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);
905 /* start offset must point into the last copy */
906 data->last_start_min += minnext * (mincount - 1);
907 data->last_start_max += is_inf ? 0 : (maxcount - 1)
908 * (minnext + data->pos_delta);
911 /* It is counted once already... */
912 data->pos_min += minnext * (mincount - counted);
913 data->pos_delta += - counted * deltanext +
914 (minnext + deltanext) * maxcount - minnext * mincount;
915 if (mincount != maxcount) {
916 /* Cannot extend fixed substrings found inside
919 if (mincount && last_str) {
920 sv_setsv(data->last_found, last_str);
921 data->last_end = data->pos_min;
922 data->last_start_min =
923 data->pos_min - CHR_SVLEN(last_str);
924 data->last_start_max = is_inf
926 : data->pos_min + data->pos_delta
927 - CHR_SVLEN(last_str);
929 data->longest = &(data->longest_float);
931 SvREFCNT_dec(last_str);
933 if (data && (fl & SF_HAS_EVAL))
934 data->flags |= SF_HAS_EVAL;
936 if (OP(oscan) != CURLYX) {
937 while (PL_regkind[(U8)OP(next = regnext(oscan))] == NOTHING
939 NEXT_OFF(oscan) += NEXT_OFF(next);
942 default: /* REF and CLUMP only? */
943 if (flags & SCF_DO_SUBSTR) {
944 scan_commit(data); /* Cannot expect anything... */
945 data->longest = &(data->longest_float);
947 is_inf = is_inf_internal = 1;
948 if (flags & SCF_DO_STCLASS_OR)
949 cl_anything(data->start_class);
950 flags &= ~SCF_DO_STCLASS;
954 else if (strchr((char*)PL_simple,OP(scan)) || PL_regkind[(U8)OP(scan)] == ANYUTF8) {
957 if (flags & SCF_DO_SUBSTR) {
962 if (flags & SCF_DO_STCLASS) {
963 data->start_class->flags &= ~ANYOF_EOS; /* No match on empty */
965 /* Some of the logic below assumes that switching
966 locale on will only add false positives. */
967 switch (PL_regkind[(U8)OP(scan)]) {
984 /* Perl_croak(aTHX_ "panic: unexpected simple REx opcode %d", OP(scan)); */
985 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
986 cl_anything(data->start_class);
989 if (OP(scan) == SANY)
991 if (flags & SCF_DO_STCLASS_OR) { /* Everything but \n */
992 value = (ANYOF_BITMAP_TEST(data->start_class,'\n')
993 || (data->start_class->flags & ANYOF_CLASS));
994 cl_anything(data->start_class);
996 if (flags & SCF_DO_STCLASS_AND || !value)
997 ANYOF_BITMAP_CLEAR(data->start_class,'\n');
1000 if (flags & SCF_DO_STCLASS_AND)
1001 cl_and(data->start_class,
1002 (struct regnode_charclass_class*)scan);
1004 cl_or(data->start_class,
1005 (struct regnode_charclass_class*)scan);
1008 if (flags & SCF_DO_STCLASS_AND) {
1009 if (!(data->start_class->flags & ANYOF_LOCALE)) {
1010 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
1011 for (value = 0; value < 256; value++)
1012 if (!isALNUM(value))
1013 ANYOF_BITMAP_CLEAR(data->start_class, value);
1017 if (data->start_class->flags & ANYOF_LOCALE)
1018 ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
1020 for (value = 0; value < 256; value++)
1022 ANYOF_BITMAP_SET(data->start_class, value);
1027 if (flags & SCF_DO_STCLASS_AND) {
1028 if (data->start_class->flags & ANYOF_LOCALE)
1029 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
1032 ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
1033 data->start_class->flags |= ANYOF_LOCALE;
1037 if (flags & SCF_DO_STCLASS_AND) {
1038 if (!(data->start_class->flags & ANYOF_LOCALE)) {
1039 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
1040 for (value = 0; value < 256; value++)
1042 ANYOF_BITMAP_CLEAR(data->start_class, value);
1046 if (data->start_class->flags & ANYOF_LOCALE)
1047 ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
1049 for (value = 0; value < 256; value++)
1050 if (!isALNUM(value))
1051 ANYOF_BITMAP_SET(data->start_class, value);
1056 if (flags & SCF_DO_STCLASS_AND) {
1057 if (data->start_class->flags & ANYOF_LOCALE)
1058 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
1061 data->start_class->flags |= ANYOF_LOCALE;
1062 ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
1066 if (flags & SCF_DO_STCLASS_AND) {
1067 if (!(data->start_class->flags & ANYOF_LOCALE)) {
1068 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
1069 for (value = 0; value < 256; value++)
1070 if (!isSPACE(value))
1071 ANYOF_BITMAP_CLEAR(data->start_class, value);
1075 if (data->start_class->flags & ANYOF_LOCALE)
1076 ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
1078 for (value = 0; value < 256; value++)
1080 ANYOF_BITMAP_SET(data->start_class, value);
1085 if (flags & SCF_DO_STCLASS_AND) {
1086 if (data->start_class->flags & ANYOF_LOCALE)
1087 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
1090 data->start_class->flags |= ANYOF_LOCALE;
1091 ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
1095 if (flags & SCF_DO_STCLASS_AND) {
1096 if (!(data->start_class->flags & ANYOF_LOCALE)) {
1097 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
1098 for (value = 0; value < 256; value++)
1100 ANYOF_BITMAP_CLEAR(data->start_class, value);
1104 if (data->start_class->flags & ANYOF_LOCALE)
1105 ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
1107 for (value = 0; value < 256; value++)
1108 if (!isSPACE(value))
1109 ANYOF_BITMAP_SET(data->start_class, value);
1114 if (flags & SCF_DO_STCLASS_AND) {
1115 if (data->start_class->flags & ANYOF_LOCALE) {
1116 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
1117 for (value = 0; value < 256; value++)
1118 if (!isSPACE(value))
1119 ANYOF_BITMAP_CLEAR(data->start_class, value);
1123 data->start_class->flags |= ANYOF_LOCALE;
1124 ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
1128 if (flags & SCF_DO_STCLASS_AND) {
1129 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NDIGIT);
1130 for (value = 0; value < 256; value++)
1131 if (!isDIGIT(value))
1132 ANYOF_BITMAP_CLEAR(data->start_class, value);
1135 if (data->start_class->flags & ANYOF_LOCALE)
1136 ANYOF_CLASS_SET(data->start_class,ANYOF_DIGIT);
1138 for (value = 0; value < 256; value++)
1140 ANYOF_BITMAP_SET(data->start_class, value);
1145 if (flags & SCF_DO_STCLASS_AND) {
1146 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_DIGIT);
1147 for (value = 0; value < 256; value++)
1149 ANYOF_BITMAP_CLEAR(data->start_class, value);
1152 if (data->start_class->flags & ANYOF_LOCALE)
1153 ANYOF_CLASS_SET(data->start_class,ANYOF_NDIGIT);
1155 for (value = 0; value < 256; value++)
1156 if (!isDIGIT(value))
1157 ANYOF_BITMAP_SET(data->start_class, value);
1162 if (flags & SCF_DO_STCLASS_OR)
1163 cl_and(data->start_class, &and_with);
1164 flags &= ~SCF_DO_STCLASS;
1167 else if (PL_regkind[(U8)OP(scan)] == EOL && flags & SCF_DO_SUBSTR) {
1168 data->flags |= (OP(scan) == MEOL
1172 else if ( PL_regkind[(U8)OP(scan)] == BRANCHJ
1173 /* Lookbehind, or need to calculate parens/evals/stclass: */
1174 && (scan->flags || data || (flags & SCF_DO_STCLASS))
1175 && (OP(scan) == IFMATCH || OP(scan) == UNLESSM)) {
1176 /* Lookahead/lookbehind */
1177 I32 deltanext, minnext;
1179 struct regnode_charclass_class intrnl;
1182 data_fake.flags = 0;
1184 data_fake.whilem_c = data->whilem_c;
1185 if ( flags & SCF_DO_STCLASS && !scan->flags
1186 && OP(scan) == IFMATCH ) { /* Lookahead */
1188 data_fake.start_class = &intrnl;
1189 f = SCF_DO_STCLASS_AND;
1191 next = regnext(scan);
1192 nscan = NEXTOPER(NEXTOPER(scan));
1193 minnext = study_chunk(&nscan, &deltanext, last, &data_fake, f);
1196 FAIL("variable length lookbehind not implemented");
1198 else if (minnext > U8_MAX) {
1199 FAIL2("lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX);
1201 scan->flags = minnext;
1203 if (data && data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
1205 if (data && (data_fake.flags & SF_HAS_EVAL))
1206 data->flags |= SF_HAS_EVAL;
1208 data->whilem_c = data_fake.whilem_c;
1210 int was = (data->start_class->flags & ANYOF_EOS);
1212 cl_and(data->start_class, &intrnl);
1214 data->start_class->flags |= ANYOF_EOS;
1217 else if (OP(scan) == OPEN) {
1220 else if (OP(scan) == CLOSE && ARG(scan) == is_par) {
1221 next = regnext(scan);
1223 if ( next && (OP(next) != WHILEM) && next < last)
1224 is_par = 0; /* Disable optimization */
1226 else if (OP(scan) == EVAL) {
1228 data->flags |= SF_HAS_EVAL;
1230 else if (OP(scan) == LOGICAL && scan->flags == 2) { /* Embedded follows */
1231 if (flags & SCF_DO_SUBSTR) {
1233 data->longest = &(data->longest_float);
1235 is_inf = is_inf_internal = 1;
1236 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
1237 cl_anything(data->start_class);
1238 flags &= ~SCF_DO_STCLASS;
1240 /* Else: zero-length, ignore. */
1241 scan = regnext(scan);
1246 *deltap = is_inf_internal ? I32_MAX : delta;
1247 if (flags & SCF_DO_SUBSTR && is_inf)
1248 data->pos_delta = I32_MAX - data->pos_min;
1249 if (is_par > U8_MAX)
1251 if (is_par && pars==1 && data) {
1252 data->flags |= SF_IN_PAR;
1253 data->flags &= ~SF_HAS_PAR;
1255 else if (pars && data) {
1256 data->flags |= SF_HAS_PAR;
1257 data->flags &= ~SF_IN_PAR;
1259 if (flags & SCF_DO_STCLASS_OR)
1260 cl_and(data->start_class, &and_with);
1265 S_add_data(pTHX_ I32 n, char *s)
1268 if (PL_regcomp_rx->data) {
1269 Renewc(PL_regcomp_rx->data,
1270 sizeof(*PL_regcomp_rx->data) + sizeof(void*) * (PL_regcomp_rx->data->count + n - 1),
1271 char, struct reg_data);
1272 Renew(PL_regcomp_rx->data->what, PL_regcomp_rx->data->count + n, U8);
1273 PL_regcomp_rx->data->count += n;
1276 Newc(1207, PL_regcomp_rx->data, sizeof(*PL_regcomp_rx->data) + sizeof(void*) * (n - 1),
1277 char, struct reg_data);
1278 New(1208, PL_regcomp_rx->data->what, n, U8);
1279 PL_regcomp_rx->data->count = n;
1281 Copy(s, PL_regcomp_rx->data->what + PL_regcomp_rx->data->count - n, n, U8);
1282 return PL_regcomp_rx->data->count - n;
1286 Perl_reginitcolors(pTHX)
1290 char *s = PerlEnv_getenv("PERL_RE_COLORS");
1293 PL_colors[0] = s = savepv(s);
1295 s = strchr(s, '\t');
1301 PL_colors[i] = s = "";
1305 PL_colors[i++] = "";
1311 - pregcomp - compile a regular expression into internal code
1313 * We can't allocate space until we know how big the compiled form will be,
1314 * but we can't compile it (and thus know how big it is) until we've got a
1315 * place to put the code. So we cheat: we compile it twice, once with code
1316 * generation turned off and size counting turned on, and once "for real".
