3 * Copyright (C) 1993, 1994, 1996, 1997, 1999, 2000, 2001, 2003,
4 * 2005, 2006 by Larry Wall and others
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.
12 * Definitions etc. for regexp(3) routines.
14 * Caveat: this is V8 regexp(3) [actually, a reimplementation thereof],
15 * not the System V one.
17 #ifndef PLUGGABLE_RE_EXTENSION
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'*/
27 typedef struct regnode regnode;
29 struct reg_substr_data;
36 struct reg_substr_datum {
39 SV *substr; /* non-utf8 variant */
40 SV *utf8_substr; /* utf8 variant */
43 struct reg_substr_data {
44 struct reg_substr_datum data[3]; /* Actual array */
47 #ifdef PERL_OLD_COPY_ON_WRITE
48 #define SV_SAVED_COPY SV *saved_copy; /* If non-NULL, SV which is COW from original */
53 typedef struct regexp_paren_pair {
59 The regexp/REGEXP struct, see L<perlreapi> for further documentation
60 on the individual fields. The struct is ordered so that the most
61 commonly used fields are placed at the start.
63 Any patch that adds items to this struct will need to include
64 changes to F<sv.c> (C<Perl_re_dup()>) and F<regcomp.c>
65 (C<pregfree()>). This involves freeing or cloning items in the
66 regexp's data array based on the data item's type.
69 typedef struct regexp {
70 /* what engine created this regexp? */
71 const struct regexp_engine* engine;
72 struct regexp* mother_re; /* what re is this a lightweight copy of? */
74 /* Information about the match that the perl core uses to manage things */
75 U32 extflags; /* Flags used both externally and internally */
76 I32 minlen; /* mininum possible length of string to match */
77 I32 minlenret; /* mininum possible length of $& */
78 U32 gofs; /* chars left of pos that we search from */
79 struct reg_substr_data *substrs; /* substring data about strings that must appear
80 in the final match, used for optimisations */
81 U32 nparens; /* number of capture buffers */
83 /* private engine specific data */
84 U32 intflags; /* Engine Specific Internal flags */
85 void *pprivate; /* Data private to the regex engine which
86 created this object. */
88 /* Data about the last/current match. These are modified during matching*/
89 U32 lastparen; /* last open paren matched */
90 U32 lastcloseparen; /* last close paren matched */
91 regexp_paren_pair *swap; /* Swap copy of *offs */
92 regexp_paren_pair *offs; /* Array of offsets for (@-) and (@+) */
94 char *subbeg; /* saved or original string
95 so \digit works forever. */
96 SV_SAVED_COPY /* If non-NULL, SV which is COW from original */
97 I32 sublen; /* Length of string pointed by subbeg */
100 /* Information about the match that isn't often used */
101 I32 prelen; /* length of precomp */
102 const char *precomp; /* pre-compilation regular expression */
103 /* wrapped can't be const char*, as it is returned by sv_2pv_flags */
104 char *wrapped; /* wrapped version of the pattern */
105 I32 wraplen; /* length of wrapped */
106 I32 seen_evals; /* number of eval groups in the pattern - for security checks */
107 HV *paren_names; /* Optional hash of paren names */
109 /* Refcount of this regexp */
110 I32 refcnt; /* Refcount of this regexp */
113 /* used for high speed searches */
114 typedef struct re_scream_pos_data_s
116 char **scream_olds; /* match pos */
117 I32 *scream_pos; /* Internal iterator of scream. */
118 } re_scream_pos_data;
120 /* regexp_engine structure. This is the dispatch table for regexes.
121 * Any regex engine implementation must be able to build one of these.
123 typedef struct regexp_engine {
124 REGEXP* (*comp) (pTHX_ const SV * const pattern, const U32 flags);
125 I32 (*exec) (pTHX_ REGEXP * const rx, char* stringarg, char* strend,
126 char* strbeg, I32 minend, SV* screamer,
127 void* data, U32 flags);
128 char* (*intuit) (pTHX_ REGEXP * const rx, SV *sv, char *strpos,
129 char *strend, const U32 flags,
130 re_scream_pos_data *data);
131 SV* (*checkstr) (pTHX_ REGEXP * const rx);
132 void (*free) (pTHX_ REGEXP * const rx);
133 void (*numbered_buff_FETCH) (pTHX_ REGEXP * const rx, const I32 paren,
135 void (*numbered_buff_STORE) (pTHX_ REGEXP * const rx, const I32 paren,
136 SV const * const value);
137 I32 (*numbered_buff_LENGTH) (pTHX_ REGEXP * const rx, const SV * const sv,
139 SV* (*named_buff_FETCH) (pTHX_ REGEXP * const rx, SV * const key,
141 SV* (*qr_package)(pTHX_ REGEXP * const rx);
143 void* (*dupe) (pTHX_ REGEXP * const rx, CLONE_PARAMS *param);
147 /* Flags stored in regexp->extflags
148 * These are used by code external to the regexp engine
150 * Note that flags starting with RXf_PMf_ have exact equivalents
151 * stored in op_pmflags and which are defined in op.h, they are defined
152 * numerically here only for clarity.
155 /* Anchor and GPOS related stuff */
156 #define RXf_ANCH_BOL 0x00000001
157 #define RXf_ANCH_MBOL 0x00000002
158 #define RXf_ANCH_SBOL 0x00000004
159 #define RXf_ANCH_GPOS 0x00000008
160 #define RXf_GPOS_SEEN 0x00000010
161 #define RXf_GPOS_FLOAT 0x00000020
163 #define RXf_ANCH (RXf_ANCH_BOL|RXf_ANCH_MBOL|RXf_ANCH_GPOS|RXf_ANCH_SBOL)
164 #define RXf_GPOS_CHECK (RXf_GPOS_SEEN|RXf_ANCH_GPOS)
165 #define RXf_ANCH_SINGLE (RXf_ANCH_SBOL|RXf_ANCH_GPOS)
167 /* Flags indicating special patterns */
168 #define RXf_SKIPWHITE 0x00000100 /* Pattern is for a split / / */
169 #define RXf_START_ONLY 0x00000200 /* Pattern is /^/ */
170 #define RXf_WHITE 0x00000400 /* Pattern is /\s+/ */
172 /* 0x1F800 of extflags is used by (RXf_)PMf_COMPILETIME */
173 #define RXf_PMf_LOCALE 0x00000800 /* use locale */
174 #define RXf_PMf_MULTILINE 0x00001000 /* /m */
175 #define RXf_PMf_SINGLELINE 0x00002000 /* /s */
176 #define RXf_PMf_FOLD 0x00004000 /* /i */
177 #define RXf_PMf_EXTENDED 0x00008000 /* /x */
178 #define RXf_PMf_KEEPCOPY 0x00010000 /* /k */
179 /* these flags are transfered from the PMOP->op_pmflags member during compilation */
180 #define RXf_PMf_STD_PMMOD (RXf_PMf_MULTILINE|RXf_PMf_SINGLELINE|RXf_PMf_FOLD|RXf_PMf_EXTENDED)
181 #define RXf_PMf_COMPILETIME (RXf_PMf_MULTILINE|RXf_PMf_SINGLELINE|RXf_PMf_LOCALE|RXf_PMf_FOLD|RXf_PMf_EXTENDED|RXf_PMf_KEEPCOPY)
183 #define CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl) \
184 case IGNORE_PAT_MOD: *(pmfl) |= RXf_PMf_FOLD; break; \
185 case MULTILINE_PAT_MOD: *(pmfl) |= RXf_PMf_MULTILINE; break; \
186 case SINGLE_PAT_MOD: *(pmfl) |= RXf_PMf_SINGLELINE; break; \
187 case XTENDED_PAT_MOD: *(pmfl) |= RXf_PMf_EXTENDED; break
189 /* chars and strings used as regex pattern modifiers
190 * Singlular is a 'c'har, plural is a "string"
192 * NOTE, KEEPCOPY was originally 'k', but was changed to 'p' for preserve
193 * for compatibility reasons with Regexp::Common which highjacked (?k:...)
194 * for its own uses. So 'k' is out as well.
196 #define EXEC_PAT_MOD 'e'
197 #define KEEPCOPY_PAT_MOD 'p'
198 #define ONCE_PAT_MOD 'o'
199 #define GLOBAL_PAT_MOD 'g'
200 #define CONTINUE_PAT_MOD 'c'
201 #define MULTILINE_PAT_MOD 'm'
202 #define SINGLE_PAT_MOD 's'
203 #define IGNORE_PAT_MOD 'i'
204 #define XTENDED_PAT_MOD 'x'
206 #define ONCE_PAT_MODS "o"
207 #define KEEPCOPY_PAT_MODS "p"
208 #define EXEC_PAT_MODS "e"
209 #define LOOP_PAT_MODS "gc"
211 #define STD_PAT_MODS "msix"
213 #define INT_PAT_MODS STD_PAT_MODS KEEPCOPY_PAT_MODS
215 #define EXT_PAT_MODS ONCE_PAT_MODS KEEPCOPY_PAT_MODS
216 #define QR_PAT_MODS STD_PAT_MODS EXT_PAT_MODS
217 #define M_PAT_MODS QR_PAT_MODS LOOP_PAT_MODS
218 #define S_PAT_MODS M_PAT_MODS EXEC_PAT_MODS
221 /* What we have seen */
222 #define RXf_LOOKBEHIND_SEEN 0x00020000
223 #define RXf_EVAL_SEEN 0x00040000
224 #define RXf_CANY_SEEN 0x00080000
227 #define RXf_NOSCAN 0x00100000
228 #define RXf_CHECK_ALL 0x00200000
231 #define RXf_UTF8 0x00400000
232 #define RXf_MATCH_UTF8 0x00800000
235 #define RXf_USE_INTUIT_NOML 0x01000000
236 #define RXf_USE_INTUIT_ML 0x02000000
237 #define RXf_INTUIT_TAIL 0x04000000
239 #define RXf_USE_INTUIT (RXf_USE_INTUIT_NOML|RXf_USE_INTUIT_ML)
241 /* Copy and tainted info */
242 #define RXf_COPY_DONE 0x10000000
243 #define RXf_TAINTED_SEEN 0x20000000
244 #define RXf_TAINTED 0x80000000 /* this pattern is tainted */
247 #define RX_HAS_CUTGROUP(prog) ((prog)->intflags & PREGf_CUTGROUP_SEEN)
248 #define RX_MATCH_TAINTED(prog) ((prog)->extflags & RXf_TAINTED_SEEN)
249 #define RX_MATCH_TAINTED_on(prog) ((prog)->extflags |= RXf_TAINTED_SEEN)
250 #define RX_MATCH_TAINTED_off(prog) ((prog)->extflags &= ~RXf_TAINTED_SEEN)
251 #define RX_MATCH_TAINTED_set(prog, t) ((t) \
252 ? RX_MATCH_TAINTED_on(prog) \
253 : RX_MATCH_TAINTED_off(prog))
255 #define RX_MATCH_COPIED(prog) ((prog)->extflags & RXf_COPY_DONE)
256 #define RX_MATCH_COPIED_on(prog) ((prog)->extflags |= RXf_COPY_DONE)
257 #define RX_MATCH_COPIED_off(prog) ((prog)->extflags &= ~RXf_COPY_DONE)
258 #define RX_MATCH_COPIED_set(prog,t) ((t) \
259 ? RX_MATCH_COPIED_on(prog) \
260 : RX_MATCH_COPIED_off(prog))
262 #endif /* PLUGGABLE_RE_EXTENSION */
264 /* Stuff that needs to be included in the plugable extension goes below here */
266 #ifdef PERL_OLD_COPY_ON_WRITE
267 #define RX_MATCH_COPY_FREE(rx) \
268 STMT_START {if (rx->saved_copy) { \
269 SV_CHECK_THINKFIRST_COW_DROP(rx->saved_copy); \
271 if (RX_MATCH_COPIED(rx)) { \
272 Safefree(rx->subbeg); \
273 RX_MATCH_COPIED_off(rx); \
276 #define RX_MATCH_COPY_FREE(rx) \
277 STMT_START {if (RX_MATCH_COPIED(rx)) { \
278 Safefree(rx->subbeg); \
279 RX_MATCH_COPIED_off(rx); \
283 #define RX_MATCH_UTF8(prog) ((prog)->extflags & RXf_MATCH_UTF8)
284 #define RX_MATCH_UTF8_on(prog) ((prog)->extflags |= RXf_MATCH_UTF8)
285 #define RX_MATCH_UTF8_off(prog) ((prog)->extflags &= ~RXf_MATCH_UTF8)
286 #define RX_MATCH_UTF8_set(prog, t) ((t) \
287 ? (RX_MATCH_UTF8_on(prog), (PL_reg_match_utf8 = 1)) \
288 : (RX_MATCH_UTF8_off(prog), (PL_reg_match_utf8 = 0)))
290 #define REXEC_COPY_STR 0x01 /* Need to copy the string. */
291 #define REXEC_CHECKED 0x02 /* check_substr already checked. */
292 #define REXEC_SCREAM 0x04 /* use scream table. */
293 #define REXEC_IGNOREPOS 0x08 /* \G matches at start. */
294 #define REXEC_NOT_FIRST 0x10 /* This is another iteration of //g. */
296 #define ReREFCNT_inc(re) ((void)(re && re->refcnt++), re)
297 #define ReREFCNT_dec(re) CALLREGFREE(re)
299 #define FBMcf_TAIL_DOLLAR 1
300 #define FBMcf_TAIL_DOLLARM 2
301 #define FBMcf_TAIL_Z 4
302 #define FBMcf_TAIL_z 8
303 #define FBMcf_TAIL (FBMcf_TAIL_DOLLAR|FBMcf_TAIL_DOLLARM|FBMcf_TAIL_Z|FBMcf_TAIL_z)
305 #define FBMrf_MULTILINE 1
307 /* an accepting state/position*/
308 struct _reg_trie_accepted {
312 typedef struct _reg_trie_accepted reg_trie_accepted;
314 /* some basic information about the current match that is created by
315 * Perl_regexec_flags and then passed to regtry(), regmatch() etc */
327 /* structures for holding and saving the state maintained by regmatch() */
329 #ifndef MAX_RECURSE_EVAL_NOCHANGE_DEPTH
330 #define MAX_RECURSE_EVAL_NOCHANGE_DEPTH 1000
333 typedef I32 CHECKPOINT;
335 typedef struct regmatch_state {
336 int resume_state; /* where to jump to on return */
337 char *locinput; /* where to backtrack in string on failure */
341 /* this is a fake union member that matches the first element
342 * of each member that needs to store positive backtrack
345 struct regmatch_state *prev_yes_state;
348 /* branchlike members */
349 /* this is a fake union member that matches the first elements
350 * of each member that needs to behave like a branch */
352 /* this first element must match u.yes */
353 struct regmatch_state *prev_yes_state;
360 /* the first elements must match u.branchlike */
361 struct regmatch_state *prev_yes_state;
365 regnode *next_branch; /* next branch node */
369 /* the first elements must match u.branchlike */
370 struct regmatch_state *prev_yes_state;
374 reg_trie_accepted *accept_buff; /* accepting states we have seen */
375 U32 accepted; /* how many accepting states we have seen */
376 U16 *jump; /* positive offsets from me */
377 regnode *B; /* node following the trie */
378 regnode *me; /* Which node am I - needed for jump tries*/
381 /* special types - these members are used to store state for special
382 regops like eval, if/then, lookaround and the markpoint state */
384 /* this first element must match u.yes */
385 struct regmatch_state *prev_yes_state;
386 struct regmatch_state *prev_eval;
387 struct regmatch_state *prev_curlyx;
389 U32 toggle_reg_flags; /* what bits in PL_reg_flags to
390 flip when transitioning between
391 inner and outer rexen */
392 CHECKPOINT cp; /* remember current savestack indexes */
394 U32 close_paren; /* which close bracket is our end */
395 regnode *B; /* the node following us */
399 /* this first element must match u.yes */
400 struct regmatch_state *prev_yes_state;
402 I32 logical; /* saved copy of 'logical' var */
403 regnode *me; /* the IFMATCH/SUSPEND/UNLESSM node */
404 } ifmatch; /* and SUSPEND/UNLESSM */
407 /* this first element must match u.yes */
408 struct regmatch_state *prev_yes_state;
409 struct regmatch_state *prev_mark;
418 /* quantifiers - these members are used for storing state for
419 for the regops used to implement quantifiers */
421 /* this first element must match u.yes */
422 struct regmatch_state *prev_yes_state;
423 struct regmatch_state *prev_curlyx; /* previous cur_curlyx */
424 regnode *A, *B; /* the nodes corresponding to /A*B/ */
425 CHECKPOINT cp; /* remember current savestack index */
427 int parenfloor;/* how far back to strip paren data */
428 int min; /* the minimal number of A's to match */
429 int max; /* the maximal number of A's to match */
431 /* these two are modified by WHILEM */
432 int count; /* how many instances of A we've matched */
433 char *lastloc;/* where previous A matched (0-len detect) */
437 /* this first element must match u.yes */
438 struct regmatch_state *prev_yes_state;
439 struct regmatch_state *save_curlyx;
440 CHECKPOINT cp; /* remember current savestack indexes */
442 char *save_lastloc; /* previous curlyx.lastloc */
448 /* this first element must match u.yes */
449 struct regmatch_state *prev_yes_state;
450 I32 c1, c2; /* case fold search */
452 I32 alen; /* length of first-matched A string */
455 regnode *A, *B; /* the nodes corresponding to /A*B/ */
456 regnode *me; /* the curlym node */
462 I32 c1, c2; /* case fold search */
463 char *maxpos; /* highest possible point in string to match */
464 char *oldloc; /* the previous locinput */
466 int min, max; /* {m,n} */
467 regnode *A, *B; /* the nodes corresponding to /A*B/ */
468 } curly; /* and CURLYN/PLUS/STAR */
473 /* how many regmatch_state structs to allocate as a single slab.
474 * We do it in 4K blocks for efficiency. The "3" is 2 for the next/prev
475 * pointers, plus 1 for any mythical malloc overhead. */
477 #define PERL_REGMATCH_SLAB_SLOTS \
478 ((4096 - 3 * sizeof (void*)) / sizeof(regmatch_state))
480 typedef struct regmatch_slab {
481 regmatch_state states[PERL_REGMATCH_SLAB_SLOTS];
482 struct regmatch_slab *prev, *next;
485 #define PL_reg_flags PL_reg_state.re_state_reg_flags
486 #define PL_bostr PL_reg_state.re_state_bostr
487 #define PL_reginput PL_reg_state.re_state_reginput
488 #define PL_regeol PL_reg_state.re_state_regeol
489 #define PL_regoffs PL_reg_state.re_state_regoffs
490 #define PL_reglastparen PL_reg_state.re_state_reglastparen
491 #define PL_reglastcloseparen PL_reg_state.re_state_reglastcloseparen
492 #define PL_reg_start_tmp PL_reg_state.re_state_reg_start_tmp
493 #define PL_reg_start_tmpl PL_reg_state.re_state_reg_start_tmpl
494 #define PL_reg_eval_set PL_reg_state.re_state_reg_eval_set
495 #define PL_reg_match_utf8 PL_reg_state.re_state_reg_match_utf8
496 #define PL_reg_magic PL_reg_state.re_state_reg_magic
497 #define PL_reg_oldpos PL_reg_state.re_state_reg_oldpos
498 #define PL_reg_oldcurpm PL_reg_state.re_state_reg_oldcurpm
499 #define PL_reg_curpm PL_reg_state.re_state_reg_curpm
500 #define PL_reg_oldsaved PL_reg_state.re_state_reg_oldsaved
501 #define PL_reg_oldsavedlen PL_reg_state.re_state_reg_oldsavedlen
502 #define PL_reg_maxiter PL_reg_state.re_state_reg_maxiter
503 #define PL_reg_leftiter PL_reg_state.re_state_reg_leftiter
504 #define PL_reg_poscache PL_reg_state.re_state_reg_poscache
505 #define PL_reg_poscache_size PL_reg_state.re_state_reg_poscache_size
506 #define PL_regsize PL_reg_state.re_state_regsize
507 #define PL_reg_starttry PL_reg_state.re_state_reg_starttry
508 #define PL_nrs PL_reg_state.re_state_nrs
510 struct re_save_state {
511 U32 re_state_reg_flags; /* from regexec.c */
512 U32 re_state_reg_start_tmpl; /* from regexec.c */
513 I32 re_state_reg_eval_set; /* from regexec.c */
514 bool re_state_reg_match_utf8; /* from regexec.c */
515 char *re_state_bostr;
516 char *re_state_reginput; /* String-input pointer. */
517 char *re_state_regeol; /* End of input, for $ check. */
518 regexp_paren_pair *re_state_regoffs; /* Pointer to start/end pairs */
519 U32 *re_state_reglastparen; /* Similarly for lastparen. */
520 U32 *re_state_reglastcloseparen; /* Similarly for lastcloseparen. */
521 char **re_state_reg_start_tmp; /* from regexec.c */
522 MAGIC *re_state_reg_magic; /* from regexec.c */
523 PMOP *re_state_reg_oldcurpm; /* from regexec.c */
524 PMOP *re_state_reg_curpm; /* from regexec.c */
525 char *re_state_reg_oldsaved; /* old saved substr during match */
526 STRLEN re_state_reg_oldsavedlen; /* old length of saved substr during match */
527 STRLEN re_state_reg_poscache_size; /* size of pos cache of WHILEM */
528 I32 re_state_reg_oldpos; /* from regexec.c */
529 I32 re_state_reg_maxiter; /* max wait until caching pos */
530 I32 re_state_reg_leftiter; /* wait until caching pos */
531 U32 re_state_regsize; /* from regexec.c */
532 char *re_state_reg_poscache; /* cache of pos of WHILEM */
533 char *re_state_reg_starttry; /* from regexec.c */
534 #ifdef PERL_OLD_COPY_ON_WRITE
535 SV *re_state_nrs; /* was placeholder: unused since 5.8.0 (5.7.2 patch #12027 for bug ID 20010815.012). Used to save rx->saved_copy */
539 #define SAVESTACK_ALLOC_FOR_RE_SAVE_STATE \
540 (1 + ((sizeof(struct re_save_state) - 1) / sizeof(*PL_savestack)))
544 * c-indentation-style: bsd
546 * indent-tabs-mode: t
549 * ex: set ts=8 sts=4 sw=4 noet: