5 * 'A fair jaw-cracker dwarf-language must be.' --Samwise Gamgee
7 * [p.285 of _The Lord of the Rings_, II/iii: "The Ring Goes South"]
10 /* This file contains functions for compiling a regular expression. See
11 * also regexec.c which funnily enough, contains functions for executing
12 * a regular expression.
14 * This file is also copied at build time to ext/re/re_comp.c, where
15 * it's built with -DPERL_EXT_RE_BUILD -DPERL_EXT_RE_DEBUG -DPERL_EXT.
16 * This causes the main functions to be compiled under new names and with
17 * debugging support added, which makes "use re 'debug'" work.
20 /* NOTE: this is derived from Henry Spencer's regexp code, and should not
21 * confused with the original package (see point 3 below). Thanks, Henry!
24 /* Additional note: this code is very heavily munged from Henry's version
25 * in places. In some spots I've traded clarity for efficiency, so don't
26 * blame Henry for some of the lack of readability.
29 /* The names of the functions have been changed from regcomp and
30 * regexec to pregcomp and pregexec in order to avoid conflicts
31 * with the POSIX routines of the same names.
34 #ifdef PERL_EXT_RE_BUILD
39 * pregcomp and pregexec -- regsub and regerror are not used in perl
41 * Copyright (c) 1986 by University of Toronto.
42 * Written by Henry Spencer. Not derived from licensed software.
44 * Permission is granted to anyone to use this software for any
45 * purpose on any computer system, and to redistribute it freely,
46 * subject to the following restrictions:
48 * 1. The author is not responsible for the consequences of use of
49 * this software, no matter how awful, even if they arise
52 * 2. The origin of this software must not be misrepresented, either
53 * by explicit claim or by omission.
55 * 3. Altered versions must be plainly marked as such, and must not
56 * be misrepresented as being the original software.
59 **** Alterations to Henry's code are...
61 **** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
62 **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
63 **** by Larry Wall and others
65 **** You may distribute under the terms of either the GNU General Public
66 **** License or the Artistic License, as specified in the README file.
69 * Beware that some of this code is subtly aware of the way operator
70 * precedence is structured in regular expressions. Serious changes in
71 * regular-expression syntax might require a total rethink.
74 #define PERL_IN_REGCOMP_C
77 #ifndef PERL_IN_XSUB_RE
82 #ifdef PERL_IN_XSUB_RE
88 #include "dquote_static.c"
89 #ifndef PERL_IN_XSUB_RE
90 # include "charclass_invlists.h"
93 #define HAS_NONLATIN1_FOLD_CLOSURE(i) _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(i)
100 # if defined(BUGGY_MSC6)
101 /* MSC 6.00A breaks on op/regexp.t test 85 unless we turn this off */
102 # pragma optimize("a",off)
103 /* But MSC 6.00A is happy with 'w', for aliases only across function calls*/
104 # pragma optimize("w",on )
105 # endif /* BUGGY_MSC6 */
109 #define STATIC static
113 typedef struct RExC_state_t {
114 U32 flags; /* RXf_* are we folding, multilining? */
115 U32 pm_flags; /* PMf_* stuff from the calling PMOP */
116 char *precomp; /* uncompiled string. */
117 REGEXP *rx_sv; /* The SV that is the regexp. */
118 regexp *rx; /* perl core regexp structure */
119 regexp_internal *rxi; /* internal data for regexp object pprivate field */
120 char *start; /* Start of input for compile */
121 char *end; /* End of input for compile */
122 char *parse; /* Input-scan pointer. */
123 I32 whilem_seen; /* number of WHILEM in this expr */
124 regnode *emit_start; /* Start of emitted-code area */
125 regnode *emit_bound; /* First regnode outside of the allocated space */
126 regnode *emit; /* Code-emit pointer; ®dummy = don't = compiling */
127 I32 naughty; /* How bad is this pattern? */
128 I32 sawback; /* Did we see \1, ...? */
130 I32 size; /* Code size. */
131 I32 npar; /* Capture buffer count, (OPEN). */
132 I32 cpar; /* Capture buffer count, (CLOSE). */
133 I32 nestroot; /* root parens we are in - used by accept */
137 regnode **open_parens; /* pointers to open parens */
138 regnode **close_parens; /* pointers to close parens */
139 regnode *opend; /* END node in program */
140 I32 utf8; /* whether the pattern is utf8 or not */
141 I32 orig_utf8; /* whether the pattern was originally in utf8 */
142 /* XXX use this for future optimisation of case
143 * where pattern must be upgraded to utf8. */
144 I32 uni_semantics; /* If a d charset modifier should use unicode
145 rules, even if the pattern is not in
147 HV *paren_names; /* Paren names */
149 regnode **recurse; /* Recurse regops */
150 I32 recurse_count; /* Number of recurse regops */
153 I32 override_recoding;
154 struct reg_code_block *code_blocks; /* positions of literal (?{})
156 int num_code_blocks; /* size of code_blocks[] */
157 int code_index; /* next code_blocks[] slot */
159 char *starttry; /* -Dr: where regtry was called. */
160 #define RExC_starttry (pRExC_state->starttry)
162 SV *runtime_code_qr; /* qr with the runtime code blocks */
164 const char *lastparse;
166 AV *paren_name_list; /* idx -> name */
167 #define RExC_lastparse (pRExC_state->lastparse)
168 #define RExC_lastnum (pRExC_state->lastnum)
169 #define RExC_paren_name_list (pRExC_state->paren_name_list)
173 #define RExC_flags (pRExC_state->flags)
174 #define RExC_pm_flags (pRExC_state->pm_flags)
175 #define RExC_precomp (pRExC_state->precomp)
176 #define RExC_rx_sv (pRExC_state->rx_sv)
177 #define RExC_rx (pRExC_state->rx)
178 #define RExC_rxi (pRExC_state->rxi)
179 #define RExC_start (pRExC_state->start)
180 #define RExC_end (pRExC_state->end)
181 #define RExC_parse (pRExC_state->parse)
182 #define RExC_whilem_seen (pRExC_state->whilem_seen)
183 #ifdef RE_TRACK_PATTERN_OFFSETS
184 #define RExC_offsets (pRExC_state->rxi->u.offsets) /* I am not like the others */
186 #define RExC_emit (pRExC_state->emit)
187 #define RExC_emit_start (pRExC_state->emit_start)
188 #define RExC_emit_bound (pRExC_state->emit_bound)
189 #define RExC_naughty (pRExC_state->naughty)
190 #define RExC_sawback (pRExC_state->sawback)
191 #define RExC_seen (pRExC_state->seen)
192 #define RExC_size (pRExC_state->size)
193 #define RExC_npar (pRExC_state->npar)
194 #define RExC_nestroot (pRExC_state->nestroot)
195 #define RExC_extralen (pRExC_state->extralen)
196 #define RExC_seen_zerolen (pRExC_state->seen_zerolen)
197 #define RExC_seen_evals (pRExC_state->seen_evals)
198 #define RExC_utf8 (pRExC_state->utf8)
199 #define RExC_uni_semantics (pRExC_state->uni_semantics)
200 #define RExC_orig_utf8 (pRExC_state->orig_utf8)
201 #define RExC_open_parens (pRExC_state->open_parens)
202 #define RExC_close_parens (pRExC_state->close_parens)
203 #define RExC_opend (pRExC_state->opend)
204 #define RExC_paren_names (pRExC_state->paren_names)
205 #define RExC_recurse (pRExC_state->recurse)
206 #define RExC_recurse_count (pRExC_state->recurse_count)
207 #define RExC_in_lookbehind (pRExC_state->in_lookbehind)
208 #define RExC_contains_locale (pRExC_state->contains_locale)
209 #define RExC_override_recoding (pRExC_state->override_recoding)
212 #define ISMULT1(c) ((c) == '*' || (c) == '+' || (c) == '?')
213 #define ISMULT2(s) ((*s) == '*' || (*s) == '+' || (*s) == '?' || \
214 ((*s) == '{' && regcurly(s)))
217 #undef SPSTART /* dratted cpp namespace... */
220 * Flags to be passed up and down.
222 #define WORST 0 /* Worst case. */
223 #define HASWIDTH 0x01 /* Known to match non-null strings. */
225 /* Simple enough to be STAR/PLUS operand, in an EXACT node must be a single
226 * character, and if utf8, must be invariant. Note that this is not the same
227 * thing as REGNODE_SIMPLE */
229 #define SPSTART 0x04 /* Starts with * or +. */
230 #define TRYAGAIN 0x08 /* Weeded out a declaration. */
231 #define POSTPONED 0x10 /* (?1),(?&name), (??{...}) or similar */
233 #define REG_NODE_NUM(x) ((x) ? (int)((x)-RExC_emit_start) : -1)
235 /* whether trie related optimizations are enabled */
236 #if PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION
237 #define TRIE_STUDY_OPT
238 #define FULL_TRIE_STUDY
244 #define PBYTE(u8str,paren) ((U8*)(u8str))[(paren) >> 3]
245 #define PBITVAL(paren) (1 << ((paren) & 7))
246 #define PAREN_TEST(u8str,paren) ( PBYTE(u8str,paren) & PBITVAL(paren))
247 #define PAREN_SET(u8str,paren) PBYTE(u8str,paren) |= PBITVAL(paren)
248 #define PAREN_UNSET(u8str,paren) PBYTE(u8str,paren) &= (~PBITVAL(paren))
250 /* If not already in utf8, do a longjmp back to the beginning */
251 #define UTF8_LONGJMP 42 /* Choose a value not likely to ever conflict */
252 #define REQUIRE_UTF8 STMT_START { \
253 if (! UTF) JMPENV_JUMP(UTF8_LONGJMP); \
256 /* About scan_data_t.
258 During optimisation we recurse through the regexp program performing
259 various inplace (keyhole style) optimisations. In addition study_chunk
260 and scan_commit populate this data structure with information about
261 what strings MUST appear in the pattern. We look for the longest
262 string that must appear at a fixed location, and we look for the
263 longest string that may appear at a floating location. So for instance
268 Both 'FOO' and 'A' are fixed strings. Both 'B' and 'BAR' are floating
269 strings (because they follow a .* construct). study_chunk will identify
270 both FOO and BAR as being the longest fixed and floating strings respectively.
272 The strings can be composites, for instance
276 will result in a composite fixed substring 'foo'.
278 For each string some basic information is maintained:
280 - offset or min_offset
281 This is the position the string must appear at, or not before.
282 It also implicitly (when combined with minlenp) tells us how many
283 characters must match before the string we are searching for.
284 Likewise when combined with minlenp and the length of the string it
285 tells us how many characters must appear after the string we have
289 Only used for floating strings. This is the rightmost point that
290 the string can appear at. If set to I32 max it indicates that the
291 string can occur infinitely far to the right.
294 A pointer to the minimum length of the pattern that the string
295 was found inside. This is important as in the case of positive
296 lookahead or positive lookbehind we can have multiple patterns
301 The minimum length of the pattern overall is 3, the minimum length
302 of the lookahead part is 3, but the minimum length of the part that
303 will actually match is 1. So 'FOO's minimum length is 3, but the
304 minimum length for the F is 1. This is important as the minimum length
305 is used to determine offsets in front of and behind the string being
306 looked for. Since strings can be composites this is the length of the
307 pattern at the time it was committed with a scan_commit. Note that
308 the length is calculated by study_chunk, so that the minimum lengths
309 are not known until the full pattern has been compiled, thus the
310 pointer to the value.
314 In the case of lookbehind the string being searched for can be
315 offset past the start point of the final matching string.
316 If this value was just blithely removed from the min_offset it would
317 invalidate some of the calculations for how many chars must match
318 before or after (as they are derived from min_offset and minlen and
319 the length of the string being searched for).
320 When the final pattern is compiled and the data is moved from the
321 scan_data_t structure into the regexp structure the information
322 about lookbehind is factored in, with the information that would
323 have been lost precalculated in the end_shift field for the
326 The fields pos_min and pos_delta are used to store the minimum offset
327 and the delta to the maximum offset at the current point in the pattern.
331 typedef struct scan_data_t {
332 /*I32 len_min; unused */
333 /*I32 len_delta; unused */
337 I32 last_end; /* min value, <0 unless valid. */
340 SV **longest; /* Either &l_fixed, or &l_float. */
341 SV *longest_fixed; /* longest fixed string found in pattern */
342 I32 offset_fixed; /* offset where it starts */
343 I32 *minlen_fixed; /* pointer to the minlen relevant to the string */
344 I32 lookbehind_fixed; /* is the position of the string modfied by LB */
345 SV *longest_float; /* longest floating string found in pattern */
346 I32 offset_float_min; /* earliest point in string it can appear */
347 I32 offset_float_max; /* latest point in string it can appear */
348 I32 *minlen_float; /* pointer to the minlen relevant to the string */
349 I32 lookbehind_float; /* is the position of the string modified by LB */
353 struct regnode_charclass_class *start_class;
357 * Forward declarations for pregcomp()'s friends.
360 static const scan_data_t zero_scan_data =
361 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ,0};
363 #define SF_BEFORE_EOL (SF_BEFORE_SEOL|SF_BEFORE_MEOL)
364 #define SF_BEFORE_SEOL 0x0001
365 #define SF_BEFORE_MEOL 0x0002
366 #define SF_FIX_BEFORE_EOL (SF_FIX_BEFORE_SEOL|SF_FIX_BEFORE_MEOL)
367 #define SF_FL_BEFORE_EOL (SF_FL_BEFORE_SEOL|SF_FL_BEFORE_MEOL)
370 # define SF_FIX_SHIFT_EOL (0+2)
371 # define SF_FL_SHIFT_EOL (0+4)
373 # define SF_FIX_SHIFT_EOL (+2)
374 # define SF_FL_SHIFT_EOL (+4)
377 #define SF_FIX_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FIX_SHIFT_EOL)
378 #define SF_FIX_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FIX_SHIFT_EOL)
380 #define SF_FL_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FL_SHIFT_EOL)
381 #define SF_FL_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FL_SHIFT_EOL) /* 0x20 */
382 #define SF_IS_INF 0x0040
383 #define SF_HAS_PAR 0x0080
384 #define SF_IN_PAR 0x0100
385 #define SF_HAS_EVAL 0x0200
386 #define SCF_DO_SUBSTR 0x0400
387 #define SCF_DO_STCLASS_AND 0x0800
388 #define SCF_DO_STCLASS_OR 0x1000
389 #define SCF_DO_STCLASS (SCF_DO_STCLASS_AND|SCF_DO_STCLASS_OR)
390 #define SCF_WHILEM_VISITED_POS 0x2000
392 #define SCF_TRIE_RESTUDY 0x4000 /* Do restudy? */
393 #define SCF_SEEN_ACCEPT 0x8000
395 #define UTF cBOOL(RExC_utf8)
397 /* The enums for all these are ordered so things work out correctly */
398 #define LOC (get_regex_charset(RExC_flags) == REGEX_LOCALE_CHARSET)
399 #define DEPENDS_SEMANTICS (get_regex_charset(RExC_flags) == REGEX_DEPENDS_CHARSET)
400 #define UNI_SEMANTICS (get_regex_charset(RExC_flags) == REGEX_UNICODE_CHARSET)
401 #define AT_LEAST_UNI_SEMANTICS (get_regex_charset(RExC_flags) >= REGEX_UNICODE_CHARSET)
402 #define ASCII_RESTRICTED (get_regex_charset(RExC_flags) == REGEX_ASCII_RESTRICTED_CHARSET)
403 #define MORE_ASCII_RESTRICTED (get_regex_charset(RExC_flags) == REGEX_ASCII_MORE_RESTRICTED_CHARSET)
404 #define AT_LEAST_ASCII_RESTRICTED (get_regex_charset(RExC_flags) >= REGEX_ASCII_RESTRICTED_CHARSET)
406 #define FOLD cBOOL(RExC_flags & RXf_PMf_FOLD)
408 #define OOB_UNICODE 12345678
409 #define OOB_NAMEDCLASS -1
411 #define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
412 #define CHR_DIST(a,b) (UTF ? utf8_distance(a,b) : a - b)
415 /* length of regex to show in messages that don't mark a position within */
416 #define RegexLengthToShowInErrorMessages 127
419 * If MARKER[12] are adjusted, be sure to adjust the constants at the top
420 * of t/op/regmesg.t, the tests in t/op/re_tests, and those in
421 * op/pragma/warn/regcomp.
423 #define MARKER1 "<-- HERE" /* marker as it appears in the description */
424 #define MARKER2 " <-- HERE " /* marker as it appears within the regex */
426 #define REPORT_LOCATION " in regex; marked by " MARKER1 " in m/%.*s" MARKER2 "%s/"
429 * Calls SAVEDESTRUCTOR_X if needed, then calls Perl_croak with the given
430 * arg. Show regex, up to a maximum length. If it's too long, chop and add
433 #define _FAIL(code) STMT_START { \
434 const char *ellipses = ""; \
435 IV len = RExC_end - RExC_precomp; \
438 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx_sv); \
439 if (len > RegexLengthToShowInErrorMessages) { \
440 /* chop 10 shorter than the max, to ensure meaning of "..." */ \
441 len = RegexLengthToShowInErrorMessages - 10; \
447 #define FAIL(msg) _FAIL( \
448 Perl_croak(aTHX_ "%s in regex m/%.*s%s/", \
449 msg, (int)len, RExC_precomp, ellipses))
451 #define FAIL2(msg,arg) _FAIL( \
452 Perl_croak(aTHX_ msg " in regex m/%.*s%s/", \
453 arg, (int)len, RExC_precomp, ellipses))
456 * Simple_vFAIL -- like FAIL, but marks the current location in the scan
458 #define Simple_vFAIL(m) STMT_START { \
459 const IV offset = RExC_parse - RExC_precomp; \
460 Perl_croak(aTHX_ "%s" REPORT_LOCATION, \
461 m, (int)offset, RExC_precomp, RExC_precomp + offset); \
465 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL()
467 #define vFAIL(m) STMT_START { \
469 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx_sv); \
474 * Like Simple_vFAIL(), but accepts two arguments.
