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
75 #undef PERL_IN_XSUB_RE
76 #define PERL_IN_XSUB_RE 1
78 #undef PERL_IN_XSUB_RE
80 #ifndef PERL_IN_XSUB_RE
85 #ifdef PERL_IN_XSUB_RE
87 EXTERN_C const struct regexp_engine my_reg_engine;
92 #include "dquote_static.c"
93 #include "charclass_invlists.h"
94 #include "inline_invlist.c"
95 #include "unicode_constants.h"
97 #define HAS_NONLATIN1_FOLD_CLOSURE(i) \
98 _HAS_NONLATIN1_FOLD_CLOSURE_ONLY_FOR_USE_BY_REGCOMP_DOT_C_AND_REGEXEC_DOT_C(i)
99 #define IS_NON_FINAL_FOLD(c) _IS_NON_FINAL_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c)
100 #define IS_IN_SOME_FOLD_L1(c) _IS_IN_SOME_FOLD_ONLY_FOR_USE_BY_REGCOMP_DOT_C(c)
103 #define STATIC static
107 struct RExC_state_t {
108 U32 flags; /* RXf_* are we folding, multilining? */
109 U32 pm_flags; /* PMf_* stuff from the calling PMOP */
110 char *precomp; /* uncompiled string. */
111 REGEXP *rx_sv; /* The SV that is the regexp. */
112 regexp *rx; /* perl core regexp structure */
113 regexp_internal *rxi; /* internal data for regexp object
115 char *start; /* Start of input for compile */
116 char *end; /* End of input for compile */
117 char *parse; /* Input-scan pointer. */
118 SSize_t whilem_seen; /* number of WHILEM in this expr */
119 regnode *emit_start; /* Start of emitted-code area */
120 regnode *emit_bound; /* First regnode outside of the
122 regnode *emit; /* Code-emit pointer; if = &emit_dummy,
123 implies compiling, so don't emit */
124 regnode_ssc emit_dummy; /* placeholder for emit to point to;
125 large enough for the largest
126 non-EXACTish node, so can use it as
128 I32 naughty; /* How bad is this pattern? */
129 I32 sawback; /* Did we see \1, ...? */
131 SSize_t size; /* Code size. */
132 I32 npar; /* Capture buffer count, (OPEN) plus
133 one. ("par" 0 is the whole
135 I32 nestroot; /* root parens we are in - used by
139 regnode **open_parens; /* pointers to open parens */
140 regnode **close_parens; /* pointers to close parens */
141 regnode *opend; /* END node in program */
142 I32 utf8; /* whether the pattern is utf8 or not */
143 I32 orig_utf8; /* whether the pattern was originally in utf8 */
144 /* XXX use this for future optimisation of case
145 * where pattern must be upgraded to utf8. */
146 I32 uni_semantics; /* If a d charset modifier should use unicode
147 rules, even if the pattern is not in
149 HV *paren_names; /* Paren names */
151 regnode **recurse; /* Recurse regops */
152 I32 recurse_count; /* Number of recurse regops */
153 U8 *study_chunk_recursed; /* bitmap of which parens we have moved
155 U32 study_chunk_recursed_bytes; /* bytes in bitmap */
159 I32 override_recoding;
160 I32 in_multi_char_class;
161 struct reg_code_block *code_blocks; /* positions of literal (?{})
163 int num_code_blocks; /* size of code_blocks[] */
164 int code_index; /* next code_blocks[] slot */
165 SSize_t maxlen; /* mininum possible number of chars in string to match */
166 #ifdef ADD_TO_REGEXEC
167 char *starttry; /* -Dr: where regtry was called. */
168 #define RExC_starttry (pRExC_state->starttry)
170 SV *runtime_code_qr; /* qr with the runtime code blocks */
172 const char *lastparse;
174 AV *paren_name_list; /* idx -> name */
175 #define RExC_lastparse (pRExC_state->lastparse)
176 #define RExC_lastnum (pRExC_state->lastnum)
177 #define RExC_paren_name_list (pRExC_state->paren_name_list)
181 #define RExC_flags (pRExC_state->flags)
182 #define RExC_pm_flags (pRExC_state->pm_flags)
183 #define RExC_precomp (pRExC_state->precomp)
184 #define RExC_rx_sv (pRExC_state->rx_sv)
185 #define RExC_rx (pRExC_state->rx)
186 #define RExC_rxi (pRExC_state->rxi)
187 #define RExC_start (pRExC_state->start)
188 #define RExC_end (pRExC_state->end)
189 #define RExC_parse (pRExC_state->parse)
190 #define RExC_whilem_seen (pRExC_state->whilem_seen)
191 #ifdef RE_TRACK_PATTERN_OFFSETS
192 #define RExC_offsets (pRExC_state->rxi->u.offsets) /* I am not like the
195 #define RExC_emit (pRExC_state->emit)
196 #define RExC_emit_dummy (pRExC_state->emit_dummy)
197 #define RExC_emit_start (pRExC_state->emit_start)
198 #define RExC_emit_bound (pRExC_state->emit_bound)
199 #define RExC_naughty (pRExC_state->naughty)
200 #define RExC_sawback (pRExC_state->sawback)
201 #define RExC_seen (pRExC_state->seen)
202 #define RExC_size (pRExC_state->size)
203 #define RExC_maxlen (pRExC_state->maxlen)
204 #define RExC_npar (pRExC_state->npar)
205 #define RExC_nestroot (pRExC_state->nestroot)
206 #define RExC_extralen (pRExC_state->extralen)
207 #define RExC_seen_zerolen (pRExC_state->seen_zerolen)
208 #define RExC_utf8 (pRExC_state->utf8)
209 #define RExC_uni_semantics (pRExC_state->uni_semantics)
210 #define RExC_orig_utf8 (pRExC_state->orig_utf8)
211 #define RExC_open_parens (pRExC_state->open_parens)
212 #define RExC_close_parens (pRExC_state->close_parens)
213 #define RExC_opend (pRExC_state->opend)
214 #define RExC_paren_names (pRExC_state->paren_names)
215 #define RExC_recurse (pRExC_state->recurse)
216 #define RExC_recurse_count (pRExC_state->recurse_count)
217 #define RExC_study_chunk_recursed (pRExC_state->study_chunk_recursed)
218 #define RExC_study_chunk_recursed_bytes \
219 (pRExC_state->study_chunk_recursed_bytes)
220 #define RExC_in_lookbehind (pRExC_state->in_lookbehind)
221 #define RExC_contains_locale (pRExC_state->contains_locale)
222 #define RExC_contains_i (pRExC_state->contains_i)
223 #define RExC_override_recoding (pRExC_state->override_recoding)
224 #define RExC_in_multi_char_class (pRExC_state->in_multi_char_class)
227 #define ISMULT1(c) ((c) == '*' || (c) == '+' || (c) == '?')
228 #define ISMULT2(s) ((*s) == '*' || (*s) == '+' || (*s) == '?' || \
229 ((*s) == '{' && regcurly(s, FALSE)))
232 * Flags to be passed up and down.
234 #define WORST 0 /* Worst case. */
235 #define HASWIDTH 0x01 /* Known to match non-null strings. */
237 /* Simple enough to be STAR/PLUS operand; in an EXACTish node must be a single
238 * character. (There needs to be a case: in the switch statement in regexec.c
239 * for any node marked SIMPLE.) Note that this is not the same thing as
242 #define SPSTART 0x04 /* Starts with * or + */
243 #define POSTPONED 0x08 /* (?1),(?&name), (??{...}) or similar */
244 #define TRYAGAIN 0x10 /* Weeded out a declaration. */
245 #define RESTART_UTF8 0x20 /* Restart, need to calcuate sizes as UTF-8 */
247 #define REG_NODE_NUM(x) ((x) ? (int)((x)-RExC_emit_start) : -1)
249 /* whether trie related optimizations are enabled */
250 #if PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION
251 #define TRIE_STUDY_OPT
252 #define FULL_TRIE_STUDY
258 #define PBYTE(u8str,paren) ((U8*)(u8str))[(paren) >> 3]
259 #define PBITVAL(paren) (1 << ((paren) & 7))
260 #define PAREN_TEST(u8str,paren) ( PBYTE(u8str,paren) & PBITVAL(paren))
261 #define PAREN_SET(u8str,paren) PBYTE(u8str,paren) |= PBITVAL(paren)
262 #define PAREN_UNSET(u8str,paren) PBYTE(u8str,paren) &= (~PBITVAL(paren))
264 #define REQUIRE_UTF8 STMT_START { \
266 *flagp = RESTART_UTF8; \
271 /* This converts the named class defined in regcomp.h to its equivalent class
272 * number defined in handy.h. */
273 #define namedclass_to_classnum(class) ((int) ((class) / 2))
274 #define classnum_to_namedclass(classnum) ((classnum) * 2)
276 #define _invlist_union_complement_2nd(a, b, output) \
277 _invlist_union_maybe_complement_2nd(a, b, TRUE, output)
278 #define _invlist_intersection_complement_2nd(a, b, output) \
279 _invlist_intersection_maybe_complement_2nd(a, b, TRUE, output)
281 /* About scan_data_t.
283 During optimisation we recurse through the regexp program performing
284 various inplace (keyhole style) optimisations. In addition study_chunk
285 and scan_commit populate this data structure with information about
286 what strings MUST appear in the pattern. We look for the longest
287 string that must appear at a fixed location, and we look for the
288 longest string that may appear at a floating location. So for instance
293 Both 'FOO' and 'A' are fixed strings. Both 'B' and 'BAR' are floating
294 strings (because they follow a .* construct). study_chunk will identify
295 both FOO and BAR as being the longest fixed and floating strings respectively.
297 The strings can be composites, for instance
301 will result in a composite fixed substring 'foo'.
303 For each string some basic information is maintained:
305 - offset or min_offset
306 This is the position the string must appear at, or not before.
307 It also implicitly (when combined with minlenp) tells us how many
308 characters must match before the string we are searching for.
309 Likewise when combined with minlenp and the length of the string it
310 tells us how many characters must appear after the string we have
314 Only used for floating strings. This is the rightmost point that
315 the string can appear at. If set to SSize_t_MAX it indicates that the
316 string can occur infinitely far to the right.
319 A pointer to the minimum number of characters of the pattern that the
320 string was found inside. This is important as in the case of positive
321 lookahead or positive lookbehind we can have multiple patterns
326 The minimum length of the pattern overall is 3, the minimum length
327 of the lookahead part is 3, but the minimum length of the part that
328 will actually match is 1. So 'FOO's minimum length is 3, but the
329 minimum length for the F is 1. This is important as the minimum length
330 is used to determine offsets in front of and behind the string being
331 looked for. Since strings can be composites this is the length of the
332 pattern at the time it was committed with a scan_commit. Note that
333 the length is calculated by study_chunk, so that the minimum lengths
334 are not known until the full pattern has been compiled, thus the
335 pointer to the value.
339 In the case of lookbehind the string being searched for can be
340 offset past the start point of the final matching string.
341 If this value was just blithely removed from the min_offset it would
342 invalidate some of the calculations for how many chars must match
343 before or after (as they are derived from min_offset and minlen and
344 the length of the string being searched for).
345 When the final pattern is compiled and the data is moved from the
346 scan_data_t structure into the regexp structure the information
347 about lookbehind is factored in, with the information that would
348 have been lost precalculated in the end_shift field for the
351 The fields pos_min and pos_delta are used to store the minimum offset
352 and the delta to the maximum offset at the current point in the pattern.
356 typedef struct scan_data_t {
357 /*I32 len_min; unused */
358 /*I32 len_delta; unused */
362 SSize_t last_end; /* min value, <0 unless valid. */
363 SSize_t last_start_min;
364 SSize_t last_start_max;
365 SV **longest; /* Either &l_fixed, or &l_float. */
366 SV *longest_fixed; /* longest fixed string found in pattern */
367 SSize_t offset_fixed; /* offset where it starts */
368 SSize_t *minlen_fixed; /* pointer to the minlen relevant to the string */
369 I32 lookbehind_fixed; /* is the position of the string modfied by LB */
370 SV *longest_float; /* longest floating string found in pattern */
371 SSize_t offset_float_min; /* earliest point in string it can appear */
372 SSize_t offset_float_max; /* latest point in string it can appear */
373 SSize_t *minlen_float; /* pointer to the minlen relevant to the string */
374 SSize_t lookbehind_float; /* is the pos of the string modified by LB */
377 SSize_t *last_closep;
378 regnode_ssc *start_class;
381 /* The below is perhaps overboard, but this allows us to save a test at the
382 * expense of a mask. This is because on both EBCDIC and ASCII machines, 'A'
383 * and 'a' differ by a single bit; the same with the upper and lower case of
384 * all other ASCII-range alphabetics. On ASCII platforms, they are 32 apart;
385 * on EBCDIC, they are 64. This uses an exclusive 'or' to find that bit and
386 * then inverts it to form a mask, with just a single 0, in the bit position
387 * where the upper- and lowercase differ. XXX There are about 40 other
388 * instances in the Perl core where this micro-optimization could be used.
389 * Should decide if maintenance cost is worse, before changing those
391 * Returns a boolean as to whether or not 'v' is either a lowercase or
392 * uppercase instance of 'c', where 'c' is in [A-Za-z]. If 'c' is a
393 * compile-time constant, the generated code is better than some optimizing
394 * compilers figure out, amounting to a mask and test. The results are
395 * meaningless if 'c' is not one of [A-Za-z] */
396 #define isARG2_lower_or_UPPER_ARG1(c, v) \
397 (((v) & ~('A' ^ 'a')) == ((c) & ~('A' ^ 'a')))
400 * Forward declarations for pregcomp()'s friends.
403 static const scan_data_t zero_scan_data =
404 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ,0};
406 #define SF_BEFORE_EOL (SF_BEFORE_SEOL|SF_BEFORE_MEOL)
407 #define SF_BEFORE_SEOL 0x0001
408 #define SF_BEFORE_MEOL 0x0002
409 #define SF_FIX_BEFORE_EOL (SF_FIX_BEFORE_SEOL|SF_FIX_BEFORE_MEOL)
410 #define SF_FL_BEFORE_EOL (SF_FL_BEFORE_SEOL|SF_FL_BEFORE_MEOL)
412 #define SF_FIX_SHIFT_EOL (+2)
413 #define SF_FL_SHIFT_EOL (+4)
415 #define SF_FIX_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FIX_SHIFT_EOL)
416 #define SF_FIX_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FIX_SHIFT_EOL)
418 #define SF_FL_BEFORE_SEOL (SF_BEFORE_SEOL << SF_FL_SHIFT_EOL)
419 #define SF_FL_BEFORE_MEOL (SF_BEFORE_MEOL << SF_FL_SHIFT_EOL) /* 0x20 */
420 #define SF_IS_INF 0x0040
421 #define SF_HAS_PAR 0x0080
422 #define SF_IN_PAR 0x0100
423 #define SF_HAS_EVAL 0x0200
424 #define SCF_DO_SUBSTR 0x0400
425 #define SCF_DO_STCLASS_AND 0x0800
426 #define SCF_DO_STCLASS_OR 0x1000
427 #define SCF_DO_STCLASS (SCF_DO_STCLASS_AND|SCF_DO_STCLASS_OR)
428 #define SCF_WHILEM_VISITED_POS 0x2000
430 #define SCF_TRIE_RESTUDY 0x4000 /* Do restudy? */
431 #define SCF_SEEN_ACCEPT 0x8000
432 #define SCF_TRIE_DOING_RESTUDY 0x10000
434 #define UTF cBOOL(RExC_utf8)
436 /* The enums for all these are ordered so things work out correctly */
437 #define LOC (get_regex_charset(RExC_flags) == REGEX_LOCALE_CHARSET)
438 #define DEPENDS_SEMANTICS (get_regex_charset(RExC_flags) \
439 == REGEX_DEPENDS_CHARSET)
440 #define UNI_SEMANTICS (get_regex_charset(RExC_flags) == REGEX_UNICODE_CHARSET)
441 #define AT_LEAST_UNI_SEMANTICS (get_regex_charset(RExC_flags) \
442 >= REGEX_UNICODE_CHARSET)
443 #define ASCII_RESTRICTED (get_regex_charset(RExC_flags) \
444 == REGEX_ASCII_RESTRICTED_CHARSET)
445 #define AT_LEAST_ASCII_RESTRICTED (get_regex_charset(RExC_flags) \
446 >= REGEX_ASCII_RESTRICTED_CHARSET)
447 #define ASCII_FOLD_RESTRICTED (get_regex_charset(RExC_flags) \
448 == REGEX_ASCII_MORE_RESTRICTED_CHARSET)
450 #define FOLD cBOOL(RExC_flags & RXf_PMf_FOLD)
452 /* For programs that want to be strictly Unicode compatible by dying if any
453 * attempt is made to match a non-Unicode code point against a Unicode
455 #define ALWAYS_WARN_SUPER ckDEAD(packWARN(WARN_NON_UNICODE))
457 #define OOB_NAMEDCLASS -1
459 /* There is no code point that is out-of-bounds, so this is problematic. But
460 * its only current use is to initialize a variable that is always set before
462 #define OOB_UNICODE 0xDEADBEEF
464 #define CHR_SVLEN(sv) (UTF ? sv_len_utf8(sv) : SvCUR(sv))
465 #define CHR_DIST(a,b) (UTF ? utf8_distance(a,b) : a - b)
468 /* length of regex to show in messages that don't mark a position within */
469 #define RegexLengthToShowInErrorMessages 127
472 * If MARKER[12] are adjusted, be sure to adjust the constants at the top
473 * of t/op/regmesg.t, the tests in t/op/re_tests, and those in
474 * op/pragma/warn/regcomp.
476 #define MARKER1 "<-- HERE" /* marker as it appears in the description */
477 #define MARKER2 " <-- HERE " /* marker as it appears within the regex */
479 #define REPORT_LOCATION " in regex; marked by " MARKER1 \
480 " in m/%"UTF8f MARKER2 "%"UTF8f"/"
482 #define REPORT_LOCATION_ARGS(offset) \
483 UTF8fARG(UTF, offset, RExC_precomp), \
484 UTF8fARG(UTF, RExC_end - RExC_precomp - offset, RExC_precomp + offset)
487 * Calls SAVEDESTRUCTOR_X if needed, then calls Perl_croak with the given
488 * arg. Show regex, up to a maximum length. If it's too long, chop and add
491 #define _FAIL(code) STMT_START { \
492 const char *ellipses = ""; \
493 IV len = RExC_end - RExC_precomp; \
496 SAVEFREESV(RExC_rx_sv); \
497 if (len > RegexLengthToShowInErrorMessages) { \
498 /* chop 10 shorter than the max, to ensure meaning of "..." */ \
499 len = RegexLengthToShowInErrorMessages - 10; \
505 #define FAIL(msg) _FAIL( \
506 Perl_croak(aTHX_ "%s in regex m/%"UTF8f"%s/", \
507 msg, UTF8fARG(UTF, len, RExC_precomp), ellipses))
509 #define FAIL2(msg,arg) _FAIL( \
510 Perl_croak(aTHX_ msg " in regex m/%"UTF8f"%s/", \
511 arg, UTF8fARG(UTF, len, RExC_precomp), ellipses))
514 * Simple_vFAIL -- like FAIL, but marks the current location in the scan
516 #define Simple_vFAIL(m) STMT_START { \
517 const IV offset = RExC_parse - RExC_precomp; \
518 Perl_croak(aTHX_ "%s" REPORT_LOCATION, \
519 m, REPORT_LOCATION_ARGS(offset)); \
523 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL()
525 #define vFAIL(m) STMT_START { \
527 SAVEFREESV(RExC_rx_sv); \
532 * Like Simple_vFAIL(), but accepts two arguments.
534 #define Simple_vFAIL2(m,a1) STMT_START { \
535 const IV offset = RExC_parse - RExC_precomp; \
536 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, \
537 REPORT_LOCATION_ARGS(offset)); \
541 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL2().
543 #define vFAIL2(m,a1) STMT_START { \
545 SAVEFREESV(RExC_rx_sv); \
546 Simple_vFAIL2(m, a1); \
551 * Like Simple_vFAIL(), but accepts three arguments.
553 #define Simple_vFAIL3(m, a1, a2) STMT_START { \
554 const IV offset = RExC_parse - RExC_precomp; \
555 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, a2, \
556 REPORT_LOCATION_ARGS(offset)); \
560 * Calls SAVEDESTRUCTOR_X if needed, then Simple_vFAIL3().
562 #define vFAIL3(m,a1,a2) STMT_START { \
564 SAVEFREESV(RExC_rx_sv); \
565 Simple_vFAIL3(m, a1, a2); \
569 * Like Simple_vFAIL(), but accepts four arguments.
571 #define Simple_vFAIL4(m, a1, a2, a3) STMT_START { \
572 const IV offset = RExC_parse - RExC_precomp; \
573 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, a2, a3, \
574 REPORT_LOCATION_ARGS(offset)); \
577 #define vFAIL4(m,a1,a2,a3) STMT_START { \
579 SAVEFREESV(RExC_rx_sv); \
580 Simple_vFAIL4(m, a1, a2, a3); \
583 /* A specialized version of vFAIL2 that works with UTF8f */
584 #define vFAIL2utf8f(m, a1) STMT_START { \
585 const IV offset = RExC_parse - RExC_precomp; \
587 SAVEFREESV(RExC_rx_sv); \
588 S_re_croak2(aTHX_ UTF, m, REPORT_LOCATION, a1, \
589 REPORT_LOCATION_ARGS(offset)); \
593 /* m is not necessarily a "literal string", in this macro */
594 #define reg_warn_non_literal_string(loc, m) STMT_START { \
595 const IV offset = loc - RExC_precomp; \
596 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "%s" REPORT_LOCATION, \
597 m, REPORT_LOCATION_ARGS(offset)); \
600 #define ckWARNreg(loc,m) STMT_START { \
601 const IV offset = loc - RExC_precomp; \
602 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
603 REPORT_LOCATION_ARGS(offset)); \
606 #define vWARN_dep(loc, m) STMT_START { \
607 const IV offset = loc - RExC_precomp; \
608 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), m REPORT_LOCATION, \
609 REPORT_LOCATION_ARGS(offset)); \
612 #define ckWARNdep(loc,m) STMT_START { \
613 const IV offset = loc - RExC_precomp; \
614 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED), \
616 REPORT_LOCATION_ARGS(offset)); \
619 #define ckWARNregdep(loc,m) STMT_START { \
620 const IV offset = loc - RExC_precomp; \
621 Perl_ck_warner_d(aTHX_ packWARN2(WARN_DEPRECATED, WARN_REGEXP), \
623 REPORT_LOCATION_ARGS(offset)); \
626 #define ckWARN2reg_d(loc,m, a1) STMT_START { \
627 const IV offset = loc - RExC_precomp; \
628 Perl_ck_warner_d(aTHX_ packWARN(WARN_REGEXP), \
630 a1, REPORT_LOCATION_ARGS(offset)); \
633 #define ckWARN2reg(loc, m, a1) STMT_START { \
634 const IV offset = loc - RExC_precomp; \
635 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
636 a1, REPORT_LOCATION_ARGS(offset)); \
639 #define vWARN3(loc, m, a1, a2) STMT_START { \
640 const IV offset = loc - RExC_precomp; \
641 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
642 a1, a2, REPORT_LOCATION_ARGS(offset)); \
645 #define ckWARN3reg(loc, m, a1, a2) STMT_START { \
646 const IV offset = loc - RExC_precomp; \
647 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
648 a1, a2, REPORT_LOCATION_ARGS(offset)); \
651 #define vWARN4(loc, m, a1, a2, a3) STMT_START { \
652 const IV offset = loc - RExC_precomp; \
653 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
654 a1, a2, a3, REPORT_LOCATION_ARGS(offset)); \
657 #define ckWARN4reg(loc, m, a1, a2, a3) STMT_START { \
658 const IV offset = loc - RExC_precomp; \
659 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
660 a1, a2, a3, REPORT_LOCATION_ARGS(offset)); \
663 #define vWARN5(loc, m, a1, a2, a3, a4) STMT_START { \
664 const IV offset = loc - RExC_precomp; \
665 Perl_warner(aTHX_ packWARN(WARN_REGEXP), m REPORT_LOCATION, \
666 a1, a2, a3, a4, REPORT_LOCATION_ARGS(offset)); \
670 /* Allow for side effects in s */
671 #define REGC(c,s) STMT_START { \
672 if (!SIZE_ONLY) *(s) = (c); else (void)(s); \
675 /* Macros for recording node offsets. 20001227 mjd@plover.com
676 * Nodes are numbered 1, 2, 3, 4. Node #n's position is recorded in
677 * element 2*n-1 of the array. Element #2n holds the byte length node #n.
678 * Element 0 holds the number n.
679 * Position is 1 indexed.
681 #ifndef RE_TRACK_PATTERN_OFFSETS
682 #define Set_Node_Offset_To_R(node,byte)
683 #define Set_Node_Offset(node,byte)
684 #define Set_Cur_Node_Offset
685 #define Set_Node_Length_To_R(node,len)
686 #define Set_Node_Length(node,len)
687 #define Set_Node_Cur_Length(node,start)
688 #define Node_Offset(n)
689 #define Node_Length(n)
690 #define Set_Node_Offset_Length(node,offset,len)
691 #define ProgLen(ri) ri->u.proglen
692 #define SetProgLen(ri,x) ri->u.proglen = x
694 #define ProgLen(ri) ri->u.offsets[0]
695 #define SetProgLen(ri,x) ri->u.offsets[0] = x
696 #define Set_Node_Offset_To_R(node,byte) STMT_START { \
698 MJD_OFFSET_DEBUG(("** (%d) offset of node %d is %d.\n", \
699 __LINE__, (int)(node), (int)(byte))); \
701 Perl_croak(aTHX_ "value of node is %d in Offset macro", \
704 RExC_offsets[2*(node)-1] = (byte); \
709 #define Set_Node_Offset(node,byte) \
710 Set_Node_Offset_To_R((node)-RExC_emit_start, (byte)-RExC_start)
711 #define Set_Cur_Node_Offset Set_Node_Offset(RExC_emit, RExC_parse)
713 #define Set_Node_Length_To_R(node,len) STMT_START { \
715 MJD_OFFSET_DEBUG(("** (%d) size of node %d is %d.\n", \
716 __LINE__, (int)(node), (int)(len))); \
718 Perl_croak(aTHX_ "value of node is %d in Length macro", \
721 RExC_offsets[2*(node)] = (len); \
726 #define Set_Node_Length(node,len) \
727 Set_Node_Length_To_R((node)-RExC_emit_start, len)
728 #define Set_Node_Cur_Length(node, start) \
729 Set_Node_Length(node, RExC_parse - start)
731 /* Get offsets and lengths */
732 #define Node_Offset(n) (RExC_offsets[2*((n)-RExC_emit_start)-1])
733 #define Node_Length(n) (RExC_offsets[2*((n)-RExC_emit_start)])
735 #define Set_Node_Offset_Length(node,offset,len) STMT_START { \
736 Set_Node_Offset_To_R((node)-RExC_emit_start, (offset)); \
737 Set_Node_Length_To_R((node)-RExC_emit_start, (len)); \
741 #if PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS
742 #define EXPERIMENTAL_INPLACESCAN
743 #endif /*PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS*/
745 #define DEBUG_RExC_seen() \
746 DEBUG_OPTIMISE_MORE_r({ \
747 PerlIO_printf(Perl_debug_log,"RExC_seen: "); \
749 if (RExC_seen & REG_ZERO_LEN_SEEN) \
750 PerlIO_printf(Perl_debug_log,"REG_ZERO_LEN_SEEN "); \
752 if (RExC_seen & REG_LOOKBEHIND_SEEN) \
753 PerlIO_printf(Perl_debug_log,"REG_LOOKBEHIND_SEEN "); \
755 if (RExC_seen & REG_GPOS_SEEN) \
756 PerlIO_printf(Perl_debug_log,"REG_GPOS_SEEN "); \
758 if (RExC_seen & REG_CANY_SEEN) \
759 PerlIO_printf(Perl_debug_log,"REG_CANY_SEEN "); \
761 if (RExC_seen & REG_RECURSE_SEEN) \
762 PerlIO_printf(Perl_debug_log,"REG_RECURSE_SEEN "); \
764 if (RExC_seen & REG_TOP_LEVEL_BRANCHES_SEEN) \
765 PerlIO_printf(Perl_debug_log,"REG_TOP_LEVEL_BRANCHES_SEEN "); \
767 if (RExC_seen & REG_VERBARG_SEEN) \
768 PerlIO_printf(Perl_debug_log,"REG_VERBARG_SEEN "); \
770 if (RExC_seen & REG_CUTGROUP_SEEN) \
771 PerlIO_printf(Perl_debug_log,"REG_CUTGROUP_SEEN "); \
773 if (RExC_seen & REG_RUN_ON_COMMENT_SEEN) \
774 PerlIO_printf(Perl_debug_log,"REG_RUN_ON_COMMENT_SEEN "); \
776 if (RExC_seen & REG_UNFOLDED_MULTI_SEEN) \
777 PerlIO_printf(Perl_debug_log,"REG_UNFOLDED_MULTI_SEEN "); \
779 if (RExC_seen & REG_GOSTART_SEEN) \
780 PerlIO_printf(Perl_debug_log,"REG_GOSTART_SEEN "); \
782 if (RExC_seen & REG_UNBOUNDED_QUANTIFIER_SEEN) \
783 PerlIO_printf(Perl_debug_log,"REG_UNBOUNDED_QUANTIFIER_SEEN "); \
785 PerlIO_printf(Perl_debug_log,"\n"); \
788 #define DEBUG_STUDYDATA(str,data,depth) \
789 DEBUG_OPTIMISE_MORE_r(if(data){ \
790 PerlIO_printf(Perl_debug_log, \
791 "%*s" str "Pos:%"IVdf"/%"IVdf \
792 " Flags: 0x%"UVXf" Whilem_c: %"IVdf" Lcp: %"IVdf" %s", \
793 (int)(depth)*2, "", \
794 (IV)((data)->pos_min), \
795 (IV)((data)->pos_delta), \
796 (UV)((data)->flags), \
797 (IV)((data)->whilem_c), \
798 (IV)((data)->last_closep ? *((data)->last_closep) : -1), \
799 is_inf ? "INF " : "" \
801 if ((data)->last_found) \
802 PerlIO_printf(Perl_debug_log, \
803 "Last:'%s' %"IVdf":%"IVdf"/%"IVdf" %sFixed:'%s' @ %"IVdf \
804 " %sFloat: '%s' @ %"IVdf"/%"IVdf"", \
805 SvPVX_const((data)->last_found), \
806 (IV)((data)->last_end), \
807 (IV)((data)->last_start_min), \
808 (IV)((data)->last_start_max), \
809 ((data)->longest && \
810 (data)->longest==&((data)->longest_fixed)) ? "*" : "", \
811 SvPVX_const((data)->longest_fixed), \
812 (IV)((data)->offset_fixed), \
813 ((data)->longest && \
814 (data)->longest==&((data)->longest_float)) ? "*" : "", \
815 SvPVX_const((data)->longest_float), \
816 (IV)((data)->offset_float_min), \
817 (IV)((data)->offset_float_max) \
819 PerlIO_printf(Perl_debug_log,"\n"); \
822 /* Mark that we cannot extend a found fixed substring at this point.
823 Update the longest found anchored substring and the longest found
824 floating substrings if needed. */
827 S_scan_commit(pTHX_ const RExC_state_t *pRExC_state, scan_data_t *data,
828 SSize_t *minlenp, int is_inf)
830 const STRLEN l = CHR_SVLEN(data->last_found);
831 const STRLEN old_l = CHR_SVLEN(*data->longest);
832 GET_RE_DEBUG_FLAGS_DECL;
834 PERL_ARGS_ASSERT_SCAN_COMMIT;
836 if ((l >= old_l) && ((l > old_l) || (data->flags & SF_BEFORE_EOL))) {
837 SvSetMagicSV(*data->longest, data->last_found);
838 if (*data->longest == data->longest_fixed) {
839 data->offset_fixed = l ? data->last_start_min : data->pos_min;
840 if (data->flags & SF_BEFORE_EOL)
842 |= ((data->flags & SF_BEFORE_EOL) << SF_FIX_SHIFT_EOL);
844 data->flags &= ~SF_FIX_BEFORE_EOL;
845 data->minlen_fixed=minlenp;
846 data->lookbehind_fixed=0;
848 else { /* *data->longest == data->longest_float */
849 data->offset_float_min = l ? data->last_start_min : data->pos_min;
850 data->offset_float_max = (l
851 ? data->last_start_max
852 : (data->pos_delta == SSize_t_MAX
854 : data->pos_min + data->pos_delta));
856 || (STRLEN)data->offset_float_max > (STRLEN)SSize_t_MAX)
857 data->offset_float_max = SSize_t_MAX;
858 if (data->flags & SF_BEFORE_EOL)
860 |= ((data->flags & SF_BEFORE_EOL) << SF_FL_SHIFT_EOL);
862 data->flags &= ~SF_FL_BEFORE_EOL;
863 data->minlen_float=minlenp;
864 data->lookbehind_float=0;
867 SvCUR_set(data->last_found, 0);
869 SV * const sv = data->last_found;
870 if (SvUTF8(sv) && SvMAGICAL(sv)) {
871 MAGIC * const mg = mg_find(sv, PERL_MAGIC_utf8);
877 data->flags &= ~SF_BEFORE_EOL;
878 DEBUG_STUDYDATA("commit: ",data,0);
881 /* An SSC is just a regnode_charclass_posix with an extra field: the inversion
882 * list that describes which code points it matches */
885 S_ssc_anything(pTHX_ regnode_ssc *ssc)
887 /* Set the SSC 'ssc' to match an empty string or any code point */
889 PERL_ARGS_ASSERT_SSC_ANYTHING;
891 assert(is_ANYOF_SYNTHETIC(ssc));
893 ssc->invlist = sv_2mortal(_new_invlist(2)); /* mortalize so won't leak */
894 _append_range_to_invlist(ssc->invlist, 0, UV_MAX);
895 ANYOF_FLAGS(ssc) |= ANYOF_EMPTY_STRING; /* Plus match empty string */
899 S_ssc_is_anything(pTHX_ const regnode_ssc *ssc)
901 /* Returns TRUE if the SSC 'ssc' can match the empty string and any code
902 * point; FALSE otherwise. Thus, this is used to see if using 'ssc' buys
903 * us anything: if the function returns TRUE, 'ssc' hasn't been restricted
904 * in any way, so there's no point in using it */
909 PERL_ARGS_ASSERT_SSC_IS_ANYTHING;
911 assert(is_ANYOF_SYNTHETIC(ssc));
913 if (! (ANYOF_FLAGS(ssc) & ANYOF_EMPTY_STRING)) {
917 /* See if the list consists solely of the range 0 - Infinity */
918 invlist_iterinit(ssc->invlist);
919 ret = invlist_iternext(ssc->invlist, &start, &end)
923 invlist_iterfinish(ssc->invlist);
929 /* If e.g., both \w and \W are set, matches everything */
930 if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
932 for (i = 0; i < ANYOF_POSIXL_MAX; i += 2) {
933 if (ANYOF_POSIXL_TEST(ssc, i) && ANYOF_POSIXL_TEST(ssc, i+1)) {
943 S_ssc_init(pTHX_ const RExC_state_t *pRExC_state, regnode_ssc *ssc)
945 /* Initializes the SSC 'ssc'. This includes setting it to match an empty
946 * string, any code point, or any posix class under locale */
948 PERL_ARGS_ASSERT_SSC_INIT;
950 Zero(ssc, 1, regnode_ssc);
951 set_ANYOF_SYNTHETIC(ssc);
952 ARG_SET(ssc, ANYOF_NONBITMAP_EMPTY);
955 /* If any portion of the regex is to operate under locale rules,
956 * initialization includes it. The reason this isn't done for all regexes
957 * is that the optimizer was written under the assumption that locale was
958 * all-or-nothing. Given the complexity and lack of documentation in the
959 * optimizer, and that there are inadequate test cases for locale, many
960 * parts of it may not work properly, it is safest to avoid locale unless
962 if (RExC_contains_locale) {
963 ANYOF_POSIXL_SETALL(ssc);
966 ANYOF_POSIXL_ZERO(ssc);
971 S_ssc_is_cp_posixl_init(pTHX_ const RExC_state_t *pRExC_state,
972 const regnode_ssc *ssc)
974 /* Returns TRUE if the SSC 'ssc' is in its initial state with regard only
975 * to the list of code points matched, and locale posix classes; hence does
976 * not check its flags) */
981 PERL_ARGS_ASSERT_SSC_IS_CP_POSIXL_INIT;
983 assert(is_ANYOF_SYNTHETIC(ssc));
985 invlist_iterinit(ssc->invlist);
986 ret = invlist_iternext(ssc->invlist, &start, &end)
990 invlist_iterfinish(ssc->invlist);
996 if (RExC_contains_locale && ! ANYOF_POSIXL_SSC_TEST_ALL_SET(ssc)) {
1004 S_get_ANYOF_cp_list_for_ssc(pTHX_ const RExC_state_t *pRExC_state,
1005 const regnode_charclass* const node)
1007 /* Returns a mortal inversion list defining which code points are matched
1008 * by 'node', which is of type ANYOF. Handles complementing the result if
1009 * appropriate. If some code points aren't knowable at this time, the
1010 * returned list must, and will, contain every code point that is a
1013 SV* invlist = sv_2mortal(_new_invlist(0));
1014 SV* only_utf8_locale_invlist = NULL;
1016 const U32 n = ARG(node);
1017 bool new_node_has_latin1 = FALSE;
1019 PERL_ARGS_ASSERT_GET_ANYOF_CP_LIST_FOR_SSC;
1021 /* Look at the data structure created by S_set_ANYOF_arg() */
1022 if (n != ANYOF_NONBITMAP_EMPTY) {
1023 SV * const rv = MUTABLE_SV(RExC_rxi->data->data[n]);
1024 AV * const av = MUTABLE_AV(SvRV(rv));
1025 SV **const ary = AvARRAY(av);
1026 assert(RExC_rxi->data->what[n] == 's');
1028 if (ary[1] && ary[1] != &PL_sv_undef) { /* Has compile-time swash */
1029 invlist = sv_2mortal(invlist_clone(_get_swash_invlist(ary[1])));
1031 else if (ary[0] && ary[0] != &PL_sv_undef) {
1033 /* Here, no compile-time swash, and there are things that won't be
1034 * known until runtime -- we have to assume it could be anything */
1035 return _add_range_to_invlist(invlist, 0, UV_MAX);
1037 else if (ary[3] && ary[3] != &PL_sv_undef) {
1039 /* Here no compile-time swash, and no run-time only data. Use the
1040 * node's inversion list */
1041 invlist = sv_2mortal(invlist_clone(ary[3]));
1044 /* Get the code points valid only under UTF-8 locales */
1045 if ((ANYOF_FLAGS(node) & ANYOF_LOC_FOLD)
1046 && ary[2] && ary[2] != &PL_sv_undef)
1048 only_utf8_locale_invlist = ary[2];
1052 /* An ANYOF node contains a bitmap for the first 256 code points, and an
1053 * inversion list for the others, but if there are code points that should
1054 * match only conditionally on the target string being UTF-8, those are
1055 * placed in the inversion list, and not the bitmap. Since there are
1056 * circumstances under which they could match, they are included in the
1057 * SSC. But if the ANYOF node is to be inverted, we have to exclude them
1058 * here, so that when we invert below, the end result actually does include
1059 * them. (Think about "\xe0" =~ /[^\xc0]/di;). We have to do this here
1060 * before we add the unconditionally matched code points */
1061 if (ANYOF_FLAGS(node) & ANYOF_INVERT) {
1062 _invlist_intersection_complement_2nd(invlist,
1067 /* Add in the points from the bit map */
1068 for (i = 0; i < 256; i++) {
1069 if (ANYOF_BITMAP_TEST(node, i)) {
1070 invlist = add_cp_to_invlist(invlist, i);
1071 new_node_has_latin1 = TRUE;
1075 /* If this can match all upper Latin1 code points, have to add them
1077 if (ANYOF_FLAGS(node) & ANYOF_NON_UTF8_NON_ASCII_ALL) {
1078 _invlist_union(invlist, PL_UpperLatin1, &invlist);
1081 /* Similarly for these */
1082 if (ANYOF_FLAGS(node) & ANYOF_ABOVE_LATIN1_ALL) {
1083 invlist = _add_range_to_invlist(invlist, 256, UV_MAX);
1086 if (ANYOF_FLAGS(node) & ANYOF_INVERT) {
1087 _invlist_invert(invlist);
1089 else if (new_node_has_latin1 && ANYOF_FLAGS(node) & ANYOF_LOC_FOLD) {
1091 /* Under /li, any 0-255 could fold to any other 0-255, depending on the
1092 * locale. We can skip this if there are no 0-255 at all. */
1093 _invlist_union(invlist, PL_Latin1, &invlist);
1096 /* Similarly add the UTF-8 locale possible matches. These have to be
1097 * deferred until after the non-UTF-8 locale ones are taken care of just
1098 * above, or it leads to wrong results under ANYOF_INVERT */
1099 if (only_utf8_locale_invlist) {
1100 _invlist_union_maybe_complement_2nd(invlist,
1101 only_utf8_locale_invlist,
1102 ANYOF_FLAGS(node) & ANYOF_INVERT,
1109 /* These two functions currently do the exact same thing */
1110 #define ssc_init_zero ssc_init
1112 #define ssc_add_cp(ssc, cp) ssc_add_range((ssc), (cp), (cp))
1113 #define ssc_match_all_cp(ssc) ssc_add_range(ssc, 0, UV_MAX)
1115 /* 'AND' a given class with another one. Can create false positives. 'ssc'
1116 * should not be inverted. 'and_with->flags & ANYOF_POSIXL' should be 0 if
1117 * 'and_with' is a regnode_charclass instead of a regnode_ssc. */
1120 S_ssc_and(pTHX_ const RExC_state_t *pRExC_state, regnode_ssc *ssc,
1121 const regnode_charclass *and_with)
1123 /* Accumulate into SSC 'ssc' its 'AND' with 'and_with', which is either
1124 * another SSC or a regular ANYOF class. Can create false positives. */
1129 PERL_ARGS_ASSERT_SSC_AND;
1131 assert(is_ANYOF_SYNTHETIC(ssc));
1133 /* 'and_with' is used as-is if it too is an SSC; otherwise have to extract
1134 * the code point inversion list and just the relevant flags */
1135 if (is_ANYOF_SYNTHETIC(and_with)) {
1136 anded_cp_list = ((regnode_ssc *)and_with)->invlist;
1137 anded_flags = ANYOF_FLAGS(and_with);
1139 /* XXX This is a kludge around what appears to be deficiencies in the
1140 * optimizer. If we make S_ssc_anything() add in the WARN_SUPER flag,
1141 * there are paths through the optimizer where it doesn't get weeded
1142 * out when it should. And if we don't make some extra provision for
1143 * it like the code just below, it doesn't get added when it should.
1144 * This solution is to add it only when AND'ing, which is here, and
1145 * only when what is being AND'ed is the pristine, original node
1146 * matching anything. Thus it is like adding it to ssc_anything() but
1147 * only when the result is to be AND'ed. Probably the same solution
1148 * could be adopted for the same problem we have with /l matching,
1149 * which is solved differently in S_ssc_init(), and that would lead to
1150 * fewer false positives than that solution has. But if this solution
1151 * creates bugs, the consequences are only that a warning isn't raised
1152 * that should be; while the consequences for having /l bugs is
1153 * incorrect matches */
1154 if (ssc_is_anything((regnode_ssc *)and_with)) {
1155 anded_flags |= ANYOF_WARN_SUPER;
1159 anded_cp_list = get_ANYOF_cp_list_for_ssc(pRExC_state, and_with);
1160 anded_flags = ANYOF_FLAGS(and_with) & ANYOF_COMMON_FLAGS;
1163 ANYOF_FLAGS(ssc) &= anded_flags;
1165 /* Below, C1 is the list of code points in 'ssc'; P1, its posix classes.
1166 * C2 is the list of code points in 'and-with'; P2, its posix classes.
1167 * 'and_with' may be inverted. When not inverted, we have the situation of
1169 * (C1 | P1) & (C2 | P2)
1170 * = (C1 & (C2 | P2)) | (P1 & (C2 | P2))
1171 * = ((C1 & C2) | (C1 & P2)) | ((P1 & C2) | (P1 & P2))
1172 * <= ((C1 & C2) | P2)) | ( P1 | (P1 & P2))
1173 * <= ((C1 & C2) | P1 | P2)
1174 * Alternatively, the last few steps could be:
1175 * = ((C1 & C2) | (C1 & P2)) | ((P1 & C2) | (P1 & P2))
1176 * <= ((C1 & C2) | C1 ) | ( C2 | (P1 & P2))
1177 * <= (C1 | C2 | (P1 & P2))
1178 * We favor the second approach if either P1 or P2 is non-empty. This is
1179 * because these components are a barrier to doing optimizations, as what
1180 * they match cannot be known until the moment of matching as they are
1181 * dependent on the current locale, 'AND"ing them likely will reduce or
1183 * But we can do better if we know that C1,P1 are in their initial state (a
1184 * frequent occurrence), each matching everything:
1185 * (<everything>) & (C2 | P2) = C2 | P2
1186 * Similarly, if C2,P2 are in their initial state (again a frequent
1187 * occurrence), the result is a no-op
1188 * (C1 | P1) & (<everything>) = C1 | P1
1191 * (C1 | P1) & ~(C2 | P2) = (C1 | P1) & (~C2 & ~P2)
1192 * = (C1 & (~C2 & ~P2)) | (P1 & (~C2 & ~P2))
1193 * <= (C1 & ~C2) | (P1 & ~P2)
1196 if ((ANYOF_FLAGS(and_with) & ANYOF_INVERT)
1197 && ! is_ANYOF_SYNTHETIC(and_with))
1201 ssc_intersection(ssc,
1203 FALSE /* Has already been inverted */
1206 /* If either P1 or P2 is empty, the intersection will be also; can skip
1208 if (! (ANYOF_FLAGS(and_with) & ANYOF_POSIXL)) {
1209 ANYOF_POSIXL_ZERO(ssc);
1211 else if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
1213 /* Note that the Posix class component P from 'and_with' actually
1215 * P = Pa | Pb | ... | Pn
1216 * where each component is one posix class, such as in [\w\s].
1218 * ~P = ~(Pa | Pb | ... | Pn)
1219 * = ~Pa & ~Pb & ... & ~Pn
1220 * <= ~Pa | ~Pb | ... | ~Pn
1221 * The last is something we can easily calculate, but unfortunately
1222 * is likely to have many false positives. We could do better
1223 * in some (but certainly not all) instances if two classes in
1224 * P have known relationships. For example
1225 * :lower: <= :alpha: <= :alnum: <= \w <= :graph: <= :print:
1227 * :lower: & :print: = :lower:
1228 * And similarly for classes that must be disjoint. For example,
1229 * since \s and \w can have no elements in common based on rules in
1230 * the POSIX standard,
1231 * \w & ^\S = nothing
1232 * Unfortunately, some vendor locales do not meet the Posix
1233 * standard, in particular almost everything by Microsoft.
1234 * The loop below just changes e.g., \w into \W and vice versa */
1236 regnode_charclass_posixl temp;
1237 int add = 1; /* To calculate the index of the complement */
1239 ANYOF_POSIXL_ZERO(&temp);
1240 for (i = 0; i < ANYOF_MAX; i++) {
1242 || ! ANYOF_POSIXL_TEST((regnode_charclass_posixl*) and_with, i)
1243 || ! ANYOF_POSIXL_TEST((regnode_charclass_posixl*) and_with, i + 1));
1245 if (ANYOF_POSIXL_TEST((regnode_charclass_posixl*) and_with, i)) {
1246 ANYOF_POSIXL_SET(&temp, i + add);
1248 add = 0 - add; /* 1 goes to -1; -1 goes to 1 */
1250 ANYOF_POSIXL_AND(&temp, ssc);
1252 } /* else ssc already has no posixes */
1253 } /* else: Not inverted. This routine is a no-op if 'and_with' is an SSC
1254 in its initial state */
1255 else if (! is_ANYOF_SYNTHETIC(and_with)
1256 || ! ssc_is_cp_posixl_init(pRExC_state, (regnode_ssc *)and_with))
1258 /* But if 'ssc' is in its initial state, the result is just 'and_with';
1259 * copy it over 'ssc' */
1260 if (ssc_is_cp_posixl_init(pRExC_state, ssc)) {
1261 if (is_ANYOF_SYNTHETIC(and_with)) {
1262 StructCopy(and_with, ssc, regnode_ssc);
1265 ssc->invlist = anded_cp_list;
1266 ANYOF_POSIXL_ZERO(ssc);
1267 if (ANYOF_FLAGS(and_with) & ANYOF_POSIXL) {
1268 ANYOF_POSIXL_OR((regnode_charclass_posixl*) and_with, ssc);
1272 else if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)
1273 || (ANYOF_FLAGS(and_with) & ANYOF_POSIXL))
1275 /* One or the other of P1, P2 is non-empty. */
1276 if (ANYOF_FLAGS(and_with) & ANYOF_POSIXL) {
1277 ANYOF_POSIXL_AND((regnode_charclass_posixl*) and_with, ssc);
1279 ssc_union(ssc, anded_cp_list, FALSE);
1281 else { /* P1 = P2 = empty */
1282 ssc_intersection(ssc, anded_cp_list, FALSE);
1288 S_ssc_or(pTHX_ const RExC_state_t *pRExC_state, regnode_ssc *ssc,
1289 const regnode_charclass *or_with)
1291 /* Accumulate into SSC 'ssc' its 'OR' with 'or_with', which is either
1292 * another SSC or a regular ANYOF class. Can create false positives if
1293 * 'or_with' is to be inverted. */
1298 PERL_ARGS_ASSERT_SSC_OR;
1300 assert(is_ANYOF_SYNTHETIC(ssc));
1302 /* 'or_with' is used as-is if it too is an SSC; otherwise have to extract
1303 * the code point inversion list and just the relevant flags */
1304 if (is_ANYOF_SYNTHETIC(or_with)) {
1305 ored_cp_list = ((regnode_ssc*) or_with)->invlist;
1306 ored_flags = ANYOF_FLAGS(or_with);
1309 ored_cp_list = get_ANYOF_cp_list_for_ssc(pRExC_state, or_with);
1310 ored_flags = ANYOF_FLAGS(or_with) & ANYOF_COMMON_FLAGS;
1313 ANYOF_FLAGS(ssc) |= ored_flags;
1315 /* Below, C1 is the list of code points in 'ssc'; P1, its posix classes.
1316 * C2 is the list of code points in 'or-with'; P2, its posix classes.
1317 * 'or_with' may be inverted. When not inverted, we have the simple
1318 * situation of computing:
1319 * (C1 | P1) | (C2 | P2) = (C1 | C2) | (P1 | P2)
1320 * If P1|P2 yields a situation with both a class and its complement are
1321 * set, like having both \w and \W, this matches all code points, and we
1322 * can delete these from the P component of the ssc going forward. XXX We
1323 * might be able to delete all the P components, but I (khw) am not certain
1324 * about this, and it is better to be safe.
1327 * (C1 | P1) | ~(C2 | P2) = (C1 | P1) | (~C2 & ~P2)
1328 * <= (C1 | P1) | ~C2
1329 * <= (C1 | ~C2) | P1
1330 * (which results in actually simpler code than the non-inverted case)
1333 if ((ANYOF_FLAGS(or_with) & ANYOF_INVERT)
1334 && ! is_ANYOF_SYNTHETIC(or_with))
1336 /* We ignore P2, leaving P1 going forward */
1337 } /* else Not inverted */
1338 else if (ANYOF_FLAGS(or_with) & ANYOF_POSIXL) {
1339 ANYOF_POSIXL_OR((regnode_charclass_posixl*)or_with, ssc);
1340 if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
1342 for (i = 0; i < ANYOF_MAX; i += 2) {
1343 if (ANYOF_POSIXL_TEST(ssc, i) && ANYOF_POSIXL_TEST(ssc, i + 1))
1345 ssc_match_all_cp(ssc);
1346 ANYOF_POSIXL_CLEAR(ssc, i);
1347 ANYOF_POSIXL_CLEAR(ssc, i+1);
1355 FALSE /* Already has been inverted */
1359 PERL_STATIC_INLINE void
1360 S_ssc_union(pTHX_ regnode_ssc *ssc, SV* const invlist, const bool invert2nd)
1362 PERL_ARGS_ASSERT_SSC_UNION;
1364 assert(is_ANYOF_SYNTHETIC(ssc));
1366 _invlist_union_maybe_complement_2nd(ssc->invlist,
1372 PERL_STATIC_INLINE void
1373 S_ssc_intersection(pTHX_ regnode_ssc *ssc,
1375 const bool invert2nd)
1377 PERL_ARGS_ASSERT_SSC_INTERSECTION;
1379 assert(is_ANYOF_SYNTHETIC(ssc));
1381 _invlist_intersection_maybe_complement_2nd(ssc->invlist,
1387 PERL_STATIC_INLINE void
1388 S_ssc_add_range(pTHX_ regnode_ssc *ssc, const UV start, const UV end)
1390 PERL_ARGS_ASSERT_SSC_ADD_RANGE;
1392 assert(is_ANYOF_SYNTHETIC(ssc));
1394 ssc->invlist = _add_range_to_invlist(ssc->invlist, start, end);
1397 PERL_STATIC_INLINE void
1398 S_ssc_cp_and(pTHX_ regnode_ssc *ssc, const UV cp)
1400 /* AND just the single code point 'cp' into the SSC 'ssc' */
1402 SV* cp_list = _new_invlist(2);
1404 PERL_ARGS_ASSERT_SSC_CP_AND;
1406 assert(is_ANYOF_SYNTHETIC(ssc));
1408 cp_list = add_cp_to_invlist(cp_list, cp);
1409 ssc_intersection(ssc, cp_list,
1410 FALSE /* Not inverted */
1412 SvREFCNT_dec_NN(cp_list);
1415 PERL_STATIC_INLINE void
1416 S_ssc_clear_locale(pTHX_ regnode_ssc *ssc)
1418 /* Set the SSC 'ssc' to not match any locale things */
1420 PERL_ARGS_ASSERT_SSC_CLEAR_LOCALE;
1422 assert(is_ANYOF_SYNTHETIC(ssc));
1424 ANYOF_POSIXL_ZERO(ssc);
1425 ANYOF_FLAGS(ssc) &= ~ANYOF_LOCALE_FLAGS;
1429 S_ssc_finalize(pTHX_ RExC_state_t *pRExC_state, regnode_ssc *ssc)
1431 /* The inversion list in the SSC is marked mortal; now we need a more
1432 * permanent copy, which is stored the same way that is done in a regular
1433 * ANYOF node, with the first 256 code points in a bit map */
1435 SV* invlist = invlist_clone(ssc->invlist);
1437 PERL_ARGS_ASSERT_SSC_FINALIZE;
1439 assert(is_ANYOF_SYNTHETIC(ssc));
1441 /* The code in this file assumes that all but these flags aren't relevant
1442 * to the SSC, except ANYOF_EMPTY_STRING, which should be cleared by the
1443 * time we reach here */
1444 assert(! (ANYOF_FLAGS(ssc) & ~ANYOF_COMMON_FLAGS));
1446 populate_ANYOF_from_invlist( (regnode *) ssc, &invlist);
1448 set_ANYOF_arg(pRExC_state, (regnode *) ssc, invlist,
1449 NULL, NULL, NULL, FALSE);
1451 /* Make sure is clone-safe */
1452 ssc->invlist = NULL;
1454 if (ANYOF_POSIXL_SSC_TEST_ANY_SET(ssc)) {
1455 ANYOF_FLAGS(ssc) |= ANYOF_POSIXL;
1458 assert(! (ANYOF_FLAGS(ssc) & ANYOF_LOCALE_FLAGS) || RExC_contains_locale);
1461 #define TRIE_LIST_ITEM(state,idx) (trie->states[state].trans.list)[ idx ]
1462 #define TRIE_LIST_CUR(state) ( TRIE_LIST_ITEM( state, 0 ).forid )
1463 #define TRIE_LIST_LEN(state) ( TRIE_LIST_ITEM( state, 0 ).newstate )
1464 #define TRIE_LIST_USED(idx) ( trie->states[state].trans.list \
1465 ? (TRIE_LIST_CUR( idx ) - 1) \
1471 dump_trie(trie,widecharmap,revcharmap)
1472 dump_trie_interim_list(trie,widecharmap,revcharmap,next_alloc)
1473 dump_trie_interim_table(trie,widecharmap,revcharmap,next_alloc)
1475 These routines dump out a trie in a somewhat readable format.
1476 The _interim_ variants are used for debugging the interim
1477 tables that are used to generate the final compressed
1478 representation which is what dump_trie expects.
1480 Part of the reason for their existence is to provide a form
1481 of documentation as to how the different representations function.
1486 Dumps the final compressed table form of the trie to Perl_debug_log.
1487 Used for debugging make_trie().
1491 S_dump_trie(pTHX_ const struct _reg_trie_data *trie, HV *widecharmap,
1492 AV *revcharmap, U32 depth)
1495 SV *sv=sv_newmortal();
1496 int colwidth= widecharmap ? 6 : 4;
1498 GET_RE_DEBUG_FLAGS_DECL;
1500 PERL_ARGS_ASSERT_DUMP_TRIE;
1502 PerlIO_printf( Perl_debug_log, "%*sChar : %-6s%-6s%-4s ",
1503 (int)depth * 2 + 2,"",
1504 "Match","Base","Ofs" );
1506 for( state = 0 ; state < trie->uniquecharcount ; state++ ) {
1507 SV ** const tmp = av_fetch( revcharmap, state, 0);
1509 PerlIO_printf( Perl_debug_log, "%*s",
1511 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
1512 PL_colors[0], PL_colors[1],
1513 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
1514 PERL_PV_ESCAPE_FIRSTCHAR
1519 PerlIO_printf( Perl_debug_log, "\n%*sState|-----------------------",
1520 (int)depth * 2 + 2,"");
1522 for( state = 0 ; state < trie->uniquecharcount ; state++ )
1523 PerlIO_printf( Perl_debug_log, "%.*s", colwidth, "--------");
1524 PerlIO_printf( Perl_debug_log, "\n");
1526 for( state = 1 ; state < trie->statecount ; state++ ) {
1527 const U32 base = trie->states[ state ].trans.base;
1529 PerlIO_printf( Perl_debug_log, "%*s#%4"UVXf"|",
1530 (int)depth * 2 + 2,"", (UV)state);
1532 if ( trie->states[ state ].wordnum ) {
1533 PerlIO_printf( Perl_debug_log, " W%4X",
1534 trie->states[ state ].wordnum );
1536 PerlIO_printf( Perl_debug_log, "%6s", "" );
1539 PerlIO_printf( Perl_debug_log, " @%4"UVXf" ", (UV)base );
1544 while( ( base + ofs < trie->uniquecharcount ) ||
1545 ( base + ofs - trie->uniquecharcount < trie->lasttrans
1546 && trie->trans[ base + ofs - trie->uniquecharcount ].check
1550 PerlIO_printf( Perl_debug_log, "+%2"UVXf"[ ", (UV)ofs);
1552 for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) {
1553 if ( ( base + ofs >= trie->uniquecharcount )
1554 && ( base + ofs - trie->uniquecharcount
1556 && trie->trans[ base + ofs
1557 - trie->uniquecharcount ].check == state )
1559 PerlIO_printf( Perl_debug_log, "%*"UVXf,
1561 (UV)trie->trans[ base + ofs
1562 - trie->uniquecharcount ].next );
1564 PerlIO_printf( Perl_debug_log, "%*s",colwidth," ." );
1568 PerlIO_printf( Perl_debug_log, "]");
1571 PerlIO_printf( Perl_debug_log, "\n" );
1573 PerlIO_printf(Perl_debug_log, "%*sword_info N:(prev,len)=",
1575 for (word=1; word <= trie->wordcount; word++) {
1576 PerlIO_printf(Perl_debug_log, " %d:(%d,%d)",
1577 (int)word, (int)(trie->wordinfo[word].prev),
1578 (int)(trie->wordinfo[word].len));
1580 PerlIO_printf(Perl_debug_log, "\n" );
1583 Dumps a fully constructed but uncompressed trie in list form.
1584 List tries normally only are used for construction when the number of
1585 possible chars (trie->uniquecharcount) is very high.
1586 Used for debugging make_trie().
1589 S_dump_trie_interim_list(pTHX_ const struct _reg_trie_data *trie,
1590 HV *widecharmap, AV *revcharmap, U32 next_alloc,
1594 SV *sv=sv_newmortal();
1595 int colwidth= widecharmap ? 6 : 4;
1596 GET_RE_DEBUG_FLAGS_DECL;
1598 PERL_ARGS_ASSERT_DUMP_TRIE_INTERIM_LIST;
1600 /* print out the table precompression. */
1601 PerlIO_printf( Perl_debug_log, "%*sState :Word | Transition Data\n%*s%s",
1602 (int)depth * 2 + 2,"", (int)depth * 2 + 2,"",
1603 "------:-----+-----------------\n" );
1605 for( state=1 ; state < next_alloc ; state ++ ) {
1608 PerlIO_printf( Perl_debug_log, "%*s %4"UVXf" :",
1609 (int)depth * 2 + 2,"", (UV)state );
1610 if ( ! trie->states[ state ].wordnum ) {
1611 PerlIO_printf( Perl_debug_log, "%5s| ","");
1613 PerlIO_printf( Perl_debug_log, "W%4x| ",
1614 trie->states[ state ].wordnum
1617 for( charid = 1 ; charid <= TRIE_LIST_USED( state ) ; charid++ ) {
1618 SV ** const tmp = av_fetch( revcharmap,
1619 TRIE_LIST_ITEM(state,charid).forid, 0);
1621 PerlIO_printf( Perl_debug_log, "%*s:%3X=%4"UVXf" | ",
1623 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp),
1625 PL_colors[0], PL_colors[1],
1626 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0)
1627 | PERL_PV_ESCAPE_FIRSTCHAR
1629 TRIE_LIST_ITEM(state,charid).forid,
1630 (UV)TRIE_LIST_ITEM(state,charid).newstate
1633 PerlIO_printf(Perl_debug_log, "\n%*s| ",
1634 (int)((depth * 2) + 14), "");
1637 PerlIO_printf( Perl_debug_log, "\n");
1642 Dumps a fully constructed but uncompressed trie in table form.
1643 This is the normal DFA style state transition table, with a few
1644 twists to facilitate compression later.
1645 Used for debugging make_trie().
1648 S_dump_trie_interim_table(pTHX_ const struct _reg_trie_data *trie,
1649 HV *widecharmap, AV *revcharmap, U32 next_alloc,
1654 SV *sv=sv_newmortal();
1655 int colwidth= widecharmap ? 6 : 4;
1656 GET_RE_DEBUG_FLAGS_DECL;
1658 PERL_ARGS_ASSERT_DUMP_TRIE_INTERIM_TABLE;
1661 print out the table precompression so that we can do a visual check
1662 that they are identical.
1665 PerlIO_printf( Perl_debug_log, "%*sChar : ",(int)depth * 2 + 2,"" );
1667 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
1668 SV ** const tmp = av_fetch( revcharmap, charid, 0);
1670 PerlIO_printf( Perl_debug_log, "%*s",
1672 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), colwidth,
1673 PL_colors[0], PL_colors[1],
1674 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
1675 PERL_PV_ESCAPE_FIRSTCHAR
1681 PerlIO_printf( Perl_debug_log, "\n%*sState+-",(int)depth * 2 + 2,"" );
1683 for( charid=0 ; charid < trie->uniquecharcount ; charid++ ) {
1684 PerlIO_printf( Perl_debug_log, "%.*s", colwidth,"--------");
1687 PerlIO_printf( Perl_debug_log, "\n" );
1689 for( state=1 ; state < next_alloc ; state += trie->uniquecharcount ) {
1691 PerlIO_printf( Perl_debug_log, "%*s%4"UVXf" : ",
1692 (int)depth * 2 + 2,"",
1693 (UV)TRIE_NODENUM( state ) );
1695 for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
1696 UV v=(UV)SAFE_TRIE_NODENUM( trie->trans[ state + charid ].next );
1698 PerlIO_printf( Perl_debug_log, "%*"UVXf, colwidth, v );
1700 PerlIO_printf( Perl_debug_log, "%*s", colwidth, "." );
1702 if ( ! trie->states[ TRIE_NODENUM( state ) ].wordnum ) {
1703 PerlIO_printf( Perl_debug_log, " (%4"UVXf")\n",
1704 (UV)trie->trans[ state ].check );
1706 PerlIO_printf( Perl_debug_log, " (%4"UVXf") W%4X\n",
1707 (UV)trie->trans[ state ].check,
1708 trie->states[ TRIE_NODENUM( state ) ].wordnum );
1716 /* make_trie(startbranch,first,last,tail,word_count,flags,depth)
1717 startbranch: the first branch in the whole branch sequence
1718 first : start branch of sequence of branch-exact nodes.
1719 May be the same as startbranch
1720 last : Thing following the last branch.
1721 May be the same as tail.
1722 tail : item following the branch sequence
1723 count : words in the sequence
1724 flags : currently the OP() type we will be building one of /EXACT(|F|Fl)/
1725 depth : indent depth
1727 Inplace optimizes a sequence of 2 or more Branch-Exact nodes into a TRIE node.
1729 A trie is an N'ary tree where the branches are determined by digital
1730 decomposition of the key. IE, at the root node you look up the 1st character and
1731 follow that branch repeat until you find the end of the branches. Nodes can be
1732 marked as "accepting" meaning they represent a complete word. Eg:
1736 would convert into the following structure. Numbers represent states, letters
1737 following numbers represent valid transitions on the letter from that state, if
1738 the number is in square brackets it represents an accepting state, otherwise it
1739 will be in parenthesis.
1741 +-h->+-e->[3]-+-r->(8)-+-s->[9]
1745 (1) +-i->(6)-+-s->[7]
1747 +-s->(3)-+-h->(4)-+-e->[5]
1749 Accept Word Mapping: 3=>1 (he),5=>2 (she), 7=>3 (his), 9=>4 (hers)
1751 This shows that when matching against the string 'hers' we will begin at state 1
1752 read 'h' and move to state 2, read 'e' and move to state 3 which is accepting,
1753 then read 'r' and go to state 8 followed by 's' which takes us to state 9 which
1754 is also accepting. Thus we know that we can match both 'he' and 'hers' with a
1755 single traverse. We store a mapping from accepting to state to which word was
1756 matched, and then when we have multiple possibilities we try to complete the
1757 rest of the regex in the order in which they occured in the alternation.
1759 The only prior NFA like behaviour that would be changed by the TRIE support is
1760 the silent ignoring of duplicate alternations which are of the form:
1762 / (DUPE|DUPE) X? (?{ ... }) Y /x
1764 Thus EVAL blocks following a trie may be called a different number of times with
1765 and without the optimisation. With the optimisations dupes will be silently
1766 ignored. This inconsistent behaviour of EVAL type nodes is well established as
1767 the following demonstrates:
1769 'words'=~/(word|word|word)(?{ print $1 })[xyz]/
1771 which prints out 'word' three times, but
1773 'words'=~/(word|word|word)(?{ print $1 })S/
1775 which doesnt print it out at all. This is due to other optimisations kicking in.
1777 Example of what happens on a structural level:
1779 The regexp /(ac|ad|ab)+/ will produce the following debug output:
1781 1: CURLYM[1] {1,32767}(18)
1792 This would be optimizable with startbranch=5, first=5, last=16, tail=16
1793 and should turn into:
1795 1: CURLYM[1] {1,32767}(18)
1797 [Words:3 Chars Stored:6 Unique Chars:4 States:5 NCP:1]
1805 Cases where tail != last would be like /(?foo|bar)baz/:
1815 which would be optimizable with startbranch=1, first=1, last=7, tail=8
1816 and would end up looking like:
1819 [Words:2 Chars Stored:6 Unique Chars:5 States:7 NCP:1]
1826 d = uvchr_to_utf8_flags(d, uv, 0);
1828 is the recommended Unicode-aware way of saying
1833 #define TRIE_STORE_REVCHAR(val) \
1836 SV *zlopp = newSV(7); /* XXX: optimize me */ \
1837 unsigned char *flrbbbbb = (unsigned char *) SvPVX(zlopp); \
1838 unsigned const char *const kapow = uvchr_to_utf8(flrbbbbb, val); \
1839 SvCUR_set(zlopp, kapow - flrbbbbb); \
1842 av_push(revcharmap, zlopp); \
1844 char ooooff = (char)val; \
1845 av_push(revcharmap, newSVpvn(&ooooff, 1)); \
1849 /* This gets the next character from the input, folding it if not already
1851 #define TRIE_READ_CHAR STMT_START { \
1854 /* if it is UTF then it is either already folded, or does not need \
1856 uvc = valid_utf8_to_uvchr( (const U8*) uc, &len); \
1858 else if (folder == PL_fold_latin1) { \
1859 /* This folder implies Unicode rules, which in the range expressible \
1860 * by not UTF is the lower case, with the two exceptions, one of \
1861 * which should have been taken care of before calling this */ \
1862 assert(*uc != LATIN_SMALL_LETTER_SHARP_S); \
1863 uvc = toLOWER_L1(*uc); \
1864 if (UNLIKELY(uvc == MICRO_SIGN)) uvc = GREEK_SMALL_LETTER_MU; \
1867 /* raw data, will be folded later if needed */ \
1875 #define TRIE_LIST_PUSH(state,fid,ns) STMT_START { \
1876 if ( TRIE_LIST_CUR( state ) >=TRIE_LIST_LEN( state ) ) { \
1877 U32 ging = TRIE_LIST_LEN( state ) *= 2; \
1878 Renew( trie->states[ state ].trans.list, ging, reg_trie_trans_le ); \
1880 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).forid = fid; \
1881 TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).newstate = ns; \
1882 TRIE_LIST_CUR( state )++; \
1885 #define TRIE_LIST_NEW(state) STMT_START { \
1886 Newxz( trie->states[ state ].trans.list, \
1887 4, reg_trie_trans_le ); \
1888 TRIE_LIST_CUR( state ) = 1; \
1889 TRIE_LIST_LEN( state ) = 4; \
1892 #define TRIE_HANDLE_WORD(state) STMT_START { \
1893 U16 dupe= trie->states[ state ].wordnum; \
1894 regnode * const noper_next = regnext( noper ); \
1897 /* store the word for dumping */ \
1899 if (OP(noper) != NOTHING) \
1900 tmp = newSVpvn_utf8(STRING(noper), STR_LEN(noper), UTF); \
1902 tmp = newSVpvn_utf8( "", 0, UTF ); \
1903 av_push( trie_words, tmp ); \
1907 trie->wordinfo[curword].prev = 0; \
1908 trie->wordinfo[curword].len = wordlen; \
1909 trie->wordinfo[curword].accept = state; \
1911 if ( noper_next < tail ) { \
1913 trie->jump = (U16 *) PerlMemShared_calloc( word_count + 1, \
1915 trie->jump[curword] = (U16)(noper_next - convert); \
1917 jumper = noper_next; \
1919 nextbranch= regnext(cur); \
1923 /* It's a dupe. Pre-insert into the wordinfo[].prev */\
1924 /* chain, so that when the bits of chain are later */\
1925 /* linked together, the dups appear in the chain */\
1926 trie->wordinfo[curword].prev = trie->wordinfo[dupe].prev; \
1927 trie->wordinfo[dupe].prev = curword; \
1929 /* we haven't inserted this word yet. */ \
1930 trie->states[ state ].wordnum = curword; \
1935 #define TRIE_TRANS_STATE(state,base,ucharcount,charid,special) \
1936 ( ( base + charid >= ucharcount \
1937 && base + charid < ubound \
1938 && state == trie->trans[ base - ucharcount + charid ].check \
1939 && trie->trans[ base - ucharcount + charid ].next ) \
1940 ? trie->trans[ base - ucharcount + charid ].next \
1941 : ( state==1 ? special : 0 ) \
1945 #define MADE_JUMP_TRIE 2
1946 #define MADE_EXACT_TRIE 4
1949 S_make_trie(pTHX_ RExC_state_t *pRExC_state, regnode *startbranch,
1950 regnode *first, regnode *last, regnode *tail,
1951 U32 word_count, U32 flags, U32 depth)
1954 /* first pass, loop through and scan words */
1955 reg_trie_data *trie;
1956 HV *widecharmap = NULL;
1957 AV *revcharmap = newAV();
1963 regnode *jumper = NULL;
1964 regnode *nextbranch = NULL;
1965 regnode *convert = NULL;
1966 U32 *prev_states; /* temp array mapping each state to previous one */
1967 /* we just use folder as a flag in utf8 */
1968 const U8 * folder = NULL;
1971 const U32 data_slot = add_data( pRExC_state, STR_WITH_LEN("tuuu"));
1972 AV *trie_words = NULL;
1973 /* along with revcharmap, this only used during construction but both are
1974 * useful during debugging so we store them in the struct when debugging.
1977 const U32 data_slot = add_data( pRExC_state, STR_WITH_LEN("tu"));
1978 STRLEN trie_charcount=0;
1980 SV *re_trie_maxbuff;
1981 GET_RE_DEBUG_FLAGS_DECL;
1983 PERL_ARGS_ASSERT_MAKE_TRIE;
1985 PERL_UNUSED_ARG(depth);
1992 case EXACTFU: folder = PL_fold_latin1; break;
1993 case EXACTF: folder = PL_fold; break;
1994 default: Perl_croak( aTHX_ "panic! In trie construction, unknown node type %u %s", (unsigned) flags, PL_reg_name[flags] );
1997 trie = (reg_trie_data *) PerlMemShared_calloc( 1, sizeof(reg_trie_data) );
1999 trie->startstate = 1;
2000 trie->wordcount = word_count;
2001 RExC_rxi->data->data[ data_slot ] = (void*)trie;
2002 trie->charmap = (U16 *) PerlMemShared_calloc( 256, sizeof(U16) );
2004 trie->bitmap = (char *) PerlMemShared_calloc( ANYOF_BITMAP_SIZE, 1 );
2005 trie->wordinfo = (reg_trie_wordinfo *) PerlMemShared_calloc(
2006 trie->wordcount+1, sizeof(reg_trie_wordinfo));
2009 trie_words = newAV();
2012 re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1);
2013 if (!SvIOK(re_trie_maxbuff)) {
2014 sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
2016 DEBUG_TRIE_COMPILE_r({
2017 PerlIO_printf( Perl_debug_log,
2018 "%*smake_trie start==%d, first==%d, last==%d, tail==%d depth=%d\n",
2019 (int)depth * 2 + 2, "",
2020 REG_NODE_NUM(startbranch),REG_NODE_NUM(first),
2021 REG_NODE_NUM(last), REG_NODE_NUM(tail), (int)depth);
2024 /* Find the node we are going to overwrite */
2025 if ( first == startbranch && OP( last ) != BRANCH ) {
2026 /* whole branch chain */
2029 /* branch sub-chain */
2030 convert = NEXTOPER( first );
2033 /* -- First loop and Setup --
2035 We first traverse the branches and scan each word to determine if it
2036 contains widechars, and how many unique chars there are, this is
2037 important as we have to build a table with at least as many columns as we
2040 We use an array of integers to represent the character codes 0..255
2041 (trie->charmap) and we use a an HV* to store Unicode characters. We use
2042 the native representation of the character value as the key and IV's for
2045 *TODO* If we keep track of how many times each character is used we can
2046 remap the columns so that the table compression later on is more
2047 efficient in terms of memory by ensuring the most common value is in the
2048 middle and the least common are on the outside. IMO this would be better
2049 than a most to least common mapping as theres a decent chance the most
2050 common letter will share a node with the least common, meaning the node
2051 will not be compressible. With a middle is most common approach the worst
2052 case is when we have the least common nodes twice.
2056 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
2057 regnode *noper = NEXTOPER( cur );
2058 const U8 *uc = (U8*)STRING( noper );
2059 const U8 *e = uc + STR_LEN( noper );
2061 U32 wordlen = 0; /* required init */
2062 STRLEN minchars = 0;
2063 STRLEN maxchars = 0;
2064 bool set_bit = trie->bitmap ? 1 : 0; /*store the first char in the
2067 if (OP(noper) == NOTHING) {
2068 regnode *noper_next= regnext(noper);
2069 if (noper_next != tail && OP(noper_next) == flags) {
2071 uc= (U8*)STRING(noper);
2072 e= uc + STR_LEN(noper);
2073 trie->minlen= STR_LEN(noper);
2080 if ( set_bit ) { /* bitmap only alloced when !(UTF&&Folding) */
2081 TRIE_BITMAP_SET(trie,*uc); /* store the raw first byte
2082 regardless of encoding */
2083 if (OP( noper ) == EXACTFU_SS) {
2084 /* false positives are ok, so just set this */
2085 TRIE_BITMAP_SET(trie, LATIN_SMALL_LETTER_SHARP_S);
2088 for ( ; uc < e ; uc += len ) { /* Look at each char in the current
2090 TRIE_CHARCOUNT(trie)++;
2093 /* TRIE_READ_CHAR returns the current character, or its fold if /i
2094 * is in effect. Under /i, this character can match itself, or
2095 * anything that folds to it. If not under /i, it can match just
2096 * itself. Most folds are 1-1, for example k, K, and KELVIN SIGN
2097 * all fold to k, and all are single characters. But some folds
2098 * expand to more than one character, so for example LATIN SMALL
2099 * LIGATURE FFI folds to the three character sequence 'ffi'. If
2100 * the string beginning at 'uc' is 'ffi', it could be matched by
2101 * three characters, or just by the one ligature character. (It
2102 * could also be matched by two characters: LATIN SMALL LIGATURE FF
2103 * followed by 'i', or by 'f' followed by LATIN SMALL LIGATURE FI).
2104 * (Of course 'I' and/or 'F' instead of 'i' and 'f' can also
2105 * match.) The trie needs to know the minimum and maximum number
2106 * of characters that could match so that it can use size alone to
2107 * quickly reject many match attempts. The max is simple: it is
2108 * the number of folded characters in this branch (since a fold is
2109 * never shorter than what folds to it. */
2113 /* And the min is equal to the max if not under /i (indicated by
2114 * 'folder' being NULL), or there are no multi-character folds. If
2115 * there is a multi-character fold, the min is incremented just
2116 * once, for the character that folds to the sequence. Each
2117 * character in the sequence needs to be added to the list below of
2118 * characters in the trie, but we count only the first towards the
2119 * min number of characters needed. This is done through the
2120 * variable 'foldlen', which is returned by the macros that look
2121 * for these sequences as the number of bytes the sequence
2122 * occupies. Each time through the loop, we decrement 'foldlen' by
2123 * how many bytes the current char occupies. Only when it reaches
2124 * 0 do we increment 'minchars' or look for another multi-character
2126 if (folder == NULL) {
2129 else if (foldlen > 0) {
2130 foldlen -= (UTF) ? UTF8SKIP(uc) : 1;
2135 /* See if *uc is the beginning of a multi-character fold. If
2136 * so, we decrement the length remaining to look at, to account
2137 * for the current character this iteration. (We can use 'uc'
2138 * instead of the fold returned by TRIE_READ_CHAR because for
2139 * non-UTF, the latin1_safe macro is smart enough to account
2140 * for all the unfolded characters, and because for UTF, the
2141 * string will already have been folded earlier in the
2142 * compilation process */
2144 if ((foldlen = is_MULTI_CHAR_FOLD_utf8_safe(uc, e))) {
2145 foldlen -= UTF8SKIP(uc);
2148 else if ((foldlen = is_MULTI_CHAR_FOLD_latin1_safe(uc, e))) {
2153 /* The current character (and any potential folds) should be added
2154 * to the possible matching characters for this position in this
2158 U8 folded= folder[ (U8) uvc ];
2159 if ( !trie->charmap[ folded ] ) {
2160 trie->charmap[ folded ]=( ++trie->uniquecharcount );
2161 TRIE_STORE_REVCHAR( folded );
2164 if ( !trie->charmap[ uvc ] ) {
2165 trie->charmap[ uvc ]=( ++trie->uniquecharcount );
2166 TRIE_STORE_REVCHAR( uvc );
2169 /* store the codepoint in the bitmap, and its folded
2171 TRIE_BITMAP_SET(trie, uvc);
2173 /* store the folded codepoint */
2174 if ( folder ) TRIE_BITMAP_SET(trie, folder[(U8) uvc ]);
2177 /* store first byte of utf8 representation of
2178 variant codepoints */
2179 if (! UVCHR_IS_INVARIANT(uvc)) {
2180 TRIE_BITMAP_SET(trie, UTF8_TWO_BYTE_HI(uvc));
2183 set_bit = 0; /* We've done our bit :-) */
2187 /* XXX We could come up with the list of code points that fold
2188 * to this using PL_utf8_foldclosures, except not for
2189 * multi-char folds, as there may be multiple combinations
2190 * there that could work, which needs to wait until runtime to
2191 * resolve (The comment about LIGATURE FFI above is such an
2196 widecharmap = newHV();
2198 svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 1 );
2201 Perl_croak( aTHX_ "error creating/fetching widecharmap entry for 0x%"UVXf, uvc );
2203 if ( !SvTRUE( *svpp ) ) {
2204 sv_setiv( *svpp, ++trie->uniquecharcount );
2205 TRIE_STORE_REVCHAR(uvc);
2208 } /* end loop through characters in this branch of the trie */
2210 /* We take the min and max for this branch and combine to find the min
2211 * and max for all branches processed so far */
2212 if( cur == first ) {
2213 trie->minlen = minchars;
2214 trie->maxlen = maxchars;
2215 } else if (minchars < trie->minlen) {
2216 trie->minlen = minchars;
2217 } else if (maxchars > trie->maxlen) {
2218 trie->maxlen = maxchars;
2220 } /* end first pass */
2221 DEBUG_TRIE_COMPILE_r(
2222 PerlIO_printf( Perl_debug_log,
2223 "%*sTRIE(%s): W:%d C:%d Uq:%d Min:%d Max:%d\n",
2224 (int)depth * 2 + 2,"",
2225 ( widecharmap ? "UTF8" : "NATIVE" ), (int)word_count,
2226 (int)TRIE_CHARCOUNT(trie), trie->uniquecharcount,
2227 (int)trie->minlen, (int)trie->maxlen )
2231 We now know what we are dealing with in terms of unique chars and
2232 string sizes so we can calculate how much memory a naive
2233 representation using a flat table will take. If it's over a reasonable
2234 limit (as specified by ${^RE_TRIE_MAXBUF}) we use a more memory
2235 conservative but potentially much slower representation using an array
2238 At the end we convert both representations into the same compressed
2239 form that will be used in regexec.c for matching with. The latter
2240 is a form that cannot be used to construct with but has memory
2241 properties similar to the list form and access properties similar
2242 to the table form making it both suitable for fast searches and
2243 small enough that its feasable to store for the duration of a program.
2245 See the comment in the code where the compressed table is produced
2246 inplace from the flat tabe representation for an explanation of how
2247 the compression works.
2252 Newx(prev_states, TRIE_CHARCOUNT(trie) + 2, U32);
2255 if ( (IV)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1)
2256 > SvIV(re_trie_maxbuff) )
2259 Second Pass -- Array Of Lists Representation
2261 Each state will be represented by a list of charid:state records
2262 (reg_trie_trans_le) the first such element holds the CUR and LEN
2263 points of the allocated array. (See defines above).
2265 We build the initial structure using the lists, and then convert
2266 it into the compressed table form which allows faster lookups
2267 (but cant be modified once converted).
2270 STRLEN transcount = 1;
2272 DEBUG_TRIE_COMPILE_MORE_r( PerlIO_printf( Perl_debug_log,
2273 "%*sCompiling trie using list compiler\n",
2274 (int)depth * 2 + 2, ""));
2276 trie->states = (reg_trie_state *)
2277 PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
2278 sizeof(reg_trie_state) );
2282 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
2284 regnode *noper = NEXTOPER( cur );
2285 U8 *uc = (U8*)STRING( noper );
2286 const U8 *e = uc + STR_LEN( noper );
2287 U32 state = 1; /* required init */
2288 U16 charid = 0; /* sanity init */
2289 U32 wordlen = 0; /* required init */
2291 if (OP(noper) == NOTHING) {
2292 regnode *noper_next= regnext(noper);
2293 if (noper_next != tail && OP(noper_next) == flags) {
2295 uc= (U8*)STRING(noper);
2296 e= uc + STR_LEN(noper);
2300 if (OP(noper) != NOTHING) {
2301 for ( ; uc < e ; uc += len ) {
2306 charid = trie->charmap[ uvc ];
2308 SV** const svpp = hv_fetch( widecharmap,
2315 charid=(U16)SvIV( *svpp );
2318 /* charid is now 0 if we dont know the char read, or
2319 * nonzero if we do */
2326 if ( !trie->states[ state ].trans.list ) {
2327 TRIE_LIST_NEW( state );
2330 check <= TRIE_LIST_USED( state );
2333 if ( TRIE_LIST_ITEM( state, check ).forid
2336 newstate = TRIE_LIST_ITEM( state, check ).newstate;
2341 newstate = next_alloc++;
2342 prev_states[newstate] = state;
2343 TRIE_LIST_PUSH( state, charid, newstate );
2348 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %"IVdf, uvc );
2352 TRIE_HANDLE_WORD(state);
2354 } /* end second pass */
2356 /* next alloc is the NEXT state to be allocated */
2357 trie->statecount = next_alloc;
2358 trie->states = (reg_trie_state *)
2359 PerlMemShared_realloc( trie->states,
2361 * sizeof(reg_trie_state) );
2363 /* and now dump it out before we compress it */
2364 DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_list(trie, widecharmap,
2365 revcharmap, next_alloc,
2369 trie->trans = (reg_trie_trans *)
2370 PerlMemShared_calloc( transcount, sizeof(reg_trie_trans) );
2377 for( state=1 ; state < next_alloc ; state ++ ) {
2381 DEBUG_TRIE_COMPILE_MORE_r(
2382 PerlIO_printf( Perl_debug_log, "tp: %d zp: %d ",tp,zp)
2386 if (trie->states[state].trans.list) {
2387 U16 minid=TRIE_LIST_ITEM( state, 1).forid;
2391 for( idx = 2 ; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
2392 const U16 forid = TRIE_LIST_ITEM( state, idx).forid;
2393 if ( forid < minid ) {
2395 } else if ( forid > maxid ) {
2399 if ( transcount < tp + maxid - minid + 1) {
2401 trie->trans = (reg_trie_trans *)
2402 PerlMemShared_realloc( trie->trans,
2404 * sizeof(reg_trie_trans) );
2405 Zero( trie->trans + (transcount / 2),
2409 base = trie->uniquecharcount + tp - minid;
2410 if ( maxid == minid ) {
2412 for ( ; zp < tp ; zp++ ) {
2413 if ( ! trie->trans[ zp ].next ) {
2414 base = trie->uniquecharcount + zp - minid;
2415 trie->trans[ zp ].next = TRIE_LIST_ITEM( state,
2417 trie->trans[ zp ].check = state;
2423 trie->trans[ tp ].next = TRIE_LIST_ITEM( state,
2425 trie->trans[ tp ].check = state;
2430 for ( idx=1; idx <= TRIE_LIST_USED( state ) ; idx++ ) {
2431 const U32 tid = base
2432 - trie->uniquecharcount
2433 + TRIE_LIST_ITEM( state, idx ).forid;
2434 trie->trans[ tid ].next = TRIE_LIST_ITEM( state,
2436 trie->trans[ tid ].check = state;
2438 tp += ( maxid - minid + 1 );
2440 Safefree(trie->states[ state ].trans.list);
2443 DEBUG_TRIE_COMPILE_MORE_r(
2444 PerlIO_printf( Perl_debug_log, " base: %d\n",base);
2447 trie->states[ state ].trans.base=base;
2449 trie->lasttrans = tp + 1;
2453 Second Pass -- Flat Table Representation.
2455 we dont use the 0 slot of either trans[] or states[] so we add 1 to
2456 each. We know that we will need Charcount+1 trans at most to store
2457 the data (one row per char at worst case) So we preallocate both
2458 structures assuming worst case.
2460 We then construct the trie using only the .next slots of the entry
2463 We use the .check field of the first entry of the node temporarily
2464 to make compression both faster and easier by keeping track of how
2465 many non zero fields are in the node.
2467 Since trans are numbered from 1 any 0 pointer in the table is a FAIL
2470 There are two terms at use here: state as a TRIE_NODEIDX() which is
2471 a number representing the first entry of the node, and state as a
2472 TRIE_NODENUM() which is the trans number. state 1 is TRIE_NODEIDX(1)
2473 and TRIE_NODENUM(1), state 2 is TRIE_NODEIDX(2) and TRIE_NODENUM(3)
2474 if there are 2 entrys per node. eg:
2482 The table is internally in the right hand, idx form. However as we
2483 also have to deal with the states array which is indexed by nodenum
2484 we have to use TRIE_NODENUM() to convert.
2487 DEBUG_TRIE_COMPILE_MORE_r( PerlIO_printf( Perl_debug_log,
2488 "%*sCompiling trie using table compiler\n",
2489 (int)depth * 2 + 2, ""));
2491 trie->trans = (reg_trie_trans *)
2492 PerlMemShared_calloc( ( TRIE_CHARCOUNT(trie) + 1 )
2493 * trie->uniquecharcount + 1,
2494 sizeof(reg_trie_trans) );
2495 trie->states = (reg_trie_state *)
2496 PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
2497 sizeof(reg_trie_state) );
2498 next_alloc = trie->uniquecharcount + 1;
2501 for ( cur = first ; cur < last ; cur = regnext( cur ) ) {
2503 regnode *noper = NEXTOPER( cur );
2504 const U8 *uc = (U8*)STRING( noper );
2505 const U8 *e = uc + STR_LEN( noper );
2507 U32 state = 1; /* required init */
2509 U16 charid = 0; /* sanity init */
2510 U32 accept_state = 0; /* sanity init */
2512 U32 wordlen = 0; /* required init */
2514 if (OP(noper) == NOTHING) {
2515 regnode *noper_next= regnext(noper);
2516 if (noper_next != tail && OP(noper_next) == flags) {
2518 uc= (U8*)STRING(noper);
2519 e= uc + STR_LEN(noper);
2523 if ( OP(noper) != NOTHING ) {
2524 for ( ; uc < e ; uc += len ) {
2529 charid = trie->charmap[ uvc ];
2531 SV* const * const svpp = hv_fetch( widecharmap,
2535 charid = svpp ? (U16)SvIV(*svpp) : 0;
2539 if ( !trie->trans[ state + charid ].next ) {
2540 trie->trans[ state + charid ].next = next_alloc;
2541 trie->trans[ state ].check++;
2542 prev_states[TRIE_NODENUM(next_alloc)]
2543 = TRIE_NODENUM(state);
2544 next_alloc += trie->uniquecharcount;
2546 state = trie->trans[ state + charid ].next;
2548 Perl_croak( aTHX_ "panic! In trie construction, no char mapping for %"IVdf, uvc );
2550 /* charid is now 0 if we dont know the char read, or
2551 * nonzero if we do */
2554 accept_state = TRIE_NODENUM( state );
2555 TRIE_HANDLE_WORD(accept_state);
2557 } /* end second pass */
2559 /* and now dump it out before we compress it */
2560 DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_table(trie, widecharmap,
2562 next_alloc, depth+1));
2566 * Inplace compress the table.*
2568 For sparse data sets the table constructed by the trie algorithm will
2569 be mostly 0/FAIL transitions or to put it another way mostly empty.
2570 (Note that leaf nodes will not contain any transitions.)
2572 This algorithm compresses the tables by eliminating most such
2573 transitions, at the cost of a modest bit of extra work during lookup:
2575 - Each states[] entry contains a .base field which indicates the
2576 index in the state[] array wheres its transition data is stored.
2578 - If .base is 0 there are no valid transitions from that node.
2580 - If .base is nonzero then charid is added to it to find an entry in
2583 -If trans[states[state].base+charid].check!=state then the
2584 transition is taken to be a 0/Fail transition. Thus if there are fail
2585 transitions at the front of the node then the .base offset will point
2586 somewhere inside the previous nodes data (or maybe even into a node
2587 even earlier), but the .check field determines if the transition is
2591 The following process inplace converts the table to the compressed
2592 table: We first do not compress the root node 1,and mark all its
2593 .check pointers as 1 and set its .base pointer as 1 as well. This
2594 allows us to do a DFA construction from the compressed table later,
2595 and ensures that any .base pointers we calculate later are greater
2598 - We set 'pos' to indicate the first entry of the second node.
2600 - We then iterate over the columns of the node, finding the first and
2601 last used entry at l and m. We then copy l..m into pos..(pos+m-l),
2602 and set the .check pointers accordingly, and advance pos
2603 appropriately and repreat for the next node. Note that when we copy
2604 the next pointers we have to convert them from the original
2605 NODEIDX form to NODENUM form as the former is not valid post
2608 - If a node has no transitions used we mark its base as 0 and do not
2609 advance the pos pointer.
2611 - If a node only has one transition we use a second pointer into the
2612 structure to fill in allocated fail transitions from other states.
2613 This pointer is independent of the main pointer and scans forward
2614 looking for null transitions that are allocated to a state. When it
2615 finds one it writes the single transition into the "hole". If the
2616 pointer doesnt find one the single transition is appended as normal.
2618 - Once compressed we can Renew/realloc the structures to release the
2621 See "Table-Compression Methods" in sec 3.9 of the Red Dragon,
2622 specifically Fig 3.47 and the associated pseudocode.
2626 const U32 laststate = TRIE_NODENUM( next_alloc );
2629 trie->statecount = laststate;
2631 for ( state = 1 ; state < laststate ; state++ ) {
2633 const U32 stateidx = TRIE_NODEIDX( state );
2634 const U32 o_used = trie->trans[ stateidx ].check;
2635 U32 used = trie->trans[ stateidx ].check;
2636 trie->trans[ stateidx ].check = 0;
2639 used && charid < trie->uniquecharcount;
2642 if ( flag || trie->trans[ stateidx + charid ].next ) {
2643 if ( trie->trans[ stateidx + charid ].next ) {
2645 for ( ; zp < pos ; zp++ ) {
2646 if ( ! trie->trans[ zp ].next ) {
2650 trie->states[ state ].trans.base
2652 + trie->uniquecharcount
2654 trie->trans[ zp ].next
2655 = SAFE_TRIE_NODENUM( trie->trans[ stateidx
2657 trie->trans[ zp ].check = state;
2658 if ( ++zp > pos ) pos = zp;
2665 trie->states[ state ].trans.base
2666 = pos + trie->uniquecharcount - charid ;
2668 trie->trans[ pos ].next
2669 = SAFE_TRIE_NODENUM(
2670 trie->trans[ stateidx + charid ].next );
2671 trie->trans[ pos ].check = state;
2676 trie->lasttrans = pos + 1;
2677 trie->states = (reg_trie_state *)
2678 PerlMemShared_realloc( trie->states, laststate
2679 * sizeof(reg_trie_state) );
2680 DEBUG_TRIE_COMPILE_MORE_r(
2681 PerlIO_printf( Perl_debug_log,
2682 "%*sAlloc: %d Orig: %"IVdf" elements, Final:%"IVdf". Savings of %%%5.2f\n",
2683 (int)depth * 2 + 2,"",
2684 (int)( ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount
2688 ( ( next_alloc - pos ) * 100 ) / (double)next_alloc );
2691 } /* end table compress */
2693 DEBUG_TRIE_COMPILE_MORE_r(
2694 PerlIO_printf(Perl_debug_log,
2695 "%*sStatecount:%"UVxf" Lasttrans:%"UVxf"\n",
2696 (int)depth * 2 + 2, "",
2697 (UV)trie->statecount,
2698 (UV)trie->lasttrans)
2700 /* resize the trans array to remove unused space */
2701 trie->trans = (reg_trie_trans *)
2702 PerlMemShared_realloc( trie->trans, trie->lasttrans
2703 * sizeof(reg_trie_trans) );
2705 { /* Modify the program and insert the new TRIE node */
2706 U8 nodetype =(U8)(flags & 0xFF);
2710 regnode *optimize = NULL;
2711 #ifdef RE_TRACK_PATTERN_OFFSETS
2714 U32 mjd_nodelen = 0;
2715 #endif /* RE_TRACK_PATTERN_OFFSETS */
2716 #endif /* DEBUGGING */
2718 This means we convert either the first branch or the first Exact,
2719 depending on whether the thing following (in 'last') is a branch
2720 or not and whther first is the startbranch (ie is it a sub part of
2721 the alternation or is it the whole thing.)
2722 Assuming its a sub part we convert the EXACT otherwise we convert
2723 the whole branch sequence, including the first.
2725 /* Find the node we are going to overwrite */
2726 if ( first != startbranch || OP( last ) == BRANCH ) {
2727 /* branch sub-chain */
2728 NEXT_OFF( first ) = (U16)(last - first);
2729 #ifdef RE_TRACK_PATTERN_OFFSETS
2731 mjd_offset= Node_Offset((convert));
2732 mjd_nodelen= Node_Length((convert));
2735 /* whole branch chain */
2737 #ifdef RE_TRACK_PATTERN_OFFSETS
2740 const regnode *nop = NEXTOPER( convert );
2741 mjd_offset= Node_Offset((nop));
2742 mjd_nodelen= Node_Length((nop));
2746 PerlIO_printf(Perl_debug_log,
2747 "%*sMJD offset:%"UVuf" MJD length:%"UVuf"\n",
2748 (int)depth * 2 + 2, "",
2749 (UV)mjd_offset, (UV)mjd_nodelen)
2752 /* But first we check to see if there is a common prefix we can
2753 split out as an EXACT and put in front of the TRIE node. */
2754 trie->startstate= 1;
2755 if ( trie->bitmap && !widecharmap && !trie->jump ) {
2757 for ( state = 1 ; state < trie->statecount-1 ; state++ ) {
2761 const U32 base = trie->states[ state ].trans.base;
2763 if ( trie->states[state].wordnum )
2766 for ( ofs = 0 ; ofs < trie->uniquecharcount ; ofs++ ) {
2767 if ( ( base + ofs >= trie->uniquecharcount ) &&
2768 ( base + ofs - trie->uniquecharcount < trie->lasttrans ) &&
2769 trie->trans[ base + ofs - trie->uniquecharcount ].check == state )
2771 if ( ++count > 1 ) {
2772 SV **tmp = av_fetch( revcharmap, ofs, 0);
2773 const U8 *ch = (U8*)SvPV_nolen_const( *tmp );
2774 if ( state == 1 ) break;
2776 Zero(trie->bitmap, ANYOF_BITMAP_SIZE, char);
2778 PerlIO_printf(Perl_debug_log,
2779 "%*sNew Start State=%"UVuf" Class: [",
2780 (int)depth * 2 + 2, "",
2783 SV ** const tmp = av_fetch( revcharmap, idx, 0);
2784 const U8 * const ch = (U8*)SvPV_nolen_const( *tmp );
2786 TRIE_BITMAP_SET(trie,*ch);
2788 TRIE_BITMAP_SET(trie, folder[ *ch ]);
2790 PerlIO_printf(Perl_debug_log, "%s", (char*)ch)
2794 TRIE_BITMAP_SET(trie,*ch);
2796 TRIE_BITMAP_SET(trie,folder[ *ch ]);
2797 DEBUG_OPTIMISE_r(PerlIO_printf( Perl_debug_log,"%s", ch));
2803 SV **tmp = av_fetch( revcharmap, idx, 0);
2805 char *ch = SvPV( *tmp, len );
2807 SV *sv=sv_newmortal();
2808 PerlIO_printf( Perl_debug_log,
2809 "%*sPrefix State: %"UVuf" Idx:%"UVuf" Char='%s'\n",
2810 (int)depth * 2 + 2, "",
2812 pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 6,
2813 PL_colors[0], PL_colors[1],
2814 (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
2815 PERL_PV_ESCAPE_FIRSTCHAR
2820 OP( convert ) = nodetype;
2821 str=STRING(convert);
2824 STR_LEN(convert) += len;
2830 DEBUG_OPTIMISE_r(PerlIO_printf( Perl_debug_log,"]\n"));
2835 trie->prefixlen = (state-1);
2837 regnode *n = convert+NODE_SZ_STR(convert);
2838 NEXT_OFF(convert) = NODE_SZ_STR(convert);
2839 trie->startstate = state;
2840 trie->minlen -= (state - 1);
2841 trie->maxlen -= (state - 1);
2843 /* At least the UNICOS C compiler choked on this
2844 * being argument to DEBUG_r(), so let's just have
2847 #ifdef PERL_EXT_RE_BUILD
2853 regnode *fix = convert;
2854 U32 word = trie->wordcount;
2856 Set_Node_Offset_Length(convert, mjd_offset, state - 1);
2857 while( ++fix < n ) {
2858 Set_Node_Offset_Length(fix, 0, 0);
2861 SV ** const tmp = av_fetch( trie_words, word, 0 );
2863 if ( STR_LEN(convert) <= SvCUR(*tmp) )
2864 sv_chop(*tmp, SvPV_nolen(*tmp) + STR_LEN(convert));
2866 sv_chop(*tmp, SvPV_nolen(*tmp) + SvCUR(*tmp));
2874 NEXT_OFF(convert) = (U16)(tail - convert);
2875 DEBUG_r(optimize= n);
2881 if ( trie->maxlen ) {
2882 NEXT_OFF( convert ) = (U16)(tail - convert);
2883 ARG_SET( convert, data_slot );
2884 /* Store the offset to the first unabsorbed branch in
2885 jump[0], which is otherwise unused by the jump logic.
2886 We use this when dumping a trie and during optimisation. */
2888 trie->jump[0] = (U16)(nextbranch - convert);
2890 /* If the start state is not accepting (meaning there is no empty string/NOTHING)
2891 * and there is a bitmap
2892 * and the first "jump target" node we found leaves enough room
2893 * then convert the TRIE node into a TRIEC node, with the bitmap
2894 * embedded inline in the opcode - this is hypothetically faster.
2896 if ( !trie->states[trie->startstate].wordnum
2898 && ( (char *)jumper - (char *)convert) >= (int)sizeof(struct regnode_charclass) )
2900 OP( convert ) = TRIEC;
2901 Copy(trie->bitmap, ((struct regnode_charclass *)convert)->bitmap, ANYOF_BITMAP_SIZE, char);
2902 PerlMemShared_free(trie->bitmap);
2905 OP( convert ) = TRIE;
2907 /* store the type in the flags */
2908 convert->flags = nodetype;
2912 + regarglen[ OP( convert ) ];
2914 /* XXX We really should free up the resource in trie now,
2915 as we won't use them - (which resources?) dmq */
2917 /* needed for dumping*/
2918 DEBUG_r(if (optimize) {
2919 regnode *opt = convert;
2921 while ( ++opt < optimize) {
2922 Set_Node_Offset_Length(opt,0,0);
2925 Try to clean up some of the debris left after the
2928 while( optimize < jumper ) {
2929 mjd_nodelen += Node_Length((optimize));
2930 OP( optimize ) = OPTIMIZED;
2931 Set_Node_Offset_Length(optimize,0,0);
2934 Set_Node_Offset_Length(convert,mjd_offset,mjd_nodelen);
2936 } /* end node insert */
2937 REH_CALL_COMP_NODE_HOOK(pRExC_state->rx, convert);
2939 /* Finish populating the prev field of the wordinfo array. Walk back
2940 * from each accept state until we find another accept state, and if
2941 * so, point the first word's .prev field at the second word. If the
2942 * second already has a .prev field set, stop now. This will be the
2943 * case either if we've already processed that word's accept state,
2944 * or that state had multiple words, and the overspill words were
2945 * already linked up earlier.
2952 for (word=1; word <= trie->wordcount; word++) {
2954 if (trie->wordinfo[word].prev)
2956 state = trie->wordinfo[word].accept;
2958 state = prev_states[state];
2961 prev = trie->states[state].wordnum;
2965 trie->wordinfo[word].prev = prev;
2967 Safefree(prev_states);
2971 /* and now dump out the compressed format */
2972 DEBUG_TRIE_COMPILE_r(dump_trie(trie, widecharmap, revcharmap, depth+1));
2974 RExC_rxi->data->data[ data_slot + 1 ] = (void*)widecharmap;
2976 RExC_rxi->data->data[ data_slot + TRIE_WORDS_OFFSET ] = (void*)trie_words;
2977 RExC_rxi->data->data[ data_slot + 3 ] = (void*)revcharmap;
2979 SvREFCNT_dec_NN(revcharmap);
2983 : trie->startstate>1
2989 S_make_trie_failtable(pTHX_ RExC_state_t *pRExC_state, regnode *source, regnode *stclass, U32 depth)
2991 /* The Trie is constructed and compressed now so we can build a fail array if
2994 This is basically the Aho-Corasick algorithm. Its from exercise 3.31 and
2996 "Red Dragon" -- Compilers, principles, techniques, and tools. Aho, Sethi,
3000 We find the fail state for each state in the trie, this state is the longest
3001 proper suffix of the current state's 'word' that is also a proper prefix of
3002 another word in our trie. State 1 represents the word '' and is thus the
3003 default fail state. This allows the DFA not to have to restart after its
3004 tried and failed a word at a given point, it simply continues as though it
3005 had been matching the other word in the first place.
3007 'abcdgu'=~/abcdefg|cdgu/
3008 When we get to 'd' we are still matching the first word, we would encounter
3009 'g' which would fail, which would bring us to the state representing 'd' in
3010 the second word where we would try 'g' and succeed, proceeding to match
3013 /* add a fail transition */
3014 const U32 trie_offset = ARG(source);
3015 reg_trie_data *trie=(reg_trie_data *)RExC_rxi->data->data[trie_offset];
3017 const U32 ucharcount = trie->uniquecharcount;
3018 const U32 numstates = trie->statecount;
3019 const U32 ubound = trie->lasttrans + ucharcount;
3023 U32 base = trie->states[ 1 ].trans.base;
3026 const U32 data_slot = add_data( pRExC_state, STR_WITH_LEN("T"));
3027 GET_RE_DEBUG_FLAGS_DECL;
3029 PERL_ARGS_ASSERT_MAKE_TRIE_FAILTABLE;
3031 PERL_UNUSED_ARG(depth);
3035 ARG_SET( stclass, data_slot );
3036 aho = (reg_ac_data *) PerlMemShared_calloc( 1, sizeof(reg_ac_data) );
3037 RExC_rxi->data->data[ data_slot ] = (void*)aho;
3038 aho->trie=trie_offset;
3039 aho->states=(reg_trie_state *)PerlMemShared_malloc( numstates * sizeof(reg_trie_state) );
3040 Copy( trie->states, aho->states, numstates, reg_trie_state );
3041 Newxz( q, numstates, U32);
3042 aho->fail = (U32 *) PerlMemShared_calloc( numstates, sizeof(U32) );
3045 /* initialize fail[0..1] to be 1 so that we always have
3046 a valid final fail state */
3047 fail[ 0 ] = fail[ 1 ] = 1;
3049 for ( charid = 0; charid < ucharcount ; charid++ ) {
3050 const U32 newstate = TRIE_TRANS_STATE( 1, base, ucharcount, charid, 0 );
3052 q[ q_write ] = newstate;
3053 /* set to point at the root */
3054 fail[ q[ q_write++ ] ]=1;
3057 while ( q_read < q_write) {
3058 const U32 cur = q[ q_read++ % numstates ];
3059 base = trie->states[ cur ].trans.base;
3061 for ( charid = 0 ; charid < ucharcount ; charid++ ) {
3062 const U32 ch_state = TRIE_TRANS_STATE( cur, base, ucharcount, charid, 1 );
3064 U32 fail_state = cur;
3067 fail_state = fail[ fail_state ];
3068 fail_base = aho->states[ fail_state ].trans.base;
3069 } while ( !TRIE_TRANS_STATE( fail_state, fail_base, ucharcount, charid, 1 ) );
3071 fail_state = TRIE_TRANS_STATE( fail_state, fail_base, ucharcount, charid, 1 );
3072 fail[ ch_state ] = fail_state;
3073 if ( !aho->states[ ch_state ].wordnum && aho->states[ fail_state ].wordnum )
3075 aho->states[ ch_state ].wordnum = aho->states[ fail_state ].wordnum;
3077 q[ q_write++ % numstates] = ch_state;
3081 /* restore fail[0..1] to 0 so that we "fall out" of the AC loop
3082 when we fail in state 1, this allows us to use the
3083 charclass scan to find a valid start char. This is based on the principle
3084 that theres a good chance the string being searched contains lots of stuff
3085 that cant be a start char.
3087 fail[ 0 ] = fail[ 1 ] = 0;
3088 DEBUG_TRIE_COMPILE_r({
3089 PerlIO_printf(Perl_debug_log,
3090 "%*sStclass Failtable (%"UVuf" states): 0",
3091 (int)(depth * 2), "", (UV)numstates
3093 for( q_read=1; q_read<numstates; q_read++ ) {
3094 PerlIO_printf(Perl_debug_log, ", %"UVuf, (UV)fail[q_read]);
3096 PerlIO_printf(Perl_debug_log, "\n");
3099 /*RExC_seen |= REG_TRIEDFA_SEEN;*/
3103 #define DEBUG_PEEP(str,scan,depth) \
3104 DEBUG_OPTIMISE_r({if (scan){ \
3105 SV * const mysv=sv_newmortal(); \
3106 regnode *Next = regnext(scan); \
3107 regprop(RExC_rx, mysv, scan, NULL); \
3108 PerlIO_printf(Perl_debug_log, "%*s" str ">%3d: %s (%d)\n", \
3109 (int)depth*2, "", REG_NODE_NUM(scan), SvPV_nolen_const(mysv),\
3110 Next ? (REG_NODE_NUM(Next)) : 0 ); \
3114 /* The below joins as many adjacent EXACTish nodes as possible into a single
3115 * one. The regop may be changed if the node(s) contain certain sequences that
3116 * require special handling. The joining is only done if:
3117 * 1) there is room in the current conglomerated node to entirely contain the
3119 * 2) they are the exact same node type
3121 * The adjacent nodes actually may be separated by NOTHING-kind nodes, and
3122 * these get optimized out
3124 * If a node is to match under /i (folded), the number of characters it matches
3125 * can be different than its character length if it contains a multi-character
3126 * fold. *min_subtract is set to the total delta number of characters of the
3129 * And *unfolded_multi_char is set to indicate whether or not the node contains
3130 * an unfolded multi-char fold. This happens when whether the fold is valid or
3131 * not won't be known until runtime; namely for EXACTF nodes that contain LATIN
3132 * SMALL LETTER SHARP S, as only if the target string being matched against
3133 * turns out to be UTF-8 is that fold valid; and also for EXACTFL nodes whose
3134 * folding rules depend on the locale in force at runtime. (Multi-char folds
3135 * whose components are all above the Latin1 range are not run-time locale
3136 * dependent, and have already been folded by the time this function is
3139 * This is as good a place as any to discuss the design of handling these
3140 * multi-character fold sequences. It's been wrong in Perl for a very long
3141 * time. There are three code points in Unicode whose multi-character folds
3142 * were long ago discovered to mess things up. The previous designs for
3143 * dealing with these involved assigning a special node for them. This
3144 * approach doesn't always work, as evidenced by this example:
3145 * "\xDFs" =~ /s\xDF/ui # Used to fail before these patches
3146 * Both sides fold to "sss", but if the pattern is parsed to create a node that
3147 * would match just the \xDF, it won't be able to handle the case where a
3148 * successful match would have to cross the node's boundary. The new approach
3149 * that hopefully generally solves the problem generates an EXACTFU_SS node
3150 * that is "sss" in this case.
3152 * It turns out that there are problems with all multi-character folds, and not
3153 * just these three. Now the code is general, for all such cases. The
3154 * approach taken is:
3155 * 1) This routine examines each EXACTFish node that could contain multi-
3156 * character folded sequences. Since a single character can fold into
3157 * such a sequence, the minimum match length for this node is less than
3158 * the number of characters in the node. This routine returns in
3159 * *min_subtract how many characters to subtract from the the actual
3160 * length of the string to get a real minimum match length; it is 0 if
3161 * there are no multi-char foldeds. This delta is used by the caller to
3162 * adjust the min length of the match, and the delta between min and max,
3163 * so that the optimizer doesn't reject these possibilities based on size
3165 * 2) For the sequence involving the Sharp s (\xDF), the node type EXACTFU_SS
3166 * is used for an EXACTFU node that contains at least one "ss" sequence in
3167 * it. For non-UTF-8 patterns and strings, this is the only case where
3168 * there is a possible fold length change. That means that a regular
3169 * EXACTFU node without UTF-8 involvement doesn't have to concern itself
3170 * with length changes, and so can be processed faster. regexec.c takes
3171 * advantage of this. Generally, an EXACTFish node that is in UTF-8 is
3172 * pre-folded by regcomp.c (except EXACTFL, some of whose folds aren't
3173 * known until runtime). This saves effort in regex matching. However,
3174 * the pre-folding isn't done for non-UTF8 patterns because the fold of
3175 * the MICRO SIGN requires UTF-8, and we don't want to slow things down by
3176 * forcing the pattern into UTF8 unless necessary. Also what EXACTF (and,
3177 * again, EXACTFL) nodes fold to isn't known until runtime. The fold
3178 * possibilities for the non-UTF8 patterns are quite simple, except for
3179 * the sharp s. All the ones that don't involve a UTF-8 target string are
3180 * members of a fold-pair, and arrays are set up for all of them so that
3181 * the other member of the pair can be found quickly. Code elsewhere in
3182 * this file makes sure that in EXACTFU nodes, the sharp s gets folded to
3183 * 'ss', even if the pattern isn't UTF-8. This avoids the issues
3184 * described in the next item.
3185 * 3) A problem remains for unfolded multi-char folds. (These occur when the
3186 * validity of the fold won't be known until runtime, and so must remain
3187 * unfolded for now. This happens for the sharp s in EXACTF and EXACTFA
3188 * nodes when the pattern isn't in UTF-8. (Note, BTW, that there cannot
3189 * be an EXACTF node with a UTF-8 pattern.) They also occur for various
3190 * folds in EXACTFL nodes, regardless of the UTF-ness of the pattern.)
3191 * The reason this is a problem is that the optimizer part of regexec.c
3192 * (probably unwittingly, in Perl_regexec_flags()) makes an assumption
3193 * that a character in the pattern corresponds to at most a single
3194 * character in the target string. (And I do mean character, and not byte
3195 * here, unlike other parts of the documentation that have never been
3196 * updated to account for multibyte Unicode.) sharp s in EXACTF and
3197 * EXACTFL nodes can match the two character string 'ss'; in EXACTFA nodes
3198 * it can match "\x{17F}\x{17F}". These, along with other ones in EXACTFL
3199 * nodes, violate the assumption, and they are the only instances where it
3200 * is violated. I'm reluctant to try to change the assumption, as the
3201 * code involved is impenetrable to me (khw), so instead the code here
3202 * punts. This routine examines EXACTFL nodes, and (when the pattern
3203 * isn't UTF-8) EXACTF and EXACTFA for such unfolded folds, and returns a
3204 * boolean indicating whether or not the node contains such a fold. When
3205 * it is true, the caller sets a flag that later causes the optimizer in
3206 * this file to not set values for the floating and fixed string lengths,
3207 * and thus avoids the optimizer code in regexec.c that makes the invalid
3208 * assumption. Thus, there is no optimization based on string lengths for
3209 * EXACTFL nodes that contain these few folds, nor for non-UTF8-pattern
3210 * EXACTF and EXACTFA nodes that contain the sharp s. (The reason the
3211 * assumption is wrong only in these cases is that all other non-UTF-8
3212 * folds are 1-1; and, for UTF-8 patterns, we pre-fold all other folds to
3213 * their expanded versions. (Again, we can't prefold sharp s to 'ss' in
3214 * EXACTF nodes because we don't know at compile time if it actually
3215 * matches 'ss' or not. For EXACTF nodes it will match iff the target
3216 * string is in UTF-8. This is in contrast to EXACTFU nodes, where it
3217 * always matches; and EXACTFA where it never does. In an EXACTFA node in
3218 * a UTF-8 pattern, sharp s is folded to "\x{17F}\x{17F}, avoiding the
3219 * problem; but in a non-UTF8 pattern, folding it to that above-Latin1
3220 * string would require the pattern to be forced into UTF-8, the overhead
3221 * of which we want to avoid. Similarly the unfolded multi-char folds in
3222 * EXACTFL nodes will match iff the locale at the time of match is a UTF-8
3225 * Similarly, the code that generates tries doesn't currently handle
3226 * not-already-folded multi-char folds, and it looks like a pain to change
3227 * that. Therefore, trie generation of EXACTFA nodes with the sharp s
3228 * doesn't work. Instead, such an EXACTFA is turned into a new regnode,
3229 * EXACTFA_NO_TRIE, which the trie code knows not to handle. Most people
3230 * using /iaa matching will be doing so almost entirely with ASCII
3231 * strings, so this should rarely be encountered in practice */
3233 #define JOIN_EXACT(scan,min_subtract,unfolded_multi_char, flags) \
3234 if (PL_regkind[OP(scan)] == EXACT) \
3235 join_exact(pRExC_state,(scan),(min_subtract),unfolded_multi_char, (flags),NULL,depth+1)
3238 S_join_exact(pTHX_ RExC_state_t *pRExC_state, regnode *scan,
3239 UV *min_subtract, bool *unfolded_multi_char,
3240 U32 flags,regnode *val, U32 depth)
3242 /* Merge several consecutive EXACTish nodes into one. */
3243 regnode *n = regnext(scan);
3245 regnode *next = scan + NODE_SZ_STR(scan);
3249 regnode *stop = scan;
3250 GET_RE_DEBUG_FLAGS_DECL;
3252 PERL_UNUSED_ARG(depth);
3255 PERL_ARGS_ASSERT_JOIN_EXACT;
3256 #ifndef EXPERIMENTAL_INPLACESCAN
3257 PERL_UNUSED_ARG(flags);
3258 PERL_UNUSED_ARG(val);
3260 DEBUG_PEEP("join",scan,depth);
3262 /* Look through the subsequent nodes in the chain. Skip NOTHING, merge
3263 * EXACT ones that are mergeable to the current one. */
3265 && (PL_regkind[OP(n)] == NOTHING
3266 || (stringok && OP(n) == OP(scan)))
3268 && NEXT_OFF(scan) + NEXT_OFF(n) < I16_MAX)
3271 if (OP(n) == TAIL || n > next)
3273 if (PL_regkind[OP(n)] == NOTHING) {
3274 DEBUG_PEEP("skip:",n,depth);
3275 NEXT_OFF(scan) += NEXT_OFF(n);
3276 next = n + NODE_STEP_REGNODE;
3283 else if (stringok) {
3284 const unsigned int oldl = STR_LEN(scan);
3285 regnode * const nnext = regnext(n);
3287 /* XXX I (khw) kind of doubt that this works on platforms (should
3288 * Perl ever run on one) where U8_MAX is above 255 because of lots
3289 * of other assumptions */
3290 /* Don't join if the sum can't fit into a single node */
3291 if (oldl + STR_LEN(n) > U8_MAX)
3294 DEBUG_PEEP("merg",n,depth);
3297 NEXT_OFF(scan) += NEXT_OFF(n);
3298 STR_LEN(scan) += STR_LEN(n);
3299 next = n + NODE_SZ_STR(n);
3300 /* Now we can overwrite *n : */
3301 Move(STRING(n), STRING(scan) + oldl, STR_LEN(n), char);
3309 #ifdef EXPERIMENTAL_INPLACESCAN
3310 if (flags && !NEXT_OFF(n)) {
3311 DEBUG_PEEP("atch", val, depth);
3312 if (reg_off_by_arg[OP(n)]) {
3313 ARG_SET(n, val - n);
3316 NEXT_OFF(n) = val - n;
3324 *unfolded_multi_char = FALSE;
3326 /* Here, all the adjacent mergeable EXACTish nodes have been merged. We
3327 * can now analyze for sequences of problematic code points. (Prior to
3328 * this final joining, sequences could have been split over boundaries, and
3329 * hence missed). The sequences only happen in folding, hence for any
3330 * non-EXACT EXACTish node */
3331 if (OP(scan) != EXACT) {
3332 U8* s0 = (U8*) STRING(scan);
3334 U8* s_end = s0 + STR_LEN(scan);
3336 int total_count_delta = 0; /* Total delta number of characters that
3337 multi-char folds expand to */
3339 /* One pass is made over the node's string looking for all the
3340 * possibilities. To avoid some tests in the loop, there are two main
3341 * cases, for UTF-8 patterns (which can't have EXACTF nodes) and
3346 if (OP(scan) == EXACTFL) {
3349 /* An EXACTFL node would already have been changed to another
3350 * node type unless there is at least one character in it that
3351 * is problematic; likely a character whose fold definition
3352 * won't be known until runtime, and so has yet to be folded.
3353 * For all but the UTF-8 locale, folds are 1-1 in length, but
3354 * to handle the UTF-8 case, we need to create a temporary
3355 * folded copy using UTF-8 locale rules in order to analyze it.
3356 * This is because our macros that look to see if a sequence is
3357 * a multi-char fold assume everything is folded (otherwise the
3358 * tests in those macros would be too complicated and slow).
3359 * Note that here, the non-problematic folds will have already
3360 * been done, so we can just copy such characters. We actually
3361 * don't completely fold the EXACTFL string. We skip the
3362 * unfolded multi-char folds, as that would just create work
3363 * below to figure out the size they already are */
3365 Newx(folded, UTF8_MAX_FOLD_CHAR_EXPAND * STR_LEN(scan) + 1, U8);
3368 STRLEN s_len = UTF8SKIP(s);
3369 if (! is_PROBLEMATIC_LOCALE_FOLD_utf8(s)) {
3370 Copy(s, d, s_len, U8);
3373 else if (is_FOLDS_TO_MULTI_utf8(s)) {
3374 *unfolded_multi_char = TRUE;
3375 Copy(s, d, s_len, U8);
3378 else if (isASCII(*s)) {
3379 *(d++) = toFOLD(*s);
3383 _to_utf8_fold_flags(s, d, &len, FOLD_FLAGS_FULL);
3389 /* Point the remainder of the routine to look at our temporary
3393 } /* End of creating folded copy of EXACTFL string */
3395 /* Examine the string for a multi-character fold sequence. UTF-8
3396 * patterns have all characters pre-folded by the time this code is
3398 while (s < s_end - 1) /* Can stop 1 before the end, as minimum
3399 length sequence we are looking for is 2 */
3401 int count = 0; /* How many characters in a multi-char fold */
3402 int len = is_MULTI_CHAR_FOLD_utf8_safe(s, s_end);
3403 if (! len) { /* Not a multi-char fold: get next char */
3408 /* Nodes with 'ss' require special handling, except for
3409 * EXACTFA-ish for which there is no multi-char fold to this */
3410 if (len == 2 && *s == 's' && *(s+1) == 's'
3411 && OP(scan) != EXACTFA
3412 && OP(scan) != EXACTFA_NO_TRIE)
3415 if (OP(scan) != EXACTFL) {
3416 OP(scan) = EXACTFU_SS;
3420 else { /* Here is a generic multi-char fold. */
3421 U8* multi_end = s + len;
3423 /* Count how many characters in it. In the case of /aa, no
3424 * folds which contain ASCII code points are allowed, so
3425 * check for those, and skip if found. */
3426 if (OP(scan) != EXACTFA && OP(scan) != EXACTFA_NO_TRIE) {
3427 count = utf8_length(s, multi_end);
3431 while (s < multi_end) {
3434 goto next_iteration;
3444 /* The delta is how long the sequence is minus 1 (1 is how long
3445 * the character that folds to the sequence is) */
3446 total_count_delta += count - 1;
3450 /* We created a temporary folded copy of the string in EXACTFL
3451 * nodes. Therefore we need to be sure it doesn't go below zero,
3452 * as the real string could be shorter */
3453 if (OP(scan) == EXACTFL) {
3454 int total_chars = utf8_length((U8*) STRING(scan),
3455 (U8*) STRING(scan) + STR_LEN(scan));
3456 if (total_count_delta > total_chars) {
3457 total_count_delta = total_chars;
3461 *min_subtract += total_count_delta;
3464 else if (OP(scan) == EXACTFA) {
3466 /* Non-UTF-8 pattern, EXACTFA node. There can't be a multi-char
3467 * fold to the ASCII range (and there are no existing ones in the
3468 * upper latin1 range). But, as outlined in the comments preceding
3469 * this function, we need to flag any occurrences of the sharp s.
3470 * This character forbids trie formation (because of added
3473 if (*s == LATIN_SMALL_LETTER_SHARP_S) {
3474 OP(scan) = EXACTFA_NO_TRIE;
3475 *unfolded_multi_char = TRUE;
3484 /* Non-UTF-8 pattern, not EXACTFA node. Look for the multi-char
3485 * folds that are all Latin1. As explained in the comments
3486 * preceding this function, we look also for the sharp s in EXACTF
3487 * and EXACTFL nodes; it can be in the final position. Otherwise
3488 * we can stop looking 1 byte earlier because have to find at least
3489 * two characters for a multi-fold */
3490 const U8* upper = (OP(scan) == EXACTF || OP(scan) == EXACTFL)
3495 int len = is_MULTI_CHAR_FOLD_latin1_safe(s, s_end);
3496 if (! len) { /* Not a multi-char fold. */
3497 if (*s == LATIN_SMALL_LETTER_SHARP_S
3498 && (OP(scan) == EXACTF || OP(scan) == EXACTFL))
3500 *unfolded_multi_char = TRUE;
3507 && isARG2_lower_or_UPPER_ARG1('s', *s)
3508 && isARG2_lower_or_UPPER_ARG1('s', *(s+1)))
3511 /* EXACTF nodes need to know that the minimum length
3512 * changed so that a sharp s in the string can match this
3513 * ss in the pattern, but they remain EXACTF nodes, as they
3514 * won't match this unless the target string is is UTF-8,
3515 * which we don't know until runtime. EXACTFL nodes can't
3516 * transform into EXACTFU nodes */
3517 if (OP(scan) != EXACTF && OP(scan) != EXACTFL) {
3518 OP(scan) = EXACTFU_SS;
3522 *min_subtract += len - 1;
3529 /* Allow dumping but overwriting the collection of skipped
3530 * ops and/or strings with fake optimized ops */
3531 n = scan + NODE_SZ_STR(scan);
3539 DEBUG_OPTIMISE_r(if (merged){DEBUG_PEEP("finl",scan,depth)});
3543 /* REx optimizer. Converts nodes into quicker variants "in place".
3544 Finds fixed substrings. */
3546 /* Stops at toplevel WHILEM as well as at "last". At end *scanp is set
3547 to the position after last scanned or to NULL. */
3549 #define INIT_AND_WITHP \
3550 assert(!and_withp); \
3551 Newx(and_withp,1, regnode_ssc); \
3552 SAVEFREEPV(and_withp)
3554 /* this is a chain of data about sub patterns we are processing that
3555 need to be handled separately/specially in study_chunk. Its so
3556 we can simulate recursion without losing state. */
3558 typedef struct scan_frame {
3559 regnode *last; /* last node to process in this frame */
3560 regnode *next; /* next node to process when last is reached */
3561 struct scan_frame *prev; /*previous frame*/
3562 U32 prev_recursed_depth;
3563 I32 stop; /* what stopparen do we use */
3568 S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp,
3569 SSize_t *minlenp, SSize_t *deltap,
3574 regnode_ssc *and_withp,
3575 U32 flags, U32 depth)
3576 /* scanp: Start here (read-write). */
3577 /* deltap: Write maxlen-minlen here. */
3578 /* last: Stop before this one. */
3579 /* data: string data about the pattern */
3580 /* stopparen: treat close N as END */
3581 /* recursed: which subroutines have we recursed into */
3582 /* and_withp: Valid if flags & SCF_DO_STCLASS_OR */
3585 /* There must be at least this number of characters to match */
3588 regnode *scan = *scanp, *next;
3590 int is_inf = (flags & SCF_DO_SUBSTR) && (data->flags & SF_IS_INF);
3591 int is_inf_internal = 0; /* The studied chunk is infinite */
3592 I32 is_par = OP(scan) == OPEN ? ARG(scan) : 0;
3593 scan_data_t data_fake;
3594 SV *re_trie_maxbuff = NULL;
3595 regnode *first_non_open = scan;
3596 SSize_t stopmin = SSize_t_MAX;
3597 scan_frame *frame = NULL;
3598 GET_RE_DEBUG_FLAGS_DECL;
3600 PERL_ARGS_ASSERT_STUDY_CHUNK;
3603 StructCopy(&zero_scan_data, &data_fake, scan_data_t);
3606 while (first_non_open && OP(first_non_open) == OPEN)
3607 first_non_open=regnext(first_non_open);
3612 while ( scan && OP(scan) != END && scan < last ){
3613 UV min_subtract = 0; /* How mmany chars to subtract from the minimum
3614 node length to get a real minimum (because
3615 the folded version may be shorter) */
3616 bool unfolded_multi_char = FALSE;
3617 /* Peephole optimizer: */
3618 DEBUG_OPTIMISE_MORE_r(
3620 PerlIO_printf(Perl_debug_log,
3621 "%*sstudy_chunk stopparen=%ld depth=%lu recursed_depth=%lu ",
3622 ((int) depth*2), "", (long)stopparen,
3623 (unsigned long)depth, (unsigned long)recursed_depth);
3624 if (recursed_depth) {
3627 for ( j = 0 ; j < recursed_depth ; j++ ) {
3628 PerlIO_printf(Perl_debug_log,"[");
3629 for ( i = 0 ; i < (U32)RExC_npar ; i++ )
3630 PerlIO_printf(Perl_debug_log,"%d",
3631 PAREN_TEST(RExC_study_chunk_recursed +
3632 (j * RExC_study_chunk_recursed_bytes), i)
3635 PerlIO_printf(Perl_debug_log,"]");
3638 PerlIO_printf(Perl_debug_log,"\n");
3641 DEBUG_STUDYDATA("Peep:", data, depth);
3642 DEBUG_PEEP("Peep", scan, depth);
3645 /* The reason we do this here we need to deal with things like /(?:f)(?:o)(?:o)/
3646 * which cant be dealt with by the normal EXACT parsing code, as each (?:..) is handled
3647 * by a different invocation of reg() -- Yves
3649 JOIN_EXACT(scan,&min_subtract, &unfolded_multi_char, 0);
3651 /* Follow the next-chain of the current node and optimize
3652 away all the NOTHINGs from it. */
3653 if (OP(scan) != CURLYX) {
3654 const int max = (reg_off_by_arg[OP(scan)]
3656 /* I32 may be smaller than U16 on CRAYs! */
3657 : (I32_MAX < U16_MAX ? I32_MAX : U16_MAX));
3658 int off = (reg_off_by_arg[OP(scan)] ? ARG(scan) : NEXT_OFF(scan));
3662 /* Skip NOTHING and LONGJMP. */
3663 while ((n = regnext(n))
3664 && ((PL_regkind[OP(n)] == NOTHING && (noff = NEXT_OFF(n)))
3665 || ((OP(n) == LONGJMP) && (noff = ARG(n))))
3666 && off + noff < max)
3668 if (reg_off_by_arg[OP(scan)])
3671 NEXT_OFF(scan) = off;
3676 /* The principal pseudo-switch. Cannot be a switch, since we
3677 look into several different things. */
3678 if (OP(scan) == BRANCH || OP(scan) == BRANCHJ
3679 || OP(scan) == IFTHEN) {
3680 next = regnext(scan);
3682 /* demq: the op(next)==code check is to see if we have
3683 * "branch-branch" AFAICT */
3685 if (OP(next) == code || code == IFTHEN) {
3686 /* NOTE - There is similar code to this block below for
3687 * handling TRIE nodes on a re-study. If you change stuff here
3688 * check there too. */
3689 SSize_t max1 = 0, min1 = SSize_t_MAX, num = 0;
3691 regnode * const startbranch=scan;
3693 if (flags & SCF_DO_SUBSTR) {
3694 /* Cannot merge strings after this. */
3695 scan_commit(pRExC_state, data, minlenp, is_inf);
3698 if (flags & SCF_DO_STCLASS)
3699 ssc_init_zero(pRExC_state, &accum);
3701 while (OP(scan) == code) {
3702 SSize_t deltanext, minnext, fake;
3704 regnode_ssc this_class;
3707 data_fake.flags = 0;
3709 data_fake.whilem_c = data->whilem_c;
3710 data_fake.last_closep = data->last_closep;
3713 data_fake.last_closep = &fake;
3715 data_fake.pos_delta = delta;
3716 next = regnext(scan);
3717 scan = NEXTOPER(scan);
3719 scan = NEXTOPER(scan);
3720 if (flags & SCF_DO_STCLASS) {
3721 ssc_init(pRExC_state, &this_class);
3722 data_fake.start_class = &this_class;
3723 f = SCF_DO_STCLASS_AND;
3725 if (flags & SCF_WHILEM_VISITED_POS)
3726 f |= SCF_WHILEM_VISITED_POS;
3728 /* we suppose the run is continuous, last=next...*/
3729 minnext = study_chunk(pRExC_state, &scan, minlenp,
3730 &deltanext, next, &data_fake, stopparen,
3731 recursed_depth, NULL, f,depth+1);
3734 if (deltanext == SSize_t_MAX) {
3735 is_inf = is_inf_internal = 1;
3737 } else if (max1 < minnext + deltanext)
3738 max1 = minnext + deltanext;
3740 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
3742 if (data_fake.flags & SCF_SEEN_ACCEPT) {
3743 if ( stopmin > minnext)
3744 stopmin = min + min1;
3745 flags &= ~SCF_DO_SUBSTR;
3747 data->flags |= SCF_SEEN_ACCEPT;
3750 if (data_fake.flags & SF_HAS_EVAL)
3751 data->flags |= SF_HAS_EVAL;
3752 data->whilem_c = data_fake.whilem_c;
3754 if (flags & SCF_DO_STCLASS)
3755 ssc_or(pRExC_state, &accum, (regnode_charclass*)&this_class);
3757 if (code == IFTHEN && num < 2) /* Empty ELSE branch */
3759 if (flags & SCF_DO_SUBSTR) {
3760 data->pos_min += min1;
3761 if (data->pos_delta >= SSize_t_MAX - (max1 - min1))
3762 data->pos_delta = SSize_t_MAX;
3764 data->pos_delta += max1 - min1;
3765 if (max1 != min1 || is_inf)
3766 data->longest = &(data->longest_float);
3769 if (delta == SSize_t_MAX
3770 || SSize_t_MAX - delta - (max1 - min1) < 0)
3771 delta = SSize_t_MAX;
3773 delta += max1 - min1;
3774 if (flags & SCF_DO_STCLASS_OR) {
3775 ssc_or(pRExC_state, data->start_class, (regnode_charclass*) &accum);
3777 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
3778 flags &= ~SCF_DO_STCLASS;
3781 else if (flags & SCF_DO_STCLASS_AND) {
3783 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) &accum);
3784 flags &= ~SCF_DO_STCLASS;
3787 /* Switch to OR mode: cache the old value of
3788 * data->start_class */
3790 StructCopy(data->start_class, and_withp, regnode_ssc);
3791 flags &= ~SCF_DO_STCLASS_AND;
3792 StructCopy(&accum, data->start_class, regnode_ssc);
3793 flags |= SCF_DO_STCLASS_OR;
3797 if (PERL_ENABLE_TRIE_OPTIMISATION &&
3798 OP( startbranch ) == BRANCH )
3802 Assuming this was/is a branch we are dealing with: 'scan'
3803 now points at the item that follows the branch sequence,
3804 whatever it is. We now start at the beginning of the
3805 sequence and look for subsequences of
3811 which would be constructed from a pattern like
3814 If we can find such a subsequence we need to turn the first
3815 element into a trie and then add the subsequent branch exact
3816 strings to the trie.
3820 1. patterns where the whole set of branches can be
3823 2. patterns where only a subset can be converted.
3825 In case 1 we can replace the whole set with a single regop
3826 for the trie. In case 2 we need to keep the start and end
3829 'BRANCH EXACT; BRANCH EXACT; BRANCH X'
3830 becomes BRANCH TRIE; BRANCH X;
3832 There is an additional case, that being where there is a
3833 common prefix, which gets split out into an EXACT like node
3834 preceding the TRIE node.
