This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
regcomp.h: Remove extraneous comment
[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
58e23c8d 12/* Convert branch sequences to more efficient trie ops? */
07be1b83 13#define PERL_ENABLE_TRIE_OPTIMISATION 1
58e23c8d 14
486ec47a 15/* Be really aggressive about optimising patterns with trie sequences? */
07be1b83 16#define PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION 1
58e23c8d
YO
17
18/* Should the optimiser take positive assertions into account? */
68ba3a3f 19#define PERL_ENABLE_POSITIVE_ASSERTION_STUDY 0
58e23c8d
YO
20
21/* Not for production use: */
07be1b83 22#define PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS 0
58e23c8d 23
7122b237
YO
24/* Activate offsets code - set to if 1 to enable */
25#ifdef DEBUGGING
26#define RE_TRACK_PATTERN_OFFSETS
27#endif
28
a687059c
LW
29/*
30 * The "internal use only" fields in regexp.h are present to pass info from
31 * compile to execute that permits the execute phase to run lots faster on
32 * simple cases. They are:
33 *
bd61b366 34 * regstart sv that must begin a match; NULL if none obvious
a687059c
LW
35 * reganch is the match anchored (at beginning-of-line only)?
36 * regmust string (pointer into program) that match must include, or NULL
79072805 37 * [regmust changed to SV* for bminstr()--law]
a687059c
LW
38 * regmlen length of regmust string
39 * [regmlen not used currently]
40 *
41 * Regstart and reganch permit very fast decisions on suitable starting points
42 * for a match, cutting down the work a lot. Regmust permits fast rejection
43 * of lines that cannot possibly match. The regmust tests are costly enough
e50aee73 44 * that pregcomp() supplies a regmust only if the r.e. contains something
a687059c
LW
45 * potentially expensive (at present, the only such thing detected is * or +
46 * at the start of the r.e., which can involve a lot of backup). Regmlen is
e50aee73 47 * supplied because the test in pregexec() needs it and pregcomp() is computing
a687059c
LW
48 * it anyway.
49 * [regmust is now supplied always. The tests that use regmust have a
50 * heuristic that disables the test if it usually matches.]
51 *
52 * [In fact, we now use regmust in many cases to locate where the search
53 * starts in the string, so if regback is >= 0, the regmust search is never
54 * wasted effort. The regback variable says how many characters back from
55 * where regmust matched is the earliest possible start of the match.
56 * For instance, /[a-z].foo/ has a regmust of 'foo' and a regback of 2.]
57 */
58
59/*
60 * Structure for regexp "program". This is essentially a linear encoding
61 * of a nondeterministic finite-state machine (aka syntax charts or
62 * "railroad normal form" in parsing technology). Each node is an opcode
63 * plus a "next" pointer, possibly plus an operand. "Next" pointers of
64 * all nodes except BRANCH implement concatenation; a "next" pointer with
65 * a BRANCH on both ends of it is connecting two alternatives. (Here we
66 * have one of the subtle syntax dependencies: an individual BRANCH (as
67 * opposed to a collection of them) is never concatenated with anything
68 * because of operator precedence.) The operand of some types of node is
69 * a literal string; for others, it is a node leading into a sub-FSM. In
70 * particular, the operand of a BRANCH node is the first node of the branch.
71 * (NB this is *not* a tree structure: the tail of the branch connects
a3621e74
YO
72 * to the thing following the set of BRANCHes.) The opcodes are defined
73 * in regnodes.h which is generated from regcomp.sym by regcomp.pl.
a687059c
LW
74 */
75
a687059c
LW
76/*
77 * A node is one char of opcode followed by two chars of "next" pointer.
78 * "Next" pointers are stored as two 8-bit pieces, high order first. The
79 * value is a positive offset from the opcode of the node containing it.
80 * An operand, if any, simply follows the node. (Note that much of the
81 * code generation knows about this implicit relationship.)
82 *
83 * Using two bytes for the "next" pointer is vast overkill for most things,
84 * but allows patterns to get big without disasters.
85 *
e91177ed 86 * [The "next" pointer is always aligned on an even
a687059c
LW
87 * boundary, and reads the offset directly as a short. Also, there is no
88 * special test to reverse the sign of BACK pointers since the offset is
89 * stored negative.]
90 */
785a26d5
YO
91
92/* This is the stuff that used to live in regexp.h that was truly
93 private to the engine itself. It now lives here. */
94
28d8d7f4 95
785a26d5
YO
96
97 typedef struct regexp_internal {
1f1031fe 98 int name_list_idx; /* Optional data index of an array of paren names */
7122b237
YO
99 union {
100 U32 *offsets; /* offset annotations 20001228 MJD
101 data about mapping the program to the
102 string -
103 offsets[0] is proglen when this is used
104 */
105 U32 proglen;
106 } u;
107
6d5c990f
RGS
108 regnode *regstclass; /* Optional startclass as identified or constructed
109 by the optimiser */
110 struct reg_data *data; /* Additional miscellaneous data used by the program.
111 Used to make it easier to clone and free arbitrary
112 data that the regops need. Often the ARG field of
113 a regop is an index into this structure */
3d2bd50a
DM
114 struct reg_code_block *code_blocks;/* positions of literal (?{}) */
115 int num_code_blocks; /* size of code_blocks[] */
6d5c990f
RGS
116 regnode program[1]; /* Unwarranted chumminess with compiler. */
117} regexp_internal;
118
119#define RXi_SET(x,y) (x)->pprivate = (void*)(y)
120#define RXi_GET(x) ((regexp_internal *)((x)->pprivate))
121#define RXi_GET_DECL(r,ri) regexp_internal *ri = RXi_GET(r)
785a26d5
YO
122/*
123 * Flags stored in regexp->intflags
124 * These are used only internally to the regexp engine
125 *
126 * See regexp.h for flags used externally to the regexp engine
127 */
e3e400ec
YO
128#define RXp_INTFLAGS(rx) ((rx)->intflags)
129#define RX_INTFLAGS(prog) RXp_INTFLAGS(ReANY(prog))
130
785a26d5
YO
131#define PREGf_SKIP 0x00000001
132#define PREGf_IMPLICIT 0x00000002 /* Converted .* to ^.* */
133#define PREGf_NAUGHTY 0x00000004 /* how exponential is this pattern? */
134#define PREGf_VERBARG_SEEN 0x00000008
135#define PREGf_CUTGROUP_SEEN 0x00000010
732caac7 136#define PREGf_USE_RE_EVAL 0x00000020 /* compiled with "use re 'eval'" */
e3e400ec
YO
137/* these used to be extflags, but are now intflags */
138#define PREGf_NOSCAN 0x00000040
58430ea8 139#define PREGf_CANY_SEEN 0x00000080
8e1490ee
YO
140#define PREGf_GPOS_SEEN 0x00000100
141#define PREGf_GPOS_FLOAT 0x00000200
785a26d5 142
8e1490ee
YO
143#define PREGf_ANCH_BOL 0x00000400
144#define PREGf_ANCH_MBOL 0x00000800
145#define PREGf_ANCH_SBOL 0x00001000
146#define PREGf_ANCH_GPOS 0x00002000
147
fec6b98a
DM
148#define PREGf_ANCH (PREGf_ANCH_SBOL | PREGf_ANCH_GPOS | \
149 PREGf_ANCH_MBOL | PREGf_ANCH_BOL )
785a26d5
YO
150
151/* this is where the old regcomp.h started */
a687059c 152
c277df42 153struct regnode_string {
cd439c50 154 U8 str_len;
c277df42
IZ
155 U8 type;
156 U16 next_off;
cd439c50 157 char string[1];
c277df42 158};
a687059c 159
6bda09f9
YO
160/* Argument bearing node - workhorse,
161 arg1 is often for the data field */
c277df42
IZ
162struct regnode_1 {
163 U8 flags;
164 U8 type;
165 U16 next_off;
166 U32 arg1;
167};
168
6bda09f9
YO
169/* Similar to a regnode_1 but with an extra signed argument */
170struct regnode_2L {
171 U8 flags;
172 U8 type;
173 U16 next_off;
174 U32 arg1;
175 I32 arg2;
176};
177
178/* 'Two field' -- Two 16 bit unsigned args */
c277df42
IZ
179struct regnode_2 {
180 U8 flags;
181 U8 type;
182 U16 next_off;
183 U16 arg1;
184 U16 arg2;
185};
186
6bda09f9 187
936ed897 188#define ANYOF_BITMAP_SIZE 32 /* 256 b/(8 b/B) */
936ed897 189
786e8c11 190/* also used by trie */
936ed897
IZ
191struct regnode_charclass {
192 U8 flags;
193 U8 type;
194 U16 next_off;
7e817188 195 U32 arg1;
19103b91 196 char bitmap[ANYOF_BITMAP_SIZE]; /* only compile-time */
936ed897
IZ
197};
198
bf3f2d85
KW
199/* has runtime (locale) \d, \w, ..., [:posix:] classes */
200struct regnode_charclass_class {
8efd3f97 201 U8 flags; /* ANYOF_POSIXL bit must go here */
936ed897
IZ
202 U8 type;
203 U16 next_off;
7e817188 204 U32 arg1;
19103b91 205 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time */
9cf78fa0 206 U32 classflags; /* and run-time */
936ed897
IZ
207};
208
31f05a37
KW
209/* like above, but also has folds that are used only if the runtime locale is
210 * UTF-8. */
211struct regnode_charclass_posixl_fold {
212 U8 flags; /* ANYOF_POSIXL bit must go here */
213 U8 type;
214 U16 next_off;
7e817188 215 U32 arg1;
31f05a37
KW
216 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time */
217 U32 classflags; /* and run-time */
218 SV* utf8_locale_list; /* list of code points matched by folds
219 in a UTF-8 locale */
220};
221
222/* A synthetic start class; is a regnode_charclass_posixl_fold, plus an extra
223 * SV*, used only during its construction and which is not used by regexec.c.
224 * Note that the 'next_off' field is unused, as the SSC stands alone, so there
225 * is never a next node. */
b8f7bb16 226struct regnode_ssc {
8efd3f97 227 U8 flags; /* ANYOF_POSIXL bit must go here */
b8f7bb16
KW
228 U8 type;
229 U16 next_off;
7e817188 230 U32 arg1;
b8f7bb16
KW
231 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time */
232 U32 classflags; /* and run-time */
31f05a37
KW
233 SV* utf8_locale_list; /* list of code points matched by folds
234 in a UTF-8 locale */
b8f7bb16
KW
235 SV* invlist; /* list of code points matched */
236};
237
71068078
KW
238/* We take advantage of 'next_off' not otherwise being used in the SSC by
239 * actually using it: by setting it to 1. This allows us to test and
240 * distinguish between an SSC and other ANYOF node types, as 'next_off' cannot
241 * otherwise be 1, because it is the offset to the next regnode expressed in
242 * units of regnodes. Since an ANYOF node contains extra fields, it adds up
243 * to 12 regnode units on 32-bit systems, (hence the minimum this can be (if
244 * not 0) is 11 there. Even if things get tightly packed on a 64-bit system,
245 * it still would be more than 1. */
246#define set_ANYOF_SYNTHETIC(n) STMT_START{ OP(n) = ANYOF; \
247 NEXT_OFF(n) = 1; \
248 } STMT_END
249#define is_ANYOF_SYNTHETIC(n) (OP(n) == ANYOF && NEXT_OFF(n) == 1)
250
9c7e81e8
AD
251/* XXX fix this description.
252 Impose a limit of REG_INFTY on various pattern matching operations
253 to limit stack growth and to avoid "infinite" recursions.
254*/
255/* The default size for REG_INFTY is I16_MAX, which is the same as
256 SHORT_MAX (see perl.h). Unfortunately I16 isn't necessarily 16 bits
257 (see handy.h). On the Cray C90, sizeof(short)==4 and hence I16_MAX is
258 ((1<<31)-1), while on the Cray T90, sizeof(short)==8 and I16_MAX is
259 ((1<<63)-1). To limit stack growth to reasonable sizes, supply a
260 smaller default.
261 --Andy Dougherty 11 June 1998
262*/
263#if SHORTSIZE > 2
264# ifndef REG_INFTY
265# define REG_INFTY ((1<<15)-1)
266# endif
267#endif
268
269#ifndef REG_INFTY
83cfe48c
IZ
270# define REG_INFTY I16_MAX
271#endif
a687059c 272
e91177ed
IZ
273#define ARG_VALUE(arg) (arg)
274#define ARG__SET(arg,val) ((arg) = (val))
c277df42 275
e8e3b0fb
JH
276#undef ARG
277#undef ARG1
278#undef ARG2
279
c277df42
IZ
280#define ARG(p) ARG_VALUE(ARG_LOC(p))
281#define ARG1(p) ARG_VALUE(ARG1_LOC(p))
282#define ARG2(p) ARG_VALUE(ARG2_LOC(p))
6bda09f9 283#define ARG2L(p) ARG_VALUE(ARG2L_LOC(p))
786e8c11 284
c277df42
IZ
285#define ARG_SET(p, val) ARG__SET(ARG_LOC(p), (val))
286#define ARG1_SET(p, val) ARG__SET(ARG1_LOC(p), (val))
287#define ARG2_SET(p, val) ARG__SET(ARG2_LOC(p), (val))
6bda09f9 288#define ARG2L_SET(p, val) ARG__SET(ARG2L_LOC(p), (val))
c277df42 289
e8e3b0fb
JH
290#undef NEXT_OFF
291#undef NODE_ALIGN
292
075abff3
AL
293#define NEXT_OFF(p) ((p)->next_off)
294#define NODE_ALIGN(node)
295#define NODE_ALIGN_FILL(node) ((node)->flags = 0xde) /* deadbeef */
a687059c 296
c277df42
IZ
297#define SIZE_ALIGN NODE_ALIGN
298
e8e3b0fb
JH
299#undef OP
300#undef OPERAND
301#undef MASK
302#undef STRING
303
e91177ed 304#define OP(p) ((p)->type)
b988e673
KW
305#define FLAGS(p) ((p)->flags) /* Caution: Doesn't apply to all \
306 regnode types. For some, it's the \
307 character set of the regnode */
e91177ed 308#define OPERAND(p) (((struct regnode_string *)p)->string)
cd439c50
IZ
309#define MASK(p) ((char*)OPERAND(p))
310#define STR_LEN(p) (((struct regnode_string *)p)->str_len)
311#define STRING(p) (((struct regnode_string *)p)->string)
312#define STR_SZ(l) ((l + sizeof(regnode) - 1) / sizeof(regnode))
313#define NODE_SZ_STR(p) (STR_SZ(STR_LEN(p))+1)
314
e8e3b0fb
JH
315#undef NODE_ALIGN
316#undef ARG_LOC
317#undef NEXTOPER
318#undef PREVOPER
319
e91177ed
IZ
320#define NODE_ALIGN(node)
321#define ARG_LOC(p) (((struct regnode_1 *)p)->arg1)
322#define ARG1_LOC(p) (((struct regnode_2 *)p)->arg1)
323#define ARG2_LOC(p) (((struct regnode_2 *)p)->arg2)
6bda09f9 324#define ARG2L_LOC(p) (((struct regnode_2L *)p)->arg2)
786e8c11 325
e91177ed
IZ
326#define NODE_STEP_REGNODE 1 /* sizeof(regnode)/sizeof(regnode) */
327#define EXTRA_STEP_2ARGS EXTRA_SIZE(struct regnode_2)
328
329#define NODE_STEP_B 4
c277df42
IZ
330
331#define NEXTOPER(p) ((p) + NODE_STEP_REGNODE)
332#define PREVOPER(p) ((p) - NODE_STEP_REGNODE)
333
e91177ed 334#define FILL_ADVANCE_NODE(ptr, op) STMT_START { \
c277df42 335 (ptr)->type = op; (ptr)->next_off = 0; (ptr)++; } STMT_END
e91177ed 336#define FILL_ADVANCE_NODE_ARG(ptr, op, arg) STMT_START { \
c277df42 337 ARG_SET(ptr, arg); FILL_ADVANCE_NODE(ptr, op); (ptr) += 1; } STMT_END
a687059c 338
22c35a8c 339#define REG_MAGIC 0234
a687059c 340
9a81a976 341#define SIZE_ONLY (RExC_emit == (regnode *) & RExC_emit_dummy)
6d15ebf3
KW
342#define PASS1 SIZE_ONLY
343#define PASS2 (! SIZE_ONLY)
c277df42 344
137165a6
KW
345/* If the bitmap doesn't fully represent what this ANYOF node can match, the
346 * ARG is set to this special value (since 0, 1, ... are legal, but will never
347 * reach this high). */
348#define ANYOF_NONBITMAP_EMPTY ((U32) -1)
349
350/* The information used to be stored as as combination of the ANYOF_UTF8 and
351 * ANYOF_NONBITMAP_NON_UTF8 bits in the flags field, but was moved out of there
352 * to free up a bit for other uses. This tries to hide the change from
353 * existing code as much as possible. Now, the data structure that goes in ARG
354 * is not allocated unless it is needed, and that is what is used to determine
355 * if there is something outside the bitmap. The code now assumes that if
356 * that structure exists, that any UTF-8 encoded string should be tried against
357 * it, but a non-UTF8-encoded string will be tried only if the
358 * ANYOF_NONBITMAP_NON_UTF8 bit is also set. */
359#define ANYOF_NONBITMAP(node) (ARG(node) != ANYOF_NONBITMAP_EMPTY)
8f536e26
KW
360/* Flags for node->flags of ANYOF. These are in short supply, with one
361 * currently available. The ABOVE_LATIN1_ALL bit could be freed up
7bc66b18
KW
362 * by resorting to creating a swash containing everything above 255. This
363 * introduces a performance penalty. An option that wouldn't slow things down
8f536e26 364 * would be to split one of the two LOC flags out into a separate
7bc66b18
KW
365 * node, like what was done with ANYOF_NON_UTF8_NON_ASCII_ALL in commit
366 * 34fdef848b1687b91892ba55e9e0c3430e0770f6 (but which was reverted because it
367 * wasn't the best option available at the time), and using a LOC flag is
368 * probably better than that commit anyway. But it could be reinstated if we
369 * need a bit. The LOC flags are only for /l nodes; the reverted commit was
8f536e26
KW
370 * only for /d, so there are no combinatorial issues. The LOC flag to use is
371 * probably the POSIXL one.
34fdef84
KW
372 * Several flags are not used in synthetic start class (SSC) nodes, so could be
373 * shared should new flags be needed for SSCs, like ANYOF_EMPTY_STRING now. */
b8c5462f 374
5dbb0c08
KW
375/* regexec.c is expecting this to be in the low bit */
376#define ANYOF_INVERT 0x01
1984f6a0 377
a0dd4231
KW
378/* For the SSC node only, which cannot be inverted, so is shared with that bit.
379 * This means "Does this SSC match an empty string?" This is used only during
380 * regex compilation. */
381#define ANYOF_EMPTY_STRING ANYOF_INVERT
382
8f536e26 383/* spare 0x02 */
b63f73da
KW
384
385/* The fold is calculated and stored in the bitmap where possible at compile
386 * time. However under locale, the actual folding varies depending on
387 * what the locale is at the time of execution, so it has to be deferred until
388 * then */
389#define ANYOF_LOC_FOLD 0x04
390
d1d040e5 391/* Set if this is a regnode_charclass_posixl vs a regnode_charclass. This
dd58aee1
KW
392 * is used for runtime \d, \w, [:posix:], ..., which are used only in locale
393 * and the optimizer's synthetic start class. Non-locale \d, etc are resolved
8f536e26 394 * at compile-time */
8efd3f97
KW
395#define ANYOF_POSIXL 0x08
396#define ANYOF_CLASS ANYOF_POSIXL
397#define ANYOF_LARGE ANYOF_POSIXL
c8d3cd88 398
ae986089
KW
399/* Should we raise a warning if matching against an above-Unicode code point?
400 * */
401#define ANYOF_WARN_SUPER 0x10
f4244008 402
ef87b810
KW
403/* Can match something outside the bitmap that isn't in utf8 */
404#define ANYOF_NONBITMAP_NON_UTF8 0x20
405
61ef9872 406/* Matches every code point 0x100 and above*/
9b1eb067
KW
407#define ANYOF_ABOVE_LATIN1_ALL 0x40
408#define ANYOF_UNICODE_ALL ANYOF_ABOVE_LATIN1_ALL
ffc61ed2 409
7bc66b18
KW
410/* Match all Latin1 characters that aren't ASCII when the target string is not
411 * in utf8. */
412#define ANYOF_NON_UTF8_NON_ASCII_ALL 0x80
413
8f536e26 414#define ANYOF_FLAGS_ALL (0xf5)
653099ff 415
1462525b 416#define ANYOF_LOCALE_FLAGS (ANYOF_LOC_FOLD | ANYOF_POSIXL)
ace6b0e4 417
eff8b7dc
KW
418/* These are the flags that apply to both regular ANYOF nodes and synthetic
419 * start class nodes during construction of the SSC. During finalization of
420 * the SSC, other of the flags could be added to it */
1462525b 421#define ANYOF_COMMON_FLAGS (ANYOF_WARN_SUPER)
eff8b7dc 422
936ed897 423/* Character classes for node->classflags of ANYOF */
653099ff 424/* Should be synchronized with a table in regprop() */
575a8a29 425/* 2n should be the normal one, paired with its complement at 2n+1 */
b8c5462f 426
08ada29e
KW
427#define ANYOF_ALPHA ((_CC_ALPHA) * 2)
428#define ANYOF_NALPHA ((ANYOF_ALPHA) + 1)
e8d596e0
KW
429#define ANYOF_ALPHANUMERIC ((_CC_ALPHANUMERIC) * 2) /* [[:alnum:]] isalnum(3), utf8::IsAlnum */
430#define ANYOF_NALPHANUMERIC ((ANYOF_ALPHANUMERIC) + 1)
08ada29e
KW
431#define ANYOF_ASCII ((_CC_ASCII) * 2)
432#define ANYOF_NASCII ((ANYOF_ASCII) + 1)
e8d596e0
KW
433#define ANYOF_BLANK ((_CC_BLANK) * 2) /* GNU extension: space and tab: non-vertical space */
434#define ANYOF_NBLANK ((ANYOF_BLANK) + 1)
b0d691b2
KW
435#define ANYOF_CASED ((_CC_CASED) * 2) /* Pseudo class for [:lower:] or
436 [:upper:] under /i */
437#define ANYOF_NCASED ((ANYOF_CASED) + 1)
08ada29e
KW
438#define ANYOF_CNTRL ((_CC_CNTRL) * 2)
439#define ANYOF_NCNTRL ((ANYOF_CNTRL) + 1)
e8d596e0
KW
440#define ANYOF_DIGIT ((_CC_DIGIT) * 2) /* \d */
441#define ANYOF_NDIGIT ((ANYOF_DIGIT) + 1)
08ada29e
KW
442#define ANYOF_GRAPH ((_CC_GRAPH) * 2)
443#define ANYOF_NGRAPH ((ANYOF_GRAPH) + 1)
444#define ANYOF_LOWER ((_CC_LOWER) * 2)
445#define ANYOF_NLOWER ((ANYOF_LOWER) + 1)
446#define ANYOF_PRINT ((_CC_PRINT) * 2)
447#define ANYOF_NPRINT ((ANYOF_PRINT) + 1)
e8d596e0
KW
448#define ANYOF_PSXSPC ((_CC_PSXSPC) * 2) /* POSIX space: \s plus the vertical tab */
449#define ANYOF_NPSXSPC ((ANYOF_PSXSPC) + 1)
08ada29e
KW
450#define ANYOF_PUNCT ((_CC_PUNCT) * 2)
451#define ANYOF_NPUNCT ((ANYOF_PUNCT) + 1)
e8d596e0
KW
452#define ANYOF_SPACE ((_CC_SPACE) * 2) /* \s */
453#define ANYOF_NSPACE ((ANYOF_SPACE) + 1)
08ada29e
KW
454#define ANYOF_UPPER ((_CC_UPPER) * 2)
455#define ANYOF_NUPPER ((ANYOF_UPPER) + 1)
e8d596e0
KW
456#define ANYOF_WORDCHAR ((_CC_WORDCHAR) * 2) /* \w, PL_utf8_alnum, utf8::IsWord, ALNUM */
457#define ANYOF_NWORDCHAR ((ANYOF_WORDCHAR) + 1)
08ada29e
KW
458#define ANYOF_XDIGIT ((_CC_XDIGIT) * 2)
459#define ANYOF_NXDIGIT ((ANYOF_XDIGIT) + 1)
b8c5462f 460
87cad4c0 461/* pseudo classes below this, not stored in the class bitmap, but used as flags
e1d1eefb
YO
462 during compilation of char classes */
463
b2ef2e34
KW
464#define ANYOF_VERTWS ((_CC_VERTSPACE) * 2)
465#define ANYOF_NVERTWS ((ANYOF_VERTWS)+1)
e1d1eefb 466
b2ef2e34
KW
467/* It is best if this is the last one, as all above it are stored as bits in a
468 * bitmap, and it isn't part of that bitmap */
469#if _CC_VERTSPACE != _HIGHEST_REGCOMP_DOT_H_SYNC
470# error Problem with handy.h _HIGHEST_REGCOMP_DOT_H_SYNC #define
471#endif
472
8efd3f97
KW
473#define ANYOF_POSIXL_MAX (ANYOF_VERTWS) /* So upper loop limit is written:
474 * '< ANYOF_MAX'
475 * Hence doesn't include VERTWS, as that
476 * is a pseudo class */
477#define ANYOF_MAX ANYOF_POSIXL_MAX
478
479#if (ANYOF_POSIXL_MAX > 32) /* Must fit in 32-bit word */
b2ef2e34 480# error Problem with handy.h _CC_foo #defines
a0947d7b
KW
481#endif
482
8efd3f97
KW
483#define ANYOF_HORIZWS ((ANYOF_POSIXL_MAX)+2) /* = (ANYOF_NVERTWS + 1) */
484#define ANYOF_NHORIZWS ((ANYOF_POSIXL_MAX)+3)
87cad4c0 485
8efd3f97
KW
486#define ANYOF_UNIPROP ((ANYOF_POSIXL_MAX)+4) /* Used to indicate a Unicode
487 property: \p{} or \P{} */
f39a4dc2 488
b8c5462f
JH
489/* Backward source code compatibility. */
490
491#define ANYOF_ALNUML ANYOF_ALNUM
492#define ANYOF_NALNUML ANYOF_NALNUM
493#define ANYOF_SPACEL ANYOF_SPACE
494#define ANYOF_NSPACEL ANYOF_NSPACE
8ebfc29a
KW
495#define ANYOF_ALNUM ANYOF_WORDCHAR
496#define ANYOF_NALNUM ANYOF_NWORDCHAR
b8c5462f
JH
497
498/* Utility macros for the bitmap and classes of ANYOF */
499
936ed897 500#define ANYOF_SIZE (sizeof(struct regnode_charclass))
d1d040e5 501#define ANYOF_POSIXL_SIZE (sizeof(regnode_charclass_posixl))
8efd3f97 502#define ANYOF_CLASS_SIZE ANYOF_POSIXL_SIZE
31f05a37 503#define ANYOF_POSIXL_FOLD_SIZE (sizeof(regnode_charclass_posixl_fold))
b8c5462f 504
936ed897 505#define ANYOF_FLAGS(p) ((p)->flags)
b8c5462f
JH
506
507#define ANYOF_BIT(c) (1 << ((c) & 7))
508
d1d040e5 509#define ANYOF_POSIXL_SET(p, c) (((regnode_charclass_posixl*) (p))->classflags |= (1U << (c)))
8efd3f97
KW
510#define ANYOF_CLASS_SET(p, c) ANYOF_POSIXL_SET((p), (c))
511
d1d040e5 512#define ANYOF_POSIXL_CLEAR(p, c) (((regnode_charclass_posixl*) (p))->classflags &= ~ (1U <<(c)))
8efd3f97
KW
513#define ANYOF_CLASS_CLEAR(p, c) ANYOF_POSIXL_CLEAR((p), (c))
514
d1d040e5 515#define ANYOF_POSIXL_TEST(p, c) (((regnode_charclass_posixl*) (p))->classflags & (1U << (c)))
8efd3f97 516#define ANYOF_CLASS_TEST(p, c) ANYOF_POSIXL_TEST((p), (c))
b8c5462f 517
d1d040e5 518#define ANYOF_POSIXL_ZERO(ret) STMT_START { ((regnode_charclass_posixl*) (ret))->classflags = 0; } STMT_END
8efd3f97 519#define ANYOF_CLASS_ZERO(ret) ANYOF_POSIXL_ZERO(ret)
9cf78fa0
KW
520
521/* Shifts a bit to get, eg. 0x4000_0000, then subtracts 1 to get 0x3FFF_FFFF */
d1d040e5 522#define ANYOF_POSIXL_SETALL(ret) STMT_START { ((regnode_charclass_posixl*) (ret))->classflags = ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1; } STMT_END
8efd3f97 523#define ANYOF_CLASS_SETALL(ret) ANYOF_POSIXL_SETALL(ret)
9cf78fa0 524
8efd3f97
KW
525#define ANYOF_POSIXL_TEST_ANY_SET(p) \
526 ((ANYOF_FLAGS(p) & ANYOF_POSIXL) \
d1d040e5 527 && (((regnode_charclass_posixl*)(p))->classflags))
8efd3f97 528#define ANYOF_CLASS_TEST_ANY_SET(p) ANYOF_POSIXL_TEST_ANY_SET(p)
e867ce45 529
e0e1be5f
KW
530/* Since an SSC always has this field, we don't have to test for that; nor do
531 * we want to because the bit isn't set for SSC during its construction */
532#define ANYOF_POSIXL_SSC_TEST_ANY_SET(p) \
533 cBOOL(((regnode_ssc*)(p))->classflags)
534#define ANYOF_POSIXL_SSC_TEST_ALL_SET(p) /* Are all bits set? */ \
535 (((regnode_ssc*) (p))->classflags \
536 == ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1)
537
538#define ANYOF_POSIXL_TEST_ALL_SET(p) \
539 ((ANYOF_FLAGS(p) & ANYOF_POSIXL) \
540 && ((regnode_charclass_posixl*) (p))->classflags \
541 == ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1)
fb38762f 542
b27fd97a 543#define ANYOF_POSIXL_OR(source, dest) STMT_START { (dest)->classflags |= (source)->classflags ; } STMT_END
8efd3f97 544#define ANYOF_CLASS_OR(source, dest) ANYOF_POSIXL_OR((source), (dest))
936ed897 545
cdd87c1d
KW
546#define ANYOF_POSIXL_AND(source, dest) STMT_START { (dest)->classflags &= (source)->classflags ; } STMT_END
547
9cf78fa0 548#define ANYOF_BITMAP_ZERO(ret) Zero(((struct regnode_charclass*)(ret))->bitmap, ANYOF_BITMAP_SIZE, char)
936ed897 549#define ANYOF_BITMAP(p) (((struct regnode_charclass*)(p))->bitmap)
786e8c11 550#define ANYOF_BITMAP_BYTE(p, c) (ANYOF_BITMAP(p)[(((U8)(c)) >> 3) & 31])
b8c5462f
JH
551#define ANYOF_BITMAP_SET(p, c) (ANYOF_BITMAP_BYTE(p, c) |= ANYOF_BIT(c))
552#define ANYOF_BITMAP_CLEAR(p,c) (ANYOF_BITMAP_BYTE(p, c) &= ~ANYOF_BIT(c))
553#define ANYOF_BITMAP_TEST(p, c) (ANYOF_BITMAP_BYTE(p, c) & ANYOF_BIT(c))
554
f8bef550
NC
555#define ANYOF_BITMAP_SETALL(p) \
556 memset (ANYOF_BITMAP(p), 255, ANYOF_BITMAP_SIZE)
557#define ANYOF_BITMAP_CLEARALL(p) \
558 Zero (ANYOF_BITMAP(p), ANYOF_BITMAP_SIZE)
559/* Check that all 256 bits are all set. Used in S_cl_is_anything() */
c355c09a 560#define ANYOF_BITMAP_TESTALLSET(p) /* Assumes sizeof(p) == 32 */ \
f8bef550
NC
561 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)
562
936ed897 563#define ANYOF_SKIP ((ANYOF_SIZE - 1)/sizeof(regnode))
8efd3f97 564#define ANYOF_POSIXL_SKIP ((ANYOF_POSIXL_SIZE - 1)/sizeof(regnode))
31f05a37 565#define ANYOF_POSIXL_FOLD_SKIP ((ANYOF_POSIXL_FOLD_SIZE - 1)/sizeof(regnode))
8efd3f97 566#define ANYOF_CLASS_SKIP ANYOF_POSIXL_SKIP
3a15e693 567
31f05a37
KW
568#define ANYOF_UTF8_LOCALE_INVLIST(node) (((regnode_charclass_posixl_fold*) (node))->utf8_locale_list)
569
a687059c
LW
570/*
571 * Utility definitions.
572 */
2304df62 573#ifndef CHARMASK
075abff3 574# define UCHARAT(p) ((int)*(const U8*)(p))
a687059c 575#else
075abff3 576# define UCHARAT(p) ((int)*(p)&CHARMASK)
a687059c 577#endif
a687059c 578
c277df42
IZ
579#define EXTRA_SIZE(guy) ((sizeof(guy)-1)/sizeof(struct regnode))
580
e384d5c1
YO
581#define REG_ZERO_LEN_SEEN 0x00000001
582#define REG_LOOKBEHIND_SEEN 0x00000002
583#define REG_GPOS_SEEN 0x00000004
471f5387 584/* spare */
e384d5c1
YO
585#define REG_CANY_SEEN 0x00000010
586#define REG_RECURSE_SEEN 0x00000020
587#define REG_TOP_LEVEL_BRANCHES_SEEN 0x00000040
588#define REG_VERBARG_SEEN 0x00000080
589#define REG_CUTGROUP_SEEN 0x00000100
590#define REG_RUN_ON_COMMENT_SEEN 0x00000200
591#define REG_UNFOLDED_MULTI_SEEN 0x00000400
592#define REG_GOSTART_SEEN 0x00000800
ee273784
YO
593#define REG_UNBOUNDED_QUANTIFIER_SEEN 0x00001000
594
d09b2d29 595
73c4f7a1
GS
596START_EXTERN_C
597
be8e71aa
YO
598#ifdef PLUGGABLE_RE_EXTENSION
599#include "re_nodes.h"
600#else
d09b2d29 601#include "regnodes.h"
be8e71aa 602#endif
d09b2d29 603
f9f4320a
YO
604#ifndef PLUGGABLE_RE_EXTENSION
605#ifndef DOINIT
606EXTCONST regexp_engine PL_core_reg_engine;
607#else /* DOINIT */
608EXTCONST regexp_engine PL_core_reg_engine = {
fe578d7f 609 Perl_re_compile,
2fdbfb4d 610 Perl_regexec_flags,
f9f4320a
YO
611 Perl_re_intuit_start,
612 Perl_re_intuit_string,
2fdbfb4d
AB
613 Perl_regfree_internal,
614 Perl_reg_numbered_buff_fetch,
615 Perl_reg_numbered_buff_store,
616 Perl_reg_numbered_buff_length,
192b9cd1
AB
617 Perl_reg_named_buff,
618 Perl_reg_named_buff_iter,
49d7dfbc 619 Perl_reg_qr_package,
f9f4320a 620#if defined(USE_ITHREADS)
3c13cae6 621 Perl_regdupe_internal,
f9f4320a 622#endif
3c13cae6 623 Perl_re_op_compile
f9f4320a
YO
624};
625#endif /* DOINIT */
626#endif /* PLUGGABLE_RE_EXTENSION */
627
628
73c4f7a1 629END_EXTERN_C
cad2e5aa 630
cad2e5aa 631
261faec3
MJD
632/* .what is a character array with one character for each member of .data
633 * The character describes the function of the corresponding .data item:
af534a04 634 * a - AV for paren_name_list under DEBUGGING
0111c4fd 635 * f - start-class data for regstclass optimization
68e2671b 636 * l - start op for literal (?{EVAL}) item
d63c20f2 637 * L - start op for literal (?{EVAL}) item, with separate CV (qr//)
b30fcab9 638 * r - pointer to an embedded code-containing qr, e.g. /ab$qr/
38a44b82 639 * s - swash for Unicode-style character class, and the multicharacter
9e55ce06
JH
640 * strings resulting from casefolding the single-character entries
641 * in the character class
a3621e74 642 * t - trie struct
55eed653 643 * u - trie struct's widecharmap (a HV, so can't share, must dup)
2b8b4781 644 * also used for revcharmap and words under DEBUGGING
07be1b83 645 * T - aho-trie struct
81714fb9 646 * S - sv for named capture lookup
261faec3 647 * 20010712 mjd@plover.com
c3613ed7 648 * (Remember to update re_dup() and pregfree() if you add any items.)
261faec3 649 */
cad2e5aa
JH
650struct reg_data {
651 U32 count;
652 U8 *what;
653 void* data[1];
654};
655
a1cac82e
NC
656/* Code in S_to_utf8_substr() and S_to_byte_substr() in regexec.c accesses
657 anchored* and float* via array indexes 0 and 1. */
cad2e5aa 658#define anchored_substr substrs->data[0].substr
33b8afdf 659#define anchored_utf8 substrs->data[0].utf8_substr
cad2e5aa 660#define anchored_offset substrs->data[0].min_offset
1de06328
YO
661#define anchored_end_shift substrs->data[0].end_shift
662
cad2e5aa 663#define float_substr substrs->data[1].substr
33b8afdf 664#define float_utf8 substrs->data[1].utf8_substr
cad2e5aa
JH
665#define float_min_offset substrs->data[1].min_offset
666#define float_max_offset substrs->data[1].max_offset
1de06328
YO
667#define float_end_shift substrs->data[1].end_shift
668
cad2e5aa 669#define check_substr substrs->data[2].substr
33b8afdf 670#define check_utf8 substrs->data[2].utf8_substr
cad2e5aa
JH
671#define check_offset_min substrs->data[2].min_offset
672#define check_offset_max substrs->data[2].max_offset
1de06328 673#define check_end_shift substrs->data[2].end_shift
a3621e74 674
8d919b0a
FC
675#define RX_ANCHORED_SUBSTR(rx) (ReANY(rx)->anchored_substr)
676#define RX_ANCHORED_UTF8(rx) (ReANY(rx)->anchored_utf8)
677#define RX_FLOAT_SUBSTR(rx) (ReANY(rx)->float_substr)
678#define RX_FLOAT_UTF8(rx) (ReANY(rx)->float_utf8)
a3621e74
YO
679
680/* trie related stuff */
5d9a96ca 681
a3621e74
YO
682/* a transition record for the state machine. the
683 check field determines which state "owns" the
684 transition. the char the transition is for is
685 determined by offset from the owning states base
686 field. the next field determines which state
687 is to be transitioned to if any.
688*/
689struct _reg_trie_trans {
690 U32 next;
691 U32 check;
692};
693
694/* a transition list element for the list based representation */
695struct _reg_trie_trans_list_elem {
696 U16 forid;
697 U32 newstate;
698};
699typedef struct _reg_trie_trans_list_elem reg_trie_trans_le;
700
701/* a state for compressed nodes. base is an offset
702 into an array of reg_trie_trans array. If wordnum is
703 nonzero the state is accepting. if base is zero then
704 the state has no children (and will be accepting)
705*/
706struct _reg_trie_state {
707 U16 wordnum;
708 union {
709 U32 base;
710 reg_trie_trans_le* list;
711 } trans;
712};
713
2e64971a
DM
714/* info per word; indexed by wordnum */
715typedef struct {
716 U16 prev; /* previous word in acceptance chain; eg in
717 * zzz|abc|ab/ after matching the chars abc, the
718 * accepted word is #2, and the previous accepted
719 * word is #3 */
720 U32 len; /* how many chars long is this word? */
721 U32 accept; /* accept state for this word */
722} reg_trie_wordinfo;
a3621e74
YO
723
724
a3621e74
YO
725typedef struct _reg_trie_state reg_trie_state;
726typedef struct _reg_trie_trans reg_trie_trans;
727
728
729/* anything in here that needs to be freed later
23eab42c
NC
730 should be dealt with in pregfree.
731 refcount is first in both this and _reg_ac_data to allow a space
732 optimisation in Perl_regdupe. */
a3621e74 733struct _reg_trie_data {
23eab42c 734 U32 refcount; /* number of times this trie is referenced */
786e8c11
YO
735 U32 lasttrans; /* last valid transition element */
736 U16 *charmap; /* byte to charid lookup array */
786e8c11
YO
737 reg_trie_state *states; /* state data */
738 reg_trie_trans *trans; /* array of transition elements */
739 char *bitmap; /* stclass bitmap */
786e8c11
YO
740 U16 *jump; /* optional 1 indexed array of offsets before tail
741 for the node following a given word. */
2e64971a 742 reg_trie_wordinfo *wordinfo; /* array of info per word */
1b5c067c
NC
743 U16 uniquecharcount; /* unique chars in trie (width of trans table) */
744 U32 startstate; /* initial state - used for common prefix optimisation */
745 STRLEN minlen; /* minimum length of words in trie - build/opt only? */
746 STRLEN maxlen; /* maximum length of words in trie - build/opt only? */
2e64971a 747 U32 prefixlen; /* #chars in common prefix */
1e2e3d02
YO
748 U32 statecount; /* Build only - number of states in the states array
749 (including the unused zero state) */
786e8c11 750 U32 wordcount; /* Build only */
a3621e74 751#ifdef DEBUGGING
786e8c11 752 STRLEN charcount; /* Build only */
a3621e74
YO
753#endif
754};
2b8b4781
NC
755/* There is one (3 under DEBUGGING) pointers that logically belong in this
756 structure, but are held outside as they need duplication on thread cloning,
757 whereas the rest of the structure can be read only:
758 HV *widecharmap; code points > 255 to charid
759#ifdef DEBUGGING
760 AV *words; Array of words contained in trie, for dumping
761 AV *revcharmap; Map of each charid back to its character representation
762#endif
763*/
764
765#define TRIE_WORDS_OFFSET 2
766
a3621e74
YO
767typedef struct _reg_trie_data reg_trie_data;
768
23eab42c
NC
769/* refcount is first in both this and _reg_trie_data to allow a space
770 optimisation in Perl_regdupe. */
07be1b83 771struct _reg_ac_data {
23eab42c 772 U32 refcount;
1b5c067c 773 U32 trie;
07be1b83
YO
774 U32 *fail;
775 reg_trie_state *states;
07be1b83
YO
776};
777typedef struct _reg_ac_data reg_ac_data;
778
486ec47a 779/* ANY_BIT doesn't use the structure, so we can borrow it here.
3dab1dad
YO
780 This is simpler than refactoring all of it as wed end up with
781 three different sets... */
782
783#define TRIE_BITMAP(p) (((reg_trie_data *)(p))->bitmap)
786e8c11 784#define TRIE_BITMAP_BYTE(p, c) (TRIE_BITMAP(p)[(((U8)(c)) >> 3) & 31])
5dfe062f
YO
785#define TRIE_BITMAP_SET(p, c) (TRIE_BITMAP_BYTE(p, c) |= ANYOF_BIT((U8)c))
786#define TRIE_BITMAP_CLEAR(p,c) (TRIE_BITMAP_BYTE(p, c) &= ~ANYOF_BIT((U8)c))
787#define TRIE_BITMAP_TEST(p, c) (TRIE_BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
3dab1dad 788
1de06328
YO
789#define IS_ANYOF_TRIE(op) ((op)==TRIEC || (op)==AHOCORASICKC)
790#define IS_TRIE_AC(op) ((op)>=AHOCORASICK)
791
792
786e8c11
YO
793#define BITMAP_BYTE(p, c) (((U8*)p)[(((U8)(c)) >> 3) & 31])
794#define BITMAP_TEST(p, c) (BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
3dab1dad 795
a3621e74
YO
796/* these defines assume uniquecharcount is the correct variable, and state may be evaluated twice */
797#define TRIE_NODENUM(state) (((state)-1)/(trie->uniquecharcount)+1)
798#define SAFE_TRIE_NODENUM(state) ((state) ? (((state)-1)/(trie->uniquecharcount)+1) : (state))
799#define TRIE_NODEIDX(state) ((state) ? (((state)-1)*(trie->uniquecharcount)+1) : (state))
800
3dab1dad 801#ifdef DEBUGGING
3dab1dad 802#define TRIE_CHARCOUNT(trie) ((trie)->charcount)
3dab1dad 803#else
3dab1dad 804#define TRIE_CHARCOUNT(trie) (trie_charcount)
3dab1dad 805#endif
a3621e74 806
0111c4fd
RGS
807#define RE_TRIE_MAXBUF_INIT 65536
808#define RE_TRIE_MAXBUF_NAME "\022E_TRIE_MAXBUF"
a3621e74
YO
809#define RE_DEBUG_FLAGS "\022E_DEBUG_FLAGS"
810
be8e71aa
YO
811/*
812
813RE_DEBUG_FLAGS is used to control what debug output is emitted
814its divided into three groups of options, some of which interact.
815The three groups are: Compile, Execute, Extra. There is room for a
816further group, as currently only the low three bytes are used.
817
818 Compile Options:
07be1b83 819
be8e71aa
YO
820 PARSE
821 PEEP
822 TRIE
823 PROGRAM
824 OFFSETS
a3621e74 825
be8e71aa 826 Execute Options:
a3621e74 827
be8e71aa
YO
828 INTUIT
829 MATCH
830 TRIE
a3621e74 831
be8e71aa 832 Extra Options
a3621e74 833
be8e71aa
YO
834 TRIE
835 OFFSETS
836
837If you modify any of these make sure you make corresponding changes to
838re.pm, especially to the documentation.
a3621e74 839
be8e71aa
YO
840*/
841
842
843/* Compile */
844#define RE_DEBUG_COMPILE_MASK 0x0000FF
845#define RE_DEBUG_COMPILE_PARSE 0x000001
846#define RE_DEBUG_COMPILE_OPTIMISE 0x000002
847#define RE_DEBUG_COMPILE_TRIE 0x000004
848#define RE_DEBUG_COMPILE_DUMP 0x000008
f7819f85 849#define RE_DEBUG_COMPILE_FLAGS 0x000010
be8e71aa
YO
850
851/* Execute */
852#define RE_DEBUG_EXECUTE_MASK 0x00FF00
853#define RE_DEBUG_EXECUTE_INTUIT 0x000100
854#define RE_DEBUG_EXECUTE_MATCH 0x000200
855#define RE_DEBUG_EXECUTE_TRIE 0x000400
856
857/* Extra */
858#define RE_DEBUG_EXTRA_MASK 0xFF0000
859#define RE_DEBUG_EXTRA_TRIE 0x010000
860#define RE_DEBUG_EXTRA_OFFSETS 0x020000
a5ca303d
YO
861#define RE_DEBUG_EXTRA_OFFDEBUG 0x040000
862#define RE_DEBUG_EXTRA_STATE 0x080000
863#define RE_DEBUG_EXTRA_OPTIMISE 0x100000
e7707071 864#define RE_DEBUG_EXTRA_BUFFERS 0x400000
2c296965 865#define RE_DEBUG_EXTRA_GPOS 0x800000
24b23f37
YO
866/* combined */
867#define RE_DEBUG_EXTRA_STACK 0x280000
be8e71aa 868
e68ec53f 869#define RE_DEBUG_FLAG(x) (re_debug_flags & x)
be8e71aa
YO
870/* Compile */
871#define DEBUG_COMPILE_r(x) DEBUG_r( \
e68ec53f 872 if (re_debug_flags & RE_DEBUG_COMPILE_MASK) x )
be8e71aa 873#define DEBUG_PARSE_r(x) DEBUG_r( \
e68ec53f 874 if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x )
be8e71aa 875#define DEBUG_OPTIMISE_r(x) DEBUG_r( \
e68ec53f 876 if (re_debug_flags & RE_DEBUG_COMPILE_OPTIMISE) x )
be8e71aa 877#define DEBUG_PARSE_r(x) DEBUG_r( \
e68ec53f 878 if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x )
be8e71aa 879#define DEBUG_DUMP_r(x) DEBUG_r( \
e68ec53f 880 if (re_debug_flags & RE_DEBUG_COMPILE_DUMP) x )
be8e71aa 881#define DEBUG_TRIE_COMPILE_r(x) DEBUG_r( \
e68ec53f 882 if (re_debug_flags & RE_DEBUG_COMPILE_TRIE) x )
f7819f85
A
883#define DEBUG_FLAGS_r(x) DEBUG_r( \
884 if (re_debug_flags & RE_DEBUG_COMPILE_FLAGS) x )
be8e71aa
YO
885/* Execute */
886#define DEBUG_EXECUTE_r(x) DEBUG_r( \
e68ec53f 887 if (re_debug_flags & RE_DEBUG_EXECUTE_MASK) x )
be8e71aa 888#define DEBUG_INTUIT_r(x) DEBUG_r( \
e68ec53f 889 if (re_debug_flags & RE_DEBUG_EXECUTE_INTUIT) x )
be8e71aa 890#define DEBUG_MATCH_r(x) DEBUG_r( \
e68ec53f 891 if (re_debug_flags & RE_DEBUG_EXECUTE_MATCH) x )
be8e71aa 892#define DEBUG_TRIE_EXECUTE_r(x) DEBUG_r( \
e68ec53f 893 if (re_debug_flags & RE_DEBUG_EXECUTE_TRIE) x )
be8e71aa
YO
894
895/* Extra */
896#define DEBUG_EXTRA_r(x) DEBUG_r( \
e68ec53f 897 if (re_debug_flags & RE_DEBUG_EXTRA_MASK) x )
a5ca303d
YO
898#define DEBUG_OFFSETS_r(x) DEBUG_r( \
899 if (re_debug_flags & RE_DEBUG_EXTRA_OFFSETS) x )
ab3bbdeb
YO
900#define DEBUG_STATE_r(x) DEBUG_r( \
901 if (re_debug_flags & RE_DEBUG_EXTRA_STATE) x )
24b23f37
YO
902#define DEBUG_STACK_r(x) DEBUG_r( \
903 if (re_debug_flags & RE_DEBUG_EXTRA_STACK) x )
e7707071
YO
904#define DEBUG_BUFFERS_r(x) DEBUG_r( \
905 if (re_debug_flags & RE_DEBUG_EXTRA_BUFFERS) x )
906
a5ca303d
YO
907#define DEBUG_OPTIMISE_MORE_r(x) DEBUG_r( \
908 if ((RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE) == \
909 (re_debug_flags & (RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE)) ) x )
07be1b83 910#define MJD_OFFSET_DEBUG(x) DEBUG_r( \
a5ca303d 911 if (re_debug_flags & RE_DEBUG_EXTRA_OFFDEBUG) \
be8e71aa
YO
912 Perl_warn_nocontext x )
913#define DEBUG_TRIE_COMPILE_MORE_r(x) DEBUG_TRIE_COMPILE_r( \
e68ec53f 914 if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x )
be8e71aa 915#define DEBUG_TRIE_EXECUTE_MORE_r(x) DEBUG_TRIE_EXECUTE_r( \
e68ec53f 916 if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x )
be8e71aa
YO
917
918#define DEBUG_TRIE_r(x) DEBUG_r( \
e68ec53f 919 if (re_debug_flags & (RE_DEBUG_COMPILE_TRIE \
be8e71aa 920 | RE_DEBUG_EXECUTE_TRIE )) x )
2c296965
YO
921#define DEBUG_GPOS_r(x) DEBUG_r( \
922 if (re_debug_flags & RE_DEBUG_EXTRA_GPOS) x )
be8e71aa
YO
923
924/* initialization */
d2d8c711 925/* get_sv() can return NULL during global destruction. */
e68ec53f
YO
926#define GET_RE_DEBUG_FLAGS DEBUG_r({ \
927 SV * re_debug_flags_sv = NULL; \
928 re_debug_flags_sv = get_sv(RE_DEBUG_FLAGS, 1); \
929 if (re_debug_flags_sv) { \
8e9a8a48 930 if (!SvIOK(re_debug_flags_sv)) \
e68ec53f 931 sv_setuv(re_debug_flags_sv, RE_DEBUG_COMPILE_DUMP | RE_DEBUG_EXECUTE_MASK ); \
8e9a8a48 932 re_debug_flags=SvIV(re_debug_flags_sv); \
e68ec53f
YO
933 }\
934})
a3621e74
YO
935
936#ifdef DEBUGGING
ab3bbdeb 937
cb65b687
DM
938#define GET_RE_DEBUG_FLAGS_DECL VOL IV re_debug_flags = 0; \
939 PERL_UNUSED_VAR(re_debug_flags); GET_RE_DEBUG_FLAGS;
ab3bbdeb
YO
940
941#define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2) \
942 const char * const rpv = \
ddc5bc0f 943 pv_pretty((dsv), (pv), (l), (m), \
ab3bbdeb 944 PL_colors[(c1)],PL_colors[(c2)], \
c89df6cf 945 PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) ); \
0df25f3d 946 const int rlen = SvCUR(dsv)
ab3bbdeb
YO
947
948#define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m) \
949 const char * const rpv = \
ddc5bc0f 950 pv_pretty((dsv), (SvPV_nolen_const(sv)), (SvCUR(sv)), (m), \
ab3bbdeb 951 PL_colors[(c1)],PL_colors[(c2)], \
c89df6cf 952 PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) )
ab3bbdeb
YO
953
954#define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m) \
955 const char * const rpv = \
ddc5bc0f 956 pv_pretty((dsv), (pv), (l), (m), \
ab3bbdeb 957 PL_colors[0], PL_colors[1], \
c89df6cf 958 ( PERL_PV_PRETTY_QUOTE | PERL_PV_ESCAPE_RE | PERL_PV_ESCAPE_NONASCII | PERL_PV_PRETTY_ELLIPSES | \
ab3bbdeb 959 ((isuni) ? PERL_PV_ESCAPE_UNI : 0)) \
0a0e9afc 960 )
ab3bbdeb
YO
961
962#define RE_SV_DUMPLEN(ItEm) (SvCUR(ItEm) - (SvTAIL(ItEm)!=0))
963#define RE_SV_TAIL(ItEm) (SvTAIL(ItEm) ? "$" : "")
964
965#else /* if not DEBUGGING */
966
a3621e74 967#define GET_RE_DEBUG_FLAGS_DECL
ab3bbdeb
YO
968#define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2)
969#define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m)
970#define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m)
971#define RE_SV_DUMPLEN(ItEm)
972#define RE_SV_TAIL(ItEm)
973
974#endif /* DEBUG RELATED DEFINES */
a3621e74 975
e9a8c099
MHM
976/*
977 * Local variables:
978 * c-indentation-style: bsd
979 * c-basic-offset: 4
14d04a33 980 * indent-tabs-mode: nil
e9a8c099
MHM
981 * End:
982 *
14d04a33 983 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 984 */