This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
regcomp.h: Remove some no-longer used #defines
[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
dcb20b36 187#define NUM_ANYOF_CODE_POINTS 256
6bda09f9 188
dcb20b36 189#define ANYOF_BITMAP_SIZE (NUM_ANYOF_CODE_POINTS / 8) /* 8 bits/Byte */
936ed897 190
1ee208c4
KW
191/* Note that these form structs which are supersets of the next smaller one, by
192 * appending fields. Alignment problems can occur if one of those optional
193 * fields requires stricter alignment than the base struct. And formal
194 * parameters that can really be two or more of the structs should be
195 * declared as the smallest one it could be. See commit message for
196 * 7dcac5f6a5195002b55c935ee1d67f67e1df280b. Regnode allocation is done
197 * without regard to alignment, and changing it to would also require changing
198 * the code that inserts and deletes regnodes. The basic single-argument
199 * regnode has a U32, which is what reganode() allocates as a unit. Therefore
200 * no field can require stricter alignment than U32. */
201
786e8c11 202/* also used by trie */
936ed897
IZ
203struct regnode_charclass {
204 U8 flags;
205 U8 type;
206 U16 next_off;
7e817188 207 U32 arg1;
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 {
8efd3f97 213 U8 flags; /* ANYOF_POSIXL bit must go here */
936ed897
IZ
214 U8 type;
215 U16 next_off;
7e817188 216 U32 arg1;
19103b91 217 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time */
9cf78fa0 218 U32 classflags; /* and run-time */
936ed897
IZ
219};
220
85c8e306
KW
221/* A synthetic start class (SSC); is a regnode_charclass_posixl_fold, plus an
222 * extra SV*, used only during its construction and which is not used by
223 * regexec.c. Note that the 'next_off' field is unused, as the SSC stands
224 * alone, so there is never a next node. Also, there is no alignment issue,
225 * becase these are declared or allocated as a complete unit so the compiler
226 * takes care of alignment. This is unlike the other regnodes which are
227 * allocated in terms of multiples of a single-argument regnode. SSC nodes can
228 * have a pointer field because there is no alignment issue, and because it is
229 * set to NULL after construction, before any cloning of the pattern */
b8f7bb16 230struct regnode_ssc {
8efd3f97 231 U8 flags; /* ANYOF_POSIXL bit must go here */
b8f7bb16
KW
232 U8 type;
233 U16 next_off;
7e817188 234 U32 arg1;
b8f7bb16
KW
235 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time */
236 U32 classflags; /* and run-time */
85c8e306
KW
237
238 /* Auxiliary, only used during construction; NULL afterwards: list of code
239 * points matched */
240 SV* invlist;
b8f7bb16
KW
241};
242
71068078
KW
243/* We take advantage of 'next_off' not otherwise being used in the SSC by
244 * actually using it: by setting it to 1. This allows us to test and
245 * distinguish between an SSC and other ANYOF node types, as 'next_off' cannot
246 * otherwise be 1, because it is the offset to the next regnode expressed in
247 * units of regnodes. Since an ANYOF node contains extra fields, it adds up
248 * to 12 regnode units on 32-bit systems, (hence the minimum this can be (if
249 * not 0) is 11 there. Even if things get tightly packed on a 64-bit system,
250 * it still would be more than 1. */
251#define set_ANYOF_SYNTHETIC(n) STMT_START{ OP(n) = ANYOF; \
252 NEXT_OFF(n) = 1; \
253 } STMT_END
254#define is_ANYOF_SYNTHETIC(n) (OP(n) == ANYOF && NEXT_OFF(n) == 1)
255
9c7e81e8
AD
256/* XXX fix this description.
257 Impose a limit of REG_INFTY on various pattern matching operations
258 to limit stack growth and to avoid "infinite" recursions.
259*/
260/* The default size for REG_INFTY is I16_MAX, which is the same as
261 SHORT_MAX (see perl.h). Unfortunately I16 isn't necessarily 16 bits
262 (see handy.h). On the Cray C90, sizeof(short)==4 and hence I16_MAX is
263 ((1<<31)-1), while on the Cray T90, sizeof(short)==8 and I16_MAX is
264 ((1<<63)-1). To limit stack growth to reasonable sizes, supply a
265 smaller default.
