This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
op_reg_common.h: Update comment
[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 */ \
08e44740
NC
105 /* Information about the match that the perl core uses to */ \
106 /* manage things */ \
107 U32 extflags; /* Flags used both externally and internally */ \
ea3daa5d
FC
108 SSize_t minlen; /* mininum possible number of chars in string to match */\
109 SSize_t minlenret; /* mininum possible number of chars in $& */ \
110 STRLEN gofs; /* chars left of pos that we search from */ \
08e44740
NC
111 /* substring data about strings that must appear in the */ \
112 /* final match, used for optimisations */ \
113 struct reg_substr_data *substrs; \
114 U32 nparens; /* number of capture buffers */ \
115 /* private engine specific data */ \
116 U32 intflags; /* Engine Specific Internal flags */ \
117 void *pprivate; /* Data private to the regex engine which */ \
118 /* created this object. */ \
119 /* Data about the last/current match. These are modified */ \
120 /* during matching */ \
121 U32 lastparen; /* last open paren matched */ \
122 U32 lastcloseparen; /* last close paren matched */ \
08e44740
NC
123 /* Array of offsets for (@-) and (@+) */ \
124 regexp_paren_pair *offs; \
125 /* saved or original string so \digit works forever. */ \
126 char *subbeg; \
127 SV_SAVED_COPY /* If non-NULL, SV which is COW from original */\
19394178 128 SSize_t sublen; /* Length of string pointed by subbeg */ \
ea3daa5d
FC
129 SSize_t suboffset; /* byte offset of subbeg from logical start of str */ \
130 SSize_t subcoffset; /* suboffset equiv, but in chars (for @-/@+) */ \
08e44740 131 /* Information about the match that isn't often used */ \
ee273784 132 SSize_t maxlen; /* mininum possible number of chars in string to match */\
08e44740 133 /* offset from wrapped to the start of precomp */ \
654eccd5 134 PERL_BITFIELD32 pre_prefix:4; \
dbc200c5
YO
135 /* original flags used to compile the pattern, may differ */ \
136 /* from extflags in various ways */ \
137 PERL_BITFIELD32 compflags:9; \
b82ad30c 138 CV *qr_anoncv /* the anon sub wrapped round qr/(?{..})/ */
08e44740 139
bbe252da 140typedef struct regexp {
288b8c02 141 _XPV_HEAD;
08e44740 142 _REGEXP_COMMON;
f8fc2ecf
YO
143} regexp;
144
89f6f287 145#define RXp_PAREN_NAMES(rx) ((rx)->paren_names)
5daac39c 146
785a26d5 147/* used for high speed searches */
f9f4320a
YO
148typedef struct re_scream_pos_data_s
149{
150 char **scream_olds; /* match pos */
ea3daa5d 151 SSize_t *scream_pos; /* Internal iterator of scream. */
f9f4320a
YO
152} re_scream_pos_data;
153
785a26d5
YO
154/* regexp_engine structure. This is the dispatch table for regexes.
155 * Any regex engine implementation must be able to build one of these.
156 */
f9f4320a 157typedef struct regexp_engine {
1593ad57 158 REGEXP* (*comp) (pTHX_ SV * const pattern, U32 flags);
49d7dfbc 159 I32 (*exec) (pTHX_ REGEXP * const rx, char* stringarg, char* strend,
ea3daa5d 160 char* strbeg, SSize_t minend, SV* sv,
49d7dfbc 161 void* data, U32 flags);
52a21eb3
DM
162 char* (*intuit) (pTHX_
163 REGEXP * const rx,
164 SV *sv,
165 const char * const strbeg,
166 char *strpos,
167 char *strend,
168 const U32 flags,
9f61653a 169 re_scream_pos_data *data);
49d7dfbc
AB
170 SV* (*checkstr) (pTHX_ REGEXP * const rx);
171 void (*free) (pTHX_ REGEXP * const rx);
2fdbfb4d 172 void (*numbered_buff_FETCH) (pTHX_ REGEXP * const rx, const I32 paren,
192b9cd1 173 SV * const sv);
2fdbfb4d
AB
174 void (*numbered_buff_STORE) (pTHX_ REGEXP * const rx, const I32 paren,
175 SV const * const value);
176 I32 (*numbered_buff_LENGTH) (pTHX_ REGEXP * const rx, const SV * const sv,
177 const I32 paren);
192b9cd1
AB
178 SV* (*named_buff) (pTHX_ REGEXP * const rx, SV * const key,
179 SV * const value, const U32 flags);
180 SV* (*named_buff_iter) (pTHX_ REGEXP * const rx, const SV * const lastkey,
181 const U32 flags);
49d7dfbc 182 SV* (*qr_package)(pTHX_ REGEXP * const rx);
f2f78491 183#ifdef USE_ITHREADS
49d7dfbc 184 void* (*dupe) (pTHX_ REGEXP * const rx, CLONE_PARAMS *param);
2fdbfb4d 185#endif
3c13cae6
DM
186 REGEXP* (*op_comp) (pTHX_ SV ** const patternp, int pat_count,
187 OP *expr, const struct regexp_engine* eng,
188 REGEXP *VOL old_re,
76ac488f 189 bool *is_bare_re, U32 orig_rx_flags, U32 pm_flags);
f9f4320a
YO
190} regexp_engine;
191
192b9cd1
AB
192/*
193 These are passed to the numbered capture variable callbacks as the
194 paren name. >= 1 is reserved for actual numbered captures, i.e. $1,
195 $2 etc.
196*/
2c7b5d76
DM
197#define RX_BUFF_IDX_CARET_PREMATCH -5 /* ${^PREMATCH} */
198#define RX_BUFF_IDX_CARET_POSTMATCH -4 /* ${^POSTMATCH} */
199#define RX_BUFF_IDX_CARET_FULLMATCH -3 /* ${^MATCH} */
200#define RX_BUFF_IDX_PREMATCH -2 /* $` */
201#define RX_BUFF_IDX_POSTMATCH -1 /* $' */
202#define RX_BUFF_IDX_FULLMATCH 0 /* $& */
192b9cd1
AB
203
204/*
205 Flags that are passed to the named_buff and named_buff_iter
206 callbacks above. Those routines are called from universal.c via the
207 Tie::Hash::NamedCapture interface for %+ and %- and the re::
208 functions in the same file.
209*/
210
211/* The Tie::Hash::NamedCapture operation this is part of, if any */
f1b875a0
YO
212#define RXapif_FETCH 0x0001
213#define RXapif_STORE 0x0002
214#define RXapif_DELETE 0x0004
215#define RXapif_CLEAR 0x0008
216#define RXapif_EXISTS 0x0010
217#define RXapif_SCALAR 0x0020
218#define RXapif_FIRSTKEY 0x0040
219#define RXapif_NEXTKEY 0x0080
192b9cd1
AB
220
221/* Whether %+ or %- is being operated on */
f1b875a0
YO
222#define RXapif_ONE 0x0100 /* %+ */
223#define RXapif_ALL 0x0200 /* %- */
192b9cd1
AB
224
225/* Whether this is being called from a re:: function */
f1b875a0
YO
226#define RXapif_REGNAME 0x0400
227#define RXapif_REGNAMES 0x0800
7948fc08 228#define RXapif_REGNAMES_COUNT 0x1000
192b9cd1 229
f7e71195
AB
230/*
231=head1 REGEXP Functions
232
233=for apidoc Am|REGEXP *|SvRX|SV *sv
234
72d33970 235Convenience macro to get the REGEXP from a SV. This is approximately
f7e71195
AB
236equivalent to the following snippet:
237
238 if (SvMAGICAL(sv))
239 mg_get(sv);
1af910ea
DL
240 if (SvROK(sv))
241 sv = MUTABLE_SV(SvRV(sv));
242 if (SvTYPE(sv) == SVt_REGEXP)
243 return (REGEXP*) sv;
f7e71195
AB
244
245NULL will be returned if a REGEXP* is not found.
246
247=for apidoc Am|bool|SvRXOK|SV* sv
248
1af910ea
DL
249Returns a boolean indicating whether the SV (or the one it references)
250is a REGEXP.
f7e71195
AB
251
252If you want to do something with the REGEXP* later use SvRX instead
253and check for NULL.
254
255=cut
256*/
257
258#define SvRX(sv) (Perl_get_re_arg(aTHX_ sv))
259#define SvRXOK(sv) (Perl_get_re_arg(aTHX_ sv) ? TRUE : FALSE)
260
261
7948fc08 262/* Flags stored in regexp->extflags
bbe252da 263 * These are used by code external to the regexp engine
e357fc67 264 *
5b126c84
KW
265 * Note that the flags whose names start with RXf_PMf_ are defined in
266 * op_reg_common.h, being copied from the parallel flags of op_pmflags
cb5027f2 267 *
eb2624c9
FC
268 * NOTE: if you modify any RXf flags you should run regen.pl or
269 * regen/regcomp.pl so that regnodes.h is updated with the changes.
cb5027f2 270 *
bbe252da
YO
271 */
272
5b126c84
KW
273#include "op_reg_common.h"
274
cde0cee5 275#define RXf_PMf_STD_PMMOD (RXf_PMf_MULTILINE|RXf_PMf_SINGLELINE|RXf_PMf_FOLD|RXf_PMf_EXTENDED)
cde0cee5 276
cc4d09e1
KW
277#define CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl, x_count) \
278 case IGNORE_PAT_MOD: *(pmfl) |= RXf_PMf_FOLD; break; \
279 case MULTILINE_PAT_MOD: *(pmfl) |= RXf_PMf_MULTILINE; break; \
280 case SINGLE_PAT_MOD: *(pmfl) |= RXf_PMf_SINGLELINE; break; \
281 case XTENDED_PAT_MOD: *(pmfl) |= RXf_PMf_EXTENDED; (x_count)++; break;
282
283#define STD_PMMOD_FLAGS_PARSE_X_WARN(x_count) \
284 if (UNLIKELY((x_count) > 1)) { \
285 Perl_ck_warner_d(aTHX_ packWARN2(WARN_DEPRECATED, WARN_REGEXP), \
286 "Having more than one /%c regexp modifier is deprecated", \
287 XTENDED_PAT_MOD); \
288 }
bbe252da 289
94b03d7d 290/* Note, includes charset ones, assumes 0 is the default for them */
fb85c044 291#define STD_PMMOD_FLAGS_CLEAR(pmfl) \
a62b1201 292 *(pmfl) &= ~(RXf_PMf_FOLD|RXf_PMf_MULTILINE|RXf_PMf_SINGLELINE|RXf_PMf_EXTENDED|RXf_PMf_CHARSET)
fb85c044 293
bcdf7404 294/* chars and strings used as regex pattern modifiers
486ec47a 295 * Singular is a 'c'har, plural is a "string"
87e95b7f
YO
296 *
297 * NOTE, KEEPCOPY was originally 'k', but was changed to 'p' for preserve
298 * for compatibility reasons with Regexp::Common which highjacked (?k:...)
299 * for its own uses. So 'k' is out as well.
bcdf7404 300 */
85508812 301#define DEFAULT_PAT_MOD '^' /* Short for all the default modifiers */
bcdf7404 302#define EXEC_PAT_MOD 'e'
87e95b7f 303#define KEEPCOPY_PAT_MOD 'p'
bcdf7404
YO
304#define ONCE_PAT_MOD 'o'
305#define GLOBAL_PAT_MOD 'g'
306#define CONTINUE_PAT_MOD 'c'
307#define MULTILINE_PAT_MOD 'm'
308#define SINGLE_PAT_MOD 's'
309#define IGNORE_PAT_MOD 'i'
310#define XTENDED_PAT_MOD 'x'
4f4d7508 311#define NONDESTRUCT_PAT_MOD 'r'
9de15fec
KW
312#define LOCALE_PAT_MOD 'l'
313#define UNICODE_PAT_MOD 'u'
50e91148 314#define DEPENDS_PAT_MOD 'd'
cfaf538b 315#define ASCII_RESTRICT_PAT_MOD 'a'
bcdf7404
YO
316
317#define ONCE_PAT_MODS "o"
87e95b7f 318#define KEEPCOPY_PAT_MODS "p"
bcdf7404
YO
319#define EXEC_PAT_MODS "e"
320#define LOOP_PAT_MODS "gc"
4f4d7508 321#define NONDESTRUCT_PAT_MODS "r"
9de15fec
KW
322#define LOCALE_PAT_MODS "l"
323#define UNICODE_PAT_MODS "u"
50e91148 324#define DEPENDS_PAT_MODS "d"
cfaf538b 325#define ASCII_RESTRICT_PAT_MODS "a"
df7a8460 326#define ASCII_MORE_RESTRICT_PAT_MODS "aa"
bcdf7404 327
78fecfc8 328/* This string is expected by regcomp.c to be ordered so that the first
c18d5d15
KW
329 * character is the flag in bit RXf_PMf_STD_PMMOD_SHIFT of extflags; the next
330 * character is bit +1, etc. */
334afb3e 331#define STD_PAT_MODS "msixx"
bcdf7404 332
94b03d7d
KW
333#define CHARSET_PAT_MODS ASCII_RESTRICT_PAT_MODS DEPENDS_PAT_MODS LOCALE_PAT_MODS UNICODE_PAT_MODS
334
78fecfc8 335/* This string is expected by XS_re_regexp_pattern() in universal.c to be ordered
c18d5d15
KW
336 * so that the first character is the flag in bit RXf_PMf_STD_PMMOD_SHIFT of
337 * extflags; the next character is in bit +1, etc. */
bcdf7404
YO
338#define INT_PAT_MODS STD_PAT_MODS KEEPCOPY_PAT_MODS
339
340#define EXT_PAT_MODS ONCE_PAT_MODS KEEPCOPY_PAT_MODS
94b03d7d 341#define QR_PAT_MODS STD_PAT_MODS EXT_PAT_MODS CHARSET_PAT_MODS
bcdf7404 342#define M_PAT_MODS QR_PAT_MODS LOOP_PAT_MODS
4f4d7508 343#define S_PAT_MODS M_PAT_MODS EXEC_PAT_MODS NONDESTRUCT_PAT_MODS
bcdf7404 344
cb5027f2 345/*
eb2624c9
FC
346 * NOTE: if you modify any RXf flags you should run regen.pl or
347 * regen/regcomp.pl so that regnodes.h is updated with the changes.
cb5027f2
YO
348 *
349 */
bcdf7404 350
dbc200c5
YO
351/*
352 Set in Perl_pmruntime if op_flags & OPf_SPECIAL, i.e. split. Will
353 be used by regex engines to check whether they should set
354 RXf_SKIPWHITE
355*/
32f62247
KW
356#define RXf_SPLIT RXf_PMf_SPLIT
357
0bd51918
KW
358/* Currently the regex flags occupy a single 32-bit word. Not all bits are
359 * currently used. The lower bits are shared with their corresponding PMf flag
360 * bits, up to but not including _RXf_PMf_SHIFT_NEXT. The unused bits
361 * immediately follow; finally the used RXf-only (unshared) bits, so that the
362 * highest bit in the word is used. This gathers all the unused bits as a pool
363 * in the middle, like so: 11111111111111110000001111111111
364 * where the '1's represent used bits, and the '0's unused. This design allows
365 * us to allocate off one end of the pool if we need to add a shared bit, and
366 * off the other end if we need a non-shared bit, without disturbing the other
367 * bits. This maximizes the likelihood of being able to change things without
368 * breaking binary compatibility.
369 *
370 * To add shared bits, do so in op_reg_common.h. This should change
371 * _RXf_PMf_SHIFT_NEXT so that things won't compile. Then come to regexp.h and
372 * op.h and adjust the constant adders in the definitions of RXf_BASE_SHIFT and
373 * Pmf_BASE_SHIFT down by the number of shared bits you added. That's it.
374 * Things should be binary compatible. But if either of these gets to having
375 * to subtract rather than add, leave at 0 and instead adjust all the entries
376 * that are in terms of it. But if the first one of those is already
377 * RXf_BASE_SHIFT+0, there are no bits left, and a redesign is in order.
378 *
379 * To remove unshared bits, just delete its entry. If you're where breaking
380 * binary compatibility is ok to do, you might want to adjust things to move
381 * the newly opened space so that it gets absorbed into the common pool.
382 *
383 * To add unshared bits, first use up any gaps in the middle. Otherwise,
384 * allocate off the low end until you get to RXf_BASE_SHIFT+0. If that isn't
385 * enough, move RXf_BASE_SHIFT down (if possible) and add the new bit at the
386 * other end instead; this preserves binary compatibility.
387 *
388 * For the regexp bits, PL_reg_extflags_name[] in regnodes.h has a comment
389 * giving which bits are used/unused */
390
334afb3e 391#define RXf_BASE_SHIFT (_RXf_PMf_SHIFT_NEXT + 4)
52d81aa8 392
bbe252da 393/* What we have seen */
3f6e6b29
KW
394#define RXf_NO_INPLACE_SUBST (1<<(RXf_BASE_SHIFT+2))
395#define RXf_EVAL_SEEN (1<<(RXf_BASE_SHIFT+3))
bbe252da
YO
396
397/* Special */
3f6e6b29
KW
398#define RXf_UNBOUNDED_QUANTIFIER_SEEN (1<<(RXf_BASE_SHIFT+4))
399#define RXf_CHECK_ALL (1<<(RXf_BASE_SHIFT+5))
bbe252da
YO
400
401/* UTF8 related */
3f6e6b29 402#define RXf_MATCH_UTF8 (1<<(RXf_BASE_SHIFT+6)) /* $1 etc are utf8 */
bbe252da
YO
403
404/* Intuit related */
3f6e6b29
KW
405#define RXf_USE_INTUIT_NOML (1<<(RXf_BASE_SHIFT+7))
406#define RXf_USE_INTUIT_ML (1<<(RXf_BASE_SHIFT+8))
407#define RXf_INTUIT_TAIL (1<<(RXf_BASE_SHIFT+9))
bbe252da
YO
408#define RXf_USE_INTUIT (RXf_USE_INTUIT_NOML|RXf_USE_INTUIT_ML)
409
a3b51d37 410/* Do we have some sort of anchor? */
3f6e6b29 411#define RXf_IS_ANCHORED (1<<(RXf_BASE_SHIFT+10))
a3b51d37 412
bbe252da 413/* Copy and tainted info */
3f6e6b29 414#define RXf_COPY_DONE (1<<(RXf_BASE_SHIFT+11))
ef07e810 415
1738e041 416/* post-execution: $1 et al are tainted */
3f6e6b29 417#define RXf_TAINTED_SEEN (1<<(RXf_BASE_SHIFT+12))
ef07e810 418/* this pattern was tainted during compilation */
3f6e6b29 419#define RXf_TAINTED (1<<(RXf_BASE_SHIFT+13))
52d81aa8
NC
420
421/* Flags indicating special patterns */
3f6e6b29
KW
422#define RXf_START_ONLY (1<<(RXf_BASE_SHIFT+14)) /* Pattern is /^/ */
423#define RXf_SKIPWHITE (1<<(RXf_BASE_SHIFT+15)) /* Pattern is for a split " " */
424#define RXf_WHITE (1<<(RXf_BASE_SHIFT+16)) /* Pattern is /\s+/ */
425#define RXf_NULL (1U<<(RXf_BASE_SHIFT+17)) /* Pattern is // */
0bd51918
KW
426
427/* See comments at the beginning of these defines about adding bits. The
428 * highest bit position should be used, so that if RXf_BASE_SHIFT gets
429 * increased, the #error below will be triggered so that you will be reminded
430 * to adjust things at the other end to keep the bit positions unchanged */
3f6e6b29 431#if RXf_BASE_SHIFT+17 > 31
0bd51918 432# error Too many RXf_PMf bits used. See comments at beginning of these for what to do
1850c8f9 433#endif
c737faaf 434
cb5027f2 435/*
eb2624c9
FC
436 * NOTE: if you modify any RXf flags you should run regen.pl or
437 * regen/regcomp.pl so that regnodes.h is updated with the changes.
cb5027f2
YO
438 *
439 */
bbe252da 440
dc6d7f5c 441#ifdef NO_TAINT_SUPPORT
284167a5
SM
442# define RX_ISTAINTED(prog) 0
443# define RX_TAINT_on(prog) NOOP
272d35c9
DM
444# define RXp_MATCH_TAINTED(prog) 0
445# define RX_MATCH_TAINTED(prog) 0
446# define RXp_MATCH_TAINTED_on(prog) NOOP
447# define RX_MATCH_TAINTED_on(prog) NOOP
448# define RX_MATCH_TAINTED_off(prog) NOOP
284167a5
SM
449#else
450# define RX_ISTAINTED(prog) (RX_EXTFLAGS(prog) & RXf_TAINTED)
451# define RX_TAINT_on(prog) (RX_EXTFLAGS(prog) |= RXf_TAINTED)
272d35c9
DM
452# define RXp_MATCH_TAINTED(prog) (RXp_EXTFLAGS(prog) & RXf_TAINTED_SEEN)
453# define RX_MATCH_TAINTED(prog) (RX_EXTFLAGS(prog) & RXf_TAINTED_SEEN)
454# define RXp_MATCH_TAINTED_on(prog) (RXp_EXTFLAGS(prog) |= RXf_TAINTED_SEEN)
455# define RX_MATCH_TAINTED_on(prog) (RX_EXTFLAGS(prog) |= RXf_TAINTED_SEEN)
456# define RX_MATCH_TAINTED_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_TAINTED_SEEN)
284167a5
SM
457#endif
458
bbe252da 459#define RX_HAS_CUTGROUP(prog) ((prog)->intflags & PREGf_CUTGROUP_SEEN)
72311751
GS
460#define RX_MATCH_TAINTED_set(prog, t) ((t) \
461 ? RX_MATCH_TAINTED_on(prog) \
462 : RX_MATCH_TAINTED_off(prog))
c277df42 463
07bc277f
NC
464#define RXp_MATCH_COPIED(prog) (RXp_EXTFLAGS(prog) & RXf_COPY_DONE)
465#define RX_MATCH_COPIED(prog) (RX_EXTFLAGS(prog) & RXf_COPY_DONE)
466#define RXp_MATCH_COPIED_on(prog) (RXp_EXTFLAGS(prog) |= RXf_COPY_DONE)
467#define RX_MATCH_COPIED_on(prog) (RX_EXTFLAGS(prog) |= RXf_COPY_DONE)
468#define RXp_MATCH_COPIED_off(prog) (RXp_EXTFLAGS(prog) &= ~RXf_COPY_DONE)
469#define RX_MATCH_COPIED_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_COPY_DONE)
cf93c79d
IZ
470#define RX_MATCH_COPIED_set(prog,t) ((t) \
471 ? RX_MATCH_COPIED_on(prog) \
472 : RX_MATCH_COPIED_off(prog))
f9f4320a 473
07bc277f 474#define RXp_EXTFLAGS(rx) ((rx)->extflags)
dbc200c5 475#define RXp_COMPFLAGS(rx) ((rx)->compflags)
866c78d1 476
07bc277f 477/* For source compatibility. We used to store these explicitly. */
8d919b0a
FC
478#define RX_PRECOMP(prog) (RX_WRAPPED(prog) + ReANY(prog)->pre_prefix)
479#define RX_PRECOMP_const(prog) (RX_WRAPPED_const(prog) + ReANY(prog)->pre_prefix)
5509d87a
NC
480/* FIXME? Are we hardcoding too much here and constraining plugin extension
481 writers? Specifically, the value 1 assumes that the wrapped version always
482 has exactly one character at the end, a ')'. Will that always be true? */
8d919b0a
FC
483#define RX_PRELEN(prog) (RX_WRAPLEN(prog) - ReANY(prog)->pre_prefix - 1)
484#define RX_WRAPPED(prog) ReANY(prog)->xpv_len_u.xpvlenu_pv
485#define RX_WRAPPED_const(prog) ((const char *)RX_WRAPPED(prog))
9d17798d 486#define RX_WRAPLEN(prog) SvCUR(prog)
8d919b0a 487#define RX_CHECK_SUBSTR(prog) (ReANY(prog)->check_substr)
288b8c02 488#define RX_REFCNT(prog) SvREFCNT(prog)
8d919b0a 489#define RX_EXTFLAGS(prog) RXp_EXTFLAGS(ReANY(prog))
dbc200c5 490#define RX_COMPFLAGS(prog) RXp_COMPFLAGS(ReANY(prog))
8d919b0a
FC
491#define RX_ENGINE(prog) (ReANY(prog)->engine)
492#define RX_SUBBEG(prog) (ReANY(prog)->subbeg)
493#define RX_SUBOFFSET(prog) (ReANY(prog)->suboffset)
494#define RX_SUBCOFFSET(prog) (ReANY(prog)->subcoffset)
495#define RX_OFFS(prog) (ReANY(prog)->offs)
496#define RX_NPARENS(prog) (ReANY(prog)->nparens)
497#define RX_SUBLEN(prog) (ReANY(prog)->sublen)
498#define RX_MINLEN(prog) (ReANY(prog)->minlen)
499#define RX_MINLENRET(prog) (ReANY(prog)->minlenret)
500#define RX_GOFS(prog) (ReANY(prog)->gofs)
501#define RX_LASTPAREN(prog) (ReANY(prog)->lastparen)
502#define RX_LASTCLOSEPAREN(prog) (ReANY(prog)->lastcloseparen)
503#define RX_SAVED_COPY(prog) (ReANY(prog)->saved_copy)
03c83e26
DM
504/* last match was zero-length */
505#define RX_ZERO_LEN(prog) \
99a90e59
FC
506 (RX_OFFS(prog)[0].start + (SSize_t)RX_GOFS(prog) \
507 == RX_OFFS(prog)[0].end)
220fc49f 508
be8e71aa
YO
509#endif /* PLUGGABLE_RE_EXTENSION */
510
486ec47a 511/* Stuff that needs to be included in the pluggable extension goes below here */
be8e71aa 512
db2c6cb3 513#ifdef PERL_ANY_COW
ed252734 514#define RX_MATCH_COPY_FREE(rx) \
bdd9a1b1
NC
515 STMT_START {if (RX_SAVED_COPY(rx)) { \
516 SV_CHECK_THINKFIRST_COW_DROP(RX_SAVED_COPY(rx)); \
ed252734
NC
517 } \
518 if (RX_MATCH_COPIED(rx)) { \
07bc277f 519 Safefree(RX_SUBBEG(rx)); \
ed252734
NC
520 RX_MATCH_COPIED_off(rx); \
521 }} STMT_END
522#else
523#define RX_MATCH_COPY_FREE(rx) \
524 STMT_START {if (RX_MATCH_COPIED(rx)) { \
07bc277f 525 Safefree(RX_SUBBEG(rx)); \
ed252734
NC
526 RX_MATCH_COPIED_off(rx); \
527 }} STMT_END
528#endif
529
07bc277f
NC
530#define RXp_MATCH_UTF8(prog) (RXp_EXTFLAGS(prog) & RXf_MATCH_UTF8)
531#define RX_MATCH_UTF8(prog) (RX_EXTFLAGS(prog) & RXf_MATCH_UTF8)
532#define RX_MATCH_UTF8_on(prog) (RX_EXTFLAGS(prog) |= RXf_MATCH_UTF8)
533#define RX_MATCH_UTF8_off(prog) (RX_EXTFLAGS(prog) &= ~RXf_MATCH_UTF8)
a30b2f1f 534#define RX_MATCH_UTF8_set(prog, t) ((t) \
0603fe5c
DM
535 ? RX_MATCH_UTF8_on(prog) \
536 : RX_MATCH_UTF8_off(prog))
efd26800
NC
537
538/* Whether the pattern stored at RX_WRAPPED is in UTF-8 */
8f6ae13c 539#define RX_UTF8(prog) SvUTF8(prog)
7948fc08 540
a340edde
DM
541
542/* bits in flags arg of Perl_regexec_flags() */
543
544#define REXEC_COPY_STR 0x01 /* Need to copy the string for captures. */
545#define REXEC_CHECKED 0x02 /* re_intuit_start() already called. */
546#define REXEC_SCREAM 0x04 /* currently unused. */
547#define REXEC_IGNOREPOS 0x08 /* use stringarg, not pos(), for \G match */
548#define REXEC_NOT_FIRST 0x10 /* This is another iteration of //g:
549 no need to copy string again */
550
551 /* under REXEC_COPY_STR, it's ok for the
552 engine (modulo PL_sawamperand etc)
553 to skip copying: ... */
554#define REXEC_COPY_SKIP_PRE 0x20 /* ...the $` part of the string, or */
555#define REXEC_COPY_SKIP_POST 0x40 /* ...the $' part of the string */
d5e7783a
DM
556#define REXEC_FAIL_ON_UNDERFLOW 0x80 /* fail the match if $& would start before
557 the start pos (so s/.\G// would fail
558 on second iteration */
c277df42 559
7fe44985 560#if defined(__GNUC__) && !defined(PERL_GCC_BRACE_GROUPS_FORBIDDEN)
288b8c02
NC
561# define ReREFCNT_inc(re) \
562 ({ \
563 /* This is here to generate a casting warning if incorrect. */ \
7948fc08
RU
564 REGEXP *const _rerefcnt_inc = (re); \
565 assert(SvTYPE(_rerefcnt_inc) == SVt_REGEXP); \
566 SvREFCNT_inc(_rerefcnt_inc); \
567 _rerefcnt_inc; \
288b8c02
NC
568 })
569# define ReREFCNT_dec(re) \
570 ({ \
571 /* This is here to generate a casting warning if incorrect. */ \
7948fc08
RU
572 REGEXP *const _rerefcnt_dec = (re); \
573 SvREFCNT_dec(_rerefcnt_dec); \
288b8c02
NC
574 })
575#else
576# define ReREFCNT_dec(re) SvREFCNT_dec(re)
b7e0bd9e 577# define ReREFCNT_inc(re) ((REGEXP *) SvREFCNT_inc(re))
288b8c02 578#endif
8d919b0a 579#define ReANY(re) S_ReANY((const REGEXP *)(re))
288b8c02
NC
580
581/* FIXME for plugins. */
cf93c79d
IZ
582
583#define FBMcf_TAIL_DOLLAR 1
cad2e5aa
JH
584#define FBMcf_TAIL_DOLLARM 2
585#define FBMcf_TAIL_Z 4
586#define FBMcf_TAIL_z 8
587#define FBMcf_TAIL (FBMcf_TAIL_DOLLAR|FBMcf_TAIL_DOLLARM|FBMcf_TAIL_Z|FBMcf_TAIL_z)
cf93c79d
IZ
588
589#define FBMrf_MULTILINE 1
cad2e5aa 590
331b2dcc
DM
591struct regmatch_state;
592struct regmatch_slab;
8adc0f72 593
bf2039a9
DM
594/* like regmatch_info_aux, but contains extra fields only needed if the
595 * pattern contains (?{}). If used, is snuck into the second slot in the
596 * regmatch_state stack at the start of execution */
8adc0f72
DM
597
598typedef struct {
599 regexp *rex;
600 PMOP *curpm; /* saved PL_curpm */
601#ifdef PERL_ANY_COW
602 SV *saved_copy; /* saved saved_copy field from rex */
603#endif
604 char *subbeg; /* saved subbeg field from rex */
605 STRLEN sublen; /* saved sublen field from rex */
606 STRLEN suboffset; /* saved suboffset field from rex */
607 STRLEN subcoffset; /* saved subcoffset field from rex */
608 MAGIC *pos_magic; /* pos() magic attached to $_ */
ea3daa5d 609 SSize_t pos; /* the original value of pos() in pos_magic */
25fdce4a 610 U8 pos_flags; /* flags to be restored; currently only MGf_BYTES*/
bf2039a9
DM
611} regmatch_info_aux_eval;
612
613
614/* fields that logically live in regmatch_info, but which need cleaning
615 * up on croak(), and so are instead are snuck into the first slot in
616 * the regmatch_state stack at the start of execution */
617
618typedef struct {
619 regmatch_info_aux_eval *info_aux_eval;
331b2dcc
DM
620 struct regmatch_state *old_regmatch_state; /* saved PL_regmatch_state */
621 struct regmatch_slab *old_regmatch_slab; /* saved PL_regmatch_slab */
2ac8ff4b 622 char *poscache; /* S-L cache of fail positions of WHILEMs */
bf2039a9
DM
623} regmatch_info_aux;
624
8adc0f72 625
3b0527fe 626/* some basic information about the current match that is created by
bf2039a9
DM
627 * Perl_regexec_flags and then passed to regtry(), regmatch() etc.
628 * It is allocated as a local var on the stack, so nothing should be
629 * stored in it that needs preserving or clearing up on croak().
630 * For that, see the aux_info and aux_info_eval members of the
631 * regmatch_state union. */
3b0527fe
DM
632
633typedef struct {
28d03b21 634 REGEXP *prog; /* the regex being executed */
9d9163fb 635 const char * strbeg; /* real start of string */
28d03b21
DM
636 char *strend; /* one byte beyond last char of match string */
637 char *till; /* matches shorter than this fail (see minlen arg) */
638 SV *sv; /* the SV string currently being matched */
639 char *ganch; /* position of \G anchor */
640 char *cutpoint; /* (*COMMIT) position (if any) */
bf2039a9
DM
641 regmatch_info_aux *info_aux; /* extra fields that need cleanup */
642 regmatch_info_aux_eval *info_aux_eval; /* extra saved state for (?{}) */
1cb48e53
DM
643 I32 poscache_maxiter; /* how many whilems todo before S-L cache kicks in */
644 I32 poscache_iter; /* current countdown from _maxiter to zero */
2ac8ff4b 645 STRLEN poscache_size; /* size of regmatch_info_aux.poscache */
02d5137b 646 bool intuit; /* re_intuit_start() is the top-level caller */
ba44c216
DM
647 bool is_utf8_pat; /* regex is utf8 */
648 bool is_utf8_target; /* string being matched is utf8 */
39819bd9 649 bool warned; /* we have issued a recursion warning; no need for more */
3b0527fe
DM
650} regmatch_info;
651
5d9a96ca
DM
652
653/* structures for holding and saving the state maintained by regmatch() */
654
d56b3014 655#ifndef MAX_RECURSE_EVAL_NOCHANGE_DEPTH
4b196cd4 656#define MAX_RECURSE_EVAL_NOCHANGE_DEPTH 1000
d56b3014 657#endif
6bda09f9 658
5d9a96ca
DM
659typedef I32 CHECKPOINT;
660
a0374537 661typedef struct regmatch_state {
40a82448 662 int resume_state; /* where to jump to on return */
24d3c4a9 663 char *locinput; /* where to backtrack in string on failure */
e822a8b4
DM
664
665 union {
77cb431f 666
bf2039a9
DM
667 /* the 'info_aux' and 'info_aux_eval' union members are cuckoos in
668 * the nest. They aren't saved backtrack state; rather they
669 * represent one or two extra chunks of data that need allocating
670 * at the start of a match. These fields would logically live in
671 * the regmatch_info struct, except that is allocated on the
672 * C stack, and these fields are all things that require cleanup
673 * after a croak(), when the stack is lost.
674 * As a convenience, we just use the first 1 or 2 regmatch_state
675 * slots to store this info, as we will be allocating a slab of
676 * these anyway. Otherwise we'd have to malloc and then free them,
677 * or allocate them on the save stack (where they will get
678 * realloced if the save stack grows).
679 * info_aux contains the extra fields that are always needed;
680 * info_aux_eval contains extra fields that only needed if
681 * the pattern contains code blocks
682 * We split them into two separate structs to avoid increasing
683 * the size of the union.
684 */
685
686 regmatch_info_aux info_aux;
687
688 regmatch_info_aux_eval info_aux_eval;
689
77cb431f
DM
690 /* this is a fake union member that matches the first element
691 * of each member that needs to store positive backtrack
692 * information */
693 struct {
694 struct regmatch_state *prev_yes_state;
695 } yes;
696
fae667d5
YO
697 /* branchlike members */
698 /* this is a fake union member that matches the first elements
699 * of each member that needs to behave like a branch */
700 struct {
5d458dd8
YO
701 /* this first element must match u.yes */
702 struct regmatch_state *prev_yes_state;
fae667d5 703 U32 lastparen;
f6033a9d 704 U32 lastcloseparen;
fae667d5
YO
705 CHECKPOINT cp;
706
707 } branchlike;
708
709 struct {
710 /* the first elements must match u.branchlike */
711 struct regmatch_state *prev_yes_state;
712 U32 lastparen;
f6033a9d 713 U32 lastcloseparen;
fae667d5
YO
714 CHECKPOINT cp;
715
716 regnode *next_branch; /* next branch node */
717 } branch;
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
2e64971a 726 U32 accepted; /* how many accepting states left */
6603fe3e 727 bool longfold;/* saw a fold with a 1->n char mapping */
7f69552c 728 U16 *jump; /* positive offsets from me */
7f69552c 729 regnode *me; /* Which node am I - needed for jump tries*/
2e64971a
DM
730 U8 *firstpos;/* pos in string of first trie match */
731 U32 firstchars;/* len in chars of firstpos from start */
732 U16 nextword;/* next word to try */
733 U16 topword; /* longest accepted word */
e822a8b4
DM
734 } trie;
735
fae667d5
YO
736 /* special types - these members are used to store state for special
737 regops like eval, if/then, lookaround and the markpoint state */
e822a8b4 738 struct {
77cb431f
DM
739 /* this first element must match u.yes */
740 struct regmatch_state *prev_yes_state;
faec1544
DM
741 struct regmatch_state *prev_eval;
742 struct regmatch_state *prev_curlyx;
288b8c02 743 REGEXP *prev_rex;
aa283a38
DM
744 CHECKPOINT cp; /* remember current savestack indexes */
745 CHECKPOINT lastcp;
6bda09f9 746 U32 close_paren; /* which close bracket is our end */
1b5c067c 747 regnode *B; /* the node following us */
e822a8b4
DM
748 } eval;
749
750 struct {
c476f425 751 /* this first element must match u.yes */
262b90c4 752 struct regmatch_state *prev_yes_state;
fae667d5
YO
753 I32 wanted;
754 I32 logical; /* saved copy of 'logical' var */
755 regnode *me; /* the IFMATCH/SUSPEND/UNLESSM node */
756 } ifmatch; /* and SUSPEND/UNLESSM */
757
758 struct {
759 /* this first element must match u.yes */
760 struct regmatch_state *prev_yes_state;
761 struct regmatch_state *prev_mark;
762 SV* mark_name;
763 char *mark_loc;
764 } mark;
765
766 struct {
767 int val;
768 } keeper;
769
770 /* quantifiers - these members are used for storing state for
771 for the regops used to implement quantifiers */
772 struct {
773 /* this first element must match u.yes */
774 struct regmatch_state *prev_yes_state;
c476f425 775 struct regmatch_state *prev_curlyx; /* previous cur_curlyx */
d02d6d97
DM
776 regnode *me; /* the CURLYX node */
777 regnode *B; /* the B node in /A*B/ */
c476f425
DM
778 CHECKPOINT cp; /* remember current savestack index */
779 bool minmod;
a0374537 780 int parenfloor;/* how far back to strip paren data */
c476f425
DM
781
782 /* these two are modified by WHILEM */
783 int count; /* how many instances of A we've matched */
784 char *lastloc;/* where previous A matched (0-len detect) */
e822a8b4
DM
785 } curlyx;
786
787 struct {
c476f425 788 /* this first element must match u.yes */
262b90c4 789 struct regmatch_state *prev_yes_state;
c476f425
DM
790 struct regmatch_state *save_curlyx;
791 CHECKPOINT cp; /* remember current savestack indexes */
792 CHECKPOINT lastcp;
793 char *save_lastloc; /* previous curlyx.lastloc */
794 I32 cache_offset;
795 I32 cache_mask;
e822a8b4
DM
796 } whilem;
797
798 struct {
5d458dd8
YO
799 /* this first element must match u.yes */
800 struct regmatch_state *prev_yes_state;
79a2a0e8 801 int c1, c2; /* case fold search */
40a82448 802 CHECKPOINT cp;
f6033a9d
DM
803 U32 lastparen;
804 U32 lastcloseparen;
40a82448
DM
805 I32 alen; /* length of first-matched A string */
806 I32 count;
0cadcf80 807 bool minmod;
40a82448
DM
808 regnode *A, *B; /* the nodes corresponding to /A*B/ */
809 regnode *me; /* the curlym node */
79a2a0e8
KW
810 U8 c1_utf8[UTF8_MAXBYTES+1]; /* */
811 U8 c2_utf8[UTF8_MAXBYTES+1];
e822a8b4
DM
812 } curlym;
813
814 struct {
3b6647e0 815 U32 paren;
c255a977 816 CHECKPOINT cp;
f6033a9d
DM
817 U32 lastparen;
818 U32 lastcloseparen;
79a2a0e8 819 int c1, c2; /* case fold search */
c255a977
DM
820 char *maxpos; /* highest possible point in string to match */
821 char *oldloc; /* the previous locinput */
e822a8b4 822 int count;
c255a977
DM
823 int min, max; /* {m,n} */
824 regnode *A, *B; /* the nodes corresponding to /A*B/ */
79a2a0e8
KW
825 U8 c1_utf8[UTF8_MAXBYTES+1]; /* */
826 U8 c2_utf8[UTF8_MAXBYTES+1];
c255a977 827 } curly; /* and CURLYN/PLUS/STAR */
dad79028 828
d8319b27 829 } u;
5d9a96ca
DM
830} regmatch_state;
831
832/* how many regmatch_state structs to allocate as a single slab.
833 * We do it in 4K blocks for efficiency. The "3" is 2 for the next/prev
834 * pointers, plus 1 for any mythical malloc overhead. */
835
836#define PERL_REGMATCH_SLAB_SLOTS \
837 ((4096 - 3 * sizeof (void*)) / sizeof(regmatch_state))
838
839typedef struct regmatch_slab {
840 regmatch_state states[PERL_REGMATCH_SLAB_SLOTS];
841 struct regmatch_slab *prev, *next;
842} regmatch_slab;
1ade1aa1 843
46ab3289 844
cde0cee5 845
1ade1aa1
NC
846/*
847 * Local variables:
848 * c-indentation-style: bsd
849 * c-basic-offset: 4
14d04a33 850 * indent-tabs-mode: nil
1ade1aa1
NC
851 * End:
852 *
14d04a33 853 * ex: set ts=8 sts=4 sw=4 et:
1ade1aa1 854 */