476 #define Simple_vFAIL2(m,a1) STMT_START { \
477 const IV offset = RExC_parse - RExC_precomp; \
478 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, \
479 (int)offset, RExC_precomp, RExC_precomp + offset); \
483 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL2().
485 #define vFAIL2(m,a1) STMT_START { \
487 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx_sv); \
488 Simple_vFAIL2(m, a1); \
493 * Like Simple_vFAIL(), but accepts three arguments.
495 #define Simple_vFAIL3(m, a1, a2) STMT_START { \
496 const IV offset = RExC_parse - RExC_precomp; \
497 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, \
498 (int)offset, RExC_precomp, RExC_precomp + offset); \
502 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL3().
504 #define vFAIL3(m,a1,a2) STMT_START { \
506 SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx_sv); \
507 Simple_vFAIL3(m, a1, a2); \
511 * Like Simple_vFAIL(), but accepts four arguments.
513 #define Simple_vFAIL4(m, a1, a2, a3) STMT_START { \
514 const IV offset = RExC_parse - RExC_precomp; \
515 S_re_croak2(aTHX_ m, REPORT_LOCATION, a1, a2, a3, \
516 (int)offset, RExC_precomp, RExC_precomp + offset); \
519 #define ckWARNreg(loc,m) STMT_START { \
520 const IV offset = loc - RExC_precomp; \
521 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
522 (int)offset, RExC_precomp, RExC_precomp + offset); \
525 #define ckWARNregdep(loc,m) STMT_START { \
526 const IV offset = loc - RExC_precomp; \
527 Perl_ck_warner_d(aTHX_ packWARN2(WARN_DEPRECATED, WARN_REGEXP), \
529 (int)offset, RExC_precomp, RExC_precomp + offset); \
532 #define ckWARN2regdep(loc,m, a1) STMT_START { \
533 const IV offset = loc - RExC_precomp; \
534 Perl_ck_warner_d(aTHX_ packWARN2(WARN_DEPRECATED, WARN_REGEXP), \
536 a1, (int)offset, RExC_precomp, RExC_precomp + offset); \
539 #define ckWARN2reg(loc, m, a1) STMT_START { \
540 const IV offset = loc - RExC_precomp; \
541 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
542 a1, (int)offset, RExC_precomp, RExC_precomp + offset); \
545 #define vWARN3(loc, m, a1, a2) STMT_START { \
546 const IV offset = loc - RExC_precomp; \
547 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
548 a1, a2, (int)offset, RExC_precomp, RExC_precomp + offset); \
551 #define ckWARN3reg(loc, m, a1, a2) STMT_START { \
552 const IV offset = loc - RExC_precomp; \
553 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
554 a1, a2, (int)offset, RExC_precomp, RExC_precomp + offset); \
557 #define vWARN4(loc, m, a1, a2, a3) STMT_START { \
558 const IV offset = loc - RExC_precomp; \
559 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
560 a1, a2, a3, (int)offset, RExC_precomp, RExC_precomp + offset); \
563 #define ckWARN4reg(loc, m, a1, a2, a3) STMT_START { \
564 const IV offset = loc - RExC_precomp; \
565 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
566 a1, a2, a3, (int)offset, RExC_precomp, RExC_precomp + offset); \
569 #define vWARN5(loc, m, a1, a2, a3, a4) STMT_START { \
570 const IV offset = loc - RExC_precomp; \
571 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
572 a1, a2, a3, a4, (int)offset, RExC_precomp, RExC_precomp + offset); \
576 /* Allow for side effects in s */
577 #define REGC(c,s) STMT_START { \
578 if (!SIZE_ONLY) *(s) = (c); else (void)(s); \
581 /* Macros for recording node offsets. 20001227 mjd@plover.com
582 * Nodes are numbered 1, 2, 3, 4. Node #n's position is recorded in
583 * element 2*n-1 of the array. Element #2n holds the byte length node #n.
584 * Element 0 holds the number n.
585 * Position is 1 indexed.
587 #ifndef RE_TRACK_PATTERN_OFFSETS
588 #define Set_Node_Offset_To_R(node,byte)
589 #define Set_Node_Offset(node,byte)
590 #define Set_Cur_Node_Offset
591 #define Set_Node_Length_To_R(node,len)
592 #define Set_Node_Length(node,len)
593 #define Set_Node_Cur_Length(node)
594 #define Node_Offset(n)
595 #define Node_Length(n)
596 #define Set_Node_Offset_Length(node,offset,len)
597 #define ProgLen(ri) ri->u.proglen
598 #define SetProgLen(ri,x) ri->u.proglen = x
600 #define ProgLen(ri) ri->u.offsets[0]
601 #define SetProgLen(ri,x) ri->u.offsets[0] = x
602 #define Set_Node_Offset_To_R(node,byte) STMT_START { \
604 MJD_OFFSET_DEBUG(("** (%d) offset of node %d is %d.\n", \
605 __LINE__, (int)(node), (int)(byte))); \
607 Perl_croak(aTHX_ "value of node is %d in Offset macro", (int)(node)); \
609 RExC_offsets[2*(node)-1] = (byte); \
614 #define Set_Node_Offset(node,byte) \
615 Set_Node_Offset_To_R((node)-RExC_emit_start, (byte)-RExC_start)
616 #define Set_Cur_Node_Offset Set_Node_Offset(RExC_emit, RExC_parse)
618 #define Set_Node_Length_To_R(node,len) STMT_START { \
620 MJD_OFFSET_DEBUG(("** (%d) size of node %d is %d.\n", \
621 __LINE__, (int)(node), (int)(len))); \
623 Perl_croak(aTHX_ "value of node is %d in Length macro", (int)(node)); \
625 RExC_offsets[2*(node)] = (len); \
630 #define Set_Node_Length(node,len) \
631 Set_Node_Length_To_R((node)-RExC_emit_start, len)
632 #define Set_Cur_Node_Length(len) Set_Node_Length(RExC_emit, len)
633 #define Set_Node_Cur_Length(node) \
634 Set_Node_Length(node, RExC_parse - parse_start)
636 /* Get offsets and lengths */
637 #define Node_Offset(n) (RExC_offsets[2*((n)-RExC_emit_start)-1])
638 #define Node_Length(n) (RExC_offsets[2*((n)-RExC_emit_start)])
640 #define Set_Node_Offset_Length(node,offset,len) STMT_START { \
641 Set_Node_Offset_To_R((node)-RExC_emit_start, (offset)); \
642 Set_Node_Length_To_R((node)-RExC_emit_start, (len)); \
646 #if PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS
647 #define EXPERIMENTAL_INPLACESCAN
648 #endif /*PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS*/
650 #define DEBUG_STUDYDATA(str,data,depth) \
651 DEBUG_OPTIMISE_MORE_r(if(data){ \
652 PerlIO_printf(Perl_debug_log, \
653 "%*s" str "Pos:%"IVdf"/%"IVdf \
654 " Flags: 0x%"UVXf" Whilem_c: %"IVdf" Lcp: %"IVdf" %s", \
655 (int)(depth)*2, "", \
656 (IV)((data)->pos_min), \
657 (IV)((data)->pos_delta), \
658 (UV)((data)->flags), \
659 (IV)((data)->whilem_c), \
660 (IV)((data)->last_closep ? *((data)->last_closep) : -1), \
661 is_inf ? "INF " : "" \
663 if ((data)->last_found) \
664 PerlIO_printf(Perl_debug_log, \
665 "Last:'%s' %"IVdf":%"IVdf"/%"IVdf" %sFixed:'%s' @ %"IVdf \
666 " %sFloat: '%s' @ %"IVdf"/%"IVdf"", \
667 SvPVX_const((data)->last_found), \
668 (IV)((data)->last_end), \
669 (IV)((data)->last_start_min), \
670 (IV)((data)->last_start_max), \
671 ((data)->longest && \
672 (data)->longest==&((data)->longest_fixed)) ? "*" : "", \
673 SvPVX_const((data)->longest_fixed), \
674 (IV)((data)->offset_fixed), \
675 ((data)->longest && \
676 (data)->longest==&((data)->longest_float)) ? "*" : "", \
677 SvPVX_const((data)->longest_float), \
678 (IV)((data)->offset_float_min), \
679 (IV)((data)->offset_float_max) \
681 PerlIO_printf(Perl_debug_log,"\n"); \
684 static void clear_re(pTHX_ void *r);
686 /* Mark that we cannot extend a found fixed substring at this point.
687 Update the longest found anchored substring and the longest found
688 floating substrings if needed. */
691 S_scan_commit(pTHX_ const RExC_state_t *pRExC_state, scan_data_t *data, I32 *minlenp, int is_inf)
693 const STRLEN l = CHR_SVLEN(data->last_found);
694 const STRLEN old_l = CHR_SVLEN(*data->longest);
695 GET_RE_DEBUG_FLAGS_DECL;
697 PERL_ARGS_ASSERT_SCAN_COMMIT;
699 if ((l >= old_l) && ((l > old_l) || (data->flags & SF_BEFORE_EOL))) {
700 SvSetMagicSV(*data->longest, data->last_found);
701 if (*data->longest == data->longest_fixed) {
702 data->offset_fixed = l ? data->last_start_min : data->pos_min;
703 if (data->flags & SF_BEFORE_EOL)
705 |= ((data->flags & SF_BEFORE_EOL) << SF_FIX_SHIFT_EOL);
707 data->flags &= ~SF_FIX_BEFORE_EOL;
708 data->minlen_fixed=minlenp;
709 data->lookbehind_fixed=0;
711 else { /* *data->longest == data->longest_float */
712 data->offset_float_min = l ? data->last_start_min : data->pos_min;
713 data->offset_float_max = (l
714 ? data->last_start_max
715 : data->pos_min + data->pos_delta);
716 if (is_inf || (U32)data->offset_float_max > (U32)I32_MAX)
717 data->offset_float_max = I32_MAX;
718 if (data->flags & SF_BEFORE_EOL)
720 |= ((data->flags & SF_BEFORE_EOL) << SF_FL_SHIFT_EOL);
722 data->flags &= ~SF_FL_BEFORE_EOL;
723 data->minlen_float=minlenp;
724 data->lookbehind_float=0;
727 SvCUR_set(data->last_found, 0);
729 SV * const sv = data->last_found;
730 if (SvUTF8(sv) && SvMAGICAL(sv)) {
731 MAGIC * const mg = mg_find(sv, PERL_MAGIC_utf8);
737 data->flags &= ~SF_BEFORE_EOL;
738 DEBUG_STUDYDATA("commit: ",data,0);
741 /* Can match anything (initialization) */
743 S_cl_anything(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
745 PERL_ARGS_ASSERT_CL_ANYTHING;
747 ANYOF_BITMAP_SETALL(cl);
748 cl->flags = ANYOF_CLASS|ANYOF_EOS|ANYOF_UNICODE_ALL
749 |ANYOF_LOC_NONBITMAP_FOLD|ANYOF_NON_UTF8_LATIN1_ALL;
751 /* If any portion of the regex is to operate under locale rules,
752 * initialization includes it. The reason this isn't done for all regexes
753 * is that the optimizer was written under the assumption that locale was
754 * all-or-nothing. Given the complexity and lack of documentation in the
755 * optimizer, and that there are inadequate test cases for locale, so many
756 * parts of it may not work properly, it is safest to avoid locale unless
758 if (RExC_contains_locale) {
759 ANYOF_CLASS_SETALL(cl); /* /l uses class */
760 cl->flags |= ANYOF_LOCALE;
763 ANYOF_CLASS_ZERO(cl); /* Only /l uses class now */
767 /* Can match anything (initialization) */
769 S_cl_is_anything(const struct regnode_charclass_class *cl)
773 PERL_ARGS_ASSERT_CL_IS_ANYTHING;
775 for (value = 0; value <= ANYOF_MAX; value += 2)
776 if (ANYOF_CLASS_TEST(cl, value) && ANYOF_CLASS_TEST(cl, value + 1))
778 if (!(cl->flags & ANYOF_UNICODE_ALL))
780 if (!ANYOF_BITMAP_TESTALLSET((const void*)cl))
785 /* Can match anything (initialization) */
787 S_cl_init(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl)
789 PERL_ARGS_ASSERT_CL_INIT;
791 Zero(cl, 1, struct regnode_charclass_class);
793 cl_anything(pRExC_state, cl);
794 ARG_SET(cl, ANYOF_NONBITMAP_EMPTY);
797 /* These two functions currently do the exact same thing */
798 #define cl_init_zero S_cl_init
800 /* 'AND' a given class with another one. Can create false positives. 'cl'
801 * should not be inverted. 'and_with->flags & ANYOF_CLASS' should be 0 if
802 * 'and_with' is a regnode_charclass instead of a regnode_charclass_class. */
804 S_cl_and(struct regnode_charclass_class *cl,
805 const struct regnode_charclass_class *and_with)
807 PERL_ARGS_ASSERT_CL_AND;
809 assert(and_with->type == ANYOF);
811 /* I (khw) am not sure all these restrictions are necessary XXX */
812 if (!(ANYOF_CLASS_TEST_ANY_SET(and_with))
813 && !(ANYOF_CLASS_TEST_ANY_SET(cl))
814 && (and_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
815 && !(and_with->flags & ANYOF_LOC_NONBITMAP_FOLD)
816 && !(cl->flags & ANYOF_LOC_NONBITMAP_FOLD)) {
819 if (and_with->flags & ANYOF_INVERT)
820 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
821 cl->bitmap[i] &= ~and_with->bitmap[i];
823 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
824 cl->bitmap[i] &= and_with->bitmap[i];
825 } /* XXXX: logic is complicated otherwise, leave it along for a moment. */
827 if (and_with->flags & ANYOF_INVERT) {
829 /* Here, the and'ed node is inverted. Get the AND of the flags that
830 * aren't affected by the inversion. Those that are affected are
831 * handled individually below */
832 U8 affected_flags = cl->flags & ~INVERSION_UNAFFECTED_FLAGS;
833 cl->flags &= (and_with->flags & INVERSION_UNAFFECTED_FLAGS);
834 cl->flags |= affected_flags;
836 /* We currently don't know how to deal with things that aren't in the
837 * bitmap, but we know that the intersection is no greater than what
838 * is already in cl, so let there be false positives that get sorted
839 * out after the synthetic start class succeeds, and the node is
840 * matched for real. */
842 /* The inversion of these two flags indicate that the resulting
843 * intersection doesn't have them */
844 if (and_with->flags & ANYOF_UNICODE_ALL) {
845 cl->flags &= ~ANYOF_UNICODE_ALL;
847 if (and_with->flags & ANYOF_NON_UTF8_LATIN1_ALL) {
848 cl->flags &= ~ANYOF_NON_UTF8_LATIN1_ALL;
851 else { /* and'd node is not inverted */
852 U8 outside_bitmap_but_not_utf8; /* Temp variable */
854 if (! ANYOF_NONBITMAP(and_with)) {
856 /* Here 'and_with' doesn't match anything outside the bitmap
857 * (except possibly ANYOF_UNICODE_ALL), which means the
858 * intersection can't either, except for ANYOF_UNICODE_ALL, in
859 * which case we don't know what the intersection is, but it's no
860 * greater than what cl already has, so can just leave it alone,
861 * with possible false positives */
862 if (! (and_with->flags & ANYOF_UNICODE_ALL)) {
863 ARG_SET(cl, ANYOF_NONBITMAP_EMPTY);
864 cl->flags &= ~ANYOF_NONBITMAP_NON_UTF8;
867 else if (! ANYOF_NONBITMAP(cl)) {
869 /* Here, 'and_with' does match something outside the bitmap, and cl
870 * doesn't have a list of things to match outside the bitmap. If
871 * cl can match all code points above 255, the intersection will
872 * be those above-255 code points that 'and_with' matches. If cl
873 * can't match all Unicode code points, it means that it can't
874 * match anything outside the bitmap (since the 'if' that got us
875 * into this block tested for that), so we leave the bitmap empty.
877 if (cl->flags & ANYOF_UNICODE_ALL) {
878 ARG_SET(cl, ARG(and_with));
880 /* and_with's ARG may match things that don't require UTF8.
