5 * "A fair jaw-cracker dwarf-language must be." --Samwise Gamgee
8 /* This file contains functions for compiling a regular expression. See
9 * also regexec.c which funnily enough, contains functions for executing
10 * a regular expression.
12 * This file is also copied at build time to ext/re/re_comp.c, where
13 * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT.
14 * This causes the main functions to be compiled under new names and with
15 * debugging support added, which makes "use re 'debug'" work.
18 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
19 * confused with the original package (see point 3 below). Thanks, Henry!
22 /* Additional note: this code is very heavily munged from Henry's version
23 * in places. In some spots I've traded clarity for efficiency, so don't
24 * blame Henry for some of the lack of readability.
27 /* The names of the functions have been changed from regcomp and
28 * regexec to pregcomp and pregexec in order to avoid conflicts
29 * with the POSIX routines of the same names.
32 #ifdef PERL_EXT_RE_BUILD
37 * pregcomp and pregexec -- regsub and regerror are not used in perl
39 * Copyright (c) 1986 by University of Toronto.
40 * Written by Henry Spencer. Not derived from licensed software.
42 * Permission is granted to anyone to use this software for any
43 * purpose on any computer system, and to redistribute it freely,
44 * subject to the following restrictions:
46 * 1. The author is not responsible for the consequences of use of
47 * this software, no matter how awful, even if they arise
50 * 2. The origin of this software must not be misrepresented, either
51 * by explicit claim or by omission.
53 * 3. Altered versions must be plainly marked as such, and must not
54 * be misrepresented as being the original software.
57 **** Alterations to Henry's code are...
59 **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
60 **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, by Larry Wall and others
62 **** You may distribute under the terms of either the GNU General Public
63 **** License or the Artistic License, as specified in the README file.
66 * Beware that some of this code is subtly aware of the way operator
67 * precedence is structured in regular expressions. Serious changes in
68 * regular-expression syntax might require a total rethink.
71 #define PERL_IN_REGCOMP_C
74 #ifndef PERL_IN_XSUB_RE
79 #ifdef PERL_IN_XSUB_RE
90 # if defined(BUGGY_MSC6)
91 /* MSC 6.00A breaks on op/regexp.t test 85 unless we turn this off */
92 # pragma optimize("a",off)
93 /* But MSC 6.00A is happy with 'w', for aliases only across function calls*/
94 # pragma optimize("w",on )
95 # endif /* BUGGY_MSC6 */
102 typedef struct RExC_state_t {
103 U32 flags; /* are we folding, multilining? */
104 char *precomp; /* uncompiled string. */
105 regexp *rx; /* perl core regexp structure */
106 regexp_internal *rxi; /* internal data for regexp object pprivate field */
107 char *start; /* Start of input for compile */
108 char *end; /* End of input for compile */
109 char *parse; /* Input-scan pointer. */
110 I32 whilem_seen; /* number of WHILEM in this expr */
111 regnode *emit_start; /* Start of emitted-code area */
112 regnode *emit; /* Code-emit pointer; ®dummy = don't = compiling */
113 I32 naughty; /* How bad is this pattern? */
114 I32 sawback; /* Did we see \1, ...? */
116 I32 size; /* Code size. */
117 I32 npar; /* Capture buffer count, (OPEN). */
118 I32 cpar; /* Capture buffer count, (CLOSE). */
119 I32 nestroot; /* root parens we are in - used by accept */
123 regnode **open_parens; /* pointers to open parens */
124 regnode **close_parens; /* pointers to close parens */
125 regnode *opend; /* END node in program */
127 HV *charnames; /* cache of named sequences */
128 HV *paren_names; /* Paren names */
129 regnode **recurse; /* Recurse regops */
130 I32 recurse_count; /* Number of recurse regops */
132 char *starttry; /* -Dr: where regtry was called. */
133 #define RExC_starttry (pRExC_state->starttry)
136 const char *lastparse;
138 #define RExC_lastparse (pRExC_state->lastparse)
139 #define RExC_lastnum (pRExC_state->lastnum)
143 #define RExC_flags (pRExC_state->flags)
144 #define RExC_precomp (pRExC_state->precomp)
145 #define RExC_rx (pRExC_state->rx)
146 #define RExC_rxi (pRExC_state->rxi)
147 #define RExC_start (pRExC_state->start)
148 #define RExC_end (pRExC_state->end)
149 #define RExC_parse (pRExC_state->parse)
150 #define RExC_whilem_seen (pRExC_state->whilem_seen)
151 #define RExC_offsets (pRExC_state->rxi->offsets) /* I am not like the others */
152 #define RExC_emit (pRExC_state->emit)
153 #define RExC_emit_start (pRExC_state->emit_start)
154 #define RExC_naughty (pRExC_state->naughty)
155 #define RExC_sawback (pRExC_state->sawback)
156 #define RExC_seen (pRExC_state->seen)
157 #define RExC_size (pRExC_state->size)
158 #define RExC_npar (pRExC_state->npar)
159 #define RExC_cpar (pRExC_state->cpar)
160 #define RExC_nestroot (pRExC_state->nestroot)
161 #define RExC_extralen (pRExC_state->extralen)
162 #define RExC_seen_zerolen (pRExC_state->seen_zerolen)
163 #define RExC_seen_evals (pRExC_state->seen_evals)
164 #define RExC_utf8 (pRExC_state->utf8)
165 #define RExC_charnames (pRExC_state->charnames)
166 #define RExC_open_parens (pRExC_state->open_parens)
167 #define RExC_close_parens (pRExC_state->close_parens)
168 #define RExC_opend (pRExC_state->opend)
169 #define RExC_paren_names (pRExC_state->paren_names)
170 #define RExC_recurse (pRExC_state->recurse)
171 #define RExC_recurse_count (pRExC_state->recurse_count)
173 #define ISMULT1(c) ((c) == '*' || (c) == '+' || (c) == '?')
174 #define ISMULT2(s) ((*s) == '*' || (*s) == '+' || (*s) == '?' || \
175 ((*s) == '{' && regcurly(s)))
178 #undef SPSTART /* dratted cpp namespace... */
181 * Flags to be passed up and down.
183 #define WORST 0 /* Worst case. */
184 #define HASWIDTH 0x1 /* Known to match non-null strings. */
185 #define SIMPLE 0x2 /* Simple enough to be STAR/PLUS operand. */
186 #define SPSTART 0x4 /* Starts with * or +. */
187 #define TRYAGAIN 0x8 /* Weeded out a declaration. */
189 #define REG_NODE_NUM(x) ((x) ? (int)((x)-RExC_emit_start) : -1)
191 /* whether trie related optimizations are enabled */
192 #if PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION
193 #define TRIE_STUDY_OPT
194 #define FULL_TRIE_STUDY
200 #define PBYTE(u8str,paren) ((U8*)(u8str))[(paren) >> 3]
201 #define PBITVAL(paren) (1 << ((paren) & 7))
202 #define PAREN_TEST(u8str,paren) ( PBYTE(u8str,paren) & PBITVAL(paren))
203 #define PAREN_SET(u8str,paren) PBYTE(u8str,paren) |= PBITVAL(paren)
204 #define PAREN_UNSET(u8str,paren) PBYTE(u8str,paren) &= (~PBITVAL(paren))
207 /* About scan_data_t.
209 During optimisation we recurse through the regexp program performing
210 various inplace (keyhole style) optimisations. In addition study_chunk
211 and scan_commit populate this data structure with information about
212 what strings MUST appear in the pattern. We look for the longest
213 string that must appear for at a fixed location, and we look for the
214 longest string that may appear at a floating location. So for instance
219 Both 'FOO' and 'A' are fixed strings. Both 'B' and 'BAR' are floating
220 strings (because they follow a .* construct). study_chunk will identify
221 both FOO and BAR as being the longest fixed and floating strings respectively.
223 The strings can be composites, for instance
227 will result in a composite fixed substring 'foo'.
229 For each string some basic information is maintained:
231 - offset or min_offset
232 This is the position the string must appear at, or not before.
233 It also implicitly (when combined with minlenp) tells us how many
234 character must match before the string we are searching.
235 Likewise when combined with minlenp and the length of the string
236 tells us how many characters must appear after the string we have
240 Only used for floating strings. This is the rightmost point that
241 the string can appear at. Ifset to I32 max it indicates that the
242 string can occur infinitely far to the right.
245 A pointer to the minimum length of the pattern that the string
246 was found inside. This is important as in the case of positive
247 lookahead or positive lookbehind we can have multiple patterns
252 The minimum length of the pattern overall is 3, the minimum length
253 of the lookahead part is 3, but the minimum length of the part that
254 will actually match is 1. So 'FOO's minimum length is 3, but the
255 minimum length for the F is 1. This is important as the minimum length
256 is used to determine offsets in front of and behind the string being
257 looked for. Since strings can be composites this is the length of the
258 pattern at the time it was commited with a scan_commit. Note that
259 the length is calculated by study_chunk, so that the minimum lengths
260 are not known until the full pattern has been compiled, thus the
261 pointer to the value.
265 In the case of lookbehind the string being searched for can be
266 offset past the start point of the final matching string.
267 If this value was just blithely removed from the min_offset it would
268 invalidate some of the calculations for how many chars must match
269 before or after (as they are derived from min_offset and minlen and
270 the length of the string being searched for).
271 When the final pattern is compiled and the data is moved from the
272 scan_data_t structure into the regexp structure the information
273 about lookbehind is factored in, with the information that would
274 have been lost precalculated in the end_shift field for the
277 The fields pos_min and pos_delta are used to store the minimum offset
278 and the delta to the maximum offset at the current point in the pattern.
282 typedef struct scan_data_t {
283 /*I32 len_min; unused */
284 /*I32 len_delta; unused */
288 I32 last_end; /* min value, <0 unless valid. */
291 SV **longest; /* Either &l_fixed, or &l_float. */
292 SV *longest_fixed; /* longest fixed string found in pattern */
293 I32 offset_fixed; /* offset where it starts */
294 I32 *minlen_fixed; /* pointer to the minlen relevent to the string */
295 I32 lookbehind_fixed; /* is the position of the string modfied by LB */
296 SV *longest_float; /* longest floating string found in pattern */
297 I32 offset_float_min; /* earliest point in string it can appear */
298 I32 offset_float_max; /* latest point in string it can appear */
299 I32 *minlen_float; /* pointer to the minlen relevent to the string */
300 I32 lookbehind_float; /* is the position of the string modified by LB */
304 struct regnode_charclass_class *start_class;
308 * Forward declarations for pregcomp()'s friends.
311 static const scan_data_t zero_scan_data =
312 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ,0};
314 #define SF_BEFORE_EOL (SF_BEFORE_SEOL|SF_BEFORE_MEOL)
315 #define SF_BEFORE_SEOL 0x0001
316 #define SF_BEFORE_MEOL 0x0002
317 #define SF_FIX_BEFORE_EOL (SF_FIX_BEFORE_SEOL|SF_FIX_BEFORE_MEOL)
318 #define SF_FL_BEFORE_EOL (SF_FL_BEFORE_SEOL|SF_FL_BEFORE_MEOL)
321 # define SF_FIX_SHIFT_EOL (0+2)
322 # define SF_FL_SHIFT_EOL (0+4)
324 # define SF_FIX_SHIFT_EOL (+2)
325 # define SF_FL_SHIFT_EOL (+4)
328 #define SF_FIX_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FIX_SHIFT_EOL)
329 #define SF_FIX_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FIX_SHIFT_EOL)
331 #define SF_FL_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FL_SHIFT_EOL)
332 #define SF_FL_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FL_SHIFT_EOL) /* 0x20 */
333 #define SF_IS_INF 0x0040
334 #define SF_HAS_PAR 0x0080
335 #define SF_IN_PAR 0x0100
336 #define SF_HAS_EVAL 0x0200
337 #define SCF_DO_SUBSTR 0x0400
338 #define SCF_DO_STCLASS_AND 0x0800
339 #define SCF_DO_STCLASS_OR 0x1000
340 #define SCF_DO_STCLASS (SCF_DO_STCLASS_AND|SCF_DO_STCLASS_OR)
341 #define SCF_WHILEM_VISITED_POS 0x2000
343 #define SCF_TRIE_RESTUDY 0x4000 /* Do restudy? */
344 #define SCF_SEEN_ACCEPT 0x8000
346 #define UTF (RExC_utf8 != 0)
347 #define LOC ((RExC_flags & RXf_PMf_LOCALE) != 0)
348 #define FOLD ((RExC_flags & RXf_PMf_FOLD) != 0)
350 #define OOB_UNICODE 12345678
351 #define OOB_NAMEDCLASS -1
353 #define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
354 #define CHR_DIST(a,b) (UTF ? utf8_distance(a,b) : a - b)
357 /* length of regex to show in messages that don't mark a position within */
358 #define RegexLengthToShowInErrorMessages 127
361 * If MARKER[12] are adjusted, be sure to adjust the constants at the top
362 * of t/op/regmesg.t, the tests in t/op/re_tests, and those in
363 * op/pragma/warn/regcomp.
365 #define MARKER1 "<-- HERE" /* marker as it appears in the description */
366 #define MARKER2 " <-- HERE " /* marker as it appears within the regex */
368 #define REPORT_LOCATION " in regex; marked by " MARKER1 " in m/%.*s" MARKER2 "%s/"
371 * Calls SAVEDESTRUCTOR_X if needed, then calls Perl_croak with the given
372 * arg. Show regex, up to a maximum length. If it's too long, chop and add
375 #define _FAIL(code) STMT_START { \
376 const char *ellipses = ""; \
377 IV len = RExC_end - RExC_precomp; \
380 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
381 if (len > RegexLengthToShowInErrorMessages) { \
382 /* chop 10 shorter than the max, to ensure meaning of "..." */ \
383 len = RegexLengthToShowInErrorMessages - 10; \
389 #define FAIL(msg) _FAIL( \
390 Perl_croak(aTHX_ "%s in regex m/%.*s%s/", \
391 msg, (int)len, RExC_precomp, ellipses))
393 #define FAIL2(msg,arg) _FAIL( \
394 Perl_croak(aTHX_ msg " in regex m/%.*s%s/", \
395 arg, (int)len, RExC_precomp, ellipses))
398 * Simple_vFAIL -- like FAIL, but marks the current location in the scan
400 #define Simple_vFAIL(m) STMT_START { \
401 const IV offset = RExC_parse - RExC_precomp; \
402 Perl_croak(aTHX_ "%s" REPORT_LOCATION, \
403 m, (int)offset, RExC_precomp, RExC_precomp + offset); \
407 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL()
409 #define vFAIL(m) STMT_START { \
411 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
416 * Like Simple_vFAIL(), but accepts two arguments.
418 #define Simple_vFAIL2(m,a1) STMT_START { \
419 const IV offset = RExC_parse - RExC_precomp; \
420 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, \
421 (int)offset, RExC_precomp, RExC_precomp + offset); \
425 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL2().
427 #define vFAIL2(m,a1) STMT_START { \
429 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
430 Simple_vFAIL2(m, a1); \
435 * Like Simple_vFAIL(), but accepts three arguments.
437 #define Simple_vFAIL3(m, a1, a2) STMT_START { \
438 const IV offset = RExC_parse - RExC_precomp; \
439 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, \
440 (int)offset, RExC_precomp, RExC_precomp + offset); \
444 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL3().
446 #define vFAIL3(m,a1,a2) STMT_START { \
448 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
449 Simple_vFAIL3(m, a1, a2); \
453 * Like Simple_vFAIL(), but accepts four arguments.
