This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
regcomp.sym: ANYOF nodes have an argument
[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
d3d47aac
YO
143#define PREGf_ANCH_MBOL 0x00000400
144#define PREGf_ANCH_SBOL 0x00000800
145#define PREGf_ANCH_GPOS 0x00001000
8e1490ee 146
d3d47aac
YO
147#define PREGf_ANCH \
148 ( PREGf_ANCH_SBOL | PREGf_ANCH_GPOS | PREGf_ANCH_MBOL )
785a26d5
YO
149
150/* this is where the old regcomp.h started */
a687059c 151
c277df42 152struct regnode_string {
cd439c50 153 U8 str_len;
c277df42
IZ
154 U8 type;
155 U16 next_off;
cd439c50 156 char string[1];
c277df42 157};
a687059c 158
6bda09f9
YO
159/* Argument bearing node - workhorse,
160 arg1 is often for the data field */
c277df42
IZ
161struct regnode_1 {
162 U8 flags;
163 U8 type;
164 U16 next_off;
165 U32 arg1;
166};
167
6bda09f9
YO
168/* Similar to a regnode_1 but with an extra signed argument */
169struct regnode_2L {
170 U8 flags;
171 U8 type;
172 U16 next_off;
173 U32 arg1;
174 I32 arg2;
175};
176
177/* 'Two field' -- Two 16 bit unsigned args */
c277df42
IZ
178struct regnode_2 {
179 U8 flags;
180 U8 type;
181 U16 next_off;
182 U16 arg1;
183 U16 arg2;
184};
185
e0a1ff7a
KW
186/* This give the number of code points that can be in the bitmap of an ANYOF
187 * node. The shift number must currently be one of: 8..12. It can't be less
188 * than 8 (256) because some code relies on it being at least that. Above 12
189 * (4096), and you start running into warnings that some data structure widths
190 * have been exceeded, though the test suite as of this writing still passes
191 * for up through 16, which is as high as anyone would ever want to go,
192 * encompassing all of the Unicode BMP, and thus including all the economically
193 * important world scripts. At 12 most of them are: including Arabic,
194 * Cyrillic, Greek, Hebrew, Indian subcontinent, Latin, and Thai; but not Han,
195 * Japanese, nor Korean. (The regarglen structure in regnodes.h is a U8, and
196 * the trie types TRIEC and AHOCORASICKC are larger than U8 for shift values
197 * below above 12.) Be sure to benchmark before changing, as larger sizes do
198 * significantly slow down the test suite */
199#define NUM_ANYOF_CODE_POINTS (1 << 8)
6bda09f9 200
dcb20b36 201#define ANYOF_BITMAP_SIZE (NUM_ANYOF_CODE_POINTS / 8) /* 8 bits/Byte */
936ed897 202
1ee208c4
KW
203/* Note that these form structs which are supersets of the next smaller one, by
204 * appending fields. Alignment problems can occur if one of those optional
205 * fields requires stricter alignment than the base struct. And formal
206 * parameters that can really be two or more of the structs should be
207 * declared as the smallest one it could be. See commit message for
208 * 7dcac5f6a5195002b55c935ee1d67f67e1df280b. Regnode allocation is done
209 * without regard to alignment, and changing it to would also require changing
210 * the code that inserts and deletes regnodes. The basic single-argument
211 * regnode has a U32, which is what reganode() allocates as a unit. Therefore
212 * no field can require stricter alignment than U32. */
213
786e8c11 214/* also used by trie */
936ed897
IZ
215struct regnode_charclass {
216 U8 flags;
217 U8 type;
218 U16 next_off;
9da99c5e 219 U32 arg1; /* set by set_ANYOF_arg() */
19103b91 220 char bitmap[ANYOF_BITMAP_SIZE]; /* only compile-time */
936ed897
IZ
221};
222
bf3f2d85
KW
223/* has runtime (locale) \d, \w, ..., [:posix:] classes */
224struct regnode_charclass_class {
93e92956 225 U8 flags; /* ANYOF_MATCHES_POSIXL bit must go here */
936ed897
IZ
226 U8 type;
227 U16 next_off;
7e817188 228 U32 arg1;
03fa83ba 229 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time ... */
9cf78fa0 230 U32 classflags; /* and run-time */
936ed897
IZ
231};
232
85c8e306
KW
233/* A synthetic start class (SSC); is a regnode_charclass_posixl_fold, plus an
234 * extra SV*, used only during its construction and which is not used by
235 * regexec.c. Note that the 'next_off' field is unused, as the SSC stands
236 * alone, so there is never a next node. Also, there is no alignment issue,
237 * becase these are declared or allocated as a complete unit so the compiler
238 * takes care of alignment. This is unlike the other regnodes which are
239 * allocated in terms of multiples of a single-argument regnode. SSC nodes can
240 * have a pointer field because there is no alignment issue, and because it is
241 * set to NULL after construction, before any cloning of the pattern */
b8f7bb16 242struct regnode_ssc {
93e92956 243 U8 flags; /* ANYOF_MATCHES_POSIXL bit must go here */
b8f7bb16
KW
244 U8 type;
245 U16 next_off;
7e817188 246 U32 arg1;
03fa83ba 247 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time ... */
b8f7bb16 248 U32 classflags; /* and run-time */
85c8e306
KW
249
250 /* Auxiliary, only used during construction; NULL afterwards: list of code
251 * points matched */
252 SV* invlist;
b8f7bb16
KW
253};
254
71068078
KW
255/* We take advantage of 'next_off' not otherwise being used in the SSC by
256 * actually using it: by setting it to 1. This allows us to test and
257 * distinguish between an SSC and other ANYOF node types, as 'next_off' cannot
258 * otherwise be 1, because it is the offset to the next regnode expressed in
259 * units of regnodes. Since an ANYOF node contains extra fields, it adds up
260 * to 12 regnode units on 32-bit systems, (hence the minimum this can be (if
261 * not 0) is 11 there. Even if things get tightly packed on a 64-bit system,
262 * it still would be more than 1. */
263#define set_ANYOF_SYNTHETIC(n) STMT_START{ OP(n) = ANYOF; \
264 NEXT_OFF(n) = 1; \
265 } STMT_END
266#define is_ANYOF_SYNTHETIC(n) (OP(n) == ANYOF && NEXT_OFF(n) == 1)
267
9c7e81e8
AD
268/* XXX fix this description.
269 Impose a limit of REG_INFTY on various pattern matching operations
270 to limit stack growth and to avoid "infinite" recursions.
271*/
272/* The default size for REG_INFTY is I16_MAX, which is the same as
273 SHORT_MAX (see perl.h). Unfortunately I16 isn't necessarily 16 bits
274 (see handy.h). On the Cray C90, sizeof(short)==4 and hence I16_MAX is
275 ((1<<31)-1), while on the Cray T90, sizeof(short)==8 and I16_MAX is
276 ((1<<63)-1). To limit stack growth to reasonable sizes, supply a
277 smaller default.
278 --Andy Dougherty 11 June 1998
279*/
280#if SHORTSIZE > 2
281# ifndef REG_INFTY
282# define REG_INFTY ((1<<15)-1)
283# endif
284#endif
285
286#ifndef REG_INFTY
83cfe48c
IZ
287# define REG_INFTY I16_MAX
288#endif
a687059c 289
e91177ed
IZ
290#define ARG_VALUE(arg) (arg)
291#define ARG__SET(arg,val) ((arg) = (val))
c277df42 292
e8e3b0fb
JH
293#undef ARG
294#undef ARG1
295#undef ARG2
296
c277df42
IZ
297#define ARG(p) ARG_VALUE(ARG_LOC(p))
298#define ARG1(p) ARG_VALUE(ARG1_LOC(p))
299#define ARG2(p) ARG_VALUE(ARG2_LOC(p))
6bda09f9 300#define ARG2L(p) ARG_VALUE(ARG2L_LOC(p))
786e8c11 301
c277df42
IZ
302#define ARG_SET(p, val) ARG__SET(ARG_LOC(p), (val))
303#define ARG1_SET(p, val) ARG__SET(ARG1_LOC(p), (val))
304#define ARG2_SET(p, val) ARG__SET(ARG2_LOC(p), (val))
6bda09f9 305#define ARG2L_SET(p, val) ARG__SET(ARG2L_LOC(p), (val))
c277df42 306
e8e3b0fb
JH
307#undef NEXT_OFF
308#undef NODE_ALIGN
309
075abff3
AL
310#define NEXT_OFF(p) ((p)->next_off)
311#define NODE_ALIGN(node)
aa48e906
YO
312/* the following define was set to 0xde in 075abff3
313 * as part of some linting logic. I have set it to 0
314 * as otherwise in every place where we /might/ set flags
315 * we have to set it 0 explicitly, which duplicates
316 * assignments and IMO adds an unacceptable level of
317 * surprise to working in the regex engine. If this
318 * is changed from 0 then at the very least make sure
319 * that SBOL for /^/ sets the flags to 0 explicitly.
320 * -- Yves */
321#define NODE_ALIGN_FILL(node) ((node)->flags = 0)
a687059c 322
c277df42
IZ
323#define SIZE_ALIGN NODE_ALIGN
324
e8e3b0fb
JH
325#undef OP
326#undef OPERAND
327#undef MASK
328#undef STRING
329
e91177ed 330#define OP(p) ((p)->type)
b988e673
KW
331#define FLAGS(p) ((p)->flags) /* Caution: Doesn't apply to all \
332 regnode types. For some, it's the \
333 character set of the regnode */
e91177ed 334#define OPERAND(p) (((struct regnode_string *)p)->string)
cd439c50
IZ
335#define MASK(p) ((char*)OPERAND(p))
336#define STR_LEN(p) (((struct regnode_string *)p)->str_len)
337#define STRING(p) (((struct regnode_string *)p)->string)
338#define STR_SZ(l) ((l + sizeof(regnode) - 1) / sizeof(regnode))
339#define NODE_SZ_STR(p) (STR_SZ(STR_LEN(p))+1)
340
e8e3b0fb
JH
341#undef NODE_ALIGN
342#undef ARG_LOC
343#undef NEXTOPER
344#undef PREVOPER
345
e91177ed
IZ
346#define NODE_ALIGN(node)
347#define ARG_LOC(p) (((struct regnode_1 *)p)->arg1)
348#define ARG1_LOC(p) (((struct regnode_2 *)p)->arg1)
349#define ARG2_LOC(p) (((struct regnode_2 *)p)->arg2)
6bda09f9 350#define ARG2L_LOC(p) (((struct regnode_2L *)p)->arg2)
786e8c11 351
e91177ed
IZ
352#define NODE_STEP_REGNODE 1 /* sizeof(regnode)/sizeof(regnode) */
353#define EXTRA_STEP_2ARGS EXTRA_SIZE(struct regnode_2)
354
355#define NODE_STEP_B 4
c277df42
IZ
356
357#define NEXTOPER(p) ((p) + NODE_STEP_REGNODE)
358#define PREVOPER(p) ((p) - NODE_STEP_REGNODE)
359
e91177ed 360#define FILL_ADVANCE_NODE(ptr, op) STMT_START { \
c277df42 361 (ptr)->type = op; (ptr)->next_off = 0; (ptr)++; } STMT_END
e91177ed 362#define FILL_ADVANCE_NODE_ARG(ptr, op, arg) STMT_START { \
c277df42 363 ARG_SET(ptr, arg); FILL_ADVANCE_NODE(ptr, op); (ptr) += 1; } STMT_END
a687059c 364
22c35a8c 365#define REG_MAGIC 0234
a687059c 366
9a81a976 367#define SIZE_ONLY (RExC_emit == (regnode *) & RExC_emit_dummy)
6d15ebf3
KW
368#define PASS1 SIZE_ONLY
369#define PASS2 (! SIZE_ONLY)
c277df42 370
6f16c8da 371/* If the bitmap fully represents what this ANYOF node can match, the
137165a6
KW
372 * ARG is set to this special value (since 0, 1, ... are legal, but will never
373 * reach this high). */
93e92956 374#define ANYOF_ONLY_HAS_BITMAP ((U32) -1)
137165a6 375
3b04b210 376/* Flags for node->flags of ANYOF. These are in short supply, with none
93e92956 377 * currently available. The ABOVE_BITMAP_ALL bit could be freed up
7bc66b18
KW
378 * by resorting to creating a swash containing everything above 255. This
379 * introduces a performance penalty. An option that wouldn't slow things down
8f536e26 380 * would be to split one of the two LOC flags out into a separate
7bc66b18
KW
381 * node, like what was done with ANYOF_NON_UTF8_NON_ASCII_ALL in commit
382 * 34fdef848b1687b91892ba55e9e0c3430e0770f6 (but which was reverted because it
383 * wasn't the best option available at the time), and using a LOC flag is
384 * probably better than that commit anyway. But it could be reinstated if we
385 * need a bit. The LOC flags are only for /l nodes; the reverted commit was
8f536e26
KW
386 * only for /d, so there are no combinatorial issues. The LOC flag to use is
387 * probably the POSIXL one.
34fdef84 388 * Several flags are not used in synthetic start class (SSC) nodes, so could be
93e92956
KW
389 * shared should new flags be needed for SSCs, like SSC_MATCHES_EMPTY_STRING
390 * now. */
b8c5462f 391
5dbb0c08 392/* regexec.c is expecting this to be in the low bit */
93e92956 393#define ANYOF_INVERT 0x01
1984f6a0 394
a0dd4231 395/* For the SSC node only, which cannot be inverted, so is shared with that bit.
93e92956
KW
396 * This is used only during regex compilation. */
397#define SSC_MATCHES_EMPTY_STRING ANYOF_INVERT
a0dd4231 398
93e92956
KW
399/* Are there things outside the bitmap that will match only if the target
400 * string is encoded in UTF-8? (This is not set if ANYOF_ABOVE_BITMAP_ALL is
401 * set) */
402#define ANYOF_HAS_UTF8_NONBITMAP_MATCHES 0x02
b63f73da
KW
403
404/* The fold is calculated and stored in the bitmap where possible at compile
405 * time. However under locale, the actual folding varies depending on
406 * what the locale is at the time of execution, so it has to be deferred until
407 * then */
93e92956 408#define ANYOF_LOC_FOLD 0x04
b63f73da 409
d1d040e5 410/* Set if this is a regnode_charclass_posixl vs a regnode_charclass. This
dd58aee1
KW
411 * is used for runtime \d, \w, [:posix:], ..., which are used only in locale
412 * and the optimizer's synthetic start class. Non-locale \d, etc are resolved
8f536e26 413 * at compile-time */
93e92956 414#define ANYOF_MATCHES_POSIXL 0x08
c8d3cd88 415
ae986089
KW
416/* Should we raise a warning if matching against an above-Unicode code point?
417 * */
93e92956 418#define ANYOF_WARN_SUPER 0x10
f4244008 419
ef87b810 420/* Can match something outside the bitmap that isn't in utf8 */
93e92956 421#define ANYOF_HAS_NONBITMAP_NON_UTF8_MATCHES 0x20
ef87b810 422
93e92956
KW
423/* Matches every code point NUM_ANYOF_CODE_POINTS and above*/
424#define ANYOF_MATCHES_ALL_ABOVE_BITMAP 0x40
ffc61ed2 425
7bc66b18
KW
426/* Match all Latin1 characters that aren't ASCII when the target string is not
427 * in utf8. */
93e92956 428#define ANYOF_MATCHES_ALL_NON_UTF8_NON_ASCII 0x80
7bc66b18 429
3b04b210 430#define ANYOF_FLAGS_ALL (0xff)
653099ff 431
93e92956 432#define ANYOF_LOCALE_FLAGS (ANYOF_LOC_FOLD | ANYOF_MATCHES_POSIXL)
ace6b0e4 433
eff8b7dc
KW
434/* These are the flags that apply to both regular ANYOF nodes and synthetic
435 * start class nodes during construction of the SSC. During finalization of
436 * the SSC, other of the flags could be added to it */
93e92956 437#define ANYOF_COMMON_FLAGS (ANYOF_WARN_SUPER|ANYOF_HAS_UTF8_NONBITMAP_MATCHES)
eff8b7dc 438
936ed897 439/* Character classes for node->classflags of ANYOF */
653099ff 440/* Should be synchronized with a table in regprop() */
575a8a29 441/* 2n should be the normal one, paired with its complement at 2n+1 */
b8c5462f 442
08ada29e
KW
443#define ANYOF_ALPHA ((_CC_ALPHA) * 2)
444#define ANYOF_NALPHA ((ANYOF_ALPHA) + 1)
e8d596e0
KW
445#define ANYOF_ALPHANUMERIC ((_CC_ALPHANUMERIC) * 2) /* [[:alnum:]] isalnum(3), utf8::IsAlnum */
446#define ANYOF_NALPHANUMERIC ((ANYOF_ALPHANUMERIC) + 1)
08ada29e
KW
447#define ANYOF_ASCII ((_CC_ASCII) * 2)
448#define ANYOF_NASCII ((ANYOF_ASCII) + 1)
e8d596e0
KW
449#define ANYOF_BLANK ((_CC_BLANK) * 2) /* GNU extension: space and tab: non-vertical space */
450#define ANYOF_NBLANK ((ANYOF_BLANK) + 1)
b0d691b2
KW
451#define ANYOF_CASED ((_CC_CASED) * 2) /* Pseudo class for [:lower:] or
452 [:upper:] under /i */
453#define ANYOF_NCASED ((ANYOF_CASED) + 1)
08ada29e
KW
454#define ANYOF_CNTRL ((_CC_CNTRL) * 2)
455#define ANYOF_NCNTRL ((ANYOF_CNTRL) + 1)
e8d596e0
KW
456#define ANYOF_DIGIT ((_CC_DIGIT) * 2) /* \d */
457#define ANYOF_NDIGIT ((ANYOF_DIGIT) + 1)
08ada29e
KW
458#define ANYOF_GRAPH ((_CC_GRAPH) * 2)
459#define ANYOF_NGRAPH ((ANYOF_GRAPH) + 1)
460#define ANYOF_LOWER ((_CC_LOWER) * 2)
461#define ANYOF_NLOWER ((ANYOF_LOWER) + 1)
462#define ANYOF_PRINT ((_CC_PRINT) * 2)
463#define ANYOF_NPRINT ((ANYOF_PRINT) + 1)
e8d596e0
KW
464#define ANYOF_PSXSPC ((_CC_PSXSPC) * 2) /* POSIX space: \s plus the vertical tab */
465#define ANYOF_NPSXSPC ((ANYOF_PSXSPC) + 1)
08ada29e
KW
466#define ANYOF_PUNCT ((_CC_PUNCT) * 2)
467#define ANYOF_NPUNCT ((ANYOF_PUNCT) + 1)
e8d596e0
KW
468#define ANYOF_SPACE ((_CC_SPACE) * 2) /* \s */
469#define ANYOF_NSPACE ((ANYOF_SPACE) + 1)
08ada29e
KW
470#define ANYOF_UPPER ((_CC_UPPER) * 2)
471#define ANYOF_NUPPER ((ANYOF_UPPER) + 1)
e8d596e0
KW
472#define ANYOF_WORDCHAR ((_CC_WORDCHAR) * 2) /* \w, PL_utf8_alnum, utf8::IsWord, ALNUM */
473#define ANYOF_NWORDCHAR ((ANYOF_WORDCHAR) + 1)
08ada29e
KW
474#define ANYOF_XDIGIT ((_CC_XDIGIT) * 2)
475#define ANYOF_NXDIGIT ((ANYOF_XDIGIT) + 1)
b8c5462f 476
87cad4c0 477/* pseudo classes below this, not stored in the class bitmap, but used as flags
e1d1eefb
YO
478 during compilation of char classes */
479
b2ef2e34
KW
480#define ANYOF_VERTWS ((_CC_VERTSPACE) * 2)
481#define ANYOF_NVERTWS ((ANYOF_VERTWS)+1)
e1d1eefb 482
b2ef2e34
KW
483/* It is best if this is the last one, as all above it are stored as bits in a
484 * bitmap, and it isn't part of that bitmap */
485#if _CC_VERTSPACE != _HIGHEST_REGCOMP_DOT_H_SYNC
486# error Problem with handy.h _HIGHEST_REGCOMP_DOT_H_SYNC #define
487#endif
488
8efd3f97
KW
489#define ANYOF_POSIXL_MAX (ANYOF_VERTWS) /* So upper loop limit is written:
490 * '< ANYOF_MAX'
491 * Hence doesn't include VERTWS, as that
492 * is a pseudo class */
493#define ANYOF_MAX ANYOF_POSIXL_MAX
494
495#if (ANYOF_POSIXL_MAX > 32) /* Must fit in 32-bit word */
b2ef2e34 496# error Problem with handy.h _CC_foo #defines
a0947d7b
KW
497#endif
498
8efd3f97
KW
499#define ANYOF_HORIZWS ((ANYOF_POSIXL_MAX)+2) /* = (ANYOF_NVERTWS + 1) */
500#define ANYOF_NHORIZWS ((ANYOF_POSIXL_MAX)+3)
87cad4c0 501
8efd3f97
KW
502#define ANYOF_UNIPROP ((ANYOF_POSIXL_MAX)+4) /* Used to indicate a Unicode
503 property: \p{} or \P{} */
f39a4dc2 504
b8c5462f
JH
505/* Backward source code compatibility. */
506
507#define ANYOF_ALNUML ANYOF_ALNUM
508#define ANYOF_NALNUML ANYOF_NALNUM
509#define ANYOF_SPACEL ANYOF_SPACE
510#define ANYOF_NSPACEL ANYOF_NSPACE
8ebfc29a
KW
511#define ANYOF_ALNUM ANYOF_WORDCHAR
512#define ANYOF_NALNUM ANYOF_NWORDCHAR
b8c5462f
JH
513
514/* Utility macros for the bitmap and classes of ANYOF */
515
936ed897 516#define ANYOF_FLAGS(p) ((p)->flags)
b8c5462f 517
bc51fd78 518#define ANYOF_BIT(c) (1U << ((c) & 7))
b8c5462f 519
d1d040e5 520#define ANYOF_POSIXL_SET(p, c) (((regnode_charclass_posixl*) (p))->classflags |= (1U << (c)))
8efd3f97
KW
521#define ANYOF_CLASS_SET(p, c) ANYOF_POSIXL_SET((p), (c))
522
d1d040e5 523#define ANYOF_POSIXL_CLEAR(p, c) (((regnode_charclass_posixl*) (p))->classflags &= ~ (1U <<(c)))
8efd3f97
KW
524#define ANYOF_CLASS_CLEAR(p, c) ANYOF_POSIXL_CLEAR((p), (c))
525
d1d040e5 526#define ANYOF_POSIXL_TEST(p, c) (((regnode_charclass_posixl*) (p))->classflags & (1U << (c)))
8efd3f97 527#define ANYOF_CLASS_TEST(p, c) ANYOF_POSIXL_TEST((p), (c))
b8c5462f 528
d1d040e5 529#define ANYOF_POSIXL_ZERO(ret) STMT_START { ((regnode_charclass_posixl*) (ret))->classflags = 0; } STMT_END
8efd3f97 530#define ANYOF_CLASS_ZERO(ret) ANYOF_POSIXL_ZERO(ret)
9cf78fa0
KW
531
532/* Shifts a bit to get, eg. 0x4000_0000, then subtracts 1 to get 0x3FFF_FFFF */
d1d040e5 533#define ANYOF_POSIXL_SETALL(ret) STMT_START { ((regnode_charclass_posixl*) (ret))->classflags = ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1; } STMT_END
8efd3f97 534#define ANYOF_CLASS_SETALL(ret) ANYOF_POSIXL_SETALL(ret)
9cf78fa0 535
8efd3f97 536#define ANYOF_POSIXL_TEST_ANY_SET(p) \
93e92956 537 ((ANYOF_FLAGS(p) & ANYOF_MATCHES_POSIXL) \
d1d040e5 538 && (((regnode_charclass_posixl*)(p))->classflags))
8efd3f97 539#define ANYOF_CLASS_TEST_ANY_SET(p) ANYOF_POSIXL_TEST_ANY_SET(p)
e867ce45 540
e0e1be5f
KW
541/* Since an SSC always has this field, we don't have to test for that; nor do
542 * we want to because the bit isn't set for SSC during its construction */
543#define ANYOF_POSIXL_SSC_TEST_ANY_SET(p) \
544 cBOOL(((regnode_ssc*)(p))->classflags)
545#define ANYOF_POSIXL_SSC_TEST_ALL_SET(p) /* Are all bits set? */ \
546 (((regnode_ssc*) (p))->classflags \
547 == ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1)
548
549#define ANYOF_POSIXL_TEST_ALL_SET(p) \
93e92956 550 ((ANYOF_FLAGS(p) & ANYOF_MATCHES_POSIXL) \
e0e1be5f
KW
551 && ((regnode_charclass_posixl*) (p))->classflags \
552 == ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1)
fb38762f 553
b27fd97a 554#define ANYOF_POSIXL_OR(source, dest) STMT_START { (dest)->classflags |= (source)->classflags ; } STMT_END
8efd3f97 555#define ANYOF_CLASS_OR(source, dest) ANYOF_POSIXL_OR((source), (dest))
936ed897 556
cdd87c1d
KW
557#define ANYOF_POSIXL_AND(source, dest) STMT_START { (dest)->classflags &= (source)->classflags ; } STMT_END
558
9cf78fa0 559#define ANYOF_BITMAP_ZERO(ret) Zero(((struct regnode_charclass*)(ret))->bitmap, ANYOF_BITMAP_SIZE, char)
936ed897 560#define ANYOF_BITMAP(p) (((struct regnode_charclass*)(p))->bitmap)
786e8c11 561#define ANYOF_BITMAP_BYTE(p, c) (ANYOF_BITMAP(p)[(((U8)(c)) >> 3) & 31])
b8c5462f
JH
562#define ANYOF_BITMAP_SET(p, c) (ANYOF_BITMAP_BYTE(p, c) |= ANYOF_BIT(c))
563#define ANYOF_BITMAP_CLEAR(p,c) (ANYOF_BITMAP_BYTE(p, c) &= ~ANYOF_BIT(c))
bc51fd78 564#define ANYOF_BITMAP_TEST(p, c) cBOOL(ANYOF_BITMAP_BYTE(p, c) & ANYOF_BIT(c))
b8c5462f 565
f8bef550
NC
566#define ANYOF_BITMAP_SETALL(p) \
567 memset (ANYOF_BITMAP(p), 255, ANYOF_BITMAP_SIZE)
568#define ANYOF_BITMAP_CLEARALL(p) \
569 Zero (ANYOF_BITMAP(p), ANYOF_BITMAP_SIZE)
f8bef550 570
888ee784
KW
571#define ANYOF_SKIP EXTRA_SIZE(struct regnode_charclass)
572#define ANYOF_POSIXL_SKIP EXTRA_SIZE(regnode_charclass_posixl)
3a15e693 573
a687059c
LW
574/*
575 * Utility definitions.
576 */
2304df62 577#ifndef CHARMASK
075abff3 578# define UCHARAT(p) ((int)*(const U8*)(p))
a687059c 579#else
075abff3 580# define UCHARAT(p) ((int)*(p)&CHARMASK)
a687059c 581#endif
a687059c 582
c277df42
IZ
583#define EXTRA_SIZE(guy) ((sizeof(guy)-1)/sizeof(struct regnode))
584
e384d5c1
YO
585#define REG_ZERO_LEN_SEEN 0x00000001
586#define REG_LOOKBEHIND_SEEN 0x00000002
587#define REG_GPOS_SEEN 0x00000004
471f5387 588/* spare */
e384d5c1
YO
589#define REG_CANY_SEEN 0x00000010
590#define REG_RECURSE_SEEN 0x00000020
591#define REG_TOP_LEVEL_BRANCHES_SEEN 0x00000040
592#define REG_VERBARG_SEEN 0x00000080
593#define REG_CUTGROUP_SEEN 0x00000100
594#define REG_RUN_ON_COMMENT_SEEN 0x00000200
595#define REG_UNFOLDED_MULTI_SEEN 0x00000400
596#define REG_GOSTART_SEEN 0x00000800
ee273784
YO
597#define REG_UNBOUNDED_QUANTIFIER_SEEN 0x00001000
598
d09b2d29 599
73c4f7a1
GS
600START_EXTERN_C
601
be8e71aa
YO
602#ifdef PLUGGABLE_RE_EXTENSION
603#include "re_nodes.h"
604#else
d09b2d29 605#include "regnodes.h"
be8e71aa 606#endif
d09b2d29 607
f9f4320a
YO
608#ifndef PLUGGABLE_RE_EXTENSION
609#ifndef DOINIT
610EXTCONST regexp_engine PL_core_reg_engine;
611#else /* DOINIT */
612EXTCONST regexp_engine PL_core_reg_engine = {
fe578d7f 613 Perl_re_compile,
2fdbfb4d 614 Perl_regexec_flags,
f9f4320a
YO
615 Perl_re_intuit_start,
616 Perl_re_intuit_string,
2fdbfb4d
AB
617 Perl_regfree_internal,
618 Perl_reg_numbered_buff_fetch,
619 Perl_reg_numbered_buff_store,
620 Perl_reg_numbered_buff_length,
192b9cd1
AB
621 Perl_reg_named_buff,
622 Perl_reg_named_buff_iter,
49d7dfbc 623 Perl_reg_qr_package,
f9f4320a 624#if defined(USE_ITHREADS)
3c13cae6 625 Perl_regdupe_internal,
f9f4320a 626#endif
3c13cae6 627 Perl_re_op_compile
f9f4320a
YO
628};
629#endif /* DOINIT */
630#endif /* PLUGGABLE_RE_EXTENSION */
631
632
73c4f7a1 633END_EXTERN_C
cad2e5aa 634
cad2e5aa 635
261faec3
MJD
636/* .what is a character array with one character for each member of .data
637 * The character describes the function of the corresponding .data item:
af534a04 638 * a - AV for paren_name_list under DEBUGGING
0111c4fd 639 * f - start-class data for regstclass optimization
68e2671b 640 * l - start op for literal (?{EVAL}) item
d63c20f2 641 * L - start op for literal (?{EVAL}) item, with separate CV (qr//)
b30fcab9 642 * r - pointer to an embedded code-containing qr, e.g. /ab$qr/
38a44b82 643 * s - swash for Unicode-style character class, and the multicharacter
9e55ce06
JH
644 * strings resulting from casefolding the single-character entries
645 * in the character class
a3621e74 646 * t - trie struct
55eed653 647 * u - trie struct's widecharmap (a HV, so can't share, must dup)
2b8b4781 648 * also used for revcharmap and words under DEBUGGING
07be1b83 649 * T - aho-trie struct
81714fb9 650 * S - sv for named capture lookup
261faec3 651 * 20010712 mjd@plover.com
c3613ed7 652 * (Remember to update re_dup() and pregfree() if you add any items.)
261faec3 653 */
cad2e5aa
JH
654struct reg_data {
655 U32 count;
656 U8 *what;
657 void* data[1];
658};
659
a1cac82e
NC
660/* Code in S_to_utf8_substr() and S_to_byte_substr() in regexec.c accesses
661 anchored* and float* via array indexes 0 and 1. */
cad2e5aa 662#define anchored_substr substrs->data[0].substr
33b8afdf 663#define anchored_utf8 substrs->data[0].utf8_substr
cad2e5aa 664#define anchored_offset substrs->data[0].min_offset
1de06328
YO
665#define anchored_end_shift substrs->data[0].end_shift
666
cad2e5aa 667#define float_substr substrs->data[1].substr
33b8afdf 668#define float_utf8 substrs->data[1].utf8_substr
cad2e5aa
JH
669#define float_min_offset substrs->data[1].min_offset
670#define float_max_offset substrs->data[1].max_offset
1de06328
YO
671#define float_end_shift substrs->data[1].end_shift
672
cad2e5aa 673#define check_substr substrs->data[2].substr
33b8afdf 674#define check_utf8 substrs->data[2].utf8_substr
cad2e5aa
JH
675#define check_offset_min substrs->data[2].min_offset
676#define check_offset_max substrs->data[2].max_offset
1de06328 677#define check_end_shift substrs->data[2].end_shift
a3621e74 678
8d919b0a
FC
679#define RX_ANCHORED_SUBSTR(rx) (ReANY(rx)->anchored_substr)
680#define RX_ANCHORED_UTF8(rx) (ReANY(rx)->anchored_utf8)
681#define RX_FLOAT_SUBSTR(rx) (ReANY(rx)->float_substr)
682#define RX_FLOAT_UTF8(rx) (ReANY(rx)->float_utf8)
a3621e74
YO
683
684/* trie related stuff */
5d9a96ca 685
a3621e74
YO
686/* a transition record for the state machine. the
687 check field determines which state "owns" the
688 transition. the char the transition is for is
689 determined by offset from the owning states base
690 field. the next field determines which state
691 is to be transitioned to if any.
692*/
693struct _reg_trie_trans {
694 U32 next;
695 U32 check;
696};
697
698/* a transition list element for the list based representation */
699struct _reg_trie_trans_list_elem {
700 U16 forid;
701 U32 newstate;
702};
703typedef struct _reg_trie_trans_list_elem reg_trie_trans_le;
704
705/* a state for compressed nodes. base is an offset
706 into an array of reg_trie_trans array. If wordnum is
707 nonzero the state is accepting. if base is zero then
708 the state has no children (and will be accepting)
709*/
710struct _reg_trie_state {
711 U16 wordnum;
712 union {
713 U32 base;
714 reg_trie_trans_le* list;
715 } trans;
716};
717
2e64971a
DM
718/* info per word; indexed by wordnum */
719typedef struct {
720 U16 prev; /* previous word in acceptance chain; eg in
721 * zzz|abc|ab/ after matching the chars abc, the
722 * accepted word is #2, and the previous accepted
723 * word is #3 */
724 U32 len; /* how many chars long is this word? */
725 U32 accept; /* accept state for this word */
726} reg_trie_wordinfo;
a3621e74
YO
727
728
a3621e74
YO
729typedef struct _reg_trie_state reg_trie_state;
730typedef struct _reg_trie_trans reg_trie_trans;
731
732
733/* anything in here that needs to be freed later
23eab42c
NC
734 should be dealt with in pregfree.
735 refcount is first in both this and _reg_ac_data to allow a space
736 optimisation in Perl_regdupe. */
a3621e74 737struct _reg_trie_data {
23eab42c 738 U32 refcount; /* number of times this trie is referenced */
786e8c11
YO
739 U32 lasttrans; /* last valid transition element */
740 U16 *charmap; /* byte to charid lookup array */
786e8c11
YO
741 reg_trie_state *states; /* state data */
742 reg_trie_trans *trans; /* array of transition elements */
743 char *bitmap; /* stclass bitmap */
786e8c11
YO
744 U16 *jump; /* optional 1 indexed array of offsets before tail
745 for the node following a given word. */
2e64971a 746 reg_trie_wordinfo *wordinfo; /* array of info per word */
1b5c067c
NC
747 U16 uniquecharcount; /* unique chars in trie (width of trans table) */
748 U32 startstate; /* initial state - used for common prefix optimisation */
749 STRLEN minlen; /* minimum length of words in trie - build/opt only? */
750 STRLEN maxlen; /* maximum length of words in trie - build/opt only? */
2e64971a 751 U32 prefixlen; /* #chars in common prefix */
1e2e3d02
YO
752 U32 statecount; /* Build only - number of states in the states array
753 (including the unused zero state) */
786e8c11 754 U32 wordcount; /* Build only */
a3621e74 755#ifdef DEBUGGING
786e8c11 756 STRLEN charcount; /* Build only */
a3621e74
YO
757#endif
758};
2b8b4781
NC
759/* There is one (3 under DEBUGGING) pointers that logically belong in this
760 structure, but are held outside as they need duplication on thread cloning,
761 whereas the rest of the structure can be read only:
762 HV *widecharmap; code points > 255 to charid
763#ifdef DEBUGGING
764 AV *words; Array of words contained in trie, for dumping
765 AV *revcharmap; Map of each charid back to its character representation
766#endif
767*/
768
769#define TRIE_WORDS_OFFSET 2
770
a3621e74
YO
771typedef struct _reg_trie_data reg_trie_data;
772
23eab42c
NC
773/* refcount is first in both this and _reg_trie_data to allow a space
774 optimisation in Perl_regdupe. */
07be1b83 775struct _reg_ac_data {
23eab42c 776 U32 refcount;
1b5c067c 777 U32 trie;
07be1b83
YO
778 U32 *fail;
779 reg_trie_state *states;
07be1b83
YO
780};
781typedef struct _reg_ac_data reg_ac_data;
782
486ec47a 783/* ANY_BIT doesn't use the structure, so we can borrow it here.
3dab1dad
YO
784 This is simpler than refactoring all of it as wed end up with
785 three different sets... */
786
787#define TRIE_BITMAP(p) (((reg_trie_data *)(p))->bitmap)
786e8c11 788#define TRIE_BITMAP_BYTE(p, c) (TRIE_BITMAP(p)[(((U8)(c)) >> 3) & 31])
5dfe062f
YO
789#define TRIE_BITMAP_SET(p, c) (TRIE_BITMAP_BYTE(p, c) |= ANYOF_BIT((U8)c))
790#define TRIE_BITMAP_CLEAR(p,c) (TRIE_BITMAP_BYTE(p, c) &= ~ANYOF_BIT((U8)c))
791#define TRIE_BITMAP_TEST(p, c) (TRIE_BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
3dab1dad 792
1de06328
YO
793#define IS_ANYOF_TRIE(op) ((op)==TRIEC || (op)==AHOCORASICKC)
794#define IS_TRIE_AC(op) ((op)>=AHOCORASICK)
795
796
786e8c11
YO
797#define BITMAP_BYTE(p, c) (((U8*)p)[(((U8)(c)) >> 3) & 31])
798#define BITMAP_TEST(p, c) (BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
3dab1dad 799
a3621e74
YO
800/* these defines assume uniquecharcount is the correct variable, and state may be evaluated twice */
801#define TRIE_NODENUM(state) (((state)-1)/(trie->uniquecharcount)+1)
802#define SAFE_TRIE_NODENUM(state) ((state) ? (((state)-1)/(trie->uniquecharcount)+1) : (state))
803#define TRIE_NODEIDX(state) ((state) ? (((state)-1)*(trie->uniquecharcount)+1) : (state))
804
3dab1dad 805#ifdef DEBUGGING
3dab1dad 806#define TRIE_CHARCOUNT(trie) ((trie)->charcount)
3dab1dad 807#else
3dab1dad 808#define TRIE_CHARCOUNT(trie) (trie_charcount)
3dab1dad 809#endif
a3621e74 810
0111c4fd
RGS
811#define RE_TRIE_MAXBUF_INIT 65536
812#define RE_TRIE_MAXBUF_NAME "\022E_TRIE_MAXBUF"
a3621e74
YO
813#define RE_DEBUG_FLAGS "\022E_DEBUG_FLAGS"
814
be8e71aa
YO
815/*
816
817RE_DEBUG_FLAGS is used to control what debug output is emitted
818its divided into three groups of options, some of which interact.
819The three groups are: Compile, Execute, Extra. There is room for a
820further group, as currently only the low three bytes are used.
821
822 Compile Options:
07be1b83 823
be8e71aa
YO
824 PARSE
825 PEEP
826 TRIE
827 PROGRAM
828 OFFSETS
a3621e74 829
be8e71aa 830 Execute Options:
a3621e74 831
be8e71aa
YO
832 INTUIT
833 MATCH
834 TRIE
a3621e74 835
be8e71aa 836 Extra Options
a3621e74 837
be8e71aa
YO
838 TRIE
839 OFFSETS
840
841If you modify any of these make sure you make corresponding changes to
842re.pm, especially to the documentation.
a3621e74 843
be8e71aa
YO
844*/
845
846
847/* Compile */
848#define RE_DEBUG_COMPILE_MASK 0x0000FF
849#define RE_DEBUG_COMPILE_PARSE 0x000001
850#define RE_DEBUG_COMPILE_OPTIMISE 0x000002
851#define RE_DEBUG_COMPILE_TRIE 0x000004
852#define RE_DEBUG_COMPILE_DUMP 0x000008
f7819f85 853#define RE_DEBUG_COMPILE_FLAGS 0x000010
d9a72fcc 854#define RE_DEBUG_COMPILE_TEST 0x000020
be8e71aa
YO
855
856/* Execute */
857#define RE_DEBUG_EXECUTE_MASK 0x00FF00
858#define RE_DEBUG_EXECUTE_INTUIT 0x000100
859#define RE_DEBUG_EXECUTE_MATCH 0x000200
860#define RE_DEBUG_EXECUTE_TRIE 0x000400
861
862/* Extra */
863#define RE_DEBUG_EXTRA_MASK 0xFF0000
864#define RE_DEBUG_EXTRA_TRIE 0x010000
865#define RE_DEBUG_EXTRA_OFFSETS 0x020000
a5ca303d
YO
866#define RE_DEBUG_EXTRA_OFFDEBUG 0x040000
867#define RE_DEBUG_EXTRA_STATE 0x080000
868#define RE_DEBUG_EXTRA_OPTIMISE 0x100000
e7707071 869#define RE_DEBUG_EXTRA_BUFFERS 0x400000
2c296965 870#define RE_DEBUG_EXTRA_GPOS 0x800000
24b23f37
YO
871/* combined */
872#define RE_DEBUG_EXTRA_STACK 0x280000
be8e71aa 873
e68ec53f 874#define RE_DEBUG_FLAG(x) (re_debug_flags & x)
be8e71aa
YO
875/* Compile */
876#define DEBUG_COMPILE_r(x) DEBUG_r( \
e68ec53f 877 if (re_debug_flags & RE_DEBUG_COMPILE_MASK) x )
be8e71aa 878#define DEBUG_PARSE_r(x) DEBUG_r( \
e68ec53f 879 if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x )
be8e71aa 880#define DEBUG_OPTIMISE_r(x) DEBUG_r( \
e68ec53f 881 if (re_debug_flags & RE_DEBUG_COMPILE_OPTIMISE) x )
be8e71aa 882#define DEBUG_PARSE_r(x) DEBUG_r( \
e68ec53f 883 if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x )
be8e71aa 884#define DEBUG_DUMP_r(x) DEBUG_r( \
e68ec53f 885 if (re_debug_flags & RE_DEBUG_COMPILE_DUMP) x )
be8e71aa 886#define DEBUG_TRIE_COMPILE_r(x) DEBUG_r( \
e68ec53f 887 if (re_debug_flags & RE_DEBUG_COMPILE_TRIE) x )
f7819f85
A
888#define DEBUG_FLAGS_r(x) DEBUG_r( \
889 if (re_debug_flags & RE_DEBUG_COMPILE_FLAGS) x )
d9a72fcc
YO
890#define DEBUG_TEST_r(x) DEBUG_r( \
891 if (re_debug_flags & RE_DEBUG_COMPILE_TEST) x )
be8e71aa
YO
892/* Execute */
893#define DEBUG_EXECUTE_r(x) DEBUG_r( \
e68ec53f 894 if (re_debug_flags & RE_DEBUG_EXECUTE_MASK) x )
be8e71aa 895#define DEBUG_INTUIT_r(x) DEBUG_r( \
e68ec53f 896 if (re_debug_flags & RE_DEBUG_EXECUTE_INTUIT) x )
be8e71aa 897#define DEBUG_MATCH_r(x) DEBUG_r( \
e68ec53f 898 if (re_debug_flags & RE_DEBUG_EXECUTE_MATCH) x )
be8e71aa 899#define DEBUG_TRIE_EXECUTE_r(x) DEBUG_r( \
e68ec53f 900 if (re_debug_flags & RE_DEBUG_EXECUTE_TRIE) x )
be8e71aa
YO
901
902/* Extra */
903#define DEBUG_EXTRA_r(x) DEBUG_r( \
e68ec53f 904 if (re_debug_flags & RE_DEBUG_EXTRA_MASK) x )
a5ca303d
YO
905#define DEBUG_OFFSETS_r(x) DEBUG_r( \
906 if (re_debug_flags & RE_DEBUG_EXTRA_OFFSETS) x )
ab3bbdeb
YO
907#define DEBUG_STATE_r(x) DEBUG_r( \
908 if (re_debug_flags & RE_DEBUG_EXTRA_STATE) x )
24b23f37
YO
909#define DEBUG_STACK_r(x) DEBUG_r( \
910 if (re_debug_flags & RE_DEBUG_EXTRA_STACK) x )
e7707071
YO
911#define DEBUG_BUFFERS_r(x) DEBUG_r( \
912 if (re_debug_flags & RE_DEBUG_EXTRA_BUFFERS) x )
913
a5ca303d
YO
914#define DEBUG_OPTIMISE_MORE_r(x) DEBUG_r( \
915 if ((RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE) == \
916 (re_debug_flags & (RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE)) ) x )
07be1b83 917#define MJD_OFFSET_DEBUG(x) DEBUG_r( \
a5ca303d 918 if (re_debug_flags & RE_DEBUG_EXTRA_OFFDEBUG) \
be8e71aa
YO
919 Perl_warn_nocontext x )
920#define DEBUG_TRIE_COMPILE_MORE_r(x) DEBUG_TRIE_COMPILE_r( \
e68ec53f 921 if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x )
be8e71aa 922#define DEBUG_TRIE_EXECUTE_MORE_r(x) DEBUG_TRIE_EXECUTE_r( \
e68ec53f 923 if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x )
be8e71aa
YO
924
925#define DEBUG_TRIE_r(x) DEBUG_r( \
e68ec53f 926 if (re_debug_flags & (RE_DEBUG_COMPILE_TRIE \
be8e71aa 927 | RE_DEBUG_EXECUTE_TRIE )) x )
2c296965
YO
928#define DEBUG_GPOS_r(x) DEBUG_r( \
929 if (re_debug_flags & RE_DEBUG_EXTRA_GPOS) x )
be8e71aa
YO
930
931/* initialization */
d2d8c711 932/* get_sv() can return NULL during global destruction. */
e68ec53f
YO
933#define GET_RE_DEBUG_FLAGS DEBUG_r({ \
934 SV * re_debug_flags_sv = NULL; \
935 re_debug_flags_sv = get_sv(RE_DEBUG_FLAGS, 1); \
936 if (re_debug_flags_sv) { \
8e9a8a48 937 if (!SvIOK(re_debug_flags_sv)) \
e68ec53f 938 sv_setuv(re_debug_flags_sv, RE_DEBUG_COMPILE_DUMP | RE_DEBUG_EXECUTE_MASK ); \
8e9a8a48 939 re_debug_flags=SvIV(re_debug_flags_sv); \
e68ec53f
YO
940 }\
941})
a3621e74
YO
942
943#ifdef DEBUGGING
ab3bbdeb 944
cb65b687
DM
945#define GET_RE_DEBUG_FLAGS_DECL VOL IV re_debug_flags = 0; \
946 PERL_UNUSED_VAR(re_debug_flags); GET_RE_DEBUG_FLAGS;
ab3bbdeb
YO
947
948#define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2) \
949 const char * const rpv = \
ddc5bc0f 950 pv_pretty((dsv), (pv), (l), (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) ); \
0df25f3d 953 const int rlen = SvCUR(dsv)
ab3bbdeb
YO
954
955#define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m) \
956 const char * const rpv = \
ddc5bc0f 957 pv_pretty((dsv), (SvPV_nolen_const(sv)), (SvCUR(sv)), (m), \
ab3bbdeb 958 PL_colors[(c1)],PL_colors[(c2)], \
c89df6cf 959 PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) )
ab3bbdeb
YO
960
961#define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m) \
962 const char * const rpv = \
ddc5bc0f 963 pv_pretty((dsv), (pv), (l), (m), \
ab3bbdeb 964 PL_colors[0], PL_colors[1], \
c89df6cf 965 ( PERL_PV_PRETTY_QUOTE | PERL_PV_ESCAPE_RE | PERL_PV_ESCAPE_NONASCII | PERL_PV_PRETTY_ELLIPSES | \
ab3bbdeb 966 ((isuni) ? PERL_PV_ESCAPE_UNI : 0)) \
0a0e9afc 967 )
ab3bbdeb
YO
968
969#define RE_SV_DUMPLEN(ItEm) (SvCUR(ItEm) - (SvTAIL(ItEm)!=0))
970#define RE_SV_TAIL(ItEm) (SvTAIL(ItEm) ? "$" : "")
971
972#else /* if not DEBUGGING */
973
a3621e74 974#define GET_RE_DEBUG_FLAGS_DECL
ab3bbdeb
YO
975#define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2)
976#define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m)
977#define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m)
978#define RE_SV_DUMPLEN(ItEm)
979#define RE_SV_TAIL(ItEm)
980
981#endif /* DEBUG RELATED DEFINES */
a3621e74 982
e9a8c099
MHM
983/*
984 * Local variables:
985 * c-indentation-style: bsd
986 * c-basic-offset: 4
14d04a33 987 * indent-tabs-mode: nil
e9a8c099
MHM
988 * End:
989 *
14d04a33 990 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 991 */