4109 TRYPAREN(paren, n, PL_reginput);
4110 REGCP_UNWIND(lastcp);
4112 /* Couldn't or didn't -- back up. */
4114 PL_reginput = locinput = HOPc(locinput, -1);
4122 c = utf8n_to_uvchr((U8*)PL_reginput,
4126 c = UCHARAT(PL_reginput);
4128 /* If it could work, try it. */
4129 if (c1 == -1000 || c == (UV)c1 || c == (UV)c2)
4131 TRYPAREN(paren, n, PL_reginput);
4132 REGCP_UNWIND(lastcp);
4134 /* Couldn't or didn't -- back up. */
4136 PL_reginput = locinput = HOPc(locinput, -1);
4143 if (PL_reg_call_cc) {
4144 re_cc_state *cur_call_cc = PL_reg_call_cc;
4145 CURCUR *cctmp = PL_regcc;
4146 regexp *re = PL_reg_re;
4147 CHECKPOINT cp, lastcp;
4149 cp = regcppush(0); /* Save *all* the positions. */
4151 regcp_set_to(PL_reg_call_cc->ss); /* Restore parens of
4153 PL_reginput = locinput; /* Make position available to
4155 cache_re(PL_reg_call_cc->re);
4156 PL_regcc = PL_reg_call_cc->cc;
4157 PL_reg_call_cc = PL_reg_call_cc->prev;
4158 if (regmatch(cur_call_cc->node)) {
4159 PL_reg_call_cc = cur_call_cc;
4163 REGCP_UNWIND(lastcp);
4165 PL_reg_call_cc = cur_call_cc;
4171 PerlIO_printf(Perl_debug_log,
4172 "%*s continuation failed...\n",
4173 REPORT_CODE_OFF+PL_regindent*2, "")
4177 if (locinput < PL_regtill) {
4178 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log,
4179 "%sMatch possible, but length=%ld is smaller than requested=%ld, failing!%s\n",
4181 (long)(locinput - PL_reg_starttry),
4182 (long)(PL_regtill - PL_reg_starttry),
4184 sayNO_FINAL; /* Cannot match: too short. */
4186 PL_reginput = locinput; /* put where regtry can find it */
4187 sayYES_FINAL; /* Success! */
4189 PL_reginput = locinput; /* put where regtry can find it */
4190 sayYES_LOUD; /* Success! */
4193 PL_reginput = locinput;
4198 s = HOPBACKc(locinput, scan->flags);
4204 PL_reginput = locinput;
4209 s = HOPBACKc(locinput, scan->flags);
4215 PL_reginput = locinput;
4218 inner = NEXTOPER(NEXTOPER(scan));
4219 if (regmatch(inner) != n) {
4234 if (OP(scan) == SUSPEND) {
4235 locinput = PL_reginput;
4236 nextchr = UCHARAT(locinput);
4241 next = scan + ARG(scan);
4246 PerlIO_printf(Perl_error_log, "%"UVxf" %d\n",
4247 PTR2UV(scan), OP(scan));
4248 Perl_croak(aTHX_ "regexp memory corruption");
4255 * We get here only if there's trouble -- normally "case END" is
4256 * the terminating point.
4258 Perl_croak(aTHX_ "corrupted regexp pointers");
4264 PerlIO_printf(Perl_debug_log,
4265 "%*s %scould match...%s\n",
4266 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5])
4270 DEBUG_EXECUTE_r(PerlIO_printf(Perl_debug_log, "%sMatch successful!%s\n",
4271 PL_colors[4], PL_colors[5]));
4277 #if 0 /* Breaks $^R */
4285 PerlIO_printf(Perl_debug_log,
4286 "%*s %sfailed...%s\n",
4287 REPORT_CODE_OFF+PL_regindent*2, "", PL_colors[4], PL_colors[5])
4293 re_unwind_t *uw = SSPTRt(unwind,re_unwind_t);
4296 case RE_UNWIND_BRANCH:
4297 case RE_UNWIND_BRANCHJ:
4299 re_unwind_branch_t *uwb = &(uw->branch);
4300 const I32 lastparen = uwb->lastparen;
4302 REGCP_UNWIND(uwb->lastcp);
4303 for (n = *PL_reglastparen; n > lastparen; n--)
4305 *PL_reglastparen = n;
4306 scan = next = uwb->next;
4308 OP(scan) != (uwb->type == RE_UNWIND_BRANCH
4309 ? BRANCH : BRANCHJ) ) { /* Failure */
4316 /* Have more choice yet. Reuse the same uwb. */
4317 if ((n = (uwb->type == RE_UNWIND_BRANCH
4318 ? NEXT_OFF(next) : ARG(next))))
4321 next = NULL; /* XXXX Needn't unwinding in this case... */
4323 next = NEXTOPER(scan);
4324 if (uwb->type == RE_UNWIND_BRANCHJ)
4325 next = NEXTOPER(next);
4326 locinput = uwb->locinput;
4327 nextchr = uwb->nextchr;
4329 PL_regindent = uwb->regindent;
4336 Perl_croak(aTHX_ "regexp unwind memory corruption");
4347 - regrepeat - repeatedly match something simple, report how many
4350 * [This routine now assumes that it will only match on things of length 1.