266 --Andy Dougherty 11 June 1998
267*/
268#if SHORTSIZE > 2
269# ifndef REG_INFTY
270# define REG_INFTY ((1<<15)-1)
271# endif
272#endif
273
274#ifndef REG_INFTY
83cfe48c
IZ
275# define REG_INFTY I16_MAX
276#endif
a687059c 277
e91177ed
IZ
278#define ARG_VALUE(arg) (arg)
279#define ARG__SET(arg,val) ((arg) = (val))
c277df42 280
e8e3b0fb
JH
281#undef ARG
282#undef ARG1
283#undef ARG2
284
c277df42
IZ
285#define ARG(p) ARG_VALUE(ARG_LOC(p))
286#define ARG1(p) ARG_VALUE(ARG1_LOC(p))
287#define ARG2(p) ARG_VALUE(ARG2_LOC(p))
6bda09f9 288#define ARG2L(p) ARG_VALUE(ARG2L_LOC(p))
786e8c11 289
c277df42
IZ
290#define ARG_SET(p, val) ARG__SET(ARG_LOC(p), (val))
291#define ARG1_SET(p, val) ARG__SET(ARG1_LOC(p), (val))
292#define ARG2_SET(p, val) ARG__SET(ARG2_LOC(p), (val))
6bda09f9 293#define ARG2L_SET(p, val) ARG__SET(ARG2L_LOC(p), (val))
c277df42 294
e8e3b0fb
JH
295#undef NEXT_OFF
296#undef NODE_ALIGN
297
075abff3
AL
298#define NEXT_OFF(p) ((p)->next_off)
299#define NODE_ALIGN(node)
300#define NODE_ALIGN_FILL(node) ((node)->flags = 0xde) /* deadbeef */
a687059c 301
c277df42
IZ
302#define SIZE_ALIGN NODE_ALIGN
303
e8e3b0fb
JH
304#undef OP
305#undef OPERAND
306#undef MASK
307#undef STRING
308
e91177ed 309#define OP(p) ((p)->type)
b988e673
KW
310#define FLAGS(p) ((p)->flags) /* Caution: Doesn't apply to all \
311 regnode types. For some, it's the \
312 character set of the regnode */
e91177ed 313#define OPERAND(p) (((struct regnode_string *)p)->string)
cd439c50
IZ
314#define MASK(p) ((char*)OPERAND(p))
315#define STR_LEN(p) (((struct regnode_string *)p)->str_len)
316#define STRING(p) (((struct regnode_string *)p)->string)
317#define STR_SZ(l) ((l + sizeof(regnode) - 1) / sizeof(regnode))
318#define NODE_SZ_STR(p) (STR_SZ(STR_LEN(p))+1)
319
e8e3b0fb
JH
320#undef NODE_ALIGN
321#undef ARG_LOC
322#undef NEXTOPER
323#undef PREVOPER
324
e91177ed
IZ
325#define NODE_ALIGN(node)
326#define ARG_LOC(p) (((struct regnode_1 *)p)->arg1)
327#define ARG1_LOC(p) (((struct regnode_2 *)p)->arg1)
328#define ARG2_LOC(p) (((struct regnode_2 *)p)->arg2)
6bda09f9 329#define ARG2L_LOC(p) (((struct regnode_2L *)p)->arg2)
786e8c11 330
e91177ed
IZ
331#define NODE_STEP_REGNODE 1 /* sizeof(regnode)/sizeof(regnode) */
332#define EXTRA_STEP_2ARGS EXTRA_SIZE(struct regnode_2)
333
334#define NODE_STEP_B 4
c277df42
IZ
335
336#define NEXTOPER(p) ((p) + NODE_STEP_REGNODE)
337#define PREVOPER(p) ((p) - NODE_STEP_REGNODE)
338
e91177ed 339#define FILL_ADVANCE_NODE(ptr, op) STMT_START { \
c277df42 340 (ptr)->type = op; (ptr)->next_off = 0; (ptr)++; } STMT_END
e91177ed 341#define FILL_ADVANCE_NODE_ARG(ptr, op, arg) STMT_START { \
c277df42 342 ARG_SET(ptr, arg); FILL_ADVANCE_NODE(ptr, op); (ptr) += 1; } STMT_END
a687059c 343
22c35a8c 344#define REG_MAGIC 0234
a687059c 345
9a81a976 346#define SIZE_ONLY (RExC_emit == (regnode *) & RExC_emit_dummy)
6d15ebf3
KW
347#define PASS1 SIZE_ONLY
348#define PASS2 (! SIZE_ONLY)
c277df42 349
6f16c8da 350/* If the bitmap fully represents what this ANYOF node can match, the
137165a6
KW
351 * ARG is set to this special value (since 0, 1, ... are legal, but will never
352 * reach this high). */
353#define ANYOF_NONBITMAP_EMPTY ((U32) -1)
354
3b04b210 355/* Flags for node->flags of ANYOF. These are in short supply, with none
8f536e26 356 * currently available. The ABOVE_LATIN1_ALL bit could be freed up
7bc66b18
KW
357 * by resorting to creating a swash containing everything above 255. This
358 * introduces a performance penalty. An option that wouldn't slow things down
8f536e26 359 * would be to split one of the two LOC flags out into a separate
7bc66b18
KW
360 * node, like what was done with ANYOF_NON_UTF8_NON_ASCII_ALL in commit
361 * 34fdef848b1687b91892ba55e9e0c3430e0770f6 (but which was reverted because it
362 * wasn't the best option available at the time), and using a LOC flag is
363 * probably better than that commit anyway. But it could be reinstated if we
364 * need a bit. The LOC flags are only for /l nodes; the reverted commit was
8f536e26
KW
365 * only for /d, so there are no combinatorial issues. The LOC flag to use is
366 * probably the POSIXL one.
34fdef84
KW
367 * Several flags are not used in synthetic start class (SSC) nodes, so could be
368 * shared should new flags be needed for SSCs, like ANYOF_EMPTY_STRING now. */
b8c5462f 369
5dbb0c08
KW
370/* regexec.c is expecting this to be in the low bit */
371#define ANYOF_INVERT 0x01
1984f6a0 372
a0dd4231
KW
373/* For the SSC node only, which cannot be inverted, so is shared with that bit.
374 * This means "Does this SSC match an empty string?" This is used only during
375 * regex compilation. */
376#define ANYOF_EMPTY_STRING ANYOF_INVERT
377
3b04b210
KW
378/* Are there things that will match only if the target string is encoded in
379 * UTF-8? (This is not set if ANYOF_AOVE_LATIN1_ALL is set) */
380#define ANYOF_UTF8 0x02
b63f73da
KW
381
382/* The fold is calculated and stored in the bitmap where possible at compile
383 * time. However under locale, the actual folding varies depending on
384 * what the locale is at the time of execution, so it has to be deferred until
385 * then */
386#define ANYOF_LOC_FOLD 0x04
387
d1d040e5 388/* Set if this is a regnode_charclass_posixl vs a regnode_charclass. This
dd58aee1
KW
389 * is used for runtime \d, \w, [:posix:], ..., which are used only in locale
390 * and the optimizer's synthetic start class. Non-locale \d, etc are resolved
8f536e26 391 * at compile-time */
8efd3f97 392#define ANYOF_POSIXL 0x08
c8d3cd88 393
ae986089
KW
394/* Should we raise a warning if matching against an above-Unicode code point?
395 * */
396#define ANYOF_WARN_SUPER 0x10
f4244008 397
ef87b810
KW
398/* Can match something outside the bitmap that isn't in utf8 */
399#define ANYOF_NONBITMAP_NON_UTF8 0x20
400
61ef9872 401/* Matches every code point 0x100 and above*/
9b1eb067 402#define ANYOF_ABOVE_LATIN1_ALL 0x40
ffc61ed2 403
7bc66b18
KW
404/* Match all Latin1 characters that aren't ASCII when the target string is not
405 * in utf8. */
406#define ANYOF_NON_UTF8_NON_ASCII_ALL 0x80
407
3b04b210 408#define ANYOF_FLAGS_ALL (0xff)
653099ff 409
1462525b 410#define ANYOF_LOCALE_FLAGS (ANYOF_LOC_FOLD | ANYOF_POSIXL)
ace6b0e4 411
eff8b7dc
KW
412/* These are the flags that apply to both regular ANYOF nodes and synthetic
413 * start class nodes during construction of the SSC. During finalization of
414 * the SSC, other of the flags could be added to it */
3b04b210 415#define ANYOF_COMMON_FLAGS (ANYOF_WARN_SUPER|ANYOF_UTF8)
eff8b7dc 416
936ed897 417/* Character classes for node->classflags of ANYOF */
653099ff 418/* Should be synchronized with a table in regprop() */
575a8a29 419/* 2n should be the normal one, paired with its complement at 2n+1 */
b8c5462f 420
08ada29e
KW
421#define ANYOF_ALPHA ((_CC_ALPHA) * 2)
422#define ANYOF_NALPHA ((ANYOF_ALPHA) + 1)
e8d596e0
KW
423#define ANYOF_ALPHANUMERIC ((_CC_ALPHANUMERIC) * 2) /* [[:alnum:]] isalnum(3), utf8::IsAlnum */
424#define ANYOF_NALPHANUMERIC ((ANYOF_ALPHANUMERIC) + 1)
08ada29e
KW
425#define ANYOF_ASCII ((_CC_ASCII) * 2)
426#define ANYOF_NASCII ((ANYOF_ASCII) + 1)
e8d596e0
KW
427#define ANYOF_BLANK ((_CC_BLANK) * 2) /* GNU extension: space and tab: non-vertical space */
428#define ANYOF_NBLANK ((ANYOF_BLANK) + 1)
b0d691b2
KW
429#define ANYOF_CASED ((_CC_CASED) * 2) /* Pseudo class for [:lower:] or
430 [:upper:] under /i */
431#define ANYOF_NCASED ((ANYOF_CASED) + 1)
08ada29e
KW
432#define ANYOF_CNTRL ((_CC_CNTRL) * 2)
433#define ANYOF_NCNTRL ((ANYOF_CNTRL) + 1)
e8d596e0
KW
434#define ANYOF_DIGIT ((_CC_DIGIT) * 2) /* \d */
435#define ANYOF_NDIGIT ((ANYOF_DIGIT) + 1)
08ada29e
KW
436#define ANYOF_GRAPH ((_CC_GRAPH) * 2)
437#define ANYOF_NGRAPH ((ANYOF_GRAPH) + 1)
438#define ANYOF_LOWER ((_CC_LOWER) * 2)
439#define ANYOF_NLOWER ((ANYOF_LOWER) + 1)
440#define ANYOF_PRINT ((_CC_PRINT) * 2)
441#define ANYOF_NPRINT ((ANYOF_PRINT) + 1)
e8d596e0
KW
442#define ANYOF_PSXSPC ((_CC_PSXSPC) * 2) /* POSIX space: \s plus the vertical tab */
443#define ANYOF_NPSXSPC ((ANYOF_PSXSPC) + 1)
08ada29e
KW
444#define ANYOF_PUNCT ((_CC_PUNCT) * 2)
445#define ANYOF_NPUNCT ((ANYOF_PUNCT) + 1)
e8d596e0
KW
446#define ANYOF_SPACE ((_CC_SPACE) * 2) /* \s */
447#define ANYOF_NSPACE ((ANYOF_SPACE) + 1)
08ada29e
KW
448#define ANYOF_UPPER ((_CC_UPPER) * 2)
449#define ANYOF_NUPPER ((ANYOF_UPPER) + 1)
e8d596e0
KW
450#define ANYOF_WORDCHAR ((_CC_WORDCHAR) * 2) /* \w, PL_utf8_alnum, utf8::IsWord, ALNUM */
451#define ANYOF_NWORDCHAR ((ANYOF_WORDCHAR) + 1)
08ada29e
KW
452#define ANYOF_XDIGIT ((_CC_XDIGIT) * 2)
453#define ANYOF_NXDIGIT ((ANYOF_XDIGIT) + 1)
b8c5462f 454
87cad4c0 455/* pseudo classes below this, not stored in the class bitmap, but used as flags
e1d1eefb
YO
456 during compilation of char classes */
457
b2ef2e34
KW
458#define ANYOF_VERTWS ((_CC_VERTSPACE) * 2)
459#define ANYOF_NVERTWS ((ANYOF_VERTWS)+1)
e1d1eefb 460
b2ef2e34
KW
461/* It is best if this is the last one, as all above it are stored as bits in a
462 * bitmap, and it isn't part of that bitmap */
463#if _CC_VERTSPACE != _HIGHEST_REGCOMP_DOT_H_SYNC
464# error Problem with handy.h _HIGHEST_REGCOMP_DOT_H_SYNC #define
465#endif
466
8efd3f97
KW
467#define ANYOF_POSIXL_MAX (ANYOF_VERTWS) /* So upper loop limit is written:
468 * '< ANYOF_MAX'
469 * Hence doesn't include VERTWS, as that
470 * is a pseudo class */
471#define ANYOF_MAX ANYOF_POSIXL_MAX
472
473#if (ANYOF_POSIXL_MAX > 32) /* Must fit in 32-bit word */
b2ef2e34 474# error Problem with handy.h _CC_foo #defines
a0947d7b
KW
475#endif
476
8efd3f97
KW
477#define ANYOF_HORIZWS ((ANYOF_POSIXL_MAX)+2) /* = (ANYOF_NVERTWS + 1) */
478#define ANYOF_NHORIZWS ((ANYOF_POSIXL_MAX)+3)
87cad4c0 479
8efd3f97
KW
480#define ANYOF_UNIPROP ((ANYOF_POSIXL_MAX)+4) /* Used to indicate a Unicode
481 property: \p{} or \P{} */
f39a4dc2 482
b8c5462f
JH
483/* Backward source code compatibility. */
484
485#define ANYOF_ALNUML ANYOF_ALNUM
486#define ANYOF_NALNUML ANYOF_NALNUM
487#define ANYOF_SPACEL ANYOF_SPACE
488#define ANYOF_NSPACEL ANYOF_NSPACE
8ebfc29a
KW
489#define ANYOF_ALNUM ANYOF_WORDCHAR
490#define ANYOF_NALNUM ANYOF_NWORDCHAR
b8c5462f
JH
491
492/* Utility macros for the bitmap and classes of ANYOF */
493
936ed897 494#define ANYOF_SIZE (sizeof(struct regnode_charclass))
d1d040e5 495#define ANYOF_POSIXL_SIZE (sizeof(regnode_charclass_posixl))
8efd3f97 496#define ANYOF_CLASS_SIZE ANYOF_POSIXL_SIZE
b8c5462f 497
936ed897 498#define ANYOF_FLAGS(p) ((p)->flags)
b8c5462f 499
bc51fd78 500#define ANYOF_BIT(c) (1U << ((c) & 7))
b8c5462f 501
d1d040e5 502#define ANYOF_POSIXL_SET(p, c) (((regnode_charclass_posixl*) (p))->classflags |= (1U << (c)))
8efd3f97
KW
503#define ANYOF_CLASS_SET(p, c) ANYOF_POSIXL_SET((p), (c))
504
d1d040e5 505#define ANYOF_POSIXL_CLEAR(p, c) (((regnode_charclass_posixl*) (p))->classflags &= ~ (1U <<(c)))
8efd3f97
KW
506#define ANYOF_CLASS_CLEAR(p, c) ANYOF_POSIXL_CLEAR((p), (c))
507
d1d040e5 508#define ANYOF_POSIXL_TEST(p, c) (((regnode_charclass_posixl*) (p))->classflags & (1U << (c)))
8efd3f97 509#define ANYOF_CLASS_TEST(p, c) ANYOF_POSIXL_TEST((p), (c))
b8c5462f 510
d1d040e5 511#define ANYOF_POSIXL_ZERO(ret) STMT_START { ((regnode_charclass_posixl*) (ret))->classflags = 0; } STMT_END
8efd3f97 512#define ANYOF_CLASS_ZERO(ret) ANYOF_POSIXL_ZERO(ret)
9cf78fa0
KW
513
514/* Shifts a bit to get, eg. 0x4000_0000, then subtracts 1 to get 0x3FFF_FFFF */
d1d040e5 515#define ANYOF_POSIXL_SETALL(ret) STMT_START { ((regnode_charclass_posixl*) (ret))->classflags = ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1; } STMT_END
8efd3f97 516#define ANYOF_CLASS_SETALL(ret) ANYOF_POSIXL_SETALL(ret)
9cf78fa0 517
8efd3f97
KW
518#define ANYOF_POSIXL_TEST_ANY_SET(p) \
519 ((ANYOF_FLAGS(p) & ANYOF_POSIXL) \
d1d040e5 520 && (((regnode_charclass_posixl*)(p))->classflags))
8efd3f97 521#define ANYOF_CLASS_TEST_ANY_SET(p) ANYOF_POSIXL_TEST_ANY_SET(p)
e867ce45 522
e0e1be5f
KW
523/* Since an SSC always has this field, we don't have to test for that; nor do
524 * we want to because the bit isn't set for SSC during its construction */
525#define ANYOF_POSIXL_SSC_TEST_ANY_SET(p) \
526 cBOOL(((regnode_ssc*)(p))->classflags)
527#define ANYOF_POSIXL_SSC_TEST_ALL_SET(p) /* Are all bits set? */ \
528 (((regnode_ssc*) (p))->classflags \
529 == ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1)
530
531#define ANYOF_POSIXL_TEST_ALL_SET(p) \
532 ((ANYOF_FLAGS(p) & ANYOF_POSIXL) \
533 && ((regnode_charclass_posixl*) (p))->classflags \
534 == ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1)
fb38762f 535
b27fd97a 536#define ANYOF_POSIXL_OR(source, dest) STMT_START { (dest)->classflags |= (source)->classflags ; } STMT_END
8efd3f97 537#define ANYOF_CLASS_OR(source, dest) ANYOF_POSIXL_OR((source), (dest))
936ed897 538
cdd87c1d
KW
539#define ANYOF_POSIXL_AND(source, dest) STMT_START { (dest)->classflags &= (source)->classflags ; } STMT_END
540
9cf78fa0 541#define ANYOF_BITMAP_ZERO(ret) Zero(((struct regnode_charclass*)(ret))->bitmap, ANYOF_BITMAP_SIZE, char)
936ed897 542#define ANYOF_BITMAP(p) (((struct regnode_charclass*)(p))->bitmap)
786e8c11 543#define ANYOF_BITMAP_BYTE(p, c) (ANYOF_BITMAP(p)[(((U8)(c)) >> 3) & 31])
b8c5462f
JH
544#define ANYOF_BITMAP_SET(p, c) (ANYOF_BITMAP_BYTE(p, c) |= ANYOF_BIT(c))
545#define ANYOF_BITMAP_CLEAR(p,c) (ANYOF_BITMAP_BYTE(p, c) &= ~ANYOF_BIT(c))
bc51fd78 546#define ANYOF_BITMAP_TEST(p, c) cBOOL(ANYOF_BITMAP_BYTE(p, c) & ANYOF_BIT(c))
b8c5462f 547
f8bef550
NC
548#define ANYOF_BITMAP_SETALL(p) \
549 memset (ANYOF_BITMAP(p), 255, ANYOF_BITMAP_SIZE)
550#define ANYOF_BITMAP_CLEARALL(p) \
551 Zero (ANYOF_BITMAP(p), ANYOF_BITMAP_SIZE)
f8bef550 552
936ed897 553#define ANYOF_SKIP ((ANYOF_SIZE - 1)/sizeof(regnode))
8efd3f97
KW
554#define ANYOF_POSIXL_SKIP ((ANYOF_POSIXL_SIZE - 1)/sizeof(regnode))
555#define ANYOF_CLASS_SKIP ANYOF_POSIXL_SKIP
3a15e693 556
a687059c
LW
557/*
558 * Utility definitions.
559 */
2304df62 560#ifndef CHARMASK
075abff3 561# define UCHARAT(p) ((int)*(const U8*)(p))
a687059c 562#else
075abff3 563# define UCHARAT(p) ((int)*(p)&CHARMASK)
a687059c 564#endif
a687059c 565
c277df42
IZ
566#define EXTRA_SIZE(guy) ((sizeof(guy)-1)/sizeof(struct regnode))
567
e384d5c1
YO
568#define REG_ZERO_LEN_SEEN 0x00000001
569#define REG_LOOKBEHIND_SEEN 0x00000002
570#define REG_GPOS_SEEN 0x00000004
471f5387 571/* spare */
e384d5c1
YO
572#define REG_CANY_SEEN 0x00000010
573#define REG_RECURSE_SEEN 0x00000020
574#define REG_TOP_LEVEL_BRANCHES_SEEN 0x00000040
575#define REG_VERBARG_SEEN 0x00000080
576#define REG_CUTGROUP_SEEN 0x00000100
577#define REG_RUN_ON_COMMENT_SEEN 0x00000200
578#define REG_UNFOLDED_MULTI_SEEN 0x00000400
579#define REG_GOSTART_SEEN 0x00000800
ee273784
YO
580#define REG_UNBOUNDED_QUANTIFIER_SEEN 0x00001000
581
d09b2d29 582
73c4f7a1
GS
583START_EXTERN_C
584
be8e71aa
YO
585#ifdef PLUGGABLE_RE_EXTENSION
586#include "re_nodes.h"
587#else
d09b2d29 588#include "regnodes.h"
be8e71aa 589#endif
d09b2d29 590
f9f4320a
YO
591#ifndef PLUGGABLE_RE_EXTENSION
592#ifndef DOINIT
593EXTCONST regexp_engine PL_core_reg_engine;
594#else /* DOINIT */
595EXTCONST regexp_engine PL_core_reg_engine = {
fe578d7f 596 Perl_re_compile,
2fdbfb4d 597 Perl_regexec_flags,
f9f4320a
YO
598 Perl_re_intuit_start,
599 Perl_re_intuit_string,
2fdbfb4d
AB
600 Perl_regfree_internal,
601 Perl_reg_numbered_buff_fetch,
602 Perl_reg_numbered_buff_store,
603 Perl_reg_numbered_buff_length,
192b9cd1
AB
604 Perl_reg_named_buff,
605 Perl_reg_named_buff_iter,
49d7dfbc 606 Perl_reg_qr_package,
f9f4320a 607#if defined(USE_ITHREADS)
3c13cae6 608 Perl_regdupe_internal,
f9f4320a 609#endif
3c13cae6 610 Perl_re_op_compile
f9f4320a
YO
611};
612#endif /* DOINIT */
613#endif /* PLUGGABLE_RE_EXTENSION */
614
615
73c4f7a1 616END_EXTERN_C
cad2e5aa 617
cad2e5aa 618
261faec3
MJD
619/* .what is a character array with one character for each member of .data
620 * The character describes the function of the corresponding .data item:
af534a04 621 * a - AV for paren_name_list under DEBUGGING
0111c4fd 622 * f - start-class data for regstclass optimization
68e2671b 623 * l - start op for literal (?{EVAL}) item
d63c20f2 624 * L - start op for literal (?{EVAL}) item, with separate CV (qr//)
b30fcab9 625 * r - pointer to an embedded code-containing qr, e.g. /ab$qr/
38a44b82 626 * s - swash for Unicode-style character class, and the multicharacter
9e55ce06
JH
627 * strings resulting from casefolding the single-character entries
628 * in the character class
a3621e74 629 * t - trie struct
55eed653 630 * u - trie struct's widecharmap (a HV, so can't share, must dup)
2b8b4781 631 * also used for revcharmap and words under DEBUGGING
07be1b83 632 * T - aho-trie struct
81714fb9 633 * S - sv for named capture lookup
261faec3 634 * 20010712 mjd@plover.com
c3613ed7 635 * (Remember to update re_dup() and pregfree() if you add any items.)
261faec3 636 */
cad2e5aa
JH
637struct reg_data {
638 U32 count;
639 U8 *what;
640 void* data[1];
641};
642
a1cac82e
NC
643/* Code in S_to_utf8_substr() and S_to_byte_substr() in regexec.c accesses
644 anchored* and float* via array indexes 0 and 1. */
cad2e5aa 645#define anchored_substr substrs->data[0].substr
33b8afdf 646#define anchored_utf8 substrs->data[0].utf8_substr
cad2e5aa 647#define anchored_offset substrs->data[0].min_offset
1de06328
YO
648#define anchored_end_shift substrs->data[0].end_shift
649
cad2e5aa 650#define float_substr substrs->data[1].substr
33b8afdf 651#define float_utf8 substrs->data[1].utf8_substr
cad2e5aa
JH
652#define float_min_offset substrs->data[1].min_offset
653#define float_max_offset substrs->data[1].max_offset
1de06328
YO
654#define float_end_shift substrs->data[1].end_shift
655
cad2e5aa 656#define check_substr substrs->data[2].substr
33b8afdf 657#define check_utf8 substrs->data[2].utf8_substr
cad2e5aa
JH
658#define check_offset_min substrs->data[2].min_offset
659#define check_offset_max substrs->data[2].max_offset
1de06328 660#define check_end_shift substrs->data[2].end_shift
a3621e74 661
8d919b0a
FC
662#define RX_ANCHORED_SUBSTR(rx) (ReANY(rx)->anchored_substr)
663#define RX_ANCHORED_UTF8(rx) (ReANY(rx)->anchored_utf8)
664#define RX_FLOAT_SUBSTR(rx) (ReANY(rx)->float_substr)
665#define RX_FLOAT_UTF8(rx) (ReANY(rx)->float_utf8)
a3621e74
YO
666
667/* trie related stuff */
5d9a96ca 668
a3621e74
YO
669/* a transition record for the state machine. the
670 check field determines which state "owns" the
671 transition. the char the transition is for is
672 determined by offset from the owning states base
673 field. the next field determines which state
674 is to be transitioned to if any.
675*/
676struct _reg_trie_trans {
677 U32 next;
678 U32 check;
679};
680
681/* a transition list element for the list based representation */
682struct _reg_trie_trans_list_elem {
683 U16 forid;
684 U32 newstate;
685};
686typedef struct _reg_trie_trans_list_elem reg_trie_trans_le;
687
688/* a state for compressed nodes. base is an offset
689 into an array of reg_trie_trans array. If wordnum is
690 nonzero the state is accepting. if base is zero then
691 the state has no children (and will be accepting)
692*/
693struct _reg_trie_state {
694 U16 wordnum;
695 union {
696 U32 base;
697 reg_trie_trans_le* list;
698 } trans;
699};
700
2e64971a
DM
701/* info per word; indexed by wordnum */
702typedef struct {
703 U16 prev; /* previous word in acceptance chain; eg in
704 * zzz|abc|ab/ after matching the chars abc, the
705 * accepted word is #2, and the previous accepted
706 * word is #3 */
707 U32 len; /* how many chars long is this word? */
708 U32 accept; /* accept state for this word */
709} reg_trie_wordinfo;
a3621e74
YO
710
711
a3621e74
YO
712typedef struct _reg_trie_state reg_trie_state;
713typedef struct _reg_trie_trans reg_trie_trans;
714
715
716/* anything in here that needs to be freed later
23eab42c
NC
717 should be dealt with in pregfree.
718 refcount is first in both this and _reg_ac_data to allow a space
719 optimisation in Perl_regdupe. */
a3621e74 720struct _reg_trie_data {
23eab42c 721 U32 refcount; /* number of times this trie is referenced */
786e8c11
YO
722 U32 lasttrans; /* last valid transition element */
723 U16 *charmap; /* byte to charid lookup array */
786e8c11
YO
724 reg_trie_state *states; /* state data */
725 reg_trie_trans *trans; /* array of transition elements */
726 char *bitmap; /* stclass bitmap */
786e8c11
YO
727 U16 *jump; /* optional 1 indexed array of offsets before tail
728 for the node following a given word. */
2e64971a 729 reg_trie_wordinfo *wordinfo; /* array of info per word */
1b5c067c
NC
730 U16 uniquecharcount; /* unique chars in trie (width of trans table) */
731 U32 startstate; /* initial state - used for common prefix optimisation */
732 STRLEN minlen; /* minimum length of words in trie - build/opt only? */
733 STRLEN maxlen; /* maximum length of words in trie - build/opt only? */
2e64971a 734 U32 prefixlen; /* #chars in common prefix */
1e2e3d02
YO
735 U32 statecount; /* Build only - number of states in the states array
736 (including the unused zero state) */
786e8c11 737 U32 wordcount; /* Build only */
a3621e74 738#ifdef DEBUGGING
786e8c11 739 STRLEN charcount; /* Build only */
a3621e74
YO
740#endif
741};
2b8b4781
NC
742/* There is one (3 under DEBUGGING) pointers that logically belong in this
743 structure, but are held outside as they need duplication on thread cloning,
744 whereas the rest of the structure can be read only:
745 HV *widecharmap; code points > 255 to charid
746#ifdef DEBUGGING
747 AV *words; Array of words contained in trie, for dumping
748 AV *revcharmap; Map of each charid back to its character representation
749#endif
750*/
751
752#define TRIE_WORDS_OFFSET 2
753
a3621e74
YO
754typedef struct _reg_trie_data reg_trie_data;
755
23eab42c
NC
756/* refcount is first in both this and _reg_trie_data to allow a space
757 optimisation in Perl_regdupe. */
07be1b83 758struct _reg_ac_data {
23eab42c 759 U32 refcount;
1b5c067c 760 U32 trie;
07be1b83
YO
761 U32 *fail;
762 reg_trie_state *states;
07be1b83
YO
763};
764typedef struct _reg_ac_data reg_ac_data;
765
486ec47a 766/* ANY_BIT doesn't use the structure, so we can borrow it here.
3dab1dad
YO
767 This is simpler than refactoring all of it as wed end up with
768 three different sets... */
769
770#define TRIE_BITMAP(p) (((reg_trie_data *)(p))->bitmap)
786e8c11 771#define TRIE_BITMAP_BYTE(p, c) (TRIE_BITMAP(p)[(((U8)(c)) >> 3) & 31])
5dfe062f
YO
772#define TRIE_BITMAP_SET(p, c) (TRIE_BITMAP_BYTE(p, c) |= ANYOF_BIT((U8)c))
773#define TRIE_BITMAP_CLEAR(p,c) (TRIE_BITMAP_BYTE(p, c) &= ~ANYOF_BIT((U8)c))
774#define TRIE_BITMAP_TEST(p, c) (TRIE_BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
3dab1dad 775
1de06328
YO
776#define IS_ANYOF_TRIE(op) ((op)==TRIEC || (op)==AHOCORASICKC)
777#define IS_TRIE_AC(op) ((op)>=AHOCORASICK)
778
779
786e8c11
YO
780#define BITMAP_BYTE(p, c) (((U8*)p)[(((U8)(c)) >> 3) & 31])
781#define BITMAP_TEST(p, c) (BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
3dab1dad 782
a3621e74
YO
783/* these defines assume uniquecharcount is the correct variable, and state may be evaluated twice */
784#define TRIE_NODENUM(state) (((state)-1)/(trie->uniquecharcount)+1)
785#define SAFE_TRIE_NODENUM(state) ((state) ? (((state)-1)/(trie->uniquecharcount)+1) : (state))
786#define TRIE_NODEIDX(state) ((state) ? (((state)-1)*(trie->uniquecharcount)+1) : (state))
787
3dab1dad 788#ifdef DEBUGGING
3dab1dad 789#define TRIE_CHARCOUNT(trie) ((trie)->charcount)
3dab1dad 790#else
3dab1dad 791#define TRIE_CHARCOUNT(trie) (trie_charcount)
3dab1dad 792#endif
a3621e74 793
0111c4fd
RGS
794#define RE_TRIE_MAXBUF_INIT 65536
795#define RE_TRIE_MAXBUF_NAME "\022E_TRIE_MAXBUF"
a3621e74
YO
796#define RE_DEBUG_FLAGS "\022E_DEBUG_FLAGS"
797
be8e71aa
YO
798/*
799
800RE_DEBUG_FLAGS is used to control what debug output is emitted
801its divided into three groups of options, some of which interact.
802The three groups are: Compile, Execute, Extra. There is room for a
803further group, as currently only the low three bytes are used.
804
805 Compile Options:
07be1b83 806
be8e71aa
YO
807 PARSE
808 PEEP
809 TRIE
810 PROGRAM
811 OFFSETS
a3621e74 812
be8e71aa 813 Execute Options:
a3621e74 814
be8e71aa
YO
815 INTUIT
816 MATCH
817 TRIE
a3621e74 818
be8e71aa 819 Extra Options
a3621e74 820
be8e71aa
YO
821 TRIE
822 OFFSETS
823
824If you modify any of these make sure you make corresponding changes to
825re.pm, especially to the documentation.
a3621e74 826
be8e71aa
YO
827*/
828
829
830/* Compile */
831#define RE_DEBUG_COMPILE_MASK 0x0000FF
832#define RE_DEBUG_COMPILE_PARSE 0x000001
833#define RE_DEBUG_COMPILE_OPTIMISE 0x000002
834#define RE_DEBUG_COMPILE_TRIE 0x000004
835#define RE_DEBUG_COMPILE_DUMP 0x000008
f7819f85 836#define RE_DEBUG_COMPILE_FLAGS 0x000010
be8e71aa
YO
837
838/* Execute */
839#define RE_DEBUG_EXECUTE_MASK 0x00FF00
840#define RE_DEBUG_EXECUTE_INTUIT 0x000100
841#define RE_DEBUG_EXECUTE_MATCH 0x000200
842#define RE_DEBUG_EXECUTE_TRIE 0x000400
843
844/* Extra */
845#define RE_DEBUG_EXTRA_MASK 0xFF0000
846#define RE_DEBUG_EXTRA_TRIE 0x010000
847#define RE_DEBUG_EXTRA_OFFSETS 0x020000
a5ca303d
YO
848#define RE_DEBUG_EXTRA_OFFDEBUG 0x040000
849#define RE_DEBUG_EXTRA_STATE 0x080000
850#define RE_DEBUG_EXTRA_OPTIMISE 0x100000
e7707071 851#define RE_DEBUG_EXTRA_BUFFERS 0x400000
2c296965 852#define RE_DEBUG_EXTRA_GPOS 0x800000
24b23f37
YO
853/* combined */
854#define RE_DEBUG_EXTRA_STACK 0x280000
be8e71aa 855
e68ec53f 856#define RE_DEBUG_FLAG(x) (re_debug_flags & x)
be8e71aa
YO
857/* Compile */
858#define DEBUG_COMPILE_r(x) DEBUG_r( \
e68ec53f 859 if (re_debug_flags & RE_DEBUG_COMPILE_MASK) x )
be8e71aa 860#define DEBUG_PARSE_r(x) DEBUG_r( \
e68ec53f 861 if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x )
be8e71aa 862#define DEBUG_OPTIMISE_r(x) DEBUG_r( \
e68ec53f 863 if (re_debug_flags & RE_DEBUG_COMPILE_OPTIMISE) x )
be8e71aa 864#define DEBUG_PARSE_r(x) DEBUG_r( \
e68ec53f 865 if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x )
be8e71aa 866#define DEBUG_DUMP_r(x) DEBUG_r( \
e68ec53f 867 if (re_debug_flags & RE_DEBUG_COMPILE_DUMP) x )
be8e71aa 868#define DEBUG_TRIE_COMPILE_r(x) DEBUG_r( \
e68ec53f 869 if (re_debug_flags & RE_DEBUG_COMPILE_TRIE) x )
f7819f85
A
870#define DEBUG_FLAGS_r(x) DEBUG_r( \
871 if (re_debug_flags & RE_DEBUG_COMPILE_FLAGS) x )
be8e71aa
YO
872/* Execute */
873#define DEBUG_EXECUTE_r(x) DEBUG_r( \
e68ec53f 874 if (re_debug_flags & RE_DEBUG_EXECUTE_MASK) x )
be8e71aa 875#define DEBUG_INTUIT_r(x) DEBUG_r( \
e68ec53f 876 if (re_debug_flags & RE_DEBUG_EXECUTE_INTUIT) x )
be8e71aa 877#define DEBUG_MATCH_r(x) DEBUG_r( \
e68ec53f 878 if (re_debug_flags & RE_DEBUG_EXECUTE_MATCH) x )
be8e71aa 879#define DEBUG_TRIE_EXECUTE_r(x) DEBUG_r( \
e68ec53f 880 if (re_debug_flags & RE_DEBUG_EXECUTE_TRIE) x )
be8e71aa
YO
881
882/* Extra */
883#define DEBUG_EXTRA_r(x) DEBUG_r( \
e68ec53f 884 if (re_debug_flags & RE_DEBUG_EXTRA_MASK) x )
a5ca303d
YO
885#define DEBUG_OFFSETS_r(x) DEBUG_r( \
886 if (re_debug_flags & RE_DEBUG_EXTRA_OFFSETS) x )
ab3bbdeb
YO
887#define DEBUG_STATE_r(x) DEBUG_r( \
888 if (re_debug_flags & RE_DEBUG_EXTRA_STATE) x )
24b23f37
YO
889#define DEBUG_STACK_r(x) DEBUG_r( \
890 if (re_debug_flags & RE_DEBUG_EXTRA_STACK) x )
e7707071
YO
891#define DEBUG_BUFFERS_r(x) DEBUG_r( \
892 if (re_debug_flags & RE_DEBUG_EXTRA_BUFFERS) x )
893
a5ca303d
YO
894#define DEBUG_OPTIMISE_MORE_r(x) DEBUG_r( \
895 if ((RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE) == \
896 (re_debug_flags & (RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE)) ) x )
07be1b83 897#define MJD_OFFSET_DEBUG(x) DEBUG_r( \
a5ca303d 898 if (re_debug_flags & RE_DEBUG_EXTRA_OFFDEBUG) \
be8e71aa
YO
899 Perl_warn_nocontext x )
900#define DEBUG_TRIE_COMPILE_MORE_r(x) DEBUG_TRIE_COMPILE_r( \
e68ec53f 901 if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x )
be8e71aa 902#define DEBUG_TRIE_EXECUTE_MORE_r(x) DEBUG_TRIE_EXECUTE_r( \
e68ec53f 903 if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x )
be8e71aa
YO
904
905#define DEBUG_TRIE_r(x) DEBUG_r( \
e68ec53f 906 if (re_debug_flags & (RE_DEBUG_COMPILE_TRIE \
be8e71aa 907 | RE_DEBUG_EXECUTE_TRIE )) x )
2c296965
YO
908#define DEBUG_GPOS_r(x) DEBUG_r( \
909 if (re_debug_flags & RE_DEBUG_EXTRA_GPOS) x )
be8e71aa
YO
910
911/* initialization */
d2d8c711 912/* get_sv() can return NULL during global destruction. */
e68ec53f
YO
913#define GET_RE_DEBUG_FLAGS DEBUG_r({ \
914 SV * re_debug_flags_sv = NULL; \
915 re_debug_flags_sv = get_sv(RE_DEBUG_FLAGS, 1); \
916 if (re_debug_flags_sv) { \
8e9a8a48 917 if (!SvIOK(re_debug_flags_sv)) \
e68ec53f 918 sv_setuv(re_debug_flags_sv, RE_DEBUG_COMPILE_DUMP | RE_DEBUG_EXECUTE_MASK ); \
8e9a8a48 919 re_debug_flags=SvIV(re_debug_flags_sv); \
e68ec53f
YO
920 }\
921})
a3621e74
YO
922
923#ifdef DEBUGGING
ab3bbdeb 924
cb65b687
DM
925#define GET_RE_DEBUG_FLAGS_DECL VOL IV re_debug_flags = 0; \
926 PERL_UNUSED_VAR(re_debug_flags); GET_RE_DEBUG_FLAGS;
ab3bbdeb
YO
927
928#define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2) \
929 const char * const rpv = \
ddc5bc0f 930 pv_pretty((dsv), (pv), (l), (m), \
ab3bbdeb 931 PL_colors[(c1)],PL_colors[(c2)], \
c89df6cf 932 PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) ); \
0df25f3d 933 const int rlen = SvCUR(dsv)
ab3bbdeb
YO
934
935#define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m) \
936 const char * const rpv = \
ddc5bc0f 937 pv_pretty((dsv), (SvPV_nolen_const(sv)), (SvCUR(sv)), (m), \
ab3bbdeb 938 PL_colors[(c1)],PL_colors[(c2)], \
c89df6cf 939 PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) )
ab3bbdeb
YO
940
941#define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m) \
942 const char * const rpv = \
ddc5bc0f 943 pv_pretty((dsv), (pv), (l), (m), \
ab3bbdeb 944 PL_colors[0], PL_colors[1], \
c89df6cf 945 ( PERL_PV_PRETTY_QUOTE | PERL_PV_ESCAPE_RE | PERL_PV_ESCAPE_NONASCII | PERL_PV_PRETTY_ELLIPSES | \
ab3bbdeb 946 ((isuni) ? PERL_PV_ESCAPE_UNI : 0)) \
0a0e9afc 947 )
ab3bbdeb
YO
948
949#define RE_SV_DUMPLEN(ItEm) (SvCUR(ItEm) - (SvTAIL(ItEm)!=0))
950#define RE_SV_TAIL(ItEm) (SvTAIL(ItEm) ? "$" : "")
951
952#else /* if not DEBUGGING */
953
a3621e74 954#define GET_RE_DEBUG_FLAGS_DECL
ab3bbdeb
YO
955#define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2)
956#define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m)
957#define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m)
958#define RE_SV_DUMPLEN(ItEm)
959#define RE_SV_TAIL(ItEm)
960
961#endif /* DEBUG RELATED DEFINES */
a3621e74 962
e9a8c099
MHM
963/*
964 * Local variables:
965 * c-indentation-style: bsd
966 * c-basic-offset: 4
14d04a33 967 * indent-tabs-mode: nil
e9a8c099
MHM
968 * End:
969 *
14d04a33 970 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 971 */