1317 * This also means that we don't allocate space until we are sure that the
1318 * thing really will compile successfully, and we never have to move the
1319 * code and thus invalidate pointers into it. (Note that it has to be in
1320 * one piece because free() must be able to free it all.) [NB: not true in perl]
1322 * Beware that the optimization-preparation code in here knows about some
1323 * of the structure of the compiled regexp. [I'll say.]
1326 Perl_pregcomp(pTHX_ char *exp, char *xend, PMOP *pm)
1339 FAIL("NULL regexp argument");
1341 if (pm->op_pmdynflags & PMdf_UTF8) {
1342 PL_reg_flags |= RF_utf8;
1347 PL_regprecomp = savepvn(exp, xend - exp);
1348 DEBUG_r(if (!PL_colorset) reginitcolors());
1349 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sCompiling REx%s `%s%*s%s'\n",
1350 PL_colors[4],PL_colors[5],PL_colors[0],
1351 (int)(xend - exp), PL_regprecomp, PL_colors[1]));
1352 PL_regflags = pm->op_pmflags;
1356 PL_seen_zerolen = *exp == '^' ? -1 : 0;
1360 /* First pass: determine size, legality. */
1361 PL_regcomp_parse = exp;
1366 PL_regcode = &PL_regdummy;
1367 PL_reg_whilem_seen = 0;
1368 #if 0 /* REGC() is (currently) a NOP at the first pass.
1369 * Clever compilers notice this and complain. --jhi */
1370 REGC((U8)REG_MAGIC, (char*)PL_regcode);
1372 if (reg(0, &flags) == NULL) {
1373 Safefree(PL_regprecomp);
1374 PL_regprecomp = Nullch;
1377 DEBUG_r(PerlIO_printf(Perl_debug_log, "size %"IVdf" ", (IV)PL_regsize));
1379 /* Small enough for pointer-storage convention?
1380 If extralen==0, this means that we will not need long jumps. */
1381 if (PL_regsize >= 0x10000L && PL_extralen)
1382 PL_regsize += PL_extralen;
1385 if (PL_reg_whilem_seen > 15)
1386 PL_reg_whilem_seen = 15;
1388 /* Allocate space and initialize. */
1389 Newc(1001, r, sizeof(regexp) + (unsigned)PL_regsize * sizeof(regnode),
1392 FAIL("regexp out of space");
1394 /* avoid reading uninitialized memory in DEBUGGING code in study_chunk() */
1395 Zero(r, sizeof(regexp) + (unsigned)PL_regsize * sizeof(regnode), char);
1398 r->prelen = xend - exp;
1399 r->precomp = PL_regprecomp;
1401 r->reganch = pm->op_pmflags & PMf_COMPILETIME;
1402 r->nparens = PL_regnpar - 1; /* set early to validate backrefs */
1404 r->substrs = 0; /* Useful during FAIL. */
1405 r->startp = 0; /* Useful during FAIL. */
1406 r->endp = 0; /* Useful during FAIL. */
1410 /* Second pass: emit code. */
1411 PL_regcomp_parse = exp;
1415 PL_regcode = r->program;
1416 /* Store the count of eval-groups for security checks: */
1417 PL_regcode->next_off = ((PL_seen_evals > U16_MAX) ? U16_MAX : PL_seen_evals);
1418 REGC((U8)REG_MAGIC, (char*) PL_regcode++);
1420 if (reg(0, &flags) == NULL)
1423 /* Dig out information for optimizations. */
1424 r->reganch = pm->op_pmflags & PMf_COMPILETIME; /* Again? */
1425 pm->op_pmflags = PL_regflags;
1427 r->reganch |= ROPT_UTF8;
1428 r->regstclass = NULL;
1429 if (PL_regnaughty >= 10) /* Probably an expensive pattern. */
1430 r->reganch |= ROPT_NAUGHTY;
1431 scan = r->program + 1; /* First BRANCH. */
1433 /* XXXX To minimize changes to RE engine we always allocate
1434 3-units-long substrs field. */
1435 Newz(1004, r->substrs, 1, struct reg_substr_data);
1437 StructCopy(&zero_scan_data, &data, scan_data_t);
1438 /* XXXX Should not we check for something else? Usually it is OPEN1... */
1439 if (OP(scan) != BRANCH) { /* Only one top-level choice. */
1441 STRLEN longest_float_length, longest_fixed_length;
1442 struct regnode_charclass_class ch_class;
1446 /* Skip introductions and multiplicators >= 1. */
1447 while ((OP(first) == OPEN && (sawopen = 1)) ||
1448 /* An OR of *one* alternative - should not happen now. */
1449 (OP(first) == BRANCH && OP(regnext(first)) != BRANCH) ||
1450 (OP(first) == PLUS) ||
1451 (OP(first) == MINMOD) ||
1452 /* An {n,m} with n>0 */
1453 (PL_regkind[(U8)OP(first)] == CURLY && ARG1(first) > 0) ) {
1454 if (OP(first) == PLUS)
1457 first += regarglen[(U8)OP(first)];
1458 first = NEXTOPER(first);
1461 /* Starting-point info. */
1463 if (PL_regkind[(U8)OP(first)] == EXACT) {
1464 if (OP(first) == EXACT); /* Empty, get anchored substr later. */
1465 else if ((OP(first) == EXACTF || OP(first) == EXACTFL)
1467 r->regstclass = first;
1469 else if (strchr((char*)PL_simple,OP(first)))
1470 r->regstclass = first;
1471 else if (PL_regkind[(U8)OP(first)] == BOUND ||
1472 PL_regkind[(U8)OP(first)] == NBOUND)
1473 r->regstclass = first;
1474 else if (PL_regkind[(U8)OP(first)] == BOL) {
1475 r->reganch |= (OP(first) == MBOL
1477 : (OP(first) == SBOL
1480 first = NEXTOPER(first);
1483 else if (OP(first) == GPOS) {
1484 r->reganch |= ROPT_ANCH_GPOS;
1485 first = NEXTOPER(first);
1488 else if ((OP(first) == STAR &&
1489 PL_regkind[(U8)OP(NEXTOPER(first))] == REG_ANY) &&
1490 !(r->reganch & ROPT_ANCH) )
1492 /* turn .* into ^.* with an implied $*=1 */
1493 int type = OP(NEXTOPER(first));
1495 if (type == REG_ANY || type == ANYUTF8)
1496 type = ROPT_ANCH_MBOL;
1498 type = ROPT_ANCH_SBOL;
1500 r->reganch |= type | ROPT_IMPLICIT;
1501 first = NEXTOPER(first);
1504 if (sawplus && (!sawopen || !PL_regsawback)
1505 && !(PL_regseen & REG_SEEN_EVAL)) /* May examine pos and $& */
1506 /* x+ must match at the 1st pos of run of x's */
1507 r->reganch |= ROPT_SKIP;
1509 /* Scan is after the zeroth branch, first is atomic matcher. */
1510 DEBUG_r(PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n",
1511 (IV)(first - scan + 1)));
1513 * If there's something expensive in the r.e., find the
1514 * longest literal string that must appear and make it the
1515 * regmust. Resolve ties in favor of later strings, since
1516 * the regstart check works with the beginning of the r.e.
1517 * and avoiding duplication strengthens checking. Not a
1518 * strong reason, but sufficient in the absence of others.
1519 * [Now we resolve ties in favor of the earlier string if
1520 * it happens that c_offset_min has been invalidated, since the
1521 * earlier string may buy us something the later one won't.]
1525 data.longest_fixed = newSVpvn("",0);
1526 data.longest_float = newSVpvn("",0);
1527 data.last_found = newSVpvn("",0);
1528 data.longest = &(data.longest_fixed);
1530 if (!r->regstclass) {
1532 data.start_class = &ch_class;
1533 stclass_flag = SCF_DO_STCLASS_AND;
1534 } else /* XXXX Check for BOUND? */
1537 minlen = study_chunk(&first, &fake, scan + PL_regsize, /* Up to end */
1538 &data, SCF_DO_SUBSTR | stclass_flag);
1539 if ( PL_regnpar == 1 && data.longest == &(data.longest_fixed)
1540 && data.last_start_min == 0 && data.last_end > 0
1542 && (!(PL_regseen & REG_SEEN_GPOS) || (r->reganch & ROPT_ANCH_GPOS)))
1543 r->reganch |= ROPT_CHECK_ALL;
1545 SvREFCNT_dec(data.last_found);
1547 longest_float_length = CHR_SVLEN(data.longest_float);
1548 if (longest_float_length
1549 || (data.flags & SF_FL_BEFORE_EOL
1550 && (!(data.flags & SF_FL_BEFORE_MEOL)
1551 || (PL_regflags & PMf_MULTILINE)))) {
1554 if (SvCUR(data.longest_fixed) /* ok to leave SvCUR */
1555 && data.offset_fixed == data.offset_float_min
1556 && SvCUR(data.longest_fixed) == SvCUR(data.longest_float))
1557 goto remove_float; /* As in (a)+. */
1559 r->float_substr = data.longest_float;
1560 r->float_min_offset = data.offset_float_min;
1561 r->float_max_offset = data.offset_float_max;
1562 t = (data.flags & SF_FL_BEFORE_EOL /* Can't have SEOL and MULTI */
1563 && (!(data.flags & SF_FL_BEFORE_MEOL)
1564 || (PL_regflags & PMf_MULTILINE)));
1565 fbm_compile(r->float_substr, t ? FBMcf_TAIL : 0);
1569 r->float_substr = Nullsv;
1570 SvREFCNT_dec(data.longest_float);
1571 longest_float_length = 0;
1574 longest_fixed_length = CHR_SVLEN(data.longest_fixed);
1575 if (longest_fixed_length
1576 || (data.flags & SF_FIX_BEFORE_EOL /* Cannot have SEOL and MULTI */
1577 && (!(data.flags & SF_FIX_BEFORE_MEOL)
1578 || (PL_regflags & PMf_MULTILINE)))) {
1581 r->anchored_substr = data.longest_fixed;
1582 r->anchored_offset = data.offset_fixed;
1583 t = (data.flags & SF_FIX_BEFORE_EOL /* Can't have SEOL and MULTI */
1584 && (!(data.flags & SF_FIX_BEFORE_MEOL)
1585 || (PL_regflags & PMf_MULTILINE)));
1586 fbm_compile(r->anchored_substr, t ? FBMcf_TAIL : 0);
1589 r->anchored_substr = Nullsv;
1590 SvREFCNT_dec(data.longest_fixed);
1591 longest_fixed_length = 0;
1594 && (OP(r->regstclass) == REG_ANY || OP(r->regstclass) == ANYUTF8
1595 || OP(r->regstclass) == SANYUTF8 || OP(r->regstclass) == SANY))
1596 r->regstclass = NULL;
1597 if ((!r->anchored_substr || r->anchored_offset) && stclass_flag
1598 && !(data.start_class->flags & ANYOF_EOS)
1599 && !cl_is_anything(data.start_class)) {
1601 I32 n = add_data(1, "f");
1603 New(1006, PL_regcomp_rx->data->data[n], 1,
1604 struct regnode_charclass_class);
1605 StructCopy(data.start_class,
1606 (struct regnode_charclass_class*)PL_regcomp_rx->data->data[n],
1607 struct regnode_charclass_class);
1608 r->regstclass = (regnode*)PL_regcomp_rx->data->data[n];
1609 r->reganch &= ~ROPT_SKIP; /* Used in find_byclass(). */
1610 DEBUG_r((sv = sv_newmortal(),
1611 regprop(sv, (regnode*)data.start_class),
1612 PerlIO_printf(Perl_debug_log, "synthetic stclass `%s'.\n",
1616 /* A temporary algorithm prefers floated substr to fixed one to dig more info. */
1617 if (longest_fixed_length > longest_float_length) {
1618 r->check_substr = r->anchored_substr;
1619 r->check_offset_min = r->check_offset_max = r->anchored_offset;
1620 if (r->reganch & ROPT_ANCH_SINGLE)
1621 r->reganch |= ROPT_NOSCAN;
1624 r->check_substr = r->float_substr;
1625 r->check_offset_min = data.offset_float_min;
1626 r->check_offset_max = data.offset_float_max;
1628 /* XXXX Currently intuiting is not compatible with ANCH_GPOS.
1629 This should be changed ASAP! */
1630 if (r->check_substr && !(r->reganch & ROPT_ANCH_GPOS)) {
1631 r->reganch |= RE_USE_INTUIT;
1632 if (SvTAIL(r->check_substr))
1633 r->reganch |= RE_INTUIT_TAIL;
1637 /* Several toplevels. Best we can is to set minlen. */
1639 struct regnode_charclass_class ch_class;
1641 DEBUG_r(PerlIO_printf(Perl_debug_log, "\n"));
1642 scan = r->program + 1;
1644 data.start_class = &ch_class;
1645 minlen = study_chunk(&scan, &fake, scan + PL_regsize, &data, SCF_DO_STCLASS_AND);
1646 r->check_substr = r->anchored_substr = r->float_substr = Nullsv;
1647 if (!(data.start_class->flags & ANYOF_EOS)
1648 && !cl_is_anything(data.start_class)) {
1650 I32 n = add_data(1, "f");
1652 New(1006, PL_regcomp_rx->data->data[n], 1,
1653 struct regnode_charclass_class);
1654 StructCopy(data.start_class,
1655 (struct regnode_charclass_class*)PL_regcomp_rx->data->data[n],
1656 struct regnode_charclass_class);
1657 r->regstclass = (regnode*)PL_regcomp_rx->data->data[n];
1658 r->reganch &= ~ROPT_SKIP; /* Used in find_byclass(). */
1659 DEBUG_r((sv = sv_newmortal(),
1660 regprop(sv, (regnode*)data.start_class),
1661 PerlIO_printf(Perl_debug_log, "synthetic stclass `%s'.\n",
1667 if (PL_regseen & REG_SEEN_GPOS)
1668 r->reganch |= ROPT_GPOS_SEEN;
1669 if (PL_regseen & REG_SEEN_LOOKBEHIND)
1670 r->reganch |= ROPT_LOOKBEHIND_SEEN;
1671 if (PL_regseen & REG_SEEN_EVAL)
1672 r->reganch |= ROPT_EVAL_SEEN;
1673 Newz(1002, r->startp, PL_regnpar, I32);
1674 Newz(1002, r->endp, PL_regnpar, I32);
1675 DEBUG_r(regdump(r));
1680 - reg - regular expression, i.e. main body or parenthesized thing
1682 * Caller must absorb opening parenthesis.
1684 * Combining parenthesis handling with the base level of regular expression
1685 * is a trifle forced, but the need to tie the tails of the branches to what
1686 * follows makes it hard to avoid.
1689 S_reg(pTHX_ I32 paren, I32 *flagp)
1690 /* paren: Parenthesized? 0=top, 1=(, inside: changed to letter. */
1693 register regnode *ret; /* Will be the head of the group. */
1694 register regnode *br;
1695 register regnode *lastbr;
1696 register regnode *ender = 0;
1697 register I32 parno = 0;
1698 I32 flags, oregflags = PL_regflags, have_branch = 0, open = 0;
1701 *flagp = 0; /* Tentatively. */
1703 /* Make an OPEN node, if parenthesized. */
1705 if (*PL_regcomp_parse == '?') {
1706 U16 posflags = 0, negflags = 0;
1707 U16 *flagsp = &posflags;
1711 paren = *PL_regcomp_parse++;
1712 ret = NULL; /* For look-ahead/behind. */
1715 PL_regseen |= REG_SEEN_LOOKBEHIND;
1716 if (*PL_regcomp_parse == '!')
1718 if (*PL_regcomp_parse != '=' && *PL_regcomp_parse != '!')
1729 FAIL2("Sequence (?%c...) not implemented", (int)paren);
1732 while (*PL_regcomp_parse && *PL_regcomp_parse != ')')
1734 if (*PL_regcomp_parse != ')')
1735 FAIL("Sequence (?#... not terminated");
1741 Perl_warner(aTHX_ WARN_REGEXP,
1742 "(?p{}) is deprecated - use (??{})");
1746 paren = *PL_regcomp_parse++;
1751 I32 count = 1, n = 0;
1753 char *s = PL_regcomp_parse;
1755 OP_4tree *sop, *rop;
1758 PL_regseen |= REG_SEEN_EVAL;
1759 while (count && (c = *PL_regcomp_parse)) {
1760 if (c == '\\' && PL_regcomp_parse[1])
1768 if (*PL_regcomp_parse != ')')
1769 FAIL("Sequence (?{...}) not terminated or not {}-balanced");
1773 if (PL_regcomp_parse - 1 - s)
1774 sv = newSVpvn(s, PL_regcomp_parse - 1 - s);
1776 sv = newSVpvn("", 0);
1778 rop = sv_compile_2op(sv, &sop, "re", &av);
1780 n = add_data(3, "nop");
1781 PL_regcomp_rx->data->data[n] = (void*)rop;
1782 PL_regcomp_rx->data->data[n+1] = (void*)sop;
1783 PL_regcomp_rx->data->data[n+2] = (void*)av;
1786 else { /* First pass */
1787 if (PL_reginterp_cnt < ++PL_seen_evals
1788 && PL_curcop != &PL_compiling)
1789 /* No compiled RE interpolated, has runtime
1790 components ===> unsafe. */
1791 FAIL("Eval-group not allowed at runtime, use re 'eval'");
1793 FAIL("Eval-group in insecure regular expression");
1798 ret = reg_node(LOGICAL);
1801 regtail(ret, reganode(EVAL, n));
1804 return reganode(EVAL, n);
1808 if (PL_regcomp_parse[0] == '?') {
1809 if (PL_regcomp_parse[1] == '=' || PL_regcomp_parse[1] == '!'
1810 || PL_regcomp_parse[1] == '<'
1811 || PL_regcomp_parse[1] == '{') { /* Lookahead or eval. */
1814 ret = reg_node(LOGICAL);
1817 regtail(ret, reg(1, &flag));
1821 else if (PL_regcomp_parse[0] >= '1' && PL_regcomp_parse[0] <= '9' ) {
1822 parno = atoi(PL_regcomp_parse++);
1824 while (isDIGIT(*PL_regcomp_parse))
1826 ret = reganode(GROUPP, parno);
1827 if ((c = *nextchar()) != ')')
1828 FAIL2("Switch (?(number%c not recognized", c);
1830 regtail(ret, reganode(IFTHEN, 0));
1831 br = regbranch(&flags, 1);
1833 br = reganode(LONGJMP, 0);
1835 regtail(br, reganode(LONGJMP, 0));
1840 lastbr = reganode(IFTHEN, 0); /* Fake one for optimizer. */
1841 regbranch(&flags, 1);
1842 regtail(ret, lastbr);
1850 FAIL("Switch (?(condition)... contains too many branches");
1851 ender = reg_node(TAIL);
1854 regtail(lastbr, ender);
1855 regtail(NEXTOPER(NEXTOPER(lastbr)), ender);
1858 regtail(ret, ender);
1862 FAIL2("Unknown condition for (?(%.2s", PL_regcomp_parse);
1866 FAIL("Sequence (? incomplete");
1871 while (*PL_regcomp_parse && strchr("iogcmsx", *PL_regcomp_parse)) {
1872 if (*PL_regcomp_parse != 'o')
1873 pmflag(flagsp, *PL_regcomp_parse);
1876 if (*PL_regcomp_parse == '-') {
1881 PL_regflags |= posflags;
1882 PL_regflags &= ~negflags;
1883 if (*PL_regcomp_parse == ':') {
1889 if (*PL_regcomp_parse != ')')
1890 FAIL2("Sequence (?%c...) not recognized", *PL_regcomp_parse);
1899 ret = reganode(OPEN, parno);
1906 /* Pick up the branches, linking them together. */
1907 br = regbranch(&flags, 1);
1910 if (*PL_regcomp_parse == '|') {
1911 if (!SIZE_ONLY && PL_extralen) {
1912 reginsert(BRANCHJ, br);
1915 reginsert(BRANCH, br);
1918 PL_extralen += 1; /* For BRANCHJ-BRANCH. */
1920 else if (paren == ':') {
1921 *flagp |= flags&SIMPLE;
1923 if (open) { /* Starts with OPEN. */
1924 regtail(ret, br); /* OPEN -> first. */
1926 else if (paren != '?') /* Not Conditional */
1930 *flagp |= flags&SPSTART;
1932 while (*PL_regcomp_parse == '|') {
1933 if (!SIZE_ONLY && PL_extralen) {
1934 ender = reganode(LONGJMP,0);
1935 regtail(NEXTOPER(NEXTOPER(lastbr)), ender); /* Append to the previous. */
1938 PL_extralen += 2; /* Account for LONGJMP. */
1940 br = regbranch(&flags, 0);
1943 regtail(lastbr, br); /* BRANCH -> BRANCH. */
1947 *flagp |= flags&SPSTART;
1950 if (have_branch || paren != ':') {
1951 /* Make a closing node, and hook it on the end. */
1954 ender = reg_node(TAIL);
1957 ender = reganode(CLOSE, parno);
1963 *flagp &= ~HASWIDTH;
1966 ender = reg_node(SUCCEED);
1969 ender = reg_node(END);
1972 regtail(lastbr, ender);
1975 /* Hook the tails of the branches to the closing node. */
1976 for (br = ret; br != NULL; br = regnext(br)) {
1977 regoptail(br, ender);
1984 static char parens[] = "=!<,>";
1986 if (paren && (p = strchr(parens, paren))) {
1987 int node = ((p - parens) % 2) ? UNLESSM : IFMATCH;
1988 int flag = (p - parens) > 1;
1991 node = SUSPEND, flag = 0;
1992 reginsert(node,ret);
1994 regtail(ret, reg_node(TAIL));
1998 /* Check for proper termination. */
2000 PL_regflags = oregflags;
2001 if (PL_regcomp_parse >= PL_regxend || *nextchar() != ')') {
2002 FAIL("unmatched () in regexp");
2005 else if (!paren && PL_regcomp_parse < PL_regxend) {
2006 if (*PL_regcomp_parse == ')') {
2007 FAIL("unmatched () in regexp");
2010 FAIL("junk on end of regexp"); /* "Can't happen". */
2018 - regbranch - one alternative of an | operator
2020 * Implements the concatenation operator.