4351 * That was true before, but now we assume scan - reginput is the count,
4352 * rather than incrementing count on every character. [Er, except utf8.]]
4355 S_regrepeat(pTHX_ const regnode *p, I32 max)
4358 register char *scan;
4360 register char *loceol = PL_regeol;
4361 register I32 hardcount = 0;
4362 register bool do_utf8 = PL_reg_match_utf8;
4365 if (max == REG_INFTY)
4367 else if (max < loceol - scan)
4368 loceol = scan + max;
4373 while (scan < loceol && hardcount < max && *scan != '\n') {
4374 scan += UTF8SKIP(scan);
4378 while (scan < loceol && *scan != '\n')
4385 while (scan < loceol && hardcount < max) {
4386 scan += UTF8SKIP(scan);
4396 case EXACT: /* length of string is 1 */
4398 while (scan < loceol && UCHARAT(scan) == c)
4401 case EXACTF: /* length of string is 1 */
4403 while (scan < loceol &&
4404 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold[c]))
4407 case EXACTFL: /* length of string is 1 */
4408 PL_reg_flags |= RF_tainted;
4410 while (scan < loceol &&
4411 (UCHARAT(scan) == c || UCHARAT(scan) == PL_fold_locale[c]))
4417 while (hardcount < max && scan < loceol &&
4418 reginclass(p, (U8*)scan, 0, do_utf8)) {
4419 scan += UTF8SKIP(scan);
4423 while (scan < loceol && REGINCLASS(p, (U8*)scan))
4430 LOAD_UTF8_CHARCLASS_ALNUM();
4431 while (hardcount < max && scan < loceol &&
4432 swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4433 scan += UTF8SKIP(scan);
4437 while (scan < loceol && isALNUM(*scan))
4442 PL_reg_flags |= RF_tainted;
4445 while (hardcount < max && scan < loceol &&
4446 isALNUM_LC_utf8((U8*)scan)) {
4447 scan += UTF8SKIP(scan);
4451 while (scan < loceol && isALNUM_LC(*scan))
4458 LOAD_UTF8_CHARCLASS_ALNUM();
4459 while (hardcount < max && scan < loceol &&
4460 !swash_fetch(PL_utf8_alnum, (U8*)scan, do_utf8)) {
4461 scan += UTF8SKIP(scan);
4465 while (scan < loceol && !isALNUM(*scan))
4470 PL_reg_flags |= RF_tainted;
4473 while (hardcount < max && scan < loceol &&
4474 !isALNUM_LC_utf8((U8*)scan)) {
4475 scan += UTF8SKIP(scan);
4479 while (scan < loceol && !isALNUM_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_SPACE();
4516 while (hardcount < max && scan < loceol &&
4518 swash_fetch(PL_utf8_space,(U8*)scan, do_utf8))) {
4519 scan += UTF8SKIP(scan);
4523 while (scan < loceol && !isSPACE(*scan))
4528 PL_reg_flags |= RF_tainted;
4531 while (hardcount < max && scan < loceol &&
4532 !(*scan == ' ' || isSPACE_LC_utf8((U8*)scan))) {
4533 scan += UTF8SKIP(scan);
4537 while (scan < loceol && !isSPACE_LC(*scan))
4544 LOAD_UTF8_CHARCLASS_DIGIT();
4545 while (hardcount < max && scan < loceol &&
4546 swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4547 scan += UTF8SKIP(scan);
4551 while (scan < loceol && isDIGIT(*scan))
4558 LOAD_UTF8_CHARCLASS_DIGIT();
4559 while (hardcount < max && scan < loceol &&
4560 !swash_fetch(PL_utf8_digit, (U8*)scan, do_utf8)) {
4561 scan += UTF8SKIP(scan);
4565 while (scan < loceol && !isDIGIT(*scan))
4569 default: /* Called on something of 0 width. */
4570 break; /* So match right here or not at all. */
4576 c = scan - PL_reginput;
4580 SV *re_debug_flags = NULL;
4581 SV *prop = sv_newmortal();
4585 PerlIO_printf(Perl_debug_log,
4586 "%*s %s can match %"IVdf" times out of %"IVdf"...\n",
4587 REPORT_CODE_OFF+1, "", SvPVX_const(prop),(IV)c,(IV)max);
4595 - regrepeat_hard - repeatedly match something, report total lenth and length
4597 * The repeater is supposed to have constant non-zero length.
4601 S_regrepeat_hard(pTHX_ regnode *p, I32 max, I32 *lp)
4603 register char *scan = Nullch;
4604 register char *start;
4605 register char *loceol = PL_regeol;
4607 I32 count = 0, res = 1;
4612 start = PL_reginput;
4613 if (PL_reg_match_utf8) {
4614 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4617 while (start < PL_reginput) {
4619 start += UTF8SKIP(start);
4630 while (PL_reginput < loceol && (scan = PL_reginput, res = regmatch(p))) {
4632 *lp = l = PL_reginput - start;
4633 if (max != REG_INFTY && l*max < loceol - scan)
4634 loceol = scan + l*max;
4647 - regclass_swash - prepare the utf8 swash
4651 Perl_regclass_swash(pTHX_ register const regnode* node, bool doinit, SV** listsvp, SV **altsvp)
4657 if (PL_regdata && PL_regdata->count) {
4658 const U32 n = ARG(node);
4660 if (PL_regdata->what[n] == 's') {
4661 SV * const rv = (SV*)PL_regdata->data[n];
4662 AV * const av = (AV*)SvRV((SV*)rv);
4663 SV **const ary = AvARRAY(av);
4666 /* See the end of regcomp.c:S_reglass() for
4667 * documentation of these array elements. */
4670 a = SvTYPE(ary[1]) == SVt_RV ? &ary[1] : 0;
4671 b = SvTYPE(ary[2]) == SVt_PVAV ? &ary[2] : 0;
4675 else if (si && doinit) {
4676 sw = swash_init("utf8", "", si, 1, 0);
4677 (void)av_store(av, 1, sw);
4693 - reginclass - determine if a character falls into a character class
4695 The n is the ANYOF regnode, the p is the target string, lenp
4696 is pointer to the maximum length of how far to go in the p
4697 (if the lenp is zero, UTF8SKIP(p) is used),
4698 do_utf8 tells whether the target string is in UTF-8.
4703 S_reginclass(pTHX_ register const regnode *n, register const U8* p, STRLEN* lenp, register bool do_utf8)
4706 const char flags = ANYOF_FLAGS(n);
4712 if (do_utf8 && !UTF8_IS_INVARIANT(c))
4713 c = utf8n_to_uvchr(p, UTF8_MAXBYTES, &len,
4714 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
4716 plen = lenp ? *lenp : UNISKIP(NATIVE_TO_UNI(c));
4717 if (do_utf8 || (flags & ANYOF_UNICODE)) {
4720 if (do_utf8 && !ANYOF_RUNTIME(n)) {
4721 if (len != (STRLEN)-1 && c < 256 && ANYOF_BITMAP_TEST(n, c))
4724 if (!match && do_utf8 && (flags & ANYOF_UNICODE_ALL) && c >= 256)
4728 SV * const sw = regclass_swash(n, TRUE, 0, (SV**)&av);
4731 if (swash_fetch(sw, p, do_utf8))
4733 else if (flags & ANYOF_FOLD) {
4734 if (!match && lenp && av) {
4736 for (i = 0; i <= av_len(av); i++) {
4737 SV* const sv = *av_fetch(av, i, FALSE);
4739 const char * const s = SvPV_const(sv, len);
4741 if (len <= plen && memEQ(s, (char*)p, len)) {
4749 U8 tmpbuf[UTF8_MAXBYTES_CASE+1];
4752 to_utf8_fold(p, tmpbuf, &tmplen);
4753 if (swash_fetch(sw, tmpbuf, do_utf8))
4759 if (match && lenp && *lenp == 0)
4760 *lenp = UNISKIP(NATIVE_TO_UNI(c));
4762 if (!match && c < 256) {
4763 if (ANYOF_BITMAP_TEST(n, c))
4765 else if (flags & ANYOF_FOLD) {
4768 if (flags & ANYOF_LOCALE) {
4769 PL_reg_flags |= RF_tainted;
4770 f = PL_fold_locale[c];
4774 if (f != c && ANYOF_BITMAP_TEST(n, f))
4778 if (!match && (flags & ANYOF_CLASS)) {
4779 PL_reg_flags |= RF_tainted;
4781 (ANYOF_CLASS_TEST(n, ANYOF_ALNUM) && isALNUM_LC(c)) ||
4782 (ANYOF_CLASS_TEST(n, ANYOF_NALNUM) && !isALNUM_LC(c)) ||
4783 (ANYOF_CLASS_TEST(n, ANYOF_SPACE) && isSPACE_LC(c)) ||
4784 (ANYOF_CLASS_TEST(n, ANYOF_NSPACE) && !isSPACE_LC(c)) ||
4785 (ANYOF_CLASS_TEST(n, ANYOF_DIGIT) && isDIGIT_LC(c)) ||
4786 (ANYOF_CLASS_TEST(n, ANYOF_NDIGIT) && !isDIGIT_LC(c)) ||
4787 (ANYOF_CLASS_TEST(n, ANYOF_ALNUMC) && isALNUMC_LC(c)) ||
4788 (ANYOF_CLASS_TEST(n, ANYOF_NALNUMC) && !isALNUMC_LC(c)) ||
4789 (ANYOF_CLASS_TEST(n, ANYOF_ALPHA) && isALPHA_LC(c)) ||
4790 (ANYOF_CLASS_TEST(n, ANYOF_NALPHA) && !isALPHA_LC(c)) ||
4791 (ANYOF_CLASS_TEST(n, ANYOF_ASCII) && isASCII(c)) ||
4792 (ANYOF_CLASS_TEST(n, ANYOF_NASCII) && !isASCII(c)) ||
4793 (ANYOF_CLASS_TEST(n, ANYOF_CNTRL) && isCNTRL_LC(c)) ||
4794 (ANYOF_CLASS_TEST(n, ANYOF_NCNTRL) && !isCNTRL_LC(c)) ||
4795 (ANYOF_CLASS_TEST(n, ANYOF_GRAPH) && isGRAPH_LC(c)) ||
4796 (ANYOF_CLASS_TEST(n, ANYOF_NGRAPH) && !isGRAPH_LC(c)) ||
4797 (ANYOF_CLASS_TEST(n, ANYOF_LOWER) && isLOWER_LC(c)) ||
4798 (ANYOF_CLASS_TEST(n, ANYOF_NLOWER) && !isLOWER_LC(c)) ||
4799 (ANYOF_CLASS_TEST(n, ANYOF_PRINT) && isPRINT_LC(c)) ||
4800 (ANYOF_CLASS_TEST(n, ANYOF_NPRINT) && !isPRINT_LC(c)) ||
4801 (ANYOF_CLASS_TEST(n, ANYOF_PUNCT) && isPUNCT_LC(c)) ||
4802 (ANYOF_CLASS_TEST(n, ANYOF_NPUNCT) && !isPUNCT_LC(c)) ||
4803 (ANYOF_CLASS_TEST(n, ANYOF_UPPER) && isUPPER_LC(c)) ||
4804 (ANYOF_CLASS_TEST(n, ANYOF_NUPPER) && !isUPPER_LC(c)) ||
4805 (ANYOF_CLASS_TEST(n, ANYOF_XDIGIT) && isXDIGIT(c)) ||
4806 (ANYOF_CLASS_TEST(n, ANYOF_NXDIGIT) && !isXDIGIT(c)) ||
4807 (ANYOF_CLASS_TEST(n, ANYOF_PSXSPC) && isPSXSPC(c)) ||
4808 (ANYOF_CLASS_TEST(n, ANYOF_NPSXSPC) && !isPSXSPC(c)) ||
4809 (ANYOF_CLASS_TEST(n, ANYOF_BLANK) && isBLANK(c)) ||
4810 (ANYOF_CLASS_TEST(n, ANYOF_NBLANK) && !isBLANK(c))
4811 ) /* How's that for a conditional? */
4818 return (flags & ANYOF_INVERT) ? !match : match;
4822 S_reghop(pTHX_ U8 *s, I32 off)
4824 return S_reghop3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4828 S_reghop3(pTHX_ U8 *s, I32 off, U8* lim)
4831 while (off-- && s < lim) {
4832 /* XXX could check well-formedness here */
4840 if (UTF8_IS_CONTINUED(*s)) {
4841 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4844 /* XXX could check well-formedness here */
4852 S_reghopmaybe(pTHX_ U8 *s, I32 off)
4854 return S_reghopmaybe3(aTHX_ s, off, (U8*)(off >= 0 ? PL_regeol : PL_bostr));
4858 S_reghopmaybe3(pTHX_ U8* s, I32 off, U8* lim)
4861 while (off-- && s < lim) {
4862 /* XXX could check well-formedness here */
4872 if (UTF8_IS_CONTINUED(*s)) {
4873 while (s > (U8*)lim && UTF8_IS_CONTINUATION(*s))
4876 /* XXX could check well-formedness here */
4888 restore_pos(pTHX_ void *arg)
4890 PERL_UNUSED_ARG(arg);
4891 if (PL_reg_eval_set) {
4892 if (PL_reg_oldsaved) {
4893 PL_reg_re->subbeg = PL_reg_oldsaved;
4894 PL_reg_re->sublen = PL_reg_oldsavedlen;
4895 #ifdef PERL_OLD_COPY_ON_WRITE
4896 PL_reg_re->saved_copy = PL_nrs;
4898 RX_MATCH_COPIED_on(PL_reg_re);
4900 PL_reg_magic->mg_len = PL_reg_oldpos;
4901 PL_reg_eval_set = 0;
4902 PL_curpm = PL_reg_oldcurpm;
4907 S_to_utf8_substr(pTHX_ register regexp *prog)
4909 if (prog->float_substr && !prog->float_utf8) {
4911 prog->float_utf8 = sv = newSVsv(prog->float_substr);
4912 sv_utf8_upgrade(sv);
4913 if (SvTAIL(prog->float_substr))
4915 if (prog->float_substr == prog->check_substr)
4916 prog->check_utf8 = sv;
4918 if (prog->anchored_substr && !prog->anchored_utf8) {
4920 prog->anchored_utf8 = sv = newSVsv(prog->anchored_substr);
4921 sv_utf8_upgrade(sv);
4922 if (SvTAIL(prog->anchored_substr))
4924 if (prog->anchored_substr == prog->check_substr)
4925 prog->check_utf8 = sv;
4930 S_to_byte_substr(pTHX_ register regexp *prog)
4932 if (prog->float_utf8 && !prog->float_substr) {
4934 prog->float_substr = sv = newSVsv(prog->float_utf8);
4935 if (sv_utf8_downgrade(sv, TRUE)) {
4936 if (SvTAIL(prog->float_utf8))
4940 prog->float_substr = sv = &PL_sv_undef;
4942 if (prog->float_utf8 == prog->check_utf8)
4943 prog->check_substr = sv;
4945 if (prog->anchored_utf8 && !prog->anchored_substr) {
4947 prog->anchored_substr = sv = newSVsv(prog->anchored_utf8);
4948 if (sv_utf8_downgrade(sv, TRUE)) {
4949 if (SvTAIL(prog->anchored_utf8))
4953 prog->anchored_substr = sv = &PL_sv_undef;
4955 if (prog->anchored_utf8 == prog->check_utf8)
4956 prog->check_substr = sv;
4962 * c-indentation-style: bsd
4964 * indent-tabs-mode: t
4967 * ex: set ts=8 sts=4 sw=4 noet: