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) 1991-2001, 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 #ifndef PERL_IN_XSUB_RE
98 # if defined(BUGGY_MSC6)
99 /* MSC 6.00A breaks on op/regexp.t test 85 unless we turn this off */
100 # pragma optimize("a",off)
101 /* But MSC 6.00A is happy with 'w', for aliases only across function calls*/
102 # pragma optimize("w",on )
103 # endif /* BUGGY_MSC6 */
107 #define STATIC static
110 typedef struct RExC_state_t {
111 U16 flags16; /* are we folding, multilining? */
112 char *precomp; /* uncompiled string. */
114 char *start; /* Start of input for compile */
115 char *end; /* End of input for compile */
116 char *parse; /* Input-scan pointer. */
117 I32 whilem_seen; /* number of WHILEM in this expr */
118 regnode *emit_start; /* Start of emitted-code area */
119 regnode *emit; /* Code-emit pointer; ®dummy = don't = compiling */
120 I32 naughty; /* How bad is this pattern? */
121 I32 sawback; /* Did we see \1, ...? */
123 I32 size; /* Code size. */
124 I32 npar; /* () count. */
130 char *starttry; /* -Dr: where regtry was called. */
131 #define RExC_starttry (pRExC_state->starttry)
135 #define RExC_flags16 (pRExC_state->flags16)
136 #define RExC_precomp (pRExC_state->precomp)
137 #define RExC_rx (pRExC_state->rx)
138 #define RExC_start (pRExC_state->start)
139 #define RExC_end (pRExC_state->end)
140 #define RExC_parse (pRExC_state->parse)
141 #define RExC_whilem_seen (pRExC_state->whilem_seen)
142 #define RExC_offsets (pRExC_state->rx->offsets) /* I am not like the others */
143 #define RExC_emit (pRExC_state->emit)
144 #define RExC_emit_start (pRExC_state->emit_start)
145 #define RExC_naughty (pRExC_state->naughty)
146 #define RExC_sawback (pRExC_state->sawback)
147 #define RExC_seen (pRExC_state->seen)
148 #define RExC_size (pRExC_state->size)
149 #define RExC_npar (pRExC_state->npar)
150 #define RExC_extralen (pRExC_state->extralen)
151 #define RExC_seen_zerolen (pRExC_state->seen_zerolen)
152 #define RExC_seen_evals (pRExC_state->seen_evals)
153 #define RExC_utf8 (pRExC_state->utf8)
155 #define ISMULT1(c) ((c) == '*' || (c) == '+' || (c) == '?')
156 #define ISMULT2(s) ((*s) == '*' || (*s) == '+' || (*s) == '?' || \
157 ((*s) == '{' && regcurly(s)))
160 #undef SPSTART /* dratted cpp namespace... */
163 * Flags to be passed up and down.
165 #define WORST 0 /* Worst case. */
166 #define HASWIDTH 0x1 /* Known to match non-null strings. */
167 #define SIMPLE 0x2 /* Simple enough to be STAR/PLUS operand. */
168 #define SPSTART 0x4 /* Starts with * or +. */
169 #define TRYAGAIN 0x8 /* Weeded out a declaration. */
171 /* Length of a variant. */
173 typedef struct scan_data_t {
179 I32 last_end; /* min value, <0 unless valid. */
182 SV **longest; /* Either &l_fixed, or &l_float. */
186 I32 offset_float_min;
187 I32 offset_float_max;
191 struct regnode_charclass_class *start_class;
195 * Forward declarations for pregcomp()'s friends.
198 static scan_data_t zero_scan_data = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
201 #define SF_BEFORE_EOL (SF_BEFORE_SEOL|SF_BEFORE_MEOL)
202 #define SF_BEFORE_SEOL 0x1
203 #define SF_BEFORE_MEOL 0x2
204 #define SF_FIX_BEFORE_EOL (SF_FIX_BEFORE_SEOL|SF_FIX_BEFORE_MEOL)
205 #define SF_FL_BEFORE_EOL (SF_FL_BEFORE_SEOL|SF_FL_BEFORE_MEOL)
208 # define SF_FIX_SHIFT_EOL (0+2)
209 # define SF_FL_SHIFT_EOL (0+4)
211 # define SF_FIX_SHIFT_EOL (+2)
212 # define SF_FL_SHIFT_EOL (+4)
215 #define SF_FIX_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FIX_SHIFT_EOL)
216 #define SF_FIX_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FIX_SHIFT_EOL)
218 #define SF_FL_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FL_SHIFT_EOL)
219 #define SF_FL_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FL_SHIFT_EOL) /* 0x20 */
220 #define SF_IS_INF 0x40
221 #define SF_HAS_PAR 0x80
222 #define SF_IN_PAR 0x100
223 #define SF_HAS_EVAL 0x200
224 #define SCF_DO_SUBSTR 0x400
225 #define SCF_DO_STCLASS_AND 0x0800
226 #define SCF_DO_STCLASS_OR 0x1000
227 #define SCF_DO_STCLASS (SCF_DO_STCLASS_AND|SCF_DO_STCLASS_OR)
228 #define SCF_WHILEM_VISITED_POS 0x2000
230 #define UTF RExC_utf8
231 #define LOC (RExC_flags16 & PMf_LOCALE)
232 #define FOLD (RExC_flags16 & PMf_FOLD)
234 #define OOB_UNICODE 12345678
235 #define OOB_NAMEDCLASS -1
237 #define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
238 #define CHR_DIST(a,b) (UTF ? utf8_distance(a,b) : a - b)
241 /* length of regex to show in messages that don't mark a position within */
242 #define RegexLengthToShowInErrorMessages 127
245 * If MARKER[12] are adjusted, be sure to adjust the constants at the top
246 * of t/op/regmesg.t, the tests in t/op/re_tests, and those in
247 * op/pragma/warn/regcomp.
249 #define MARKER1 "<-- HERE" /* marker as it appears in the description */
250 #define MARKER2 " <-- HERE " /* marker as it appears within the regex */
252 #define REPORT_LOCATION " in regex; marked by " MARKER1 " in m/%.*s" MARKER2 "%s/"
255 * Calls SAVEDESTRUCTOR_X if needed, then calls Perl_croak with the given
256 * arg. Show regex, up to a maximum length. If it's too long, chop and add
261 char *ellipses = ""; \
262 IV len = RExC_end - RExC_precomp; \
265 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
267 if (len > RegexLengthToShowInErrorMessages) { \
268 /* chop 10 shorter than the max, to ensure meaning of "..." */ \
269 len = RegexLengthToShowInErrorMessages - 10; \
272 Perl_croak(aTHX_ "%s in regex m/%.*s%s/", \
273 msg, (int)len, RExC_precomp, ellipses); \
277 * Calls SAVEDESTRUCTOR_X if needed, then calls Perl_croak with the given
278 * args. Show regex, up to a maximum length. If it's too long, chop and add
281 #define FAIL2(pat,msg) \
283 char *ellipses = ""; \
284 IV len = RExC_end - RExC_precomp; \
287 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
289 if (len > RegexLengthToShowInErrorMessages) { \
290 /* chop 10 shorter than the max, to ensure meaning of "..." */ \
291 len = RegexLengthToShowInErrorMessages - 10; \
294 S_re_croak2(aTHX_ pat, " in regex m/%.*s%s/", \
295 msg, (int)len, RExC_precomp, ellipses); \
300 * Simple_vFAIL -- like FAIL, but marks the current location in the scan
302 #define Simple_vFAIL(m) \
304 IV offset = RExC_parse - RExC_precomp; \
306 Perl_croak(aTHX_ "%s" REPORT_LOCATION, \
307 m, (int)offset, RExC_precomp, RExC_precomp + offset); \
311 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL()
316 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
321 * Like Simple_vFAIL(), but accepts two arguments.
323 #define Simple_vFAIL2(m,a1) \
325 IV offset = RExC_parse - RExC_precomp; \
327 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, \
328 (int)offset, RExC_precomp, RExC_precomp + offset); \
332 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL2().
334 #define vFAIL2(m,a1) \
337 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
338 Simple_vFAIL2(m, a1); \
343 * Like Simple_vFAIL(), but accepts three arguments.
345 #define Simple_vFAIL3(m, a1, a2) \
347 IV offset = RExC_parse - RExC_precomp; \
349 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, \
350 (int)offset, RExC_precomp, RExC_precomp + offset); \
354 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL3().
356 #define vFAIL3(m,a1,a2) \
359 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
360 Simple_vFAIL3(m, a1, a2); \
364 * Like Simple_vFAIL(), but accepts four arguments.
366 #define Simple_vFAIL4(m, a1, a2, a3) \
368 IV offset = RExC_parse - RExC_precomp; \
370 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, a3,\
371 (int)offset, RExC_precomp, RExC_precomp + offset); \
375 * Like Simple_vFAIL(), but accepts five arguments.
377 #define Simple_vFAIL5(m, a1, a2, a3, a4) \
379 IV offset = RExC_parse - RExC_precomp; \
380 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, a3, a4,\
381 (int)offset, RExC_precomp, RExC_precomp + offset); \
385 #define vWARN(loc,m) \
387 IV offset = loc - RExC_precomp; \
388 Perl_warner(aTHX_ WARN_REGEXP, "%s" REPORT_LOCATION,\
389 m, (int)offset, RExC_precomp, RExC_precomp + offset); \
392 #define vWARNdep(loc,m) \
394 IV offset = loc - RExC_precomp; \
395 int warn_cat = ckWARN(WARN_REGEXP) ? WARN_REGEXP : WARN_DEPRECATED; \
396 Perl_warner(aTHX_ warn_cat, "%s" REPORT_LOCATION,\
397 m, (int)offset, RExC_precomp, RExC_precomp + offset); \
401 #define vWARN2(loc, m, a1) \
403 IV offset = loc - RExC_precomp; \
404 Perl_warner(aTHX_ WARN_REGEXP, m REPORT_LOCATION,\
406 (int)offset, RExC_precomp, RExC_precomp + offset); \
409 #define vWARN3(loc, m, a1, a2) \
411 IV offset = loc - RExC_precomp; \
412 Perl_warner(aTHX_ WARN_REGEXP, m REPORT_LOCATION, \
414 (int)offset, RExC_precomp, RExC_precomp + offset); \
417 #define vWARN4(loc, m, a1, a2, a3) \
419 IV offset = loc - RExC_precomp; \
420 Perl_warner(aTHX_ WARN_REGEXP, m REPORT_LOCATION,\
422 (int)offset, RExC_precomp, RExC_precomp + offset); \
425 /* used for the parse_flags section for (?c) -- japhy */
426 #define vWARN5(loc, m, a1, a2, a3, a4) \
428 IV offset = loc - RExC_precomp; \
429 Perl_warner(aTHX_ WARN_REGEXP, m REPORT_LOCATION, \
431 (int)offset, RExC_precomp, RExC_precomp + offset); \
435 /* Allow for side effects in s */
436 #define REGC(c,s) STMT_START { if (!SIZE_ONLY) *(s) = (c); else (void)(s);} STMT_END
438 /* Macros for recording node offsets. 20001227 mjd@plover.com
439 * Nodes are numbered 1, 2, 3, 4. Node #n's position is recorded in
440 * element 2*n-1 of the array. Element #2n holds the byte length node #n.
441 * Element 0 holds the number n.
444 #define MJD_OFFSET_DEBUG(x)
445 /* #define MJD_OFFSET_DEBUG(x) fprintf x */
448 # define Set_Node_Offset_To_R(node,byte) \
452 Perl_croak(aTHX_ "value of node is %d in Offset macro", node); \
454 RExC_offsets[2*(node)-1] = (byte); \
459 # define Set_Node_Offset(node,byte) Set_Node_Offset_To_R((node)-RExC_emit_start, (byte)-RExC_start)
460 # define Set_Cur_Node_Offset Set_Node_Offset(RExC_emit, RExC_parse)
462 # define Set_Node_Length_To_R(node,len) \
465 MJD_OFFSET_DEBUG((stderr, "** (%d) size of node %d is %d.\n", __LINE__, (node), (len))); \
467 Perl_croak(aTHX_ "value of node is %d in Length macro", node); \
469 RExC_offsets[2*(node)] = (len); \
474 # define Set_Node_Length(node,len) Set_Node_Length_To_R((node)-RExC_emit_start, len)
475 # define Set_Cur_Node_Length(len) Set_Node_Length(RExC_emit, len)
476 # define Set_Node_Cur_Length(node) Set_Node_Length(node, RExC_parse - parse_start)
478 /* Get offsets and lengths */
479 #define Node_Offset(n) (RExC_offsets[2*((n)-RExC_emit_start)-1])
480 #define Node_Length(n) (RExC_offsets[2*((n)-RExC_emit_start)])
482 static void clear_re(pTHX_ void *r);
484 /* Mark that we cannot extend a found fixed substring at this point.
485 Updata the longest found anchored substring and the longest found
486 floating substrings if needed. */
489 S_scan_commit(pTHX_ RExC_state_t *pRExC_state, scan_data_t *data)
491 STRLEN l = CHR_SVLEN(data->last_found);
492 STRLEN old_l = CHR_SVLEN(*data->longest);
494 if ((l >= old_l) && ((l > old_l) || (data->flags & SF_BEFORE_EOL))) {
495 sv_setsv(*data->longest, data->last_found);
496 if (*data->longest == data->longest_fixed) {
497 data->offset_fixed = l ? data->last_start_min : data->pos_min;
498 if (data->flags & SF_BEFORE_EOL)
500 |= ((data->flags & SF_BEFORE_EOL) << SF_FIX_SHIFT_EOL);
502 data->flags &= ~SF_FIX_BEFORE_EOL;
505 data->offset_float_min = l ? data->last_start_min : data->pos_min;
506 data->offset_float_max = (l
507 ? data->last_start_max
508 : data->pos_min + data->pos_delta);
509 if (data->flags & SF_BEFORE_EOL)
511 |= ((data->flags & SF_BEFORE_EOL) << SF_FL_SHIFT_EOL);
513 data->flags &= ~SF_FL_BEFORE_EOL;
516 SvCUR_set(data->last_found, 0);
518 data->flags &= ~SF_BEFORE_EOL;
521 /* Can match anything (initialization) */
523 S_cl_anything(pTHX_ RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
525 ANYOF_CLASS_ZERO(cl);
526 ANYOF_BITMAP_SETALL(cl);
527 cl->flags = ANYOF_EOS|ANYOF_UNICODE_ALL;
529 cl->flags |= ANYOF_LOCALE;
532 /* Can match anything (initialization) */
534 S_cl_is_anything(pTHX_ struct regnode_charclass_class *cl)
538 for (value = 0; value <= ANYOF_MAX; value += 2)
539 if (ANYOF_CLASS_TEST(cl, value) && ANYOF_CLASS_TEST(cl, value + 1))
541 if (!(cl->flags & ANYOF_UNICODE_ALL))
543 if (!ANYOF_BITMAP_TESTALLSET(cl))
548 /* Can match anything (initialization) */
550 S_cl_init(pTHX_ RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
552 Zero(cl, 1, struct regnode_charclass_class);
554 cl_anything(pRExC_state, cl);
558 S_cl_init_zero(pTHX_ RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
560 Zero(cl, 1, struct regnode_charclass_class);
562 cl_anything(pRExC_state, cl);
564 cl->flags |= ANYOF_LOCALE;
567 /* 'And' a given class with another one. Can create false positives */
568 /* We assume that cl is not inverted */
570 S_cl_and(pTHX_ struct regnode_charclass_class *cl,
571 struct regnode_charclass_class *and_with)
573 if (!(and_with->flags & ANYOF_CLASS)
574 && !(cl->flags & ANYOF_CLASS)
575 && (and_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
576 && !(and_with->flags & ANYOF_FOLD)
577 && !(cl->flags & ANYOF_FOLD)) {
580 if (and_with->flags & ANYOF_INVERT)
581 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
582 cl->bitmap[i] &= ~and_with->bitmap[i];
584 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
585 cl->bitmap[i] &= and_with->bitmap[i];
586 } /* XXXX: logic is complicated otherwise, leave it along for a moment. */
587 if (!(and_with->flags & ANYOF_EOS))
588 cl->flags &= ~ANYOF_EOS;
590 if (cl->flags & ANYOF_UNICODE_ALL && and_with->flags & ANYOF_UNICODE) {
591 cl->flags &= ~ANYOF_UNICODE_ALL;
592 cl->flags |= ANYOF_UNICODE;
593 ARG_SET(cl, ARG(and_with));
595 if (!(and_with->flags & ANYOF_UNICODE_ALL))
596 cl->flags &= ~ANYOF_UNICODE_ALL;
597 if (!(and_with->flags & (ANYOF_UNICODE|ANYOF_UNICODE_ALL)))
598 cl->flags &= ~ANYOF_UNICODE;
601 /* 'OR' a given class with another one. Can create false positives */
602 /* We assume that cl is not inverted */
604 S_cl_or(pTHX_ RExC_state_t *pRExC_state, struct regnode_charclass_class *cl, struct regnode_charclass_class *or_with)
606 if (or_with->flags & ANYOF_INVERT) {
608 * (B1 | CL1) | (!B2 & !CL2) = (B1 | !B2 & !CL2) | (CL1 | (!B2 & !CL2))
609 * <= (B1 | !B2) | (CL1 | !CL2)
610 * which is wasteful if CL2 is small, but we ignore CL2:
611 * (B1 | CL1) | (!B2 & !CL2) <= (B1 | CL1) | !B2 = (B1 | !B2) | CL1
612 * XXXX Can we handle case-fold? Unclear:
613 * (OK1(i) | OK1(i')) | !(OK1(i) | OK1(i')) =
614 * (OK1(i) | OK1(i')) | (!OK1(i) & !OK1(i'))
616 if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
617 && !(or_with->flags & ANYOF_FOLD)
618 && !(cl->flags & ANYOF_FOLD) ) {
621 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
622 cl->bitmap[i] |= ~or_with->bitmap[i];
623 } /* XXXX: logic is complicated otherwise */
625 cl_anything(pRExC_state, cl);
628 /* (B1 | CL1) | (B2 | CL2) = (B1 | B2) | (CL1 | CL2)) */
629 if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
630 && (!(or_with->flags & ANYOF_FOLD)
631 || (cl->flags & ANYOF_FOLD)) ) {
634 /* OR char bitmap and class bitmap separately */
635 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
636 cl->bitmap[i] |= or_with->bitmap[i];
637 if (or_with->flags & ANYOF_CLASS) {
638 for (i = 0; i < ANYOF_CLASSBITMAP_SIZE; i++)
639 cl->classflags[i] |= or_with->classflags[i];
640 cl->flags |= ANYOF_CLASS;
643 else { /* XXXX: logic is complicated, leave it along for a moment. */
644 cl_anything(pRExC_state, cl);
647 if (or_with->flags & ANYOF_EOS)
648 cl->flags |= ANYOF_EOS;
650 if (cl->flags & ANYOF_UNICODE && or_with->flags & ANYOF_UNICODE &&
651 ARG(cl) != ARG(or_with)) {
652 cl->flags |= ANYOF_UNICODE_ALL;
653 cl->flags &= ~ANYOF_UNICODE;
655 if (or_with->flags & ANYOF_UNICODE_ALL) {
656 cl->flags |= ANYOF_UNICODE_ALL;
657 cl->flags &= ~ANYOF_UNICODE;
661 /* REx optimizer. Converts nodes into quickier variants "in place".
662 Finds fixed substrings. */
664 /* Stops at toplevel WHILEM as well as at `last'. At end *scanp is set
665 to the position after last scanned or to NULL. */
668 S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp, I32 *deltap, regnode *last, scan_data_t *data, U32 flags)
669 /* scanp: Start here (read-write). */
670 /* deltap: Write maxlen-minlen here. */
671 /* last: Stop before this one. */
673 I32 min = 0, pars = 0, code;
674 regnode *scan = *scanp, *next;
676 int is_inf = (flags & SCF_DO_SUBSTR) && (data->flags & SF_IS_INF);
677 int is_inf_internal = 0; /* The studied chunk is infinite */
678 I32 is_par = OP(scan) == OPEN ? ARG(scan) : 0;
679 scan_data_t data_fake;
680 struct regnode_charclass_class and_with; /* Valid if flags & SCF_DO_STCLASS_OR */
682 while (scan && OP(scan) != END && scan < last) {
683 /* Peephole optimizer: */
685 if (PL_regkind[(U8)OP(scan)] == EXACT) {
686 /* Merge several consecutive EXACTish nodes into one. */
687 regnode *n = regnext(scan);
690 regnode *stop = scan;
693 next = scan + NODE_SZ_STR(scan);
694 /* Skip NOTHING, merge EXACT*. */
696 ( PL_regkind[(U8)OP(n)] == NOTHING ||
697 (stringok && (OP(n) == OP(scan))))
699 && NEXT_OFF(scan) + NEXT_OFF(n) < I16_MAX) {
700 if (OP(n) == TAIL || n > next)
702 if (PL_regkind[(U8)OP(n)] == NOTHING) {
703 NEXT_OFF(scan) += NEXT_OFF(n);
704 next = n + NODE_STEP_REGNODE;
712 int oldl = STR_LEN(scan);
713 regnode *nnext = regnext(n);
715 if (oldl + STR_LEN(n) > U8_MAX)
717 NEXT_OFF(scan) += NEXT_OFF(n);
718 STR_LEN(scan) += STR_LEN(n);
719 next = n + NODE_SZ_STR(n);
720 /* Now we can overwrite *n : */
721 Move(STRING(n), STRING(scan) + oldl, STR_LEN(n), char);
730 n = scan + NODE_SZ_STR(scan);
732 if (PL_regkind[(U8)OP(n)] != NOTHING || OP(n) == NOTHING) {
740 /* Follow the next-chain of the current node and optimize
741 away all the NOTHINGs from it. */
742 if (OP(scan) != CURLYX) {
743 int max = (reg_off_by_arg[OP(scan)]
745 /* I32 may be smaller than U16 on CRAYs! */
746 : (I32_MAX < U16_MAX ? I32_MAX : U16_MAX));
747 int off = (reg_off_by_arg[OP(scan)] ? ARG(scan) : NEXT_OFF(scan));
751 /* Skip NOTHING and LONGJMP. */
752 while ((n = regnext(n))
753 && ((PL_regkind[(U8)OP(n)] == NOTHING && (noff = NEXT_OFF(n)))
754 || ((OP(n) == LONGJMP) && (noff = ARG(n))))
757 if (reg_off_by_arg[OP(scan)])
760 NEXT_OFF(scan) = off;
762 /* The principal pseudo-switch. Cannot be a switch, since we
763 look into several different things. */
764 if (OP(scan) == BRANCH || OP(scan) == BRANCHJ
765 || OP(scan) == IFTHEN || OP(scan) == SUSPEND) {
766 next = regnext(scan);
769 if (OP(next) == code || code == IFTHEN || code == SUSPEND) {
770 I32 max1 = 0, min1 = I32_MAX, num = 0;
771 struct regnode_charclass_class accum;
773 if (flags & SCF_DO_SUBSTR) /* XXXX Add !SUSPEND? */
774 scan_commit(pRExC_state, data); /* Cannot merge strings after this. */
775 if (flags & SCF_DO_STCLASS)
776 cl_init_zero(pRExC_state, &accum);
777 while (OP(scan) == code) {
778 I32 deltanext, minnext, f = 0, fake;
779 struct regnode_charclass_class this_class;
784 data_fake.whilem_c = data->whilem_c;
785 data_fake.last_closep = data->last_closep;
788 data_fake.last_closep = &fake;
789 next = regnext(scan);
790 scan = NEXTOPER(scan);
792 scan = NEXTOPER(scan);
793 if (flags & SCF_DO_STCLASS) {
794 cl_init(pRExC_state, &this_class);
795 data_fake.start_class = &this_class;
796 f = SCF_DO_STCLASS_AND;
798 if (flags & SCF_WHILEM_VISITED_POS)
799 f |= SCF_WHILEM_VISITED_POS;
800 /* we suppose the run is continuous, last=next...*/
801 minnext = study_chunk(pRExC_state, &scan, &deltanext,
802 next, &data_fake, f);
805 if (max1 < minnext + deltanext)
806 max1 = minnext + deltanext;
807 if (deltanext == I32_MAX)
808 is_inf = is_inf_internal = 1;
810 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
812 if (data && (data_fake.flags & SF_HAS_EVAL))
813 data->flags |= SF_HAS_EVAL;
815 data->whilem_c = data_fake.whilem_c;
816 if (flags & SCF_DO_STCLASS)
817 cl_or(pRExC_state, &accum, &this_class);
821 if (code == IFTHEN && num < 2) /* Empty ELSE branch */
823 if (flags & SCF_DO_SUBSTR) {
824 data->pos_min += min1;
825 data->pos_delta += max1 - min1;
826 if (max1 != min1 || is_inf)
827 data->longest = &(data->longest_float);
830 delta += max1 - min1;
831 if (flags & SCF_DO_STCLASS_OR) {
832 cl_or(pRExC_state, data->start_class, &accum);
834 cl_and(data->start_class, &and_with);
835 flags &= ~SCF_DO_STCLASS;
838 else if (flags & SCF_DO_STCLASS_AND) {
840 cl_and(data->start_class, &accum);
841 flags &= ~SCF_DO_STCLASS;
844 /* Switch to OR mode: cache the old value of
845 * data->start_class */
846 StructCopy(data->start_class, &and_with,
847 struct regnode_charclass_class);
848 flags &= ~SCF_DO_STCLASS_AND;
849 StructCopy(&accum, data->start_class,
850 struct regnode_charclass_class);
851 flags |= SCF_DO_STCLASS_OR;
852 data->start_class->flags |= ANYOF_EOS;
856 else if (code == BRANCHJ) /* single branch is optimized. */
857 scan = NEXTOPER(NEXTOPER(scan));
858 else /* single branch is optimized. */
859 scan = NEXTOPER(scan);
862 else if (OP(scan) == EXACT) {
863 I32 l = STR_LEN(scan);
864 UV uc = *((U8*)STRING(scan));
866 U8 *s = (U8*)STRING(scan);
867 l = utf8_length(s, s + l);
868 uc = utf8_to_uvchr(s, NULL);
871 if (flags & SCF_DO_SUBSTR) { /* Update longest substr. */
872 /* The code below prefers earlier match for fixed
873 offset, later match for variable offset. */
874 if (data->last_end == -1) { /* Update the start info. */
875 data->last_start_min = data->pos_min;
876 data->last_start_max = is_inf
877 ? I32_MAX : data->pos_min + data->pos_delta;
879 sv_catpvn(data->last_found, STRING(scan), STR_LEN(scan));
880 data->last_end = data->pos_min + l;
881 data->pos_min += l; /* As in the first entry. */
882 data->flags &= ~SF_BEFORE_EOL;
884 if (flags & SCF_DO_STCLASS_AND) {
885 /* Check whether it is compatible with what we know already! */
889 (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
890 && !ANYOF_BITMAP_TEST(data->start_class, uc)
891 && (!(data->start_class->flags & ANYOF_FOLD)
892 || !ANYOF_BITMAP_TEST(data->start_class, PL_fold[uc])))
895 ANYOF_CLASS_ZERO(data->start_class);
896 ANYOF_BITMAP_ZERO(data->start_class);
898 ANYOF_BITMAP_SET(data->start_class, uc);
899 data->start_class->flags &= ~ANYOF_EOS;
901 data->start_class->flags &= ~ANYOF_UNICODE_ALL;
903 else if (flags & SCF_DO_STCLASS_OR) {
904 /* false positive possible if the class is case-folded */
906 ANYOF_BITMAP_SET(data->start_class, uc);
908 data->start_class->flags |= ANYOF_UNICODE_ALL;
909 data->start_class->flags &= ~ANYOF_EOS;
910 cl_and(data->start_class, &and_with);
912 flags &= ~SCF_DO_STCLASS;
914 else if (PL_regkind[(U8)OP(scan)] == EXACT) { /* But OP != EXACT! */
915 I32 l = STR_LEN(scan);
916 UV uc = *((U8*)STRING(scan));
918 /* Search for fixed substrings supports EXACT only. */
919 if (flags & SCF_DO_SUBSTR)
920 scan_commit(pRExC_state, data);
922 U8 *s = (U8 *)STRING(scan);
923 l = utf8_length(s, s + l);
924 uc = utf8_to_uvchr(s, NULL);
927 if (data && (flags & SCF_DO_SUBSTR))
929 if (flags & SCF_DO_STCLASS_AND) {
930 /* Check whether it is compatible with what we know already! */
934 (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
935 && !ANYOF_BITMAP_TEST(data->start_class, uc)
936 && !ANYOF_BITMAP_TEST(data->start_class, PL_fold[uc])))
938 ANYOF_CLASS_ZERO(data->start_class);
939 ANYOF_BITMAP_ZERO(data->start_class);
941 ANYOF_BITMAP_SET(data->start_class, uc);
942 data->start_class->flags &= ~ANYOF_EOS;
943 data->start_class->flags |= ANYOF_FOLD;
944 if (OP(scan) == EXACTFL)
945 data->start_class->flags |= ANYOF_LOCALE;
948 else if (flags & SCF_DO_STCLASS_OR) {
949 if (data->start_class->flags & ANYOF_FOLD) {
950 /* false positive possible if the class is case-folded.
951 Assume that the locale settings are the same... */
953 ANYOF_BITMAP_SET(data->start_class, uc);
954 data->start_class->flags &= ~ANYOF_EOS;
956 cl_and(data->start_class, &and_with);
958 flags &= ~SCF_DO_STCLASS;
960 else if (strchr((char*)PL_varies,OP(scan))) {
961 I32 mincount, maxcount, minnext, deltanext, fl = 0;
962 I32 f = flags, pos_before = 0;
963 regnode *oscan = scan;
964 struct regnode_charclass_class this_class;
965 struct regnode_charclass_class *oclass = NULL;
966 I32 next_is_eval = 0;
968 switch (PL_regkind[(U8)OP(scan)]) {
969 case WHILEM: /* End of (?:...)* . */
970 scan = NEXTOPER(scan);
973 if (flags & (SCF_DO_SUBSTR | SCF_DO_STCLASS)) {
974 next = NEXTOPER(scan);
975 if (OP(next) == EXACT || (flags & SCF_DO_STCLASS)) {
977 maxcount = REG_INFTY;
978 next = regnext(scan);
979 scan = NEXTOPER(scan);
983 if (flags & SCF_DO_SUBSTR)
988 if (flags & SCF_DO_STCLASS) {
990 maxcount = REG_INFTY;
991 next = regnext(scan);
992 scan = NEXTOPER(scan);
995 is_inf = is_inf_internal = 1;
996 scan = regnext(scan);
997 if (flags & SCF_DO_SUBSTR) {
998 scan_commit(pRExC_state, data); /* Cannot extend fixed substrings */
999 data->longest = &(data->longest_float);
1001 goto optimize_curly_tail;
1003 mincount = ARG1(scan);
1004 maxcount = ARG2(scan);
1005 next = regnext(scan);
1006 if (OP(scan) == CURLYX) {
1007 I32 lp = (data ? *(data->last_closep) : 0);
1009 scan->flags = ((lp <= U8_MAX) ? lp : U8_MAX);
1011 scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
1012 next_is_eval = (OP(scan) == EVAL);
1014 if (flags & SCF_DO_SUBSTR) {
1015 if (mincount == 0) scan_commit(pRExC_state,data); /* Cannot extend fixed substrings */
1016 pos_before = data->pos_min;
1020 data->flags &= ~(SF_HAS_PAR|SF_IN_PAR|SF_HAS_EVAL);
1022 data->flags |= SF_IS_INF;
1024 if (flags & SCF_DO_STCLASS) {
1025 cl_init(pRExC_state, &this_class);
1026 oclass = data->start_class;
1027 data->start_class = &this_class;
1028 f |= SCF_DO_STCLASS_AND;
1029 f &= ~SCF_DO_STCLASS_OR;
1031 /* These are the cases when once a subexpression
1032 fails at a particular position, it cannot succeed
1033 even after backtracking at the enclosing scope.
1035 XXXX what if minimal match and we are at the
1036 initial run of {n,m}? */
1037 if ((mincount != maxcount - 1) && (maxcount != REG_INFTY))
1038 f &= ~SCF_WHILEM_VISITED_POS;
1040 /* This will finish on WHILEM, setting scan, or on NULL: */
1041 minnext = study_chunk(pRExC_state, &scan, &deltanext, last, data,
1043 ? (f & ~SCF_DO_SUBSTR) : f);
1045 if (flags & SCF_DO_STCLASS)
1046 data->start_class = oclass;
1047 if (mincount == 0 || minnext == 0) {
1048 if (flags & SCF_DO_STCLASS_OR) {
1049 cl_or(pRExC_state, data->start_class, &this_class);
1051 else if (flags & SCF_DO_STCLASS_AND) {
1052 /* Switch to OR mode: cache the old value of
1053 * data->start_class */
1054 StructCopy(data->start_class, &and_with,
1055 struct regnode_charclass_class);
1056 flags &= ~SCF_DO_STCLASS_AND;
1057 StructCopy(&this_class, data->start_class,
1058 struct regnode_charclass_class);
1059 flags |= SCF_DO_STCLASS_OR;
1060 data->start_class->flags |= ANYOF_EOS;
1062 } else { /* Non-zero len */
1063 if (flags & SCF_DO_STCLASS_OR) {
1064 cl_or(pRExC_state, data->start_class, &this_class);
1065 cl_and(data->start_class, &and_with);
1067 else if (flags & SCF_DO_STCLASS_AND)
1068 cl_and(data->start_class, &this_class);
1069 flags &= ~SCF_DO_STCLASS;
1071 if (!scan) /* It was not CURLYX, but CURLY. */
1073 if (ckWARN(WARN_REGEXP)
1074 /* ? quantifier ok, except for (?{ ... }) */
1075 && (next_is_eval || !(mincount == 0 && maxcount == 1))
1076 && (minnext == 0) && (deltanext == 0)
1077 && data && !(data->flags & (SF_HAS_PAR|SF_IN_PAR))
1078 && maxcount <= REG_INFTY/3) /* Complement check for big count */
1081 "Quantifier unexpected on zero-length expression");
1084 min += minnext * mincount;
1085 is_inf_internal |= ((maxcount == REG_INFTY
1086 && (minnext + deltanext) > 0)
1087 || deltanext == I32_MAX);
1088 is_inf |= is_inf_internal;
1089 delta += (minnext + deltanext) * maxcount - minnext * mincount;
1091 /* Try powerful optimization CURLYX => CURLYN. */
1092 if ( OP(oscan) == CURLYX && data
1093 && data->flags & SF_IN_PAR
1094 && !(data->flags & SF_HAS_EVAL)
1095 && !deltanext && minnext == 1 ) {
1096 /* Try to optimize to CURLYN. */
1097 regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS;
1098 regnode *nxt1 = nxt;
1105 if (!strchr((char*)PL_simple,OP(nxt))
1106 && !(PL_regkind[(U8)OP(nxt)] == EXACT
1107 && STR_LEN(nxt) == 1))
1113 if (OP(nxt) != CLOSE)
1115 /* Now we know that nxt2 is the only contents: */
1116 oscan->flags = ARG(nxt);
1118 OP(nxt1) = NOTHING; /* was OPEN. */
1120 OP(nxt1 + 1) = OPTIMIZED; /* was count. */
1121 NEXT_OFF(nxt1+ 1) = 0; /* just for consistancy. */
1122 NEXT_OFF(nxt2) = 0; /* just for consistancy with CURLY. */
1123 OP(nxt) = OPTIMIZED; /* was CLOSE. */
1124 OP(nxt + 1) = OPTIMIZED; /* was count. */
1125 NEXT_OFF(nxt+ 1) = 0; /* just for consistancy. */
1130 /* Try optimization CURLYX => CURLYM. */
1131 if ( OP(oscan) == CURLYX && data
1132 && !(data->flags & SF_HAS_PAR)
1133 && !(data->flags & SF_HAS_EVAL)
1135 /* XXXX How to optimize if data == 0? */
1136 /* Optimize to a simpler form. */
1137 regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN */
1141 while ( (nxt2 = regnext(nxt)) /* skip over embedded stuff*/
1142 && (OP(nxt2) != WHILEM))
1144 OP(nxt2) = SUCCEED; /* Whas WHILEM */
1145 /* Need to optimize away parenths. */
1146 if (data->flags & SF_IN_PAR) {
1147 /* Set the parenth number. */
1148 regnode *nxt1 = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN*/
1150 if (OP(nxt) != CLOSE)
1151 FAIL("Panic opt close");
1152 oscan->flags = ARG(nxt);
1153 OP(nxt1) = OPTIMIZED; /* was OPEN. */
1154 OP(nxt) = OPTIMIZED; /* was CLOSE. */
1156 OP(nxt1 + 1) = OPTIMIZED; /* was count. */
1157 OP(nxt + 1) = OPTIMIZED; /* was count. */
1158 NEXT_OFF(nxt1 + 1) = 0; /* just for consistancy. */
1159 NEXT_OFF(nxt + 1) = 0; /* just for consistancy. */
1162 while ( nxt1 && (OP(nxt1) != WHILEM)) {
1163 regnode *nnxt = regnext(nxt1);
1166 if (reg_off_by_arg[OP(nxt1)])
1167 ARG_SET(nxt1, nxt2 - nxt1);
1168 else if (nxt2 - nxt1 < U16_MAX)
1169 NEXT_OFF(nxt1) = nxt2 - nxt1;
1171 OP(nxt) = NOTHING; /* Cannot beautify */
1176 /* Optimize again: */
1177 study_chunk(pRExC_state, &nxt1, &deltanext, nxt,
1183 else if ((OP(oscan) == CURLYX)
1184 && (flags & SCF_WHILEM_VISITED_POS)
1185 /* See the comment on a similar expression above.
1186 However, this time it not a subexpression
1187 we care about, but the expression itself. */
1188 && (maxcount == REG_INFTY)
1189 && data && ++data->whilem_c < 16) {
1190 /* This stays as CURLYX, we can put the count/of pair. */
1191 /* Find WHILEM (as in regexec.c) */
1192 regnode *nxt = oscan + NEXT_OFF(oscan);
1194 if (OP(PREVOPER(nxt)) == NOTHING) /* LONGJMP */
1196 PREVOPER(nxt)->flags = data->whilem_c
1197 | (RExC_whilem_seen << 4); /* On WHILEM */
1199 if (data && fl & (SF_HAS_PAR|SF_IN_PAR))
1201 if (flags & SCF_DO_SUBSTR) {
1202 SV *last_str = Nullsv;
1203 int counted = mincount != 0;
1205 if (data->last_end > 0 && mincount != 0) { /* Ends with a string. */
1206 I32 b = pos_before >= data->last_start_min
1207 ? pos_before : data->last_start_min;
1209 char *s = SvPV(data->last_found, l);
1210 I32 old = b - data->last_start_min;
1213 old = utf8_hop((U8*)s, old) - (U8*)s;
1216 /* Get the added string: */
1217 last_str = newSVpvn(s + old, l);
1218 if (deltanext == 0 && pos_before == b) {
1219 /* What was added is a constant string */
1221 SvGROW(last_str, (mincount * l) + 1);
1222 repeatcpy(SvPVX(last_str) + l,
1223 SvPVX(last_str), l, mincount - 1);
1224 SvCUR(last_str) *= mincount;
1225 /* Add additional parts. */
1226 SvCUR_set(data->last_found,
1227 SvCUR(data->last_found) - l);
1228 sv_catsv(data->last_found, last_str);
1229 data->last_end += l * (mincount - 1);
1232 /* start offset must point into the last copy */
1233 data->last_start_min += minnext * (mincount - 1);
1234 data->last_start_max += is_inf ? 0 : (maxcount - 1)
1235 * (minnext + data->pos_delta);
1238 /* It is counted once already... */
1239 data->pos_min += minnext * (mincount - counted);
1240 data->pos_delta += - counted * deltanext +
1241 (minnext + deltanext) * maxcount - minnext * mincount;
1242 if (mincount != maxcount) {
1243 /* Cannot extend fixed substrings found inside
1245 scan_commit(pRExC_state,data);
1246 if (mincount && last_str) {
1247 sv_setsv(data->last_found, last_str);
1248 data->last_end = data->pos_min;
1249 data->last_start_min =
1250 data->pos_min - CHR_SVLEN(last_str);
1251 data->last_start_max = is_inf
1253 : data->pos_min + data->pos_delta
1254 - CHR_SVLEN(last_str);
1256 data->longest = &(data->longest_float);
1258 SvREFCNT_dec(last_str);
1260 if (data && (fl & SF_HAS_EVAL))
1261 data->flags |= SF_HAS_EVAL;
1262 optimize_curly_tail:
1263 if (OP(oscan) != CURLYX) {
1264 while (PL_regkind[(U8)OP(next = regnext(oscan))] == NOTHING
1266 NEXT_OFF(oscan) += NEXT_OFF(next);
1269 default: /* REF and CLUMP only? */
1270 if (flags & SCF_DO_SUBSTR) {
1271 scan_commit(pRExC_state,data); /* Cannot expect anything... */
1272 data->longest = &(data->longest_float);
1274 is_inf = is_inf_internal = 1;
1275 if (flags & SCF_DO_STCLASS_OR)
1276 cl_anything(pRExC_state, data->start_class);
1277 flags &= ~SCF_DO_STCLASS;
1281 else if (strchr((char*)PL_simple,OP(scan))) {
1284 if (flags & SCF_DO_SUBSTR) {
1285 scan_commit(pRExC_state,data);
1289 if (flags & SCF_DO_STCLASS) {
1290 data->start_class->flags &= ~ANYOF_EOS; /* No match on empty */
1292 /* Some of the logic below assumes that switching
1293 locale on will only add false positives. */
1294 switch (PL_regkind[(U8)OP(scan)]) {
1298 /* Perl_croak(aTHX_ "panic: unexpected simple REx opcode %d", OP(scan)); */
1299 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
1300 cl_anything(pRExC_state, data->start_class);
1303 if (OP(scan) == SANY)
1305 if (flags & SCF_DO_STCLASS_OR) { /* Everything but \n */
1306 value = (ANYOF_BITMAP_TEST(data->start_class,'\n')
1307 || (data->start_class->flags & ANYOF_CLASS));
1308 cl_anything(pRExC_state, data->start_class);
1310 if (flags & SCF_DO_STCLASS_AND || !value)
1311 ANYOF_BITMAP_CLEAR(data->start_class,'\n');
1314 if (flags & SCF_DO_STCLASS_AND)
1315 cl_and(data->start_class,
1316 (struct regnode_charclass_class*)scan);
1318 cl_or(pRExC_state, data->start_class,
1319 (struct regnode_charclass_class*)scan);
1322 if (flags & SCF_DO_STCLASS_AND) {
1323 if (!(data->start_class->flags & ANYOF_LOCALE)) {
1324 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
1325 for (value = 0; value < 256; value++)
1326 if (!isALNUM(value))
1327 ANYOF_BITMAP_CLEAR(data->start_class, value);
1331 if (data->start_class->flags & ANYOF_LOCALE)
1332 ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
1334 for (value = 0; value < 256; value++)
1336 ANYOF_BITMAP_SET(data->start_class, value);
1341 if (flags & SCF_DO_STCLASS_AND) {
1342 if (data->start_class->flags & ANYOF_LOCALE)
1343 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
1346 ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
1347 data->start_class->flags |= ANYOF_LOCALE;
1351 if (flags & SCF_DO_STCLASS_AND) {
1352 if (!(data->start_class->flags & ANYOF_LOCALE)) {
1353 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
1354 for (value = 0; value < 256; value++)
1356 ANYOF_BITMAP_CLEAR(data->start_class, value);
1360 if (data->start_class->flags & ANYOF_LOCALE)
1361 ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
1363 for (value = 0; value < 256; value++)
1364 if (!isALNUM(value))
1365 ANYOF_BITMAP_SET(data->start_class, value);
1370 if (flags & SCF_DO_STCLASS_AND) {
1371 if (data->start_class->flags & ANYOF_LOCALE)
1372 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
1375 data->start_class->flags |= ANYOF_LOCALE;
1376 ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
1380 if (flags & SCF_DO_STCLASS_AND) {
1381 if (!(data->start_class->flags & ANYOF_LOCALE)) {
1382 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
1383 for (value = 0; value < 256; value++)
1384 if (!isSPACE(value))
1385 ANYOF_BITMAP_CLEAR(data->start_class, value);
1389 if (data->start_class->flags & ANYOF_LOCALE)
1390 ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
1392 for (value = 0; value < 256; value++)
1394 ANYOF_BITMAP_SET(data->start_class, value);
1399 if (flags & SCF_DO_STCLASS_AND) {
1400 if (data->start_class->flags & ANYOF_LOCALE)
1401 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
1404 data->start_class->flags |= ANYOF_LOCALE;
1405 ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
1409 if (flags & SCF_DO_STCLASS_AND) {
1410 if (!(data->start_class->flags & ANYOF_LOCALE)) {
1411 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
1412 for (value = 0; value < 256; value++)
1414 ANYOF_BITMAP_CLEAR(data->start_class, value);
1418 if (data->start_class->flags & ANYOF_LOCALE)
1419 ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
1421 for (value = 0; value < 256; value++)
1422 if (!isSPACE(value))
1423 ANYOF_BITMAP_SET(data->start_class, value);
1428 if (flags & SCF_DO_STCLASS_AND) {
1429 if (data->start_class->flags & ANYOF_LOCALE) {
1430 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
1431 for (value = 0; value < 256; value++)
1432 if (!isSPACE(value))
1433 ANYOF_BITMAP_CLEAR(data->start_class, value);
1437 data->start_class->flags |= ANYOF_LOCALE;
1438 ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
1442 if (flags & SCF_DO_STCLASS_AND) {
1443 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NDIGIT);
1444 for (value = 0; value < 256; value++)
1445 if (!isDIGIT(value))
1446 ANYOF_BITMAP_CLEAR(data->start_class, value);
1449 if (data->start_class->flags & ANYOF_LOCALE)
1450 ANYOF_CLASS_SET(data->start_class,ANYOF_DIGIT);
1452 for (value = 0; value < 256; value++)
1454 ANYOF_BITMAP_SET(data->start_class, value);
1459 if (flags & SCF_DO_STCLASS_AND) {
1460 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_DIGIT);
1461 for (value = 0; value < 256; value++)
1463 ANYOF_BITMAP_CLEAR(data->start_class, value);
1466 if (data->start_class->flags & ANYOF_LOCALE)
1467 ANYOF_CLASS_SET(data->start_class,ANYOF_NDIGIT);
1469 for (value = 0; value < 256; value++)
1470 if (!isDIGIT(value))
1471 ANYOF_BITMAP_SET(data->start_class, value);
1476 if (flags & SCF_DO_STCLASS_OR)
1477 cl_and(data->start_class, &and_with);
1478 flags &= ~SCF_DO_STCLASS;
1481 else if (PL_regkind[(U8)OP(scan)] == EOL && flags & SCF_DO_SUBSTR) {
1482 data->flags |= (OP(scan) == MEOL
1486 else if ( PL_regkind[(U8)OP(scan)] == BRANCHJ
1487 /* Lookbehind, or need to calculate parens/evals/stclass: */
1488 && (scan->flags || data || (flags & SCF_DO_STCLASS))
1489 && (OP(scan) == IFMATCH || OP(scan) == UNLESSM)) {
1490 /* Lookahead/lookbehind */
1491 I32 deltanext, minnext, fake = 0;
1493 struct regnode_charclass_class intrnl;
1496 data_fake.flags = 0;
1498 data_fake.whilem_c = data->whilem_c;
1499 data_fake.last_closep = data->last_closep;
1502 data_fake.last_closep = &fake;
1503 if ( flags & SCF_DO_STCLASS && !scan->flags
1504 && OP(scan) == IFMATCH ) { /* Lookahead */
1505 cl_init(pRExC_state, &intrnl);
1506 data_fake.start_class = &intrnl;
1507 f |= SCF_DO_STCLASS_AND;
1509 if (flags & SCF_WHILEM_VISITED_POS)
1510 f |= SCF_WHILEM_VISITED_POS;
1511 next = regnext(scan);
1512 nscan = NEXTOPER(NEXTOPER(scan));
1513 minnext = study_chunk(pRExC_state, &nscan, &deltanext, last, &data_fake, f);
1516 vFAIL("Variable length lookbehind not implemented");
1518 else if (minnext > U8_MAX) {
1519 vFAIL2("Lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX);
1521 scan->flags = minnext;
1523 if (data && data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
1525 if (data && (data_fake.flags & SF_HAS_EVAL))
1526 data->flags |= SF_HAS_EVAL;
1528 data->whilem_c = data_fake.whilem_c;
1529 if (f & SCF_DO_STCLASS_AND) {
1530 int was = (data->start_class->flags & ANYOF_EOS);
1532 cl_and(data->start_class, &intrnl);
1534 data->start_class->flags |= ANYOF_EOS;
1537 else if (OP(scan) == OPEN) {
1540 else if (OP(scan) == CLOSE) {
1541 if (ARG(scan) == is_par) {
1542 next = regnext(scan);
1544 if ( next && (OP(next) != WHILEM) && next < last)
1545 is_par = 0; /* Disable optimization */
1548 *(data->last_closep) = ARG(scan);
1550 else if (OP(scan) == EVAL) {
1552 data->flags |= SF_HAS_EVAL;
1554 else if (OP(scan) == LOGICAL && scan->flags == 2) { /* Embedded follows */
1555 if (flags & SCF_DO_SUBSTR) {
1556 scan_commit(pRExC_state,data);
1557 data->longest = &(data->longest_float);
1559 is_inf = is_inf_internal = 1;
1560 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
1561 cl_anything(pRExC_state, data->start_class);
1562 flags &= ~SCF_DO_STCLASS;
1564 /* Else: zero-length, ignore. */
1565 scan = regnext(scan);
1570 *deltap = is_inf_internal ? I32_MAX : delta;
1571 if (flags & SCF_DO_SUBSTR && is_inf)
1572 data->pos_delta = I32_MAX - data->pos_min;
1573 if (is_par > U8_MAX)
1575 if (is_par && pars==1 && data) {
1576 data->flags |= SF_IN_PAR;
1577 data->flags &= ~SF_HAS_PAR;
1579 else if (pars && data) {
1580 data->flags |= SF_HAS_PAR;
1581 data->flags &= ~SF_IN_PAR;
1583 if (flags & SCF_DO_STCLASS_OR)
1584 cl_and(data->start_class, &and_with);
1589 S_add_data(pTHX_ RExC_state_t *pRExC_state, I32 n, char *s)
1591 if (RExC_rx->data) {
1592 Renewc(RExC_rx->data,
1593 sizeof(*RExC_rx->data) + sizeof(void*) * (RExC_rx->data->count + n - 1),
1594 char, struct reg_data);
1595 Renew(RExC_rx->data->what, RExC_rx->data->count + n, U8);
1596 RExC_rx->data->count += n;
1599 Newc(1207, RExC_rx->data, sizeof(*RExC_rx->data) + sizeof(void*) * (n - 1),
1600 char, struct reg_data);
1601 New(1208, RExC_rx->data->what, n, U8);
1602 RExC_rx->data->count = n;
1604 Copy(s, RExC_rx->data->what + RExC_rx->data->count - n, n, U8);
1605 return RExC_rx->data->count - n;
1609 Perl_reginitcolors(pTHX)
1612 char *s = PerlEnv_getenv("PERL_RE_COLORS");
1615 PL_colors[0] = s = savepv(s);
1617 s = strchr(s, '\t');
1623 PL_colors[i] = s = "";
1627 PL_colors[i++] = "";
1634 - pregcomp - compile a regular expression into internal code
1636 * We can't allocate space until we know how big the compiled form will be,
1637 * but we can't compile it (and thus know how big it is) until we've got a
1638 * place to put the code. So we cheat: we compile it twice, once with code
1639 * generation turned off and size counting turned on, and once "for real".
1640 * This also means that we don't allocate space until we are sure that the
1641 * thing really will compile successfully, and we never have to move the
1642 * code and thus invalidate pointers into it. (Note that it has to be in
1643 * one piece because free() must be able to free it all.) [NB: not true in perl]
1645 * Beware that the optimization-preparation code in here knows about some
1646 * of the structure of the compiled regexp. [I'll say.]
1649 Perl_pregcomp(pTHX_ char *exp, char *xend, PMOP *pm)
1659 RExC_state_t RExC_state;
1660 RExC_state_t *pRExC_state = &RExC_state;
1663 FAIL("NULL regexp argument");
1665 /* XXXX This looks very suspicious... */
1666 if (pm->op_pmdynflags & PMdf_CMP_UTF8)
1672 DEBUG_r(if (!PL_colorset) reginitcolors());
1673 DEBUG_r(PerlIO_printf(Perl_debug_log, "%sCompiling REx%s `%s%*s%s'\n",
1674 PL_colors[4],PL_colors[5],PL_colors[0],
1675 (int)(xend - exp), RExC_precomp, PL_colors[1]));
1676 RExC_flags16 = pm->op_pmflags;
1680 RExC_seen_zerolen = *exp == '^' ? -1 : 0;
1681 RExC_seen_evals = 0;
1684 /* First pass: determine size, legality. */
1691 RExC_emit = &PL_regdummy;
1692 RExC_whilem_seen = 0;
1693 #if 0 /* REGC() is (currently) a NOP at the first pass.
1694 * Clever compilers notice this and complain. --jhi */
1695 REGC((U8)REG_MAGIC, (char*)RExC_emit);
1697 if (reg(pRExC_state, 0, &flags) == NULL) {
1698 RExC_precomp = Nullch;
1701 DEBUG_r(PerlIO_printf(Perl_debug_log, "size %"IVdf" ", (IV)RExC_size));
1703 /* Small enough for pointer-storage convention?
1704 If extralen==0, this means that we will not need long jumps. */
1705 if (RExC_size >= 0x10000L && RExC_extralen)
1706 RExC_size += RExC_extralen;
1709 if (RExC_whilem_seen > 15)
1710 RExC_whilem_seen = 15;
1712 /* Allocate space and initialize. */
1713 Newc(1001, r, sizeof(regexp) + (unsigned)RExC_size * sizeof(regnode),
1716 FAIL("Regexp out of space");
1719 /* avoid reading uninitialized memory in DEBUGGING code in study_chunk() */
1720 Zero(r, sizeof(regexp) + (unsigned)RExC_size * sizeof(regnode), char);
1723 r->prelen = xend - exp;
1724 r->precomp = savepvn(RExC_precomp, r->prelen);
1726 r->reganch = pm->op_pmflags & PMf_COMPILETIME;
1727 r->nparens = RExC_npar - 1; /* set early to validate backrefs */
1729 r->substrs = 0; /* Useful during FAIL. */
1730 r->startp = 0; /* Useful during FAIL. */
1731 r->endp = 0; /* Useful during FAIL. */
1733 Newz(1304, r->offsets, 2*RExC_size+1, U32); /* MJD 20001228 */
1735 r->offsets[0] = RExC_size;
1737 DEBUG_r(PerlIO_printf(Perl_debug_log,
1738 "%s %"UVuf" bytes for offset annotations.\n",
1739 r->offsets ? "Got" : "Couldn't get",
1740 (UV)((2*RExC_size+1) * sizeof(U32))));
1744 /* Second pass: emit code. */
1745 RExC_flags16 = pm->op_pmflags; /* don't let top level (?i) bleed */
1750 RExC_emit_start = r->program;
1751 RExC_emit = r->program;
1752 /* Store the count of eval-groups for security checks: */
1753 RExC_emit->next_off = ((RExC_seen_evals > U16_MAX) ? U16_MAX : RExC_seen_evals);
1754 REGC((U8)REG_MAGIC, (char*) RExC_emit++);
1756 if (reg(pRExC_state, 0, &flags) == NULL)
1759 /* Dig out information for optimizations. */
1760 r->reganch = pm->op_pmflags & PMf_COMPILETIME; /* Again? */
1761 pm->op_pmflags = RExC_flags16;
1763 r->reganch |= ROPT_UTF8; /* Unicode in it? */
1764 r->regstclass = NULL;
1765 if (RExC_naughty >= 10) /* Probably an expensive pattern. */
1766 r->reganch |= ROPT_NAUGHTY;
1767 scan = r->program + 1; /* First BRANCH. */
1769 /* XXXX To minimize changes to RE engine we always allocate
1770 3-units-long substrs field. */
1771 Newz(1004, r->substrs, 1, struct reg_substr_data);
1773 StructCopy(&zero_scan_data, &data, scan_data_t);
1774 /* XXXX Should not we check for something else? Usually it is OPEN1... */
1775 if (OP(scan) != BRANCH) { /* Only one top-level choice. */
1777 STRLEN longest_float_length, longest_fixed_length;
1778 struct regnode_charclass_class ch_class;
1783 /* Skip introductions and multiplicators >= 1. */
1784 while ((OP(first) == OPEN && (sawopen = 1)) ||
1785 /* An OR of *one* alternative - should not happen now. */
1786 (OP(first) == BRANCH && OP(regnext(first)) != BRANCH) ||
1787 (OP(first) == PLUS) ||
1788 (OP(first) == MINMOD) ||
1789 /* An {n,m} with n>0 */
1790 (PL_regkind[(U8)OP(first)] == CURLY && ARG1(first) > 0) ) {
1791 if (OP(first) == PLUS)
1794 first += regarglen[(U8)OP(first)];
1795 first = NEXTOPER(first);
1798 /* Starting-point info. */
1800 if (PL_regkind[(U8)OP(first)] == EXACT) {
1801 if (OP(first) == EXACT)
1802 ; /* Empty, get anchored substr later. */
1803 else if ((OP(first) == EXACTF || OP(first) == EXACTFL))
1804 r->regstclass = first;
1806 else if (strchr((char*)PL_simple,OP(first)))
1807 r->regstclass = first;
1808 else if (PL_regkind[(U8)OP(first)] == BOUND ||
1809 PL_regkind[(U8)OP(first)] == NBOUND)
1810 r->regstclass = first;
1811 else if (PL_regkind[(U8)OP(first)] == BOL) {
1812 r->reganch |= (OP(first) == MBOL
1814 : (OP(first) == SBOL
1817 first = NEXTOPER(first);
1820 else if (OP(first) == GPOS) {
1821 r->reganch |= ROPT_ANCH_GPOS;
1822 first = NEXTOPER(first);
1825 else if ((OP(first) == STAR &&
1826 PL_regkind[(U8)OP(NEXTOPER(first))] == REG_ANY) &&
1827 !(r->reganch & ROPT_ANCH) )
1829 /* turn .* into ^.* with an implied $*=1 */
1830 int type = OP(NEXTOPER(first));
1832 if (type == REG_ANY)
1833 type = ROPT_ANCH_MBOL;
1835 type = ROPT_ANCH_SBOL;
1837 r->reganch |= type | ROPT_IMPLICIT;
1838 first = NEXTOPER(first);
1841 if (sawplus && (!sawopen || !RExC_sawback)
1842 && !(RExC_seen & REG_SEEN_EVAL)) /* May examine pos and $& */
1843 /* x+ must match at the 1st pos of run of x's */
1844 r->reganch |= ROPT_SKIP;
1846 /* Scan is after the zeroth branch, first is atomic matcher. */
1847 DEBUG_r(PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n",
1848 (IV)(first - scan + 1)));
1850 * If there's something expensive in the r.e., find the
1851 * longest literal string that must appear and make it the
1852 * regmust. Resolve ties in favor of later strings, since
1853 * the regstart check works with the beginning of the r.e.
1854 * and avoiding duplication strengthens checking. Not a
1855 * strong reason, but sufficient in the absence of others.
1856 * [Now we resolve ties in favor of the earlier string if
1857 * it happens that c_offset_min has been invalidated, since the
1858 * earlier string may buy us something the later one won't.]
1862 data.longest_fixed = newSVpvn("",0);
1863 data.longest_float = newSVpvn("",0);
1864 data.last_found = newSVpvn("",0);
1865 data.longest = &(data.longest_fixed);
1867 if (!r->regstclass) {
1868 cl_init(pRExC_state, &ch_class);
1869 data.start_class = &ch_class;
1870 stclass_flag = SCF_DO_STCLASS_AND;
1871 } else /* XXXX Check for BOUND? */
1873 data.last_closep = &last_close;
1875 minlen = study_chunk(pRExC_state, &first, &fake, scan + RExC_size, /* Up to end */
1876 &data, SCF_DO_SUBSTR | SCF_WHILEM_VISITED_POS | stclass_flag);
1877 if ( RExC_npar == 1 && data.longest == &(data.longest_fixed)
1878 && data.last_start_min == 0 && data.last_end > 0
1879 && !RExC_seen_zerolen
1880 && (!(RExC_seen & REG_SEEN_GPOS) || (r->reganch & ROPT_ANCH_GPOS)))
1881 r->reganch |= ROPT_CHECK_ALL;
1882 scan_commit(pRExC_state, &data);
1883 SvREFCNT_dec(data.last_found);
1885 longest_float_length = CHR_SVLEN(data.longest_float);
1886 if (longest_float_length
1887 || (data.flags & SF_FL_BEFORE_EOL
1888 && (!(data.flags & SF_FL_BEFORE_MEOL)
1889 || (RExC_flags16 & PMf_MULTILINE)))) {
1892 if (SvCUR(data.longest_fixed) /* ok to leave SvCUR */
1893 && data.offset_fixed == data.offset_float_min
1894 && SvCUR(data.longest_fixed) == SvCUR(data.longest_float))
1895 goto remove_float; /* As in (a)+. */
1897 r->float_substr = data.longest_float;
1898 r->float_min_offset = data.offset_float_min;
1899 r->float_max_offset = data.offset_float_max;
1900 t = (data.flags & SF_FL_BEFORE_EOL /* Can't have SEOL and MULTI */
1901 && (!(data.flags & SF_FL_BEFORE_MEOL)
1902 || (RExC_flags16 & PMf_MULTILINE)));
1903 fbm_compile(r->float_substr, t ? FBMcf_TAIL : 0);
1907 r->float_substr = Nullsv;
1908 SvREFCNT_dec(data.longest_float);
1909 longest_float_length = 0;
1912 longest_fixed_length = CHR_SVLEN(data.longest_fixed);
1913 if (longest_fixed_length
1914 || (data.flags & SF_FIX_BEFORE_EOL /* Cannot have SEOL and MULTI */
1915 && (!(data.flags & SF_FIX_BEFORE_MEOL)
1916 || (RExC_flags16 & PMf_MULTILINE)))) {
1919 r->anchored_substr = data.longest_fixed;
1920 r->anchored_offset = data.offset_fixed;
1921 t = (data.flags & SF_FIX_BEFORE_EOL /* Can't have SEOL and MULTI */
1922 && (!(data.flags & SF_FIX_BEFORE_MEOL)
1923 || (RExC_flags16 & PMf_MULTILINE)));
1924 fbm_compile(r->anchored_substr, t ? FBMcf_TAIL : 0);
1927 r->anchored_substr = Nullsv;
1928 SvREFCNT_dec(data.longest_fixed);
1929 longest_fixed_length = 0;
1932 && (OP(r->regstclass) == REG_ANY || OP(r->regstclass) == SANY))
1933 r->regstclass = NULL;
1934 if ((!r->anchored_substr || r->anchored_offset) && stclass_flag
1935 && !(data.start_class->flags & ANYOF_EOS)
1936 && !cl_is_anything(data.start_class)) {
1937 I32 n = add_data(pRExC_state, 1, "f");
1939 New(1006, RExC_rx->data->data[n], 1,
1940 struct regnode_charclass_class);
1941 StructCopy(data.start_class,
1942 (struct regnode_charclass_class*)RExC_rx->data->data[n],
1943 struct regnode_charclass_class);
1944 r->regstclass = (regnode*)RExC_rx->data->data[n];
1945 r->reganch &= ~ROPT_SKIP; /* Used in find_byclass(). */
1946 PL_regdata = r->data; /* for regprop() */
1947 DEBUG_r({ SV *sv = sv_newmortal();
1948 regprop(sv, (regnode*)data.start_class);
1949 PerlIO_printf(Perl_debug_log,
1950 "synthetic stclass `%s'.\n",
1954 /* A temporary algorithm prefers floated substr to fixed one to dig more info. */
1955 if (longest_fixed_length > longest_float_length) {
1956 r->check_substr = r->anchored_substr;
1957 r->check_offset_min = r->check_offset_max = r->anchored_offset;
1958 if (r->reganch & ROPT_ANCH_SINGLE)
1959 r->reganch |= ROPT_NOSCAN;
1962 r->check_substr = r->float_substr;
1963 r->check_offset_min = data.offset_float_min;
1964 r->check_offset_max = data.offset_float_max;
1966 /* XXXX Currently intuiting is not compatible with ANCH_GPOS.
1967 This should be changed ASAP! */
1968 if (r->check_substr && !(r->reganch & ROPT_ANCH_GPOS)) {
1969 r->reganch |= RE_USE_INTUIT;
1970 if (SvTAIL(r->check_substr))
1971 r->reganch |= RE_INTUIT_TAIL;
1975 /* Several toplevels. Best we can is to set minlen. */
1977 struct regnode_charclass_class ch_class;
1980 DEBUG_r(PerlIO_printf(Perl_debug_log, "\n"));
1981 scan = r->program + 1;
1982 cl_init(pRExC_state, &ch_class);
1983 data.start_class = &ch_class;
1984 data.last_closep = &last_close;
1985 minlen = study_chunk(pRExC_state, &scan, &fake, scan + RExC_size, &data, SCF_DO_STCLASS_AND|SCF_WHILEM_VISITED_POS);
1986 r->check_substr = r->anchored_substr = r->float_substr = Nullsv;
1987 if (!(data.start_class->flags & ANYOF_EOS)
1988 && !cl_is_anything(data.start_class)) {
1989 I32 n = add_data(pRExC_state, 1, "f");
1991 New(1006, RExC_rx->data->data[n], 1,
1992 struct regnode_charclass_class);
1993 StructCopy(data.start_class,
1994 (struct regnode_charclass_class*)RExC_rx->data->data[n],
1995 struct regnode_charclass_class);
1996 r->regstclass = (regnode*)RExC_rx->data->data[n];
1997 r->reganch &= ~ROPT_SKIP; /* Used in find_byclass(). */
1998 DEBUG_r({ SV* sv = sv_newmortal();
1999 regprop(sv, (regnode*)data.start_class);
2000 PerlIO_printf(Perl_debug_log,
2001 "synthetic stclass `%s'.\n",
2007 if (RExC_seen & REG_SEEN_GPOS)
2008 r->reganch |= ROPT_GPOS_SEEN;
2009 if (RExC_seen & REG_SEEN_LOOKBEHIND)
2010 r->reganch |= ROPT_LOOKBEHIND_SEEN;
2011 if (RExC_seen & REG_SEEN_EVAL)
2012 r->reganch |= ROPT_EVAL_SEEN;
2013 if (RExC_seen & REG_SEEN_CANY)
2014 r->reganch |= ROPT_CANY_SEEN;
2015 Newz(1002, r->startp, RExC_npar, I32);
2016 Newz(1002, r->endp, RExC_npar, I32);
2017 PL_regdata = r->data; /* for regprop() */
2018 DEBUG_r(regdump(r));
2023 - reg - regular expression, i.e. main body or parenthesized thing
2025 * Caller must absorb opening parenthesis.
2027 * Combining parenthesis handling with the base level of regular expression
2028 * is a trifle forced, but the need to tie the tails of the branches to what
2029 * follows makes it hard to avoid.
2032 S_reg(pTHX_ RExC_state_t *pRExC_state, I32 paren, I32 *flagp)
2033 /* paren: Parenthesized? 0=top, 1=(, inside: changed to letter. */
2035 register regnode *ret; /* Will be the head of the group. */
2036 register regnode *br;
2037 register regnode *lastbr;
2038 register regnode *ender = 0;
2039 register I32 parno = 0;
2040 I32 flags, oregflags = RExC_flags16, have_branch = 0, open = 0;
2042 /* for (?g), (?gc), and (?o) warnings; warning
2043 about (?c) will warn about (?g) -- japhy */
2045 I32 wastedflags = 0x00,
2048 wasted_gc = 0x02 | 0x04,
2051 char * parse_start = RExC_parse; /* MJD */
2052 char *oregcomp_parse = RExC_parse;
2055 *flagp = 0; /* Tentatively. */
2058 /* Make an OPEN node, if parenthesized. */
2060 if (*RExC_parse == '?') { /* (?...) */
2061 U16 posflags = 0, negflags = 0;
2062 U16 *flagsp = &posflags;
2064 char *seqstart = RExC_parse;
2067 paren = *RExC_parse++;
2068 ret = NULL; /* For look-ahead/behind. */
2070 case '<': /* (?<...) */
2071 RExC_seen |= REG_SEEN_LOOKBEHIND;
2072 if (*RExC_parse == '!')
2074 if (*RExC_parse != '=' && *RExC_parse != '!')
2077 case '=': /* (?=...) */
2078 case '!': /* (?!...) */
2079 RExC_seen_zerolen++;
2080 case ':': /* (?:...) */
2081 case '>': /* (?>...) */
2083 case '$': /* (?$...) */
2084 case '@': /* (?@...) */
2085 vFAIL2("Sequence (?%c...) not implemented", (int)paren);
2087 case '#': /* (?#...) */
2088 while (*RExC_parse && *RExC_parse != ')')
2090 if (*RExC_parse != ')')
2091 FAIL("Sequence (?#... not terminated");
2092 nextchar(pRExC_state);
2095 case 'p': /* (?p...) */
2096 if (SIZE_ONLY && ckWARN2(WARN_DEPRECATED, WARN_REGEXP))
2097 vWARNdep(RExC_parse, "(?p{}) is deprecated - use (??{})");
2099 case '?': /* (??...) */
2101 paren = *RExC_parse++;
2103 case '{': /* (?{...}) */
2105 I32 count = 1, n = 0;
2107 char *s = RExC_parse;
2109 OP_4tree *sop, *rop;
2111 RExC_seen_zerolen++;
2112 RExC_seen |= REG_SEEN_EVAL;
2113 while (count && (c = *RExC_parse)) {
2114 if (c == '\\' && RExC_parse[1])
2122 if (*RExC_parse != ')')
2125 vFAIL("Sequence (?{...}) not terminated or not {}-balanced");
2130 if (RExC_parse - 1 - s)
2131 sv = newSVpvn(s, RExC_parse - 1 - s);
2133 sv = newSVpvn("", 0);
2136 Perl_save_re_context(aTHX);
2137 rop = sv_compile_2op(sv, &sop, "re", &av);
2138 sop->op_private |= OPpREFCOUNTED;
2139 /* re_dup will OpREFCNT_inc */
2140 OpREFCNT_set(sop, 1);
2143 n = add_data(pRExC_state, 3, "nop");
2144 RExC_rx->data->data[n] = (void*)rop;
2145 RExC_rx->data->data[n+1] = (void*)sop;
2146 RExC_rx->data->data[n+2] = (void*)av;
2149 else { /* First pass */
2150 if (PL_reginterp_cnt < ++RExC_seen_evals
2151 && PL_curcop != &PL_compiling)
2152 /* No compiled RE interpolated, has runtime
2153 components ===> unsafe. */
2154 FAIL("Eval-group not allowed at runtime, use re 'eval'");
2155 if (PL_tainting && PL_tainted)
2156 FAIL("Eval-group in insecure regular expression");
2159 nextchar(pRExC_state);
2161 ret = reg_node(pRExC_state, LOGICAL);
2164 regtail(pRExC_state, ret, reganode(pRExC_state, EVAL, n));
2165 /* deal with the length of this later - MJD */
2168 return reganode(pRExC_state, EVAL, n);
2170 case '(': /* (?(?{...})...) and (?(?=...)...) */
2172 if (RExC_parse[0] == '?') { /* (?(?...)) */
2173 if (RExC_parse[1] == '=' || RExC_parse[1] == '!'
2174 || RExC_parse[1] == '<'
2175 || RExC_parse[1] == '{') { /* Lookahead or eval. */
2178 ret = reg_node(pRExC_state, LOGICAL);
2181 regtail(pRExC_state, ret, reg(pRExC_state, 1, &flag));
2185 else if (RExC_parse[0] >= '1' && RExC_parse[0] <= '9' ) {
2187 parno = atoi(RExC_parse++);
2189 while (isDIGIT(*RExC_parse))
2191 ret = reganode(pRExC_state, GROUPP, parno);
2193 if ((c = *nextchar(pRExC_state)) != ')')
2194 vFAIL("Switch condition not recognized");
2196 regtail(pRExC_state, ret, reganode(pRExC_state, IFTHEN, 0));
2197 br = regbranch(pRExC_state, &flags, 1);
2199 br = reganode(pRExC_state, LONGJMP, 0);
2201 regtail(pRExC_state, br, reganode(pRExC_state, LONGJMP, 0));
2202 c = *nextchar(pRExC_state);
2206 lastbr = reganode(pRExC_state, IFTHEN, 0); /* Fake one for optimizer. */
2207 regbranch(pRExC_state, &flags, 1);
2208 regtail(pRExC_state, ret, lastbr);
2211 c = *nextchar(pRExC_state);
2216 vFAIL("Switch (?(condition)... contains too many branches");
2217 ender = reg_node(pRExC_state, TAIL);
2218 regtail(pRExC_state, br, ender);
2220 regtail(pRExC_state, lastbr, ender);
2221 regtail(pRExC_state, NEXTOPER(NEXTOPER(lastbr)), ender);
2224 regtail(pRExC_state, ret, ender);
2228 vFAIL2("Unknown switch condition (?(%.2s", RExC_parse);
2232 RExC_parse--; /* for vFAIL to print correctly */
2233 vFAIL("Sequence (? incomplete");
2237 parse_flags: /* (?i) */
2238 while (*RExC_parse && strchr("iogcmsx", *RExC_parse)) {
2239 /* (?g), (?gc) and (?o) are useless here
2240 and must be globally applied -- japhy */
2242 if (*RExC_parse == 'o' || *RExC_parse == 'g') {
2243 if (SIZE_ONLY && ckWARN(WARN_REGEXP)) {
2244 I32 wflagbit = *RExC_parse == 'o' ? wasted_o : wasted_g;
2245 if (! (wastedflags & wflagbit) ) {
2246 wastedflags |= wflagbit;
2249 "Useless (%s%c) - %suse /%c modifier",
2250 flagsp == &negflags ? "?-" : "?",
2252 flagsp == &negflags ? "don't " : "",
2258 else if (*RExC_parse == 'c') {
2259 if (SIZE_ONLY && ckWARN(WARN_REGEXP)) {
2260 if (! (wastedflags & wasted_c) ) {
2261 wastedflags |= wasted_gc;
2264 "Useless (%sc) - %suse /gc modifier",
2265 flagsp == &negflags ? "?-" : "?",
2266 flagsp == &negflags ? "don't " : ""
2271 else { pmflag(flagsp, *RExC_parse); }
2275 if (*RExC_parse == '-') {
2277 wastedflags = 0; /* reset so (?g-c) warns twice */
2281 RExC_flags16 |= posflags;
2282 RExC_flags16 &= ~negflags;
2283 if (*RExC_parse == ':') {
2289 if (*RExC_parse != ')') {
2291 vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
2293 nextchar(pRExC_state);
2301 ret = reganode(pRExC_state, OPEN, parno);
2302 Set_Node_Length(ret, 1); /* MJD */
2303 Set_Node_Offset(ret, RExC_parse); /* MJD */
2310 /* Pick up the branches, linking them together. */
2311 parse_start = RExC_parse; /* MJD */
2312 br = regbranch(pRExC_state, &flags, 1);
2313 /* branch_len = (paren != 0); */
2317 if (*RExC_parse == '|') {
2318 if (!SIZE_ONLY && RExC_extralen) {
2319 reginsert(pRExC_state, BRANCHJ, br);
2322 reginsert(pRExC_state, BRANCH, br);
2323 Set_Node_Length(br, paren != 0);
2324 Set_Node_Offset_To_R(br-RExC_emit_start, parse_start-RExC_start);
2328 RExC_extralen += 1; /* For BRANCHJ-BRANCH. */
2330 else if (paren == ':') {
2331 *flagp |= flags&SIMPLE;
2333 if (open) { /* Starts with OPEN. */
2334 regtail(pRExC_state, ret, br); /* OPEN -> first. */
2336 else if (paren != '?') /* Not Conditional */
2340 *flagp |= flags&SPSTART;
2342 while (*RExC_parse == '|') {
2343 if (!SIZE_ONLY && RExC_extralen) {
2344 ender = reganode(pRExC_state, LONGJMP,0);
2345 regtail(pRExC_state, NEXTOPER(NEXTOPER(lastbr)), ender); /* Append to the previous. */
2348 RExC_extralen += 2; /* Account for LONGJMP. */
2349 nextchar(pRExC_state);
2350 br = regbranch(pRExC_state, &flags, 0);
2354 regtail(pRExC_state, lastbr, br); /* BRANCH -> BRANCH. */
2358 *flagp |= flags&SPSTART;
2361 if (have_branch || paren != ':') {
2362 /* Make a closing node, and hook it on the end. */
2365 ender = reg_node(pRExC_state, TAIL);
2368 ender = reganode(pRExC_state, CLOSE, parno);
2369 Set_Node_Offset(ender,RExC_parse+1); /* MJD */
2370 Set_Node_Length(ender,1); /* MJD */
2376 *flagp &= ~HASWIDTH;
2379 ender = reg_node(pRExC_state, SUCCEED);
2382 ender = reg_node(pRExC_state, END);
2385 regtail(pRExC_state, lastbr, ender);
2388 /* Hook the tails of the branches to the closing node. */
2389 for (br = ret; br != NULL; br = regnext(br)) {
2390 regoptail(pRExC_state, br, ender);
2397 static char parens[] = "=!<,>";
2399 if (paren && (p = strchr(parens, paren))) {
2400 int node = ((p - parens) % 2) ? UNLESSM : IFMATCH;
2401 int flag = (p - parens) > 1;
2404 node = SUSPEND, flag = 0;
2405 reginsert(pRExC_state, node,ret);
2407 regtail(pRExC_state, ret, reg_node(pRExC_state, TAIL));
2411 /* Check for proper termination. */
2413 RExC_flags16 = oregflags;
2414 if (RExC_parse >= RExC_end || *nextchar(pRExC_state) != ')') {
2415 RExC_parse = oregcomp_parse;
2416 vFAIL("Unmatched (");
2419 else if (!paren && RExC_parse < RExC_end) {
2420 if (*RExC_parse == ')') {
2422 vFAIL("Unmatched )");
2425 FAIL("Junk on end of regexp"); /* "Can't happen". */
2433 - regbranch - one alternative of an | operator
2435 * Implements the concatenation operator.
2438 S_regbranch(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, I32 first)
2440 register regnode *ret;
2441 register regnode *chain = NULL;
2442 register regnode *latest;
2443 I32 flags = 0, c = 0;
2448 if (!SIZE_ONLY && RExC_extralen)
2449 ret = reganode(pRExC_state, BRANCHJ,0);
2451 ret = reg_node(pRExC_state, BRANCH);
2452 Set_Node_Length(ret, 1);
2456 if (!first && SIZE_ONLY)
2457 RExC_extralen += 1; /* BRANCHJ */
2459 *flagp = WORST; /* Tentatively. */
2462 nextchar(pRExC_state);
2463 while (RExC_parse < RExC_end && *RExC_parse != '|' && *RExC_parse != ')') {
2465 latest = regpiece(pRExC_state, &flags);
2466 if (latest == NULL) {
2467 if (flags & TRYAGAIN)
2471 else if (ret == NULL)
2473 *flagp |= flags&HASWIDTH;
2474 if (chain == NULL) /* First piece. */
2475 *flagp |= flags&SPSTART;
2478 regtail(pRExC_state, chain, latest);
2483 if (chain == NULL) { /* Loop ran zero times. */
2484 chain = reg_node(pRExC_state, NOTHING);
2489 *flagp |= flags&SIMPLE;
2496 - regpiece - something followed by possible [*+?]
2498 * Note that the branching code sequences used for ? and the general cases
2499 * of * and + are somewhat optimized: they use the same NOTHING node as
2500 * both the endmarker for their branch list and the body of the last branch.
2501 * It might seem that this node could be dispensed with entirely, but the
2502 * endmarker role is not redundant.
2505 S_regpiece(pTHX_ RExC_state_t *pRExC_state, I32 *flagp)
2507 register regnode *ret;
2509 register char *next;
2511 char *origparse = RExC_parse;
2514 I32 max = REG_INFTY;
2517 ret = regatom(pRExC_state, &flags);
2519 if (flags & TRYAGAIN)
2526 if (op == '{' && regcurly(RExC_parse)) {
2527 parse_start = RExC_parse; /* MJD */
2528 next = RExC_parse + 1;
2530 while (isDIGIT(*next) || *next == ',') {
2539 if (*next == '}') { /* got one */
2543 min = atoi(RExC_parse);
2547 maxpos = RExC_parse;
2549 if (!max && *maxpos != '0')
2550 max = REG_INFTY; /* meaning "infinity" */
2551 else if (max >= REG_INFTY)
2552 vFAIL2("Quantifier in {,} bigger than %d", REG_INFTY - 1);
2554 nextchar(pRExC_state);
2557 if ((flags&SIMPLE)) {
2558 RExC_naughty += 2 + RExC_naughty / 2;
2559 reginsert(pRExC_state, CURLY, ret);
2560 Set_Node_Offset(ret, parse_start+1); /* MJD */
2561 Set_Node_Cur_Length(ret);
2564 regnode *w = reg_node(pRExC_state, WHILEM);
2567 regtail(pRExC_state, ret, w);
2568 if (!SIZE_ONLY && RExC_extralen) {
2569 reginsert(pRExC_state, LONGJMP,ret);
2570 reginsert(pRExC_state, NOTHING,ret);
2571 NEXT_OFF(ret) = 3; /* Go over LONGJMP. */
2573 reginsert(pRExC_state, CURLYX,ret);
2575 Set_Node_Offset(ret, parse_start+1);
2576 Set_Node_Length(ret,
2577 op == '{' ? (RExC_parse - parse_start) : 1);
2579 if (!SIZE_ONLY && RExC_extralen)
2580 NEXT_OFF(ret) = 3; /* Go over NOTHING to LONGJMP. */
2581 regtail(pRExC_state, ret, reg_node(pRExC_state, NOTHING));
2583 RExC_whilem_seen++, RExC_extralen += 3;
2584 RExC_naughty += 4 + RExC_naughty; /* compound interest */
2592 if (max && max < min)
2593 vFAIL("Can't do {n,m} with n > m");
2608 #if 0 /* Now runtime fix should be reliable. */
2610 /* if this is reinstated, don't forget to put this back into perldiag:
2612 =item Regexp *+ operand could be empty at {#} in regex m/%s/
2614 (F) The part of the regexp subject to either the * or + quantifier
2615 could match an empty string. The {#} shows in the regular
2616 expression about where the problem was discovered.
2620 if (!(flags&HASWIDTH) && op != '?')
2621 vFAIL("Regexp *+ operand could be empty");
2624 parse_start = RExC_parse;
2625 nextchar(pRExC_state);
2627 *flagp = (op != '+') ? (WORST|SPSTART|HASWIDTH) : (WORST|HASWIDTH);
2629 if (op == '*' && (flags&SIMPLE)) {
2630 reginsert(pRExC_state, STAR, ret);
2634 else if (op == '*') {
2638 else if (op == '+' && (flags&SIMPLE)) {
2639 reginsert(pRExC_state, PLUS, ret);
2643 else if (op == '+') {
2647 else if (op == '?') {
2652 if (ckWARN(WARN_REGEXP) && !SIZE_ONLY && !(flags&HASWIDTH) && max > REG_INFTY/3) {
2654 "%.*s matches null string many times",
2655 RExC_parse - origparse,
2659 if (*RExC_parse == '?') {
2660 nextchar(pRExC_state);
2661 reginsert(pRExC_state, MINMOD, ret);
2662 regtail(pRExC_state, ret, ret + NODE_STEP_REGNODE);
2664 if (ISMULT2(RExC_parse)) {
2666 vFAIL("Nested quantifiers");
2673 - regatom - the lowest level
2675 * Optimization: gobbles an entire sequence of ordinary characters so that
2676 * it can turn them into a single node, which is smaller to store and
2677 * faster to run. Backslashed characters are exceptions, each becoming a
2678 * separate node; the code is simpler that way and it's not worth fixing.
2680 * [Yes, it is worth fixing, some scripts can run twice the speed.] */
2682 S_regatom(pTHX_ RExC_state_t *pRExC_state, I32 *flagp)
2684 register regnode *ret = 0;
2686 char *parse_start = 0;
2688 *flagp = WORST; /* Tentatively. */
2691 switch (*RExC_parse) {
2693 RExC_seen_zerolen++;
2694 nextchar(pRExC_state);
2695 if (RExC_flags16 & PMf_MULTILINE)
2696 ret = reg_node(pRExC_state, MBOL);
2697 else if (RExC_flags16 & PMf_SINGLELINE)
2698 ret = reg_node(pRExC_state, SBOL);
2700 ret = reg_node(pRExC_state, BOL);
2701 Set_Node_Length(ret, 1); /* MJD */
2704 nextchar(pRExC_state);
2706 RExC_seen_zerolen++;
2707 if (RExC_flags16 & PMf_MULTILINE)
2708 ret = reg_node(pRExC_state, MEOL);
2709 else if (RExC_flags16 & PMf_SINGLELINE)
2710 ret = reg_node(pRExC_state, SEOL);
2712 ret = reg_node(pRExC_state, EOL);
2713 Set_Node_Length(ret, 1); /* MJD */
2716 nextchar(pRExC_state);
2717 if (RExC_flags16 & PMf_SINGLELINE)
2718 ret = reg_node(pRExC_state, SANY);
2720 ret = reg_node(pRExC_state, REG_ANY);
2721 *flagp |= HASWIDTH|SIMPLE;
2723 Set_Node_Length(ret, 1); /* MJD */
2727 char *oregcomp_parse = ++RExC_parse;
2728 ret = regclass(pRExC_state);
2729 if (*RExC_parse != ']') {
2730 RExC_parse = oregcomp_parse;
2731 vFAIL("Unmatched [");
2733 nextchar(pRExC_state);
2734 *flagp |= HASWIDTH|SIMPLE;
2735 Set_Node_Length(ret, RExC_parse - oregcomp_parse + 1); /* MJD */
2739 nextchar(pRExC_state);
2740 ret = reg(pRExC_state, 1, &flags);
2742 if (flags & TRYAGAIN) {
2743 if (RExC_parse == RExC_end) {
2744 /* Make parent create an empty node if needed. */
2752 *flagp |= flags&(HASWIDTH|SPSTART|SIMPLE);
2756 if (flags & TRYAGAIN) {
2760 vFAIL("Internal urp");
2761 /* Supposed to be caught earlier. */
2764 if (!regcurly(RExC_parse)) {
2773 vFAIL("Quantifier follows nothing");
2776 switch (*++RExC_parse) {
2778 RExC_seen_zerolen++;
2779 ret = reg_node(pRExC_state, SBOL);
2781 nextchar(pRExC_state);
2782 Set_Node_Length(ret, 2); /* MJD */
2785 ret = reg_node(pRExC_state, GPOS);
2786 RExC_seen |= REG_SEEN_GPOS;
2788 nextchar(pRExC_state);
2789 Set_Node_Length(ret, 2); /* MJD */
2792 ret = reg_node(pRExC_state, SEOL);
2794 RExC_seen_zerolen++; /* Do not optimize RE away */
2795 nextchar(pRExC_state);
2798 ret = reg_node(pRExC_state, EOS);
2800 RExC_seen_zerolen++; /* Do not optimize RE away */
2801 nextchar(pRExC_state);
2802 Set_Node_Length(ret, 2); /* MJD */
2805 ret = reg_node(pRExC_state, CANY);
2806 RExC_seen |= REG_SEEN_CANY;
2807 *flagp |= HASWIDTH|SIMPLE;
2808 nextchar(pRExC_state);
2809 Set_Node_Length(ret, 2); /* MJD */
2812 ret = reg_node(pRExC_state, CLUMP);
2814 nextchar(pRExC_state);
2815 Set_Node_Length(ret, 2); /* MJD */
2818 ret = reg_node(pRExC_state, LOC ? ALNUML : ALNUM);
2819 *flagp |= HASWIDTH|SIMPLE;
2820 nextchar(pRExC_state);
2821 Set_Node_Length(ret, 2); /* MJD */
2824 ret = reg_node(pRExC_state, LOC ? NALNUML : NALNUM);
2825 *flagp |= HASWIDTH|SIMPLE;
2826 nextchar(pRExC_state);
2827 Set_Node_Length(ret, 2); /* MJD */
2830 RExC_seen_zerolen++;
2831 RExC_seen |= REG_SEEN_LOOKBEHIND;
2832 ret = reg_node(pRExC_state, LOC ? BOUNDL : BOUND);
2834 nextchar(pRExC_state);
2835 Set_Node_Length(ret, 2); /* MJD */
2838 RExC_seen_zerolen++;
2839 RExC_seen |= REG_SEEN_LOOKBEHIND;
2840 ret = reg_node(pRExC_state, LOC ? NBOUNDL : NBOUND);
2842 nextchar(pRExC_state);
2843 Set_Node_Length(ret, 2); /* MJD */
2846 ret = reg_node(pRExC_state, LOC ? SPACEL : SPACE);
2847 *flagp |= HASWIDTH|SIMPLE;
2848 nextchar(pRExC_state);
2849 Set_Node_Length(ret, 2); /* MJD */
2852 ret = reg_node(pRExC_state, LOC ? NSPACEL : NSPACE);
2853 *flagp |= HASWIDTH|SIMPLE;
2854 nextchar(pRExC_state);
2855 Set_Node_Length(ret, 2); /* MJD */
2858 ret = reg_node(pRExC_state, DIGIT);
2859 *flagp |= HASWIDTH|SIMPLE;
2860 nextchar(pRExC_state);
2861 Set_Node_Length(ret, 2); /* MJD */
2864 ret = reg_node(pRExC_state, NDIGIT);
2865 *flagp |= HASWIDTH|SIMPLE;
2866 nextchar(pRExC_state);
2867 Set_Node_Length(ret, 2); /* MJD */
2872 char* oldregxend = RExC_end;
2873 char* parse_start = RExC_parse;
2875 if (RExC_parse[1] == '{') {
2876 /* a lovely hack--pretend we saw [\pX] instead */
2877 RExC_end = strchr(RExC_parse, '}');
2879 U8 c = (U8)*RExC_parse;
2881 RExC_end = oldregxend;
2882 vFAIL2("Missing right brace on \\%c{}", c);
2887 RExC_end = RExC_parse + 2;
2890 ret = regclass(pRExC_state);
2892 RExC_end = oldregxend;
2894 Set_Node_Cur_Length(ret); /* MJD */
2895 nextchar(pRExC_state);
2896 *flagp |= HASWIDTH|SIMPLE;
2909 case '1': case '2': case '3': case '4':
2910 case '5': case '6': case '7': case '8': case '9':
2912 I32 num = atoi(RExC_parse);
2914 if (num > 9 && num >= RExC_npar)
2917 char * parse_start = RExC_parse - 1; /* MJD */
2918 while (isDIGIT(*RExC_parse))
2921 if (!SIZE_ONLY && num > RExC_rx->nparens)
2922 vFAIL("Reference to nonexistent group");
2924 ret = reganode(pRExC_state, FOLD
2925 ? (LOC ? REFFL : REFF)
2929 /* override incorrect value set in reganode MJD */
2930 Set_Node_Offset(ret, parse_start+1);
2931 Set_Node_Cur_Length(ret); /* MJD */
2933 nextchar(pRExC_state);
2938 if (RExC_parse >= RExC_end)
2939 FAIL("Trailing \\");
2942 /* Do not generate `unrecognized' warnings here, we fall
2943 back into the quick-grab loop below */
2949 if (RExC_flags16 & PMf_EXTENDED) {
2950 while (RExC_parse < RExC_end && *RExC_parse != '\n') RExC_parse++;
2951 if (RExC_parse < RExC_end)
2957 register STRLEN len;
2963 U8 tmpbuf[UTF8_MAXLEN*2+1];
2965 parse_start = RExC_parse - 1;
2970 ret = reg_node(pRExC_state, FOLD
2971 ? (LOC ? EXACTFL : EXACTF)
2974 for (len = 0, p = RExC_parse - 1;
2975 len < 127 && p < RExC_end;
2980 if (RExC_flags16 & PMf_EXTENDED)
2981 p = regwhite(p, RExC_end);
3026 ender = ASCII_TO_NATIVE('\033');
3030 ender = ASCII_TO_NATIVE('\007');
3035 char* e = strchr(p, '}');
3039 vFAIL("Missing right brace on \\x{}");
3042 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3043 | PERL_SCAN_DISALLOW_PREFIX;
3045 ender = grok_hex(p + 1, &numlen, &flags, NULL);
3048 /* numlen is generous */
3049 if (numlen + len >= 127) {
3057 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
3059 ender = grok_hex(p, &numlen, &flags, NULL);
3065 ender = UCHARAT(p++);
3066 ender = toCTRL(ender);
3068 case '0': case '1': case '2': case '3':case '4':
3069 case '5': case '6': case '7': case '8':case '9':
3071 (isDIGIT(p[1]) && atoi(p) >= RExC_npar) ) {
3074 ender = grok_oct(p, &numlen, &flags, NULL);
3084 FAIL("Trailing \\");
3087 if (!SIZE_ONLY && ckWARN(WARN_REGEXP) && isALPHA(*p))
3088 vWARN2(p + 1, "Unrecognized escape \\%c passed through", UCHARAT(p));
3089 goto normal_default;
3094 if (UTF8_IS_START(*p) && UTF) {
3095 ender = utf8n_to_uvchr((U8*)p, RExC_end - p,
3103 if (RExC_flags16 & PMf_EXTENDED)
3104 p = regwhite(p, RExC_end);
3106 toLOWER_uni(ender, tmpbuf, &ulen);
3107 ender = utf8_to_uvchr(tmpbuf, 0);
3109 if (ISMULT2(p)) { /* Back off on ?+*. */
3112 else if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(ender)) && UTF) {
3113 reguni(pRExC_state, ender, s, &numlen);
3123 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(ender)) && UTF) {
3124 reguni(pRExC_state, ender, s, &numlen);
3133 Set_Node_Cur_Length(ret); /* MJD */
3134 nextchar(pRExC_state);
3136 /* len is STRLEN which is unsigned, need to copy to signed */
3139 vFAIL("Internal disaster");
3148 RExC_size += STR_SZ(len);
3150 RExC_emit += STR_SZ(len);
3155 if (PL_encoding && PL_regkind[(U8)OP(ret)] == EXACT && !RExC_utf8) {
3156 STRLEN oldlen = STR_LEN(ret);
3157 SV *sv = sv_2mortal(newSVpvn(STRING(ret), oldlen));
3158 char *s = Perl_sv_recode_to_utf8(aTHX_ sv, PL_encoding);
3159 STRLEN newlen = SvCUR(sv);
3161 DEBUG_r(PerlIO_printf(Perl_debug_log, "recode %*s to %*s\n",
3162 oldlen, STRING(ret), newlen, s));
3163 Copy(s, STRING(ret), newlen, char);
3164 STR_LEN(ret) += newlen - oldlen;
3165 RExC_emit += STR_SZ(newlen) - STR_SZ(oldlen);
3167 RExC_size += STR_SZ(newlen) - STR_SZ(oldlen);
3175 S_regwhite(pTHX_ char *p, char *e)
3180 else if (*p == '#') {
3183 } while (p < e && *p != '\n');
3191 /* Parse POSIX character classes: [[:foo:]], [[=foo=]], [[.foo.]].
3192 Character classes ([:foo:]) can also be negated ([:^foo:]).
3193 Returns a named class id (ANYOF_XXX) if successful, -1 otherwise.
3194 Equivalence classes ([=foo=]) and composites ([.foo.]) are parsed,
3195 but trigger failures because they are currently unimplemented. */
3197 #define POSIXCC_DONE(c) ((c) == ':')
3198 #define POSIXCC_NOTYET(c) ((c) == '=' || (c) == '.')
3199 #define POSIXCC(c) (POSIXCC_DONE(c) || POSIXCC_NOTYET(c))
3202 S_regpposixcc(pTHX_ RExC_state_t *pRExC_state, I32 value)
3205 I32 namedclass = OOB_NAMEDCLASS;
3207 if (value == '[' && RExC_parse + 1 < RExC_end &&
3208 /* I smell either [: or [= or [. -- POSIX has been here, right? */
3209 POSIXCC(UCHARAT(RExC_parse))) {
3210 char c = UCHARAT(RExC_parse);
3211 char* s = RExC_parse++;
3213 while (RExC_parse < RExC_end && UCHARAT(RExC_parse) != c)
3215 if (RExC_parse == RExC_end)
3216 /* Grandfather lone [:, [=, [. */
3219 char* t = RExC_parse++; /* skip over the c */
3221 if (UCHARAT(RExC_parse) == ']') {
3222 RExC_parse++; /* skip over the ending ] */
3225 I32 complement = *posixcc == '^' ? *posixcc++ : 0;
3226 I32 skip = 5; /* the most common skip */
3230 if (strnEQ(posixcc, "alnum", 5))
3232 complement ? ANYOF_NALNUMC : ANYOF_ALNUMC;
3233 else if (strnEQ(posixcc, "alpha", 5))
3235 complement ? ANYOF_NALPHA : ANYOF_ALPHA;
3236 else if (strnEQ(posixcc, "ascii", 5))
3238 complement ? ANYOF_NASCII : ANYOF_ASCII;
3241 if (strnEQ(posixcc, "blank", 5))
3243 complement ? ANYOF_NBLANK : ANYOF_BLANK;
3246 if (strnEQ(posixcc, "cntrl", 5))
3248 complement ? ANYOF_NCNTRL : ANYOF_CNTRL;
3251 if (strnEQ(posixcc, "digit", 5))
3253 complement ? ANYOF_NDIGIT : ANYOF_DIGIT;
3256 if (strnEQ(posixcc, "graph", 5))
3258 complement ? ANYOF_NGRAPH : ANYOF_GRAPH;
3261 if (strnEQ(posixcc, "lower", 5))
3263 complement ? ANYOF_NLOWER : ANYOF_LOWER;
3266 if (strnEQ(posixcc, "print", 5))
3268 complement ? ANYOF_NPRINT : ANYOF_PRINT;
3269 else if (strnEQ(posixcc, "punct", 5))
3271 complement ? ANYOF_NPUNCT : ANYOF_PUNCT;
3274 if (strnEQ(posixcc, "space", 5))
3276 complement ? ANYOF_NPSXSPC : ANYOF_PSXSPC;
3279 if (strnEQ(posixcc, "upper", 5))
3281 complement ? ANYOF_NUPPER : ANYOF_UPPER;
3283 case 'w': /* this is not POSIX, this is the Perl \w */
3284 if (strnEQ(posixcc, "word", 4)) {
3286 complement ? ANYOF_NALNUM : ANYOF_ALNUM;
3291 if (strnEQ(posixcc, "xdigit", 6)) {
3293 complement ? ANYOF_NXDIGIT : ANYOF_XDIGIT;
3298 if (namedclass == OOB_NAMEDCLASS ||
3299 posixcc[skip] != ':' ||
3300 posixcc[skip+1] != ']')
3302 Simple_vFAIL3("POSIX class [:%.*s:] unknown",
3305 } else if (!SIZE_ONLY) {
3306 /* [[=foo=]] and [[.foo.]] are still future. */
3308 /* adjust RExC_parse so the warning shows after
3310 while (UCHARAT(RExC_parse) && UCHARAT(RExC_parse) != ']')
3312 Simple_vFAIL3("POSIX syntax [%c %c] is reserved for future extensions", c, c);
3315 /* Maternal grandfather:
3316 * "[:" ending in ":" but not in ":]" */
3326 S_checkposixcc(pTHX_ RExC_state_t *pRExC_state)
3328 if (!SIZE_ONLY && ckWARN(WARN_REGEXP) &&
3329 POSIXCC(UCHARAT(RExC_parse))) {
3330 char *s = RExC_parse;
3333 while(*s && isALNUM(*s))
3335 if (*s && c == *s && s[1] == ']') {
3336 vWARN3(s+2, "POSIX syntax [%c %c] belongs inside character classes", c, c);
3338 /* [[=foo=]] and [[.foo.]] are still future. */
3339 if (POSIXCC_NOTYET(c)) {
3340 /* adjust RExC_parse so the error shows after
3342 while (UCHARAT(RExC_parse) && UCHARAT(RExC_parse++) != ']')
3344 Simple_vFAIL3("POSIX syntax [%c %c] is reserved for future extensions", c, c);
3351 S_regclass(pTHX_ RExC_state_t *pRExC_state)
3354 register UV nextvalue;
3355 register IV prevvalue = OOB_UNICODE;
3356 register IV range = 0;
3357 register regnode *ret;
3360 char *rangebegin = 0;
3361 bool need_class = 0;
3362 SV *listsv = Nullsv;
3365 bool optimize_invert = TRUE;
3367 ret = reganode(pRExC_state, ANYOF, 0);
3370 ANYOF_FLAGS(ret) = 0;
3372 if (UCHARAT(RExC_parse) == '^') { /* Complement of range. */
3376 ANYOF_FLAGS(ret) |= ANYOF_INVERT;
3380 RExC_size += ANYOF_SKIP;
3382 RExC_emit += ANYOF_SKIP;
3384 ANYOF_FLAGS(ret) |= ANYOF_FOLD;
3386 ANYOF_FLAGS(ret) |= ANYOF_LOCALE;
3387 ANYOF_BITMAP_ZERO(ret);
3388 listsv = newSVpvn("# comment\n", 10);
3391 nextvalue = RExC_parse < RExC_end ? UCHARAT(RExC_parse) : 0;
3393 if (!SIZE_ONLY && ckWARN(WARN_REGEXP) && POSIXCC(nextvalue))
3394 checkposixcc(pRExC_state);
3396 if (UCHARAT(RExC_parse) == ']' || UCHARAT(RExC_parse) == '-')
3397 goto charclassloop; /* allow 1st char to be ] or - */
3399 while (RExC_parse < RExC_end && UCHARAT(RExC_parse) != ']') {
3403 namedclass = OOB_NAMEDCLASS; /* initialize as illegal */
3406 rangebegin = RExC_parse;
3408 value = utf8n_to_uvchr((U8*)RExC_parse,
3409 RExC_end - RExC_parse,
3411 RExC_parse += numlen;
3414 value = UCHARAT(RExC_parse++);
3415 nextvalue = RExC_parse < RExC_end ? UCHARAT(RExC_parse) : 0;
3416 if (value == '[' && POSIXCC(nextvalue))
3417 namedclass = regpposixcc(pRExC_state, value);
3418 else if (value == '\\') {
3420 value = utf8n_to_uvchr((U8*)RExC_parse,
3421 RExC_end - RExC_parse,
3423 RExC_parse += numlen;
3426 value = UCHARAT(RExC_parse++);
3427 /* Some compilers cannot handle switching on 64-bit integer
3428 * values, therefore value cannot be an UV. Yes, this will
3429 * be a problem later if we want switch on Unicode.
3430 * A similar issue a little bit later when switching on
3431 * namedclass. --jhi */
3432 switch ((I32)value) {
3433 case 'w': namedclass = ANYOF_ALNUM; break;
3434 case 'W': namedclass = ANYOF_NALNUM; break;
3435 case 's': namedclass = ANYOF_SPACE; break;
3436 case 'S': namedclass = ANYOF_NSPACE; break;
3437 case 'd': namedclass = ANYOF_DIGIT; break;
3438 case 'D': namedclass = ANYOF_NDIGIT; break;
3441 if (*RExC_parse == '{') {
3443 e = strchr(RExC_parse++, '}');
3445 vFAIL2("Missing right brace on \\%c{}", c);
3446 while (isSPACE(UCHARAT(RExC_parse)))
3448 if (e == RExC_parse)
3449 vFAIL2("Empty \\%c{}", c);
3451 while (isSPACE(UCHARAT(RExC_parse + n - 1)))
3459 if (UCHARAT(RExC_parse) == '^') {
3462 value = value == 'p' ? 'P' : 'p'; /* toggle */
3463 while (isSPACE(UCHARAT(RExC_parse))) {
3469 Perl_sv_catpvf(aTHX_ listsv,
3470 "+utf8::%.*s\n", (int)n, RExC_parse);
3472 Perl_sv_catpvf(aTHX_ listsv,
3473 "!utf8::%.*s\n", (int)n, RExC_parse);
3476 ANYOF_FLAGS(ret) |= ANYOF_UNICODE;
3478 case 'n': value = '\n'; break;
3479 case 'r': value = '\r'; break;
3480 case 't': value = '\t'; break;
3481 case 'f': value = '\f'; break;
3482 case 'b': value = '\b'; break;
3483 case 'e': value = ASCII_TO_NATIVE('\033');break;
3484 case 'a': value = ASCII_TO_NATIVE('\007');break;
3486 if (*RExC_parse == '{') {
3487 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3488 | PERL_SCAN_DISALLOW_PREFIX;
3489 e = strchr(RExC_parse++, '}');
3491 vFAIL("Missing right brace on \\x{}");
3493 numlen = e - RExC_parse;
3494 value = grok_hex(RExC_parse, &numlen, &flags, NULL);
3498 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
3500 value = grok_hex(RExC_parse, &numlen, &flags, NULL);
3501 RExC_parse += numlen;
3505 value = UCHARAT(RExC_parse++);
3506 value = toCTRL(value);
3508 case '0': case '1': case '2': case '3': case '4':
3509 case '5': case '6': case '7': case '8': case '9':
3513 value = grok_oct(--RExC_parse, &numlen, &flags, NULL);
3514 RExC_parse += numlen;
3518 if (!SIZE_ONLY && ckWARN(WARN_REGEXP) && isALPHA(value))
3520 "Unrecognized escape \\%c in character class passed through",
3524 } /* end of \blah */
3526 if (namedclass > OOB_NAMEDCLASS) { /* this is a named class \blah */
3528 if (!SIZE_ONLY && !need_class)
3529 ANYOF_CLASS_ZERO(ret);
3533 /* a bad range like a-\d, a-[:digit:] ? */
3536 if (ckWARN(WARN_REGEXP))
3538 "False [] range \"%*.*s\"",
3539 RExC_parse - rangebegin,
3540 RExC_parse - rangebegin,
3542 if (prevvalue < 256) {
3543 ANYOF_BITMAP_SET(ret, prevvalue);
3544 ANYOF_BITMAP_SET(ret, '-');
3547 ANYOF_FLAGS(ret) |= ANYOF_UNICODE;
3548 Perl_sv_catpvf(aTHX_ listsv,
3549 "%04"UVxf"\n%04"UVxf"\n", (UV)prevvalue, (UV) '-');
3553 range = 0; /* this was not a true range */
3557 if (namedclass > OOB_NAMEDCLASS)
3558 optimize_invert = FALSE;
3559 /* Possible truncation here but in some 64-bit environments
3560 * the compiler gets heartburn about switch on 64-bit values.
3561 * A similar issue a little earlier when switching on value.
3563 switch ((I32)namedclass) {
3566 ANYOF_CLASS_SET(ret, ANYOF_ALNUM);
3568 for (value = 0; value < 256; value++)
3570 ANYOF_BITMAP_SET(ret, value);
3572 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsWord\n");
3576 ANYOF_CLASS_SET(ret, ANYOF_NALNUM);
3578 for (value = 0; value < 256; value++)
3579 if (!isALNUM(value))
3580 ANYOF_BITMAP_SET(ret, value);
3582 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsWord\n");
3586 ANYOF_CLASS_SET(ret, ANYOF_ALNUMC);
3588 for (value = 0; value < 256; value++)
3589 if (isALNUMC(value))
3590 ANYOF_BITMAP_SET(ret, value);
3592 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsAlnum\n");
3596 ANYOF_CLASS_SET(ret, ANYOF_NALNUMC);
3598 for (value = 0; value < 256; value++)
3599 if (!isALNUMC(value))
3600 ANYOF_BITMAP_SET(ret, value);
3602 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsAlnum\n");
3606 ANYOF_CLASS_SET(ret, ANYOF_ALPHA);
3608 for (value = 0; value < 256; value++)
3610 ANYOF_BITMAP_SET(ret, value);
3612 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsAlpha\n");
3616 ANYOF_CLASS_SET(ret, ANYOF_NALPHA);
3618 for (value = 0; value < 256; value++)
3619 if (!isALPHA(value))
3620 ANYOF_BITMAP_SET(ret, value);
3622 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsAlpha\n");
3626 ANYOF_CLASS_SET(ret, ANYOF_ASCII);
3629 for (value = 0; value < 128; value++)
3630 ANYOF_BITMAP_SET(ret, value);
3632 for (value = 0; value < 256; value++) {
3634 ANYOF_BITMAP_SET(ret, value);
3638 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsASCII\n");
3642 ANYOF_CLASS_SET(ret, ANYOF_NASCII);
3645 for (value = 128; value < 256; value++)
3646 ANYOF_BITMAP_SET(ret, value);
3648 for (value = 0; value < 256; value++) {
3649 if (!isASCII(value))
3650 ANYOF_BITMAP_SET(ret, value);
3654 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsASCII\n");
3658 ANYOF_CLASS_SET(ret, ANYOF_BLANK);
3660 for (value = 0; value < 256; value++)
3662 ANYOF_BITMAP_SET(ret, value);
3664 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsBlank\n");
3668 ANYOF_CLASS_SET(ret, ANYOF_NBLANK);
3670 for (value = 0; value < 256; value++)
3671 if (!isBLANK(value))
3672 ANYOF_BITMAP_SET(ret, value);
3674 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsBlank\n");
3678 ANYOF_CLASS_SET(ret, ANYOF_CNTRL);
3680 for (value = 0; value < 256; value++)
3682 ANYOF_BITMAP_SET(ret, value);
3684 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsCntrl\n");
3688 ANYOF_CLASS_SET(ret, ANYOF_NCNTRL);
3690 for (value = 0; value < 256; value++)
3691 if (!isCNTRL(value))
3692 ANYOF_BITMAP_SET(ret, value);
3694 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsCntrl\n");
3698 ANYOF_CLASS_SET(ret, ANYOF_DIGIT);
3700 /* consecutive digits assumed */
3701 for (value = '0'; value <= '9'; value++)
3702 ANYOF_BITMAP_SET(ret, value);
3704 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsDigit\n");
3708 ANYOF_CLASS_SET(ret, ANYOF_NDIGIT);
3710 /* consecutive digits assumed */
3711 for (value = 0; value < '0'; value++)
3712 ANYOF_BITMAP_SET(ret, value);
3713 for (value = '9' + 1; value < 256; value++)
3714 ANYOF_BITMAP_SET(ret, value);
3716 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsDigit\n");
3720 ANYOF_CLASS_SET(ret, ANYOF_GRAPH);
3722 for (value = 0; value < 256; value++)
3724 ANYOF_BITMAP_SET(ret, value);
3726 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsGraph\n");
3730 ANYOF_CLASS_SET(ret, ANYOF_NGRAPH);
3732 for (value = 0; value < 256; value++)
3733 if (!isGRAPH(value))
3734 ANYOF_BITMAP_SET(ret, value);
3736 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsGraph\n");
3740 ANYOF_CLASS_SET(ret, ANYOF_LOWER);
3742 for (value = 0; value < 256; value++)
3744 ANYOF_BITMAP_SET(ret, value);
3746 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsLower\n");
3750 ANYOF_CLASS_SET(ret, ANYOF_NLOWER);
3752 for (value = 0; value < 256; value++)
3753 if (!isLOWER(value))
3754 ANYOF_BITMAP_SET(ret, value);
3756 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsLower\n");
3760 ANYOF_CLASS_SET(ret, ANYOF_PRINT);
3762 for (value = 0; value < 256; value++)
3764 ANYOF_BITMAP_SET(ret, value);
3766 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsPrint\n");
3770 ANYOF_CLASS_SET(ret, ANYOF_NPRINT);
3772 for (value = 0; value < 256; value++)
3773 if (!isPRINT(value))
3774 ANYOF_BITMAP_SET(ret, value);
3776 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsPrint\n");
3780 ANYOF_CLASS_SET(ret, ANYOF_PSXSPC);
3782 for (value = 0; value < 256; value++)
3783 if (isPSXSPC(value))
3784 ANYOF_BITMAP_SET(ret, value);
3786 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsSpace\n");
3790 ANYOF_CLASS_SET(ret, ANYOF_NPSXSPC);
3792 for (value = 0; value < 256; value++)
3793 if (!isPSXSPC(value))
3794 ANYOF_BITMAP_SET(ret, value);
3796 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsSpace\n");
3800 ANYOF_CLASS_SET(ret, ANYOF_PUNCT);
3802 for (value = 0; value < 256; value++)
3804 ANYOF_BITMAP_SET(ret, value);
3806 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsPunct\n");
3810 ANYOF_CLASS_SET(ret, ANYOF_NPUNCT);
3812 for (value = 0; value < 256; value++)
3813 if (!isPUNCT(value))
3814 ANYOF_BITMAP_SET(ret, value);
3816 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsPunct\n");
3820 ANYOF_CLASS_SET(ret, ANYOF_SPACE);
3822 for (value = 0; value < 256; value++)
3824 ANYOF_BITMAP_SET(ret, value);
3826 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsSpacePerl\n");
3830 ANYOF_CLASS_SET(ret, ANYOF_NSPACE);
3832 for (value = 0; value < 256; value++)
3833 if (!isSPACE(value))
3834 ANYOF_BITMAP_SET(ret, value);
3836 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsSpacePerl\n");
3840 ANYOF_CLASS_SET(ret, ANYOF_UPPER);
3842 for (value = 0; value < 256; value++)
3844 ANYOF_BITMAP_SET(ret, value);
3846 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsUpper\n");
3850 ANYOF_CLASS_SET(ret, ANYOF_NUPPER);
3852 for (value = 0; value < 256; value++)
3853 if (!isUPPER(value))
3854 ANYOF_BITMAP_SET(ret, value);
3856 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsUpper\n");
3860 ANYOF_CLASS_SET(ret, ANYOF_XDIGIT);
3862 for (value = 0; value < 256; value++)
3863 if (isXDIGIT(value))
3864 ANYOF_BITMAP_SET(ret, value);
3866 Perl_sv_catpvf(aTHX_ listsv, "+utf8::IsXDigit\n");
3870 ANYOF_CLASS_SET(ret, ANYOF_NXDIGIT);
3872 for (value = 0; value < 256; value++)
3873 if (!isXDIGIT(value))
3874 ANYOF_BITMAP_SET(ret, value);
3876 Perl_sv_catpvf(aTHX_ listsv, "!utf8::IsXDigit\n");
3879 vFAIL("Invalid [::] class");
3883 ANYOF_FLAGS(ret) |= ANYOF_CLASS;
3886 } /* end of namedclass \blah */
3889 if (prevvalue > value) /* b-a */ {
3890 Simple_vFAIL4("Invalid [] range \"%*.*s\"",
3891 RExC_parse - rangebegin,
3892 RExC_parse - rangebegin,
3894 range = 0; /* not a valid range */
3898 prevvalue = value; /* save the beginning of the range */
3899 if (*RExC_parse == '-' && RExC_parse+1 < RExC_end &&
3900 RExC_parse[1] != ']') {
3903 /* a bad range like \w-, [:word:]- ? */
3904 if (namedclass > OOB_NAMEDCLASS) {
3905 if (ckWARN(WARN_REGEXP))
3907 "False [] range \"%*.*s\"",
3908 RExC_parse - rangebegin,
3909 RExC_parse - rangebegin,
3912 ANYOF_BITMAP_SET(ret, '-');
3914 range = 1; /* yeah, it's a range! */
3915 continue; /* but do it the next time */
3919 /* now is the next time */
3923 if (prevvalue < 256) {
3924 IV ceilvalue = value < 256 ? value : 255;
3927 if ((isLOWER(prevvalue) && isLOWER(ceilvalue)) ||
3928 (isUPPER(prevvalue) && isUPPER(ceilvalue)))
3930 if (isLOWER(prevvalue)) {
3931 for (i = prevvalue; i <= ceilvalue; i++)
3933 ANYOF_BITMAP_SET(ret, i);
3935 for (i = prevvalue; i <= ceilvalue; i++)
3937 ANYOF_BITMAP_SET(ret, i);
3942 for (i = prevvalue; i <= ceilvalue; i++)
3943 ANYOF_BITMAP_SET(ret, i);
3946 ANYOF_FLAGS(ret) |= ANYOF_UNICODE;
3947 if (prevvalue < value)
3948 Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\t%04"UVxf"\n",
3949 (UV)prevvalue, (UV)value);
3950 else if (prevvalue == value)
3951 Perl_sv_catpvf(aTHX_ listsv, "%04"UVxf"\n",
3956 range = 0; /* this range (if it was one) is done now */
3960 ANYOF_FLAGS(ret) |= ANYOF_LARGE;
3962 RExC_size += ANYOF_CLASS_ADD_SKIP;
3964 RExC_emit += ANYOF_CLASS_ADD_SKIP;
3967 /* optimize case-insensitive simple patterns (e.g. /[a-z]/i) */
3969 /* If the only flag is folding (plus possibly inversion). */
3970 ((ANYOF_FLAGS(ret) & (ANYOF_FLAGS_ALL ^ ANYOF_INVERT)) == ANYOF_FOLD)
3972 for (value = 0; value < 256; ++value) {
3973 if (ANYOF_BITMAP_TEST(ret, value)) {
3974 IV fold = PL_fold[value];
3977 ANYOF_BITMAP_SET(ret, fold);
3980 ANYOF_FLAGS(ret) &= ~ANYOF_FOLD;
3983 /* optimize inverted simple patterns (e.g. [^a-z]) */
3984 if (!SIZE_ONLY && optimize_invert &&
3985 /* If the only flag is inversion. */
3986 (ANYOF_FLAGS(ret) & ANYOF_FLAGS_ALL) == ANYOF_INVERT) {
3987 for (value = 0; value < ANYOF_BITMAP_SIZE; ++value)
3988 ANYOF_BITMAP(ret)[value] ^= ANYOF_FLAGS_ALL;
3989 ANYOF_FLAGS(ret) = ANYOF_UNICODE_ALL;
3996 av_store(av, 0, listsv);
3997 av_store(av, 1, NULL);
3998 rv = newRV_noinc((SV*)av);
3999 n = add_data(pRExC_state, 1, "s");
4000 RExC_rx->data->data[n] = (void*)rv;
4008 S_nextchar(pTHX_ RExC_state_t *pRExC_state)
4010 char* retval = RExC_parse++;
4013 if (*RExC_parse == '(' && RExC_parse[1] == '?' &&
4014 RExC_parse[2] == '#') {
4015 while (*RExC_parse && *RExC_parse != ')')
4020 if (RExC_flags16 & PMf_EXTENDED) {
4021 if (isSPACE(*RExC_parse)) {
4025 else if (*RExC_parse == '#') {
4026 while (*RExC_parse && *RExC_parse != '\n')
4037 - reg_node - emit a node
4039 STATIC regnode * /* Location. */
4040 S_reg_node(pTHX_ RExC_state_t *pRExC_state, U8 op)
4042 register regnode *ret;
4043 register regnode *ptr;
4047 SIZE_ALIGN(RExC_size);
4052 NODE_ALIGN_FILL(ret);
4054 FILL_ADVANCE_NODE(ptr, op);
4055 if (RExC_offsets) { /* MJD */
4056 MJD_OFFSET_DEBUG((stderr, "%s:%u: (op %s) %s %u <- %u (len %u) (max %u).\n",
4057 "reg_node", __LINE__,
4059 RExC_emit - RExC_emit_start > RExC_offsets[0]
4060 ? "Overwriting end of array!\n" : "OK",
4061 RExC_emit - RExC_emit_start,
4062 RExC_parse - RExC_start,
4064 Set_Node_Offset(RExC_emit, RExC_parse + (op == END));
4073 - reganode - emit a node with an argument
4075 STATIC regnode * /* Location. */
4076 S_reganode(pTHX_ RExC_state_t *pRExC_state, U8 op, U32 arg)
4078 register regnode *ret;
4079 register regnode *ptr;
4083 SIZE_ALIGN(RExC_size);
4088 NODE_ALIGN_FILL(ret);
4090 FILL_ADVANCE_NODE_ARG(ptr, op, arg);
4091 if (RExC_offsets) { /* MJD */
4092 MJD_OFFSET_DEBUG((stderr, "%s: %s %u <- %u (max %u).\n",
4094 RExC_emit - RExC_emit_start > RExC_offsets[0] ?
4095 "Overwriting end of array!\n" : "OK",
4096 RExC_emit - RExC_emit_start,
4097 RExC_parse - RExC_start,
4099 Set_Cur_Node_Offset;
4108 - reguni - emit (if appropriate) a Unicode character
4111 S_reguni(pTHX_ RExC_state_t *pRExC_state, UV uv, char* s, STRLEN* lenp)
4113 *lenp = SIZE_ONLY ? UNISKIP(uv) : (uvchr_to_utf8((U8*)s, uv) - (U8*)s);
4117 - reginsert - insert an operator in front of already-emitted operand
4119 * Means relocating the operand.
4122 S_reginsert(pTHX_ RExC_state_t *pRExC_state, U8 op, regnode *opnd)
4124 register regnode *src;
4125 register regnode *dst;
4126 register regnode *place;
4127 register int offset = regarglen[(U8)op];
4129 /* (PL_regkind[(U8)op] == CURLY ? EXTRA_STEP_2ARGS : 0); */
4132 RExC_size += NODE_STEP_REGNODE + offset;
4137 RExC_emit += NODE_STEP_REGNODE + offset;
4139 while (src > opnd) {
4140 StructCopy(--src, --dst, regnode);
4141 if (RExC_offsets) { /* MJD 20010112 */
4142 MJD_OFFSET_DEBUG((stderr, "%s: %s copy %u -> %u (max %u).\n",
4144 dst - RExC_emit_start > RExC_offsets[0]
4145 ? "Overwriting end of array!\n" : "OK",
4146 src - RExC_emit_start,
4147 dst - RExC_emit_start,
4149 Set_Node_Offset_To_R(dst-RExC_emit_start, Node_Offset(src));
4150 Set_Node_Length_To_R(dst-RExC_emit_start, Node_Length(src));
4155 place = opnd; /* Op node, where operand used to be. */
4156 if (RExC_offsets) { /* MJD */
4157 MJD_OFFSET_DEBUG((stderr, "%s: %s %u <- %u (max %u).\n",
4159 place - RExC_emit_start > RExC_offsets[0]
4160 ? "Overwriting end of array!\n" : "OK",
4161 place - RExC_emit_start,
4162 RExC_parse - RExC_start,
4164 Set_Node_Offset(place, RExC_parse);
4166 src = NEXTOPER(place);
4167 FILL_ADVANCE_NODE(place, op);
4168 Zero(src, offset, regnode);
4172 - regtail - set the next-pointer at the end of a node chain of p to val.
4175 S_regtail(pTHX_ RExC_state_t *pRExC_state, regnode *p, regnode *val)
4177 register regnode *scan;
4178 register regnode *temp;
4183 /* Find last node. */
4186 temp = regnext(scan);
4192 if (reg_off_by_arg[OP(scan)]) {
4193 ARG_SET(scan, val - scan);
4196 NEXT_OFF(scan) = val - scan;
4201 - regoptail - regtail on operand of first argument; nop if operandless
4204 S_regoptail(pTHX_ RExC_state_t *pRExC_state, regnode *p, regnode *val)
4206 /* "Operandless" and "op != BRANCH" are synonymous in practice. */
4207 if (p == NULL || SIZE_ONLY)
4209 if (PL_regkind[(U8)OP(p)] == BRANCH) {
4210 regtail(pRExC_state, NEXTOPER(p), val);
4212 else if ( PL_regkind[(U8)OP(p)] == BRANCHJ) {
4213 regtail(pRExC_state, NEXTOPER(NEXTOPER(p)), val);
4220 - regcurly - a little FSA that accepts {\d+,?\d*}
4223 S_regcurly(pTHX_ register char *s)
4244 S_dumpuntil(pTHX_ regnode *start, regnode *node, regnode *last, SV* sv, I32 l)
4246 register U8 op = EXACT; /* Arbitrary non-END op. */
4247 register regnode *next;
4249 while (op != END && (!last || node < last)) {
4250 /* While that wasn't END last time... */
4256 next = regnext(node);
4258 if (OP(node) == OPTIMIZED)
4261 PerlIO_printf(Perl_debug_log, "%4"IVdf":%*s%s", (IV)(node - start),
4262 (int)(2*l + 1), "", SvPVX(sv));
4263 if (next == NULL) /* Next ptr. */
4264 PerlIO_printf(Perl_debug_log, "(0)");
4266 PerlIO_printf(Perl_debug_log, "(%"IVdf")", (IV)(next - start));
4267 (void)PerlIO_putc(Perl_debug_log, '\n');
4269 if (PL_regkind[(U8)op] == BRANCHJ) {
4270 register regnode *nnode = (OP(next) == LONGJMP
4273 if (last && nnode > last)
4275 node = dumpuntil(start, NEXTOPER(NEXTOPER(node)), nnode, sv, l + 1);
4277 else if (PL_regkind[(U8)op] == BRANCH) {
4278 node = dumpuntil(start, NEXTOPER(node), next, sv, l + 1);
4280 else if ( op == CURLY) { /* `next' might be very big: optimizer */
4281 node = dumpuntil(start, NEXTOPER(node) + EXTRA_STEP_2ARGS,
4282 NEXTOPER(node) + EXTRA_STEP_2ARGS + 1, sv, l + 1);
4284 else if (PL_regkind[(U8)op] == CURLY && op != CURLYX) {
4285 node = dumpuntil(start, NEXTOPER(node) + EXTRA_STEP_2ARGS,
4288 else if ( op == PLUS || op == STAR) {
4289 node = dumpuntil(start, NEXTOPER(node), NEXTOPER(node) + 1, sv, l + 1);
4291 else if (op == ANYOF) {
4292 /* arglen 1 + class block */
4293 node += 1 + ((ANYOF_FLAGS(node) & ANYOF_LARGE)
4294 ? ANYOF_CLASS_SKIP : ANYOF_SKIP);
4295 node = NEXTOPER(node);
4297 else if (PL_regkind[(U8)op] == EXACT) {
4298 /* Literal string, where present. */
4299 node += NODE_SZ_STR(node) - 1;
4300 node = NEXTOPER(node);
4303 node = NEXTOPER(node);
4304 node += regarglen[(U8)op];
4306 if (op == CURLYX || op == OPEN)
4308 else if (op == WHILEM)
4314 #endif /* DEBUGGING */
4317 - regdump - dump a regexp onto Perl_debug_log in vaguely comprehensible form
4320 Perl_regdump(pTHX_ regexp *r)
4323 SV *sv = sv_newmortal();
4325 (void)dumpuntil(r->program, r->program + 1, NULL, sv, 0);
4327 /* Header fields of interest. */
4328 if (r->anchored_substr)
4329 PerlIO_printf(Perl_debug_log,
4330 "anchored `%s%.*s%s'%s at %"IVdf" ",
4332 (int)(SvCUR(r->anchored_substr) - (SvTAIL(r->anchored_substr)!=0)),
4333 SvPVX(r->anchored_substr),
4335 SvTAIL(r->anchored_substr) ? "$" : "",
4336 (IV)r->anchored_offset);
4337 if (r->float_substr)
4338 PerlIO_printf(Perl_debug_log,
4339 "floating `%s%.*s%s'%s at %"IVdf"..%"UVuf" ",
4341 (int)(SvCUR(r->float_substr) - (SvTAIL(r->float_substr)!=0)),
4342 SvPVX(r->float_substr),
4344 SvTAIL(r->float_substr) ? "$" : "",
4345 (IV)r->float_min_offset, (UV)r->float_max_offset);
4346 if (r->check_substr)
4347 PerlIO_printf(Perl_debug_log,
4348 r->check_substr == r->float_substr
4349 ? "(checking floating" : "(checking anchored");
4350 if (r->reganch & ROPT_NOSCAN)
4351 PerlIO_printf(Perl_debug_log, " noscan");
4352 if (r->reganch & ROPT_CHECK_ALL)
4353 PerlIO_printf(Perl_debug_log, " isall");
4354 if (r->check_substr)
4355 PerlIO_printf(Perl_debug_log, ") ");
4357 if (r->regstclass) {
4358 regprop(sv, r->regstclass);
4359 PerlIO_printf(Perl_debug_log, "stclass `%s' ", SvPVX(sv));
4361 if (r->reganch & ROPT_ANCH) {
4362 PerlIO_printf(Perl_debug_log, "anchored");
4363 if (r->reganch & ROPT_ANCH_BOL)
4364 PerlIO_printf(Perl_debug_log, "(BOL)");
4365 if (r->reganch & ROPT_ANCH_MBOL)
4366 PerlIO_printf(Perl_debug_log, "(MBOL)");
4367 if (r->reganch & ROPT_ANCH_SBOL)
4368 PerlIO_printf(Perl_debug_log, "(SBOL)");
4369 if (r->reganch & ROPT_ANCH_GPOS)
4370 PerlIO_printf(Perl_debug_log, "(GPOS)");
4371 PerlIO_putc(Perl_debug_log, ' ');
4373 if (r->reganch & ROPT_GPOS_SEEN)
4374 PerlIO_printf(Perl_debug_log, "GPOS ");
4375 if (r->reganch & ROPT_SKIP)
4376 PerlIO_printf(Perl_debug_log, "plus ");
4377 if (r->reganch & ROPT_IMPLICIT)
4378 PerlIO_printf(Perl_debug_log, "implicit ");
4379 PerlIO_printf(Perl_debug_log, "minlen %ld ", (long) r->minlen);
4380 if (r->reganch & ROPT_EVAL_SEEN)
4381 PerlIO_printf(Perl_debug_log, "with eval ");
4382 PerlIO_printf(Perl_debug_log, "\n");
4385 U32 len = r->offsets[0];
4386 PerlIO_printf(Perl_debug_log, "Offsets: [%"UVuf"]\n\t", (UV)r->offsets[0]);
4387 for (i = 1; i <= len; i++)
4388 PerlIO_printf(Perl_debug_log, "%"UVuf"[%"UVuf"] ",
4389 (UV)r->offsets[i*2-1],
4390 (UV)r->offsets[i*2]);
4391 PerlIO_printf(Perl_debug_log, "\n");
4393 #endif /* DEBUGGING */
4399 S_put_byte(pTHX_ SV *sv, int c)
4401 if (isCNTRL(c) || c == 255 || !isPRINT(c))
4402 Perl_sv_catpvf(aTHX_ sv, "\\%o", c);
4403 else if (c == '-' || c == ']' || c == '\\' || c == '^')
4404 Perl_sv_catpvf(aTHX_ sv, "\\%c", c);
4406 Perl_sv_catpvf(aTHX_ sv, "%c", c);
4409 #endif /* DEBUGGING */
4412 - regprop - printable representation of opcode
4415 Perl_regprop(pTHX_ SV *sv, regnode *o)
4420 sv_setpvn(sv, "", 0);
4421 if (OP(o) >= reg_num) /* regnode.type is unsigned */
4422 /* It would be nice to FAIL() here, but this may be called from
4423 regexec.c, and it would be hard to supply pRExC_state. */
4424 Perl_croak(aTHX_ "Corrupted regexp opcode");
4425 sv_catpv(sv, (char*)reg_name[OP(o)]); /* Take off const! */
4427 k = PL_regkind[(U8)OP(o)];
4430 SV *dsv = sv_2mortal(newSVpvn("", 0));
4431 bool do_utf8 = DO_UTF8(sv);
4433 pv_uni_display(dsv, (U8*)STRING(o), STR_LEN(o), 60, 0) :
4438 Perl_sv_catpvf(aTHX_ sv, " <%s%.*s%s>",
4443 else if (k == CURLY) {
4444 if (OP(o) == CURLYM || OP(o) == CURLYN || OP(o) == CURLYX)
4445 Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* Parenth number */
4446 Perl_sv_catpvf(aTHX_ sv, " {%d,%d}", ARG1(o), ARG2(o));
4448 else if (k == WHILEM && o->flags) /* Ordinal/of */
4449 Perl_sv_catpvf(aTHX_ sv, "[%d/%d]", o->flags & 0xf, o->flags>>4);
4450 else if (k == REF || k == OPEN || k == CLOSE || k == GROUPP )
4451 Perl_sv_catpvf(aTHX_ sv, "%d", (int)ARG(o)); /* Parenth number */
4452 else if (k == LOGICAL)
4453 Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* 2: embedded, otherwise 1 */
4454 else if (k == ANYOF) {
4455 int i, rangestart = -1;
4456 U8 flags = ANYOF_FLAGS(o);
4457 const char * const anyofs[] = { /* Should be syncronized with
4458 * ANYOF_ #xdefines in regcomp.h */
4491 if (flags & ANYOF_LOCALE)
4492 sv_catpv(sv, "{loc}");
4493 if (flags & ANYOF_FOLD)
4494 sv_catpv(sv, "{i}");
4495 Perl_sv_catpvf(aTHX_ sv, "[%s", PL_colors[0]);
4496 if (flags & ANYOF_INVERT)
4498 for (i = 0; i <= 256; i++) {
4499 if (i < 256 && ANYOF_BITMAP_TEST(o,i)) {
4500 if (rangestart == -1)
4502 } else if (rangestart != -1) {
4503 if (i <= rangestart + 3)
4504 for (; rangestart < i; rangestart++)
4505 put_byte(sv, rangestart);
4507 put_byte(sv, rangestart);
4509 put_byte(sv, i - 1);
4515 if (o->flags & ANYOF_CLASS)
4516 for (i = 0; i < sizeof(anyofs)/sizeof(char*); i++)
4517 if (ANYOF_CLASS_TEST(o,i))
4518 sv_catpv(sv, anyofs[i]);
4520 if (flags & ANYOF_UNICODE)
4521 sv_catpv(sv, "{unicode}");
4522 else if (flags & ANYOF_UNICODE_ALL)
4523 sv_catpv(sv, "{unicode_all}");
4527 SV *sw = regclass_swash(o, FALSE, &lv);
4532 U8 s[UTF8_MAXLEN+1];
4534 for (i = 0; i <= 256; i++) { /* just the first 256 */
4535 U8 *e = uvchr_to_utf8(s, i);
4537 if (i < 256 && swash_fetch(sw, s, TRUE)) {
4538 if (rangestart == -1)
4540 } else if (rangestart != -1) {
4543 if (i <= rangestart + 3)
4544 for (; rangestart < i; rangestart++) {
4545 for(e = uvchr_to_utf8(s, rangestart), p = s; p < e; p++)
4549 for (e = uvchr_to_utf8(s, rangestart), p = s; p < e; p++)
4552 for (e = uvchr_to_utf8(s, i - 1), p = s; p < e; p++)
4559 sv_catpv(sv, "..."); /* et cetera */
4563 char *s = savepv(SvPVX(lv));
4566 while(*s && *s != '\n') s++;
4587 Perl_sv_catpvf(aTHX_ sv, "%s]", PL_colors[1]);
4589 else if (k == BRANCHJ && (OP(o) == UNLESSM || OP(o) == IFMATCH))
4590 Perl_sv_catpvf(aTHX_ sv, "[-%d]", o->flags);
4591 #endif /* DEBUGGING */
4595 Perl_re_intuit_string(pTHX_ regexp *prog)
4596 { /* Assume that RE_INTUIT is set */
4599 char *s = SvPV(prog->check_substr,n_a);
4601 if (!PL_colorset) reginitcolors();
4602 PerlIO_printf(Perl_debug_log,
4603 "%sUsing REx substr:%s `%s%.60s%s%s'\n",
4604 PL_colors[4],PL_colors[5],PL_colors[0],
4607 (strlen(s) > 60 ? "..." : ""));
4610 return prog->check_substr;
4614 Perl_pregfree(pTHX_ struct regexp *r)
4616 DEBUG_r(if (!PL_colorset) reginitcolors());
4618 if (!r || (--r->refcnt > 0))
4620 DEBUG_r(PerlIO_printf(Perl_debug_log,
4621 "%sFreeing REx:%s `%s%.60s%s%s'\n",
4622 PL_colors[4],PL_colors[5],PL_colors[0],
4625 (strlen(r->precomp) > 60 ? "..." : "")));
4628 Safefree(r->precomp);
4629 if (r->offsets) /* 20010421 MJD */
4630 Safefree(r->offsets);
4631 if (RX_MATCH_COPIED(r))
4632 Safefree(r->subbeg);
4634 if (r->anchored_substr)
4635 SvREFCNT_dec(r->anchored_substr);
4636 if (r->float_substr)
4637 SvREFCNT_dec(r->float_substr);
4638 Safefree(r->substrs);
4641 int n = r->data->count;
4642 AV* new_comppad = NULL;
4647 /* If you add a ->what type here, update the comment in regcomp.h */
4648 switch (r->data->what[n]) {
4650 SvREFCNT_dec((SV*)r->data->data[n]);
4653 Safefree(r->data->data[n]);
4656 new_comppad = (AV*)r->data->data[n];
4659 if (new_comppad == NULL)
4660 Perl_croak(aTHX_ "panic: pregfree comppad");
4661 old_comppad = PL_comppad;
4662 old_curpad = PL_curpad;
4663 /* Watch out for global destruction's random ordering. */
4664 if (SvTYPE(new_comppad) == SVt_PVAV) {
4665 PL_comppad = new_comppad;
4666 PL_curpad = AvARRAY(new_comppad);
4671 if (!OpREFCNT_dec((OP_4tree*)r->data->data[n])) {
4672 op_free((OP_4tree*)r->data->data[n]);
4675 PL_comppad = old_comppad;
4676 PL_curpad = old_curpad;
4677 SvREFCNT_dec((SV*)new_comppad);
4683 Perl_croak(aTHX_ "panic: regfree data code '%c'", r->data->what[n]);
4686 Safefree(r->data->what);
4689 Safefree(r->startp);
4695 - regnext - dig the "next" pointer out of a node
4697 * [Note, when REGALIGN is defined there are two places in regmatch()
4698 * that bypass this code for speed.]
4701 Perl_regnext(pTHX_ register regnode *p)
4703 register I32 offset;
4705 if (p == &PL_regdummy)
4708 offset = (reg_off_by_arg[OP(p)] ? ARG(p) : NEXT_OFF(p));
4716 S_re_croak2(pTHX_ const char* pat1,const char* pat2,...)
4719 STRLEN l1 = strlen(pat1);
4720 STRLEN l2 = strlen(pat2);
4729 Copy(pat1, buf, l1 , char);
4730 Copy(pat2, buf + l1, l2 , char);
4731 buf[l1 + l2] = '\n';
4732 buf[l1 + l2 + 1] = '\0';
4734 /* ANSI variant takes additional second argument */
4735 va_start(args, pat2);
4739 msv = vmess(buf, &args);
4741 message = SvPV(msv,l1);
4744 Copy(message, buf, l1 , char);
4745 buf[l1] = '\0'; /* Overwrite \n */
4746 Perl_croak(aTHX_ "%s", buf);
4749 /* XXX Here's a total kludge. But we need to re-enter for swash routines. */
4752 Perl_save_re_context(pTHX)
4755 SAVEPPTR(RExC_precomp); /* uncompiled string. */
4756 SAVEI32(RExC_npar); /* () count. */
4757 SAVEI32(RExC_size); /* Code size. */
4758 SAVEI16(RExC_flags16); /* are we folding, multilining? */
4759 SAVEVPTR(RExC_rx); /* from regcomp.c */
4760 SAVEI32(RExC_seen); /* from regcomp.c */
4761 SAVEI32(RExC_sawback); /* Did we see \1, ...? */
4762 SAVEI32(RExC_naughty); /* How bad is this pattern? */
4763 SAVEVPTR(RExC_emit); /* Code-emit pointer; ®dummy = don't */
4764 SAVEPPTR(RExC_end); /* End of input for compile */
4765 SAVEPPTR(RExC_parse); /* Input-scan pointer. */
4768 SAVEI32(PL_reg_flags); /* from regexec.c */
4770 SAVEPPTR(PL_reginput); /* String-input pointer. */
4771 SAVEPPTR(PL_regbol); /* Beginning of input, for ^ check. */
4772 SAVEPPTR(PL_regeol); /* End of input, for $ check. */
4773 SAVEVPTR(PL_regstartp); /* Pointer to startp array. */
4774 SAVEVPTR(PL_regendp); /* Ditto for endp. */
4775 SAVEVPTR(PL_reglastparen); /* Similarly for lastparen. */
4776 SAVEPPTR(PL_regtill); /* How far we are required to go. */
4777 SAVEGENERICPV(PL_reg_start_tmp); /* from regexec.c */
4778 PL_reg_start_tmp = 0;
4779 SAVEI32(PL_reg_start_tmpl); /* from regexec.c */
4780 PL_reg_start_tmpl = 0;
4781 SAVEVPTR(PL_regdata);
4782 SAVEI32(PL_reg_eval_set); /* from regexec.c */
4783 SAVEI32(PL_regnarrate); /* from regexec.c */
4784 SAVEVPTR(PL_regprogram); /* from regexec.c */
4785 SAVEINT(PL_regindent); /* from regexec.c */
4786 SAVEVPTR(PL_regcc); /* from regexec.c */
4787 SAVEVPTR(PL_curcop);
4788 SAVEVPTR(PL_reg_call_cc); /* from regexec.c */
4789 SAVEVPTR(PL_reg_re); /* from regexec.c */
4790 SAVEPPTR(PL_reg_ganch); /* from regexec.c */
4791 SAVESPTR(PL_reg_sv); /* from regexec.c */
4792 SAVEI8(PL_reg_match_utf8); /* from regexec.c */
4793 SAVEVPTR(PL_reg_magic); /* from regexec.c */
4794 SAVEI32(PL_reg_oldpos); /* from regexec.c */
4795 SAVEVPTR(PL_reg_oldcurpm); /* from regexec.c */
4796 SAVEVPTR(PL_reg_curpm); /* from regexec.c */
4797 SAVEI32(PL_regnpar); /* () count. */
4798 SAVEI32(PL_regsize); /* from regexec.c */
4800 SAVEPPTR(PL_reg_starttry); /* from regexec.c */
4805 clear_re(pTHX_ void *r)
4807 ReREFCNT_dec((regexp *)r);
https://perl5.git.perl.org / perl5.git / blob