2023 S_regbranch(pTHX_ I32 *flagp, I32 first)
2026 register regnode *ret;
2027 register regnode *chain = NULL;
2028 register regnode *latest;
2029 I32 flags = 0, c = 0;
2034 if (!SIZE_ONLY && PL_extralen)
2035 ret = reganode(BRANCHJ,0);
2037 ret = reg_node(BRANCH);
2040 if (!first && SIZE_ONLY)
2041 PL_extralen += 1; /* BRANCHJ */
2043 *flagp = WORST; /* Tentatively. */
2047 while (PL_regcomp_parse < PL_regxend && *PL_regcomp_parse != '|' && *PL_regcomp_parse != ')') {
2049 latest = regpiece(&flags);
2050 if (latest == NULL) {
2051 if (flags & TRYAGAIN)
2055 else if (ret == NULL)
2057 *flagp |= flags&HASWIDTH;
2058 if (chain == NULL) /* First piece. */
2059 *flagp |= flags&SPSTART;
2062 regtail(chain, latest);
2067 if (chain == NULL) { /* Loop ran zero times. */
2068 chain = reg_node(NOTHING);
2073 *flagp |= flags&SIMPLE;
2080 - regpiece - something followed by possible [*+?]
2082 * Note that the branching code sequences used for ? and the general cases
2083 * of * and + are somewhat optimized: they use the same NOTHING node as
2084 * both the endmarker for their branch list and the body of the last branch.
2085 * It might seem that this node could be dispensed with entirely, but the
2086 * endmarker role is not redundant.
2089 S_regpiece(pTHX_ I32 *flagp)
2092 register regnode *ret;
2094 register char *next;
2096 char *origparse = PL_regcomp_parse;
2099 I32 max = REG_INFTY;
2101 ret = regatom(&flags);
2103 if (flags & TRYAGAIN)
2108 op = *PL_regcomp_parse;
2110 if (op == '{' && regcurly(PL_regcomp_parse)) {
2111 next = PL_regcomp_parse + 1;
2113 while (isDIGIT(*next) || *next == ',') {
2122 if (*next == '}') { /* got one */
2126 min = atoi(PL_regcomp_parse);
2130 maxpos = PL_regcomp_parse;
2132 if (!max && *maxpos != '0')
2133 max = REG_INFTY; /* meaning "infinity" */
2134 else if (max >= REG_INFTY)
2135 FAIL2("Quantifier in {,} bigger than %d", REG_INFTY - 1);
2136 PL_regcomp_parse = next;
2140 if ((flags&SIMPLE)) {
2141 PL_regnaughty += 2 + PL_regnaughty / 2;
2142 reginsert(CURLY, ret);
2145 regnode *w = reg_node(WHILEM);
2149 if (!SIZE_ONLY && PL_extralen) {
2150 reginsert(LONGJMP,ret);
2151 reginsert(NOTHING,ret);
2152 NEXT_OFF(ret) = 3; /* Go over LONGJMP. */
2154 reginsert(CURLYX,ret);
2155 if (!SIZE_ONLY && PL_extralen)
2156 NEXT_OFF(ret) = 3; /* Go over NOTHING to LONGJMP. */
2157 regtail(ret, reg_node(NOTHING));
2159 PL_reg_whilem_seen++, PL_extralen += 3;
2160 PL_regnaughty += 4 + PL_regnaughty; /* compound interest */
2168 if (max && max < min)
2169 FAIL("Can't do {n,m} with n > m");
2184 #if 0 /* Now runtime fix should be reliable. */
2185 if (!(flags&HASWIDTH) && op != '?')
2186 FAIL("regexp *+ operand could be empty");
2191 *flagp = (op != '+') ? (WORST|SPSTART|HASWIDTH) : (WORST|HASWIDTH);
2193 if (op == '*' && (flags&SIMPLE)) {
2194 reginsert(STAR, ret);
2198 else if (op == '*') {
2202 else if (op == '+' && (flags&SIMPLE)) {
2203 reginsert(PLUS, ret);
2207 else if (op == '+') {
2211 else if (op == '?') {
2216 if (ckWARN(WARN_REGEXP) && !SIZE_ONLY && !(flags&HASWIDTH) && max > REG_INFTY/3) {
2217 Perl_warner(aTHX_ WARN_REGEXP, "%.*s matches null string many times",
2218 PL_regcomp_parse - origparse, origparse);
2221 if (*PL_regcomp_parse == '?') {
2223 reginsert(MINMOD, ret);
2224 regtail(ret, ret + NODE_STEP_REGNODE);
2226 if (ISMULT2(PL_regcomp_parse))
2227 FAIL("nested *?+ in regexp");
2233 - regatom - the lowest level
2235 * Optimization: gobbles an entire sequence of ordinary characters so that
2236 * it can turn them into a single node, which is smaller to store and
2237 * faster to run. Backslashed characters are exceptions, each becoming a
2238 * separate node; the code is simpler that way and it's not worth fixing.
2240 * [Yes, it is worth fixing, some scripts can run twice the speed.]
2243 S_regatom(pTHX_ I32 *flagp)
2246 register regnode *ret = 0;
2249 *flagp = WORST; /* Tentatively. */
2252 switch (*PL_regcomp_parse) {
2256 if (PL_regflags & PMf_MULTILINE)
2257 ret = reg_node(MBOL);
2258 else if (PL_regflags & PMf_SINGLELINE)
2259 ret = reg_node(SBOL);
2261 ret = reg_node(BOL);
2264 if (PL_regcomp_parse[1])
2267 if (PL_regflags & PMf_MULTILINE)
2268 ret = reg_node(MEOL);
2269 else if (PL_regflags & PMf_SINGLELINE)
2270 ret = reg_node(SEOL);
2272 ret = reg_node(EOL);
2277 if (PL_regflags & PMf_SINGLELINE)
2278 ret = reg_node(SANYUTF8);
2280 ret = reg_node(ANYUTF8);
2284 if (PL_regflags & PMf_SINGLELINE)
2285 ret = reg_node(SANY);
2287 ret = reg_node(REG_ANY);
2288 *flagp |= HASWIDTH|SIMPLE;
2294 ret = (UTF ? regclassutf8() : regclass());
2295 if (*PL_regcomp_parse != ']')
2296 FAIL("unmatched [] in regexp");
2298 *flagp |= HASWIDTH|SIMPLE;
2302 ret = reg(1, &flags);
2304 if (flags & TRYAGAIN) {
2305 if (PL_regcomp_parse == PL_regxend) {
2306 /* Make parent create an empty node if needed. */
2314 *flagp |= flags&(HASWIDTH|SPSTART|SIMPLE);
2318 if (flags & TRYAGAIN) {
2322 FAIL2("internal urp in regexp at /%s/", PL_regcomp_parse);
2323 /* Supposed to be caught earlier. */
2326 if (!regcurly(PL_regcomp_parse)) {
2334 FAIL("?+*{} follows nothing in regexp");
2337 switch (*++PL_regcomp_parse) {
2340 ret = reg_node(SBOL);
2345 ret = reg_node(GPOS);
2346 PL_regseen |= REG_SEEN_GPOS;
2351 ret = reg_node(SEOL);
2356 ret = reg_node(EOS);
2358 PL_seen_zerolen++; /* Do not optimize RE away */
2362 ret = reg_node(SANY);
2363 *flagp |= HASWIDTH|SIMPLE;
2367 ret = reg_node(CLUMP);
2370 if (UTF && !PL_utf8_mark)
2371 is_utf8_mark((U8*)"~"); /* preload table */
2376 ? (LOC ? ALNUMLUTF8 : ALNUMUTF8)
2377 : (LOC ? ALNUML : ALNUM));
2378 *flagp |= HASWIDTH|SIMPLE;
2380 if (UTF && !PL_utf8_alnum)
2381 is_utf8_alnum((U8*)"a"); /* preload table */
2386 ? (LOC ? NALNUMLUTF8 : NALNUMUTF8)
2387 : (LOC ? NALNUML : NALNUM));
2388 *flagp |= HASWIDTH|SIMPLE;
2390 if (UTF && !PL_utf8_alnum)
2391 is_utf8_alnum((U8*)"a"); /* preload table */
2395 PL_regseen |= REG_SEEN_LOOKBEHIND;
2398 ? (LOC ? BOUNDLUTF8 : BOUNDUTF8)
2399 : (LOC ? BOUNDL : BOUND));
2402 if (UTF && !PL_utf8_alnum)
2403 is_utf8_alnum((U8*)"a"); /* preload table */
2407 PL_regseen |= REG_SEEN_LOOKBEHIND;
2410 ? (LOC ? NBOUNDLUTF8 : NBOUNDUTF8)
2411 : (LOC ? NBOUNDL : NBOUND));
2414 if (UTF && !PL_utf8_alnum)
2415 is_utf8_alnum((U8*)"a"); /* preload table */
2420 ? (LOC ? SPACELUTF8 : SPACEUTF8)
2421 : (LOC ? SPACEL : SPACE));
2422 *flagp |= HASWIDTH|SIMPLE;
2424 if (UTF && !PL_utf8_space)
2425 is_utf8_space((U8*)" "); /* preload table */
2430 ? (LOC ? NSPACELUTF8 : NSPACEUTF8)
2431 : (LOC ? NSPACEL : NSPACE));
2432 *flagp |= HASWIDTH|SIMPLE;
2434 if (UTF && !PL_utf8_space)
2435 is_utf8_space((U8*)" "); /* preload table */
2438 ret = reg_node(UTF ? DIGITUTF8 : DIGIT);
2439 *flagp |= HASWIDTH|SIMPLE;
2441 if (UTF && !PL_utf8_digit)
2442 is_utf8_digit((U8*)"1"); /* preload table */
2445 ret = reg_node(UTF ? NDIGITUTF8 : NDIGIT);
2446 *flagp |= HASWIDTH|SIMPLE;
2448 if (UTF && !PL_utf8_digit)
2449 is_utf8_digit((U8*)"1"); /* preload table */
2453 { /* a lovely hack--pretend we saw [\pX] instead */
2454 char* oldregxend = PL_regxend;
2456 if (PL_regcomp_parse[1] == '{') {
2457 PL_regxend = strchr(PL_regcomp_parse, '}');
2459 FAIL("Missing right brace on \\p{}");
2463 PL_regxend = PL_regcomp_parse + 2;
2466 ret = regclassutf8();
2468 PL_regxend = oldregxend;
2471 *flagp |= HASWIDTH|SIMPLE;
2484 case '1': case '2': case '3': case '4':
2485 case '5': case '6': case '7': case '8': case '9':
2487 I32 num = atoi(PL_regcomp_parse);
2489 if (num > 9 && num >= PL_regnpar)
2492 if (!SIZE_ONLY && num > PL_regcomp_rx->nparens)
2493 FAIL("reference to nonexistent group");
2496 ? (LOC ? REFFL : REFF)
2499 while (isDIGIT(*PL_regcomp_parse))
2507 if (PL_regcomp_parse >= PL_regxend)
2508 FAIL("trailing \\ in regexp");
2511 /* Do not generate `unrecognized' warnings here, we fall
2512 back into the quick-grab loop below */
2518 if (PL_regflags & PMf_EXTENDED) {
2519 while (PL_regcomp_parse < PL_regxend && *PL_regcomp_parse != '\n') PL_regcomp_parse++;
2520 if (PL_regcomp_parse < PL_regxend)
2536 ? (LOC ? EXACTFL : EXACTF)
2539 for (len = 0, p = PL_regcomp_parse - 1;
2540 len < 127 && p < PL_regxend;
2545 if (PL_regflags & PMf_EXTENDED)
2546 p = regwhite(p, PL_regxend);
2608 char* e = strchr(p, '}');
2611 FAIL("Missing right brace on \\x{}");
2613 numlen = 1; /* allow underscores */
2614 ender = (UV)scan_hex(p + 1, e - p - 1, &numlen);
2615 /* numlen is generous */
2616 if (numlen + len >= 127) {
2623 FAIL("Can't use \\x{} without 'use utf8' declaration");
2626 numlen = 0; /* disallow underscores */
2627 ender = (UV)scan_hex(p, 2, &numlen);
2633 ender = UCHARAT(p++);
2634 ender = toCTRL(ender);
2636 case '0': case '1': case '2': case '3':case '4':
2637 case '5': case '6': case '7': case '8':case '9':
2639 (isDIGIT(p[1]) && atoi(p) >= PL_regnpar) ) {
2640 numlen = 0; /* disallow underscores */
2641 ender = (UV)scan_oct(p, 3, &numlen);
2650 if (p >= PL_regxend)
2651 FAIL("trailing \\ in regexp");
2654 if (!SIZE_ONLY && ckWARN(WARN_REGEXP) && isALPHA(*p))
2655 Perl_warner(aTHX_ WARN_REGEXP,
2656 "/%.127s/: Unrecognized escape \\%c passed through",
2659 goto normal_default;
2664 if ((*p & 0xc0) == 0xc0 && UTF) {
2665 ender = utf8_to_uv((U8*)p, &numlen);
2672 if (PL_regflags & PMf_EXTENDED)
2673 p = regwhite(p, PL_regxend);
2676 ender = toLOWER_LC_uni(ender);
2678 ender = toLOWER_uni(ender);
2680 if (ISMULT2(p)) { /* Back off on ?+*. */
2683 else if (ender >= 0x80 && UTF) {
2684 reguni(ender, s, &numlen);
2694 if (ender >= 0x80 && UTF) {
2695 reguni(ender, s, &numlen);
2703 PL_regcomp_parse = p - 1;
2706 FAIL("internal disaster in regexp");
2714 PL_regsize += STR_SZ(len);
2716 PL_regcode += STR_SZ(len);
2725 S_regwhite(pTHX_ char *p, char *e)
2730 else if (*p == '#') {
2733 } while (p < e && *p != '\n');
2741 /* Parse POSIX character classes: [[:foo:]], [[=foo=]], [[.foo.]].
2742 Character classes ([:foo:]) can also be negated ([:^foo:]).
2743 Returns a named class id (ANYOF_XXX) if successful, -1 otherwise.
2744 Equivalence classes ([=foo=]) and composites ([.foo.]) are parsed,
2745 but trigger warnings because they are currently unimplemented. */
2747 S_regpposixcc(pTHX_ I32 value)
2751 I32 namedclass = OOB_NAMEDCLASS;
2753 if (value == '[' && PL_regcomp_parse + 1 < PL_regxend &&
2754 /* I smell either [: or [= or [. -- POSIX has been here, right? */
2755 (*PL_regcomp_parse == ':' ||
2756 *PL_regcomp_parse == '=' ||
2757 *PL_regcomp_parse == '.')) {
2758 char c = *PL_regcomp_parse;
2759 char* s = PL_regcomp_parse++;
2761 while (PL_regcomp_parse < PL_regxend && *PL_regcomp_parse != c)
2763 if (PL_regcomp_parse == PL_regxend)
2764 /* Grandfather lone [:, [=, [. */
2765 PL_regcomp_parse = s;
2767 char* t = PL_regcomp_parse++; /* skip over the c */
2769 if (*PL_regcomp_parse == ']') {
2770 PL_regcomp_parse++; /* skip over the ending ] */
2773 I32 complement = *posixcc == '^' ? *posixcc++ : 0;
2774 I32 skip = 5; /* the most common skip */
2778 if (strnEQ(posixcc, "alnum", 5))
2780 complement ? ANYOF_NALNUMC : ANYOF_ALNUMC;
2781 else if (strnEQ(posixcc, "alpha", 5))
2783 complement ? ANYOF_NALPHA : ANYOF_ALPHA;
2784 else if (strnEQ(posixcc, "ascii", 5))
2786 complement ? ANYOF_NASCII : ANYOF_ASCII;
2789 if (strnEQ(posixcc, "cntrl", 5))
2791 complement ? ANYOF_NCNTRL : ANYOF_CNTRL;
2794 if (strnEQ(posixcc, "digit", 5))
2796 complement ? ANYOF_NDIGIT : ANYOF_DIGIT;
2799 if (strnEQ(posixcc, "graph", 5))
2801 complement ? ANYOF_NGRAPH : ANYOF_GRAPH;
2804 if (strnEQ(posixcc, "lower", 5))
2806 complement ? ANYOF_NLOWER : ANYOF_LOWER;
2809 if (strnEQ(posixcc, "print", 5))
2811 complement ? ANYOF_NPRINT : ANYOF_PRINT;
2812 else if (strnEQ(posixcc, "punct", 5))
2814 complement ? ANYOF_NPUNCT : ANYOF_PUNCT;
2817 if (strnEQ(posixcc, "space", 5))
2819 complement ? ANYOF_NSPACE : ANYOF_SPACE;
2822 if (strnEQ(posixcc, "upper", 5))
2824 complement ? ANYOF_NUPPER : ANYOF_UPPER;
2826 case 'w': /* this is not POSIX, this is the Perl \w */
2827 if (strnEQ(posixcc, "word", 4)) {
2829 complement ? ANYOF_NALNUM : ANYOF_ALNUM;
2834 if (strnEQ(posixcc, "xdigit", 6)) {
2836 complement ? ANYOF_NXDIGIT : ANYOF_XDIGIT;
2841 if (namedclass == OOB_NAMEDCLASS ||
2842 posixcc[skip] != ':' ||
2843 posixcc[skip+1] != ']')
2845 "Character class [:%.*s:] unknown",
2847 } else if (!SIZE_ONLY)
2848 /* [[=foo=]] and [[.foo.]] are still future. */
2850 "Character class syntax [%c %c] is reserved for future extensions", c, c);
2852 /* Maternal grandfather:
2853 * "[:" ending in ":" but not in ":]" */
2854 PL_regcomp_parse = s;
2863 S_checkposixcc(pTHX)
2865 if (!SIZE_ONLY && ckWARN(WARN_REGEXP) &&
2866 (*PL_regcomp_parse == ':' ||
2867 *PL_regcomp_parse == '=' ||
2868 *PL_regcomp_parse == '.')) {
2869 char *s = PL_regcomp_parse;
2872 while(*s && isALNUM(*s))
2874 if (*s && c == *s && s[1] == ']') {
2875 Perl_warner(aTHX_ WARN_REGEXP,
2876 "Character class syntax [%c %c] belongs inside character classes", c, c);
2877 if (c == '=' || c == '.')
2879 "Character class syntax [%c %c] is reserved for future extensions", c, c);
2889 register I32 lastvalue = OOB_CHAR8;
2890 register I32 range = 0;
2891 register regnode *ret;
2895 bool need_class = 0;
2897 ret = reg_node(ANYOF);
2899 PL_regsize += ANYOF_SKIP;
2902 ANYOF_BITMAP_ZERO(ret);
2903 PL_regcode += ANYOF_SKIP;
2905 ANYOF_FLAGS(ret) |= ANYOF_FOLD;
2907 ANYOF_FLAGS(ret) |= ANYOF_LOCALE;
2909 if (*PL_regcomp_parse == '^') { /* Complement of range. */
2913 ANYOF_FLAGS(ret) |= ANYOF_INVERT;
2916 if (!SIZE_ONLY && ckWARN(WARN_REGEXP))
2919 if (*PL_regcomp_parse == ']' || *PL_regcomp_parse == '-')
2920 goto skipcond; /* allow 1st char to be ] or - */
2921 while (PL_regcomp_parse < PL_regxend && *PL_regcomp_parse != ']') {
2923 namedclass = OOB_NAMEDCLASS;
2925 rangebegin = PL_regcomp_parse;
2926 value = UCHARAT(PL_regcomp_parse++);
2928 namedclass = regpposixcc(value);
2929 else if (value == '\\') {
2930 value = UCHARAT(PL_regcomp_parse++);
2931 /* Some compilers cannot handle switching on 64-bit integer
2932 * values, therefore value cannot be an UV. --jhi */
2934 case 'w': namedclass = ANYOF_ALNUM; break;
2935 case 'W': namedclass = ANYOF_NALNUM; break;
2936 case 's': namedclass = ANYOF_SPACE; break;
2937 case 'S': namedclass = ANYOF_NSPACE; break;
2938 case 'd': namedclass = ANYOF_DIGIT; break;
2939 case 'D': namedclass = ANYOF_NDIGIT; break;
2940 case 'n': value = '\n'; break;
2941 case 'r': value = '\r'; break;
2942 case 't': value = '\t'; break;
2943 case 'f': value = '\f'; break;
2944 case 'b': value = '\b'; break;
2946 case 'e': value = '\033'; break;
2947 case 'a': value = '\007'; break;
2949 case 'e': value = '\047'; break;
2950 case 'a': value = '\057'; break;
2953 numlen = 0; /* disallow underscores */
2954 value = (UV)scan_hex(PL_regcomp_parse, 2, &numlen);
2955 PL_regcomp_parse += numlen;
2958 value = UCHARAT(PL_regcomp_parse++);
2959 value = toCTRL(value);
2961 case '0': case '1': case '2': case '3': case '4':
2962 case '5': case '6': case '7': case '8': case '9':
2963 numlen = 0; /* disallow underscores */
2964 value = (UV)scan_oct(--PL_regcomp_parse, 3, &numlen);
2965 PL_regcomp_parse += numlen;
2968 if (!SIZE_ONLY && ckWARN(WARN_REGEXP) && isALPHA(value))
2969 Perl_warner(aTHX_ WARN_REGEXP,
2970 "/%.127s/: Unrecognized escape \\%c in character class passed through",
2976 if (namedclass > OOB_NAMEDCLASS) {
2977 if (!need_class && !SIZE_ONLY)
2978 ANYOF_CLASS_ZERO(ret);
2980 if (range) { /* a-\d, a-[:digit:] */
2982 if (ckWARN(WARN_REGEXP))
2983 Perl_warner(aTHX_ WARN_REGEXP,
2984 "/%.127s/: false [] range \"%*.*s\" in regexp",
2986 PL_regcomp_parse - rangebegin,
2987 PL_regcomp_parse - rangebegin,
2989 ANYOF_BITMAP_SET(ret, lastvalue);
2990 ANYOF_BITMAP_SET(ret, '-');
2992 range = 0; /* this is not a true range */
2995 switch (namedclass) {
2998 ANYOF_CLASS_SET(ret, ANYOF_ALNUM);
3000 for (value = 0; value < 256; value++)
3002 ANYOF_BITMAP_SET(ret, value);
3007 ANYOF_CLASS_SET(ret, ANYOF_NALNUM);
3009 for (value = 0; value < 256; value++)
3010 if (!isALNUM(value))
3011 ANYOF_BITMAP_SET(ret, value);
3016 ANYOF_CLASS_SET(ret, ANYOF_SPACE);
3018 for (value = 0; value < 256; value++)
3020 ANYOF_BITMAP_SET(ret, value);
3025 ANYOF_CLASS_SET(ret, ANYOF_NSPACE);
3027 for (value = 0; value < 256; value++)
3028 if (!isSPACE(value))
3029 ANYOF_BITMAP_SET(ret, value);
3034 ANYOF_CLASS_SET(ret, ANYOF_DIGIT);
3036 for (value = '0'; value <= '9'; value++)
3037 ANYOF_BITMAP_SET(ret, value);
3042 ANYOF_CLASS_SET(ret, ANYOF_NDIGIT);
3044 for (value = 0; value < '0'; value++)
3045 ANYOF_BITMAP_SET(ret, value);
3046 for (value = '9' + 1; value < 256; value++)
3047 ANYOF_BITMAP_SET(ret, value);
3052 ANYOF_CLASS_SET(ret, ANYOF_NALNUMC);
3054 for (value = 0; value < 256; value++)
3055 if (!isALNUMC(value))
3056 ANYOF_BITMAP_SET(ret, value);
3061 ANYOF_CLASS_SET(ret, ANYOF_ALNUMC);
3063 for (value = 0; value < 256; value++)
3064 if (isALNUMC(value))
3065 ANYOF_BITMAP_SET(ret, value);
3070 ANYOF_CLASS_SET(ret, ANYOF_ALPHA);
3072 for (value = 0; value < 256; value++)
3074 ANYOF_BITMAP_SET(ret, value);
3079 ANYOF_CLASS_SET(ret, ANYOF_NALPHA);
3081 for (value = 0; value < 256; value++)
3082 if (!isALPHA(value))
3083 ANYOF_BITMAP_SET(ret, value);
3088 ANYOF_CLASS_SET(ret, ANYOF_ASCII);
3091 for (value = 0; value < 128; value++)
3092 ANYOF_BITMAP_SET(ret, value);
3094 for (value = 0; value < 256; value++)
3096 ANYOF_BITMAP_SET(ret, value);
3102 ANYOF_CLASS_SET(ret, ANYOF_NASCII);
3105 for (value = 128; value < 256; value++)
3106 ANYOF_BITMAP_SET(ret, value);
3108 for (value = 0; value < 256; value++)
3109 if (!isASCII(value))
3110 ANYOF_BITMAP_SET(ret, value);
3116 ANYOF_CLASS_SET(ret, ANYOF_CNTRL);
3118 for (value = 0; value < 256; value++)
3120 ANYOF_BITMAP_SET(ret, value);