455 #define Simple_vFAIL4(m, a1, a2, a3) STMT_START { \
456 const IV offset = RExC_parse - RExC_precomp; \
457 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, a3, \
458 (int)offset, RExC_precomp, RExC_precomp + offset); \
461 #define vWARN(loc,m) STMT_START { \
462 const IV offset = loc - RExC_precomp; \
463 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s" REPORT_LOCATION, \
464 m, (int)offset, RExC_precomp, RExC_precomp + offset); \
467 #define vWARNdep(loc,m) STMT_START { \
468 const IV offset = loc - RExC_precomp; \
469 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_REGEXP), \
470 "%s" REPORT_LOCATION, \
471 m, (int)offset, RExC_precomp, RExC_precomp + offset); \
475 #define vWARN2(loc, m, a1) STMT_START { \
476 const IV offset = loc - RExC_precomp; \
477 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
478 a1, (int)offset, RExC_precomp, RExC_precomp + offset); \
481 #define vWARN3(loc, m, a1, a2) STMT_START { \
482 const IV offset = loc - RExC_precomp; \
483 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
484 a1, a2, (int)offset, RExC_precomp, RExC_precomp + offset); \
487 #define vWARN4(loc, m, a1, a2, a3) STMT_START { \
488 const IV offset = loc - RExC_precomp; \
489 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
490 a1, a2, a3, (int)offset, RExC_precomp, RExC_precomp + offset); \
493 #define vWARN5(loc, m, a1, a2, a3, a4) STMT_START { \
494 const IV offset = loc - RExC_precomp; \
495 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
496 a1, a2, a3, a4, (int)offset, RExC_precomp, RExC_precomp + offset); \
500 /* Allow for side effects in s */
501 #define REGC(c,s) STMT_START { \
502 if (!SIZE_ONLY) *(s) = (c); else (void)(s); \
505 /* Macros for recording node offsets. 20001227 mjd@plover.com
506 * Nodes are numbered 1, 2, 3, 4. Node #n's position is recorded in
507 * element 2*n-1 of the array. Element #2n holds the byte length node #n.
508 * Element 0 holds the number n.
509 * Position is 1 indexed.
512 #define Set_Node_Offset_To_R(node,byte) STMT_START { \
514 MJD_OFFSET_DEBUG(("** (%d) offset of node %d is %d.\n", \
515 __LINE__, (int)(node), (int)(byte))); \
517 Perl_croak(aTHX_ "value of node is %d in Offset macro", (int)(node)); \
519 RExC_offsets[2*(node)-1] = (byte); \
524 #define Set_Node_Offset(node,byte) \
525 Set_Node_Offset_To_R((node)-RExC_emit_start, (byte)-RExC_start)
526 #define Set_Cur_Node_Offset Set_Node_Offset(RExC_emit, RExC_parse)
528 #define Set_Node_Length_To_R(node,len) STMT_START { \
530 MJD_OFFSET_DEBUG(("** (%d) size of node %d is %d.\n", \
531 __LINE__, (int)(node), (int)(len))); \
533 Perl_croak(aTHX_ "value of node is %d in Length macro", (int)(node)); \
535 RExC_offsets[2*(node)] = (len); \
540 #define Set_Node_Length(node,len) \
541 Set_Node_Length_To_R((node)-RExC_emit_start, len)
542 #define Set_Cur_Node_Length(len) Set_Node_Length(RExC_emit, len)
543 #define Set_Node_Cur_Length(node) \
544 Set_Node_Length(node, RExC_parse - parse_start)
546 /* Get offsets and lengths */
547 #define Node_Offset(n) (RExC_offsets[2*((n)-RExC_emit_start)-1])
548 #define Node_Length(n) (RExC_offsets[2*((n)-RExC_emit_start)])
550 #define Set_Node_Offset_Length(node,offset,len) STMT_START { \
551 Set_Node_Offset_To_R((node)-RExC_emit_start, (offset)); \
552 Set_Node_Length_To_R((node)-RExC_emit_start, (len)); \
556 #if PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS
557 #define EXPERIMENTAL_INPLACESCAN
560 #define DEBUG_STUDYDATA(data,depth) \
561 DEBUG_OPTIMISE_MORE_r(if(data){ \
562 PerlIO_printf(Perl_debug_log, \
563 "%*s"/* Len:%"IVdf"/%"IVdf" */"Pos:%"IVdf"/%"IVdf \
564 " Flags: %"IVdf" Whilem_c: %"IVdf" Lcp: %"IVdf" ", \
565 (int)(depth)*2, "", \
566 (IV)((data)->pos_min), \
567 (IV)((data)->pos_delta), \
568 (IV)((data)->flags), \
569 (IV)((data)->whilem_c), \
570 (IV)((data)->last_closep ? *((data)->last_closep) : -1) \
572 if ((data)->last_found) \
573 PerlIO_printf(Perl_debug_log, \
574 "Last:'%s' %"IVdf":%"IVdf"/%"IVdf" %sFixed:'%s' @ %"IVdf \
575 " %sFloat: '%s' @ %"IVdf"/%"IVdf"", \
576 SvPVX_const((data)->last_found), \
577 (IV)((data)->last_end), \
578 (IV)((data)->last_start_min), \
579 (IV)((data)->last_start_max), \
580 ((data)->longest && \
581 (data)->longest==&((data)->longest_fixed)) ? "*" : "", \
582 SvPVX_const((data)->longest_fixed), \
583 (IV)((data)->offset_fixed), \
584 ((data)->longest && \
585 (data)->longest==&((data)->longest_float)) ? "*" : "", \
586 SvPVX_const((data)->longest_float), \
587 (IV)((data)->offset_float_min), \
588 (IV)((data)->offset_float_max) \
590 PerlIO_printf(Perl_debug_log,"\n"); \
593 static void clear_re(pTHX_ void *r);
595 /* Mark that we cannot extend a found fixed substring at this point.
596 Update the longest found anchored substring and the longest found
597 floating substrings if needed. */
600 S_scan_commit(pTHX_ const RExC_state_t *pRExC_state, scan_data_t *data, I32 *minlenp)
602 const STRLEN l = CHR_SVLEN(data->last_found);
603 const STRLEN old_l = CHR_SVLEN(*data->longest);
604 GET_RE_DEBUG_FLAGS_DECL;
606 if ((l >= old_l) && ((l > old_l) || (data->flags & SF_BEFORE_EOL))) {
607 SvSetMagicSV(*data->longest, data->last_found);
608 if (*data->longest == data->longest_fixed) {
609 data->offset_fixed = l ? data->last_start_min : data->pos_min;
610 if (data->flags & SF_BEFORE_EOL)
612 |= ((data->flags & SF_BEFORE_EOL) << SF_FIX_SHIFT_EOL);
614 data->flags &= ~SF_FIX_BEFORE_EOL;
615 data->minlen_fixed=minlenp;
616 data->lookbehind_fixed=0;
619 data->offset_float_min = l ? data->last_start_min : data->pos_min;
620 data->offset_float_max = (l
621 ? data->last_start_max
622 : data->pos_min + data->pos_delta);
623 if ((U32)data->offset_float_max > (U32)I32_MAX)
624 data->offset_float_max = I32_MAX;
625 if (data->flags & SF_BEFORE_EOL)
627 |= ((data->flags & SF_BEFORE_EOL) << SF_FL_SHIFT_EOL);
629 data->flags &= ~SF_FL_BEFORE_EOL;
630 data->minlen_float=minlenp;
631 data->lookbehind_float=0;
634 SvCUR_set(data->last_found, 0);
636 SV * const sv = data->last_found;
637 if (SvUTF8(sv) && SvMAGICAL(sv)) {
638 MAGIC * const mg = mg_find(sv, PERL_MAGIC_utf8);
644 data->flags &= ~SF_BEFORE_EOL;
645 DEBUG_STUDYDATA(data,0);
648 /* Can match anything (initialization) */
650 S_cl_anything(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
652 ANYOF_CLASS_ZERO(cl);
653 ANYOF_BITMAP_SETALL(cl);
654 cl->flags = ANYOF_EOS|ANYOF_UNICODE_ALL;
656 cl->flags |= ANYOF_LOCALE;
659 /* Can match anything (initialization) */
661 S_cl_is_anything(const struct regnode_charclass_class *cl)
665 for (value = 0; value <= ANYOF_MAX; value += 2)
666 if (ANYOF_CLASS_TEST(cl, value) && ANYOF_CLASS_TEST(cl, value + 1))
668 if (!(cl->flags & ANYOF_UNICODE_ALL))
670 if (!ANYOF_BITMAP_TESTALLSET((const void*)cl))
675 /* Can match anything (initialization) */
677 S_cl_init(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
679 Zero(cl, 1, struct regnode_charclass_class);
681 cl_anything(pRExC_state, cl);
685 S_cl_init_zero(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
687 Zero(cl, 1, struct regnode_charclass_class);
689 cl_anything(pRExC_state, cl);
691 cl->flags |= ANYOF_LOCALE;
694 /* 'And' a given class with another one. Can create false positives */
695 /* We assume that cl is not inverted */
697 S_cl_and(struct regnode_charclass_class *cl,
698 const struct regnode_charclass_class *and_with)
701 assert(and_with->type == ANYOF);
702 if (!(and_with->flags & ANYOF_CLASS)
703 && !(cl->flags & ANYOF_CLASS)
704 && (and_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
705 && !(and_with->flags & ANYOF_FOLD)
706 && !(cl->flags & ANYOF_FOLD)) {
709 if (and_with->flags & ANYOF_INVERT)
710 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
711 cl->bitmap[i] &= ~and_with->bitmap[i];
713 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
714 cl->bitmap[i] &= and_with->bitmap[i];
715 } /* XXXX: logic is complicated otherwise, leave it along for a moment. */
716 if (!(and_with->flags & ANYOF_EOS))
717 cl->flags &= ~ANYOF_EOS;
719 if (cl->flags & ANYOF_UNICODE_ALL && and_with->flags & ANYOF_UNICODE &&
720 !(and_with->flags & ANYOF_INVERT)) {
721 cl->flags &= ~ANYOF_UNICODE_ALL;
722 cl->flags |= ANYOF_UNICODE;
723 ARG_SET(cl, ARG(and_with));
725 if (!(and_with->flags & ANYOF_UNICODE_ALL) &&
726 !(and_with->flags & ANYOF_INVERT))
727 cl->flags &= ~ANYOF_UNICODE_ALL;
728 if (!(and_with->flags & (ANYOF_UNICODE|ANYOF_UNICODE_ALL)) &&
729 !(and_with->flags & ANYOF_INVERT))
730 cl->flags &= ~ANYOF_UNICODE;
733 /* 'OR' a given class with another one. Can create false positives */
734 /* We assume that cl is not inverted */
736 S_cl_or(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl, const struct regnode_charclass_class *or_with)
738 if (or_with->flags & ANYOF_INVERT) {
740 * (B1 | CL1) | (!B2 & !CL2) = (B1 | !B2 & !CL2) | (CL1 | (!B2 & !CL2))
741 * <= (B1 | !B2) | (CL1 | !CL2)
742 * which is wasteful if CL2 is small, but we ignore CL2:
743 * (B1 | CL1) | (!B2 & !CL2) <= (B1 | CL1) | !B2 = (B1 | !B2) | CL1
744 * XXXX Can we handle case-fold? Unclear:
745 * (OK1(i) | OK1(i')) | !(OK1(i) | OK1(i')) =
746 * (OK1(i) | OK1(i')) | (!OK1(i) & !OK1(i'))
748 if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
749 && !(or_with->flags & ANYOF_FOLD)
750 && !(cl->flags & ANYOF_FOLD) ) {
753 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
754 cl->bitmap[i] |= ~or_with->bitmap[i];
755 } /* XXXX: logic is complicated otherwise */
757 cl_anything(pRExC_state, cl);
760 /* (B1 | CL1) | (B2 | CL2) = (B1 | B2) | (CL1 | CL2)) */
761 if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
762 && (!(or_with->flags & ANYOF_FOLD)
763 || (cl->flags & ANYOF_FOLD)) ) {
766 /* OR char bitmap and class bitmap separately */
767 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
768 cl->bitmap[i] |= or_with->bitmap[i];
769 if (or_with->flags & ANYOF_CLASS) {
770 for (i = 0; i < ANYOF_CLASSBITMAP_SIZE; i++)
771 cl->classflags[i] |= or_with->classflags[i];
772 cl->flags |= ANYOF_CLASS;
775 else { /* XXXX: logic is complicated, leave it along for a moment. */
776 cl_anything(pRExC_state, cl);
779 if (or_with->flags & ANYOF_EOS)
780 cl->flags |= ANYOF_EOS;
782 if (cl->flags & ANYOF_UNICODE && or_with->flags & ANYOF_UNICODE &&
783 ARG(cl) != ARG(or_with)) {
784 cl->flags |= ANYOF_UNICODE_ALL;
785 cl->flags &= ~ANYOF_UNICODE;
787 if (or_with->flags & ANYOF_UNICODE_ALL) {
788 cl->flags |= ANYOF_UNICODE_ALL;
789 cl->flags &= ~ANYOF_UNICODE;
793 #define TRIE_LIST_ITEM(state,idx) (trie->states[state].trans.list)[ idx ]
794 #define TRIE_LIST_CUR(state) ( TRIE_LIST_ITEM( state, 0 ).forid )
795 #define TRIE_LIST_LEN(state) ( TRIE_LIST_ITEM( state, 0 ).newstate )
796 #define TRIE_LIST_USED(idx) ( trie->states[state].trans.list ? (TRIE_LIST_CUR( idx ) - 1) : 0 )
802 dump_trie_interim_list(trie,next_alloc)
803 dump_trie_interim_table(trie,next_alloc)
805 These routines dump out a trie in a somewhat readable format.
806 The _interim_ variants are used for debugging the interim
807 tables that are used to generate the final compressed
808 representation which is what dump_trie expects.
810 Part of the reason for their existance is to provide a form
811 of documentation as to how the different representations function.
817 Dumps the final compressed table form of the trie to Perl_debug_log.
818 Used for debugging make_trie().
822 S_dump_trie(pTHX_ const struct _reg_trie_data *trie,U32 depth)
825 SV *sv=sv_newmortal();
826 int colwidth= trie->widecharmap ? 6 : 4;
827 GET_RE_DEBUG_FLAGS_DECL;
830 PerlIO_printf( Perl_debug_log, "%*sChar : %-6s%-6s%-4s ",
831 (int)depth * 2 + 2,"",
832 "Match","Base","Ofs" );
834 for( state = 0 ; state < trie->uniquecharcount ; state++ ) {
835 SV ** const tmp = av_fetch( trie->revcharmap, state, 0);
837 PerlIO_printf( Perl_debug_log, "%*s",
839 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
840 PL_colors[0], PL_colors[1],
841 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
842 PERL_PV_ESCAPE_FIRSTCHAR
847 PerlIO_printf( Perl_debug_log, "\n%*sState|-----------------------",
848 (int)depth * 2 + 2,"");
850 for( state = 0 ; state < trie->uniquecharcount ; state++ )
851 PerlIO_printf( Perl_debug_log, "%.*s", colwidth, "--------");
852 PerlIO_printf( Perl_debug_log, "\n");
854 for( state = 1 ; state < trie->statecount ; state++ ) {
855 const U32 base = trie->states[ state ].trans.base;
857 PerlIO_printf( Perl_debug_log, "%*s#%4"UVXf"|", (int)depth * 2 + 2,"", (UV)state);
859 if ( trie->states[ state ].wordnum ) {
860 PerlIO_printf( Perl_debug_log, " W%4X", trie->states[ state ].wordnum );
862 PerlIO_printf( Perl_debug_log, "%6s", "" );
865 PerlIO_printf( Perl_debug_log, " @%4"UVXf" ", (UV)base );
870 while( ( base + ofs < trie->uniquecharcount ) ||
871 ( base + ofs - trie->uniquecharcount < trie->lasttrans
872 && trie->trans[ base + ofs - trie->uniquecharcount ].check != state))
875 PerlIO_printf( Perl_debug_log, "+%2"UVXf"[ ", (UV)ofs);
877 for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) {
878 if ( ( base + ofs >= trie->uniquecharcount ) &&
879 ( base + ofs - trie->uniquecharcount < trie->lasttrans ) &&
880 trie->trans[ base + ofs - trie->uniquecharcount ].check == state )
882 PerlIO_printf( Perl_debug_log, "%*"UVXf,
884 (UV)trie->trans[ base + ofs - trie->uniquecharcount ].next );
886 PerlIO_printf( Perl_debug_log, "%*s",colwidth," ." );
890 PerlIO_printf( Perl_debug_log, "]");
893 PerlIO_printf( Perl_debug_log, "\n" );
897 dump_trie_interim_list(trie,next_alloc)
898 Dumps a fully constructed but uncompressed trie in list form.
899 List tries normally only are used for construction when the number of
900 possible chars (trie->uniquecharcount) is very high.