3836 If x(1..n)==tail then we can do a simple trie, if not we make
3837 a "jump" trie, such that when we match the appropriate word
3838 we "jump" to the appropriate tail node. Essentially we turn
3839 a nested if into a case structure of sorts.
3844 if (!re_trie_maxbuff) {
3845 re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1);
3846 if (!SvIOK(re_trie_maxbuff))
3847 sv_setiv(re_trie_maxbuff, RE_TRIE_MAXBUF_INIT);
3849 if ( SvIV(re_trie_maxbuff)>=0 ) {
3851 regnode *first = (regnode *)NULL;
3852 regnode *last = (regnode *)NULL;
3853 regnode *tail = scan;
3858 SV * const mysv = sv_newmortal(); /* for dumping */
3860 /* var tail is used because there may be a TAIL
3861 regop in the way. Ie, the exacts will point to the
3862 thing following the TAIL, but the last branch will
3863 point at the TAIL. So we advance tail. If we
3864 have nested (?:) we may have to move through several
3868 while ( OP( tail ) == TAIL ) {
3869 /* this is the TAIL generated by (?:) */
3870 tail = regnext( tail );
3874 DEBUG_TRIE_COMPILE_r({
3875 regprop(RExC_rx, mysv, tail, NULL);
3876 PerlIO_printf( Perl_debug_log, "%*s%s%s\n",
3877 (int)depth * 2 + 2, "",
3878 "Looking for TRIE'able sequences. Tail node is: ",
3879 SvPV_nolen_const( mysv )
3885 Step through the branches
3886 cur represents each branch,
3887 noper is the first thing to be matched as part
3889 noper_next is the regnext() of that node.
3891 We normally handle a case like this
3892 /FOO[xyz]|BAR[pqr]/ via a "jump trie" but we also
3893 support building with NOJUMPTRIE, which restricts
3894 the trie logic to structures like /FOO|BAR/.
3896 If noper is a trieable nodetype then the branch is
3897 a possible optimization target. If we are building
3898 under NOJUMPTRIE then we require that noper_next is
3899 the same as scan (our current position in the regex
3902 Once we have two or more consecutive such branches
3903 we can create a trie of the EXACT's contents and
3904 stitch it in place into the program.
3906 If the sequence represents all of the branches in
3907 the alternation we replace the entire thing with a
3910 Otherwise when it is a subsequence we need to
3911 stitch it in place and replace only the relevant
3912 branches. This means the first branch has to remain
3913 as it is used by the alternation logic, and its
3914 next pointer, and needs to be repointed at the item
3915 on the branch chain following the last branch we
3916 have optimized away.
3918 This could be either a BRANCH, in which case the
3919 subsequence is internal, or it could be the item
3920 following the branch sequence in which case the
3921 subsequence is at the end (which does not
3922 necessarily mean the first node is the start of the
3925 TRIE_TYPE(X) is a define which maps the optype to a
3929 ----------------+-----------
3933 EXACTFU_SS | EXACTFU
3938 #define TRIE_TYPE(X) ( ( NOTHING == (X) ) ? NOTHING : \
3939 ( EXACT == (X) ) ? EXACT : \
3940 ( EXACTFU == (X) || EXACTFU_SS == (X) ) ? EXACTFU : \
3941 ( EXACTFA == (X) ) ? EXACTFA : \
3944 /* dont use tail as the end marker for this traverse */
3945 for ( cur = startbranch ; cur != scan ; cur = regnext( cur ) ) {
3946 regnode * const noper = NEXTOPER( cur );
3947 U8 noper_type = OP( noper );
3948 U8 noper_trietype = TRIE_TYPE( noper_type );
3949 #if defined(DEBUGGING) || defined(NOJUMPTRIE)
3950 regnode * const noper_next = regnext( noper );
3951 U8 noper_next_type = (noper_next && noper_next != tail) ? OP(noper_next) : 0;
3952 U8 noper_next_trietype = (noper_next && noper_next != tail) ? TRIE_TYPE( noper_next_type ) :0;
3955 DEBUG_TRIE_COMPILE_r({
3956 regprop(RExC_rx, mysv, cur, NULL);
3957 PerlIO_printf( Perl_debug_log, "%*s- %s (%d)",
3958 (int)depth * 2 + 2,"", SvPV_nolen_const( mysv ), REG_NODE_NUM(cur) );
3960 regprop(RExC_rx, mysv, noper, NULL);
3961 PerlIO_printf( Perl_debug_log, " -> %s",
3962 SvPV_nolen_const(mysv));
3965 regprop(RExC_rx, mysv, noper_next, NULL);
3966 PerlIO_printf( Perl_debug_log,"\t=> %s\t",
3967 SvPV_nolen_const(mysv));
3969 PerlIO_printf( Perl_debug_log, "(First==%d,Last==%d,Cur==%d,tt==%s,nt==%s,nnt==%s)\n",
3970 REG_NODE_NUM(first), REG_NODE_NUM(last), REG_NODE_NUM(cur),
3971 PL_reg_name[trietype], PL_reg_name[noper_trietype], PL_reg_name[noper_next_trietype]
3975 /* Is noper a trieable nodetype that can be merged
3976 * with the current trie (if there is one)? */
3980 ( noper_trietype == NOTHING)
3981 || ( trietype == NOTHING )
3982 || ( trietype == noper_trietype )
3985 && noper_next == tail
3989 /* Handle mergable triable node Either we are
3990 * the first node in a new trieable sequence,
3991 * in which case we do some bookkeeping,
3992 * otherwise we update the end pointer. */
3995 if ( noper_trietype == NOTHING ) {
3996 #if !defined(DEBUGGING) && !defined(NOJUMPTRIE)
3997 regnode * const noper_next = regnext( noper );
3998 U8 noper_next_type = (noper_next && noper_next!=tail) ? OP(noper_next) : 0;
3999 U8 noper_next_trietype = noper_next_type ? TRIE_TYPE( noper_next_type ) :0;
4002 if ( noper_next_trietype ) {
4003 trietype = noper_next_trietype;
4004 } else if (noper_next_type) {
4005 /* a NOTHING regop is 1 regop wide.
4006 * We need at least two for a trie
4007 * so we can't merge this in */
4011 trietype = noper_trietype;
4014 if ( trietype == NOTHING )
4015 trietype = noper_trietype;
4020 } /* end handle mergable triable node */
4022 /* handle unmergable node -
4023 * noper may either be a triable node which can
4024 * not be tried together with the current trie,
4025 * or a non triable node */
4027 /* If last is set and trietype is not
4028 * NOTHING then we have found at least two
4029 * triable branch sequences in a row of a
4030 * similar trietype so we can turn them
4031 * into a trie. If/when we allow NOTHING to
4032 * start a trie sequence this condition
4033 * will be required, and it isn't expensive
4034 * so we leave it in for now. */
4035 if ( trietype && trietype != NOTHING )
4036 make_trie( pRExC_state,
4037 startbranch, first, cur, tail,
4038 count, trietype, depth+1 );
4039 last = NULL; /* note: we clear/update
4040 first, trietype etc below,
4041 so we dont do it here */
4045 && noper_next == tail
4048 /* noper is triable, so we can start a new
4052 trietype = noper_trietype;
4054 /* if we already saw a first but the
4055 * current node is not triable then we have
4056 * to reset the first information. */
4061 } /* end handle unmergable node */
4062 } /* loop over branches */
4063 DEBUG_TRIE_COMPILE_r({
4064 regprop(RExC_rx, mysv, cur, NULL);
4065 PerlIO_printf( Perl_debug_log,
4066 "%*s- %s (%d) <SCAN FINISHED>\n",
4068 "", SvPV_nolen_const( mysv ),REG_NODE_NUM(cur));
4071 if ( last && trietype ) {
4072 if ( trietype != NOTHING ) {
4073 /* the last branch of the sequence was part of
4074 * a trie, so we have to construct it here
4075 * outside of the loop */
4076 made= make_trie( pRExC_state, startbranch,
4077 first, scan, tail, count,
4078 trietype, depth+1 );
4079 #ifdef TRIE_STUDY_OPT
4080 if ( ((made == MADE_EXACT_TRIE &&
4081 startbranch == first)
4082 || ( first_non_open == first )) &&
4084 flags |= SCF_TRIE_RESTUDY;
4085 if ( startbranch == first
4088 RExC_seen &=~REG_TOP_LEVEL_BRANCHES_SEEN;
4093 /* at this point we know whatever we have is a
4094 * NOTHING sequence/branch AND if 'startbranch'
4095 * is 'first' then we can turn the whole thing
4098 if ( startbranch == first ) {
4100 /* the entire thing is a NOTHING sequence,
4101 * something like this: (?:|) So we can
4102 * turn it into a plain NOTHING op. */
4103 DEBUG_TRIE_COMPILE_r({
4104 regprop(RExC_rx, mysv, cur, NULL);
4105 PerlIO_printf( Perl_debug_log,
4106 "%*s- %s (%d) <NOTHING BRANCH SEQUENCE>\n", (int)depth * 2 + 2,
4107 "", SvPV_nolen_const( mysv ),REG_NODE_NUM(cur));
4110 OP(startbranch)= NOTHING;
4111 NEXT_OFF(startbranch)= tail - startbranch;
4112 for ( opt= startbranch + 1; opt < tail ; opt++ )
4116 } /* end if ( last) */
4117 } /* TRIE_MAXBUF is non zero */
4122 else if ( code == BRANCHJ ) { /* single branch is optimized. */
4123 scan = NEXTOPER(NEXTOPER(scan));
4124 } else /* single branch is optimized. */
4125 scan = NEXTOPER(scan);
4127 } else if (OP(scan) == SUSPEND || OP(scan) == GOSUB || OP(scan) == GOSTART) {
4128 scan_frame *newframe = NULL;
4132 U32 my_recursed_depth= recursed_depth;
4134 if (OP(scan) != SUSPEND) {
4135 /* set the pointer */
4136 if (OP(scan) == GOSUB) {
4138 RExC_recurse[ARG2L(scan)] = scan;
4139 start = RExC_open_parens[paren-1];
4140 end = RExC_close_parens[paren-1];
4143 start = RExC_rxi->program + 1;
4148 !PAREN_TEST(RExC_study_chunk_recursed + ((recursed_depth-1) * RExC_study_chunk_recursed_bytes), paren)
4150 if (!recursed_depth) {
4151 Zero(RExC_study_chunk_recursed, RExC_study_chunk_recursed_bytes, U8);
4153 Copy(RExC_study_chunk_recursed + ((recursed_depth-1) * RExC_study_chunk_recursed_bytes),
4154 RExC_study_chunk_recursed + (recursed_depth * RExC_study_chunk_recursed_bytes),
4155 RExC_study_chunk_recursed_bytes, U8);
4157 /* we havent recursed into this paren yet, so recurse into it */
4158 DEBUG_STUDYDATA("set:", data,depth);
4159 PAREN_SET(RExC_study_chunk_recursed + (recursed_depth * RExC_study_chunk_recursed_bytes), paren);
4160 my_recursed_depth= recursed_depth + 1;
4161 Newx(newframe,1,scan_frame);
4163 DEBUG_STUDYDATA("inf:", data,depth);
4164 /* some form of infinite recursion, assume infinite length
4166 if (flags & SCF_DO_SUBSTR) {
4167 scan_commit(pRExC_state, data, minlenp, is_inf);
4168 data->longest = &(data->longest_float);
4170 is_inf = is_inf_internal = 1;
4171 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
4172 ssc_anything(data->start_class);
4173 flags &= ~SCF_DO_STCLASS;
4176 Newx(newframe,1,scan_frame);
4179 end = regnext(scan);
4184 SAVEFREEPV(newframe);
4185 newframe->next = regnext(scan);
4186 newframe->last = last;
4187 newframe->stop = stopparen;
4188 newframe->prev = frame;
4189 newframe->prev_recursed_depth = recursed_depth;
4191 DEBUG_STUDYDATA("frame-new:",data,depth);
4192 DEBUG_PEEP("fnew", scan, depth);
4199 recursed_depth= my_recursed_depth;
4204 else if (OP(scan) == EXACT) {
4205 SSize_t l = STR_LEN(scan);
4208 const U8 * const s = (U8*)STRING(scan);
4209 uc = utf8_to_uvchr_buf(s, s + l, NULL);
4210 l = utf8_length(s, s + l);
4212 uc = *((U8*)STRING(scan));
4215 if (flags & SCF_DO_SUBSTR) { /* Update longest substr. */
4216 /* The code below prefers earlier match for fixed
4217 offset, later match for variable offset. */
4218 if (data->last_end == -1) { /* Update the start info. */
4219 data->last_start_min = data->pos_min;
4220 data->last_start_max = is_inf
4221 ? SSize_t_MAX : data->pos_min + data->pos_delta;
4223 sv_catpvn(data->last_found, STRING(scan), STR_LEN(scan));
4225 SvUTF8_on(data->last_found);
4227 SV * const sv = data->last_found;
4228 MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ?
4229 mg_find(sv, PERL_MAGIC_utf8) : NULL;
4230 if (mg && mg->mg_len >= 0)
4231 mg->mg_len += utf8_length((U8*)STRING(scan),
4232 (U8*)STRING(scan)+STR_LEN(scan));
4234 data->last_end = data->pos_min + l;
4235 data->pos_min += l; /* As in the first entry. */
4236 data->flags &= ~SF_BEFORE_EOL;
4239 /* ANDing the code point leaves at most it, and not in locale, and
4240 * can't match null string */
4241 if (flags & SCF_DO_STCLASS_AND) {
4242 ssc_cp_and(data->start_class, uc);
4243 ANYOF_FLAGS(data->start_class) &= ~ANYOF_EMPTY_STRING;
4244 ssc_clear_locale(data->start_class);
4246 else if (flags & SCF_DO_STCLASS_OR) {
4247 ssc_add_cp(data->start_class, uc);
4248 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
4250 /* See commit msg 749e076fceedeb708a624933726e7989f2302f6a */
4251 ANYOF_FLAGS(data->start_class) &= ~ANYOF_EMPTY_STRING;
4253 flags &= ~SCF_DO_STCLASS;
4255 else if (PL_regkind[OP(scan)] == EXACT) { /* But OP != EXACT! */
4256 SSize_t l = STR_LEN(scan);
4257 UV uc = *((U8*)STRING(scan));
4258 SV* EXACTF_invlist = _new_invlist(4); /* Start out big enough for 2
4259 separate code points */
4261 /* Search for fixed substrings supports EXACT only. */
4262 if (flags & SCF_DO_SUBSTR) {
4264 scan_commit(pRExC_state, data, minlenp, is_inf);
4267 const U8 * const s = (U8 *)STRING(scan);
4268 uc = utf8_to_uvchr_buf(s, s + l, NULL);
4269 l = utf8_length(s, s + l);
4271 if (unfolded_multi_char) {
4272 RExC_seen |= REG_UNFOLDED_MULTI_SEEN;
4274 min += l - min_subtract;
4276 delta += min_subtract;
4277 if (flags & SCF_DO_SUBSTR) {
4278 data->pos_min += l - min_subtract;
4279 if (data->pos_min < 0) {
4282 data->pos_delta += min_subtract;
4284 data->longest = &(data->longest_float);
4287 if (OP(scan) == EXACTFL) {
4289 /* We don't know what the folds are; it could be anything. XXX
4290 * Actually, we only support UTF-8 encoding for code points
4291 * above Latin1, so we could know what those folds are. */
4292 EXACTF_invlist = _add_range_to_invlist(EXACTF_invlist,
4296 else { /* Non-locale EXACTFish */
4297 EXACTF_invlist = add_cp_to_invlist(EXACTF_invlist, uc);
4298 if (flags & SCF_DO_STCLASS_AND) {
4299 ssc_clear_locale(data->start_class);
4301 if (uc < 256) { /* We know what the Latin1 folds are ... */
4302 if (IS_IN_SOME_FOLD_L1(uc)) { /* For instance, we
4303 know if anything folds
4305 EXACTF_invlist = add_cp_to_invlist(EXACTF_invlist,
4306 PL_fold_latin1[uc]);
4307 if (OP(scan) != EXACTFA) { /* The folds below aren't
4309 if (isARG2_lower_or_UPPER_ARG1('s', uc)) {
4311 = add_cp_to_invlist(EXACTF_invlist,
4312 LATIN_SMALL_LETTER_SHARP_S);
4314 else if (uc == LATIN_SMALL_LETTER_SHARP_S) {
4316 = add_cp_to_invlist(EXACTF_invlist, 's');
4318 = add_cp_to_invlist(EXACTF_invlist, 'S');
4322 /* We also know if there are above-Latin1 code points
4323 * that fold to this (none legal for ASCII and /iaa) */
4324 if ((! isASCII(uc) || OP(scan) != EXACTFA)
4325 && HAS_NONLATIN1_FOLD_CLOSURE(uc))
4327 /* XXX We could know exactly what does fold to this
4328 * if the reverse folds are loaded, as currently in
4330 _invlist_union(EXACTF_invlist,
4336 else { /* Non-locale, above Latin1. XXX We don't currently
4337 know what participates in folds with this, so have
4338 to assume anything could */
4340 /* XXX We could know exactly what does fold to this if the
4341 * reverse folds are loaded, as currently in S_regclass().
4342 * But we do know that under /iaa nothing in the ASCII
4343 * range can participate */
4344 if (OP(scan) == EXACTFA) {
4345 _invlist_union_complement_2nd(EXACTF_invlist,
4346 PL_XPosix_ptrs[_CC_ASCII],
4350 EXACTF_invlist = _add_range_to_invlist(EXACTF_invlist,
4355 if (flags & SCF_DO_STCLASS_AND) {
4356 ANYOF_FLAGS(data->start_class) &= ~ANYOF_EMPTY_STRING;
4357 ANYOF_POSIXL_ZERO(data->start_class);
4358 ssc_intersection(data->start_class, EXACTF_invlist, FALSE);
4360 else if (flags & SCF_DO_STCLASS_OR) {
4361 ssc_union(data->start_class, EXACTF_invlist, FALSE);
4362 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
4364 /* See commit msg 749e076fceedeb708a624933726e7989f2302f6a */
4365 ANYOF_FLAGS(data->start_class) &= ~ANYOF_EMPTY_STRING;
4367 flags &= ~SCF_DO_STCLASS;
4368 SvREFCNT_dec(EXACTF_invlist);
4370 else if (REGNODE_VARIES(OP(scan))) {
4371 SSize_t mincount, maxcount, minnext, deltanext, pos_before = 0;
4372 I32 fl = 0, f = flags;
4373 regnode * const oscan = scan;
4374 regnode_ssc this_class;
4375 regnode_ssc *oclass = NULL;
4376 I32 next_is_eval = 0;
4378 switch (PL_regkind[OP(scan)]) {
4379 case WHILEM: /* End of (?:...)* . */
4380 scan = NEXTOPER(scan);
4383 if (flags & (SCF_DO_SUBSTR | SCF_DO_STCLASS)) {
4384 next = NEXTOPER(scan);
4385 if (OP(next) == EXACT || (flags & SCF_DO_STCLASS)) {
4387 maxcount = REG_INFTY;
4388 next = regnext(scan);
4389 scan = NEXTOPER(scan);
4393 if (flags & SCF_DO_SUBSTR)
4398 if (flags & SCF_DO_STCLASS) {
4400 maxcount = REG_INFTY;
4401 next = regnext(scan);
4402 scan = NEXTOPER(scan);
4405 if (flags & SCF_DO_SUBSTR) {
4406 scan_commit(pRExC_state, data, minlenp, is_inf);
4407 /* Cannot extend fixed substrings */
4408 data->longest = &(data->longest_float);
4410 is_inf = is_inf_internal = 1;
4411 scan = regnext(scan);
4412 goto optimize_curly_tail;
4414 if (stopparen>0 && (OP(scan)==CURLYN || OP(scan)==CURLYM)
4415 && (scan->flags == stopparen))
4420 mincount = ARG1(scan);
4421 maxcount = ARG2(scan);
4423 next = regnext(scan);
4424 if (OP(scan) == CURLYX) {
4425 I32 lp = (data ? *(data->last_closep) : 0);
4426 scan->flags = ((lp <= (I32)U8_MAX) ? (U8)lp : U8_MAX);
4428 scan = NEXTOPER(scan) + EXTRA_STEP_2ARGS;
4429 next_is_eval = (OP(scan) == EVAL);
4431 if (flags & SCF_DO_SUBSTR) {
4433 scan_commit(pRExC_state, data, minlenp, is_inf);
4434 /* Cannot extend fixed substrings */
4435 pos_before = data->pos_min;
4439 data->flags &= ~(SF_HAS_PAR|SF_IN_PAR|SF_HAS_EVAL);
4441 data->flags |= SF_IS_INF;
4443 if (flags & SCF_DO_STCLASS) {
4444 ssc_init(pRExC_state, &this_class);
4445 oclass = data->start_class;
4446 data->start_class = &this_class;
4447 f |= SCF_DO_STCLASS_AND;
4448 f &= ~SCF_DO_STCLASS_OR;
4450 /* Exclude from super-linear cache processing any {n,m}
4451 regops for which the combination of input pos and regex
4452 pos is not enough information to determine if a match
4455 For example, in the regex /foo(bar\s*){4,8}baz/ with the
4456 regex pos at the \s*, the prospects for a match depend not
4457 only on the input position but also on how many (bar\s*)
4458 repeats into the {4,8} we are. */
4459 if ((mincount > 1) || (maxcount > 1 && maxcount != REG_INFTY))
4460 f &= ~SCF_WHILEM_VISITED_POS;
4462 /* This will finish on WHILEM, setting scan, or on NULL: */
4463 minnext = study_chunk(pRExC_state, &scan, minlenp, &deltanext,
4464 last, data, stopparen, recursed_depth, NULL,
4466 ? (f & ~SCF_DO_SUBSTR)
4470 if (flags & SCF_DO_STCLASS)
4471 data->start_class = oclass;
4472 if (mincount == 0 || minnext == 0) {
4473 if (flags & SCF_DO_STCLASS_OR) {
4474 ssc_or(pRExC_state, data->start_class, (regnode_charclass *) &this_class);
4476 else if (flags & SCF_DO_STCLASS_AND) {
4477 /* Switch to OR mode: cache the old value of
4478 * data->start_class */
4480 StructCopy(data->start_class, and_withp, regnode_ssc);
4481 flags &= ~SCF_DO_STCLASS_AND;
4482 StructCopy(&this_class, data->start_class, regnode_ssc);
4483 flags |= SCF_DO_STCLASS_OR;
4484 ANYOF_FLAGS(data->start_class) |= ANYOF_EMPTY_STRING;
4486 } else { /* Non-zero len */
4487 if (flags & SCF_DO_STCLASS_OR) {
4488 ssc_or(pRExC_state, data->start_class, (regnode_charclass *) &this_class);
4489 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
4491 else if (flags & SCF_DO_STCLASS_AND)
4492 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) &this_class);
4493 flags &= ~SCF_DO_STCLASS;
4495 if (!scan) /* It was not CURLYX, but CURLY. */
4497 if (!(flags & SCF_TRIE_DOING_RESTUDY)
4498 /* ? quantifier ok, except for (?{ ... }) */
4499 && (next_is_eval || !(mincount == 0 && maxcount == 1))
4500 && (minnext == 0) && (deltanext == 0)
4501 && data && !(data->flags & (SF_HAS_PAR|SF_IN_PAR))
4502 && maxcount <= REG_INFTY/3) /* Complement check for big
4505 /* Fatal warnings may leak the regexp without this: */
4506 SAVEFREESV(RExC_rx_sv);
4507 ckWARNreg(RExC_parse,
4508 "Quantifier unexpected on zero-length expression");
4509 (void)ReREFCNT_inc(RExC_rx_sv);
4512 min += minnext * mincount;
4513 is_inf_internal |= deltanext == SSize_t_MAX
4514 || (maxcount == REG_INFTY && minnext + deltanext > 0);
4515 is_inf |= is_inf_internal;
4517 delta = SSize_t_MAX;
4519 delta += (minnext + deltanext) * maxcount
4520 - minnext * mincount;
4522 /* Try powerful optimization CURLYX => CURLYN. */
4523 if ( OP(oscan) == CURLYX && data
4524 && data->flags & SF_IN_PAR
4525 && !(data->flags & SF_HAS_EVAL)
4526 && !deltanext && minnext == 1 ) {
4527 /* Try to optimize to CURLYN. */
4528 regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS;
4529 regnode * const nxt1 = nxt;
4536 if (!REGNODE_SIMPLE(OP(nxt))
4537 && !(PL_regkind[OP(nxt)] == EXACT
4538 && STR_LEN(nxt) == 1))
4544 if (OP(nxt) != CLOSE)
4546 if (RExC_open_parens) {
4547 RExC_open_parens[ARG(nxt1)-1]=oscan; /*open->CURLYM*/
4548 RExC_close_parens[ARG(nxt1)-1]=nxt+2; /*close->while*/
4550 /* Now we know that nxt2 is the only contents: */
4551 oscan->flags = (U8)ARG(nxt);
4553 OP(nxt1) = NOTHING; /* was OPEN. */
4556 OP(nxt1 + 1) = OPTIMIZED; /* was count. */
4557 NEXT_OFF(nxt1+ 1) = 0; /* just for consistency. */
4558 NEXT_OFF(nxt2) = 0; /* just for consistency with CURLY. */
4559 OP(nxt) = OPTIMIZED; /* was CLOSE. */
4560 OP(nxt + 1) = OPTIMIZED; /* was count. */
4561 NEXT_OFF(nxt+ 1) = 0; /* just for consistency. */
4566 /* Try optimization CURLYX => CURLYM. */
4567 if ( OP(oscan) == CURLYX && data
4568 && !(data->flags & SF_HAS_PAR)
4569 && !(data->flags & SF_HAS_EVAL)
4570 && !deltanext /* atom is fixed width */
4571 && minnext != 0 /* CURLYM can't handle zero width */
4573 /* Nor characters whose fold at run-time may be
4574 * multi-character */
4575 && ! (RExC_seen & REG_UNFOLDED_MULTI_SEEN)
4577 /* XXXX How to optimize if data == 0? */
4578 /* Optimize to a simpler form. */
4579 regnode *nxt = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN */
4583 while ( (nxt2 = regnext(nxt)) /* skip over embedded stuff*/
4584 && (OP(nxt2) != WHILEM))
4586 OP(nxt2) = SUCCEED; /* Whas WHILEM */
4587 /* Need to optimize away parenths. */
4588 if ((data->flags & SF_IN_PAR) && OP(nxt) == CLOSE) {
4589 /* Set the parenth number. */
4590 regnode *nxt1 = NEXTOPER(oscan) + EXTRA_STEP_2ARGS; /* OPEN*/
4592 oscan->flags = (U8)ARG(nxt);
4593 if (RExC_open_parens) {
4594 RExC_open_parens[ARG(nxt1)-1]=oscan; /*open->CURLYM*/
4595 RExC_close_parens[ARG(nxt1)-1]=nxt2+1; /*close->NOTHING*/
4597 OP(nxt1) = OPTIMIZED; /* was OPEN. */
4598 OP(nxt) = OPTIMIZED; /* was CLOSE. */
4601 OP(nxt1 + 1) = OPTIMIZED; /* was count. */
4602 OP(nxt + 1) = OPTIMIZED; /* was count. */
4603 NEXT_OFF(nxt1 + 1) = 0; /* just for consistency. */
4604 NEXT_OFF(nxt + 1) = 0; /* just for consistency. */
4607 while ( nxt1 && (OP(nxt1) != WHILEM)) {
4608 regnode *nnxt = regnext(nxt1);
4610 if (reg_off_by_arg[OP(nxt1)])
4611 ARG_SET(nxt1, nxt2 - nxt1);
4612 else if (nxt2 - nxt1 < U16_MAX)
4613 NEXT_OFF(nxt1) = nxt2 - nxt1;
4615 OP(nxt) = NOTHING; /* Cannot beautify */
4620 /* Optimize again: */
4621 study_chunk(pRExC_state, &nxt1, minlenp, &deltanext, nxt,
4622 NULL, stopparen, recursed_depth, NULL, 0,depth+1);
4627 else if ((OP(oscan) == CURLYX)
4628 && (flags & SCF_WHILEM_VISITED_POS)
4629 /* See the comment on a similar expression above.
4630 However, this time it's not a subexpression
4631 we care about, but the expression itself. */
4632 && (maxcount == REG_INFTY)
4633 && data && ++data->whilem_c < 16) {
4634 /* This stays as CURLYX, we can put the count/of pair. */
4635 /* Find WHILEM (as in regexec.c) */
4636 regnode *nxt = oscan + NEXT_OFF(oscan);
4638 if (OP(PREVOPER(nxt)) == NOTHING) /* LONGJMP */
4640 PREVOPER(nxt)->flags = (U8)(data->whilem_c
4641 | (RExC_whilem_seen << 4)); /* On WHILEM */
4643 if (data && fl & (SF_HAS_PAR|SF_IN_PAR))
4645 if (flags & SCF_DO_SUBSTR) {
4646 SV *last_str = NULL;
4647 STRLEN last_chrs = 0;
4648 int counted = mincount != 0;
4650 if (data->last_end > 0 && mincount != 0) { /* Ends with a
4652 SSize_t b = pos_before >= data->last_start_min
4653 ? pos_before : data->last_start_min;
4655 const char * const s = SvPV_const(data->last_found, l);
4656 SSize_t old = b - data->last_start_min;
4659 old = utf8_hop((U8*)s, old) - (U8*)s;
4661 /* Get the added string: */
4662 last_str = newSVpvn_utf8(s + old, l, UTF);
4663 last_chrs = UTF ? utf8_length((U8*)(s + old),
4664 (U8*)(s + old + l)) : l;
4665 if (deltanext == 0 && pos_before == b) {
4666 /* What was added is a constant string */
4669 SvGROW(last_str, (mincount * l) + 1);
4670 repeatcpy(SvPVX(last_str) + l,
4671 SvPVX_const(last_str), l,
4673 SvCUR_set(last_str, SvCUR(last_str) * mincount);
4674 /* Add additional parts. */
4675 SvCUR_set(data->last_found,
4676 SvCUR(data->last_found) - l);
4677 sv_catsv(data->last_found, last_str);
4679 SV * sv = data->last_found;
4681 SvUTF8(sv) && SvMAGICAL(sv) ?
4682 mg_find(sv, PERL_MAGIC_utf8) : NULL;
4683 if (mg && mg->mg_len >= 0)
4684 mg->mg_len += last_chrs * (mincount-1);
4686 last_chrs *= mincount;
4687 data->last_end += l * (mincount - 1);
4690 /* start offset must point into the last copy */
4691 data->last_start_min += minnext * (mincount - 1);
4692 data->last_start_max += is_inf ? SSize_t_MAX
4693 : (maxcount - 1) * (minnext + data->pos_delta);
4696 /* It is counted once already... */
4697 data->pos_min += minnext * (mincount - counted);
4699 PerlIO_printf(Perl_debug_log, "counted=%"UVdf" deltanext=%"UVdf
4700 " SSize_t_MAX=%"UVdf" minnext=%"UVdf
4701 " maxcount=%"UVdf" mincount=%"UVdf"\n",
4702 (UV)counted, (UV)deltanext, (UV)SSize_t_MAX, (UV)minnext, (UV)maxcount,
4704 if (deltanext != SSize_t_MAX)
4705 PerlIO_printf(Perl_debug_log, "LHS=%"UVdf" RHS=%"UVdf"\n",
4706 (UV)(-counted * deltanext + (minnext + deltanext) * maxcount
4707 - minnext * mincount), (UV)(SSize_t_MAX - data->pos_delta));
4709 if (deltanext == SSize_t_MAX
4710 || -counted * deltanext + (minnext + deltanext) * maxcount - minnext * mincount >= SSize_t_MAX - data->pos_delta)
4711 data->pos_delta = SSize_t_MAX;
4713 data->pos_delta += - counted * deltanext +
4714 (minnext + deltanext) * maxcount - minnext * mincount;
4715 if (mincount != maxcount) {
4716 /* Cannot extend fixed substrings found inside
4718 scan_commit(pRExC_state, data, minlenp, is_inf);
4719 if (mincount && last_str) {
4720 SV * const sv = data->last_found;
4721 MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ?
4722 mg_find(sv, PERL_MAGIC_utf8) : NULL;
4726 sv_setsv(sv, last_str);
4727 data->last_end = data->pos_min;
4728 data->last_start_min = data->pos_min - last_chrs;
4729 data->last_start_max = is_inf
4731 : data->pos_min + data->pos_delta - last_chrs;
4733 data->longest = &(data->longest_float);
4735 SvREFCNT_dec(last_str);
4737 if (data && (fl & SF_HAS_EVAL))
4738 data->flags |= SF_HAS_EVAL;
4739 optimize_curly_tail:
4740 if (OP(oscan) != CURLYX) {
4741 while (PL_regkind[OP(next = regnext(oscan))] == NOTHING
4743 NEXT_OFF(oscan) += NEXT_OFF(next);
4749 Perl_croak(aTHX_ "panic: unexpected varying REx opcode %d",
4754 if (flags & SCF_DO_SUBSTR) {
4755 /* Cannot expect anything... */
4756 scan_commit(pRExC_state, data, minlenp, is_inf);
4757 data->longest = &(data->longest_float);
4759 is_inf = is_inf_internal = 1;
4760 if (flags & SCF_DO_STCLASS_OR) {
4761 if (OP(scan) == CLUMP) {
4762 /* Actually is any start char, but very few code points
4763 * aren't start characters */
4764 ssc_match_all_cp(data->start_class);
4767 ssc_anything(data->start_class);
4770 flags &= ~SCF_DO_STCLASS;
4774 else if (OP(scan) == LNBREAK) {
4775 if (flags & SCF_DO_STCLASS) {
4776 if (flags & SCF_DO_STCLASS_AND) {
4777 ssc_intersection(data->start_class,
4778 PL_XPosix_ptrs[_CC_VERTSPACE], FALSE);
4779 ssc_clear_locale(data->start_class);
4780 ANYOF_FLAGS(data->start_class) &= ~ANYOF_EMPTY_STRING;
4782 else if (flags & SCF_DO_STCLASS_OR) {
4783 ssc_union(data->start_class,
4784 PL_XPosix_ptrs[_CC_VERTSPACE],
4786 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
4788 /* See commit msg for
4789 * 749e076fceedeb708a624933726e7989f2302f6a */
4790 ANYOF_FLAGS(data->start_class) &= ~ANYOF_EMPTY_STRING;
4792 flags &= ~SCF_DO_STCLASS;
4795 delta++; /* Because of the 2 char string cr-lf */
4796 if (flags & SCF_DO_SUBSTR) {
4797 /* Cannot expect anything... */
4798 scan_commit(pRExC_state, data, minlenp, is_inf);
4800 data->pos_delta += 1;
4801 data->longest = &(data->longest_float);
4804 else if (REGNODE_SIMPLE(OP(scan))) {
4806 if (flags & SCF_DO_SUBSTR) {
4807 scan_commit(pRExC_state, data, minlenp, is_inf);
4811 if (flags & SCF_DO_STCLASS) {
4813 SV* my_invlist = NULL;
4816 /* See commit msg 749e076fceedeb708a624933726e7989f2302f6a */
4817 ANYOF_FLAGS(data->start_class) &= ~ANYOF_EMPTY_STRING;
4819 /* Some of the logic below assumes that switching
4820 locale on will only add false positives. */
4825 Perl_croak(aTHX_ "panic: unexpected simple REx opcode %d",
4830 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
4831 ssc_match_all_cp(data->start_class);
4836 SV* REG_ANY_invlist = _new_invlist(2);
4837 REG_ANY_invlist = add_cp_to_invlist(REG_ANY_invlist,
4839 if (flags & SCF_DO_STCLASS_OR) {
4840 ssc_union(data->start_class,
4842 TRUE /* TRUE => invert, hence all but \n
4846 else if (flags & SCF_DO_STCLASS_AND) {
4847 ssc_intersection(data->start_class,
4849 TRUE /* TRUE => invert */
4851 ssc_clear_locale(data->start_class);
4853 SvREFCNT_dec_NN(REG_ANY_invlist);
4858 if (flags & SCF_DO_STCLASS_AND)
4859 ssc_and(pRExC_state, data->start_class,
4860 (regnode_charclass *) scan);
4862 ssc_or(pRExC_state, data->start_class,
4863 (regnode_charclass *) scan);
4871 namedclass = classnum_to_namedclass(FLAGS(scan)) + invert;
4872 if (flags & SCF_DO_STCLASS_AND) {
4873 bool was_there = cBOOL(
4874 ANYOF_POSIXL_TEST(data->start_class,
4876 ANYOF_POSIXL_ZERO(data->start_class);
4877 if (was_there) { /* Do an AND */
4878 ANYOF_POSIXL_SET(data->start_class, namedclass);
4880 /* No individual code points can now match */
4881 data->start_class->invlist
4882 = sv_2mortal(_new_invlist(0));
4885 int complement = namedclass + ((invert) ? -1 : 1);
4887 assert(flags & SCF_DO_STCLASS_OR);
4889 /* If the complement of this class was already there,
4890 * the result is that they match all code points,
4891 * (\d + \D == everything). Remove the classes from
4892 * future consideration. Locale is not relevant in
4894 if (ANYOF_POSIXL_TEST(data->start_class, complement)) {
4895 ssc_match_all_cp(data->start_class);
4896 ANYOF_POSIXL_CLEAR(data->start_class, namedclass);
4897 ANYOF_POSIXL_CLEAR(data->start_class, complement);
4899 else { /* The usual case; just add this class to the
4901 ANYOF_POSIXL_SET(data->start_class, namedclass);
4906 case NPOSIXA: /* For these, we always know the exact set of
4911 if (FLAGS(scan) == _CC_ASCII) {
4912 my_invlist = invlist_clone(PL_XPosix_ptrs[_CC_ASCII]);
4915 _invlist_intersection(PL_XPosix_ptrs[FLAGS(scan)],
4916 PL_XPosix_ptrs[_CC_ASCII],
4927 my_invlist = invlist_clone(PL_XPosix_ptrs[FLAGS(scan)]);
4929 /* NPOSIXD matches all upper Latin1 code points unless the
4930 * target string being matched is UTF-8, which is
4931 * unknowable until match time. Since we are going to
4932 * invert, we want to get rid of all of them so that the
4933 * inversion will match all */
4934 if (OP(scan) == NPOSIXD) {
4935 _invlist_subtract(my_invlist, PL_UpperLatin1,
4941 if (flags & SCF_DO_STCLASS_AND) {
4942 ssc_intersection(data->start_class, my_invlist, invert);
4943 ssc_clear_locale(data->start_class);
4946 assert(flags & SCF_DO_STCLASS_OR);
4947 ssc_union(data->start_class, my_invlist, invert);
4949 SvREFCNT_dec(my_invlist);
4951 if (flags & SCF_DO_STCLASS_OR)
4952 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
4953 flags &= ~SCF_DO_STCLASS;
4956 else if (PL_regkind[OP(scan)] == EOL && flags & SCF_DO_SUBSTR) {
4957 data->flags |= (OP(scan) == MEOL
4960 scan_commit(pRExC_state, data, minlenp, is_inf);
4963 else if ( PL_regkind[OP(scan)] == BRANCHJ
4964 /* Lookbehind, or need to calculate parens/evals/stclass: */
4965 && (scan->flags || data || (flags & SCF_DO_STCLASS))
4966 && (OP(scan) == IFMATCH || OP(scan) == UNLESSM)) {
4967 if ( OP(scan) == UNLESSM &&
4969 OP(NEXTOPER(NEXTOPER(scan))) == NOTHING &&
4970 OP(regnext(NEXTOPER(NEXTOPER(scan)))) == SUCCEED
4973 regnode *upto= regnext(scan);
4975 SV * const mysv_val=sv_newmortal();
4976 DEBUG_STUDYDATA("OPFAIL",data,depth);
4978 /*DEBUG_PARSE_MSG("opfail");*/
4979 regprop(RExC_rx, mysv_val, upto, NULL);
4980 PerlIO_printf(Perl_debug_log,
4981 "~ replace with OPFAIL pointed at %s (%"IVdf") offset %"IVdf"\n",
4982 SvPV_nolen_const(mysv_val),
4983 (IV)REG_NODE_NUM(upto),
4988 NEXT_OFF(scan) = upto - scan;
4989 for (opt= scan + 1; opt < upto ; opt++)
4990 OP(opt) = OPTIMIZED;
4994 if ( !PERL_ENABLE_POSITIVE_ASSERTION_STUDY
4995 || OP(scan) == UNLESSM )
4997 /* Negative Lookahead/lookbehind
4998 In this case we can't do fixed string optimisation.
5001 SSize_t deltanext, minnext, fake = 0;
5006 data_fake.flags = 0;
5008 data_fake.whilem_c = data->whilem_c;
5009 data_fake.last_closep = data->last_closep;
5012 data_fake.last_closep = &fake;
5013 data_fake.pos_delta = delta;
5014 if ( flags & SCF_DO_STCLASS && !scan->flags
5015 && OP(scan) == IFMATCH ) { /* Lookahead */
5016 ssc_init(pRExC_state, &intrnl);
5017 data_fake.start_class = &intrnl;
5018 f |= SCF_DO_STCLASS_AND;
5020 if (flags & SCF_WHILEM_VISITED_POS)
5021 f |= SCF_WHILEM_VISITED_POS;
5022 next = regnext(scan);
5023 nscan = NEXTOPER(NEXTOPER(scan));
5024 minnext = study_chunk(pRExC_state, &nscan, minlenp, &deltanext,
5025 last, &data_fake, stopparen,
5026 recursed_depth, NULL, f, depth+1);
5029 FAIL("Variable length lookbehind not implemented");
5031 else if (minnext > (I32)U8_MAX) {
5032 FAIL2("Lookbehind longer than %"UVuf" not implemented",
5035 scan->flags = (U8)minnext;
5038 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
5040 if (data_fake.flags & SF_HAS_EVAL)
5041 data->flags |= SF_HAS_EVAL;
5042 data->whilem_c = data_fake.whilem_c;
5044 if (f & SCF_DO_STCLASS_AND) {
5045 if (flags & SCF_DO_STCLASS_OR) {
5046 /* OR before, AND after: ideally we would recurse with
5047 * data_fake to get the AND applied by study of the
5048 * remainder of the pattern, and then derecurse;
5049 * *** HACK *** for now just treat as "no information".
5050 * See [perl #56690].
5052 ssc_init(pRExC_state, data->start_class);
5054 /* AND before and after: combine and continue. These
5055 * assertions are zero-length, so can match an EMPTY
5057 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) &intrnl);
5058 ANYOF_FLAGS(data->start_class) |= ANYOF_EMPTY_STRING;
5062 #if PERL_ENABLE_POSITIVE_ASSERTION_STUDY
5064 /* Positive Lookahead/lookbehind
5065 In this case we can do fixed string optimisation,
5066 but we must be careful about it. Note in the case of
5067 lookbehind the positions will be offset by the minimum
5068 length of the pattern, something we won't know about
5069 until after the recurse.
5071 SSize_t deltanext, fake = 0;
5075 /* We use SAVEFREEPV so that when the full compile
5076 is finished perl will clean up the allocated
5077 minlens when it's all done. This way we don't
5078 have to worry about freeing them when we know
5079 they wont be used, which would be a pain.
5082 Newx( minnextp, 1, SSize_t );
5083 SAVEFREEPV(minnextp);
5086 StructCopy(data, &data_fake, scan_data_t);
5087 if ((flags & SCF_DO_SUBSTR) && data->last_found) {
5090 scan_commit(pRExC_state, &data_fake, minlenp, is_inf);
5091 data_fake.last_found=newSVsv(data->last_found);
5095 data_fake.last_closep = &fake;
5096 data_fake.flags = 0;
5097 data_fake.pos_delta = delta;
5099 data_fake.flags |= SF_IS_INF;
5100 if ( flags & SCF_DO_STCLASS && !scan->flags
5101 && OP(scan) == IFMATCH ) { /* Lookahead */
5102 ssc_init(pRExC_state, &intrnl);
5103 data_fake.start_class = &intrnl;
5104 f |= SCF_DO_STCLASS_AND;
5106 if (flags & SCF_WHILEM_VISITED_POS)
5107 f |= SCF_WHILEM_VISITED_POS;
5108 next = regnext(scan);
5109 nscan = NEXTOPER(NEXTOPER(scan));
5111 *minnextp = study_chunk(pRExC_state, &nscan, minnextp,
5112 &deltanext, last, &data_fake,
5113 stopparen, recursed_depth, NULL,
5117 FAIL("Variable length lookbehind not implemented");
5119 else if (*minnextp > (I32)U8_MAX) {
5120 FAIL2("Lookbehind longer than %"UVuf" not implemented",
5123 scan->flags = (U8)*minnextp;
5128 if (f & SCF_DO_STCLASS_AND) {
5129 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) &intrnl);
5130 ANYOF_FLAGS(data->start_class) |= ANYOF_EMPTY_STRING;
5133 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
5135 if (data_fake.flags & SF_HAS_EVAL)
5136 data->flags |= SF_HAS_EVAL;
5137 data->whilem_c = data_fake.whilem_c;
5138 if ((flags & SCF_DO_SUBSTR) && data_fake.last_found) {
5139 if (RExC_rx->minlen<*minnextp)
5140 RExC_rx->minlen=*minnextp;
5141 scan_commit(pRExC_state, &data_fake, minnextp, is_inf);
5142 SvREFCNT_dec_NN(data_fake.last_found);
5144 if ( data_fake.minlen_fixed != minlenp )
5146 data->offset_fixed= data_fake.offset_fixed;
5147 data->minlen_fixed= data_fake.minlen_fixed;
5148 data->lookbehind_fixed+= scan->flags;
5150 if ( data_fake.minlen_float != minlenp )
5152 data->minlen_float= data_fake.minlen_float;
5153 data->offset_float_min=data_fake.offset_float_min;
5154 data->offset_float_max=data_fake.offset_float_max;
5155 data->lookbehind_float+= scan->flags;
5162 else if (OP(scan) == OPEN) {
5163 if (stopparen != (I32)ARG(scan))
5166 else if (OP(scan) == CLOSE) {
5167 if (stopparen == (I32)ARG(scan)) {
5170 if ((I32)ARG(scan) == is_par) {
5171 next = regnext(scan);
5173 if ( next && (OP(next) != WHILEM) && next < last)
5174 is_par = 0; /* Disable optimization */
5177 *(data->last_closep) = ARG(scan);
5179 else if (OP(scan) == EVAL) {
5181 data->flags |= SF_HAS_EVAL;
5183 else if ( PL_regkind[OP(scan)] == ENDLIKE ) {
5184 if (flags & SCF_DO_SUBSTR) {
5185 scan_commit(pRExC_state, data, minlenp, is_inf);
5186 flags &= ~SCF_DO_SUBSTR;
5188 if (data && OP(scan)==ACCEPT) {
5189 data->flags |= SCF_SEEN_ACCEPT;
5194 else if (OP(scan) == LOGICAL && scan->flags == 2) /* Embedded follows */
5196 if (flags & SCF_DO_SUBSTR) {
5197 scan_commit(pRExC_state, data, minlenp, is_inf);
5198 data->longest = &(data->longest_float);
5200 is_inf = is_inf_internal = 1;
5201 if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
5202 ssc_anything(data->start_class);
5203 flags &= ~SCF_DO_STCLASS;
5205 else if (OP(scan) == GPOS) {
5206 if (!(RExC_rx->intflags & PREGf_GPOS_FLOAT) &&
5207 !(delta || is_inf || (data && data->pos_delta)))
5209 if (!(RExC_rx->intflags & PREGf_ANCH) && (flags & SCF_DO_SUBSTR))
5210 RExC_rx->intflags |= PREGf_ANCH_GPOS;
5211 if (RExC_rx->gofs < (STRLEN)min)
5212 RExC_rx->gofs = min;
5214 RExC_rx->intflags |= PREGf_GPOS_FLOAT;
5218 #ifdef TRIE_STUDY_OPT
5219 #ifdef FULL_TRIE_STUDY
5220 else if (PL_regkind[OP(scan)] == TRIE) {
5221 /* NOTE - There is similar code to this block above for handling
5222 BRANCH nodes on the initial study. If you change stuff here
5224 regnode *trie_node= scan;
5225 regnode *tail= regnext(scan);
5226 reg_trie_data *trie = (reg_trie_data*)RExC_rxi->data->data[ ARG(scan) ];
5227 SSize_t max1 = 0, min1 = SSize_t_MAX;
5230 if (flags & SCF_DO_SUBSTR) { /* XXXX Add !SUSPEND? */
5231 /* Cannot merge strings after this. */
5232 scan_commit(pRExC_state, data, minlenp, is_inf);
5234 if (flags & SCF_DO_STCLASS)
5235 ssc_init_zero(pRExC_state, &accum);
5241 const regnode *nextbranch= NULL;
5244 for ( word=1 ; word <= trie->wordcount ; word++)
5246 SSize_t deltanext=0, minnext=0, f = 0, fake;
5247 regnode_ssc this_class;
5249 data_fake.flags = 0;
5251 data_fake.whilem_c = data->whilem_c;
5252 data_fake.last_closep = data->last_closep;
5255 data_fake.last_closep = &fake;
5256 data_fake.pos_delta = delta;
5257 if (flags & SCF_DO_STCLASS) {
5258 ssc_init(pRExC_state, &this_class);
5259 data_fake.start_class = &this_class;
5260 f = SCF_DO_STCLASS_AND;
5262 if (flags & SCF_WHILEM_VISITED_POS)
5263 f |= SCF_WHILEM_VISITED_POS;
5265 if (trie->jump[word]) {
5267 nextbranch = trie_node + trie->jump[0];
5268 scan= trie_node + trie->jump[word];
5269 /* We go from the jump point to the branch that follows
5270 it. Note this means we need the vestigal unused
5271 branches even though they arent otherwise used. */
5272 minnext = study_chunk(pRExC_state, &scan, minlenp,
5273 &deltanext, (regnode *)nextbranch, &data_fake,
5274 stopparen, recursed_depth, NULL, f,depth+1);
5276 if (nextbranch && PL_regkind[OP(nextbranch)]==BRANCH)
5277 nextbranch= regnext((regnode*)nextbranch);
5279 if (min1 > (SSize_t)(minnext + trie->minlen))
5280 min1 = minnext + trie->minlen;
5281 if (deltanext == SSize_t_MAX) {
5282 is_inf = is_inf_internal = 1;
5284 } else if (max1 < (SSize_t)(minnext + deltanext + trie->maxlen))
5285 max1 = minnext + deltanext + trie->maxlen;
5287 if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
5289 if (data_fake.flags & SCF_SEEN_ACCEPT) {
5290 if ( stopmin > min + min1)
5291 stopmin = min + min1;
5292 flags &= ~SCF_DO_SUBSTR;
5294 data->flags |= SCF_SEEN_ACCEPT;
5297 if (data_fake.flags & SF_HAS_EVAL)
5298 data->flags |= SF_HAS_EVAL;
5299 data->whilem_c = data_fake.whilem_c;
5301 if (flags & SCF_DO_STCLASS)
5302 ssc_or(pRExC_state, &accum, (regnode_charclass *) &this_class);
5305 if (flags & SCF_DO_SUBSTR) {
5306 data->pos_min += min1;
5307 data->pos_delta += max1 - min1;
5308 if (max1 != min1 || is_inf)
5309 data->longest = &(data->longest_float);
5312 delta += max1 - min1;
5313 if (flags & SCF_DO_STCLASS_OR) {
5314 ssc_or(pRExC_state, data->start_class, (regnode_charclass *) &accum);
5316 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
5317 flags &= ~SCF_DO_STCLASS;
5320 else if (flags & SCF_DO_STCLASS_AND) {
5322 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) &accum);
5323 flags &= ~SCF_DO_STCLASS;
5326 /* Switch to OR mode: cache the old value of
5327 * data->start_class */
5329 StructCopy(data->start_class, and_withp, regnode_ssc);
5330 flags &= ~SCF_DO_STCLASS_AND;
5331 StructCopy(&accum, data->start_class, regnode_ssc);
5332 flags |= SCF_DO_STCLASS_OR;
5339 else if (PL_regkind[OP(scan)] == TRIE) {
5340 reg_trie_data *trie = (reg_trie_data*)RExC_rxi->data->data[ ARG(scan) ];
5343 min += trie->minlen;
5344 delta += (trie->maxlen - trie->minlen);
5345 flags &= ~SCF_DO_STCLASS; /* xxx */
5346 if (flags & SCF_DO_SUBSTR) {
5347 /* Cannot expect anything... */
5348 scan_commit(pRExC_state, data, minlenp, is_inf);
5349 data->pos_min += trie->minlen;
5350 data->pos_delta += (trie->maxlen - trie->minlen);
5351 if (trie->maxlen != trie->minlen)
5352 data->longest = &(data->longest_float);
5354 if (trie->jump) /* no more substrings -- for now /grr*/
5355 flags &= ~SCF_DO_SUBSTR;
5357 #endif /* old or new */
5358 #endif /* TRIE_STUDY_OPT */
5360 /* Else: zero-length, ignore. */
5361 scan = regnext(scan);
5363 /* If we are exiting a recursion we can unset its recursed bit
5364 * and allow ourselves to enter it again - no danger of an
5365 * infinite loop there.
5366 if (stopparen > -1 && recursed) {
5367 DEBUG_STUDYDATA("unset:", data,depth);
5368 PAREN_UNSET( recursed, stopparen);
5372 DEBUG_STUDYDATA("frame-end:",data,depth);
5373 DEBUG_PEEP("fend", scan, depth);
5374 /* restore previous context */
5377 stopparen = frame->stop;
5378 recursed_depth = frame->prev_recursed_depth;
5381 frame = frame->prev;
5382 goto fake_study_recurse;
5387 DEBUG_STUDYDATA("pre-fin:",data,depth);
5390 *deltap = is_inf_internal ? SSize_t_MAX : delta;
5392 if (flags & SCF_DO_SUBSTR && is_inf)
5393 data->pos_delta = SSize_t_MAX - data->pos_min;
5394 if (is_par > (I32)U8_MAX)
5396 if (is_par && pars==1 && data) {
5397 data->flags |= SF_IN_PAR;
5398 data->flags &= ~SF_HAS_PAR;
5400 else if (pars && data) {
5401 data->flags |= SF_HAS_PAR;
5402 data->flags &= ~SF_IN_PAR;
5404 if (flags & SCF_DO_STCLASS_OR)
5405 ssc_and(pRExC_state, data->start_class, (regnode_charclass *) and_withp);
5406 if (flags & SCF_TRIE_RESTUDY)
5407 data->flags |= SCF_TRIE_RESTUDY;
5409 DEBUG_STUDYDATA("post-fin:",data,depth);
5412 SSize_t final_minlen= min < stopmin ? min : stopmin;
5414 if (!(RExC_seen & REG_UNBOUNDED_QUANTIFIER_SEEN) && (RExC_maxlen < final_minlen + delta)) {
5415 RExC_maxlen = final_minlen + delta;
5417 return final_minlen;
5423 S_add_data(RExC_state_t* const pRExC_state, const char* const s, const U32 n)
5425 U32 count = RExC_rxi->data ? RExC_rxi->data->count : 0;
5427 PERL_ARGS_ASSERT_ADD_DATA;
5429 Renewc(RExC_rxi->data,
5430 sizeof(*RExC_rxi->data) + sizeof(void*) * (count + n - 1),
5431 char, struct reg_data);
5433 Renew(RExC_rxi->data->what, count + n, U8);
5435 Newx(RExC_rxi->data->what, n, U8);
5436 RExC_rxi->data->count = count + n;
5437 Copy(s, RExC_rxi->data->what + count, n, U8);
5441 /*XXX: todo make this not included in a non debugging perl */
5442 #ifndef PERL_IN_XSUB_RE
5444 Perl_reginitcolors(pTHX)
5447 const char * const s = PerlEnv_getenv("PERL_RE_COLORS");
5449 char *t = savepv(s);
5453 t = strchr(t, '\t');
5459 PL_colors[i] = t = (char *)"";
5464 PL_colors[i++] = (char *)"";
5471 #ifdef TRIE_STUDY_OPT
5472 #define CHECK_RESTUDY_GOTO_butfirst(dOsomething) \
5475 (data.flags & SCF_TRIE_RESTUDY) \
5483 #define CHECK_RESTUDY_GOTO_butfirst
5487 * pregcomp - compile a regular expression into internal code
5489 * Decides which engine's compiler to call based on the hint currently in
5493 #ifndef PERL_IN_XSUB_RE
5495 /* return the currently in-scope regex engine (or the default if none) */
5497 regexp_engine const *
5498 Perl_current_re_engine(pTHX)
5502 if (IN_PERL_COMPILETIME) {
5503 HV * const table = GvHV(PL_hintgv);
5506 if (!table || !(PL_hints & HINT_LOCALIZE_HH))
5507 return &reh_regexp_engine;
5508 ptr = hv_fetchs(table, "regcomp", FALSE);
5509 if ( !(ptr && SvIOK(*ptr) && SvIV(*ptr)))
5510 return &reh_regexp_engine;
5511 return INT2PTR(regexp_engine*,SvIV(*ptr));
5515 if (!PL_curcop->cop_hints_hash)
5516 return &reh_regexp_engine;
5517 ptr = cop_hints_fetch_pvs(PL_curcop, "regcomp", 0);
5518 if ( !(ptr && SvIOK(ptr) && SvIV(ptr)))
5519 return &reh_regexp_engine;
5520 return INT2PTR(regexp_engine*,SvIV(ptr));
5526 Perl_pregcomp(pTHX_ SV * const pattern, const U32 flags)
5529 regexp_engine const *eng = current_re_engine();
5530 GET_RE_DEBUG_FLAGS_DECL;
5532 PERL_ARGS_ASSERT_PREGCOMP;
5534 /* Dispatch a request to compile a regexp to correct regexp engine. */
5536 PerlIO_printf(Perl_debug_log, "Using engine %"UVxf"\n",
5539 return CALLREGCOMP_ENG(eng, pattern, flags);
5543 /* public(ish) entry point for the perl core's own regex compiling code.
5544 * It's actually a wrapper for Perl_re_op_compile that only takes an SV
5545 * pattern rather than a list of OPs, and uses the internal engine rather
5546 * than the current one */
5549 Perl_re_compile(pTHX_ SV * const pattern, U32 rx_flags)
5551 SV *pat = pattern; /* defeat constness! */
5552 PERL_ARGS_ASSERT_RE_COMPILE;
5553 return Perl_re_op_compile(aTHX_ &pat, 1, NULL,
5554 #ifdef PERL_IN_XSUB_RE
5559 NULL, NULL, rx_flags, 0);
5563 /* upgrade pattern pat_p of length plen_p to UTF8, and if there are code
5564 * blocks, recalculate the indices. Update pat_p and plen_p in-place to
5565 * point to the realloced string and length.
5567 * This is essentially a copy of Perl_bytes_to_utf8() with the code index
5571 S_pat_upgrade_to_utf8(pTHX_ RExC_state_t * const pRExC_state,
5572 char **pat_p, STRLEN *plen_p, int num_code_blocks)
5574 U8 *const src = (U8*)*pat_p;
5577 STRLEN s = 0, d = 0;
5579 GET_RE_DEBUG_FLAGS_DECL;
5581 DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log,
5582 "UTF8 mismatch! Converting to utf8 for resizing and compile\n"));
5584 Newx(dst, *plen_p * 2 + 1, U8);
5586 while (s < *plen_p) {
5587 if (NATIVE_BYTE_IS_INVARIANT(src[s]))
5590 dst[d++] = UTF8_EIGHT_BIT_HI(src[s]);
5591 dst[d] = UTF8_EIGHT_BIT_LO(src[s]);
5593 if (n < num_code_blocks) {
5594 if (!do_end && pRExC_state->code_blocks[n].start == s) {
5595 pRExC_state->code_blocks[n].start = d;
5596 assert(dst[d] == '(');
5599 else if (do_end && pRExC_state->code_blocks[n].end == s) {
5600 pRExC_state->code_blocks[n].end = d;
5601 assert(dst[d] == ')');
5611 *pat_p = (char*) dst;
5613 RExC_orig_utf8 = RExC_utf8 = 1;
5618 /* S_concat_pat(): concatenate a list of args to the pattern string pat,
5619 * while recording any code block indices, and handling overloading,
5620 * nested qr// objects etc. If pat is null, it will allocate a new
5621 * string, or just return the first arg, if there's only one.
5623 * Returns the malloced/updated pat.
5624 * patternp and pat_count is the array of SVs to be concatted;
5625 * oplist is the optional list of ops that generated the SVs;
5626 * recompile_p is a pointer to a boolean that will be set if
5627 * the regex will need to be recompiled.
5628 * delim, if non-null is an SV that will be inserted between each element
5632 S_concat_pat(pTHX_ RExC_state_t * const pRExC_state,
5633 SV *pat, SV ** const patternp, int pat_count,
5634 OP *oplist, bool *recompile_p, SV *delim)
5638 bool use_delim = FALSE;
5639 bool alloced = FALSE;
5641 /* if we know we have at least two args, create an empty string,
5642 * then concatenate args to that. For no args, return an empty string */
5643 if (!pat && pat_count != 1) {
5644 pat = newSVpvn("", 0);
5649 for (svp = patternp; svp < patternp + pat_count; svp++) {
5652 STRLEN orig_patlen = 0;
5654 SV *msv = use_delim ? delim : *svp;
5655 if (!msv) msv = &PL_sv_undef;
5657 /* if we've got a delimiter, we go round the loop twice for each
5658 * svp slot (except the last), using the delimiter the second
5667 if (SvTYPE(msv) == SVt_PVAV) {
5668 /* we've encountered an interpolated array within
5669 * the pattern, e.g. /...@a..../. Expand the list of elements,
5670 * then recursively append elements.
5671 * The code in this block is based on S_pushav() */
5673 AV *const av = (AV*)msv;
5674 const SSize_t maxarg = AvFILL(av) + 1;
5678 assert(oplist->op_type == OP_PADAV
5679 || oplist->op_type == OP_RV2AV);
5680 oplist = oplist->op_sibling;;
5683 if (SvRMAGICAL(av)) {
5686 Newx(array, maxarg, SV*);
5688 for (i=0; i < maxarg; i++) {
5689 SV ** const svp = av_fetch(av, i, FALSE);
5690 array[i] = svp ? *svp : &PL_sv_undef;
5694 array = AvARRAY(av);
5696 pat = S_concat_pat(aTHX_ pRExC_state, pat,
5697 array, maxarg, NULL, recompile_p,
5699 GvSV((gv_fetchpvs("\"", GV_ADDMULTI, SVt_PV))));
5705 /* we make the assumption here that each op in the list of
5706 * op_siblings maps to one SV pushed onto the stack,
5707 * except for code blocks, with have both an OP_NULL and
5709 * This allows us to match up the list of SVs against the
5710 * list of OPs to find the next code block.
5712 * Note that PUSHMARK PADSV PADSV ..
5714 * PADRANGE PADSV PADSV ..
5715 * so the alignment still works. */
5718 if (oplist->op_type == OP_NULL
5719 && (oplist->op_flags & OPf_SPECIAL))
5721 assert(n < pRExC_state->num_code_blocks);
5722 pRExC_state->code_blocks[n].start = pat ? SvCUR(pat) : 0;
5723 pRExC_state->code_blocks[n].block = oplist;
5724 pRExC_state->code_blocks[n].src_regex = NULL;
5727 oplist = oplist->op_sibling; /* skip CONST */
5730 oplist = oplist->op_sibling;;
5733 /* apply magic and QR overloading to arg */
5736 if (SvROK(msv) && SvAMAGIC(msv)) {
5737 SV *sv = AMG_CALLunary(msv, regexp_amg);
5741 if (SvTYPE(sv) != SVt_REGEXP)
5742 Perl_croak(aTHX_ "Overloaded qr did not return a REGEXP");
5747 /* try concatenation overload ... */
5748 if (pat && (SvAMAGIC(pat) || SvAMAGIC(msv)) &&
5749 (sv = amagic_call(pat, msv, concat_amg, AMGf_assign)))
5752 /* overloading involved: all bets are off over literal
5753 * code. Pretend we haven't seen it */
5754 pRExC_state->num_code_blocks -= n;
5758 /* ... or failing that, try "" overload */
5759 while (SvAMAGIC(msv)
5760 && (sv = AMG_CALLunary(msv, string_amg))
5764 && SvRV(msv) == SvRV(sv))
5769 if (SvROK(msv) && SvTYPE(SvRV(msv)) == SVt_REGEXP)
5773 /* this is a partially unrolled
5774 * sv_catsv_nomg(pat, msv);
5775 * that allows us to adjust code block indices if
5778 char *dst = SvPV_force_nomg(pat, dlen);
5780 if (SvUTF8(msv) && !SvUTF8(pat)) {
5781 S_pat_upgrade_to_utf8(aTHX_ pRExC_state, &dst, &dlen, n);
5782 sv_setpvn(pat, dst, dlen);
5785 sv_catsv_nomg(pat, msv);
5792 pRExC_state->code_blocks[n-1].end = SvCUR(pat)-1;
5795 /* extract any code blocks within any embedded qr//'s */
5796 if (rx && SvTYPE(rx) == SVt_REGEXP
5797 && RX_ENGINE((REGEXP*)rx)->op_comp)
5800 RXi_GET_DECL(ReANY((REGEXP *)rx), ri);
5801 if (ri->num_code_blocks) {
5803 /* the presence of an embedded qr// with code means
5804 * we should always recompile: the text of the
5805 * qr// may not have changed, but it may be a
5806 * different closure than last time */
5808 Renew(pRExC_state->code_blocks,
5809 pRExC_state->num_code_blocks + ri->num_code_blocks,
5810 struct reg_code_block);
5811 pRExC_state->num_code_blocks += ri->num_code_blocks;
5813 for (i=0; i < ri->num_code_blocks; i++) {
5814 struct reg_code_block *src, *dst;
5815 STRLEN offset = orig_patlen
5816 + ReANY((REGEXP *)rx)->pre_prefix;
5817 assert(n < pRExC_state->num_code_blocks);
5818 src = &ri->code_blocks[i];
5819 dst = &pRExC_state->code_blocks[n];
5820 dst->start = src->start + offset;
5821 dst->end = src->end + offset;
5822 dst->block = src->block;
5823 dst->src_regex = (REGEXP*) SvREFCNT_inc( (SV*)
5832 /* avoid calling magic multiple times on a single element e.g. =~ $qr */
5841 /* see if there are any run-time code blocks in the pattern.
5842 * False positives are allowed */
5845 S_has_runtime_code(pTHX_ RExC_state_t * const pRExC_state,
5846 char *pat, STRLEN plen)
5851 for (s = 0; s < plen; s++) {
5852 if (n < pRExC_state->num_code_blocks
5853 && s == pRExC_state->code_blocks[n].start)
5855 s = pRExC_state->code_blocks[n].end;
5859 /* TODO ideally should handle [..], (#..), /#.../x to reduce false
5861 if (pat[s] == '(' && s+2 <= plen && pat[s+1] == '?' &&
5863 || (s + 2 <= plen && pat[s+2] == '?' && pat[s+3] == '{'))
5870 /* Handle run-time code blocks. We will already have compiled any direct
5871 * or indirect literal code blocks. Now, take the pattern 'pat' and make a
5872 * copy of it, but with any literal code blocks blanked out and
5873 * appropriate chars escaped; then feed it into
5875 * eval "qr'modified_pattern'"
5879 * a\bc(?{"this was literal"})def'ghi\\jkl(?{"this is runtime"})mno
5883 * qr'a\\bc_______________________def\'ghi\\\\jkl(?{"this is runtime"})mno'
5885 * After eval_sv()-ing that, grab any new code blocks from the returned qr
5886 * and merge them with any code blocks of the original regexp.
5888 * If the pat is non-UTF8, while the evalled qr is UTF8, don't merge;
5889 * instead, just save the qr and return FALSE; this tells our caller that
5890 * the original pattern needs upgrading to utf8.
5894 S_compile_runtime_code(pTHX_ RExC_state_t * const pRExC_state,
5895 char *pat, STRLEN plen)
5899 GET_RE_DEBUG_FLAGS_DECL;
5901 if (pRExC_state->runtime_code_qr) {
5902 /* this is the second time we've been called; this should
5903 * only happen if the main pattern got upgraded to utf8
5904 * during compilation; re-use the qr we compiled first time
5905 * round (which should be utf8 too)
5907 qr = pRExC_state->runtime_code_qr;
5908 pRExC_state->runtime_code_qr = NULL;
5909 assert(RExC_utf8 && SvUTF8(qr));
5915 int newlen = plen + 6; /* allow for "qr''x\0" extra chars */
5919 /* determine how many extra chars we need for ' and \ escaping */
5920 for (s = 0; s < plen; s++) {
5921 if (pat[s] == '\'' || pat[s] == '\\')
5925 Newx(newpat, newlen, char);
5927 *p++ = 'q'; *p++ = 'r'; *p++ = '\'';
5929 for (s = 0; s < plen; s++) {
5930 if (n < pRExC_state->num_code_blocks
5931 && s == pRExC_state->code_blocks[n].start)
5933 /* blank out literal code block */
5934 assert(pat[s] == '(');
5935 while (s <= pRExC_state->code_blocks[n].end) {
5943 if (pat[s] == '\'' || pat[s] == '\\')
5948 if (pRExC_state->pm_flags & RXf_PMf_EXTENDED)
5952 PerlIO_printf(Perl_debug_log,
5953 "%sre-parsing pattern for runtime code:%s %s\n",
5954 PL_colors[4],PL_colors[5],newpat);
5957 sv = newSVpvn_flags(newpat, p-newpat-1, RExC_utf8 ? SVf_UTF8 : 0);
5963 PUSHSTACKi(PERLSI_REQUIRE);
5964 /* G_RE_REPARSING causes the toker to collapse \\ into \ when
5965 * parsing qr''; normally only q'' does this. It also alters
5967 eval_sv(sv, G_SCALAR|G_RE_REPARSING);
5968 SvREFCNT_dec_NN(sv);
5973 SV * const errsv = ERRSV;
5974 if (SvTRUE_NN(errsv))
5976 Safefree(pRExC_state->code_blocks);
5977 /* use croak_sv ? */
5978 Perl_croak_nocontext("%"SVf, SVfARG(errsv));
5981 assert(SvROK(qr_ref));
5983 assert(SvTYPE(qr) == SVt_REGEXP && RX_ENGINE((REGEXP*)qr)->op_comp);
5984 /* the leaving below frees the tmp qr_ref.
5985 * Give qr a life of its own */
5993 if (!RExC_utf8 && SvUTF8(qr)) {
5994 /* first time through; the pattern got upgraded; save the
5995 * qr for the next time through */
5996 assert(!pRExC_state->runtime_code_qr);
5997 pRExC_state->runtime_code_qr = qr;
6002 /* extract any code blocks within the returned qr// */
6005 /* merge the main (r1) and run-time (r2) code blocks into one */
6007 RXi_GET_DECL(ReANY((REGEXP *)qr), r2);
6008 struct reg_code_block *new_block, *dst;
6009 RExC_state_t * const r1 = pRExC_state; /* convenient alias */
6012 if (!r2->num_code_blocks) /* we guessed wrong */
6014 SvREFCNT_dec_NN(qr);
6019 r1->num_code_blocks + r2->num_code_blocks,
6020 struct reg_code_block);
6023 while ( i1 < r1->num_code_blocks
6024 || i2 < r2->num_code_blocks)
6026 struct reg_code_block *src;
6029 if (i1 == r1->num_code_blocks) {
6030 src = &r2->code_blocks[i2++];
6033 else if (i2 == r2->num_code_blocks)
6034 src = &r1->code_blocks[i1++];
6035 else if ( r1->code_blocks[i1].start
6036 < r2->code_blocks[i2].start)
6038 src = &r1->code_blocks[i1++];
6039 assert(src->end < r2->code_blocks[i2].start);
6042 assert( r1->code_blocks[i1].start
6043 > r2->code_blocks[i2].start);
6044 src = &r2->code_blocks[i2++];
6046 assert(src->end < r1->code_blocks[i1].start);
6049 assert(pat[src->start] == '(');
6050 assert(pat[src->end] == ')');
6051 dst->start = src->start;
6052 dst->end = src->end;
6053 dst->block = src->block;
6054 dst->src_regex = is_qr ? (REGEXP*) SvREFCNT_inc( (SV*) qr)
6058 r1->num_code_blocks += r2->num_code_blocks;
6059 Safefree(r1->code_blocks);
6060 r1->code_blocks = new_block;
6063 SvREFCNT_dec_NN(qr);
6069 S_setup_longest(pTHX_ RExC_state_t *pRExC_state, SV* sv_longest,
6070 SV** rx_utf8, SV** rx_substr, SSize_t* rx_end_shift,
6071 SSize_t lookbehind, SSize_t offset, SSize_t *minlen,
6072 STRLEN longest_length, bool eol, bool meol)
6074 /* This is the common code for setting up the floating and fixed length
6075 * string data extracted from Perl_re_op_compile() below. Returns a boolean
6076 * as to whether succeeded or not */
6081 if (! (longest_length
6082 || (eol /* Can't have SEOL and MULTI */
6083 && (! meol || (RExC_flags & RXf_PMf_MULTILINE)))
6085 /* See comments for join_exact for why REG_UNFOLDED_MULTI_SEEN */
6086 || (RExC_seen & REG_UNFOLDED_MULTI_SEEN))
6091 /* copy the information about the longest from the reg_scan_data
6092 over to the program. */
6093 if (SvUTF8(sv_longest)) {
6094 *rx_utf8 = sv_longest;
6097 *rx_substr = sv_longest;
6100 /* end_shift is how many chars that must be matched that
6101 follow this item. We calculate it ahead of time as once the
6102 lookbehind offset is added in we lose the ability to correctly
6104 ml = minlen ? *(minlen) : (SSize_t)longest_length;
6105 *rx_end_shift = ml - offset
6106 - longest_length + (SvTAIL(sv_longest) != 0)
6109 t = (eol/* Can't have SEOL and MULTI */
6110 && (! meol || (RExC_flags & RXf_PMf_MULTILINE)));
6111 fbm_compile(sv_longest, t ? FBMcf_TAIL : 0);
6117 * Perl_re_op_compile - the perl internal RE engine's function to compile a
6118 * regular expression into internal code.
6119 * The pattern may be passed either as:
6120 * a list of SVs (patternp plus pat_count)
6121 * a list of OPs (expr)
6122 * If both are passed, the SV list is used, but the OP list indicates
6123 * which SVs are actually pre-compiled code blocks
6125 * The SVs in the list have magic and qr overloading applied to them (and
6126 * the list may be modified in-place with replacement SVs in the latter
6129 * If the pattern hasn't changed from old_re, then old_re will be
6132 * eng is the current engine. If that engine has an op_comp method, then
6133 * handle directly (i.e. we assume that op_comp was us); otherwise, just
6134 * do the initial concatenation of arguments and pass on to the external
6137 * If is_bare_re is not null, set it to a boolean indicating whether the
6138 * arg list reduced (after overloading) to a single bare regex which has
6139 * been returned (i.e. /$qr/).
6141 * orig_rx_flags contains RXf_* flags. See perlreapi.pod for more details.
6143 * pm_flags contains the PMf_* flags, typically based on those from the
6144 * pm_flags field of the related PMOP. Currently we're only interested in
6145 * PMf_HAS_CV, PMf_IS_QR, PMf_USE_RE_EVAL.
6147 * We can't allocate space until we know how big the compiled form will be,
6148 * but we can't compile it (and thus know how big it is) until we've got a
6149 * place to put the code. So we cheat: we compile it twice, once with code
6150 * generation turned off and size counting turned on, and once "for real".
6151 * This also means that we don't allocate space until we are sure that the
6152 * thing really will compile successfully, and we never have to move the
6153 * code and thus invalidate pointers into it. (Note that it has to be in
6154 * one piece because free() must be able to free it all.) [NB: not true in perl]
6156 * Beware that the optimization-preparation code in here knows about some
6157 * of the structure of the compiled regexp. [I'll say.]
6161 Perl_re_op_compile(pTHX_ SV ** const patternp, int pat_count,
6162 OP *expr, const regexp_engine* eng, REGEXP *old_re,
6163 bool *is_bare_re, U32 orig_rx_flags, U32 pm_flags)
6168 regexp_internal *ri;
6176 SV *code_blocksv = NULL;
6177 SV** new_patternp = patternp;
6179 /* these are all flags - maybe they should be turned
6180 * into a single int with different bit masks */
6181 I32 sawlookahead = 0;
6186 regex_charset initial_charset = get_regex_charset(orig_rx_flags);
6188 bool runtime_code = 0;
6190 RExC_state_t RExC_state;
6191 RExC_state_t * const pRExC_state = &RExC_state;
6192 #ifdef TRIE_STUDY_OPT
6194 RExC_state_t copyRExC_state;
6196 GET_RE_DEBUG_FLAGS_DECL;
6198 PERL_ARGS_ASSERT_RE_OP_COMPILE;
6200 DEBUG_r(if (!PL_colorset) reginitcolors());
6202 #ifndef PERL_IN_XSUB_RE
6203 /* Initialize these here instead of as-needed, as is quick and avoids
6204 * having to test them each time otherwise */
6205 if (! PL_AboveLatin1) {
6206 PL_AboveLatin1 = _new_invlist_C_array(AboveLatin1_invlist);
6207 PL_Latin1 = _new_invlist_C_array(Latin1_invlist);
6208 PL_UpperLatin1 = _new_invlist_C_array(UpperLatin1_invlist);
6209 PL_utf8_foldable = _new_invlist_C_array(_Perl_Any_Folds_invlist);
6210 PL_HasMultiCharFold =
6211 _new_invlist_C_array(_Perl_Folds_To_Multi_Char_invlist);
6215 pRExC_state->code_blocks = NULL;
6216 pRExC_state->num_code_blocks = 0;
6219 *is_bare_re = FALSE;
6221 if (expr && (expr->op_type == OP_LIST ||
6222 (expr->op_type == OP_NULL && expr->op_targ == OP_LIST))) {
6223 /* allocate code_blocks if needed */
6227 for (o = cLISTOPx(expr)->op_first; o; o = o->op_sibling)
6228 if (o->op_type == OP_NULL && (o->op_flags & OPf_SPECIAL))
6229 ncode++; /* count of DO blocks */
6231 pRExC_state->num_code_blocks = ncode;
6232 Newx(pRExC_state->code_blocks, ncode, struct reg_code_block);
6237 /* compile-time pattern with just OP_CONSTs and DO blocks */
6242 /* find how many CONSTs there are */
6245 if (expr->op_type == OP_CONST)
6248 for (o = cLISTOPx(expr)->op_first; o; o = o->op_sibling) {
6249 if (o->op_type == OP_CONST)
6253 /* fake up an SV array */
6255 assert(!new_patternp);
6256 Newx(new_patternp, n, SV*);
6257 SAVEFREEPV(new_patternp);
6261 if (expr->op_type == OP_CONST)
6262 new_patternp[n] = cSVOPx_sv(expr);
6264 for (o = cLISTOPx(expr)->op_first; o; o = o->op_sibling) {
6265 if (o->op_type == OP_CONST)
6266 new_patternp[n++] = cSVOPo_sv;
6271 DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log,
6272 "Assembling pattern from %d elements%s\n", pat_count,
6273 orig_rx_flags & RXf_SPLIT ? " for split" : ""));
6275 /* set expr to the first arg op */
6277 if (pRExC_state->num_code_blocks
6278 && expr->op_type != OP_CONST)
6280 expr = cLISTOPx(expr)->op_first;
6281 assert( expr->op_type == OP_PUSHMARK
6282 || (expr->op_type == OP_NULL && expr->op_targ == OP_PUSHMARK)
6283 || expr->op_type == OP_PADRANGE);
6284 expr = expr->op_sibling;
6287 pat = S_concat_pat(aTHX_ pRExC_state, NULL, new_patternp, pat_count,
6288 expr, &recompile, NULL);
6290 /* handle bare (possibly after overloading) regex: foo =~ $re */
6295 if (SvTYPE(re) == SVt_REGEXP) {
6299 Safefree(pRExC_state->code_blocks);
6300 DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log,
6301 "Precompiled pattern%s\n",
6302 orig_rx_flags & RXf_SPLIT ? " for split" : ""));
6308 exp = SvPV_nomg(pat, plen);
6310 if (!eng->op_comp) {
6311 if ((SvUTF8(pat) && IN_BYTES)
6312 || SvGMAGICAL(pat) || SvAMAGIC(pat))
6314 /* make a temporary copy; either to convert to bytes,
6315 * or to avoid repeating get-magic / overloaded stringify */
6316 pat = newSVpvn_flags(exp, plen, SVs_TEMP |
6317 (IN_BYTES ? 0 : SvUTF8(pat)));
6319 Safefree(pRExC_state->code_blocks);
6320 return CALLREGCOMP_ENG(eng, pat, orig_rx_flags);
6323 /* ignore the utf8ness if the pattern is 0 length */
6324 RExC_utf8 = RExC_orig_utf8 = (plen == 0 || IN_BYTES) ? 0 : SvUTF8(pat);
6325 RExC_uni_semantics = 0;
6326 RExC_contains_locale = 0;
6327 RExC_contains_i = 0;
6328 pRExC_state->runtime_code_qr = NULL;
6331 SV *dsv= sv_newmortal();
6332 RE_PV_QUOTED_DECL(s, RExC_utf8, dsv, exp, plen, 60);
6333 PerlIO_printf(Perl_debug_log, "%sCompiling REx%s %s\n",
6334 PL_colors[4],PL_colors[5],s);
6338 /* we jump here if we upgrade the pattern to utf8 and have to
6341 if ((pm_flags & PMf_USE_RE_EVAL)
6342 /* this second condition covers the non-regex literal case,
6343 * i.e. $foo =~ '(?{})'. */
6344 || (IN_PERL_COMPILETIME && (PL_hints & HINT_RE_EVAL))
6346 runtime_code = S_has_runtime_code(aTHX_ pRExC_state, exp, plen);
6348 /* return old regex if pattern hasn't changed */
6349 /* XXX: note in the below we have to check the flags as well as the
6352 * Things get a touch tricky as we have to compare the utf8 flag
6353 * independently from the compile flags. */
6357 && !!RX_UTF8(old_re) == !!RExC_utf8
6358 && ( RX_COMPFLAGS(old_re) == ( orig_rx_flags & RXf_PMf_FLAGCOPYMASK ) )
6359 && RX_PRECOMP(old_re)
6360 && RX_PRELEN(old_re) == plen
6361 && memEQ(RX_PRECOMP(old_re), exp, plen)
6362 && !runtime_code /* with runtime code, always recompile */ )
6364 Safefree(pRExC_state->code_blocks);
6368 rx_flags = orig_rx_flags;
6370 if (rx_flags & PMf_FOLD) {
6371 RExC_contains_i = 1;
6373 if (RExC_utf8 && initial_charset == REGEX_DEPENDS_CHARSET) {
6375 /* Set to use unicode semantics if the pattern is in utf8 and has the
6376 * 'depends' charset specified, as it means unicode when utf8 */
6377 set_regex_charset(&rx_flags, REGEX_UNICODE_CHARSET);
6381 RExC_flags = rx_flags;
6382 RExC_pm_flags = pm_flags;
6385 if (TAINTING_get && TAINT_get)
6386 Perl_croak(aTHX_ "Eval-group in insecure regular expression");
6388 if (!S_compile_runtime_code(aTHX_ pRExC_state, exp, plen)) {
6389 /* whoops, we have a non-utf8 pattern, whilst run-time code
6390 * got compiled as utf8. Try again with a utf8 pattern */
6391 S_pat_upgrade_to_utf8(aTHX_ pRExC_state, &exp, &plen,
6392 pRExC_state->num_code_blocks);
6393 goto redo_first_pass;
6396 assert(!pRExC_state->runtime_code_qr);
6402 RExC_in_lookbehind = 0;
6403 RExC_seen_zerolen = *exp == '^' ? -1 : 0;
6405 RExC_override_recoding = 0;
6406 RExC_in_multi_char_class = 0;
6408 /* First pass: determine size, legality. */
6411 RExC_end = exp + plen;
6416 RExC_emit = (regnode *) &RExC_emit_dummy;
6417 RExC_whilem_seen = 0;
6418 RExC_open_parens = NULL;
6419 RExC_close_parens = NULL;
6421 RExC_paren_names = NULL;
6423 RExC_paren_name_list = NULL;
6425 RExC_recurse = NULL;
6426 RExC_study_chunk_recursed = NULL;
6427 RExC_study_chunk_recursed_bytes= 0;
6428 RExC_recurse_count = 0;
6429 pRExC_state->code_index = 0;
6431 #if 0 /* REGC() is (currently) a NOP at the first pass.
6432 * Clever compilers notice this and complain. --jhi */
6433 REGC((U8)REG_MAGIC, (char*)RExC_emit);
6436 PerlIO_printf(Perl_debug_log, "Starting first pass (sizing)\n");
6438 RExC_lastparse=NULL;
6440 /* reg may croak on us, not giving us a chance to free
6441 pRExC_state->code_blocks. We cannot SAVEFREEPV it now, as we may
6442 need it to survive as long as the regexp (qr/(?{})/).
6443 We must check that code_blocksv is not already set, because we may
6444 have jumped back to restart the sizing pass. */
6445 if (pRExC_state->code_blocks && !code_blocksv) {
6446 code_blocksv = newSV_type(SVt_PV);
6447 SAVEFREESV(code_blocksv);
6448 SvPV_set(code_blocksv, (char *)pRExC_state->code_blocks);
6449 SvLEN_set(code_blocksv, 1); /*sufficient to make sv_clear free it*/
6451 if (reg(pRExC_state, 0, &flags,1) == NULL) {
6452 /* It's possible to write a regexp in ascii that represents Unicode
6453 codepoints outside of the byte range, such as via \x{100}. If we
6454 detect such a sequence we have to convert the entire pattern to utf8
6455 and then recompile, as our sizing calculation will have been based
6456 on 1 byte == 1 character, but we will need to use utf8 to encode
6457 at least some part of the pattern, and therefore must convert the whole
6460 if (flags & RESTART_UTF8) {
6461 S_pat_upgrade_to_utf8(aTHX_ pRExC_state, &exp, &plen,
6462 pRExC_state->num_code_blocks);
6463 goto redo_first_pass;
6465 Perl_croak(aTHX_ "panic: reg returned NULL to re_op_compile for sizing pass, flags=%#"UVxf"", (UV) flags);
6468 SvLEN_set(code_blocksv,0); /* no you can't have it, sv_clear */
6471 PerlIO_printf(Perl_debug_log,
6472 "Required size %"IVdf" nodes\n"
6473 "Starting second pass (creation)\n",
6476 RExC_lastparse=NULL;
6479 /* The first pass could have found things that force Unicode semantics */
6480 if ((RExC_utf8 || RExC_uni_semantics)
6481 && get_regex_charset(rx_flags) == REGEX_DEPENDS_CHARSET)
6483 set_regex_charset(&rx_flags, REGEX_UNICODE_CHARSET);
6486 /* Small enough for pointer-storage convention?
6487 If extralen==0, this means that we will not need long jumps. */
6488 if (RExC_size >= 0x10000L && RExC_extralen)
6489 RExC_size += RExC_extralen;
6492 if (RExC_whilem_seen > 15)
6493 RExC_whilem_seen = 15;
6495 /* Allocate space and zero-initialize. Note, the two step process
6496 of zeroing when in debug mode, thus anything assigned has to
6497 happen after that */
6498 rx = (REGEXP*) newSV_type(SVt_REGEXP);
6500 Newxc(ri, sizeof(regexp_internal) + (unsigned)RExC_size * sizeof(regnode),
6501 char, regexp_internal);
6502 if ( r == NULL || ri == NULL )
6503 FAIL("Regexp out of space");
6505 /* avoid reading uninitialized memory in DEBUGGING code in study_chunk() */
6506 Zero(ri, sizeof(regexp_internal) + (unsigned)RExC_size * sizeof(regnode),
6509 /* bulk initialize base fields with 0. */
6510 Zero(ri, sizeof(regexp_internal), char);
6513 /* non-zero initialization begins here */
6516 r->extflags = rx_flags;
6517 RXp_COMPFLAGS(r) = orig_rx_flags & RXf_PMf_FLAGCOPYMASK;
6519 if (pm_flags & PMf_IS_QR) {
6520 ri->code_blocks = pRExC_state->code_blocks;
6521 ri->num_code_blocks = pRExC_state->num_code_blocks;
6526 for (n = 0; n < pRExC_state->num_code_blocks; n++)
6527 if (pRExC_state->code_blocks[n].src_regex)
6528 SAVEFREESV(pRExC_state->code_blocks[n].src_regex);
6529 SAVEFREEPV(pRExC_state->code_blocks);
6533 bool has_p = ((r->extflags & RXf_PMf_KEEPCOPY) == RXf_PMf_KEEPCOPY);
6534 bool has_charset = (get_regex_charset(r->extflags)
6535 != REGEX_DEPENDS_CHARSET);
6537 /* The caret is output if there are any defaults: if not all the STD
6538 * flags are set, or if no character set specifier is needed */
6540 (((r->extflags & RXf_PMf_STD_PMMOD) != RXf_PMf_STD_PMMOD)
6542 bool has_runon = ((RExC_seen & REG_RUN_ON_COMMENT_SEEN)
6543 == REG_RUN_ON_COMMENT_SEEN);
6544 U16 reganch = (U16)((r->extflags & RXf_PMf_STD_PMMOD)
6545 >> RXf_PMf_STD_PMMOD_SHIFT);
6546 const char *fptr = STD_PAT_MODS; /*"msix"*/
6548 /* Allocate for the worst case, which is all the std flags are turned
6549 * on. If more precision is desired, we could do a population count of
6550 * the flags set. This could be done with a small lookup table, or by
6551 * shifting, masking and adding, or even, when available, assembly
6552 * language for a machine-language population count.
6553 * We never output a minus, as all those are defaults, so are
6554 * covered by the caret */
6555 const STRLEN wraplen = plen + has_p + has_runon
6556 + has_default /* If needs a caret */
6558 /* If needs a character set specifier */
6559 + ((has_charset) ? MAX_CHARSET_NAME_LENGTH : 0)
6560 + (sizeof(STD_PAT_MODS) - 1)
6561 + (sizeof("(?:)") - 1);
6563 Newx(p, wraplen + 1, char); /* +1 for the ending NUL */
6564 r->xpv_len_u.xpvlenu_pv = p;
6566 SvFLAGS(rx) |= SVf_UTF8;
6569 /* If a default, cover it using the caret */
6571 *p++= DEFAULT_PAT_MOD;
6575 const char* const name = get_regex_charset_name(r->extflags, &len);
6576 Copy(name, p, len, char);
6580 *p++ = KEEPCOPY_PAT_MOD; /*'p'*/
6583 while((ch = *fptr++)) {
6591 Copy(RExC_precomp, p, plen, char);
6592 assert ((RX_WRAPPED(rx) - p) < 16);
6593 r->pre_prefix = p - RX_WRAPPED(rx);
6599 SvCUR_set(rx, p - RX_WRAPPED(rx));
6603 r->nparens = RExC_npar - 1; /* set early to validate backrefs */
6605 /* setup various meta data about recursion, this all requires
6606 * RExC_npar to be correctly set, and a bit later on we clear it */
6607 if (RExC_seen & REG_RECURSE_SEEN) {
6608 Newxz(RExC_open_parens, RExC_npar,regnode *);
6609 SAVEFREEPV(RExC_open_parens);
6610 Newxz(RExC_close_parens,RExC_npar,regnode *);
6611 SAVEFREEPV(RExC_close_parens);
6613 if (RExC_seen & (REG_RECURSE_SEEN | REG_GOSTART_SEEN)) {
6614 /* Note, RExC_npar is 1 + the number of parens in a pattern.
6615 * So its 1 if there are no parens. */
6616 RExC_study_chunk_recursed_bytes= (RExC_npar >> 3) +
6617 ((RExC_npar & 0x07) != 0);
6618 Newx(RExC_study_chunk_recursed,
6619 RExC_study_chunk_recursed_bytes * RExC_npar, U8);
6620 SAVEFREEPV(RExC_study_chunk_recursed);
6623 /* Useful during FAIL. */
6624 #ifdef RE_TRACK_PATTERN_OFFSETS
6625 Newxz(ri->u.offsets, 2*RExC_size+1, U32); /* MJD 20001228 */
6626 DEBUG_OFFSETS_r(PerlIO_printf(Perl_debug_log,
6627 "%s %"UVuf" bytes for offset annotations.\n",
6628 ri->u.offsets ? "Got" : "Couldn't get",
6629 (UV)((2*RExC_size+1) * sizeof(U32))));
6631 SetProgLen(ri,RExC_size);
6635 REH_CALL_COMP_BEGIN_HOOK(pRExC_state->rx);
6637 /* Second pass: emit code. */
6638 RExC_flags = rx_flags; /* don't let top level (?i) bleed */
6639 RExC_pm_flags = pm_flags;
6641 RExC_end = exp + plen;
6644 RExC_emit_start = ri->program;
6645 RExC_emit = ri->program;
6646 RExC_emit_bound = ri->program + RExC_size + 1;
6647 pRExC_state->code_index = 0;
6649 REGC((U8)REG_MAGIC, (char*) RExC_emit++);
6650 if (reg(pRExC_state, 0, &flags,1) == NULL) {
6652 Perl_croak(aTHX_ "panic: reg returned NULL to re_op_compile for generation pass, flags=%#"UVxf"", (UV) flags);
6654 /* XXXX To minimize changes to RE engine we always allocate
6655 3-units-long substrs field. */
6656 Newx(r->substrs, 1, struct reg_substr_data);
6657 if (RExC_recurse_count) {
6658 Newxz(RExC_recurse,RExC_recurse_count,regnode *);
6659 SAVEFREEPV(RExC_recurse);
6663 r->minlen = minlen = sawlookahead = sawplus = sawopen = sawminmod = 0;
6664 Zero(r->substrs, 1, struct reg_substr_data);
6665 if (RExC_study_chunk_recursed)
6666 Zero(RExC_study_chunk_recursed,
6667 RExC_study_chunk_recursed_bytes * RExC_npar, U8);
6669 #ifdef TRIE_STUDY_OPT
6671 StructCopy(&zero_scan_data, &data, scan_data_t);
6672 copyRExC_state = RExC_state;
6675 DEBUG_OPTIMISE_r(PerlIO_printf(Perl_debug_log,"Restudying\n"));
6677 RExC_state = copyRExC_state;
6678 if (seen & REG_TOP_LEVEL_BRANCHES_SEEN)
6679 RExC_seen |= REG_TOP_LEVEL_BRANCHES_SEEN;
6681 RExC_seen &= ~REG_TOP_LEVEL_BRANCHES_SEEN;
6682 StructCopy(&zero_scan_data, &data, scan_data_t);
6685 StructCopy(&zero_scan_data, &data, scan_data_t);
6688 /* Dig out information for optimizations. */
6689 r->extflags = RExC_flags; /* was pm_op */
6690 /*dmq: removed as part of de-PMOP: pm->op_pmflags = RExC_flags; */
6693 SvUTF8_on(rx); /* Unicode in it? */
6694 ri->regstclass = NULL;
6695 if (RExC_naughty >= 10) /* Probably an expensive pattern. */
6696 r->intflags |= PREGf_NAUGHTY;
6697 scan = ri->program + 1; /* First BRANCH. */
6699 /* testing for BRANCH here tells us whether there is "must appear"
6700 data in the pattern. If there is then we can use it for optimisations */
6701 if (!(RExC_seen & REG_TOP_LEVEL_BRANCHES_SEEN)) { /* Only one top-level choice.
6704 STRLEN longest_float_length, longest_fixed_length;
6705 regnode_ssc ch_class; /* pointed to by data */
6707 SSize_t last_close = 0; /* pointed to by data */
6708 regnode *first= scan;
6709 regnode *first_next= regnext(first);
6711 * Skip introductions and multiplicators >= 1
6712 * so that we can extract the 'meat' of the pattern that must
6713 * match in the large if() sequence following.
6714 * NOTE that EXACT is NOT covered here, as it is normally
6715 * picked up by the optimiser separately.
6717 * This is unfortunate as the optimiser isnt handling lookahead
6718 * properly currently.
6721 while ((OP(first) == OPEN && (sawopen = 1)) ||
6722 /* An OR of *one* alternative - should not happen now. */
6723 (OP(first) == BRANCH && OP(first_next) != BRANCH) ||
6724 /* for now we can't handle lookbehind IFMATCH*/
6725 (OP(first) == IFMATCH && !first->flags && (sawlookahead = 1)) ||
6726 (OP(first) == PLUS) ||
6727 (OP(first) == MINMOD) ||
6728 /* An {n,m} with n>0 */
6729 (PL_regkind[OP(first)] == CURLY && ARG1(first) > 0) ||
6730 (OP(first) == NOTHING && PL_regkind[OP(first_next)] != END ))
6733 * the only op that could be a regnode is PLUS, all the rest
6734 * will be regnode_1 or regnode_2.
6736 * (yves doesn't think this is true)
6738 if (OP(first) == PLUS)
6741 if (OP(first) == MINMOD)
6743 first += regarglen[OP(first)];
6745 first = NEXTOPER(first);
6746 first_next= regnext(first);
6749 /* Starting-point info. */
6751 DEBUG_PEEP("first:",first,0);
6752 /* Ignore EXACT as we deal with it later. */
6753 if (PL_regkind[OP(first)] == EXACT) {
6754 if (OP(first) == EXACT)
6755 NOOP; /* Empty, get anchored substr later. */
6757 ri->regstclass = first;
6760 else if (PL_regkind[OP(first)] == TRIE &&
6761 ((reg_trie_data *)ri->data->data[ ARG(first) ])->minlen>0)
6764 /* this can happen only on restudy */
6765 if ( OP(first) == TRIE ) {
6766 struct regnode_1 *trieop = (struct regnode_1 *)
6767 PerlMemShared_calloc(1, sizeof(struct regnode_1));
6768 StructCopy(first,trieop,struct regnode_1);
6769 trie_op=(regnode *)trieop;
6771 struct regnode_charclass *trieop = (struct regnode_charclass *)
6772 PerlMemShared_calloc(1, sizeof(struct regnode_charclass));
6773 StructCopy(first,trieop,struct regnode_charclass);
6774 trie_op=(regnode *)trieop;
6777 make_trie_failtable(pRExC_state, (regnode *)first, trie_op, 0);
6778 ri->regstclass = trie_op;
6781 else if (REGNODE_SIMPLE(OP(first)))
6782 ri->regstclass = first;
6783 else if (PL_regkind[OP(first)] == BOUND ||
6784 PL_regkind[OP(first)] == NBOUND)
6785 ri->regstclass = first;
6786 else if (PL_regkind[OP(first)] == BOL) {
6787 r->intflags |= (OP(first) == MBOL
6789 : (OP(first) == SBOL
6792 first = NEXTOPER(first);
6795 else if (OP(first) == GPOS) {
6796 r->intflags |= PREGf_ANCH_GPOS;
6797 first = NEXTOPER(first);
6800 else if ((!sawopen || !RExC_sawback) &&
6801 (OP(first) == STAR &&
6802 PL_regkind[OP(NEXTOPER(first))] == REG_ANY) &&
6803 !(r->intflags & PREGf_ANCH) && !pRExC_state->num_code_blocks)
6805 /* turn .* into ^.* with an implied $*=1 */
6807 (OP(NEXTOPER(first)) == REG_ANY)
6810 r->intflags |= (type | PREGf_IMPLICIT);
6811 first = NEXTOPER(first);
6814 if (sawplus && !sawminmod && !sawlookahead
6815 && (!sawopen || !RExC_sawback)
6816 && !pRExC_state->num_code_blocks) /* May examine pos and $& */
6817 /* x+ must match at the 1st pos of run of x's */
6818 r->intflags |= PREGf_SKIP;
6820 /* Scan is after the zeroth branch, first is atomic matcher. */
6821 #ifdef TRIE_STUDY_OPT
6824 PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n",
6825 (IV)(first - scan + 1))
6829 PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n",
6830 (IV)(first - scan + 1))
6836 * If there's something expensive in the r.e., find the
6837 * longest literal string that must appear and make it the
6838 * regmust. Resolve ties in favor of later strings, since
6839 * the regstart check works with the beginning of the r.e.
6840 * and avoiding duplication strengthens checking. Not a
6841 * strong reason, but sufficient in the absence of others.
6842 * [Now we resolve ties in favor of the earlier string if
6843 * it happens that c_offset_min has been invalidated, since the
6844 * earlier string may buy us something the later one won't.]
6847 data.longest_fixed = newSVpvs("");
6848 data.longest_float = newSVpvs("");
6849 data.last_found = newSVpvs("");
6850 data.longest = &(data.longest_fixed);
6851 ENTER_with_name("study_chunk");
6852 SAVEFREESV(data.longest_fixed);
6853 SAVEFREESV(data.longest_float);
6854 SAVEFREESV(data.last_found);
6856 if (!ri->regstclass) {
6857 ssc_init(pRExC_state, &ch_class);
6858 data.start_class = &ch_class;
6859 stclass_flag = SCF_DO_STCLASS_AND;
6860 } else /* XXXX Check for BOUND? */
6862 data.last_closep = &last_close;
6865 minlen = study_chunk(pRExC_state, &first, &minlen, &fake,
6866 scan + RExC_size, /* Up to end */
6868 SCF_DO_SUBSTR | SCF_WHILEM_VISITED_POS | stclass_flag
6869 | (restudied ? SCF_TRIE_DOING_RESTUDY : 0),
6873 CHECK_RESTUDY_GOTO_butfirst(LEAVE_with_name("study_chunk"));
6876 if ( RExC_npar == 1 && data.longest == &(data.longest_fixed)
6877 && data.last_start_min == 0 && data.last_end > 0
6878 && !RExC_seen_zerolen
6879 && !(RExC_seen & REG_VERBARG_SEEN)
6880 && !(RExC_seen & REG_GPOS_SEEN)
6882 r->extflags |= RXf_CHECK_ALL;
6884 scan_commit(pRExC_state, &data,&minlen,0);
6886 longest_float_length = CHR_SVLEN(data.longest_float);
6888 if (! ((SvCUR(data.longest_fixed) /* ok to leave SvCUR */
6889 && data.offset_fixed == data.offset_float_min
6890 && SvCUR(data.longest_fixed) == SvCUR(data.longest_float)))
6891 && S_setup_longest (aTHX_ pRExC_state,
6895 &(r->float_end_shift),
6896 data.lookbehind_float,
6897 data.offset_float_min,
6899 longest_float_length,
6900 cBOOL(data.flags & SF_FL_BEFORE_EOL),
6901 cBOOL(data.flags & SF_FL_BEFORE_MEOL)))
6903 r->float_min_offset = data.offset_float_min - data.lookbehind_float;
6904 r->float_max_offset = data.offset_float_max;
6905 if (data.offset_float_max < SSize_t_MAX) /* Don't offset infinity */
6906 r->float_max_offset -= data.lookbehind_float;
6907 SvREFCNT_inc_simple_void_NN(data.longest_float);
6910 r->float_substr = r->float_utf8 = NULL;
6911 longest_float_length = 0;
6914 longest_fixed_length = CHR_SVLEN(data.longest_fixed);
6916 if (S_setup_longest (aTHX_ pRExC_state,
6918 &(r->anchored_utf8),
6919 &(r->anchored_substr),
6920 &(r->anchored_end_shift),
6921 data.lookbehind_fixed,
6924 longest_fixed_length,
6925 cBOOL(data.flags & SF_FIX_BEFORE_EOL),
6926 cBOOL(data.flags & SF_FIX_BEFORE_MEOL)))
6928 r->anchored_offset = data.offset_fixed - data.lookbehind_fixed;
6929 SvREFCNT_inc_simple_void_NN(data.longest_fixed);
6932 r->anchored_substr = r->anchored_utf8 = NULL;
6933 longest_fixed_length = 0;
6935 LEAVE_with_name("study_chunk");
6938 && (OP(ri->regstclass) == REG_ANY || OP(ri->regstclass) == SANY))
6939 ri->regstclass = NULL;
6941 if ((!(r->anchored_substr || r->anchored_utf8) || r->anchored_offset)
6943 && ! (ANYOF_FLAGS(data.start_class) & ANYOF_EMPTY_STRING)
6944 && !ssc_is_anything(data.start_class))
6946 const U32 n = add_data(pRExC_state, STR_WITH_LEN("f"));
6948 ssc_finalize(pRExC_state, data.start_class);
6950 Newx(RExC_rxi->data->data[n], 1, regnode_ssc);
6951 StructCopy(data.start_class,
6952 (regnode_ssc*)RExC_rxi->data->data[n],
6954 ri->regstclass = (regnode*)RExC_rxi->data->data[n];
6955 r->intflags &= ~PREGf_SKIP; /* Used in find_byclass(). */
6956 DEBUG_COMPILE_r({ SV *sv = sv_newmortal();
6957 regprop(r, sv, (regnode*)data.start_class, NULL);
6958 PerlIO_printf(Perl_debug_log,
6959 "synthetic stclass \"%s\".\n",
6960 SvPVX_const(sv));});
6961 data.start_class = NULL;
6964 /* A temporary algorithm prefers floated substr to fixed one to dig
6966 if (longest_fixed_length > longest_float_length) {
6967 r->substrs->check_ix = 0;
6968 r->check_end_shift = r->anchored_end_shift;
6969 r->check_substr = r->anchored_substr;
6970 r->check_utf8 = r->anchored_utf8;
6971 r->check_offset_min = r->check_offset_max = r->anchored_offset;
6972 if (r->intflags & (PREGf_ANCH_SBOL|PREGf_ANCH_GPOS))
6973 r->intflags |= PREGf_NOSCAN;
6976 r->substrs->check_ix = 1;
6977 r->check_end_shift = r->float_end_shift;
6978 r->check_substr = r->float_substr;
6979 r->check_utf8 = r->float_utf8;
6980 r->check_offset_min = r->float_min_offset;
6981 r->check_offset_max = r->float_max_offset;
6983 if ((r->check_substr || r->check_utf8) ) {
6984 r->extflags |= RXf_USE_INTUIT;
6985 if (SvTAIL(r->check_substr ? r->check_substr : r->check_utf8))
6986 r->extflags |= RXf_INTUIT_TAIL;
6988 r->substrs->data[0].max_offset = r->substrs->data[0].min_offset;
6990 /* XXX Unneeded? dmq (shouldn't as this is handled elsewhere)
6991 if ( (STRLEN)minlen < longest_float_length )
6992 minlen= longest_float_length;
6993 if ( (STRLEN)minlen < longest_fixed_length )
6994 minlen= longest_fixed_length;
6998 /* Several toplevels. Best we can is to set minlen. */
7000 regnode_ssc ch_class;
7001 SSize_t last_close = 0;
7003 DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log, "\nMulti Top Level\n"));
7005 scan = ri->program + 1;
7006 ssc_init(pRExC_state, &ch_class);
7007 data.start_class = &ch_class;
7008 data.last_closep = &last_close;
7011 minlen = study_chunk(pRExC_state,
7012 &scan, &minlen, &fake, scan + RExC_size, &data, -1, 0, NULL,
7013 SCF_DO_STCLASS_AND|SCF_WHILEM_VISITED_POS|(restudied
7014 ? SCF_TRIE_DOING_RESTUDY
7018 CHECK_RESTUDY_GOTO_butfirst(NOOP);
7020 r->check_substr = r->check_utf8 = r->anchored_substr = r->anchored_utf8
7021 = r->float_substr = r->float_utf8 = NULL;
7023 if (! (ANYOF_FLAGS(data.start_class) & ANYOF_EMPTY_STRING)
7024 && ! ssc_is_anything(data.start_class))
7026 const U32 n = add_data(pRExC_state, STR_WITH_LEN("f"));
7028 ssc_finalize(pRExC_state, data.start_class);
7030 Newx(RExC_rxi->data->data[n], 1, regnode_ssc);
7031 StructCopy(data.start_class,
7032 (regnode_ssc*)RExC_rxi->data->data[n],
7034 ri->regstclass = (regnode*)RExC_rxi->data->data[n];
7035 r->intflags &= ~PREGf_SKIP; /* Used in find_byclass(). */
7036 DEBUG_COMPILE_r({ SV* sv = sv_newmortal();
7037 regprop(r, sv, (regnode*)data.start_class, NULL);
7038 PerlIO_printf(Perl_debug_log,
7039 "synthetic stclass \"%s\".\n",
7040 SvPVX_const(sv));});
7041 data.start_class = NULL;
7045 if (RExC_seen & REG_UNBOUNDED_QUANTIFIER_SEEN) {
7046 r->extflags |= RXf_UNBOUNDED_QUANTIFIER_SEEN;
7047 r->maxlen = REG_INFTY;
7050 r->maxlen = RExC_maxlen;
7053 /* Guard against an embedded (?=) or (?<=) with a longer minlen than
7054 the "real" pattern. */
7056 PerlIO_printf(Perl_debug_log,"minlen: %"IVdf" r->minlen:%"IVdf" maxlen:%ld\n",
7057 (IV)minlen, (IV)r->minlen, RExC_maxlen);
7059 r->minlenret = minlen;
7060 if (r->minlen < minlen)
7063 if (RExC_seen & REG_GPOS_SEEN)
7064 r->intflags |= PREGf_GPOS_SEEN;
7065 if (RExC_seen & REG_LOOKBEHIND_SEEN)
7066 r->extflags |= RXf_NO_INPLACE_SUBST; /* inplace might break the
7068 if (pRExC_state->num_code_blocks)
7069 r->extflags |= RXf_EVAL_SEEN;
7070 if (RExC_seen & REG_CANY_SEEN)
7071 r->intflags |= PREGf_CANY_SEEN;
7072 if (RExC_seen & REG_VERBARG_SEEN)
7074 r->intflags |= PREGf_VERBARG_SEEN;
7075 r->extflags |= RXf_NO_INPLACE_SUBST; /* don't understand this! Yves */
7077 if (RExC_seen & REG_CUTGROUP_SEEN)
7078 r->intflags |= PREGf_CUTGROUP_SEEN;
7079 if (pm_flags & PMf_USE_RE_EVAL)
7080 r->intflags |= PREGf_USE_RE_EVAL;
7081 if (RExC_paren_names)
7082 RXp_PAREN_NAMES(r) = MUTABLE_HV(SvREFCNT_inc(RExC_paren_names));
7084 RXp_PAREN_NAMES(r) = NULL;
7086 /* If we have seen an anchor in our pattern then we set the extflag RXf_IS_ANCHORED
7087 * so it can be used in pp.c */
7088 if (r->intflags & PREGf_ANCH)
7089 r->extflags |= RXf_IS_ANCHORED;
7093 /* this is used to identify "special" patterns that might result
7094 * in Perl NOT calling the regex engine and instead doing the match "itself",
7095 * particularly special cases in split//. By having the regex compiler
7096 * do this pattern matching at a regop level (instead of by inspecting the pattern)
7097 * we avoid weird issues with equivalent patterns resulting in different behavior,
7098 * AND we allow non Perl engines to get the same optimizations by the setting the
7099 * flags appropriately - Yves */
7100 regnode *first = ri->program + 1;
7102 regnode *next = NEXTOPER(first);
7105 if (PL_regkind[fop] == NOTHING && nop == END)
7106 r->extflags |= RXf_NULL;
7107 else if (PL_regkind[fop] == BOL && nop == END)
7108 r->extflags |= RXf_START_ONLY;
7109 else if (fop == PLUS
7110 && PL_regkind[nop] == POSIXD && FLAGS(next) == _CC_SPACE
7111 && OP(regnext(first)) == END)
7112 r->extflags |= RXf_WHITE;
7113 else if ( r->extflags & RXf_SPLIT
7115 && STR_LEN(first) == 1
7116 && *(STRING(first)) == ' '
7117 && OP(regnext(first)) == END )
7118 r->extflags |= (RXf_SKIPWHITE|RXf_WHITE);
7122 if (RExC_contains_locale) {
7123 RXp_EXTFLAGS(r) |= RXf_TAINTED;
7127 if (RExC_paren_names) {
7128 ri->name_list_idx = add_data( pRExC_state, STR_WITH_LEN("a"));
7129 ri->data->data[ri->name_list_idx]
7130 = (void*)SvREFCNT_inc(RExC_paren_name_list);
7133 ri->name_list_idx = 0;
7135 if (RExC_recurse_count) {
7136 for ( ; RExC_recurse_count ; RExC_recurse_count-- ) {
7137 const regnode *scan = RExC_recurse[RExC_recurse_count-1];
7138 ARG2L_SET( scan, RExC_open_parens[ARG(scan)-1] - scan );
7141 Newxz(r->offs, RExC_npar, regexp_paren_pair);
7142 /* assume we don't need to swap parens around before we match */
7146 PerlIO_printf(Perl_debug_log,"Final program:\n");
7149 #ifdef RE_TRACK_PATTERN_OFFSETS
7150 DEBUG_OFFSETS_r(if (ri->u.offsets) {
7151 const STRLEN len = ri->u.offsets[0];
7153 GET_RE_DEBUG_FLAGS_DECL;
7154 PerlIO_printf(Perl_debug_log,
7155 "Offsets: [%"UVuf"]\n\t", (UV)ri->u.offsets[0]);
7156 for (i = 1; i <= len; i++) {
7157 if (ri->u.offsets[i*2-1] || ri->u.offsets[i*2])
7158 PerlIO_printf(Perl_debug_log, "%"UVuf":%"UVuf"[%"UVuf"] ",
7159 (UV)i, (UV)ri->u.offsets[i*2-1], (UV)ri->u.offsets[i*2]);
7161 PerlIO_printf(Perl_debug_log, "\n");
7166 /* under ithreads the ?pat? PMf_USED flag on the pmop is simulated
7167 * by setting the regexp SV to readonly-only instead. If the
7168 * pattern's been recompiled, the USEDness should remain. */
7169 if (old_re && SvREADONLY(old_re))
7177 Perl_reg_named_buff(pTHX_ REGEXP * const rx, SV * const key, SV * const value,
7180 PERL_ARGS_ASSERT_REG_NAMED_BUFF;
7182 PERL_UNUSED_ARG(value);
7184 if (flags & RXapif_FETCH) {
7185 return reg_named_buff_fetch(rx, key, flags);
7186 } else if (flags & (RXapif_STORE | RXapif_DELETE | RXapif_CLEAR)) {
7187 Perl_croak_no_modify();
7189 } else if (flags & RXapif_EXISTS) {
7190 return reg_named_buff_exists(rx, key, flags)
7193 } else if (flags & RXapif_REGNAMES) {
7194 return reg_named_buff_all(rx, flags);
7195 } else if (flags & (RXapif_SCALAR | RXapif_REGNAMES_COUNT)) {
7196 return reg_named_buff_scalar(rx, flags);
7198 Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff", (int)flags);
7204 Perl_reg_named_buff_iter(pTHX_ REGEXP * const rx, const SV * const lastkey,
7207 PERL_ARGS_ASSERT_REG_NAMED_BUFF_ITER;
7208 PERL_UNUSED_ARG(lastkey);
7210 if (flags & RXapif_FIRSTKEY)
7211 return reg_named_buff_firstkey(rx, flags);
7212 else if (flags & RXapif_NEXTKEY)
7213 return reg_named_buff_nextkey(rx, flags);
7215 Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff_iter",
7222 Perl_reg_named_buff_fetch(pTHX_ REGEXP * const r, SV * const namesv,
7225 AV *retarray = NULL;
7227 struct regexp *const rx = ReANY(r);
7229 PERL_ARGS_ASSERT_REG_NAMED_BUFF_FETCH;
7231 if (flags & RXapif_ALL)
7234 if (rx && RXp_PAREN_NAMES(rx)) {
7235 HE *he_str = hv_fetch_ent( RXp_PAREN_NAMES(rx), namesv, 0, 0 );
7238 SV* sv_dat=HeVAL(he_str);
7239 I32 *nums=(I32*)SvPVX(sv_dat);
7240 for ( i=0; i<SvIVX(sv_dat); i++ ) {
7241 if ((I32)(rx->nparens) >= nums[i]
7242 && rx->offs[nums[i]].start != -1
7243 && rx->offs[nums[i]].end != -1)
7246 CALLREG_NUMBUF_FETCH(r,nums[i],ret);
7251 ret = newSVsv(&PL_sv_undef);
7254 av_push(retarray, ret);
7257 return newRV_noinc(MUTABLE_SV(retarray));
7264 Perl_reg_named_buff_exists(pTHX_ REGEXP * const r, SV * const key,
7267 struct regexp *const rx = ReANY(r);
7269 PERL_ARGS_ASSERT_REG_NAMED_BUFF_EXISTS;
7271 if (rx && RXp_PAREN_NAMES(rx)) {
7272 if (flags & RXapif_ALL) {
7273 return hv_exists_ent(RXp_PAREN_NAMES(rx), key, 0);
7275 SV *sv = CALLREG_NAMED_BUFF_FETCH(r, key, flags);
7277 SvREFCNT_dec_NN(sv);
7289 Perl_reg_named_buff_firstkey(pTHX_ REGEXP * const r, const U32 flags)
7291 struct regexp *const rx = ReANY(r);
7293 PERL_ARGS_ASSERT_REG_NAMED_BUFF_FIRSTKEY;
7295 if ( rx && RXp_PAREN_NAMES(rx) ) {
7296 (void)hv_iterinit(RXp_PAREN_NAMES(rx));
7298 return CALLREG_NAMED_BUFF_NEXTKEY(r, NULL, flags & ~RXapif_FIRSTKEY);
7305 Perl_reg_named_buff_nextkey(pTHX_ REGEXP * const r, const U32 flags)
7307 struct regexp *const rx = ReANY(r);
7308 GET_RE_DEBUG_FLAGS_DECL;
7310 PERL_ARGS_ASSERT_REG_NAMED_BUFF_NEXTKEY;
7312 if (rx && RXp_PAREN_NAMES(rx)) {
7313 HV *hv = RXp_PAREN_NAMES(rx);
7315 while ( (temphe = hv_iternext_flags(hv,0)) ) {
7318 SV* sv_dat = HeVAL(temphe);
7319 I32 *nums = (I32*)SvPVX(sv_dat);
7320 for ( i = 0; i < SvIVX(sv_dat); i++ ) {
7321 if ((I32)(rx->lastparen) >= nums[i] &&
7322 rx->offs[nums[i]].start != -1 &&
7323 rx->offs[nums[i]].end != -1)
7329 if (parno || flags & RXapif_ALL) {
7330 return newSVhek(HeKEY_hek(temphe));
7338 Perl_reg_named_buff_scalar(pTHX_ REGEXP * const r, const U32 flags)
7343 struct regexp *const rx = ReANY(r);
7345 PERL_ARGS_ASSERT_REG_NAMED_BUFF_SCALAR;
7347 if (rx && RXp_PAREN_NAMES(rx)) {
7348 if (flags & (RXapif_ALL | RXapif_REGNAMES_COUNT)) {
7349 return newSViv(HvTOTALKEYS(RXp_PAREN_NAMES(rx)));
7350 } else if (flags & RXapif_ONE) {
7351 ret = CALLREG_NAMED_BUFF_ALL(r, (flags | RXapif_REGNAMES));
7352 av = MUTABLE_AV(SvRV(ret));
7353 length = av_tindex(av);
7354 SvREFCNT_dec_NN(ret);
7355 return newSViv(length + 1);
7357 Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff_scalar",
7362 return &PL_sv_undef;
7366 Perl_reg_named_buff_all(pTHX_ REGEXP * const r, const U32 flags)
7368 struct regexp *const rx = ReANY(r);
7371 PERL_ARGS_ASSERT_REG_NAMED_BUFF_ALL;
7373 if (rx && RXp_PAREN_NAMES(rx)) {
7374 HV *hv= RXp_PAREN_NAMES(rx);
7376 (void)hv_iterinit(hv);
7377 while ( (temphe = hv_iternext_flags(hv,0)) ) {
7380 SV* sv_dat = HeVAL(temphe);
7381 I32 *nums = (I32*)SvPVX(sv_dat);
7382 for ( i = 0; i < SvIVX(sv_dat); i++ ) {
7383 if ((I32)(rx->lastparen) >= nums[i] &&
7384 rx->offs[nums[i]].start != -1 &&
7385 rx->offs[nums[i]].end != -1)
7391 if (parno || flags & RXapif_ALL) {
7392 av_push(av, newSVhek(HeKEY_hek(temphe)));
7397 return newRV_noinc(MUTABLE_SV(av));
7401 Perl_reg_numbered_buff_fetch(pTHX_ REGEXP * const r, const I32 paren,
7404 struct regexp *const rx = ReANY(r);
7410 PERL_ARGS_ASSERT_REG_NUMBERED_BUFF_FETCH;
7412 if ( n == RX_BUFF_IDX_CARET_PREMATCH
7413 || n == RX_BUFF_IDX_CARET_FULLMATCH
7414 || n == RX_BUFF_IDX_CARET_POSTMATCH
7417 bool keepcopy = cBOOL(rx->extflags & RXf_PMf_KEEPCOPY);
7419 /* on something like
7422 * the KEEPCOPY is set on the PMOP rather than the regex */
7423 if (PL_curpm && r == PM_GETRE(PL_curpm))
7424 keepcopy = cBOOL(PL_curpm->op_pmflags & PMf_KEEPCOPY);
7433 if (n == RX_BUFF_IDX_CARET_FULLMATCH)
7434 /* no need to distinguish between them any more */
7435 n = RX_BUFF_IDX_FULLMATCH;
7437 if ((n == RX_BUFF_IDX_PREMATCH || n == RX_BUFF_IDX_CARET_PREMATCH)
7438 && rx->offs[0].start != -1)
7440 /* $`, ${^PREMATCH} */
7441 i = rx->offs[0].start;
7445 if ((n == RX_BUFF_IDX_POSTMATCH || n == RX_BUFF_IDX_CARET_POSTMATCH)
7446 && rx->offs[0].end != -1)
7448 /* $', ${^POSTMATCH} */
7449 s = rx->subbeg - rx->suboffset + rx->offs[0].end;
7450 i = rx->sublen + rx->suboffset - rx->offs[0].end;
7453 if ( 0 <= n && n <= (I32)rx->nparens &&
7454 (s1 = rx->offs[n].start) != -1 &&
7455 (t1 = rx->offs[n].end) != -1)
7457 /* $&, ${^MATCH}, $1 ... */
7459 s = rx->subbeg + s1 - rx->suboffset;
7464 assert(s >= rx->subbeg);
7465 assert((STRLEN)rx->sublen >= (STRLEN)((s - rx->subbeg) + i) );
7467 #ifdef NO_TAINT_SUPPORT
7468 sv_setpvn(sv, s, i);
7470 const int oldtainted = TAINT_get;
7472 sv_setpvn(sv, s, i);
7473 TAINT_set(oldtainted);
7475 if ( (rx->intflags & PREGf_CANY_SEEN)
7476 ? (RXp_MATCH_UTF8(rx)
7477 && (!i || is_utf8_string((U8*)s, i)))
7478 : (RXp_MATCH_UTF8(rx)) )
7485 if (RXp_MATCH_TAINTED(rx)) {
7486 if (SvTYPE(sv) >= SVt_PVMG) {
7487 MAGIC* const mg = SvMAGIC(sv);
7490 SvMAGIC_set(sv, mg->mg_moremagic);
7492 if ((mgt = SvMAGIC(sv))) {
7493 mg->mg_moremagic = mgt;
7494 SvMAGIC_set(sv, mg);
7505 sv_setsv(sv,&PL_sv_undef);
7511 Perl_reg_numbered_buff_store(pTHX_ REGEXP * const rx, const I32 paren,
7512 SV const * const value)
7514 PERL_ARGS_ASSERT_REG_NUMBERED_BUFF_STORE;
7516 PERL_UNUSED_ARG(rx);
7517 PERL_UNUSED_ARG(paren);
7518 PERL_UNUSED_ARG(value);
7521 Perl_croak_no_modify();
7525 Perl_reg_numbered_buff_length(pTHX_ REGEXP * const r, const SV * const sv,
7528 struct regexp *const rx = ReANY(r);
7532 PERL_ARGS_ASSERT_REG_NUMBERED_BUFF_LENGTH;
7534 if ( paren == RX_BUFF_IDX_CARET_PREMATCH
7535 || paren == RX_BUFF_IDX_CARET_FULLMATCH
7536 || paren == RX_BUFF_IDX_CARET_POSTMATCH
7539 bool keepcopy = cBOOL(rx->extflags & RXf_PMf_KEEPCOPY);
7541 /* on something like
7544 * the KEEPCOPY is set on the PMOP rather than the regex */
7545 if (PL_curpm && r == PM_GETRE(PL_curpm))
7546 keepcopy = cBOOL(PL_curpm->op_pmflags & PMf_KEEPCOPY);
7552 /* Some of this code was originally in C<Perl_magic_len> in F<mg.c> */
7554 case RX_BUFF_IDX_CARET_PREMATCH: /* ${^PREMATCH} */
7555 case RX_BUFF_IDX_PREMATCH: /* $` */
7556 if (rx->offs[0].start != -1) {
7557 i = rx->offs[0].start;
7566 case RX_BUFF_IDX_CARET_POSTMATCH: /* ${^POSTMATCH} */
7567 case RX_BUFF_IDX_POSTMATCH: /* $' */
7568 if (rx->offs[0].end != -1) {
7569 i = rx->sublen - rx->offs[0].end;
7571 s1 = rx->offs[0].end;
7578 default: /* $& / ${^MATCH}, $1, $2, ... */
7579 if (paren <= (I32)rx->nparens &&
7580 (s1 = rx->offs[paren].start) != -1 &&
7581 (t1 = rx->offs[paren].end) != -1)
7587 if (ckWARN(WARN_UNINITIALIZED))
7588 report_uninit((const SV *)sv);
7593 if (i > 0 && RXp_MATCH_UTF8(rx)) {
7594 const char * const s = rx->subbeg - rx->suboffset + s1;
7599 if (is_utf8_string_loclen((U8*)s, i, &ep, &el))
7606 Perl_reg_qr_package(pTHX_ REGEXP * const rx)
7608 PERL_ARGS_ASSERT_REG_QR_PACKAGE;
7609 PERL_UNUSED_ARG(rx);
7613 return newSVpvs("Regexp");
7616 /* Scans the name of a named buffer from the pattern.
7617 * If flags is REG_RSN_RETURN_NULL returns null.
7618 * If flags is REG_RSN_RETURN_NAME returns an SV* containing the name
7619 * If flags is REG_RSN_RETURN_DATA returns the data SV* corresponding
7620 * to the parsed name as looked up in the RExC_paren_names hash.
7621 * If there is an error throws a vFAIL().. type exception.
7624 #define REG_RSN_RETURN_NULL 0
7625 #define REG_RSN_RETURN_NAME 1
7626 #define REG_RSN_RETURN_DATA 2
7629 S_reg_scan_name(pTHX_ RExC_state_t *pRExC_state, U32 flags)
7631 char *name_start = RExC_parse;
7633 PERL_ARGS_ASSERT_REG_SCAN_NAME;
7635 assert (RExC_parse <= RExC_end);
7636 if (RExC_parse == RExC_end) NOOP;
7637 else if (isIDFIRST_lazy_if(RExC_parse, UTF)) {
7638 /* skip IDFIRST by using do...while */
7641 RExC_parse += UTF8SKIP(RExC_parse);
7642 } while (isWORDCHAR_utf8((U8*)RExC_parse));
7646 } while (isWORDCHAR(*RExC_parse));
7648 RExC_parse++; /* so the <- from the vFAIL is after the offending
7650 vFAIL("Group name must start with a non-digit word character");
7654 = newSVpvn_flags(name_start, (int)(RExC_parse - name_start),
7655 SVs_TEMP | (UTF ? SVf_UTF8 : 0));
7656 if ( flags == REG_RSN_RETURN_NAME)
7658 else if (flags==REG_RSN_RETURN_DATA) {
7661 if ( ! sv_name ) /* should not happen*/
7662 Perl_croak(aTHX_ "panic: no svname in reg_scan_name");
7663 if (RExC_paren_names)
7664 he_str = hv_fetch_ent( RExC_paren_names, sv_name, 0, 0 );
7666 sv_dat = HeVAL(he_str);
7668 vFAIL("Reference to nonexistent named group");
7672 Perl_croak(aTHX_ "panic: bad flag %lx in reg_scan_name",
7673 (unsigned long) flags);
7675 assert(0); /* NOT REACHED */
7680 #define DEBUG_PARSE_MSG(funcname) DEBUG_PARSE_r({ \
7681 int rem=(int)(RExC_end - RExC_parse); \
7690 if (RExC_lastparse!=RExC_parse) \
7691 PerlIO_printf(Perl_debug_log," >%.*s%-*s", \
7694 iscut ? "..." : "<" \
7697 PerlIO_printf(Perl_debug_log,"%16s",""); \
7700 num = RExC_size + 1; \
7702 num=REG_NODE_NUM(RExC_emit); \
7703 if (RExC_lastnum!=num) \
7704 PerlIO_printf(Perl_debug_log,"|%4d",num); \
7706 PerlIO_printf(Perl_debug_log,"|%4s",""); \
7707 PerlIO_printf(Perl_debug_log,"|%*s%-4s", \
7708 (int)((depth*2)), "", \
7712 RExC_lastparse=RExC_parse; \
7717 #define DEBUG_PARSE(funcname) DEBUG_PARSE_r({ \
7718 DEBUG_PARSE_MSG((funcname)); \
7719 PerlIO_printf(Perl_debug_log,"%4s","\n"); \
7721 #define DEBUG_PARSE_FMT(funcname,fmt,args) DEBUG_PARSE_r({ \
7722 DEBUG_PARSE_MSG((funcname)); \
7723 PerlIO_printf(Perl_debug_log,fmt "\n",args); \
7726 /* This section of code defines the inversion list object and its methods. The
7727 * interfaces are highly subject to change, so as much as possible is static to
7728 * this file. An inversion list is here implemented as a malloc'd C UV array
7729 * as an SVt_INVLIST scalar.
7731 * An inversion list for Unicode is an array of code points, sorted by ordinal
7732 * number. The zeroth element is the first code point in the list. The 1th
7733 * element is the first element beyond that not in the list. In other words,
7734 * the first range is
7735 * invlist[0]..(invlist[1]-1)
7736 * The other ranges follow. Thus every element whose index is divisible by two
7737 * marks the beginning of a range that is in the list, and every element not
7738 * divisible by two marks the beginning of a range not in the list. A single
7739 * element inversion list that contains the single code point N generally
7740 * consists of two elements
7743 * (The exception is when N is the highest representable value on the
7744 * machine, in which case the list containing just it would be a single
7745 * element, itself. By extension, if the last range in the list extends to
7746 * infinity, then the first element of that range will be in the inversion list
7747 * at a position that is divisible by two, and is the final element in the
7749 * Taking the complement (inverting) an inversion list is quite simple, if the
7750 * first element is 0, remove it; otherwise add a 0 element at the beginning.
7751 * This implementation reserves an element at the beginning of each inversion
7752 * list to always contain 0; there is an additional flag in the header which
7753 * indicates if the list begins at the 0, or is offset to begin at the next
7756 * More about inversion lists can be found in "Unicode Demystified"
7757 * Chapter 13 by Richard Gillam, published by Addison-Wesley.
7758 * More will be coming when functionality is added later.
7760 * The inversion list data structure is currently implemented as an SV pointing
7761 * to an array of UVs that the SV thinks are bytes. This allows us to have an
7762 * array of UV whose memory management is automatically handled by the existing
7763 * facilities for SV's.
7765 * Some of the methods should always be private to the implementation, and some
7766 * should eventually be made public */
7768 /* The header definitions are in F<inline_invlist.c> */
7770 PERL_STATIC_INLINE UV*
7771 S__invlist_array_init(pTHX_ SV* const invlist, const bool will_have_0)
7773 /* Returns a pointer to the first element in the inversion list's array.
7774 * This is called upon initialization of an inversion list. Where the
7775 * array begins depends on whether the list has the code point U+0000 in it
7776 * or not. The other parameter tells it whether the code that follows this
7777 * call is about to put a 0 in the inversion list or not. The first
7778 * element is either the element reserved for 0, if TRUE, or the element
7779 * after it, if FALSE */
7781 bool* offset = get_invlist_offset_addr(invlist);
7782 UV* zero_addr = (UV *) SvPVX(invlist);
7784 PERL_ARGS_ASSERT__INVLIST_ARRAY_INIT;
7787 assert(! _invlist_len(invlist));
7791 /* 1^1 = 0; 1^0 = 1 */
7792 *offset = 1 ^ will_have_0;
7793 return zero_addr + *offset;
7796 PERL_STATIC_INLINE UV*
7797 S_invlist_array(pTHX_ SV* const invlist)
7799 /* Returns the pointer to the inversion list's array. Every time the
7800 * length changes, this needs to be called in case malloc or realloc moved
7803 PERL_ARGS_ASSERT_INVLIST_ARRAY;
7805 /* Must not be empty. If these fail, you probably didn't check for <len>
7806 * being non-zero before trying to get the array */
7807 assert(_invlist_len(invlist));
7809 /* The very first element always contains zero, The array begins either
7810 * there, or if the inversion list is offset, at the element after it.
7811 * The offset header field determines which; it contains 0 or 1 to indicate
7812 * how much additionally to add */
7813 assert(0 == *(SvPVX(invlist)));
7814 return ((UV *) SvPVX(invlist) + *get_invlist_offset_addr(invlist));
7817 PERL_STATIC_INLINE void
7818 S_invlist_set_len(pTHX_ SV* const invlist, const UV len, const bool offset)
7820 /* Sets the current number of elements stored in the inversion list.
7821 * Updates SvCUR correspondingly */
7823 PERL_ARGS_ASSERT_INVLIST_SET_LEN;
7825 assert(SvTYPE(invlist) == SVt_INVLIST);
7830 : TO_INTERNAL_SIZE(len + offset));
7831 assert(SvLEN(invlist) == 0 || SvCUR(invlist) <= SvLEN(invlist));
7834 PERL_STATIC_INLINE IV*
7835 S_get_invlist_previous_index_addr(pTHX_ SV* invlist)
7837 /* Return the address of the IV that is reserved to hold the cached index
7840 PERL_ARGS_ASSERT_GET_INVLIST_PREVIOUS_INDEX_ADDR;
7842 assert(SvTYPE(invlist) == SVt_INVLIST);
7844 return &(((XINVLIST*) SvANY(invlist))->prev_index);
7847 PERL_STATIC_INLINE IV
7848 S_invlist_previous_index(pTHX_ SV* const invlist)
7850 /* Returns cached index of previous search */
7852 PERL_ARGS_ASSERT_INVLIST_PREVIOUS_INDEX;
7854 return *get_invlist_previous_index_addr(invlist);
7857 PERL_STATIC_INLINE void
7858 S_invlist_set_previous_index(pTHX_ SV* const invlist, const IV index)
7860 /* Caches <index> for later retrieval */
7862 PERL_ARGS_ASSERT_INVLIST_SET_PREVIOUS_INDEX;
7864 assert(index == 0 || index < (int) _invlist_len(invlist));
7866 *get_invlist_previous_index_addr(invlist) = index;
7869 PERL_STATIC_INLINE UV
7870 S_invlist_max(pTHX_ SV* const invlist)
7872 /* Returns the maximum number of elements storable in the inversion list's
7873 * array, without having to realloc() */
7875 PERL_ARGS_ASSERT_INVLIST_MAX;
7877 assert(SvTYPE(invlist) == SVt_INVLIST);
7879 /* Assumes worst case, in which the 0 element is not counted in the
7880 * inversion list, so subtracts 1 for that */
7881 return SvLEN(invlist) == 0 /* This happens under _new_invlist_C_array */
7882 ? FROM_INTERNAL_SIZE(SvCUR(invlist)) - 1
7883 : FROM_INTERNAL_SIZE(SvLEN(invlist)) - 1;
7886 #ifndef PERL_IN_XSUB_RE
7888 Perl__new_invlist(pTHX_ IV initial_size)
7891 /* Return a pointer to a newly constructed inversion list, with enough
7892 * space to store 'initial_size' elements. If that number is negative, a
7893 * system default is used instead */
7897 if (initial_size < 0) {
7901 /* Allocate the initial space */
7902 new_list = newSV_type(SVt_INVLIST);
7904 /* First 1 is in case the zero element isn't in the list; second 1 is for
7906 SvGROW(new_list, TO_INTERNAL_SIZE(initial_size + 1) + 1);
7907 invlist_set_len(new_list, 0, 0);
7909 /* Force iterinit() to be used to get iteration to work */
7910 *get_invlist_iter_addr(new_list) = (STRLEN) UV_MAX;
7912 *get_invlist_previous_index_addr(new_list) = 0;
7918 Perl__new_invlist_C_array(pTHX_ const UV* const list)
7920 /* Return a pointer to a newly constructed inversion list, initialized to
7921 * point to <list>, which has to be in the exact correct inversion list
7922 * form, including internal fields. Thus this is a dangerous routine that
7923 * should not be used in the wrong hands. The passed in 'list' contains
7924 * several header fields at the beginning that are not part of the
7925 * inversion list body proper */
7927 const STRLEN length = (STRLEN) list[0];
7928 const UV version_id = list[1];
7929 const bool offset = cBOOL(list[2]);
7930 #define HEADER_LENGTH 3
7931 /* If any of the above changes in any way, you must change HEADER_LENGTH
7932 * (if appropriate) and regenerate INVLIST_VERSION_ID by running
7933 * perl -E 'say int(rand 2**31-1)'
7935 #define INVLIST_VERSION_ID 148565664 /* This is a combination of a version and
7936 data structure type, so that one being
7937 passed in can be validated to be an
7938 inversion list of the correct vintage.
7941 SV* invlist = newSV_type(SVt_INVLIST);
7943 PERL_ARGS_ASSERT__NEW_INVLIST_C_ARRAY;
7945 if (version_id != INVLIST_VERSION_ID) {
7946 Perl_croak(aTHX_ "panic: Incorrect version for previously generated inversion list");
7949 /* The generated array passed in includes header elements that aren't part
7950 * of the list proper, so start it just after them */
7951 SvPV_set(invlist, (char *) (list + HEADER_LENGTH));
7953 SvLEN_set(invlist, 0); /* Means we own the contents, and the system
7954 shouldn't touch it */
7956 *(get_invlist_offset_addr(invlist)) = offset;
7958 /* The 'length' passed to us is the physical number of elements in the
7959 * inversion list. But if there is an offset the logical number is one
7961 invlist_set_len(invlist, length - offset, offset);
7963 invlist_set_previous_index(invlist, 0);
7965 /* Initialize the iteration pointer. */
7966 invlist_iterfinish(invlist);
7968 SvREADONLY_on(invlist);
7972 #endif /* ifndef PERL_IN_XSUB_RE */
7975 S_invlist_extend(pTHX_ SV* const invlist, const UV new_max)
7977 /* Grow the maximum size of an inversion list */
7979 PERL_ARGS_ASSERT_INVLIST_EXTEND;
7981 assert(SvTYPE(invlist) == SVt_INVLIST);
7983 /* Add one to account for the zero element at the beginning which may not
7984 * be counted by the calling parameters */
7985 SvGROW((SV *)invlist, TO_INTERNAL_SIZE(new_max + 1));
7988 PERL_STATIC_INLINE void
7989 S_invlist_trim(pTHX_ SV* const invlist)
7991 PERL_ARGS_ASSERT_INVLIST_TRIM;
7993 assert(SvTYPE(invlist) == SVt_INVLIST);
7995 /* Change the length of the inversion list to how many entries it currently
7997 SvPV_shrink_to_cur((SV *) invlist);
8001 S__append_range_to_invlist(pTHX_ SV* const invlist,
8002 const UV start, const UV end)
8004 /* Subject to change or removal. Append the range from 'start' to 'end' at
8005 * the end of the inversion list. The range must be above any existing
8009 UV max = invlist_max(invlist);
8010 UV len = _invlist_len(invlist);
8013 PERL_ARGS_ASSERT__APPEND_RANGE_TO_INVLIST;
8015 if (len == 0) { /* Empty lists must be initialized */
8016 offset = start != 0;
8017 array = _invlist_array_init(invlist, ! offset);
8020 /* Here, the existing list is non-empty. The current max entry in the
8021 * list is generally the first value not in the set, except when the
8022 * set extends to the end of permissible values, in which case it is
8023 * the first entry in that final set, and so this call is an attempt to
8024 * append out-of-order */
8026 UV final_element = len - 1;
8027 array = invlist_array(invlist);
8028 if (array[final_element] > start
8029 || ELEMENT_RANGE_MATCHES_INVLIST(final_element))
8031 Perl_croak(aTHX_ "panic: attempting to append to an inversion list, but wasn't at the end of the list, final=%"UVuf", start=%"UVuf", match=%c",
8032 array[final_element], start,
8033 ELEMENT_RANGE_MATCHES_INVLIST(final_element) ? 't' : 'f');
8036 /* Here, it is a legal append. If the new range begins with the first
8037 * value not in the set, it is extending the set, so the new first
8038 * value not in the set is one greater than the newly extended range.
8040 offset = *get_invlist_offset_addr(invlist);
8041 if (array[final_element] == start) {
8042 if (end != UV_MAX) {
8043 array[final_element] = end + 1;
8046 /* But if the end is the maximum representable on the machine,
8047 * just let the range that this would extend to have no end */
8048 invlist_set_len(invlist, len - 1, offset);
8054 /* Here the new range doesn't extend any existing set. Add it */
8056 len += 2; /* Includes an element each for the start and end of range */
8058 /* If wll overflow the existing space, extend, which may cause the array to
8061 invlist_extend(invlist, len);
8063 /* Have to set len here to avoid assert failure in invlist_array() */
8064 invlist_set_len(invlist, len, offset);
8066 array = invlist_array(invlist);
8069 invlist_set_len(invlist, len, offset);
8072 /* The next item on the list starts the range, the one after that is
8073 * one past the new range. */
8074 array[len - 2] = start;
8075 if (end != UV_MAX) {
8076 array[len - 1] = end + 1;
8079 /* But if the end is the maximum representable on the machine, just let
8080 * the range have no end */
8081 invlist_set_len(invlist, len - 1, offset);
8085 #ifndef PERL_IN_XSUB_RE
8088 Perl__invlist_search(pTHX_ SV* const invlist, const UV cp)
8090 /* Searches the inversion list for the entry that contains the input code
8091 * point <cp>. If <cp> is not in the list, -1 is returned. Otherwise, the
8092 * return value is the index into the list's array of the range that
8097 IV high = _invlist_len(invlist);
8098 const IV highest_element = high - 1;
8101 PERL_ARGS_ASSERT__INVLIST_SEARCH;
8103 /* If list is empty, return failure. */
8108 /* (We can't get the array unless we know the list is non-empty) */
8109 array = invlist_array(invlist);
8111 mid = invlist_previous_index(invlist);
8112 assert(mid >=0 && mid <= highest_element);
8114 /* <mid> contains the cache of the result of the previous call to this
8115 * function (0 the first time). See if this call is for the same result,
8116 * or if it is for mid-1. This is under the theory that calls to this
8117 * function will often be for related code points that are near each other.
8118 * And benchmarks show that caching gives better results. We also test
8119 * here if the code point is within the bounds of the list. These tests
8120 * replace others that would have had to be made anyway to make sure that
8121 * the array bounds were not exceeded, and these give us extra information
8122 * at the same time */
8123 if (cp >= array[mid]) {
8124 if (cp >= array[highest_element]) {
8125 return highest_element;
8128 /* Here, array[mid] <= cp < array[highest_element]. This means that
8129 * the final element is not the answer, so can exclude it; it also
8130 * means that <mid> is not the final element, so can refer to 'mid + 1'
8132 if (cp < array[mid + 1]) {
8138 else { /* cp < aray[mid] */
8139 if (cp < array[0]) { /* Fail if outside the array */
8143 if (cp >= array[mid - 1]) {
8148 /* Binary search. What we are looking for is <i> such that
8149 * array[i] <= cp < array[i+1]
8150 * The loop below converges on the i+1. Note that there may not be an
8151 * (i+1)th element in the array, and things work nonetheless */
8152 while (low < high) {
8153 mid = (low + high) / 2;
8154 assert(mid <= highest_element);
8155 if (array[mid] <= cp) { /* cp >= array[mid] */
8158 /* We could do this extra test to exit the loop early.
8159 if (cp < array[low]) {
8164 else { /* cp < array[mid] */
8171 invlist_set_previous_index(invlist, high);
8176 Perl__invlist_populate_swatch(pTHX_ SV* const invlist,
8177 const UV start, const UV end, U8* swatch)
8179 /* populates a swatch of a swash the same way swatch_get() does in utf8.c,
8180 * but is used when the swash has an inversion list. This makes this much
8181 * faster, as it uses a binary search instead of a linear one. This is
8182 * intimately tied to that function, and perhaps should be in utf8.c,
8183 * except it is intimately tied to inversion lists as well. It assumes
8184 * that <swatch> is all 0's on input */
8187 const IV len = _invlist_len(invlist);
8191 PERL_ARGS_ASSERT__INVLIST_POPULATE_SWATCH;
8193 if (len == 0) { /* Empty inversion list */
8197 array = invlist_array(invlist);
8199 /* Find which element it is */
8200 i = _invlist_search(invlist, start);
8202 /* We populate from <start> to <end> */
8203 while (current < end) {
8206 /* The inversion list gives the results for every possible code point
8207 * after the first one in the list. Only those ranges whose index is
8208 * even are ones that the inversion list matches. For the odd ones,
8209 * and if the initial code point is not in the list, we have to skip
8210 * forward to the next element */
8211 if (i == -1 || ! ELEMENT_RANGE_MATCHES_INVLIST(i)) {
8213 if (i >= len) { /* Finished if beyond the end of the array */
8217 if (current >= end) { /* Finished if beyond the end of what we
8219 if (LIKELY(end < UV_MAX)) {
8223 /* We get here when the upper bound is the maximum
8224 * representable on the machine, and we are looking for just
8225 * that code point. Have to special case it */
8227 goto join_end_of_list;
8230 assert(current >= start);
8232 /* The current range ends one below the next one, except don't go past
8235 upper = (i < len && array[i] < end) ? array[i] : end;
8237 /* Here we are in a range that matches. Populate a bit in the 3-bit U8
8238 * for each code point in it */
8239 for (; current < upper; current++) {
8240 const STRLEN offset = (STRLEN)(current - start);
8241 swatch[offset >> 3] |= 1 << (offset & 7);
8246 /* Quit if at the end of the list */
8249 /* But first, have to deal with the highest possible code point on
8250 * the platform. The previous code assumes that <end> is one
8251 * beyond where we want to populate, but that is impossible at the
8252 * platform's infinity, so have to handle it specially */
8253 if (UNLIKELY(end == UV_MAX && ELEMENT_RANGE_MATCHES_INVLIST(len-1)))
8255 const STRLEN offset = (STRLEN)(end - start);
8256 swatch[offset >> 3] |= 1 << (offset & 7);
8261 /* Advance to the next range, which will be for code points not in the
8270 Perl__invlist_union_maybe_complement_2nd(pTHX_ SV* const a, SV* const b,
8271 const bool complement_b, SV** output)
8273 /* Take the union of two inversion lists and point <output> to it. *output
8274 * SHOULD BE DEFINED upon input, and if it points to one of the two lists,
8275 * the reference count to that list will be decremented if not already a
8276 * temporary (mortal); otherwise *output will be made correspondingly
8277 * mortal. The first list, <a>, may be NULL, in which case a copy of the
8278 * second list is returned. If <complement_b> is TRUE, the union is taken
8279 * of the complement (inversion) of <b> instead of b itself.
8281 * The basis for this comes from "Unicode Demystified" Chapter 13 by
8282 * Richard Gillam, published by Addison-Wesley, and explained at some
8283 * length there. The preface says to incorporate its examples into your
8284 * code at your own risk.
8286 * The algorithm is like a merge sort.
8288 * XXX A potential performance improvement is to keep track as we go along
8289 * if only one of the inputs contributes to the result, meaning the other
8290 * is a subset of that one. In that case, we can skip the final copy and
8291 * return the larger of the input lists, but then outside code might need
8292 * to keep track of whether to free the input list or not */
8294 const UV* array_a; /* a's array */
8296 UV len_a; /* length of a's array */
8299 SV* u; /* the resulting union */
8303 UV i_a = 0; /* current index into a's array */
8307 /* running count, as explained in the algorithm source book; items are
8308 * stopped accumulating and are output when the count changes to/from 0.
8309 * The count is incremented when we start a range that's in the set, and
8310 * decremented when we start a range that's not in the set. So its range
8311 * is 0 to 2. Only when the count is zero is something not in the set.
8315 PERL_ARGS_ASSERT__INVLIST_UNION_MAYBE_COMPLEMENT_2ND;
8318 /* If either one is empty, the union is the other one */
8319 if (a == NULL || ((len_a = _invlist_len(a)) == 0)) {
8320 bool make_temp = FALSE; /* Should we mortalize the result? */
8324 if (! (make_temp = cBOOL(SvTEMP(a)))) {
8330 *output = invlist_clone(b);
8332 _invlist_invert(*output);
8334 } /* else *output already = b; */
8337 sv_2mortal(*output);
8341 else if ((len_b = _invlist_len(b)) == 0) {
8342 bool make_temp = FALSE;
8344 if (! (make_temp = cBOOL(SvTEMP(b)))) {
8349 /* The complement of an empty list is a list that has everything in it,
8350 * so the union with <a> includes everything too */
8353 if (! (make_temp = cBOOL(SvTEMP(a)))) {
8357 *output = _new_invlist(1);
8358 _append_range_to_invlist(*output, 0, UV_MAX);
8360 else if (*output != a) {
8361 *output = invlist_clone(a);
8363 /* else *output already = a; */
8366 sv_2mortal(*output);
8371 /* Here both lists exist and are non-empty */
8372 array_a = invlist_array(a);
8373 array_b = invlist_array(b);
8375 /* If are to take the union of 'a' with the complement of b, set it
8376 * up so are looking at b's complement. */
8379 /* To complement, we invert: if the first element is 0, remove it. To
8380 * do this, we just pretend the array starts one later */
8381 if (array_b[0] == 0) {
8387 /* But if the first element is not zero, we pretend the list starts
8388 * at the 0 that is always stored immediately before the array. */
8394 /* Size the union for the worst case: that the sets are completely
8396 u = _new_invlist(len_a + len_b);
8398 /* Will contain U+0000 if either component does */
8399 array_u = _invlist_array_init(u, (len_a > 0 && array_a[0] == 0)
8400 || (len_b > 0 && array_b[0] == 0));
8402 /* Go through each list item by item, stopping when exhausted one of
8404 while (i_a < len_a && i_b < len_b) {
8405 UV cp; /* The element to potentially add to the union's array */
8406 bool cp_in_set; /* is it in the the input list's set or not */
8408 /* We need to take one or the other of the two inputs for the union.
8409 * Since we are merging two sorted lists, we take the smaller of the
8410 * next items. In case of a tie, we take the one that is in its set
8411 * first. If we took one not in the set first, it would decrement the
8412 * count, possibly to 0 which would cause it to be output as ending the
8413 * range, and the next time through we would take the same number, and
8414 * output it again as beginning the next range. By doing it the
8415 * opposite way, there is no possibility that the count will be
8416 * momentarily decremented to 0, and thus the two adjoining ranges will
8417 * be seamlessly merged. (In a tie and both are in the set or both not
8418 * in the set, it doesn't matter which we take first.) */
8419 if (array_a[i_a] < array_b[i_b]
8420 || (array_a[i_a] == array_b[i_b]
8421 && ELEMENT_RANGE_MATCHES_INVLIST(i_a)))
8423 cp_in_set = ELEMENT_RANGE_MATCHES_INVLIST(i_a);
8427 cp_in_set = ELEMENT_RANGE_MATCHES_INVLIST(i_b);
8428 cp = array_b[i_b++];
8431 /* Here, have chosen which of the two inputs to look at. Only output
8432 * if the running count changes to/from 0, which marks the
8433 * beginning/end of a range in that's in the set */
8436 array_u[i_u++] = cp;
8443 array_u[i_u++] = cp;
8448 /* Here, we are finished going through at least one of the lists, which
8449 * means there is something remaining in at most one. We check if the list
8450 * that hasn't been exhausted is positioned such that we are in the middle
8451 * of a range in its set or not. (i_a and i_b point to the element beyond
8452 * the one we care about.) If in the set, we decrement 'count'; if 0, there
8453 * is potentially more to output.
8454 * There are four cases:
8455 * 1) Both weren't in their sets, count is 0, and remains 0. What's left
8456 * in the union is entirely from the non-exhausted set.
8457 * 2) Both were in their sets, count is 2. Nothing further should
8458 * be output, as everything that remains will be in the exhausted
8459 * list's set, hence in the union; decrementing to 1 but not 0 insures
8461 * 3) the exhausted was in its set, non-exhausted isn't, count is 1.
8462 * Nothing further should be output because the union includes
8463 * everything from the exhausted set. Not decrementing ensures that.
8464 * 4) the exhausted wasn't in its set, non-exhausted is, count is 1;
8465 * decrementing to 0 insures that we look at the remainder of the
8466 * non-exhausted set */
8467 if ((i_a != len_a && PREV_RANGE_MATCHES_INVLIST(i_a))
8468 || (i_b != len_b && PREV_RANGE_MATCHES_INVLIST(i_b)))
8473 /* The final length is what we've output so far, plus what else is about to
8474 * be output. (If 'count' is non-zero, then the input list we exhausted
8475 * has everything remaining up to the machine's limit in its set, and hence
8476 * in the union, so there will be no further output. */
8479 /* At most one of the subexpressions will be non-zero */
8480 len_u += (len_a - i_a) + (len_b - i_b);
8483 /* Set result to final length, which can change the pointer to array_u, so
8485 if (len_u != _invlist_len(u)) {
8486 invlist_set_len(u, len_u, *get_invlist_offset_addr(u));
8488 array_u = invlist_array(u);
8491 /* When 'count' is 0, the list that was exhausted (if one was shorter than
8492 * the other) ended with everything above it not in its set. That means
8493 * that the remaining part of the union is precisely the same as the
8494 * non-exhausted list, so can just copy it unchanged. (If both list were
8495 * exhausted at the same time, then the operations below will be both 0.)
8498 IV copy_count; /* At most one will have a non-zero copy count */
8499 if ((copy_count = len_a - i_a) > 0) {
8500 Copy(array_a + i_a, array_u + i_u, copy_count, UV);
8502 else if ((copy_count = len_b - i_b) > 0) {
8503 Copy(array_b + i_b, array_u + i_u, copy_count, UV);
8507 /* We may be removing a reference to one of the inputs. If so, the output
8508 * is made mortal if the input was. (Mortal SVs shouldn't have their ref
8509 * count decremented) */
8510 if (a == *output || b == *output) {
8511 assert(! invlist_is_iterating(*output));
8512 if ((SvTEMP(*output))) {
8516 SvREFCNT_dec_NN(*output);
8526 Perl__invlist_intersection_maybe_complement_2nd(pTHX_ SV* const a, SV* const b,
8527 const bool complement_b, SV** i)
8529 /* Take the intersection of two inversion lists and point <i> to it. *i
8530 * SHOULD BE DEFINED upon input, and if it points to one of the two lists,
8531 * the reference count to that list will be decremented if not already a
8532 * temporary (mortal); otherwise *i will be made correspondingly mortal.
8533 * The first list, <a>, may be NULL, in which case an empty list is
8534 * returned. If <complement_b> is TRUE, the result will be the
8535 * intersection of <a> and the complement (or inversion) of <b> instead of
8538 * The basis for this comes from "Unicode Demystified" Chapter 13 by
8539 * Richard Gillam, published by Addison-Wesley, and explained at some
8540 * length there. The preface says to incorporate its examples into your
8541 * code at your own risk. In fact, it had bugs
8543 * The algorithm is like a merge sort, and is essentially the same as the
8547 const UV* array_a; /* a's array */
8549 UV len_a; /* length of a's array */
8552 SV* r; /* the resulting intersection */
8556 UV i_a = 0; /* current index into a's array */
8560 /* running count, as explained in the algorithm source book; items are
8561 * stopped accumulating and are output when the count changes to/from 2.
8562 * The count is incremented when we start a range that's in the set, and
8563 * decremented when we start a range that's not in the set. So its range
8564 * is 0 to 2. Only when the count is 2 is something in the intersection.
8568 PERL_ARGS_ASSERT__INVLIST_INTERSECTION_MAYBE_COMPLEMENT_2ND;
8571 /* Special case if either one is empty */
8572 len_a = (a == NULL) ? 0 : _invlist_len(a);
8573 if ((len_a == 0) || ((len_b = _invlist_len(b)) == 0)) {
8574 bool make_temp = FALSE;
8576 if (len_a != 0 && complement_b) {
8578 /* Here, 'a' is not empty, therefore from the above 'if', 'b' must
8579 * be empty. Here, also we are using 'b's complement, which hence
8580 * must be every possible code point. Thus the intersection is
8584 if (! (make_temp = cBOOL(SvTEMP(b)))) {
8589 *i = invlist_clone(a);
8591 /* else *i is already 'a' */
8599 /* Here, 'a' or 'b' is empty and not using the complement of 'b'. The
8600 * intersection must be empty */
8602 if (! (make_temp = cBOOL(SvTEMP(a)))) {
8607 if (! (make_temp = cBOOL(SvTEMP(b)))) {
8611 *i = _new_invlist(0);
8619 /* Here both lists exist and are non-empty */
8620 array_a = invlist_array(a);
8621 array_b = invlist_array(b);
8623 /* If are to take the intersection of 'a' with the complement of b, set it
8624 * up so are looking at b's complement. */
8627 /* To complement, we invert: if the first element is 0, remove it. To
8628 * do this, we just pretend the array starts one later */
8629 if (array_b[0] == 0) {
8635 /* But if the first element is not zero, we pretend the list starts
8636 * at the 0 that is always stored immediately before the array. */
8642 /* Size the intersection for the worst case: that the intersection ends up
8643 * fragmenting everything to be completely disjoint */
8644 r= _new_invlist(len_a + len_b);
8646 /* Will contain U+0000 iff both components do */
8647 array_r = _invlist_array_init(r, len_a > 0 && array_a[0] == 0
8648 && len_b > 0 && array_b[0] == 0);
8650 /* Go through each list item by item, stopping when exhausted one of
8652 while (i_a < len_a && i_b < len_b) {
8653 UV cp; /* The element to potentially add to the intersection's
8655 bool cp_in_set; /* Is it in the input list's set or not */
8657 /* We need to take one or the other of the two inputs for the
8658 * intersection. Since we are merging two sorted lists, we take the
8659 * smaller of the next items. In case of a tie, we take the one that
8660 * is not in its set first (a difference from the union algorithm). If
8661 * we took one in the set first, it would increment the count, possibly
8662 * to 2 which would cause it to be output as starting a range in the
8663 * intersection, and the next time through we would take that same
8664 * number, and output it again as ending the set. By doing it the
8665 * opposite of this, there is no possibility that the count will be
8666 * momentarily incremented to 2. (In a tie and both are in the set or
8667 * both not in the set, it doesn't matter which we take first.) */
8668 if (array_a[i_a] < array_b[i_b]
8669 || (array_a[i_a] == array_b[i_b]
8670 && ! ELEMENT_RANGE_MATCHES_INVLIST(i_a)))
8672 cp_in_set = ELEMENT_RANGE_MATCHES_INVLIST(i_a);
8676 cp_in_set = ELEMENT_RANGE_MATCHES_INVLIST(i_b);
8680 /* Here, have chosen which of the two inputs to look at. Only output
8681 * if the running count changes to/from 2, which marks the
8682 * beginning/end of a range that's in the intersection */
8686 array_r[i_r++] = cp;
8691 array_r[i_r++] = cp;
8697 /* Here, we are finished going through at least one of the lists, which
8698 * means there is something remaining in at most one. We check if the list
8699 * that has been exhausted is positioned such that we are in the middle
8700 * of a range in its set or not. (i_a and i_b point to elements 1 beyond
8701 * the ones we care about.) There are four cases:
8702 * 1) Both weren't in their sets, count is 0, and remains 0. There's
8703 * nothing left in the intersection.
8704 * 2) Both were in their sets, count is 2 and perhaps is incremented to
8705 * above 2. What should be output is exactly that which is in the
8706 * non-exhausted set, as everything it has is also in the intersection
8707 * set, and everything it doesn't have can't be in the intersection
8708 * 3) The exhausted was in its set, non-exhausted isn't, count is 1, and
8709 * gets incremented to 2. Like the previous case, the intersection is
8710 * everything that remains in the non-exhausted set.
8711 * 4) the exhausted wasn't in its set, non-exhausted is, count is 1, and
8712 * remains 1. And the intersection has nothing more. */
8713 if ((i_a == len_a && PREV_RANGE_MATCHES_INVLIST(i_a))
8714 || (i_b == len_b && PREV_RANGE_MATCHES_INVLIST(i_b)))
8719 /* The final length is what we've output so far plus what else is in the
8720 * intersection. At most one of the subexpressions below will be non-zero
8724 len_r += (len_a - i_a) + (len_b - i_b);
8727 /* Set result to final length, which can change the pointer to array_r, so
8729 if (len_r != _invlist_len(r)) {
8730 invlist_set_len(r, len_r, *get_invlist_offset_addr(r));
8732 array_r = invlist_array(r);
8735 /* Finish outputting any remaining */
8736 if (count >= 2) { /* At most one will have a non-zero copy count */
8738 if ((copy_count = len_a - i_a) > 0) {
8739 Copy(array_a + i_a, array_r + i_r, copy_count, UV);
8741 else if ((copy_count = len_b - i_b) > 0) {
8742 Copy(array_b + i_b, array_r + i_r, copy_count, UV);
8746 /* We may be removing a reference to one of the inputs. If so, the output
8747 * is made mortal if the input was. (Mortal SVs shouldn't have their ref
8748 * count decremented) */
8749 if (a == *i || b == *i) {
8750 assert(! invlist_is_iterating(*i));
8755 SvREFCNT_dec_NN(*i);
8765 Perl__add_range_to_invlist(pTHX_ SV* invlist, const UV start, const UV end)
8767 /* Add the range from 'start' to 'end' inclusive to the inversion list's
8768 * set. A pointer to the inversion list is returned. This may actually be
8769 * a new list, in which case the passed in one has been destroyed. The
8770 * passed in inversion list can be NULL, in which case a new one is created
8771 * with just the one range in it */
8776 if (invlist == NULL) {
8777 invlist = _new_invlist(2);
8781 len = _invlist_len(invlist);
8784 /* If comes after the final entry actually in the list, can just append it
8787 || (! ELEMENT_RANGE_MATCHES_INVLIST(len - 1)
8788 && start >= invlist_array(invlist)[len - 1]))
8790 _append_range_to_invlist(invlist, start, end);
8794 /* Here, can't just append things, create and return a new inversion list
8795 * which is the union of this range and the existing inversion list */
8796 range_invlist = _new_invlist(2);
8797 _append_range_to_invlist(range_invlist, start, end);
8799 _invlist_union(invlist, range_invlist, &invlist);
8801 /* The temporary can be freed */
8802 SvREFCNT_dec_NN(range_invlist);
8808 Perl__setup_canned_invlist(pTHX_ const STRLEN size, const UV element0,
8809 UV** other_elements_ptr)
8811 /* Create and return an inversion list whose contents are to be populated
8812 * by the caller. The caller gives the number of elements (in 'size') and
8813 * the very first element ('element0'). This function will set
8814 * '*other_elements_ptr' to an array of UVs, where the remaining elements
8817 * Obviously there is some trust involved that the caller will properly
8818 * fill in the other elements of the array.
8820 * (The first element needs to be passed in, as the underlying code does
8821 * things differently depending on whether it is zero or non-zero) */
8823 SV* invlist = _new_invlist(size);
8826 PERL_ARGS_ASSERT__SETUP_CANNED_INVLIST;
8828 _append_range_to_invlist(invlist, element0, element0);
8829 offset = *get_invlist_offset_addr(invlist);
8831 invlist_set_len(invlist, size, offset);
8832 *other_elements_ptr = invlist_array(invlist) + 1;
8838 PERL_STATIC_INLINE SV*
8839 S_add_cp_to_invlist(pTHX_ SV* invlist, const UV cp) {
8840 return _add_range_to_invlist(invlist, cp, cp);
8843 #ifndef PERL_IN_XSUB_RE
8845 Perl__invlist_invert(pTHX_ SV* const invlist)
8847 /* Complement the input inversion list. This adds a 0 if the list didn't
8848 * have a zero; removes it otherwise. As described above, the data
8849 * structure is set up so that this is very efficient */
8851 PERL_ARGS_ASSERT__INVLIST_INVERT;
8853 assert(! invlist_is_iterating(invlist));
8855 /* The inverse of matching nothing is matching everything */
8856 if (_invlist_len(invlist) == 0) {
8857 _append_range_to_invlist(invlist, 0, UV_MAX);
8861 *get_invlist_offset_addr(invlist) = ! *get_invlist_offset_addr(invlist);
8866 PERL_STATIC_INLINE SV*
8867 S_invlist_clone(pTHX_ SV* const invlist)
8870 /* Return a new inversion list that is a copy of the input one, which is
8871 * unchanged. The new list will not be mortal even if the old one was. */
8873 /* Need to allocate extra space to accommodate Perl's addition of a
8874 * trailing NUL to SvPV's, since it thinks they are always strings */
8875 SV* new_invlist = _new_invlist(_invlist_len(invlist) + 1);
8876 STRLEN physical_length = SvCUR(invlist);
8877 bool offset = *(get_invlist_offset_addr(invlist));
8879 PERL_ARGS_ASSERT_INVLIST_CLONE;
8881 *(get_invlist_offset_addr(new_invlist)) = offset;
8882 invlist_set_len(new_invlist, _invlist_len(invlist), offset);
8883 Copy(SvPVX(invlist), SvPVX(new_invlist), physical_length, char);
8888 PERL_STATIC_INLINE STRLEN*
8889 S_get_invlist_iter_addr(pTHX_ SV* invlist)
8891 /* Return the address of the UV that contains the current iteration
8894 PERL_ARGS_ASSERT_GET_INVLIST_ITER_ADDR;
8896 assert(SvTYPE(invlist) == SVt_INVLIST);
8898 return &(((XINVLIST*) SvANY(invlist))->iterator);
8901 PERL_STATIC_INLINE void
8902 S_invlist_iterinit(pTHX_ SV* invlist) /* Initialize iterator for invlist */
8904 PERL_ARGS_ASSERT_INVLIST_ITERINIT;
8906 *get_invlist_iter_addr(invlist) = 0;
8909 PERL_STATIC_INLINE void
8910 S_invlist_iterfinish(pTHX_ SV* invlist)
8912 /* Terminate iterator for invlist. This is to catch development errors.
8913 * Any iteration that is interrupted before completed should call this
8914 * function. Functions that add code points anywhere else but to the end
8915 * of an inversion list assert that they are not in the middle of an
8916 * iteration. If they were, the addition would make the iteration
8917 * problematical: if the iteration hadn't reached the place where things
8918 * were being added, it would be ok */
8920 PERL_ARGS_ASSERT_INVLIST_ITERFINISH;
8922 *get_invlist_iter_addr(invlist) = (STRLEN) UV_MAX;
8926 S_invlist_iternext(pTHX_ SV* invlist, UV* start, UV* end)
8928 /* An C<invlist_iterinit> call on <invlist> must be used to set this up.
8929 * This call sets in <*start> and <*end>, the next range in <invlist>.
8930 * Returns <TRUE> if successful and the next call will return the next
8931 * range; <FALSE> if was already at the end of the list. If the latter,
8932 * <*start> and <*end> are unchanged, and the next call to this function
8933 * will start over at the beginning of the list */
8935 STRLEN* pos = get_invlist_iter_addr(invlist);
8936 UV len = _invlist_len(invlist);
8939 PERL_ARGS_ASSERT_INVLIST_ITERNEXT;
8942 *pos = (STRLEN) UV_MAX; /* Force iterinit() to be required next time */
8946 array = invlist_array(invlist);
8948 *start = array[(*pos)++];
8954 *end = array[(*pos)++] - 1;
8960 PERL_STATIC_INLINE bool
8961 S_invlist_is_iterating(pTHX_ SV* const invlist)
8963 PERL_ARGS_ASSERT_INVLIST_IS_ITERATING;
8965 return *(get_invlist_iter_addr(invlist)) < (STRLEN) UV_MAX;
8968 PERL_STATIC_INLINE UV
8969 S_invlist_highest(pTHX_ SV* const invlist)
8971 /* Returns the highest code point that matches an inversion list. This API
8972 * has an ambiguity, as it returns 0 under either the highest is actually
8973 * 0, or if the list is empty. If this distinction matters to you, check
8974 * for emptiness before calling this function */
8976 UV len = _invlist_len(invlist);
8979 PERL_ARGS_ASSERT_INVLIST_HIGHEST;
8985 array = invlist_array(invlist);
8987 /* The last element in the array in the inversion list always starts a
8988 * range that goes to infinity. That range may be for code points that are
8989 * matched in the inversion list, or it may be for ones that aren't
8990 * matched. In the latter case, the highest code point in the set is one
8991 * less than the beginning of this range; otherwise it is the final element
8992 * of this range: infinity */
8993 return (ELEMENT_RANGE_MATCHES_INVLIST(len - 1))
8995 : array[len - 1] - 1;
8998 #ifndef PERL_IN_XSUB_RE
9000 Perl__invlist_contents(pTHX_ SV* const invlist)
9002 /* Get the contents of an inversion list into a string SV so that they can
9003 * be printed out. It uses the format traditionally done for debug tracing
9007 SV* output = newSVpvs("\n");
9009 PERL_ARGS_ASSERT__INVLIST_CONTENTS;
9011 assert(! invlist_is_iterating(invlist));
9013 invlist_iterinit(invlist);
9014 while (invlist_iternext(invlist, &start, &end)) {
9015 if (end == UV_MAX) {
9016 Perl_sv_catpvf(aTHX_ output, "%04"UVXf"\tINFINITY\n", start);
9018 else if (end != start) {
9019 Perl_sv_catpvf(aTHX_ output, "%04"UVXf"\t%04"UVXf"\n",
9023 Perl_sv_catpvf(aTHX_ output, "%04"UVXf"\n", start);
9031 #ifndef PERL_IN_XSUB_RE
9033 Perl__invlist_dump(pTHX_ PerlIO *file, I32 level,
9034 const char * const indent, SV* const invlist)
9036 /* Designed to be called only by do_sv_dump(). Dumps out the ranges of the
9037 * inversion list 'invlist' to 'file' at 'level' Each line is prefixed by
9038 * the string 'indent'. The output looks like this:
9039 [0] 0x000A .. 0x000D
9041 [4] 0x2028 .. 0x2029
9042 [6] 0x3104 .. INFINITY
9043 * This means that the first range of code points matched by the list are
9044 * 0xA through 0xD; the second range contains only the single code point
9045 * 0x85, etc. An inversion list is an array of UVs. Two array elements
9046 * are used to define each range (except if the final range extends to
9047 * infinity, only a single element is needed). The array index of the
9048 * first element for the corresponding range is given in brackets. */
9053 PERL_ARGS_ASSERT__INVLIST_DUMP;
9055 if (invlist_is_iterating(invlist)) {
9056 Perl_dump_indent(aTHX_ level, file,
9057 "%sCan't dump inversion list because is in middle of iterating\n",
9062 invlist_iterinit(invlist);
9063 while (invlist_iternext(invlist, &start, &end)) {
9064 if (end == UV_MAX) {
9065 Perl_dump_indent(aTHX_ level, file,
9066 "%s[%"UVuf"] 0x%04"UVXf" .. INFINITY\n",
9067 indent, (UV)count, start);
9069 else if (end != start) {
9070 Perl_dump_indent(aTHX_ level, file,
9071 "%s[%"UVuf"] 0x%04"UVXf" .. 0x%04"UVXf"\n",
9072 indent, (UV)count, start, end);
9075 Perl_dump_indent(aTHX_ level, file, "%s[%"UVuf"] 0x%04"UVXf"\n",
9076 indent, (UV)count, start);
9083 #ifdef PERL_ARGS_ASSERT__INVLISTEQ
9085 S__invlistEQ(pTHX_ SV* const a, SV* const b, const bool complement_b)
9087 /* Return a boolean as to if the two passed in inversion lists are
9088 * identical. The final argument, if TRUE, says to take the complement of
9089 * the second inversion list before doing the comparison */
9091 const UV* array_a = invlist_array(a);
9092 const UV* array_b = invlist_array(b);
9093 UV len_a = _invlist_len(a);
9094 UV len_b = _invlist_len(b);
9096 UV i = 0; /* current index into the arrays */
9097 bool retval = TRUE; /* Assume are identical until proven otherwise */
9099 PERL_ARGS_ASSERT__INVLISTEQ;
9101 /* If are to compare 'a' with the complement of b, set it
9102 * up so are looking at b's complement. */
9105 /* The complement of nothing is everything, so <a> would have to have
9106 * just one element, starting at zero (ending at infinity) */
9108 return (len_a == 1 && array_a[0] == 0);
9110 else if (array_b[0] == 0) {
9112 /* Otherwise, to complement, we invert. Here, the first element is
9113 * 0, just remove it. To do this, we just pretend the array starts
9121 /* But if the first element is not zero, we pretend the list starts
9122 * at the 0 that is always stored immediately before the array. */
9128 /* Make sure that the lengths are the same, as well as the final element
9129 * before looping through the remainder. (Thus we test the length, final,
9130 * and first elements right off the bat) */
9131 if (len_a != len_b || array_a[len_a-1] != array_b[len_a-1]) {
9134 else for (i = 0; i < len_a - 1; i++) {
9135 if (array_a[i] != array_b[i]) {
9145 #undef HEADER_LENGTH
9146 #undef TO_INTERNAL_SIZE
9147 #undef FROM_INTERNAL_SIZE
9148 #undef INVLIST_VERSION_ID
9150 /* End of inversion list object */
9153 S_parse_lparen_question_flags(pTHX_ RExC_state_t *pRExC_state)
9155 /* This parses the flags that are in either the '(?foo)' or '(?foo:bar)'
9156 * constructs, and updates RExC_flags with them. On input, RExC_parse
9157 * should point to the first flag; it is updated on output to point to the
9158 * final ')' or ':'. There needs to be at least one flag, or this will
9161 /* for (?g), (?gc), and (?o) warnings; warning
9162 about (?c) will warn about (?g) -- japhy */
9164 #define WASTED_O 0x01
9165 #define WASTED_G 0x02
9166 #define WASTED_C 0x04
9167 #define WASTED_GC (WASTED_G|WASTED_C)
9168 I32 wastedflags = 0x00;
9169 U32 posflags = 0, negflags = 0;
9170 U32 *flagsp = &posflags;
9171 char has_charset_modifier = '\0';
9173 bool has_use_defaults = FALSE;
9174 const char* const seqstart = RExC_parse - 1; /* Point to the '?' */
9176 PERL_ARGS_ASSERT_PARSE_LPAREN_QUESTION_FLAGS;
9178 /* '^' as an initial flag sets certain defaults */
9179 if (UCHARAT(RExC_parse) == '^') {
9181 has_use_defaults = TRUE;
9182 STD_PMMOD_FLAGS_CLEAR(&RExC_flags);
9183 set_regex_charset(&RExC_flags, (RExC_utf8 || RExC_uni_semantics)
9184 ? REGEX_UNICODE_CHARSET
9185 : REGEX_DEPENDS_CHARSET);
9188 cs = get_regex_charset(RExC_flags);
9189 if (cs == REGEX_DEPENDS_CHARSET
9190 && (RExC_utf8 || RExC_uni_semantics))
9192 cs = REGEX_UNICODE_CHARSET;
9195 while (*RExC_parse) {
9196 /* && strchr("iogcmsx", *RExC_parse) */
9197 /* (?g), (?gc) and (?o) are useless here
9198 and must be globally applied -- japhy */
9199 switch (*RExC_parse) {
9201 /* Code for the imsx flags */
9202 CASE_STD_PMMOD_FLAGS_PARSE_SET(flagsp);
9204 case LOCALE_PAT_MOD:
9205 if (has_charset_modifier) {
9206 goto excess_modifier;
9208 else if (flagsp == &negflags) {
9211 cs = REGEX_LOCALE_CHARSET;
9212 has_charset_modifier = LOCALE_PAT_MOD;
9214 case UNICODE_PAT_MOD:
9215 if (has_charset_modifier) {
9216 goto excess_modifier;
9218 else if (flagsp == &negflags) {
9221 cs = REGEX_UNICODE_CHARSET;
9222 has_charset_modifier = UNICODE_PAT_MOD;
9224 case ASCII_RESTRICT_PAT_MOD:
9225 if (flagsp == &negflags) {
9228 if (has_charset_modifier) {
9229 if (cs != REGEX_ASCII_RESTRICTED_CHARSET) {
9230 goto excess_modifier;
9232 /* Doubled modifier implies more restricted */
9233 cs = REGEX_ASCII_MORE_RESTRICTED_CHARSET;
9236 cs = REGEX_ASCII_RESTRICTED_CHARSET;
9238 has_charset_modifier = ASCII_RESTRICT_PAT_MOD;
9240 case DEPENDS_PAT_MOD:
9241 if (has_use_defaults) {
9242 goto fail_modifiers;
9244 else if (flagsp == &negflags) {
9247 else if (has_charset_modifier) {
9248 goto excess_modifier;
9251 /* The dual charset means unicode semantics if the
9252 * pattern (or target, not known until runtime) are
9253 * utf8, or something in the pattern indicates unicode
9255 cs = (RExC_utf8 || RExC_uni_semantics)
9256 ? REGEX_UNICODE_CHARSET
9257 : REGEX_DEPENDS_CHARSET;
9258 has_charset_modifier = DEPENDS_PAT_MOD;
9262 if (has_charset_modifier == ASCII_RESTRICT_PAT_MOD) {
9263 vFAIL2("Regexp modifier \"%c\" may appear a maximum of twice", ASCII_RESTRICT_PAT_MOD);
9265 else if (has_charset_modifier == *(RExC_parse - 1)) {
9266 vFAIL2("Regexp modifier \"%c\" may not appear twice",
9270 vFAIL3("Regexp modifiers \"%c\" and \"%c\" are mutually exclusive", has_charset_modifier, *(RExC_parse - 1));
9275 vFAIL2("Regexp modifier \"%c\" may not appear after the \"-\"",
9278 case ONCE_PAT_MOD: /* 'o' */
9279 case GLOBAL_PAT_MOD: /* 'g' */
9280 if (SIZE_ONLY && ckWARN(WARN_REGEXP)) {
9281 const I32 wflagbit = *RExC_parse == 'o'
9284 if (! (wastedflags & wflagbit) ) {
9285 wastedflags |= wflagbit;
9286 /* diag_listed_as: Useless (?-%s) - don't use /%s modifier in regex; marked by <-- HERE in m/%s/ */
9289 "Useless (%s%c) - %suse /%c modifier",
9290 flagsp == &negflags ? "?-" : "?",
9292 flagsp == &negflags ? "don't " : "",
9299 case CONTINUE_PAT_MOD: /* 'c' */
9300 if (SIZE_ONLY && ckWARN(WARN_REGEXP)) {
9301 if (! (wastedflags & WASTED_C) ) {
9302 wastedflags |= WASTED_GC;
9303 /* diag_listed_as: Useless (?-%s) - don't use /%s modifier in regex; marked by <-- HERE in m/%s/ */
9306 "Useless (%sc) - %suse /gc modifier",
9307 flagsp == &negflags ? "?-" : "?",
9308 flagsp == &negflags ? "don't " : ""
9313 case KEEPCOPY_PAT_MOD: /* 'p' */
9314 if (flagsp == &negflags) {
9316 ckWARNreg(RExC_parse + 1,"Useless use of (?-p)");
9318 *flagsp |= RXf_PMf_KEEPCOPY;
9322 /* A flag is a default iff it is following a minus, so
9323 * if there is a minus, it means will be trying to
9324 * re-specify a default which is an error */
9325 if (has_use_defaults || flagsp == &negflags) {
9326 goto fail_modifiers;
9329 wastedflags = 0; /* reset so (?g-c) warns twice */
9333 RExC_flags |= posflags;
9334 RExC_flags &= ~negflags;
9335 set_regex_charset(&RExC_flags, cs);
9336 if (RExC_flags & RXf_PMf_FOLD) {
9337 RExC_contains_i = 1;
9343 RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
9344 /* diag_listed_as: Sequence (?%s...) not recognized in regex; marked by <-- HERE in m/%s/ */
9345 vFAIL2utf8f("Sequence (%"UTF8f"...) not recognized",
9346 UTF8fARG(UTF, RExC_parse-seqstart, seqstart));
9355 - reg - regular expression, i.e. main body or parenthesized thing
9357 * Caller must absorb opening parenthesis.
9359 * Combining parenthesis handling with the base level of regular expression
9360 * is a trifle forced, but the need to tie the tails of the branches to what
9361 * follows makes it hard to avoid.
9363 #define REGTAIL(x,y,z) regtail((x),(y),(z),depth+1)
9365 #define REGTAIL_STUDY(x,y,z) regtail_study((x),(y),(z),depth+1)
9367 #define REGTAIL_STUDY(x,y,z) regtail((x),(y),(z),depth+1)
9370 /* Returns NULL, setting *flagp to TRYAGAIN at the end of (?) that only sets
9371 flags. Returns NULL, setting *flagp to RESTART_UTF8 if the sizing scan
9372 needs to be restarted.
9373 Otherwise would only return NULL if regbranch() returns NULL, which
9376 S_reg(pTHX_ RExC_state_t *pRExC_state, I32 paren, I32 *flagp,U32 depth)
9377 /* paren: Parenthesized? 0=top; 1,2=inside '(': changed to letter.
9378 * 2 is like 1, but indicates that nextchar() has been called to advance
9379 * RExC_parse beyond the '('. Things like '(?' are indivisible tokens, and
9380 * this flag alerts us to the need to check for that */
9383 regnode *ret; /* Will be the head of the group. */
9386 regnode *ender = NULL;
9389 U32 oregflags = RExC_flags;
9390 bool have_branch = 0;
9392 I32 freeze_paren = 0;
9393 I32 after_freeze = 0;
9395 char * parse_start = RExC_parse; /* MJD */
9396 char * const oregcomp_parse = RExC_parse;
9398 GET_RE_DEBUG_FLAGS_DECL;
9400 PERL_ARGS_ASSERT_REG;
9401 DEBUG_PARSE("reg ");
9403 *flagp = 0; /* Tentatively. */
9406 /* Make an OPEN node, if parenthesized. */
9409 /* Under /x, space and comments can be gobbled up between the '(' and
9410 * here (if paren ==2). The forms '(*VERB' and '(?...' disallow such
9411 * intervening space, as the sequence is a token, and a token should be
9413 bool has_intervening_patws = paren == 2 && *(RExC_parse - 1) != '(';
9415 if ( *RExC_parse == '*') { /* (*VERB:ARG) */
9416 char *start_verb = RExC_parse;
9417 STRLEN verb_len = 0;
9418 char *start_arg = NULL;
9419 unsigned char op = 0;
9421 int internal_argval = 0; /* internal_argval is only useful if
9424 if (has_intervening_patws && SIZE_ONLY) {
9425 ckWARNregdep(RExC_parse + 1, "In '(*VERB...)', splitting the initial '(*' is deprecated");
9427 while ( *RExC_parse && *RExC_parse != ')' ) {
9428 if ( *RExC_parse == ':' ) {
9429 start_arg = RExC_parse + 1;
9435 verb_len = RExC_parse - start_verb;
9438 while ( *RExC_parse && *RExC_parse != ')' )
9440 if ( *RExC_parse != ')' )
9441 vFAIL("Unterminated verb pattern argument");
9442 if ( RExC_parse == start_arg )
9445 if ( *RExC_parse != ')' )
9446 vFAIL("Unterminated verb pattern");
9449 switch ( *start_verb ) {
9450 case 'A': /* (*ACCEPT) */
9451 if ( memEQs(start_verb,verb_len,"ACCEPT") ) {
9453 internal_argval = RExC_nestroot;
9456 case 'C': /* (*COMMIT) */
9457 if ( memEQs(start_verb,verb_len,"COMMIT") )
9460 case 'F': /* (*FAIL) */
9461 if ( verb_len==1 || memEQs(start_verb,verb_len,"FAIL") ) {
9466 case ':': /* (*:NAME) */
9467 case 'M': /* (*MARK:NAME) */
9468 if ( verb_len==0 || memEQs(start_verb,verb_len,"MARK") ) {
9473 case 'P': /* (*PRUNE) */
9474 if ( memEQs(start_verb,verb_len,"PRUNE") )
9477 case 'S': /* (*SKIP) */
9478 if ( memEQs(start_verb,verb_len,"SKIP") )
9481 case 'T': /* (*THEN) */
9482 /* [19:06] <TimToady> :: is then */
9483 if ( memEQs(start_verb,verb_len,"THEN") ) {
9485 RExC_seen |= REG_CUTGROUP_SEEN;
9490 RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
9492 "Unknown verb pattern '%"UTF8f"'",
9493 UTF8fARG(UTF, verb_len, start_verb));
9496 if ( start_arg && internal_argval ) {
9497 vFAIL3("Verb pattern '%.*s' may not have an argument",
9498 verb_len, start_verb);
9499 } else if ( argok < 0 && !start_arg ) {
9500 vFAIL3("Verb pattern '%.*s' has a mandatory argument",
9501 verb_len, start_verb);
9503 ret = reganode(pRExC_state, op, internal_argval);
9504 if ( ! internal_argval && ! SIZE_ONLY ) {
9506 SV *sv = newSVpvn( start_arg,
9507 RExC_parse - start_arg);
9508 ARG(ret) = add_data( pRExC_state,
9510 RExC_rxi->data->data[ARG(ret)]=(void*)sv;
9517 if (!internal_argval)
9518 RExC_seen |= REG_VERBARG_SEEN;
9519 } else if ( start_arg ) {
9520 vFAIL3("Verb pattern '%.*s' may not have an argument",
9521 verb_len, start_verb);
9523 ret = reg_node(pRExC_state, op);
9525 nextchar(pRExC_state);
9528 else if (*RExC_parse == '?') { /* (?...) */
9529 bool is_logical = 0;
9530 const char * const seqstart = RExC_parse;
9531 if (has_intervening_patws && SIZE_ONLY) {
9532 ckWARNregdep(RExC_parse + 1, "In '(?...)', splitting the initial '(?' is deprecated");
9536 paren = *RExC_parse++;
9537 ret = NULL; /* For look-ahead/behind. */
9540 case 'P': /* (?P...) variants for those used to PCRE/Python */
9541 paren = *RExC_parse++;
9542 if ( paren == '<') /* (?P<...>) named capture */
9544 else if (paren == '>') { /* (?P>name) named recursion */
9545 goto named_recursion;
9547 else if (paren == '=') { /* (?P=...) named backref */
9548 /* this pretty much dupes the code for \k<NAME> in
9549 * regatom(), if you change this make sure you change that
9551 char* name_start = RExC_parse;
9553 SV *sv_dat = reg_scan_name(pRExC_state,
9554 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
9555 if (RExC_parse == name_start || *RExC_parse != ')')
9556 /* diag_listed_as: Sequence ?P=... not terminated in regex; marked by <-- HERE in m/%s/ */
9557 vFAIL2("Sequence %.3s... not terminated",parse_start);
9560 num = add_data( pRExC_state, STR_WITH_LEN("S"));
9561 RExC_rxi->data->data[num]=(void*)sv_dat;
9562 SvREFCNT_inc_simple_void(sv_dat);
9565 ret = reganode(pRExC_state,
9568 : (ASCII_FOLD_RESTRICTED)
9570 : (AT_LEAST_UNI_SEMANTICS)
9578 Set_Node_Offset(ret, parse_start+1);
9579 Set_Node_Cur_Length(ret, parse_start);
9581 nextchar(pRExC_state);
9585 /* diag_listed_as: Sequence (?%s...) not recognized in regex; marked by <-- HERE in m/%s/ */
9586 vFAIL3("Sequence (%.*s...) not recognized",
9587 RExC_parse-seqstart, seqstart);
9589 case '<': /* (?<...) */
9590 if (*RExC_parse == '!')
9592 else if (*RExC_parse != '=')
9598 case '\'': /* (?'...') */
9599 name_start= RExC_parse;
9600 svname = reg_scan_name(pRExC_state,
9601 SIZE_ONLY /* reverse test from the others */
9602 ? REG_RSN_RETURN_NAME
9603 : REG_RSN_RETURN_NULL);
9604 if (RExC_parse == name_start || *RExC_parse != paren)
9605 vFAIL2("Sequence (?%c... not terminated",
9606 paren=='>' ? '<' : paren);
9610 if (!svname) /* shouldn't happen */
9612 "panic: reg_scan_name returned NULL");
9613 if (!RExC_paren_names) {
9614 RExC_paren_names= newHV();
9615 sv_2mortal(MUTABLE_SV(RExC_paren_names));
9617 RExC_paren_name_list= newAV();
9618 sv_2mortal(MUTABLE_SV(RExC_paren_name_list));
9621 he_str = hv_fetch_ent( RExC_paren_names, svname, 1, 0 );
9623 sv_dat = HeVAL(he_str);
9625 /* croak baby croak */
9627 "panic: paren_name hash element allocation failed");
9628 } else if ( SvPOK(sv_dat) ) {
9629 /* (?|...) can mean we have dupes so scan to check
9630 its already been stored. Maybe a flag indicating
9631 we are inside such a construct would be useful,
9632 but the arrays are likely to be quite small, so
9633 for now we punt -- dmq */
9634 IV count = SvIV(sv_dat);
9635 I32 *pv = (I32*)SvPVX(sv_dat);
9637 for ( i = 0 ; i < count ; i++ ) {
9638 if ( pv[i] == RExC_npar ) {
9644 pv = (I32*)SvGROW(sv_dat,
9645 SvCUR(sv_dat) + sizeof(I32)+1);
9646 SvCUR_set(sv_dat, SvCUR(sv_dat) + sizeof(I32));
9647 pv[count] = RExC_npar;
9648 SvIV_set(sv_dat, SvIVX(sv_dat) + 1);
9651 (void)SvUPGRADE(sv_dat,SVt_PVNV);
9652 sv_setpvn(sv_dat, (char *)&(RExC_npar),
9655 SvIV_set(sv_dat, 1);
9658 /* Yes this does cause a memory leak in debugging Perls
9660 if (!av_store(RExC_paren_name_list,
9661 RExC_npar, SvREFCNT_inc(svname)))
9662 SvREFCNT_dec_NN(svname);
9665 /*sv_dump(sv_dat);*/
9667 nextchar(pRExC_state);
9669 goto capturing_parens;
9671 RExC_seen |= REG_LOOKBEHIND_SEEN;
9672 RExC_in_lookbehind++;
9674 case '=': /* (?=...) */
9675 RExC_seen_zerolen++;
9677 case '!': /* (?!...) */
9678 RExC_seen_zerolen++;
9679 if (*RExC_parse == ')') {
9680 ret=reg_node(pRExC_state, OPFAIL);
9681 nextchar(pRExC_state);
9685 case '|': /* (?|...) */
9686 /* branch reset, behave like a (?:...) except that
9687 buffers in alternations share the same numbers */
9689 after_freeze = freeze_paren = RExC_npar;
9691 case ':': /* (?:...) */
9692 case '>': /* (?>...) */
9694 case '$': /* (?$...) */
9695 case '@': /* (?@...) */
9696 vFAIL2("Sequence (?%c...) not implemented", (int)paren);
9698 case '#': /* (?#...) */
9699 /* XXX As soon as we disallow separating the '?' and '*' (by
9700 * spaces or (?#...) comment), it is believed that this case
9701 * will be unreachable and can be removed. See
9703 while (*RExC_parse && *RExC_parse != ')')
9705 if (*RExC_parse != ')')
9706 FAIL("Sequence (?#... not terminated");
9707 nextchar(pRExC_state);
9710 case '0' : /* (?0) */
9711 case 'R' : /* (?R) */
9712 if (*RExC_parse != ')')
9713 FAIL("Sequence (?R) not terminated");
9714 ret = reg_node(pRExC_state, GOSTART);
9715 RExC_seen |= REG_GOSTART_SEEN;
9716 *flagp |= POSTPONED;
9717 nextchar(pRExC_state);
9720 { /* named and numeric backreferences */
9722 case '&': /* (?&NAME) */
9723 parse_start = RExC_parse - 1;
9726 SV *sv_dat = reg_scan_name(pRExC_state,
9727 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
9728 num = sv_dat ? *((I32 *)SvPVX(sv_dat)) : 0;
9730 if (RExC_parse == RExC_end || *RExC_parse != ')')
9731 vFAIL("Sequence (?&... not terminated");
9732 goto gen_recurse_regop;
9733 assert(0); /* NOT REACHED */
9735 if (!(RExC_parse[0] >= '1' && RExC_parse[0] <= '9')) {
9737 vFAIL("Illegal pattern");
9739 goto parse_recursion;
9741 case '-': /* (?-1) */
9742 if (!(RExC_parse[0] >= '1' && RExC_parse[0] <= '9')) {
9743 RExC_parse--; /* rewind to let it be handled later */
9747 case '1': case '2': case '3': case '4': /* (?1) */
9748 case '5': case '6': case '7': case '8': case '9':
9751 num = atoi(RExC_parse);
9752 parse_start = RExC_parse - 1; /* MJD */
9753 if (*RExC_parse == '-')
9755 while (isDIGIT(*RExC_parse))
9757 if (*RExC_parse!=')')
9758 vFAIL("Expecting close bracket");
9761 if ( paren == '-' ) {
9763 Diagram of capture buffer numbering.
9764 Top line is the normal capture buffer numbers
9765 Bottom line is the negative indexing as from
9769 /(a(x)y)(a(b(c(?-2)d)e)f)(g(h))/
9773 num = RExC_npar + num;
9776 vFAIL("Reference to nonexistent group");
9778 } else if ( paren == '+' ) {
9779 num = RExC_npar + num - 1;
9782 ret = reganode(pRExC_state, GOSUB, num);
9784 if (num > (I32)RExC_rx->nparens) {
9786 vFAIL("Reference to nonexistent group");
9788 ARG2L_SET( ret, RExC_recurse_count++);
9790 DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log,
9791 "Recurse #%"UVuf" to %"IVdf"\n",
9792 (UV)ARG(ret), (IV)ARG2L(ret)));
9796 RExC_seen |= REG_RECURSE_SEEN;
9797 Set_Node_Length(ret, 1 + regarglen[OP(ret)]); /* MJD */
9798 Set_Node_Offset(ret, parse_start); /* MJD */
9800 *flagp |= POSTPONED;
9801 nextchar(pRExC_state);
9803 } /* named and numeric backreferences */
9804 assert(0); /* NOT REACHED */
9806 case '?': /* (??...) */
9808 if (*RExC_parse != '{') {
9810 /* diag_listed_as: Sequence (?%s...) not recognized in regex; marked by <-- HERE in m/%s/ */
9812 "Sequence (%"UTF8f"...) not recognized",
9813 UTF8fARG(UTF, RExC_parse-seqstart, seqstart));
9816 *flagp |= POSTPONED;
9817 paren = *RExC_parse++;
9819 case '{': /* (?{...}) */
9822 struct reg_code_block *cb;
9824 RExC_seen_zerolen++;
9826 if ( !pRExC_state->num_code_blocks
9827 || pRExC_state->code_index >= pRExC_state->num_code_blocks
9828 || pRExC_state->code_blocks[pRExC_state->code_index].start
9829 != (STRLEN)((RExC_parse -3 - (is_logical ? 1 : 0))
9832 if (RExC_pm_flags & PMf_USE_RE_EVAL)
9833 FAIL("panic: Sequence (?{...}): no code block found\n");
9834 FAIL("Eval-group not allowed at runtime, use re 'eval'");
9836 /* this is a pre-compiled code block (?{...}) */
9837 cb = &pRExC_state->code_blocks[pRExC_state->code_index];
9838 RExC_parse = RExC_start + cb->end;
9841 if (cb->src_regex) {
9842 n = add_data(pRExC_state, STR_WITH_LEN("rl"));
9843 RExC_rxi->data->data[n] =
9844 (void*)SvREFCNT_inc((SV*)cb->src_regex);
9845 RExC_rxi->data->data[n+1] = (void*)o;
9848 n = add_data(pRExC_state,
9849 (RExC_pm_flags & PMf_HAS_CV) ? "L" : "l", 1);
9850 RExC_rxi->data->data[n] = (void*)o;
9853 pRExC_state->code_index++;
9854 nextchar(pRExC_state);
9858 ret = reg_node(pRExC_state, LOGICAL);
9859 eval = reganode(pRExC_state, EVAL, n);
9862 /* for later propagation into (??{}) return value */
9863 eval->flags = (U8) (RExC_flags & RXf_PMf_COMPILETIME);
9865 REGTAIL(pRExC_state, ret, eval);
9866 /* deal with the length of this later - MJD */
9869 ret = reganode(pRExC_state, EVAL, n);
9870 Set_Node_Length(ret, RExC_parse - parse_start + 1);
9871 Set_Node_Offset(ret, parse_start);
9874 case '(': /* (?(?{...})...) and (?(?=...)...) */
9877 if (RExC_parse[0] == '?') { /* (?(?...)) */
9878 if (RExC_parse[1] == '=' || RExC_parse[1] == '!'
9879 || RExC_parse[1] == '<'
9880 || RExC_parse[1] == '{') { /* Lookahead or eval. */
9884 ret = reg_node(pRExC_state, LOGICAL);
9888 tail = reg(pRExC_state, 1, &flag, depth+1);
9889 if (flag & RESTART_UTF8) {
9890 *flagp = RESTART_UTF8;
9893 REGTAIL(pRExC_state, ret, tail);
9897 else if ( RExC_parse[0] == '<' /* (?(<NAME>)...) */
9898 || RExC_parse[0] == '\'' ) /* (?('NAME')...) */
9900 char ch = RExC_parse[0] == '<' ? '>' : '\'';
9901 char *name_start= RExC_parse++;
9903 SV *sv_dat=reg_scan_name(pRExC_state,
9904 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
9905 if (RExC_parse == name_start || *RExC_parse != ch)
9906 vFAIL2("Sequence (?(%c... not terminated",
9907 (ch == '>' ? '<' : ch));
9910 num = add_data( pRExC_state, STR_WITH_LEN("S"));
9911 RExC_rxi->data->data[num]=(void*)sv_dat;
9912 SvREFCNT_inc_simple_void(sv_dat);
9914 ret = reganode(pRExC_state,NGROUPP,num);
9915 goto insert_if_check_paren;
9917 else if (RExC_parse[0] == 'D' &&
9918 RExC_parse[1] == 'E' &&
9919 RExC_parse[2] == 'F' &&
9920 RExC_parse[3] == 'I' &&
9921 RExC_parse[4] == 'N' &&
9922 RExC_parse[5] == 'E')
9924 ret = reganode(pRExC_state,DEFINEP,0);
9927 goto insert_if_check_paren;
9929 else if (RExC_parse[0] == 'R') {
9932 if (RExC_parse[0] >= '1' && RExC_parse[0] <= '9' ) {
9933 parno = atoi(RExC_parse++);
9934 while (isDIGIT(*RExC_parse))
9936 } else if (RExC_parse[0] == '&') {
9939 sv_dat = reg_scan_name(pRExC_state,
9941 ? REG_RSN_RETURN_NULL
9942 : REG_RSN_RETURN_DATA);
9943 parno = sv_dat ? *((I32 *)SvPVX(sv_dat)) : 0;
9945 ret = reganode(pRExC_state,INSUBP,parno);
9946 goto insert_if_check_paren;
9948 else if (RExC_parse[0] >= '1' && RExC_parse[0] <= '9' ) {
9952 parno = atoi(RExC_parse++);
9954 while (isDIGIT(*RExC_parse))
9956 ret = reganode(pRExC_state, GROUPP, parno);
9958 insert_if_check_paren:
9959 if (*(tmp = nextchar(pRExC_state)) != ')') {
9960 /* nextchar also skips comments, so undo its work
9961 * and skip over the the next character.
9964 RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
9965 vFAIL("Switch condition not recognized");
9968 REGTAIL(pRExC_state, ret, reganode(pRExC_state, IFTHEN, 0));
9969 br = regbranch(pRExC_state, &flags, 1,depth+1);
9971 if (flags & RESTART_UTF8) {
9972 *flagp = RESTART_UTF8;
9975 FAIL2("panic: regbranch returned NULL, flags=%#"UVxf"",
9978 REGTAIL(pRExC_state, br, reganode(pRExC_state,
9980 c = *nextchar(pRExC_state);
9985 vFAIL("(?(DEFINE)....) does not allow branches");
9987 /* Fake one for optimizer. */
9988 lastbr = reganode(pRExC_state, IFTHEN, 0);
9990 if (!regbranch(pRExC_state, &flags, 1,depth+1)) {
9991 if (flags & RESTART_UTF8) {
9992 *flagp = RESTART_UTF8;
9995 FAIL2("panic: regbranch returned NULL, flags=%#"UVxf"",
9998 REGTAIL(pRExC_state, ret, lastbr);
10000 *flagp |= HASWIDTH;
10001 c = *nextchar(pRExC_state);
10006 vFAIL("Switch (?(condition)... contains too many branches");
10007 ender = reg_node(pRExC_state, TAIL);
10008 REGTAIL(pRExC_state, br, ender);
10010 REGTAIL(pRExC_state, lastbr, ender);
10011 REGTAIL(pRExC_state, NEXTOPER(NEXTOPER(lastbr)), ender);
10014 REGTAIL(pRExC_state, ret, ender);
10015 RExC_size++; /* XXX WHY do we need this?!!
10016 For large programs it seems to be required
10017 but I can't figure out why. -- dmq*/
10021 RExC_parse += UTF ? UTF8SKIP(RExC_parse) : 1;
10022 vFAIL("Unknown switch condition (?(...))");
10025 case '[': /* (?[ ... ]) */
10026 return handle_regex_sets(pRExC_state, NULL, flagp, depth,
10029 RExC_parse--; /* for vFAIL to print correctly */
10030 vFAIL("Sequence (? incomplete");
10032 default: /* e.g., (?i) */
10035 parse_lparen_question_flags(pRExC_state);
10036 if (UCHARAT(RExC_parse) != ':') {
10037 nextchar(pRExC_state);
10042 nextchar(pRExC_state);
10052 ret = reganode(pRExC_state, OPEN, parno);
10054 if (!RExC_nestroot)
10055 RExC_nestroot = parno;
10056 if (RExC_seen & REG_RECURSE_SEEN
10057 && !RExC_open_parens[parno-1])
10059 DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log,
10060 "Setting open paren #%"IVdf" to %d\n",
10061 (IV)parno, REG_NODE_NUM(ret)));
10062 RExC_open_parens[parno-1]= ret;
10065 Set_Node_Length(ret, 1); /* MJD */
10066 Set_Node_Offset(ret, RExC_parse); /* MJD */
10074 /* Pick up the branches, linking them together. */
10075 parse_start = RExC_parse; /* MJD */
10076 br = regbranch(pRExC_state, &flags, 1,depth+1);
10078 /* branch_len = (paren != 0); */
10081 if (flags & RESTART_UTF8) {
10082 *flagp = RESTART_UTF8;
10085 FAIL2("panic: regbranch returned NULL, flags=%#"UVxf"", (UV) flags);
10087 if (*RExC_parse == '|') {
10088 if (!SIZE_ONLY && RExC_extralen) {
10089 reginsert(pRExC_state, BRANCHJ, br, depth+1);
10092 reginsert(pRExC_state, BRANCH, br, depth+1);
10093 Set_Node_Length(br, paren != 0);
10094 Set_Node_Offset_To_R(br-RExC_emit_start, parse_start-RExC_start);
10098 RExC_extralen += 1; /* For BRANCHJ-BRANCH. */
10100 else if (paren == ':') {
10101 *flagp |= flags&SIMPLE;
10103 if (is_open) { /* Starts with OPEN. */
10104 REGTAIL(pRExC_state, ret, br); /* OPEN -> first. */
10106 else if (paren != '?') /* Not Conditional */
10108 *flagp |= flags & (SPSTART | HASWIDTH | POSTPONED);
10110 while (*RExC_parse == '|') {
10111 if (!SIZE_ONLY && RExC_extralen) {
10112 ender = reganode(pRExC_state, LONGJMP,0);
10114 /* Append to the previous. */
10115 REGTAIL(pRExC_state, NEXTOPER(NEXTOPER(lastbr)), ender);
10118 RExC_extralen += 2; /* Account for LONGJMP. */
10119 nextchar(pRExC_state);
10120 if (freeze_paren) {
10121 if (RExC_npar > after_freeze)
10122 after_freeze = RExC_npar;
10123 RExC_npar = freeze_paren;
10125 br = regbranch(pRExC_state, &flags, 0, depth+1);
10128 if (flags & RESTART_UTF8) {
10129 *flagp = RESTART_UTF8;
10132 FAIL2("panic: regbranch returned NULL, flags=%#"UVxf"", (UV) flags);
10134 REGTAIL(pRExC_state, lastbr, br); /* BRANCH -> BRANCH. */
10136 *flagp |= flags & (SPSTART | HASWIDTH | POSTPONED);
10139 if (have_branch || paren != ':') {
10140 /* Make a closing node, and hook it on the end. */
10143 ender = reg_node(pRExC_state, TAIL);
10146 ender = reganode(pRExC_state, CLOSE, parno);
10147 if (!SIZE_ONLY && RExC_seen & REG_RECURSE_SEEN) {
10148 DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log,
10149 "Setting close paren #%"IVdf" to %d\n",
10150 (IV)parno, REG_NODE_NUM(ender)));
10151 RExC_close_parens[parno-1]= ender;
10152 if (RExC_nestroot == parno)
10155 Set_Node_Offset(ender,RExC_parse+1); /* MJD */
10156 Set_Node_Length(ender,1); /* MJD */
10162 *flagp &= ~HASWIDTH;
10165 ender = reg_node(pRExC_state, SUCCEED);
10168 ender = reg_node(pRExC_state, END);
10170 assert(!RExC_opend); /* there can only be one! */
10171 RExC_opend = ender;
10175 DEBUG_PARSE_r(if (!SIZE_ONLY) {
10176 SV * const mysv_val1=sv_newmortal();
10177 SV * const mysv_val2=sv_newmortal();
10178 DEBUG_PARSE_MSG("lsbr");
10179 regprop(RExC_rx, mysv_val1, lastbr, NULL);
10180 regprop(RExC_rx, mysv_val2, ender, NULL);
10181 PerlIO_printf(Perl_debug_log, "~ tying lastbr %s (%"IVdf") to ender %s (%"IVdf") offset %"IVdf"\n",
10182 SvPV_nolen_const(mysv_val1),
10183 (IV)REG_NODE_NUM(lastbr),
10184 SvPV_nolen_const(mysv_val2),
10185 (IV)REG_NODE_NUM(ender),
10186 (IV)(ender - lastbr)
10189 REGTAIL(pRExC_state, lastbr, ender);
10191 if (have_branch && !SIZE_ONLY) {
10192 char is_nothing= 1;
10194 RExC_seen |= REG_TOP_LEVEL_BRANCHES_SEEN;
10196 /* Hook the tails of the branches to the closing node. */
10197 for (br = ret; br; br = regnext(br)) {
10198 const U8 op = PL_regkind[OP(br)];
10199 if (op == BRANCH) {
10200 REGTAIL_STUDY(pRExC_state, NEXTOPER(br), ender);
10201 if ( OP(NEXTOPER(br)) != NOTHING
10202 || regnext(NEXTOPER(br)) != ender)
10205 else if (op == BRANCHJ) {
10206 REGTAIL_STUDY(pRExC_state, NEXTOPER(NEXTOPER(br)), ender);
10207 /* for now we always disable this optimisation * /
10208 if ( OP(NEXTOPER(NEXTOPER(br))) != NOTHING
10209 || regnext(NEXTOPER(NEXTOPER(br))) != ender)
10215 br= PL_regkind[OP(ret)] != BRANCH ? regnext(ret) : ret;
10216 DEBUG_PARSE_r(if (!SIZE_ONLY) {
10217 SV * const mysv_val1=sv_newmortal();
10218 SV * const mysv_val2=sv_newmortal();
10219 DEBUG_PARSE_MSG("NADA");
10220 regprop(RExC_rx, mysv_val1, ret, NULL);
10221 regprop(RExC_rx, mysv_val2, ender, NULL);
10222 PerlIO_printf(Perl_debug_log, "~ converting ret %s (%"IVdf") to ender %s (%"IVdf") offset %"IVdf"\n",
10223 SvPV_nolen_const(mysv_val1),
10224 (IV)REG_NODE_NUM(ret),
10225 SvPV_nolen_const(mysv_val2),
10226 (IV)REG_NODE_NUM(ender),
10231 if (OP(ender) == TAIL) {
10236 for ( opt= br + 1; opt < ender ; opt++ )
10237 OP(opt)= OPTIMIZED;
10238 NEXT_OFF(br)= ender - br;
10246 static const char parens[] = "=!<,>";
10248 if (paren && (p = strchr(parens, paren))) {
10249 U8 node = ((p - parens) % 2) ? UNLESSM : IFMATCH;
10250 int flag = (p - parens) > 1;
10253 node = SUSPEND, flag = 0;
10254 reginsert(pRExC_state, node,ret, depth+1);
10255 Set_Node_Cur_Length(ret, parse_start);
10256 Set_Node_Offset(ret, parse_start + 1);
10258 REGTAIL_STUDY(pRExC_state, ret, reg_node(pRExC_state, TAIL));
10262 /* Check for proper termination. */
10264 /* restore original flags, but keep (?p) */
10265 RExC_flags = oregflags | (RExC_flags & RXf_PMf_KEEPCOPY);
10266 if (RExC_parse >= RExC_end || *nextchar(pRExC_state) != ')') {
10267 RExC_parse = oregcomp_parse;
10268 vFAIL("Unmatched (");
10271 else if (!paren && RExC_parse < RExC_end) {
10272 if (*RExC_parse == ')') {
10274 vFAIL("Unmatched )");
10277 FAIL("Junk on end of regexp"); /* "Can't happen". */
10278 assert(0); /* NOTREACHED */
10281 if (RExC_in_lookbehind) {
10282 RExC_in_lookbehind--;
10284 if (after_freeze > RExC_npar)
10285 RExC_npar = after_freeze;
10290 - regbranch - one alternative of an | operator
10292 * Implements the concatenation operator.
10294 * Returns NULL, setting *flagp to RESTART_UTF8 if the sizing scan needs to be
10298 S_regbranch(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, I32 first, U32 depth)
10302 regnode *chain = NULL;
10304 I32 flags = 0, c = 0;
10305 GET_RE_DEBUG_FLAGS_DECL;
10307 PERL_ARGS_ASSERT_REGBRANCH;
10309 DEBUG_PARSE("brnc");
10314 if (!SIZE_ONLY && RExC_extralen)
10315 ret = reganode(pRExC_state, BRANCHJ,0);
10317 ret = reg_node(pRExC_state, BRANCH);
10318 Set_Node_Length(ret, 1);
10322 if (!first && SIZE_ONLY)
10323 RExC_extralen += 1; /* BRANCHJ */
10325 *flagp = WORST; /* Tentatively. */
10328 nextchar(pRExC_state);
10329 while (RExC_parse < RExC_end && *RExC_parse != '|' && *RExC_parse != ')') {
10330 flags &= ~TRYAGAIN;
10331 latest = regpiece(pRExC_state, &flags,depth+1);
10332 if (latest == NULL) {
10333 if (flags & TRYAGAIN)
10335 if (flags & RESTART_UTF8) {
10336 *flagp = RESTART_UTF8;
10339 FAIL2("panic: regpiece returned NULL, flags=%#"UVxf"", (UV) flags);
10341 else if (ret == NULL)
10343 *flagp |= flags&(HASWIDTH|POSTPONED);
10344 if (chain == NULL) /* First piece. */
10345 *flagp |= flags&SPSTART;
10348 REGTAIL(pRExC_state, chain, latest);
10353 if (chain == NULL) { /* Loop ran zero times. */
10354 chain = reg_node(pRExC_state, NOTHING);
10359 *flagp |= flags&SIMPLE;
10366 - regpiece - something followed by possible [*+?]
10368 * Note that the branching code sequences used for ? and the general cases
10369 * of * and + are somewhat optimized: they use the same NOTHING node as
10370 * both the endmarker for their branch list and the body of the last branch.
10371 * It might seem that this node could be dispensed with entirely, but the
10372 * endmarker role is not redundant.
10374 * Returns NULL, setting *flagp to TRYAGAIN if regatom() returns NULL with
10376 * Returns NULL, setting *flagp to RESTART_UTF8 if the sizing scan needs to be
10380 S_regpiece(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth)
10387 const char * const origparse = RExC_parse;
10389 I32 max = REG_INFTY;
10390 #ifdef RE_TRACK_PATTERN_OFFSETS
10393 const char *maxpos = NULL;
10395 /* Save the original in case we change the emitted regop to a FAIL. */
10396 regnode * const orig_emit = RExC_emit;
10398 GET_RE_DEBUG_FLAGS_DECL;
10400 PERL_ARGS_ASSERT_REGPIECE;
10402 DEBUG_PARSE("piec");
10404 ret = regatom(pRExC_state, &flags,depth+1);
10406 if (flags & (TRYAGAIN|RESTART_UTF8))
10407 *flagp |= flags & (TRYAGAIN|RESTART_UTF8);
10409 FAIL2("panic: regatom returned NULL, flags=%#"UVxf"", (UV) flags);
10415 if (op == '{' && regcurly(RExC_parse, FALSE)) {
10417 #ifdef RE_TRACK_PATTERN_OFFSETS
10418 parse_start = RExC_parse; /* MJD */
10420 next = RExC_parse + 1;
10421 while (isDIGIT(*next) || *next == ',') {
10422 if (*next == ',') {
10430 if (*next == '}') { /* got one */
10434 min = atoi(RExC_parse);
10435 if (*maxpos == ',')
10438 maxpos = RExC_parse;
10439 max = atoi(maxpos);
10440 if (!max && *maxpos != '0')
10441 max = REG_INFTY; /* meaning "infinity" */
10442 else if (max >= REG_INFTY)
10443 vFAIL2("Quantifier in {,} bigger than %d", REG_INFTY - 1);
10445 nextchar(pRExC_state);
10446 if (max < min) { /* If can't match, warn and optimize to fail
10449 ckWARNreg(RExC_parse, "Quantifier {n,m} with n > m can't match");
10451 /* We can't back off the size because we have to reserve
10452 * enough space for all the things we are about to throw
10453 * away, but we can shrink it by the ammount we are about
10454 * to re-use here */
10455 RExC_size = PREVOPER(RExC_size) - regarglen[(U8)OPFAIL];
10458 RExC_emit = orig_emit;
10460 ret = reg_node(pRExC_state, OPFAIL);
10463 else if (min == max
10464 && RExC_parse < RExC_end
10465 && (*RExC_parse == '?' || *RExC_parse == '+'))
10468 ckWARN2reg(RExC_parse + 1,
10469 "Useless use of greediness modifier '%c'",
10472 /* Absorb the modifier, so later code doesn't see nor use
10474 nextchar(pRExC_state);
10478 if ((flags&SIMPLE)) {
10479 RExC_naughty += 2 + RExC_naughty / 2;
10480 reginsert(pRExC_state, CURLY, ret, depth+1);
10481 Set_Node_Offset(ret, parse_start+1); /* MJD */
10482 Set_Node_Cur_Length(ret, parse_start);
10485 regnode * const w = reg_node(pRExC_state, WHILEM);
10488 REGTAIL(pRExC_state, ret, w);
10489 if (!SIZE_ONLY && RExC_extralen) {
10490 reginsert(pRExC_state, LONGJMP,ret, depth+1);
10491 reginsert(pRExC_state, NOTHING,ret, depth+1);
10492 NEXT_OFF(ret) = 3; /* Go over LONGJMP. */
10494 reginsert(pRExC_state, CURLYX,ret, depth+1);
10496 Set_Node_Offset(ret, parse_start+1);
10497 Set_Node_Length(ret,
10498 op == '{' ? (RExC_parse - parse_start) : 1);
10500 if (!SIZE_ONLY && RExC_extralen)
10501 NEXT_OFF(ret) = 3; /* Go over NOTHING to LONGJMP. */
10502 REGTAIL(pRExC_state, ret, reg_node(pRExC_state, NOTHING));
10504 RExC_whilem_seen++, RExC_extralen += 3;
10505 RExC_naughty += 4 + RExC_naughty; /* compound interest */
10512 *flagp |= HASWIDTH;
10514 ARG1_SET(ret, (U16)min);
10515 ARG2_SET(ret, (U16)max);
10517 if (max == REG_INFTY)
10518 RExC_seen |= REG_UNBOUNDED_QUANTIFIER_SEEN;
10524 if (!ISMULT1(op)) {
10529 #if 0 /* Now runtime fix should be reliable. */
10531 /* if this is reinstated, don't forget to put this back into perldiag:
10533 =item Regexp *+ operand could be empty at {#} in regex m/%s/
10535 (F) The part of the regexp subject to either the * or + quantifier
10536 could match an empty string. The {#} shows in the regular
10537 expression about where the problem was discovered.
10541 if (!(flags&HASWIDTH) && op != '?')
10542 vFAIL("Regexp *+ operand could be empty");
10545 #ifdef RE_TRACK_PATTERN_OFFSETS
10546 parse_start = RExC_parse;
10548 nextchar(pRExC_state);
10550 *flagp = (op != '+') ? (WORST|SPSTART|HASWIDTH) : (WORST|HASWIDTH);
10552 if (op == '*' && (flags&SIMPLE)) {
10553 reginsert(pRExC_state, STAR, ret, depth+1);
10556 RExC_seen |= REG_UNBOUNDED_QUANTIFIER_SEEN;
10558 else if (op == '*') {
10562 else if (op == '+' && (flags&SIMPLE)) {
10563 reginsert(pRExC_state, PLUS, ret, depth+1);
10566 RExC_seen |= REG_UNBOUNDED_QUANTIFIER_SEEN;
10568 else if (op == '+') {
10572 else if (op == '?') {
10577 if (!SIZE_ONLY && !(flags&(HASWIDTH|POSTPONED)) && max > REG_INFTY/3) {
10578 SAVEFREESV(RExC_rx_sv); /* in case of fatal warnings */
10579 ckWARN2reg(RExC_parse,
10580 "%"UTF8f" matches null string many times",
10581 UTF8fARG(UTF, (RExC_parse >= origparse
10582 ? RExC_parse - origparse
10585 (void)ReREFCNT_inc(RExC_rx_sv);
10588 if (RExC_parse < RExC_end && *RExC_parse == '?') {
10589 nextchar(pRExC_state);
10590 reginsert(pRExC_state, MINMOD, ret, depth+1);
10591 REGTAIL(pRExC_state, ret, ret + NODE_STEP_REGNODE);
10594 if (RExC_parse < RExC_end && *RExC_parse == '+') {
10596 nextchar(pRExC_state);
10597 ender = reg_node(pRExC_state, SUCCEED);
10598 REGTAIL(pRExC_state, ret, ender);
10599 reginsert(pRExC_state, SUSPEND, ret, depth+1);
10601 ender = reg_node(pRExC_state, TAIL);
10602 REGTAIL(pRExC_state, ret, ender);
10605 if (RExC_parse < RExC_end && ISMULT2(RExC_parse)) {
10607 vFAIL("Nested quantifiers");
10614 S_grok_bslash_N(pTHX_ RExC_state_t *pRExC_state, regnode** node_p,
10615 UV *valuep, I32 *flagp, U32 depth, bool in_char_class,
10616 const bool strict /* Apply stricter parsing rules? */
10620 /* This is expected to be called by a parser routine that has recognized '\N'
10621 and needs to handle the rest. RExC_parse is expected to point at the first
10622 char following the N at the time of the call. On successful return,
10623 RExC_parse has been updated to point to just after the sequence identified
10624 by this routine, and <*flagp> has been updated.
10626 The \N may be inside (indicated by the boolean <in_char_class>) or outside a
10629 \N may begin either a named sequence, or if outside a character class, mean
10630 to match a non-newline. For non single-quoted regexes, the tokenizer has
10631 attempted to decide which, and in the case of a named sequence, converted it
10632 into one of the forms: \N{} (if the sequence is null), or \N{U+c1.c2...},
10633 where c1... are the characters in the sequence. For single-quoted regexes,
10634 the tokenizer passes the \N sequence through unchanged; this code will not
10635 attempt to determine this nor expand those, instead raising a syntax error.
10636 The net effect is that if the beginning of the passed-in pattern isn't '{U+'
10637 or there is no '}', it signals that this \N occurrence means to match a
10640 Only the \N{U+...} form should occur in a character class, for the same
10641 reason that '.' inside a character class means to just match a period: it
10642 just doesn't make sense.
10644 The function raises an error (via vFAIL), and doesn't return for various
10645 syntax errors. Otherwise it returns TRUE and sets <node_p> or <valuep> on
10646 success; it returns FALSE otherwise. Returns FALSE, setting *flagp to
10647 RESTART_UTF8 if the sizing scan needs to be restarted. Such a restart is
10648 only possible if node_p is non-NULL.
10651 If <valuep> is non-null, it means the caller can accept an input sequence
10652 consisting of a just a single code point; <*valuep> is set to that value
10653 if the input is such.
10655 If <node_p> is non-null it signifies that the caller can accept any other
10656 legal sequence (i.e., one that isn't just a single code point). <*node_p>
10658 1) \N means not-a-NL: points to a newly created REG_ANY node;
10659 2) \N{}: points to a new NOTHING node;
10660 3) otherwise: points to a new EXACT node containing the resolved
10662 Note that FALSE is returned for single code point sequences if <valuep> is
10666 char * endbrace; /* '}' following the name */
10668 char *endchar; /* Points to '.' or '}' ending cur char in the input
10670 bool has_multiple_chars; /* true if the input stream contains a sequence of
10671 more than one character */
10673 GET_RE_DEBUG_FLAGS_DECL;
10675 PERL_ARGS_ASSERT_GROK_BSLASH_N;
10677 GET_RE_DEBUG_FLAGS;
10679 assert(cBOOL(node_p) ^ cBOOL(valuep)); /* Exactly one should be set */
10681 /* The [^\n] meaning of \N ignores spaces and comments under the /x
10682 * modifier. The other meaning does not, so use a temporary until we find
10683 * out which we are being called with */
10684 p = (RExC_flags & RXf_PMf_EXTENDED)
10685 ? regwhite( pRExC_state, RExC_parse )
10688 /* Disambiguate between \N meaning a named character versus \N meaning
10689 * [^\n]. The former is assumed when it can't be the latter. */
10690 if (*p != '{' || regcurly(p, FALSE)) {
10693 /* no bare \N allowed in a charclass */
10694 if (in_char_class) {
10695 vFAIL("\\N in a character class must be a named character: \\N{...}");
10699 RExC_parse--; /* Need to back off so nextchar() doesn't skip the
10701 nextchar(pRExC_state);
10702 *node_p = reg_node(pRExC_state, REG_ANY);
10703 *flagp |= HASWIDTH|SIMPLE;
10705 Set_Node_Length(*node_p, 1); /* MJD */
10709 /* Here, we have decided it should be a named character or sequence */
10711 /* The test above made sure that the next real character is a '{', but
10712 * under the /x modifier, it could be separated by space (or a comment and
10713 * \n) and this is not allowed (for consistency with \x{...} and the
10714 * tokenizer handling of \N{NAME}). */
10715 if (*RExC_parse != '{') {
10716 vFAIL("Missing braces on \\N{}");
10719 RExC_parse++; /* Skip past the '{' */
10721 if (! (endbrace = strchr(RExC_parse, '}')) /* no trailing brace */
10722 || ! (endbrace == RExC_parse /* nothing between the {} */
10723 || (endbrace - RExC_parse >= 2 /* U+ (bad hex is checked below
10725 && strnEQ(RExC_parse, "U+", 2)))) /* for a better error msg)
10728 if (endbrace) RExC_parse = endbrace; /* position msg's '<--HERE' */
10729 vFAIL("\\N{NAME} must be resolved by the lexer");
10732 if (endbrace == RExC_parse) { /* empty: \N{} */
10735 *node_p = reg_node(pRExC_state,NOTHING);
10737 else if (in_char_class) {
10738 if (SIZE_ONLY && in_char_class) {
10740 RExC_parse++; /* Position after the "}" */
10741 vFAIL("Zero length \\N{}");
10744 ckWARNreg(RExC_parse,
10745 "Ignoring zero length \\N{} in character class");
10753 nextchar(pRExC_state);
10757 RExC_uni_semantics = 1; /* Unicode named chars imply Unicode semantics */
10758 RExC_parse += 2; /* Skip past the 'U+' */
10760 endchar = RExC_parse + strcspn(RExC_parse, ".}");
10762 /* Code points are separated by dots. If none, there is only one code
10763 * point, and is terminated by the brace */
10764 has_multiple_chars = (endchar < endbrace);
10766 if (valuep && (! has_multiple_chars || in_char_class)) {
10767 /* We only pay attention to the first char of
10768 multichar strings being returned in char classes. I kinda wonder
10769 if this makes sense as it does change the behaviour
10770 from earlier versions, OTOH that behaviour was broken
10771 as well. XXX Solution is to recharacterize as
10772 [rest-of-class]|multi1|multi2... */
10774 STRLEN length_of_hex = (STRLEN)(endchar - RExC_parse);
10775 I32 grok_hex_flags = PERL_SCAN_ALLOW_UNDERSCORES
10776 | PERL_SCAN_DISALLOW_PREFIX
10777 | (SIZE_ONLY ? PERL_SCAN_SILENT_ILLDIGIT : 0);
10779 *valuep = grok_hex(RExC_parse, &length_of_hex, &grok_hex_flags, NULL);
10781 /* The tokenizer should have guaranteed validity, but it's possible to
10782 * bypass it by using single quoting, so check */
10783 if (length_of_hex == 0
10784 || length_of_hex != (STRLEN)(endchar - RExC_parse) )
10786 RExC_parse += length_of_hex; /* Includes all the valid */
10787 RExC_parse += (RExC_orig_utf8) /* point to after 1st invalid */
10788 ? UTF8SKIP(RExC_parse)
10790 /* Guard against malformed utf8 */
10791 if (RExC_parse >= endchar) {
10792 RExC_parse = endchar;
10794 vFAIL("Invalid hexadecimal number in \\N{U+...}");
10797 if (in_char_class && has_multiple_chars) {
10799 RExC_parse = endbrace;
10800 vFAIL("\\N{} in character class restricted to one character");
10803 ckWARNreg(endchar, "Using just the first character returned by \\N{} in character class");
10807 RExC_parse = endbrace + 1;
10809 else if (! node_p || ! has_multiple_chars) {
10811 /* Here, the input is legal, but not according to the caller's
10812 * options. We fail without advancing the parse, so that the
10813 * caller can try again */
10819 /* What is done here is to convert this to a sub-pattern of the form
10820 * (?:\x{char1}\x{char2}...)
10821 * and then call reg recursively. That way, it retains its atomicness,
10822 * while not having to worry about special handling that some code
10823 * points may have. toke.c has converted the original Unicode values
10824 * to native, so that we can just pass on the hex values unchanged. We
10825 * do have to set a flag to keep recoding from happening in the
10828 SV * substitute_parse = newSVpvn_flags("?:", 2, SVf_UTF8|SVs_TEMP);
10830 char *orig_end = RExC_end;
10833 while (RExC_parse < endbrace) {
10835 /* Convert to notation the rest of the code understands */
10836 sv_catpv(substitute_parse, "\\x{");
10837 sv_catpvn(substitute_parse, RExC_parse, endchar - RExC_parse);
10838 sv_catpv(substitute_parse, "}");
10840 /* Point to the beginning of the next character in the sequence. */
10841 RExC_parse = endchar + 1;
10842 endchar = RExC_parse + strcspn(RExC_parse, ".}");
10844 sv_catpv(substitute_parse, ")");
10846 RExC_parse = SvPV(substitute_parse, len);
10848 /* Don't allow empty number */
10850 vFAIL("Invalid hexadecimal number in \\N{U+...}");
10852 RExC_end = RExC_parse + len;
10854 /* The values are Unicode, and therefore not subject to recoding */
10855 RExC_override_recoding = 1;
10857 if (!(*node_p = reg(pRExC_state, 1, &flags, depth+1))) {
10858 if (flags & RESTART_UTF8) {
10859 *flagp = RESTART_UTF8;
10862 FAIL2("panic: reg returned NULL to grok_bslash_N, flags=%#"UVxf"",
10865 *flagp |= flags&(HASWIDTH|SPSTART|SIMPLE|POSTPONED);
10867 RExC_parse = endbrace;
10868 RExC_end = orig_end;
10869 RExC_override_recoding = 0;
10871 nextchar(pRExC_state);
10881 * It returns the code point in utf8 for the value in *encp.
10882 * value: a code value in the source encoding
10883 * encp: a pointer to an Encode object
10885 * If the result from Encode is not a single character,
10886 * it returns U+FFFD (Replacement character) and sets *encp to NULL.
10889 S_reg_recode(pTHX_ const char value, SV **encp)
10892 SV * const sv = newSVpvn_flags(&value, numlen, SVs_TEMP);
10893 const char * const s = *encp ? sv_recode_to_utf8(sv, *encp) : SvPVX(sv);
10894 const STRLEN newlen = SvCUR(sv);
10895 UV uv = UNICODE_REPLACEMENT;
10897 PERL_ARGS_ASSERT_REG_RECODE;
10901 ? utf8n_to_uvchr((U8*)s, newlen, &numlen, UTF8_ALLOW_DEFAULT)
10904 if (!newlen || numlen != newlen) {
10905 uv = UNICODE_REPLACEMENT;
10911 PERL_STATIC_INLINE U8
10912 S_compute_EXACTish(pTHX_ RExC_state_t *pRExC_state)
10916 PERL_ARGS_ASSERT_COMPUTE_EXACTISH;
10922 op = get_regex_charset(RExC_flags);
10923 if (op >= REGEX_ASCII_RESTRICTED_CHARSET) {
10924 op--; /* /a is same as /u, and map /aa's offset to what /a's would have
10925 been, so there is no hole */
10928 return op + EXACTF;
10931 PERL_STATIC_INLINE void
10932 S_alloc_maybe_populate_EXACT(pTHX_ RExC_state_t *pRExC_state,
10933 regnode *node, I32* flagp, STRLEN len, UV code_point,
10936 /* This knows the details about sizing an EXACTish node, setting flags for
10937 * it (by setting <*flagp>, and potentially populating it with a single
10940 * If <len> (the length in bytes) is non-zero, this function assumes that
10941 * the node has already been populated, and just does the sizing. In this
10942 * case <code_point> should be the final code point that has already been
10943 * placed into the node. This value will be ignored except that under some
10944 * circumstances <*flagp> is set based on it.
10946 * If <len> is zero, the function assumes that the node is to contain only
10947 * the single character given by <code_point> and calculates what <len>
10948 * should be. In pass 1, it sizes the node appropriately. In pass 2, it
10949 * additionally will populate the node's STRING with <code_point> or its
10952 * In both cases <*flagp> is appropriately set
10954 * It knows that under FOLD, the Latin Sharp S and UTF characters above
10955 * 255, must be folded (the former only when the rules indicate it can
10958 * When it does the populating, it looks at the flag 'downgradable'. If
10959 * true with a node that folds, it checks if the single code point
10960 * participates in a fold, and if not downgrades the node to an EXACT.
10961 * This helps the optimizer */
10963 bool len_passed_in = cBOOL(len != 0);
10964 U8 character[UTF8_MAXBYTES_CASE+1];
10966 PERL_ARGS_ASSERT_ALLOC_MAYBE_POPULATE_EXACT;
10968 /* Don't bother to check for downgrading in PASS1, as it doesn't make any
10969 * sizing difference, and is extra work that is thrown away */
10970 if (downgradable && ! PASS2) {
10971 downgradable = FALSE;
10974 if (! len_passed_in) {
10976 if (UNI_IS_INVARIANT(code_point)) {
10977 if (LOC || ! FOLD) { /* /l defers folding until runtime */
10978 *character = (U8) code_point;
10980 else { /* Here is /i and not /l (toFOLD() is defined on just
10981 ASCII, which isn't the same thing as INVARIANT on
10982 EBCDIC, but it works there, as the extra invariants
10983 fold to themselves) */
10984 *character = toFOLD((U8) code_point);
10986 /* We can downgrade to an EXACT node if this character
10987 * isn't a folding one. Note that this assumes that
10988 * nothing above Latin1 folds to some other invariant than
10989 * one of these alphabetics; otherwise we would also have
10991 * && (! HAS_NONLATIN1_FOLD_CLOSURE(code_point)
10992 * || ASCII_FOLD_RESTRICTED))
10994 if (downgradable && PL_fold[code_point] == code_point) {
11000 else if (FOLD && (! LOC
11001 || ! is_PROBLEMATIC_LOCALE_FOLD_cp(code_point)))
11002 { /* Folding, and ok to do so now */
11003 UV folded = _to_uni_fold_flags(
11007 FOLD_FLAGS_FULL | ((ASCII_FOLD_RESTRICTED)
11008 ? FOLD_FLAGS_NOMIX_ASCII
11011 && folded == code_point
11012 && ! _invlist_contains_cp(PL_utf8_foldable, code_point))
11017 else if (code_point <= MAX_UTF8_TWO_BYTE) {
11019 /* Not folding this cp, and can output it directly */
11020 *character = UTF8_TWO_BYTE_HI(code_point);
11021 *(character + 1) = UTF8_TWO_BYTE_LO(code_point);
11025 uvchr_to_utf8( character, code_point);
11026 len = UTF8SKIP(character);
11028 } /* Else pattern isn't UTF8. */
11030 *character = (U8) code_point;
11032 } /* Else is folded non-UTF8 */
11033 else if (LIKELY(code_point != LATIN_SMALL_LETTER_SHARP_S)) {
11035 /* We don't fold any non-UTF8 except possibly the Sharp s (see
11036 * comments at join_exact()); */
11037 *character = (U8) code_point;
11040 /* Can turn into an EXACT node if we know the fold at compile time,
11041 * and it folds to itself and doesn't particpate in other folds */
11044 && PL_fold_latin1[code_point] == code_point
11045 && (! HAS_NONLATIN1_FOLD_CLOSURE(code_point)
11046 || (isASCII(code_point) && ASCII_FOLD_RESTRICTED)))
11050 } /* else is Sharp s. May need to fold it */
11051 else if (AT_LEAST_UNI_SEMANTICS && ! ASCII_FOLD_RESTRICTED) {
11053 *(character + 1) = 's';
11057 *character = LATIN_SMALL_LETTER_SHARP_S;
11063 RExC_size += STR_SZ(len);
11066 RExC_emit += STR_SZ(len);
11067 STR_LEN(node) = len;
11068 if (! len_passed_in) {
11069 Copy((char *) character, STRING(node), len, char);
11073 *flagp |= HASWIDTH;
11075 /* A single character node is SIMPLE, except for the special-cased SHARP S
11077 if ((len == 1 || (UTF && len == UNISKIP(code_point)))
11078 && (code_point != LATIN_SMALL_LETTER_SHARP_S
11079 || ! FOLD || ! DEPENDS_SEMANTICS))
11084 /* The OP may not be well defined in PASS1 */
11085 if (PASS2 && OP(node) == EXACTFL) {
11086 RExC_contains_locale = 1;
11091 /* return atoi(p), unless it's too big to sensibly be a backref,
11092 * in which case return I32_MAX (rather than possibly 32-bit wrapping) */
11095 S_backref_value(char *p)
11099 for (;isDIGIT(*q); q++); /* calculate length of num */
11100 if (q - p == 0 || q - p > 9)
11107 - regatom - the lowest level
11109 Try to identify anything special at the start of the pattern. If there
11110 is, then handle it as required. This may involve generating a single regop,
11111 such as for an assertion; or it may involve recursing, such as to
11112 handle a () structure.
11114 If the string doesn't start with something special then we gobble up
11115 as much literal text as we can.
11117 Once we have been able to handle whatever type of thing started the
11118 sequence, we return.
11120 Note: we have to be careful with escapes, as they can be both literal
11121 and special, and in the case of \10 and friends, context determines which.
11123 A summary of the code structure is:
11125 switch (first_byte) {
11126 cases for each special:
11127 handle this special;
11130 switch (2nd byte) {
11131 cases for each unambiguous special:
11132 handle this special;
11134 cases for each ambigous special/literal:
11136 if (special) handle here
11138 default: // unambiguously literal:
11141 default: // is a literal char
11144 create EXACTish node for literal;
11145 while (more input and node isn't full) {
11146 switch (input_byte) {
11147 cases for each special;
11148 make sure parse pointer is set so that the next call to
11149 regatom will see this special first
11150 goto loopdone; // EXACTish node terminated by prev. char
11152 append char to EXACTISH node;
11154 get next input byte;
11158 return the generated node;
11160 Specifically there are two separate switches for handling
11161 escape sequences, with the one for handling literal escapes requiring
11162 a dummy entry for all of the special escapes that are actually handled
11165 Returns NULL, setting *flagp to TRYAGAIN if reg() returns NULL with
11167 Returns NULL, setting *flagp to RESTART_UTF8 if the sizing scan needs to be
11169 Otherwise does not return NULL.
11173 S_regatom(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth)
11176 regnode *ret = NULL;
11178 char *parse_start = RExC_parse;
11182 GET_RE_DEBUG_FLAGS_DECL;
11184 *flagp = WORST; /* Tentatively. */
11186 DEBUG_PARSE("atom");
11188 PERL_ARGS_ASSERT_REGATOM;
11191 switch ((U8)*RExC_parse) {
11193 RExC_seen_zerolen++;
11194 nextchar(pRExC_state);
11195 if (RExC_flags & RXf_PMf_MULTILINE)
11196 ret = reg_node(pRExC_state, MBOL);
11197 else if (RExC_flags & RXf_PMf_SINGLELINE)
11198 ret = reg_node(pRExC_state, SBOL);
11200 ret = reg_node(pRExC_state, BOL);
11201 Set_Node_Length(ret, 1); /* MJD */
11204 nextchar(pRExC_state);
11206 RExC_seen_zerolen++;
11207 if (RExC_flags & RXf_PMf_MULTILINE)
11208 ret = reg_node(pRExC_state, MEOL);
11209 else if (RExC_flags & RXf_PMf_SINGLELINE)
11210 ret = reg_node(pRExC_state, SEOL);
11212 ret = reg_node(pRExC_state, EOL);
11213 Set_Node_Length(ret, 1); /* MJD */
11216 nextchar(pRExC_state);
11217 if (RExC_flags & RXf_PMf_SINGLELINE)
11218 ret = reg_node(pRExC_state, SANY);
11220 ret = reg_node(pRExC_state, REG_ANY);
11221 *flagp |= HASWIDTH|SIMPLE;
11223 Set_Node_Length(ret, 1); /* MJD */
11227 char * const oregcomp_parse = ++RExC_parse;
11228 ret = regclass(pRExC_state, flagp,depth+1,
11229 FALSE, /* means parse the whole char class */
11230 TRUE, /* allow multi-char folds */
11231 FALSE, /* don't silence non-portable warnings. */
11233 if (*RExC_parse != ']') {
11234 RExC_parse = oregcomp_parse;
11235 vFAIL("Unmatched [");
11238 if (*flagp & RESTART_UTF8)
11240 FAIL2("panic: regclass returned NULL to regatom, flags=%#"UVxf"",
11243 nextchar(pRExC_state);
11244 Set_Node_Length(ret, RExC_parse - oregcomp_parse + 1); /* MJD */
11248 nextchar(pRExC_state);
11249 ret = reg(pRExC_state, 2, &flags,depth+1);
11251 if (flags & TRYAGAIN) {
11252 if (RExC_parse == RExC_end) {
11253 /* Make parent create an empty node if needed. */
11254 *flagp |= TRYAGAIN;
11259 if (flags & RESTART_UTF8) {
11260 *flagp = RESTART_UTF8;
11263 FAIL2("panic: reg returned NULL to regatom, flags=%#"UVxf"",
11266 *flagp |= flags&(HASWIDTH|SPSTART|SIMPLE|POSTPONED);
11270 if (flags & TRYAGAIN) {
11271 *flagp |= TRYAGAIN;
11274 vFAIL("Internal urp");
11275 /* Supposed to be caught earlier. */
11278 if (!regcurly(RExC_parse, FALSE)) {
11287 vFAIL("Quantifier follows nothing");
11292 This switch handles escape sequences that resolve to some kind
11293 of special regop and not to literal text. Escape sequnces that
11294 resolve to literal text are handled below in the switch marked
11297 Every entry in this switch *must* have a corresponding entry
11298 in the literal escape switch. However, the opposite is not
11299 required, as the default for this switch is to jump to the
11300 literal text handling code.
11302 switch ((U8)*++RExC_parse) {
11304 /* Special Escapes */
11306 RExC_seen_zerolen++;
11307 ret = reg_node(pRExC_state, SBOL);
11309 goto finish_meta_pat;
11311 ret = reg_node(pRExC_state, GPOS);
11312 RExC_seen |= REG_GPOS_SEEN;
11314 goto finish_meta_pat;
11316 RExC_seen_zerolen++;
11317 ret = reg_node(pRExC_state, KEEPS);
11319 /* XXX:dmq : disabling in-place substitution seems to
11320 * be necessary here to avoid cases of memory corruption, as
11321 * with: C<$_="x" x 80; s/x\K/y/> -- rgs
11323 RExC_seen |= REG_LOOKBEHIND_SEEN;
11324 goto finish_meta_pat;
11326 ret = reg_node(pRExC_state, SEOL);
11328 RExC_seen_zerolen++; /* Do not optimize RE away */
11329 goto finish_meta_pat;
11331 ret = reg_node(pRExC_state, EOS);
11333 RExC_seen_zerolen++; /* Do not optimize RE away */
11334 goto finish_meta_pat;
11336 ret = reg_node(pRExC_state, CANY);
11337 RExC_seen |= REG_CANY_SEEN;
11338 *flagp |= HASWIDTH|SIMPLE;
11339 goto finish_meta_pat;
11341 ret = reg_node(pRExC_state, CLUMP);
11342 *flagp |= HASWIDTH;
11343 goto finish_meta_pat;
11349 arg = ANYOF_WORDCHAR;
11353 RExC_seen_zerolen++;
11354 RExC_seen |= REG_LOOKBEHIND_SEEN;
11355 op = BOUND + get_regex_charset(RExC_flags);
11356 if (op > BOUNDA) { /* /aa is same as /a */
11359 else if (op == BOUNDL) {
11360 RExC_contains_locale = 1;
11362 ret = reg_node(pRExC_state, op);
11363 FLAGS(ret) = get_regex_charset(RExC_flags);
11365 if (! SIZE_ONLY && (U8) *(RExC_parse + 1) == '{') {
11366 /* diag_listed_as: Use "%s" instead of "%s" */
11367 vFAIL("Use \"\\b\\{\" instead of \"\\b{\"");
11369 goto finish_meta_pat;
11371 RExC_seen_zerolen++;
11372 RExC_seen |= REG_LOOKBEHIND_SEEN;
11373 op = NBOUND + get_regex_charset(RExC_flags);
11374 if (op > NBOUNDA) { /* /aa is same as /a */
11377 else if (op == NBOUNDL) {
11378 RExC_contains_locale = 1;
11380 ret = reg_node(pRExC_state, op);
11381 FLAGS(ret) = get_regex_charset(RExC_flags);
11383 if (! SIZE_ONLY && (U8) *(RExC_parse + 1) == '{') {
11384 /* diag_listed_as: Use "%s" instead of "%s" */
11385 vFAIL("Use \"\\B\\{\" instead of \"\\B{\"");
11387 goto finish_meta_pat;
11397 ret = reg_node(pRExC_state, LNBREAK);
11398 *flagp |= HASWIDTH|SIMPLE;
11399 goto finish_meta_pat;
11407 goto join_posix_op_known;
11413 arg = ANYOF_VERTWS;
11415 goto join_posix_op_known;
11425 op = POSIXD + get_regex_charset(RExC_flags);
11426 if (op > POSIXA) { /* /aa is same as /a */
11429 else if (op == POSIXL) {
11430 RExC_contains_locale = 1;
11433 join_posix_op_known:
11436 op += NPOSIXD - POSIXD;
11439 ret = reg_node(pRExC_state, op);
11441 FLAGS(ret) = namedclass_to_classnum(arg);
11444 *flagp |= HASWIDTH|SIMPLE;
11448 nextchar(pRExC_state);
11449 Set_Node_Length(ret, 2); /* MJD */
11455 char* parse_start = RExC_parse - 2;
11460 ret = regclass(pRExC_state, flagp,depth+1,
11461 TRUE, /* means just parse this element */
11462 FALSE, /* don't allow multi-char folds */
11463 FALSE, /* don't silence non-portable warnings.
11464 It would be a bug if these returned
11467 /* regclass() can only return RESTART_UTF8 if multi-char folds
11470 FAIL2("panic: regclass returned NULL to regatom, flags=%#"UVxf"",
11475 Set_Node_Offset(ret, parse_start + 2);
11476 Set_Node_Cur_Length(ret, parse_start);
11477 nextchar(pRExC_state);
11481 /* Handle \N and \N{NAME} with multiple code points here and not
11482 * below because it can be multicharacter. join_exact() will join
11483 * them up later on. Also this makes sure that things like
11484 * /\N{BLAH}+/ and \N{BLAH} being multi char Just Happen. dmq.
11485 * The options to the grok function call causes it to fail if the
11486 * sequence is just a single code point. We then go treat it as
11487 * just another character in the current EXACT node, and hence it
11488 * gets uniform treatment with all the other characters. The
11489 * special treatment for quantifiers is not needed for such single
11490 * character sequences */
11492 if (! grok_bslash_N(pRExC_state, &ret, NULL, flagp, depth, FALSE,
11493 FALSE /* not strict */ )) {
11494 if (*flagp & RESTART_UTF8)
11500 case 'k': /* Handle \k<NAME> and \k'NAME' */
11503 char ch= RExC_parse[1];
11504 if (ch != '<' && ch != '\'' && ch != '{') {
11506 /* diag_listed_as: Sequence \%s... not terminated in regex; marked by <-- HERE in m/%s/ */
11507 vFAIL2("Sequence %.2s... not terminated",parse_start);
11509 /* this pretty much dupes the code for (?P=...) in reg(), if
11510 you change this make sure you change that */
11511 char* name_start = (RExC_parse += 2);
11513 SV *sv_dat = reg_scan_name(pRExC_state,
11514 SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
11515 ch= (ch == '<') ? '>' : (ch == '{') ? '}' : '\'';
11516 if (RExC_parse == name_start || *RExC_parse != ch)
11517 /* diag_listed_as: Sequence \%s... not terminated in regex; marked by <-- HERE in m/%s/ */
11518 vFAIL2("Sequence %.3s... not terminated",parse_start);
11521 num = add_data( pRExC_state, STR_WITH_LEN("S"));
11522 RExC_rxi->data->data[num]=(void*)sv_dat;
11523 SvREFCNT_inc_simple_void(sv_dat);
11527 ret = reganode(pRExC_state,
11530 : (ASCII_FOLD_RESTRICTED)
11532 : (AT_LEAST_UNI_SEMANTICS)
11538 *flagp |= HASWIDTH;
11540 /* override incorrect value set in reganode MJD */
11541 Set_Node_Offset(ret, parse_start+1);
11542 Set_Node_Cur_Length(ret, parse_start);
11543 nextchar(pRExC_state);
11549 case '1': case '2': case '3': case '4':
11550 case '5': case '6': case '7': case '8': case '9':
11555 if (*RExC_parse == 'g') {
11559 if (*RExC_parse == '{') {
11563 if (*RExC_parse == '-') {
11567 if (hasbrace && !isDIGIT(*RExC_parse)) {
11568 if (isrel) RExC_parse--;
11570 goto parse_named_seq;
11573 num = S_backref_value(RExC_parse);
11575 vFAIL("Reference to invalid group 0");
11576 else if (num == I32_MAX) {
11577 if (isDIGIT(*RExC_parse))
11578 vFAIL("Reference to nonexistent group");
11580 vFAIL("Unterminated \\g... pattern");
11584 num = RExC_npar - num;
11586 vFAIL("Reference to nonexistent or unclosed group");
11590 num = S_backref_value(RExC_parse);
11591 /* bare \NNN might be backref or octal - if it is larger than or equal
11592 * RExC_npar then it is assumed to be and octal escape.
11593 * Note RExC_npar is +1 from the actual number of parens*/
11594 if (num == I32_MAX || (num > 9 && num >= RExC_npar
11595 && *RExC_parse != '8' && *RExC_parse != '9'))
11597 /* Probably a character specified in octal, e.g. \35 */
11602 /* at this point RExC_parse definitely points to a backref
11605 #ifdef RE_TRACK_PATTERN_OFFSETS
11606 char * const parse_start = RExC_parse - 1; /* MJD */
11608 while (isDIGIT(*RExC_parse))
11611 if (*RExC_parse != '}')
11612 vFAIL("Unterminated \\g{...} pattern");
11616 if (num > (I32)RExC_rx->nparens)
11617 vFAIL("Reference to nonexistent group");
11620 ret = reganode(pRExC_state,
11623 : (ASCII_FOLD_RESTRICTED)
11625 : (AT_LEAST_UNI_SEMANTICS)
11631 *flagp |= HASWIDTH;
11633 /* override incorrect value set in reganode MJD */
11634 Set_Node_Offset(ret, parse_start+1);
11635 Set_Node_Cur_Length(ret, parse_start);
11637 nextchar(pRExC_state);
11642 if (RExC_parse >= RExC_end)
11643 FAIL("Trailing \\");
11646 /* Do not generate "unrecognized" warnings here, we fall
11647 back into the quick-grab loop below */
11654 if (RExC_flags & RXf_PMf_EXTENDED) {
11655 if ( reg_skipcomment( pRExC_state ) )
11662 parse_start = RExC_parse - 1;
11671 #define MAX_NODE_STRING_SIZE 127
11672 char foldbuf[MAX_NODE_STRING_SIZE+UTF8_MAXBYTES_CASE];
11674 U8 upper_parse = MAX_NODE_STRING_SIZE;
11675 U8 node_type = compute_EXACTish(pRExC_state);
11676 bool next_is_quantifier;
11677 char * oldp = NULL;
11679 /* We can convert EXACTF nodes to EXACTFU if they contain only
11680 * characters that match identically regardless of the target
11681 * string's UTF8ness. The reason to do this is that EXACTF is not
11682 * trie-able, EXACTFU is.
11684 * Similarly, we can convert EXACTFL nodes to EXACTFU if they
11685 * contain only above-Latin1 characters (hence must be in UTF8),
11686 * which don't participate in folds with Latin1-range characters,
11687 * as the latter's folds aren't known until runtime. (We don't
11688 * need to figure this out until pass 2) */
11689 bool maybe_exactfu = PASS2
11690 && (node_type == EXACTF || node_type == EXACTFL);
11692 /* If a folding node contains only code points that don't
11693 * participate in folds, it can be changed into an EXACT node,
11694 * which allows the optimizer more things to look for */
11697 ret = reg_node(pRExC_state, node_type);
11699 /* In pass1, folded, we use a temporary buffer instead of the
11700 * actual node, as the node doesn't exist yet */
11701 s = (SIZE_ONLY && FOLD) ? foldbuf : STRING(ret);
11707 /* We do the EXACTFish to EXACT node only if folding. (And we
11708 * don't need to figure this out until pass 2) */
11709 maybe_exact = FOLD && PASS2;
11711 /* XXX The node can hold up to 255 bytes, yet this only goes to
11712 * 127. I (khw) do not know why. Keeping it somewhat less than
11713 * 255 allows us to not have to worry about overflow due to
11714 * converting to utf8 and fold expansion, but that value is
11715 * 255-UTF8_MAXBYTES_CASE. join_exact() may join adjacent nodes
11716 * split up by this limit into a single one using the real max of
11717 * 255. Even at 127, this breaks under rare circumstances. If
11718 * folding, we do not want to split a node at a character that is a
11719 * non-final in a multi-char fold, as an input string could just
11720 * happen to want to match across the node boundary. The join
11721 * would solve that problem if the join actually happens. But a
11722 * series of more than two nodes in a row each of 127 would cause
11723 * the first join to succeed to get to 254, but then there wouldn't
11724 * be room for the next one, which could at be one of those split
11725 * multi-char folds. I don't know of any fool-proof solution. One
11726 * could back off to end with only a code point that isn't such a
11727 * non-final, but it is possible for there not to be any in the
11729 for (p = RExC_parse - 1;
11730 len < upper_parse && p < RExC_end;
11735 if (RExC_flags & RXf_PMf_EXTENDED)
11736 p = regwhite( pRExC_state, p );
11747 /* Literal Escapes Switch
11749 This switch is meant to handle escape sequences that
11750 resolve to a literal character.
11752 Every escape sequence that represents something
11753 else, like an assertion or a char class, is handled
11754 in the switch marked 'Special Escapes' above in this
11755 routine, but also has an entry here as anything that
11756 isn't explicitly mentioned here will be treated as
11757 an unescaped equivalent literal.
11760 switch ((U8)*++p) {
11761 /* These are all the special escapes. */
11762 case 'A': /* Start assertion */
11763 case 'b': case 'B': /* Word-boundary assertion*/
11764 case 'C': /* Single char !DANGEROUS! */
11765 case 'd': case 'D': /* digit class */
11766 case 'g': case 'G': /* generic-backref, pos assertion */
11767 case 'h': case 'H': /* HORIZWS */
11768 case 'k': case 'K': /* named backref, keep marker */
11769 case 'p': case 'P': /* Unicode property */
11770 case 'R': /* LNBREAK */
11771 case 's': case 'S': /* space class */
11772 case 'v': case 'V': /* VERTWS */
11773 case 'w': case 'W': /* word class */
11774 case 'X': /* eXtended Unicode "combining
11775 character sequence" */
11776 case 'z': case 'Z': /* End of line/string assertion */
11780 /* Anything after here is an escape that resolves to a
11781 literal. (Except digits, which may or may not)
11787 case 'N': /* Handle a single-code point named character. */
11788 /* The options cause it to fail if a multiple code
11789 * point sequence. Handle those in the switch() above
11791 RExC_parse = p + 1;
11792 if (! grok_bslash_N(pRExC_state, NULL, &ender,
11793 flagp, depth, FALSE,
11794 FALSE /* not strict */ ))
11796 if (*flagp & RESTART_UTF8)
11797 FAIL("panic: grok_bslash_N set RESTART_UTF8");
11798 RExC_parse = p = oldp;
11802 if (ender > 0xff) {
11819 ender = ASCII_TO_NATIVE('\033');
11829 const char* error_msg;
11831 bool valid = grok_bslash_o(&p,
11834 TRUE, /* out warnings */
11835 FALSE, /* not strict */
11836 TRUE, /* Output warnings
11841 RExC_parse = p; /* going to die anyway; point
11842 to exact spot of failure */
11846 if (PL_encoding && ender < 0x100) {
11847 goto recode_encoding;
11849 if (ender > 0xff) {
11856 UV result = UV_MAX; /* initialize to erroneous
11858 const char* error_msg;
11860 bool valid = grok_bslash_x(&p,
11863 TRUE, /* out warnings */
11864 FALSE, /* not strict */
11865 TRUE, /* Output warnings
11870 RExC_parse = p; /* going to die anyway; point
11871 to exact spot of failure */
11876 if (PL_encoding && ender < 0x100) {
11877 goto recode_encoding;
11879 if (ender > 0xff) {
11886 ender = grok_bslash_c(*p++, SIZE_ONLY);
11888 case '8': case '9': /* must be a backreference */
11891 case '1': case '2': case '3':case '4':
11892 case '5': case '6': case '7':
11893 /* When we parse backslash escapes there is ambiguity
11894 * between backreferences and octal escapes. Any escape
11895 * from \1 - \9 is a backreference, any multi-digit
11896 * escape which does not start with 0 and which when
11897 * evaluated as decimal could refer to an already
11898 * parsed capture buffer is a backslash. Anything else
11901 * Note this implies that \118 could be interpreted as
11902 * 118 OR as "\11" . "8" depending on whether there
11903 * were 118 capture buffers defined already in the
11906 /* NOTE, RExC_npar is 1 more than the actual number of
11907 * parens we have seen so far, hence the < RExC_npar below. */
11909 if ( !isDIGIT(p[1]) || S_backref_value(p) < RExC_npar)
11910 { /* Not to be treated as an octal constant, go
11917 I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
11919 ender = grok_oct(p, &numlen, &flags, NULL);
11920 if (ender > 0xff) {
11924 if (SIZE_ONLY /* like \08, \178 */
11927 && isDIGIT(*p) && ckWARN(WARN_REGEXP))
11929 reg_warn_non_literal_string(
11931 form_short_octal_warning(p, numlen));
11934 if (PL_encoding && ender < 0x100)
11935 goto recode_encoding;
11938 if (! RExC_override_recoding) {
11939 SV* enc = PL_encoding;
11940 ender = reg_recode((const char)(U8)ender, &enc);
11941 if (!enc && SIZE_ONLY)
11942 ckWARNreg(p, "Invalid escape in the specified encoding");
11948 FAIL("Trailing \\");
11951 if (!SIZE_ONLY&& isALPHANUMERIC(*p)) {
11952 /* Include any { following the alpha to emphasize
11953 * that it could be part of an escape at some point
11955 int len = (isALPHA(*p) && *(p + 1) == '{') ? 2 : 1;
11956 ckWARN3reg(p + len, "Unrecognized escape \\%.*s passed through", len, p);
11958 goto normal_default;
11959 } /* End of switch on '\' */
11961 default: /* A literal character */
11964 && RExC_flags & RXf_PMf_EXTENDED
11965 && ckWARN_d(WARN_DEPRECATED)
11966 && is_PATWS_non_low_safe(p, RExC_end, UTF))
11968 vWARN_dep(p + ((UTF) ? UTF8SKIP(p) : 1),
11969 "Escape literal pattern white space under /x");
11973 if (UTF8_IS_START(*p) && UTF) {
11975 ender = utf8n_to_uvchr((U8*)p, RExC_end - p,
11976 &numlen, UTF8_ALLOW_DEFAULT);
11982 } /* End of switch on the literal */
11984 /* Here, have looked at the literal character and <ender>
11985 * contains its ordinal, <p> points to the character after it
11988 if ( RExC_flags & RXf_PMf_EXTENDED)
11989 p = regwhite( pRExC_state, p );
11991 /* If the next thing is a quantifier, it applies to this
11992 * character only, which means that this character has to be in
11993 * its own node and can't just be appended to the string in an
11994 * existing node, so if there are already other characters in
11995 * the node, close the node with just them, and set up to do
11996 * this character again next time through, when it will be the
11997 * only thing in its new node */
11998 if ((next_is_quantifier = (p < RExC_end && ISMULT2(p))) && len)
12004 if (! FOLD /* The simple case, just append the literal */
12005 || (LOC /* Also don't fold for tricky chars under /l */
12006 && is_PROBLEMATIC_LOCALE_FOLD_cp(ender)))
12010 /* Normally, we don't need the representation of the
12011 * character in the sizing pass--just its size, but if
12012 * folding, we have to actually put the character out
12013 * even in the sizing pass, because the size could
12014 * change as we juggle things at the end of this loop
12015 * to avoid splitting a too-full node in the middle of
12016 * a potential multi-char fold [perl #123539] */
12017 const STRLEN unilen = (SIZE_ONLY && ! FOLD)
12019 : (uvchr_to_utf8((U8*)s, ender) - (U8*)s);
12025 /* The loop increments <len> each time, as all but this
12026 * path (and one other) through it add a single byte to
12027 * the EXACTish node. But this one has changed len to
12028 * be the correct final value, so subtract one to
12029 * cancel out the increment that follows */
12033 /* See comment above for [perl #123539] */
12034 *(s++) = (char) ender;
12037 REGC((char)ender, s++);
12040 /* Can get here if folding only if is one of the /l
12041 * characters whose fold depends on the locale. The
12042 * occurrence of any of these indicate that we can't
12043 * simplify things */
12045 maybe_exact = FALSE;
12046 maybe_exactfu = FALSE;
12051 /* See comments for join_exact() as to why we fold this
12052 * non-UTF at compile time */
12053 || (node_type == EXACTFU
12054 && ender == LATIN_SMALL_LETTER_SHARP_S)))
12056 /* Here, are folding and are not UTF-8 encoded; therefore
12057 * the character must be in the range 0-255, and is not /l
12058 * (Not /l because we already handled these under /l in
12059 * is_PROBLEMATIC_LOCALE_FOLD_cp */
12060 if (IS_IN_SOME_FOLD_L1(ender)) {
12061 maybe_exact = FALSE;
12063 /* See if the character's fold differs between /d and
12064 * /u. This includes the multi-char fold SHARP S to
12067 && (PL_fold[ender] != PL_fold_latin1[ender]
12068 || ender == LATIN_SMALL_LETTER_SHARP_S
12070 && isARG2_lower_or_UPPER_ARG1('s', ender)
12071 && isARG2_lower_or_UPPER_ARG1('s',
12074 maybe_exactfu = FALSE;
12078 /* Even when folding, we store just the input character, as
12079 * we have an array that finds its fold quickly */
12080 *(s++) = (char) ender;
12082 else { /* FOLD and UTF */
12083 /* Unlike the non-fold case, we do actually have to
12084 * calculate the results here in pass 1. This is for two
12085 * reasons, the folded length may be longer than the
12086 * unfolded, and we have to calculate how many EXACTish
12087 * nodes it will take; and we may run out of room in a node
12088 * in the middle of a potential multi-char fold, and have
12089 * to back off accordingly. (Hence we can't use REGC for
12090 * the simple case just below.) */
12093 if (isASCII(ender)) {
12094 folded = toFOLD(ender);
12095 *(s)++ = (U8) folded;
12100 folded = _to_uni_fold_flags(
12104 FOLD_FLAGS_FULL | ((ASCII_FOLD_RESTRICTED)
12105 ? FOLD_FLAGS_NOMIX_ASCII
12109 /* The loop increments <len> each time, as all but this
12110 * path (and one other) through it add a single byte to
12111 * the EXACTish node. But this one has changed len to
12112 * be the correct final value, so subtract one to
12113 * cancel out the increment that follows */
12114 len += foldlen - 1;
12116 /* If this node only contains non-folding code points so
12117 * far, see if this new one is also non-folding */
12119 if (folded != ender) {
12120 maybe_exact = FALSE;
12123 /* Here the fold is the original; we have to check
12124 * further to see if anything folds to it */
12125 if (_invlist_contains_cp(PL_utf8_foldable,
12128 maybe_exact = FALSE;
12135 if (next_is_quantifier) {
12137 /* Here, the next input is a quantifier, and to get here,
12138 * the current character is the only one in the node.
12139 * Also, here <len> doesn't include the final byte for this
12145 } /* End of loop through literal characters */
12147 /* Here we have either exhausted the input or ran out of room in
12148 * the node. (If we encountered a character that can't be in the
12149 * node, transfer is made directly to <loopdone>, and so we
12150 * wouldn't have fallen off the end of the loop.) In the latter
12151 * case, we artificially have to split the node into two, because
12152 * we just don't have enough space to hold everything. This
12153 * creates a problem if the final character participates in a
12154 * multi-character fold in the non-final position, as a match that
12155 * should have occurred won't, due to the way nodes are matched,
12156 * and our artificial boundary. So back off until we find a non-
12157 * problematic character -- one that isn't at the beginning or
12158 * middle of such a fold. (Either it doesn't participate in any
12159 * folds, or appears only in the final position of all the folds it
12160 * does participate in.) A better solution with far fewer false
12161 * positives, and that would fill the nodes more completely, would
12162 * be to actually have available all the multi-character folds to
12163 * test against, and to back-off only far enough to be sure that
12164 * this node isn't ending with a partial one. <upper_parse> is set
12165 * further below (if we need to reparse the node) to include just
12166 * up through that final non-problematic character that this code
12167 * identifies, so when it is set to less than the full node, we can
12168 * skip the rest of this */
12169 if (FOLD && p < RExC_end && upper_parse == MAX_NODE_STRING_SIZE) {
12171 const STRLEN full_len = len;
12173 assert(len >= MAX_NODE_STRING_SIZE);
12175 /* Here, <s> points to the final byte of the final character.
12176 * Look backwards through the string until find a non-
12177 * problematic character */
12181 /* This has no multi-char folds to non-UTF characters */
12182 if (ASCII_FOLD_RESTRICTED) {
12186 while (--s >= s0 && IS_NON_FINAL_FOLD(*s)) { }
12190 if (! PL_NonL1NonFinalFold) {
12191 PL_NonL1NonFinalFold = _new_invlist_C_array(
12192 NonL1_Perl_Non_Final_Folds_invlist);
12195 /* Point to the first byte of the final character */
12196 s = (char *) utf8_hop((U8 *) s, -1);
12198 while (s >= s0) { /* Search backwards until find
12199 non-problematic char */
12200 if (UTF8_IS_INVARIANT(*s)) {
12202 /* There are no ascii characters that participate
12203 * in multi-char folds under /aa. In EBCDIC, the
12204 * non-ascii invariants are all control characters,
12205 * so don't ever participate in any folds. */
12206 if (ASCII_FOLD_RESTRICTED
12207 || ! IS_NON_FINAL_FOLD(*s))
12212 else if (UTF8_IS_DOWNGRADEABLE_START(*s)) {
12213 if (! IS_NON_FINAL_FOLD(TWO_BYTE_UTF8_TO_NATIVE(
12219 else if (! _invlist_contains_cp(
12220 PL_NonL1NonFinalFold,
12221 valid_utf8_to_uvchr((U8 *) s, NULL)))
12226 /* Here, the current character is problematic in that
12227 * it does occur in the non-final position of some
12228 * fold, so try the character before it, but have to
12229 * special case the very first byte in the string, so
12230 * we don't read outside the string */
12231 s = (s == s0) ? s -1 : (char *) utf8_hop((U8 *) s, -1);
12232 } /* End of loop backwards through the string */
12234 /* If there were only problematic characters in the string,
12235 * <s> will point to before s0, in which case the length
12236 * should be 0, otherwise include the length of the
12237 * non-problematic character just found */
12238 len = (s < s0) ? 0 : s - s0 + UTF8SKIP(s);
12241 /* Here, have found the final character, if any, that is
12242 * non-problematic as far as ending the node without splitting
12243 * it across a potential multi-char fold. <len> contains the
12244 * number of bytes in the node up-to and including that
12245 * character, or is 0 if there is no such character, meaning
12246 * the whole node contains only problematic characters. In
12247 * this case, give up and just take the node as-is. We can't
12252 /* If the node ends in an 's' we make sure it stays EXACTF,
12253 * as if it turns into an EXACTFU, it could later get
12254 * joined with another 's' that would then wrongly match
12256 if (maybe_exactfu && isARG2_lower_or_UPPER_ARG1('s', ender))
12258 maybe_exactfu = FALSE;
12262 /* Here, the node does contain some characters that aren't
12263 * problematic. If one such is the final character in the
12264 * node, we are done */
12265 if (len == full_len) {
12268 else if (len + ((UTF) ? UTF8SKIP(s) : 1) == full_len) {
12270 /* If the final character is problematic, but the
12271 * penultimate is not, back-off that last character to
12272 * later start a new node with it */
12277 /* Here, the final non-problematic character is earlier
12278 * in the input than the penultimate character. What we do
12279 * is reparse from the beginning, going up only as far as
12280 * this final ok one, thus guaranteeing that the node ends
12281 * in an acceptable character. The reason we reparse is
12282 * that we know how far in the character is, but we don't
12283 * know how to correlate its position with the input parse.
12284 * An alternate implementation would be to build that
12285 * correlation as we go along during the original parse,
12286 * but that would entail extra work for every node, whereas
12287 * this code gets executed only when the string is too
12288 * large for the node, and the final two characters are
12289 * problematic, an infrequent occurrence. Yet another
12290 * possible strategy would be to save the tail of the
12291 * string, and the next time regatom is called, initialize
12292 * with that. The problem with this is that unless you
12293 * back off one more character, you won't be guaranteed
12294 * regatom will get called again, unless regbranch,
12295 * regpiece ... are also changed. If you do back off that
12296 * extra character, so that there is input guaranteed to
12297 * force calling regatom, you can't handle the case where
12298 * just the first character in the node is acceptable. I
12299 * (khw) decided to try this method which doesn't have that
12300 * pitfall; if performance issues are found, we can do a
12301 * combination of the current approach plus that one */
12307 } /* End of verifying node ends with an appropriate char */
12309 loopdone: /* Jumped to when encounters something that shouldn't be in
12312 /* I (khw) don't know if you can get here with zero length, but the
12313 * old code handled this situation by creating a zero-length EXACT
12314 * node. Might as well be NOTHING instead */
12320 /* If 'maybe_exact' is still set here, means there are no
12321 * code points in the node that participate in folds;
12322 * similarly for 'maybe_exactfu' and code points that match
12323 * differently depending on UTF8ness of the target string
12324 * (for /u), or depending on locale for /l */
12328 else if (maybe_exactfu) {
12332 alloc_maybe_populate_EXACT(pRExC_state, ret, flagp, len, ender,
12333 FALSE /* Don't look to see if could
12334 be turned into an EXACT
12335 node, as we have already
12340 RExC_parse = p - 1;
12341 Set_Node_Cur_Length(ret, parse_start);
12342 nextchar(pRExC_state);
12344 /* len is STRLEN which is unsigned, need to copy to signed */
12347 vFAIL("Internal disaster");
12350 } /* End of label 'defchar:' */
12352 } /* End of giant switch on input character */
12358 S_regwhite( RExC_state_t *pRExC_state, char *p )
12360 const char *e = RExC_end;
12362 PERL_ARGS_ASSERT_REGWHITE;
12367 else if (*p == '#') {
12370 if (*p++ == '\n') {
12376 RExC_seen |= REG_RUN_ON_COMMENT_SEEN;
12385 S_regpatws( RExC_state_t *pRExC_state, char *p , const bool recognize_comment )
12387 /* Returns the next non-pattern-white space, non-comment character (the
12388 * latter only if 'recognize_comment is true) in the string p, which is
12389 * ended by RExC_end. If there is no line break ending a comment,
12390 * RExC_seen has added the REG_RUN_ON_COMMENT_SEEN flag; */
12391 const char *e = RExC_end;
12393 PERL_ARGS_ASSERT_REGPATWS;
12397 if ((len = is_PATWS_safe(p, e, UTF))) {
12400 else if (recognize_comment && *p == '#') {
12404 if (is_LNBREAK_safe(p, e, UTF)) {
12410 RExC_seen |= REG_RUN_ON_COMMENT_SEEN;
12419 S_populate_ANYOF_from_invlist(pTHX_ regnode *node, SV** invlist_ptr)
12421 /* Uses the inversion list '*invlist_ptr' to populate the ANYOF 'node'. It
12422 * sets up the bitmap and any flags, removing those code points from the
12423 * inversion list, setting it to NULL should it become completely empty */
12425 PERL_ARGS_ASSERT_POPULATE_ANYOF_FROM_INVLIST;
12426 assert(PL_regkind[OP(node)] == ANYOF);
12428 ANYOF_BITMAP_ZERO(node);
12429 if (*invlist_ptr) {
12431 /* This gets set if we actually need to modify things */
12432 bool change_invlist = FALSE;
12436 /* Start looking through *invlist_ptr */
12437 invlist_iterinit(*invlist_ptr);
12438 while (invlist_iternext(*invlist_ptr, &start, &end)) {
12442 if (end == UV_MAX && start <= 256) {
12443 ANYOF_FLAGS(node) |= ANYOF_ABOVE_LATIN1_ALL;
12445 else if (end >= 256) {
12446 ANYOF_FLAGS(node) |= ANYOF_UTF8;
12449 /* Quit if are above what we should change */
12454 change_invlist = TRUE;
12456 /* Set all the bits in the range, up to the max that we are doing */
12457 high = (end < 255) ? end : 255;
12458 for (i = start; i <= (int) high; i++) {
12459 if (! ANYOF_BITMAP_TEST(node, i)) {
12460 ANYOF_BITMAP_SET(node, i);
12464 invlist_iterfinish(*invlist_ptr);
12466 /* Done with loop; remove any code points that are in the bitmap from
12467 * *invlist_ptr; similarly for code points above latin1 if we have a
12468 * flag to match all of them anyways */
12469 if (change_invlist) {
12470 _invlist_subtract(*invlist_ptr, PL_Latin1, invlist_ptr);
12472 if (ANYOF_FLAGS(node) & ANYOF_ABOVE_LATIN1_ALL) {
12473 _invlist_intersection(*invlist_ptr, PL_Latin1, invlist_ptr);
12476 /* If have completely emptied it, remove it completely */
12477 if (_invlist_len(*invlist_ptr) == 0) {
12478 SvREFCNT_dec_NN(*invlist_ptr);
12479 *invlist_ptr = NULL;
12484 /* Parse POSIX character classes: [[:foo:]], [[=foo=]], [[.foo.]].
12485 Character classes ([:foo:]) can also be negated ([:^foo:]).
12486 Returns a named class id (ANYOF_XXX) if successful, -1 otherwise.
12487 Equivalence classes ([=foo=]) and composites ([.foo.]) are parsed,
12488 but trigger failures because they are currently unimplemented. */
12490 #define POSIXCC_DONE(c) ((c) == ':')
12491 #define POSIXCC_NOTYET(c) ((c) == '=' || (c) == '.')
12492 #define POSIXCC(c) (POSIXCC_DONE(c) || POSIXCC_NOTYET(c))
12494 PERL_STATIC_INLINE I32
12495 S_regpposixcc(pTHX_ RExC_state_t *pRExC_state, I32 value, const bool strict)
12498 I32 namedclass = OOB_NAMEDCLASS;
12500 PERL_ARGS_ASSERT_REGPPOSIXCC;
12502 if (value == '[' && RExC_parse + 1 < RExC_end &&
12503 /* I smell either [: or [= or [. -- POSIX has been here, right? */
12504 POSIXCC(UCHARAT(RExC_parse)))
12506 const char c = UCHARAT(RExC_parse);
12507 char* const s = RExC_parse++;
12509 while (RExC_parse < RExC_end && UCHARAT(RExC_parse) != c)
12511 if (RExC_parse == RExC_end) {
12514 /* Try to give a better location for the error (than the end of
12515 * the string) by looking for the matching ']' */
12517 while (RExC_parse < RExC_end && UCHARAT(RExC_parse) != ']') {
12520 vFAIL2("Unmatched '%c' in POSIX class", c);
12522 /* Grandfather lone [:, [=, [. */
12526 const char* const t = RExC_parse++; /* skip over the c */
12529 if (UCHARAT(RExC_parse) == ']') {
12530 const char *posixcc = s + 1;
12531 RExC_parse++; /* skip over the ending ] */
12534 const I32 complement = *posixcc == '^' ? *posixcc++ : 0;
12535 const I32 skip = t - posixcc;
12537 /* Initially switch on the length of the name. */
12540 if (memEQ(posixcc, "word", 4)) /* this is not POSIX,
12541 this is the Perl \w
12543 namedclass = ANYOF_WORDCHAR;
12546 /* Names all of length 5. */
12547 /* alnum alpha ascii blank cntrl digit graph lower
12548 print punct space upper */
12549 /* Offset 4 gives the best switch position. */
12550 switch (posixcc[4]) {
12552 if (memEQ(posixcc, "alph", 4)) /* alpha */
12553 namedclass = ANYOF_ALPHA;
12556 if (memEQ(posixcc, "spac", 4)) /* space */
12557 namedclass = ANYOF_PSXSPC;
12560 if (memEQ(posixcc, "grap", 4)) /* graph */
12561 namedclass = ANYOF_GRAPH;
12564 if (memEQ(posixcc, "asci", 4)) /* ascii */
12565 namedclass = ANYOF_ASCII;
12568 if (memEQ(posixcc, "blan", 4)) /* blank */
12569 namedclass = ANYOF_BLANK;
12572 if (memEQ(posixcc, "cntr", 4)) /* cntrl */
12573 namedclass = ANYOF_CNTRL;
12576 if (memEQ(posixcc, "alnu", 4)) /* alnum */
12577 namedclass = ANYOF_ALPHANUMERIC;
12580 if (memEQ(posixcc, "lowe", 4)) /* lower */
12581 namedclass = (FOLD) ? ANYOF_CASED : ANYOF_LOWER;
12582 else if (memEQ(posixcc, "uppe", 4)) /* upper */
12583 namedclass = (FOLD) ? ANYOF_CASED : ANYOF_UPPER;
12586 if (memEQ(posixcc, "digi", 4)) /* digit */
12587 namedclass = ANYOF_DIGIT;
12588 else if (memEQ(posixcc, "prin", 4)) /* print */
12589 namedclass = ANYOF_PRINT;
12590 else if (memEQ(posixcc, "punc", 4)) /* punct */
12591 namedclass = ANYOF_PUNCT;
12596 if (memEQ(posixcc, "xdigit", 6))
12597 namedclass = ANYOF_XDIGIT;
12601 if (namedclass == OOB_NAMEDCLASS)
12603 "POSIX class [:%"UTF8f":] unknown",
12604 UTF8fARG(UTF, t - s - 1, s + 1));
12606 /* The #defines are structured so each complement is +1 to
12607 * the normal one */
12611 assert (posixcc[skip] == ':');
12612 assert (posixcc[skip+1] == ']');
12613 } else if (!SIZE_ONLY) {
12614 /* [[=foo=]] and [[.foo.]] are still future. */
12616 /* adjust RExC_parse so the warning shows after
12617 the class closes */
12618 while (UCHARAT(RExC_parse) && UCHARAT(RExC_parse) != ']')
12620 vFAIL3("POSIX syntax [%c %c] is reserved for future extensions", c, c);
12623 /* Maternal grandfather:
12624 * "[:" ending in ":" but not in ":]" */
12626 vFAIL("Unmatched '[' in POSIX class");
12629 /* Grandfather lone [:, [=, [. */
12639 S_could_it_be_a_POSIX_class(pTHX_ RExC_state_t *pRExC_state)
12641 /* This applies some heuristics at the current parse position (which should
12642 * be at a '[') to see if what follows might be intended to be a [:posix:]
12643 * class. It returns true if it really is a posix class, of course, but it
12644 * also can return true if it thinks that what was intended was a posix
12645 * class that didn't quite make it.
12647 * It will return true for
12649 * [:alphanumerics] (as long as the ] isn't followed immediately by a
12650 * ')' indicating the end of the (?[
12651 * [:any garbage including %^&$ punctuation:]
12653 * This is designed to be called only from S_handle_regex_sets; it could be
12654 * easily adapted to be called from the spot at the beginning of regclass()
12655 * that checks to see in a normal bracketed class if the surrounding []
12656 * have been omitted ([:word:] instead of [[:word:]]). But doing so would
12657 * change long-standing behavior, so I (khw) didn't do that */
12658 char* p = RExC_parse + 1;
12659 char first_char = *p;
12661 PERL_ARGS_ASSERT_COULD_IT_BE_A_POSIX_CLASS;
12663 assert(*(p - 1) == '[');
12665 if (! POSIXCC(first_char)) {
12670 while (p < RExC_end && isWORDCHAR(*p)) p++;
12672 if (p >= RExC_end) {
12676 if (p - RExC_parse > 2 /* Got at least 1 word character */
12677 && (*p == first_char
12678 || (*p == ']' && p + 1 < RExC_end && *(p + 1) != ')')))
12683 p = (char *) memchr(RExC_parse, ']', RExC_end - RExC_parse);
12686 && p - RExC_parse > 2 /* [:] evaluates to colon;
12687 [::] is a bad posix class. */
12688 && first_char == *(p - 1));
12692 S_handle_regex_sets(pTHX_ RExC_state_t *pRExC_state, SV** return_invlist,
12693 I32 *flagp, U32 depth,
12694 char * const oregcomp_parse)
12696 /* Handle the (?[...]) construct to do set operations */
12699 UV start, end; /* End points of code point ranges */
12701 char *save_end, *save_parse;
12706 const bool save_fold = FOLD;
12708 GET_RE_DEBUG_FLAGS_DECL;
12710 PERL_ARGS_ASSERT_HANDLE_REGEX_SETS;
12713 vFAIL("(?[...]) not valid in locale");
12715 RExC_uni_semantics = 1;
12717 /* This will return only an ANYOF regnode, or (unlikely) something smaller
12718 * (such as EXACT). Thus we can skip most everything if just sizing. We
12719 * call regclass to handle '[]' so as to not have to reinvent its parsing
12720 * rules here (throwing away the size it computes each time). And, we exit
12721 * upon an unescaped ']' that isn't one ending a regclass. To do both
12722 * these things, we need to realize that something preceded by a backslash
12723 * is escaped, so we have to keep track of backslashes */
12725 UV depth = 0; /* how many nested (?[...]) constructs */
12727 Perl_ck_warner_d(aTHX_
12728 packWARN(WARN_EXPERIMENTAL__REGEX_SETS),
12729 "The regex_sets feature is experimental" REPORT_LOCATION,
12730 UTF8fARG(UTF, (RExC_parse - RExC_precomp), RExC_precomp),
12732 RExC_end - RExC_start - (RExC_parse - RExC_precomp),
12733 RExC_precomp + (RExC_parse - RExC_precomp)));
12735 while (RExC_parse < RExC_end) {
12736 SV* current = NULL;
12737 RExC_parse = regpatws(pRExC_state, RExC_parse,
12738 TRUE); /* means recognize comments */
12739 switch (*RExC_parse) {
12741 if (RExC_parse[1] == '[') depth++, RExC_parse++;
12746 /* Skip the next byte (which could cause us to end up in
12747 * the middle of a UTF-8 character, but since none of those
12748 * are confusable with anything we currently handle in this
12749 * switch (invariants all), it's safe. We'll just hit the
12750 * default: case next time and keep on incrementing until
12751 * we find one of the invariants we do handle. */
12756 /* If this looks like it is a [:posix:] class, leave the
12757 * parse pointer at the '[' to fool regclass() into
12758 * thinking it is part of a '[[:posix:]]'. That function
12759 * will use strict checking to force a syntax error if it
12760 * doesn't work out to a legitimate class */
12761 bool is_posix_class
12762 = could_it_be_a_POSIX_class(pRExC_state);
12763 if (! is_posix_class) {
12767 /* regclass() can only return RESTART_UTF8 if multi-char
12768 folds are allowed. */
12769 if (!regclass(pRExC_state, flagp,depth+1,
12770 is_posix_class, /* parse the whole char
12771 class only if not a
12773 FALSE, /* don't allow multi-char folds */
12774 TRUE, /* silence non-portable warnings. */
12776 FAIL2("panic: regclass returned NULL to handle_sets, flags=%#"UVxf"",
12779 /* function call leaves parse pointing to the ']', except
12780 * if we faked it */
12781 if (is_posix_class) {
12785 SvREFCNT_dec(current); /* In case it returned something */
12790 if (depth--) break;
12792 if (RExC_parse < RExC_end
12793 && *RExC_parse == ')')
12795 node = reganode(pRExC_state, ANYOF, 0);
12796 RExC_size += ANYOF_SKIP;
12797 nextchar(pRExC_state);
12798 Set_Node_Length(node,
12799 RExC_parse - oregcomp_parse + 1); /* MJD */
12808 FAIL("Syntax error in (?[...])");
12811 /* Pass 2 only after this. Everything in this construct is a
12812 * metacharacter. Operands begin with either a '\' (for an escape
12813 * sequence), or a '[' for a bracketed character class. Any other
12814 * character should be an operator, or parenthesis for grouping. Both
12815 * types of operands are handled by calling regclass() to parse them. It
12816 * is called with a parameter to indicate to return the computed inversion
12817 * list. The parsing here is implemented via a stack. Each entry on the
12818 * stack is a single character representing one of the operators, or the
12819 * '('; or else a pointer to an operand inversion list. */
12821 #define IS_OPERAND(a) (! SvIOK(a))
12823 /* The stack starts empty. It is a syntax error if the first thing parsed
12824 * is a binary operator; everything else is pushed on the stack. When an
12825 * operand is parsed, the top of the stack is examined. If it is a binary
12826 * operator, the item before it should be an operand, and both are replaced
12827 * by the result of doing that operation on the new operand and the one on
12828 * the stack. Thus a sequence of binary operands is reduced to a single
12829 * one before the next one is parsed.
12831 * A unary operator may immediately follow a binary in the input, for
12834 * When an operand is parsed and the top of the stack is a unary operator,
12835 * the operation is performed, and then the stack is rechecked to see if
12836 * this new operand is part of a binary operation; if so, it is handled as
12839 * A '(' is simply pushed on the stack; it is valid only if the stack is
12840 * empty, or the top element of the stack is an operator or another '('
12841 * (for which the parenthesized expression will become an operand). By the
12842 * time the corresponding ')' is parsed everything in between should have
12843 * been parsed and evaluated to a single operand (or else is a syntax
12844 * error), and is handled as a regular operand */
12846 sv_2mortal((SV *)(stack = newAV()));
12848 while (RExC_parse < RExC_end) {
12849 I32 top_index = av_tindex(stack);
12851 SV* current = NULL;
12853 /* Skip white space */
12854 RExC_parse = regpatws(pRExC_state, RExC_parse,
12855 TRUE); /* means recognize comments */
12856 if (RExC_parse >= RExC_end) {
12857 Perl_croak(aTHX_ "panic: Read past end of '(?[ ])'");
12859 if ((curchar = UCHARAT(RExC_parse)) == ']') {
12866 if (av_tindex(stack) >= 0 /* This makes sure that we can
12867 safely subtract 1 from
12868 RExC_parse in the next clause.
12869 If we have something on the
12870 stack, we have parsed something
12872 && UCHARAT(RExC_parse - 1) == '('
12873 && RExC_parse < RExC_end)
12875 /* If is a '(?', could be an embedded '(?flags:(?[...])'.
12876 * This happens when we have some thing like
12878 * my $thai_or_lao = qr/(?[ \p{Thai} + \p{Lao} ])/;
12880 * qr/(?[ \p{Digit} & $thai_or_lao ])/;
12882 * Here we would be handling the interpolated
12883 * '$thai_or_lao'. We handle this by a recursive call to
12884 * ourselves which returns the inversion list the
12885 * interpolated expression evaluates to. We use the flags
12886 * from the interpolated pattern. */
12887 U32 save_flags = RExC_flags;
12888 const char * const save_parse = ++RExC_parse;
12890 parse_lparen_question_flags(pRExC_state);
12892 if (RExC_parse == save_parse /* Makes sure there was at
12893 least one flag (or this
12894 embedding wasn't compiled)
12896 || RExC_parse >= RExC_end - 4
12897 || UCHARAT(RExC_parse) != ':'
12898 || UCHARAT(++RExC_parse) != '('
12899 || UCHARAT(++RExC_parse) != '?'
12900 || UCHARAT(++RExC_parse) != '[')
12903 /* In combination with the above, this moves the
12904 * pointer to the point just after the first erroneous
12905 * character (or if there are no flags, to where they
12906 * should have been) */
12907 if (RExC_parse >= RExC_end - 4) {
12908 RExC_parse = RExC_end;
12910 else if (RExC_parse != save_parse) {
12911 RExC_parse += (UTF) ? UTF8SKIP(RExC_parse) : 1;
12913 vFAIL("Expecting '(?flags:(?[...'");
12916 (void) handle_regex_sets(pRExC_state, ¤t, flagp,
12917 depth+1, oregcomp_parse);
12919 /* Here, 'current' contains the embedded expression's
12920 * inversion list, and RExC_parse points to the trailing
12921 * ']'; the next character should be the ')' which will be
12922 * paired with the '(' that has been put on the stack, so
12923 * the whole embedded expression reduces to '(operand)' */
12926 RExC_flags = save_flags;
12927 goto handle_operand;
12932 RExC_parse += (UTF) ? UTF8SKIP(RExC_parse) : 1;
12933 vFAIL("Unexpected character");
12936 /* regclass() can only return RESTART_UTF8 if multi-char
12937 folds are allowed. */
12938 if (!regclass(pRExC_state, flagp,depth+1,
12939 TRUE, /* means parse just the next thing */
12940 FALSE, /* don't allow multi-char folds */
12941 FALSE, /* don't silence non-portable warnings. */
12943 FAIL2("panic: regclass returned NULL to handle_sets, flags=%#"UVxf"",
12945 /* regclass() will return with parsing just the \ sequence,
12946 * leaving the parse pointer at the next thing to parse */
12948 goto handle_operand;
12950 case '[': /* Is a bracketed character class */
12952 bool is_posix_class = could_it_be_a_POSIX_class(pRExC_state);
12954 if (! is_posix_class) {
12958 /* regclass() can only return RESTART_UTF8 if multi-char
12959 folds are allowed. */
12960 if(!regclass(pRExC_state, flagp,depth+1,
12961 is_posix_class, /* parse the whole char class
12962 only if not a posix class */
12963 FALSE, /* don't allow multi-char folds */
12964 FALSE, /* don't silence non-portable warnings. */
12966 FAIL2("panic: regclass returned NULL to handle_sets, flags=%#"UVxf"",
12968 /* function call leaves parse pointing to the ']', except if we
12970 if (is_posix_class) {
12974 goto handle_operand;
12983 || ( ! (top_ptr = av_fetch(stack, top_index, FALSE)))
12984 || ! IS_OPERAND(*top_ptr))
12987 vFAIL2("Unexpected binary operator '%c' with no preceding operand", curchar);
12989 av_push(stack, newSVuv(curchar));
12993 av_push(stack, newSVuv(curchar));
12997 if (top_index >= 0) {
12998 top_ptr = av_fetch(stack, top_index, FALSE);
13000 if (IS_OPERAND(*top_ptr)) {
13002 vFAIL("Unexpected '(' with no preceding operator");
13005 av_push(stack, newSVuv(curchar));
13012 || ! (current = av_pop(stack))
13013 || ! IS_OPERAND(current)
13014 || ! (lparen = av_pop(stack))
13015 || IS_OPERAND(lparen)
13016 || SvUV(lparen) != '(')
13018 SvREFCNT_dec(current);
13020 vFAIL("Unexpected ')'");
13023 SvREFCNT_dec_NN(lparen);
13030 /* Here, we have an operand to process, in 'current' */
13032 if (top_index < 0) { /* Just push if stack is empty */
13033 av_push(stack, current);
13036 SV* top = av_pop(stack);
13038 char current_operator;
13040 if (IS_OPERAND(top)) {
13041 SvREFCNT_dec_NN(top);
13042 SvREFCNT_dec_NN(current);
13043 vFAIL("Operand with no preceding operator");
13045 current_operator = (char) SvUV(top);
13046 switch (current_operator) {
13047 case '(': /* Push the '(' back on followed by the new
13049 av_push(stack, top);
13050 av_push(stack, current);
13051 SvREFCNT_inc(top); /* Counters the '_dec' done
13052 just after the 'break', so
13053 it doesn't get wrongly freed
13058 _invlist_invert(current);
13060 /* Unlike binary operators, the top of the stack,
13061 * now that this unary one has been popped off, may
13062 * legally be an operator, and we now have operand
13065 SvREFCNT_dec_NN(top);
13066 goto handle_operand;
13069 prev = av_pop(stack);
13070 _invlist_intersection(prev,
13073 av_push(stack, current);
13078 prev = av_pop(stack);
13079 _invlist_union(prev, current, ¤t);
13080 av_push(stack, current);
13084 prev = av_pop(stack);;
13085 _invlist_subtract(prev, current, ¤t);
13086 av_push(stack, current);
13089 case '^': /* The union minus the intersection */
13095 prev = av_pop(stack);
13096 _invlist_union(prev, current, &u);
13097 _invlist_intersection(prev, current, &i);
13098 /* _invlist_subtract will overwrite current
13099 without freeing what it already contains */
13101 _invlist_subtract(u, i, ¤t);
13102 av_push(stack, current);
13103 SvREFCNT_dec_NN(i);
13104 SvREFCNT_dec_NN(u);
13105 SvREFCNT_dec_NN(element);
13110 Perl_croak(aTHX_ "panic: Unexpected item on '(?[ ])' stack");
13112 SvREFCNT_dec_NN(top);
13113 SvREFCNT_dec(prev);
13117 RExC_parse += (UTF) ? UTF8SKIP(RExC_parse) : 1;
13120 if (av_tindex(stack) < 0 /* Was empty */
13121 || ((final = av_pop(stack)) == NULL)
13122 || ! IS_OPERAND(final)
13123 || av_tindex(stack) >= 0) /* More left on stack */
13125 vFAIL("Incomplete expression within '(?[ ])'");
13128 /* Here, 'final' is the resultant inversion list from evaluating the
13129 * expression. Return it if so requested */
13130 if (return_invlist) {
13131 *return_invlist = final;
13135 /* Otherwise generate a resultant node, based on 'final'. regclass() is
13136 * expecting a string of ranges and individual code points */
13137 invlist_iterinit(final);
13138 result_string = newSVpvs("");
13139 while (invlist_iternext(final, &start, &end)) {
13140 if (start == end) {
13141 Perl_sv_catpvf(aTHX_ result_string, "\\x{%"UVXf"}", start);
13144 Perl_sv_catpvf(aTHX_ result_string, "\\x{%"UVXf"}-\\x{%"UVXf"}",
13149 save_parse = RExC_parse;
13150 RExC_parse = SvPV(result_string, len);
13151 save_end = RExC_end;
13152 RExC_end = RExC_parse + len;
13154 /* We turn off folding around the call, as the class we have constructed
13155 * already has all folding taken into consideration, and we don't want
13156 * regclass() to add to that */
13157 RExC_flags &= ~RXf_PMf_FOLD;
13158 /* regclass() can only return RESTART_UTF8 if multi-char folds are allowed.
13160 node = regclass(pRExC_state, flagp,depth+1,
13161 FALSE, /* means parse the whole char class */
13162 FALSE, /* don't allow multi-char folds */
13163 TRUE, /* silence non-portable warnings. The above may very
13164 well have generated non-portable code points, but
13165 they're valid on this machine */
13168 FAIL2("panic: regclass returned NULL to handle_sets, flags=%#"UVxf,
13171 RExC_flags |= RXf_PMf_FOLD;
13173 RExC_parse = save_parse + 1;
13174 RExC_end = save_end;
13175 SvREFCNT_dec_NN(final);
13176 SvREFCNT_dec_NN(result_string);
13178 nextchar(pRExC_state);
13179 Set_Node_Length(node, RExC_parse - oregcomp_parse + 1); /* MJD */
13184 /* The names of properties whose definitions are not known at compile time are
13185 * stored in this SV, after a constant heading. So if the length has been
13186 * changed since initialization, then there is a run-time definition. */
13187 #define HAS_NONLOCALE_RUNTIME_PROPERTY_DEFINITION \
13188 (SvCUR(listsv) != initial_listsv_len)
13191 S_regclass(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth,
13192 const bool stop_at_1, /* Just parse the next thing, don't
13193 look for a full character class */
13194 bool allow_multi_folds,
13195 const bool silence_non_portable, /* Don't output warnings
13198 SV** ret_invlist) /* Return an inversion list, not a node */
13200 /* parse a bracketed class specification. Most of these will produce an
13201 * ANYOF node; but something like [a] will produce an EXACT node; [aA], an
13202 * EXACTFish node; [[:ascii:]], a POSIXA node; etc. It is more complex
13203 * under /i with multi-character folds: it will be rewritten following the
13204 * paradigm of this example, where the <multi-fold>s are characters which
13205 * fold to multiple character sequences:
13206 * /[abc\x{multi-fold1}def\x{multi-fold2}ghi]/i
13207 * gets effectively rewritten as:
13208 * /(?:\x{multi-fold1}|\x{multi-fold2}|[abcdefghi]/i
13209 * reg() gets called (recursively) on the rewritten version, and this
13210 * function will return what it constructs. (Actually the <multi-fold>s
13211 * aren't physically removed from the [abcdefghi], it's just that they are
13212 * ignored in the recursion by means of a flag:
13213 * <RExC_in_multi_char_class>.)
13215 * ANYOF nodes contain a bit map for the first 256 characters, with the
13216 * corresponding bit set if that character is in the list. For characters
13217 * above 255, a range list or swash is used. There are extra bits for \w,
13218 * etc. in locale ANYOFs, as what these match is not determinable at
13221 * Returns NULL, setting *flagp to RESTART_UTF8 if the sizing scan needs
13222 * to be restarted. This can only happen if ret_invlist is non-NULL.
13226 UV prevvalue = OOB_UNICODE, save_prevvalue = OOB_UNICODE;
13228 UV value = OOB_UNICODE, save_value = OOB_UNICODE;
13231 IV namedclass = OOB_NAMEDCLASS;
13232 char *rangebegin = NULL;
13233 bool need_class = 0;
13235 STRLEN initial_listsv_len = 0; /* Kind of a kludge to see if it is more
13236 than just initialized. */
13237 SV* properties = NULL; /* Code points that match \p{} \P{} */
13238 SV* posixes = NULL; /* Code points that match classes like [:word:],
13239 extended beyond the Latin1 range. These have to
13240 be kept separate from other code points for much
13241 of this function because their handling is
13242 different under /i, and for most classes under
13244 SV* nposixes = NULL; /* Similarly for [:^word:]. These are kept
13245 separate for a while from the non-complemented
13246 versions because of complications with /d
13248 UV element_count = 0; /* Number of distinct elements in the class.
13249 Optimizations may be possible if this is tiny */
13250 AV * multi_char_matches = NULL; /* Code points that fold to more than one
13251 character; used under /i */
13253 char * stop_ptr = RExC_end; /* where to stop parsing */
13254 const bool skip_white = cBOOL(ret_invlist); /* ignore unescaped white
13256 const bool strict = cBOOL(ret_invlist); /* Apply strict parsing rules? */
13258 /* Unicode properties are stored in a swash; this holds the current one
13259 * being parsed. If this swash is the only above-latin1 component of the
13260 * character class, an optimization is to pass it directly on to the
13261 * execution engine. Otherwise, it is set to NULL to indicate that there
13262 * are other things in the class that have to be dealt with at execution
13264 SV* swash = NULL; /* Code points that match \p{} \P{} */
13266 /* Set if a component of this character class is user-defined; just passed
13267 * on to the engine */
13268 bool has_user_defined_property = FALSE;
13270 /* inversion list of code points this node matches only when the target
13271 * string is in UTF-8. (Because is under /d) */
13272 SV* depends_list = NULL;
13274 /* Inversion list of code points this node matches regardless of things
13275 * like locale, folding, utf8ness of the target string */
13276 SV* cp_list = NULL;
13278 /* Like cp_list, but code points on this list need to be checked for things
13279 * that fold to/from them under /i */
13280 SV* cp_foldable_list = NULL;
13282 /* Like cp_list, but code points on this list are valid only when the
13283 * runtime locale is UTF-8 */
13284 SV* only_utf8_locale_list = NULL;
13287 /* In a range, counts how many 0-2 of the ends of it came from literals,
13288 * not escapes. Thus we can tell if 'A' was input vs \x{C1} */
13289 UV literal_endpoint = 0;
13291 bool invert = FALSE; /* Is this class to be complemented */
13293 bool warn_super = ALWAYS_WARN_SUPER;
13295 regnode * const orig_emit = RExC_emit; /* Save the original RExC_emit in
13296 case we need to change the emitted regop to an EXACT. */
13297 const char * orig_parse = RExC_parse;
13298 const SSize_t orig_size = RExC_size;
13299 bool posixl_matches_all = FALSE; /* Does /l class have both e.g. \W,\w ? */
13300 GET_RE_DEBUG_FLAGS_DECL;
13302 PERL_ARGS_ASSERT_REGCLASS;
13304 PERL_UNUSED_ARG(depth);
13307 DEBUG_PARSE("clas");
13309 /* Assume we are going to generate an ANYOF node. */
13310 ret = reganode(pRExC_state, ANYOF, 0);
13313 RExC_size += ANYOF_SKIP;
13314 listsv = &PL_sv_undef; /* For code scanners: listsv always non-NULL. */
13317 ANYOF_FLAGS(ret) = 0;
13319 RExC_emit += ANYOF_SKIP;
13320 listsv = newSVpvs_flags("# comment\n", SVs_TEMP);
13321 initial_listsv_len = SvCUR(listsv);
13322 SvTEMP_off(listsv); /* Grr, TEMPs and mortals are conflated. */
13326 RExC_parse = regpatws(pRExC_state, RExC_parse,
13327 FALSE /* means don't recognize comments */);
13330 if (UCHARAT(RExC_parse) == '^') { /* Complement of range. */
13333 allow_multi_folds = FALSE;
13336 RExC_parse = regpatws(pRExC_state, RExC_parse,
13337 FALSE /* means don't recognize comments */);
13341 /* Check that they didn't say [:posix:] instead of [[:posix:]] */
13342 if (!SIZE_ONLY && RExC_parse < RExC_end && POSIXCC(UCHARAT(RExC_parse))) {
13343 const char *s = RExC_parse;
13344 const char c = *s++;
13346 while (isWORDCHAR(*s))
13348 if (*s && c == *s && s[1] == ']') {
13349 SAVEFREESV(RExC_rx_sv);
13351 "POSIX syntax [%c %c] belongs inside character classes",
13353 (void)ReREFCNT_inc(RExC_rx_sv);
13357 /* If the caller wants us to just parse a single element, accomplish this
13358 * by faking the loop ending condition */
13359 if (stop_at_1 && RExC_end > RExC_parse) {
13360 stop_ptr = RExC_parse + 1;
13363 /* allow 1st char to be ']' (allowing it to be '-' is dealt with later) */
13364 if (UCHARAT(RExC_parse) == ']')
13365 goto charclassloop;
13369 if (RExC_parse >= stop_ptr) {
13374 RExC_parse = regpatws(pRExC_state, RExC_parse,
13375 FALSE /* means don't recognize comments */);
13378 if (UCHARAT(RExC_parse) == ']') {
13384 namedclass = OOB_NAMEDCLASS; /* initialize as illegal */
13385 save_value = value;
13386 save_prevvalue = prevvalue;
13389 rangebegin = RExC_parse;
13393 value = utf8n_to_uvchr((U8*)RExC_parse,
13394 RExC_end - RExC_parse,
13395 &numlen, UTF8_ALLOW_DEFAULT);
13396 RExC_parse += numlen;
13399 value = UCHARAT(RExC_parse++);
13402 && RExC_parse < RExC_end
13403 && POSIXCC(UCHARAT(RExC_parse)))
13405 namedclass = regpposixcc(pRExC_state, value, strict);
13407 else if (value == '\\') {
13409 value = utf8n_to_uvchr((U8*)RExC_parse,
13410 RExC_end - RExC_parse,
13411 &numlen, UTF8_ALLOW_DEFAULT);
13412 RExC_parse += numlen;
13415 value = UCHARAT(RExC_parse++);
13417 /* Some compilers cannot handle switching on 64-bit integer
13418 * values, therefore value cannot be an UV. Yes, this will
13419 * be a problem later if we want switch on Unicode.
13420 * A similar issue a little bit later when switching on
13421 * namedclass. --jhi */
13423 /* If the \ is escaping white space when white space is being
13424 * skipped, it means that that white space is wanted literally, and
13425 * is already in 'value'. Otherwise, need to translate the escape
13426 * into what it signifies. */
13427 if (! skip_white || ! is_PATWS_cp(value)) switch ((I32)value) {
13429 case 'w': namedclass = ANYOF_WORDCHAR; break;
13430 case 'W': namedclass = ANYOF_NWORDCHAR; break;
13431 case 's': namedclass = ANYOF_SPACE; break;
13432 case 'S': namedclass = ANYOF_NSPACE; break;
13433 case 'd': namedclass = ANYOF_DIGIT; break;
13434 case 'D': namedclass = ANYOF_NDIGIT; break;
13435 case 'v': namedclass = ANYOF_VERTWS; break;
13436 case 'V': namedclass = ANYOF_NVERTWS; break;
13437 case 'h': namedclass = ANYOF_HORIZWS; break;
13438 case 'H': namedclass = ANYOF_NHORIZWS; break;
13439 case 'N': /* Handle \N{NAME} in class */
13441 /* We only pay attention to the first char of
13442 multichar strings being returned. I kinda wonder
13443 if this makes sense as it does change the behaviour
13444 from earlier versions, OTOH that behaviour was broken
13446 if (! grok_bslash_N(pRExC_state, NULL, &value, flagp, depth,
13447 TRUE, /* => charclass */
13450 if (*flagp & RESTART_UTF8)
13451 FAIL("panic: grok_bslash_N set RESTART_UTF8");
13461 /* We will handle any undefined properties ourselves */
13462 U8 swash_init_flags = _CORE_SWASH_INIT_RETURN_IF_UNDEF
13463 /* And we actually would prefer to get
13464 * the straight inversion list of the
13465 * swash, since we will be accessing it
13466 * anyway, to save a little time */
13467 |_CORE_SWASH_INIT_ACCEPT_INVLIST;
13469 if (RExC_parse >= RExC_end)
13470 vFAIL2("Empty \\%c{}", (U8)value);
13471 if (*RExC_parse == '{') {
13472 const U8 c = (U8)value;
13473 e = strchr(RExC_parse++, '}');
13475 vFAIL2("Missing right brace on \\%c{}", c);
13476 while (isSPACE(UCHARAT(RExC_parse)))
13478 if (e == RExC_parse)
13479 vFAIL2("Empty \\%c{}", c);
13480 n = e - RExC_parse;
13481 while (isSPACE(UCHARAT(RExC_parse + n - 1)))
13493 if (UCHARAT(RExC_parse) == '^') {
13496 /* toggle. (The rhs xor gets the single bit that
13497 * differs between P and p; the other xor inverts just
13499 value ^= 'P' ^ 'p';
13501 while (isSPACE(UCHARAT(RExC_parse))) {
13506 /* Try to get the definition of the property into
13507 * <invlist>. If /i is in effect, the effective property
13508 * will have its name be <__NAME_i>. The design is
13509 * discussed in commit
13510 * 2f833f5208e26b208886e51e09e2c072b5eabb46 */
13511 formatted = Perl_form(aTHX_
13513 (FOLD) ? "__" : "",
13518 name = savepvn(formatted, strlen(formatted));
13520 /* Look up the property name, and get its swash and
13521 * inversion list, if the property is found */
13523 SvREFCNT_dec_NN(swash);
13525 swash = _core_swash_init("utf8", name, &PL_sv_undef,
13528 NULL, /* No inversion list */
13531 if (! swash || ! (invlist = _get_swash_invlist(swash))) {
13533 SvREFCNT_dec_NN(swash);
13537 /* Here didn't find it. It could be a user-defined
13538 * property that will be available at run-time. If we
13539 * accept only compile-time properties, is an error;
13540 * otherwise add it to the list for run-time look up */
13542 RExC_parse = e + 1;
13544 "Property '%"UTF8f"' is unknown",
13545 UTF8fARG(UTF, n, name));
13547 Perl_sv_catpvf(aTHX_ listsv, "%cutf8::%"UTF8f"\n",
13548 (value == 'p' ? '+' : '!'),
13549 UTF8fARG(UTF, n, name));
13550 has_user_defined_property = TRUE;
13552 /* We don't know yet, so have to assume that the
13553 * property could match something in the Latin1 range,
13554 * hence something that isn't utf8. Note that this
13555 * would cause things in <depends_list> to match
13556 * inappropriately, except that any \p{}, including
13557 * this one forces Unicode semantics, which means there
13558 * is no <depends_list> */
13559 ANYOF_FLAGS(ret) |= ANYOF_NONBITMAP_NON_UTF8;
13563 /* Here, did get the swash and its inversion list. If
13564 * the swash is from a user-defined property, then this
13565 * whole character class should be regarded as such */
13566 if (swash_init_flags
13567 & _CORE_SWASH_INIT_USER_DEFINED_PROPERTY)
13569 has_user_defined_property = TRUE;
13572 /* We warn on matching an above-Unicode code point
13573 * if the match would return true, except don't
13574 * warn for \p{All}, which has exactly one element
13576 (_invlist_contains_cp(invlist, 0x110000)
13577 && (! (_invlist_len(invlist) == 1
13578 && *invlist_array(invlist) == 0)))
13584 /* Invert if asking for the complement */
13585 if (value == 'P') {
13586 _invlist_union_complement_2nd(properties,
13590 /* The swash can't be used as-is, because we've
13591 * inverted things; delay removing it to here after
13592 * have copied its invlist above */
13593 SvREFCNT_dec_NN(swash);
13597 _invlist_union(properties, invlist, &properties);
13602 RExC_parse = e + 1;
13603 namedclass = ANYOF_UNIPROP; /* no official name, but it's
13606 /* \p means they want Unicode semantics */
13607 RExC_uni_semantics = 1;
13610 case 'n': value = '\n'; break;
13611 case 'r': value = '\r'; break;
13612 case 't': value = '\t'; break;
13613 case 'f': value = '\f'; break;
13614 case 'b': value = '\b'; break;
13615 case 'e': value = ASCII_TO_NATIVE('\033');break;
13616 case 'a': value = '\a'; break;
13618 RExC_parse--; /* function expects to be pointed at the 'o' */
13620 const char* error_msg;
13621 bool valid = grok_bslash_o(&RExC_parse,
13624 SIZE_ONLY, /* warnings in pass
13627 silence_non_portable,
13633 if (PL_encoding && value < 0x100) {
13634 goto recode_encoding;
13638 RExC_parse--; /* function expects to be pointed at the 'x' */
13640 const char* error_msg;
13641 bool valid = grok_bslash_x(&RExC_parse,
13644 TRUE, /* Output warnings */
13646 silence_non_portable,
13652 if (PL_encoding && value < 0x100)
13653 goto recode_encoding;
13656 value = grok_bslash_c(*RExC_parse++, SIZE_ONLY);
13658 case '0': case '1': case '2': case '3': case '4':
13659 case '5': case '6': case '7':
13661 /* Take 1-3 octal digits */
13662 I32 flags = PERL_SCAN_SILENT_ILLDIGIT;
13663 numlen = (strict) ? 4 : 3;
13664 value = grok_oct(--RExC_parse, &numlen, &flags, NULL);
13665 RExC_parse += numlen;
13668 RExC_parse += (UTF) ? UTF8SKIP(RExC_parse) : 1;
13669 vFAIL("Need exactly 3 octal digits");
13671 else if (! SIZE_ONLY /* like \08, \178 */
13673 && RExC_parse < RExC_end
13674 && isDIGIT(*RExC_parse)
13675 && ckWARN(WARN_REGEXP))
13677 SAVEFREESV(RExC_rx_sv);
13678 reg_warn_non_literal_string(
13680 form_short_octal_warning(RExC_parse, numlen));
13681 (void)ReREFCNT_inc(RExC_rx_sv);
13684 if (PL_encoding && value < 0x100)
13685 goto recode_encoding;
13689 if (! RExC_override_recoding) {
13690 SV* enc = PL_encoding;
13691 value = reg_recode((const char)(U8)value, &enc);
13694 vFAIL("Invalid escape in the specified encoding");
13696 else if (SIZE_ONLY) {
13697 ckWARNreg(RExC_parse,
13698 "Invalid escape in the specified encoding");
13704 /* Allow \_ to not give an error */
13705 if (!SIZE_ONLY && isWORDCHAR(value) && value != '_') {
13707 vFAIL2("Unrecognized escape \\%c in character class",
13711 SAVEFREESV(RExC_rx_sv);
13712 ckWARN2reg(RExC_parse,
13713 "Unrecognized escape \\%c in character class passed through",
13715 (void)ReREFCNT_inc(RExC_rx_sv);
13719 } /* End of switch on char following backslash */
13720 } /* end of handling backslash escape sequences */
13723 literal_endpoint++;
13726 /* Here, we have the current token in 'value' */
13728 if (namedclass > OOB_NAMEDCLASS) { /* this is a named class \blah */
13731 /* a bad range like a-\d, a-[:digit:]. The '-' is taken as a
13732 * literal, as is the character that began the false range, i.e.
13733 * the 'a' in the examples */
13736 const int w = (RExC_parse >= rangebegin)
13737 ? RExC_parse - rangebegin
13741 "False [] range \"%"UTF8f"\"",
13742 UTF8fARG(UTF, w, rangebegin));
13745 SAVEFREESV(RExC_rx_sv); /* in case of fatal warnings */
13746 ckWARN2reg(RExC_parse,
13747 "False [] range \"%"UTF8f"\"",
13748 UTF8fARG(UTF, w, rangebegin));
13749 (void)ReREFCNT_inc(RExC_rx_sv);
13750 cp_list = add_cp_to_invlist(cp_list, '-');
13751 cp_foldable_list = add_cp_to_invlist(cp_foldable_list,
13756 range = 0; /* this was not a true range */
13757 element_count += 2; /* So counts for three values */
13760 classnum = namedclass_to_classnum(namedclass);
13762 if (LOC && namedclass < ANYOF_POSIXL_MAX
13763 #ifndef HAS_ISASCII
13764 && classnum != _CC_ASCII
13767 /* What the Posix classes (like \w, [:space:]) match in locale
13768 * isn't knowable under locale until actual match time. Room
13769 * must be reserved (one time per outer bracketed class) to
13770 * store such classes. The space will contain a bit for each
13771 * named class that is to be matched against. This isn't
13772 * needed for \p{} and pseudo-classes, as they are not affected
13773 * by locale, and hence are dealt with separately */
13774 if (! need_class) {
13777 RExC_size += ANYOF_POSIXL_SKIP - ANYOF_SKIP;
13780 RExC_emit += ANYOF_POSIXL_SKIP - ANYOF_SKIP;
13782 ANYOF_FLAGS(ret) |= ANYOF_POSIXL;
13783 ANYOF_POSIXL_ZERO(ret);
13786 /* See if it already matches the complement of this POSIX
13788 if ((ANYOF_FLAGS(ret) & ANYOF_POSIXL)
13789 && ANYOF_POSIXL_TEST(ret, namedclass + ((namedclass % 2)
13793 posixl_matches_all = TRUE;
13794 break; /* No need to continue. Since it matches both
13795 e.g., \w and \W, it matches everything, and the
13796 bracketed class can be optimized into qr/./s */
13799 /* Add this class to those that should be checked at runtime */
13800 ANYOF_POSIXL_SET(ret, namedclass);
13802 /* The above-Latin1 characters are not subject to locale rules.
13803 * Just add them, in the second pass, to the
13804 * unconditionally-matched list */
13806 SV* scratch_list = NULL;
13808 /* Get the list of the above-Latin1 code points this
13810 _invlist_intersection_maybe_complement_2nd(PL_AboveLatin1,
13811 PL_XPosix_ptrs[classnum],
13813 /* Odd numbers are complements, like
13814 * NDIGIT, NASCII, ... */
13815 namedclass % 2 != 0,
13817 /* Checking if 'cp_list' is NULL first saves an extra
13818 * clone. Its reference count will be decremented at the
13819 * next union, etc, or if this is the only instance, at the
13820 * end of the routine */
13822 cp_list = scratch_list;
13825 _invlist_union(cp_list, scratch_list, &cp_list);
13826 SvREFCNT_dec_NN(scratch_list);
13828 continue; /* Go get next character */
13831 else if (! SIZE_ONLY) {
13833 /* Here, not in pass1 (in that pass we skip calculating the
13834 * contents of this class), and is /l, or is a POSIX class for
13835 * which /l doesn't matter (or is a Unicode property, which is
13836 * skipped here). */
13837 if (namedclass >= ANYOF_POSIXL_MAX) { /* If a special class */
13838 if (namedclass != ANYOF_UNIPROP) { /* UNIPROP = \p and \P */
13840 /* Here, should be \h, \H, \v, or \V. None of /d, /i
13841 * nor /l make a difference in what these match,
13842 * therefore we just add what they match to cp_list. */
13843 if (classnum != _CC_VERTSPACE) {
13844 assert( namedclass == ANYOF_HORIZWS
13845 || namedclass == ANYOF_NHORIZWS);
13847 /* It turns out that \h is just a synonym for
13849 classnum = _CC_BLANK;
13852 _invlist_union_maybe_complement_2nd(
13854 PL_XPosix_ptrs[classnum],
13855 namedclass % 2 != 0, /* Complement if odd
13856 (NHORIZWS, NVERTWS)
13861 else { /* Garden variety class. If is NASCII, NDIGIT, ...
13862 complement and use nposixes */
13863 SV** posixes_ptr = namedclass % 2 == 0
13866 SV** source_ptr = &PL_XPosix_ptrs[classnum];
13867 _invlist_union_maybe_complement_2nd(
13870 namedclass % 2 != 0,
13873 continue; /* Go get next character */
13875 } /* end of namedclass \blah */
13877 /* Here, we have a single value. If 'range' is set, it is the ending
13878 * of a range--check its validity. Later, we will handle each
13879 * individual code point in the range. If 'range' isn't set, this
13880 * could be the beginning of a range, so check for that by looking
13881 * ahead to see if the next real character to be processed is the range
13882 * indicator--the minus sign */
13885 RExC_parse = regpatws(pRExC_state, RExC_parse,
13886 FALSE /* means don't recognize comments */);
13890 if (prevvalue > value) /* b-a */ {
13891 const int w = RExC_parse - rangebegin;
13893 "Invalid [] range \"%"UTF8f"\"",
13894 UTF8fARG(UTF, w, rangebegin));
13895 range = 0; /* not a valid range */
13899 prevvalue = value; /* save the beginning of the potential range */
13900 if (! stop_at_1 /* Can't be a range if parsing just one thing */
13901 && *RExC_parse == '-')
13903 char* next_char_ptr = RExC_parse + 1;
13904 if (skip_white) { /* Get the next real char after the '-' */
13905 next_char_ptr = regpatws(pRExC_state,
13907 FALSE); /* means don't recognize
13911 /* If the '-' is at the end of the class (just before the ']',
13912 * it is a literal minus; otherwise it is a range */
13913 if (next_char_ptr < RExC_end && *next_char_ptr != ']') {
13914 RExC_parse = next_char_ptr;
13916 /* a bad range like \w-, [:word:]- ? */
13917 if (namedclass > OOB_NAMEDCLASS) {
13918 if (strict || ckWARN(WARN_REGEXP)) {
13920 RExC_parse >= rangebegin ?
13921 RExC_parse - rangebegin : 0;
13923 vFAIL4("False [] range \"%*.*s\"",
13928 "False [] range \"%*.*s\"",
13933 cp_list = add_cp_to_invlist(cp_list, '-');
13937 range = 1; /* yeah, it's a range! */
13938 continue; /* but do it the next time */
13943 /* Here, <prevvalue> is the beginning of the range, if any; or <value>
13946 /* non-Latin1 code point implies unicode semantics. Must be set in
13947 * pass1 so is there for the whole of pass 2 */
13949 RExC_uni_semantics = 1;
13952 /* Ready to process either the single value, or the completed range.
13953 * For single-valued non-inverted ranges, we consider the possibility
13954 * of multi-char folds. (We made a conscious decision to not do this
13955 * for the other cases because it can often lead to non-intuitive
13956 * results. For example, you have the peculiar case that:
13957 * "s s" =~ /^[^\xDF]+$/i => Y
13958 * "ss" =~ /^[^\xDF]+$/i => N
13960 * See [perl #89750] */
13961 if (FOLD && allow_multi_folds && value == prevvalue) {
13962 if (value == LATIN_SMALL_LETTER_SHARP_S
13963 || (value > 255 && _invlist_contains_cp(PL_HasMultiCharFold,
13966 /* Here <value> is indeed a multi-char fold. Get what it is */
13968 U8 foldbuf[UTF8_MAXBYTES_CASE];
13971 UV folded = _to_uni_fold_flags(
13975 FOLD_FLAGS_FULL | (ASCII_FOLD_RESTRICTED
13976 ? FOLD_FLAGS_NOMIX_ASCII
13980 /* Here, <folded> should be the first character of the
13981 * multi-char fold of <value>, with <foldbuf> containing the
13982 * whole thing. But, if this fold is not allowed (because of
13983 * the flags), <fold> will be the same as <value>, and should
13984 * be processed like any other character, so skip the special
13986 if (folded != value) {
13988 /* Skip if we are recursed, currently parsing the class
13989 * again. Otherwise add this character to the list of
13990 * multi-char folds. */
13991 if (! RExC_in_multi_char_class) {
13992 AV** this_array_ptr;
13994 STRLEN cp_count = utf8_length(foldbuf,
13995 foldbuf + foldlen);
13996 SV* multi_fold = sv_2mortal(newSVpvn("", 0));
13998 Perl_sv_catpvf(aTHX_ multi_fold, "\\x{%"UVXf"}", value);
14001 if (! multi_char_matches) {
14002 multi_char_matches = newAV();
14005 /* <multi_char_matches> is actually an array of arrays.
14006 * There will be one or two top-level elements: [2],
14007 * and/or [3]. The [2] element is an array, each
14008 * element thereof is a character which folds to TWO
14009 * characters; [3] is for folds to THREE characters.
14010 * (Unicode guarantees a maximum of 3 characters in any
14011 * fold.) When we rewrite the character class below,
14012 * we will do so such that the longest folds are
14013 * written first, so that it prefers the longest
14014 * matching strings first. This is done even if it
14015 * turns out that any quantifier is non-greedy, out of
14016 * programmer laziness. Tom Christiansen has agreed
14017 * that this is ok. This makes the test for the
14018 * ligature 'ffi' come before the test for 'ff' */
14019 if (av_exists(multi_char_matches, cp_count)) {
14020 this_array_ptr = (AV**) av_fetch(multi_char_matches,
14022 this_array = *this_array_ptr;
14025 this_array = newAV();
14026 av_store(multi_char_matches, cp_count,
14029 av_push(this_array, multi_fold);
14032 /* This element should not be processed further in this
14035 value = save_value;
14036 prevvalue = save_prevvalue;
14042 /* Deal with this element of the class */
14045 cp_foldable_list = _add_range_to_invlist(cp_foldable_list,
14048 SV* this_range = _new_invlist(1);
14049 _append_range_to_invlist(this_range, prevvalue, value);
14051 /* In EBCDIC, the ranges 'A-Z' and 'a-z' are each not contiguous.
14052 * If this range was specified using something like 'i-j', we want
14053 * to include only the 'i' and the 'j', and not anything in
14054 * between, so exclude non-ASCII, non-alphabetics from it.
14055 * However, if the range was specified with something like
14056 * [\x89-\x91] or [\x89-j], all code points within it should be
14057 * included. literal_endpoint==2 means both ends of the range used
14058 * a literal character, not \x{foo} */
14059 if (literal_endpoint == 2
14060 && ((prevvalue >= 'a' && value <= 'z')
14061 || (prevvalue >= 'A' && value <= 'Z')))
14063 _invlist_intersection(this_range, PL_ASCII,
14066 /* Since this above only contains ascii, the intersection of it
14067 * with anything will still yield only ascii */
14068 _invlist_intersection(this_range, PL_XPosix_ptrs[_CC_ALPHA],
14071 _invlist_union(cp_foldable_list, this_range, &cp_foldable_list);
14072 literal_endpoint = 0;
14076 range = 0; /* this range (if it was one) is done now */
14077 } /* End of loop through all the text within the brackets */
14079 /* If anything in the class expands to more than one character, we have to
14080 * deal with them by building up a substitute parse string, and recursively
14081 * calling reg() on it, instead of proceeding */
14082 if (multi_char_matches) {
14083 SV * substitute_parse = newSVpvn_flags("?:", 2, SVs_TEMP);
14086 char *save_end = RExC_end;
14087 char *save_parse = RExC_parse;
14088 bool first_time = TRUE; /* First multi-char occurrence doesn't get
14093 #if 0 /* Have decided not to deal with multi-char folds in inverted classes,
14094 because too confusing */
14096 sv_catpv(substitute_parse, "(?:");
14100 /* Look at the longest folds first */
14101 for (cp_count = av_tindex(multi_char_matches); cp_count > 0; cp_count--) {
14103 if (av_exists(multi_char_matches, cp_count)) {
14104 AV** this_array_ptr;
14107 this_array_ptr = (AV**) av_fetch(multi_char_matches,
14109 while ((this_sequence = av_pop(*this_array_ptr)) !=
14112 if (! first_time) {
14113 sv_catpv(substitute_parse, "|");
14115 first_time = FALSE;
14117 sv_catpv(substitute_parse, SvPVX(this_sequence));
14122 /* If the character class contains anything else besides these
14123 * multi-character folds, have to include it in recursive parsing */
14124 if (element_count) {
14125 sv_catpv(substitute_parse, "|[");
14126 sv_catpvn(substitute_parse, orig_parse, RExC_parse - orig_parse);
14127 sv_catpv(substitute_parse, "]");
14130 sv_catpv(substitute_parse, ")");
14133 /* This is a way to get the parse to skip forward a whole named
14134 * sequence instead of matching the 2nd character when it fails the
14136 sv_catpv(substitute_parse, "(*THEN)(*SKIP)(*FAIL)|.)");
14140 RExC_parse = SvPV(substitute_parse, len);
14141 RExC_end = RExC_parse + len;
14142 RExC_in_multi_char_class = 1;
14143 RExC_emit = (regnode *)orig_emit;
14145 ret = reg(pRExC_state, 1, ®_flags, depth+1);
14147 *flagp |= reg_flags&(HASWIDTH|SIMPLE|SPSTART|POSTPONED|RESTART_UTF8);
14149 RExC_parse = save_parse;
14150 RExC_end = save_end;
14151 RExC_in_multi_char_class = 0;
14152 SvREFCNT_dec_NN(multi_char_matches);
14156 /* Here, we've gone through the entire class and dealt with multi-char
14157 * folds. We are now in a position that we can do some checks to see if we
14158 * can optimize this ANYOF node into a simpler one, even in Pass 1.
14159 * Currently we only do two checks:
14160 * 1) is in the unlikely event that the user has specified both, eg. \w and
14161 * \W under /l, then the class matches everything. (This optimization
14162 * is done only to make the optimizer code run later work.)
14163 * 2) if the character class contains only a single element (including a
14164 * single range), we see if there is an equivalent node for it.
14165 * Other checks are possible */
14166 if (! ret_invlist /* Can't optimize if returning the constructed
14168 && (UNLIKELY(posixl_matches_all) || element_count == 1))
14173 if (UNLIKELY(posixl_matches_all)) {
14176 else if (namedclass > OOB_NAMEDCLASS) { /* this is a named class, like
14177 \w or [:digit:] or \p{foo}
14180 /* All named classes are mapped into POSIXish nodes, with its FLAG
14181 * argument giving which class it is */
14182 switch ((I32)namedclass) {
14183 case ANYOF_UNIPROP:
14186 /* These don't depend on the charset modifiers. They always
14187 * match under /u rules */
14188 case ANYOF_NHORIZWS:
14189 case ANYOF_HORIZWS:
14190 namedclass = ANYOF_BLANK + namedclass - ANYOF_HORIZWS;
14193 case ANYOF_NVERTWS:
14198 /* The actual POSIXish node for all the rest depends on the
14199 * charset modifier. The ones in the first set depend only on
14200 * ASCII or, if available on this platform, locale */
14204 op = (LOC) ? POSIXL : POSIXA;
14215 /* under /a could be alpha */
14217 if (ASCII_RESTRICTED) {
14218 namedclass = ANYOF_ALPHA + (namedclass % 2);
14226 /* The rest have more possibilities depending on the charset.
14227 * We take advantage of the enum ordering of the charset
14228 * modifiers to get the exact node type, */
14230 op = POSIXD + get_regex_charset(RExC_flags);
14231 if (op > POSIXA) { /* /aa is same as /a */
14236 /* The odd numbered ones are the complements of the
14237 * next-lower even number one */
14238 if (namedclass % 2 == 1) {
14242 arg = namedclass_to_classnum(namedclass);
14246 else if (value == prevvalue) {
14248 /* Here, the class consists of just a single code point */
14251 if (! LOC && value == '\n') {
14252 op = REG_ANY; /* Optimize [^\n] */
14253 *flagp |= HASWIDTH|SIMPLE;
14257 else if (value < 256 || UTF) {
14259 /* Optimize a single value into an EXACTish node, but not if it
14260 * would require converting the pattern to UTF-8. */
14261 op = compute_EXACTish(pRExC_state);
14263 } /* Otherwise is a range */
14264 else if (! LOC) { /* locale could vary these */
14265 if (prevvalue == '0') {
14266 if (value == '9') {
14273 /* Here, we have changed <op> away from its initial value iff we found
14274 * an optimization */
14277 /* Throw away this ANYOF regnode, and emit the calculated one,
14278 * which should correspond to the beginning, not current, state of
14280 const char * cur_parse = RExC_parse;
14281 RExC_parse = (char *)orig_parse;
14285 /* To get locale nodes to not use the full ANYOF size would
14286 * require moving the code above that writes the portions
14287 * of it that aren't in other nodes to after this point.
14288 * e.g. ANYOF_POSIXL_SET */
14289 RExC_size = orig_size;
14293 RExC_emit = (regnode *)orig_emit;
14294 if (PL_regkind[op] == POSIXD) {
14295 if (op == POSIXL) {
14296 RExC_contains_locale = 1;
14299 op += NPOSIXD - POSIXD;
14304 ret = reg_node(pRExC_state, op);
14306 if (PL_regkind[op] == POSIXD || PL_regkind[op] == NPOSIXD) {
14310 *flagp |= HASWIDTH|SIMPLE;
14312 else if (PL_regkind[op] == EXACT) {
14313 alloc_maybe_populate_EXACT(pRExC_state, ret, flagp, 0, value,
14314 TRUE /* downgradable to EXACT */
14318 RExC_parse = (char *) cur_parse;
14320 SvREFCNT_dec(posixes);
14321 SvREFCNT_dec(nposixes);
14322 SvREFCNT_dec(cp_list);
14323 SvREFCNT_dec(cp_foldable_list);
14330 /****** !SIZE_ONLY (Pass 2) AFTER HERE *********/
14332 /* If folding, we calculate all characters that could fold to or from the
14333 * ones already on the list */
14334 if (cp_foldable_list) {
14336 UV start, end; /* End points of code point ranges */
14338 SV* fold_intersection = NULL;
14341 /* Our calculated list will be for Unicode rules. For locale
14342 * matching, we have to keep a separate list that is consulted at
14343 * runtime only when the locale indicates Unicode rules. For
14344 * non-locale, we just use to the general list */
14346 use_list = &only_utf8_locale_list;
14349 use_list = &cp_list;
14352 /* Only the characters in this class that participate in folds need
14353 * be checked. Get the intersection of this class and all the
14354 * possible characters that are foldable. This can quickly narrow
14355 * down a large class */
14356 _invlist_intersection(PL_utf8_foldable, cp_foldable_list,
14357 &fold_intersection);
14359 /* The folds for all the Latin1 characters are hard-coded into this
14360 * program, but we have to go out to disk to get the others. */
14361 if (invlist_highest(cp_foldable_list) >= 256) {
14363 /* This is a hash that for a particular fold gives all
14364 * characters that are involved in it */
14365 if (! PL_utf8_foldclosures) {
14367 /* If the folds haven't been read in, call a fold function
14369 if (! PL_utf8_tofold) {
14370 U8 dummy[UTF8_MAXBYTES_CASE+1];
14372 /* This string is just a short named one above \xff */
14373 to_utf8_fold((U8*) HYPHEN_UTF8, dummy, NULL);
14374 assert(PL_utf8_tofold); /* Verify that worked */
14376 PL_utf8_foldclosures
14377 = _swash_inversion_hash(PL_utf8_tofold);
14381 /* Now look at the foldable characters in this class individually */
14382 invlist_iterinit(fold_intersection);
14383 while (invlist_iternext(fold_intersection, &start, &end)) {
14386 /* Look at every character in the range */
14387 for (j = start; j <= end; j++) {
14388 U8 foldbuf[UTF8_MAXBYTES_CASE+1];
14394 /* We have the latin1 folding rules hard-coded here so
14395 * that an innocent-looking character class, like
14396 * /[ks]/i won't have to go out to disk to find the
14397 * possible matches. XXX It would be better to
14398 * generate these via regen, in case a new version of
14399 * the Unicode standard adds new mappings, though that
14400 * is not really likely, and may be caught by the
14401 * default: case of the switch below. */
14403 if (IS_IN_SOME_FOLD_L1(j)) {
14405 /* ASCII is always matched; non-ASCII is matched
14406 * only under Unicode rules (which could happen
14407 * under /l if the locale is a UTF-8 one */
14408 if (isASCII(j) || ! DEPENDS_SEMANTICS) {
14409 *use_list = add_cp_to_invlist(*use_list,
14410 PL_fold_latin1[j]);
14414 add_cp_to_invlist(depends_list,
14415 PL_fold_latin1[j]);
14419 if (HAS_NONLATIN1_FOLD_CLOSURE(j)
14420 && (! isASCII(j) || ! ASCII_FOLD_RESTRICTED))
14422 /* Certain Latin1 characters have matches outside
14423 * Latin1. To get here, <j> is one of those
14424 * characters. None of these matches is valid for
14425 * ASCII characters under /aa, which is why the 'if'
14426 * just above excludes those. These matches only
14427 * happen when the target string is utf8. The code
14428 * below adds the single fold closures for <j> to the
14429 * inversion list. */
14435 add_cp_to_invlist(*use_list, KELVIN_SIGN);
14439 *use_list = add_cp_to_invlist(*use_list,
14440 LATIN_SMALL_LETTER_LONG_S);
14443 *use_list = add_cp_to_invlist(*use_list,
14444 GREEK_CAPITAL_LETTER_MU);
14445 *use_list = add_cp_to_invlist(*use_list,
14446 GREEK_SMALL_LETTER_MU);
14448 case LATIN_CAPITAL_LETTER_A_WITH_RING_ABOVE:
14449 case LATIN_SMALL_LETTER_A_WITH_RING_ABOVE:
14451 add_cp_to_invlist(*use_list, ANGSTROM_SIGN);
14453 case LATIN_SMALL_LETTER_Y_WITH_DIAERESIS:
14454 *use_list = add_cp_to_invlist(*use_list,
14455 LATIN_CAPITAL_LETTER_Y_WITH_DIAERESIS);
14457 case LATIN_SMALL_LETTER_SHARP_S:
14458 *use_list = add_cp_to_invlist(*use_list,
14459 LATIN_CAPITAL_LETTER_SHARP_S);
14461 case 'F': case 'f':
14462 case 'I': case 'i':
14463 case 'L': case 'l':
14464 case 'T': case 't':
14465 case 'A': case 'a':
14466 case 'H': case 'h':
14467 case 'J': case 'j':
14468 case 'N': case 'n':
14469 case 'W': case 'w':
14470 case 'Y': case 'y':
14471 /* These all are targets of multi-character
14472 * folds from code points that require UTF8
14473 * to express, so they can't match unless
14474 * the target string is in UTF-8, so no
14475 * action here is necessary, as regexec.c
14476 * properly handles the general case for
14477 * UTF-8 matching and multi-char folds */
14480 /* Use deprecated warning to increase the
14481 * chances of this being output */
14482 ckWARN2reg_d(RExC_parse, "Perl folding rules are not up-to-date for 0x%"UVXf"; please use the perlbug utility to report;", j);
14489 /* Here is an above Latin1 character. We don't have the
14490 * rules hard-coded for it. First, get its fold. This is
14491 * the simple fold, as the multi-character folds have been
14492 * handled earlier and separated out */
14493 _to_uni_fold_flags(j, foldbuf, &foldlen,
14494 (ASCII_FOLD_RESTRICTED)
14495 ? FOLD_FLAGS_NOMIX_ASCII
14498 /* Single character fold of above Latin1. Add everything in
14499 * its fold closure to the list that this node should match.
14500 * The fold closures data structure is a hash with the keys
14501 * being the UTF-8 of every character that is folded to, like
14502 * 'k', and the values each an array of all code points that
14503 * fold to its key. e.g. [ 'k', 'K', KELVIN_SIGN ].
14504 * Multi-character folds are not included */
14505 if ((listp = hv_fetch(PL_utf8_foldclosures,
14506 (char *) foldbuf, foldlen, FALSE)))
14508 AV* list = (AV*) *listp;
14510 for (k = 0; k <= av_tindex(list); k++) {
14511 SV** c_p = av_fetch(list, k, FALSE);
14514 Perl_croak(aTHX_ "panic: invalid PL_utf8_foldclosures structure");
14518 /* /aa doesn't allow folds between ASCII and non- */
14519 if ((ASCII_FOLD_RESTRICTED
14520 && (isASCII(c) != isASCII(j))))
14525 /* Folds under /l which cross the 255/256 boundary
14526 * are added to a separate list. (These are valid
14527 * only when the locale is UTF-8.) */
14528 if (c < 256 && LOC) {
14529 *use_list = add_cp_to_invlist(*use_list, c);
14533 if (isASCII(c) || c > 255 || AT_LEAST_UNI_SEMANTICS)
14535 cp_list = add_cp_to_invlist(cp_list, c);
14538 /* Similarly folds involving non-ascii Latin1
14539 * characters under /d are added to their list */
14540 depends_list = add_cp_to_invlist(depends_list,
14547 SvREFCNT_dec_NN(fold_intersection);
14550 /* Now that we have finished adding all the folds, there is no reason
14551 * to keep the foldable list separate */
14552 _invlist_union(cp_list, cp_foldable_list, &cp_list);
14553 SvREFCNT_dec_NN(cp_foldable_list);
14556 /* And combine the result (if any) with any inversion list from posix
14557 * classes. The lists are kept separate up to now because we don't want to
14558 * fold the classes (folding of those is automatically handled by the swash
14559 * fetching code) */
14560 if (posixes || nposixes) {
14561 if (posixes && AT_LEAST_ASCII_RESTRICTED) {
14562 /* Under /a and /aa, nothing above ASCII matches these */
14563 _invlist_intersection(posixes,
14564 PL_XPosix_ptrs[_CC_ASCII],
14568 if (DEPENDS_SEMANTICS) {
14569 /* Under /d, everything in the upper half of the Latin1 range
14570 * matches these complements */
14571 ANYOF_FLAGS(ret) |= ANYOF_NON_UTF8_NON_ASCII_ALL;
14573 else if (AT_LEAST_ASCII_RESTRICTED) {
14574 /* Under /a and /aa, everything above ASCII matches these
14576 _invlist_union_complement_2nd(nposixes,
14577 PL_XPosix_ptrs[_CC_ASCII],
14581 _invlist_union(posixes, nposixes, &posixes);
14582 SvREFCNT_dec_NN(nposixes);
14585 posixes = nposixes;
14588 if (! DEPENDS_SEMANTICS) {
14590 _invlist_union(cp_list, posixes, &cp_list);
14591 SvREFCNT_dec_NN(posixes);
14598 /* Under /d, we put into a separate list the Latin1 things that
14599 * match only when the target string is utf8 */
14600 SV* nonascii_but_latin1_properties = NULL;
14601 _invlist_intersection(posixes, PL_UpperLatin1,
14602 &nonascii_but_latin1_properties);
14603 _invlist_subtract(posixes, nonascii_but_latin1_properties,
14606 _invlist_union(cp_list, posixes, &cp_list);
14607 SvREFCNT_dec_NN(posixes);
14613 if (depends_list) {
14614 _invlist_union(depends_list, nonascii_but_latin1_properties,
14616 SvREFCNT_dec_NN(nonascii_but_latin1_properties);
14619 depends_list = nonascii_but_latin1_properties;
14624 /* And combine the result (if any) with any inversion list from properties.
14625 * The lists are kept separate up to now so that we can distinguish the two
14626 * in regards to matching above-Unicode. A run-time warning is generated
14627 * if a Unicode property is matched against a non-Unicode code point. But,
14628 * we allow user-defined properties to match anything, without any warning,
14629 * and we also suppress the warning if there is a portion of the character
14630 * class that isn't a Unicode property, and which matches above Unicode, \W
14631 * or [\x{110000}] for example.