3122 lastvalue = OOB_CHAR8;
3126 ANYOF_CLASS_SET(ret, ANYOF_NCNTRL);
3128 for (value = 0; value < 256; value++)
3129 if (!isCNTRL(value))
3130 ANYOF_BITMAP_SET(ret, value);
3135 ANYOF_CLASS_SET(ret, ANYOF_GRAPH);
3137 for (value = 0; value < 256; value++)
3139 ANYOF_BITMAP_SET(ret, value);
3144 ANYOF_CLASS_SET(ret, ANYOF_NGRAPH);
3146 for (value = 0; value < 256; value++)
3147 if (!isGRAPH(value))
3148 ANYOF_BITMAP_SET(ret, value);
3153 ANYOF_CLASS_SET(ret, ANYOF_LOWER);
3155 for (value = 0; value < 256; value++)
3157 ANYOF_BITMAP_SET(ret, value);
3162 ANYOF_CLASS_SET(ret, ANYOF_NLOWER);
3164 for (value = 0; value < 256; value++)
3165 if (!isLOWER(value))
3166 ANYOF_BITMAP_SET(ret, value);
3171 ANYOF_CLASS_SET(ret, ANYOF_PRINT);
3173 for (value = 0; value < 256; value++)
3175 ANYOF_BITMAP_SET(ret, value);
3180 ANYOF_CLASS_SET(ret, ANYOF_NPRINT);
3182 for (value = 0; value < 256; value++)
3183 if (!isPRINT(value))
3184 ANYOF_BITMAP_SET(ret, value);
3189 ANYOF_CLASS_SET(ret, ANYOF_PUNCT);
3191 for (value = 0; value < 256; value++)
3193 ANYOF_BITMAP_SET(ret, value);
3198 ANYOF_CLASS_SET(ret, ANYOF_NPUNCT);
3200 for (value = 0; value < 256; value++)
3201 if (!isPUNCT(value))
3202 ANYOF_BITMAP_SET(ret, value);
3207 ANYOF_CLASS_SET(ret, ANYOF_UPPER);
3209 for (value = 0; value < 256; value++)
3211 ANYOF_BITMAP_SET(ret, value);
3216 ANYOF_CLASS_SET(ret, ANYOF_NUPPER);
3218 for (value = 0; value < 256; value++)
3219 if (!isUPPER(value))
3220 ANYOF_BITMAP_SET(ret, value);
3225 ANYOF_CLASS_SET(ret, ANYOF_XDIGIT);
3227 for (value = 0; value < 256; value++)
3228 if (isXDIGIT(value))
3229 ANYOF_BITMAP_SET(ret, value);
3234 ANYOF_CLASS_SET(ret, ANYOF_NXDIGIT);
3236 for (value = 0; value < 256; value++)
3237 if (!isXDIGIT(value))
3238 ANYOF_BITMAP_SET(ret, value);
3242 FAIL("invalid [::] class in regexp");
3246 ANYOF_FLAGS(ret) |= ANYOF_CLASS;
3251 if (lastvalue > value) /* b-a */ {
3253 "/%.127s/: invalid [] range \"%*.*s\" in regexp",
3255 PL_regcomp_parse - rangebegin,
3256 PL_regcomp_parse - rangebegin,
3263 if (*PL_regcomp_parse == '-' && PL_regcomp_parse+1 < PL_regxend &&
3264 PL_regcomp_parse[1] != ']') {
3266 if (namedclass > OOB_NAMEDCLASS) { /* \w-, [:word:]- */
3267 if (ckWARN(WARN_REGEXP))
3268 Perl_warner(aTHX_ WARN_REGEXP,
3269 "/%.127s/: false [] range \"%*.*s\" in regexp",
3271 PL_regcomp_parse - rangebegin,
3272 PL_regcomp_parse - rangebegin,
3275 ANYOF_BITMAP_SET(ret, '-');
3278 continue; /* do it next time */
3281 /* now is the next time */
3283 #ifndef ASCIIish /* EBCDIC, for example. */
3284 if ((isLOWER(lastvalue) && isLOWER(value)) ||
3285 (isUPPER(lastvalue) && isUPPER(value)))
3288 if (isLOWER(lastvalue)) {
3289 for (i = lastvalue; i <= value; i++)
3291 ANYOF_BITMAP_SET(ret, i);
3293 for (i = lastvalue; i <= value; i++)
3295 ANYOF_BITMAP_SET(ret, i);
3300 for ( ; lastvalue <= value; lastvalue++)
3301 ANYOF_BITMAP_SET(ret, lastvalue);
3307 PL_regsize += ANYOF_CLASS_ADD_SKIP;
3309 PL_regcode += ANYOF_CLASS_ADD_SKIP;
3311 /* optimize case-insensitive simple patterns (e.g. /[a-z]/i) */
3313 (ANYOF_FLAGS(ret) & (ANYOF_FLAGS_ALL ^ ANYOF_INVERT)) == ANYOF_FOLD) {
3314 for (value = 0; value < 256; ++value) {
3315 if (ANYOF_BITMAP_TEST(ret, value)) {
3316 I32 cf = PL_fold[value];
3317 ANYOF_BITMAP_SET(ret, cf);
3320 ANYOF_FLAGS(ret) &= ~ANYOF_FOLD;
3322 /* optimize inverted simple patterns (e.g. [^a-z]) */
3323 if (!SIZE_ONLY && (ANYOF_FLAGS(ret) & ANYOF_FLAGS_ALL) == ANYOF_INVERT) {
3324 for (value = 0; value < ANYOF_BITMAP_SIZE; ++value)
3325 ANYOF_BITMAP(ret)[value] ^= ANYOF_FLAGS_ALL;
3326 ANYOF_FLAGS(ret) = 0;
3332 S_regclassutf8(pTHX)
3337 register U32 lastvalue = OOB_UTF8;
3338 register I32 range = 0;
3339 register regnode *ret;
3347 if (*PL_regcomp_parse == '^') { /* Complement of range. */
3351 flags |= ANYOF_INVERT;
3355 flags |= ANYOF_FOLD;
3357 flags |= ANYOF_LOCALE;
3358 listsv = newSVpvn("# comment\n",10);
3361 if (!SIZE_ONLY && ckWARN(WARN_REGEXP))
3364 if (*PL_regcomp_parse == ']' || *PL_regcomp_parse == '-')
3365 goto skipcond; /* allow 1st char to be ] or - */
3367 while (PL_regcomp_parse < PL_regxend && *PL_regcomp_parse != ']') {
3369 namedclass = OOB_NAMEDCLASS;
3371 rangebegin = PL_regcomp_parse;
3372 value = utf8_to_uv((U8*)PL_regcomp_parse, &numlen);
3373 PL_regcomp_parse += numlen;
3375 namedclass = regpposixcc(value);
3376 else if (value == '\\') {
3377 value = (U32)utf8_to_uv((U8*)PL_regcomp_parse, &numlen);
3378 PL_regcomp_parse += numlen;
3379 /* Some compilers cannot handle switching on 64-bit integer
3380 * values, therefore value cannot be an UV. Yes, this will
3381 * be a problem later if we want switch on Unicode. --jhi */
3383 case 'w': namedclass = ANYOF_ALNUM; break;
3384 case 'W': namedclass = ANYOF_NALNUM; break;
3385 case 's': namedclass = ANYOF_SPACE; break;
3386 case 'S': namedclass = ANYOF_NSPACE; break;
3387 case 'd': namedclass = ANYOF_DIGIT; break;
3388 case 'D': namedclass = ANYOF_NDIGIT; break;
3391 if (*PL_regcomp_parse == '{') {
3392 e = strchr(PL_regcomp_parse++, '}');
3394 FAIL("Missing right brace on \\p{}");
3395 n = e - PL_regcomp_parse;
3398 e = PL_regcomp_parse;
3403 Perl_sv_catpvf(aTHX_ listsv,
3404 "+utf8::%.*s\n", (int)n, PL_regcomp_parse);
3406 Perl_sv_catpvf(aTHX_ listsv,
3407 "!utf8::%.*s\n", (int)n, PL_regcomp_parse);
3409 PL_regcomp_parse = e + 1;
3410 lastvalue = OOB_UTF8;
3412 case 'n': value = '\n'; break;
3413 case 'r': value = '\r'; break;
3414 case 't': value = '\t'; break;
3415 case 'f': value = '\f'; break;
3416 case 'b': value = '\b'; break;
3418 case 'e': value = '\033'; break;
3419 case 'a': value = '\007'; break;
3421 case 'e': value = '\047'; break;
3422 case 'a': value = '\057'; break;
3425 if (*PL_regcomp_parse == '{') {
3426 e = strchr(PL_regcomp_parse++, '}');
3428 FAIL("Missing right brace on \\x{}");
3429 numlen = 1; /* allow underscores */
3430 value = (UV)scan_hex(PL_regcomp_parse,
3431 e - PL_regcomp_parse,
3433 PL_regcomp_parse = e + 1;
3436 numlen = 0; /* disallow underscores */
3437 value = (UV)scan_hex(PL_regcomp_parse, 2, &numlen);
3438 PL_regcomp_parse += numlen;
3442 value = UCHARAT(PL_regcomp_parse++);
3443 value = toCTRL(value);
3445 case '0': case '1': case '2': case '3': case '4':
3446 case '5': case '6': case '7': case '8': case '9':
3447 numlen = 0; /* disallow underscores */
3448 value = (UV)scan_oct(--PL_regcomp_parse, 3, &numlen);
3449 PL_regcomp_parse += numlen;
3452 if (!SIZE_ONLY && ckWARN(WARN_REGEXP) && isALPHA(value))
3453 Perl_warner(aTHX_ WARN_REGEXP,
3454 "/%.127s/: Unrecognized escape \\%c in character class passed through",
3460 if (namedclass > OOB_NAMEDCLASS) {
3461 if (range) { /* a-\d, a-[:digit:] */
3463 if (ckWARN(WARN_REGEXP))
3464 Perl_warner(aTHX_ WARN_REGEXP,
3465 "/%.127s/: false [] range \"%*.*s\" in regexp",
3467 PL_regcomp_parse - rangebegin,
3468 PL_regcomp_parse - rangebegin,
3470 Perl_sv_catpvf(aTHX_ listsv,
3471 /* 0x002D is Unicode for '-' */
3472 "%04"UVxf"\n002D\n", (UV)lastvalue);
3477 switch (namedclass) {
3479 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsWord\n"); break;
3481 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsWord\n"); break;
3483 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsAlnum\n"); break;
3485 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsAlnum\n"); break;
3487 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsAlpha\n"); break;
3489 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsAlpha\n"); break;
3491 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsASCII\n"); break;
3493 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsASCII\n"); break;
3495 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsCntrl\n"); break;
3497 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsCntrl\n"); break;
3499 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsGraph\n"); break;
3501 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsGraph\n"); break;
3503 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsDigit\n"); break;
3505 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsDigit\n"); break;
3507 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsLower\n"); break;
3509 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsLower\n"); break;
3511 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsPrint\n"); break;
3513 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsPrint\n"); break;
3515 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsPunct\n"); break;
3517 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsPunct\n"); break;
3519 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsSpace\n"); break;
3521 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsSpace\n"); break;
3523 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsUpper\n"); break;
3525 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsUpper\n"); break;
3527 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsXDigit\n"); break;
3529 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsXDigit\n"); break;
3535 if (lastvalue > value) { /* b-a */
3537 "/%.127s/: invalid [] range \"%*.*s\" in regexp",
3539 PL_regcomp_parse - rangebegin,
3540 PL_regcomp_parse - rangebegin,
3547 if (*PL_regcomp_parse == '-' && PL_regcomp_parse+1 < PL_regxend &&
3548 PL_regcomp_parse[1] != ']') {
3550 if (namedclass > OOB_NAMEDCLASS) { /* \w-, [:word:]- */
3551 if (ckWARN(WARN_REGEXP))
3552 Perl_warner(aTHX_ WARN_REGEXP,
3553 "/%.127s/: false [] range \"%*.*s\" in regexp",
3555 PL_regcomp_parse - rangebegin,
3556 PL_regcomp_parse - rangebegin,
3559 Perl_sv_catpvf(aTHX_ listsv,
3560 /* 0x002D is Unicode for '-' */
3564 continue; /* do it next time */
3567 /* now is the next time */
3569 Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\t%04"UVxf"\n",
3570 (UV)lastvalue, (UV)value);
3574 ret = reganode(ANYOFUTF8, 0);
3577 SV *rv = swash_init("utf8", "", listsv, 1, 0);
3578 SvREFCNT_dec(listsv);
3579 n = add_data(1,"s");
3580 PL_regcomp_rx->data->data[n] = (void*)rv;
3581 ARG1_SET(ret, flags);
3592 char* retval = PL_regcomp_parse++;
3595 if (*PL_regcomp_parse == '(' && PL_regcomp_parse[1] == '?' &&
3596 PL_regcomp_parse[2] == '#') {
3597 while (*PL_regcomp_parse && *PL_regcomp_parse != ')')
3602 if (PL_regflags & PMf_EXTENDED) {
3603 if (isSPACE(*PL_regcomp_parse)) {
3607 else if (*PL_regcomp_parse == '#') {
3608 while (*PL_regcomp_parse && *PL_regcomp_parse != '\n')
3619 - reg_node - emit a node
3621 STATIC regnode * /* Location. */
3622 S_reg_node(pTHX_ U8 op)
3625 register regnode *ret;
3626 register regnode *ptr;
3630 SIZE_ALIGN(PL_regsize);
3635 NODE_ALIGN_FILL(ret);
3637 FILL_ADVANCE_NODE(ptr, op);
3644 - reganode - emit a node with an argument
3646 STATIC regnode * /* Location. */
3647 S_reganode(pTHX_ U8 op, U32 arg)
3650 register regnode *ret;
3651 register regnode *ptr;
3655 SIZE_ALIGN(PL_regsize);
3660 NODE_ALIGN_FILL(ret);
3662 FILL_ADVANCE_NODE_ARG(ptr, op, arg);
3669 - reguni - emit (if appropriate) a Unicode character
3672 S_reguni(pTHX_ UV uv, char* s, I32* lenp)
3676 U8 tmpbuf[UTF8_MAXLEN];
3677 *lenp = uv_to_utf8(tmpbuf, uv) - tmpbuf;
3680 *lenp = uv_to_utf8((U8*)s, uv) - (U8*)s;
3685 - reginsert - insert an operator in front of already-emitted operand
3687 * Means relocating the operand.
3690 S_reginsert(pTHX_ U8 op, regnode *opnd)
3693 register regnode *src;
3694 register regnode *dst;
3695 register regnode *place;
3696 register int offset = regarglen[(U8)op];
3698 /* (PL_regkind[(U8)op] == CURLY ? EXTRA_STEP_2ARGS : 0); */
3701 PL_regsize += NODE_STEP_REGNODE + offset;
3706 PL_regcode += NODE_STEP_REGNODE + offset;
3709 StructCopy(--src, --dst, regnode);
3711 place = opnd; /* Op node, where operand used to be. */
3712 src = NEXTOPER(place);
3713 FILL_ADVANCE_NODE(place, op);
3714 Zero(src, offset, regnode);
3718 - regtail - set the next-pointer at the end of a node chain of p to val.
3721 S_regtail(pTHX_ regnode *p, regnode *val)
3724 register regnode *scan;
3725 register regnode *temp;
3730 /* Find last node. */
3733 temp = regnext(scan);
3739 if (reg_off_by_arg[OP(scan)]) {
3740 ARG_SET(scan, val - scan);
3743 NEXT_OFF(scan) = val - scan;
3748 - regoptail - regtail on operand of first argument; nop if operandless
3751 S_regoptail(pTHX_ regnode *p, regnode *val)
3754 /* "Operandless" and "op != BRANCH" are synonymous in practice. */
3755 if (p == NULL || SIZE_ONLY)
3757 if (PL_regkind[(U8)OP(p)] == BRANCH) {
3758 regtail(NEXTOPER(p), val);
3760 else if ( PL_regkind[(U8)OP(p)] == BRANCHJ) {
3761 regtail(NEXTOPER(NEXTOPER(p)), val);
3768 - regcurly - a little FSA that accepts {\d+,?\d*}
3771 S_regcurly(pTHX_ register char *s)
3790 S_dumpuntil(pTHX_ regnode *start, regnode *node, regnode *last, SV* sv, I32 l)
3793 register U8 op = EXACT; /* Arbitrary non-END op. */
3794 register regnode *next;
3796 while (op != END && (!last || node < last)) {
3797 /* While that wasn't END last time... */
3803 next = regnext(node);
3805 if (OP(node) == OPTIMIZED)
3808 PerlIO_printf(Perl_debug_log, "%4"IVdf":%*s%s", (IV)(node - start),
3809 (int)(2*l + 1), "", SvPVX(sv));
3810 if (next == NULL) /* Next ptr. */
3811 PerlIO_printf(Perl_debug_log, "(0)");
3813 PerlIO_printf(Perl_debug_log, "(%"IVdf")", (IV)(next - start));
3814 (void)PerlIO_putc(Perl_debug_log, '\n');
3816 if (PL_regkind[(U8)op] == BRANCHJ) {
3817 register regnode *nnode = (OP(next) == LONGJMP
3820 if (last && nnode > last)
3822 node = dumpuntil(start, NEXTOPER(NEXTOPER(node)), nnode, sv, l + 1);
3824 else if (PL_regkind[(U8)op] == BRANCH) {
3825 node = dumpuntil(start, NEXTOPER(node), next, sv, l + 1);
3827 else if ( op == CURLY) { /* `next' might be very big: optimizer */
3828 node = dumpuntil(start, NEXTOPER(node) + EXTRA_STEP_2ARGS,
3829 NEXTOPER(node) + EXTRA_STEP_2ARGS + 1, sv, l + 1);
3831 else if (PL_regkind[(U8)op] == CURLY && op != CURLYX) {
3832 node = dumpuntil(start, NEXTOPER(node) + EXTRA_STEP_2ARGS,
3835 else if ( op == PLUS || op == STAR) {
3836 node = dumpuntil(start, NEXTOPER(node), NEXTOPER(node) + 1, sv, l + 1);
3838 else if (op == ANYOF) {
3839 node = NEXTOPER(node);
3842 else if (PL_regkind[(U8)op] == EXACT) {
3843 /* Literal string, where present. */
3844 node += NODE_SZ_STR(node) - 1;
3845 node = NEXTOPER(node);
3848 node = NEXTOPER(node);
3849 node += regarglen[(U8)op];
3851 if (op == CURLYX || op == OPEN)
3853 else if (op == WHILEM)
3856 #endif /* DEBUGGING */
3861 - regdump - dump a regexp onto Perl_debug_log in vaguely comprehensible form
3864 Perl_regdump(pTHX_ regexp *r)
3868 SV *sv = sv_newmortal();
3870 (void)dumpuntil(r->program, r->program + 1, NULL, sv, 0);
3872 /* Header fields of interest. */
3873 if (r->anchored_substr)
3874 PerlIO_printf(Perl_debug_log,
3875 "anchored `%s%.*s%s'%s at %"IVdf" ",
3877 (int)(SvCUR(r->anchored_substr) - (SvTAIL(r->anchored_substr)!=0)),
3878 SvPVX(r->anchored_substr),
3880 SvTAIL(r->anchored_substr) ? "$" : "",
3881 (IV)r->anchored_offset);
3882 if (r->float_substr)
3883 PerlIO_printf(Perl_debug_log,
3884 "floating `%s%.*s%s'%s at %"IVdf"..%"UVuf" ",
3886 (int)(SvCUR(r->float_substr) - (SvTAIL(r->float_substr)!=0)),
3887 SvPVX(r->float_substr),
3889 SvTAIL(r->float_substr) ? "$" : "",
3890 (IV)r->float_min_offset, (UV)r->float_max_offset);
3891 if (r->check_substr)
3892 PerlIO_printf(Perl_debug_log,
3893 r->check_substr == r->float_substr
3894 ? "(checking floating" : "(checking anchored");
3895 if (r->reganch & ROPT_NOSCAN)
3896 PerlIO_printf(Perl_debug_log, " noscan");
3897 if (r->reganch & ROPT_CHECK_ALL)
3898 PerlIO_printf(Perl_debug_log, " isall");
3899 if (r->check_substr)
3900 PerlIO_printf(Perl_debug_log, ") ");
3902 if (r->regstclass) {
3903 regprop(sv, r->regstclass);
3904 PerlIO_printf(Perl_debug_log, "stclass `%s' ", SvPVX(sv));
3906 if (r->reganch & ROPT_ANCH) {
3907 PerlIO_printf(Perl_debug_log, "anchored");
3908 if (r->reganch & ROPT_ANCH_BOL)
3909 PerlIO_printf(Perl_debug_log, "(BOL)");
3910 if (r->reganch & ROPT_ANCH_MBOL)
3911 PerlIO_printf(Perl_debug_log, "(MBOL)");
3912 if (r->reganch & ROPT_ANCH_SBOL)
3913 PerlIO_printf(Perl_debug_log, "(SBOL)");
3914 if (r->reganch & ROPT_ANCH_GPOS)
3915 PerlIO_printf(Perl_debug_log, "(GPOS)");
3916 PerlIO_putc(Perl_debug_log, ' ');
3918 if (r->reganch & ROPT_GPOS_SEEN)
3919 PerlIO_printf(Perl_debug_log, "GPOS ");
3920 if (r->reganch & ROPT_SKIP)
3921 PerlIO_printf(Perl_debug_log, "plus ");
3922 if (r->reganch & ROPT_IMPLICIT)
3923 PerlIO_printf(Perl_debug_log, "implicit ");
3924 PerlIO_printf(Perl_debug_log, "minlen %ld ", (long) r->minlen);
3925 if (r->reganch & ROPT_EVAL_SEEN)
3926 PerlIO_printf(Perl_debug_log, "with eval ");
3927 PerlIO_printf(Perl_debug_log, "\n");
3928 #endif /* DEBUGGING */
3932 S_put_byte(pTHX_ SV *sv, int c)
3934 if (c <= ' ' || c == 127 || c == 255)
3935 Perl_sv_catpvf(aTHX_ sv, "\\%o", c);
3936 else if (c == '-' || c == ']' || c == '\\' || c == '^')
3937 Perl_sv_catpvf(aTHX_ sv, "\\%c", c);
3939 Perl_sv_catpvf(aTHX_ sv, "%c", c);
3943 - regprop - printable representation of opcode
3946 Perl_regprop(pTHX_ SV *sv, regnode *o)
3952 sv_setpvn(sv, "", 0);
3953 if (OP(o) >= reg_num) /* regnode.type is unsigned */
3954 FAIL("corrupted regexp opcode");
3955 sv_catpv(sv, (char*)reg_name[OP(o)]); /* Take off const! */
3957 k = PL_regkind[(U8)OP(o)];
3960 Perl_sv_catpvf(aTHX_ sv, " <%s%.*s%s>", PL_colors[0],
3961 STR_LEN(o), STRING(o), PL_colors[1]);
3962 else if (k == CURLY) {
3963 if (OP(o) == CURLYM || OP(o) == CURLYN)
3964 Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* Parenth number */
3965 Perl_sv_catpvf(aTHX_ sv, " {%d,%d}", ARG1(o), ARG2(o));
3967 else if (k == WHILEM && o->flags) /* Ordinal/of */
3968 Perl_sv_catpvf(aTHX_ sv, "[%d/%d]", o->flags & 0xf, o->flags>>4);
3969 else if (k == REF || k == OPEN || k == CLOSE || k == GROUPP )
3970 Perl_sv_catpvf(aTHX_ sv, "%d", (int)ARG(o)); /* Parenth number */
3971 else if (k == LOGICAL)
3972 Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* 2: embedded, otherwise 1 */
3973 else if (k == ANYOF) {
3974 int i, rangestart = -1;
3975 const char * const out[] = { /* Should be syncronized with
3976 a table in regcomp.h */
4005 if (o->flags & ANYOF_LOCALE)
4006 sv_catpv(sv, "{loc}");
4007 if (o->flags & ANYOF_FOLD)
4008 sv_catpv(sv, "{i}");
4009 Perl_sv_catpvf(aTHX_ sv, "[%s", PL_colors[0]);
4010 if (o->flags & ANYOF_INVERT)
4012 for (i = 0; i <= 256; i++) {
4013 if (i < 256 && ANYOF_BITMAP_TEST(o,i)) {
4014 if (rangestart == -1)
4016 } else if (rangestart != -1) {
4017 if (i <= rangestart + 3)
4018 for (; rangestart < i; rangestart++)
4019 put_byte(sv, rangestart);
4021 put_byte(sv, rangestart);
4023 put_byte(sv, i - 1);
4028 if (o->flags & ANYOF_CLASS)
4029 for (i = 0; i < sizeof(out)/sizeof(char*); i++)
4030 if (ANYOF_CLASS_TEST(o,i))
4031 sv_catpv(sv, out[i]);
4032 Perl_sv_catpvf(aTHX_ sv, "%s]", PL_colors[1]);
4034 else if (k == BRANCHJ && (OP(o) == UNLESSM || OP(o) == IFMATCH))
4035 Perl_sv_catpvf(aTHX_ sv, "[-%d]", o->flags);
4036 #endif /* DEBUGGING */
4040 Perl_re_intuit_string(pTHX_ regexp *prog)
4041 { /* Assume that RE_INTUIT is set */
4044 char *s = SvPV(prog->check_substr,n_a);
4046 if (!PL_colorset) reginitcolors();
4047 PerlIO_printf(Perl_debug_log,
4048 "%sUsing REx substr:%s `%s%.60s%s%s'\n",
4049 PL_colors[4],PL_colors[5],PL_colors[0],
4052 (strlen(s) > 60 ? "..." : ""));
4055 return prog->check_substr;
4059 Perl_pregfree(pTHX_ struct regexp *r)
4062 DEBUG_r(if (!PL_colorset) reginitcolors());
4064 if (!r || (--r->refcnt > 0))
4066 DEBUG_r(PerlIO_printf(Perl_debug_log,
4067 "%sFreeing REx:%s `%s%.60s%s%s'\n",
4068 PL_colors[4],PL_colors[5],PL_colors[0],
4071 (strlen(r->precomp) > 60 ? "..." : "")));
4074 Safefree(r->precomp);
4075 if (RX_MATCH_COPIED(r))
4076 Safefree(r->subbeg);
4078 if (r->anchored_substr)
4079 SvREFCNT_dec(r->anchored_substr);
4080 if (r->float_substr)
4081 SvREFCNT_dec(r->float_substr);
4082 Safefree(r->substrs);
4085 int n = r->data->count;
4086 AV* new_comppad = NULL;
4091 switch (r->data->what[n]) {
4093 SvREFCNT_dec((SV*)r->data->data[n]);
4096 Safefree(r->data->data[n]);
4099 new_comppad = (AV*)r->data->data[n];
4102 if (new_comppad == NULL)
4103 Perl_croak(aTHX_ "panic: pregfree comppad");
4104 old_comppad = PL_comppad;
4105 old_curpad = PL_curpad;
4106 PL_comppad = new_comppad;
4107 PL_curpad = AvARRAY(new_comppad);
4108 op_free((OP_4tree*)r->data->data[n]);
4109 PL_comppad = old_comppad;
4110 PL_curpad = old_curpad;
4111 SvREFCNT_dec((SV*)new_comppad);
4117 FAIL2("panic: regfree data code '%c'", r->data->what[n]);
4120 Safefree(r->data->what);
4123 Safefree(r->startp);
4129 - regnext - dig the "next" pointer out of a node
4131 * [Note, when REGALIGN is defined there are two places in regmatch()
4132 * that bypass this code for speed.]
4135 Perl_regnext(pTHX_ register regnode *p)
4138 register I32 offset;
4140 if (p == &PL_regdummy)
4143 offset = (reg_off_by_arg[OP(p)] ? ARG(p) : NEXT_OFF(p));
4151 S_re_croak2(pTHX_ const char* pat1,const char* pat2,...)
4154 STRLEN l1 = strlen(pat1);
4155 STRLEN l2 = strlen(pat2);
4164 Copy(pat1, buf, l1 , char);
4165 Copy(pat2, buf + l1, l2 , char);
4166 buf[l1 + l2] = '\n';
4167 buf[l1 + l2 + 1] = '\0';
4169 /* ANSI variant takes additional second argument */
4170 va_start(args, pat2);
4174 msv = vmess(buf, &args);
4176 message = SvPV(msv,l1);
4179 Copy(message, buf, l1 , char);
4180 buf[l1] = '\0'; /* Overwrite \n */
4181 Perl_croak(aTHX_ "%s", buf);
4184 /* XXX Here's a total kludge. But we need to re-enter for swash routines. */
4187 Perl_save_re_context(pTHX)
4191 SAVEPPTR(PL_regprecomp); /* uncompiled string. */
4192 SAVEI32(PL_regnpar); /* () count. */
4193 SAVEI32(PL_regsize); /* Code size. */
4194 SAVEI16(PL_regflags); /* are we folding, multilining? */
4195 SAVEPPTR(PL_reginput); /* String-input pointer. */
4196 SAVEPPTR(PL_regbol); /* Beginning of input, for ^ check. */
4197 SAVEPPTR(PL_regeol); /* End of input, for $ check. */
4198 SAVEVPTR(PL_regstartp); /* Pointer to startp array. */
4199 SAVEVPTR(PL_regendp); /* Ditto for endp. */
4200 SAVEVPTR(PL_reglastparen); /* Similarly for lastparen. */
4201 SAVEPPTR(PL_regtill); /* How far we are required to go. */
4202 SAVEI8(PL_regprev); /* char before regbol, \n if none */
4203 SAVEVPTR(PL_reg_start_tmp); /* from regexec.c */
4204 PL_reg_start_tmp = 0;
4205 SAVEFREEPV(PL_reg_start_tmp);
4206 SAVEI32(PL_reg_start_tmpl); /* from regexec.c */
4207 PL_reg_start_tmpl = 0;
4208 SAVEVPTR(PL_regdata);
4209 SAVEI32(PL_reg_flags); /* from regexec.c */
4210 SAVEI32(PL_reg_eval_set); /* from regexec.c */
4211 SAVEI32(PL_regnarrate); /* from regexec.c */
4212 SAVEVPTR(PL_regprogram); /* from regexec.c */
4213 SAVEINT(PL_regindent); /* from regexec.c */
4214 SAVEVPTR(PL_regcc); /* from regexec.c */
4215 SAVEVPTR(PL_curcop);
4216 SAVEVPTR(PL_regcomp_rx); /* from regcomp.c */
4217 SAVEI32(PL_regseen); /* from regcomp.c */
4218 SAVEI32(PL_regsawback); /* Did we see \1, ...? */
4219 SAVEI32(PL_regnaughty); /* How bad is this pattern? */
4220 SAVEVPTR(PL_regcode); /* Code-emit pointer; ®dummy = don't */
4221 SAVEPPTR(PL_regxend); /* End of input for compile */
4222 SAVEPPTR(PL_regcomp_parse); /* Input-scan pointer. */
4223 SAVEVPTR(PL_reg_call_cc); /* from regexec.c */
4224 SAVEVPTR(PL_reg_re); /* from regexec.c */
4225 SAVEPPTR(PL_reg_ganch); /* from regexec.c */
4226 SAVESPTR(PL_reg_sv); /* from regexec.c */
4227 SAVEVPTR(PL_reg_magic); /* from regexec.c */
4228 SAVEI32(PL_reg_oldpos); /* from regexec.c */
4229 SAVEVPTR(PL_reg_oldcurpm); /* from regexec.c */
4230 SAVEVPTR(PL_reg_curpm); /* from regexec.c */
4232 SAVEPPTR(PL_reg_starttry); /* from regexec.c */
4243 clear_re(pTHXo_ void *r)
4245 ReREFCNT_dec((regexp *)r);