This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Fix \xa0 matching both [\s] [\S], et.al.
[perl5.git] / regcomp.h
CommitLineData
a0d0e21e 1/* regcomp.h
d6376244 2 *
4bb101f2 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
663f364b 4 * 2000, 2001, 2002, 2003, 2005, 2006, 2007, 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 *
a687059c 9 */
e1d1eefb 10#include "regcharclass.h"
a687059c 11
c277df42
IZ
12typedef OP OP_4tree; /* Will be redefined later. */
13
07be1b83 14
58e23c8d 15/* Convert branch sequences to more efficient trie ops? */
07be1b83 16#define PERL_ENABLE_TRIE_OPTIMISATION 1
58e23c8d 17
486ec47a 18/* Be really aggressive about optimising patterns with trie sequences? */
07be1b83 19#define PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION 1
58e23c8d 20
da7fcca4
YO
21/* Use old style unicode mappings for perl and posix character classes
22 *
23 * NOTE: Enabling this essentially breaks character class matching against unicode
24 * strings, so that POSIX char classes match when they shouldn't, and \d matches
25 * way more than 10 characters, and sometimes a charclass and its complement either
26 * both match or neither match.
27 * NOTE: Disabling this will cause various backwards compatibility issues to rear
28 * their head, and tests to fail. However it will make the charclass behaviour
486ec47a
PAR
29 * consistent regardless of internal string type, and make character class inversions
30 * consistent. The tests that fail in the regex engine are basically broken tests.
da7fcca4
YO
31 *
32 * Personally I think 5.12 should disable this for sure. Its a bit more debatable for
33 * 5.10, so for now im leaving it enabled.
ba9ac175 34 * XXX: It is now enabled for 5.11/5.12
da7fcca4
YO
35 *
36 * -demerphq
37 */
2e84be61 38#define PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS 1
da7fcca4 39
58e23c8d 40/* Should the optimiser take positive assertions into account? */
68ba3a3f 41#define PERL_ENABLE_POSITIVE_ASSERTION_STUDY 0
58e23c8d
YO
42
43/* Not for production use: */
07be1b83 44#define PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS 0
58e23c8d 45
7122b237
YO
46/* Activate offsets code - set to if 1 to enable */
47#ifdef DEBUGGING
48#define RE_TRACK_PATTERN_OFFSETS
49#endif
50
b9b4dddf
YO
51/* Unless the next line is uncommented it is illegal to combine lazy
52 matching with possessive matching. Frankly it doesn't make much sense
53 to allow it as X*?+ matches nothing, X+?+ matches a single char only,
54 and X{min,max}?+ matches min times only.
55 */
56/* #define REG_ALLOW_MINMOD_SUSPEND */
07be1b83 57
a687059c
LW
58/*
59 * The "internal use only" fields in regexp.h are present to pass info from
60 * compile to execute that permits the execute phase to run lots faster on
61 * simple cases. They are:
62 *
bd61b366 63 * regstart sv that must begin a match; NULL if none obvious
a687059c
LW
64 * reganch is the match anchored (at beginning-of-line only)?
65 * regmust string (pointer into program) that match must include, or NULL
79072805 66 * [regmust changed to SV* for bminstr()--law]
a687059c
LW
67 * regmlen length of regmust string
68 * [regmlen not used currently]
69 *
70 * Regstart and reganch permit very fast decisions on suitable starting points
71 * for a match, cutting down the work a lot. Regmust permits fast rejection
72 * of lines that cannot possibly match. The regmust tests are costly enough
e50aee73 73 * that pregcomp() supplies a regmust only if the r.e. contains something
a687059c
LW
74 * potentially expensive (at present, the only such thing detected is * or +
75 * at the start of the r.e., which can involve a lot of backup). Regmlen is
e50aee73 76 * supplied because the test in pregexec() needs it and pregcomp() is computing
a687059c
LW
77 * it anyway.
78 * [regmust is now supplied always. The tests that use regmust have a
79 * heuristic that disables the test if it usually matches.]
80 *
81 * [In fact, we now use regmust in many cases to locate where the search
82 * starts in the string, so if regback is >= 0, the regmust search is never
83 * wasted effort. The regback variable says how many characters back from
84 * where regmust matched is the earliest possible start of the match.
85 * For instance, /[a-z].foo/ has a regmust of 'foo' and a regback of 2.]
86 */
87
88/*
89 * Structure for regexp "program". This is essentially a linear encoding
90 * of a nondeterministic finite-state machine (aka syntax charts or
91 * "railroad normal form" in parsing technology). Each node is an opcode
92 * plus a "next" pointer, possibly plus an operand. "Next" pointers of
93 * all nodes except BRANCH implement concatenation; a "next" pointer with
94 * a BRANCH on both ends of it is connecting two alternatives. (Here we
95 * have one of the subtle syntax dependencies: an individual BRANCH (as
96 * opposed to a collection of them) is never concatenated with anything
97 * because of operator precedence.) The operand of some types of node is
98 * a literal string; for others, it is a node leading into a sub-FSM. In
99 * particular, the operand of a BRANCH node is the first node of the branch.
100 * (NB this is *not* a tree structure: the tail of the branch connects
a3621e74
YO
101 * to the thing following the set of BRANCHes.) The opcodes are defined
102 * in regnodes.h which is generated from regcomp.sym by regcomp.pl.
a687059c
LW
103 */
104
a687059c
LW
105/*
106 * A node is one char of opcode followed by two chars of "next" pointer.
107 * "Next" pointers are stored as two 8-bit pieces, high order first. The
108 * value is a positive offset from the opcode of the node containing it.
109 * An operand, if any, simply follows the node. (Note that much of the
110 * code generation knows about this implicit relationship.)
111 *
112 * Using two bytes for the "next" pointer is vast overkill for most things,
113 * but allows patterns to get big without disasters.
114 *
e91177ed 115 * [The "next" pointer is always aligned on an even
a687059c
LW
116 * boundary, and reads the offset directly as a short. Also, there is no
117 * special test to reverse the sign of BACK pointers since the offset is
118 * stored negative.]
119 */
785a26d5
YO
120
121/* This is the stuff that used to live in regexp.h that was truly
122 private to the engine itself. It now lives here. */
123
28d8d7f4 124
785a26d5
YO
125
126 typedef struct regexp_internal {
1f1031fe 127 int name_list_idx; /* Optional data index of an array of paren names */
7122b237
YO
128 union {
129 U32 *offsets; /* offset annotations 20001228 MJD
130 data about mapping the program to the
131 string -
132 offsets[0] is proglen when this is used
133 */
134 U32 proglen;
135 } u;
136
6d5c990f
RGS
137 regnode *regstclass; /* Optional startclass as identified or constructed
138 by the optimiser */
139 struct reg_data *data; /* Additional miscellaneous data used by the program.
140 Used to make it easier to clone and free arbitrary
141 data that the regops need. Often the ARG field of
142 a regop is an index into this structure */
143 regnode program[1]; /* Unwarranted chumminess with compiler. */
144} regexp_internal;
145
146#define RXi_SET(x,y) (x)->pprivate = (void*)(y)
147#define RXi_GET(x) ((regexp_internal *)((x)->pprivate))
148#define RXi_GET_DECL(r,ri) regexp_internal *ri = RXi_GET(r)
785a26d5
YO
149/*
150 * Flags stored in regexp->intflags
151 * These are used only internally to the regexp engine
152 *
153 * See regexp.h for flags used externally to the regexp engine
154 */
155#define PREGf_SKIP 0x00000001
156#define PREGf_IMPLICIT 0x00000002 /* Converted .* to ^.* */
157#define PREGf_NAUGHTY 0x00000004 /* how exponential is this pattern? */
158#define PREGf_VERBARG_SEEN 0x00000008
159#define PREGf_CUTGROUP_SEEN 0x00000010
160
161
162/* this is where the old regcomp.h started */
a687059c 163
c277df42 164struct regnode_string {
cd439c50 165 U8 str_len;
c277df42
IZ
166 U8 type;
167 U16 next_off;
cd439c50 168 char string[1];
c277df42 169};
a687059c 170
6bda09f9
YO
171/* Argument bearing node - workhorse,
172 arg1 is often for the data field */
c277df42
IZ
173struct regnode_1 {
174 U8 flags;
175 U8 type;
176 U16 next_off;
177 U32 arg1;
178};
179
6bda09f9
YO
180/* Similar to a regnode_1 but with an extra signed argument */
181struct regnode_2L {
182 U8 flags;
183 U8 type;
184 U16 next_off;
185 U32 arg1;
186 I32 arg2;
187};
188
189/* 'Two field' -- Two 16 bit unsigned args */
c277df42
IZ
190struct regnode_2 {
191 U8 flags;
192 U8 type;
193 U16 next_off;
194 U16 arg1;
195 U16 arg2;
196};
197
6bda09f9 198
936ed897 199#define ANYOF_BITMAP_SIZE 32 /* 256 b/(8 b/B) */
d1eb3177 200#define ANYOF_CLASSBITMAP_SIZE 4 /* up to 32 (8*4) named classes */
936ed897 201
786e8c11 202/* also used by trie */
936ed897
IZ
203struct regnode_charclass {
204 U8 flags;
205 U8 type;
206 U16 next_off;
bf3f2d85 207 U32 arg1; /* used as ptr in S_regclass */
19103b91 208 char bitmap[ANYOF_BITMAP_SIZE]; /* only compile-time */
936ed897
IZ
209};
210
bf3f2d85
KW
211/* has runtime (locale) \d, \w, ..., [:posix:] classes */
212struct regnode_charclass_class {
213 U8 flags; /* ANYOF_CLASS bit must go here */
936ed897
IZ
214 U8 type;
215 U16 next_off;
bf3f2d85 216 U32 arg1; /* used as ptr in S_regclass */
19103b91
JH
217 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time */
218 char classflags[ANYOF_CLASSBITMAP_SIZE]; /* and run-time */
936ed897
IZ
219};
220
9c7e81e8
AD
221/* XXX fix this description.
222 Impose a limit of REG_INFTY on various pattern matching operations
223 to limit stack growth and to avoid "infinite" recursions.
224*/
225/* The default size for REG_INFTY is I16_MAX, which is the same as
226 SHORT_MAX (see perl.h). Unfortunately I16 isn't necessarily 16 bits
227 (see handy.h). On the Cray C90, sizeof(short)==4 and hence I16_MAX is
228 ((1<<31)-1), while on the Cray T90, sizeof(short)==8 and I16_MAX is
229 ((1<<63)-1). To limit stack growth to reasonable sizes, supply a
230 smaller default.
231 --Andy Dougherty 11 June 1998
232*/
233#if SHORTSIZE > 2
234# ifndef REG_INFTY
235# define REG_INFTY ((1<<15)-1)
236# endif
237#endif
238
239#ifndef REG_INFTY
83cfe48c
IZ
240# define REG_INFTY I16_MAX
241#endif
a687059c 242
e91177ed
IZ
243#define ARG_VALUE(arg) (arg)
244#define ARG__SET(arg,val) ((arg) = (val))
c277df42 245
e8e3b0fb
JH
246#undef ARG
247#undef ARG1
248#undef ARG2
249
c277df42
IZ
250#define ARG(p) ARG_VALUE(ARG_LOC(p))
251#define ARG1(p) ARG_VALUE(ARG1_LOC(p))
252#define ARG2(p) ARG_VALUE(ARG2_LOC(p))
6bda09f9 253#define ARG2L(p) ARG_VALUE(ARG2L_LOC(p))
786e8c11 254
c277df42
IZ
255#define ARG_SET(p, val) ARG__SET(ARG_LOC(p), (val))
256#define ARG1_SET(p, val) ARG__SET(ARG1_LOC(p), (val))
257#define ARG2_SET(p, val) ARG__SET(ARG2_LOC(p), (val))
6bda09f9 258#define ARG2L_SET(p, val) ARG__SET(ARG2L_LOC(p), (val))
c277df42 259
e8e3b0fb
JH
260#undef NEXT_OFF
261#undef NODE_ALIGN
262
075abff3
AL
263#define NEXT_OFF(p) ((p)->next_off)
264#define NODE_ALIGN(node)
265#define NODE_ALIGN_FILL(node) ((node)->flags = 0xde) /* deadbeef */
a687059c 266
c277df42
IZ
267#define SIZE_ALIGN NODE_ALIGN
268
e8e3b0fb
JH
269#undef OP
270#undef OPERAND
271#undef MASK
272#undef STRING
273
e91177ed 274#define OP(p) ((p)->type)
bf3f2d85 275#define FLAGS(p) ((p)->flags) /* Caution: Doesn't apply to all \
b4eac3bb 276 regnode types */
e91177ed 277#define OPERAND(p) (((struct regnode_string *)p)->string)
cd439c50
IZ
278#define MASK(p) ((char*)OPERAND(p))
279#define STR_LEN(p) (((struct regnode_string *)p)->str_len)
280#define STRING(p) (((struct regnode_string *)p)->string)
281#define STR_SZ(l) ((l + sizeof(regnode) - 1) / sizeof(regnode))
282#define NODE_SZ_STR(p) (STR_SZ(STR_LEN(p))+1)
283
e8e3b0fb
JH
284#undef NODE_ALIGN
285#undef ARG_LOC
286#undef NEXTOPER
287#undef PREVOPER
288
e91177ed
IZ
289#define NODE_ALIGN(node)
290#define ARG_LOC(p) (((struct regnode_1 *)p)->arg1)
291#define ARG1_LOC(p) (((struct regnode_2 *)p)->arg1)
292#define ARG2_LOC(p) (((struct regnode_2 *)p)->arg2)
6bda09f9 293#define ARG2L_LOC(p) (((struct regnode_2L *)p)->arg2)
786e8c11 294
e91177ed
IZ
295#define NODE_STEP_REGNODE 1 /* sizeof(regnode)/sizeof(regnode) */
296#define EXTRA_STEP_2ARGS EXTRA_SIZE(struct regnode_2)
297
298#define NODE_STEP_B 4
c277df42
IZ
299
300#define NEXTOPER(p) ((p) + NODE_STEP_REGNODE)
301#define PREVOPER(p) ((p) - NODE_STEP_REGNODE)
302
e91177ed 303#define FILL_ADVANCE_NODE(ptr, op) STMT_START { \
c277df42 304 (ptr)->type = op; (ptr)->next_off = 0; (ptr)++; } STMT_END
e91177ed 305#define FILL_ADVANCE_NODE_ARG(ptr, op, arg) STMT_START { \
c277df42 306 ARG_SET(ptr, arg); FILL_ADVANCE_NODE(ptr, op); (ptr) += 1; } STMT_END
a687059c 307
22c35a8c 308#define REG_MAGIC 0234
a687059c 309
830247a4 310#define SIZE_ONLY (RExC_emit == &PL_regdummy)
c277df42 311
a12cf05f
KW
312/* Flags for node->flags of several of the node types */
313#define USE_UNI 0x01
314
d5788240 315/* Flags for node->flags of ANYOF. These are in short supply, so some games
f4244008
KW
316 * are done to share them, as described below. If necessary, the ANYOF_LOCALE
317 * and ANYOF_CLASS bits could be shared with a space penalty for locale nodes
318 * (and the code at the time this comment was written, is written so that all
319 * that is necessary to make the change would be to redefine the ANYOF_CLASS
320 * define). Once the planned change to compile all the above-latin1 code points
321 * is done, then the UNICODE_ALL bit can be freed up. If flags need to be
322 * added that are applicable to the synthetic start class only, with some work,
323 * they could be put in the next-node field, or in an unused bit of the
324 * classflags field. */
b8c5462f 325
39065660
KW
326#define ANYOF_LOCALE 0x01
327
328/* The fold is calculated and stored in the bitmap where possible at compile
329 * time. However there are two cases where it isn't possible. These share
330 * this bit: 1) under locale, where the actual folding varies depending on
331 * what the locale is at the time of execution; and 2) where the folding is
332 * specified in a swash, not the bitmap, such as characters which aren't
333 * specified in the bitmap, or properties that aren't looked at at compile time
334 */
335#define ANYOF_LOC_NONBITMAP_FOLD 0x02
336
337#define ANYOF_INVERT 0x04
1984f6a0 338
f4244008
KW
339/* EOS, meaning that it can match an empty string too, is used for the
340 * synthetic start class (ssc) only. It can share the INVERT bit, as the ssc
341 * is never inverted. The bit just needs to be turned off before regexec.c
342 * gets a hold of it so that regexec.c doesn't think it's inverted, but this
343 * happens automatically, as if the ssc can match an EOS, the ssc is discarded,
344 * and never passed to regexec.c */
345#define ANYOF_EOS ANYOF_INVERT
346
d5788240
KW
347/* CLASS is never set unless LOCALE is too: has runtime \d, \w, [:posix:], ...
348 * The non-locale ones are resolved at compile-time */
1eb41b0a 349#define ANYOF_CLASS 0x08
1984f6a0 350#define ANYOF_LARGE ANYOF_CLASS /* Same; name retained for back compat */
b8c5462f 351
ef87b810
KW
352/* Can match something outside the bitmap that is expressible only in utf8 */
353#define ANYOF_UTF8 0x10
354
355/* Can match something outside the bitmap that isn't in utf8 */
356#define ANYOF_NONBITMAP_NON_UTF8 0x20
357
3ff7ceb3 358/* Set if the bitmap doesn't fully represent what this node can match */
ef87b810 359#define ANYOF_NONBITMAP (ANYOF_UTF8|ANYOF_NONBITMAP_NON_UTF8)
3ff7ceb3
KW
360#define ANYOF_UNICODE ANYOF_NONBITMAP /* old name, for back compat */
361
61ef9872
KW
362/* Matches every code point 0x100 and above*/
363#define ANYOF_UNICODE_ALL 0x40
ffc61ed2 364
11454c59
KW
365/* Match all Latin1 characters that aren't ASCII when the target string is not
366 * in utf8. */
367#define ANYOF_NON_UTF8_LATIN1_ALL 0x80
368
ffc61ed2 369#define ANYOF_FLAGS_ALL 0xff
653099ff 370
936ed897 371/* Character classes for node->classflags of ANYOF */
653099ff
GS
372/* Should be synchronized with a table in regprop() */
373/* 2n should pair with 2n+1 */
b8c5462f 374
289d4f09 375#define ANYOF_ALNUM 0 /* \w, PL_utf8_alnum, utf8::IsWord, ALNUM */
b8c5462f 376#define ANYOF_NALNUM 1
aaa51d5e 377#define ANYOF_SPACE 2 /* \s */
b8c5462f 378#define ANYOF_NSPACE 3
24d786f4 379#define ANYOF_DIGIT 4 /* \d */
b8c5462f 380#define ANYOF_NDIGIT 5
24d786f4 381#define ANYOF_ALNUMC 6 /* [[:alnum:]] isalnum(3), utf8::IsAlnum, ALNUMC */
b8c5462f
JH
382#define ANYOF_NALNUMC 7
383#define ANYOF_ALPHA 8
384#define ANYOF_NALPHA 9
385#define ANYOF_ASCII 10
386#define ANYOF_NASCII 11
387#define ANYOF_CNTRL 12
388#define ANYOF_NCNTRL 13
389#define ANYOF_GRAPH 14
390#define ANYOF_NGRAPH 15
391#define ANYOF_LOWER 16
392#define ANYOF_NLOWER 17
393#define ANYOF_PRINT 18
394#define ANYOF_NPRINT 19
395#define ANYOF_PUNCT 20
396#define ANYOF_NPUNCT 21
397#define ANYOF_UPPER 22
398#define ANYOF_NUPPER 23
399#define ANYOF_XDIGIT 24
400#define ANYOF_NXDIGIT 25
aaa51d5e
JF
401#define ANYOF_PSXSPC 26 /* POSIX space: \s plus the vertical tab */
402#define ANYOF_NPSXSPC 27
ffc61ed2 403#define ANYOF_BLANK 28 /* GNU extension: space and tab: non-vertical space */
289d4f09 404#define ANYOF_NBLANK 29
b8c5462f 405
aaa51d5e 406#define ANYOF_MAX 32
b8c5462f 407
e1d1eefb
YO
408/* pseudo classes, not stored in the class bitmap, but used as flags
409 during compilation of char classes */
410
411#define ANYOF_VERTWS (ANYOF_MAX+1)
412#define ANYOF_NVERTWS (ANYOF_MAX+2)
413#define ANYOF_HORIZWS (ANYOF_MAX+3)
414#define ANYOF_NHORIZWS (ANYOF_MAX+4)
415
b8c5462f
JH
416/* Backward source code compatibility. */
417
418#define ANYOF_ALNUML ANYOF_ALNUM
419#define ANYOF_NALNUML ANYOF_NALNUM
420#define ANYOF_SPACEL ANYOF_SPACE
421#define ANYOF_NSPACEL ANYOF_NSPACE
422
423/* Utility macros for the bitmap and classes of ANYOF */
424
936ed897
IZ
425#define ANYOF_SIZE (sizeof(struct regnode_charclass))
426#define ANYOF_CLASS_SIZE (sizeof(struct regnode_charclass_class))
b8c5462f 427
936ed897 428#define ANYOF_FLAGS(p) ((p)->flags)
b8c5462f
JH
429
430#define ANYOF_BIT(c) (1 << ((c) & 7))
431
936ed897 432#define ANYOF_CLASS_BYTE(p, c) (((struct regnode_charclass_class*)(p))->classflags[((c) >> 3) & 3])
b8c5462f
JH
433#define ANYOF_CLASS_SET(p, c) (ANYOF_CLASS_BYTE(p, c) |= ANYOF_BIT(c))
434#define ANYOF_CLASS_CLEAR(p, c) (ANYOF_CLASS_BYTE(p, c) &= ~ANYOF_BIT(c))
435#define ANYOF_CLASS_TEST(p, c) (ANYOF_CLASS_BYTE(p, c) & ANYOF_BIT(c))
436
936ed897
IZ
437#define ANYOF_CLASS_ZERO(ret) Zero(((struct regnode_charclass_class*)(ret))->classflags, ANYOF_CLASSBITMAP_SIZE, char)
438#define ANYOF_BITMAP_ZERO(ret) Zero(((struct regnode_charclass*)(ret))->bitmap, ANYOF_BITMAP_SIZE, char)
439
440#define ANYOF_BITMAP(p) (((struct regnode_charclass*)(p))->bitmap)
786e8c11 441#define ANYOF_BITMAP_BYTE(p, c) (ANYOF_BITMAP(p)[(((U8)(c)) >> 3) & 31])
b8c5462f
JH
442#define ANYOF_BITMAP_SET(p, c) (ANYOF_BITMAP_BYTE(p, c) |= ANYOF_BIT(c))
443#define ANYOF_BITMAP_CLEAR(p,c) (ANYOF_BITMAP_BYTE(p, c) &= ~ANYOF_BIT(c))
444#define ANYOF_BITMAP_TEST(p, c) (ANYOF_BITMAP_BYTE(p, c) & ANYOF_BIT(c))
445
f8bef550
NC
446#define ANYOF_BITMAP_SETALL(p) \
447 memset (ANYOF_BITMAP(p), 255, ANYOF_BITMAP_SIZE)
448#define ANYOF_BITMAP_CLEARALL(p) \
449 Zero (ANYOF_BITMAP(p), ANYOF_BITMAP_SIZE)
450/* Check that all 256 bits are all set. Used in S_cl_is_anything() */
c355c09a 451#define ANYOF_BITMAP_TESTALLSET(p) /* Assumes sizeof(p) == 32 */ \
f8bef550
NC
452 memEQ (ANYOF_BITMAP(p), "\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377\377", ANYOF_BITMAP_SIZE)
453
936ed897
IZ
454#define ANYOF_SKIP ((ANYOF_SIZE - 1)/sizeof(regnode))
455#define ANYOF_CLASS_SKIP ((ANYOF_CLASS_SIZE - 1)/sizeof(regnode))
3a15e693
KW
456
457/* The class bit can be set to the locale one if necessary to save bits at the
458 * expense of having locale ANYOF nodes always have a class bit map, and hence
459 * take up extra space. This allows convenient changing it as development
460 * proceeds on this */
461#if ANYOF_CLASS == ANYOF_LOCALE
462# undef ANYOF_CLASS_ADD_SKIP
463# define ANYOF_ADD_LOC_SKIP (ANYOF_CLASS_SKIP - ANYOF_SKIP)
464
465 /* Quicker way to see if there are actually any tests. This is because
466 * currently the set of tests can be empty even when the class bitmap is
467 * allocated */
468# if ANYOF_CLASSBITMAP_SIZE != 4
469# error ANYOF_CLASSBITMAP_SIZE is expected to be 4
470# endif
471# define ANYOF_CLASS_TEST_ANY_SET(p) /* assumes sizeof(p) = 4 */ \
472 memNE (((struct regnode_charclass_class*)(p))->classflags, \
473 "\0\0\0\0", ANYOF_CLASSBITMAP_SIZE)
474#else
475# define ANYOF_CLASS_ADD_SKIP (ANYOF_CLASS_SKIP - ANYOF_SKIP)
476# undef ANYOF_ADD_LOC_SKIP
477# define ANYOF_CLASS_TEST_ANY_SET(p) (ANYOF_FLAGS(p) & ANYOF_CLASS)
478#endif
c277df42 479
786e8c11 480
a687059c
LW
481/*
482 * Utility definitions.
483 */
2304df62 484#ifndef CHARMASK
075abff3 485# define UCHARAT(p) ((int)*(const U8*)(p))
a687059c 486#else
075abff3 487# define UCHARAT(p) ((int)*(p)&CHARMASK)
a687059c 488#endif
a687059c 489
c277df42
IZ
490#define EXTRA_SIZE(guy) ((sizeof(guy)-1)/sizeof(struct regnode))
491
3dab1dad
YO
492#define REG_SEEN_ZERO_LEN 0x00000001
493#define REG_SEEN_LOOKBEHIND 0x00000002
494#define REG_SEEN_GPOS 0x00000004
495#define REG_SEEN_EVAL 0x00000008
496#define REG_SEEN_CANY 0x00000010
cce850e4 497#define REG_SEEN_SANY REG_SEEN_CANY /* src bckwrd cmpt */
6bda09f9 498#define REG_SEEN_RECURSE 0x00000020
eaf3ca90 499#define REG_TOP_LEVEL_BRANCHES 0x00000040
e2e6a0f1 500#define REG_SEEN_VERBARG 0x00000080
5d458dd8 501#define REG_SEEN_CUTGROUP 0x00000100
bcdf7404 502#define REG_SEEN_RUN_ON_COMMENT 0x00000200
d09b2d29 503
73c4f7a1
GS
504START_EXTERN_C
505
be8e71aa
YO
506#ifdef PLUGGABLE_RE_EXTENSION
507#include "re_nodes.h"
508#else
d09b2d29 509#include "regnodes.h"
be8e71aa 510#endif
d09b2d29 511
f9f4320a
YO
512#ifndef PLUGGABLE_RE_EXTENSION
513#ifndef DOINIT
514EXTCONST regexp_engine PL_core_reg_engine;
515#else /* DOINIT */
516EXTCONST regexp_engine PL_core_reg_engine = {
fe578d7f 517 Perl_re_compile,
2fdbfb4d 518 Perl_regexec_flags,
f9f4320a
YO
519 Perl_re_intuit_start,
520 Perl_re_intuit_string,
2fdbfb4d
AB
521 Perl_regfree_internal,
522 Perl_reg_numbered_buff_fetch,
523 Perl_reg_numbered_buff_store,
524 Perl_reg_numbered_buff_length,
192b9cd1
AB
525 Perl_reg_named_buff,
526 Perl_reg_named_buff_iter,
49d7dfbc 527 Perl_reg_qr_package,
f9f4320a 528#if defined(USE_ITHREADS)
f8149455 529 Perl_regdupe_internal
f9f4320a
YO
530#endif
531};
532#endif /* DOINIT */
533#endif /* PLUGGABLE_RE_EXTENSION */
534
535
73c4f7a1 536END_EXTERN_C
cad2e5aa 537
cad2e5aa 538
261faec3
MJD
539/* .what is a character array with one character for each member of .data
540 * The character describes the function of the corresponding .data item:
af534a04 541 * a - AV for paren_name_list under DEBUGGING
0111c4fd 542 * f - start-class data for regstclass optimization
261faec3
MJD
543 * n - Root of op tree for (?{EVAL}) item
544 * o - Start op for (?{EVAL}) item
3dab1dad 545 * p - Pad for (?{EVAL}) item
38a44b82 546 * s - swash for Unicode-style character class, and the multicharacter
9e55ce06
JH
547 * strings resulting from casefolding the single-character entries
548 * in the character class
a3621e74 549 * t - trie struct
55eed653 550 * u - trie struct's widecharmap (a HV, so can't share, must dup)
2b8b4781 551 * also used for revcharmap and words under DEBUGGING
07be1b83 552 * T - aho-trie struct
81714fb9 553 * S - sv for named capture lookup
261faec3 554 * 20010712 mjd@plover.com
c3613ed7 555 * (Remember to update re_dup() and pregfree() if you add any items.)
261faec3 556 */
cad2e5aa
JH
557struct reg_data {
558 U32 count;
559 U8 *what;
560 void* data[1];
561};
562
a1cac82e
NC
563/* Code in S_to_utf8_substr() and S_to_byte_substr() in regexec.c accesses
564 anchored* and float* via array indexes 0 and 1. */
cad2e5aa 565#define anchored_substr substrs->data[0].substr
33b8afdf 566#define anchored_utf8 substrs->data[0].utf8_substr
cad2e5aa 567#define anchored_offset substrs->data[0].min_offset
1de06328
YO
568#define anchored_end_shift substrs->data[0].end_shift
569
cad2e5aa 570#define float_substr substrs->data[1].substr
33b8afdf 571#define float_utf8 substrs->data[1].utf8_substr
cad2e5aa
JH
572#define float_min_offset substrs->data[1].min_offset
573#define float_max_offset substrs->data[1].max_offset
1de06328
YO
574#define float_end_shift substrs->data[1].end_shift
575
cad2e5aa 576#define check_substr substrs->data[2].substr
33b8afdf 577#define check_utf8 substrs->data[2].utf8_substr
cad2e5aa
JH
578#define check_offset_min substrs->data[2].min_offset
579#define check_offset_max substrs->data[2].max_offset
1de06328 580#define check_end_shift substrs->data[2].end_shift
a3621e74 581
288b8c02
NC
582#define RX_ANCHORED_SUBSTR(rx) (((struct regexp *)SvANY(rx))->anchored_substr)
583#define RX_ANCHORED_UTF8(rx) (((struct regexp *)SvANY(rx))->anchored_utf8)
584#define RX_FLOAT_SUBSTR(rx) (((struct regexp *)SvANY(rx))->float_substr)
585#define RX_FLOAT_UTF8(rx) (((struct regexp *)SvANY(rx))->float_utf8)
a3621e74
YO
586
587/* trie related stuff */
5d9a96ca 588
a3621e74
YO
589/* a transition record for the state machine. the
590 check field determines which state "owns" the
591 transition. the char the transition is for is
592 determined by offset from the owning states base
593 field. the next field determines which state
594 is to be transitioned to if any.
595*/
596struct _reg_trie_trans {
597 U32 next;
598 U32 check;
599};
600
601/* a transition list element for the list based representation */
602struct _reg_trie_trans_list_elem {
603 U16 forid;
604 U32 newstate;
605};
606typedef struct _reg_trie_trans_list_elem reg_trie_trans_le;
607
608/* a state for compressed nodes. base is an offset
609 into an array of reg_trie_trans array. If wordnum is
610 nonzero the state is accepting. if base is zero then
611 the state has no children (and will be accepting)
612*/
613struct _reg_trie_state {
614 U16 wordnum;
615 union {
616 U32 base;
617 reg_trie_trans_le* list;
618 } trans;
619};
620
2e64971a
DM
621/* info per word; indexed by wordnum */
622typedef struct {
623 U16 prev; /* previous word in acceptance chain; eg in
624 * zzz|abc|ab/ after matching the chars abc, the
625 * accepted word is #2, and the previous accepted
626 * word is #3 */
627 U32 len; /* how many chars long is this word? */
628 U32 accept; /* accept state for this word */
629} reg_trie_wordinfo;
a3621e74
YO
630
631
a3621e74
YO
632typedef struct _reg_trie_state reg_trie_state;
633typedef struct _reg_trie_trans reg_trie_trans;
634
635
636/* anything in here that needs to be freed later
23eab42c
NC
637 should be dealt with in pregfree.
638 refcount is first in both this and _reg_ac_data to allow a space
639 optimisation in Perl_regdupe. */
a3621e74 640struct _reg_trie_data {
23eab42c 641 U32 refcount; /* number of times this trie is referenced */
786e8c11
YO
642 U32 lasttrans; /* last valid transition element */
643 U16 *charmap; /* byte to charid lookup array */
786e8c11
YO
644 reg_trie_state *states; /* state data */
645 reg_trie_trans *trans; /* array of transition elements */
646 char *bitmap; /* stclass bitmap */
786e8c11
YO
647 U16 *jump; /* optional 1 indexed array of offsets before tail
648 for the node following a given word. */
2e64971a 649 reg_trie_wordinfo *wordinfo; /* array of info per word */
1b5c067c
NC
650 U16 uniquecharcount; /* unique chars in trie (width of trans table) */
651 U32 startstate; /* initial state - used for common prefix optimisation */
652 STRLEN minlen; /* minimum length of words in trie - build/opt only? */
653 STRLEN maxlen; /* maximum length of words in trie - build/opt only? */
2e64971a 654 U32 prefixlen; /* #chars in common prefix */
1e2e3d02
YO
655 U32 statecount; /* Build only - number of states in the states array
656 (including the unused zero state) */
786e8c11 657 U32 wordcount; /* Build only */
a3621e74 658#ifdef DEBUGGING
786e8c11 659 STRLEN charcount; /* Build only */
a3621e74
YO
660#endif
661};
2b8b4781
NC
662/* There is one (3 under DEBUGGING) pointers that logically belong in this
663 structure, but are held outside as they need duplication on thread cloning,
664 whereas the rest of the structure can be read only:
665 HV *widecharmap; code points > 255 to charid
666#ifdef DEBUGGING
667 AV *words; Array of words contained in trie, for dumping
668 AV *revcharmap; Map of each charid back to its character representation
669#endif
670*/
671
672#define TRIE_WORDS_OFFSET 2
673
a3621e74
YO
674typedef struct _reg_trie_data reg_trie_data;
675
23eab42c
NC
676/* refcount is first in both this and _reg_trie_data to allow a space
677 optimisation in Perl_regdupe. */
07be1b83 678struct _reg_ac_data {
23eab42c 679 U32 refcount;
1b5c067c 680 U32 trie;
07be1b83
YO
681 U32 *fail;
682 reg_trie_state *states;
07be1b83
YO
683};
684typedef struct _reg_ac_data reg_ac_data;
685
486ec47a 686/* ANY_BIT doesn't use the structure, so we can borrow it here.
3dab1dad
YO
687 This is simpler than refactoring all of it as wed end up with
688 three different sets... */
689
690#define TRIE_BITMAP(p) (((reg_trie_data *)(p))->bitmap)
786e8c11 691#define TRIE_BITMAP_BYTE(p, c) (TRIE_BITMAP(p)[(((U8)(c)) >> 3) & 31])
5dfe062f
YO
692#define TRIE_BITMAP_SET(p, c) (TRIE_BITMAP_BYTE(p, c) |= ANYOF_BIT((U8)c))
693#define TRIE_BITMAP_CLEAR(p,c) (TRIE_BITMAP_BYTE(p, c) &= ~ANYOF_BIT((U8)c))
694#define TRIE_BITMAP_TEST(p, c) (TRIE_BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
3dab1dad 695
1de06328
YO
696#define IS_ANYOF_TRIE(op) ((op)==TRIEC || (op)==AHOCORASICKC)
697#define IS_TRIE_AC(op) ((op)>=AHOCORASICK)
698
699
786e8c11
YO
700#define BITMAP_BYTE(p, c) (((U8*)p)[(((U8)(c)) >> 3) & 31])
701#define BITMAP_TEST(p, c) (BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
3dab1dad 702
a3621e74
YO
703/* these defines assume uniquecharcount is the correct variable, and state may be evaluated twice */
704#define TRIE_NODENUM(state) (((state)-1)/(trie->uniquecharcount)+1)
705#define SAFE_TRIE_NODENUM(state) ((state) ? (((state)-1)/(trie->uniquecharcount)+1) : (state))
706#define TRIE_NODEIDX(state) ((state) ? (((state)-1)*(trie->uniquecharcount)+1) : (state))
707
3dab1dad 708#ifdef DEBUGGING
3dab1dad 709#define TRIE_CHARCOUNT(trie) ((trie)->charcount)
3dab1dad 710#else
3dab1dad 711#define TRIE_CHARCOUNT(trie) (trie_charcount)
3dab1dad 712#endif
a3621e74 713
0111c4fd
RGS
714#define RE_TRIE_MAXBUF_INIT 65536
715#define RE_TRIE_MAXBUF_NAME "\022E_TRIE_MAXBUF"
a3621e74
YO
716#define RE_DEBUG_FLAGS "\022E_DEBUG_FLAGS"
717
be8e71aa
YO
718/*
719
720RE_DEBUG_FLAGS is used to control what debug output is emitted
721its divided into three groups of options, some of which interact.
722The three groups are: Compile, Execute, Extra. There is room for a
723further group, as currently only the low three bytes are used.
724
725 Compile Options:
07be1b83 726
be8e71aa
YO
727 PARSE
728 PEEP
729 TRIE
730 PROGRAM
731 OFFSETS
a3621e74 732
be8e71aa 733 Execute Options:
a3621e74 734
be8e71aa
YO
735 INTUIT
736 MATCH
737 TRIE
a3621e74 738
be8e71aa 739 Extra Options
a3621e74 740
be8e71aa
YO
741 TRIE
742 OFFSETS
743
744If you modify any of these make sure you make corresponding changes to
745re.pm, especially to the documentation.
a3621e74 746
be8e71aa
YO
747*/
748
749
750/* Compile */
751#define RE_DEBUG_COMPILE_MASK 0x0000FF
752#define RE_DEBUG_COMPILE_PARSE 0x000001
753#define RE_DEBUG_COMPILE_OPTIMISE 0x000002
754#define RE_DEBUG_COMPILE_TRIE 0x000004
755#define RE_DEBUG_COMPILE_DUMP 0x000008
f7819f85 756#define RE_DEBUG_COMPILE_FLAGS 0x000010
be8e71aa
YO
757
758/* Execute */
759#define RE_DEBUG_EXECUTE_MASK 0x00FF00
760#define RE_DEBUG_EXECUTE_INTUIT 0x000100
761#define RE_DEBUG_EXECUTE_MATCH 0x000200
762#define RE_DEBUG_EXECUTE_TRIE 0x000400
763
764/* Extra */
765#define RE_DEBUG_EXTRA_MASK 0xFF0000
766#define RE_DEBUG_EXTRA_TRIE 0x010000
767#define RE_DEBUG_EXTRA_OFFSETS 0x020000
a5ca303d
YO
768#define RE_DEBUG_EXTRA_OFFDEBUG 0x040000
769#define RE_DEBUG_EXTRA_STATE 0x080000
770#define RE_DEBUG_EXTRA_OPTIMISE 0x100000
e7707071 771#define RE_DEBUG_EXTRA_BUFFERS 0x400000
2c296965 772#define RE_DEBUG_EXTRA_GPOS 0x800000
24b23f37
YO
773/* combined */
774#define RE_DEBUG_EXTRA_STACK 0x280000
be8e71aa 775
e68ec53f 776#define RE_DEBUG_FLAG(x) (re_debug_flags & x)
be8e71aa
YO
777/* Compile */
778#define DEBUG_COMPILE_r(x) DEBUG_r( \
e68ec53f 779 if (re_debug_flags & RE_DEBUG_COMPILE_MASK) x )
be8e71aa 780#define DEBUG_PARSE_r(x) DEBUG_r( \
e68ec53f 781 if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x )
be8e71aa 782#define DEBUG_OPTIMISE_r(x) DEBUG_r( \
e68ec53f 783 if (re_debug_flags & RE_DEBUG_COMPILE_OPTIMISE) x )
be8e71aa 784#define DEBUG_PARSE_r(x) DEBUG_r( \
e68ec53f 785 if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x )
be8e71aa 786#define DEBUG_DUMP_r(x) DEBUG_r( \
e68ec53f 787 if (re_debug_flags & RE_DEBUG_COMPILE_DUMP) x )
be8e71aa 788#define DEBUG_TRIE_COMPILE_r(x) DEBUG_r( \
e68ec53f 789 if (re_debug_flags & RE_DEBUG_COMPILE_TRIE) x )
f7819f85
A
790#define DEBUG_FLAGS_r(x) DEBUG_r( \
791 if (re_debug_flags & RE_DEBUG_COMPILE_FLAGS) x )
be8e71aa
YO
792/* Execute */
793#define DEBUG_EXECUTE_r(x) DEBUG_r( \
e68ec53f 794 if (re_debug_flags & RE_DEBUG_EXECUTE_MASK) x )
be8e71aa 795#define DEBUG_INTUIT_r(x) DEBUG_r( \
e68ec53f 796 if (re_debug_flags & RE_DEBUG_EXECUTE_INTUIT) x )
be8e71aa 797#define DEBUG_MATCH_r(x) DEBUG_r( \
e68ec53f 798 if (re_debug_flags & RE_DEBUG_EXECUTE_MATCH) x )
be8e71aa 799#define DEBUG_TRIE_EXECUTE_r(x) DEBUG_r( \
e68ec53f 800 if (re_debug_flags & RE_DEBUG_EXECUTE_TRIE) x )
be8e71aa
YO
801
802/* Extra */
803#define DEBUG_EXTRA_r(x) DEBUG_r( \
e68ec53f 804 if (re_debug_flags & RE_DEBUG_EXTRA_MASK) x )
a5ca303d
YO
805#define DEBUG_OFFSETS_r(x) DEBUG_r( \
806 if (re_debug_flags & RE_DEBUG_EXTRA_OFFSETS) x )
ab3bbdeb
YO
807#define DEBUG_STATE_r(x) DEBUG_r( \
808 if (re_debug_flags & RE_DEBUG_EXTRA_STATE) x )
24b23f37
YO
809#define DEBUG_STACK_r(x) DEBUG_r( \
810 if (re_debug_flags & RE_DEBUG_EXTRA_STACK) x )
e7707071
YO
811#define DEBUG_BUFFERS_r(x) DEBUG_r( \
812 if (re_debug_flags & RE_DEBUG_EXTRA_BUFFERS) x )
813
a5ca303d
YO
814#define DEBUG_OPTIMISE_MORE_r(x) DEBUG_r( \
815 if ((RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE) == \
816 (re_debug_flags & (RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE)) ) x )
07be1b83 817#define MJD_OFFSET_DEBUG(x) DEBUG_r( \
a5ca303d 818 if (re_debug_flags & RE_DEBUG_EXTRA_OFFDEBUG) \
be8e71aa
YO
819 Perl_warn_nocontext x )
820#define DEBUG_TRIE_COMPILE_MORE_r(x) DEBUG_TRIE_COMPILE_r( \
e68ec53f 821 if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x )
be8e71aa 822#define DEBUG_TRIE_EXECUTE_MORE_r(x) DEBUG_TRIE_EXECUTE_r( \
e68ec53f 823 if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x )
be8e71aa
YO
824
825#define DEBUG_TRIE_r(x) DEBUG_r( \
e68ec53f 826 if (re_debug_flags & (RE_DEBUG_COMPILE_TRIE \
be8e71aa 827 | RE_DEBUG_EXECUTE_TRIE )) x )
2c296965
YO
828#define DEBUG_GPOS_r(x) DEBUG_r( \
829 if (re_debug_flags & RE_DEBUG_EXTRA_GPOS) x )
be8e71aa
YO
830
831/* initialization */
595598ee
KW
832/* get_sv() can return NULL during global destruction. re_debug_flags can get
833 * clobbered by a longjmp, so must be initialized */
e68ec53f
YO
834#define GET_RE_DEBUG_FLAGS DEBUG_r({ \
835 SV * re_debug_flags_sv = NULL; \
595598ee 836 re_debug_flags = 0; \
e68ec53f
YO
837 re_debug_flags_sv = get_sv(RE_DEBUG_FLAGS, 1); \
838 if (re_debug_flags_sv) { \
8e9a8a48 839 if (!SvIOK(re_debug_flags_sv)) \
e68ec53f 840 sv_setuv(re_debug_flags_sv, RE_DEBUG_COMPILE_DUMP | RE_DEBUG_EXECUTE_MASK ); \
8e9a8a48 841 re_debug_flags=SvIV(re_debug_flags_sv); \
e68ec53f
YO
842 }\
843})
a3621e74
YO
844
845#ifdef DEBUGGING
ab3bbdeb 846
d53d27f9 847#define GET_RE_DEBUG_FLAGS_DECL VOL IV re_debug_flags = 0; GET_RE_DEBUG_FLAGS;
ab3bbdeb
YO
848
849#define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2) \
850 const char * const rpv = \
ddc5bc0f 851 pv_pretty((dsv), (pv), (l), (m), \
ab3bbdeb 852 PL_colors[(c1)],PL_colors[(c2)], \
c89df6cf 853 PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) ); \
0df25f3d 854 const int rlen = SvCUR(dsv)
ab3bbdeb
YO
855
856#define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m) \
857 const char * const rpv = \
ddc5bc0f 858 pv_pretty((dsv), (SvPV_nolen_const(sv)), (SvCUR(sv)), (m), \
ab3bbdeb 859 PL_colors[(c1)],PL_colors[(c2)], \
c89df6cf 860 PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) )
ab3bbdeb
YO
861
862#define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m) \
863 const char * const rpv = \
ddc5bc0f 864 pv_pretty((dsv), (pv), (l), (m), \
ab3bbdeb 865 PL_colors[0], PL_colors[1], \
c89df6cf 866 ( PERL_PV_PRETTY_QUOTE | PERL_PV_ESCAPE_RE | PERL_PV_ESCAPE_NONASCII | PERL_PV_PRETTY_ELLIPSES | \
ab3bbdeb 867 ((isuni) ? PERL_PV_ESCAPE_UNI : 0)) \
0a0e9afc 868 )
ab3bbdeb
YO
869
870#define RE_SV_DUMPLEN(ItEm) (SvCUR(ItEm) - (SvTAIL(ItEm)!=0))
871#define RE_SV_TAIL(ItEm) (SvTAIL(ItEm) ? "$" : "")
872
873#else /* if not DEBUGGING */
874
a3621e74 875#define GET_RE_DEBUG_FLAGS_DECL
ab3bbdeb
YO
876#define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2)
877#define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m)
878#define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m)
879#define RE_SV_DUMPLEN(ItEm)
880#define RE_SV_TAIL(ItEm)
881
882#endif /* DEBUG RELATED DEFINES */
a3621e74 883
e9a8c099
MHM
884/*
885 * Local variables:
886 * c-indentation-style: bsd
887 * c-basic-offset: 4
888 * indent-tabs-mode: t
889 * End:
890 *
891 * ex: set ts=8 sts=4 sw=4 noet:
892 */