5 * "A fair jaw-cracker dwarf-language must be." --Samwise Gamgee
8 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
9 * confused with the original package (see point 3 below). Thanks, Henry!
12 /* Additional note: this code is very heavily munged from Henry's version
13 * in places. In some spots I've traded clarity for efficiency, so don't
14 * blame Henry for some of the lack of readability.
17 /* The names of the functions have been changed from regcomp and
18 * regexec to pregcomp and pregexec in order to avoid conflicts
19 * with the POSIX routines of the same names.
22 #ifdef PERL_EXT_RE_BUILD
23 /* need to replace pregcomp et al, so enable that */
24 # ifndef PERL_IN_XSUB_RE
25 # define PERL_IN_XSUB_RE
27 /* need access to debugger hooks */
28 # if defined(PERL_EXT_RE_DEBUG) && !defined(DEBUGGING)
33 #ifdef PERL_IN_XSUB_RE
34 /* We *really* need to overwrite these symbols: */
35 # define Perl_pregcomp my_regcomp
36 # define Perl_regdump my_regdump
37 # define Perl_regprop my_regprop
38 # define Perl_pregfree my_regfree
39 # define Perl_re_intuit_string my_re_intuit_string
40 /* *These* symbols are masked to allow static link. */
41 # define Perl_regnext my_regnext
42 # define Perl_save_re_context my_save_re_context
43 # define Perl_reginitcolors my_reginitcolors
45 # define PERL_NO_GET_CONTEXT
50 * pregcomp and pregexec -- regsub and regerror are not used in perl
52 * Copyright (c) 1986 by University of Toronto.
53 * Written by Henry Spencer. Not derived from licensed software.
55 * Permission is granted to anyone to use this software for any
56 * purpose on any computer system, and to redistribute it freely,
57 * subject to the following restrictions:
59 * 1. The author is not responsible for the consequences of use of
60 * this software, no matter how awful, even if they arise
63 * 2. The origin of this software must not be misrepresented, either
64 * by explicit claim or by omission.
66 * 3. Altered versions must be plainly marked as such, and must not
67 * be misrepresented as being the original software.
70 **** Alterations to Henry's code are...
72 **** Copyright (c) 1991-2000, Larry Wall
74 **** You may distribute under the terms of either the GNU General Public
75 **** License or the Artistic License, as specified in the README file.
78 * Beware that some of this code is subtly aware of the way operator
79 * precedence is structured in regular expressions. Serious changes in
80 * regular-expression syntax might require a total rethink.
83 #define PERL_IN_REGCOMP_C
86 #ifdef PERL_IN_XSUB_RE
87 # if defined(PERL_CAPI) || defined(PERL_OBJECT)
102 # if defined(BUGGY_MSC6)
103 /* MSC 6.00A breaks on op/regexp.t test 85 unless we turn this off */
104 # pragma optimize("a",off)
105 /* But MSC 6.00A is happy with 'w', for aliases only across function calls*/
106 # pragma optimize("w",on )
107 # endif /* BUGGY_MSC6 */
111 #define STATIC static
114 #define ISMULT1(c) ((c) == '*' || (c) == '+' || (c) == '?')
115 #define ISMULT2(s) ((*s) == '*' || (*s) == '+' || (*s) == '?' || \
116 ((*s) == '{' && regcurly(s)))
118 #define PERL_META "^$.[()|?+*\\"
120 #define META "^$.[()|?+*\\"
124 #undef SPSTART /* dratted cpp namespace... */
127 * Flags to be passed up and down.
129 #define WORST 0 /* Worst case. */
130 #define HASWIDTH 0x1 /* Known to match non-null strings. */
131 #define SIMPLE 0x2 /* Simple enough to be STAR/PLUS operand. */
132 #define SPSTART 0x4 /* Starts with * or +. */
133 #define TRYAGAIN 0x8 /* Weeded out a declaration. */
135 /* Length of a variant. */
137 typedef struct scan_data_t {
143 I32 last_end; /* min value, <0 unless valid. */
146 SV **longest; /* Either &l_fixed, or &l_float. */
150 I32 offset_float_min;
151 I32 offset_float_max;
154 struct regnode_charclass_class *start_class;
158 * Forward declarations for pregcomp()'s friends.
161 static scan_data_t zero_scan_data = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
164 #define SF_BEFORE_EOL (SF_BEFORE_SEOL|SF_BEFORE_MEOL)
165 #define SF_BEFORE_SEOL 0x1
166 #define SF_BEFORE_MEOL 0x2
167 #define SF_FIX_BEFORE_EOL (SF_FIX_BEFORE_SEOL|SF_FIX_BEFORE_MEOL)
168 #define SF_FL_BEFORE_EOL (SF_FL_BEFORE_SEOL|SF_FL_BEFORE_MEOL)
171 # define SF_FIX_SHIFT_EOL (0+2)
172 # define SF_FL_SHIFT_EOL (0+4)
174 # define SF_FIX_SHIFT_EOL (+2)
175 # define SF_FL_SHIFT_EOL (+4)
178 #define SF_FIX_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FIX_SHIFT_EOL)
179 #define SF_FIX_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FIX_SHIFT_EOL)
181 #define SF_FL_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FL_SHIFT_EOL)
182 #define SF_FL_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FL_SHIFT_EOL) /* 0x20 */
183 #define SF_IS_INF 0x40
184 #define SF_HAS_PAR 0x80
185 #define SF_IN_PAR 0x100
186 #define SF_HAS_EVAL 0x200
187 #define SCF_DO_SUBSTR 0x400
188 #define SCF_DO_STCLASS_AND 0x0800
189 #define SCF_DO_STCLASS_OR 0x1000
190 #define SCF_DO_STCLASS (SCF_DO_STCLASS_AND|SCF_DO_STCLASS_OR)
193 #define UTF (PL_reg_flags & RF_utf8)
194 #define LOC (PL_regflags & PMf_LOCALE)
195 #define FOLD (PL_regflags & PMf_FOLD)
197 #define OOB_CHAR8 1234
198 #define OOB_UTF8 123456
199 #define OOB_NAMEDCLASS -1
201 #define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
202 #define CHR_DIST(a,b) (UTF ? utf8_distance(a,b) : a - b)
204 /* Allow for side effects in s */
205 #define REGC(c,s) STMT_START { if (!SIZE_ONLY) *(s) = (c); else (s);} STMT_END
207 static void clear_re(pTHXo_ void *r);
209 /* Mark that we cannot extend a found fixed substring at this point.
210 Updata the longest found anchored substring and the longest found
211 floating substrings if needed. */
214 S_scan_commit(pTHX_ scan_data_t *data)
217 STRLEN l = CHR_SVLEN(data->last_found);
218 STRLEN old_l = CHR_SVLEN(*data->longest);
220 if ((l >= old_l) && ((l > old_l) || (data->flags & SF_BEFORE_EOL))) {
221 sv_setsv(*data->longest, data->last_found);
222 if (*data->longest == data->longest_fixed) {
223 data->offset_fixed = l ? data->last_start_min : data->pos_min;
224 if (data->flags & SF_BEFORE_EOL)
226 |= ((data->flags & SF_BEFORE_EOL) << SF_FIX_SHIFT_EOL);
228 data->flags &= ~SF_FIX_BEFORE_EOL;
231 data->offset_float_min = l ? data->last_start_min : data->pos_min;
232 data->offset_float_max = (l
233 ? data->last_start_max
234 : data->pos_min + data->pos_delta);
235 if (data->flags & SF_BEFORE_EOL)
237 |= ((data->flags & SF_BEFORE_EOL) << SF_FL_SHIFT_EOL);
239 data->flags &= ~SF_FL_BEFORE_EOL;
242 SvCUR_set(data->last_found, 0);
244 data->flags &= ~SF_BEFORE_EOL;
247 /* Can match anything (initialization) */
249 S_cl_anything(pTHX_ struct regnode_charclass_class *cl)
253 ANYOF_CLASS_ZERO(cl);
254 for (value = 0; value < 256; ++value)
255 ANYOF_BITMAP_SET(cl, value);
256 cl->flags = ANYOF_EOS;
258 cl->flags |= ANYOF_LOCALE;
261 /* Can match anything (initialization) */
263 S_cl_is_anything(pTHX_ struct regnode_charclass_class *cl)
267 for (value = 0; value < ANYOF_MAX; value += 2)
268 if (ANYOF_CLASS_TEST(cl, value) && ANYOF_CLASS_TEST(cl, value + 1))
270 for (value = 0; value < 256; ++value)
271 if (!ANYOF_BITMAP_TEST(cl, value))
276 /* Can match anything (initialization) */
278 S_cl_init(pTHX_ struct regnode_charclass_class *cl)
280 Zero(cl, 1, struct regnode_charclass_class);
286 S_cl_init_zero(pTHX_ struct regnode_charclass_class *cl)
288 Zero(cl, 1, struct regnode_charclass_class);
292 cl->flags |= ANYOF_LOCALE;
295 /* 'And' a given class with another one. Can create false positives */
296 /* We assume that cl is not inverted */
298 S_cl_and(pTHX_ struct regnode_charclass_class *cl,
299 struct regnode_charclass_class *and_with)
301 if (!(and_with->flags & ANYOF_CLASS)
302 && !(cl->flags & ANYOF_CLASS)
303 && (and_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
304 && !(and_with->flags & ANYOF_FOLD)
305 && !(cl->flags & ANYOF_FOLD)) {
308 if (and_with->flags & ANYOF_INVERT)
309 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
310 cl->bitmap[i] &= ~and_with->bitmap[i];
312 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
313 cl->bitmap[i] &= and_with->bitmap[i];
314 } /* XXXX: logic is complicated otherwise, leave it along for a moment. */
315 if (!(and_with->flags & ANYOF_EOS))
316 cl->flags &= ~ANYOF_EOS;
319 /* 'OR' a given class with another one. Can create false positives */
320 /* We assume that cl is not inverted */
322 S_cl_or(pTHX_ struct regnode_charclass_class *cl, struct regnode_charclass_class *or_with)
324 if (or_with->flags & ANYOF_INVERT) {
326 * (B1 | CL1) | (!B2 & !CL2) = (B1 | !B2 & !CL2) | (CL1 | (!B2 & !CL2))
327 * <= (B1 | !B2) | (CL1 | !CL2)
328 * which is wasteful if CL2 is small, but we ignore CL2:
329 * (B1 | CL1) | (!B2 & !CL2) <= (B1 | CL1) | !B2 = (B1 | !B2) | CL1
330 * XXXX Can we handle case-fold? Unclear:
331 * (OK1(i) | OK1(i')) | !(OK1(i) | OK1(i')) =
332 * (OK1(i) | OK1(i')) | (!OK1(i) & !OK1(i'))
334 if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
335 && !(or_with->flags & ANYOF_FOLD)
336 && !(cl->flags & ANYOF_FOLD) ) {
339 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
340 cl->bitmap[i] |= ~or_with->bitmap[i];
341 } /* XXXX: logic is complicated otherwise */
346 /* (B1 | CL1) | (B2 | CL2) = (B1 | B2) | (CL1 | CL2)) */
347 if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
348 && (!(or_with->flags & ANYOF_FOLD)
349 || (cl->flags & ANYOF_FOLD)) ) {
352 /* OR char bitmap and class bitmap separately */
353 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
354 cl->bitmap[i] |= or_with->bitmap[i];
355 if (or_with->flags & ANYOF_CLASS) {
356 for (i = 0; i < ANYOF_CLASSBITMAP_SIZE; i++)
357 cl->classflags[i] |= or_with->classflags[i];
358 cl->flags |= ANYOF_CLASS;
361 else { /* XXXX: logic is complicated, leave it along for a moment. */
365 if (or_with->flags & ANYOF_EOS)
366 cl->flags |= ANYOF_EOS;
369 /* REx optimizer. Converts nodes into quickier variants "in place".
370 Finds fixed substrings. */
372 /* Stops at toplevel WHILEM as well as at `last'. At end *scanp is set
373 to the position after last scanned or to NULL. */
376 S_study_chunk(pTHX_ regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 flags)
377 /* scanp: Start here (read-write). */
378 /* deltap: Write maxlen-minlen here. */
379 /* last: Stop before this one. */
382 I32 min = 0, pars = 0, code;
383 regnode *scan = *scanp, *next;
385 int is_inf = (flags & SCF_DO_SUBSTR) && (data->flags & SF_IS_INF);
386 int is_inf_internal = 0; /* The studied chunk is infinite */
387 I32 is_par = OP(scan) == OPEN ? ARG(scan) : 0;
388 scan_data_t data_fake;
389 struct regnode_charclass_class and_with; /* Valid if flags & SCF_DO_STCLASS_OR */
391 while (scan && OP(scan) != END && scan < last) {
392 /* Peephole optimizer: */
394 if (PL_regkind[(U8)OP(scan)] == EXACT) {
395 /* Merge several consecutive EXACTish nodes into one. */
396 regnode *n = regnext(scan);
399 regnode *stop = scan;
402 next = scan + NODE_SZ_STR(scan);
403 /* Skip NOTHING, merge EXACT*. */
405 ( PL_regkind[(U8)OP(n)] == NOTHING ||
406 (stringok && (OP(n) == OP(scan))))
408 && NEXT_OFF(scan) + NEXT_OFF(n) < I16_MAX) {
409 if (OP(n) == TAIL || n > next)
411 if (PL_regkind[(U8)OP(n)] == NOTHING) {
412 NEXT_OFF(scan) += NEXT_OFF(n);
413 next = n + NODE_STEP_REGNODE;
421 int oldl = STR_LEN(scan);
422 regnode *nnext = regnext(n);
424 if (oldl + STR_LEN(n) > U8_MAX)
426 NEXT_OFF(scan) += NEXT_OFF(n);
427 STR_LEN(scan) += STR_LEN(n);
428 next = n + NODE_SZ_STR(n);
429 /* Now we can overwrite *n : */
430 Move(STRING(n), STRING(scan) + oldl,
441 n = scan + NODE_SZ_STR(scan);
443 if (PL_regkind[(U8)OP(n)] != NOTHING || OP(n) == NOTHING) {
451 /* Follow the next-chain of the current node and optimize
452 away all the NOTHINGs from it. */
453 if (OP(scan) != CURLYX) {
454 int max = (reg_off_by_arg[OP(scan)]
456 /* I32 may be smaller than U16 on CRAYs! */
457 : (I32_MAX < U16_MAX ? I32_MAX : U16_MAX));
458 int off = (reg_off_by_arg[OP(scan)] ? ARG(scan) : NEXT_OFF(scan));
462 /* Skip NOTHING and LONGJMP. */
463 while ((n = regnext(n))
464 && ((PL_regkind[(U8)OP(n)] == NOTHING && (noff = NEXT_OFF(n)))
465 || ((OP(n) == LONGJMP) && (noff = ARG(n))))
468 if (reg_off_by_arg[OP(scan)])
471 NEXT_OFF(scan) = off;
473 /* The principal pseudo-switch. Cannot be a switch, since we
474 look into several different things. */
475 if (OP(scan) == BRANCH || OP(scan) == BRANCHJ
476 || OP(scan) == IFTHEN || OP(scan) == SUSPEND) {
477 next = regnext(scan);
480 if (OP(next) == code || code == IFTHEN || code == SUSPEND) {
481 I32 max1 = 0, min1 = I32_MAX, num = 0;
482 struct regnode_charclass_class accum;
484 if (flags & SCF_DO_SUBSTR) /* XXXX Add !SUSPEND? */
485 scan_commit(data); /* Cannot merge strings after this. */
486 if (flags & SCF_DO_STCLASS)
487 cl_init_zero(&accum);
488 while (OP(scan) == code) {
489 I32 deltanext, minnext, f = 0;
490 struct regnode_charclass_class this_class;
495 data_fake.whilem_c = data->whilem_c;
496 next = regnext(scan);
497 scan = NEXTOPER(scan);
499 scan = NEXTOPER(scan);
500 if (flags & SCF_DO_STCLASS) {
501 cl_init(&this_class);
502 data_fake.start_class = &this_class;
503 f = SCF_DO_STCLASS_AND;
505 /* we suppose the run is continuous, last=next...*/
506 minnext = study_chunk(&scan, &deltanext, next,
510 if (max1 < minnext + deltanext)
511 max1 = minnext + deltanext;
512 if (deltanext == I32_MAX)
513 is_inf = is_inf_internal = 1;
515 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
517 if (data && (data_fake.flags & SF_HAS_EVAL))
518 data->flags |= SF_HAS_EVAL;
520 data->whilem_c = data_fake.whilem_c;
521 if (flags & SCF_DO_STCLASS)
522 cl_or(&accum, &this_class);
526 if (code == IFTHEN && num < 2) /* Empty ELSE branch */
528 if (flags & SCF_DO_SUBSTR) {
529 data->pos_min += min1;
530 data->pos_delta += max1 - min1;
531 if (max1 != min1 || is_inf)
532 data->longest = &(data->longest_float);
535 delta += max1 - min1;
536 if (flags & SCF_DO_STCLASS_OR) {
537 cl_or(data->start_class, &accum);
539 cl_and(data->start_class, &and_with);
540 flags &= ~SCF_DO_STCLASS;
543 else if (flags & SCF_DO_STCLASS_AND) {
545 cl_and(data->start_class, &accum);
546 flags &= ~SCF_DO_STCLASS;
549 /* Switch to OR mode: cache the old value of
550 * data->start_class */
551 StructCopy(data->start_class, &and_with,
552 struct regnode_charclass_class);
553 flags &= ~SCF_DO_STCLASS_AND;
554 StructCopy(&accum, data->start_class,
555 struct regnode_charclass_class);
556 flags |= SCF_DO_STCLASS_OR;
557 data->start_class->flags |= ANYOF_EOS;
561 else if (code == BRANCHJ) /* single branch is optimized. */
562 scan = NEXTOPER(NEXTOPER(scan));
563 else /* single branch is optimized. */
564 scan = NEXTOPER(scan);
567 else if (OP(scan) == EXACT) {
568 I32 l = STR_LEN(scan);
570 unsigned char *s = (unsigned char *)STRING(scan);
571 unsigned char *e = s + l;
580 if (flags & SCF_DO_SUBSTR) { /* Update longest substr. */
581 /* The code below prefers earlier match for fixed
582 offset, later match for variable offset. */
583 if (data->last_end == -1) { /* Update the start info. */
584 data->last_start_min = data->pos_min;
585 data->last_start_max = is_inf
586 ? I32_MAX : data->pos_min + data->pos_delta;
588 sv_catpvn(data->last_found, STRING(scan), STR_LEN(scan));
589 data->last_end = data->pos_min + l;
590 data->pos_min += l; /* As in the first entry. */
591 data->flags &= ~SF_BEFORE_EOL;
593 if (flags & SCF_DO_STCLASS_AND) {
594 /* Check whether it is compatible with what we know already! */
597 if (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
598 && !ANYOF_BITMAP_TEST(data->start_class, *STRING(scan))
599 && (!(data->start_class->flags & ANYOF_FOLD)
600 || !ANYOF_BITMAP_TEST(data->start_class,
601 PL_fold[*(U8*)STRING(scan)])))
603 ANYOF_CLASS_ZERO(data->start_class);
604 ANYOF_BITMAP_ZERO(data->start_class);
606 ANYOF_BITMAP_SET(data->start_class, *STRING(scan));
607 data->start_class->flags &= ~ANYOF_EOS;
609 else if (flags & SCF_DO_STCLASS_OR) {
610 /* false positive possible if the class is case-folded */
611 ANYOF_BITMAP_SET(data->start_class, *STRING(scan));
612 data->start_class->flags &= ~ANYOF_EOS;
613 cl_and(data->start_class, &and_with);
615 flags &= ~SCF_DO_STCLASS;
617 else if (PL_regkind[(U8)OP(scan)] == EXACT) { /* But OP != EXACT! */
618 I32 l = STR_LEN(scan);
620 /* Search for fixed substrings supports EXACT only. */
621 if (flags & SCF_DO_SUBSTR)
624 unsigned char *s = (unsigned char *)STRING(scan);
625 unsigned char *e = s + l;
634 if (data && (flags & SCF_DO_SUBSTR))
636 if (flags & SCF_DO_STCLASS_AND) {
637 /* Check whether it is compatible with what we know already! */
640 if (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
641 && !ANYOF_BITMAP_TEST(data->start_class, *STRING(scan))
642 && !ANYOF_BITMAP_TEST(data->start_class,
643 PL_fold[*(U8*)STRING(scan)]))
645 ANYOF_CLASS_ZERO(data->start_class);
646 ANYOF_BITMAP_ZERO(data->start_class);
648 ANYOF_BITMAP_SET(data->start_class, *STRING(scan));
649 data->start_class->flags &= ~ANYOF_EOS;
650 data->start_class->flags |= ANYOF_FOLD;
651 if (OP(scan) == EXACTFL)
652 data->start_class->flags |= ANYOF_LOCALE;
655 else if (flags & SCF_DO_STCLASS_OR) {
656 if (data->start_class->flags & ANYOF_FOLD) {
657 /* false positive possible if the class is case-folded.
658 Assume that the locale settings are the same... */
659 ANYOF_BITMAP_SET(data->start_class, *STRING(scan));
660 data->start_class->flags &= ~ANYOF_EOS;
662 cl_and(data->start_class, &and_with);
664 flags &= ~SCF_DO_STCLASS;
666 else if (strchr((char*)PL_varies,OP(scan))) {
667 I32 mincount, maxcount, minnext, deltanext, pos_before, fl;
669 regnode *oscan = scan;
670 struct regnode_charclass_class this_class;
671 struct regnode_charclass_class *oclass = NULL;
673 switch (PL_regkind[(U8)OP(scan)]) {
674 case WHILEM: /* End of (?:...)* . */
675 scan = NEXTOPER(scan);
678 if (flags & (SCF_DO_SUBSTR | SCF_DO_STCLASS)) {
679 next = NEXTOPER(scan);
680 if (OP(next) == EXACT || (flags & SCF_DO_STCLASS)) {
682 maxcount = REG_INFTY;
683 next = regnext(scan);
684 scan = NEXTOPER(scan);
688 if (flags & SCF_DO_SUBSTR)
693 if (flags & SCF_DO_STCLASS) {
695 maxcount = REG_INFTY;
696 next = regnext(scan);
697 scan = NEXTOPER(scan);
700 is_inf = is_inf_internal = 1;
701 scan = regnext(scan);
702 if (flags & SCF_DO_SUBSTR) {
703 scan_commit(data); /* Cannot extend fixed substrings */
704 data->longest = &(data->longest_float);
706 goto optimize_curly_tail;
708 mincount = ARG1(scan);
709 maxcount = ARG2(scan);
710 next = regnext(scan);
711 scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
713 if (flags & SCF_DO_SUBSTR) {
714 if (mincount == 0) scan_commit(data); /* Cannot extend fixed substrings */
715 pos_before = data->pos_min;
719 data->flags &= ~(SF_HAS_PAR|SF_IN_PAR|SF_HAS_EVAL);
721 data->flags |= SF_IS_INF;
723 if (flags & SCF_DO_STCLASS) {
724 cl_init(&this_class);
725 oclass = data->start_class;
726 data->start_class = &this_class;
727 f |= SCF_DO_STCLASS_AND;
728 f &= ~SCF_DO_STCLASS_OR;
731 /* This will finish on WHILEM, setting scan, or on NULL: */
732 minnext = study_chunk(&scan, &deltanext, last, data,
734 ? (f & ~SCF_DO_SUBSTR) : f);
736 if (flags & SCF_DO_STCLASS)
737 data->start_class = oclass;
738 if (mincount == 0 || minnext == 0) {
739 if (flags & SCF_DO_STCLASS_OR) {
740 cl_or(data->start_class, &this_class);
742 else if (flags & SCF_DO_STCLASS_AND) {
743 /* Switch to OR mode: cache the old value of
744 * data->start_class */
745 StructCopy(data->start_class, &and_with,
746 struct regnode_charclass_class);
747 flags &= ~SCF_DO_STCLASS_AND;
748 StructCopy(&this_class, data->start_class,
749 struct regnode_charclass_class);
750 flags |= SCF_DO_STCLASS_OR;
751 data->start_class->flags |= ANYOF_EOS;
753 } else { /* Non-zero len */
754 if (flags & SCF_DO_STCLASS_OR) {
755 cl_or(data->start_class, &this_class);
756 cl_and(data->start_class, &and_with);
758 else if (flags & SCF_DO_STCLASS_AND)
759 cl_and(data->start_class, &this_class);
760 flags &= ~SCF_DO_STCLASS;
762 if (!scan) /* It was not CURLYX, but CURLY. */
764 if (ckWARN(WARN_REGEXP) && (minnext + deltanext == 0)
765 && !(data->flags & (SF_HAS_PAR|SF_IN_PAR))
766 && maxcount <= REG_INFTY/3) /* Complement check for big count */
767 Perl_warner(aTHX_ WARN_REGEXP,
768 "Strange *+?{} on zero-length expression");
769 min += minnext * mincount;
770 is_inf_internal |= ((maxcount == REG_INFTY
771 && (minnext + deltanext) > 0)
772 || deltanext == I32_MAX);
773 is_inf |= is_inf_internal;
774 delta += (minnext + deltanext) * maxcount - minnext * mincount;
776 /* Try powerful optimization CURLYX => CURLYN. */
777 if ( OP(oscan) == CURLYX && data
778 && data->flags & SF_IN_PAR
779 && !(data->flags & SF_HAS_EVAL)
780 && !deltanext && minnext == 1 ) {
781 /* Try to optimize to CURLYN. */
782 regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS;
783 regnode *nxt1 = nxt, *nxt2;
787 if (!strchr((char*)PL_simple,OP(nxt))
788 && !(PL_regkind[(U8)OP(nxt)] == EXACT
789 && STR_LEN(nxt) == 1))
793 if (OP(nxt) != CLOSE)
795 /* Now we know that nxt2 is the only contents: */
796 oscan->flags = ARG(nxt);
798 OP(nxt1) = NOTHING; /* was OPEN. */
800 OP(nxt1 + 1) = OPTIMIZED; /* was count. */
801 NEXT_OFF(nxt1+ 1) = 0; /* just for consistancy. */
802 NEXT_OFF(nxt2) = 0; /* just for consistancy with CURLY. */
803 OP(nxt) = OPTIMIZED; /* was CLOSE. */
804 OP(nxt + 1) = OPTIMIZED; /* was count. */
805 NEXT_OFF(nxt+ 1) = 0; /* just for consistancy. */
810 /* Try optimization CURLYX => CURLYM. */
811 if ( OP(oscan) == CURLYX && data
812 && !(data->flags & SF_HAS_PAR)
813 && !(data->flags & SF_HAS_EVAL)
815 /* XXXX How to optimize if data == 0? */
816 /* Optimize to a simpler form. */
817 regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN */
821 while ( (nxt2 = regnext(nxt)) /* skip over embedded stuff*/
822 && (OP(nxt2) != WHILEM))
824 OP(nxt2) = SUCCEED; /* Whas WHILEM */
825 /* Need to optimize away parenths. */
826 if (data->flags & SF_IN_PAR) {
827 /* Set the parenth number. */
828 regnode *nxt1 = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN*/
830 if (OP(nxt) != CLOSE)
831 FAIL("panic opt close");
832 oscan->flags = ARG(nxt);
833 OP(nxt1) = OPTIMIZED; /* was OPEN. */
834 OP(nxt) = OPTIMIZED; /* was CLOSE. */
836 OP(nxt1 + 1) = OPTIMIZED; /* was count. */
837 OP(nxt + 1) = OPTIMIZED; /* was count. */
838 NEXT_OFF(nxt1 + 1) = 0; /* just for consistancy. */
839 NEXT_OFF(nxt + 1) = 0; /* just for consistancy. */
842 while ( nxt1 && (OP(nxt1) != WHILEM)) {
843 regnode *nnxt = regnext(nxt1);
846 if (reg_off_by_arg[OP(nxt1)])
847 ARG_SET(nxt1, nxt2 - nxt1);
848 else if (nxt2 - nxt1 < U16_MAX)
849 NEXT_OFF(nxt1) = nxt2 - nxt1;
851 OP(nxt) = NOTHING; /* Cannot beautify */
856 /* Optimize again: */
857 study_chunk(&nxt1, &deltanext, nxt, NULL, 0);
862 else if (OP(oscan) == CURLYX && data && ++data->whilem_c < 16) {
863 /* This stays as CURLYX, and can put the count/of pair. */
864 /* Find WHILEM (as in regexec.c) */
865 regnode *nxt = oscan + NEXT_OFF(oscan);
867 if (OP(PREVOPER(nxt)) == NOTHING) /* LONGJMP */
869 PREVOPER(nxt)->flags = data->whilem_c
870 | (PL_reg_whilem_seen << 4); /* On WHILEM */
872 if (data && fl & (SF_HAS_PAR|SF_IN_PAR))
874 if (flags & SCF_DO_SUBSTR) {
875 SV *last_str = Nullsv;
876 int counted = mincount != 0;
878 if (data->last_end > 0 && mincount != 0) { /* Ends with a string. */
879 I32 b = pos_before >= data->last_start_min
880 ? pos_before : data->last_start_min;
882 char *s = SvPV(data->last_found, l);
883 I32 old = b - data->last_start_min;
886 old = utf8_hop((U8*)s, old) - (U8*)s;
889 /* Get the added string: */
890 last_str = newSVpvn(s + old, l);
891 if (deltanext == 0 && pos_before == b) {
892 /* What was added is a constant string */
894 SvGROW(last_str, (mincount * l) + 1);
895 repeatcpy(SvPVX(last_str) + l,
896 SvPVX(last_str), l, mincount - 1);
897 SvCUR(last_str) *= mincount;
898 /* Add additional parts. */
899 SvCUR_set(data->last_found,
900 SvCUR(data->last_found) - l);
901 sv_catsv(data->last_found, last_str);
902 data->last_end += l * (mincount - 1);
906 /* It is counted once already... */
907 data->pos_min += minnext * (mincount - counted);
908 data->pos_delta += - counted * deltanext +
909 (minnext + deltanext) * maxcount - minnext * mincount;
910 if (mincount != maxcount) {
911 /* Cannot extend fixed substrings found inside
914 if (mincount && last_str) {
915 sv_setsv(data->last_found, last_str);
916 data->last_end = data->pos_min;
917 data->last_start_min =
918 data->pos_min - CHR_SVLEN(last_str);
919 data->last_start_max = is_inf
921 : data->pos_min + data->pos_delta
922 - CHR_SVLEN(last_str);
924 data->longest = &(data->longest_float);
926 SvREFCNT_dec(last_str);
928 if (data && (fl & SF_HAS_EVAL))
929 data->flags |= SF_HAS_EVAL;
931 if (OP(oscan) != CURLYX) {
932 while (PL_regkind[(U8)OP(next = regnext(oscan))] == NOTHING
934 NEXT_OFF(oscan) += NEXT_OFF(next);
937 default: /* REF and CLUMP only? */
938 if (flags & SCF_DO_SUBSTR) {
939 scan_commit(data); /* Cannot expect anything... */
940 data->longest = &(data->longest_float);
942 is_inf = is_inf_internal = 1;
943 if (flags & SCF_DO_STCLASS_OR)
944 cl_anything(data->start_class);
945 flags &= ~SCF_DO_STCLASS;
949 else if (strchr((char*)PL_simple,OP(scan)) || PL_regkind[(U8)OP(scan)] == ANYUTF8) {
952 if (flags & SCF_DO_SUBSTR) {
957 if (flags & SCF_DO_STCLASS) {
958 data->start_class->flags &= ~ANYOF_EOS; /* No match on empty */
960 /* Some of the logic below assumes that switching
961 locale on will only add false positives. */
962 switch (PL_regkind[(U8)OP(scan)]) {
979 /* Perl_croak(aTHX_ "panic: unexpected simple REx opcode %d", OP(scan)); */
980 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
981 cl_anything(data->start_class);
984 if (OP(scan) == SANY)
986 if (flags & SCF_DO_STCLASS_OR) { /* Everything but \n */
987 value = (ANYOF_BITMAP_TEST(data->start_class,'\n')
988 || (data->start_class->flags & ANYOF_CLASS));
989 cl_anything(data->start_class);
991 if (flags & SCF_DO_STCLASS_AND || !value)
992 ANYOF_BITMAP_CLEAR(data->start_class,'\n');
995 if (flags & SCF_DO_STCLASS_AND)
996 cl_and(data->start_class,
997 (struct regnode_charclass_class*)scan);
999 cl_or(data->start_class,
1000 (struct regnode_charclass_class*)scan);
1003 if (flags & SCF_DO_STCLASS_AND) {
1004 if (!(data->start_class->flags & ANYOF_LOCALE)) {
1005 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
1006 for (value = 0; value < 256; value++)
1007 if (!isALNUM(value))
1008 ANYOF_BITMAP_CLEAR(data->start_class, value);
1012 if (data->start_class->flags & ANYOF_LOCALE)
1013 ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
1015 for (value = 0; value < 256; value++)
1017 ANYOF_BITMAP_SET(data->start_class, value);
1022 if (flags & SCF_DO_STCLASS_AND) {
1023 if (data->start_class->flags & ANYOF_LOCALE)
1024 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
1027 ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
1028 data->start_class->flags |= ANYOF_LOCALE;
1032 if (flags & SCF_DO_STCLASS_AND) {
1033 if (!(data->start_class->flags & ANYOF_LOCALE)) {
1034 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
1035 for (value = 0; value < 256; value++)
1037 ANYOF_BITMAP_CLEAR(data->start_class, value);
1041 if (data->start_class->flags & ANYOF_LOCALE)
1042 ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
1044 for (value = 0; value < 256; value++)
1045 if (!isALNUM(value))
1046 ANYOF_BITMAP_SET(data->start_class, value);
1051 if (flags & SCF_DO_STCLASS_AND) {
1052 if (data->start_class->flags & ANYOF_LOCALE)
1053 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
1056 data->start_class->flags |= ANYOF_LOCALE;
1057 ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
1061 if (flags & SCF_DO_STCLASS_AND) {
1062 if (!(data->start_class->flags & ANYOF_LOCALE)) {
1063 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
1064 for (value = 0; value < 256; value++)
1065 if (!isSPACE(value))
1066 ANYOF_BITMAP_CLEAR(data->start_class, value);
1070 if (data->start_class->flags & ANYOF_LOCALE)
1071 ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
1073 for (value = 0; value < 256; value++)
1075 ANYOF_BITMAP_SET(data->start_class, value);
1080 if (flags & SCF_DO_STCLASS_AND) {
1081 if (data->start_class->flags & ANYOF_LOCALE)
1082 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
1085 data->start_class->flags |= ANYOF_LOCALE;
1086 ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
1090 if (flags & SCF_DO_STCLASS_AND) {
1091 if (!(data->start_class->flags & ANYOF_LOCALE)) {
1092 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
1093 for (value = 0; value < 256; value++)
1095 ANYOF_BITMAP_CLEAR(data->start_class, value);
1099 if (data->start_class->flags & ANYOF_LOCALE)
1100 ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
1102 for (value = 0; value < 256; value++)
1103 if (!isSPACE(value))
1104 ANYOF_BITMAP_SET(data->start_class, value);
1109 if (flags & SCF_DO_STCLASS_AND) {
1110 if (data->start_class->flags & ANYOF_LOCALE) {
1111 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
1112 for (value = 0; value < 256; value++)
1113 if (!isSPACE(value))
1114 ANYOF_BITMAP_CLEAR(data->start_class, value);
1118 data->start_class->flags |= ANYOF_LOCALE;
1119 ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
1123 if (flags & SCF_DO_STCLASS_AND) {
1124 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NDIGIT);
1125 for (value = 0; value < 256; value++)
1126 if (!isDIGIT(value))
1127 ANYOF_BITMAP_CLEAR(data->start_class, value);
1130 if (data->start_class->flags & ANYOF_LOCALE)
1131 ANYOF_CLASS_SET(data->start_class,ANYOF_DIGIT);
1133 for (value = 0; value < 256; value++)
1135 ANYOF_BITMAP_SET(data->start_class, value);
1140 if (flags & SCF_DO_STCLASS_AND) {
1141 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_DIGIT);
1142 for (value = 0; value < 256; value++)
1144 ANYOF_BITMAP_CLEAR(data->start_class, value);
1147 if (data->start_class->flags & ANYOF_LOCALE)
1148 ANYOF_CLASS_SET(data->start_class,ANYOF_NDIGIT);
1150 for (value = 0; value < 256; value++)
1151 if (!isDIGIT(value))
1152 ANYOF_BITMAP_SET(data->start_class, value);
1157 if (flags & SCF_DO_STCLASS_OR)
1158 cl_and(data->start_class, &and_with);
1159 flags &= ~SCF_DO_STCLASS;
1162 else if (PL_regkind[(U8)OP(scan)] == EOL && flags & SCF_DO_SUBSTR) {
1163 data->flags |= (OP(scan) == MEOL
1167 else if ( PL_regkind[(U8)OP(scan)] == BRANCHJ
1168 /* Lookbehind, or need to calculate parens/evals/stclass: */
1169 && (scan->flags || data || (flags & SCF_DO_STCLASS))
1170 && (OP(scan) == IFMATCH || OP(scan) == UNLESSM)) {
1171 /* Lookahead/lookbehind */
1172 I32 deltanext, minnext;
1174 struct regnode_charclass_class intrnl;
1177 data_fake.flags = 0;
1179 data_fake.whilem_c = data->whilem_c;
1180 if ( flags & SCF_DO_STCLASS && !scan->flags
1181 && OP(scan) == IFMATCH ) { /* Lookahead */
1183 data_fake.start_class = &intrnl;
1184 f = SCF_DO_STCLASS_AND;
1186 next = regnext(scan);
1187 nscan = NEXTOPER(NEXTOPER(scan));
1188 minnext = study_chunk(&nscan, &deltanext, last, &data_fake, f);
1191 FAIL("variable length lookbehind not implemented");
1193 else if (minnext > U8_MAX) {
1194 FAIL2("lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX);
1196 scan->flags = minnext;
1198 if (data && data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
1200 if (data && (data_fake.flags & SF_HAS_EVAL))
1201 data->flags |= SF_HAS_EVAL;
1203 data->whilem_c = data_fake.whilem_c;
1205 int was = (data->start_class->flags & ANYOF_EOS);
1207 cl_and(data->start_class, &intrnl);
1209 data->start_class->flags |= ANYOF_EOS;
1212 else if (OP(scan) == OPEN) {
1215 else if (OP(scan) == CLOSE && ARG(scan) == is_par) {
1216 next = regnext(scan);
1218 if ( next && (OP(next) != WHILEM) && next < last)
1219 is_par = 0; /* Disable optimization */
1221 else if (OP(scan) == EVAL) {
1223 data->flags |= SF_HAS_EVAL;
1225 else if (OP(scan) == LOGICAL && scan->flags == 2) { /* Embedded */
1226 if (flags & SCF_DO_SUBSTR) {
1228 data->longest = &(data->longest_float);
1230 is_inf = is_inf_internal = 1;
1231 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
1232 cl_anything(data->start_class);
1234 /* Else: zero-length, ignore. */
1235 scan = regnext(scan);
1240 *deltap = is_inf_internal ? I32_MAX : delta;
1241 if (flags & SCF_DO_SUBSTR && is_inf)
1242 data->pos_delta = I32_MAX - data->pos_min;
1243 if (is_par > U8_MAX)
1245 if (is_par && pars==1 && data) {
1246 data->flags |= SF_IN_PAR;
1247 data->flags &= ~SF_HAS_PAR;
1249 else if (pars && data) {
1250 data->flags |= SF_HAS_PAR;
1251 data->flags &= ~SF_IN_PAR;
1253 if (flags & SCF_DO_STCLASS_OR)
1254 cl_and(data->start_class, &and_with);
1259 S_add_data(pTHX_ I32 n, char *s)
1262 if (PL_regcomp_rx->data) {
1263 Renewc(PL_regcomp_rx->data,
1264 sizeof(*PL_regcomp_rx->data) + sizeof(void*) * (PL_regcomp_rx->data->count + n - 1),
1265 char, struct reg_data);
1266 Renew(PL_regcomp_rx->data->what, PL_regcomp_rx->data->count + n, U8);
1267 PL_regcomp_rx->data->count += n;
1270 Newc(1207, PL_regcomp_rx->data, sizeof(*PL_regcomp_rx->data) + sizeof(void*) * (n - 1),
1271 char, struct reg_data);
1272 New(1208, PL_regcomp_rx->data->what, n, U8);
1273 PL_regcomp_rx->data->count = n;
1275 Copy(s, PL_regcomp_rx->data->what + PL_regcomp_rx->data->count - n, n, U8);
1276 return PL_regcomp_rx->data->count - n;
1280 Perl_reginitcolors(pTHX)
1284 char *s = PerlEnv_getenv("PERL_RE_COLORS");
1287 PL_colors[0] = s = savepv(s);
1289 s = strchr(s, '\t');
1295 PL_colors[i] = s = "";
1299 PL_colors[i++] = "";
1305 - pregcomp - compile a regular expression into internal code
1307 * We can't allocate space until we know how big the compiled form will be,
1308 * but we can't compile it (and thus know how big it is) until we've got a
1309 * place to put the code. So we cheat: we compile it twice, once with code
1310 * generation turned off and size counting turned on, and once "for real".
1311 * This also means that we don't allocate space until we are sure that the
1312 * thing really will compile successfully, and we never have to move the
1313 * code and thus invalidate pointers into it. (Note that it has to be in
1314 * one piece because free() must be able to free it all.) [NB: not true in perl]
1316 * Beware that the optimization-preparation code in here knows about some
1317 * of the structure of the compiled regexp. [I'll say.]
1320 Perl_pregcomp(pTHX_ char *exp, char *xend, PMOP *pm)
1333 FAIL("NULL regexp argument");
1335 if (pm->op_pmdynflags & PMdf_UTF8) {
1336 PL_reg_flags |= RF_utf8;
1341 PL_regprecomp = savepvn(exp, xend - exp);
1342 DEBUG_r(if (!PL_colorset) reginitcolors());
1343 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sCompiling REx%s `%s%*s%s'\n",
1344 PL_colors[4],PL_colors[5],PL_colors[0],
1345 (int)(xend - exp), PL_regprecomp, PL_colors[1]));
1346 PL_regflags = pm->op_pmflags;
1350 PL_seen_zerolen = *exp == '^' ? -1 : 0;
1354 /* First pass: determine size, legality. */
1355 PL_regcomp_parse = exp;
1360 PL_regcode = &PL_regdummy;
1361 PL_reg_whilem_seen = 0;
1362 #if 0 /* REGC() is (currently) a NOP at the first pass.
1363 * Clever compilers notice this and complain. --jhi */
1364 REGC((U8)REG_MAGIC, (char*)PL_regcode);
1366 if (reg(0, &flags) == NULL) {
1367 Safefree(PL_regprecomp);
1368 PL_regprecomp = Nullch;
1371 DEBUG_r(PerlIO_printf(Perl_debug_log, "size %"IVdf" ", (IV)PL_regsize));
1373 /* Small enough for pointer-storage convention?
1374 If extralen==0, this means that we will not need long jumps. */
1375 if (PL_regsize >= 0x10000L && PL_extralen)
1376 PL_regsize += PL_extralen;
1379 if (PL_reg_whilem_seen > 15)
1380 PL_reg_whilem_seen = 15;
1382 /* Allocate space and initialize. */
1383 Newc(1001, r, sizeof(regexp) + (unsigned)PL_regsize * sizeof(regnode),
1386 FAIL("regexp out of space");
1388 /* avoid reading uninitialized memory in DEBUGGING code in study_chunk() */
1389 Zero(r, sizeof(regexp) + (unsigned)PL_regsize * sizeof(regnode), char);
1392 r->prelen = xend - exp;
1393 r->precomp = PL_regprecomp;
1395 r->reganch = pm->op_pmflags & PMf_COMPILETIME;
1396 r->nparens = PL_regnpar - 1; /* set early to validate backrefs */
1398 r->substrs = 0; /* Useful during FAIL. */
1399 r->startp = 0; /* Useful during FAIL. */
1400 r->endp = 0; /* Useful during FAIL. */
1404 /* Second pass: emit code. */
1405 PL_regcomp_parse = exp;
1409 PL_regcode = r->program;
1410 /* Store the count of eval-groups for security checks: */
1411 PL_regcode->next_off = ((PL_seen_evals > U16_MAX) ? U16_MAX : PL_seen_evals);
1412 REGC((U8)REG_MAGIC, (char*) PL_regcode++);
1414 if (reg(0, &flags) == NULL)
1417 /* Dig out information for optimizations. */
1418 r->reganch = pm->op_pmflags & PMf_COMPILETIME; /* Again? */
1419 pm->op_pmflags = PL_regflags;
1421 r->reganch |= ROPT_UTF8;
1422 r->regstclass = NULL;
1423 if (PL_regnaughty >= 10) /* Probably an expensive pattern. */
1424 r->reganch |= ROPT_NAUGHTY;
1425 scan = r->program + 1; /* First BRANCH. */
1427 /* XXXX To minimize changes to RE engine we always allocate
1428 3-units-long substrs field. */
1429 Newz(1004, r->substrs, 1, struct reg_substr_data);
1431 StructCopy(&zero_scan_data, &data, scan_data_t);
1432 /* XXXX Should not we check for something else? Usually it is OPEN1... */
1433 if (OP(scan) != BRANCH) { /* Only one top-level choice. */
1435 STRLEN longest_float_length, longest_fixed_length;
1436 struct regnode_charclass_class ch_class;
1440 /* Skip introductions and multiplicators >= 1. */
1441 while ((OP(first) == OPEN && (sawopen = 1)) ||
1442 /* An OR of *one* alternative - should not happen now. */
1443 (OP(first) == BRANCH && OP(regnext(first)) != BRANCH) ||
1444 (OP(first) == PLUS) ||
1445 (OP(first) == MINMOD) ||
1446 /* An {n,m} with n>0 */
1447 (PL_regkind[(U8)OP(first)] == CURLY && ARG1(first) > 0) ) {
1448 if (OP(first) == PLUS)
1451 first += regarglen[(U8)OP(first)];
1452 first = NEXTOPER(first);
1455 /* Starting-point info. */
1457 if (PL_regkind[(U8)OP(first)] == EXACT) {
1458 if (OP(first) == EXACT); /* Empty, get anchored substr later. */
1459 else if ((OP(first) == EXACTF || OP(first) == EXACTFL)
1461 r->regstclass = first;
1463 else if (strchr((char*)PL_simple,OP(first)))
1464 r->regstclass = first;
1465 else if (PL_regkind[(U8)OP(first)] == BOUND ||
1466 PL_regkind[(U8)OP(first)] == NBOUND)
1467 r->regstclass = first;
1468 else if (PL_regkind[(U8)OP(first)] == BOL) {
1469 r->reganch |= (OP(first) == MBOL
1471 : (OP(first) == SBOL
1474 first = NEXTOPER(first);
1477 else if (OP(first) == GPOS) {
1478 r->reganch |= ROPT_ANCH_GPOS;
1479 first = NEXTOPER(first);
1482 else if ((OP(first) == STAR &&
1483 PL_regkind[(U8)OP(NEXTOPER(first))] == REG_ANY) &&
1484 !(r->reganch & ROPT_ANCH) )
1486 /* turn .* into ^.* with an implied $*=1 */
1487 int type = OP(NEXTOPER(first));
1489 if (type == REG_ANY || type == ANYUTF8)
1490 type = ROPT_ANCH_MBOL;
1492 type = ROPT_ANCH_SBOL;
1494 r->reganch |= type | ROPT_IMPLICIT;
1495 first = NEXTOPER(first);
1498 if (sawplus && (!sawopen || !PL_regsawback)
1499 && !(PL_regseen & REG_SEEN_EVAL)) /* May examine pos and $& */
1500 /* x+ must match at the 1st pos of run of x's */
1501 r->reganch |= ROPT_SKIP;
1503 /* Scan is after the zeroth branch, first is atomic matcher. */
1504 DEBUG_r(PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n",
1505 (IV)(first - scan + 1)));
1507 * If there's something expensive in the r.e., find the
1508 * longest literal string that must appear and make it the
1509 * regmust. Resolve ties in favor of later strings, since
1510 * the regstart check works with the beginning of the r.e.
1511 * and avoiding duplication strengthens checking. Not a
1512 * strong reason, but sufficient in the absence of others.
1513 * [Now we resolve ties in favor of the earlier string if
1514 * it happens that c_offset_min has been invalidated, since the
1515 * earlier string may buy us something the later one won't.]
1519 data.longest_fixed = newSVpvn("",0);
1520 data.longest_float = newSVpvn("",0);
1521 data.last_found = newSVpvn("",0);
1522 data.longest = &(data.longest_fixed);
1524 if (!r->regstclass) {
1526 data.start_class = &ch_class;
1527 stclass_flag = SCF_DO_STCLASS_AND;
1528 } else /* XXXX Check for BOUND? */
1531 minlen = study_chunk(&first, &fake, scan + PL_regsize, /* Up to end */
1532 &data, SCF_DO_SUBSTR | stclass_flag);
1533 if ( PL_regnpar == 1 && data.longest == &(data.longest_fixed)
1534 && data.last_start_min == 0 && data.last_end > 0
1536 && (!(PL_regseen & REG_SEEN_GPOS) || (r->reganch & ROPT_ANCH_GPOS)))
1537 r->reganch |= ROPT_CHECK_ALL;
1539 SvREFCNT_dec(data.last_found);
1541 longest_float_length = CHR_SVLEN(data.longest_float);
1542 if (longest_float_length
1543 || (data.flags & SF_FL_BEFORE_EOL
1544 && (!(data.flags & SF_FL_BEFORE_MEOL)
1545 || (PL_regflags & PMf_MULTILINE)))) {
1548 if (SvCUR(data.longest_fixed) /* ok to leave SvCUR */
1549 && data.offset_fixed == data.offset_float_min
1550 && SvCUR(data.longest_fixed) == SvCUR(data.longest_float))
1551 goto remove_float; /* As in (a)+. */
1553 r->float_substr = data.longest_float;
1554 r->float_min_offset = data.offset_float_min;
1555 r->float_max_offset = data.offset_float_max;
1556 t = (data.flags & SF_FL_BEFORE_EOL /* Can't have SEOL and MULTI */
1557 && (!(data.flags & SF_FL_BEFORE_MEOL)
1558 || (PL_regflags & PMf_MULTILINE)));
1559 fbm_compile(r->float_substr, t ? FBMcf_TAIL : 0);
1563 r->float_substr = Nullsv;
1564 SvREFCNT_dec(data.longest_float);
1565 longest_float_length = 0;
1568 longest_fixed_length = CHR_SVLEN(data.longest_fixed);
1569 if (longest_fixed_length
1570 || (data.flags & SF_FIX_BEFORE_EOL /* Cannot have SEOL and MULTI */
1571 && (!(data.flags & SF_FIX_BEFORE_MEOL)
1572 || (PL_regflags & PMf_MULTILINE)))) {
1575 r->anchored_substr = data.longest_fixed;
1576 r->anchored_offset = data.offset_fixed;
1577 t = (data.flags & SF_FIX_BEFORE_EOL /* Can't have SEOL and MULTI */
1578 && (!(data.flags & SF_FIX_BEFORE_MEOL)
1579 || (PL_regflags & PMf_MULTILINE)));
1580 fbm_compile(r->anchored_substr, t ? FBMcf_TAIL : 0);
1583 r->anchored_substr = Nullsv;
1584 SvREFCNT_dec(data.longest_fixed);
1585 longest_fixed_length = 0;
1588 && (OP(r->regstclass) == REG_ANY || OP(r->regstclass) == ANYUTF8
1589 || OP(r->regstclass) == SANYUTF8 || OP(r->regstclass) == SANY))
1590 r->regstclass = NULL;
1591 if ((!r->anchored_substr || r->anchored_offset) && stclass_flag
1592 && !(data.start_class->flags & ANYOF_EOS)
1593 && !cl_is_anything(data.start_class)) {
1595 I32 n = add_data(1, "f");
1597 New(1006, PL_regcomp_rx->data->data[n], 1,
1598 struct regnode_charclass_class);
1599 StructCopy(data.start_class,
1600 (struct regnode_charclass_class*)PL_regcomp_rx->data->data[n],
1601 struct regnode_charclass_class);
1602 r->regstclass = (regnode*)PL_regcomp_rx->data->data[n];
1603 r->reganch &= ~ROPT_SKIP; /* Used in find_byclass(). */
1604 DEBUG_r((sv = sv_newmortal(),
1605 regprop(sv, (regnode*)data.start_class),
1606 PerlIO_printf(Perl_debug_log, "synthetic stclass `%s'.\n",
1610 /* A temporary algorithm prefers floated substr to fixed one to dig more info. */
1611 if (longest_fixed_length > longest_float_length) {
1612 r->check_substr = r->anchored_substr;
1613 r->check_offset_min = r->check_offset_max = r->anchored_offset;
1614 if (r->reganch & ROPT_ANCH_SINGLE)
1615 r->reganch |= ROPT_NOSCAN;
1618 r->check_substr = r->float_substr;
1619 r->check_offset_min = data.offset_float_min;
1620 r->check_offset_max = data.offset_float_max;
1622 /* XXXX Currently intuiting is not compatible with ANCH_GPOS.
1623 This should be changed ASAP! */
1624 if (r->check_substr && !(r->reganch & ROPT_ANCH_GPOS)) {
1625 r->reganch |= RE_USE_INTUIT;
1626 if (SvTAIL(r->check_substr))
1627 r->reganch |= RE_INTUIT_TAIL;
1631 /* Several toplevels. Best we can is to set minlen. */
1633 struct regnode_charclass_class ch_class;
1635 DEBUG_r(PerlIO_printf(Perl_debug_log, "\n"));
1636 scan = r->program + 1;
1638 data.start_class = &ch_class;
1639 minlen = study_chunk(&scan, &fake, scan + PL_regsize, &data, SCF_DO_STCLASS_AND);
1640 r->check_substr = r->anchored_substr = r->float_substr = Nullsv;
1641 if (!(data.start_class->flags & ANYOF_EOS)
1642 && !cl_is_anything(data.start_class)) {
1644 I32 n = add_data(1, "f");
1646 New(1006, PL_regcomp_rx->data->data[n], 1,
1647 struct regnode_charclass_class);
1648 StructCopy(data.start_class,
1649 (struct regnode_charclass_class*)PL_regcomp_rx->data->data[n],
1650 struct regnode_charclass_class);
1651 r->regstclass = (regnode*)PL_regcomp_rx->data->data[n];
1652 r->reganch &= ~ROPT_SKIP; /* Used in find_byclass(). */
1653 DEBUG_r((sv = sv_newmortal(),
1654 regprop(sv, (regnode*)data.start_class),
1655 PerlIO_printf(Perl_debug_log, "synthetic stclass `%s'.\n",
1661 if (PL_regseen & REG_SEEN_GPOS)
1662 r->reganch |= ROPT_GPOS_SEEN;
1663 if (PL_regseen & REG_SEEN_LOOKBEHIND)
1664 r->reganch |= ROPT_LOOKBEHIND_SEEN;
1665 if (PL_regseen & REG_SEEN_EVAL)
1666 r->reganch |= ROPT_EVAL_SEEN;
1667 Newz(1002, r->startp, PL_regnpar, I32);
1668 Newz(1002, r->endp, PL_regnpar, I32);
1669 DEBUG_r(regdump(r));
1674 - reg - regular expression, i.e. main body or parenthesized thing
1676 * Caller must absorb opening parenthesis.
1678 * Combining parenthesis handling with the base level of regular expression
1679 * is a trifle forced, but the need to tie the tails of the branches to what
1680 * follows makes it hard to avoid.
1683 S_reg(pTHX_ I32 paren, I32 *flagp)
1684 /* paren: Parenthesized? 0=top, 1=(, inside: changed to letter. */
1687 register regnode *ret; /* Will be the head of the group. */
1688 register regnode *br;
1689 register regnode *lastbr;
1690 register regnode *ender = 0;
1691 register I32 parno = 0;
1692 I32 flags, oregflags = PL_regflags, have_branch = 0, open = 0;
1695 *flagp = 0; /* Tentatively. */
1697 /* Make an OPEN node, if parenthesized. */
1699 if (*PL_regcomp_parse == '?') {
1700 U16 posflags = 0, negflags = 0;
1701 U16 *flagsp = &posflags;
1705 paren = *PL_regcomp_parse++;
1706 ret = NULL; /* For look-ahead/behind. */
1709 PL_regseen |= REG_SEEN_LOOKBEHIND;
1710 if (*PL_regcomp_parse == '!')
1712 if (*PL_regcomp_parse != '=' && *PL_regcomp_parse != '!')
1723 FAIL2("Sequence (?%c...) not implemented", (int)paren);
1726 while (*PL_regcomp_parse && *PL_regcomp_parse != ')')
1728 if (*PL_regcomp_parse != ')')
1729 FAIL("Sequence (?#... not terminated");
1734 Perl_warner(aTHX_ WARN_REGEXP, "(?p{}) is deprecated - use (??{})");
1738 paren = *PL_regcomp_parse++;
1743 I32 count = 1, n = 0;
1745 char *s = PL_regcomp_parse;
1747 OP_4tree *sop, *rop;
1750 PL_regseen |= REG_SEEN_EVAL;
1751 while (count && (c = *PL_regcomp_parse)) {
1752 if (c == '\\' && PL_regcomp_parse[1])
1760 if (*PL_regcomp_parse != ')')
1761 FAIL("Sequence (?{...}) not terminated or not {}-balanced");
1765 if (PL_regcomp_parse - 1 - s)
1766 sv = newSVpvn(s, PL_regcomp_parse - 1 - s);
1768 sv = newSVpvn("", 0);
1770 rop = sv_compile_2op(sv, &sop, "re", &av);
1772 n = add_data(3, "nop");
1773 PL_regcomp_rx->data->data[n] = (void*)rop;
1774 PL_regcomp_rx->data->data[n+1] = (void*)sop;
1775 PL_regcomp_rx->data->data[n+2] = (void*)av;
1778 else { /* First pass */
1779 if (PL_reginterp_cnt < ++PL_seen_evals
1780 && PL_curcop != &PL_compiling)
1781 /* No compiled RE interpolated, has runtime
1782 components ===> unsafe. */
1783 FAIL("Eval-group not allowed at runtime, use re 'eval'");
1785 FAIL("Eval-group in insecure regular expression");
1790 ret = reg_node(LOGICAL);
1793 regtail(ret, reganode(EVAL, n));
1796 return reganode(EVAL, n);
1800 if (PL_regcomp_parse[0] == '?') {
1801 if (PL_regcomp_parse[1] == '=' || PL_regcomp_parse[1] == '!'
1802 || PL_regcomp_parse[1] == '<'
1803 || PL_regcomp_parse[1] == '{') { /* Lookahead or eval. */
1806 ret = reg_node(LOGICAL);
1809 regtail(ret, reg(1, &flag));
1813 else if (PL_regcomp_parse[0] >= '1' && PL_regcomp_parse[0] <= '9' ) {
1814 parno = atoi(PL_regcomp_parse++);
1816 while (isDIGIT(*PL_regcomp_parse))
1818 ret = reganode(GROUPP, parno);
1819 if ((c = *nextchar()) != ')')
1820 FAIL2("Switch (?(number%c not recognized", c);
1822 regtail(ret, reganode(IFTHEN, 0));
1823 br = regbranch(&flags, 1);
1825 br = reganode(LONGJMP, 0);
1827 regtail(br, reganode(LONGJMP, 0));
1832 lastbr = reganode(IFTHEN, 0); /* Fake one for optimizer. */
1833 regbranch(&flags, 1);
1834 regtail(ret, lastbr);
1842 FAIL("Switch (?(condition)... contains too many branches");
1843 ender = reg_node(TAIL);
1846 regtail(lastbr, ender);
1847 regtail(NEXTOPER(NEXTOPER(lastbr)), ender);
1850 regtail(ret, ender);
1854 FAIL2("Unknown condition for (?(%.2s", PL_regcomp_parse);
1858 FAIL("Sequence (? incomplete");
1863 while (*PL_regcomp_parse && strchr("iogcmsx", *PL_regcomp_parse)) {
1864 if (*PL_regcomp_parse != 'o')
1865 pmflag(flagsp, *PL_regcomp_parse);
1868 if (*PL_regcomp_parse == '-') {
1873 PL_regflags |= posflags;
1874 PL_regflags &= ~negflags;
1875 if (*PL_regcomp_parse == ':') {
1881 if (*PL_regcomp_parse != ')')
1882 FAIL2("Sequence (?%c...) not recognized", *PL_regcomp_parse);
1891 ret = reganode(OPEN, parno);
1898 /* Pick up the branches, linking them together. */
1899 br = regbranch(&flags, 1);
1902 if (*PL_regcomp_parse == '|') {
1903 if (!SIZE_ONLY && PL_extralen) {
1904 reginsert(BRANCHJ, br);
1907 reginsert(BRANCH, br);
1910 PL_extralen += 1; /* For BRANCHJ-BRANCH. */
1912 else if (paren == ':') {
1913 *flagp |= flags&SIMPLE;
1915 if (open) { /* Starts with OPEN. */
1916 regtail(ret, br); /* OPEN -> first. */
1918 else if (paren != '?') /* Not Conditional */
1922 *flagp |= flags&SPSTART;
1924 while (*PL_regcomp_parse == '|') {
1925 if (!SIZE_ONLY && PL_extralen) {
1926 ender = reganode(LONGJMP,0);
1927 regtail(NEXTOPER(NEXTOPER(lastbr)), ender); /* Append to the previous. */
1930 PL_extralen += 2; /* Account for LONGJMP. */
1932 br = regbranch(&flags, 0);
1935 regtail(lastbr, br); /* BRANCH -> BRANCH. */
1939 *flagp |= flags&SPSTART;
1942 if (have_branch || paren != ':') {
1943 /* Make a closing node, and hook it on the end. */
1946 ender = reg_node(TAIL);
1949 ender = reganode(CLOSE, parno);
1955 *flagp &= ~HASWIDTH;
1958 ender = reg_node(SUCCEED);
1961 ender = reg_node(END);
1964 regtail(lastbr, ender);
1967 /* Hook the tails of the branches to the closing node. */
1968 for (br = ret; br != NULL; br = regnext(br)) {
1969 regoptail(br, ender);
1976 static char parens[] = "=!<,>";
1978 if (paren && (p = strchr(parens, paren))) {
1979 int node = ((p - parens) % 2) ? UNLESSM : IFMATCH;
1980 int flag = (p - parens) > 1;
1983 node = SUSPEND, flag = 0;
1984 reginsert(node,ret);
1986 regtail(ret, reg_node(TAIL));
1990 /* Check for proper termination. */
1992 PL_regflags = oregflags;
1993 if (PL_regcomp_parse >= PL_regxend || *nextchar() != ')') {
1994 FAIL("unmatched () in regexp");
1997 else if (!paren && PL_regcomp_parse < PL_regxend) {
1998 if (*PL_regcomp_parse == ')') {
1999 FAIL("unmatched () in regexp");
2002 FAIL("junk on end of regexp"); /* "Can't happen". */
2010 - regbranch - one alternative of an | operator
2012 * Implements the concatenation operator.
2015 S_regbranch(pTHX_ I32 *flagp, I32 first)
2018 register regnode *ret;
2019 register regnode *chain = NULL;
2020 register regnode *latest;
2021 I32 flags = 0, c = 0;
2026 if (!SIZE_ONLY && PL_extralen)
2027 ret = reganode(BRANCHJ,0);
2029 ret = reg_node(BRANCH);
2032 if (!first && SIZE_ONLY)
2033 PL_extralen += 1; /* BRANCHJ */
2035 *flagp = WORST; /* Tentatively. */
2039 while (PL_regcomp_parse < PL_regxend && *PL_regcomp_parse != '|' && *PL_regcomp_parse != ')') {
2041 latest = regpiece(&flags);
2042 if (latest == NULL) {
2043 if (flags & TRYAGAIN)
2047 else if (ret == NULL)
2049 *flagp |= flags&HASWIDTH;
2050 if (chain == NULL) /* First piece. */
2051 *flagp |= flags&SPSTART;
2054 regtail(chain, latest);
2059 if (chain == NULL) { /* Loop ran zero times. */
2060 chain = reg_node(NOTHING);
2065 *flagp |= flags&SIMPLE;
2072 - regpiece - something followed by possible [*+?]
2074 * Note that the branching code sequences used for ? and the general cases
2075 * of * and + are somewhat optimized: they use the same NOTHING node as
2076 * both the endmarker for their branch list and the body of the last branch.
2077 * It might seem that this node could be dispensed with entirely, but the
2078 * endmarker role is not redundant.
2081 S_regpiece(pTHX_ I32 *flagp)
2084 register regnode *ret;
2086 register char *next;
2088 char *origparse = PL_regcomp_parse;
2091 I32 max = REG_INFTY;
2093 ret = regatom(&flags);
2095 if (flags & TRYAGAIN)
2100 op = *PL_regcomp_parse;
2102 if (op == '{' && regcurly(PL_regcomp_parse)) {
2103 next = PL_regcomp_parse + 1;
2105 while (isDIGIT(*next) || *next == ',') {
2114 if (*next == '}') { /* got one */
2118 min = atoi(PL_regcomp_parse);
2122 maxpos = PL_regcomp_parse;
2124 if (!max && *maxpos != '0')
2125 max = REG_INFTY; /* meaning "infinity" */
2126 else if (max >= REG_INFTY)
2127 FAIL2("Quantifier in {,} bigger than %d", REG_INFTY - 1);
2128 PL_regcomp_parse = next;
2132 if ((flags&SIMPLE)) {
2133 PL_regnaughty += 2 + PL_regnaughty / 2;
2134 reginsert(CURLY, ret);
2137 regnode *w = reg_node(WHILEM);
2141 if (!SIZE_ONLY && PL_extralen) {
2142 reginsert(LONGJMP,ret);
2143 reginsert(NOTHING,ret);
2144 NEXT_OFF(ret) = 3; /* Go over LONGJMP. */
2146 reginsert(CURLYX,ret);
2147 if (!SIZE_ONLY && PL_extralen)
2148 NEXT_OFF(ret) = 3; /* Go over NOTHING to LONGJMP. */
2149 regtail(ret, reg_node(NOTHING));
2151 PL_reg_whilem_seen++, PL_extralen += 3;
2152 PL_regnaughty += 4 + PL_regnaughty; /* compound interest */
2160 if (max && max < min)
2161 FAIL("Can't do {n,m} with n > m");
2176 #if 0 /* Now runtime fix should be reliable. */
2177 if (!(flags&HASWIDTH) && op != '?')
2178 FAIL("regexp *+ operand could be empty");
2183 *flagp = (op != '+') ? (WORST|SPSTART|HASWIDTH) : (WORST|HASWIDTH);
2185 if (op == '*' && (flags&SIMPLE)) {
2186 reginsert(STAR, ret);
2190 else if (op == '*') {
2194 else if (op == '+' && (flags&SIMPLE)) {
2195 reginsert(PLUS, ret);
2199 else if (op == '+') {
2203 else if (op == '?') {
2208 if (ckWARN(WARN_REGEXP) && !SIZE_ONLY && !(flags&HASWIDTH) && max > REG_INFTY/3) {
2209 Perl_warner(aTHX_ WARN_REGEXP, "%.*s matches null string many times",
2210 PL_regcomp_parse - origparse, origparse);
2213 if (*PL_regcomp_parse == '?') {
2215 reginsert(MINMOD, ret);
2216 regtail(ret, ret + NODE_STEP_REGNODE);
2218 if (ISMULT2(PL_regcomp_parse))
2219 FAIL("nested *?+ in regexp");
2225 - regatom - the lowest level
2227 * Optimization: gobbles an entire sequence of ordinary characters so that
2228 * it can turn them into a single node, which is smaller to store and
2229 * faster to run. Backslashed characters are exceptions, each becoming a
2230 * separate node; the code is simpler that way and it's not worth fixing.
2232 * [Yes, it is worth fixing, some scripts can run twice the speed.]
2235 S_regatom(pTHX_ I32 *flagp)
2238 register regnode *ret = 0;
2241 *flagp = WORST; /* Tentatively. */
2244 switch (*PL_regcomp_parse) {
2248 if (PL_regflags & PMf_MULTILINE)
2249 ret = reg_node(MBOL);
2250 else if (PL_regflags & PMf_SINGLELINE)
2251 ret = reg_node(SBOL);
2253 ret = reg_node(BOL);
2256 if (PL_regcomp_parse[1])
2259 if (PL_regflags & PMf_MULTILINE)
2260 ret = reg_node(MEOL);
2261 else if (PL_regflags & PMf_SINGLELINE)
2262 ret = reg_node(SEOL);
2264 ret = reg_node(EOL);
2269 if (PL_regflags & PMf_SINGLELINE)
2270 ret = reg_node(SANYUTF8);
2272 ret = reg_node(ANYUTF8);
2276 if (PL_regflags & PMf_SINGLELINE)
2277 ret = reg_node(SANY);
2279 ret = reg_node(REG_ANY);
2280 *flagp |= HASWIDTH|SIMPLE;
2286 ret = (UTF ? regclassutf8() : regclass());
2287 if (*PL_regcomp_parse != ']')
2288 FAIL("unmatched [] in regexp");
2290 *flagp |= HASWIDTH|SIMPLE;
2294 ret = reg(1, &flags);
2296 if (flags & TRYAGAIN)
2300 *flagp |= flags&(HASWIDTH|SPSTART|SIMPLE);
2304 if (flags & TRYAGAIN) {
2308 FAIL2("internal urp in regexp at /%s/", PL_regcomp_parse);
2309 /* Supposed to be caught earlier. */
2312 if (!regcurly(PL_regcomp_parse)) {
2320 FAIL("?+*{} follows nothing in regexp");
2323 switch (*++PL_regcomp_parse) {
2326 ret = reg_node(SBOL);
2331 ret = reg_node(GPOS);
2332 PL_regseen |= REG_SEEN_GPOS;
2337 ret = reg_node(SEOL);
2342 ret = reg_node(EOS);
2344 PL_seen_zerolen++; /* Do not optimize RE away */
2348 ret = reg_node(SANY);
2349 *flagp |= HASWIDTH|SIMPLE;
2353 ret = reg_node(CLUMP);
2356 if (UTF && !PL_utf8_mark)
2357 is_utf8_mark((U8*)"~"); /* preload table */
2362 ? (LOC ? ALNUMLUTF8 : ALNUMUTF8)
2363 : (LOC ? ALNUML : ALNUM));
2364 *flagp |= HASWIDTH|SIMPLE;
2366 if (UTF && !PL_utf8_alnum)
2367 is_utf8_alnum((U8*)"a"); /* preload table */
2372 ? (LOC ? NALNUMLUTF8 : NALNUMUTF8)
2373 : (LOC ? NALNUML : NALNUM));
2374 *flagp |= HASWIDTH|SIMPLE;
2376 if (UTF && !PL_utf8_alnum)
2377 is_utf8_alnum((U8*)"a"); /* preload table */
2381 PL_regseen |= REG_SEEN_LOOKBEHIND;
2384 ? (LOC ? BOUNDLUTF8 : BOUNDUTF8)
2385 : (LOC ? BOUNDL : BOUND));
2388 if (UTF && !PL_utf8_alnum)
2389 is_utf8_alnum((U8*)"a"); /* preload table */
2393 PL_regseen |= REG_SEEN_LOOKBEHIND;
2396 ? (LOC ? NBOUNDLUTF8 : NBOUNDUTF8)
2397 : (LOC ? NBOUNDL : NBOUND));
2400 if (UTF && !PL_utf8_alnum)
2401 is_utf8_alnum((U8*)"a"); /* preload table */
2406 ? (LOC ? SPACELUTF8 : SPACEUTF8)
2407 : (LOC ? SPACEL : SPACE));
2408 *flagp |= HASWIDTH|SIMPLE;
2410 if (UTF && !PL_utf8_space)
2411 is_utf8_space((U8*)" "); /* preload table */
2416 ? (LOC ? NSPACELUTF8 : NSPACEUTF8)
2417 : (LOC ? NSPACEL : NSPACE));
2418 *flagp |= HASWIDTH|SIMPLE;
2420 if (UTF && !PL_utf8_space)
2421 is_utf8_space((U8*)" "); /* preload table */
2424 ret = reg_node(UTF ? DIGITUTF8 : DIGIT);
2425 *flagp |= HASWIDTH|SIMPLE;
2427 if (UTF && !PL_utf8_digit)
2428 is_utf8_digit((U8*)"1"); /* preload table */
2431 ret = reg_node(UTF ? NDIGITUTF8 : NDIGIT);
2432 *flagp |= HASWIDTH|SIMPLE;
2434 if (UTF && !PL_utf8_digit)
2435 is_utf8_digit((U8*)"1"); /* preload table */
2439 { /* a lovely hack--pretend we saw [\pX] instead */
2440 char* oldregxend = PL_regxend;
2442 if (PL_regcomp_parse[1] == '{') {
2443 PL_regxend = strchr(PL_regcomp_parse, '}');
2445 FAIL("Missing right brace on \\p{}");
2449 PL_regxend = PL_regcomp_parse + 2;
2452 ret = regclassutf8();
2454 PL_regxend = oldregxend;
2457 *flagp |= HASWIDTH|SIMPLE;
2470 case '1': case '2': case '3': case '4':
2471 case '5': case '6': case '7': case '8': case '9':
2473 I32 num = atoi(PL_regcomp_parse);
2475 if (num > 9 && num >= PL_regnpar)
2478 if (!SIZE_ONLY && num > PL_regcomp_rx->nparens)
2479 FAIL("reference to nonexistent group");
2482 ? (LOC ? REFFL : REFF)
2485 while (isDIGIT(*PL_regcomp_parse))
2493 if (PL_regcomp_parse >= PL_regxend)
2494 FAIL("trailing \\ in regexp");
2497 /* Do not generate `unrecognized' warnings here, we fall
2498 back into the quick-grab loop below */
2504 if (PL_regflags & PMf_EXTENDED) {
2505 while (PL_regcomp_parse < PL_regxend && *PL_regcomp_parse != '\n') PL_regcomp_parse++;
2506 if (PL_regcomp_parse < PL_regxend)
2522 ? (LOC ? EXACTFL : EXACTF)
2525 for (len = 0, p = PL_regcomp_parse - 1;
2526 len < 127 && p < PL_regxend;
2531 if (PL_regflags & PMf_EXTENDED)
2532 p = regwhite(p, PL_regxend);
2594 char* e = strchr(p, '}');
2597 FAIL("Missing right brace on \\x{}");
2599 ender = (UV)scan_hex(p + 1, e - p, &numlen);
2600 if (numlen + len >= 127) { /* numlen is generous */
2607 FAIL("Can't use \\x{} without 'use utf8' declaration");
2610 ender = (UV)scan_hex(p, 2, &numlen);
2616 ender = UCHARAT(p++);
2617 ender = toCTRL(ender);
2619 case '0': case '1': case '2': case '3':case '4':
2620 case '5': case '6': case '7': case '8':case '9':
2622 (isDIGIT(p[1]) && atoi(p) >= PL_regnpar) ) {
2623 ender = (UV)scan_oct(p, 3, &numlen);
2632 if (p >= PL_regxend)
2633 FAIL("trailing \\ in regexp");
2636 if (!SIZE_ONLY && ckWARN(WARN_REGEXP) && isALPHA(*p))
2637 Perl_warner(aTHX_ WARN_REGEXP,
2638 "/%.127s/: Unrecognized escape \\%c passed through",
2641 goto normal_default;
2646 if ((*p & 0xc0) == 0xc0 && UTF) {
2647 ender = utf8_to_uv((U8*)p, &numlen);
2654 if (PL_regflags & PMf_EXTENDED)
2655 p = regwhite(p, PL_regxend);
2658 ender = toLOWER_LC_uni(ender);
2660 ender = toLOWER_uni(ender);
2662 if (ISMULT2(p)) { /* Back off on ?+*. */
2665 else if (ender >= 0x80 && UTF) {
2666 reguni(ender, s, &numlen);
2676 if (ender >= 0x80 && UTF) {
2677 reguni(ender, s, &numlen);
2685 PL_regcomp_parse = p - 1;
2688 FAIL("internal disaster in regexp");
2696 PL_regsize += STR_SZ(len);
2698 PL_regcode += STR_SZ(len);
2707 S_regwhite(pTHX_ char *p, char *e)
2712 else if (*p == '#') {
2715 } while (p < e && *p != '\n');
2723 /* Parse POSIX character classes: [[:foo:]], [[=foo=]], [[.foo.]].
2724 Character classes ([:foo:]) can also be negated ([:^foo:]).
2725 Returns a named class id (ANYOF_XXX) if successful, -1 otherwise.
2726 Equivalence classes ([=foo=]) and composites ([.foo.]) are parsed,
2727 but trigger warnings because they are currently unimplemented. */
2729 S_regpposixcc(pTHX_ I32 value)
2733 I32 namedclass = OOB_NAMEDCLASS;
2735 if (value == '[' && PL_regcomp_parse + 1 < PL_regxend &&
2736 /* I smell either [: or [= or [. -- POSIX has been here, right? */
2737 (*PL_regcomp_parse == ':' ||
2738 *PL_regcomp_parse == '=' ||
2739 *PL_regcomp_parse == '.')) {
2740 char c = *PL_regcomp_parse;
2741 char* s = PL_regcomp_parse++;
2743 while (PL_regcomp_parse < PL_regxend && *PL_regcomp_parse != c)
2745 if (PL_regcomp_parse == PL_regxend)
2746 /* Grandfather lone [:, [=, [. */
2747 PL_regcomp_parse = s;
2749 char* t = PL_regcomp_parse++; /* skip over the c */
2751 if (*PL_regcomp_parse == ']') {
2752 PL_regcomp_parse++; /* skip over the ending ] */
2755 I32 complement = *posixcc == '^' ? *posixcc++ : 0;
2756 I32 skip = 5; /* the most common skip */
2760 if (strnEQ(posixcc, "alnum", 5))
2762 complement ? ANYOF_NALNUMC : ANYOF_ALNUMC;
2763 else if (strnEQ(posixcc, "alpha", 5))
2765 complement ? ANYOF_NALPHA : ANYOF_ALPHA;
2766 else if (strnEQ(posixcc, "ascii", 5))
2768 complement ? ANYOF_NASCII : ANYOF_ASCII;
2771 if (strnEQ(posixcc, "cntrl", 5))
2773 complement ? ANYOF_NCNTRL : ANYOF_CNTRL;
2776 if (strnEQ(posixcc, "digit", 5))
2778 complement ? ANYOF_NDIGIT : ANYOF_DIGIT;
2781 if (strnEQ(posixcc, "graph", 5))
2783 complement ? ANYOF_NGRAPH : ANYOF_GRAPH;
2786 if (strnEQ(posixcc, "lower", 5))
2788 complement ? ANYOF_NLOWER : ANYOF_LOWER;
2791 if (strnEQ(posixcc, "print", 5))
2793 complement ? ANYOF_NPRINT : ANYOF_PRINT;
2794 else if (strnEQ(posixcc, "punct", 5))
2796 complement ? ANYOF_NPUNCT : ANYOF_PUNCT;
2799 if (strnEQ(posixcc, "space", 5))
2801 complement ? ANYOF_NSPACE : ANYOF_SPACE;
2803 if (strnEQ(posixcc, "upper", 5))
2805 complement ? ANYOF_NUPPER : ANYOF_UPPER;
2807 case 'w': /* this is not POSIX, this is the Perl \w */
2808 if (strnEQ(posixcc, "word", 4)) {
2810 complement ? ANYOF_NALNUM : ANYOF_ALNUM;
2815 if (strnEQ(posixcc, "xdigit", 6)) {
2817 complement ? ANYOF_NXDIGIT : ANYOF_XDIGIT;
2822 if (namedclass == OOB_NAMEDCLASS ||
2823 posixcc[skip] != ':' ||
2824 posixcc[skip+1] != ']')
2826 "Character class [:%.*s:] unknown",
2828 } else if (ckWARN(WARN_REGEXP) && !SIZE_ONLY)
2829 /* [[=foo=]] and [[.foo.]] are still future. */
2830 Perl_warner(aTHX_ WARN_REGEXP,
2831 "Character class syntax [%c %c] is reserved for future extensions", c, c);
2833 /* Maternal grandfather:
2834 * "[:" ending in ":" but not in ":]" */
2835 PL_regcomp_parse = s;
2844 S_checkposixcc(pTHX)
2846 if (!SIZE_ONLY && ckWARN(WARN_REGEXP) &&
2847 (*PL_regcomp_parse == ':' ||
2848 *PL_regcomp_parse == '=' ||
2849 *PL_regcomp_parse == '.')) {
2850 char *s = PL_regcomp_parse;
2853 while(*s && isALNUM(*s))
2855 if (*s && c == *s && s[1] == ']') {
2856 Perl_warner(aTHX_ WARN_REGEXP,
2857 "Character class syntax [%c %c] belongs inside character classes", c, c);
2858 if (c == '=' || c == '.')
2859 Perl_warner(aTHX_ WARN_REGEXP,
2860 "Character class syntax [%c %c] is reserved for future extensions", c, c);
2870 register I32 lastvalue = OOB_CHAR8;
2871 register I32 range = 0;
2872 register regnode *ret;
2876 bool need_class = 0;
2878 ret = reg_node(ANYOF);
2880 PL_regsize += ANYOF_SKIP;
2883 ANYOF_BITMAP_ZERO(ret);
2884 PL_regcode += ANYOF_SKIP;
2886 ANYOF_FLAGS(ret) |= ANYOF_FOLD;
2888 ANYOF_FLAGS(ret) |= ANYOF_LOCALE;
2890 if (*PL_regcomp_parse == '^') { /* Complement of range. */
2894 ANYOF_FLAGS(ret) |= ANYOF_INVERT;
2897 if (!SIZE_ONLY && ckWARN(WARN_REGEXP))
2900 if (*PL_regcomp_parse == ']' || *PL_regcomp_parse == '-')
2901 goto skipcond; /* allow 1st char to be ] or - */
2902 while (PL_regcomp_parse < PL_regxend && *PL_regcomp_parse != ']') {
2904 namedclass = OOB_NAMEDCLASS;
2906 rangebegin = PL_regcomp_parse;
2907 value = UCHARAT(PL_regcomp_parse++);
2909 namedclass = regpposixcc(value);
2910 else if (value == '\\') {
2911 value = UCHARAT(PL_regcomp_parse++);
2912 /* Some compilers cannot handle switching on 64-bit integer
2913 * values, therefore value cannot be an UV. --jhi */
2915 case 'w': namedclass = ANYOF_ALNUM; break;
2916 case 'W': namedclass = ANYOF_NALNUM; break;
2917 case 's': namedclass = ANYOF_SPACE; break;
2918 case 'S': namedclass = ANYOF_NSPACE; break;
2919 case 'd': namedclass = ANYOF_DIGIT; break;
2920 case 'D': namedclass = ANYOF_NDIGIT; break;
2921 case 'n': value = '\n'; break;
2922 case 'r': value = '\r'; break;
2923 case 't': value = '\t'; break;
2924 case 'f': value = '\f'; break;
2925 case 'b': value = '\b'; break;
2927 case 'e': value = '\033'; break;
2928 case 'a': value = '\007'; break;
2930 case 'e': value = '\047'; break;
2931 case 'a': value = '\057'; break;
2934 value = (UV)scan_hex(PL_regcomp_parse, 2, &numlen);
2935 PL_regcomp_parse += numlen;
2938 value = UCHARAT(PL_regcomp_parse++);
2939 value = toCTRL(value);
2941 case '0': case '1': case '2': case '3': case '4':
2942 case '5': case '6': case '7': case '8': case '9':
2943 value = (UV)scan_oct(--PL_regcomp_parse, 3, &numlen);
2944 PL_regcomp_parse += numlen;
2947 if (!SIZE_ONLY && ckWARN(WARN_REGEXP) && isALPHA(value))
2948 Perl_warner(aTHX_ WARN_REGEXP,
2949 "/%.127s/: Unrecognized escape \\%c in character class passed through",
2955 if (namedclass > OOB_NAMEDCLASS) {
2956 if (!need_class && !SIZE_ONLY)
2957 ANYOF_CLASS_ZERO(ret);
2959 if (range) { /* a-\d, a-[:digit:] */
2961 if (ckWARN(WARN_REGEXP))
2962 Perl_warner(aTHX_ WARN_REGEXP,
2963 "/%.127s/: false [] range \"%*.*s\" in regexp",
2965 PL_regcomp_parse - rangebegin,
2966 PL_regcomp_parse - rangebegin,
2968 ANYOF_BITMAP_SET(ret, lastvalue);
2969 ANYOF_BITMAP_SET(ret, '-');
2971 range = 0; /* this is not a true range */
2974 switch (namedclass) {
2977 ANYOF_CLASS_SET(ret, ANYOF_ALNUM);
2979 for (value = 0; value < 256; value++)
2981 ANYOF_BITMAP_SET(ret, value);
2986 ANYOF_CLASS_SET(ret, ANYOF_NALNUM);
2988 for (value = 0; value < 256; value++)
2989 if (!isALNUM(value))
2990 ANYOF_BITMAP_SET(ret, value);
2995 ANYOF_CLASS_SET(ret, ANYOF_SPACE);
2997 for (value = 0; value < 256; value++)
2999 ANYOF_BITMAP_SET(ret, value);
3004 ANYOF_CLASS_SET(ret, ANYOF_NSPACE);
3006 for (value = 0; value < 256; value++)
3007 if (!isSPACE(value))
3008 ANYOF_BITMAP_SET(ret, value);
3013 ANYOF_CLASS_SET(ret, ANYOF_DIGIT);
3015 for (value = '0'; value <= '9'; value++)
3016 ANYOF_BITMAP_SET(ret, value);
3021 ANYOF_CLASS_SET(ret, ANYOF_NDIGIT);
3023 for (value = 0; value < '0'; value++)
3024 ANYOF_BITMAP_SET(ret, value);
3025 for (value = '9' + 1; value < 256; value++)
3026 ANYOF_BITMAP_SET(ret, value);
3031 ANYOF_CLASS_SET(ret, ANYOF_NALNUMC);
3033 for (value = 0; value < 256; value++)
3034 if (!isALNUMC(value))
3035 ANYOF_BITMAP_SET(ret, value);
3040 ANYOF_CLASS_SET(ret, ANYOF_ALNUMC);
3042 for (value = 0; value < 256; value++)
3043 if (isALNUMC(value))
3044 ANYOF_BITMAP_SET(ret, value);
3049 ANYOF_CLASS_SET(ret, ANYOF_ALPHA);
3051 for (value = 0; value < 256; value++)
3053 ANYOF_BITMAP_SET(ret, value);
3058 ANYOF_CLASS_SET(ret, ANYOF_NALPHA);
3060 for (value = 0; value < 256; value++)
3061 if (!isALPHA(value))
3062 ANYOF_BITMAP_SET(ret, value);
3067 ANYOF_CLASS_SET(ret, ANYOF_ASCII);
3070 for (value = 0; value < 128; value++)
3071 ANYOF_BITMAP_SET(ret, value);
3073 for (value = 0; value < 256; value++)
3075 ANYOF_BITMAP_SET(ret, value);
3081 ANYOF_CLASS_SET(ret, ANYOF_NASCII);
3084 for (value = 128; value < 256; value++)
3085 ANYOF_BITMAP_SET(ret, value);
3087 for (value = 0; value < 256; value++)
3088 if (!isASCII(value))
3089 ANYOF_BITMAP_SET(ret, value);
3095 ANYOF_CLASS_SET(ret, ANYOF_CNTRL);
3097 for (value = 0; value < 256; value++)
3099 ANYOF_BITMAP_SET(ret, value);
3101 lastvalue = OOB_CHAR8;
3105 ANYOF_CLASS_SET(ret, ANYOF_NCNTRL);
3107 for (value = 0; value < 256; value++)
3108 if (!isCNTRL(value))
3109 ANYOF_BITMAP_SET(ret, value);
3114 ANYOF_CLASS_SET(ret, ANYOF_GRAPH);
3116 for (value = 0; value < 256; value++)
3118 ANYOF_BITMAP_SET(ret, value);
3123 ANYOF_CLASS_SET(ret, ANYOF_NGRAPH);
3125 for (value = 0; value < 256; value++)
3126 if (!isGRAPH(value))
3127 ANYOF_BITMAP_SET(ret, value);
3132 ANYOF_CLASS_SET(ret, ANYOF_LOWER);
3134 for (value = 0; value < 256; value++)
3136 ANYOF_BITMAP_SET(ret, value);
3141 ANYOF_CLASS_SET(ret, ANYOF_NLOWER);
3143 for (value = 0; value < 256; value++)
3144 if (!isLOWER(value))
3145 ANYOF_BITMAP_SET(ret, value);
3150 ANYOF_CLASS_SET(ret, ANYOF_PRINT);
3152 for (value = 0; value < 256; value++)
3154 ANYOF_BITMAP_SET(ret, value);
3159 ANYOF_CLASS_SET(ret, ANYOF_NPRINT);
3161 for (value = 0; value < 256; value++)
3162 if (!isPRINT(value))
3163 ANYOF_BITMAP_SET(ret, value);
3168 ANYOF_CLASS_SET(ret, ANYOF_PUNCT);
3170 for (value = 0; value < 256; value++)
3172 ANYOF_BITMAP_SET(ret, value);
3177 ANYOF_CLASS_SET(ret, ANYOF_NPUNCT);
3179 for (value = 0; value < 256; value++)
3180 if (!isPUNCT(value))
3181 ANYOF_BITMAP_SET(ret, value);
3186 ANYOF_CLASS_SET(ret, ANYOF_UPPER);
3188 for (value = 0; value < 256; value++)
3190 ANYOF_BITMAP_SET(ret, value);
3195 ANYOF_CLASS_SET(ret, ANYOF_NUPPER);
3197 for (value = 0; value < 256; value++)
3198 if (!isUPPER(value))
3199 ANYOF_BITMAP_SET(ret, value);
3204 ANYOF_CLASS_SET(ret, ANYOF_XDIGIT);
3206 for (value = 0; value < 256; value++)
3207 if (isXDIGIT(value))
3208 ANYOF_BITMAP_SET(ret, value);
3213 ANYOF_CLASS_SET(ret, ANYOF_NXDIGIT);
3215 for (value = 0; value < 256; value++)
3216 if (!isXDIGIT(value))
3217 ANYOF_BITMAP_SET(ret, value);
3221 FAIL("invalid [::] class in regexp");
3225 ANYOF_FLAGS(ret) |= ANYOF_CLASS;
3230 if (lastvalue > value) /* b-a */ {
3232 "/%.127s/: invalid [] range \"%*.*s\" in regexp",
3234 PL_regcomp_parse - rangebegin,
3235 PL_regcomp_parse - rangebegin,
3242 if (*PL_regcomp_parse == '-' && PL_regcomp_parse+1 < PL_regxend &&
3243 PL_regcomp_parse[1] != ']') {
3245 if (namedclass > OOB_NAMEDCLASS) { /* \w-, [:word:]- */
3246 if (ckWARN(WARN_REGEXP))
3247 Perl_warner(aTHX_ WARN_REGEXP,
3248 "/%.127s/: false [] range \"%*.*s\" in regexp",
3250 PL_regcomp_parse - rangebegin,
3251 PL_regcomp_parse - rangebegin,
3254 ANYOF_BITMAP_SET(ret, '-');
3257 continue; /* do it next time */
3260 /* now is the next time */
3262 #ifndef ASCIIish /* EBCDIC, for example. */
3263 if ((isLOWER(lastvalue) && isLOWER(value)) ||
3264 (isUPPER(lastvalue) && isUPPER(value)))
3267 if (isLOWER(lastvalue)) {
3268 for (i = lastvalue; i <= value; i++)
3270 ANYOF_BITMAP_SET(ret, i);
3272 for (i = lastvalue; i <= value; i++)
3274 ANYOF_BITMAP_SET(ret, i);
3279 for ( ; lastvalue <= value; lastvalue++)
3280 ANYOF_BITMAP_SET(ret, lastvalue);
3286 PL_regsize += ANYOF_CLASS_ADD_SKIP;
3288 PL_regcode += ANYOF_CLASS_ADD_SKIP;
3290 /* optimize case-insensitive simple patterns (e.g. /[a-z]/i) */
3292 (ANYOF_FLAGS(ret) & (ANYOF_FLAGS_ALL ^ ANYOF_INVERT)) == ANYOF_FOLD) {
3293 for (value = 0; value < 256; ++value) {
3294 if (ANYOF_BITMAP_TEST(ret, value)) {
3295 I32 cf = PL_fold[value];
3296 ANYOF_BITMAP_SET(ret, cf);
3299 ANYOF_FLAGS(ret) &= ~ANYOF_FOLD;
3301 /* optimize inverted simple patterns (e.g. [^a-z]) */
3302 if (!SIZE_ONLY && (ANYOF_FLAGS(ret) & ANYOF_FLAGS_ALL) == ANYOF_INVERT) {
3303 for (value = 0; value < ANYOF_BITMAP_SIZE; ++value)
3304 ANYOF_BITMAP(ret)[value] ^= ANYOF_FLAGS_ALL;
3305 ANYOF_FLAGS(ret) = 0;
3311 S_regclassutf8(pTHX)
3316 register U32 lastvalue = OOB_UTF8;
3317 register I32 range = 0;
3318 register regnode *ret;
3326 if (*PL_regcomp_parse == '^') { /* Complement of range. */
3330 flags |= ANYOF_INVERT;
3334 flags |= ANYOF_FOLD;
3336 flags |= ANYOF_LOCALE;
3337 listsv = newSVpvn("# comment\n",10);
3340 if (!SIZE_ONLY && ckWARN(WARN_REGEXP))
3343 if (*PL_regcomp_parse == ']' || *PL_regcomp_parse == '-')
3344 goto skipcond; /* allow 1st char to be ] or - */
3346 while (PL_regcomp_parse < PL_regxend && *PL_regcomp_parse != ']') {
3348 namedclass = OOB_NAMEDCLASS;
3350 rangebegin = PL_regcomp_parse;
3351 value = utf8_to_uv((U8*)PL_regcomp_parse, &numlen);
3352 PL_regcomp_parse += numlen;
3354 namedclass = regpposixcc(value);
3355 else if (value == '\\') {
3356 value = (U32)utf8_to_uv((U8*)PL_regcomp_parse, &numlen);
3357 PL_regcomp_parse += numlen;
3358 /* Some compilers cannot handle switching on 64-bit integer
3359 * values, therefore value cannot be an UV. Yes, this will
3360 * be a problem later if we want switch on Unicode. --jhi */
3362 case 'w': namedclass = ANYOF_ALNUM; break;
3363 case 'W': namedclass = ANYOF_NALNUM; break;
3364 case 's': namedclass = ANYOF_SPACE; break;
3365 case 'S': namedclass = ANYOF_NSPACE; break;
3366 case 'd': namedclass = ANYOF_DIGIT; break;
3367 case 'D': namedclass = ANYOF_NDIGIT; break;
3370 if (*PL_regcomp_parse == '{') {
3371 e = strchr(PL_regcomp_parse++, '}');
3373 FAIL("Missing right brace on \\p{}");
3374 n = e - PL_regcomp_parse;
3377 e = PL_regcomp_parse;
3382 Perl_sv_catpvf(aTHX_ listsv,
3383 "+utf8::%.*s\n", (int)n, PL_regcomp_parse);
3385 Perl_sv_catpvf(aTHX_ listsv,
3386 "!utf8::%.*s\n", (int)n, PL_regcomp_parse);
3388 PL_regcomp_parse = e + 1;
3389 lastvalue = OOB_UTF8;
3391 case 'n': value = '\n'; break;
3392 case 'r': value = '\r'; break;
3393 case 't': value = '\t'; break;
3394 case 'f': value = '\f'; break;
3395 case 'b': value = '\b'; break;
3397 case 'e': value = '\033'; break;
3398 case 'a': value = '\007'; break;
3400 case 'e': value = '\047'; break;
3401 case 'a': value = '\057'; break;
3404 if (*PL_regcomp_parse == '{') {
3405 e = strchr(PL_regcomp_parse++, '}');
3407 FAIL("Missing right brace on \\x{}");
3408 value = (UV)scan_hex(PL_regcomp_parse,
3409 e - PL_regcomp_parse,
3411 PL_regcomp_parse = e + 1;
3414 value = (UV)scan_hex(PL_regcomp_parse, 2, &numlen);
3415 PL_regcomp_parse += numlen;
3419 value = UCHARAT(PL_regcomp_parse++);
3420 value = toCTRL(value);
3422 case '0': case '1': case '2': case '3': case '4':
3423 case '5': case '6': case '7': case '8': case '9':
3424 value = (UV)scan_oct(--PL_regcomp_parse, 3, &numlen);
3425 PL_regcomp_parse += numlen;
3428 if (!SIZE_ONLY && ckWARN(WARN_REGEXP) && isALPHA(value))
3429 Perl_warner(aTHX_ WARN_REGEXP,
3430 "/%.127s/: Unrecognized escape \\%c in character class passed through",
3436 if (namedclass > OOB_NAMEDCLASS) {
3437 if (range) { /* a-\d, a-[:digit:] */
3439 if (ckWARN(WARN_REGEXP))
3440 Perl_warner(aTHX_ WARN_REGEXP,
3441 "/%.127s/: false [] range \"%*.*s\" in regexp",
3443 PL_regcomp_parse - rangebegin,
3444 PL_regcomp_parse - rangebegin,
3446 Perl_sv_catpvf(aTHX_ listsv,
3447 /* 0x002D is Unicode for '-' */
3448 "%04"UVxf"\n002D\n", (UV)lastvalue);
3453 switch (namedclass) {
3455 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsWord\n"); break;
3457 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsWord\n"); break;
3459 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsAlnum\n"); break;
3461 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsAlnum\n"); break;
3463 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsAlpha\n"); break;
3465 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsAlpha\n"); break;
3467 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsASCII\n"); break;
3469 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsASCII\n"); break;
3471 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsCntrl\n"); break;
3473 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsCntrl\n"); break;
3475 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsGraph\n"); break;
3477 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsGraph\n"); break;
3479 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsDigit\n"); break;
3481 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsDigit\n"); break;
3483 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsLower\n"); break;
3485 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsLower\n"); break;
3487 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsPrint\n"); break;
3489 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsPrint\n"); break;
3491 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsPunct\n"); break;
3493 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsPunct\n"); break;
3495 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsSpace\n"); break;
3497 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsSpace\n"); break;
3499 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsUpper\n"); break;
3501 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsUpper\n"); break;
3503 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsXDigit\n"); break;
3505 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsXDigit\n"); break;
3511 if (lastvalue > value) { /* b-a */
3513 "/%.127s/: invalid [] range \"%*.*s\" in regexp",
3515 PL_regcomp_parse - rangebegin,
3516 PL_regcomp_parse - rangebegin,
3523 if (*PL_regcomp_parse == '-' && PL_regcomp_parse+1 < PL_regxend &&
3524 PL_regcomp_parse[1] != ']') {
3526 if (namedclass > OOB_NAMEDCLASS) { /* \w-, [:word:]- */
3527 if (ckWARN(WARN_REGEXP))
3528 Perl_warner(aTHX_ WARN_REGEXP,
3529 "/%.127s/: false [] range \"%*.*s\" in regexp",
3531 PL_regcomp_parse - rangebegin,
3532 PL_regcomp_parse - rangebegin,
3535 Perl_sv_catpvf(aTHX_ listsv,
3536 /* 0x002D is Unicode for '-' */
3540 continue; /* do it next time */
3543 /* now is the next time */
3545 Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\t%04"UVxf"\n",
3546 (UV)lastvalue, (UV)value);
3550 ret = reganode(ANYOFUTF8, 0);
3553 SV *rv = swash_init("utf8", "", listsv, 1, 0);
3554 SvREFCNT_dec(listsv);
3555 n = add_data(1,"s");
3556 PL_regcomp_rx->data->data[n] = (void*)rv;
3557 ARG1_SET(ret, flags);
3568 char* retval = PL_regcomp_parse++;
3571 if (*PL_regcomp_parse == '(' && PL_regcomp_parse[1] == '?' &&
3572 PL_regcomp_parse[2] == '#') {
3573 while (*PL_regcomp_parse && *PL_regcomp_parse != ')')
3578 if (PL_regflags & PMf_EXTENDED) {
3579 if (isSPACE(*PL_regcomp_parse)) {
3583 else if (*PL_regcomp_parse == '#') {
3584 while (*PL_regcomp_parse && *PL_regcomp_parse != '\n')
3595 - reg_node - emit a node
3597 STATIC regnode * /* Location. */
3598 S_reg_node(pTHX_ U8 op)
3601 register regnode *ret;
3602 register regnode *ptr;
3606 SIZE_ALIGN(PL_regsize);
3611 NODE_ALIGN_FILL(ret);
3613 FILL_ADVANCE_NODE(ptr, op);
3620 - reganode - emit a node with an argument
3622 STATIC regnode * /* Location. */
3623 S_reganode(pTHX_ U8 op, U32 arg)
3626 register regnode *ret;
3627 register regnode *ptr;
3631 SIZE_ALIGN(PL_regsize);
3636 NODE_ALIGN_FILL(ret);
3638 FILL_ADVANCE_NODE_ARG(ptr, op, arg);
3645 - reguni - emit (if appropriate) a Unicode character
3648 S_reguni(pTHX_ UV uv, char* s, I32* lenp)
3652 U8 tmpbuf[UTF8_MAXLEN];
3653 *lenp = uv_to_utf8(tmpbuf, uv) - tmpbuf;
3656 *lenp = uv_to_utf8((U8*)s, uv) - (U8*)s;
3661 - reginsert - insert an operator in front of already-emitted operand
3663 * Means relocating the operand.
3666 S_reginsert(pTHX_ U8 op, regnode *opnd)
3669 register regnode *src;
3670 register regnode *dst;
3671 register regnode *place;
3672 register int offset = regarglen[(U8)op];
3674 /* (PL_regkind[(U8)op] == CURLY ? EXTRA_STEP_2ARGS : 0); */
3677 PL_regsize += NODE_STEP_REGNODE + offset;
3682 PL_regcode += NODE_STEP_REGNODE + offset;
3685 StructCopy(--src, --dst, regnode);
3687 place = opnd; /* Op node, where operand used to be. */
3688 src = NEXTOPER(place);
3689 FILL_ADVANCE_NODE(place, op);
3690 Zero(src, offset, regnode);
3694 - regtail - set the next-pointer at the end of a node chain of p to val.
3697 S_regtail(pTHX_ regnode *p, regnode *val)
3700 register regnode *scan;
3701 register regnode *temp;
3706 /* Find last node. */
3709 temp = regnext(scan);
3715 if (reg_off_by_arg[OP(scan)]) {
3716 ARG_SET(scan, val - scan);
3719 NEXT_OFF(scan) = val - scan;
3724 - regoptail - regtail on operand of first argument; nop if operandless
3727 S_regoptail(pTHX_ regnode *p, regnode *val)
3730 /* "Operandless" and "op != BRANCH" are synonymous in practice. */
3731 if (p == NULL || SIZE_ONLY)
3733 if (PL_regkind[(U8)OP(p)] == BRANCH) {
3734 regtail(NEXTOPER(p), val);
3736 else if ( PL_regkind[(U8)OP(p)] == BRANCHJ) {
3737 regtail(NEXTOPER(NEXTOPER(p)), val);
3744 - regcurly - a little FSA that accepts {\d+,?\d*}
3747 S_regcurly(pTHX_ register char *s)
3766 S_dumpuntil(pTHX_ regnode *start, regnode *node, regnode *last, SV* sv, I32 l)
3769 register U8 op = EXACT; /* Arbitrary non-END op. */
3770 register regnode *next;
3772 while (op != END && (!last || node < last)) {
3773 /* While that wasn't END last time... */
3779 next = regnext(node);
3781 if (OP(node) == OPTIMIZED)
3784 PerlIO_printf(Perl_debug_log, "%4"IVdf":%*s%s", (IV)(node - start),
3785 (int)(2*l + 1), "", SvPVX(sv));
3786 if (next == NULL) /* Next ptr. */
3787 PerlIO_printf(Perl_debug_log, "(0)");
3789 PerlIO_printf(Perl_debug_log, "(%"IVdf")", (IV)(next - start));
3790 (void)PerlIO_putc(Perl_debug_log, '\n');
3792 if (PL_regkind[(U8)op] == BRANCHJ) {
3793 register regnode *nnode = (OP(next) == LONGJMP
3796 if (last && nnode > last)
3798 node = dumpuntil(start, NEXTOPER(NEXTOPER(node)), nnode, sv, l + 1);
3800 else if (PL_regkind[(U8)op] == BRANCH) {
3801 node = dumpuntil(start, NEXTOPER(node), next, sv, l + 1);
3803 else if ( op == CURLY) { /* `next' might be very big: optimizer */
3804 node = dumpuntil(start, NEXTOPER(node) + EXTRA_STEP_2ARGS,
3805 NEXTOPER(node) + EXTRA_STEP_2ARGS + 1, sv, l + 1);
3807 else if (PL_regkind[(U8)op] == CURLY && op != CURLYX) {
3808 node = dumpuntil(start, NEXTOPER(node) + EXTRA_STEP_2ARGS,
3811 else if ( op == PLUS || op == STAR) {
3812 node = dumpuntil(start, NEXTOPER(node), NEXTOPER(node) + 1, sv, l + 1);
3814 else if (op == ANYOF) {
3815 node = NEXTOPER(node);
3818 else if (PL_regkind[(U8)op] == EXACT) {
3819 /* Literal string, where present. */
3820 node += NODE_SZ_STR(node) - 1;
3821 node = NEXTOPER(node);
3824 node = NEXTOPER(node);
3825 node += regarglen[(U8)op];
3827 if (op == CURLYX || op == OPEN)
3829 else if (op == WHILEM)
3832 #endif /* DEBUGGING */
3837 - regdump - dump a regexp onto Perl_debug_log in vaguely comprehensible form
3840 Perl_regdump(pTHX_ regexp *r)
3844 SV *sv = sv_newmortal();
3846 (void)dumpuntil(r->program, r->program + 1, NULL, sv, 0);
3848 /* Header fields of interest. */
3849 if (r->anchored_substr)
3850 PerlIO_printf(Perl_debug_log,
3851 "anchored `%s%.*s%s'%s at %"IVdf" ",
3853 (int)(SvCUR(r->anchored_substr) - (SvTAIL(r->anchored_substr)!=0)),
3854 SvPVX(r->anchored_substr),
3856 SvTAIL(r->anchored_substr) ? "$" : "",
3857 (IV)r->anchored_offset);
3858 if (r->float_substr)
3859 PerlIO_printf(Perl_debug_log,
3860 "floating `%s%.*s%s'%s at %"IVdf"..%"UVuf" ",
3862 (int)(SvCUR(r->float_substr) - (SvTAIL(r->float_substr)!=0)),
3863 SvPVX(r->float_substr),
3865 SvTAIL(r->float_substr) ? "$" : "",
3866 (IV)r->float_min_offset, (UV)r->float_max_offset);
3867 if (r->check_substr)
3868 PerlIO_printf(Perl_debug_log,
3869 r->check_substr == r->float_substr
3870 ? "(checking floating" : "(checking anchored");
3871 if (r->reganch & ROPT_NOSCAN)
3872 PerlIO_printf(Perl_debug_log, " noscan");
3873 if (r->reganch & ROPT_CHECK_ALL)
3874 PerlIO_printf(Perl_debug_log, " isall");
3875 if (r->check_substr)
3876 PerlIO_printf(Perl_debug_log, ") ");
3878 if (r->regstclass) {
3879 regprop(sv, r->regstclass);
3880 PerlIO_printf(Perl_debug_log, "stclass `%s' ", SvPVX(sv));
3882 if (r->reganch & ROPT_ANCH) {
3883 PerlIO_printf(Perl_debug_log, "anchored");
3884 if (r->reganch & ROPT_ANCH_BOL)
3885 PerlIO_printf(Perl_debug_log, "(BOL)");
3886 if (r->reganch & ROPT_ANCH_MBOL)
3887 PerlIO_printf(Perl_debug_log, "(MBOL)");
3888 if (r->reganch & ROPT_ANCH_SBOL)
3889 PerlIO_printf(Perl_debug_log, "(SBOL)");
3890 if (r->reganch & ROPT_ANCH_GPOS)
3891 PerlIO_printf(Perl_debug_log, "(GPOS)");
3892 PerlIO_putc(Perl_debug_log, ' ');
3894 if (r->reganch & ROPT_GPOS_SEEN)
3895 PerlIO_printf(Perl_debug_log, "GPOS ");
3896 if (r->reganch & ROPT_SKIP)
3897 PerlIO_printf(Perl_debug_log, "plus ");
3898 if (r->reganch & ROPT_IMPLICIT)
3899 PerlIO_printf(Perl_debug_log, "implicit ");
3900 PerlIO_printf(Perl_debug_log, "minlen %ld ", (long) r->minlen);
3901 if (r->reganch & ROPT_EVAL_SEEN)
3902 PerlIO_printf(Perl_debug_log, "with eval ");
3903 PerlIO_printf(Perl_debug_log, "\n");
3904 #endif /* DEBUGGING */
3908 S_put_byte(pTHX_ SV *sv, int c)
3910 if (c <= ' ' || c == 127 || c == 255)
3911 Perl_sv_catpvf(aTHX_ sv, "\\%o", c);
3912 else if (c == '-' || c == ']' || c == '\\' || c == '^')
3913 Perl_sv_catpvf(aTHX_ sv, "\\%c", c);
3915 Perl_sv_catpvf(aTHX_ sv, "%c", c);
3919 - regprop - printable representation of opcode
3922 Perl_regprop(pTHX_ SV *sv, regnode *o)
3928 sv_setpvn(sv, "", 0);
3929 if (OP(o) >= reg_num) /* regnode.type is unsigned */
3930 FAIL("corrupted regexp opcode");
3931 sv_catpv(sv, (char*)reg_name[OP(o)]); /* Take off const! */
3933 k = PL_regkind[(U8)OP(o)];
3936 Perl_sv_catpvf(aTHX_ sv, " <%s%.*s%s>", PL_colors[0],
3937 STR_LEN(o), STRING(o), PL_colors[1]);
3938 else if (k == CURLY) {
3939 if (OP(o) == CURLYM || OP(o) == CURLYN)
3940 Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* Parenth number */
3941 Perl_sv_catpvf(aTHX_ sv, " {%d,%d}", ARG1(o), ARG2(o));
3943 else if (k == WHILEM && o->flags) /* Ordinal/of */
3944 Perl_sv_catpvf(aTHX_ sv, "[%d/%d]", o->flags & 0xf, o->flags>>4);
3945 else if (k == REF || k == OPEN || k == CLOSE || k == GROUPP )
3946 Perl_sv_catpvf(aTHX_ sv, "%d", (int)ARG(o)); /* Parenth number */
3947 else if (k == LOGICAL)
3948 Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* 2: embedded, otherwise 1 */
3949 else if (k == ANYOF) {
3950 int i, rangestart = -1;
3951 const char * const out[] = { /* Should be syncronized with
3952 a table in regcomp.h */
3981 if (o->flags & ANYOF_LOCALE)
3982 sv_catpv(sv, "{loc}");
3983 if (o->flags & ANYOF_FOLD)
3984 sv_catpv(sv, "{i}");
3985 Perl_sv_catpvf(aTHX_ sv, "[%s", PL_colors[0]);
3986 if (o->flags & ANYOF_INVERT)
3988 for (i = 0; i <= 256; i++) {
3989 if (i < 256 && ANYOF_BITMAP_TEST(o,i)) {
3990 if (rangestart == -1)
3992 } else if (rangestart != -1) {
3993 if (i <= rangestart + 3)
3994 for (; rangestart < i; rangestart++)
3995 put_byte(sv, rangestart);
3997 put_byte(sv, rangestart);
3999 put_byte(sv, i - 1);
4004 if (o->flags & ANYOF_CLASS)
4005 for (i = 0; i < sizeof(out)/sizeof(char*); i++)
4006 if (ANYOF_CLASS_TEST(o,i))
4007 sv_catpv(sv, out[i]);
4008 Perl_sv_catpvf(aTHX_ sv, "%s]", PL_colors[1]);
4010 else if (k == BRANCHJ && (OP(o) == UNLESSM || OP(o) == IFMATCH))
4011 Perl_sv_catpvf(aTHX_ sv, "[-%d]", o->flags);
4012 #endif /* DEBUGGING */
4016 Perl_re_intuit_string(pTHX_ regexp *prog)
4017 { /* Assume that RE_INTUIT is set */
4020 char *s = SvPV(prog->check_substr,n_a);
4022 if (!PL_colorset) reginitcolors();
4023 PerlIO_printf(Perl_debug_log,
4024 "%sUsing REx substr:%s `%s%.60s%s%s'\n",
4025 PL_colors[4],PL_colors[5],PL_colors[0],
4028 (strlen(s) > 60 ? "..." : ""));
4031 return prog->check_substr;
4035 Perl_pregfree(pTHX_ struct regexp *r)
4038 DEBUG_r(if (!PL_colorset) reginitcolors());
4040 if (!r || (--r->refcnt > 0))
4042 DEBUG_r(PerlIO_printf(Perl_debug_log,
4043 "%sFreeing REx:%s `%s%.60s%s%s'\n",
4044 PL_colors[4],PL_colors[5],PL_colors[0],
4047 (strlen(r->precomp) > 60 ? "..." : "")));
4050 Safefree(r->precomp);
4051 if (RX_MATCH_COPIED(r))
4052 Safefree(r->subbeg);
4054 if (r->anchored_substr)
4055 SvREFCNT_dec(r->anchored_substr);
4056 if (r->float_substr)
4057 SvREFCNT_dec(r->float_substr);
4058 Safefree(r->substrs);
4061 int n = r->data->count;
4062 AV* new_comppad = NULL;
4067 switch (r->data->what[n]) {
4069 SvREFCNT_dec((SV*)r->data->data[n]);
4072 Safefree(r->data->data[n]);
4075 new_comppad = (AV*)r->data->data[n];
4078 if (new_comppad == NULL)
4079 Perl_croak(aTHX_ "panic: pregfree comppad");
4080 old_comppad = PL_comppad;
4081 old_curpad = PL_curpad;
4082 PL_comppad = new_comppad;
4083 PL_curpad = AvARRAY(new_comppad);
4084 op_free((OP_4tree*)r->data->data[n]);
4085 PL_comppad = old_comppad;
4086 PL_curpad = old_curpad;
4087 SvREFCNT_dec((SV*)new_comppad);
4093 FAIL2("panic: regfree data code '%c'", r->data->what[n]);
4096 Safefree(r->data->what);
4099 Safefree(r->startp);
4105 - regnext - dig the "next" pointer out of a node
4107 * [Note, when REGALIGN is defined there are two places in regmatch()
4108 * that bypass this code for speed.]
4111 Perl_regnext(pTHX_ register regnode *p)
4114 register I32 offset;
4116 if (p == &PL_regdummy)
4119 offset = (reg_off_by_arg[OP(p)] ? ARG(p) : NEXT_OFF(p));
4127 S_re_croak2(pTHX_ const char* pat1,const char* pat2,...)
4130 STRLEN l1 = strlen(pat1);
4131 STRLEN l2 = strlen(pat2);
4140 Copy(pat1, buf, l1 , char);
4141 Copy(pat2, buf + l1, l2 , char);
4142 buf[l1 + l2] = '\n';
4143 buf[l1 + l2 + 1] = '\0';
4145 /* ANSI variant takes additional second argument */
4146 va_start(args, pat2);
4150 msv = vmess(buf, &args);
4152 message = SvPV(msv,l1);
4155 Copy(message, buf, l1 , char);
4156 buf[l1] = '\0'; /* Overwrite \n */
4157 Perl_croak(aTHX_ "%s", buf);
4160 /* XXX Here's a total kludge. But we need to re-enter for swash routines. */
4163 Perl_save_re_context(pTHX)
4167 SAVEPPTR(PL_regprecomp); /* uncompiled string. */
4168 SAVEI32(PL_regnpar); /* () count. */
4169 SAVEI32(PL_regsize); /* Code size. */
4170 SAVEI16(PL_regflags); /* are we folding, multilining? */
4171 SAVEPPTR(PL_reginput); /* String-input pointer. */
4172 SAVEPPTR(PL_regbol); /* Beginning of input, for ^ check. */
4173 SAVEPPTR(PL_regeol); /* End of input, for $ check. */
4174 SAVEVPTR(PL_regstartp); /* Pointer to startp array. */
4175 SAVEVPTR(PL_regendp); /* Ditto for endp. */
4176 SAVEVPTR(PL_reglastparen); /* Similarly for lastparen. */
4177 SAVEPPTR(PL_regtill); /* How far we are required to go. */
4178 SAVEI8(PL_regprev); /* char before regbol, \n if none */
4179 SAVEVPTR(PL_reg_start_tmp); /* from regexec.c */
4180 PL_reg_start_tmp = 0;
4181 SAVEFREEPV(PL_reg_start_tmp);
4182 SAVEI32(PL_reg_start_tmpl); /* from regexec.c */
4183 PL_reg_start_tmpl = 0;
4184 SAVEVPTR(PL_regdata);
4185 SAVEI32(PL_reg_flags); /* from regexec.c */
4186 SAVEI32(PL_reg_eval_set); /* from regexec.c */
4187 SAVEI32(PL_regnarrate); /* from regexec.c */
4188 SAVEVPTR(PL_regprogram); /* from regexec.c */
4189 SAVEINT(PL_regindent); /* from regexec.c */
4190 SAVEVPTR(PL_regcc); /* from regexec.c */
4191 SAVEVPTR(PL_curcop);
4192 SAVEVPTR(PL_regcomp_rx); /* from regcomp.c */
4193 SAVEI32(PL_regseen); /* from regcomp.c */
4194 SAVEI32(PL_regsawback); /* Did we see \1, ...? */
4195 SAVEI32(PL_regnaughty); /* How bad is this pattern? */
4196 SAVEVPTR(PL_regcode); /* Code-emit pointer; ®dummy = don't */
4197 SAVEPPTR(PL_regxend); /* End of input for compile */
4198 SAVEPPTR(PL_regcomp_parse); /* Input-scan pointer. */
4199 SAVEVPTR(PL_reg_call_cc); /* from regexec.c */
4200 SAVEVPTR(PL_reg_re); /* from regexec.c */
4201 SAVEPPTR(PL_reg_ganch); /* from regexec.c */
4202 SAVESPTR(PL_reg_sv); /* from regexec.c */
4203 SAVEVPTR(PL_reg_magic); /* from regexec.c */
4204 SAVEI32(PL_reg_oldpos); /* from regexec.c */
4205 SAVEVPTR(PL_reg_oldcurpm); /* from regexec.c */
4206 SAVEVPTR(PL_reg_curpm); /* from regexec.c */
4208 SAVEPPTR(PL_reg_starttry); /* from regexec.c */
4219 clear_re(pTHXo_ void *r)
4221 ReREFCNT_dec((regexp *)r);