901 Used for debugging make_trie().
904 S_dump_trie_interim_list(pTHX_ const struct _reg_trie_data *trie, U32 next_alloc,U32 depth)
907 SV *sv=sv_newmortal();
908 int colwidth= trie->widecharmap ? 6 : 4;
909 GET_RE_DEBUG_FLAGS_DECL;
910 /* print out the table precompression. */
911 PerlIO_printf( Perl_debug_log, "%*sState :Word | Transition Data\n%*s%s",
912 (int)depth * 2 + 2,"", (int)depth * 2 + 2,"",
913 "------:-----+-----------------\n" );
915 for( state=1 ; state < next_alloc ; state ++ ) {
918 PerlIO_printf( Perl_debug_log, "%*s %4"UVXf" :",
919 (int)depth * 2 + 2,"", (UV)state );
920 if ( ! trie->states[ state ].wordnum ) {
921 PerlIO_printf( Perl_debug_log, "%5s| ","");
923 PerlIO_printf( Perl_debug_log, "W%4x| ",
924 trie->states[ state ].wordnum
927 for( charid = 1 ; charid <= TRIE_LIST_USED( state ) ; charid++ ) {
928 SV ** const tmp = av_fetch( trie->revcharmap, TRIE_LIST_ITEM(state,charid).forid, 0);
930 PerlIO_printf( Perl_debug_log, "%*s:%3X=%4"UVXf" | ",
932 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
933 PL_colors[0], PL_colors[1],
934 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
935 PERL_PV_ESCAPE_FIRSTCHAR
937 TRIE_LIST_ITEM(state,charid).forid,
938 (UV)TRIE_LIST_ITEM(state,charid).newstate
941 PerlIO_printf(Perl_debug_log, "\n%*s| ",
942 (int)((depth * 2) + 14), "");
945 PerlIO_printf( Perl_debug_log, "\n");
950 dump_trie_interim_table(trie,next_alloc)
951 Dumps a fully constructed but uncompressed trie in table form.
952 This is the normal DFA style state transition table, with a few
953 twists to facilitate compression later.
954 Used for debugging make_trie().
957 S_dump_trie_interim_table(pTHX_ const struct _reg_trie_data *trie, U32 next_alloc, U32 depth)
961 SV *sv=sv_newmortal();
962 int colwidth= trie->widecharmap ? 6 : 4;
963 GET_RE_DEBUG_FLAGS_DECL;
966 print out the table precompression so that we can do a visual check
967 that they are identical.
970 PerlIO_printf( Perl_debug_log, "%*sChar : ",(int)depth * 2 + 2,"" );
972 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
973 SV ** const tmp = av_fetch( trie->revcharmap, charid, 0);
975 PerlIO_printf( Perl_debug_log, "%*s",
977 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
978 PL_colors[0], PL_colors[1],
979 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
980 PERL_PV_ESCAPE_FIRSTCHAR
986 PerlIO_printf( Perl_debug_log, "\n%*sState+-",(int)depth * 2 + 2,"" );
988 for( charid=0 ; charid < trie->uniquecharcount ; charid++ ) {
989 PerlIO_printf( Perl_debug_log, "%.*s", colwidth,"--------");
992 PerlIO_printf( Perl_debug_log, "\n" );
994 for( state=1 ; state < next_alloc ; state += trie->uniquecharcount ) {
996 PerlIO_printf( Perl_debug_log, "%*s%4"UVXf" : ",
997 (int)depth * 2 + 2,"",
998 (UV)TRIE_NODENUM( state ) );
1000 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
1001 UV v=(UV)SAFE_TRIE_NODENUM( trie->trans[ state + charid ].next );
1003 PerlIO_printf( Perl_debug_log, "%*"UVXf, colwidth, v );
1005 PerlIO_printf( Perl_debug_log, "%*s", colwidth, "." );
1007 if ( ! trie->states[ TRIE_NODENUM( state ) ].wordnum ) {
1008 PerlIO_printf( Perl_debug_log, " (%4"UVXf")\n", (UV)trie->trans[ state ].check );
1010 PerlIO_printf( Perl_debug_log, " (%4"UVXf") W%4X\n", (UV)trie->trans[ state ].check,
1011 trie->states[ TRIE_NODENUM( state ) ].wordnum );
1018 /* make_trie(startbranch,first,last,tail,word_count,flags,depth)
1019 startbranch: the first branch in the whole branch sequence
1020 first : start branch of sequence of branch-exact nodes.
1021 May be the same as startbranch
1022 last : Thing following the last branch.
1023 May be the same as tail.
1024 tail : item following the branch sequence
1025 count : words in the sequence
1026 flags : currently the OP() type we will be building one of /EXACT(|F|Fl)/
1027 depth : indent depth
1029 Inplace optimizes a sequence of 2 or more Branch-Exact nodes into a TRIE node.
1031 A trie is an N'ary tree where the branches are determined by digital
1032 decomposition of the key. IE, at the root node you look up the 1st character and
1033 follow that branch repeat until you find the end of the branches. Nodes can be
1034 marked as "accepting" meaning they represent a complete word. Eg:
1038 would convert into the following structure. Numbers represent states, letters
1039 following numbers represent valid transitions on the letter from that state, if
1040 the number is in square brackets it represents an accepting state, otherwise it
1041 will be in parenthesis.
1043 +-h->+-e->[3]-+-r->(8)-+-s->[9]
1047 (1) +-i->(6)-+-s->[7]
1049 +-s->(3)-+-h->(4)-+-e->[5]
1051 Accept Word Mapping: 3=>1 (he),5=>2 (she), 7=>3 (his), 9=>4 (hers)
1053 This shows that when matching against the string 'hers' we will begin at state 1
1054 read 'h' and move to state 2, read 'e' and move to state 3 which is accepting,
1055 then read 'r' and go to state 8 followed by 's' which takes us to state 9 which
1056 is also accepting. Thus we know that we can match both 'he' and 'hers' with a
1057 single traverse. We store a mapping from accepting to state to which word was
1058 matched, and then when we have multiple possibilities we try to complete the
1059 rest of the regex in the order in which they occured in the alternation.
1061 The only prior NFA like behaviour that would be changed by the TRIE support is
1062 the silent ignoring of duplicate alternations which are of the form:
1064 / (DUPE|DUPE) X? (?{ ... }) Y /x
1066 Thus EVAL blocks follwing a trie may be called a different number of times with
1067 and without the optimisation. With the optimisations dupes will be silently
1068 ignored. This inconsistant behaviour of EVAL type nodes is well established as
1069 the following demonstrates:
1071 'words'=~/(word|word|word)(?{ print $1 })[xyz]/
1073 which prints out 'word' three times, but
1075 'words'=~/(word|word|word)(?{ print $1 })S/
1077 which doesnt print it out at all. This is due to other optimisations kicking in.
1079 Example of what happens on a structural level:
1081 The regexp /(ac|ad|ab)+/ will produce the folowing debug output:
1083 1: CURLYM[1] {1,32767}(18)
1094 This would be optimizable with startbranch=5, first=5, last=16, tail=16
1095 and should turn into:
1097 1: CURLYM[1] {1,32767}(18)
1099 [Words:3 Chars Stored:6 Unique Chars:4 States:5 NCP:1]
1107 Cases where tail != last would be like /(?foo|bar)baz/:
1117 which would be optimizable with startbranch=1, first=1, last=7, tail=8
1118 and would end up looking like:
1121 [Words:2 Chars Stored:6 Unique Chars:5 States:7 NCP:1]
1128 d = uvuni_to_utf8_flags(d, uv, 0);
1130 is the recommended Unicode-aware way of saying
1135 #define TRIE_STORE_REVCHAR \
1137 SV *tmp = newSVpvs(""); \
1138 if (UTF) SvUTF8_on(tmp); \
1139 Perl_sv_catpvf( aTHX_ tmp, "%c", (int)uvc ); \
1140 av_push( TRIE_REVCHARMAP(trie), tmp ); \
1143 #define TRIE_READ_CHAR STMT_START { \
1147 if ( foldlen > 0 ) { \
1148 uvc = utf8n_to_uvuni( scan, UTF8_MAXLEN, &len, uniflags ); \
1153 uvc = utf8n_to_uvuni( (const U8*)uc, UTF8_MAXLEN, &len, uniflags);\
1154 uvc = to_uni_fold( uvc, foldbuf, &foldlen ); \
1155 foldlen -= UNISKIP( uvc ); \
1156 scan = foldbuf + UNISKIP( uvc ); \
1159 uvc = utf8n_to_uvuni( (const U8*)uc, UTF8_MAXLEN, &len, uniflags);\
1169 #define TRIE_LIST_PUSH(state,fid,ns) STMT_START { \
1170 if ( TRIE_LIST_CUR( state ) >=TRIE_LIST_LEN( state ) ) { \
1171 U32 ging = TRIE_LIST_LEN( state ) *= 2; \
1172 Renew( trie->states[ state ].trans.list, ging, reg_trie_trans_le ); \
1174 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).forid = fid; \
1175 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).newstate = ns; \
1176 TRIE_LIST_CUR( state )++; \
1179 #define TRIE_LIST_NEW(state) STMT_START { \
1180 Newxz( trie->states[ state ].trans.list, \
1181 4, reg_trie_trans_le ); \
1182 TRIE_LIST_CUR( state ) = 1; \
1183 TRIE_LIST_LEN( state ) = 4; \
1186 #define TRIE_HANDLE_WORD(state) STMT_START { \
1187 U16 dupe= trie->states[ state ].wordnum; \
1188 regnode * const noper_next = regnext( noper ); \
1190 if (trie->wordlen) \
1191 trie->wordlen[ curword ] = wordlen; \
1193 /* store the word for dumping */ \
1195 if (OP(noper) != NOTHING) \
1196 tmp = newSVpvn(STRING(noper), STR_LEN(noper)); \
1198 tmp = newSVpvn( "", 0 ); \
1199 if ( UTF ) SvUTF8_on( tmp ); \
1200 av_push( trie->words, tmp ); \
1205 if ( noper_next < tail ) { \
1207 trie->jump = PerlMemShared_calloc( word_count + 1, sizeof(U16) ); \
1208 trie->jump[curword] = (U16)(noper_next - convert); \
1210 jumper = noper_next; \
1212 nextbranch= regnext(cur); \
1216 /* So it's a dupe. This means we need to maintain a */\
1217 /* linked-list from the first to the next. */\
1218 /* we only allocate the nextword buffer when there */\
1219 /* a dupe, so first time we have to do the allocation */\
1220 if (!trie->nextword) \
1222 PerlMemShared_calloc( word_count + 1, sizeof(U16)); \
1223 while ( trie->nextword[dupe] ) \
1224 dupe= trie->nextword[dupe]; \
1225 trie->nextword[dupe]= curword; \
1227 /* we haven't inserted this word yet. */ \
1228 trie->states[ state ].wordnum = curword; \
1233 #define TRIE_TRANS_STATE(state,base,ucharcount,charid,special) \
1234 ( ( base + charid >= ucharcount \
1235 && base + charid < ubound \
1236 && state == trie->trans[ base - ucharcount + charid ].check \
1237 && trie->trans[ base - ucharcount + charid ].next ) \
1238 ? trie->trans[ base - ucharcount + charid ].next \
1239 : ( state==1 ? special : 0 ) \
1243 #define MADE_JUMP_TRIE 2
1244 #define MADE_EXACT_TRIE 4
1247 S_make_trie(pTHX_ RExC_state_t *pRExC_state, regnode *startbranch, regnode *first, regnode *last, regnode *tail, U32 word_count, U32 flags, U32 depth)
1250 /* first pass, loop through and scan words */
1251 reg_trie_data *trie;
1253 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1258 regnode *jumper = NULL;
1259 regnode *nextbranch = NULL;
1260 regnode *convert = NULL;
1261 /* we just use folder as a flag in utf8 */
1262 const U8 * const folder = ( flags == EXACTF
1264 : ( flags == EXACTFL
1270 const U32 data_slot = add_data( pRExC_state, 1, "t" );
1271 SV *re_trie_maxbuff;
1273 /* these are only used during construction but are useful during
1274 * debugging so we store them in the struct when debugging.
1276 STRLEN trie_charcount=0;
1277 AV *trie_revcharmap;
1279 GET_RE_DEBUG_FLAGS_DECL;
1281 PERL_UNUSED_ARG(depth);
1284 trie = PerlMemShared_calloc( 1, sizeof(reg_trie_data) );
1286 trie->startstate = 1;
1287 trie->wordcount = word_count;
1288 RExC_rxi->data->data[ data_slot ] = (void*)trie;
1289 trie->charmap = PerlMemShared_calloc( 256, sizeof(U16) );
1290 if (!(UTF && folder))
1291 trie->bitmap = PerlMemShared_calloc( ANYOF_BITMAP_SIZE, 1 );
1293 trie->words = newAV();
1295 TRIE_REVCHARMAP(trie) = newAV();
1297 re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1);
1298 if (!SvIOK(re_trie_maxbuff)) {
1299 sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
1302 PerlIO_printf( Perl_debug_log,
1303 "%*smake_trie start==%d, first==%d, last==%d, tail==%d depth=%d\n",
1304 (int)depth * 2 + 2, "",
1305 REG_NODE_NUM(startbranch),REG_NODE_NUM(first),
1306 REG_NODE_NUM(last), REG_NODE_NUM(tail),
1310 /* Find the node we are going to overwrite */
1311 if ( first == startbranch && OP( last ) != BRANCH ) {
1312 /* whole branch chain */
1315 /* branch sub-chain */
1316 convert = NEXTOPER( first );
1319 /* -- First loop and Setup --
1321 We first traverse the branches and scan each word to determine if it
1322 contains widechars, and how many unique chars there are, this is
1323 important as we have to build a table with at least as many columns as we
1326 We use an array of integers to represent the character codes 0..255
1327 (trie->charmap) and we use a an HV* to store unicode characters. We use the
1328 native representation of the character value as the key and IV's for the
1331 *TODO* If we keep track of how many times each character is used we can
1332 remap the columns so that the table compression later on is more
1333 efficient in terms of memory by ensuring most common value is in the
1334 middle and the least common are on the outside. IMO this would be better
1335 than a most to least common mapping as theres a decent chance the most
1336 common letter will share a node with the least common, meaning the node
1337 will not be compressable. With a middle is most common approach the worst
1338 case is when we have the least common nodes twice.
1342 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
1343 regnode * const noper = NEXTOPER( cur );
1344 const U8 *uc = (U8*)STRING( noper );
1345 const U8 * const e = uc + STR_LEN( noper );
1347 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1348 const U8 *scan = (U8*)NULL;
1349 U32 wordlen = 0; /* required init */
1352 if (OP(noper) == NOTHING) {
1357 TRIE_BITMAP_SET(trie,*uc);
1358 if ( folder ) TRIE_BITMAP_SET(trie,folder[ *uc ]);
1360 for ( ; uc < e ; uc += len ) {
1361 TRIE_CHARCOUNT(trie)++;
1365 if ( !trie->charmap[ uvc ] ) {
1366 trie->charmap[ uvc ]=( ++trie->uniquecharcount );
1368 trie->charmap[ folder[ uvc ] ] = trie->charmap[ uvc ];
1373 if ( !trie->widecharmap )
1374 trie->widecharmap = newHV();
1376 svpp = hv_fetch( trie->widecharmap, (char*)&uvc, sizeof( UV ), 1 );
1379 Perl_croak( aTHX_ "error creating/fetching widecharmap entry for 0x%"UVXf, uvc );
1381 if ( !SvTRUE( *svpp ) ) {
1382 sv_setiv( *svpp, ++trie->uniquecharcount );
1387 if( cur == first ) {
1390 } else if (chars < trie->minlen) {
1392 } else if (chars > trie->maxlen) {
1396 } /* end first pass */
1397 DEBUG_TRIE_COMPILE_r(
1398 PerlIO_printf( Perl_debug_log, "%*sTRIE(%s): W:%d C:%d Uq:%d Min:%d Max:%d\n",
1399 (int)depth * 2 + 2,"",
1400 ( trie->widecharmap ? "UTF8" : "NATIVE" ), (int)word_count,
1401 (int)TRIE_CHARCOUNT(trie), trie->uniquecharcount,
1402 (int)trie->minlen, (int)trie->maxlen )
1404 trie->wordlen = PerlMemShared_calloc( word_count, sizeof(U32) );
1407 We now know what we are dealing with in terms of unique chars and
1408 string sizes so we can calculate how much memory a naive
1409 representation using a flat table will take. If it's over a reasonable
1410 limit (as specified by ${^RE_TRIE_MAXBUF}) we use a more memory
1411 conservative but potentially much slower representation using an array
1414 At the end we convert both representations into the same compressed
1415 form that will be used in regexec.c for matching with. The latter
1416 is a form that cannot be used to construct with but has memory
1417 properties similar to the list form and access properties similar
1418 to the table form making it both suitable for fast searches and
1419 small enough that its feasable to store for the duration of a program.