14632 * (Note that in this case, unlike the Posix one above, there is no
14633 * <depends_list>, because having a Unicode property forces Unicode
14638 /* If it matters to the final outcome, see if a non-property
14639 * component of the class matches above Unicode. If so, the
14640 * warning gets suppressed. This is true even if just a single
14641 * such code point is specified, as though not strictly correct if
14642 * another such code point is matched against, the fact that they
14643 * are using above-Unicode code points indicates they should know
14644 * the issues involved */
14646 warn_super = ! (invert
14647 ^ (invlist_highest(cp_list) > PERL_UNICODE_MAX));
14650 _invlist_union(properties, cp_list, &cp_list);
14651 SvREFCNT_dec_NN(properties);
14654 cp_list = properties;
14658 ANYOF_FLAGS(ret) |= ANYOF_WARN_SUPER;
14662 /* Here, we have calculated what code points should be in the character
14665 * Now we can see about various optimizations. Fold calculation (which we
14666 * did above) needs to take place before inversion. Otherwise /[^k]/i
14667 * would invert to include K, which under /i would match k, which it
14668 * shouldn't. Therefore we can't invert folded locale now, as it won't be
14669 * folded until runtime */
14671 /* If we didn't do folding, it's because some information isn't available
14672 * until runtime; set the run-time fold flag for these. (We don't have to
14673 * worry about properties folding, as that is taken care of by the swash
14674 * fetching). We know to set the flag if we have a non-NULL list for UTF-8
14675 * locales, or the class matches at least one 0-255 range code point */
14677 if (only_utf8_locale_list) {
14678 ANYOF_FLAGS(ret) |= ANYOF_LOC_FOLD;
14680 else if (cp_list) { /* Look to see if there a 0-255 code point is in
14683 invlist_iterinit(cp_list);
14684 if (invlist_iternext(cp_list, &start, &end) && start < 256) {
14685 ANYOF_FLAGS(ret) |= ANYOF_LOC_FOLD;
14687 invlist_iterfinish(cp_list);
14691 /* Optimize inverted simple patterns (e.g. [^a-z]) when everything is known
14692 * at compile time. Besides not inverting folded locale now, we can't
14693 * invert if there are things such as \w, which aren't known until runtime
14697 && ! (ANYOF_FLAGS(ret) & (ANYOF_LOCALE_FLAGS))
14699 && ! HAS_NONLOCALE_RUNTIME_PROPERTY_DEFINITION)
14701 _invlist_invert(cp_list);
14703 /* Any swash can't be used as-is, because we've inverted things */
14705 SvREFCNT_dec_NN(swash);
14709 /* Clear the invert flag since have just done it here */
14714 *ret_invlist = cp_list;
14715 SvREFCNT_dec(swash);
14717 /* Discard the generated node */
14719 RExC_size = orig_size;
14722 RExC_emit = orig_emit;
14727 /* Some character classes are equivalent to other nodes. Such nodes take
14728 * up less room and generally fewer operations to execute than ANYOF nodes.
14729 * Above, we checked for and optimized into some such equivalents for
14730 * certain common classes that are easy to test. Getting to this point in
14731 * the code means that the class didn't get optimized there. Since this
14732 * code is only executed in Pass 2, it is too late to save space--it has
14733 * been allocated in Pass 1, and currently isn't given back. But turning
14734 * things into an EXACTish node can allow the optimizer to join it to any
14735 * adjacent such nodes. And if the class is equivalent to things like /./,
14736 * expensive run-time swashes can be avoided. Now that we have more
14737 * complete information, we can find things necessarily missed by the
14738 * earlier code. I (khw) am not sure how much to look for here. It would
14739 * be easy, but perhaps too slow, to check any candidates against all the
14740 * node types they could possibly match using _invlistEQ(). */
14745 && ! (ANYOF_FLAGS(ret) & (ANYOF_LOCALE_FLAGS))
14746 && ! HAS_NONLOCALE_RUNTIME_PROPERTY_DEFINITION
14748 /* We don't optimize if we are supposed to make sure all non-Unicode
14749 * code points raise a warning, as only ANYOF nodes have this check.
14751 && ! ((ANYOF_FLAGS(ret) | ANYOF_WARN_SUPER) && ALWAYS_WARN_SUPER))
14754 U8 op = END; /* The optimzation node-type */
14755 const char * cur_parse= RExC_parse;
14757 invlist_iterinit(cp_list);
14758 if (! invlist_iternext(cp_list, &start, &end)) {
14760 /* Here, the list is empty. This happens, for example, when a
14761 * Unicode property is the only thing in the character class, and
14762 * it doesn't match anything. (perluniprops.pod notes such
14765 *flagp |= HASWIDTH|SIMPLE;
14767 else if (start == end) { /* The range is a single code point */
14768 if (! invlist_iternext(cp_list, &start, &end)
14770 /* Don't do this optimization if it would require changing
14771 * the pattern to UTF-8 */
14772 && (start < 256 || UTF))
14774 /* Here, the list contains a single code point. Can optimize
14775 * into an EXACTish node */
14784 /* A locale node under folding with one code point can be
14785 * an EXACTFL, as its fold won't be calculated until
14791 /* Here, we are generally folding, but there is only one
14792 * code point to match. If we have to, we use an EXACT
14793 * node, but it would be better for joining with adjacent
14794 * nodes in the optimization pass if we used the same
14795 * EXACTFish node that any such are likely to be. We can
14796 * do this iff the code point doesn't participate in any
14797 * folds. For example, an EXACTF of a colon is the same as
14798 * an EXACT one, since nothing folds to or from a colon. */
14800 if (IS_IN_SOME_FOLD_L1(value)) {
14805 if (_invlist_contains_cp(PL_utf8_foldable, value)) {
14810 /* If we haven't found the node type, above, it means we
14811 * can use the prevailing one */
14813 op = compute_EXACTish(pRExC_state);
14818 else if (start == 0) {
14819 if (end == UV_MAX) {
14821 *flagp |= HASWIDTH|SIMPLE;
14824 else if (end == '\n' - 1
14825 && invlist_iternext(cp_list, &start, &end)
14826 && start == '\n' + 1 && end == UV_MAX)
14829 *flagp |= HASWIDTH|SIMPLE;
14833 invlist_iterfinish(cp_list);
14836 RExC_parse = (char *)orig_parse;
14837 RExC_emit = (regnode *)orig_emit;
14839 ret = reg_node(pRExC_state, op);
14841 RExC_parse = (char *)cur_parse;
14843 if (PL_regkind[op] == EXACT) {
14844 alloc_maybe_populate_EXACT(pRExC_state, ret, flagp, 0, value,
14845 TRUE /* downgradable to EXACT */
14849 SvREFCNT_dec_NN(cp_list);
14854 /* Here, <cp_list> contains all the code points we can determine at
14855 * compile time that match under all conditions. Go through it, and
14856 * for things that belong in the bitmap, put them there, and delete from
14857 * <cp_list>. While we are at it, see if everything above 255 is in the
14858 * list, and if so, set a flag to speed up execution */
14860 populate_ANYOF_from_invlist(ret, &cp_list);
14863 ANYOF_FLAGS(ret) |= ANYOF_INVERT;
14866 /* Here, the bitmap has been populated with all the Latin1 code points that
14867 * always match. Can now add to the overall list those that match only
14868 * when the target string is UTF-8 (<depends_list>). */
14869 if (depends_list) {
14871 _invlist_union(cp_list, depends_list, &cp_list);
14872 SvREFCNT_dec_NN(depends_list);
14875 cp_list = depends_list;
14877 ANYOF_FLAGS(ret) |= ANYOF_UTF8;
14880 /* If there is a swash and more than one element, we can't use the swash in
14881 * the optimization below. */
14882 if (swash && element_count > 1) {
14883 SvREFCNT_dec_NN(swash);
14887 set_ANYOF_arg(pRExC_state, ret, cp_list,
14888 (HAS_NONLOCALE_RUNTIME_PROPERTY_DEFINITION)
14890 only_utf8_locale_list,
14891 swash, has_user_defined_property);
14893 *flagp |= HASWIDTH|SIMPLE;
14895 if (ANYOF_FLAGS(ret) & ANYOF_LOCALE_FLAGS) {
14896 RExC_contains_locale = 1;
14902 #undef HAS_NONLOCALE_RUNTIME_PROPERTY_DEFINITION
14905 S_set_ANYOF_arg(pTHX_ RExC_state_t* const pRExC_state,
14906 regnode* const node,
14908 SV* const runtime_defns,
14909 SV* const only_utf8_locale_list,
14911 const bool has_user_defined_property)
14913 /* Sets the arg field of an ANYOF-type node 'node', using information about
14914 * the node passed-in. If there is nothing outside the node's bitmap, the
14915 * arg is set to ANYOF_NONBITMAP_EMPTY. Otherwise, it sets the argument to
14916 * the count returned by add_data(), having allocated and stored an array,
14917 * av, that that count references, as follows:
14918 * av[0] stores the character class description in its textual form.
14919 * This is used later (regexec.c:Perl_regclass_swash()) to
14920 * initialize the appropriate swash, and is also useful for dumping
14921 * the regnode. This is set to &PL_sv_undef if the textual
14922 * description is not needed at run-time (as happens if the other
14923 * elements completely define the class)
14924 * av[1] if &PL_sv_undef, is a placeholder to later contain the swash
14925 * computed from av[0]. But if no further computation need be done,
14926 * the swash is stored here now (and av[0] is &PL_sv_undef).
14927 * av[2] stores the inversion list of code points that match only if the
14928 * current locale is UTF-8
14929 * av[3] stores the cp_list inversion list for use in addition or instead
14930 * of av[0]; used only if cp_list exists and av[1] is &PL_sv_undef.
14931 * (Otherwise everything needed is already in av[0] and av[1])
14932 * av[4] is set if any component of the class is from a user-defined
14933 * property; used only if av[3] exists */
14937 PERL_ARGS_ASSERT_SET_ANYOF_ARG;
14939 if (! cp_list && ! runtime_defns && ! only_utf8_locale_list) {
14940 assert(! (ANYOF_FLAGS(node)
14941 & (ANYOF_UTF8|ANYOF_NONBITMAP_NON_UTF8)));
14942 ARG_SET(node, ANYOF_NONBITMAP_EMPTY);
14945 AV * const av = newAV();
14948 assert(ANYOF_FLAGS(node)
14949 & (ANYOF_UTF8|ANYOF_NONBITMAP_NON_UTF8|ANYOF_LOC_FOLD));
14951 av_store(av, 0, (runtime_defns)
14952 ? SvREFCNT_inc(runtime_defns) : &PL_sv_undef);
14954 av_store(av, 1, swash);
14955 SvREFCNT_dec_NN(cp_list);
14958 av_store(av, 1, &PL_sv_undef);
14960 av_store(av, 3, cp_list);
14961 av_store(av, 4, newSVuv(has_user_defined_property));
14965 if (only_utf8_locale_list) {
14966 av_store(av, 2, only_utf8_locale_list);
14969 av_store(av, 2, &PL_sv_undef);
14972 rv = newRV_noinc(MUTABLE_SV(av));
14973 n = add_data(pRExC_state, STR_WITH_LEN("s"));
14974 RExC_rxi->data->data[n] = (void*)rv;
14980 /* reg_skipcomment()
14982 Absorbs an /x style # comments from the input stream.
14983 Returns true if there is more text remaining in the stream.
14984 Will set the REG_RUN_ON_COMMENT_SEEN flag if the comment
14985 terminates the pattern without including a newline.
14987 Note its the callers responsibility to ensure that we are
14988 actually in /x mode
14993 S_reg_skipcomment(pTHX_ RExC_state_t *pRExC_state)
14997 PERL_ARGS_ASSERT_REG_SKIPCOMMENT;
14999 while (RExC_parse < RExC_end)
15000 if (*RExC_parse++ == '\n') {
15005 /* we ran off the end of the pattern without ending
15006 the comment, so we have to add an \n when wrapping */
15007 RExC_seen |= REG_RUN_ON_COMMENT_SEEN;
15015 Advances the parse position, and optionally absorbs
15016 "whitespace" from the inputstream.
15018 Without /x "whitespace" means (?#...) style comments only,
15019 with /x this means (?#...) and # comments and whitespace proper.
15021 Returns the RExC_parse point from BEFORE the scan occurs.
15023 This is the /x friendly way of saying RExC_parse++.
15027 S_nextchar(pTHX_ RExC_state_t *pRExC_state)
15029 char* const retval = RExC_parse++;
15031 PERL_ARGS_ASSERT_NEXTCHAR;
15034 if (RExC_end - RExC_parse >= 3
15035 && *RExC_parse == '('
15036 && RExC_parse[1] == '?'
15037 && RExC_parse[2] == '#')
15039 while (*RExC_parse != ')') {
15040 if (RExC_parse == RExC_end)
15041 FAIL("Sequence (?#... not terminated");
15047 if (RExC_flags & RXf_PMf_EXTENDED) {
15048 if (isSPACE(*RExC_parse)) {
15052 else if (*RExC_parse == '#') {
15053 if ( reg_skipcomment( pRExC_state ) )
15062 - reg_node - emit a node
15064 STATIC regnode * /* Location. */
15065 S_reg_node(pTHX_ RExC_state_t *pRExC_state, U8 op)
15069 regnode * const ret = RExC_emit;
15070 GET_RE_DEBUG_FLAGS_DECL;
15072 PERL_ARGS_ASSERT_REG_NODE;
15075 SIZE_ALIGN(RExC_size);
15079 if (RExC_emit >= RExC_emit_bound)
15080 Perl_croak(aTHX_ "panic: reg_node overrun trying to emit %d, %p>=%p",
15081 op, RExC_emit, RExC_emit_bound);
15083 NODE_ALIGN_FILL(ret);
15085 FILL_ADVANCE_NODE(ptr, op);
15086 REH_CALL_COMP_NODE_HOOK(pRExC_state->rx, (ptr) - 1);
15087 #ifdef RE_TRACK_PATTERN_OFFSETS
15088 if (RExC_offsets) { /* MJD */
15090 ("%s:%d: (op %s) %s %"UVuf" (len %"UVuf") (max %"UVuf").\n",
15091 "reg_node", __LINE__,
15093 (UV)(RExC_emit - RExC_emit_start) > RExC_offsets[0]
15094 ? "Overwriting end of array!\n" : "OK",
15095 (UV)(RExC_emit - RExC_emit_start),
15096 (UV)(RExC_parse - RExC_start),
15097 (UV)RExC_offsets[0]));
15098 Set_Node_Offset(RExC_emit, RExC_parse + (op == END));
15106 - reganode - emit a node with an argument
15108 STATIC regnode * /* Location. */
15109 S_reganode(pTHX_ RExC_state_t *pRExC_state, U8 op, U32 arg)
15113 regnode * const ret = RExC_emit;
15114 GET_RE_DEBUG_FLAGS_DECL;
15116 PERL_ARGS_ASSERT_REGANODE;
15119 SIZE_ALIGN(RExC_size);
15124 assert(2==regarglen[op]+1);
15126 Anything larger than this has to allocate the extra amount.
15127 If we changed this to be:
15129 RExC_size += (1 + regarglen[op]);
15131 then it wouldn't matter. Its not clear what side effect
15132 might come from that so its not done so far.
15137 if (RExC_emit >= RExC_emit_bound)
15138 Perl_croak(aTHX_ "panic: reg_node overrun trying to emit %d, %p>=%p",
15139 op, RExC_emit, RExC_emit_bound);
15141 NODE_ALIGN_FILL(ret);
15143 FILL_ADVANCE_NODE_ARG(ptr, op, arg);
15144 REH_CALL_COMP_NODE_HOOK(pRExC_state->rx, (ptr) - 2);
15145 #ifdef RE_TRACK_PATTERN_OFFSETS
15146 if (RExC_offsets) { /* MJD */
15148 ("%s(%d): (op %s) %s %"UVuf" <- %"UVuf" (max %"UVuf").\n",
15152 (UV)(RExC_emit - RExC_emit_start) > RExC_offsets[0] ?
15153 "Overwriting end of array!\n" : "OK",
15154 (UV)(RExC_emit - RExC_emit_start),
15155 (UV)(RExC_parse - RExC_start),
15156 (UV)RExC_offsets[0]));
15157 Set_Cur_Node_Offset;
15165 - reguni - emit (if appropriate) a Unicode character
15167 PERL_STATIC_INLINE STRLEN
15168 S_reguni(pTHX_ const RExC_state_t *pRExC_state, UV uv, char* s)
15172 PERL_ARGS_ASSERT_REGUNI;
15174 return SIZE_ONLY ? UNISKIP(uv) : (uvchr_to_utf8((U8*)s, uv) - (U8*)s);
15178 - reginsert - insert an operator in front of already-emitted operand
15180 * Means relocating the operand.
15183 S_reginsert(pTHX_ RExC_state_t *pRExC_state, U8 op, regnode *opnd, U32 depth)
15189 const int offset = regarglen[(U8)op];
15190 const int size = NODE_STEP_REGNODE + offset;
15191 GET_RE_DEBUG_FLAGS_DECL;
15193 PERL_ARGS_ASSERT_REGINSERT;
15194 PERL_UNUSED_ARG(depth);
15195 /* (PL_regkind[(U8)op] == CURLY ? EXTRA_STEP_2ARGS : 0); */
15196 DEBUG_PARSE_FMT("inst"," - %s",PL_reg_name[op]);
15205 if (RExC_open_parens) {
15207 /*DEBUG_PARSE_FMT("inst"," - %"IVdf, (IV)RExC_npar);*/
15208 for ( paren=0 ; paren < RExC_npar ; paren++ ) {
15209 if ( RExC_open_parens[paren] >= opnd ) {
15210 /*DEBUG_PARSE_FMT("open"," - %d",size);*/
15211 RExC_open_parens[paren] += size;
15213 /*DEBUG_PARSE_FMT("open"," - %s","ok");*/
15215 if ( RExC_close_parens[paren] >= opnd ) {
15216 /*DEBUG_PARSE_FMT("close"," - %d",size);*/
15217 RExC_close_parens[paren] += size;
15219 /*DEBUG_PARSE_FMT("close"," - %s","ok");*/
15224 while (src > opnd) {
15225 StructCopy(--src, --dst, regnode);
15226 #ifdef RE_TRACK_PATTERN_OFFSETS
15227 if (RExC_offsets) { /* MJD 20010112 */
15229 ("%s(%d): (op %s) %s copy %"UVuf" -> %"UVuf" (max %"UVuf").\n",
15233 (UV)(dst - RExC_emit_start) > RExC_offsets[0]
15234 ? "Overwriting end of array!\n" : "OK",
15235 (UV)(src - RExC_emit_start),
15236 (UV)(dst - RExC_emit_start),
15237 (UV)RExC_offsets[0]));
15238 Set_Node_Offset_To_R(dst-RExC_emit_start, Node_Offset(src));
15239 Set_Node_Length_To_R(dst-RExC_emit_start, Node_Length(src));
15245 place = opnd; /* Op node, where operand used to be. */
15246 #ifdef RE_TRACK_PATTERN_OFFSETS
15247 if (RExC_offsets) { /* MJD */
15249 ("%s(%d): (op %s) %s %"UVuf" <- %"UVuf" (max %"UVuf").\n",
15253 (UV)(place - RExC_emit_start) > RExC_offsets[0]
15254 ? "Overwriting end of array!\n" : "OK",
15255 (UV)(place - RExC_emit_start),
15256 (UV)(RExC_parse - RExC_start),
15257 (UV)RExC_offsets[0]));
15258 Set_Node_Offset(place, RExC_parse);
15259 Set_Node_Length(place, 1);
15262 src = NEXTOPER(place);
15263 FILL_ADVANCE_NODE(place, op);
15264 REH_CALL_COMP_NODE_HOOK(pRExC_state->rx, (place) - 1);
15265 Zero(src, offset, regnode);
15269 - regtail - set the next-pointer at the end of a node chain of p to val.
15270 - SEE ALSO: regtail_study
15272 /* TODO: All three parms should be const */
15274 S_regtail(pTHX_ RExC_state_t *pRExC_state, regnode *p,
15275 const regnode *val,U32 depth)
15279 GET_RE_DEBUG_FLAGS_DECL;
15281 PERL_ARGS_ASSERT_REGTAIL;
15283 PERL_UNUSED_ARG(depth);
15289 /* Find last node. */
15292 regnode * const temp = regnext(scan);
15294 SV * const mysv=sv_newmortal();
15295 DEBUG_PARSE_MSG((scan==p ? "tail" : ""));
15296 regprop(RExC_rx, mysv, scan, NULL);
15297 PerlIO_printf(Perl_debug_log, "~ %s (%d) %s %s\n",
15298 SvPV_nolen_const(mysv), REG_NODE_NUM(scan),
15299 (temp == NULL ? "->" : ""),
15300 (temp == NULL ? PL_reg_name[OP(val)] : "")
15308 if (reg_off_by_arg[OP(scan)]) {
15309 ARG_SET(scan, val - scan);
15312 NEXT_OFF(scan) = val - scan;
15318 - regtail_study - set the next-pointer at the end of a node chain of p to val.
15319 - Look for optimizable sequences at the same time.
15320 - currently only looks for EXACT chains.
15322 This is experimental code. The idea is to use this routine to perform
15323 in place optimizations on branches and groups as they are constructed,
15324 with the long term intention of removing optimization from study_chunk so
15325 that it is purely analytical.
15327 Currently only used when in DEBUG mode. The macro REGTAIL_STUDY() is used
15328 to control which is which.
15331 /* TODO: All four parms should be const */
15334 S_regtail_study(pTHX_ RExC_state_t *pRExC_state, regnode *p,
15335 const regnode *val,U32 depth)
15340 #ifdef EXPERIMENTAL_INPLACESCAN
15343 GET_RE_DEBUG_FLAGS_DECL;
15345 PERL_ARGS_ASSERT_REGTAIL_STUDY;
15351 /* Find last node. */
15355 regnode * const temp = regnext(scan);
15356 #ifdef EXPERIMENTAL_INPLACESCAN
15357 if (PL_regkind[OP(scan)] == EXACT) {
15358 bool unfolded_multi_char; /* Unexamined in this routine */
15359 if (join_exact(pRExC_state, scan, &min,
15360 &unfolded_multi_char, 1, val, depth+1))
15365 switch (OP(scan)) {
15368 case EXACTFA_NO_TRIE:
15373 if( exact == PSEUDO )
15375 else if ( exact != OP(scan) )
15384 SV * const mysv=sv_newmortal();
15385 DEBUG_PARSE_MSG((scan==p ? "tsdy" : ""));
15386 regprop(RExC_rx, mysv, scan, NULL);
15387 PerlIO_printf(Perl_debug_log, "~ %s (%d) -> %s\n",
15388 SvPV_nolen_const(mysv),
15389 REG_NODE_NUM(scan),
15390 PL_reg_name[exact]);
15397 SV * const mysv_val=sv_newmortal();
15398 DEBUG_PARSE_MSG("");
15399 regprop(RExC_rx, mysv_val, val, NULL);
15400 PerlIO_printf(Perl_debug_log,
15401 "~ attach to %s (%"IVdf") offset to %"IVdf"\n",
15402 SvPV_nolen_const(mysv_val),
15403 (IV)REG_NODE_NUM(val),
15407 if (reg_off_by_arg[OP(scan)]) {
15408 ARG_SET(scan, val - scan);
15411 NEXT_OFF(scan) = val - scan;
15419 - regdump - dump a regexp onto Perl_debug_log in vaguely comprehensible form
15424 S_regdump_intflags(pTHX_ const char *lead, const U32 flags)
15429 ASSUME(REG_INTFLAGS_NAME_SIZE <= sizeof(flags)*8);
15431 for (bit=0; bit<REG_INTFLAGS_NAME_SIZE; bit++) {
15432 if (flags & (1<<bit)) {
15433 if (!set++ && lead)
15434 PerlIO_printf(Perl_debug_log, "%s",lead);
15435 PerlIO_printf(Perl_debug_log, "%s ",PL_reg_intflags_name[bit]);
15440 PerlIO_printf(Perl_debug_log, "\n");
15442 PerlIO_printf(Perl_debug_log, "%s[none-set]\n",lead);
15447 S_regdump_extflags(pTHX_ const char *lead, const U32 flags)
15453 ASSUME(REG_EXTFLAGS_NAME_SIZE <= sizeof(flags)*8);
15455 for (bit=0; bit<REG_EXTFLAGS_NAME_SIZE; bit++) {
15456 if (flags & (1<<bit)) {
15457 if ((1<<bit) & RXf_PMf_CHARSET) { /* Output separately, below */
15460 if (!set++ && lead)
15461 PerlIO_printf(Perl_debug_log, "%s",lead);
15462 PerlIO_printf(Perl_debug_log, "%s ",PL_reg_extflags_name[bit]);
15465 if ((cs = get_regex_charset(flags)) != REGEX_DEPENDS_CHARSET) {
15466 if (!set++ && lead) {
15467 PerlIO_printf(Perl_debug_log, "%s",lead);
15470 case REGEX_UNICODE_CHARSET:
15471 PerlIO_printf(Perl_debug_log, "UNICODE");
15473 case REGEX_LOCALE_CHARSET:
15474 PerlIO_printf(Perl_debug_log, "LOCALE");
15476 case REGEX_ASCII_RESTRICTED_CHARSET:
15477 PerlIO_printf(Perl_debug_log, "ASCII-RESTRICTED");
15479 case REGEX_ASCII_MORE_RESTRICTED_CHARSET:
15480 PerlIO_printf(Perl_debug_log, "ASCII-MORE_RESTRICTED");
15483 PerlIO_printf(Perl_debug_log, "UNKNOWN CHARACTER SET");
15489 PerlIO_printf(Perl_debug_log, "\n");
15491 PerlIO_printf(Perl_debug_log, "%s[none-set]\n",lead);
15497 Perl_regdump(pTHX_ const regexp *r)
15501 SV * const sv = sv_newmortal();
15502 SV *dsv= sv_newmortal();
15503 RXi_GET_DECL(r,ri);
15504 GET_RE_DEBUG_FLAGS_DECL;
15506 PERL_ARGS_ASSERT_REGDUMP;
15508 (void)dumpuntil(r, ri->program, ri->program + 1, NULL, NULL, sv, 0, 0);
15510 /* Header fields of interest. */
15511 if (r->anchored_substr) {
15512 RE_PV_QUOTED_DECL(s, 0, dsv, SvPVX_const(r->anchored_substr),
15513 RE_SV_DUMPLEN(r->anchored_substr), 30);
15514 PerlIO_printf(Perl_debug_log,
15515 "anchored %s%s at %"IVdf" ",
15516 s, RE_SV_TAIL(r->anchored_substr),
15517 (IV)r->anchored_offset);
15518 } else if (r->anchored_utf8) {
15519 RE_PV_QUOTED_DECL(s, 1, dsv, SvPVX_const(r->anchored_utf8),
15520 RE_SV_DUMPLEN(r->anchored_utf8), 30);
15521 PerlIO_printf(Perl_debug_log,
15522 "anchored utf8 %s%s at %"IVdf" ",
15523 s, RE_SV_TAIL(r->anchored_utf8),
15524 (IV)r->anchored_offset);
15526 if (r->float_substr) {
15527 RE_PV_QUOTED_DECL(s, 0, dsv, SvPVX_const(r->float_substr),
15528 RE_SV_DUMPLEN(r->float_substr), 30);
15529 PerlIO_printf(Perl_debug_log,
15530 "floating %s%s at %"IVdf"..%"UVuf" ",
15531 s, RE_SV_TAIL(r->float_substr),
15532 (IV)r->float_min_offset, (UV)r->float_max_offset);
15533 } else if (r->float_utf8) {
15534 RE_PV_QUOTED_DECL(s, 1, dsv, SvPVX_const(r->float_utf8),
15535 RE_SV_DUMPLEN(r->float_utf8), 30);
15536 PerlIO_printf(Perl_debug_log,
15537 "floating utf8 %s%s at %"IVdf"..%"UVuf" ",
15538 s, RE_SV_TAIL(r->float_utf8),
15539 (IV)r->float_min_offset, (UV)r->float_max_offset);
15541 if (r->check_substr || r->check_utf8)
15542 PerlIO_printf(Perl_debug_log,
15544 (r->check_substr == r->float_substr
15545 && r->check_utf8 == r->float_utf8
15546 ? "(checking floating" : "(checking anchored"));
15547 if (r->intflags & PREGf_NOSCAN)
15548 PerlIO_printf(Perl_debug_log, " noscan");
15549 if (r->extflags & RXf_CHECK_ALL)
15550 PerlIO_printf(Perl_debug_log, " isall");
15551 if (r->check_substr || r->check_utf8)
15552 PerlIO_printf(Perl_debug_log, ") ");
15554 if (ri->regstclass) {
15555 regprop(r, sv, ri->regstclass, NULL);
15556 PerlIO_printf(Perl_debug_log, "stclass %s ", SvPVX_const(sv));
15558 if (r->intflags & PREGf_ANCH) {
15559 PerlIO_printf(Perl_debug_log, "anchored");
15560 if (r->intflags & PREGf_ANCH_BOL)
15561 PerlIO_printf(Perl_debug_log, "(BOL)");
15562 if (r->intflags & PREGf_ANCH_MBOL)
15563 PerlIO_printf(Perl_debug_log, "(MBOL)");
15564 if (r->intflags & PREGf_ANCH_SBOL)
15565 PerlIO_printf(Perl_debug_log, "(SBOL)");
15566 if (r->intflags & PREGf_ANCH_GPOS)
15567 PerlIO_printf(Perl_debug_log, "(GPOS)");
15568 PerlIO_putc(Perl_debug_log, ' ');
15570 if (r->intflags & PREGf_GPOS_SEEN)
15571 PerlIO_printf(Perl_debug_log, "GPOS:%"UVuf" ", (UV)r->gofs);
15572 if (r->intflags & PREGf_SKIP)
15573 PerlIO_printf(Perl_debug_log, "plus ");
15574 if (r->intflags & PREGf_IMPLICIT)
15575 PerlIO_printf(Perl_debug_log, "implicit ");
15576 PerlIO_printf(Perl_debug_log, "minlen %"IVdf" ", (IV)r->minlen);
15577 if (r->extflags & RXf_EVAL_SEEN)
15578 PerlIO_printf(Perl_debug_log, "with eval ");
15579 PerlIO_printf(Perl_debug_log, "\n");
15581 regdump_extflags("r->extflags: ",r->extflags);
15582 regdump_intflags("r->intflags: ",r->intflags);
15585 PERL_ARGS_ASSERT_REGDUMP;
15586 PERL_UNUSED_CONTEXT;
15587 PERL_UNUSED_ARG(r);
15588 #endif /* DEBUGGING */
15592 - regprop - printable representation of opcode, with run time support
15596 Perl_regprop(pTHX_ const regexp *prog, SV *sv, const regnode *o, const regmatch_info *reginfo)
15602 /* Should be synchronized with * ANYOF_ #xdefines in regcomp.h */
15603 static const char * const anyofs[] = {
15604 #if _CC_WORDCHAR != 0 || _CC_DIGIT != 1 || _CC_ALPHA != 2 || _CC_LOWER != 3 \
15605 || _CC_UPPER != 4 || _CC_PUNCT != 5 || _CC_PRINT != 6 \
15606 || _CC_ALPHANUMERIC != 7 || _CC_GRAPH != 8 || _CC_CASED != 9 \
15607 || _CC_SPACE != 10 || _CC_BLANK != 11 || _CC_XDIGIT != 12 \
15608 || _CC_PSXSPC != 13 || _CC_CNTRL != 14 || _CC_ASCII != 15 \
15609 || _CC_VERTSPACE != 16
15610 #error Need to adjust order of anyofs[]
15647 RXi_GET_DECL(prog,progi);
15648 GET_RE_DEBUG_FLAGS_DECL;
15650 PERL_ARGS_ASSERT_REGPROP;
15654 if (OP(o) > REGNODE_MAX) /* regnode.type is unsigned */
15655 /* It would be nice to FAIL() here, but this may be called from
15656 regexec.c, and it would be hard to supply pRExC_state. */
15657 Perl_croak(aTHX_ "Corrupted regexp opcode %d > %d",
15658 (int)OP(o), (int)REGNODE_MAX);
15659 sv_catpv(sv, PL_reg_name[OP(o)]); /* Take off const! */
15661 k = PL_regkind[OP(o)];
15664 sv_catpvs(sv, " ");
15665 /* Using is_utf8_string() (via PERL_PV_UNI_DETECT)
15666 * is a crude hack but it may be the best for now since
15667 * we have no flag "this EXACTish node was UTF-8"
15669 pv_pretty(sv, STRING(o), STR_LEN(o), 60, PL_colors[0], PL_colors[1],
15670 PERL_PV_ESCAPE_UNI_DETECT |
15671 PERL_PV_ESCAPE_NONASCII |
15672 PERL_PV_PRETTY_ELLIPSES |
15673 PERL_PV_PRETTY_LTGT |
15674 PERL_PV_PRETTY_NOCLEAR
15676 } else if (k == TRIE) {
15677 /* print the details of the trie in dumpuntil instead, as
15678 * progi->data isn't available here */
15679 const char op = OP(o);
15680 const U32 n = ARG(o);
15681 const reg_ac_data * const ac = IS_TRIE_AC(op) ?
15682 (reg_ac_data *)progi->data->data[n] :
15684 const reg_trie_data * const trie
15685 = (reg_trie_data*)progi->data->data[!IS_TRIE_AC(op) ? n : ac->trie];
15687 Perl_sv_catpvf(aTHX_ sv, "-%s",PL_reg_name[o->flags]);
15688 DEBUG_TRIE_COMPILE_r(
15689 Perl_sv_catpvf(aTHX_ sv,
15690 "<S:%"UVuf"/%"IVdf" W:%"UVuf" L:%"UVuf"/%"UVuf" C:%"UVuf"/%"UVuf">",
15691 (UV)trie->startstate,
15692 (IV)trie->statecount-1, /* -1 because of the unused 0 element */
15693 (UV)trie->wordcount,
15696 (UV)TRIE_CHARCOUNT(trie),
15697 (UV)trie->uniquecharcount
15700 if ( IS_ANYOF_TRIE(op) || trie->bitmap ) {
15701 sv_catpvs(sv, "[");
15702 (void) put_latin1_charclass_innards(sv, IS_ANYOF_TRIE(op)
15704 : TRIE_BITMAP(trie));
15705 sv_catpvs(sv, "]");
15708 } else if (k == CURLY) {
15709 if (OP(o) == CURLYM || OP(o) == CURLYN || OP(o) == CURLYX)
15710 Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* Parenth number */
15711 Perl_sv_catpvf(aTHX_ sv, " {%d,%d}", ARG1(o), ARG2(o));
15713 else if (k == WHILEM && o->flags) /* Ordinal/of */
15714 Perl_sv_catpvf(aTHX_ sv, "[%d/%d]", o->flags & 0xf, o->flags>>4);
15715 else if (k == REF || k == OPEN || k == CLOSE
15716 || k == GROUPP || OP(o)==ACCEPT)
15718 Perl_sv_catpvf(aTHX_ sv, "%d", (int)ARG(o)); /* Parenth number */
15719 if ( RXp_PAREN_NAMES(prog) ) {
15720 if ( k != REF || (OP(o) < NREF)) {
15721 AV *list= MUTABLE_AV(progi->data->data[progi->name_list_idx]);
15722 SV **name= av_fetch(list, ARG(o), 0 );
15724 Perl_sv_catpvf(aTHX_ sv, " '%"SVf"'", SVfARG(*name));
15727 AV *list= MUTABLE_AV(progi->data->data[ progi->name_list_idx ]);
15728 SV *sv_dat= MUTABLE_SV(progi->data->data[ ARG( o ) ]);
15729 I32 *nums=(I32*)SvPVX(sv_dat);
15730 SV **name= av_fetch(list, nums[0], 0 );
15733 for ( n=0; n<SvIVX(sv_dat); n++ ) {
15734 Perl_sv_catpvf(aTHX_ sv, "%s%"IVdf,
15735 (n ? "," : ""), (IV)nums[n]);
15737 Perl_sv_catpvf(aTHX_ sv, " '%"SVf"'", SVfARG(*name));
15741 if ( k == REF && reginfo) {
15742 U32 n = ARG(o); /* which paren pair */
15743 I32 ln = prog->offs[n].start;
15744 if (prog->lastparen < n || ln == -1)
15745 Perl_sv_catpvf(aTHX_ sv, ": FAIL");
15746 else if (ln == prog->offs[n].end)
15747 Perl_sv_catpvf(aTHX_ sv, ": ACCEPT - EMPTY STRING");
15749 const char *s = reginfo->strbeg + ln;
15750 Perl_sv_catpvf(aTHX_ sv, ": ");
15751 Perl_pv_pretty( aTHX_ sv, s, prog->offs[n].end - prog->offs[n].start, 32, 0, 0,
15752 PERL_PV_ESCAPE_UNI_DETECT|PERL_PV_PRETTY_NOCLEAR|PERL_PV_PRETTY_ELLIPSES|PERL_PV_PRETTY_QUOTE );
15755 } else if (k == GOSUB)
15756 /* Paren and offset */
15757 Perl_sv_catpvf(aTHX_ sv, "%d[%+d]", (int)ARG(o),(int)ARG2L(o));
15758 else if (k == VERB) {
15760 Perl_sv_catpvf(aTHX_ sv, ":%"SVf,
15761 SVfARG((MUTABLE_SV(progi->data->data[ ARG( o ) ]))));
15762 } else if (k == LOGICAL)
15763 /* 2: embedded, otherwise 1 */
15764 Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags);
15765 else if (k == ANYOF) {
15766 const U8 flags = ANYOF_FLAGS(o);
15770 if (flags & ANYOF_LOCALE_FLAGS)
15771 sv_catpvs(sv, "{loc}");
15772 if (flags & ANYOF_LOC_FOLD)
15773 sv_catpvs(sv, "{i}");
15774 Perl_sv_catpvf(aTHX_ sv, "[%s", PL_colors[0]);
15775 if (flags & ANYOF_INVERT)
15776 sv_catpvs(sv, "^");
15778 /* output what the standard cp 0-255 bitmap matches */
15779 do_sep = put_latin1_charclass_innards(sv, ANYOF_BITMAP(o));
15781 /* output any special charclass tests (used entirely under use
15783 if (ANYOF_POSIXL_TEST_ANY_SET(o)) {
15785 for (i = 0; i < ANYOF_POSIXL_MAX; i++) {
15786 if (ANYOF_POSIXL_TEST(o,i)) {
15787 sv_catpv(sv, anyofs[i]);
15793 if ((flags & (ANYOF_ABOVE_LATIN1_ALL
15795 |ANYOF_NONBITMAP_NON_UTF8
15799 Perl_sv_catpvf(aTHX_ sv,"%s][%s",PL_colors[1],PL_colors[0]);
15800 if (flags & ANYOF_INVERT)
15801 /*make sure the invert info is in each */
15802 sv_catpvs(sv, "^");
15805 if (flags & ANYOF_NON_UTF8_NON_ASCII_ALL) {
15806 sv_catpvs(sv, "{non-utf8-latin1-all}");
15809 /* output information about the unicode matching */
15810 if (flags & ANYOF_ABOVE_LATIN1_ALL)
15811 sv_catpvs(sv, "{unicode_all}");
15812 else if (ARG(o) != ANYOF_NONBITMAP_EMPTY) {
15813 SV *lv; /* Set if there is something outside the bit map. */
15814 bool byte_output = FALSE; /* If something in the bitmap has
15816 SV *only_utf8_locale;
15818 /* Get the stuff that wasn't in the bitmap */
15819 (void) _get_regclass_nonbitmap_data(prog, o, FALSE,
15820 &lv, &only_utf8_locale);
15821 if (lv && lv != &PL_sv_undef) {
15822 char *s = savesvpv(lv);
15823 char * const origs = s;
15825 while (*s && *s != '\n')
15829 const char * const t = ++s;
15831 if (flags & ANYOF_NONBITMAP_NON_UTF8) {
15832 sv_catpvs(sv, "{outside bitmap}");
15835 sv_catpvs(sv, "{utf8}");
15839 sv_catpvs(sv, " ");
15845 /* Truncate very long output */
15846 if (s - origs > 256) {
15847 Perl_sv_catpvf(aTHX_ sv,
15849 (int) (s - origs - 1),
15855 else if (*s == '\t') {
15869 SvREFCNT_dec_NN(lv);
15872 if ((flags & ANYOF_LOC_FOLD)
15873 && only_utf8_locale
15874 && only_utf8_locale != &PL_sv_undef)
15877 int max_entries = 256;
15879 sv_catpvs(sv, "{utf8 locale}");
15880 invlist_iterinit(only_utf8_locale);
15881 while (invlist_iternext(only_utf8_locale,
15883 put_range(sv, start, end);
15885 if (max_entries < 0) {
15886 sv_catpvs(sv, "...");
15890 invlist_iterfinish(only_utf8_locale);
15895 Perl_sv_catpvf(aTHX_ sv, "%s]", PL_colors[1]);
15897 else if (k == POSIXD || k == NPOSIXD) {
15898 U8 index = FLAGS(o) * 2;
15899 if (index < C_ARRAY_LENGTH(anyofs)) {
15900 if (*anyofs[index] != '[') {
15903 sv_catpv(sv, anyofs[index]);
15904 if (*anyofs[index] != '[') {
15909 Perl_sv_catpvf(aTHX_ sv, "[illegal type=%d])", index);
15912 else if (k == BRANCHJ && (OP(o) == UNLESSM || OP(o) == IFMATCH))
15913 Perl_sv_catpvf(aTHX_ sv, "[%d]", -(o->flags));
15915 PERL_UNUSED_CONTEXT;
15916 PERL_UNUSED_ARG(sv);
15917 PERL_UNUSED_ARG(o);
15918 PERL_UNUSED_ARG(prog);
15919 PERL_UNUSED_ARG(reginfo);
15920 #endif /* DEBUGGING */
15926 Perl_re_intuit_string(pTHX_ REGEXP * const r)
15927 { /* Assume that RE_INTUIT is set */
15929 struct regexp *const prog = ReANY(r);
15930 GET_RE_DEBUG_FLAGS_DECL;
15932 PERL_ARGS_ASSERT_RE_INTUIT_STRING;
15933 PERL_UNUSED_CONTEXT;
15937 const char * const s = SvPV_nolen_const(prog->check_substr
15938 ? prog->check_substr : prog->check_utf8);
15940 if (!PL_colorset) reginitcolors();
15941 PerlIO_printf(Perl_debug_log,
15942 "%sUsing REx %ssubstr:%s \"%s%.60s%s%s\"\n",
15944 prog->check_substr ? "" : "utf8 ",
15945 PL_colors[5],PL_colors[0],
15948 (strlen(s) > 60 ? "..." : ""));
15951 return prog->check_substr ? prog->check_substr : prog->check_utf8;
15957 handles refcounting and freeing the perl core regexp structure. When
15958 it is necessary to actually free the structure the first thing it
15959 does is call the 'free' method of the regexp_engine associated to
15960 the regexp, allowing the handling of the void *pprivate; member
15961 first. (This routine is not overridable by extensions, which is why
15962 the extensions free is called first.)
15964 See regdupe and regdupe_internal if you change anything here.
15966 #ifndef PERL_IN_XSUB_RE
15968 Perl_pregfree(pTHX_ REGEXP *r)
15974 Perl_pregfree2(pTHX_ REGEXP *rx)
15977 struct regexp *const r = ReANY(rx);
15978 GET_RE_DEBUG_FLAGS_DECL;
15980 PERL_ARGS_ASSERT_PREGFREE2;
15982 if (r->mother_re) {
15983 ReREFCNT_dec(r->mother_re);
15985 CALLREGFREE_PVT(rx); /* free the private data */
15986 SvREFCNT_dec(RXp_PAREN_NAMES(r));
15987 Safefree(r->xpv_len_u.xpvlenu_pv);
15990 SvREFCNT_dec(r->anchored_substr);
15991 SvREFCNT_dec(r->anchored_utf8);
15992 SvREFCNT_dec(r->float_substr);
15993 SvREFCNT_dec(r->float_utf8);
15994 Safefree(r->substrs);
15996 RX_MATCH_COPY_FREE(rx);
15997 #ifdef PERL_ANY_COW
15998 SvREFCNT_dec(r->saved_copy);
16001 SvREFCNT_dec(r->qr_anoncv);
16002 rx->sv_u.svu_rx = 0;
16007 This is a hacky workaround to the structural issue of match results
16008 being stored in the regexp structure which is in turn stored in
16009 PL_curpm/PL_reg_curpm. The problem is that due to qr// the pattern
16010 could be PL_curpm in multiple contexts, and could require multiple
16011 result sets being associated with the pattern simultaneously, such
16012 as when doing a recursive match with (??{$qr})
16014 The solution is to make a lightweight copy of the regexp structure
16015 when a qr// is returned from the code executed by (??{$qr}) this
16016 lightweight copy doesn't actually own any of its data except for
16017 the starp/end and the actual regexp structure itself.
16023 Perl_reg_temp_copy (pTHX_ REGEXP *ret_x, REGEXP *rx)
16025 struct regexp *ret;
16026 struct regexp *const r = ReANY(rx);
16027 const bool islv = ret_x && SvTYPE(ret_x) == SVt_PVLV;
16029 PERL_ARGS_ASSERT_REG_TEMP_COPY;
16032 ret_x = (REGEXP*) newSV_type(SVt_REGEXP);
16034 SvOK_off((SV *)ret_x);
16036 /* For PVLVs, SvANY points to the xpvlv body while sv_u points
16037 to the regexp. (For SVt_REGEXPs, sv_upgrade has already
16038 made both spots point to the same regexp body.) */
16039 REGEXP *temp = (REGEXP *)newSV_type(SVt_REGEXP);
16040 assert(!SvPVX(ret_x));
16041 ret_x->sv_u.svu_rx = temp->sv_any;
16042 temp->sv_any = NULL;
16043 SvFLAGS(temp) = (SvFLAGS(temp) & ~SVTYPEMASK) | SVt_NULL;
16044 SvREFCNT_dec_NN(temp);
16045 /* SvCUR still resides in the xpvlv struct, so the regexp copy-
16046 ing below will not set it. */
16047 SvCUR_set(ret_x, SvCUR(rx));
16050 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
16051 sv_force_normal(sv) is called. */
16053 ret = ReANY(ret_x);
16055 SvFLAGS(ret_x) |= SvUTF8(rx);
16056 /* We share the same string buffer as the original regexp, on which we
16057 hold a reference count, incremented when mother_re is set below.
16058 The string pointer is copied here, being part of the regexp struct.
16060 memcpy(&(ret->xpv_cur), &(r->xpv_cur),
16061 sizeof(regexp) - STRUCT_OFFSET(regexp, xpv_cur));
16063 const I32 npar = r->nparens+1;
16064 Newx(ret->offs, npar, regexp_paren_pair);
16065 Copy(r->offs, ret->offs, npar, regexp_paren_pair);
16068 Newx(ret->substrs, 1, struct reg_substr_data);
16069 StructCopy(r->substrs, ret->substrs, struct reg_substr_data);
16071 SvREFCNT_inc_void(ret->anchored_substr);
16072 SvREFCNT_inc_void(ret->anchored_utf8);
16073 SvREFCNT_inc_void(ret->float_substr);
16074 SvREFCNT_inc_void(ret->float_utf8);
16076 /* check_substr and check_utf8, if non-NULL, point to either their
16077 anchored or float namesakes, and don't hold a second reference. */
16079 RX_MATCH_COPIED_off(ret_x);
16080 #ifdef PERL_ANY_COW
16081 ret->saved_copy = NULL;
16083 ret->mother_re = ReREFCNT_inc(r->mother_re ? r->mother_re : rx);
16084 SvREFCNT_inc_void(ret->qr_anoncv);
16090 /* regfree_internal()
16092 Free the private data in a regexp. This is overloadable by
16093 extensions. Perl takes care of the regexp structure in pregfree(),
16094 this covers the *pprivate pointer which technically perl doesn't
16095 know about, however of course we have to handle the
16096 regexp_internal structure when no extension is in use.
16098 Note this is called before freeing anything in the regexp
16103 Perl_regfree_internal(pTHX_ REGEXP * const rx)
16106 struct regexp *const r = ReANY(rx);
16107 RXi_GET_DECL(r,ri);
16108 GET_RE_DEBUG_FLAGS_DECL;
16110 PERL_ARGS_ASSERT_REGFREE_INTERNAL;
16116 SV *dsv= sv_newmortal();
16117 RE_PV_QUOTED_DECL(s, RX_UTF8(rx),
16118 dsv, RX_PRECOMP(rx), RX_PRELEN(rx), 60);
16119 PerlIO_printf(Perl_debug_log,"%sFreeing REx:%s %s\n",
16120 PL_colors[4],PL_colors[5],s);
16123 #ifdef RE_TRACK_PATTERN_OFFSETS
16125 Safefree(ri->u.offsets); /* 20010421 MJD */
16127 if (ri->code_blocks) {
16129 for (n = 0; n < ri->num_code_blocks; n++)
16130 SvREFCNT_dec(ri->code_blocks[n].src_regex);
16131 Safefree(ri->code_blocks);
16135 int n = ri->data->count;
16138 /* If you add a ->what type here, update the comment in regcomp.h */
16139 switch (ri->data->what[n]) {
16145 SvREFCNT_dec(MUTABLE_SV(ri->data->data[n]));
16148 Safefree(ri->data->data[n]);
16154 { /* Aho Corasick add-on structure for a trie node.
16155 Used in stclass optimization only */
16157 reg_ac_data *aho=(reg_ac_data*)ri->data->data[n];
16159 refcount = --aho->refcount;
16162 PerlMemShared_free(aho->states);
16163 PerlMemShared_free(aho->fail);
16164 /* do this last!!!! */
16165 PerlMemShared_free(ri->data->data[n]);
16166 PerlMemShared_free(ri->regstclass);
16172 /* trie structure. */
16174 reg_trie_data *trie=(reg_trie_data*)ri->data->data[n];
16176 refcount = --trie->refcount;
16179 PerlMemShared_free(trie->charmap);
16180 PerlMemShared_free(trie->states);
16181 PerlMemShared_free(trie->trans);
16183 PerlMemShared_free(trie->bitmap);
16185 PerlMemShared_free(trie->jump);
16186 PerlMemShared_free(trie->wordinfo);
16187 /* do this last!!!! */
16188 PerlMemShared_free(ri->data->data[n]);
16193 Perl_croak(aTHX_ "panic: regfree data code '%c'",
16194 ri->data->what[n]);
16197 Safefree(ri->data->what);
16198 Safefree(ri->data);
16204 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
16205 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
16206 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
16209 re_dup - duplicate a regexp.
16211 This routine is expected to clone a given regexp structure. It is only
16212 compiled under USE_ITHREADS.
16214 After all of the core data stored in struct regexp is duplicated
16215 the regexp_engine.dupe method is used to copy any private data
16216 stored in the *pprivate pointer. This allows extensions to handle
16217 any duplication it needs to do.
16219 See pregfree() and regfree_internal() if you change anything here.
16221 #if defined(USE_ITHREADS)
16222 #ifndef PERL_IN_XSUB_RE
16224 Perl_re_dup_guts(pTHX_ const REGEXP *sstr, REGEXP *dstr, CLONE_PARAMS *param)
16228 const struct regexp *r = ReANY(sstr);
16229 struct regexp *ret = ReANY(dstr);
16231 PERL_ARGS_ASSERT_RE_DUP_GUTS;
16233 npar = r->nparens+1;
16234 Newx(ret->offs, npar, regexp_paren_pair);
16235 Copy(r->offs, ret->offs, npar, regexp_paren_pair);
16237 if (ret->substrs) {
16238 /* Do it this way to avoid reading from *r after the StructCopy().
16239 That way, if any of the sv_dup_inc()s dislodge *r from the L1
16240 cache, it doesn't matter. */
16241 const bool anchored = r->check_substr
16242 ? r->check_substr == r->anchored_substr
16243 : r->check_utf8 == r->anchored_utf8;
16244 Newx(ret->substrs, 1, struct reg_substr_data);
16245 StructCopy(r->substrs, ret->substrs, struct reg_substr_data);
16247 ret->anchored_substr = sv_dup_inc(ret->anchored_substr, param);
16248 ret->anchored_utf8 = sv_dup_inc(ret->anchored_utf8, param);
16249 ret->float_substr = sv_dup_inc(ret->float_substr, param);
16250 ret->float_utf8 = sv_dup_inc(ret->float_utf8, param);
16252 /* check_substr and check_utf8, if non-NULL, point to either their
16253 anchored or float namesakes, and don't hold a second reference. */
16255 if (ret->check_substr) {
16257 assert(r->check_utf8 == r->anchored_utf8);
16258 ret->check_substr = ret->anchored_substr;
16259 ret->check_utf8 = ret->anchored_utf8;
16261 assert(r->check_substr == r->float_substr);
16262 assert(r->check_utf8 == r->float_utf8);
16263 ret->check_substr = ret->float_substr;
16264 ret->check_utf8 = ret->float_utf8;
16266 } else if (ret->check_utf8) {
16268 ret->check_utf8 = ret->anchored_utf8;
16270 ret->check_utf8 = ret->float_utf8;
16275 RXp_PAREN_NAMES(ret) = hv_dup_inc(RXp_PAREN_NAMES(ret), param);
16276 ret->qr_anoncv = MUTABLE_CV(sv_dup_inc((const SV *)ret->qr_anoncv, param));
16279 RXi_SET(ret,CALLREGDUPE_PVT(dstr,param));
16281 if (RX_MATCH_COPIED(dstr))
16282 ret->subbeg = SAVEPVN(ret->subbeg, ret->sublen);
16284 ret->subbeg = NULL;
16285 #ifdef PERL_ANY_COW
16286 ret->saved_copy = NULL;
16289 /* Whether mother_re be set or no, we need to copy the string. We
16290 cannot refrain from copying it when the storage points directly to
16291 our mother regexp, because that's
16292 1: a buffer in a different thread
16293 2: something we no longer hold a reference on
16294 so we need to copy it locally. */
16295 RX_WRAPPED(dstr) = SAVEPVN(RX_WRAPPED(sstr), SvCUR(sstr)+1);
16296 ret->mother_re = NULL;
16298 #endif /* PERL_IN_XSUB_RE */
16303 This is the internal complement to regdupe() which is used to copy
16304 the structure pointed to by the *pprivate pointer in the regexp.
16305 This is the core version of the extension overridable cloning hook.
16306 The regexp structure being duplicated will be copied by perl prior
16307 to this and will be provided as the regexp *r argument, however
16308 with the /old/ structures pprivate pointer value. Thus this routine
16309 may override any copying normally done by perl.
16311 It returns a pointer to the new regexp_internal structure.
16315 Perl_regdupe_internal(pTHX_ REGEXP * const rx, CLONE_PARAMS *param)
16318 struct regexp *const r = ReANY(rx);
16319 regexp_internal *reti;
16321 RXi_GET_DECL(r,ri);
16323 PERL_ARGS_ASSERT_REGDUPE_INTERNAL;
16327 Newxc(reti, sizeof(regexp_internal) + len*sizeof(regnode),
16328 char, regexp_internal);
16329 Copy(ri->program, reti->program, len+1, regnode);
16331 reti->num_code_blocks = ri->num_code_blocks;
16332 if (ri->code_blocks) {
16334 Newxc(reti->code_blocks, ri->num_code_blocks, struct reg_code_block,
16335 struct reg_code_block);
16336 Copy(ri->code_blocks, reti->code_blocks, ri->num_code_blocks,
16337 struct reg_code_block);
16338 for (n = 0; n < ri->num_code_blocks; n++)
16339 reti->code_blocks[n].src_regex = (REGEXP*)
16340 sv_dup_inc((SV*)(ri->code_blocks[n].src_regex), param);
16343 reti->code_blocks = NULL;
16345 reti->regstclass = NULL;
16348 struct reg_data *d;
16349 const int count = ri->data->count;
16352 Newxc(d, sizeof(struct reg_data) + count*sizeof(void *),
16353 char, struct reg_data);
16354 Newx(d->what, count, U8);
16357 for (i = 0; i < count; i++) {
16358 d->what[i] = ri->data->what[i];
16359 switch (d->what[i]) {
16360 /* see also regcomp.h and regfree_internal() */
16361 case 'a': /* actually an AV, but the dup function is identical. */
16365 case 'u': /* actually an HV, but the dup function is identical. */
16366 d->data[i] = sv_dup_inc((const SV *)ri->data->data[i], param);
16369 /* This is cheating. */
16370 Newx(d->data[i], 1, regnode_ssc);
16371 StructCopy(ri->data->data[i], d->data[i], regnode_ssc);
16372 reti->regstclass = (regnode*)d->data[i];
16375 /* Trie stclasses are readonly and can thus be shared
16376 * without duplication. We free the stclass in pregfree
16377 * when the corresponding reg_ac_data struct is freed.
16379 reti->regstclass= ri->regstclass;
16383 ((reg_trie_data*)ri->data->data[i])->refcount++;
16388 d->data[i] = ri->data->data[i];
16391 Perl_croak(aTHX_ "panic: re_dup unknown data code '%c'",
16392 ri->data->what[i]);
16401 reti->name_list_idx = ri->name_list_idx;
16403 #ifdef RE_TRACK_PATTERN_OFFSETS
16404 if (ri->u.offsets) {
16405 Newx(reti->u.offsets, 2*len+1, U32);
16406 Copy(ri->u.offsets, reti->u.offsets, 2*len+1, U32);
16409 SetProgLen(reti,len);
16412 return (void*)reti;
16415 #endif /* USE_ITHREADS */
16417 #ifndef PERL_IN_XSUB_RE
16420 - regnext - dig the "next" pointer out of a node
16423 Perl_regnext(pTHX_ regnode *p)
16431 if (OP(p) > REGNODE_MAX) { /* regnode.type is unsigned */
16432 Perl_croak(aTHX_ "Corrupted regexp opcode %d > %d",
16433 (int)OP(p), (int)REGNODE_MAX);
16436 offset = (reg_off_by_arg[OP(p)] ? ARG(p) : NEXT_OFF(p));
16445 S_re_croak2(pTHX_ bool utf8, const char* pat1,const char* pat2,...)
16448 STRLEN l1 = strlen(pat1);
16449 STRLEN l2 = strlen(pat2);
16452 const char *message;
16454 PERL_ARGS_ASSERT_RE_CROAK2;
16460 Copy(pat1, buf, l1 , char);
16461 Copy(pat2, buf + l1, l2 , char);
16462 buf[l1 + l2] = '\n';
16463 buf[l1 + l2 + 1] = '\0';
16464 va_start(args, pat2);
16465 msv = vmess(buf, &args);
16467 message = SvPV_const(msv,l1);
16470 Copy(message, buf, l1 , char);
16471 /* l1-1 to avoid \n */
16472 Perl_croak(aTHX_ "%"UTF8f, UTF8fARG(utf8, l1-1, buf));
16475 /* XXX Here's a total kludge. But we need to re-enter for swash routines. */
16477 #ifndef PERL_IN_XSUB_RE
16479 Perl_save_re_context(pTHX)
16483 /* Save $1..$n (#18107: UTF-8 s/(\w+)/uc($1)/e); AMS 20021106. */
16485 const REGEXP * const rx = PM_GETRE(PL_curpm);
16488 for (i = 1; i <= RX_NPARENS(rx); i++) {
16489 char digits[TYPE_CHARS(long)];
16490 const STRLEN len = my_snprintf(digits, sizeof(digits),
16492 GV *const *const gvp
16493 = (GV**)hv_fetch(PL_defstash, digits, len, 0);
16496 GV * const gv = *gvp;
16497 if (SvTYPE(gv) == SVt_PVGV && GvSV(gv))
16509 S_put_byte(pTHX_ SV *sv, int c)
16511 PERL_ARGS_ASSERT_PUT_BYTE;
16515 case '\r': Perl_sv_catpvf(aTHX_ sv, "\\r"); break;
16516 case '\n': Perl_sv_catpvf(aTHX_ sv, "\\n"); break;
16517 case '\t': Perl_sv_catpvf(aTHX_ sv, "\\t"); break;
16518 case '\f': Perl_sv_catpvf(aTHX_ sv, "\\f"); break;
16519 case '\a': Perl_sv_catpvf(aTHX_ sv, "\\a"); break;
16522 Perl_sv_catpvf(aTHX_ sv, "\\x{%x}", c);
16527 const char string = c;
16528 if (c == '-' || c == ']' || c == '\\' || c == '^')
16529 sv_catpvs(sv, "\\");
16530 sv_catpvn(sv, &string, 1);
16535 S_put_range(pTHX_ SV *sv, UV start, UV end)
16538 /* Appends to 'sv' a displayable version of the range of code points from
16539 * 'start' to 'end' */
16541 assert(start <= end);
16543 PERL_ARGS_ASSERT_PUT_RANGE;
16545 if (end - start < 3) { /* Individual chars in short ranges */
16546 for (; start <= end; start++)
16547 put_byte(sv, start);
16549 else if ( end > 255
16550 || ! isALPHANUMERIC(start)
16551 || ! isALPHANUMERIC(end)
16552 || isDIGIT(start) != isDIGIT(end)
16553 || isUPPER(start) != isUPPER(end)
16554 || isLOWER(start) != isLOWER(end)
16556 /* This final test should get optimized out except on EBCDIC
16557 * platforms, where it causes ranges that cross discontinuities
16558 * like i/j to be shown as hex instead of the misleading,
16559 * e.g. H-K (since that range includes more than H, I, J, K).
16561 || (end - start) != NATIVE_TO_ASCII(end) - NATIVE_TO_ASCII(start))
16563 Perl_sv_catpvf(aTHX_ sv, "\\x{%02" UVXf "}-\\x{%02" UVXf "}",
16565 (end < 256) ? end : 255);
16567 else { /* Here, the ends of the range are both digits, or both uppercase,
16568 or both lowercase; and there's no discontinuity in the range
16569 (which could happen on EBCDIC platforms) */
16570 put_byte(sv, start);
16571 sv_catpvs(sv, "-");
16577 S_put_latin1_charclass_innards(pTHX_ SV *sv, char *bitmap)
16579 /* Appends to 'sv' a displayable version of the innards of the bracketed
16580 * character class whose bitmap is 'bitmap'; Returns 'TRUE' if it actually
16581 * output anything */
16584 bool has_output_anything = FALSE;
16586 PERL_ARGS_ASSERT_PUT_LATIN1_CHARCLASS_INNARDS;
16588 for (i = 0; i < 256; i++) {
16589 if (BITMAP_TEST((U8 *) bitmap,i)) {
16591 /* The character at index i should be output. Find the next
16592 * character that should NOT be output */
16594 for (j = i + 1; j < 256; j++) {
16595 if (! BITMAP_TEST((U8 *) bitmap, j)) {
16600 /* Everything between them is a single range that should be output
16602 put_range(sv, i, j - 1);
16603 has_output_anything = TRUE;
16608 return has_output_anything;
16611 #define CLEAR_OPTSTART \
16612 if (optstart) STMT_START { \
16613 DEBUG_OPTIMISE_r(PerlIO_printf(Perl_debug_log, \
16614 " (%"IVdf" nodes)\n", (IV)(node - optstart))); \
16618 #define DUMPUNTIL(b,e) \
16620 node=dumpuntil(r,start,(b),(e),last,sv,indent+1,depth+1);
16622 STATIC const regnode *
16623 S_dumpuntil(pTHX_ const regexp *r, const regnode *start, const regnode *node,
16624 const regnode *last, const regnode *plast,
16625 SV* sv, I32 indent, U32 depth)
16628 U8 op = PSEUDO; /* Arbitrary non-END op. */
16629 const regnode *next;
16630 const regnode *optstart= NULL;
16632 RXi_GET_DECL(r,ri);
16633 GET_RE_DEBUG_FLAGS_DECL;
16635 PERL_ARGS_ASSERT_DUMPUNTIL;
16637 #ifdef DEBUG_DUMPUNTIL
16638 PerlIO_printf(Perl_debug_log, "--- %d : %d - %d - %d\n",indent,node-start,
16639 last ? last-start : 0,plast ? plast-start : 0);
16642 if (plast && plast < last)
16645 while (PL_regkind[op] != END && (!last || node < last)) {
16646 /* While that wasn't END last time... */
16649 if (op == CLOSE || op == WHILEM)
16651 next = regnext((regnode *)node);
16654 if (OP(node) == OPTIMIZED) {
16655 if (!optstart && RE_DEBUG_FLAG(RE_DEBUG_COMPILE_OPTIMISE))
16662 regprop(r, sv, node, NULL);
16663 PerlIO_printf(Perl_debug_log, "%4"IVdf":%*s%s", (IV)(node - start),
16664 (int)(2*indent + 1), "", SvPVX_const(sv));
16666 if (OP(node) != OPTIMIZED) {
16667 if (next == NULL) /* Next ptr. */
16668 PerlIO_printf(Perl_debug_log, " (0)");
16669 else if (PL_regkind[(U8)op] == BRANCH
16670 && PL_regkind[OP(next)] != BRANCH )
16671 PerlIO_printf(Perl_debug_log, " (FAIL)");
16673 PerlIO_printf(Perl_debug_log, " (%"IVdf")", (IV)(next - start));
16674 (void)PerlIO_putc(Perl_debug_log, '\n');
16678 if (PL_regkind[(U8)op] == BRANCHJ) {
16681 const regnode *nnode = (OP(next) == LONGJMP
16682 ? regnext((regnode *)next)
16684 if (last && nnode > last)
16686 DUMPUNTIL(NEXTOPER(NEXTOPER(node)), nnode);
16689 else if (PL_regkind[(U8)op] == BRANCH) {
16691 DUMPUNTIL(NEXTOPER(node), next);
16693 else if ( PL_regkind[(U8)op] == TRIE ) {
16694 const regnode *this_trie = node;
16695 const char op = OP(node);
16696 const U32 n = ARG(node);
16697 const reg_ac_data * const ac = op>=AHOCORASICK ?
16698 (reg_ac_data *)ri->data->data[n] :
16700 const reg_trie_data * const trie =
16701 (reg_trie_data*)ri->data->data[op<AHOCORASICK ? n : ac->trie];
16703 AV *const trie_words
16704 = MUTABLE_AV(ri->data->data[n + TRIE_WORDS_OFFSET]);
16706 const regnode *nextbranch= NULL;
16709 for (word_idx= 0; word_idx < (I32)trie->wordcount; word_idx++) {
16710 SV ** const elem_ptr = av_fetch(trie_words,word_idx,0);
16712 PerlIO_printf(Perl_debug_log, "%*s%s ",
16713 (int)(2*(indent+3)), "",
16715 ? pv_pretty(sv, SvPV_nolen_const(*elem_ptr),
16716 SvCUR(*elem_ptr), 60,
16717 PL_colors[0], PL_colors[1],
16719 ? PERL_PV_ESCAPE_UNI
16721 | PERL_PV_PRETTY_ELLIPSES
16722 | PERL_PV_PRETTY_LTGT
16727 U16 dist= trie->jump[word_idx+1];
16728 PerlIO_printf(Perl_debug_log, "(%"UVuf")\n",
16729 (UV)((dist ? this_trie + dist : next) - start));
16732 nextbranch= this_trie + trie->jump[0];
16733 DUMPUNTIL(this_trie + dist, nextbranch);
16735 if (nextbranch && PL_regkind[OP(nextbranch)]==BRANCH)
16736 nextbranch= regnext((regnode *)nextbranch);
16738 PerlIO_printf(Perl_debug_log, "\n");
16741 if (last && next > last)
16746 else if ( op == CURLY ) { /* "next" might be very big: optimizer */
16747 DUMPUNTIL(NEXTOPER(node) + EXTRA_STEP_2ARGS,
16748 NEXTOPER(node) + EXTRA_STEP_2ARGS + 1);
16750 else if (PL_regkind[(U8)op] == CURLY && op != CURLYX) {
16752 DUMPUNTIL(NEXTOPER(node) + EXTRA_STEP_2ARGS, next);
16754 else if ( op == PLUS || op == STAR) {
16755 DUMPUNTIL(NEXTOPER(node), NEXTOPER(node) + 1);
16757 else if (PL_regkind[(U8)op] == ANYOF) {
16758 /* arglen 1 + class block */
16759 node += 1 + ((ANYOF_FLAGS(node) & ANYOF_POSIXL)
16760 ? ANYOF_POSIXL_SKIP
16762 node = NEXTOPER(node);
16764 else if (PL_regkind[(U8)op] == EXACT) {
16765 /* Literal string, where present. */
16766 node += NODE_SZ_STR(node) - 1;
16767 node = NEXTOPER(node);
16770 node = NEXTOPER(node);
16771 node += regarglen[(U8)op];
16773 if (op == CURLYX || op == OPEN)
16777 #ifdef DEBUG_DUMPUNTIL
16778 PerlIO_printf(Perl_debug_log, "--- %d\n", (int)indent);
16783 #endif /* DEBUGGING */
16787 * c-indentation-style: bsd
16788 * c-basic-offset: 4
16789 * indent-tabs-mode: nil
16792 * ex: set ts=8 sts=4 sw=4 et: