This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
perldelta entries for op_class() and op_dump()
[perl5.git] / regexp.h
CommitLineData
a0d0e21e 1/* regexp.h
d6376244 2 *
4bb101f2 3 * Copyright (C) 1993, 1994, 1996, 1997, 1999, 2000, 2001, 2003,
1129b882 4 * 2005, 2006, 2007, 2008 by Larry Wall and others
d6376244
JH
5 *
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8 *
a0d0e21e
LW
9 */
10
378cc40b
LW
11/*
12 * Definitions etc. for regexp(3) routines.
13 *
14 * Caveat: this is V8 regexp(3) [actually, a reimplementation thereof],
15 * not the System V one.
16 */
be8e71aa 17#ifndef PLUGGABLE_RE_EXTENSION
785a26d5
YO
18/* we don't want to include this stuff if we are inside of
19 an external regex engine based on the core one - like re 'debug'*/
378cc40b 20
79a2a0e8
KW
21#include "utf8.h"
22
c277df42
IZ
23struct regnode {
24 U8 flags;
25 U8 type;
26 U16 next_off;
27};
28
29typedef struct regnode regnode;
30
cad2e5aa 31struct reg_substr_data;
2779dcf1 32
0ee3c8d0
JH
33struct reg_data;
34
f9f4320a 35struct regexp_engine;
28d8d7f4 36struct regexp;
bbe252da 37
785a26d5 38struct reg_substr_datum {
c4828ca0
DM
39 SSize_t min_offset; /* min pos (in chars) that substr must appear */
40 SSize_t max_offset /* max pos (in chars) that substr must appear */;
785a26d5
YO
41 SV *substr; /* non-utf8 variant */
42 SV *utf8_substr; /* utf8 variant */
c4828ca0 43 SSize_t end_shift; /* how many fixed chars must end the string */
785a26d5
YO
44};
45struct reg_substr_data {
6480a6c4 46 U8 check_ix; /* index into data[] of check substr */
785a26d5
YO
47 struct reg_substr_datum data[3]; /* Actual array */
48};
f9f4320a 49
db2c6cb3 50#ifdef PERL_ANY_COW
bbe252da
YO
51#define SV_SAVED_COPY SV *saved_copy; /* If non-NULL, SV which is COW from original */
52#else
53#define SV_SAVED_COPY
ed252734 54#endif
28d8d7f4 55
b3fd53f3
DM
56/* offsets within a string of a particular /(.)/ capture */
57
f0ab9afb 58typedef struct regexp_paren_pair {
99a90e59
FC
59 SSize_t start;
60 SSize_t end;
b3fd53f3
DM
61 /* 'start_tmp' records a new opening position before the matching end
62 * has been found, so that the old start and end values are still
63 * valid, e.g.
64 * "abc" =~ /(.(?{print "[$1]"}))+/
65 *outputs [][a][b]
66 * This field is not part of the API. */
ea3daa5d 67 SSize_t start_tmp;
f0ab9afb 68} regexp_paren_pair;
28d8d7f4 69
52ae8f7e 70#if defined(PERL_IN_REGCOMP_C) || defined(PERL_IN_UTF8_C)
164173a2 71#define _invlist_union(a, b, output) _invlist_union_maybe_complement_2nd(a, b, FALSE, output)
52ae8f7e 72#define _invlist_intersection(a, b, output) _invlist_intersection_maybe_complement_2nd(a, b, FALSE, output)
3f80b571
KW
73
74/* Subtracting b from a leaves in a everything that was there that isn't in b,
75 * that is the intersection of a with b's complement */
76#define _invlist_subtract(a, b, output) _invlist_intersection_maybe_complement_2nd(a, b, TRUE, output)
52ae8f7e
KW
77#endif
78
3d2bd50a
DM
79/* record the position of a (?{...}) within a pattern */
80
81struct reg_code_block {
82 STRLEN start;
83 STRLEN end;
84 OP *block;
b30fcab9 85 REGEXP *src_regex;
3d2bd50a
DM
86};
87
88
882227b7
AB
89/*
90 The regexp/REGEXP struct, see L<perlreapi> for further documentation
91 on the individual fields. The struct is ordered so that the most
92 commonly used fields are placed at the start.
93
94 Any patch that adds items to this struct will need to include
95 changes to F<sv.c> (C<Perl_re_dup()>) and F<regcomp.c>
96 (C<pregfree()>). This involves freeing or cloning items in the
97 regexp's data array based on the data item's type.
98*/
99
08e44740
NC
100#define _REGEXP_COMMON \
101 /* what engine created this regexp? */ \
102 const struct regexp_engine* engine; \
103 REGEXP *mother_re; /* what re is this a lightweight copy of? */ \
89f6f287 104 HV *paren_names; /* Optional hash of paren names */ \
ba6840fb 105 /*--------------------------------------------------------*/ \
08e44740
NC
106 /* Information about the match that the perl core uses to */ \
107 /* manage things */ \
108 U32 extflags; /* Flags used both externally and internally */ \
ea3daa5d
FC
109 SSize_t minlen; /* mininum possible number of chars in string to match */\
110 SSize_t minlenret; /* mininum possible number of chars in $& */ \
111 STRLEN gofs; /* chars left of pos that we search from */ \
08e44740
NC
112 /* substring data about strings that must appear in the */ \
113 /* final match, used for optimisations */ \
114 struct reg_substr_data *substrs; \
115 U32 nparens; /* number of capture buffers */ \
116 /* private engine specific data */ \
117 U32 intflags; /* Engine Specific Internal flags */ \
118 void *pprivate; /* Data private to the regex engine which */ \
119 /* created this object. */ \
ba6840fb 120 /*--------------------------------------------------------*/ \
08e44740
NC
121 /* Data about the last/current match. These are modified */ \
122 /* during matching */ \
123 U32 lastparen; /* last open paren matched */ \
124 U32 lastcloseparen; /* last close paren matched */ \
08e44740
NC
125 /* Array of offsets for (@-) and (@+) */ \
126 regexp_paren_pair *offs; \
ba6840fb
YO
127 char **recurse_locinput; /* used to detect infinite recursion, XXX: move to internal */ \
128 /*--------------------------------------------------------*/ \
08e44740
NC
129 /* saved or original string so \digit works forever. */ \
130 char *subbeg; \
131 SV_SAVED_COPY /* If non-NULL, SV which is COW from original */\
19394178 132 SSize_t sublen; /* Length of string pointed by subbeg */ \
ea3daa5d
FC
133 SSize_t suboffset; /* byte offset of subbeg from logical start of str */ \
134 SSize_t subcoffset; /* suboffset equiv, but in chars (for @-/@+) */ \
08e44740 135 /* Information about the match that isn't often used */ \
ee273784 136 SSize_t maxlen; /* mininum possible number of chars in string to match */\
ba6840fb 137 /*--------------------------------------------------------*/ \
08e44740 138 /* offset from wrapped to the start of precomp */ \
654eccd5 139 PERL_BITFIELD32 pre_prefix:4; \
dbc200c5
YO
140 /* original flags used to compile the pattern, may differ */ \
141 /* from extflags in various ways */ \
142 PERL_BITFIELD32 compflags:9; \
ba6840fb 143 /*--------------------------------------------------------*/ \
b82ad30c 144 CV *qr_anoncv /* the anon sub wrapped round qr/(?{..})/ */
08e44740 145
bbe252da 146typedef struct regexp {
288b8c02 147 _XPV_HEAD;
08e44740 148 _REGEXP_COMMON;
f8fc2ecf
YO
149} regexp;
150
89f6f287 151#define RXp_PAREN_NAMES(rx) ((rx)->paren_names)
5daac39c 152
785a26d5 153/* used for high speed searches */
f9f4320a
YO
154typedef struct re_scream_pos_data_s
155{
156 char **scream_olds; /* match pos */
ea3daa5d 157 SSize_t *scream_pos; /* Internal iterator of scream. */
f9f4320a
YO
158} re_scream_pos_data;
159
785a26d5
YO
160/* regexp_engine structure. This is the dispatch table for regexes.
161 * Any regex engine implementation must be able to build one of these.
162 */
f9f4320a 163typedef struct regexp_engine {
1593ad57 164 REGEXP* (*comp) (pTHX_ SV * const pattern, U32 flags);
49d7dfbc 165 I32 (*exec) (pTHX_ REGEXP * const rx, char* stringarg, char* strend,
ea3daa5d 166 char* strbeg, SSize_t minend, SV* sv,
49d7dfbc 167 void* data, U32 flags);
52a21eb3
DM
168 char* (*intuit) (pTHX_
169 REGEXP * const rx,
170 SV *sv,
171 const char * const strbeg,
172 char *strpos,
173 char *strend,
174 const U32 flags,
9f61653a 175 re_scream_pos_data *data);
49d7dfbc 176 SV* (*checkstr) (pTHX_ REGEXP * const rx);
fc6bde6f 177 void (*rxfree) (pTHX_ REGEXP * const rx);
2fdbfb4d 178 void (*numbered_buff_FETCH) (pTHX_ REGEXP * const rx, const I32 paren,
192b9cd1 179 SV * const sv);
2fdbfb4d
AB
180 void (*numbered_buff_STORE) (pTHX_ REGEXP * const rx, const I32 paren,
181 SV const * const value);
182 I32 (*numbered_buff_LENGTH) (pTHX_ REGEXP * const rx, const SV * const sv,
183 const I32 paren);
192b9cd1
AB
184 SV* (*named_buff) (pTHX_ REGEXP * const rx, SV * const key,
185 SV * const value, const U32 flags);
186 SV* (*named_buff_iter) (pTHX_ REGEXP * const rx, const SV * const lastkey,
187 const U32 flags);
49d7dfbc 188 SV* (*qr_package)(pTHX_ REGEXP * const rx);
f2f78491 189#ifdef USE_ITHREADS
49d7dfbc 190 void* (*dupe) (pTHX_ REGEXP * const rx, CLONE_PARAMS *param);
2fdbfb4d 191#endif
3c13cae6
DM
192 REGEXP* (*op_comp) (pTHX_ SV ** const patternp, int pat_count,
193 OP *expr, const struct regexp_engine* eng,
7984ae89 194 REGEXP *old_re,
76ac488f 195 bool *is_bare_re, U32 orig_rx_flags, U32 pm_flags);
f9f4320a
YO
196} regexp_engine;
197
192b9cd1
AB
198/*
199 These are passed to the numbered capture variable callbacks as the
200 paren name. >= 1 is reserved for actual numbered captures, i.e. $1,
201 $2 etc.
202*/
2c7b5d76
DM
203#define RX_BUFF_IDX_CARET_PREMATCH -5 /* ${^PREMATCH} */
204#define RX_BUFF_IDX_CARET_POSTMATCH -4 /* ${^POSTMATCH} */
205#define RX_BUFF_IDX_CARET_FULLMATCH -3 /* ${^MATCH} */
206#define RX_BUFF_IDX_PREMATCH -2 /* $` */
207#define RX_BUFF_IDX_POSTMATCH -1 /* $' */
208#define RX_BUFF_IDX_FULLMATCH 0 /* $& */
192b9cd1
AB
209
210/*
211 Flags that are passed to the named_buff and named_buff_iter
212 callbacks above. Those routines are called from universal.c via the
213 Tie::Hash::NamedCapture interface for %+ and %- and the re::
214 functions in the same file.
215*/
216
217/* The Tie::Hash::NamedCapture operation this is part of, if any */
f1b875a0
YO
218#define RXapif_FETCH 0x0001
219#define RXapif_STORE 0x0002
220#define RXapif_DELETE 0x0004
221#define RXapif_CLEAR 0x0008
222#define RXapif_EXISTS 0x0010
223#define RXapif_SCALAR 0x0020
224#define RXapif_FIRSTKEY 0x0040
225#define RXapif_NEXTKEY 0x0080
192b9cd1
AB
226
227/* Whether %+ or %- is being operated on */
f1b875a0
YO
228#define RXapif_ONE 0x0100 /* %+ */
229#define RXapif_ALL 0x0200 /* %- */
192b9cd1
AB
230
231/* Whether this is being called from a re:: function */
f1b875a0
YO
232#define RXapif_REGNAME 0x0400
233#define RXapif_REGNAMES 0x0800
7948fc08 234#define RXapif_REGNAMES_COUNT 0x1000
192b9cd1 235
f7e71195
AB
236/*
237=head1 REGEXP Functions
238
239=for apidoc Am|REGEXP *|SvRX|SV *sv
240
72d33970 241Convenience macro to get the REGEXP from a SV. This is approximately
f7e71195
AB
242equivalent to the following snippet:
243
244 if (SvMAGICAL(sv))
245 mg_get(sv);
1af910ea
DL
246 if (SvROK(sv))
247 sv = MUTABLE_SV(SvRV(sv));
248 if (SvTYPE(sv) == SVt_REGEXP)
249 return (REGEXP*) sv;
f7e71195 250
796b6530 251C<NULL> will be returned if a REGEXP* is not found.
f7e71195
AB
252
253=for apidoc Am|bool|SvRXOK|SV* sv
254
1af910ea
DL
255Returns a boolean indicating whether the SV (or the one it references)
256is a REGEXP.
f7e71195
AB
257
258If you want to do something with the REGEXP* later use SvRX instead
259and check for NULL.
260
261=cut
262*/
263
264#define SvRX(sv) (Perl_get_re_arg(aTHX_ sv))
265#define SvRXOK(sv) (Perl_get_re_arg(aTHX_ sv) ? TRUE : FALSE)
266
267
7948fc08 268/* Flags stored in regexp->extflags
bbe252da 269 * These are used by code external to the regexp engine
e357fc67 270 *
5b126c84
KW
271 * Note that the flags whose names start with RXf_PMf_ are defined in
272 * op_reg_common.h, being copied from the parallel flags of op_pmflags
cb5027f2 273 *
eb2624c9
FC
274 * NOTE: if you modify any RXf flags you should run regen.pl or
275 * regen/regcomp.pl so that regnodes.h is updated with the changes.
cb5027f2 276 *
bbe252da
YO
277 */
278
5b126c84
KW
279#include "op_reg_common.h"
280
77c8f263 281#define RXf_PMf_STD_PMMOD (RXf_PMf_MULTILINE|RXf_PMf_SINGLELINE|RXf_PMf_FOLD|RXf_PMf_EXTENDED|RXf_PMf_EXTENDED_MORE|RXf_PMf_NOCAPTURE)
cde0cee5 282
cc4d09e1
KW
283#define CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl, x_count) \
284 case IGNORE_PAT_MOD: *(pmfl) |= RXf_PMf_FOLD; break; \
285 case MULTILINE_PAT_MOD: *(pmfl) |= RXf_PMf_MULTILINE; break; \
286 case SINGLE_PAT_MOD: *(pmfl) |= RXf_PMf_SINGLELINE; break; \
77c8f263
KW
287 case XTENDED_PAT_MOD: if (x_count == 0) { \
288 *(pmfl) |= RXf_PMf_EXTENDED; \
289 *(pmfl) &= ~RXf_PMf_EXTENDED_MORE; \
290 } \
291 else { \
292 *(pmfl) |= RXf_PMf_EXTENDED \
293 |RXf_PMf_EXTENDED_MORE; \
294 } \
295 (x_count)++; break; \
41d7c59e 296 case NOCAPTURE_PAT_MOD: *(pmfl) |= RXf_PMf_NOCAPTURE; break;
cc4d09e1 297
94b03d7d 298/* Note, includes charset ones, assumes 0 is the default for them */
fb85c044 299#define STD_PMMOD_FLAGS_CLEAR(pmfl) \
77c8f263 300 *(pmfl) &= ~(RXf_PMf_FOLD|RXf_PMf_MULTILINE|RXf_PMf_SINGLELINE|RXf_PMf_EXTENDED|RXf_PMf_EXTENDED_MORE|RXf_PMf_CHARSET|RXf_PMf_NOCAPTURE)
fb85c044 301
bcdf7404 302/* chars and strings used as regex pattern modifiers
486ec47a 303 * Singular is a 'c'har, plural is a "string"
87e95b7f
YO
304 *
305 * NOTE, KEEPCOPY was originally 'k', but was changed to 'p' for preserve
306 * for compatibility reasons with Regexp::Common which highjacked (?k:...)
307 * for its own uses. So 'k' is out as well.
bcdf7404 308 */
85508812 309#define DEFAULT_PAT_MOD '^' /* Short for all the default modifiers */
bcdf7404 310#define EXEC_PAT_MOD 'e'
87e95b7f 311#define KEEPCOPY_PAT_MOD 'p'
41d7c59e 312#define NOCAPTURE_PAT_MOD 'n'
bcdf7404
YO
313#define ONCE_PAT_MOD 'o'
314#define GLOBAL_PAT_MOD 'g'
315#define CONTINUE_PAT_MOD 'c'
316#define MULTILINE_PAT_MOD 'm'
317#define SINGLE_PAT_MOD 's'
318#define IGNORE_PAT_MOD 'i'
319#define XTENDED_PAT_MOD 'x'
4f4d7508 320#define NONDESTRUCT_PAT_MOD 'r'
9de15fec
KW
321#define LOCALE_PAT_MOD 'l'
322#define UNICODE_PAT_MOD 'u'
50e91148 323#define DEPENDS_PAT_MOD 'd'
cfaf538b 324#define ASCII_RESTRICT_PAT_MOD 'a'
bcdf7404
YO
325
326#define ONCE_PAT_MODS "o"
87e95b7f 327#define KEEPCOPY_PAT_MODS "p"
41d7c59e 328#define NOCAPTURE_PAT_MODS "n"
bcdf7404
YO
329#define EXEC_PAT_MODS "e"
330#define LOOP_PAT_MODS "gc"
4f4d7508 331#define NONDESTRUCT_PAT_MODS "r"
9de15fec
KW
332#define LOCALE_PAT_MODS "l"
333#define UNICODE_PAT_MODS "u"
50e91148 334#define DEPENDS_PAT_MODS "d"
cfaf538b 335#define ASCII_RESTRICT_PAT_MODS "a"
df7a8460 336#define ASCII_MORE_RESTRICT_PAT_MODS "aa"
bcdf7404 337
78fecfc8 338/* This string is expected by regcomp.c to be ordered so that the first
c18d5d15
KW
339 * character is the flag in bit RXf_PMf_STD_PMMOD_SHIFT of extflags; the next
340 * character is bit +1, etc. */
41d7c59e 341#define STD_PAT_MODS "msixxn"
bcdf7404 342
94b03d7d
KW
343#define CHARSET_PAT_MODS ASCII_RESTRICT_PAT_MODS DEPENDS_PAT_MODS LOCALE_PAT_MODS UNICODE_PAT_MODS
344
78fecfc8 345/* This string is expected by XS_re_regexp_pattern() in universal.c to be ordered
c18d5d15
KW
346 * so that the first character is the flag in bit RXf_PMf_STD_PMMOD_SHIFT of
347 * extflags; the next character is in bit +1, etc. */
bcdf7404
YO
348#define INT_PAT_MODS STD_PAT_MODS KEEPCOPY_PAT_MODS
349
41d7c59e 350#define EXT_PAT_MODS ONCE_PAT_MODS KEEPCOPY_PAT_MODS NOCAPTURE_PAT_MODS
94b03d7d 351#define QR_PAT_MODS STD_PAT_MODS EXT_PAT_MODS CHARSET_PAT_MODS
bcdf7404 352#define M_PAT_MODS QR_PAT_MODS LOOP_PAT_MODS
4f4d7508 353#define S_PAT_MODS M_PAT_MODS EXEC_PAT_MODS NONDESTRUCT_PAT_MODS
bcdf7404 354
cb5027f2 355/*
eb2624c9
FC
356 * NOTE: if you modify any RXf flags you should run regen.pl or
357 * regen/regcomp.pl so that regnodes.h is updated with the changes.
cb5027f2
YO
358 *
359 */
bcdf7404 360
dbc200c5 361/*
5012eebe
DM
362 Set in Perl_pmruntime for a split. Will be used by regex engines to
363 check whether they should set RXf_SKIPWHITE
dbc200c5 364*/
32f62247
KW
365#define RXf_SPLIT RXf_PMf_SPLIT
366
0bd51918
KW
367/* Currently the regex flags occupy a single 32-bit word. Not all bits are
368 * currently used. The lower bits are shared with their corresponding PMf flag
369 * bits, up to but not including _RXf_PMf_SHIFT_NEXT. The unused bits
370 * immediately follow; finally the used RXf-only (unshared) bits, so that the
371 * highest bit in the word is used. This gathers all the unused bits as a pool
372 * in the middle, like so: 11111111111111110000001111111111
373 * where the '1's represent used bits, and the '0's unused. This design allows
374 * us to allocate off one end of the pool if we need to add a shared bit, and
375 * off the other end if we need a non-shared bit, without disturbing the other
376 * bits. This maximizes the likelihood of being able to change things without
377 * breaking binary compatibility.
378 *
379 * To add shared bits, do so in op_reg_common.h. This should change
380 * _RXf_PMf_SHIFT_NEXT so that things won't compile. Then come to regexp.h and
381 * op.h and adjust the constant adders in the definitions of RXf_BASE_SHIFT and
382 * Pmf_BASE_SHIFT down by the number of shared bits you added. That's it.
383 * Things should be binary compatible. But if either of these gets to having
384 * to subtract rather than add, leave at 0 and instead adjust all the entries
385 * that are in terms of it. But if the first one of those is already
386 * RXf_BASE_SHIFT+0, there are no bits left, and a redesign is in order.
387 *
388 * To remove unshared bits, just delete its entry. If you're where breaking
389 * binary compatibility is ok to do, you might want to adjust things to move
390 * the newly opened space so that it gets absorbed into the common pool.
391 *
392 * To add unshared bits, first use up any gaps in the middle. Otherwise,
393 * allocate off the low end until you get to RXf_BASE_SHIFT+0. If that isn't
394 * enough, move RXf_BASE_SHIFT down (if possible) and add the new bit at the
395 * other end instead; this preserves binary compatibility.
396 *
397 * For the regexp bits, PL_reg_extflags_name[] in regnodes.h has a comment
398 * giving which bits are used/unused */
399
d262c0c7 400#define RXf_BASE_SHIFT (_RXf_PMf_SHIFT_NEXT + 2)
52d81aa8 401
bbe252da 402/* What we have seen */
9cba692b
KW
403#define RXf_NO_INPLACE_SUBST (1U<<(RXf_BASE_SHIFT+2))
404#define RXf_EVAL_SEEN (1U<<(RXf_BASE_SHIFT+3))
bbe252da
YO
405
406/* Special */
9cba692b
KW
407#define RXf_UNBOUNDED_QUANTIFIER_SEEN (1U<<(RXf_BASE_SHIFT+4))
408#define RXf_CHECK_ALL (1U<<(RXf_BASE_SHIFT+5))
bbe252da
YO
409
410/* UTF8 related */
9cba692b 411#define RXf_MATCH_UTF8 (1U<<(RXf_BASE_SHIFT+6)) /* $1 etc are utf8 */
bbe252da
YO
412
413/* Intuit related */
9cba692b
KW
414#define RXf_USE_INTUIT_NOML (1U<<(RXf_BASE_SHIFT+7))
415#define RXf_USE_INTUIT_ML (1U<<(RXf_BASE_SHIFT+8))
416#define RXf_INTUIT_TAIL (1U<<(RXf_BASE_SHIFT+9))
bbe252da
YO
417#define RXf_USE_INTUIT (RXf_USE_INTUIT_NOML|RXf_USE_INTUIT_ML)
418
a3b51d37 419/* Do we have some sort of anchor? */
9cba692b 420#define RXf_IS_ANCHORED (1U<<(RXf_BASE_SHIFT+10))
a3b51d37 421
bbe252da 422/* Copy and tainted info */
9cba692b 423#define RXf_COPY_DONE (1U<<(RXf_BASE_SHIFT+11))
ef07e810 424
1738e041 425/* post-execution: $1 et al are tainted */
9cba692b 426#define RXf_TAINTED_SEEN (1U<<(RXf_BASE_SHIFT+12))
ef07e810 427/* this pattern was tainted during compilation */
9cba692b 428#define RXf_TAINTED (1U<<(RXf_BASE_SHIFT+13))
52d81aa8
NC
429
430/* Flags indicating special patterns */
9cba692b
KW
431#define RXf_START_ONLY (1U<<(RXf_BASE_SHIFT+14)) /* Pattern is /^/ */
432#define RXf_SKIPWHITE (1U<<(RXf_BASE_SHIFT+15)) /* Pattern is for a */
433 /* split " " */
434#define RXf_WHITE (1U<<(RXf_BASE_SHIFT+16)) /* Pattern is /\s+/ */
3f6e6b29 435#define RXf_NULL (1U<<(RXf_BASE_SHIFT+17)) /* Pattern is // */
0bd51918
KW
436
437/* See comments at the beginning of these defines about adding bits. The
438 * highest bit position should be used, so that if RXf_BASE_SHIFT gets
439 * increased, the #error below will be triggered so that you will be reminded
440 * to adjust things at the other end to keep the bit positions unchanged */
3f6e6b29 441#if RXf_BASE_SHIFT+17 > 31
0bd51918 442# error Too many RXf_PMf bits used. See comments at beginning of these for what to do
1850c8f9 443#endif
c737faaf 444
cb5027f2 445/*
eb2624c9
FC
446 * NOTE: if you modify any RXf flags you should run regen.pl or
447 * regen/regcomp.pl so that regnodes.h is updated with the changes.
cb5027f2
YO
448 *
449 */
bbe252da 450
dc6d7f5c 451#ifdef NO_TAINT_SUPPORT
284167a5
SM
452# define RX_ISTAINTED(prog) 0
453# define RX_TAINT_on(prog) NOOP
272d35c9
DM
454# define RXp_MATCH_TAINTED(prog) 0
455# define RX_MATCH_TAINTED(prog) 0
456# define RXp_MATCH_TAINTED_on(prog) NOOP
457# define RX_MATCH_TAINTED_on(prog) NOOP
458# define RX_MATCH_TAINTED_off(prog) NOOP
284167a5
SM
459#else
460# define RX_ISTAINTED(prog) (RX_EXTFLAGS(prog) & RXf_TAINTED)
461# define RX_TAINT_on(prog) (RX_EXTFLAGS(prog) |= RXf_TAINTED)
272d35c9
DM
462# define RXp_MATCH_TAINTED(prog) (RXp_EXTFLAGS(prog) & RXf_TAINTED_SEEN)
463# define RX_MATCH_TAINTED(prog) (RX_EXTFLAGS(prog) & RXf_TAINTED_SEEN)
464# define RXp_MATCH_TAINTED_on(prog) (RXp_EXTFLAGS(prog) |= RXf_TAINTED_SEEN)
465# define RX_MATCH_TAINTED_on(prog) (RX_EXTFLAGS(prog) |= RXf_TAINTED_SEEN)
466# define RX_MATCH_TAINTED_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_TAINTED_SEEN)
284167a5
SM
467#endif
468
bbe252da 469#define RX_HAS_CUTGROUP(prog) ((prog)->intflags & PREGf_CUTGROUP_SEEN)
72311751
GS
470#define RX_MATCH_TAINTED_set(prog, t) ((t) \
471 ? RX_MATCH_TAINTED_on(prog) \
472 : RX_MATCH_TAINTED_off(prog))
c277df42 473
07bc277f
NC
474#define RXp_MATCH_COPIED(prog) (RXp_EXTFLAGS(prog) & RXf_COPY_DONE)
475#define RX_MATCH_COPIED(prog) (RX_EXTFLAGS(prog) & RXf_COPY_DONE)
476#define RXp_MATCH_COPIED_on(prog) (RXp_EXTFLAGS(prog) |= RXf_COPY_DONE)
477#define RX_MATCH_COPIED_on(prog) (RX_EXTFLAGS(prog) |= RXf_COPY_DONE)
478#define RXp_MATCH_COPIED_off(prog) (RXp_EXTFLAGS(prog) &= ~RXf_COPY_DONE)
479#define RX_MATCH_COPIED_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_COPY_DONE)
cf93c79d
IZ
480#define RX_MATCH_COPIED_set(prog,t) ((t) \
481 ? RX_MATCH_COPIED_on(prog) \
482 : RX_MATCH_COPIED_off(prog))
f9f4320a 483
07bc277f 484#define RXp_EXTFLAGS(rx) ((rx)->extflags)
dbc200c5 485#define RXp_COMPFLAGS(rx) ((rx)->compflags)
866c78d1 486
07bc277f 487/* For source compatibility. We used to store these explicitly. */
8d919b0a
FC
488#define RX_PRECOMP(prog) (RX_WRAPPED(prog) + ReANY(prog)->pre_prefix)
489#define RX_PRECOMP_const(prog) (RX_WRAPPED_const(prog) + ReANY(prog)->pre_prefix)
5509d87a
NC
490/* FIXME? Are we hardcoding too much here and constraining plugin extension
491 writers? Specifically, the value 1 assumes that the wrapped version always
492 has exactly one character at the end, a ')'. Will that always be true? */
8d919b0a
FC
493#define RX_PRELEN(prog) (RX_WRAPLEN(prog) - ReANY(prog)->pre_prefix - 1)
494#define RX_WRAPPED(prog) ReANY(prog)->xpv_len_u.xpvlenu_pv
495#define RX_WRAPPED_const(prog) ((const char *)RX_WRAPPED(prog))
9d17798d 496#define RX_WRAPLEN(prog) SvCUR(prog)
8d919b0a 497#define RX_CHECK_SUBSTR(prog) (ReANY(prog)->check_substr)
288b8c02 498#define RX_REFCNT(prog) SvREFCNT(prog)
8d919b0a 499#define RX_EXTFLAGS(prog) RXp_EXTFLAGS(ReANY(prog))
dbc200c5 500#define RX_COMPFLAGS(prog) RXp_COMPFLAGS(ReANY(prog))
8d919b0a
FC
501#define RX_ENGINE(prog) (ReANY(prog)->engine)
502#define RX_SUBBEG(prog) (ReANY(prog)->subbeg)
503#define RX_SUBOFFSET(prog) (ReANY(prog)->suboffset)
504#define RX_SUBCOFFSET(prog) (ReANY(prog)->subcoffset)
505#define RX_OFFS(prog) (ReANY(prog)->offs)
506#define RX_NPARENS(prog) (ReANY(prog)->nparens)
507#define RX_SUBLEN(prog) (ReANY(prog)->sublen)
508#define RX_MINLEN(prog) (ReANY(prog)->minlen)
509#define RX_MINLENRET(prog) (ReANY(prog)->minlenret)
510#define RX_GOFS(prog) (ReANY(prog)->gofs)
511#define RX_LASTPAREN(prog) (ReANY(prog)->lastparen)
512#define RX_LASTCLOSEPAREN(prog) (ReANY(prog)->lastcloseparen)
513#define RX_SAVED_COPY(prog) (ReANY(prog)->saved_copy)
03c83e26
DM
514/* last match was zero-length */
515#define RX_ZERO_LEN(prog) \
99a90e59
FC
516 (RX_OFFS(prog)[0].start + (SSize_t)RX_GOFS(prog) \
517 == RX_OFFS(prog)[0].end)
220fc49f 518
be8e71aa
YO
519#endif /* PLUGGABLE_RE_EXTENSION */
520
486ec47a 521/* Stuff that needs to be included in the pluggable extension goes below here */
be8e71aa 522
db2c6cb3 523#ifdef PERL_ANY_COW
ed252734 524#define RX_MATCH_COPY_FREE(rx) \
bdd9a1b1
NC
525 STMT_START {if (RX_SAVED_COPY(rx)) { \
526 SV_CHECK_THINKFIRST_COW_DROP(RX_SAVED_COPY(rx)); \
ed252734
NC
527 } \
528 if (RX_MATCH_COPIED(rx)) { \
07bc277f 529 Safefree(RX_SUBBEG(rx)); \
ed252734
NC
530 RX_MATCH_COPIED_off(rx); \
531 }} STMT_END
532#else
533#define RX_MATCH_COPY_FREE(rx) \
534 STMT_START {if (RX_MATCH_COPIED(rx)) { \
07bc277f 535 Safefree(RX_SUBBEG(rx)); \
ed252734
NC
536 RX_MATCH_COPIED_off(rx); \
537 }} STMT_END
538#endif
539
07bc277f
NC
540#define RXp_MATCH_UTF8(prog) (RXp_EXTFLAGS(prog) & RXf_MATCH_UTF8)
541#define RX_MATCH_UTF8(prog) (RX_EXTFLAGS(prog) & RXf_MATCH_UTF8)
542#define RX_MATCH_UTF8_on(prog) (RX_EXTFLAGS(prog) |= RXf_MATCH_UTF8)
543#define RX_MATCH_UTF8_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_MATCH_UTF8)
a30b2f1f 544#define RX_MATCH_UTF8_set(prog, t) ((t) \
0603fe5c
DM
545 ? RX_MATCH_UTF8_on(prog) \
546 : RX_MATCH_UTF8_off(prog))
efd26800
NC
547
548/* Whether the pattern stored at RX_WRAPPED is in UTF-8 */
8f6ae13c 549#define RX_UTF8(prog) SvUTF8(prog)
7948fc08 550
a340edde
DM
551
552/* bits in flags arg of Perl_regexec_flags() */
553
554#define REXEC_COPY_STR 0x01 /* Need to copy the string for captures. */
555#define REXEC_CHECKED 0x02 /* re_intuit_start() already called. */
556#define REXEC_SCREAM 0x04 /* currently unused. */
557#define REXEC_IGNOREPOS 0x08 /* use stringarg, not pos(), for \G match */
558#define REXEC_NOT_FIRST 0x10 /* This is another iteration of //g:
559 no need to copy string again */
560
561 /* under REXEC_COPY_STR, it's ok for the
562 engine (modulo PL_sawamperand etc)
563 to skip copying: ... */
564#define REXEC_COPY_SKIP_PRE 0x20 /* ...the $` part of the string, or */
565#define REXEC_COPY_SKIP_POST 0x40 /* ...the $' part of the string */
d5e7783a
DM
566#define REXEC_FAIL_ON_UNDERFLOW 0x80 /* fail the match if $& would start before
567 the start pos (so s/.\G// would fail
568 on second iteration */
c277df42 569
7fe44985 570#if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN)
288b8c02
NC
571# define ReREFCNT_inc(re) \
572 ({ \
573 /* This is here to generate a casting warning if incorrect. */ \
7948fc08
RU
574 REGEXP *const _rerefcnt_inc = (re); \
575 assert(SvTYPE(_rerefcnt_inc) == SVt_REGEXP); \
576 SvREFCNT_inc(_rerefcnt_inc); \
577 _rerefcnt_inc; \
288b8c02
NC
578 })
579# define ReREFCNT_dec(re) \
580 ({ \
581 /* This is here to generate a casting warning if incorrect. */ \
7948fc08
RU
582 REGEXP *const _rerefcnt_dec = (re); \
583 SvREFCNT_dec(_rerefcnt_dec); \
288b8c02
NC
584 })
585#else
586# define ReREFCNT_dec(re) SvREFCNT_dec(re)
b7e0bd9e 587# define ReREFCNT_inc(re) ((REGEXP *) SvREFCNT_inc(re))
288b8c02 588#endif
8d919b0a 589#define ReANY(re) S_ReANY((const REGEXP *)(re))
288b8c02
NC
590
591/* FIXME for plugins. */
cf93c79d
IZ
592
593#define FBMcf_TAIL_DOLLAR 1
cad2e5aa
JH
594#define FBMcf_TAIL_DOLLARM 2
595#define FBMcf_TAIL_Z 4
596#define FBMcf_TAIL_z 8
597#define FBMcf_TAIL (FBMcf_TAIL_DOLLAR|FBMcf_TAIL_DOLLARM|FBMcf_TAIL_Z|FBMcf_TAIL_z)
cf93c79d
IZ
598
599#define FBMrf_MULTILINE 1
cad2e5aa 600
331b2dcc
DM
601struct regmatch_state;
602struct regmatch_slab;
8adc0f72 603
bf2039a9
DM
604/* like regmatch_info_aux, but contains extra fields only needed if the
605 * pattern contains (?{}). If used, is snuck into the second slot in the
606 * regmatch_state stack at the start of execution */
8adc0f72
DM
607
608typedef struct {
609 regexp *rex;
610 PMOP *curpm; /* saved PL_curpm */
611#ifdef PERL_ANY_COW
612 SV *saved_copy; /* saved saved_copy field from rex */
613#endif
614 char *subbeg; /* saved subbeg field from rex */
615 STRLEN sublen; /* saved sublen field from rex */
616 STRLEN suboffset; /* saved suboffset field from rex */
617 STRLEN subcoffset; /* saved subcoffset field from rex */
618 MAGIC *pos_magic; /* pos() magic attached to $_ */
ea3daa5d 619 SSize_t pos; /* the original value of pos() in pos_magic */
25fdce4a 620 U8 pos_flags; /* flags to be restored; currently only MGf_BYTES*/
bf2039a9
DM
621} regmatch_info_aux_eval;
622
623
624/* fields that logically live in regmatch_info, but which need cleaning
625 * up on croak(), and so are instead are snuck into the first slot in
626 * the regmatch_state stack at the start of execution */
627
628typedef struct {
629 regmatch_info_aux_eval *info_aux_eval;
331b2dcc
DM
630 struct regmatch_state *old_regmatch_state; /* saved PL_regmatch_state */
631 struct regmatch_slab *old_regmatch_slab; /* saved PL_regmatch_slab */
2ac8ff4b 632 char *poscache; /* S-L cache of fail positions of WHILEMs */
bf2039a9
DM
633} regmatch_info_aux;
634
8adc0f72 635
3b0527fe 636/* some basic information about the current match that is created by
bf2039a9
DM
637 * Perl_regexec_flags and then passed to regtry(), regmatch() etc.
638 * It is allocated as a local var on the stack, so nothing should be
639 * stored in it that needs preserving or clearing up on croak().
640 * For that, see the aux_info and aux_info_eval members of the
641 * regmatch_state union. */
3b0527fe
DM
642
643typedef struct {
28d03b21 644 REGEXP *prog; /* the regex being executed */
9d9163fb 645 const char * strbeg; /* real start of string */
28d03b21
DM
646 char *strend; /* one byte beyond last char of match string */
647 char *till; /* matches shorter than this fail (see minlen arg) */
648 SV *sv; /* the SV string currently being matched */
649 char *ganch; /* position of \G anchor */
650 char *cutpoint; /* (*COMMIT) position (if any) */
bf2039a9
DM
651 regmatch_info_aux *info_aux; /* extra fields that need cleanup */
652 regmatch_info_aux_eval *info_aux_eval; /* extra saved state for (?{}) */
1cb48e53
DM
653 I32 poscache_maxiter; /* how many whilems todo before S-L cache kicks in */
654 I32 poscache_iter; /* current countdown from _maxiter to zero */
2ac8ff4b 655 STRLEN poscache_size; /* size of regmatch_info_aux.poscache */
02d5137b 656 bool intuit; /* re_intuit_start() is the top-level caller */
ba44c216
DM
657 bool is_utf8_pat; /* regex is utf8 */
658 bool is_utf8_target; /* string being matched is utf8 */
39819bd9 659 bool warned; /* we have issued a recursion warning; no need for more */
3b0527fe
DM
660} regmatch_info;
661
5d9a96ca
DM
662
663/* structures for holding and saving the state maintained by regmatch() */
664
d56b3014 665#ifndef MAX_RECURSE_EVAL_NOCHANGE_DEPTH
ba6840fb 666#define MAX_RECURSE_EVAL_NOCHANGE_DEPTH 10
d56b3014 667#endif
6bda09f9 668
5d9a96ca
DM
669typedef I32 CHECKPOINT;
670
a0374537 671typedef struct regmatch_state {
40a82448 672 int resume_state; /* where to jump to on return */
24d3c4a9 673 char *locinput; /* where to backtrack in string on failure */
e822a8b4
DM
674
675 union {
77cb431f 676
bf2039a9
DM
677 /* the 'info_aux' and 'info_aux_eval' union members are cuckoos in
678 * the nest. They aren't saved backtrack state; rather they
679 * represent one or two extra chunks of data that need allocating
680 * at the start of a match. These fields would logically live in
681 * the regmatch_info struct, except that is allocated on the
682 * C stack, and these fields are all things that require cleanup
683 * after a croak(), when the stack is lost.
684 * As a convenience, we just use the first 1 or 2 regmatch_state
685 * slots to store this info, as we will be allocating a slab of
686 * these anyway. Otherwise we'd have to malloc and then free them,
687 * or allocate them on the save stack (where they will get
688 * realloced if the save stack grows).
689 * info_aux contains the extra fields that are always needed;
690 * info_aux_eval contains extra fields that only needed if
691 * the pattern contains code blocks
692 * We split them into two separate structs to avoid increasing
693 * the size of the union.
694 */
695
696 regmatch_info_aux info_aux;
697
698 regmatch_info_aux_eval info_aux_eval;
699
77cb431f
DM
700 /* this is a fake union member that matches the first element
701 * of each member that needs to store positive backtrack
702 * information */
703 struct {
704 struct regmatch_state *prev_yes_state;
705 } yes;
706
fae667d5
YO
707 /* branchlike members */
708 /* this is a fake union member that matches the first elements
709 * of each member that needs to behave like a branch */
710 struct {
5d458dd8
YO
711 /* this first element must match u.yes */
712 struct regmatch_state *prev_yes_state;
fae667d5 713 U32 lastparen;
f6033a9d 714 U32 lastcloseparen;
fae667d5
YO
715 CHECKPOINT cp;
716
717 } branchlike;
718
719 struct {
720 /* the first elements must match u.branchlike */
721 struct regmatch_state *prev_yes_state;
722 U32 lastparen;
f6033a9d 723 U32 lastcloseparen;
fae667d5
YO
724 CHECKPOINT cp;
725
726 regnode *next_branch; /* next branch node */
727 } branch;
728
729 struct {
730 /* the first elements must match u.branchlike */
731 struct regmatch_state *prev_yes_state;
732 U32 lastparen;
f6033a9d 733 U32 lastcloseparen;
fae667d5
YO
734 CHECKPOINT cp;
735
2e64971a 736 U32 accepted; /* how many accepting states left */
6603fe3e 737 bool longfold;/* saw a fold with a 1->n char mapping */
7f69552c 738 U16 *jump; /* positive offsets from me */
7f69552c 739 regnode *me; /* Which node am I - needed for jump tries*/
2e64971a
DM
740 U8 *firstpos;/* pos in string of first trie match */
741 U32 firstchars;/* len in chars of firstpos from start */
742 U16 nextword;/* next word to try */
743 U16 topword; /* longest accepted word */
e822a8b4
DM
744 } trie;
745
fae667d5
YO
746 /* special types - these members are used to store state for special
747 regops like eval, if/then, lookaround and the markpoint state */
e822a8b4 748 struct {
77cb431f
DM
749 /* this first element must match u.yes */
750 struct regmatch_state *prev_yes_state;
faec1544 751 struct regmatch_state *prev_curlyx;
d1c49ad5 752 struct regmatch_state *prev_eval;
288b8c02 753 REGEXP *prev_rex;
aa283a38
DM
754 CHECKPOINT cp; /* remember current savestack indexes */
755 CHECKPOINT lastcp;
24be3102 756 U32 close_paren; /* which close bracket is our end (+1) */
1b5c067c 757 regnode *B; /* the node following us */
ba6840fb 758 char *prev_recurse_locinput;
e822a8b4
DM
759 } eval;
760
761 struct {
c476f425 762 /* this first element must match u.yes */
262b90c4 763 struct regmatch_state *prev_yes_state;
fae667d5
YO
764 I32 wanted;
765 I32 logical; /* saved copy of 'logical' var */
766 regnode *me; /* the IFMATCH/SUSPEND/UNLESSM node */
767 } ifmatch; /* and SUSPEND/UNLESSM */
768
769 struct {
770 /* this first element must match u.yes */
771 struct regmatch_state *prev_yes_state;
772 struct regmatch_state *prev_mark;
773 SV* mark_name;
774 char *mark_loc;
775 } mark;
776
777 struct {
778 int val;
779 } keeper;
780
781 /* quantifiers - these members are used for storing state for
782 for the regops used to implement quantifiers */
783 struct {
784 /* this first element must match u.yes */
785 struct regmatch_state *prev_yes_state;
c476f425 786 struct regmatch_state *prev_curlyx; /* previous cur_curlyx */
d02d6d97
DM
787 regnode *me; /* the CURLYX node */
788 regnode *B; /* the B node in /A*B/ */
c476f425
DM
789 CHECKPOINT cp; /* remember current savestack index */
790 bool minmod;
a0374537 791 int parenfloor;/* how far back to strip paren data */
c476f425
DM
792
793 /* these two are modified by WHILEM */
794 int count; /* how many instances of A we've matched */
795 char *lastloc;/* where previous A matched (0-len detect) */
e822a8b4
DM
796 } curlyx;
797
798 struct {
c476f425 799 /* this first element must match u.yes */
262b90c4 800 struct regmatch_state *prev_yes_state;
c476f425
DM
801 struct regmatch_state *save_curlyx;
802 CHECKPOINT cp; /* remember current savestack indexes */
803 CHECKPOINT lastcp;
804 char *save_lastloc; /* previous curlyx.lastloc */
805 I32 cache_offset;
806 I32 cache_mask;
e822a8b4
DM
807 } whilem;
808
809 struct {
5d458dd8
YO
810 /* this first element must match u.yes */
811 struct regmatch_state *prev_yes_state;
79a2a0e8 812 int c1, c2; /* case fold search */
40a82448 813 CHECKPOINT cp;
f6033a9d
DM
814 U32 lastparen;
815 U32 lastcloseparen;
40a82448
DM
816 I32 alen; /* length of first-matched A string */
817 I32 count;
0cadcf80 818 bool minmod;
40a82448
DM
819 regnode *A, *B; /* the nodes corresponding to /A*B/ */
820 regnode *me; /* the curlym node */
79a2a0e8
KW
821 U8 c1_utf8[UTF8_MAXBYTES+1]; /* */
822 U8 c2_utf8[UTF8_MAXBYTES+1];
e822a8b4
DM
823 } curlym;
824
825 struct {
3b6647e0 826 U32 paren;
c255a977 827 CHECKPOINT cp;
f6033a9d
DM
828 U32 lastparen;
829 U32 lastcloseparen;
79a2a0e8 830 int c1, c2; /* case fold search */
c255a977
DM
831 char *maxpos; /* highest possible point in string to match */
832 char *oldloc; /* the previous locinput */
e822a8b4 833 int count;
c255a977
DM
834 int min, max; /* {m,n} */
835 regnode *A, *B; /* the nodes corresponding to /A*B/ */
79a2a0e8
KW
836 U8 c1_utf8[UTF8_MAXBYTES+1]; /* */
837 U8 c2_utf8[UTF8_MAXBYTES+1];
c255a977 838 } curly; /* and CURLYN/PLUS/STAR */
dad79028 839
d8319b27 840 } u;
5d9a96ca
DM
841} regmatch_state;
842
ce12e254 843
d5a00e4a 844
5d9a96ca
DM
845/* how many regmatch_state structs to allocate as a single slab.
846 * We do it in 4K blocks for efficiency. The "3" is 2 for the next/prev
847 * pointers, plus 1 for any mythical malloc overhead. */
848
849#define PERL_REGMATCH_SLAB_SLOTS \
850 ((4096 - 3 * sizeof (void*)) / sizeof(regmatch_state))
851
852typedef struct regmatch_slab {
853 regmatch_state states[PERL_REGMATCH_SLAB_SLOTS];
854 struct regmatch_slab *prev, *next;
855} regmatch_slab;
1ade1aa1 856
46ab3289 857
cde0cee5 858
1ade1aa1 859/*
14d04a33 860 * ex: set ts=8 sts=4 sw=4 et:
1ade1aa1 861 */