881 * And now cl's will too, in spite of this being an 'and'. See
882 * the comments below about the kludge */
883 cl->flags |= and_with->flags & ANYOF_NONBITMAP_NON_UTF8;
887 /* Here, both 'and_with' and cl match something outside the
888 * bitmap. Currently we do not do the intersection, so just match
889 * whatever cl had at the beginning. */
893 /* Take the intersection of the two sets of flags. However, the
894 * ANYOF_NONBITMAP_NON_UTF8 flag is treated as an 'or'. This is a
895 * kludge around the fact that this flag is not treated like the others
896 * which are initialized in cl_anything(). The way the optimizer works
897 * is that the synthetic start class (SSC) is initialized to match
898 * anything, and then the first time a real node is encountered, its
899 * values are AND'd with the SSC's with the result being the values of
900 * the real node. However, there are paths through the optimizer where
901 * the AND never gets called, so those initialized bits are set
902 * inappropriately, which is not usually a big deal, as they just cause
903 * false positives in the SSC, which will just mean a probably
904 * imperceptible slow down in execution. However this bit has a
905 * higher false positive consequence in that it can cause utf8.pm,
906 * utf8_heavy.pl ... to be loaded when not necessary, which is a much
907 * bigger slowdown and also causes significant extra memory to be used.
908 * In order to prevent this, the code now takes a different tack. The
909 * bit isn't set unless some part of the regular expression needs it,
910 * but once set it won't get cleared. This means that these extra
911 * modules won't get loaded unless there was some path through the
912 * pattern that would have required them anyway, and so any false
913 * positives that occur by not ANDing them out when they could be
914 * aren't as severe as they would be if we treated this bit like all
916 outside_bitmap_but_not_utf8 = (cl->flags | and_with->flags)
917 & ANYOF_NONBITMAP_NON_UTF8;
918 cl->flags &= and_with->flags;
919 cl->flags |= outside_bitmap_but_not_utf8;
923 /* 'OR' a given class with another one. Can create false positives. 'cl'
924 * should not be inverted. 'or_with->flags & ANYOF_CLASS' should be 0 if
925 * 'or_with' is a regnode_charclass instead of a regnode_charclass_class. */
927 S_cl_or(const RExC_state_t *pRExC_state, struct regnode_charclass_class *cl, const struct regnode_charclass_class *or_with)
929 PERL_ARGS_ASSERT_CL_OR;
931 if (or_with->flags & ANYOF_INVERT) {
933 /* Here, the or'd node is to be inverted. This means we take the
934 * complement of everything not in the bitmap, but currently we don't
935 * know what that is, so give up and match anything */
936 if (ANYOF_NONBITMAP(or_with)) {
937 cl_anything(pRExC_state, cl);
940 * (B1 | CL1) | (!B2 & !CL2) = (B1 | !B2 & !CL2) | (CL1 | (!B2 & !CL2))
941 * <= (B1 | !B2) | (CL1 | !CL2)
942 * which is wasteful if CL2 is small, but we ignore CL2:
943 * (B1 | CL1) | (!B2 & !CL2) <= (B1 | CL1) | !B2 = (B1 | !B2) | CL1
944 * XXXX Can we handle case-fold? Unclear:
945 * (OK1(i) | OK1(i')) | !(OK1(i) | OK1(i')) =
946 * (OK1(i) | OK1(i')) | (!OK1(i) & !OK1(i'))
948 else if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
949 && !(or_with->flags & ANYOF_LOC_NONBITMAP_FOLD)
950 && !(cl->flags & ANYOF_LOC_NONBITMAP_FOLD) ) {
953 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
954 cl->bitmap[i] |= ~or_with->bitmap[i];
955 } /* XXXX: logic is complicated otherwise */
957 cl_anything(pRExC_state, cl);
960 /* And, we can just take the union of the flags that aren't affected
961 * by the inversion */
962 cl->flags |= or_with->flags & INVERSION_UNAFFECTED_FLAGS;
964 /* For the remaining flags:
965 ANYOF_UNICODE_ALL and inverted means to not match anything above
966 255, which means that the union with cl should just be
967 what cl has in it, so can ignore this flag
968 ANYOF_NON_UTF8_LATIN1_ALL and inverted means if not utf8 and ord
969 is 127-255 to match them, but then invert that, so the
970 union with cl should just be what cl has in it, so can
973 } else { /* 'or_with' is not inverted */
974 /* (B1 | CL1) | (B2 | CL2) = (B1 | B2) | (CL1 | CL2)) */
975 if ( (or_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
976 && (!(or_with->flags & ANYOF_LOC_NONBITMAP_FOLD)
977 || (cl->flags & ANYOF_LOC_NONBITMAP_FOLD)) ) {
980 /* OR char bitmap and class bitmap separately */
981 for (i = 0; i < ANYOF_BITMAP_SIZE; i++)
982 cl->bitmap[i] |= or_with->bitmap[i];
983 if (ANYOF_CLASS_TEST_ANY_SET(or_with)) {
984 for (i = 0; i < ANYOF_CLASSBITMAP_SIZE; i++)
985 cl->classflags[i] |= or_with->classflags[i];
986 cl->flags |= ANYOF_CLASS;
989 else { /* XXXX: logic is complicated, leave it along for a moment. */
990 cl_anything(pRExC_state, cl);
993 if (ANYOF_NONBITMAP(or_with)) {
995 /* Use the added node's outside-the-bit-map match if there isn't a
996 * conflict. If there is a conflict (both nodes match something
997 * outside the bitmap, but what they match outside is not the same
998 * pointer, and hence not easily compared until XXX we extend
999 * inversion lists this far), give up and allow the start class to
1000 * match everything outside the bitmap. If that stuff is all above
1001 * 255, can just set UNICODE_ALL, otherwise caould be anything. */
1002 if (! ANYOF_NONBITMAP(cl)) {
1003 ARG_SET(cl, ARG(or_with));
1005 else if (ARG(cl) != ARG(or_with)) {
1007 if ((or_with->flags & ANYOF_NONBITMAP_NON_UTF8)) {
1008 cl_anything(pRExC_state, cl);
1011 cl->flags |= ANYOF_UNICODE_ALL;
1016 /* Take the union */
1017 cl->flags |= or_with->flags;
1021 #define TRIE_LIST_ITEM(state,idx) (trie->states[state].trans.list)[ idx ]
1022 #define TRIE_LIST_CUR(state) ( TRIE_LIST_ITEM( state, 0 ).forid )
1023 #define TRIE_LIST_LEN(state) ( TRIE_LIST_ITEM( state, 0 ).newstate )
1024 #define TRIE_LIST_USED(idx) ( trie->states[state].trans.list ? (TRIE_LIST_CUR( idx ) - 1) : 0 )
1029 dump_trie(trie,widecharmap,revcharmap)
1030 dump_trie_interim_list(trie,widecharmap,revcharmap,next_alloc)
1031 dump_trie_interim_table(trie,widecharmap,revcharmap,next_alloc)
1033 These routines dump out a trie in a somewhat readable format.
1034 The _interim_ variants are used for debugging the interim
1035 tables that are used to generate the final compressed
1036 representation which is what dump_trie expects.
1038 Part of the reason for their existence is to provide a form
1039 of documentation as to how the different representations function.
1044 Dumps the final compressed table form of the trie to Perl_debug_log.
1045 Used for debugging make_trie().
1049 S_dump_trie(pTHX_ const struct _reg_trie_data *trie, HV *widecharmap,
1050 AV *revcharmap, U32 depth)
1053 SV *sv=sv_newmortal();
1054 int colwidth= widecharmap ? 6 : 4;
1056 GET_RE_DEBUG_FLAGS_DECL;
1058 PERL_ARGS_ASSERT_DUMP_TRIE;
1060 PerlIO_printf( Perl_debug_log, "%*sChar : %-6s%-6s%-4s ",
1061 (int)depth * 2 + 2,"",
1062 "Match","Base","Ofs" );
1064 for( state = 0 ; state < trie->uniquecharcount ; state++ ) {
1065 SV ** const tmp = av_fetch( revcharmap, state, 0);
1067 PerlIO_printf( Perl_debug_log, "%*s",
1069 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
1070 PL_colors[0], PL_colors[1],
1071 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
1072 PERL_PV_ESCAPE_FIRSTCHAR
1077 PerlIO_printf( Perl_debug_log, "\n%*sState|-----------------------",
1078 (int)depth * 2 + 2,"");
1080 for( state = 0 ; state < trie->uniquecharcount ; state++ )
1081 PerlIO_printf( Perl_debug_log, "%.*s", colwidth, "--------");
1082 PerlIO_printf( Perl_debug_log, "\n");
1084 for( state = 1 ; state < trie->statecount ; state++ ) {
1085 const U32 base = trie->states[ state ].trans.base;
1087 PerlIO_printf( Perl_debug_log, "%*s#%4"UVXf"|", (int)depth * 2 + 2,"", (UV)state);
1089 if ( trie->states[ state ].wordnum ) {
1090 PerlIO_printf( Perl_debug_log, " W%4X", trie->states[ state ].wordnum );
1092 PerlIO_printf( Perl_debug_log, "%6s", "" );
1095 PerlIO_printf( Perl_debug_log, " @%4"UVXf" ", (UV)base );
1100 while( ( base + ofs < trie->uniquecharcount ) ||
1101 ( base + ofs - trie->uniquecharcount < trie->lasttrans
1102 && trie->trans[ base + ofs - trie->uniquecharcount ].check != state))
1105 PerlIO_printf( Perl_debug_log, "+%2"UVXf"[ ", (UV)ofs);
1107 for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) {
1108 if ( ( base + ofs >= trie->uniquecharcount ) &&
1109 ( base + ofs - trie->uniquecharcount < trie->lasttrans ) &&
1110 trie->trans[ base + ofs - trie->uniquecharcount ].check == state )
1112 PerlIO_printf( Perl_debug_log, "%*"UVXf,
1114 (UV)trie->trans[ base + ofs - trie->uniquecharcount ].next );
1116 PerlIO_printf( Perl_debug_log, "%*s",colwidth," ." );
1120 PerlIO_printf( Perl_debug_log, "]");
1123 PerlIO_printf( Perl_debug_log, "\n" );
1125 PerlIO_printf(Perl_debug_log, "%*sword_info N:(prev,len)=", (int)depth*2, "");
1126 for (word=1; word <= trie->wordcount; word++) {
1127 PerlIO_printf(Perl_debug_log, " %d:(%d,%d)",
1128 (int)word, (int)(trie->wordinfo[word].prev),
1129 (int)(trie->wordinfo[word].len));
1131 PerlIO_printf(Perl_debug_log, "\n" );
1134 Dumps a fully constructed but uncompressed trie in list form.
1135 List tries normally only are used for construction when the number of
1136 possible chars (trie->uniquecharcount) is very high.
1137 Used for debugging make_trie().
1140 S_dump_trie_interim_list(pTHX_ const struct _reg_trie_data *trie,
1141 HV *widecharmap, AV *revcharmap, U32 next_alloc,
1145 SV *sv=sv_newmortal();
1146 int colwidth= widecharmap ? 6 : 4;
1147 GET_RE_DEBUG_FLAGS_DECL;
1149 PERL_ARGS_ASSERT_DUMP_TRIE_INTERIM_LIST;
1151 /* print out the table precompression. */
1152 PerlIO_printf( Perl_debug_log, "%*sState :Word | Transition Data\n%*s%s",
1153 (int)depth * 2 + 2,"", (int)depth * 2 + 2,"",
1154 "------:-----+-----------------\n" );
1156 for( state=1 ; state < next_alloc ; state ++ ) {
1159 PerlIO_printf( Perl_debug_log, "%*s %4"UVXf" :",
1160 (int)depth * 2 + 2,"", (UV)state );
1161 if ( ! trie->states[ state ].wordnum ) {
1162 PerlIO_printf( Perl_debug_log, "%5s| ","");
1164 PerlIO_printf( Perl_debug_log, "W%4x| ",
1165 trie->states[ state ].wordnum
1168 for( charid = 1 ; charid <= TRIE_LIST_USED( state ) ; charid++ ) {
1169 SV ** const tmp = av_fetch( revcharmap, TRIE_LIST_ITEM(state,charid).forid, 0);
1171 PerlIO_printf( Perl_debug_log, "%*s:%3X=%4"UVXf" | ",
1173 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
1174 PL_colors[0], PL_colors[1],
1175 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
1176 PERL_PV_ESCAPE_FIRSTCHAR
1178 TRIE_LIST_ITEM(state,charid).forid,
1179 (UV)TRIE_LIST_ITEM(state,charid).newstate
1182 PerlIO_printf(Perl_debug_log, "\n%*s| ",
1183 (int)((depth * 2) + 14), "");
1186 PerlIO_printf( Perl_debug_log, "\n");
1191 Dumps a fully constructed but uncompressed trie in table form.
1192 This is the normal DFA style state transition table, with a few
1193 twists to facilitate compression later.
1194 Used for debugging make_trie().
1197 S_dump_trie_interim_table(pTHX_ const struct _reg_trie_data *trie,
1198 HV *widecharmap, AV *revcharmap, U32 next_alloc,
1203 SV *sv=sv_newmortal();
1204 int colwidth= widecharmap ? 6 : 4;
1205 GET_RE_DEBUG_FLAGS_DECL;
1207 PERL_ARGS_ASSERT_DUMP_TRIE_INTERIM_TABLE;
1210 print out the table precompression so that we can do a visual check
1211 that they are identical.
1214 PerlIO_printf( Perl_debug_log, "%*sChar : ",(int)depth * 2 + 2,"" );
1216 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
1217 SV ** const tmp = av_fetch( revcharmap, charid, 0);
1219 PerlIO_printf( Perl_debug_log, "%*s",
1221 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
1222 PL_colors[0], PL_colors[1],
1223 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
1224 PERL_PV_ESCAPE_FIRSTCHAR
1230 PerlIO_printf( Perl_debug_log, "\n%*sState+-",(int)depth * 2 + 2,"" );
1232 for( charid=0 ; charid < trie->uniquecharcount ; charid++ ) {
1233 PerlIO_printf( Perl_debug_log, "%.*s", colwidth,"--------");
1236 PerlIO_printf( Perl_debug_log, "\n" );
1238 for( state=1 ; state < next_alloc ; state += trie->uniquecharcount ) {
1240 PerlIO_printf( Perl_debug_log, "%*s%4"UVXf" : ",
1241 (int)depth * 2 + 2,"",
1242 (UV)TRIE_NODENUM( state ) );
1244 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
1245 UV v=(UV)SAFE_TRIE_NODENUM( trie->trans[ state + charid ].next );
1247 PerlIO_printf( Perl_debug_log, "%*"UVXf, colwidth, v );
1249 PerlIO_printf( Perl_debug_log, "%*s", colwidth, "." );
1251 if ( ! trie->states[ TRIE_NODENUM( state ) ].wordnum ) {
1252 PerlIO_printf( Perl_debug_log, " (%4"UVXf")\n", (UV)trie->trans[ state ].check );
1254 PerlIO_printf( Perl_debug_log, " (%4"UVXf") W%4X\n", (UV)trie->trans[ state ].check,
1255 trie->states[ TRIE_NODENUM( state ) ].wordnum );
1263 /* make_trie(startbranch,first,last,tail,word_count,flags,depth)
1264 startbranch: the first branch in the whole branch sequence
1265 first : start branch of sequence of branch-exact nodes.
1266 May be the same as startbranch
1267 last : Thing following the last branch.
1268 May be the same as tail.
1269 tail : item following the branch sequence
1270 count : words in the sequence
1271 flags : currently the OP() type we will be building one of /EXACT(|F|Fl)/
1272 depth : indent depth
1274 Inplace optimizes a sequence of 2 or more Branch-Exact nodes into a TRIE node.
1276 A trie is an N'ary tree where the branches are determined by digital
1277 decomposition of the key. IE, at the root node you look up the 1st character and
1278 follow that branch repeat until you find the end of the branches. Nodes can be
1279 marked as "accepting" meaning they represent a complete word. Eg:
1283 would convert into the following structure. Numbers represent states, letters
1284 following numbers represent valid transitions on the letter from that state, if
1285 the number is in square brackets it represents an accepting state, otherwise it
1286 will be in parenthesis.
1288 +-h->+-e->[3]-+-r->(8)-+-s->[9]
1292 (1) +-i->(6)-+-s->[7]
1294 +-s->(3)-+-h->(4)-+-e->[5]
1296 Accept Word Mapping: 3=>1 (he),5=>2 (she), 7=>3 (his), 9=>4 (hers)
1298 This shows that when matching against the string 'hers' we will begin at state 1
1299 read 'h' and move to state 2, read 'e' and move to state 3 which is accepting,
1300 then read 'r' and go to state 8 followed by 's' which takes us to state 9 which
1301 is also accepting. Thus we know that we can match both 'he' and 'hers' with a
1302 single traverse. We store a mapping from accepting to state to which word was
1303 matched, and then when we have multiple possibilities we try to complete the
1304 rest of the regex in the order in which they occured in the alternation.
1306 The only prior NFA like behaviour that would be changed by the TRIE support is
1307 the silent ignoring of duplicate alternations which are of the form:
1309 / (DUPE|DUPE) X? (?{ ... }) Y /x
1311 Thus EVAL blocks following a trie may be called a different number of times with
1312 and without the optimisation. With the optimisations dupes will be silently
1313 ignored. This inconsistent behaviour of EVAL type nodes is well established as
1314 the following demonstrates:
1316 'words'=~/(word|word|word)(?{ print $1 })[xyz]/
1318 which prints out 'word' three times, but
1320 'words'=~/(word|word|word)(?{ print $1 })S/
1322 which doesnt print it out at all. This is due to other optimisations kicking in.
1324 Example of what happens on a structural level:
1326 The regexp /(ac|ad|ab)+/ will produce the following debug output:
1328 1: CURLYM[1] {1,32767}(18)
1339 This would be optimizable with startbranch=5, first=5, last=16, tail=16
1340 and should turn into:
1342 1: CURLYM[1] {1,32767}(18)
1344 [Words:3 Chars Stored:6 Unique Chars:4 States:5 NCP:1]
1352 Cases where tail != last would be like /(?foo|bar)baz/:
1362 which would be optimizable with startbranch=1, first=1, last=7, tail=8
1363 and would end up looking like:
1366 [Words:2 Chars Stored:6 Unique Chars:5 States:7 NCP:1]
1373 d = uvuni_to_utf8_flags(d, uv, 0);
1375 is the recommended Unicode-aware way of saying
1380 #define TRIE_STORE_REVCHAR(val) \
1383 SV *zlopp = newSV(7); /* XXX: optimize me */ \
1384 unsigned char *flrbbbbb = (unsigned char *) SvPVX(zlopp); \
1385 unsigned const char *const kapow = uvuni_to_utf8(flrbbbbb, val); \
1386 SvCUR_set(zlopp, kapow - flrbbbbb); \
1389 av_push(revcharmap, zlopp); \
1391 char ooooff = (char)val; \
1392 av_push(revcharmap, newSVpvn(&ooooff, 1)); \
1396 #define TRIE_READ_CHAR STMT_START { \
1399 /* if it is UTF then it is either already folded, or does not need folding */ \
1400 uvc = utf8n_to_uvuni( (const U8*) uc, UTF8_MAXLEN, &len, uniflags); \
1402 else if (folder == PL_fold_latin1) { \
1403 /* if we use this folder we have to obey unicode rules on latin-1 data */ \
1404 if ( foldlen > 0 ) { \
1405 uvc = utf8n_to_uvuni( (const U8*) scan, UTF8_MAXLEN, &len, uniflags ); \
1411 uvc = _to_fold_latin1( (U8) *uc, foldbuf, &foldlen, 1); \
1412 skiplen = UNISKIP(uvc); \
1413 foldlen -= skiplen; \
1414 scan = foldbuf + skiplen; \
1417 /* raw data, will be folded later if needed */ \
1425 #define TRIE_LIST_PUSH(state,fid,ns) STMT_START { \
1426 if ( TRIE_LIST_CUR( state ) >=TRIE_LIST_LEN( state ) ) { \
1427 U32 ging = TRIE_LIST_LEN( state ) *= 2; \
1428 Renew( trie->states[ state ].trans.list, ging, reg_trie_trans_le ); \
1430 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).forid = fid; \
1431 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).newstate = ns; \
1432 TRIE_LIST_CUR( state )++; \
1435 #define TRIE_LIST_NEW(state) STMT_START { \
1436 Newxz( trie->states[ state ].trans.list, \
1437 4, reg_trie_trans_le ); \
1438 TRIE_LIST_CUR( state ) = 1; \
1439 TRIE_LIST_LEN( state ) = 4; \
1442 #define TRIE_HANDLE_WORD(state) STMT_START { \
1443 U16 dupe= trie->states[ state ].wordnum; \
1444 regnode * const noper_next = regnext( noper ); \
1447 /* store the word for dumping */ \
1449 if (OP(noper) != NOTHING) \
1450 tmp = newSVpvn_utf8(STRING(noper), STR_LEN(noper), UTF); \
1452 tmp = newSVpvn_utf8( "", 0, UTF ); \
1453 av_push( trie_words, tmp ); \
1457 trie->wordinfo[curword].prev = 0; \
1458 trie->wordinfo[curword].len = wordlen; \
1459 trie->wordinfo[curword].accept = state; \
1461 if ( noper_next < tail ) { \
1463 trie->jump = (U16 *) PerlMemShared_calloc( word_count + 1, sizeof(U16) ); \
1464 trie->jump[curword] = (U16)(noper_next - convert); \
1466 jumper = noper_next; \
1468 nextbranch= regnext(cur); \
1472 /* It's a dupe. Pre-insert into the wordinfo[].prev */\
1473 /* chain, so that when the bits of chain are later */\
1474 /* linked together, the dups appear in the chain */\
1475 trie->wordinfo[curword].prev = trie->wordinfo[dupe].prev; \
1476 trie->wordinfo[dupe].prev = curword; \
1478 /* we haven't inserted this word yet. */ \
1479 trie->states[ state ].wordnum = curword; \
1484 #define TRIE_TRANS_STATE(state,base,ucharcount,charid,special) \
1485 ( ( base + charid >= ucharcount \
1486 && base + charid < ubound \
1487 && state == trie->trans[ base - ucharcount + charid ].check \
1488 && trie->trans[ base - ucharcount + charid ].next ) \
1489 ? trie->trans[ base - ucharcount + charid ].next \
1490 : ( state==1 ? special : 0 ) \
1494 #define MADE_JUMP_TRIE 2
1495 #define MADE_EXACT_TRIE 4
1498 S_make_trie(pTHX_ RExC_state_t *pRExC_state, regnode *startbranch, regnode *first, regnode *last, regnode *tail, U32 word_count, U32 flags, U32 depth)
1501 /* first pass, loop through and scan words */
1502 reg_trie_data *trie;
1503 HV *widecharmap = NULL;
1504 AV *revcharmap = newAV();
1506 const U32 uniflags = UTF8_ALLOW_DEFAULT;
1511 regnode *jumper = NULL;
1512 regnode *nextbranch = NULL;
1513 regnode *convert = NULL;
1514 U32 *prev_states; /* temp array mapping each state to previous one */
1515 /* we just use folder as a flag in utf8 */
1516 const U8 * folder = NULL;
1519 const U32 data_slot = add_data( pRExC_state, 4, "tuuu" );
1520 AV *trie_words = NULL;
1521 /* along with revcharmap, this only used during construction but both are
1522 * useful during debugging so we store them in the struct when debugging.
1525 const U32 data_slot = add_data( pRExC_state, 2, "tu" );
1526 STRLEN trie_charcount=0;
1528 SV *re_trie_maxbuff;
1529 GET_RE_DEBUG_FLAGS_DECL;
1531 PERL_ARGS_ASSERT_MAKE_TRIE;
1533 PERL_UNUSED_ARG(depth);
1540 case EXACTFU_TRICKYFOLD:
1541 case EXACTFU: folder = PL_fold_latin1; break;
1542 case EXACTF: folder = PL_fold; break;
1543 case EXACTFL: folder = PL_fold_locale; break;
1544 default: Perl_croak( aTHX_ "panic! In trie construction, unknown node type %u %s", (unsigned) flags, PL_reg_name[flags] );
1547 trie = (reg_trie_data *) PerlMemShared_calloc( 1, sizeof(reg_trie_data) );
1549 trie->startstate = 1;
1550 trie->wordcount = word_count;
1551 RExC_rxi->data->data[ data_slot ] = (void*)trie;
1552 trie->charmap = (U16 *) PerlMemShared_calloc( 256, sizeof(U16) );
1554 trie->bitmap = (char *) PerlMemShared_calloc( ANYOF_BITMAP_SIZE, 1 );
1555 trie->wordinfo = (reg_trie_wordinfo *) PerlMemShared_calloc(
1556 trie->wordcount+1, sizeof(reg_trie_wordinfo));
1559 trie_words = newAV();
1562 re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1);
1563 if (!SvIOK(re_trie_maxbuff)) {
1564 sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
1566 DEBUG_TRIE_COMPILE_r({
1567 PerlIO_printf( Perl_debug_log,
1568 "%*smake_trie start==%d, first==%d, last==%d, tail==%d depth=%d\n",
1569 (int)depth * 2 + 2, "",
1570 REG_NODE_NUM(startbranch),REG_NODE_NUM(first),
1571 REG_NODE_NUM(last), REG_NODE_NUM(tail),
1575 /* Find the node we are going to overwrite */
1576 if ( first == startbranch && OP( last ) != BRANCH ) {
1577 /* whole branch chain */
1580 /* branch sub-chain */
1581 convert = NEXTOPER( first );
1584 /* -- First loop and Setup --
1586 We first traverse the branches and scan each word to determine if it
1587 contains widechars, and how many unique chars there are, this is
1588 important as we have to build a table with at least as many columns as we
1591 We use an array of integers to represent the character codes 0..255
1592 (trie->charmap) and we use a an HV* to store Unicode characters. We use the
1593 native representation of the character value as the key and IV's for the
1596 *TODO* If we keep track of how many times each character is used we can
1597 remap the columns so that the table compression later on is more
1598 efficient in terms of memory by ensuring the most common value is in the
1599 middle and the least common are on the outside. IMO this would be better
1600 than a most to least common mapping as theres a decent chance the most
1601 common letter will share a node with the least common, meaning the node
1602 will not be compressible. With a middle is most common approach the worst
1603 case is when we have the least common nodes twice.
1607 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
1608 regnode *noper = NEXTOPER( cur );
1609 const U8 *uc = (U8*)STRING( noper );
1610 const U8 *e = uc + STR_LEN( noper );
1612 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1614 const U8 *scan = (U8*)NULL;
1615 U32 wordlen = 0; /* required init */
1617 bool set_bit = trie->bitmap ? 1 : 0; /*store the first char in the bitmap?*/
1619 if (OP(noper) == NOTHING) {
1620 regnode *noper_next= regnext(noper);
1621 if (noper_next != tail && OP(noper_next) == flags) {
1623 uc= (U8*)STRING(noper);
1624 e= uc + STR_LEN(noper);
1625 trie->minlen= STR_LEN(noper);
1632 if ( set_bit ) { /* bitmap only alloced when !(UTF&&Folding) */
1633 TRIE_BITMAP_SET(trie,*uc); /* store the raw first byte
1634 regardless of encoding */
1635 if (OP( noper ) == EXACTFU_SS) {
1636 /* false positives are ok, so just set this */
1637 TRIE_BITMAP_SET(trie,0xDF);
1640 for ( ; uc < e ; uc += len ) {
1641 TRIE_CHARCOUNT(trie)++;
1646 U8 folded= folder[ (U8) uvc ];
1647 if ( !trie->charmap[ folded ] ) {
1648 trie->charmap[ folded ]=( ++trie->uniquecharcount );
1649 TRIE_STORE_REVCHAR( folded );
1652 if ( !trie->charmap[ uvc ] ) {
1653 trie->charmap[ uvc ]=( ++trie->uniquecharcount );
1654 TRIE_STORE_REVCHAR( uvc );
1657 /* store the codepoint in the bitmap, and its folded
1659 TRIE_BITMAP_SET(trie, uvc);
1661 /* store the folded codepoint */
1662 if ( folder ) TRIE_BITMAP_SET(trie, folder[(U8) uvc ]);
1665 /* store first byte of utf8 representation of
1666 variant codepoints */
1667 if (! UNI_IS_INVARIANT(uvc)) {
1668 TRIE_BITMAP_SET(trie, UTF8_TWO_BYTE_HI(uvc));
1671 set_bit = 0; /* We've done our bit :-) */
1676 widecharmap = newHV();
1678 svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 1 );
1681 Perl_croak( aTHX_ "error creating/fetching widecharmap entry for 0x%"UVXf, uvc );
1683 if ( !SvTRUE( *svpp ) ) {
1684 sv_setiv( *svpp, ++trie->uniquecharcount );
1685 TRIE_STORE_REVCHAR(uvc);
1689 if( cur == first ) {
1690 trie->minlen = chars;
1691 trie->maxlen = chars;
1692 } else if (chars < trie->minlen) {
1693 trie->minlen = chars;
1694 } else if (chars > trie->maxlen) {
1695 trie->maxlen = chars;
1697 if (OP( noper ) == EXACTFU_SS) {
1698 /* XXX: workaround - 'ss' could match "\x{DF}" so minlen could be 1 and not 2*/
1699 if (trie->minlen > 1)
1702 if (OP( noper ) == EXACTFU_TRICKYFOLD) {
1703 /* XXX: workround - things like "\x{1FBE}\x{0308}\x{0301}" can match "\x{0390}"
1704 * - We assume that any such sequence might match a 2 byte string */
1705 if (trie->minlen > 2 )
1709 } /* end first pass */
1710 DEBUG_TRIE_COMPILE_r(
1711 PerlIO_printf( Perl_debug_log, "%*sTRIE(%s): W:%d C:%d Uq:%d Min:%d Max:%d\n",
1712 (int)depth * 2 + 2,"",
1713 ( widecharmap ? "UTF8" : "NATIVE" ), (int)word_count,
1714 (int)TRIE_CHARCOUNT(trie), trie->uniquecharcount,
1715 (int)trie->minlen, (int)trie->maxlen )
1719 We now know what we are dealing with in terms of unique chars and
1720 string sizes so we can calculate how much memory a naive
1721 representation using a flat table will take. If it's over a reasonable
1722 limit (as specified by ${^RE_TRIE_MAXBUF}) we use a more memory
1723 conservative but potentially much slower representation using an array
1726 At the end we convert both representations into the same compressed
1727 form that will be used in regexec.c for matching with. The latter
1728 is a form that cannot be used to construct with but has memory
1729 properties similar to the list form and access properties similar
1730 to the table form making it both suitable for fast searches and
1731 small enough that its feasable to store for the duration of a program.
1733 See the comment in the code where the compressed table is produced
1734 inplace from the flat tabe representation for an explanation of how
1735 the compression works.
1740 Newx(prev_states, TRIE_CHARCOUNT(trie) + 2, U32);
1743 if ( (IV)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1) > SvIV(re_trie_maxbuff) ) {
1745 Second Pass -- Array Of Lists Representation
1747 Each state will be represented by a list of charid:state records
1748 (reg_trie_trans_le) the first such element holds the CUR and LEN
1749 points of the allocated array. (See defines above).
1751 We build the initial structure using the lists, and then convert
1752 it into the compressed table form which allows faster lookups
1753 (but cant be modified once converted).
1756 STRLEN transcount = 1;
1758 DEBUG_TRIE_COMPILE_MORE_r( PerlIO_printf( Perl_debug_log,
1759 "%*sCompiling trie using list compiler\n",
1760 (int)depth * 2 + 2, ""));
1762 trie->states = (reg_trie_state *)
1763 PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
1764 sizeof(reg_trie_state) );
1768 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
1770 regnode *noper = NEXTOPER( cur );
1771 U8 *uc = (U8*)STRING( noper );
1772 const U8 *e = uc + STR_LEN( noper );
1773 U32 state = 1; /* required init */
1774 U16 charid = 0; /* sanity init */
1775 U8 *scan = (U8*)NULL; /* sanity init */
1776 STRLEN foldlen = 0; /* required init */
1777 U32 wordlen = 0; /* required init */
1778 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1781 if (OP(noper) == NOTHING) {
1782 regnode *noper_next= regnext(noper);
1783 if (noper_next != tail && OP(noper_next) == flags) {
1785 uc= (U8*)STRING(noper);
1786 e= uc + STR_LEN(noper);
1790 if (OP(noper) != NOTHING) {
1791 for ( ; uc < e ; uc += len ) {
1796 charid = trie->charmap[ uvc ];
1798 SV** const svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 0);
1802 charid=(U16)SvIV( *svpp );
1805 /* charid is now 0 if we dont know the char read, or nonzero if we do */
1812 if ( !trie->states[ state ].trans.list ) {
1813 TRIE_LIST_NEW( state );
1815 for ( check = 1; check <= TRIE_LIST_USED( state ); check++ ) {
1816 if ( TRIE_LIST_ITEM( state, check ).forid == charid ) {
1817 newstate = TRIE_LIST_ITEM( state, check ).newstate;
1822 newstate = next_alloc++;
1823 prev_states[newstate] = state;
1824 TRIE_LIST_PUSH( state, charid, newstate );
1829 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %"IVdf, uvc );
1833 TRIE_HANDLE_WORD(state);
1835 } /* end second pass */
1837 /* next alloc is the NEXT state to be allocated */
1838 trie->statecount = next_alloc;
1839 trie->states = (reg_trie_state *)
1840 PerlMemShared_realloc( trie->states,
1842 * sizeof(reg_trie_state) );
1844 /* and now dump it out before we compress it */
1845 DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_list(trie, widecharmap,
1846 revcharmap, next_alloc,
1850 trie->trans = (reg_trie_trans *)
1851 PerlMemShared_calloc( transcount, sizeof(reg_trie_trans) );
1858 for( state=1 ; state < next_alloc ; state ++ ) {
1862 DEBUG_TRIE_COMPILE_MORE_r(
1863 PerlIO_printf( Perl_debug_log, "tp: %d zp: %d ",tp,zp)
1867 if (trie->states[state].trans.list) {
1868 U16 minid=TRIE_LIST_ITEM( state, 1).forid;
1872 for( idx = 2 ; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
1873 const U16 forid = TRIE_LIST_ITEM( state, idx).forid;
1874 if ( forid < minid ) {
1876 } else if ( forid > maxid ) {
1880 if ( transcount < tp + maxid - minid + 1) {
1882 trie->trans = (reg_trie_trans *)
1883 PerlMemShared_realloc( trie->trans,
1885 * sizeof(reg_trie_trans) );
1886 Zero( trie->trans + (transcount / 2), transcount / 2 , reg_trie_trans );
1888 base = trie->uniquecharcount + tp - minid;
1889 if ( maxid == minid ) {
1891 for ( ; zp < tp ; zp++ ) {
1892 if ( ! trie->trans[ zp ].next ) {
1893 base = trie->uniquecharcount + zp - minid;
1894 trie->trans[ zp ].next = TRIE_LIST_ITEM( state, 1).newstate;
1895 trie->trans[ zp ].check = state;
1901 trie->trans[ tp ].next = TRIE_LIST_ITEM( state, 1).newstate;
1902 trie->trans[ tp ].check = state;
1907 for ( idx=1; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
1908 const U32 tid = base - trie->uniquecharcount + TRIE_LIST_ITEM( state, idx ).forid;
1909 trie->trans[ tid ].next = TRIE_LIST_ITEM( state, idx ).newstate;
1910 trie->trans[ tid ].check = state;
1912 tp += ( maxid - minid + 1 );
1914 Safefree(trie->states[ state ].trans.list);
1917 DEBUG_TRIE_COMPILE_MORE_r(
1918 PerlIO_printf( Perl_debug_log, " base: %d\n",base);
1921 trie->states[ state ].trans.base=base;
1923 trie->lasttrans = tp + 1;
1927 Second Pass -- Flat Table Representation.
1929 we dont use the 0 slot of either trans[] or states[] so we add 1 to each.
1930 We know that we will need Charcount+1 trans at most to store the data
1931 (one row per char at worst case) So we preallocate both structures
1932 assuming worst case.
1934 We then construct the trie using only the .next slots of the entry
1937 We use the .check field of the first entry of the node temporarily to
1938 make compression both faster and easier by keeping track of how many non
1939 zero fields are in the node.
1941 Since trans are numbered from 1 any 0 pointer in the table is a FAIL
1944 There are two terms at use here: state as a TRIE_NODEIDX() which is a
1945 number representing the first entry of the node, and state as a
1946 TRIE_NODENUM() which is the trans number. state 1 is TRIE_NODEIDX(1) and
1947 TRIE_NODENUM(1), state 2 is TRIE_NODEIDX(2) and TRIE_NODENUM(3) if there
1948 are 2 entrys per node. eg:
1956 The table is internally in the right hand, idx form. However as we also
1957 have to deal with the states array which is indexed by nodenum we have to
1958 use TRIE_NODENUM() to convert.
1961 DEBUG_TRIE_COMPILE_MORE_r( PerlIO_printf( Perl_debug_log,
1962 "%*sCompiling trie using table compiler\n",
1963 (int)depth * 2 + 2, ""));
1965 trie->trans = (reg_trie_trans *)
1966 PerlMemShared_calloc( ( TRIE_CHARCOUNT(trie) + 1 )
1967 * trie->uniquecharcount + 1,
1968 sizeof(reg_trie_trans) );
1969 trie->states = (reg_trie_state *)
1970 PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
1971 sizeof(reg_trie_state) );
1972 next_alloc = trie->uniquecharcount + 1;
1975 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
1977 regnode *noper = NEXTOPER( cur );
1978 const U8 *uc = (U8*)STRING( noper );
1979 const U8 *e = uc + STR_LEN( noper );
1981 U32 state = 1; /* required init */
1983 U16 charid = 0; /* sanity init */
1984 U32 accept_state = 0; /* sanity init */
1985 U8 *scan = (U8*)NULL; /* sanity init */
1987 STRLEN foldlen = 0; /* required init */
1988 U32 wordlen = 0; /* required init */
1990 U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
1992 if (OP(noper) == NOTHING) {
1993 regnode *noper_next= regnext(noper);
1994 if (noper_next != tail && OP(noper_next) == flags) {
1996 uc= (U8*)STRING(noper);
1997 e= uc + STR_LEN(noper);
2001 if ( OP(noper) != NOTHING ) {
2002 for ( ; uc < e ; uc += len ) {
2007 charid = trie->charmap[ uvc ];
2009 SV* const * const svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 0);
2010 charid = svpp ? (U16)SvIV(*svpp) : 0;
2014 if ( !trie->trans[ state + charid ].next ) {
2015 trie->trans[ state + charid ].next = next_alloc;
2016 trie->trans[ state ].check++;
2017 prev_states[TRIE_NODENUM(next_alloc)]
2018 = TRIE_NODENUM(state);
2019 next_alloc += trie->uniquecharcount;
2021 state = trie->trans[ state + charid ].next;
2023 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %"IVdf, uvc );
2025 /* charid is now 0 if we dont know the char read, or nonzero if we do */
2028 accept_state = TRIE_NODENUM( state );
2029 TRIE_HANDLE_WORD(accept_state);
2031 } /* end second pass */
2033 /* and now dump it out before we compress it */
2034 DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_table(trie, widecharmap,
2036 next_alloc, depth+1));
2040 * Inplace compress the table.*
2042 For sparse data sets the table constructed by the trie algorithm will
2043 be mostly 0/FAIL transitions or to put it another way mostly empty.
2044 (Note that leaf nodes will not contain any transitions.)
2046 This algorithm compresses the tables by eliminating most such
2047 transitions, at the cost of a modest bit of extra work during lookup:
2049 - Each states[] entry contains a .base field which indicates the
2050 index in the state[] array wheres its transition data is stored.
2052 - If .base is 0 there are no valid transitions from that node.
2054 - If .base is nonzero then charid is added to it to find an entry in
2057 -If trans[states[state].base+charid].check!=state then the
2058 transition is taken to be a 0/Fail transition. Thus if there are fail
2059 transitions at the front of the node then the .base offset will point
2060 somewhere inside the previous nodes data (or maybe even into a node
2061 even earlier), but the .check field determines if the transition is
2065 The following process inplace converts the table to the compressed
2066 table: We first do not compress the root node 1,and mark all its
2067 .check pointers as 1 and set its .base pointer as 1 as well. This
2068 allows us to do a DFA construction from the compressed table later,
2069 and ensures that any .base pointers we calculate later are greater
2072 - We set 'pos' to indicate the first entry of the second node.
2074 - We then iterate over the columns of the node, finding the first and
2075 last used entry at l and m. We then copy l..m into pos..(pos+m-l),
2076 and set the .check pointers accordingly, and advance pos
2077 appropriately and repreat for the next node. Note that when we copy
2078 the next pointers we have to convert them from the original
2079 NODEIDX form to NODENUM form as the former is not valid post
2082 - If a node has no transitions used we mark its base as 0 and do not
2083 advance the pos pointer.
2085 - If a node only has one transition we use a second pointer into the
2086 structure to fill in allocated fail transitions from other states.
2087 This pointer is independent of the main pointer and scans forward
2088 looking for null transitions that are allocated to a state. When it
2089 finds one it writes the single transition into the "hole". If the
2090 pointer doesnt find one the single transition is appended as normal.
2092 - Once compressed we can Renew/realloc the structures to release the
2095 See "Table-Compression Methods" in sec 3.9 of the Red Dragon,
2096 specifically Fig 3.47 and the associated pseudocode.
2100 const U32 laststate = TRIE_NODENUM( next_alloc );
2103 trie->statecount = laststate;
2105 for ( state = 1 ; state < laststate ; state++ ) {
2107 const U32 stateidx = TRIE_NODEIDX( state );
2108 const U32 o_used = trie->trans[ stateidx ].check;
2109 U32 used = trie->trans[ stateidx ].check;
2110 trie->trans[ stateidx ].check = 0;
2112 for ( charid = 0 ; used && charid < trie->uniquecharcount ; charid++ ) {
2113 if ( flag || trie->trans[ stateidx + charid ].next ) {
2114 if ( trie->trans[ stateidx + charid ].next ) {
2116 for ( ; zp < pos ; zp++ ) {
2117 if ( ! trie->trans[ zp ].next ) {
2121 trie->states[ state ].trans.base = zp + trie->uniquecharcount - charid ;
2122 trie->trans[ zp ].next = SAFE_TRIE_NODENUM( trie->trans[ stateidx + charid ].next );
2123 trie->trans[ zp ].check = state;
2124 if ( ++zp > pos ) pos = zp;
2131 trie->states[ state ].trans.base = pos + trie->uniquecharcount - charid ;
2133 trie->trans[ pos ].next = SAFE_TRIE_NODENUM( trie->trans[ stateidx + charid ].next );
2134 trie->trans[ pos ].check = state;
2139 trie->lasttrans = pos + 1;
2140 trie->states = (reg_trie_state *)
2141 PerlMemShared_realloc( trie->states, laststate
2142 * sizeof(reg_trie_state) );
2143 DEBUG_TRIE_COMPILE_MORE_r(
2144 PerlIO_printf( Perl_debug_log,
2145 "%*sAlloc: %d Orig: %"IVdf" elements, Final:%"IVdf". Savings of %%%5.2f\n",
2146 (int)depth * 2 + 2,"",
2147 (int)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1 ),
2150 ( ( next_alloc - pos ) * 100 ) / (double)next_alloc );
2153 } /* end table compress */
2155 DEBUG_TRIE_COMPILE_MORE_r(
2156 PerlIO_printf(Perl_debug_log, "%*sStatecount:%"UVxf" Lasttrans:%"UVxf"\n",
2157 (int)depth * 2 + 2, "",
2158 (UV)trie->statecount,
2159 (UV)trie->lasttrans)
2161 /* resize the trans array to remove unused space */
2162 trie->trans = (reg_trie_trans *)
2163 PerlMemShared_realloc( trie->trans, trie->lasttrans
2164 * sizeof(reg_trie_trans) );
2166 { /* Modify the program and insert the new TRIE node */
2167 U8 nodetype =(U8)(flags & 0xFF);
2171 regnode *optimize = NULL;
2172 #ifdef RE_TRACK_PATTERN_OFFSETS
2175 U32 mjd_nodelen = 0;
2176 #endif /* RE_TRACK_PATTERN_OFFSETS */
2177 #endif /* DEBUGGING */
2179 This means we convert either the first branch or the first Exact,
2180 depending on whether the thing following (in 'last') is a branch
2181 or not and whther first is the startbranch (ie is it a sub part of
2182 the alternation or is it the whole thing.)
2183 Assuming its a sub part we convert the EXACT otherwise we convert
2184 the whole branch sequence, including the first.
2186 /* Find the node we are going to overwrite */
2187 if ( first != startbranch || OP( last ) == BRANCH ) {
2188 /* branch sub-chain */
2189 NEXT_OFF( first ) = (U16)(last - first);
2190 #ifdef RE_TRACK_PATTERN_OFFSETS
2192 mjd_offset= Node_Offset((convert));
2193 mjd_nodelen= Node_Length((convert));
2196 /* whole branch chain */
2198 #ifdef RE_TRACK_PATTERN_OFFSETS
2201 const regnode *nop = NEXTOPER( convert );
2202 mjd_offset= Node_Offset((nop));
2203 mjd_nodelen= Node_Length((nop));
2207 PerlIO_printf(Perl_debug_log, "%*sMJD offset:%"UVuf" MJD length:%"UVuf"\n",
2208 (int)depth * 2 + 2, "",
2209 (UV)mjd_offset, (UV)mjd_nodelen)
2212 /* But first we check to see if there is a common prefix we can
2213 split out as an EXACT and put in front of the TRIE node. */
2214 trie->startstate= 1;
2215 if ( trie->bitmap && !widecharmap && !trie->jump ) {
2217 for ( state = 1 ; state < trie->statecount-1 ; state++ ) {
2221 const U32 base = trie->states[ state ].trans.base;
2223 if ( trie->states[state].wordnum )
2226 for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) {
2227 if ( ( base + ofs >= trie->uniquecharcount ) &&
2228 ( base + ofs - trie->uniquecharcount < trie->lasttrans ) &&
2229 trie->trans[ base + ofs - trie->uniquecharcount ].check == state )
2231 if ( ++count > 1 ) {
2232 SV **tmp = av_fetch( revcharmap, ofs, 0);
2233 const U8 *ch = (U8*)SvPV_nolen_const( *tmp );
2234 if ( state == 1 ) break;
2236 Zero(trie->bitmap, ANYOF_BITMAP_SIZE, char);
2238 PerlIO_printf(Perl_debug_log,
2239 "%*sNew Start State=%"UVuf" Class: [",
2240 (int)depth * 2 + 2, "",
2243 SV ** const tmp = av_fetch( revcharmap, idx, 0);
2244 const U8 * const ch = (U8*)SvPV_nolen_const( *tmp );
2246 TRIE_BITMAP_SET(trie,*ch);
2248 TRIE_BITMAP_SET(trie, folder[ *ch ]);
2250 PerlIO_printf(Perl_debug_log, "%s", (char*)ch)
2254 TRIE_BITMAP_SET(trie,*ch);
2256 TRIE_BITMAP_SET(trie,folder[ *ch ]);
2257 DEBUG_OPTIMISE_r(PerlIO_printf( Perl_debug_log,"%s", ch));
2263 SV **tmp = av_fetch( revcharmap, idx, 0);
2265 char *ch = SvPV( *tmp, len );
2267 SV *sv=sv_newmortal();
2268 PerlIO_printf( Perl_debug_log,
2269 "%*sPrefix State: %"UVuf" Idx:%"UVuf" Char='%s'\n",
2270 (int)depth * 2 + 2, "",
2272 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 6,
2273 PL_colors[0], PL_colors[1],
2274 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
2275 PERL_PV_ESCAPE_FIRSTCHAR
2280 OP( convert ) = nodetype;
2281 str=STRING(convert);
2284 STR_LEN(convert) += len;
2290 DEBUG_OPTIMISE_r(PerlIO_printf( Perl_debug_log,"]\n"));
2295 trie->prefixlen = (state-1);
2297 regnode *n = convert+NODE_SZ_STR(convert);
2298 NEXT_OFF(convert) = NODE_SZ_STR(convert);
2299 trie->startstate = state;
2300 trie->minlen -= (state - 1);
2301 trie->maxlen -= (state - 1);
2303 /* At least the UNICOS C compiler choked on this
2304 * being argument to DEBUG_r(), so let's just have
2307 #ifdef PERL_EXT_RE_BUILD
2313 regnode *fix = convert;
2314 U32 word = trie->wordcount;
2316 Set_Node_Offset_Length(convert, mjd_offset, state - 1);
2317 while( ++fix < n ) {
2318 Set_Node_Offset_Length(fix, 0, 0);
2321 SV ** const tmp = av_fetch( trie_words, word, 0 );
2323 if ( STR_LEN(convert) <= SvCUR(*tmp) )
2324 sv_chop(*tmp, SvPV_nolen(*tmp) + STR_LEN(convert));
2326 sv_chop(*tmp, SvPV_nolen(*tmp) + SvCUR(*tmp));
2334 NEXT_OFF(convert) = (U16)(tail - convert);
2335 DEBUG_r(optimize= n);
2341 if ( trie->maxlen ) {
2342 NEXT_OFF( convert ) = (U16)(tail - convert);
2343 ARG_SET( convert, data_slot );
2344 /* Store the offset to the first unabsorbed branch in
2345 jump[0], which is otherwise unused by the jump logic.
2346 We use this when dumping a trie and during optimisation. */
2348 trie->jump[0] = (U16)(nextbranch - convert);
2350 /* If the start state is not accepting (meaning there is no empty string/NOTHING)
2351 * and there is a bitmap
2352 * and the first "jump target" node we found leaves enough room
2353 * then convert the TRIE node into a TRIEC node, with the bitmap
2354 * embedded inline in the opcode - this is hypothetically faster.
2356 if ( !trie->states[trie->startstate].wordnum
2358 && ( (char *)jumper - (char *)convert) >= (int)sizeof(struct regnode_charclass) )
2360 OP( convert ) = TRIEC;
2361 Copy(trie->bitmap, ((struct regnode_charclass *)convert)->bitmap, ANYOF_BITMAP_SIZE, char);
2362 PerlMemShared_free(trie->bitmap);
2365 OP( convert ) = TRIE;
2367 /* store the type in the flags */
2368 convert->flags = nodetype;
2372 + regarglen[ OP( convert ) ];
2374 /* XXX We really should free up the resource in trie now,
2375 as we won't use them - (which resources?) dmq */
2377 /* needed for dumping*/
2378 DEBUG_r(if (optimize) {
2379 regnode *opt = convert;
2381 while ( ++opt < optimize) {
2382 Set_Node_Offset_Length(opt,0,0);
2385 Try to clean up some of the debris left after the
2388 while( optimize < jumper ) {
2389 mjd_nodelen += Node_Length((optimize));
2390 OP( optimize ) = OPTIMIZED;
2391 Set_Node_Offset_Length(optimize,0,0);
2394 Set_Node_Offset_Length(convert,mjd_offset,mjd_nodelen);
2396 } /* end node insert */
2398 /* Finish populating the prev field of the wordinfo array. Walk back
2399 * from each accept state until we find another accept state, and if
2400 * so, point the first word's .prev field at the second word. If the
2401 * second already has a .prev field set, stop now. This will be the
2402 * case either if we've already processed that word's accept state,
2403 * or that state had multiple words, and the overspill words were
2404 * already linked up earlier.
2411 for (word=1; word <= trie->wordcount; word++) {
2413 if (trie->wordinfo[word].prev)
2415 state = trie->wordinfo[word].accept;
2417 state = prev_states[state];
2420 prev = trie->states[state].wordnum;
2424 trie->wordinfo[word].prev = prev;
2426 Safefree(prev_states);
2430 /* and now dump out the compressed format */
2431 DEBUG_TRIE_COMPILE_r(dump_trie(trie, widecharmap, revcharmap, depth+1));
2433 RExC_rxi->data->data[ data_slot + 1 ] = (void*)widecharmap;
2435 RExC_rxi->data->data[ data_slot + TRIE_WORDS_OFFSET ] = (void*)trie_words;
2436 RExC_rxi->data->data[ data_slot + 3 ] = (void*)revcharmap;
2438 SvREFCNT_dec(revcharmap);
2442 : trie->startstate>1
2448 S_make_trie_failtable(pTHX_ RExC_state_t *pRExC_state, regnode *source, regnode *stclass, U32 depth)
2450 /* The Trie is constructed and compressed now so we can build a fail array if it's needed
2452 This is basically the Aho-Corasick algorithm. Its from exercise 3.31 and 3.32 in the
2453 "Red Dragon" -- Compilers, principles, techniques, and tools. Aho, Sethi, Ullman 1985/88
2456 We find the fail state for each state in the trie, this state is the longest proper
2457 suffix of the current state's 'word' that is also a proper prefix of another word in our
2458 trie. State 1 represents the word '' and is thus the default fail state. This allows
2459 the DFA not to have to restart after its tried and failed a word at a given point, it
2460 simply continues as though it had been matching the other word in the first place.
2462 'abcdgu'=~/abcdefg|cdgu/
2463 When we get to 'd' we are still matching the first word, we would encounter 'g' which would
2464 fail, which would bring us to the state representing 'd' in the second word where we would
2465 try 'g' and succeed, proceeding to match 'cdgu'.
2467 /* add a fail transition */
2468 const U32 trie_offset = ARG(source);
2469 reg_trie_data *trie=(reg_trie_data *)RExC_rxi->data->data[trie_offset];
2471 const U32 ucharcount = trie->uniquecharcount;
2472 const U32 numstates = trie->statecount;
2473 const U32 ubound = trie->lasttrans + ucharcount;
2477 U32 base = trie->states[ 1 ].trans.base;
2480 const U32 data_slot = add_data( pRExC_state, 1, "T" );
2481 GET_RE_DEBUG_FLAGS_DECL;
2483 PERL_ARGS_ASSERT_MAKE_TRIE_FAILTABLE;
2485 PERL_UNUSED_ARG(depth);
2489 ARG_SET( stclass, data_slot );
2490 aho = (reg_ac_data *) PerlMemShared_calloc( 1, sizeof(reg_ac_data) );
2491 RExC_rxi->data->data[ data_slot ] = (void*)aho;
2492 aho->trie=trie_offset;
2493 aho->states=(reg_trie_state *)PerlMemShared_malloc( numstates * sizeof(reg_trie_state) );
2494 Copy( trie->states, aho->states, numstates, reg_trie_state );
2495 Newxz( q, numstates, U32);
2496 aho->fail = (U32 *) PerlMemShared_calloc( numstates, sizeof(U32) );
2499 /* initialize fail[0..1] to be 1 so that we always have
2500 a valid final fail state */
2501 fail[ 0 ] = fail[ 1 ] = 1;
2503 for ( charid = 0; charid < ucharcount ; charid++ ) {
2504 const U32 newstate = TRIE_TRANS_STATE( 1, base, ucharcount, charid, 0 );
2506 q[ q_write ] = newstate;
2507 /* set to point at the root */
2508 fail[ q[ q_write++ ] ]=1;
2511 while ( q_read < q_write) {
2512 const U32 cur = q[ q_read++ % numstates ];
2513 base = trie->states[ cur ].trans.base;
2515 for ( charid = 0 ; charid < ucharcount ; charid++ ) {
2516 const U32 ch_state = TRIE_TRANS_STATE( cur, base, ucharcount, charid, 1 );
2518 U32 fail_state = cur;
2521 fail_state = fail[ fail_state ];
2522 fail_base = aho->states[ fail_state ].trans.base;
2523 } while ( !TRIE_TRANS_STATE( fail_state, fail_base, ucharcount, charid, 1 ) );
2525 fail_state = TRIE_TRANS_STATE( fail_state, fail_base, ucharcount, charid, 1 );
2526 fail[ ch_state ] = fail_state;
2527 if ( !aho->states[ ch_state ].wordnum && aho->states[ fail_state ].wordnum )
2529 aho->states[ ch_state ].wordnum = aho->states[ fail_state ].wordnum;
2531 q[ q_write++ % numstates] = ch_state;
2535 /* restore fail[0..1] to 0 so that we "fall out" of the AC loop
2536 when we fail in state 1, this allows us to use the
2537 charclass scan to find a valid start char. This is based on the principle
2538 that theres a good chance the string being searched contains lots of stuff
2539 that cant be a start char.
2541 fail[ 0 ] = fail[ 1 ] = 0;
2542 DEBUG_TRIE_COMPILE_r({
2543 PerlIO_printf(Perl_debug_log,
2544 "%*sStclass Failtable (%"UVuf" states): 0",
2545 (int)(depth * 2), "", (UV)numstates
2547 for( q_read=1; q_read<numstates; q_read++ ) {
2548 PerlIO_printf(Perl_debug_log, ", %"UVuf, (UV)fail[q_read]);
2550 PerlIO_printf(Perl_debug_log, "\n");
2553 /*RExC_seen |= REG_SEEN_TRIEDFA;*/
2558 * There are strange code-generation bugs caused on sparc64 by gcc-2.95.2.
2559 * These need to be revisited when a newer toolchain becomes available.
2561 #if defined(__sparc64__) && defined(__GNUC__)
2562 # if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 96)
2563 # undef SPARC64_GCC_WORKAROUND
2564 # define SPARC64_GCC_WORKAROUND 1
2568 #define DEBUG_PEEP(str,scan,depth) \
2569 DEBUG_OPTIMISE_r({if (scan){ \
2570 SV * const mysv=sv_newmortal(); \
2571 regnode *Next = regnext(scan); \
2572 regprop(RExC_rx, mysv, scan); \
2573 PerlIO_printf(Perl_debug_log, "%*s" str ">%3d: %s (%d)\n", \
2574 (int)depth*2, "", REG_NODE_NUM(scan), SvPV_nolen_const(mysv),\
2575 Next ? (REG_NODE_NUM(Next)) : 0 ); \
2579 /* The below joins as many adjacent EXACTish nodes as possible into a single
2580 * one, and looks for problematic sequences of characters whose folds vs.
2581 * non-folds have sufficiently different lengths, that the optimizer would be
2582 * fooled into rejecting legitimate matches of them, and the trie construction
2583 * code can't cope with them. The joining is only done if:
2584 * 1) there is room in the current conglomerated node to entirely contain the
2586 * 2) they are the exact same node type
2588 * The adjacent nodes actually may be separated by NOTHING kind nodes, and
2589 * these get optimized out
2591 * If there are problematic code sequences, *min_subtract is set to the delta
2592 * that the minimum size of the node can be less than its actual size. And,
2593 * the node type of the result is changed to reflect that it contains these
2596 * And *has_exactf_sharp_s is set to indicate whether or not the node is EXACTF
2597 * and contains LATIN SMALL LETTER SHARP S
2599 * This is as good a place as any to discuss the design of handling these
2600 * problematic sequences. It's been wrong in Perl for a very long time. There
2601 * are three code points in Unicode whose folded lengths differ so much from
2602 * the un-folded lengths that it causes problems for the optimizer and trie
2603 * construction. Why only these are problematic, and not others where lengths
2604 * also differ is something I (khw) do not understand. New versions of Unicode
2605 * might add more such code points. Hopefully the logic in fold_grind.t that
2606 * figures out what to test (in part by verifying that each size-combination
2607 * gets tested) will catch any that do come along, so they can be added to the
2608 * special handling below. The chances of new ones are actually rather small,
2609 * as most, if not all, of the world's scripts that have casefolding have
2610 * already been encoded by Unicode. Also, a number of Unicode's decisions were
2611 * made to allow compatibility with pre-existing standards, and almost all of
2612 * those have already been dealt with. These would otherwise be the most
2613 * likely candidates for generating further tricky sequences. In other words,
2614 * Unicode by itself is unlikely to add new ones unless it is for compatibility
2615 * with pre-existing standards, and there aren't many of those left.
2617 * The previous designs for dealing with these involved assigning a special
2618 * node for them. This approach doesn't work, as evidenced by this example:
2619 * "\xDFs" =~ /s\xDF/ui # Used to fail before these patches
2620 * Both these fold to "sss", but if the pattern is parsed to create a node of
2621 * that would match just the \xDF, it won't be able to handle the case where a
2622 * successful match would have to cross the node's boundary. The new approach
2623 * that hopefully generally solves the problem generates an EXACTFU_SS node
2626 * There are a number of components to the approach (a lot of work for just
2627 * three code points!):
2628 * 1) This routine examines each EXACTFish node that could contain the
2629 * problematic sequences. It returns in *min_subtract how much to
2630 * subtract from the the actual length of the string to get a real minimum
2631 * for one that could match it. This number is usually 0 except for the
2632 * problematic sequences. This delta is used by the caller to adjust the
2633 * min length of the match, and the delta between min and max, so that the
2634 * optimizer doesn't reject these possibilities based on size constraints.
2635 * 2) These sequences are not currently correctly handled by the trie code
2636 * either, so it changes the joined node type to ops that are not handled
2637 * by trie's, those new ops being EXACTFU_SS and EXACTFU_TRICKYFOLD.
2638 * 3) This is sufficient for the two Greek sequences (described below), but
2639 * the one involving the Sharp s (\xDF) needs more. The node type
2640 * EXACTFU_SS is used for an EXACTFU node that contains at least one "ss"
2641 * sequence in it. For non-UTF-8 patterns and strings, this is the only
2642 * case where there is a possible fold length change. That means that a
2643 * regular EXACTFU node without UTF-8 involvement doesn't have to concern
2644 * itself with length changes, and so can be processed faster. regexec.c
2645 * takes advantage of this. Generally, an EXACTFish node that is in UTF-8
2646 * is pre-folded by regcomp.c. This saves effort in regex matching.
2647 * However, probably mostly for historical reasons, the pre-folding isn't
2648 * done for non-UTF8 patterns (and it can't be for EXACTF and EXACTFL
2649 * nodes, as what they fold to isn't known until runtime.) The fold
2650 * possibilities for the non-UTF8 patterns are quite simple, except for
2651 * the sharp s. All the ones that don't involve a UTF-8 target string
2652 * are members of a fold-pair, and arrays are set up for all of them
2653 * that quickly find the other member of the pair. It might actually
2654 * be faster to pre-fold these, but it isn't currently done, except for
2655 * the sharp s. Code elsewhere in this file makes sure that it gets
2656 * folded to 'ss', even if the pattern isn't UTF-8. This avoids the
2657 * issues described in the next item.
2658 * 4) A problem remains for the sharp s in EXACTF nodes. Whether it matches
2659 * 'ss' or not is not knowable at compile time. It will match iff the
2660 * target string is in UTF-8, unlike the EXACTFU nodes, where it always
2661 * matches; and the EXACTFL and EXACTFA nodes where it never does. Thus
2662 * it can't be folded to "ss" at compile time, unlike EXACTFU does as
2663 * described in item 3). An assumption that the optimizer part of
2664 * regexec.c (probably unwittingly) makes is that a character in the
2665 * pattern corresponds to at most a single character in the target string.
2666 * (And I do mean character, and not byte here, unlike other parts of the
2667 * documentation that have never been updated to account for multibyte
2668 * Unicode.) This assumption is wrong only in this case, as all other
2669 * cases are either 1-1 folds when no UTF-8 is involved; or is true by
2670 * virtue of having this file pre-fold UTF-8 patterns. I'm
2671 * reluctant to try to change this assumption, so instead the code punts.
2672 * This routine examines EXACTF nodes for the sharp s, and returns a
2673 * boolean indicating whether or not the node is an EXACTF node that
2674 * contains a sharp s. When it is true, the caller sets a flag that later
2675 * causes the optimizer in this file to not set values for the floating
2676 * and fixed string lengths, and thus avoids the optimizer code in
2677 * regexec.c that makes the invalid assumption. Thus, there is no
2678 * optimization based on string lengths for EXACTF nodes that contain the
2679 * sharp s. This only happens for /id rules (which means the pattern
2683 #define JOIN_EXACT(scan,min_subtract,has_exactf_sharp_s, flags) \
2684 if (PL_regkind[OP(scan)] == EXACT) \
2685 join_exact(pRExC_state,(scan),(min_subtract),has_exactf_sharp_s, (flags),NULL,depth+1)
2688 S_join_exact(pTHX_ RExC_state_t *pRExC_state, regnode *scan, UV *min_subtract, bool *has_exactf_sharp_s, U32 flags,regnode *val, U32 depth) {
2689 /* Merge several consecutive EXACTish nodes into one. */
2690 regnode *n = regnext(scan);
2692 regnode *next = scan + NODE_SZ_STR(scan);
2696 regnode *stop = scan;
2697 GET_RE_DEBUG_FLAGS_DECL;
2699 PERL_UNUSED_ARG(depth);
2702 PERL_ARGS_ASSERT_JOIN_EXACT;
2703 #ifndef EXPERIMENTAL_INPLACESCAN
2704 PERL_UNUSED_ARG(flags);
2705 PERL_UNUSED_ARG(val);
2707 DEBUG_PEEP("join",scan,depth);
2709 /* Look through the subsequent nodes in the chain. Skip NOTHING, merge
2710 * EXACT ones that are mergeable to the current one. */
2712 && (PL_regkind[OP(n)] == NOTHING
2713 || (stringok && OP(n) == OP(scan)))
2715 && NEXT_OFF(scan) + NEXT_OFF(n) < I16_MAX)
2718 if (OP(n) == TAIL || n > next)
2720 if (PL_regkind[OP(n)] == NOTHING) {
2721 DEBUG_PEEP("skip:",n,depth);
2722 NEXT_OFF(scan) += NEXT_OFF(n);
2723 next = n + NODE_STEP_REGNODE;
2730 else if (stringok) {
2731 const unsigned int oldl = STR_LEN(scan);
2732 regnode * const nnext = regnext(n);
2734 if (oldl + STR_LEN(n) > U8_MAX)
2737 DEBUG_PEEP("merg",n,depth);
2740 NEXT_OFF(scan) += NEXT_OFF(n);
2741 STR_LEN(scan) += STR_LEN(n);
2742 next = n + NODE_SZ_STR(n);
2743 /* Now we can overwrite *n : */
2744 Move(STRING(n), STRING(scan) + oldl, STR_LEN(n), char);
2752 #ifdef EXPERIMENTAL_INPLACESCAN
2753 if (flags && !NEXT_OFF(n)) {
2754 DEBUG_PEEP("atch", val, depth);
2755 if (reg_off_by_arg[OP(n)]) {
2756 ARG_SET(n, val - n);
2759 NEXT_OFF(n) = val - n;
2767 *has_exactf_sharp_s = FALSE;
2769 /* Here, all the adjacent mergeable EXACTish nodes have been merged. We
2770 * can now analyze for sequences of problematic code points. (Prior to
2771 * this final joining, sequences could have been split over boundaries, and
2772 * hence missed). The sequences only happen in folding, hence for any
2773 * non-EXACT EXACTish node */
2774 if (OP(scan) != EXACT) {
2776 U8 * s0 = (U8*) STRING(scan);
2777 U8 * const s_end = s0 + STR_LEN(scan);
2779 /* The below is perhaps overboard, but this allows us to save a test
2780 * each time through the loop at the expense of a mask. This is
2781 * because on both EBCDIC and ASCII machines, 'S' and 's' differ by a
2782 * single bit. On ASCII they are 32 apart; on EBCDIC, they are 64.
2783 * This uses an exclusive 'or' to find that bit and then inverts it to
2784 * form a mask, with just a single 0, in the bit position where 'S' and
2786 const U8 S_or_s_mask = (U8) ~ ('S' ^ 's');
2787 const U8 s_masked = 's' & S_or_s_mask;
2789 /* One pass is made over the node's string looking for all the
2790 * possibilities. to avoid some tests in the loop, there are two main
2791 * cases, for UTF-8 patterns (which can't have EXACTF nodes) and
2795 /* There are two problematic Greek code points in Unicode
2798 * U+0390 - GREEK SMALL LETTER IOTA WITH DIALYTIKA AND TONOS
2799 * U+03B0 - GREEK SMALL LETTER UPSILON WITH DIALYTIKA AND TONOS
2805 * U+03B9 U+0308 U+0301 0xCE 0xB9 0xCC 0x88 0xCC 0x81
2806 * U+03C5 U+0308 U+0301 0xCF 0x85 0xCC 0x88 0xCC 0x81
2808 * This means that in case-insensitive matching (or "loose
2809 * matching", as Unicode calls it), an EXACTF of length six (the
2810 * UTF-8 encoded byte length of the above casefolded versions) can
2811 * match a target string of length two (the byte length of UTF-8
2812 * encoded U+0390 or U+03B0). This would rather mess up the
2813 * minimum length computation. (there are other code points that
2814 * also fold to these two sequences, but the delta is smaller)
2816 * If these sequences are found, the minimum length is decreased by
2817 * four (six minus two).
2819 * Similarly, 'ss' may match the single char and byte LATIN SMALL
2820 * LETTER SHARP S. We decrease the min length by 1 for each
2821 * occurrence of 'ss' found */
2823 #ifdef EBCDIC /* RD tunifold greek 0390 and 03B0 */
2824 # define U390_first_byte 0xb4
2825 const U8 U390_tail[] = "\x68\xaf\x49\xaf\x42";
2826 # define U3B0_first_byte 0xb5
2827 const U8 U3B0_tail[] = "\x46\xaf\x49\xaf\x42";
2829 # define U390_first_byte 0xce
2830 const U8 U390_tail[] = "\xb9\xcc\x88\xcc\x81";
2831 # define U3B0_first_byte 0xcf
2832 const U8 U3B0_tail[] = "\x85\xcc\x88\xcc\x81";
2834 const U8 len = sizeof(U390_tail); /* (-1 for NUL; +1 for 1st byte;
2835 yields a net of 0 */
2836 /* Examine the string for one of the problematic sequences */
2838 s < s_end - 1; /* Can stop 1 before the end, as minimum length
2839 * sequence we are looking for is 2 */
2843 /* Look for the first byte in each problematic sequence */
2845 /* We don't have to worry about other things that fold to
2846 * 's' (such as the long s, U+017F), as all above-latin1
2847 * code points have been pre-folded */
2851 /* Current character is an 's' or 'S'. If next one is
2852 * as well, we have the dreaded sequence */
2853 if (((*(s+1) & S_or_s_mask) == s_masked)
2854 /* These two node types don't have special handling
2856 && OP(scan) != EXACTFL && OP(scan) != EXACTFA)
2859 OP(scan) = EXACTFU_SS;
2860 s++; /* No need to look at this character again */
2864 case U390_first_byte:
2865 if (s_end - s >= len
2867 /* The 1's are because are skipping comparing the
2869 && memEQ(s + 1, U390_tail, len - 1))
2871 goto greek_sequence;
2875 case U3B0_first_byte:
2876 if (! (s_end - s >= len
2877 && memEQ(s + 1, U3B0_tail, len - 1)))
2884 /* This can't currently be handled by trie's, so change
2885 * the node type to indicate this. If EXACTFA and
2886 * EXACTFL were ever to be handled by trie's, this
2887 * would have to be changed. If this node has already
2888 * been changed to EXACTFU_SS in this loop, leave it as
2889 * is. (I (khw) think it doesn't matter in regexec.c
2890 * for UTF patterns, but no need to change it */
2891 if (OP(scan) == EXACTFU) {
2892 OP(scan) = EXACTFU_TRICKYFOLD;
2894 s += 6; /* We already know what this sequence is. Skip
2900 else if (OP(scan) != EXACTFL && OP(scan) != EXACTFA) {
2902 /* Here, the pattern is not UTF-8. We need to look only for the
2903 * 'ss' sequence, and in the EXACTF case, the sharp s, which can be
2904 * in the final position. Otherwise we can stop looking 1 byte
2905 * earlier because have to find both the first and second 's' */
2906 const U8* upper = (OP(scan) == EXACTF) ? s_end : s_end -1;
2908 for (s = s0; s < upper; s++) {
2913 && ((*(s+1) & S_or_s_mask) == s_masked))
2917 /* EXACTF nodes need to know that the minimum
2918 * length changed so that a sharp s in the string
2919 * can match this ss in the pattern, but they
2920 * remain EXACTF nodes, as they are not trie'able,
2921 * so don't have to invent a new node type to
2922 * exclude them from the trie code */
2923 if (OP(scan) != EXACTF) {
2924 OP(scan) = EXACTFU_SS;
2929 case LATIN_SMALL_LETTER_SHARP_S:
2930 if (OP(scan) == EXACTF) {
2931 *has_exactf_sharp_s = TRUE;
2940 /* Allow dumping but overwriting the collection of skipped
2941 * ops and/or strings with fake optimized ops */
2942 n = scan + NODE_SZ_STR(scan);
2950 DEBUG_OPTIMISE_r(if (merged){DEBUG_PEEP("finl",scan,depth)});
2954 /* REx optimizer. Converts nodes into quicker variants "in place".
2955 Finds fixed substrings. */
2957 /* Stops at toplevel WHILEM as well as at "last". At end *scanp is set
2958 to the position after last scanned or to NULL. */
2960 #define INIT_AND_WITHP \
2961 assert(!and_withp); \
2962 Newx(and_withp,1,struct regnode_charclass_class); \
2963 SAVEFREEPV(and_withp)
2965 /* this is a chain of data about sub patterns we are processing that
2966 need to be handled separately/specially in study_chunk. Its so
2967 we can simulate recursion without losing state. */
2969 typedef struct scan_frame {
2970 regnode *last; /* last node to process in this frame */
2971 regnode *next; /* next node to process when last is reached */
2972 struct scan_frame *prev; /*previous frame*/
2973 I32 stop; /* what stopparen do we use */
2977 #define SCAN_COMMIT(s, data, m) scan_commit(s, data, m, is_inf)
2979 #define CASE_SYNST_FNC(nAmE) \
2981 if (flags & SCF_DO_STCLASS_AND) { \
2982 for (value = 0; value < 256; value++) \
2983 if (!is_ ## nAmE ## _cp(value)) \
2984 ANYOF_BITMAP_CLEAR(data->start_class, value); \
2987 for (value = 0; value < 256; value++) \
2988 if (is_ ## nAmE ## _cp(value)) \
2989 ANYOF_BITMAP_SET(data->start_class, value); \
2993 if (flags & SCF_DO_STCLASS_AND) { \
2994 for (value = 0; value < 256; value++) \
2995 if (is_ ## nAmE ## _cp(value)) \
2996 ANYOF_BITMAP_CLEAR(data->start_class, value); \
2999 for (value = 0; value < 256; value++) \
3000 if (!is_ ## nAmE ## _cp(value)) \
3001 ANYOF_BITMAP_SET(data->start_class, value); \
3008 S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp,
3009 I32 *minlenp, I32 *deltap,
3014 struct regnode_charclass_class *and_withp,
3015 U32 flags, U32 depth)
3016 /* scanp: Start here (read-write). */
3017 /* deltap: Write maxlen-minlen here. */
3018 /* last: Stop before this one. */
3019 /* data: string data about the pattern */
3020 /* stopparen: treat close N as END */
3021 /* recursed: which subroutines have we recursed into */
3022 /* and_withp: Valid if flags & SCF_DO_STCLASS_OR */
3025 I32 min = 0, pars = 0, code;
3026 regnode *scan = *scanp, *next;
3028 int is_inf = (flags & SCF_DO_SUBSTR) && (data->flags & SF_IS_INF);
3029 int is_inf_internal = 0; /* The studied chunk is infinite */
3030 I32 is_par = OP(scan) == OPEN ? ARG(scan) : 0;
3031 scan_data_t data_fake;
3032 SV *re_trie_maxbuff = NULL;
3033 regnode *first_non_open = scan;
3034 I32 stopmin = I32_MAX;
3035 scan_frame *frame = NULL;
3036 GET_RE_DEBUG_FLAGS_DECL;
3038 PERL_ARGS_ASSERT_STUDY_CHUNK;
3041 StructCopy(&zero_scan_data, &data_fake, scan_data_t);
3045 while (first_non_open && OP(first_non_open) == OPEN)
3046 first_non_open=regnext(first_non_open);
3051 while ( scan && OP(scan) != END && scan < last ){
3052 UV min_subtract = 0; /* How much to subtract from the minimum node
3053 length to get a real minimum (because the
3054 folded version may be shorter) */
3055 bool has_exactf_sharp_s = FALSE;
3056 /* Peephole optimizer: */
3057 DEBUG_STUDYDATA("Peep:", data,depth);
3058 DEBUG_PEEP("Peep",scan,depth);
3060 /* Its not clear to khw or hv why this is done here, and not in the
3061 * clauses that deal with EXACT nodes. khw's guess is that it's
3062 * because of a previous design */
3063 JOIN_EXACT(scan,&min_subtract, &has_exactf_sharp_s, 0);
3065 /* Follow the next-chain of the current node and optimize
3066 away all the NOTHINGs from it. */
3067 if (OP(scan) != CURLYX) {
3068 const int max = (reg_off_by_arg[OP(scan)]
3070 /* I32 may be smaller than U16 on CRAYs! */
3071 : (I32_MAX < U16_MAX ? I32_MAX : U16_MAX));
3072 int off = (reg_off_by_arg[OP(scan)] ? ARG(scan) : NEXT_OFF(scan));
3076 /* Skip NOTHING and LONGJMP. */
3077 while ((n = regnext(n))
3078 && ((PL_regkind[OP(n)] == NOTHING && (noff = NEXT_OFF(n)))
3079 || ((OP(n) == LONGJMP) && (noff = ARG(n))))
3080 && off + noff < max)
3082 if (reg_off_by_arg[OP(scan)])
3085 NEXT_OFF(scan) = off;
3090 /* The principal pseudo-switch. Cannot be a switch, since we
3091 look into several different things. */
3092 if (OP(scan) == BRANCH || OP(scan) == BRANCHJ
3093 || OP(scan) == IFTHEN) {
3094 next = regnext(scan);
3096 /* demq: the op(next)==code check is to see if we have "branch-branch" AFAICT */
3098 if (OP(next) == code || code == IFTHEN) {
3099 /* NOTE - There is similar code to this block below for handling
3100 TRIE nodes on a re-study. If you change stuff here check there
3102 I32 max1 = 0, min1 = I32_MAX, num = 0;
3103 struct regnode_charclass_class accum;
3104 regnode * const startbranch=scan;
3106 if (flags & SCF_DO_SUBSTR)
3107 SCAN_COMMIT(pRExC_state, data, minlenp); /* Cannot merge strings after this. */
3108 if (flags & SCF_DO_STCLASS)
3109 cl_init_zero(pRExC_state, &accum);
3111 while (OP(scan) == code) {
3112 I32 deltanext, minnext, f = 0, fake;
3113 struct regnode_charclass_class this_class;
3116 data_fake.flags = 0;
3118 data_fake.whilem_c = data->whilem_c;
3119 data_fake.last_closep = data->last_closep;
3122 data_fake.last_closep = &fake;
3124 data_fake.pos_delta = delta;
3125 next = regnext(scan);
3126 scan = NEXTOPER(scan);
3128 scan = NEXTOPER(scan);
3129 if (flags & SCF_DO_STCLASS) {
3130 cl_init(pRExC_state, &this_class);
3131 data_fake.start_class = &this_class;
3132 f = SCF_DO_STCLASS_AND;
3134 if (flags & SCF_WHILEM_VISITED_POS)
3135 f |= SCF_WHILEM_VISITED_POS;
3137 /* we suppose the run is continuous, last=next...*/
3138 minnext = study_chunk(pRExC_state, &scan, minlenp, &deltanext,
3140 stopparen, recursed, NULL, f,depth+1);
3143 if (max1 < minnext + deltanext)
3144 max1 = minnext + deltanext;
3145 if (deltanext == I32_MAX)
3146 is_inf = is_inf_internal = 1;
3148 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
3150 if (data_fake.flags & SCF_SEEN_ACCEPT) {
3151 if ( stopmin > minnext)
3152 stopmin = min + min1;
3153 flags &= ~SCF_DO_SUBSTR;
3155 data->flags |= SCF_SEEN_ACCEPT;
3158 if (data_fake.flags & SF_HAS_EVAL)
3159 data->flags |= SF_HAS_EVAL;
3160 data->whilem_c = data_fake.whilem_c;
3162 if (flags & SCF_DO_STCLASS)
3163 cl_or(pRExC_state, &accum, &this_class);
3165 if (code == IFTHEN && num < 2) /* Empty ELSE branch */
3167 if (flags & SCF_DO_SUBSTR) {
3168 data->pos_min += min1;
3169 data->pos_delta += max1 - min1;
3170 if (max1 != min1 || is_inf)
3171 data->longest = &(data->longest_float);
3174 delta += max1 - min1;
3175 if (flags & SCF_DO_STCLASS_OR) {
3176 cl_or(pRExC_state, data->start_class, &accum);
3178 cl_and(data->start_class, and_withp);
3179 flags &= ~SCF_DO_STCLASS;
3182 else if (flags & SCF_DO_STCLASS_AND) {
3184 cl_and(data->start_class, &accum);
3185 flags &= ~SCF_DO_STCLASS;
3188 /* Switch to OR mode: cache the old value of
3189 * data->start_class */
3191 StructCopy(data->start_class, and_withp,
3192 struct regnode_charclass_class);
3193 flags &= ~SCF_DO_STCLASS_AND;
3194 StructCopy(&accum, data->start_class,
3195 struct regnode_charclass_class);
3196 flags |= SCF_DO_STCLASS_OR;
3197 data->start_class->flags |= ANYOF_EOS;
3201 if (PERL_ENABLE_TRIE_OPTIMISATION && OP( startbranch ) == BRANCH ) {
3204 Assuming this was/is a branch we are dealing with: 'scan' now
3205 points at the item that follows the branch sequence, whatever
3206 it is. We now start at the beginning of the sequence and look
3213 which would be constructed from a pattern like /A|LIST|OF|WORDS/
3215 If we can find such a subsequence we need to turn the first
3216 element into a trie and then add the subsequent branch exact
3217 strings to the trie.
3221 1. patterns where the whole set of branches can be converted.
3223 2. patterns where only a subset can be converted.
3225 In case 1 we can replace the whole set with a single regop
3226 for the trie. In case 2 we need to keep the start and end
3229 'BRANCH EXACT; BRANCH EXACT; BRANCH X'
3230 becomes BRANCH TRIE; BRANCH X;
3232 There is an additional case, that being where there is a
3233 common prefix, which gets split out into an EXACT like node
3234 preceding the TRIE node.
3236 If x(1..n)==tail then we can do a simple trie, if not we make
3237 a "jump" trie, such that when we match the appropriate word
3238 we "jump" to the appropriate tail node. Essentially we turn
3239 a nested if into a case structure of sorts.
3244 if (!re_trie_maxbuff) {
3245 re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1);
3246 if (!SvIOK(re_trie_maxbuff))
3247 sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
3249 if ( SvIV(re_trie_maxbuff)>=0 ) {
3251 regnode *first = (regnode *)NULL;
3252 regnode *last = (regnode *)NULL;
3253 regnode *tail = scan;
3258 SV * const mysv = sv_newmortal(); /* for dumping */
3260 /* var tail is used because there may be a TAIL
3261 regop in the way. Ie, the exacts will point to the
3262 thing following the TAIL, but the last branch will
3263 point at the TAIL. So we advance tail. If we
3264 have nested (?:) we may have to move through several
3268 while ( OP( tail ) == TAIL ) {
3269 /* this is the TAIL generated by (?:) */
3270 tail = regnext( tail );
3274 DEBUG_TRIE_COMPILE_r({
3275 regprop(RExC_rx, mysv, tail );
3276 PerlIO_printf( Perl_debug_log, "%*s%s%s\n",
3277 (int)depth * 2 + 2, "",
3278 "Looking for TRIE'able sequences. Tail node is: ",
3279 SvPV_nolen_const( mysv )
3285 Step through the branches
3286 cur represents each branch,
3287 noper is the first thing to be matched as part of that branch
3288 noper_next is the regnext() of that node.
3290 We normally handle a case like this /FOO[xyz]|BAR[pqr]/
3291 via a "jump trie" but we also support building with NOJUMPTRIE,
3292 which restricts the trie logic to structures like /FOO|BAR/.
3294 If noper is a trieable nodetype then the branch is a possible optimization
3295 target. If we are building under NOJUMPTRIE then we require that noper_next
3296 is the same as scan (our current position in the regex program).
3298 Once we have two or more consecutive such branches we can create a
3299 trie of the EXACT's contents and stitch it in place into the program.
3301 If the sequence represents all of the branches in the alternation we
3302 replace the entire thing with a single TRIE node.
3304 Otherwise when it is a subsequence we need to stitch it in place and
3305 replace only the relevant branches. This means the first branch has
3306 to remain as it is used by the alternation logic, and its next pointer,
3307 and needs to be repointed at the item on the branch chain following
3308 the last branch we have optimized away.
3310 This could be either a BRANCH, in which case the subsequence is internal,
3311 or it could be the item following the branch sequence in which case the
3312 subsequence is at the end (which does not necessarily mean the first node
3313 is the start of the alternation).
3315 TRIE_TYPE(X) is a define which maps the optype to a trietype.
3318 ----------------+-----------
3322 EXACTFU_SS | EXACTFU
3323 EXACTFU_TRICKYFOLD | EXACTFU
3328 #define TRIE_TYPE(X) ( ( NOTHING == (X) ) ? NOTHING : \
3329 ( EXACT == (X) ) ? EXACT : \
3330 ( EXACTFU == (X) || EXACTFU_SS == (X) || EXACTFU_TRICKYFOLD == (X) ) ? EXACTFU : \
3333 /* dont use tail as the end marker for this traverse */
3334 for ( cur = startbranch ; cur != scan ; cur = regnext( cur ) ) {
3335 regnode * const noper = NEXTOPER( cur );
3336 U8 noper_type = OP( noper );
3337 U8 noper_trietype = TRIE_TYPE( noper_type );
3338 #if defined(DEBUGGING) || defined(NOJUMPTRIE)
3339 regnode * const noper_next = regnext( noper );
3340 U8 noper_next_type = (noper_next && noper_next != tail) ? OP(noper_next) : 0;
3341 U8 noper_next_trietype = (noper_next && noper_next != tail) ? TRIE_TYPE( noper_next_type ) :0;
3344 DEBUG_TRIE_COMPILE_r({
3345 regprop(RExC_rx, mysv, cur);
3346 PerlIO_printf( Perl_debug_log, "%*s- %s (%d)",
3347 (int)depth * 2 + 2,"", SvPV_nolen_const( mysv ), REG_NODE_NUM(cur) );
3349 regprop(RExC_rx, mysv, noper);
3350 PerlIO_printf( Perl_debug_log, " -> %s",
3351 SvPV_nolen_const(mysv));
3354 regprop(RExC_rx, mysv, noper_next );
3355 PerlIO_printf( Perl_debug_log,"\t=> %s\t",
3356 SvPV_nolen_const(mysv));
3358 PerlIO_printf( Perl_debug_log, "(First==%d,Last==%d,Cur==%d,tt==%s,nt==%s,nnt==%s)\n",
3359 REG_NODE_NUM(first), REG_NODE_NUM(last), REG_NODE_NUM(cur),
3360 PL_reg_name[trietype], PL_reg_name[noper_trietype], PL_reg_name[noper_next_trietype]
3364 /* Is noper a trieable nodetype that can be merged with the
3365 * current trie (if there is one)? */
3369 ( noper_trietype == NOTHING)
3370 || ( trietype == NOTHING )
3371 || ( trietype == noper_trietype )
3374 && noper_next == tail
3378 /* Handle mergable triable node
3379 * Either we are the first node in a new trieable sequence,
3380 * in which case we do some bookkeeping, otherwise we update
3381 * the end pointer. */
3384 trietype = noper_trietype;
3385 if ( noper_trietype == NOTHING ) {
3386 #if !defined(DEBUGGING) && !defined(NOJUMPTRIE)
3387 regnode * const noper_next = regnext( noper );
3388 U8 noper_next_type = (noper_next && noper_next!=tail) ? OP(noper_next) : 0;
3389 U8 noper_next_trietype = noper_next_type ? TRIE_TYPE( noper_next_type ) :0;
3392 if ( noper_next_trietype )
3393 trietype = noper_next_trietype;
3396 if ( trietype == NOTHING )
3397 trietype = noper_trietype;
3402 } /* end handle mergable triable node */
3404 /* handle unmergable node -
3405 * noper may either be a triable node which can not be tried
3406 * together with the current trie, or a non triable node */
3408 /* If last is set and trietype is not NOTHING then we have found
3409 * at least two triable branch sequences in a row of a similar
3410 * trietype so we can turn them into a trie. If/when we
3411 * allow NOTHING to start a trie sequence this condition will be
3412 * required, and it isn't expensive so we leave it in for now. */
3413 if ( trietype != NOTHING )
3414 make_trie( pRExC_state,
3415 startbranch, first, cur, tail, count,
3416 trietype, depth+1 );
3417 last = NULL; /* note: we clear/update first, trietype etc below, so we dont do it here */
3421 && noper_next == tail
3424 /* noper is triable, so we can start a new trie sequence */
3427 trietype = noper_trietype;
3429 /* if we already saw a first but the current node is not triable then we have
3430 * to reset the first information. */
3435 } /* end handle unmergable node */
3436 } /* loop over branches */
3437 DEBUG_TRIE_COMPILE_r({
3438 regprop(RExC_rx, mysv, cur);
3439 PerlIO_printf( Perl_debug_log,
3440 "%*s- %s (%d) <SCAN FINISHED>\n", (int)depth * 2 + 2,
3441 "", SvPV_nolen_const( mysv ),REG_NODE_NUM(cur));
3445 if ( trietype != NOTHING ) {
3446 /* the last branch of the sequence was part of a trie,
3447 * so we have to construct it here outside of the loop
3449 made= make_trie( pRExC_state, startbranch, first, scan, tail, count, trietype, depth+1 );
3450 #ifdef TRIE_STUDY_OPT
3451 if ( ((made == MADE_EXACT_TRIE &&
3452 startbranch == first)
3453 || ( first_non_open == first )) &&
3455 flags |= SCF_TRIE_RESTUDY;
3456 if ( startbranch == first
3459 RExC_seen &=~REG_TOP_LEVEL_BRANCHES;
3464 /* at this point we know whatever we have is a NOTHING sequence/branch
3465 * AND if 'startbranch' is 'first' then we can turn the whole thing into a NOTHING
3467 if ( startbranch == first ) {
3469 /* the entire thing is a NOTHING sequence, something like this:
3470 * (?:|) So we can turn it into a plain NOTHING op. */
3471 DEBUG_TRIE_COMPILE_r({
3472 regprop(RExC_rx, mysv, cur);
3473 PerlIO_printf( Perl_debug_log,
3474 "%*s- %s (%d) <NOTHING BRANCH SEQUENCE>\n", (int)depth * 2 + 2,
3475 "", SvPV_nolen_const( mysv ),REG_NODE_NUM(cur));
3478 OP(startbranch)= NOTHING;
3479 NEXT_OFF(startbranch)= tail - startbranch;
3480 for ( opt= startbranch + 1; opt < tail ; opt++ )
3484 } /* end if ( last) */
3485 } /* TRIE_MAXBUF is non zero */
3490 else if ( code == BRANCHJ ) { /* single branch is optimized. */
3491 scan = NEXTOPER(NEXTOPER(scan));
3492 } else /* single branch is optimized. */
3493 scan = NEXTOPER(scan);
3495 } else if (OP(scan) == SUSPEND || OP(scan) == GOSUB || OP(scan) == GOSTART) {
3496 scan_frame *newframe = NULL;
3501 if (OP(scan) != SUSPEND) {
3502 /* set the pointer */
3503 if (OP(scan) == GOSUB) {
3505 RExC_recurse[ARG2L(scan)] = scan;
3506 start = RExC_open_parens[paren-1];
3507 end = RExC_close_parens[paren-1];
3510 start = RExC_rxi->program + 1;
3514 Newxz(recursed, (((RExC_npar)>>3) +1), U8);
3515 SAVEFREEPV(recursed);
3517 if (!PAREN_TEST(recursed,paren+1)) {
3518 PAREN_SET(recursed,paren+1);
3519 Newx(newframe,1,scan_frame);
3521 if (flags & SCF_DO_SUBSTR) {
3522 SCAN_COMMIT(pRExC_state,data,minlenp);
3523 data->longest = &(data->longest_float);
3525 is_inf = is_inf_internal = 1;
3526 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
3527 cl_anything(pRExC_state, data->start_class);
3528 flags &= ~SCF_DO_STCLASS;
3531 Newx(newframe,1,scan_frame);
3534 end = regnext(scan);
3539 SAVEFREEPV(newframe);
3540 newframe->next = regnext(scan);
3541 newframe->last = last;
3542 newframe->stop = stopparen;
3543 newframe->prev = frame;
3553 else if (OP(scan) == EXACT) {
3554 I32 l = STR_LEN(scan);
3557 const U8 * const s = (U8*)STRING(scan);
3558 uc = utf8_to_uvchr_buf(s, s + l, NULL);
3559 l = utf8_length(s, s + l);
3561 uc = *((U8*)STRING(scan));
3564 if (flags & SCF_DO_SUBSTR) { /* Update longest substr. */
3565 /* The code below prefers earlier match for fixed
3566 offset, later match for variable offset. */
3567 if (data->last_end == -1) { /* Update the start info. */
3568 data->last_start_min = data->pos_min;
3569 data->last_start_max = is_inf
3570 ? I32_MAX : data->pos_min + data->pos_delta;
3572 sv_catpvn(data->last_found, STRING(scan), STR_LEN(scan));
3574 SvUTF8_on(data->last_found);
3576 SV * const sv = data->last_found;
3577 MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ?
3578 mg_find(sv, PERL_MAGIC_utf8) : NULL;
3579 if (mg && mg->mg_len >= 0)
3580 mg->mg_len += utf8_length((U8*)STRING(scan),
3581 (U8*)STRING(scan)+STR_LEN(scan));
3583 data->last_end = data->pos_min + l;
3584 data->pos_min += l; /* As in the first entry. */
3585 data->flags &= ~SF_BEFORE_EOL;
3587 if (flags & SCF_DO_STCLASS_AND) {
3588 /* Check whether it is compatible with what we know already! */
3592 /* If compatible, we or it in below. It is compatible if is
3593 * in the bitmp and either 1) its bit or its fold is set, or 2)
3594 * it's for a locale. Even if there isn't unicode semantics
3595 * here, at runtime there may be because of matching against a
3596 * utf8 string, so accept a possible false positive for
3597 * latin1-range folds */
3599 (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
3600 && !ANYOF_BITMAP_TEST(data->start_class, uc)
3601 && (!(data->start_class->flags & ANYOF_LOC_NONBITMAP_FOLD)
3602 || !ANYOF_BITMAP_TEST(data->start_class, PL_fold_latin1[uc])))
3607 ANYOF_CLASS_ZERO(data->start_class);
3608 ANYOF_BITMAP_ZERO(data->start_class);
3610 ANYOF_BITMAP_SET(data->start_class, uc);
3611 else if (uc >= 0x100) {
3614 /* Some Unicode code points fold to the Latin1 range; as
3615 * XXX temporary code, instead of figuring out if this is
3616 * one, just assume it is and set all the start class bits
3617 * that could be some such above 255 code point's fold
3618 * which will generate fals positives. As the code
3619 * elsewhere that does compute the fold settles down, it
3620 * can be extracted out and re-used here */
3621 for (i = 0; i < 256; i++){
3622 if (HAS_NONLATIN1_FOLD_CLOSURE(i)) {
3623 ANYOF_BITMAP_SET(data->start_class, i);
3627 data->start_class->flags &= ~ANYOF_EOS;
3629 data->start_class->flags &= ~ANYOF_UNICODE_ALL;
3631 else if (flags & SCF_DO_STCLASS_OR) {
3632 /* false positive possible if the class is case-folded */
3634 ANYOF_BITMAP_SET(data->start_class, uc);
3636 data->start_class->flags |= ANYOF_UNICODE_ALL;
3637 data->start_class->flags &= ~ANYOF_EOS;
3638 cl_and(data->start_class, and_withp);
3640 flags &= ~SCF_DO_STCLASS;
3642 else if (PL_regkind[OP(scan)] == EXACT) { /* But OP != EXACT! */
3643 I32 l = STR_LEN(scan);
3644 UV uc = *((U8*)STRING(scan));
3646 /* Search for fixed substrings supports EXACT only. */
3647 if (flags & SCF_DO_SUBSTR) {
3649 SCAN_COMMIT(pRExC_state, data, minlenp);
3652 const U8 * const s = (U8 *)STRING(scan);
3653 uc = utf8_to_uvchr_buf(s, s + l, NULL);
3654 l = utf8_length(s, s + l);
3656 else if (has_exactf_sharp_s) {
3657 RExC_seen |= REG_SEEN_EXACTF_SHARP_S;
3659 min += l - min_subtract;
3663 delta += min_subtract;
3664 if (flags & SCF_DO_SUBSTR) {
3665 data->pos_min += l - min_subtract;
3666 if (data->pos_min < 0) {
3669 data->pos_delta += min_subtract;
3671 data->longest = &(data->longest_float);
3674 if (flags & SCF_DO_STCLASS_AND) {
3675 /* Check whether it is compatible with what we know already! */
3678 (!(data->start_class->flags & (ANYOF_CLASS | ANYOF_LOCALE))
3679 && !ANYOF_BITMAP_TEST(data->start_class, uc)
3680 && !ANYOF_BITMAP_TEST(data->start_class, PL_fold_latin1[uc])))
3684 ANYOF_CLASS_ZERO(data->start_class);
3685 ANYOF_BITMAP_ZERO(data->start_class);
3687 ANYOF_BITMAP_SET(data->start_class, uc);
3688 data->start_class->flags &= ~ANYOF_EOS;
3689 data->start_class->flags |= ANYOF_LOC_NONBITMAP_FOLD;
3690 if (OP(scan) == EXACTFL) {
3691 /* XXX This set is probably no longer necessary, and
3692 * probably wrong as LOCALE now is on in the initial
3694 data->start_class->flags |= ANYOF_LOCALE;
3698 /* Also set the other member of the fold pair. In case
3699 * that unicode semantics is called for at runtime, use
3700 * the full latin1 fold. (Can't do this for locale,
3701 * because not known until runtime) */
3702 ANYOF_BITMAP_SET(data->start_class, PL_fold_latin1[uc]);
3704 /* All other (EXACTFL handled above) folds except under
3705 * /iaa that include s, S, and sharp_s also may include
3707 if (OP(scan) != EXACTFA) {
3708 if (uc == 's' || uc == 'S') {
3709 ANYOF_BITMAP_SET(data->start_class,
3710 LATIN_SMALL_LETTER_SHARP_S);
3712 else if (uc == LATIN_SMALL_LETTER_SHARP_S) {
3713 ANYOF_BITMAP_SET(data->start_class, 's');
3714 ANYOF_BITMAP_SET(data->start_class, 'S');
3719 else if (uc >= 0x100) {
3721 for (i = 0; i < 256; i++){
3722 if (_HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(i)) {
3723 ANYOF_BITMAP_SET(data->start_class, i);
3728 else if (flags & SCF_DO_STCLASS_OR) {
3729 if (data->start_class->flags & ANYOF_LOC_NONBITMAP_FOLD) {
3730 /* false positive possible if the class is case-folded.
3731 Assume that the locale settings are the same... */
3733 ANYOF_BITMAP_SET(data->start_class, uc);
3734 if (OP(scan) != EXACTFL) {
3736 /* And set the other member of the fold pair, but
3737 * can't do that in locale because not known until
3739 ANYOF_BITMAP_SET(data->start_class,
3740 PL_fold_latin1[uc]);
3742 /* All folds except under /iaa that include s, S,
3743 * and sharp_s also may include the others */
3744 if (OP(scan) != EXACTFA) {
3745 if (uc == 's' || uc == 'S') {
3746 ANYOF_BITMAP_SET(data->start_class,
3747 LATIN_SMALL_LETTER_SHARP_S);
3749 else if (uc == LATIN_SMALL_LETTER_SHARP_S) {
3750 ANYOF_BITMAP_SET(data->start_class, 's');
3751 ANYOF_BITMAP_SET(data->start_class, 'S');
3756 data->start_class->flags &= ~ANYOF_EOS;
3758 cl_and(data->start_class, and_withp);
3760 flags &= ~SCF_DO_STCLASS;
3762 else if (REGNODE_VARIES(OP(scan))) {
3763 I32 mincount, maxcount, minnext, deltanext, fl = 0;
3764 I32 f = flags, pos_before = 0;
3765 regnode * const oscan = scan;
3766 struct regnode_charclass_class this_class;
3767 struct regnode_charclass_class *oclass = NULL;
3768 I32 next_is_eval = 0;
3770 switch (PL_regkind[OP(scan)]) {
3771 case WHILEM: /* End of (?:...)* . */
3772 scan = NEXTOPER(scan);
3775 if (flags & (SCF_DO_SUBSTR | SCF_DO_STCLASS)) {
3776 next = NEXTOPER(scan);
3777 if (OP(next) == EXACT || (flags & SCF_DO_STCLASS)) {
3779 maxcount = REG_INFTY;
3780 next = regnext(scan);
3781 scan = NEXTOPER(scan);
3785 if (flags & SCF_DO_SUBSTR)
3790 if (flags & SCF_DO_STCLASS) {
3792 maxcount = REG_INFTY;
3793 next = regnext(scan);
3794 scan = NEXTOPER(scan);
3797 is_inf = is_inf_internal = 1;
3798 scan = regnext(scan);
3799 if (flags & SCF_DO_SUBSTR) {
3800 SCAN_COMMIT(pRExC_state, data, minlenp); /* Cannot extend fixed substrings */
3801 data->longest = &(data->longest_float);
3803 goto optimize_curly_tail;
3805 if (stopparen>0 && (OP(scan)==CURLYN || OP(scan)==CURLYM)
3806 && (scan->flags == stopparen))
3811 mincount = ARG1(scan);
3812 maxcount = ARG2(scan);
3814 next = regnext(scan);
3815 if (OP(scan) == CURLYX) {
3816 I32 lp = (data ? *(data->last_closep) : 0);
3817 scan->flags = ((lp <= (I32)U8_MAX) ? (U8)lp : U8_MAX);
3819 scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
3820 next_is_eval = (OP(scan) == EVAL);
3822 if (flags & SCF_DO_SUBSTR) {
3823 if (mincount == 0) SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot extend fixed substrings */
3824 pos_before = data->pos_min;
3828 data->flags &= ~(SF_HAS_PAR|SF_IN_PAR|SF_HAS_EVAL);
3830 data->flags |= SF_IS_INF;
3832 if (flags & SCF_DO_STCLASS) {
3833 cl_init(pRExC_state, &this_class);
https://perl5.git.perl.org / perl5.git / blob