1421 See the comment in the code where the compressed table is produced
1422 inplace from the flat tabe representation for an explanation of how
1423 the compression works.
1428 if ( (IV)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1) > SvIV(re_trie_maxbuff) ) {
1430 Second Pass -- Array Of Lists Representation
1432 Each state will be represented by a list of charid:state records
1433 (reg_trie_trans_le) the first such element holds the CUR and LEN
1434 points of the allocated array. (See defines above).
1436 We build the initial structure using the lists, and then convert
1437 it into the compressed table form which allows faster lookups
1438 (but cant be modified once converted).
1441 STRLEN transcount = 1;
1443 DEBUG_TRIE_COMPILE_MORE_r( PerlIO_printf( Perl_debug_log,
1444 "%*sCompiling trie using list compiler\n",
1445 (int)depth * 2 + 2, ""));
1447 trie->states = PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
1448 sizeof(reg_trie_state) );
1452 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
1454 regnode * const noper = NEXTOPER( cur );
1455 U8 *uc = (U8*)STRING( noper );
1456 const U8 * const e = uc + STR_LEN( noper );
1457 U32 state = 1; /* required init */
1458 U16 charid = 0; /* sanity init */
1459 U8 *scan = (U8*)NULL; /* sanity init */
1460 STRLEN foldlen = 0; /* required init */
1461 U32 wordlen = 0; /* required init */
1462 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1464 if (OP(noper) != NOTHING) {
1465 for ( ; uc < e ; uc += len ) {
1470 charid = trie->charmap[ uvc ];
1472 SV** const svpp = hv_fetch( trie->widecharmap, (char*)&uvc, sizeof( UV ), 0);
1476 charid=(U16)SvIV( *svpp );
1479 /* charid is now 0 if we dont know the char read, or nonzero if we do */
1486 if ( !trie->states[ state ].trans.list ) {
1487 TRIE_LIST_NEW( state );
1489 for ( check = 1; check <= TRIE_LIST_USED( state ); check++ ) {
1490 if ( TRIE_LIST_ITEM( state, check ).forid == charid ) {
1491 newstate = TRIE_LIST_ITEM( state, check ).newstate;
1496 newstate = next_alloc++;
1497 TRIE_LIST_PUSH( state, charid, newstate );
1502 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %"IVdf, uvc );
1506 TRIE_HANDLE_WORD(state);
1508 } /* end second pass */
1510 /* next alloc is the NEXT state to be allocated */
1511 trie->statecount = next_alloc;
1512 trie->states = PerlMemShared_realloc( trie->states, next_alloc
1513 * sizeof(reg_trie_state) );
1515 /* and now dump it out before we compress it */
1516 DEBUG_TRIE_COMPILE_MORE_r(
1517 dump_trie_interim_list(trie,next_alloc,depth+1)
1521 = PerlMemShared_calloc( transcount, sizeof(reg_trie_trans) );
1528 for( state=1 ; state < next_alloc ; state ++ ) {
1532 DEBUG_TRIE_COMPILE_MORE_r(
1533 PerlIO_printf( Perl_debug_log, "tp: %d zp: %d ",tp,zp)
1537 if (trie->states[state].trans.list) {
1538 U16 minid=TRIE_LIST_ITEM( state, 1).forid;
1542 for( idx = 2 ; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
1543 const U16 forid = TRIE_LIST_ITEM( state, idx).forid;
1544 if ( forid < minid ) {
1546 } else if ( forid > maxid ) {
1550 if ( transcount < tp + maxid - minid + 1) {
1553 = PerlMemShared_realloc( trie->trans,
1555 * sizeof(reg_trie_trans) );
1556 Zero( trie->trans + (transcount / 2), transcount / 2 , reg_trie_trans );
1558 base = trie->uniquecharcount + tp - minid;
1559 if ( maxid == minid ) {
1561 for ( ; zp < tp ; zp++ ) {
1562 if ( ! trie->trans[ zp ].next ) {
1563 base = trie->uniquecharcount + zp - minid;
1564 trie->trans[ zp ].next = TRIE_LIST_ITEM( state, 1).newstate;
1565 trie->trans[ zp ].check = state;
1571 trie->trans[ tp ].next = TRIE_LIST_ITEM( state, 1).newstate;
1572 trie->trans[ tp ].check = state;
1577 for ( idx=1; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
1578 const U32 tid = base - trie->uniquecharcount + TRIE_LIST_ITEM( state, idx ).forid;
1579 trie->trans[ tid ].next = TRIE_LIST_ITEM( state, idx ).newstate;
1580 trie->trans[ tid ].check = state;
1582 tp += ( maxid - minid + 1 );
1584 Safefree(trie->states[ state ].trans.list);
1587 DEBUG_TRIE_COMPILE_MORE_r(
1588 PerlIO_printf( Perl_debug_log, " base: %d\n",base);
1591 trie->states[ state ].trans.base=base;
1593 trie->lasttrans = tp + 1;
1597 Second Pass -- Flat Table Representation.
1599 we dont use the 0 slot of either trans[] or states[] so we add 1 to each.
1600 We know that we will need Charcount+1 trans at most to store the data
1601 (one row per char at worst case) So we preallocate both structures
1602 assuming worst case.
1604 We then construct the trie using only the .next slots of the entry
1607 We use the .check field of the first entry of the node temporarily to
1608 make compression both faster and easier by keeping track of how many non
1609 zero fields are in the node.
1611 Since trans are numbered from 1 any 0 pointer in the table is a FAIL
1614 There are two terms at use here: state as a TRIE_NODEIDX() which is a
1615 number representing the first entry of the node, and state as a
1616 TRIE_NODENUM() which is the trans number. state 1 is TRIE_NODEIDX(1) and
1617 TRIE_NODENUM(1), state 2 is TRIE_NODEIDX(2) and TRIE_NODENUM(3) if there
1618 are 2 entrys per node. eg:
1626 The table is internally in the right hand, idx form. However as we also
1627 have to deal with the states array which is indexed by nodenum we have to
1628 use TRIE_NODENUM() to convert.
1631 DEBUG_TRIE_COMPILE_MORE_r( PerlIO_printf( Perl_debug_log,
1632 "%*sCompiling trie using table compiler\n",
1633 (int)depth * 2 + 2, ""));
1635 trie->trans = PerlMemShared_calloc( ( TRIE_CHARCOUNT(trie) + 1 )
1636 * trie->uniquecharcount + 1,
1637 sizeof(reg_trie_trans) );
1638 trie->states = PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
1639 sizeof(reg_trie_state) );
1640 next_alloc = trie->uniquecharcount + 1;
1643 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
1645 regnode * const noper = NEXTOPER( cur );
1646 const U8 *uc = (U8*)STRING( noper );
1647 const U8 * const e = uc + STR_LEN( noper );
1649 U32 state = 1; /* required init */
1651 U16 charid = 0; /* sanity init */
1652 U32 accept_state = 0; /* sanity init */
1653 U8 *scan = (U8*)NULL; /* sanity init */
1655 STRLEN foldlen = 0; /* required init */
1656 U32 wordlen = 0; /* required init */
1657 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1659 if ( OP(noper) != NOTHING ) {
1660 for ( ; uc < e ; uc += len ) {
1665 charid = trie->charmap[ uvc ];
1667 SV* const * const svpp = hv_fetch( trie->widecharmap, (char*)&uvc, sizeof( UV ), 0);
1668 charid = svpp ? (U16)SvIV(*svpp) : 0;
1672 if ( !trie->trans[ state + charid ].next ) {
1673 trie->trans[ state + charid ].next = next_alloc;
1674 trie->trans[ state ].check++;
1675 next_alloc += trie->uniquecharcount;
1677 state = trie->trans[ state + charid ].next;
1679 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %"IVdf, uvc );
1681 /* charid is now 0 if we dont know the char read, or nonzero if we do */
1684 accept_state = TRIE_NODENUM( state );
1685 TRIE_HANDLE_WORD(accept_state);
1687 } /* end second pass */
1689 /* and now dump it out before we compress it */
1690 DEBUG_TRIE_COMPILE_MORE_r(
1691 dump_trie_interim_table(trie,next_alloc,depth+1)
1696 * Inplace compress the table.*
1698 For sparse data sets the table constructed by the trie algorithm will
1699 be mostly 0/FAIL transitions or to put it another way mostly empty.
1700 (Note that leaf nodes will not contain any transitions.)
1702 This algorithm compresses the tables by eliminating most such
1703 transitions, at the cost of a modest bit of extra work during lookup:
1705 - Each states[] entry contains a .base field which indicates the
1706 index in the state[] array wheres its transition data is stored.
1708 - If .base is 0 there are no valid transitions from that node.
1710 - If .base is nonzero then charid is added to it to find an entry in
1713 -If trans[states[state].base+charid].check!=state then the
1714 transition is taken to be a 0/Fail transition. Thus if there are fail
1715 transitions at the front of the node then the .base offset will point
1716 somewhere inside the previous nodes data (or maybe even into a node
1717 even earlier), but the .check field determines if the transition is
1721 The following process inplace converts the table to the compressed
1722 table: We first do not compress the root node 1,and mark its all its
1723 .check pointers as 1 and set its .base pointer as 1 as well. This
1724 allows to do a DFA construction from the compressed table later, and
1725 ensures that any .base pointers we calculate later are greater than
1728 - We set 'pos' to indicate the first entry of the second node.
1730 - We then iterate over the columns of the node, finding the first and
1731 last used entry at l and m. We then copy l..m into pos..(pos+m-l),
1732 and set the .check pointers accordingly, and advance pos
1733 appropriately and repreat for the next node. Note that when we copy
1734 the next pointers we have to convert them from the original
1735 NODEIDX form to NODENUM form as the former is not valid post
1738 - If a node has no transitions used we mark its base as 0 and do not
1739 advance the pos pointer.
1741 - If a node only has one transition we use a second pointer into the
1742 structure to fill in allocated fail transitions from other states.
1743 This pointer is independent of the main pointer and scans forward
1744 looking for null transitions that are allocated to a state. When it
1745 finds one it writes the single transition into the "hole". If the
1746 pointer doesnt find one the single transition is appended as normal.
1748 - Once compressed we can Renew/realloc the structures to release the
1751 See "Table-Compression Methods" in sec 3.9 of the Red Dragon,
1752 specifically Fig 3.47 and the associated pseudocode.
1756 const U32 laststate = TRIE_NODENUM( next_alloc );
1759 trie->statecount = laststate;
1761 for ( state = 1 ; state < laststate ; state++ ) {
1763 const U32 stateidx = TRIE_NODEIDX( state );
1764 const U32 o_used = trie->trans[ stateidx ].check;
1765 U32 used = trie->trans[ stateidx ].check;
1766 trie->trans[ stateidx ].check = 0;
1768 for ( charid = 0 ; used && charid < trie->uniquecharcount ; charid++ ) {
1769 if ( flag || trie->trans[ stateidx + charid ].next ) {
1770 if ( trie->trans[ stateidx + charid ].next ) {
1772 for ( ; zp < pos ; zp++ ) {
1773 if ( ! trie->trans[ zp ].next ) {
1777 trie->states[ state ].trans.base = zp + trie->uniquecharcount - charid ;
1778 trie->trans[ zp ].next = SAFE_TRIE_NODENUM( trie->trans[ stateidx + charid ].next );
1779 trie->trans[ zp ].check = state;
1780 if ( ++zp > pos ) pos = zp;
1787 trie->states[ state ].trans.base = pos + trie->uniquecharcount - charid ;
1789 trie->trans[ pos ].next = SAFE_TRIE_NODENUM( trie->trans[ stateidx + charid ].next );
1790 trie->trans[ pos ].check = state;
1795 trie->lasttrans = pos + 1;
1796 trie->states = PerlMemShared_realloc( trie->states, laststate
1797 * sizeof(reg_trie_state) );
1798 DEBUG_TRIE_COMPILE_MORE_r(
1799 PerlIO_printf( Perl_debug_log,
1800 "%*sAlloc: %d Orig: %"IVdf" elements, Final:%"IVdf". Savings of %%%5.2f\n",
1801 (int)depth * 2 + 2,"",
1802 (int)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1 ),
1805 ( ( next_alloc - pos ) * 100 ) / (double)next_alloc );
1808 } /* end table compress */
1810 DEBUG_TRIE_COMPILE_MORE_r(
1811 PerlIO_printf(Perl_debug_log, "%*sStatecount:%"UVxf" Lasttrans:%"UVxf"\n",
1812 (int)depth * 2 + 2, "",
1813 (UV)trie->statecount,
1814 (UV)trie->lasttrans)
1816 /* resize the trans array to remove unused space */
1817 trie->trans = PerlMemShared_realloc( trie->trans, trie->lasttrans
1818 * sizeof(reg_trie_trans) );
1820 /* and now dump out the compressed format */
1821 DEBUG_TRIE_COMPILE_r(
1822 dump_trie(trie,depth+1)
1825 { /* Modify the program and insert the new TRIE node*/
1826 U8 nodetype =(U8)(flags & 0xFF);
1830 regnode *optimize = NULL;
1832 U32 mjd_nodelen = 0;
1835 This means we convert either the first branch or the first Exact,
1836 depending on whether the thing following (in 'last') is a branch
1837 or not and whther first is the startbranch (ie is it a sub part of
1838 the alternation or is it the whole thing.)
1839 Assuming its a sub part we conver the EXACT otherwise we convert
1840 the whole branch sequence, including the first.
1842 /* Find the node we are going to overwrite */
1843 if ( first != startbranch || OP( last ) == BRANCH ) {
1844 /* branch sub-chain */
1845 NEXT_OFF( first ) = (U16)(last - first);
1847 mjd_offset= Node_Offset((convert));
1848 mjd_nodelen= Node_Length((convert));
1850 /* whole branch chain */
1853 const regnode *nop = NEXTOPER( convert );
1854 mjd_offset= Node_Offset((nop));
1855 mjd_nodelen= Node_Length((nop));
1860 PerlIO_printf(Perl_debug_log, "%*sMJD offset:%"UVuf" MJD length:%"UVuf"\n",
1861 (int)depth * 2 + 2, "",
1862 (UV)mjd_offset, (UV)mjd_nodelen)
1865 /* But first we check to see if there is a common prefix we can
1866 split out as an EXACT and put in front of the TRIE node. */
1867 trie->startstate= 1;
1868 if ( trie->bitmap && !trie->widecharmap && !trie->jump ) {
1870 for ( state = 1 ; state < trie->statecount-1 ; state++ ) {
1874 const U32 base = trie->states[ state ].trans.base;
1876 if ( trie->states[state].wordnum )
1879 for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) {
1880 if ( ( base + ofs >= trie->uniquecharcount ) &&
1881 ( base + ofs - trie->uniquecharcount < trie->lasttrans ) &&
1882 trie->trans[ base + ofs - trie->uniquecharcount ].check == state )
1884 if ( ++count > 1 ) {
1885 SV **tmp = av_fetch( TRIE_REVCHARMAP(trie), ofs, 0);
1886 const U8 *ch = (U8*)SvPV_nolen_const( *tmp );
1887 if ( state == 1 ) break;
1889 Zero(trie->bitmap, ANYOF_BITMAP_SIZE, char);
1891 PerlIO_printf(Perl_debug_log,
1892 "%*sNew Start State=%"UVuf" Class: [",
1893 (int)depth * 2 + 2, "",
1896 SV ** const tmp = av_fetch( TRIE_REVCHARMAP(trie), idx, 0);
1897 const U8 * const ch = (U8*)SvPV_nolen_const( *tmp );
1899 TRIE_BITMAP_SET(trie,*ch);
1901 TRIE_BITMAP_SET(trie, folder[ *ch ]);
1903 PerlIO_printf(Perl_debug_log, (char*)ch)
1907 TRIE_BITMAP_SET(trie,*ch);
1909 TRIE_BITMAP_SET(trie,folder[ *ch ]);
1910 DEBUG_OPTIMISE_r(PerlIO_printf( Perl_debug_log,"%s", ch));
1916 SV **tmp = av_fetch( TRIE_REVCHARMAP(trie), idx, 0);
1917 char *ch = SvPV_nolen( *tmp );
1919 SV *sv=sv_newmortal();
1920 PerlIO_printf( Perl_debug_log,
1921 "%*sPrefix State: %"UVuf" Idx:%"UVuf" Char='%s'\n",
1922 (int)depth * 2 + 2, "",
1924 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 6,
1925 PL_colors[0], PL_colors[1],
1926 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
1927 PERL_PV_ESCAPE_FIRSTCHAR
1932 OP( convert ) = nodetype;
1933 str=STRING(convert);
1944 DEBUG_OPTIMISE_r(PerlIO_printf( Perl_debug_log,"]\n"));
1950 regnode *n = convert+NODE_SZ_STR(convert);
1951 NEXT_OFF(convert) = NODE_SZ_STR(convert);
1952 trie->startstate = state;
1953 trie->minlen -= (state - 1);
1954 trie->maxlen -= (state - 1);
1956 regnode *fix = convert;
1957 U32 word = trie->wordcount;
1959 Set_Node_Offset_Length(convert, mjd_offset, state - 1);
1960 while( ++fix < n ) {
1961 Set_Node_Offset_Length(fix, 0, 0);
1964 SV ** const tmp = av_fetch( trie->words, word, 0 );
1966 if ( STR_LEN(convert) <= SvCUR(*tmp) )
1967 sv_chop(*tmp, SvPV_nolen(*tmp) + STR_LEN(convert));
1969 sv_chop(*tmp, SvPV_nolen(*tmp) + SvCUR(*tmp));
1976 NEXT_OFF(convert) = (U16)(tail - convert);
1977 DEBUG_r(optimize= n);
1983 if ( trie->maxlen ) {
1984 NEXT_OFF( convert ) = (U16)(tail - convert);
1985 ARG_SET( convert, data_slot );
1986 /* Store the offset to the first unabsorbed branch in
1987 jump[0], which is otherwise unused by the jump logic.
1988 We use this when dumping a trie and during optimisation. */
1990 trie->jump[0] = (U16)(nextbranch - convert);
1993 if ( !trie->states[trie->startstate].wordnum && trie->bitmap &&
1994 ( (char *)jumper - (char *)convert) >= (int)sizeof(struct regnode_charclass) )
1996 OP( convert ) = TRIEC;
1997 Copy(trie->bitmap, ((struct regnode_charclass *)convert)->bitmap, ANYOF_BITMAP_SIZE, char);
1998 PerlMemShared_free(trie->bitmap);
2001 OP( convert ) = TRIE;
2003 /* store the type in the flags */
2004 convert->flags = nodetype;
2008 + regarglen[ OP( convert ) ];
2010 /* XXX We really should free up the resource in trie now,
2011 as we won't use them - (which resources?) dmq */
2013 /* needed for dumping*/
2014 DEBUG_r(if (optimize) {
2015 regnode *opt = convert;
2016 while ( ++opt < optimize) {
2017 Set_Node_Offset_Length(opt,0,0);
2020 Try to clean up some of the debris left after the
2023 while( optimize < jumper ) {
2024 mjd_nodelen += Node_Length((optimize));
2025 OP( optimize ) = OPTIMIZED;
2026 Set_Node_Offset_Length(optimize,0,0);
2029 Set_Node_Offset_Length(convert,mjd_offset,mjd_nodelen);
2031 } /* end node insert */
2033 SvREFCNT_dec(TRIE_REVCHARMAP(trie));
2037 : trie->startstate>1
2043 S_make_trie_failtable(pTHX_ RExC_state_t *pRExC_state, regnode *source, regnode *stclass, U32 depth)
2045 /* The Trie is constructed and compressed now so we can build a fail array now if its needed
2047 This is basically the Aho-Corasick algorithm. Its from exercise 3.31 and 3.32 in the
2048 "Red Dragon" -- Compilers, principles, techniques, and tools. Aho, Sethi, Ullman 1985/88
2051 We find the fail state for each state in the trie, this state is the longest proper
2052 suffix of the current states 'word' that is also a proper prefix of another word in our
2053 trie. State 1 represents the word '' and is the thus the default fail state. This allows
2054 the DFA not to have to restart after its tried and failed a word at a given point, it
2055 simply continues as though it had been matching the other word in the first place.
2057 'abcdgu'=~/abcdefg|cdgu/
2058 When we get to 'd' we are still matching the first word, we would encounter 'g' which would
2059 fail, which would bring use to the state representing 'd' in the second word where we would
2060 try 'g' and succeed, prodceding to match 'cdgu'.
2062 /* add a fail transition */
2063 reg_trie_data *trie=(reg_trie_data *)RExC_rxi->data->data[ARG(source)];
2065 const U32 ucharcount = trie->uniquecharcount;
2066 const U32 numstates = trie->statecount;
2067 const U32 ubound = trie->lasttrans + ucharcount;
2071 U32 base = trie->states[ 1 ].trans.base;
2074 const U32 data_slot = add_data( pRExC_state, 1, "T" );
2075 GET_RE_DEBUG_FLAGS_DECL;
2077 PERL_UNUSED_ARG(depth);
2081 ARG_SET( stclass, data_slot );
2082 aho = PerlMemShared_calloc( 1, sizeof(reg_ac_data) );
2083 RExC_rxi->data->data[ data_slot ] = (void*)aho;
2085 aho->states=(reg_trie_state *)PerlMemShared_malloc( numstates * sizeof(reg_trie_state) );
2086 Copy( trie->states, aho->states, numstates, reg_trie_state );
2087 Newxz( q, numstates, U32);
2088 aho->fail = PerlMemShared_calloc( numstates, sizeof(U32) );
2091 /* initialize fail[0..1] to be 1 so that we always have
2092 a valid final fail state */
2093 fail[ 0 ] = fail[ 1 ] = 1;
2095 for ( charid = 0; charid < ucharcount ; charid++ ) {
2096 const U32 newstate = TRIE_TRANS_STATE( 1, base, ucharcount, charid, 0 );
2098 q[ q_write ] = newstate;
2099 /* set to point at the root */
2100 fail[ q[ q_write++ ] ]=1;
2103 while ( q_read < q_write) {
2104 const U32 cur = q[ q_read++ % numstates ];
2105 base = trie->states[ cur ].trans.base;
2107 for ( charid = 0 ; charid < ucharcount ; charid++ ) {
2108 const U32 ch_state = TRIE_TRANS_STATE( cur, base, ucharcount, charid, 1 );
2110 U32 fail_state = cur;
2113 fail_state = fail[ fail_state ];
2114 fail_base = aho->states[ fail_state ].trans.base;
2115 } while ( !TRIE_TRANS_STATE( fail_state, fail_base, ucharcount, charid, 1 ) );
2117 fail_state = TRIE_TRANS_STATE( fail_state, fail_base, ucharcount, charid, 1 );
2118 fail[ ch_state ] = fail_state;
2119 if ( !aho->states[ ch_state ].wordnum && aho->states[ fail_state ].wordnum )
2121 aho->states[ ch_state ].wordnum = aho->states[ fail_state ].wordnum;
2123 q[ q_write++ % numstates] = ch_state;
2127 /* restore fail[0..1] to 0 so that we "fall out" of the AC loop
2128 when we fail in state 1, this allows us to use the
2129 charclass scan to find a valid start char. This is based on the principle
2130 that theres a good chance the string being searched contains lots of stuff
2131 that cant be a start char.
2133 fail[ 0 ] = fail[ 1 ] = 0;
2134 DEBUG_TRIE_COMPILE_r({
2135 PerlIO_printf(Perl_debug_log,
2136 "%*sStclass Failtable (%"UVuf" states): 0",
2137 (int)(depth * 2), "", (UV)numstates
2139 for( q_read=1; q_read<numstates; q_read++ ) {
2140 PerlIO_printf(Perl_debug_log, ", %"UVuf, (UV)fail[q_read]);
2142 PerlIO_printf(Perl_debug_log, "\n");
2145 /*RExC_seen |= REG_SEEN_TRIEDFA;*/
2150 * There are strange code-generation bugs caused on sparc64 by gcc-2.95.2.
2151 * These need to be revisited when a newer toolchain becomes available.
2153 #if defined(__sparc64__) && defined(__GNUC__)
2154 # if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 96)
2155 # undef SPARC64_GCC_WORKAROUND
2156 # define SPARC64_GCC_WORKAROUND 1
2160 #define DEBUG_PEEP(str,scan,depth) \
2161 DEBUG_OPTIMISE_r({if (scan){ \
2162 SV * const mysv=sv_newmortal(); \
2163 regnode *Next = regnext(scan); \
2164 regprop(RExC_rx, mysv, scan); \
2165 PerlIO_printf(Perl_debug_log, "%*s" str ">%3d: %s (%d)\n", \
2166 (int)depth*2, "", REG_NODE_NUM(scan), SvPV_nolen_const(mysv),\
2167 Next ? (REG_NODE_NUM(Next)) : 0 ); \
2174 #define JOIN_EXACT(scan,min,flags) \
2175 if (PL_regkind[OP(scan)] == EXACT) \
2176 join_exact(pRExC_state,(scan),(min),(flags),NULL,depth+1)
2179 S_join_exact(pTHX_ RExC_state_t *pRExC_state, regnode *scan, I32 *min, U32 flags,regnode *val, U32 depth) {
2180 /* Merge several consecutive EXACTish nodes into one. */
2181 regnode *n = regnext(scan);
2183 regnode *next = scan + NODE_SZ_STR(scan);
2187 regnode *stop = scan;
2188 GET_RE_DEBUG_FLAGS_DECL;
2190 PERL_UNUSED_ARG(depth);
2192 #ifndef EXPERIMENTAL_INPLACESCAN
2193 PERL_UNUSED_ARG(flags);
2194 PERL_UNUSED_ARG(val);
2196 DEBUG_PEEP("join",scan,depth);
2198 /* Skip NOTHING, merge EXACT*. */
2200 ( PL_regkind[OP(n)] == NOTHING ||
2201 (stringok && (OP(n) == OP(scan))))
2203 && NEXT_OFF(scan) + NEXT_OFF(n) < I16_MAX) {
2205 if (OP(n) == TAIL || n > next)
2207 if (PL_regkind[OP(n)] == NOTHING) {
2208 DEBUG_PEEP("skip:",n,depth);
2209 NEXT_OFF(scan) += NEXT_OFF(n);
2210 next = n + NODE_STEP_REGNODE;
2217 else if (stringok) {
2218 const unsigned int oldl = STR_LEN(scan);
2219 regnode * const nnext = regnext(n);
2221 DEBUG_PEEP("merg",n,depth);
2224 if (oldl + STR_LEN(n) > U8_MAX)
2226 NEXT_OFF(scan) += NEXT_OFF(n);
2227 STR_LEN(scan) += STR_LEN(n);
2228 next = n + NODE_SZ_STR(n);
2229 /* Now we can overwrite *n : */
2230 Move(STRING(n), STRING(scan) + oldl, STR_LEN(n), char);
2238 #ifdef EXPERIMENTAL_INPLACESCAN
2239 if (flags && !NEXT_OFF(n)) {
2240 DEBUG_PEEP("atch", val, depth);
2241 if (reg_off_by_arg[OP(n)]) {
2242 ARG_SET(n, val - n);
2245 NEXT_OFF(n) = val - n;
2252 if (UTF && ( OP(scan) == EXACTF ) && ( STR_LEN(scan) >= 6 ) ) {
2254 Two problematic code points in Unicode casefolding of EXACT nodes:
2256 U+0390 - GREEK SMALL LETTER IOTA WITH DIALYTIKA AND TONOS
2257 U+03B0 - GREEK SMALL LETTER UPSILON WITH DIALYTIKA AND TONOS
2263 U+03B9 U+0308 U+0301 0xCE 0xB9 0xCC 0x88 0xCC 0x81
2264 U+03C5 U+0308 U+0301 0xCF 0x85 0xCC 0x88 0xCC 0x81
2266 This means that in case-insensitive matching (or "loose matching",
2267 as Unicode calls it), an EXACTF of length six (the UTF-8 encoded byte
2268 length of the above casefolded versions) can match a target string
2269 of length two (the byte length of UTF-8 encoded U+0390 or U+03B0).
2270 This would rather mess up the minimum length computation.
2272 What we'll do is to look for the tail four bytes, and then peek
2273 at the preceding two bytes to see whether we need to decrease
2274 the minimum length by four (six minus two).
2276 Thanks to the design of UTF-8, there cannot be false matches:
2277 A sequence of valid UTF-8 bytes cannot be a subsequence of
2278 another valid sequence of UTF-8 bytes.
2281 char * const s0 = STRING(scan), *s, *t;
2282 char * const s1 = s0 + STR_LEN(scan) - 1;
2283 char * const s2 = s1 - 4;
2284 #ifdef EBCDIC /* RD tunifold greek 0390 and 03B0 */
2285 const char t0[] = "\xaf\x49\xaf\x42";
2287 const char t0[] = "\xcc\x88\xcc\x81";
2289 const char * const t1 = t0 + 3;
2292 s < s2 && (t = ninstr(s, s1, t0, t1));
2295 if (((U8)t[-1] == 0x68 && (U8)t[-2] == 0xB4) ||
2296 ((U8)t[-1] == 0x46 && (U8)t[-2] == 0xB5))
2298 if (((U8)t[-1] == 0xB9 && (U8)t[-2] == 0xCE) ||
2299 ((U8)t[-1] == 0x85 && (U8)t[-2] == 0xCF))
2307 n = scan + NODE_SZ_STR(scan);
2309 if (PL_regkind[OP(n)] != NOTHING || OP(n) == NOTHING) {
2316 DEBUG_OPTIMISE_r(if (merged){DEBUG_PEEP("finl",scan,depth)});
2320 /* REx optimizer. Converts nodes into quickier variants "in place".
2321 Finds fixed substrings. */
2323 /* Stops at toplevel WHILEM as well as at "last". At end *scanp is set
2324 to the position after last scanned or to NULL. */
2326 #define INIT_AND_WITHP \
2327 assert(!and_withp); \
2328 Newx(and_withp,1,struct regnode_charclass_class); \
2329 SAVEFREEPV(and_withp)
2331 /* this is a chain of data about sub patterns we are processing that
2332 need to be handled seperately/specially in study_chunk. Its so
2333 we can simulate recursion without losing state. */
2335 typedef struct scan_frame {
2336 regnode *last; /* last node to process in this frame */
2337 regnode *next; /* next node to process when last is reached */
2338 struct scan_frame *prev; /*previous frame*/
2339 I32 stop; /* what stopparen do we use */
2343 S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp,
2344 I32 *minlenp, I32 *deltap,
2349 struct regnode_charclass_class *and_withp,
2350 U32 flags, U32 depth)
2351 /* scanp: Start here (read-write). */
2352 /* deltap: Write maxlen-minlen here. */
2353 /* last: Stop before this one. */
2354 /* data: string data about the pattern */
2355 /* stopparen: treat close N as END */
2356 /* recursed: which subroutines have we recursed into */
2357 /* and_withp: Valid if flags & SCF_DO_STCLASS_OR */
2360 I32 min = 0, pars = 0, code;
2361 regnode *scan = *scanp, *next;
2363 int is_inf = (flags & SCF_DO_SUBSTR) && (data->flags & SF_IS_INF);
2364 int is_inf_internal = 0; /* The studied chunk is infinite */
2365 I32 is_par = OP(scan) == OPEN ? ARG(scan) : 0;
2366 scan_data_t data_fake;
2367 SV *re_trie_maxbuff = NULL;
2368 regnode *first_non_open = scan;
2369 I32 stopmin = I32_MAX;
2370 scan_frame *frame = NULL;
2372 GET_RE_DEBUG_FLAGS_DECL;
2375 StructCopy(&zero_scan_data, &data_fake, scan_data_t);
2379 while (first_non_open && OP(first_non_open) == OPEN)
2380 first_non_open=regnext(first_non_open);
2385 while ( scan && OP(scan) != END && scan < last ){
2386 /* Peephole optimizer: */
2387 DEBUG_STUDYDATA(data,depth);
2388 DEBUG_PEEP("Peep",scan,depth);
2389 JOIN_EXACT(scan,&min,0);
2391 /* Follow the next-chain of the current node and optimize
2392 away all the NOTHINGs from it. */
2393 if (OP(scan) != CURLYX) {
2394 const int max = (reg_off_by_arg[OP(scan)]
2396 /* I32 may be smaller than U16 on CRAYs! */
2397 : (I32_MAX < U16_MAX ? I32_MAX : U16_MAX));
2398 int off = (reg_off_by_arg[OP(scan)] ? ARG(scan) : NEXT_OFF(scan));
2402 /* Skip NOTHING and LONGJMP. */
2403 while ((n = regnext(n))
2404 && ((PL_regkind[OP(n)] == NOTHING && (noff = NEXT_OFF(n)))
2405 || ((OP(n) == LONGJMP) && (noff = ARG(n))))
2406 && off + noff < max)
2408 if (reg_off_by_arg[OP(scan)])
2411 NEXT_OFF(scan) = off;
2416 /* The principal pseudo-switch. Cannot be a switch, since we
2417 look into several different things. */
2418 if (OP(scan) == BRANCH || OP(scan) == BRANCHJ
2419 || OP(scan) == IFTHEN) {
2420 next = regnext(scan);
2422 /* demq: the op(next)==code check is to see if we have "branch-branch" AFAICT */
2424 if (OP(next) == code || code == IFTHEN) {
2425 /* NOTE - There is similar code to this block below for handling
2426 TRIE nodes on a re-study. If you change stuff here check there
2428 I32 max1 = 0, min1 = I32_MAX, num = 0;
2429 struct regnode_charclass_class accum;
2430 regnode * const startbranch=scan;
2432 if (flags & SCF_DO_SUBSTR)
2433 scan_commit(pRExC_state, data, minlenp); /* Cannot merge strings after this. */
2434 if (flags & SCF_DO_STCLASS)
2435 cl_init_zero(pRExC_state, &accum);
2437 while (OP(scan) == code) {
2438 I32 deltanext, minnext, f = 0, fake;
2439 struct regnode_charclass_class this_class;
2442 data_fake.flags = 0;
2444 data_fake.whilem_c = data->whilem_c;
2445 data_fake.last_closep = data->last_closep;
2448 data_fake.last_closep = &fake;
2450 data_fake.pos_delta = delta;
2451 next = regnext(scan);
2452 scan = NEXTOPER(scan);
2454 scan = NEXTOPER(scan);
2455 if (flags & SCF_DO_STCLASS) {
2456 cl_init(pRExC_state, &this_class);
2457 data_fake.start_class = &this_class;
2458 f = SCF_DO_STCLASS_AND;
2460 if (flags & SCF_WHILEM_VISITED_POS)
2461 f |= SCF_WHILEM_VISITED_POS;
2463 /* we suppose the run is continuous, last=next...*/
2464 minnext = study_chunk(pRExC_state, &scan, minlenp, &deltanext,
2466 stopparen, recursed, NULL, f,depth+1);
2469 if (max1 < minnext + deltanext)
2470 max1 = minnext + deltanext;
2471 if (deltanext == I32_MAX)
2472 is_inf = is_inf_internal = 1;
2474 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
2476 if (data_fake.flags & SCF_SEEN_ACCEPT) {
2477 if ( stopmin > minnext)
2478 stopmin = min + min1;
2479 flags &= ~SCF_DO_SUBSTR;
2481 data->flags |= SCF_SEEN_ACCEPT;
2484 if (data_fake.flags & SF_HAS_EVAL)
2485 data->flags |= SF_HAS_EVAL;
2486 data->whilem_c = data_fake.whilem_c;
2488 if (flags & SCF_DO_STCLASS)
2489 cl_or(pRExC_state, &accum, &this_class);
2491 if (code == IFTHEN && num < 2) /* Empty ELSE branch */
2493 if (flags & SCF_DO_SUBSTR) {
2494 data->pos_min += min1;
2495 data->pos_delta += max1 - min1;
2496 if (max1 != min1 || is_inf)
2497 data->longest = &(data->longest_float);
2500 delta += max1 - min1;
2501 if (flags & SCF_DO_STCLASS_OR) {
2502 cl_or(pRExC_state, data->start_class, &accum);
2504 cl_and(data->start_class, and_withp);
2505 flags &= ~SCF_DO_STCLASS;
2508 else if (flags & SCF_DO_STCLASS_AND) {
2510 cl_and(data->start_class, &accum);
2511 flags &= ~SCF_DO_STCLASS;
2514 /* Switch to OR mode: cache the old value of
2515 * data->start_class */
2517 StructCopy(data->start_class, and_withp,
2518 struct regnode_charclass_class);
2519 flags &= ~SCF_DO_STCLASS_AND;
2520 StructCopy(&accum, data->start_class,
2521 struct regnode_charclass_class);
2522 flags |= SCF_DO_STCLASS_OR;
2523 data->start_class->flags |= ANYOF_EOS;
2527 if (PERL_ENABLE_TRIE_OPTIMISATION && OP( startbranch ) == BRANCH ) {
2530 Assuming this was/is a branch we are dealing with: 'scan' now
2531 points at the item that follows the branch sequence, whatever
2532 it is. We now start at the beginning of the sequence and look
2539 which would be constructed from a pattern like /A|LIST|OF|WORDS/
2541 If we can find such a subseqence we need to turn the first
2542 element into a trie and then add the subsequent branch exact
2543 strings to the trie.
2547 1. patterns where the whole set of branch can be converted.
2549 2. patterns where only a subset can be converted.
2551 In case 1 we can replace the whole set with a single regop
2552 for the trie. In case 2 we need to keep the start and end
2555 'BRANCH EXACT; BRANCH EXACT; BRANCH X'
2556 becomes BRANCH TRIE; BRANCH X;
2558 There is an additional case, that being where there is a
2559 common prefix, which gets split out into an EXACT like node
2560 preceding the TRIE node.
2562 If x(1..n)==tail then we can do a simple trie, if not we make
2563 a "jump" trie, such that when we match the appropriate word
2564 we "jump" to the appopriate tail node. Essentailly we turn
2565 a nested if into a case structure of sorts.
2570 if (!re_trie_maxbuff) {
2571 re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1);
2572 if (!SvIOK(re_trie_maxbuff))
2573 sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
2575 if ( SvIV(re_trie_maxbuff)>=0 ) {
2577 regnode *first = (regnode *)NULL;
2578 regnode *last = (regnode *)NULL;
2579 regnode *tail = scan;
2584 SV * const mysv = sv_newmortal(); /* for dumping */
2586 /* var tail is used because there may be a TAIL
2587 regop in the way. Ie, the exacts will point to the
2588 thing following the TAIL, but the last branch will
2589 point at the TAIL. So we advance tail. If we
2590 have nested (?:) we may have to move through several
2594 while ( OP( tail ) == TAIL ) {
2595 /* this is the TAIL generated by (?:) */
2596 tail = regnext( tail );
2601 regprop(RExC_rx, mysv, tail );
2602 PerlIO_printf( Perl_debug_log, "%*s%s%s\n",
2603 (int)depth * 2 + 2, "",
2604 "Looking for TRIE'able sequences. Tail node is: ",
2605 SvPV_nolen_const( mysv )
2611 step through the branches, cur represents each
2612 branch, noper is the first thing to be matched
2613 as part of that branch and noper_next is the
2614 regnext() of that node. if noper is an EXACT
2615 and noper_next is the same as scan (our current
2616 position in the regex) then the EXACT branch is
2617 a possible optimization target. Once we have
2618 two or more consequetive such branches we can
2619 create a trie of the EXACT's contents and stich
2620 it in place. If the sequence represents all of
2621 the branches we eliminate the whole thing and
2622 replace it with a single TRIE. If it is a
2623 subsequence then we need to stitch it in. This
2624 means the first branch has to remain, and needs
2625 to be repointed at the item on the branch chain
2626 following the last branch optimized. This could
2627 be either a BRANCH, in which case the
2628 subsequence is internal, or it could be the
2629 item following the branch sequence in which
2630 case the subsequence is at the end.
2634 /* dont use tail as the end marker for this traverse */
2635 for ( cur = startbranch ; cur != scan ; cur = regnext( cur ) ) {
2636 regnode * const noper = NEXTOPER( cur );
2637 #if defined(DEBUGGING) || defined(NOJUMPTRIE)
2638 regnode * const noper_next = regnext( noper );
2642 regprop(RExC_rx, mysv, cur);
2643 PerlIO_printf( Perl_debug_log, "%*s- %s (%d)",
2644 (int)depth * 2 + 2,"", SvPV_nolen_const( mysv ), REG_NODE_NUM(cur) );
2646 regprop(RExC_rx, mysv, noper);
2647 PerlIO_printf( Perl_debug_log, " -> %s",
2648 SvPV_nolen_const(mysv));
2651 regprop(RExC_rx, mysv, noper_next );
2652 PerlIO_printf( Perl_debug_log,"\t=> %s\t",
2653 SvPV_nolen_const(mysv));
2655 PerlIO_printf( Perl_debug_log, "(First==%d,Last==%d,Cur==%d)\n",
2656 REG_NODE_NUM(first), REG_NODE_NUM(last), REG_NODE_NUM(cur) );
2658 if ( (((first && optype!=NOTHING) ? OP( noper ) == optype
2659 : PL_regkind[ OP( noper ) ] == EXACT )
2660 || OP(noper) == NOTHING )
2662 && noper_next == tail
2667 if ( !first || optype == NOTHING ) {
2668 if (!first) first = cur;
2669 optype = OP( noper );
2675 make_trie( pRExC_state,
2676 startbranch, first, cur, tail, count,
2679 if ( PL_regkind[ OP( noper ) ] == EXACT
2681 && noper_next == tail
2686 optype = OP( noper );
2696 regprop(RExC_rx, mysv, cur);
2697 PerlIO_printf( Perl_debug_log,
2698 "%*s- %s (%d) <SCAN FINISHED>\n", (int)depth * 2 + 2,
2699 "", SvPV_nolen_const( mysv ),REG_NODE_NUM(cur));
2703 made= make_trie( pRExC_state, startbranch, first, scan, tail, count, optype, depth+1 );
2704 #ifdef TRIE_STUDY_OPT
2705 if ( ((made == MADE_EXACT_TRIE &&
2706 startbranch == first)
2707 || ( first_non_open == first )) &&
2709 flags |= SCF_TRIE_RESTUDY;
2710 if ( startbranch == first
2713 RExC_seen &=~REG_TOP_LEVEL_BRANCHES;
2723 else if ( code == BRANCHJ ) { /* single branch is optimized. */
2724 scan = NEXTOPER(NEXTOPER(scan));
2725 } else /* single branch is optimized. */
2726 scan = NEXTOPER(scan);
2728 } else if (OP(scan) == SUSPEND || OP(scan) == GOSUB || OP(scan) == GOSTART) {
2729 scan_frame *newframe = NULL;
2734 if (OP(scan) != SUSPEND) {
2735 /* set the pointer */
2736 if (OP(scan) == GOSUB) {
2738 RExC_recurse[ARG2L(scan)] = scan;
2739 start = RExC_open_parens[paren-1];
2740 end = RExC_close_parens[paren-1];
2743 start = RExC_rxi->program + 1;
2747 Newxz(recursed, (((RExC_npar)>>3) +1), U8);
2748 SAVEFREEPV(recursed);
2750 if (!PAREN_TEST(recursed,paren+1)) {
2751 PAREN_SET(recursed,paren+1);
2752 Newx(newframe,1,scan_frame);
2754 if (flags & SCF_DO_SUBSTR) {
2755 scan_commit(pRExC_state,data,minlenp);
2756 data->longest = &(data->longest_float);
2758 is_inf = is_inf_internal = 1;
2759 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
2760 cl_anything(pRExC_state, data->start_class);
2761 flags &= ~SCF_DO_STCLASS;
2764 Newx(newframe,1,scan_frame);
2767 end = regnext(scan);
2772 SAVEFREEPV(newframe);
2773 newframe->next = regnext(scan);
2774 newframe->last = last;
2775 newframe->stop = stopparen;
2776 newframe->prev = frame;
2786 else if (OP(scan) == EXACT) {
2787 I32 l = STR_LEN(scan);
2790 const U8 * const s = (U8*)STRING(scan);
2791 l = utf8_length(s, s + l);
2792 uc = utf8_to_uvchr(s, NULL);
2794 uc = *((U8*)STRING(scan));
2797 if (flags & SCF_DO_SUBSTR) { /* Update longest substr. */
2798 /* The code below prefers earlier match for fixed
2799 offset, later match for variable offset. */
2800 if (data->last_end == -1) { /* Update the start info. */
2801 data->last_start_min = data->pos_min;
2802 data->last_start_max = is_inf
2803 ? I32_MAX : data->pos_min + data->pos_delta;
2805 sv_catpvn(data->last_found, STRING(scan), STR_LEN(scan));
2807 SvUTF8_on(data->last_found);
2809 SV * const sv = data->last_found;
2810 MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ?
2811 mg_find(sv, PERL_MAGIC_utf8) : NULL;
2812 if (mg && mg->mg_len >= 0)
2813 mg->mg_len += utf8_length((U8*)STRING(scan),
2814 (U8*)STRING(scan)+STR_LEN(scan));
2816 data->last_end = data->pos_min + l;
2817 data->pos_min += l; /* As in the first entry. */
2818 data->flags &= ~SF_BEFORE_EOL;
2820 if (flags & SCF_DO_STCLASS_AND) {
2821 /* Check whether it is compatible with what we know already! */
2825 (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
2826 && !ANYOF_BITMAP_TEST(data->start_class, uc)
2827 && (!(data->start_class->flags & ANYOF_FOLD)
2828 || !ANYOF_BITMAP_TEST(data->start_class, PL_fold[uc])))
2831 ANYOF_CLASS_ZERO(data->start_class);
2832 ANYOF_BITMAP_ZERO(data->start_class);
2834 ANYOF_BITMAP_SET(data->start_class, uc);
2835 data->start_class->flags &= ~ANYOF_EOS;
2837 data->start_class->flags &= ~ANYOF_UNICODE_ALL;
2839 else if (flags & SCF_DO_STCLASS_OR) {
2840 /* false positive possible if the class is case-folded */
2842 ANYOF_BITMAP_SET(data->start_class, uc);
2844 data->start_class->flags |= ANYOF_UNICODE_ALL;
2845 data->start_class->flags &= ~ANYOF_EOS;
2846 cl_and(data->start_class, and_withp);
2848 flags &= ~SCF_DO_STCLASS;
2850 else if (PL_regkind[OP(scan)] == EXACT) { /* But OP != EXACT! */
2851 I32 l = STR_LEN(scan);
2852 UV uc = *((U8*)STRING(scan));
2854 /* Search for fixed substrings supports EXACT only. */
2855 if (flags & SCF_DO_SUBSTR) {
2857 scan_commit(pRExC_state, data, minlenp);
2860 const U8 * const s = (U8 *)STRING(scan);
2861 l = utf8_length(s, s + l);
2862 uc = utf8_to_uvchr(s, NULL);
2865 if (flags & SCF_DO_SUBSTR)
2867 if (flags & SCF_DO_STCLASS_AND) {
2868 /* Check whether it is compatible with what we know already! */
2872 (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
2873 && !ANYOF_BITMAP_TEST(data->start_class, uc)
2874 && !ANYOF_BITMAP_TEST(data->start_class, PL_fold[uc])))
2876 ANYOF_CLASS_ZERO(data->start_class);
2877 ANYOF_BITMAP_ZERO(data->start_class);
2879 ANYOF_BITMAP_SET(data->start_class, uc);
2880 data->start_class->flags &= ~ANYOF_EOS;
2881 data->start_class->flags |= ANYOF_FOLD;
2882 if (OP(scan) == EXACTFL)
2883 data->start_class->flags |= ANYOF_LOCALE;
2886 else if (flags & SCF_DO_STCLASS_OR) {
2887 if (data->start_class->flags & ANYOF_FOLD) {
2888 /* false positive possible if the class is case-folded.
2889 Assume that the locale settings are the same... */
2891 ANYOF_BITMAP_SET(data->start_class, uc);
2892 data->start_class->flags &= ~ANYOF_EOS;
2894 cl_and(data->start_class, and_withp);
2896 flags &= ~SCF_DO_STCLASS;
2898 else if (strchr((const char*)PL_varies,OP(scan))) {
2899 I32 mincount, maxcount, minnext, deltanext, fl = 0;
2900 I32 f = flags, pos_before = 0;
2901 regnode * const oscan = scan;
2902 struct regnode_charclass_class this_class;
2903 struct regnode_charclass_class *oclass = NULL;
2904 I32 next_is_eval = 0;
2906 switch (PL_regkind[OP(scan)]) {
2907 case WHILEM: /* End of (?:...)* . */
2908 scan = NEXTOPER(scan);
2911 if (flags & (SCF_DO_SUBSTR | SCF_DO_STCLASS)) {
2912 next = NEXTOPER(scan);
2913 if (OP(next) == EXACT || (flags & SCF_DO_STCLASS)) {
2915 maxcount = REG_INFTY;
2916 next = regnext(scan);
2917 scan = NEXTOPER(scan);
2921 if (flags & SCF_DO_SUBSTR)
2926 if (flags & SCF_DO_STCLASS) {
2928 maxcount = REG_INFTY;
2929 next = regnext(scan);
2930 scan = NEXTOPER(scan);
2933 is_inf = is_inf_internal = 1;
2934 scan = regnext(scan);
2935 if (flags & SCF_DO_SUBSTR) {
2936 scan_commit(pRExC_state, data, minlenp); /* Cannot extend fixed substrings */
2937 data->longest = &(data->longest_float);
2939 goto optimize_curly_tail;
2941 if (stopparen>0 && (OP(scan)==CURLYN || OP(scan)==CURLYM)
2942 && (scan->flags == stopparen))
2947 mincount = ARG1(scan);
2948 maxcount = ARG2(scan);
2950 next = regnext(scan);
2951 if (OP(scan) == CURLYX) {
2952 I32 lp = (data ? *(data->last_closep) : 0);
2953 scan->flags = ((lp <= (I32)U8_MAX) ? (U8)lp : U8_MAX);
2955 scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
2956 next_is_eval = (OP(scan) == EVAL);
2958 if (flags & SCF_DO_SUBSTR) {
2959 if (mincount == 0) scan_commit(pRExC_state,data,minlenp); /* Cannot extend fixed substrings */
2960 pos_before = data->pos_min;
2964 data->flags &= ~(SF_HAS_PAR|SF_IN_PAR|SF_HAS_EVAL);
2966 data->flags |= SF_IS_INF;
2968 if (flags & SCF_DO_STCLASS) {
2969 cl_init(pRExC_state, &this_class);
2970 oclass = data->start_class;
2971 data->start_class = &this_class;
2972 f |= SCF_DO_STCLASS_AND;
2973 f &= ~SCF_DO_STCLASS_OR;
2975 /* These are the cases when once a subexpression
2976 fails at a particular position, it cannot succeed
2977 even after backtracking at the enclosing scope.
2979 XXXX what if minimal match and we are at the
2980 initial run of {n,m}? */
2981 if ((mincount != maxcount - 1) && (maxcount != REG_INFTY))
2982 f &= ~SCF_WHILEM_VISITED_POS;
2984 /* This will finish on WHILEM, setting scan, or on NULL: */
2985 minnext = study_chunk(pRExC_state, &scan, minlenp, &deltanext,
2986 last, data, stopparen, recursed, NULL,
2988 ? (f & ~SCF_DO_SUBSTR) : f),depth+1);
2990 if (flags & SCF_DO_STCLASS)
2991 data->start_class = oclass;
2992 if (mincount == 0 || minnext == 0) {
2993 if (flags & SCF_DO_STCLASS_OR) {
2994 cl_or(pRExC_state, data->start_class, &this_class);
2996 else if (flags & SCF_DO_STCLASS_AND) {
2997 /* Switch to OR mode: cache the old value of
2998 * data->start_class */
3000 StructCopy(data->start_class, and_withp,
3001 struct regnode_charclass_class);
3002 flags &= ~SCF_DO_STCLASS_AND;
3003 StructCopy(&this_class, data->start_class,
3004 struct regnode_charclass_class);
3005 flags |= SCF_DO_STCLASS_OR;
3006 data->start_class->flags |= ANYOF_EOS;
3008 } else { /* Non-zero len */
3009 if (flags & SCF_DO_STCLASS_OR) {
3010 cl_or(pRExC_state, data->start_class, &this_class);
3011 cl_and(data->start_class, and_withp);
3013 else if (flags & SCF_DO_STCLASS_AND)
3014 cl_and(data->start_class, &this_class);
3015 flags &= ~SCF_DO_STCLASS;
3017 if (!scan) /* It was not CURLYX, but CURLY. */
3019 if ( /* ? quantifier ok, except for (?{ ... }) */
3020 (next_is_eval || !(mincount == 0 && maxcount == 1))
3021 && (minnext == 0) && (deltanext == 0)
3022 && data && !(data->flags & (SF_HAS_PAR|SF_IN_PAR))
3023 && maxcount <= REG_INFTY/3 /* Complement check for big count */
3024 && ckWARN(WARN_REGEXP))
3027 "Quantifier unexpected on zero-length expression");
3030 min += minnext * mincount;
3031 is_inf_internal |= ((maxcount == REG_INFTY
3032 && (minnext + deltanext) > 0)
3033 || deltanext == I32_MAX);
3034 is_inf |= is_inf_internal;
3035 delta += (minnext + deltanext) * maxcount - minnext * mincount;
3037 /* Try powerful optimization CURLYX => CURLYN. */
3038 if ( OP(oscan) == CURLYX && data
3039 && data->flags & SF_IN_PAR
3040 && !(data->flags & SF_HAS_EVAL)
3041 && !deltanext && minnext == 1 ) {
3042 /* Try to optimize to CURLYN. */
3043 regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS;
3044 regnode * const nxt1 = nxt;
3051 if (!strchr((const char*)PL_simple,OP(nxt))
3052 && !(PL_regkind[OP(nxt)] == EXACT
3053 && STR_LEN(nxt) == 1))
3059 if (OP(nxt) != CLOSE)
3061 if (RExC_open_parens) {
3062 RExC_open_parens[ARG(nxt1)-1]=oscan; /*open->CURLYM*/
3063 RExC_close_parens[ARG(nxt1)-1]=nxt+2; /*close->while*/
3065 /* Now we know that nxt2 is the only contents: */
3066 oscan->flags = (U8)ARG(nxt);
3068 OP(nxt1) = NOTHING; /* was OPEN. */
3071 OP(nxt1 + 1) = OPTIMIZED; /* was count. */
3072 NEXT_OFF(nxt1+ 1) = 0; /* just for consistancy. */
3073 NEXT_OFF(nxt2) = 0; /* just for consistancy with CURLY. */
3074 OP(nxt) = OPTIMIZED; /* was CLOSE. */
3075 OP(nxt + 1) = OPTIMIZED; /* was count. */
3076 NEXT_OFF(nxt+ 1) = 0; /* just for consistancy. */
3081 /* Try optimization CURLYX => CURLYM. */
3082 if ( OP(oscan) == CURLYX && data
3083 && !(data->flags & SF_HAS_PAR)
3084 && !(data->flags & SF_HAS_EVAL)
3085 && !deltanext /* atom is fixed width */
3086 && minnext != 0 /* CURLYM can't handle zero width */
3088 /* XXXX How to optimize if data == 0? */
3089 /* Optimize to a simpler form. */
3090 regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN */
3094 while ( (nxt2 = regnext(nxt)) /* skip over embedded stuff*/
3095 && (OP(nxt2) != WHILEM))
3097 OP(nxt2) = SUCCEED; /* Whas WHILEM */
3098 /* Need to optimize away parenths. */
3099 if (data->flags & SF_IN_PAR) {
3100 /* Set the parenth number. */
3101 regnode *nxt1 = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN*/
3103 if (OP(nxt) != CLOSE)
3104 FAIL("Panic opt close");
3105 oscan->flags = (U8)ARG(nxt);
3106 if (RExC_open_parens) {
3107 RExC_open_parens[ARG(nxt1)-1]=oscan; /*open->CURLYM*/
3108 RExC_close_parens[ARG(nxt1)-1]=nxt2+1; /*close->NOTHING*/
3110 OP(nxt1) = OPTIMIZED; /* was OPEN. */
3111 OP(nxt) = OPTIMIZED; /* was CLOSE. */
3114 OP(nxt1 + 1) = OPTIMIZED; /* was count. */
3115 OP(nxt + 1) = OPTIMIZED; /* was count. */
3116 NEXT_OFF(nxt1 + 1) = 0; /* just for consistancy. */
3117 NEXT_OFF(nxt + 1) = 0; /* just for consistancy. */
3120 while ( nxt1 && (OP(nxt1) != WHILEM)) {
3121 regnode *nnxt = regnext(nxt1);
3124 if (reg_off_by_arg[OP(nxt1)])
3125 ARG_SET(nxt1, nxt2 - nxt1);
3126 else if (nxt2 - nxt1 < U16_MAX)
3127 NEXT_OFF(nxt1) = nxt2 - nxt1;
3129 OP(nxt) = NOTHING; /* Cannot beautify */
3134 /* Optimize again: */
3135 study_chunk(pRExC_state, &nxt1, minlenp, &deltanext, nxt,
3136 NULL, stopparen, recursed, NULL, 0,depth+1);
3141 else if ((OP(oscan) == CURLYX)
3142 && (flags & SCF_WHILEM_VISITED_POS)
3143 /* See the comment on a similar expression above.
3144 However, this time it not a subexpression
3145 we care about, but the expression itself. */
3146 && (maxcount == REG_INFTY)
3147 && data && ++data->whilem_c < 16) {
3148 /* This stays as CURLYX, we can put the count/of pair. */
3149 /* Find WHILEM (as in regexec.c) */
3150 regnode *nxt = oscan + NEXT_OFF(oscan);
3152 if (OP(PREVOPER(nxt)) == NOTHING) /* LONGJMP */
3154 PREVOPER(nxt)->flags = (U8)(data->whilem_c
3155 | (RExC_whilem_seen << 4)); /* On WHILEM */
3157 if (data && fl & (SF_HAS_PAR|SF_IN_PAR))
3159 if (flags & SCF_DO_SUBSTR) {
3160 SV *last_str = NULL;
3161 int counted = mincount != 0;
3163 if (data->last_end > 0 && mincount != 0) { /* Ends with a string. */
3164 #if defined(SPARC64_GCC_WORKAROUND)
3167 const char *s = NULL;
3170 if (pos_before >= data->last_start_min)
3173 b = data->last_start_min;
3176 s = SvPV_const(data->last_found, l);
3177 old = b - data->last_start_min;
3180 I32 b = pos_before >= data->last_start_min
3181 ? pos_before : data->last_start_min;
3183 const char * const s = SvPV_const(data->last_found, l);
3184 I32 old = b - data->last_start_min;
3188 old = utf8_hop((U8*)s, old) - (U8*)s;
3191 /* Get the added string: */
3192 last_str = newSVpvn(s + old, l);
3194 SvUTF8_on(last_str);
3195 if (deltanext == 0 && pos_before == b) {
3196 /* What was added is a constant string */
3198 SvGROW(last_str, (mincount * l) + 1);
3199 repeatcpy(SvPVX(last_str) + l,
3200 SvPVX_const(last_str), l, mincount - 1);
3201 SvCUR_set(last_str, SvCUR(last_str) * mincount);
3202 /* Add additional parts. */
3203 SvCUR_set(data->last_found,
3204 SvCUR(data->last_found) - l);
3205 sv_catsv(data->last_found, last_str);
3207 SV * sv = data->last_found;
3209 SvUTF8(sv) && SvMAGICAL(sv) ?
3210 mg_find(sv, PERL_MAGIC_utf8) : NULL;
3211 if (mg && mg->mg_len >= 0)
3212 mg->mg_len += CHR_SVLEN(last_str);
3214 data->last_end += l * (mincount - 1);
3217 /* start offset must point into the last copy */
3218 data->last_start_min += minnext * (mincount - 1);
3219 data->last_start_max += is_inf ? I32_MAX
3220 : (maxcount - 1) * (minnext + data->pos_delta);
3223 /* It is counted once already... */
3224 data->pos_min += minnext * (mincount - counted);
3225 data->pos_delta += - counted * deltanext +
3226 (minnext + deltanext) * maxcount - minnext * mincount;
3227 if (mincount != maxcount) {
3228 /* Cannot extend fixed substrings found inside
3230 scan_commit(pRExC_state,data,minlenp);
3231 if (mincount && last_str) {
3232 SV * const sv = data->last_found;
3233 MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ?
3234 mg_find(sv, PERL_MAGIC_utf8) : NULL;
3238 sv_setsv(sv, last_str);
3239 data->last_end = data->pos_min;
3240 data->last_start_min =
3241 data->pos_min - CHR_SVLEN(last_str);
3242 data->last_start_max = is_inf
3244 : data->pos_min + data->pos_delta
3245 - CHR_SVLEN(last_str);
3247 data->longest = &(data->longest_float);
3249 SvREFCNT_dec(last_str);
3251 if (data && (fl & SF_HAS_EVAL))
3252 data->flags |= SF_HAS_EVAL;
3253 optimize_curly_tail:
3254 if (OP(oscan) != CURLYX) {
3255 while (PL_regkind[OP(next = regnext(oscan))] == NOTHING
3257 NEXT_OFF(oscan) += NEXT_OFF(next);
3260 default: /* REF and CLUMP only? */
3261 if (flags & SCF_DO_SUBSTR) {
3262 scan_commit(pRExC_state,data,minlenp); /* Cannot expect anything... */
3263 data->longest = &(data->longest_float);
3265 is_inf = is_inf_internal = 1;
3266 if (flags & SCF_DO_STCLASS_OR)
3267 cl_anything(pRExC_state, data->start_class);
3268 flags &= ~SCF_DO_STCLASS;
3272 else if (strchr((const char*)PL_simple,OP(scan))) {
3275 if (flags & SCF_DO_SUBSTR) {
3276 scan_commit(pRExC_state,data,minlenp);
3280 if (flags & SCF_DO_STCLASS) {
3281 data->start_class->flags &= ~ANYOF_EOS; /* No match on empty */
3283 /* Some of the logic below assumes that switching
3284 locale on will only add false positives. */
3285 switch (PL_regkind[OP(scan)]) {
3289 /* Perl_croak(aTHX_ "panic: unexpected simple REx opcode %d", OP(scan)); */
3290 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
3291 cl_anything(pRExC_state, data->start_class);
3294 if (OP(scan) == SANY)
3296 if (flags & SCF_DO_STCLASS_OR) { /* Everything but \n */
3297 value = (ANYOF_BITMAP_TEST(data->start_class,'\n')
3298 || (data->start_class->flags & ANYOF_CLASS));
3299 cl_anything(pRExC_state, data->start_class);
3301 if (flags & SCF_DO_STCLASS_AND || !value)
3302 ANYOF_BITMAP_CLEAR(data->start_class,'\n');
3305 if (flags & SCF_DO_STCLASS_AND)
3306 cl_and(data->start_class,
3307 (struct regnode_charclass_class*)scan);
3309 cl_or(pRExC_state, data->start_class,
3310 (struct regnode_charclass_class*)scan);
3313 if (flags & SCF_DO_STCLASS_AND) {
3314 if (!(data->start_class->flags & ANYOF_LOCALE)) {
3315 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
3316 for (value = 0; value < 256; value++)
3317 if (!isALNUM(value))
3318 ANYOF_BITMAP_CLEAR(data->start_class, value);
3322 if (data->start_class->flags & ANYOF_LOCALE)
3323 ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
3325 for (value = 0; value < 256; value++)
3327 ANYOF_BITMAP_SET(data->start_class, value);
3332 if (flags & SCF_DO_STCLASS_AND) {
3333 if (data->start_class->flags & ANYOF_LOCALE)
3334 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NALNUM);
3337 ANYOF_CLASS_SET(data->start_class,ANYOF_ALNUM);
3338 data->start_class->flags |= ANYOF_LOCALE;
3342 if (flags & SCF_DO_STCLASS_AND) {
3343 if (!(data->start_class->flags & ANYOF_LOCALE)) {
3344 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
3345 for (value = 0; value < 256; value++)
3347 ANYOF_BITMAP_CLEAR(data->start_class, value);
3351 if (data->start_class->flags & ANYOF_LOCALE)
3352 ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
3354 for (value = 0; value < 256; value++)
3355 if (!isALNUM(value))
3356 ANYOF_BITMAP_SET(data->start_class, value);
3361 if (flags & SCF_DO_STCLASS_AND) {
3362 if (data->start_class->flags & ANYOF_LOCALE)
3363 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_ALNUM);
3366 data->start_class->flags |= ANYOF_LOCALE;
3367 ANYOF_CLASS_SET(data->start_class,ANYOF_NALNUM);
3371 if (flags & SCF_DO_STCLASS_AND) {
3372 if (!(data->start_class->flags & ANYOF_LOCALE)) {
3373 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
3374 for (value = 0; value < 256; value++)
3375 if (!isSPACE(value))
3376 ANYOF_BITMAP_CLEAR(data->start_class, value);
3380 if (data->start_class->flags & ANYOF_LOCALE)
3381 ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
3383 for (value = 0; value < 256; value++)
3385 ANYOF_BITMAP_SET(data->start_class, value);
3390 if (flags & SCF_DO_STCLASS_AND) {
3391 if (data->start_class->flags & ANYOF_LOCALE)
3392 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NSPACE);
3395 data->start_class->flags |= ANYOF_LOCALE;
3396 ANYOF_CLASS_SET(data->start_class,ANYOF_SPACE);
3400 if (flags & SCF_DO_STCLASS_AND) {
3401 if (!(data->start_class->flags & ANYOF_LOCALE)) {
3402 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
3403 for (value = 0; value < 256; value++)
3405 ANYOF_BITMAP_CLEAR(data->start_class, value);
3409 if (data->start_class->flags & ANYOF_LOCALE)
3410 ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
3412 for (value = 0; value < 256; value++)
3413 if (!isSPACE(value))
3414 ANYOF_BITMAP_SET(data->start_class, value);
3419 if (flags & SCF_DO_STCLASS_AND) {
3420 if (data->start_class->flags & ANYOF_LOCALE) {
3421 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_SPACE);
3422 for (value = 0; value < 256; value++)
3423 if (!isSPACE(value))
3424 ANYOF_BITMAP_CLEAR(data->start_class, value);
3428 data->start_class->flags |= ANYOF_LOCALE;
3429 ANYOF_CLASS_SET(data->start_class,ANYOF_NSPACE);
3433 if (flags & SCF_DO_STCLASS_AND) {
3434 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_NDIGIT);
3435 for (value = 0; value < 256; value++)
3436 if (!isDIGIT(value))
3437 ANYOF_BITMAP_CLEAR(data->start_class, value);
3440 if (data->start_class->flags & ANYOF_LOCALE)
3441 ANYOF_CLASS_SET(data->start_class,ANYOF_DIGIT);
3443 for (value = 0; value < 256; value++)
3445 ANYOF_BITMAP_SET(data->start_class, value);
3450 if (flags & SCF_DO_STCLASS_AND) {
3451 ANYOF_CLASS_CLEAR(data->start_class,ANYOF_DIGIT);
3452 for (value = 0; value < 256; value++)
3454 ANYOF_BITMAP_CLEAR(data->start_class, value);
3457 if (data->start_class->flags & ANYOF_LOCALE)
3458 ANYOF_CLASS_SET(data->start_class,ANYOF_NDIGIT);
3460 for (value = 0; value < 256; value++)
3461 if (!isDIGIT(value))
3462 ANYOF_BITMAP_SET(data->start_class, value);
3467 if (flags & SCF_DO_STCLASS_OR)
3468 cl_and(data->start_class, and_withp);
3469 flags &= ~SCF_DO_STCLASS;
3472 else if (PL_regkind[OP(scan)] == EOL && flags & SCF_DO_SUBSTR) {
3473 data->flags |= (OP(scan) == MEOL
3477 else if ( PL_regkind[OP(scan)] == BRANCHJ
3478 /* Lookbehind, or need to calculate parens/evals/stclass: */
3479 && (scan->flags || data || (flags & SCF_DO_STCLASS))
3480 && (OP(scan) == IFMATCH || OP(scan) == UNLESSM)) {
3481 if ( !PERL_ENABLE_POSITIVE_ASSERTION_STUDY
3482 || OP(scan) == UNLESSM )
3484 /* Negative Lookahead/lookbehind
3485 In this case we can't do fixed string optimisation.
3488 I32 deltanext, minnext, fake = 0;
3490 struct regnode_charclass_class intrnl;
3493 data_fake.flags = 0;
3495 data_fake.whilem_c = data->whilem_c;
3496 data_fake.last_closep = data->last_closep;
3499 data_fake.last_closep = &fake;
3500 data_fake.pos_delta = delta;
3501 if ( flags & SCF_DO_STCLASS && !scan->flags
3502 && OP(scan) == IFMATCH ) { /* Lookahead */
3503 cl_init(pRExC_state, &intrnl);
3504 data_fake.start_class = &intrnl;
3505 f |= SCF_DO_STCLASS_AND;
3507 if (flags & SCF_WHILEM_VISITED_POS)
3508 f |= SCF_WHILEM_VISITED_POS;
3509 next = regnext(scan);
3510 nscan = NEXTOPER(NEXTOPER(scan));
3511 minnext = study_chunk(pRExC_state, &nscan, minlenp, &deltanext,
3512 last, &data_fake, stopparen, recursed, NULL, f, depth+1);
3515 FAIL("Variable length lookbehind not implemented");
3517 else if (minnext > (I32)U8_MAX) {
3518 FAIL2("Lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX);
3520 scan->flags = (U8)minnext;
3523 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
3525 if (data_fake.flags & SF_HAS_EVAL)
3526 data->flags |= SF_HAS_EVAL;
3527 data->whilem_c = data_fake.whilem_c;
3529 if (f & SCF_DO_STCLASS_AND) {
3530 const int was = (data->start_class->flags & ANYOF_EOS);
3532 cl_and(data->start_class, &intrnl);
3534 data->start_class->flags |= ANYOF_EOS;
3537 #if PERL_ENABLE_POSITIVE_ASSERTION_STUDY
3539 /* Positive Lookahead/lookbehind
3540 In this case we can do fixed string optimisation,
3541 but we must be careful about it. Note in the case of
3542 lookbehind the positions will be offset by the minimum
3543 length of the pattern, something we won't know about
3544 until after the recurse.
3546 I32 deltanext, fake = 0;
3548 struct regnode_charclass_class intrnl;
3550 /* We use SAVEFREEPV so that when the full compile
3551 is finished perl will clean up the allocated
3552 minlens when its all done. This was we don't
3553 have to worry about freeing them when we know
3554 they wont be used, which would be a pain.
3557 Newx( minnextp, 1, I32 );
3558 SAVEFREEPV(minnextp);
3561 StructCopy(data, &data_fake, scan_data_t);
3562 if ((flags & SCF_DO_SUBSTR) && data->last_found) {
3565 scan_commit(pRExC_state, &data_fake,minlenp);
3566 data_fake.last_found=newSVsv(data->last_found);
3570 data_fake.last_closep = &fake;
3571 data_fake.flags = 0;
3572 data_fake.pos_delta = delta;
3574 data_fake.flags |= SF_IS_INF;
3575 if ( flags & SCF_DO_STCLASS && !scan->flags
3576 && OP(scan) == IFMATCH ) { /* Lookahead */
3577 cl_init(pRExC_state, &intrnl);
3578 data_fake.start_class = &intrnl;
3579 f |= SCF_DO_STCLASS_AND;
3581 if (flags & SCF_WHILEM_VISITED_POS)
3582 f |= SCF_WHILEM_VISITED_POS;
3583 next = regnext(scan);
3584 nscan = NEXTOPER(NEXTOPER(scan));
3586 *minnextp = study_chunk(pRExC_state, &nscan, minnextp, &deltanext,
3587 last, &data_fake, stopparen, recursed, NULL, f,depth+1);
3590 FAIL("Variable length lookbehind not implemented");
3592 else if (*minnextp > (I32)U8_MAX) {
3593 FAIL2("Lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX);
3595 scan->flags = (U8)*minnextp;
3600 if (f & SCF_DO_STCLASS_AND) {
3601 const int was = (data->start_class->flags & ANYOF_EOS);
3603 cl_and(data->start_class, &intrnl);
3605 data->start_class->flags |= ANYOF_EOS;
3608 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
3610 if (data_fake.flags & SF_HAS_EVAL)
3611 data->flags |= SF_HAS_EVAL;
3612 data->whilem_c = data_fake.whilem_c;
3613 if ((flags & SCF_DO_SUBSTR) && data_fake.last_found) {
3614 if (RExC_rx->minlen<*minnextp)
3615 RExC_rx->minlen=*minnextp;
3616 scan_commit(pRExC_state, &data_fake, minnextp);
3617 SvREFCNT_dec(data_fake.last_found);
3619 if ( data_fake.minlen_fixed != minlenp )
3621 data->offset_fixed= data_fake.offset_fixed;
3622 data->minlen_fixed= data_fake.minlen_fixed;
3623 data->lookbehind_fixed+= scan->flags;
3625 if ( data_fake.minlen_float != minlenp )
3627 data->minlen_float= data_fake.minlen_float;
3628 data->offset_float_min=data_fake.offset_float_min;
3629 data->offset_float_max=data_fake.offset_float_max;
3630 data->lookbehind_float+= scan->flags;
3639 else if (OP(scan) == OPEN) {
3640 if (stopparen != (I32)ARG(scan))
3643 else if (OP(scan) == CLOSE) {
3644 if (stopparen == (I32)ARG(scan)) {
3647 if ((I32)ARG(scan) == is_par) {
3648 next = regnext(scan);
3650 if ( next && (OP(next) != WHILEM) && next < last)
3651 is_par = 0; /* Disable optimization */
3654 *(data->last_closep) = ARG(scan);
3656 else if (OP(scan) == EVAL) {
3658 data->flags |= SF_HAS_EVAL;
3660 else if ( PL_regkind[OP(scan)] == ENDLIKE ) {
3661 if (flags & SCF_DO_SUBSTR) {
3662 scan_commit(pRExC_state,data,minlenp);
3663 flags &= ~SCF_DO_SUBSTR;
3665 if (data && OP(scan)==ACCEPT) {
3666 data->flags |= SCF_SEEN_ACCEPT;
3671 else if (OP(scan) == LOGICAL && scan->flags == 2) /* Embedded follows */
3673 if (flags & SCF_DO_SUBSTR) {
3674 scan_commit(pRExC_state,data,minlenp);
3675 data->longest = &(data->longest_float);
3677 is_inf = is_inf_internal = 1;
3678 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
3679 cl_anything(pRExC_state, data->start_class);
3680 flags &= ~SCF_DO_STCLASS;
3682 else if (OP(scan) == GPOS) {
3683 if (!(RExC_rx->extflags & RXf_GPOS_FLOAT) &&
3684 !(delta || is_inf || (data && data->pos_delta)))
3686 if (!(RExC_rx->extflags & RXf_ANCH) && (flags & SCF_DO_SUBSTR))
3687 RExC_rx->extflags |= RXf_ANCH_GPOS;
3688 if (RExC_rx->gofs < (U32)min)
3689 RExC_rx->gofs = min;
3691 RExC_rx->extflags |= RXf_GPOS_FLOAT;
3695 #ifdef TRIE_STUDY_OPT
3696 #ifdef FULL_TRIE_STUDY
3697 else if (PL_regkind[OP(scan)] == TRIE) {
3698 /* NOTE - There is similar code to this block above for handling
3699 BRANCH nodes on the initial study. If you change stuff here
3701 regnode *trie_node= scan;
3702 regnode *tail= regnext(scan);
3703 reg_trie_data *trie = (reg_trie_data*)RExC_rxi->data->data[ ARG(scan) ];
3704 I32 max1 = 0, min1 = I32_MAX;
3705 struct regnode_charclass_class accum;
3707 if (flags & SCF_DO_SUBSTR) /* XXXX Add !SUSPEND? */
3708 scan_commit(pRExC_state, data,minlenp); /* Cannot merge strings after this. */
3709 if (flags & SCF_DO_STCLASS)
3710 cl_init_zero(pRExC_state, &accum);
3716 const regnode *nextbranch= NULL;
3719 for ( word=1 ; word <= trie->wordcount ; word++)
3721 I32 deltanext=0, minnext=0, f = 0, fake;
3722 struct regnode_charclass_class this_class;
3724 data_fake.flags = 0;
3726 data_fake.whilem_c = data->whilem_c;
3727 data_fake.last_closep = data->last_closep;
3730 data_fake.last_closep = &fake;
3731 data_fake.pos_delta = delta;
3732 if (flags & SCF_DO_STCLASS) {
3733 cl_init(pRExC_state, &this_class);
3734 data_fake.start_class = &this_class;
3735 f = SCF_DO_STCLASS_AND;
3737 if (flags & SCF_WHILEM_VISITED_POS)
3738 f |= SCF_WHILEM_VISITED_POS;
3740 if (trie->jump[word]) {
3742 nextbranch = trie_node + trie->jump[0];
3743 scan= trie_node + trie->jump[word];
3744 /* We go from the jump point to the branch that follows
3745 it. Note this means we need the vestigal unused branches
3746 even though they arent otherwise used.
3748 minnext = study_chunk(pRExC_state, &scan, minlenp,
3749 &deltanext, (regnode *)nextbranch, &data_fake,
3750 stopparen, recursed, NULL, f,depth+1);
3752 if (nextbranch && PL_regkind[OP(nextbranch)]==BRANCH)
3753 nextbranch= regnext((regnode*)nextbranch);
3755 if (min1 > (I32)(minnext + trie->minlen))
3756 min1 = minnext + trie->minlen;
3757 if (max1 < (I32)(minnext + deltanext + trie->maxlen))
3758 max1 = minnext + deltanext + trie->maxlen;
3759 if (deltanext == I32_MAX)
3760 is_inf = is_inf_internal = 1;
3762 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
3764 if (data_fake.flags & SCF_SEEN_ACCEPT) {
3765 if ( stopmin > min + min1)
3766 stopmin = min + min1;
3767 flags &= ~SCF_DO_SUBSTR;
3769 data->flags |= SCF_SEEN_ACCEPT;
3772 if (data_fake.flags & SF_HAS_EVAL)
3773 data->flags |= SF_HAS_EVAL;
3774 data->whilem_c = data_fake.whilem_c;
3776 if (flags & SCF_DO_STCLASS)
3777 cl_or(pRExC_state, &accum, &this_class);
3780 if (flags & SCF_DO_SUBSTR) {
3781 data->pos_min += min1;
3782 data->pos_delta += max1 - min1;
3783 if (max1 != min1 || is_inf)
3784 data->longest = &(data->longest_float);
3787 delta += max1 - min1;
3788 if (flags & SCF_DO_STCLASS_OR) {
3789 cl_or(pRExC_state, data->start_class, &accum);
3791 cl_and(data->start_class, and_withp);
3792 flags &= ~SCF_DO_STCLASS;
3795 else if (flags & SCF_DO_STCLASS_AND) {
3797 cl_and(data->start_class, &accum);
3798 flags &= ~SCF_DO_STCLASS;
3801 /* Switch to OR mode: cache the old value of
3802 * data->start_class */
3804 StructCopy(data->start_class, and_withp,
3805 struct regnode_charclass_class);
3806 flags &= ~SCF_DO_STCLASS_AND;
3807 StructCopy(&accum, data->start_class,
3808 struct regnode_charclass_class);
3809 flags |= SCF_DO_STCLASS_OR;
3810 data->start_class->flags |= ANYOF_EOS;
3817 else if (PL_regkind[OP(scan)] == TRIE) {
3818 reg_trie_data *trie = (reg_trie_data*)RExC_rxi->data->data[ ARG(scan) ];
3821 min += trie->minlen;
3822 delta += (trie->maxlen - trie->minlen);
3823 flags &= ~SCF_DO_STCLASS; /* xxx */
3824 if (flags & SCF_DO_SUBSTR) {
3825 scan_commit(pRExC_state,data,minlenp); /* Cannot expect any
https://perl5.git.perl.org / perl5.git / blob