This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Rename an internal flag
[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 */
128#define PREGf_SKIP 0x00000001
129#define PREGf_IMPLICIT 0x00000002 /* Converted .* to ^.* */
130#define PREGf_NAUGHTY 0x00000004 /* how exponential is this pattern? */
131#define PREGf_VERBARG_SEEN 0x00000008
132#define PREGf_CUTGROUP_SEEN 0x00000010
732caac7 133#define PREGf_USE_RE_EVAL 0x00000020 /* compiled with "use re 'eval'" */
785a26d5
YO
134
135
136/* this is where the old regcomp.h started */
a687059c 137
c277df42 138struct regnode_string {
cd439c50 139 U8 str_len;
c277df42
IZ
140 U8 type;
141 U16 next_off;
cd439c50 142 char string[1];
c277df42 143};
a687059c 144
6bda09f9
YO
145/* Argument bearing node - workhorse,
146 arg1 is often for the data field */
c277df42
IZ
147struct regnode_1 {
148 U8 flags;
149 U8 type;
150 U16 next_off;
151 U32 arg1;
152};
153
6bda09f9
YO
154/* Similar to a regnode_1 but with an extra signed argument */
155struct regnode_2L {
156 U8 flags;
157 U8 type;
158 U16 next_off;
159 U32 arg1;
160 I32 arg2;
161};
162
163/* 'Two field' -- Two 16 bit unsigned args */
c277df42
IZ
164struct regnode_2 {
165 U8 flags;
166 U8 type;
167 U16 next_off;
168 U16 arg1;
169 U16 arg2;
170};
171
6bda09f9 172
936ed897 173#define ANYOF_BITMAP_SIZE 32 /* 256 b/(8 b/B) */
936ed897 174
786e8c11 175/* also used by trie */
936ed897
IZ
176struct regnode_charclass {
177 U8 flags;
178 U8 type;
179 U16 next_off;
bf3f2d85 180 U32 arg1; /* used as ptr in S_regclass */
19103b91 181 char bitmap[ANYOF_BITMAP_SIZE]; /* only compile-time */
936ed897
IZ
182};
183
bf3f2d85
KW
184/* has runtime (locale) \d, \w, ..., [:posix:] classes */
185struct regnode_charclass_class {
8efd3f97 186 U8 flags; /* ANYOF_POSIXL bit must go here */
936ed897
IZ
187 U8 type;
188 U16 next_off;
bf3f2d85 189 U32 arg1; /* used as ptr in S_regclass */
19103b91 190 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time */
9cf78fa0 191 U32 classflags; /* and run-time */
936ed897
IZ
192};
193
a0dd4231
KW
194/* Synthetic start class; is a regnode_charclass_class plus an SV*. Note that
195 * the 'next_off' field is unused, as the SSC stands alone, so there is never a
196 * next node. */
b8f7bb16 197struct regnode_ssc {
8efd3f97 198 U8 flags; /* ANYOF_POSIXL bit must go here */
b8f7bb16
KW
199 U8 type;
200 U16 next_off;
201 U32 arg1; /* used as ptr in S_regclass */
202 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time */
203 U32 classflags; /* and run-time */
204 SV* invlist; /* list of code points matched */
205};
206
71068078
KW
207/* We take advantage of 'next_off' not otherwise being used in the SSC by
208 * actually using it: by setting it to 1. This allows us to test and
209 * distinguish between an SSC and other ANYOF node types, as 'next_off' cannot
210 * otherwise be 1, because it is the offset to the next regnode expressed in
211 * units of regnodes. Since an ANYOF node contains extra fields, it adds up
212 * to 12 regnode units on 32-bit systems, (hence the minimum this can be (if
213 * not 0) is 11 there. Even if things get tightly packed on a 64-bit system,
214 * it still would be more than 1. */
215#define set_ANYOF_SYNTHETIC(n) STMT_START{ OP(n) = ANYOF; \
216 NEXT_OFF(n) = 1; \
217 } STMT_END
218#define is_ANYOF_SYNTHETIC(n) (OP(n) == ANYOF && NEXT_OFF(n) == 1)
219
9c7e81e8
AD
220/* XXX fix this description.
221 Impose a limit of REG_INFTY on various pattern matching operations
222 to limit stack growth and to avoid "infinite" recursions.
223*/
224/* The default size for REG_INFTY is I16_MAX, which is the same as
225 SHORT_MAX (see perl.h). Unfortunately I16 isn't necessarily 16 bits
226 (see handy.h). On the Cray C90, sizeof(short)==4 and hence I16_MAX is
227 ((1<<31)-1), while on the Cray T90, sizeof(short)==8 and I16_MAX is
228 ((1<<63)-1). To limit stack growth to reasonable sizes, supply a
229 smaller default.
230 --Andy Dougherty 11 June 1998
231*/
232#if SHORTSIZE > 2
233# ifndef REG_INFTY
234# define REG_INFTY ((1<<15)-1)
235# endif
236#endif
237
238#ifndef REG_INFTY
83cfe48c
IZ
239# define REG_INFTY I16_MAX
240#endif
a687059c 241
e91177ed
IZ
242#define ARG_VALUE(arg) (arg)
243#define ARG__SET(arg,val) ((arg) = (val))
c277df42 244
e8e3b0fb
JH
245#undef ARG
246#undef ARG1
247#undef ARG2
248
c277df42
IZ
249#define ARG(p) ARG_VALUE(ARG_LOC(p))
250#define ARG1(p) ARG_VALUE(ARG1_LOC(p))
251#define ARG2(p) ARG_VALUE(ARG2_LOC(p))
6bda09f9 252#define ARG2L(p) ARG_VALUE(ARG2L_LOC(p))
786e8c11 253
c277df42
IZ
254#define ARG_SET(p, val) ARG__SET(ARG_LOC(p), (val))
255#define ARG1_SET(p, val) ARG__SET(ARG1_LOC(p), (val))
256#define ARG2_SET(p, val) ARG__SET(ARG2_LOC(p), (val))
6bda09f9 257#define ARG2L_SET(p, val) ARG__SET(ARG2L_LOC(p), (val))
c277df42 258
e8e3b0fb
JH
259#undef NEXT_OFF
260#undef NODE_ALIGN
261
075abff3
AL
262#define NEXT_OFF(p) ((p)->next_off)
263#define NODE_ALIGN(node)
264#define NODE_ALIGN_FILL(node) ((node)->flags = 0xde) /* deadbeef */
a687059c 265
c277df42
IZ
266#define SIZE_ALIGN NODE_ALIGN
267
e8e3b0fb
JH
268#undef OP
269#undef OPERAND
270#undef MASK
271#undef STRING
272
e91177ed 273#define OP(p) ((p)->type)
b988e673
KW
274#define FLAGS(p) ((p)->flags) /* Caution: Doesn't apply to all \
275 regnode types. For some, it's the \
276 character set of the regnode */
e91177ed 277#define OPERAND(p) (((struct regnode_string *)p)->string)
cd439c50
IZ
278#define MASK(p) ((char*)OPERAND(p))
279#define STR_LEN(p) (((struct regnode_string *)p)->str_len)
280#define STRING(p) (((struct regnode_string *)p)->string)
281#define STR_SZ(l) ((l + sizeof(regnode) - 1) / sizeof(regnode))
282#define NODE_SZ_STR(p) (STR_SZ(STR_LEN(p))+1)
283
e8e3b0fb
JH
284#undef NODE_ALIGN
285#undef ARG_LOC
286#undef NEXTOPER
287#undef PREVOPER
288
e91177ed
IZ
289#define NODE_ALIGN(node)
290#define ARG_LOC(p) (((struct regnode_1 *)p)->arg1)
291#define ARG1_LOC(p) (((struct regnode_2 *)p)->arg1)
292#define ARG2_LOC(p) (((struct regnode_2 *)p)->arg2)
6bda09f9 293#define ARG2L_LOC(p) (((struct regnode_2L *)p)->arg2)
786e8c11 294
e91177ed
IZ
295#define NODE_STEP_REGNODE 1 /* sizeof(regnode)/sizeof(regnode) */
296#define EXTRA_STEP_2ARGS EXTRA_SIZE(struct regnode_2)
297
298#define NODE_STEP_B 4
c277df42
IZ
299
300#define NEXTOPER(p) ((p) + NODE_STEP_REGNODE)
301#define PREVOPER(p) ((p) - NODE_STEP_REGNODE)
302
e91177ed 303#define FILL_ADVANCE_NODE(ptr, op) STMT_START { \
c277df42 304 (ptr)->type = op; (ptr)->next_off = 0; (ptr)++; } STMT_END
e91177ed 305#define FILL_ADVANCE_NODE_ARG(ptr, op, arg) STMT_START { \
c277df42 306 ARG_SET(ptr, arg); FILL_ADVANCE_NODE(ptr, op); (ptr) += 1; } STMT_END
a687059c 307
22c35a8c 308#define REG_MAGIC 0234
a687059c 309
9a81a976 310#define SIZE_ONLY (RExC_emit == (regnode *) & RExC_emit_dummy)
6d15ebf3
KW
311#define PASS1 SIZE_ONLY
312#define PASS2 (! SIZE_ONLY)
c277df42 313
137165a6
KW
314/* If the bitmap doesn't fully represent what this ANYOF node can match, the
315 * ARG is set to this special value (since 0, 1, ... are legal, but will never
316 * reach this high). */
317#define ANYOF_NONBITMAP_EMPTY ((U32) -1)
318
319/* The information used to be stored as as combination of the ANYOF_UTF8 and
320 * ANYOF_NONBITMAP_NON_UTF8 bits in the flags field, but was moved out of there
321 * to free up a bit for other uses. This tries to hide the change from
322 * existing code as much as possible. Now, the data structure that goes in ARG
323 * is not allocated unless it is needed, and that is what is used to determine
324 * if there is something outside the bitmap. The code now assumes that if
325 * that structure exists, that any UTF-8 encoded string should be tried against
326 * it, but a non-UTF8-encoded string will be tried only if the
327 * ANYOF_NONBITMAP_NON_UTF8 bit is also set. */
328#define ANYOF_NONBITMAP(node) (ARG(node) != ANYOF_NONBITMAP_EMPTY)
329
ae986089
KW
330/* Flags for node->flags of ANYOF. These are in short supply, with none
331 * currently available. If more are needed, the ANYOF_LOCALE and
a0dd4231
KW
332 * ANYOF_POSIXL bits could be shared, making a space penalty for all locale
333 * nodes. Also, the ABOVE_LATIN1_ALL bit could be freed up by resorting to
334 * creating a swash containing everything above 255. This introduces a
335 * performance penalty. Better would be to split it off into a separate node,
336 * which actually would improve performance a bit by allowing regexec.c to test
337 * for a UTF-8 character being above 255 without having to call a function nor
71068078
KW
338 * calculate its code point value. Several flags are not used in synthetic
339 * start class (SSC) nodes, so could be shared should new flags be needed for
340 * SSCs. */
b8c5462f 341
5dbb0c08
KW
342/* regexec.c is expecting this to be in the low bit */
343#define ANYOF_INVERT 0x01
1984f6a0 344
a0dd4231
KW
345/* For the SSC node only, which cannot be inverted, so is shared with that bit.
346 * This means "Does this SSC match an empty string?" This is used only during
347 * regex compilation. */
348#define ANYOF_EMPTY_STRING ANYOF_INVERT
349
b63f73da
KW
350#define ANYOF_LOCALE 0x02 /* /l modifier */
351
352/* The fold is calculated and stored in the bitmap where possible at compile
353 * time. However under locale, the actual folding varies depending on
354 * what the locale is at the time of execution, so it has to be deferred until
355 * then */
356#define ANYOF_LOC_FOLD 0x04
357
d1d040e5 358/* Set if this is a regnode_charclass_posixl vs a regnode_charclass. This
dd58aee1
KW
359 * is used for runtime \d, \w, [:posix:], ..., which are used only in locale
360 * and the optimizer's synthetic start class. Non-locale \d, etc are resolved
9ffd0d41
KW
361 * at compile-time. Could be shared with ANYOF_LOCALE, forcing all locale
362 * nodes to be large */
8efd3f97
KW
363#define ANYOF_POSIXL 0x08
364#define ANYOF_CLASS ANYOF_POSIXL
365#define ANYOF_LARGE ANYOF_POSIXL
c8d3cd88 366
ae986089
KW
367/* Should we raise a warning if matching against an above-Unicode code point?
368 * */
369#define ANYOF_WARN_SUPER 0x10
f4244008 370
ef87b810
KW
371/* Can match something outside the bitmap that isn't in utf8 */
372#define ANYOF_NONBITMAP_NON_UTF8 0x20
373
61ef9872 374/* Matches every code point 0x100 and above*/
9b1eb067
KW
375#define ANYOF_ABOVE_LATIN1_ALL 0x40
376#define ANYOF_UNICODE_ALL ANYOF_ABOVE_LATIN1_ALL
ffc61ed2 377
11454c59
KW
378/* Match all Latin1 characters that aren't ASCII when the target string is not
379 * in utf8. */
51ca120f 380#define ANYOF_NON_UTF8_NON_ASCII_ALL 0x80
11454c59 381
ae986089 382#define ANYOF_FLAGS_ALL (0xff)
653099ff 383
fb38762f
KW
384#define ANYOF_LOCALE_FLAGS (ANYOF_LOCALE \
385 |ANYOF_LOC_FOLD \
386 |ANYOF_POSIXL)
ace6b0e4 387
eff8b7dc
KW
388/* These are the flags that apply to both regular ANYOF nodes and synthetic
389 * start class nodes during construction of the SSC. During finalization of
390 * the SSC, other of the flags could be added to it */
e9b08962 391#define ANYOF_COMMON_FLAGS (ANYOF_LOCALE_FLAGS | ANYOF_WARN_SUPER)
eff8b7dc 392
936ed897 393/* Character classes for node->classflags of ANYOF */
653099ff 394/* Should be synchronized with a table in regprop() */
575a8a29 395/* 2n should be the normal one, paired with its complement at 2n+1 */
b8c5462f 396
08ada29e
KW
397#define ANYOF_ALPHA ((_CC_ALPHA) * 2)
398#define ANYOF_NALPHA ((ANYOF_ALPHA) + 1)
e8d596e0
KW
399#define ANYOF_ALPHANUMERIC ((_CC_ALPHANUMERIC) * 2) /* [[:alnum:]] isalnum(3), utf8::IsAlnum */
400#define ANYOF_NALPHANUMERIC ((ANYOF_ALPHANUMERIC) + 1)
08ada29e
KW
401#define ANYOF_ASCII ((_CC_ASCII) * 2)
402#define ANYOF_NASCII ((ANYOF_ASCII) + 1)
e8d596e0
KW
403#define ANYOF_BLANK ((_CC_BLANK) * 2) /* GNU extension: space and tab: non-vertical space */
404#define ANYOF_NBLANK ((ANYOF_BLANK) + 1)
b0d691b2
KW
405#define ANYOF_CASED ((_CC_CASED) * 2) /* Pseudo class for [:lower:] or
406 [:upper:] under /i */
407#define ANYOF_NCASED ((ANYOF_CASED) + 1)
08ada29e
KW
408#define ANYOF_CNTRL ((_CC_CNTRL) * 2)
409#define ANYOF_NCNTRL ((ANYOF_CNTRL) + 1)
e8d596e0
KW
410#define ANYOF_DIGIT ((_CC_DIGIT) * 2) /* \d */
411#define ANYOF_NDIGIT ((ANYOF_DIGIT) + 1)
08ada29e
KW
412#define ANYOF_GRAPH ((_CC_GRAPH) * 2)
413#define ANYOF_NGRAPH ((ANYOF_GRAPH) + 1)
414#define ANYOF_LOWER ((_CC_LOWER) * 2)
415#define ANYOF_NLOWER ((ANYOF_LOWER) + 1)
416#define ANYOF_PRINT ((_CC_PRINT) * 2)
417#define ANYOF_NPRINT ((ANYOF_PRINT) + 1)
e8d596e0
KW
418#define ANYOF_PSXSPC ((_CC_PSXSPC) * 2) /* POSIX space: \s plus the vertical tab */
419#define ANYOF_NPSXSPC ((ANYOF_PSXSPC) + 1)
08ada29e
KW
420#define ANYOF_PUNCT ((_CC_PUNCT) * 2)
421#define ANYOF_NPUNCT ((ANYOF_PUNCT) + 1)
e8d596e0
KW
422#define ANYOF_SPACE ((_CC_SPACE) * 2) /* \s */
423#define ANYOF_NSPACE ((ANYOF_SPACE) + 1)
08ada29e
KW
424#define ANYOF_UPPER ((_CC_UPPER) * 2)
425#define ANYOF_NUPPER ((ANYOF_UPPER) + 1)
e8d596e0
KW
426#define ANYOF_WORDCHAR ((_CC_WORDCHAR) * 2) /* \w, PL_utf8_alnum, utf8::IsWord, ALNUM */
427#define ANYOF_NWORDCHAR ((ANYOF_WORDCHAR) + 1)
08ada29e
KW
428#define ANYOF_XDIGIT ((_CC_XDIGIT) * 2)
429#define ANYOF_NXDIGIT ((ANYOF_XDIGIT) + 1)
b8c5462f 430
87cad4c0 431/* pseudo classes below this, not stored in the class bitmap, but used as flags
e1d1eefb
YO
432 during compilation of char classes */
433
b2ef2e34
KW
434#define ANYOF_VERTWS ((_CC_VERTSPACE) * 2)
435#define ANYOF_NVERTWS ((ANYOF_VERTWS)+1)
e1d1eefb 436
b2ef2e34
KW
437/* It is best if this is the last one, as all above it are stored as bits in a
438 * bitmap, and it isn't part of that bitmap */
439#if _CC_VERTSPACE != _HIGHEST_REGCOMP_DOT_H_SYNC
440# error Problem with handy.h _HIGHEST_REGCOMP_DOT_H_SYNC #define
441#endif
442
8efd3f97
KW
443#define ANYOF_POSIXL_MAX (ANYOF_VERTWS) /* So upper loop limit is written:
444 * '< ANYOF_MAX'
445 * Hence doesn't include VERTWS, as that
446 * is a pseudo class */
447#define ANYOF_MAX ANYOF_POSIXL_MAX
448
449#if (ANYOF_POSIXL_MAX > 32) /* Must fit in 32-bit word */
b2ef2e34 450# error Problem with handy.h _CC_foo #defines
a0947d7b
KW
451#endif
452
8efd3f97
KW
453#define ANYOF_HORIZWS ((ANYOF_POSIXL_MAX)+2) /* = (ANYOF_NVERTWS + 1) */
454#define ANYOF_NHORIZWS ((ANYOF_POSIXL_MAX)+3)
87cad4c0 455
8efd3f97
KW
456#define ANYOF_UNIPROP ((ANYOF_POSIXL_MAX)+4) /* Used to indicate a Unicode
457 property: \p{} or \P{} */
f39a4dc2 458
b8c5462f
JH
459/* Backward source code compatibility. */
460
461#define ANYOF_ALNUML ANYOF_ALNUM
462#define ANYOF_NALNUML ANYOF_NALNUM
463#define ANYOF_SPACEL ANYOF_SPACE
464#define ANYOF_NSPACEL ANYOF_NSPACE
8ebfc29a
KW
465#define ANYOF_ALNUM ANYOF_WORDCHAR
466#define ANYOF_NALNUM ANYOF_NWORDCHAR
b8c5462f
JH
467
468/* Utility macros for the bitmap and classes of ANYOF */
469
936ed897 470#define ANYOF_SIZE (sizeof(struct regnode_charclass))
d1d040e5 471#define ANYOF_POSIXL_SIZE (sizeof(regnode_charclass_posixl))
8efd3f97 472#define ANYOF_CLASS_SIZE ANYOF_POSIXL_SIZE
b8c5462f 473
936ed897 474#define ANYOF_FLAGS(p) ((p)->flags)
b8c5462f
JH
475
476#define ANYOF_BIT(c) (1 << ((c) & 7))
477
d1d040e5 478#define ANYOF_POSIXL_SET(p, c) (((regnode_charclass_posixl*) (p))->classflags |= (1U << (c)))
8efd3f97
KW
479#define ANYOF_CLASS_SET(p, c) ANYOF_POSIXL_SET((p), (c))
480
d1d040e5 481#define ANYOF_POSIXL_CLEAR(p, c) (((regnode_charclass_posixl*) (p))->classflags &= ~ (1U <<(c)))
8efd3f97
KW
482#define ANYOF_CLASS_CLEAR(p, c) ANYOF_POSIXL_CLEAR((p), (c))
483
d1d040e5 484#define ANYOF_POSIXL_TEST(p, c) (((regnode_charclass_posixl*) (p))->classflags & (1U << (c)))
8efd3f97 485#define ANYOF_CLASS_TEST(p, c) ANYOF_POSIXL_TEST((p), (c))
b8c5462f 486
d1d040e5 487#define ANYOF_POSIXL_ZERO(ret) STMT_START { ((regnode_charclass_posixl*) (ret))->classflags = 0; } STMT_END
8efd3f97 488#define ANYOF_CLASS_ZERO(ret) ANYOF_POSIXL_ZERO(ret)
9cf78fa0
KW
489
490/* Shifts a bit to get, eg. 0x4000_0000, then subtracts 1 to get 0x3FFF_FFFF */
d1d040e5 491#define ANYOF_POSIXL_SETALL(ret) STMT_START { ((regnode_charclass_posixl*) (ret))->classflags = ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1; } STMT_END
8efd3f97 492#define ANYOF_CLASS_SETALL(ret) ANYOF_POSIXL_SETALL(ret)
9cf78fa0 493
8efd3f97
KW
494#define ANYOF_POSIXL_TEST_ANY_SET(p) \
495 ((ANYOF_FLAGS(p) & ANYOF_POSIXL) \
d1d040e5 496 && (((regnode_charclass_posixl*)(p))->classflags))
8efd3f97 497#define ANYOF_CLASS_TEST_ANY_SET(p) ANYOF_POSIXL_TEST_ANY_SET(p)
e867ce45 498
fb38762f
KW
499#define ANYOF_POSIXL_TEST_ALL_SET(p) \
500 ((ANYOF_FLAGS(p) & ANYOF_POSIXL) \
501 && ((regnode_charclass_posixl*) (p))->classflags == ((1U << ((ANYOF_POSIXL_MAX) - 1))) - 1)
502
b27fd97a 503#define ANYOF_POSIXL_OR(source, dest) STMT_START { (dest)->classflags |= (source)->classflags ; } STMT_END
8efd3f97 504#define ANYOF_CLASS_OR(source, dest) ANYOF_POSIXL_OR((source), (dest))
936ed897 505
cdd87c1d
KW
506#define ANYOF_POSIXL_AND(source, dest) STMT_START { (dest)->classflags &= (source)->classflags ; } STMT_END
507
9cf78fa0 508#define ANYOF_BITMAP_ZERO(ret) Zero(((struct regnode_charclass*)(ret))->bitmap, ANYOF_BITMAP_SIZE, char)
936ed897 509#define ANYOF_BITMAP(p) (((struct regnode_charclass*)(p))->bitmap)
786e8c11 510#define ANYOF_BITMAP_BYTE(p, c) (ANYOF_BITMAP(p)[(((U8)(c)) >> 3) & 31])
b8c5462f
JH
511#define ANYOF_BITMAP_SET(p, c) (ANYOF_BITMAP_BYTE(p, c) |= ANYOF_BIT(c))
512#define ANYOF_BITMAP_CLEAR(p,c) (ANYOF_BITMAP_BYTE(p, c) &= ~ANYOF_BIT(c))
513#define ANYOF_BITMAP_TEST(p, c) (ANYOF_BITMAP_BYTE(p, c) & ANYOF_BIT(c))
514
f8bef550
NC
515#define ANYOF_BITMAP_SETALL(p) \
516 memset (ANYOF_BITMAP(p), 255, ANYOF_BITMAP_SIZE)
517#define ANYOF_BITMAP_CLEARALL(p) \
518 Zero (ANYOF_BITMAP(p), ANYOF_BITMAP_SIZE)
519/* Check that all 256 bits are all set. Used in S_cl_is_anything() */
c355c09a 520#define ANYOF_BITMAP_TESTALLSET(p) /* Assumes sizeof(p) == 32 */ \
f8bef550
NC
521 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)
522
936ed897 523#define ANYOF_SKIP ((ANYOF_SIZE - 1)/sizeof(regnode))
8efd3f97
KW
524#define ANYOF_POSIXL_SKIP ((ANYOF_POSIXL_SIZE - 1)/sizeof(regnode))
525#define ANYOF_CLASS_SKIP ANYOF_POSIXL_SKIP
3a15e693 526
a687059c
LW
527/*
528 * Utility definitions.
529 */
2304df62 530#ifndef CHARMASK
075abff3 531# define UCHARAT(p) ((int)*(const U8*)(p))
a687059c 532#else
075abff3 533# define UCHARAT(p) ((int)*(p)&CHARMASK)
a687059c 534#endif
a687059c 535
c277df42
IZ
536#define EXTRA_SIZE(guy) ((sizeof(guy)-1)/sizeof(struct regnode))
537
3dab1dad
YO
538#define REG_SEEN_ZERO_LEN 0x00000001
539#define REG_SEEN_LOOKBEHIND 0x00000002
540#define REG_SEEN_GPOS 0x00000004
471f5387 541/* spare */
3dab1dad 542#define REG_SEEN_CANY 0x00000010
cce850e4 543#define REG_SEEN_SANY REG_SEEN_CANY /* src bckwrd cmpt */
6bda09f9 544#define REG_SEEN_RECURSE 0x00000020
eaf3ca90 545#define REG_TOP_LEVEL_BRANCHES 0x00000040
e2e6a0f1 546#define REG_SEEN_VERBARG 0x00000080
5d458dd8 547#define REG_SEEN_CUTGROUP 0x00000100
bcdf7404 548#define REG_SEEN_RUN_ON_COMMENT 0x00000200
bb914485 549#define REG_SEEN_EXACTF_SHARP_S 0x00000400
09a65838 550#define REG_SEEN_GOSTART 0x00000800
d09b2d29 551
73c4f7a1
GS
552START_EXTERN_C
553
be8e71aa
YO
554#ifdef PLUGGABLE_RE_EXTENSION
555#include "re_nodes.h"
556#else
d09b2d29 557#include "regnodes.h"
be8e71aa 558#endif
d09b2d29 559
f9f4320a
YO
560#ifndef PLUGGABLE_RE_EXTENSION
561#ifndef DOINIT
562EXTCONST regexp_engine PL_core_reg_engine;
563#else /* DOINIT */
564EXTCONST regexp_engine PL_core_reg_engine = {
fe578d7f 565 Perl_re_compile,
2fdbfb4d 566 Perl_regexec_flags,
f9f4320a
YO
567 Perl_re_intuit_start,
568 Perl_re_intuit_string,
2fdbfb4d
AB
569 Perl_regfree_internal,
570 Perl_reg_numbered_buff_fetch,
571 Perl_reg_numbered_buff_store,
572 Perl_reg_numbered_buff_length,
192b9cd1
AB
573 Perl_reg_named_buff,
574 Perl_reg_named_buff_iter,
49d7dfbc 575 Perl_reg_qr_package,
f9f4320a 576#if defined(USE_ITHREADS)
3c13cae6 577 Perl_regdupe_internal,
f9f4320a 578#endif
3c13cae6 579 Perl_re_op_compile
f9f4320a
YO
580};
581#endif /* DOINIT */
582#endif /* PLUGGABLE_RE_EXTENSION */
583
584
73c4f7a1 585END_EXTERN_C
cad2e5aa 586
cad2e5aa 587
261faec3
MJD
588/* .what is a character array with one character for each member of .data
589 * The character describes the function of the corresponding .data item:
af534a04 590 * a - AV for paren_name_list under DEBUGGING
0111c4fd 591 * f - start-class data for regstclass optimization
68e2671b 592 * l - start op for literal (?{EVAL}) item
d63c20f2 593 * L - start op for literal (?{EVAL}) item, with separate CV (qr//)
b30fcab9 594 * r - pointer to an embedded code-containing qr, e.g. /ab$qr/
38a44b82 595 * s - swash for Unicode-style character class, and the multicharacter
9e55ce06
JH
596 * strings resulting from casefolding the single-character entries
597 * in the character class
a3621e74 598 * t - trie struct
55eed653 599 * u - trie struct's widecharmap (a HV, so can't share, must dup)
2b8b4781 600 * also used for revcharmap and words under DEBUGGING
07be1b83 601 * T - aho-trie struct
81714fb9 602 * S - sv for named capture lookup
261faec3 603 * 20010712 mjd@plover.com
c3613ed7 604 * (Remember to update re_dup() and pregfree() if you add any items.)
261faec3 605 */
cad2e5aa
JH
606struct reg_data {
607 U32 count;
608 U8 *what;
609 void* data[1];
610};
611
a1cac82e
NC
612/* Code in S_to_utf8_substr() and S_to_byte_substr() in regexec.c accesses
613 anchored* and float* via array indexes 0 and 1. */
cad2e5aa 614#define anchored_substr substrs->data[0].substr
33b8afdf 615#define anchored_utf8 substrs->data[0].utf8_substr
cad2e5aa 616#define anchored_offset substrs->data[0].min_offset
1de06328
YO
617#define anchored_end_shift substrs->data[0].end_shift
618
cad2e5aa 619#define float_substr substrs->data[1].substr
33b8afdf 620#define float_utf8 substrs->data[1].utf8_substr
cad2e5aa
JH
621#define float_min_offset substrs->data[1].min_offset
622#define float_max_offset substrs->data[1].max_offset
1de06328
YO
623#define float_end_shift substrs->data[1].end_shift
624
cad2e5aa 625#define check_substr substrs->data[2].substr
33b8afdf 626#define check_utf8 substrs->data[2].utf8_substr
cad2e5aa
JH
627#define check_offset_min substrs->data[2].min_offset
628#define check_offset_max substrs->data[2].max_offset
1de06328 629#define check_end_shift substrs->data[2].end_shift
a3621e74 630
8d919b0a
FC
631#define RX_ANCHORED_SUBSTR(rx) (ReANY(rx)->anchored_substr)
632#define RX_ANCHORED_UTF8(rx) (ReANY(rx)->anchored_utf8)
633#define RX_FLOAT_SUBSTR(rx) (ReANY(rx)->float_substr)
634#define RX_FLOAT_UTF8(rx) (ReANY(rx)->float_utf8)
a3621e74
YO
635
636/* trie related stuff */
5d9a96ca 637
a3621e74
YO
638/* a transition record for the state machine. the
639 check field determines which state "owns" the
640 transition. the char the transition is for is
641 determined by offset from the owning states base
642 field. the next field determines which state
643 is to be transitioned to if any.
644*/
645struct _reg_trie_trans {
646 U32 next;
647 U32 check;
648};
649
650/* a transition list element for the list based representation */
651struct _reg_trie_trans_list_elem {
652 U16 forid;
653 U32 newstate;
654};
655typedef struct _reg_trie_trans_list_elem reg_trie_trans_le;
656
657/* a state for compressed nodes. base is an offset
658 into an array of reg_trie_trans array. If wordnum is
659 nonzero the state is accepting. if base is zero then
660 the state has no children (and will be accepting)
661*/
662struct _reg_trie_state {
663 U16 wordnum;
664 union {
665 U32 base;
666 reg_trie_trans_le* list;
667 } trans;
668};
669
2e64971a
DM
670/* info per word; indexed by wordnum */
671typedef struct {
672 U16 prev; /* previous word in acceptance chain; eg in
673 * zzz|abc|ab/ after matching the chars abc, the
674 * accepted word is #2, and the previous accepted
675 * word is #3 */
676 U32 len; /* how many chars long is this word? */
677 U32 accept; /* accept state for this word */
678} reg_trie_wordinfo;
a3621e74
YO
679
680
a3621e74
YO
681typedef struct _reg_trie_state reg_trie_state;
682typedef struct _reg_trie_trans reg_trie_trans;
683
684
685/* anything in here that needs to be freed later
23eab42c
NC
686 should be dealt with in pregfree.
687 refcount is first in both this and _reg_ac_data to allow a space
688 optimisation in Perl_regdupe. */
a3621e74 689struct _reg_trie_data {
23eab42c 690 U32 refcount; /* number of times this trie is referenced */
786e8c11
YO
691 U32 lasttrans; /* last valid transition element */
692 U16 *charmap; /* byte to charid lookup array */
786e8c11
YO
693 reg_trie_state *states; /* state data */
694 reg_trie_trans *trans; /* array of transition elements */
695 char *bitmap; /* stclass bitmap */
786e8c11
YO
696 U16 *jump; /* optional 1 indexed array of offsets before tail
697 for the node following a given word. */
2e64971a 698 reg_trie_wordinfo *wordinfo; /* array of info per word */
1b5c067c
NC
699 U16 uniquecharcount; /* unique chars in trie (width of trans table) */
700 U32 startstate; /* initial state - used for common prefix optimisation */
701 STRLEN minlen; /* minimum length of words in trie - build/opt only? */
702 STRLEN maxlen; /* maximum length of words in trie - build/opt only? */
2e64971a 703 U32 prefixlen; /* #chars in common prefix */
1e2e3d02
YO
704 U32 statecount; /* Build only - number of states in the states array
705 (including the unused zero state) */
786e8c11 706 U32 wordcount; /* Build only */
a3621e74 707#ifdef DEBUGGING
786e8c11 708 STRLEN charcount; /* Build only */
a3621e74
YO
709#endif
710};
2b8b4781
NC
711/* There is one (3 under DEBUGGING) pointers that logically belong in this
712 structure, but are held outside as they need duplication on thread cloning,
713 whereas the rest of the structure can be read only:
714 HV *widecharmap; code points > 255 to charid
715#ifdef DEBUGGING
716 AV *words; Array of words contained in trie, for dumping
717 AV *revcharmap; Map of each charid back to its character representation
718#endif
719*/
720
721#define TRIE_WORDS_OFFSET 2
722
a3621e74
YO
723typedef struct _reg_trie_data reg_trie_data;
724
23eab42c
NC
725/* refcount is first in both this and _reg_trie_data to allow a space
726 optimisation in Perl_regdupe. */
07be1b83 727struct _reg_ac_data {
23eab42c 728 U32 refcount;
1b5c067c 729 U32 trie;
07be1b83
YO
730 U32 *fail;
731 reg_trie_state *states;
07be1b83
YO
732};
733typedef struct _reg_ac_data reg_ac_data;
734
486ec47a 735/* ANY_BIT doesn't use the structure, so we can borrow it here.
3dab1dad
YO
736 This is simpler than refactoring all of it as wed end up with
737 three different sets... */
738
739#define TRIE_BITMAP(p) (((reg_trie_data *)(p))->bitmap)
786e8c11 740#define TRIE_BITMAP_BYTE(p, c) (TRIE_BITMAP(p)[(((U8)(c)) >> 3) & 31])
5dfe062f
YO
741#define TRIE_BITMAP_SET(p, c) (TRIE_BITMAP_BYTE(p, c) |= ANYOF_BIT((U8)c))
742#define TRIE_BITMAP_CLEAR(p,c) (TRIE_BITMAP_BYTE(p, c) &= ~ANYOF_BIT((U8)c))
743#define TRIE_BITMAP_TEST(p, c) (TRIE_BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
3dab1dad 744
1de06328
YO
745#define IS_ANYOF_TRIE(op) ((op)==TRIEC || (op)==AHOCORASICKC)
746#define IS_TRIE_AC(op) ((op)>=AHOCORASICK)
747
748
786e8c11
YO
749#define BITMAP_BYTE(p, c) (((U8*)p)[(((U8)(c)) >> 3) & 31])
750#define BITMAP_TEST(p, c) (BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c))
3dab1dad 751
a3621e74
YO
752/* these defines assume uniquecharcount is the correct variable, and state may be evaluated twice */
753#define TRIE_NODENUM(state) (((state)-1)/(trie->uniquecharcount)+1)
754#define SAFE_TRIE_NODENUM(state) ((state) ? (((state)-1)/(trie->uniquecharcount)+1) : (state))
755#define TRIE_NODEIDX(state) ((state) ? (((state)-1)*(trie->uniquecharcount)+1) : (state))
756
3dab1dad 757#ifdef DEBUGGING
3dab1dad 758#define TRIE_CHARCOUNT(trie) ((trie)->charcount)
3dab1dad 759#else
3dab1dad 760#define TRIE_CHARCOUNT(trie) (trie_charcount)
3dab1dad 761#endif
a3621e74 762
0111c4fd
RGS
763#define RE_TRIE_MAXBUF_INIT 65536
764#define RE_TRIE_MAXBUF_NAME "\022E_TRIE_MAXBUF"
a3621e74
YO
765#define RE_DEBUG_FLAGS "\022E_DEBUG_FLAGS"
766
be8e71aa
YO
767/*
768
769RE_DEBUG_FLAGS is used to control what debug output is emitted
770its divided into three groups of options, some of which interact.
771The three groups are: Compile, Execute, Extra. There is room for a
772further group, as currently only the low three bytes are used.
773
774 Compile Options:
07be1b83 775
be8e71aa
YO
776 PARSE
777 PEEP
778 TRIE
779 PROGRAM
780 OFFSETS
a3621e74 781
be8e71aa 782 Execute Options:
a3621e74 783
be8e71aa
YO
784 INTUIT
785 MATCH
786 TRIE
a3621e74 787
be8e71aa 788 Extra Options
a3621e74 789
be8e71aa
YO
790 TRIE
791 OFFSETS
792
793If you modify any of these make sure you make corresponding changes to
794re.pm, especially to the documentation.
a3621e74 795
be8e71aa
YO
796*/
797
798
799/* Compile */
800#define RE_DEBUG_COMPILE_MASK 0x0000FF
801#define RE_DEBUG_COMPILE_PARSE 0x000001
802#define RE_DEBUG_COMPILE_OPTIMISE 0x000002
803#define RE_DEBUG_COMPILE_TRIE 0x000004
804#define RE_DEBUG_COMPILE_DUMP 0x000008
f7819f85 805#define RE_DEBUG_COMPILE_FLAGS 0x000010
be8e71aa
YO
806
807/* Execute */
808#define RE_DEBUG_EXECUTE_MASK 0x00FF00
809#define RE_DEBUG_EXECUTE_INTUIT 0x000100
810#define RE_DEBUG_EXECUTE_MATCH 0x000200
811#define RE_DEBUG_EXECUTE_TRIE 0x000400
812
813/* Extra */
814#define RE_DEBUG_EXTRA_MASK 0xFF0000
815#define RE_DEBUG_EXTRA_TRIE 0x010000
816#define RE_DEBUG_EXTRA_OFFSETS 0x020000
a5ca303d
YO
817#define RE_DEBUG_EXTRA_OFFDEBUG 0x040000
818#define RE_DEBUG_EXTRA_STATE 0x080000
819#define RE_DEBUG_EXTRA_OPTIMISE 0x100000
e7707071 820#define RE_DEBUG_EXTRA_BUFFERS 0x400000
2c296965 821#define RE_DEBUG_EXTRA_GPOS 0x800000
24b23f37
YO
822/* combined */
823#define RE_DEBUG_EXTRA_STACK 0x280000
be8e71aa 824
e68ec53f 825#define RE_DEBUG_FLAG(x) (re_debug_flags & x)
be8e71aa
YO
826/* Compile */
827#define DEBUG_COMPILE_r(x) DEBUG_r( \
e68ec53f 828 if (re_debug_flags & RE_DEBUG_COMPILE_MASK) x )
be8e71aa 829#define DEBUG_PARSE_r(x) DEBUG_r( \
e68ec53f 830 if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x )
be8e71aa 831#define DEBUG_OPTIMISE_r(x) DEBUG_r( \
e68ec53f 832 if (re_debug_flags & RE_DEBUG_COMPILE_OPTIMISE) x )
be8e71aa 833#define DEBUG_PARSE_r(x) DEBUG_r( \
e68ec53f 834 if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x )
be8e71aa 835#define DEBUG_DUMP_r(x) DEBUG_r( \
e68ec53f 836 if (re_debug_flags & RE_DEBUG_COMPILE_DUMP) x )
be8e71aa 837#define DEBUG_TRIE_COMPILE_r(x) DEBUG_r( \
e68ec53f 838 if (re_debug_flags & RE_DEBUG_COMPILE_TRIE) x )
f7819f85
A
839#define DEBUG_FLAGS_r(x) DEBUG_r( \
840 if (re_debug_flags & RE_DEBUG_COMPILE_FLAGS) x )
be8e71aa
YO
841/* Execute */
842#define DEBUG_EXECUTE_r(x) DEBUG_r( \
e68ec53f 843 if (re_debug_flags & RE_DEBUG_EXECUTE_MASK) x )
be8e71aa 844#define DEBUG_INTUIT_r(x) DEBUG_r( \
e68ec53f 845 if (re_debug_flags & RE_DEBUG_EXECUTE_INTUIT) x )
be8e71aa 846#define DEBUG_MATCH_r(x) DEBUG_r( \
e68ec53f 847 if (re_debug_flags & RE_DEBUG_EXECUTE_MATCH) x )
be8e71aa 848#define DEBUG_TRIE_EXECUTE_r(x) DEBUG_r( \
e68ec53f 849 if (re_debug_flags & RE_DEBUG_EXECUTE_TRIE) x )
be8e71aa
YO
850
851/* Extra */
852#define DEBUG_EXTRA_r(x) DEBUG_r( \
e68ec53f 853 if (re_debug_flags & RE_DEBUG_EXTRA_MASK) x )
a5ca303d
YO
854#define DEBUG_OFFSETS_r(x) DEBUG_r( \
855 if (re_debug_flags & RE_DEBUG_EXTRA_OFFSETS) x )
ab3bbdeb
YO
856#define DEBUG_STATE_r(x) DEBUG_r( \
857 if (re_debug_flags & RE_DEBUG_EXTRA_STATE) x )
24b23f37
YO
858#define DEBUG_STACK_r(x) DEBUG_r( \
859 if (re_debug_flags & RE_DEBUG_EXTRA_STACK) x )
e7707071
YO
860#define DEBUG_BUFFERS_r(x) DEBUG_r( \
861 if (re_debug_flags & RE_DEBUG_EXTRA_BUFFERS) x )
862
a5ca303d
YO
863#define DEBUG_OPTIMISE_MORE_r(x) DEBUG_r( \
864 if ((RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE) == \
865 (re_debug_flags & (RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE)) ) x )
07be1b83 866#define MJD_OFFSET_DEBUG(x) DEBUG_r( \
a5ca303d 867 if (re_debug_flags & RE_DEBUG_EXTRA_OFFDEBUG) \
be8e71aa
YO
868 Perl_warn_nocontext x )
869#define DEBUG_TRIE_COMPILE_MORE_r(x) DEBUG_TRIE_COMPILE_r( \
e68ec53f 870 if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x )
be8e71aa 871#define DEBUG_TRIE_EXECUTE_MORE_r(x) DEBUG_TRIE_EXECUTE_r( \
e68ec53f 872 if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x )
be8e71aa
YO
873
874#define DEBUG_TRIE_r(x) DEBUG_r( \
e68ec53f 875 if (re_debug_flags & (RE_DEBUG_COMPILE_TRIE \
be8e71aa 876 | RE_DEBUG_EXECUTE_TRIE )) x )
2c296965
YO
877#define DEBUG_GPOS_r(x) DEBUG_r( \
878 if (re_debug_flags & RE_DEBUG_EXTRA_GPOS) x )
be8e71aa
YO
879
880/* initialization */
d2d8c711 881/* get_sv() can return NULL during global destruction. */
e68ec53f
YO
882#define GET_RE_DEBUG_FLAGS DEBUG_r({ \
883 SV * re_debug_flags_sv = NULL; \
884 re_debug_flags_sv = get_sv(RE_DEBUG_FLAGS, 1); \
885 if (re_debug_flags_sv) { \
8e9a8a48 886 if (!SvIOK(re_debug_flags_sv)) \
e68ec53f 887 sv_setuv(re_debug_flags_sv, RE_DEBUG_COMPILE_DUMP | RE_DEBUG_EXECUTE_MASK ); \
8e9a8a48 888 re_debug_flags=SvIV(re_debug_flags_sv); \
e68ec53f
YO
889 }\
890})
a3621e74
YO
891
892#ifdef DEBUGGING
ab3bbdeb 893
cb65b687
DM
894#define GET_RE_DEBUG_FLAGS_DECL VOL IV re_debug_flags = 0; \
895 PERL_UNUSED_VAR(re_debug_flags); GET_RE_DEBUG_FLAGS;
ab3bbdeb
YO
896
897#define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2) \
898 const char * const rpv = \
ddc5bc0f 899 pv_pretty((dsv), (pv), (l), (m), \
ab3bbdeb 900 PL_colors[(c1)],PL_colors[(c2)], \
c89df6cf 901 PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) ); \
0df25f3d 902 const int rlen = SvCUR(dsv)
ab3bbdeb
YO
903
904#define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m) \
905 const char * const rpv = \
ddc5bc0f 906 pv_pretty((dsv), (SvPV_nolen_const(sv)), (SvCUR(sv)), (m), \
ab3bbdeb 907 PL_colors[(c1)],PL_colors[(c2)], \
c89df6cf 908 PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) )
ab3bbdeb
YO
909
910#define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m) \
911 const char * const rpv = \
ddc5bc0f 912 pv_pretty((dsv), (pv), (l), (m), \
ab3bbdeb 913 PL_colors[0], PL_colors[1], \
c89df6cf 914 ( PERL_PV_PRETTY_QUOTE | PERL_PV_ESCAPE_RE | PERL_PV_ESCAPE_NONASCII | PERL_PV_PRETTY_ELLIPSES | \
ab3bbdeb 915 ((isuni) ? PERL_PV_ESCAPE_UNI : 0)) \
0a0e9afc 916 )
ab3bbdeb
YO
917
918#define RE_SV_DUMPLEN(ItEm) (SvCUR(ItEm) - (SvTAIL(ItEm)!=0))
919#define RE_SV_TAIL(ItEm) (SvTAIL(ItEm) ? "$" : "")
920
921#else /* if not DEBUGGING */
922
a3621e74 923#define GET_RE_DEBUG_FLAGS_DECL
ab3bbdeb
YO
924#define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2)
925#define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m)
926#define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m)
927#define RE_SV_DUMPLEN(ItEm)
928#define RE_SV_TAIL(ItEm)
929
930#endif /* DEBUG RELATED DEFINES */
a3621e74 931
e9a8c099
MHM
932/*
933 * Local variables:
934 * c-indentation-style: bsd
935 * c-basic-offset: 4
14d04a33 936 * indent-tabs-mode: nil
e9a8c099
MHM
937 * End:
938 *
14d04a33 939 * ex: set ts=8 sts=4 sw=4 et:
e9a8c099 940 */