Commit | Line | Data |
---|---|---|
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 | */ |
3dd7db29 | 10 | |
21d0beeb KW |
11 | #if ! defined(PERL_REGCOMP_H_) && ( defined(PERL_CORE) \ |
12 | || defined(PERL_EXT_RE_BUILD)) | |
3dd7db29 JK |
13 | #define PERL_REGCOMP_H_ |
14 | ||
e1d1eefb | 15 | #include "regcharclass.h" |
a687059c | 16 | |
58e23c8d | 17 | /* Convert branch sequences to more efficient trie ops? */ |
07be1b83 | 18 | #define PERL_ENABLE_TRIE_OPTIMISATION 1 |
58e23c8d | 19 | |
486ec47a | 20 | /* Be really aggressive about optimising patterns with trie sequences? */ |
07be1b83 | 21 | #define PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION 1 |
58e23c8d YO |
22 | |
23 | /* Should the optimiser take positive assertions into account? */ | |
68ba3a3f | 24 | #define PERL_ENABLE_POSITIVE_ASSERTION_STUDY 0 |
58e23c8d YO |
25 | |
26 | /* Not for production use: */ | |
07be1b83 | 27 | #define PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS 0 |
58e23c8d | 28 | |
a687059c | 29 | /* |
a687059c LW |
30 | * Structure for regexp "program". This is essentially a linear encoding |
31 | * of a nondeterministic finite-state machine (aka syntax charts or | |
32 | * "railroad normal form" in parsing technology). Each node is an opcode | |
33 | * plus a "next" pointer, possibly plus an operand. "Next" pointers of | |
34 | * all nodes except BRANCH implement concatenation; a "next" pointer with | |
35 | * a BRANCH on both ends of it is connecting two alternatives. (Here we | |
36 | * have one of the subtle syntax dependencies: an individual BRANCH (as | |
37 | * opposed to a collection of them) is never concatenated with anything | |
38 | * because of operator precedence.) The operand of some types of node is | |
39 | * a literal string; for others, it is a node leading into a sub-FSM. In | |
40 | * particular, the operand of a BRANCH node is the first node of the branch. | |
41 | * (NB this is *not* a tree structure: the tail of the branch connects | |
a3621e74 YO |
42 | * to the thing following the set of BRANCHes.) The opcodes are defined |
43 | * in regnodes.h which is generated from regcomp.sym by regcomp.pl. | |
a687059c LW |
44 | */ |
45 | ||
a687059c LW |
46 | /* |
47 | * A node is one char of opcode followed by two chars of "next" pointer. | |
48 | * "Next" pointers are stored as two 8-bit pieces, high order first. The | |
49 | * value is a positive offset from the opcode of the node containing it. | |
50 | * An operand, if any, simply follows the node. (Note that much of the | |
51 | * code generation knows about this implicit relationship.) | |
52 | * | |
53 | * Using two bytes for the "next" pointer is vast overkill for most things, | |
54 | * but allows patterns to get big without disasters. | |
55 | * | |
e91177ed | 56 | * [The "next" pointer is always aligned on an even |
62e6ef33 | 57 | * boundary, and reads the offset directly as a short.] |
a687059c | 58 | */ |
785a26d5 YO |
59 | |
60 | /* This is the stuff that used to live in regexp.h that was truly | |
61 | private to the engine itself. It now lives here. */ | |
62 | ||
c45b4541 | 63 | typedef struct regexp_internal { |
6d5c990f RGS |
64 | regnode *regstclass; /* Optional startclass as identified or constructed |
65 | by the optimiser */ | |
66 | struct reg_data *data; /* Additional miscellaneous data used by the program. | |
67 | Used to make it easier to clone and free arbitrary | |
68 | data that the regops need. Often the ARG field of | |
c45b4541 YO |
69 | a regop is an index into this structure. NOTE the |
70 | 0th element of this structure is NEVER used and is | |
71 | strictly reserved for internal purposes. */ | |
1604cfb0 | 72 | struct reg_code_blocks *code_blocks;/* positions of literal (?{}) */ |
c45b4541 | 73 | U32 proglen; /* size of the compiled program in regnodes */ |
f08cf407 YO |
74 | U32 name_list_idx; /* Optional data index of an array of paren names, |
75 | only valid when RXp_PAREN_NAMES(prog) is true, | |
76 | 0 means "no value" like any other index into the | |
77 | data array.*/ | |
1604cfb0 | 78 | regnode program[1]; /* Unwarranted chumminess with compiler. */ |
6d5c990f RGS |
79 | } regexp_internal; |
80 | ||
81 | #define RXi_SET(x,y) (x)->pprivate = (void*)(y) | |
82 | #define RXi_GET(x) ((regexp_internal *)((x)->pprivate)) | |
83 | #define RXi_GET_DECL(r,ri) regexp_internal *ri = RXi_GET(r) | |
785a26d5 YO |
84 | /* |
85 | * Flags stored in regexp->intflags | |
86 | * These are used only internally to the regexp engine | |
87 | * | |
88 | * See regexp.h for flags used externally to the regexp engine | |
89 | */ | |
e3e400ec YO |
90 | #define RXp_INTFLAGS(rx) ((rx)->intflags) |
91 | #define RX_INTFLAGS(prog) RXp_INTFLAGS(ReANY(prog)) | |
92 | ||
785a26d5 YO |
93 | #define PREGf_SKIP 0x00000001 |
94 | #define PREGf_IMPLICIT 0x00000002 /* Converted .* to ^.* */ | |
95 | #define PREGf_NAUGHTY 0x00000004 /* how exponential is this pattern? */ | |
96 | #define PREGf_VERBARG_SEEN 0x00000008 | |
97 | #define PREGf_CUTGROUP_SEEN 0x00000010 | |
732caac7 | 98 | #define PREGf_USE_RE_EVAL 0x00000020 /* compiled with "use re 'eval'" */ |
e3e400ec YO |
99 | /* these used to be extflags, but are now intflags */ |
100 | #define PREGf_NOSCAN 0x00000040 | |
33c28ab2 | 101 | /* spare */ |
8e1490ee YO |
102 | #define PREGf_GPOS_SEEN 0x00000100 |
103 | #define PREGf_GPOS_FLOAT 0x00000200 | |
785a26d5 | 104 | |
d3d47aac YO |
105 | #define PREGf_ANCH_MBOL 0x00000400 |
106 | #define PREGf_ANCH_SBOL 0x00000800 | |
107 | #define PREGf_ANCH_GPOS 0x00001000 | |
d5a00e4a | 108 | #define PREGf_RECURSE_SEEN 0x00002000 |
8e1490ee | 109 | |
d3d47aac YO |
110 | #define PREGf_ANCH \ |
111 | ( PREGf_ANCH_SBOL | PREGf_ANCH_GPOS | PREGf_ANCH_MBOL ) | |
785a26d5 YO |
112 | |
113 | /* this is where the old regcomp.h started */ | |
a687059c | 114 | |
c277df42 | 115 | struct regnode_string { |
cd439c50 | 116 | U8 str_len; |
c277df42 IZ |
117 | U8 type; |
118 | U16 next_off; | |
cd439c50 | 119 | char string[1]; |
c277df42 | 120 | }; |
a687059c | 121 | |
ae06e581 KW |
122 | struct regnode_lstring { /* Constructed this way to keep the string aligned. */ |
123 | U8 flags; | |
124 | U8 type; | |
125 | U16 next_off; | |
1d425dad | 126 | U32 str_len; /* Only 18 bits allowed before would overflow 'next_off' */ |
ae06e581 KW |
127 | char string[1]; |
128 | }; | |
129 | ||
34924db0 | 130 | struct regnode_anyofhs { /* Constructed this way to keep the string aligned. */ |
53d42e43 | 131 | U8 str_len; |
34924db0 KW |
132 | U8 type; |
133 | U16 next_off; | |
134 | U32 arg1; /* set by set_ANYOF_arg() */ | |
135 | char string[1]; | |
136 | }; | |
137 | ||
6bda09f9 YO |
138 | /* Argument bearing node - workhorse, |
139 | arg1 is often for the data field */ | |
c277df42 IZ |
140 | struct regnode_1 { |
141 | U8 flags; | |
142 | U8 type; | |
143 | U16 next_off; | |
144 | U32 arg1; | |
145 | }; | |
146 | ||
d3e3ece8 KW |
147 | /* Node whose argument is 'SV *'. This needs to be used very carefully in |
148 | * situations where pointers won't become invalid because of, say re-mallocs */ | |
86451f01 KW |
149 | struct regnode_p { |
150 | U8 flags; | |
151 | U8 type; | |
152 | U16 next_off; | |
d3e3ece8 | 153 | SV * arg1; |
86451f01 KW |
154 | }; |
155 | ||
6bda09f9 YO |
156 | /* Similar to a regnode_1 but with an extra signed argument */ |
157 | struct regnode_2L { | |
158 | U8 flags; | |
159 | U8 type; | |
160 | U16 next_off; | |
161 | U32 arg1; | |
162 | I32 arg2; | |
163 | }; | |
164 | ||
165 | /* 'Two field' -- Two 16 bit unsigned args */ | |
c277df42 IZ |
166 | struct regnode_2 { |
167 | U8 flags; | |
168 | U8 type; | |
169 | U16 next_off; | |
170 | U16 arg1; | |
171 | U16 arg2; | |
172 | }; | |
173 | ||
dcb20b36 | 174 | #define ANYOF_BITMAP_SIZE (NUM_ANYOF_CODE_POINTS / 8) /* 8 bits/Byte */ |
936ed897 | 175 | |
1ee208c4 KW |
176 | /* Note that these form structs which are supersets of the next smaller one, by |
177 | * appending fields. Alignment problems can occur if one of those optional | |
178 | * fields requires stricter alignment than the base struct. And formal | |
179 | * parameters that can really be two or more of the structs should be | |
180 | * declared as the smallest one it could be. See commit message for | |
181 | * 7dcac5f6a5195002b55c935ee1d67f67e1df280b. Regnode allocation is done | |
182 | * without regard to alignment, and changing it to would also require changing | |
183 | * the code that inserts and deletes regnodes. The basic single-argument | |
184 | * regnode has a U32, which is what reganode() allocates as a unit. Therefore | |
185 | * no field can require stricter alignment than U32. */ | |
186 | ||
786e8c11 | 187 | /* also used by trie */ |
936ed897 IZ |
188 | struct regnode_charclass { |
189 | U8 flags; | |
190 | U8 type; | |
191 | U16 next_off; | |
9da99c5e | 192 | U32 arg1; /* set by set_ANYOF_arg() */ |
19103b91 | 193 | char bitmap[ANYOF_BITMAP_SIZE]; /* only compile-time */ |
936ed897 IZ |
194 | }; |
195 | ||
bf3f2d85 | 196 | /* has runtime (locale) \d, \w, ..., [:posix:] classes */ |
ea461dd6 | 197 | struct regnode_charclass_posixl { |
93e92956 | 198 | U8 flags; /* ANYOF_MATCHES_POSIXL bit must go here */ |
936ed897 IZ |
199 | U8 type; |
200 | U16 next_off; | |
7e817188 | 201 | U32 arg1; |
03fa83ba | 202 | char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time ... */ |
9cf78fa0 | 203 | U32 classflags; /* and run-time */ |
936ed897 IZ |
204 | }; |
205 | ||
85c8e306 KW |
206 | /* A synthetic start class (SSC); is a regnode_charclass_posixl_fold, plus an |
207 | * extra SV*, used only during its construction and which is not used by | |
208 | * regexec.c. Note that the 'next_off' field is unused, as the SSC stands | |
209 | * alone, so there is never a next node. Also, there is no alignment issue, | |
b8fda935 | 210 | * because these are declared or allocated as a complete unit so the compiler |
85c8e306 KW |
211 | * takes care of alignment. This is unlike the other regnodes which are |
212 | * allocated in terms of multiples of a single-argument regnode. SSC nodes can | |
213 | * have a pointer field because there is no alignment issue, and because it is | |
214 | * set to NULL after construction, before any cloning of the pattern */ | |
b8f7bb16 | 215 | struct regnode_ssc { |
93e92956 | 216 | U8 flags; /* ANYOF_MATCHES_POSIXL bit must go here */ |
b8f7bb16 KW |
217 | U8 type; |
218 | U16 next_off; | |
7e817188 | 219 | U32 arg1; |
03fa83ba | 220 | char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time ... */ |
631e9161 | 221 | U32 classflags; /* ... and run-time */ |
85c8e306 KW |
222 | |
223 | /* Auxiliary, only used during construction; NULL afterwards: list of code | |
224 | * points matched */ | |
225 | SV* invlist; | |
b8f7bb16 KW |
226 | }; |
227 | ||
71068078 KW |
228 | /* We take advantage of 'next_off' not otherwise being used in the SSC by |
229 | * actually using it: by setting it to 1. This allows us to test and | |
230 | * distinguish between an SSC and other ANYOF node types, as 'next_off' cannot | |
231 | * otherwise be 1, because it is the offset to the next regnode expressed in | |
232 | * units of regnodes. Since an ANYOF node contains extra fields, it adds up | |
233 | * to 12 regnode units on 32-bit systems, (hence the minimum this can be (if | |
234 | * not 0) is 11 there. Even if things get tightly packed on a 64-bit system, | |
235 | * it still would be more than 1. */ | |
236 | #define set_ANYOF_SYNTHETIC(n) STMT_START{ OP(n) = ANYOF; \ | |
237 | NEXT_OFF(n) = 1; \ | |
238 | } STMT_END | |
c1071334 | 239 | #define is_ANYOF_SYNTHETIC(n) (PL_regkind[OP(n)] == ANYOF && NEXT_OFF(n) == 1) |
71068078 | 240 | |
9c7e81e8 AD |
241 | /* XXX fix this description. |
242 | Impose a limit of REG_INFTY on various pattern matching operations | |
243 | to limit stack growth and to avoid "infinite" recursions. | |
244 | */ | |
4288c5b9 KW |
245 | /* The default size for REG_INFTY is U16_MAX, which is the same as |
246 | USHORT_MAX (see perl.h). Unfortunately U16 isn't necessarily 16 bits | |
247 | (see handy.h). On the Cray C90, sizeof(short)==4 and hence U16_MAX is | |
248 | ((1<<32)-1), while on the Cray T90, sizeof(short)==8 and U16_MAX is | |
249 | ((1<<64)-1). To limit stack growth to reasonable sizes, supply a | |
9c7e81e8 | 250 | smaller default. |
1604cfb0 | 251 | --Andy Dougherty 11 June 1998 |
9c7e81e8 AD |
252 | */ |
253 | #if SHORTSIZE > 2 | |
254 | # ifndef REG_INFTY | |
1381ccb1 | 255 | # define REG_INFTY nBIT_UMAX(16) |
9c7e81e8 AD |
256 | # endif |
257 | #endif | |
258 | ||
259 | #ifndef REG_INFTY | |
4288c5b9 | 260 | # define REG_INFTY U16_MAX |
83cfe48c | 261 | #endif |
a687059c | 262 | |
e91177ed IZ |
263 | #define ARG_VALUE(arg) (arg) |
264 | #define ARG__SET(arg,val) ((arg) = (val)) | |
c277df42 | 265 | |
e8e3b0fb JH |
266 | #undef ARG |
267 | #undef ARG1 | |
268 | #undef ARG2 | |
269 | ||
c277df42 | 270 | #define ARG(p) ARG_VALUE(ARG_LOC(p)) |
86451f01 | 271 | #define ARGp(p) ARG_VALUE(ARGp_LOC(p)) |
c277df42 IZ |
272 | #define ARG1(p) ARG_VALUE(ARG1_LOC(p)) |
273 | #define ARG2(p) ARG_VALUE(ARG2_LOC(p)) | |
6bda09f9 | 274 | #define ARG2L(p) ARG_VALUE(ARG2L_LOC(p)) |
786e8c11 | 275 | |
c277df42 | 276 | #define ARG_SET(p, val) ARG__SET(ARG_LOC(p), (val)) |
86451f01 | 277 | #define ARGp_SET(p, val) ARG__SET(ARGp_LOC(p), (val)) |
c277df42 IZ |
278 | #define ARG1_SET(p, val) ARG__SET(ARG1_LOC(p), (val)) |
279 | #define ARG2_SET(p, val) ARG__SET(ARG2_LOC(p), (val)) | |
6bda09f9 | 280 | #define ARG2L_SET(p, val) ARG__SET(ARG2L_LOC(p), (val)) |
c277df42 | 281 | |
e8e3b0fb JH |
282 | #undef NEXT_OFF |
283 | #undef NODE_ALIGN | |
284 | ||
075abff3 AL |
285 | #define NEXT_OFF(p) ((p)->next_off) |
286 | #define NODE_ALIGN(node) | |
aa48e906 YO |
287 | /* the following define was set to 0xde in 075abff3 |
288 | * as part of some linting logic. I have set it to 0 | |
289 | * as otherwise in every place where we /might/ set flags | |
290 | * we have to set it 0 explicitly, which duplicates | |
291 | * assignments and IMO adds an unacceptable level of | |
292 | * surprise to working in the regex engine. If this | |
293 | * is changed from 0 then at the very least make sure | |
294 | * that SBOL for /^/ sets the flags to 0 explicitly. | |
295 | * -- Yves */ | |
296 | #define NODE_ALIGN_FILL(node) ((node)->flags = 0) | |
a687059c | 297 | |
c277df42 IZ |
298 | #define SIZE_ALIGN NODE_ALIGN |
299 | ||
e8e3b0fb JH |
300 | #undef OP |
301 | #undef OPERAND | |
e8e3b0fb JH |
302 | #undef STRING |
303 | ||
e91177ed | 304 | #define OP(p) ((p)->type) |
b988e673 | 305 | #define FLAGS(p) ((p)->flags) /* Caution: Doesn't apply to all \ |
1604cfb0 MS |
306 | regnode types. For some, it's the \ |
307 | character set of the regnode */ | |
3f2416ae | 308 | #define STR_LENs(p) (__ASSERT_(OP(p) != LEXACT && OP(p) != LEXACT_REQ8) \ |
5cd61b66 | 309 | ((struct regnode_string *)p)->str_len) |
3f2416ae | 310 | #define STRINGs(p) (__ASSERT_(OP(p) != LEXACT && OP(p) != LEXACT_REQ8) \ |
5cd61b66 | 311 | ((struct regnode_string *)p)->string) |
ae06e581 KW |
312 | #define OPERANDs(p) STRINGs(p) |
313 | ||
314 | /* Long strings. Currently limited to length 18 bits, which handles a 262000 | |
315 | * byte string. The limiting factor is the 16 bit 'next_off' field, which | |
316 | * points to the next regnode, so the furthest away it can be is 2**16. On | |
317 | * most architectures, regnodes are 2**2 bytes long, so that yields 2**18 | |
318 | * bytes. Should a longer string be desired, we could increase it to 26 bits | |
319 | * fairly easily, by changing this node to have longj type which causes the ARG | |
320 | * field to be used for the link to the next regnode (although code would have | |
321 | * to be changed to account for this), and then use a combination of the flags | |
322 | * and next_off fields for the length. To get 34 bit length, also change the | |
323 | * node to be an ARG2L, using the second 32 bit field for the length, and not | |
324 | * using the flags nor next_off fields at all. One could have an llstring node | |
325 | * and even an lllstring type. */ | |
3f2416ae | 326 | #define STR_LENl(p) (__ASSERT_(OP(p) == LEXACT || OP(p) == LEXACT_REQ8) \ |
5cd61b66 | 327 | (((struct regnode_lstring *)p)->str_len)) |
3f2416ae | 328 | #define STRINGl(p) (__ASSERT_(OP(p) == LEXACT || OP(p) == LEXACT_REQ8) \ |
5cd61b66 | 329 | (((struct regnode_lstring *)p)->string)) |
ae06e581 KW |
330 | #define OPERANDl(p) STRINGl(p) |
331 | ||
3f2416ae | 332 | #define STR_LEN(p) ((OP(p) == LEXACT || OP(p) == LEXACT_REQ8) \ |
5cd61b66 | 333 | ? STR_LENl(p) : STR_LENs(p)) |
3f2416ae | 334 | #define STRING(p) ((OP(p) == LEXACT || OP(p) == LEXACT_REQ8) \ |
5cd61b66 | 335 | ? STRINGl(p) : STRINGs(p)) |
c5184715 KW |
336 | #define OPERAND(p) STRING(p) |
337 | ||
c45abc0a KW |
338 | /* The number of (smallest) regnode equivalents that a string of length l bytes |
339 | * occupies */ | |
59e54936 | 340 | #define STR_SZ(l) (((l) + sizeof(regnode) - 1) / sizeof(regnode)) |
c45abc0a KW |
341 | |
342 | /* The number of (smallest) regnode equivalents that the EXACTISH node 'p' | |
343 | * occupies */ | |
ae06e581 | 344 | #define NODE_SZ_STR(p) (STR_SZ(STR_LEN(p)) + 1 + regarglen[(p)->type]) |
c45abc0a | 345 | |
c7fd9c67 | 346 | #define setSTR_LEN(p,v) \ |
ae06e581 | 347 | STMT_START{ \ |
3f2416ae | 348 | if (OP(p) == LEXACT || OP(p) == LEXACT_REQ8) \ |
ae06e581 KW |
349 | ((struct regnode_lstring *)(p))->str_len = (v); \ |
350 | else \ | |
351 | ((struct regnode_string *)(p))->str_len = (v); \ | |
352 | } STMT_END | |
cd439c50 | 353 | |
13fcf652 | 354 | #define ANYOFR_BASE_BITS 20 |
d223e1ea | 355 | #define ANYOFRbase(p) (ARG(p) & nBIT_MASK(ANYOFR_BASE_BITS)) |
13fcf652 KW |
356 | #define ANYOFRdelta(p) (ARG(p) >> ANYOFR_BASE_BITS) |
357 | ||
e8e3b0fb JH |
358 | #undef NODE_ALIGN |
359 | #undef ARG_LOC | |
360 | #undef NEXTOPER | |
361 | #undef PREVOPER | |
362 | ||
e91177ed IZ |
363 | #define NODE_ALIGN(node) |
364 | #define ARG_LOC(p) (((struct regnode_1 *)p)->arg1) | |
86451f01 | 365 | #define ARGp_LOC(p) (((struct regnode_p *)p)->arg1) |
e91177ed IZ |
366 | #define ARG1_LOC(p) (((struct regnode_2 *)p)->arg1) |
367 | #define ARG2_LOC(p) (((struct regnode_2 *)p)->arg2) | |
6bda09f9 | 368 | #define ARG2L_LOC(p) (((struct regnode_2L *)p)->arg2) |
786e8c11 | 369 | |
e91177ed IZ |
370 | #define NODE_STEP_REGNODE 1 /* sizeof(regnode)/sizeof(regnode) */ |
371 | #define EXTRA_STEP_2ARGS EXTRA_SIZE(struct regnode_2) | |
372 | ||
c277df42 IZ |
373 | #define NEXTOPER(p) ((p) + NODE_STEP_REGNODE) |
374 | #define PREVOPER(p) ((p) - NODE_STEP_REGNODE) | |
375 | ||
f55b7b26 KW |
376 | #define FILL_NODE(offset, op) \ |
377 | STMT_START { \ | |
378 | OP(REGNODE_p(offset)) = op; \ | |
379 | NEXT_OFF(REGNODE_p(offset)) = 0; \ | |
d2a88850 | 380 | } STMT_END |
f55b7b26 KW |
381 | #define FILL_ADVANCE_NODE(offset, op) \ |
382 | STMT_START { \ | |
383 | FILL_NODE(offset, op); \ | |
384 | (offset)++; \ | |
d2a88850 | 385 | } STMT_END |
f55b7b26 KW |
386 | #define FILL_ADVANCE_NODE_ARG(offset, op, arg) \ |
387 | STMT_START { \ | |
388 | ARG_SET(REGNODE_p(offset), arg); \ | |
389 | FILL_ADVANCE_NODE(offset, op); \ | |
390 | /* This is used generically for other operations \ | |
391 | * that have a longer argument */ \ | |
392 | (offset) += regarglen[op]; \ | |
86451f01 KW |
393 | } STMT_END |
394 | #define FILL_ADVANCE_NODE_ARGp(offset, op, arg) \ | |
395 | STMT_START { \ | |
396 | ARGp_SET(REGNODE_p(offset), arg); \ | |
397 | FILL_ADVANCE_NODE(offset, op); \ | |
398 | (offset) += regarglen[op]; \ | |
93539530 | 399 | } STMT_END |
f55b7b26 KW |
400 | #define FILL_ADVANCE_NODE_2L_ARG(offset, op, arg1, arg2) \ |
401 | STMT_START { \ | |
402 | ARG_SET(REGNODE_p(offset), arg1); \ | |
403 | ARG2L_SET(REGNODE_p(offset), arg2); \ | |
404 | FILL_ADVANCE_NODE(offset, op); \ | |
405 | (offset) += 2; \ | |
93539530 | 406 | } STMT_END |
a687059c | 407 | |
22c35a8c | 408 | #define REG_MAGIC 0234 |
a687059c | 409 | |
fe8d1b7c KW |
410 | /* An ANYOF node is basically a bitmap with the index being a code point. If |
411 | * the bit for that code point is 1, the code point matches; if 0, it doesn't | |
412 | * match (complemented if inverted). There is an additional mechanism to deal | |
413 | * with cases where the bitmap is insufficient in and of itself. This #define | |
414 | * indicates if the bitmap does fully represent what this ANYOF node can match. | |
415 | * The ARG is set to this special value (since 0, 1, ... are legal, but will | |
416 | * never reach this high). */ | |
93e92956 | 417 | #define ANYOF_ONLY_HAS_BITMAP ((U32) -1) |
137165a6 | 418 | |
93539530 | 419 | /* When the bitmap isn't completely sufficient for handling the ANYOF node, |
fe8d1b7c KW |
420 | * flags (in node->flags of the ANYOF node) get set to indicate this. These |
421 | * are perennially in short supply. Beyond several cases where warnings need | |
422 | * to be raised under certain circumstances, currently, there are six cases | |
423 | * where the bitmap alone isn't sufficient. We could use six flags to | |
424 | * represent the 6 cases, but to save flags bits, we play some games. The | |
425 | * cases are: | |
d2f465e9 | 426 | * |
fe8d1b7c KW |
427 | * 1) The bitmap has a compiled-in very finite size. So something else needs |
428 | * to be used to specify if a code point that is too large for the bitmap | |
7a94a5fe | 429 | * actually matches. The mechanism currently is an inversion |
fe8d1b7c KW |
430 | * list. ANYOF_ONLY_HAS_BITMAP, described above, being TRUE indicates |
431 | * there are no matches of too-large code points. But if it is FALSE, | |
432 | * then almost certainly there are matches too large for the bitmap. (The | |
433 | * other cases, described below, either imply this one or are extremely | |
434 | * rare in practice.) So we can just assume that a too-large code point | |
435 | * will need something beyond the bitmap if ANYOF_ONLY_HAS_BITMAP is | |
436 | * FALSE, instead of having a separate flag for this. | |
437 | * 2) A subset of item 1) is if all possible code points outside the bitmap | |
438 | * match. This is a common occurrence when the class is complemented, | |
439 | * like /[^ij]/. Therefore a bit is reserved to indicate this, | |
42d0f81a | 440 | * rather than having an inversion list created, |
d1c40ef5 | 441 | * ANYOF_MATCHES_ALL_ABOVE_BITMAP. |
fe8d1b7c KW |
442 | * 3) Under /d rules, it can happen that code points that are in the upper |
443 | * latin1 range (\x80-\xFF or their equivalents on EBCDIC platforms) match | |
444 | * only if the runtime target string being matched against is UTF-8. For | |
445 | * example /[\w[:punct:]]/d. This happens only for posix classes (with a | |
d1c40ef5 KW |
446 | * couple of exceptions, like \d where it doesn't happen), and all such |
447 | * ones also have above-bitmap matches. Thus, 3) implies 1) as well. | |
448 | * Note that /d rules are no longer encouraged; 'use 5.14' or higher | |
449 | * deselects them. But a flag is required so that they can be properly | |
450 | * handled. But it can be a shared flag: see 5) below. | |
451 | * 4) Also under /d rules, something like /[\Wfoo]/ will match everything in | |
fe8d1b7c | 452 | * the \x80-\xFF range, unless the string being matched against is UTF-8. |
7a94a5fe KW |
453 | * An inversion list could be created for this case, but this is |
454 | * relatively common, and it turns out that it's all or nothing: if any | |
455 | * one of these code points matches, they all do. Hence a single bit | |
456 | * suffices. We use a shared flag that doesn't take up space by itself: | |
457 | * ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER. This | |
458 | * also implies 1), with one exception: [:^cntrl:]. | |
fe8d1b7c KW |
459 | * 5) A user-defined \p{} property may not have been defined by the time the |
460 | * regex is compiled. In this case, we don't know until runtime what it | |
461 | * will match, so we have to assume it could match anything, including | |
462 | * code points that ordinarily would be in the bitmap. A flag bit is | |
874c6b29 | 463 | * necessary to indicate this, though it can be shared with the item 3) |
fe8d1b7c KW |
464 | * flag, as that only occurs under /d, and this only occurs under non-d. |
465 | * This case is quite uncommon in the field, and the /(?[ ...])/ construct | |
466 | * is a better way to accomplish what this feature does. This case also | |
467 | * implies 1). | |
468 | * ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP | |
e2f5e63d | 469 | * is the shared flag. |
fe8d1b7c KW |
470 | * 6) /[foo]/il may have folds that are only valid if the runtime locale is a |
471 | * UTF-8 one. These are quite rare, so it would be good to avoid the | |
472 | * expense of looking for them. But /l matching is slow anyway, and we've | |
e2f5e63d | 473 | * traditionally not worried too much about its performance. And this |
037715a6 | 474 | * condition requires the ANYOFL_FOLD flag to be set, so testing for |
fe8d1b7c | 475 | * that flag would be sufficient to rule out most cases of this. So it is |
d1c40ef5 KW |
476 | * unclear if this should have a flag or not. But, this flag can be |
477 | * shared with another, so it doesn't occupy extra space. | |
d2f465e9 | 478 | * |
d1c40ef5 | 479 | * At the moment, there is one spare bit, but this could be increased by |
93539530 | 480 | * various tricks: |
e2f5e63d | 481 | * |
93539530 | 482 | * If just one more bit is needed, as of this writing it seems to khw that the |
d1c40ef5 KW |
483 | * best choice would be to make ANYOF_MATCHES_ALL_ABOVE_BITMAP not a flag, but |
484 | * something like | |
d2f465e9 | 485 | * |
d1c40ef5 | 486 | * #define ANYOF_MATCHES_ALL_ABOVE_BITMAP ((U32) -2) |
fe8d1b7c | 487 | * |
d1c40ef5 KW |
488 | * and access it through the ARG like ANYOF_ONLY_HAS_BITMAP is. This flag is |
489 | * used by all ANYOF node types, and it could be used to avoid calling the | |
490 | * handler function, as the macro REGINCLASS in regexec.c does now for other | |
491 | * cases. | |
492 | * | |
46fc0c43 KW |
493 | * Another possibility is based on the fact that ANYOF_MATCHES_POSIXL is |
494 | * redundant with the node type ANYOFPOSIXL. That flag could be removed, but | |
495 | * at the expense of extra code in regexec.c. The flag has been retained | |
874c6b29 | 496 | * because it allows us to see if we need to call reginclass, or just use the |
46fc0c43 | 497 | * bitmap in one test. |
fe8d1b7c | 498 | * |
46fc0c43 KW |
499 | * If this is done, an extension would be to make all ANYOFL nodes contain the |
500 | * extra 32 bits that ANYOFPOSIXL ones do. The posix flags only occupy 30 | |
501 | * bits, so the ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD flags | |
502 | * and ANYOFL_FOLD could be moved to that extra space, but it would mean extra | |
503 | * instructions, as there are currently places in the code that assume those | |
504 | * two bits are zero. | |
d1c40ef5 KW |
505 | * |
506 | * All told, 5 bits could be available for other uses if all of the above were | |
507 | * done. | |
a4525e78 | 508 | * |
d1c40ef5 | 509 | * Some flags are not used in synthetic start class (SSC) nodes, so could be |
93e92956 KW |
510 | * shared should new flags be needed for SSCs, like SSC_MATCHES_EMPTY_STRING |
511 | * now. */ | |
b8c5462f | 512 | |
d2f465e9 KW |
513 | /* If this is set, the result of the match should be complemented. regexec.c |
514 | * is expecting this to be in the low bit. Never in an SSC */ | |
93e92956 | 515 | #define ANYOF_INVERT 0x01 |
1984f6a0 | 516 | |
a0dd4231 | 517 | /* For the SSC node only, which cannot be inverted, so is shared with that bit. |
93e92956 KW |
518 | * This is used only during regex compilation. */ |
519 | #define SSC_MATCHES_EMPTY_STRING ANYOF_INVERT | |
a0dd4231 | 520 | |
509d4de0 KW |
521 | /* Set if this is a regnode_charclass_posixl vs a regnode_charclass. This |
522 | * is used for runtime \d, \w, [:posix:], ..., which are used only in locale | |
523 | * and the optimizer's synthetic start class. Non-locale \d, etc are resolved | |
524 | * at compile-time. Only set under /l; can be in SSC */ | |
525 | #define ANYOF_MATCHES_POSIXL 0x02 | |
b63f73da KW |
526 | |
527 | /* The fold is calculated and stored in the bitmap where possible at compile | |
528 | * time. However under locale, the actual folding varies depending on | |
529 | * what the locale is at the time of execution, so it has to be deferred until | |
d2f465e9 | 530 | * then. Only set under /l; never in an SSC */ |
037715a6 | 531 | #define ANYOFL_FOLD 0x04 |
b63f73da | 532 | |
d1c40ef5 | 533 | /* Shared bit set only with ANYOFL and SSC nodes: |
269bb65b KW |
534 | * If ANYOFL_FOLD is set, this flag indicates there are potential matches |
535 | * valid only if the locale is a UTF-8 one. | |
536 | * If ANYOFL_FOLD is NOT set, this flag means to warn if the runtime locale | |
d1c40ef5 KW |
537 | * isn't a UTF-8 one (and the generated node assumes a UTF-8 locale). |
538 | * None of INVERT, POSIXL, | |
539 | * ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP | |
540 | * can be set. */ | |
541 | #define ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD 0x08 | |
542 | ||
543 | /* Convenience macros for teasing apart the meanings when reading the above bit | |
544 | * */ | |
545 | #define ANYOFL_SOME_FOLDS_ONLY_IN_UTF8_LOCALE(flags) \ | |
546 | ((flags & ( ANYOFL_FOLD /* Both bits are set */ \ | |
547 | |ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD)) \ | |
548 | == ( ANYOFL_FOLD \ | |
549 | |ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD)) | |
550 | ||
551 | #define ANYOFL_UTF8_LOCALE_REQD(flags) \ | |
552 | ((flags & ( ANYOFL_FOLD /* Only REQD bit is set */ \ | |
553 | |ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD)) \ | |
554 | == ANYOFL_SHARED_UTF8_LOCALE_fold_HAS_MATCHES_nonfold_REQD) | |
555 | ||
1744d18d | 556 | /* Spare: Be sure to change ANYOF_FLAGS_ALL if this gets used 0x10 */ |
c8d3cd88 | 557 | |
509d4de0 KW |
558 | /* If set, the node matches every code point NUM_ANYOF_CODE_POINTS and above. |
559 | * Can be in an SSC */ | |
b3b1cf17 | 560 | #define ANYOF_MATCHES_ALL_ABOVE_BITMAP 0x20 |
ef87b810 | 561 | |
108316fb KW |
562 | /* Shared bit: |
563 | * Under /d it means the ANYOFD node matches more things if the target | |
564 | * string is encoded in UTF-8; any such things will be non-ASCII, | |
7a94a5fe KW |
565 | * characters that are < 256, and can be accessed via the inversion |
566 | * list. | |
108316fb KW |
567 | * When not under /d, it means the ANYOF node contains a user-defined |
568 | * property that wasn't yet defined at the time the regex was compiled, | |
7a94a5fe | 569 | * and so must be looked up at runtime, by creating an inversion list. |
108316fb KW |
570 | * (These uses are mutually exclusive because a user-defined property is |
571 | * specified by \p{}, and \p{} implies /u which deselects /d). The long macro | |
572 | * name is to make sure that you are cautioned about its shared nature. Only | |
573 | * the non-/d meaning can be in an SSC */ | |
574 | #define ANYOF_SHARED_d_UPPER_LATIN1_UTF8_STRING_MATCHES_non_d_RUNTIME_USER_PROP 0x40 | |
f240c685 KW |
575 | |
576 | /* Shared bit: | |
d2f465e9 | 577 | * Under /d it means the ANYOFD node matches all non-ASCII Latin1 |
f240c685 KW |
578 | * characters when the target string is not in utf8. |
579 | * When not under /d, it means the ANYOF node should raise a warning if | |
580 | * matching against an above-Unicode code point. | |
581 | * (These uses are mutually exclusive because the warning requires a \p{}, and | |
582 | * \p{} implies /u which deselects /d). An SSC node only has this bit set if | |
583 | * what is meant is the warning. The long macro name is to make sure that you | |
584 | * are cautioned about its shared nature */ | |
585 | #define ANYOF_SHARED_d_MATCHES_ALL_NON_UTF8_NON_ASCII_non_d_WARN_SUPER 0x80 | |
7bc66b18 | 586 | |
031c7092 | 587 | #define ANYOF_FLAGS_ALL ((U8) ~0x10) |
653099ff | 588 | |
037715a6 | 589 | #define ANYOF_LOCALE_FLAGS (ANYOFL_FOLD | ANYOF_MATCHES_POSIXL) |
ace6b0e4 | 590 | |
eff8b7dc KW |
591 | /* These are the flags that apply to both regular ANYOF nodes and synthetic |
592 | * start class nodes during construction of the SSC. During finalization of | |
d2f465e9 | 593 | * the SSC, other of the flags may get added to it */ |
d1c40ef5 | 594 | #define ANYOF_COMMON_FLAGS 0 |
eff8b7dc | 595 | |
936ed897 | 596 | /* Character classes for node->classflags of ANYOF */ |
653099ff | 597 | /* Should be synchronized with a table in regprop() */ |
575a8a29 | 598 | /* 2n should be the normal one, paired with its complement at 2n+1 */ |
b8c5462f | 599 | |
08ada29e KW |
600 | #define ANYOF_ALPHA ((_CC_ALPHA) * 2) |
601 | #define ANYOF_NALPHA ((ANYOF_ALPHA) + 1) | |
e8d596e0 KW |
602 | #define ANYOF_ALPHANUMERIC ((_CC_ALPHANUMERIC) * 2) /* [[:alnum:]] isalnum(3), utf8::IsAlnum */ |
603 | #define ANYOF_NALPHANUMERIC ((ANYOF_ALPHANUMERIC) + 1) | |
08ada29e KW |
604 | #define ANYOF_ASCII ((_CC_ASCII) * 2) |
605 | #define ANYOF_NASCII ((ANYOF_ASCII) + 1) | |
e8d596e0 KW |
606 | #define ANYOF_BLANK ((_CC_BLANK) * 2) /* GNU extension: space and tab: non-vertical space */ |
607 | #define ANYOF_NBLANK ((ANYOF_BLANK) + 1) | |
b0d691b2 KW |
608 | #define ANYOF_CASED ((_CC_CASED) * 2) /* Pseudo class for [:lower:] or |
609 | [:upper:] under /i */ | |
610 | #define ANYOF_NCASED ((ANYOF_CASED) + 1) | |
08ada29e KW |
611 | #define ANYOF_CNTRL ((_CC_CNTRL) * 2) |
612 | #define ANYOF_NCNTRL ((ANYOF_CNTRL) + 1) | |
e8d596e0 KW |
613 | #define ANYOF_DIGIT ((_CC_DIGIT) * 2) /* \d */ |
614 | #define ANYOF_NDIGIT ((ANYOF_DIGIT) + 1) | |
08ada29e KW |
615 | #define ANYOF_GRAPH ((_CC_GRAPH) * 2) |
616 | #define ANYOF_NGRAPH ((ANYOF_GRAPH) + 1) | |
617 | #define ANYOF_LOWER ((_CC_LOWER) * 2) | |
618 | #define ANYOF_NLOWER ((ANYOF_LOWER) + 1) | |
619 | #define ANYOF_PRINT ((_CC_PRINT) * 2) | |
620 | #define ANYOF_NPRINT ((ANYOF_PRINT) + 1) | |
621 | #define ANYOF_PUNCT ((_CC_PUNCT) * 2) | |
622 | #define ANYOF_NPUNCT ((ANYOF_PUNCT) + 1) | |
e8d596e0 KW |
623 | #define ANYOF_SPACE ((_CC_SPACE) * 2) /* \s */ |
624 | #define ANYOF_NSPACE ((ANYOF_SPACE) + 1) | |
08ada29e KW |
625 | #define ANYOF_UPPER ((_CC_UPPER) * 2) |
626 | #define ANYOF_NUPPER ((ANYOF_UPPER) + 1) | |
e8d596e0 KW |
627 | #define ANYOF_WORDCHAR ((_CC_WORDCHAR) * 2) /* \w, PL_utf8_alnum, utf8::IsWord, ALNUM */ |
628 | #define ANYOF_NWORDCHAR ((ANYOF_WORDCHAR) + 1) | |
08ada29e KW |
629 | #define ANYOF_XDIGIT ((_CC_XDIGIT) * 2) |
630 | #define ANYOF_NXDIGIT ((ANYOF_XDIGIT) + 1) | |
b8c5462f | 631 | |
87cad4c0 | 632 | /* pseudo classes below this, not stored in the class bitmap, but used as flags |
e1d1eefb YO |
633 | during compilation of char classes */ |
634 | ||
b2ef2e34 KW |
635 | #define ANYOF_VERTWS ((_CC_VERTSPACE) * 2) |
636 | #define ANYOF_NVERTWS ((ANYOF_VERTWS)+1) | |
e1d1eefb | 637 | |
b2ef2e34 KW |
638 | /* It is best if this is the last one, as all above it are stored as bits in a |
639 | * bitmap, and it isn't part of that bitmap */ | |
640 | #if _CC_VERTSPACE != _HIGHEST_REGCOMP_DOT_H_SYNC | |
641 | # error Problem with handy.h _HIGHEST_REGCOMP_DOT_H_SYNC #define | |
642 | #endif | |
643 | ||
8efd3f97 KW |
644 | #define ANYOF_POSIXL_MAX (ANYOF_VERTWS) /* So upper loop limit is written: |
645 | * '< ANYOF_MAX' | |
646 | * Hence doesn't include VERTWS, as that | |
647 | * is a pseudo class */ | |
648 | #define ANYOF_MAX ANYOF_POSIXL_MAX | |
649 | ||
650 | #if (ANYOF_POSIXL_MAX > 32) /* Must fit in 32-bit word */ | |
b2ef2e34 | 651 | # error Problem with handy.h _CC_foo #defines |
a0947d7b KW |
652 | #endif |
653 | ||
8efd3f97 KW |
654 | #define ANYOF_HORIZWS ((ANYOF_POSIXL_MAX)+2) /* = (ANYOF_NVERTWS + 1) */ |
655 | #define ANYOF_NHORIZWS ((ANYOF_POSIXL_MAX)+3) | |
87cad4c0 | 656 | |
8efd3f97 KW |
657 | #define ANYOF_UNIPROP ((ANYOF_POSIXL_MAX)+4) /* Used to indicate a Unicode |
658 | property: \p{} or \P{} */ | |
f39a4dc2 | 659 | |
b8c5462f JH |
660 | /* Backward source code compatibility. */ |
661 | ||
662 | #define ANYOF_ALNUML ANYOF_ALNUM | |
663 | #define ANYOF_NALNUML ANYOF_NALNUM | |
664 | #define ANYOF_SPACEL ANYOF_SPACE | |
665 | #define ANYOF_NSPACEL ANYOF_NSPACE | |
8ebfc29a KW |
666 | #define ANYOF_ALNUM ANYOF_WORDCHAR |
667 | #define ANYOF_NALNUM ANYOF_NWORDCHAR | |
b8c5462f JH |
668 | |
669 | /* Utility macros for the bitmap and classes of ANYOF */ | |
670 | ||
936ed897 | 671 | #define ANYOF_FLAGS(p) ((p)->flags) |
b8c5462f | 672 | |
bc51fd78 | 673 | #define ANYOF_BIT(c) (1U << ((c) & 7)) |
b8c5462f | 674 | |
ad494d15 KW |
675 | #define ANYOF_POSIXL_BITMAP(p) (((regnode_charclass_posixl*) (p))->classflags) |
676 | ||
14ed74f6 | 677 | #define POSIXL_SET(field, c) ((field) |= (1U << (c))) |
ad494d15 | 678 | #define ANYOF_POSIXL_SET(p, c) POSIXL_SET(ANYOF_POSIXL_BITMAP(p), (c)) |
8efd3f97 | 679 | |
14ed74f6 | 680 | #define POSIXL_CLEAR(field, c) ((field) &= ~ (1U <<(c))) |
ad494d15 | 681 | #define ANYOF_POSIXL_CLEAR(p, c) POSIXL_CLEAR(ANYOF_POSIXL_BITMAP(p), (c)) |
8efd3f97 | 682 | |
14ed74f6 | 683 | #define POSIXL_TEST(field, c) ((field) & (1U << (c))) |
ad494d15 | 684 | #define ANYOF_POSIXL_TEST(p, c) POSIXL_TEST(ANYOF_POSIXL_BITMAP(p), (c)) |
b8c5462f | 685 | |
14ed74f6 | 686 | #define POSIXL_ZERO(field) STMT_START { (field) = 0; } STMT_END |
ad494d15 | 687 | #define ANYOF_POSIXL_ZERO(ret) POSIXL_ZERO(ANYOF_POSIXL_BITMAP(ret)) |
14ed74f6 KW |
688 | |
689 | #define ANYOF_POSIXL_SET_TO_BITMAP(p, bits) \ | |
ad494d15 | 690 | STMT_START { ANYOF_POSIXL_BITMAP(p) = (bits); } STMT_END |
9cf78fa0 KW |
691 | |
692 | /* Shifts a bit to get, eg. 0x4000_0000, then subtracts 1 to get 0x3FFF_FFFF */ | |
ad494d15 KW |
693 | #define ANYOF_POSIXL_SETALL(ret) \ |
694 | STMT_START { \ | |
695 | ANYOF_POSIXL_BITMAP(ret) = nBIT_MASK(ANYOF_POSIXL_MAX); \ | |
696 | } STMT_END | |
8efd3f97 | 697 | #define ANYOF_CLASS_SETALL(ret) ANYOF_POSIXL_SETALL(ret) |
9cf78fa0 | 698 | |
8efd3f97 | 699 | #define ANYOF_POSIXL_TEST_ANY_SET(p) \ |
ad494d15 | 700 | ((ANYOF_FLAGS(p) & ANYOF_MATCHES_POSIXL) && ANYOF_POSIXL_BITMAP(p)) |
8efd3f97 | 701 | #define ANYOF_CLASS_TEST_ANY_SET(p) ANYOF_POSIXL_TEST_ANY_SET(p) |
e867ce45 | 702 | |
e0e1be5f KW |
703 | /* Since an SSC always has this field, we don't have to test for that; nor do |
704 | * we want to because the bit isn't set for SSC during its construction */ | |
705 | #define ANYOF_POSIXL_SSC_TEST_ANY_SET(p) \ | |
706 | cBOOL(((regnode_ssc*)(p))->classflags) | |
707 | #define ANYOF_POSIXL_SSC_TEST_ALL_SET(p) /* Are all bits set? */ \ | |
708 | (((regnode_ssc*) (p))->classflags \ | |
d223e1ea | 709 | == nBIT_MASK(ANYOF_POSIXL_MAX)) |
e0e1be5f KW |
710 | |
711 | #define ANYOF_POSIXL_TEST_ALL_SET(p) \ | |
d223e1ea | 712 | ((ANYOF_FLAGS(p) & ANYOF_MATCHES_POSIXL) \ |
ad494d15 | 713 | && ANYOF_POSIXL_BITMAP(p) == nBIT_MASK(ANYOF_POSIXL_MAX)) |
fb38762f | 714 | |
b27fd97a | 715 | #define ANYOF_POSIXL_OR(source, dest) STMT_START { (dest)->classflags |= (source)->classflags ; } STMT_END |
8efd3f97 | 716 | #define ANYOF_CLASS_OR(source, dest) ANYOF_POSIXL_OR((source), (dest)) |
936ed897 | 717 | |
cdd87c1d KW |
718 | #define ANYOF_POSIXL_AND(source, dest) STMT_START { (dest)->classflags &= (source)->classflags ; } STMT_END |
719 | ||
428c1aff KW |
720 | #define ANYOF_BITMAP_ZERO(ret) Zero(((regnode_charclass*)(ret))->bitmap, ANYOF_BITMAP_SIZE, char) |
721 | #define ANYOF_BITMAP(p) ((regnode_charclass*)(p))->bitmap | |
4cbce0a6 | 722 | #define ANYOF_BITMAP_BYTE(p, c) BITMAP_BYTE(ANYOF_BITMAP(p), c) |
b8c5462f JH |
723 | #define ANYOF_BITMAP_SET(p, c) (ANYOF_BITMAP_BYTE(p, c) |= ANYOF_BIT(c)) |
724 | #define ANYOF_BITMAP_CLEAR(p,c) (ANYOF_BITMAP_BYTE(p, c) &= ~ANYOF_BIT(c)) | |
bc51fd78 | 725 | #define ANYOF_BITMAP_TEST(p, c) cBOOL(ANYOF_BITMAP_BYTE(p, c) & ANYOF_BIT(c)) |
b8c5462f | 726 | |
f8bef550 | 727 | #define ANYOF_BITMAP_SETALL(p) \ |
1604cfb0 | 728 | memset (ANYOF_BITMAP(p), 255, ANYOF_BITMAP_SIZE) |
f8bef550 | 729 | #define ANYOF_BITMAP_CLEARALL(p) \ |
1604cfb0 | 730 | Zero (ANYOF_BITMAP(p), ANYOF_BITMAP_SIZE) |
f8bef550 | 731 | |
a687059c LW |
732 | /* |
733 | * Utility definitions. | |
734 | */ | |
2304df62 | 735 | #ifndef CHARMASK |
075abff3 | 736 | # define UCHARAT(p) ((int)*(const U8*)(p)) |
a687059c | 737 | #else |
075abff3 | 738 | # define UCHARAT(p) ((int)*(p)&CHARMASK) |
a687059c | 739 | #endif |
a687059c | 740 | |
c45abc0a KW |
741 | /* Number of regnode equivalents that 'guy' occupies beyond the size of the |
742 | * smallest regnode. */ | |
c277df42 IZ |
743 | #define EXTRA_SIZE(guy) ((sizeof(guy)-1)/sizeof(struct regnode)) |
744 | ||
e384d5c1 YO |
745 | #define REG_ZERO_LEN_SEEN 0x00000001 |
746 | #define REG_LOOKBEHIND_SEEN 0x00000002 | |
747 | #define REG_GPOS_SEEN 0x00000004 | |
471f5387 | 748 | /* spare */ |
e384d5c1 YO |
749 | #define REG_RECURSE_SEEN 0x00000020 |
750 | #define REG_TOP_LEVEL_BRANCHES_SEEN 0x00000040 | |
751 | #define REG_VERBARG_SEEN 0x00000080 | |
752 | #define REG_CUTGROUP_SEEN 0x00000100 | |
753 | #define REG_RUN_ON_COMMENT_SEEN 0x00000200 | |
754 | #define REG_UNFOLDED_MULTI_SEEN 0x00000400 | |
d5a00e4a | 755 | /* spare */ |
ee273784 YO |
756 | #define REG_UNBOUNDED_QUANTIFIER_SEEN 0x00001000 |
757 | ||
d09b2d29 | 758 | |
73c4f7a1 GS |
759 | START_EXTERN_C |
760 | ||
be8e71aa YO |
761 | #ifdef PLUGGABLE_RE_EXTENSION |
762 | #include "re_nodes.h" | |
763 | #else | |
d09b2d29 | 764 | #include "regnodes.h" |
be8e71aa | 765 | #endif |
d09b2d29 | 766 | |
f9f4320a YO |
767 | #ifndef PLUGGABLE_RE_EXTENSION |
768 | #ifndef DOINIT | |
769 | EXTCONST regexp_engine PL_core_reg_engine; | |
770 | #else /* DOINIT */ | |
771 | EXTCONST regexp_engine PL_core_reg_engine = { | |
fe578d7f | 772 | Perl_re_compile, |
2fdbfb4d | 773 | Perl_regexec_flags, |
f9f4320a YO |
774 | Perl_re_intuit_start, |
775 | Perl_re_intuit_string, | |
2fdbfb4d AB |
776 | Perl_regfree_internal, |
777 | Perl_reg_numbered_buff_fetch, | |
778 | Perl_reg_numbered_buff_store, | |
779 | Perl_reg_numbered_buff_length, | |
192b9cd1 AB |
780 | Perl_reg_named_buff, |
781 | Perl_reg_named_buff_iter, | |
49d7dfbc | 782 | Perl_reg_qr_package, |
f9f4320a | 783 | #if defined(USE_ITHREADS) |
3c13cae6 | 784 | Perl_regdupe_internal, |
f9f4320a | 785 | #endif |
3c13cae6 | 786 | Perl_re_op_compile |
f9f4320a YO |
787 | }; |
788 | #endif /* DOINIT */ | |
789 | #endif /* PLUGGABLE_RE_EXTENSION */ | |
790 | ||
791 | ||
73c4f7a1 | 792 | END_EXTERN_C |
cad2e5aa | 793 | |
cad2e5aa | 794 | |
261faec3 MJD |
795 | /* .what is a character array with one character for each member of .data |
796 | * The character describes the function of the corresponding .data item: | |
af534a04 | 797 | * a - AV for paren_name_list under DEBUGGING |
0111c4fd | 798 | * f - start-class data for regstclass optimization |
68e2671b | 799 | * l - start op for literal (?{EVAL}) item |
d63c20f2 | 800 | * L - start op for literal (?{EVAL}) item, with separate CV (qr//) |
b30fcab9 | 801 | * r - pointer to an embedded code-containing qr, e.g. /ab$qr/ |
7a94a5fe KW |
802 | * s - inversion list for Unicode-style character class, and the |
803 | * multicharacter strings resulting from casefolding the single-character | |
804 | * entries in the character class | |
a3621e74 | 805 | * t - trie struct |
55eed653 | 806 | * u - trie struct's widecharmap (a HV, so can't share, must dup) |
2b8b4781 | 807 | * also used for revcharmap and words under DEBUGGING |
07be1b83 | 808 | * T - aho-trie struct |
81714fb9 | 809 | * S - sv for named capture lookup |
261faec3 | 810 | * 20010712 mjd@plover.com |
c3613ed7 | 811 | * (Remember to update re_dup() and pregfree() if you add any items.) |
261faec3 | 812 | */ |
cad2e5aa JH |
813 | struct reg_data { |
814 | U32 count; | |
815 | U8 *what; | |
816 | void* data[1]; | |
817 | }; | |
818 | ||
a1cac82e NC |
819 | /* Code in S_to_utf8_substr() and S_to_byte_substr() in regexec.c accesses |
820 | anchored* and float* via array indexes 0 and 1. */ | |
cad2e5aa | 821 | #define anchored_substr substrs->data[0].substr |
33b8afdf | 822 | #define anchored_utf8 substrs->data[0].utf8_substr |
cad2e5aa | 823 | #define anchored_offset substrs->data[0].min_offset |
1de06328 YO |
824 | #define anchored_end_shift substrs->data[0].end_shift |
825 | ||
cad2e5aa | 826 | #define float_substr substrs->data[1].substr |
33b8afdf | 827 | #define float_utf8 substrs->data[1].utf8_substr |
cad2e5aa JH |
828 | #define float_min_offset substrs->data[1].min_offset |
829 | #define float_max_offset substrs->data[1].max_offset | |
1de06328 YO |
830 | #define float_end_shift substrs->data[1].end_shift |
831 | ||
cad2e5aa | 832 | #define check_substr substrs->data[2].substr |
33b8afdf | 833 | #define check_utf8 substrs->data[2].utf8_substr |
cad2e5aa JH |
834 | #define check_offset_min substrs->data[2].min_offset |
835 | #define check_offset_max substrs->data[2].max_offset | |
1de06328 | 836 | #define check_end_shift substrs->data[2].end_shift |
a3621e74 | 837 | |
8d919b0a FC |
838 | #define RX_ANCHORED_SUBSTR(rx) (ReANY(rx)->anchored_substr) |
839 | #define RX_ANCHORED_UTF8(rx) (ReANY(rx)->anchored_utf8) | |
840 | #define RX_FLOAT_SUBSTR(rx) (ReANY(rx)->float_substr) | |
841 | #define RX_FLOAT_UTF8(rx) (ReANY(rx)->float_utf8) | |
a3621e74 YO |
842 | |
843 | /* trie related stuff */ | |
5d9a96ca | 844 | |
a3621e74 YO |
845 | /* a transition record for the state machine. the |
846 | check field determines which state "owns" the | |
847 | transition. the char the transition is for is | |
848 | determined by offset from the owning states base | |
849 | field. the next field determines which state | |
850 | is to be transitioned to if any. | |
851 | */ | |
852 | struct _reg_trie_trans { | |
853 | U32 next; | |
854 | U32 check; | |
855 | }; | |
856 | ||
857 | /* a transition list element for the list based representation */ | |
858 | struct _reg_trie_trans_list_elem { | |
859 | U16 forid; | |
860 | U32 newstate; | |
861 | }; | |
862 | typedef struct _reg_trie_trans_list_elem reg_trie_trans_le; | |
863 | ||
864 | /* a state for compressed nodes. base is an offset | |
865 | into an array of reg_trie_trans array. If wordnum is | |
866 | nonzero the state is accepting. if base is zero then | |
867 | the state has no children (and will be accepting) | |
868 | */ | |
869 | struct _reg_trie_state { | |
870 | U16 wordnum; | |
871 | union { | |
872 | U32 base; | |
873 | reg_trie_trans_le* list; | |
874 | } trans; | |
875 | }; | |
876 | ||
2e64971a DM |
877 | /* info per word; indexed by wordnum */ |
878 | typedef struct { | |
879 | U16 prev; /* previous word in acceptance chain; eg in | |
1604cfb0 MS |
880 | * zzz|abc|ab/ after matching the chars abc, the |
881 | * accepted word is #2, and the previous accepted | |
882 | * word is #3 */ | |
2e64971a DM |
883 | U32 len; /* how many chars long is this word? */ |
884 | U32 accept; /* accept state for this word */ | |
885 | } reg_trie_wordinfo; | |
a3621e74 YO |
886 | |
887 | ||
a3621e74 YO |
888 | typedef struct _reg_trie_state reg_trie_state; |
889 | typedef struct _reg_trie_trans reg_trie_trans; | |
890 | ||
891 | ||
892 | /* anything in here that needs to be freed later | |
23eab42c NC |
893 | should be dealt with in pregfree. |
894 | refcount is first in both this and _reg_ac_data to allow a space | |
895 | optimisation in Perl_regdupe. */ | |
a3621e74 | 896 | struct _reg_trie_data { |
23eab42c | 897 | U32 refcount; /* number of times this trie is referenced */ |
786e8c11 YO |
898 | U32 lasttrans; /* last valid transition element */ |
899 | U16 *charmap; /* byte to charid lookup array */ | |
786e8c11 YO |
900 | reg_trie_state *states; /* state data */ |
901 | reg_trie_trans *trans; /* array of transition elements */ | |
902 | char *bitmap; /* stclass bitmap */ | |
786e8c11 YO |
903 | U16 *jump; /* optional 1 indexed array of offsets before tail |
904 | for the node following a given word. */ | |
2e64971a | 905 | reg_trie_wordinfo *wordinfo; /* array of info per word */ |
1b5c067c NC |
906 | U16 uniquecharcount; /* unique chars in trie (width of trans table) */ |
907 | U32 startstate; /* initial state - used for common prefix optimisation */ | |
908 | STRLEN minlen; /* minimum length of words in trie - build/opt only? */ | |
909 | STRLEN maxlen; /* maximum length of words in trie - build/opt only? */ | |
2e64971a | 910 | U32 prefixlen; /* #chars in common prefix */ |
1e2e3d02 YO |
911 | U32 statecount; /* Build only - number of states in the states array |
912 | (including the unused zero state) */ | |
786e8c11 | 913 | U32 wordcount; /* Build only */ |
a3621e74 | 914 | #ifdef DEBUGGING |
786e8c11 | 915 | STRLEN charcount; /* Build only */ |
a3621e74 YO |
916 | #endif |
917 | }; | |
2b8b4781 NC |
918 | /* There is one (3 under DEBUGGING) pointers that logically belong in this |
919 | structure, but are held outside as they need duplication on thread cloning, | |
920 | whereas the rest of the structure can be read only: | |
921 | HV *widecharmap; code points > 255 to charid | |
922 | #ifdef DEBUGGING | |
923 | AV *words; Array of words contained in trie, for dumping | |
924 | AV *revcharmap; Map of each charid back to its character representation | |
925 | #endif | |
926 | */ | |
927 | ||
928 | #define TRIE_WORDS_OFFSET 2 | |
929 | ||
a3621e74 YO |
930 | typedef struct _reg_trie_data reg_trie_data; |
931 | ||
23eab42c NC |
932 | /* refcount is first in both this and _reg_trie_data to allow a space |
933 | optimisation in Perl_regdupe. */ | |
07be1b83 | 934 | struct _reg_ac_data { |
23eab42c | 935 | U32 refcount; |
1b5c067c | 936 | U32 trie; |
07be1b83 YO |
937 | U32 *fail; |
938 | reg_trie_state *states; | |
07be1b83 YO |
939 | }; |
940 | typedef struct _reg_ac_data reg_ac_data; | |
941 | ||
486ec47a | 942 | /* ANY_BIT doesn't use the structure, so we can borrow it here. |
3dab1dad YO |
943 | This is simpler than refactoring all of it as wed end up with |
944 | three different sets... */ | |
945 | ||
946 | #define TRIE_BITMAP(p) (((reg_trie_data *)(p))->bitmap) | |
4cbce0a6 | 947 | #define TRIE_BITMAP_BYTE(p, c) BITMAP_BYTE(TRIE_BITMAP(p), c) |
5dfe062f YO |
948 | #define TRIE_BITMAP_SET(p, c) (TRIE_BITMAP_BYTE(p, c) |= ANYOF_BIT((U8)c)) |
949 | #define TRIE_BITMAP_CLEAR(p,c) (TRIE_BITMAP_BYTE(p, c) &= ~ANYOF_BIT((U8)c)) | |
950 | #define TRIE_BITMAP_TEST(p, c) (TRIE_BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c)) | |
3dab1dad | 951 | |
1de06328 YO |
952 | #define IS_ANYOF_TRIE(op) ((op)==TRIEC || (op)==AHOCORASICKC) |
953 | #define IS_TRIE_AC(op) ((op)>=AHOCORASICK) | |
954 | ||
955 | ||
0effac60 | 956 | #define BITMAP_BYTE(p, c) (( (U8*) p) [ ( ( (UV) (c)) >> 3) ] ) |
786e8c11 | 957 | #define BITMAP_TEST(p, c) (BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c)) |
3dab1dad | 958 | |
a3621e74 YO |
959 | /* these defines assume uniquecharcount is the correct variable, and state may be evaluated twice */ |
960 | #define TRIE_NODENUM(state) (((state)-1)/(trie->uniquecharcount)+1) | |
961 | #define SAFE_TRIE_NODENUM(state) ((state) ? (((state)-1)/(trie->uniquecharcount)+1) : (state)) | |
962 | #define TRIE_NODEIDX(state) ((state) ? (((state)-1)*(trie->uniquecharcount)+1) : (state)) | |
963 | ||
3dab1dad | 964 | #ifdef DEBUGGING |
3dab1dad | 965 | #define TRIE_CHARCOUNT(trie) ((trie)->charcount) |
3dab1dad | 966 | #else |
3dab1dad | 967 | #define TRIE_CHARCOUNT(trie) (trie_charcount) |
3dab1dad | 968 | #endif |
a3621e74 | 969 | |
0111c4fd RGS |
970 | #define RE_TRIE_MAXBUF_INIT 65536 |
971 | #define RE_TRIE_MAXBUF_NAME "\022E_TRIE_MAXBUF" | |
a3621e74 YO |
972 | #define RE_DEBUG_FLAGS "\022E_DEBUG_FLAGS" |
973 | ||
6ef7fe53 KW |
974 | #define RE_COMPILE_RECURSION_INIT 1000 |
975 | #define RE_COMPILE_RECURSION_LIMIT "\022E_COMPILE_RECURSION_LIMIT" | |
976 | ||
be8e71aa YO |
977 | /* |
978 | ||
979 | RE_DEBUG_FLAGS is used to control what debug output is emitted | |
980 | its divided into three groups of options, some of which interact. | |
981 | The three groups are: Compile, Execute, Extra. There is room for a | |
982 | further group, as currently only the low three bytes are used. | |
983 | ||
984 | Compile Options: | |
07be1b83 | 985 | |
be8e71aa YO |
986 | PARSE |
987 | PEEP | |
988 | TRIE | |
989 | PROGRAM | |
a3621e74 | 990 | |
be8e71aa | 991 | Execute Options: |
a3621e74 | 992 | |
be8e71aa YO |
993 | INTUIT |
994 | MATCH | |
995 | TRIE | |
a3621e74 | 996 | |
be8e71aa | 997 | Extra Options |
a3621e74 | 998 | |
be8e71aa | 999 | TRIE |
be8e71aa YO |
1000 | |
1001 | If you modify any of these make sure you make corresponding changes to | |
1002 | re.pm, especially to the documentation. | |
a3621e74 | 1003 | |
be8e71aa YO |
1004 | */ |
1005 | ||
1006 | ||
1007 | /* Compile */ | |
1008 | #define RE_DEBUG_COMPILE_MASK 0x0000FF | |
1009 | #define RE_DEBUG_COMPILE_PARSE 0x000001 | |
1010 | #define RE_DEBUG_COMPILE_OPTIMISE 0x000002 | |
1011 | #define RE_DEBUG_COMPILE_TRIE 0x000004 | |
1012 | #define RE_DEBUG_COMPILE_DUMP 0x000008 | |
f7819f85 | 1013 | #define RE_DEBUG_COMPILE_FLAGS 0x000010 |
d9a72fcc | 1014 | #define RE_DEBUG_COMPILE_TEST 0x000020 |
be8e71aa YO |
1015 | |
1016 | /* Execute */ | |
1017 | #define RE_DEBUG_EXECUTE_MASK 0x00FF00 | |
1018 | #define RE_DEBUG_EXECUTE_INTUIT 0x000100 | |
1019 | #define RE_DEBUG_EXECUTE_MATCH 0x000200 | |
1020 | #define RE_DEBUG_EXECUTE_TRIE 0x000400 | |
1021 | ||
1022 | /* Extra */ | |
51bd1941 | 1023 | #define RE_DEBUG_EXTRA_MASK 0x3FF0000 |
fd41b2d1 | 1024 | #define RE_DEBUG_EXTRA_TRIE 0x0010000 |
fd41b2d1 KW |
1025 | #define RE_DEBUG_EXTRA_STATE 0x0080000 |
1026 | #define RE_DEBUG_EXTRA_OPTIMISE 0x0100000 | |
1027 | #define RE_DEBUG_EXTRA_BUFFERS 0x0400000 | |
1028 | #define RE_DEBUG_EXTRA_GPOS 0x0800000 | |
1029 | #define RE_DEBUG_EXTRA_DUMP_PRE_OPTIMIZE 0x1000000 | |
51bd1941 | 1030 | #define RE_DEBUG_EXTRA_WILDCARD 0x2000000 |
24b23f37 | 1031 | /* combined */ |
fd41b2d1 | 1032 | #define RE_DEBUG_EXTRA_STACK 0x0280000 |
be8e71aa | 1033 | |
55a0b497 | 1034 | #define RE_DEBUG_FLAG(x) (re_debug_flags & (x)) |
be8e71aa YO |
1035 | /* Compile */ |
1036 | #define DEBUG_COMPILE_r(x) DEBUG_r( \ | |
55a0b497 | 1037 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_COMPILE_MASK)) x ) |
be8e71aa | 1038 | #define DEBUG_PARSE_r(x) DEBUG_r( \ |
55a0b497 | 1039 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_COMPILE_PARSE)) x ) |
be8e71aa | 1040 | #define DEBUG_OPTIMISE_r(x) DEBUG_r( \ |
55a0b497 | 1041 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_COMPILE_OPTIMISE)) x ) |
be8e71aa | 1042 | #define DEBUG_DUMP_r(x) DEBUG_r( \ |
55a0b497 | 1043 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_COMPILE_DUMP)) x ) |
be8e71aa | 1044 | #define DEBUG_TRIE_COMPILE_r(x) DEBUG_r( \ |
55a0b497 | 1045 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_COMPILE_TRIE)) x ) |
f7819f85 | 1046 | #define DEBUG_FLAGS_r(x) DEBUG_r( \ |
55a0b497 | 1047 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_COMPILE_FLAGS)) x ) |
d9a72fcc | 1048 | #define DEBUG_TEST_r(x) DEBUG_r( \ |
55a0b497 | 1049 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_COMPILE_TEST)) x ) |
be8e71aa YO |
1050 | /* Execute */ |
1051 | #define DEBUG_EXECUTE_r(x) DEBUG_r( \ | |
55a0b497 | 1052 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXECUTE_MASK)) x ) |
be8e71aa | 1053 | #define DEBUG_INTUIT_r(x) DEBUG_r( \ |
55a0b497 | 1054 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXECUTE_INTUIT)) x ) |
be8e71aa | 1055 | #define DEBUG_MATCH_r(x) DEBUG_r( \ |
55a0b497 | 1056 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXECUTE_MATCH)) x ) |
be8e71aa | 1057 | #define DEBUG_TRIE_EXECUTE_r(x) DEBUG_r( \ |
55a0b497 | 1058 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXECUTE_TRIE)) x ) |
be8e71aa YO |
1059 | |
1060 | /* Extra */ | |
1061 | #define DEBUG_EXTRA_r(x) DEBUG_r( \ | |
55a0b497 | 1062 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXTRA_MASK)) x ) |
ab3bbdeb | 1063 | #define DEBUG_STATE_r(x) DEBUG_r( \ |
55a0b497 | 1064 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXTRA_STATE)) x ) |
24b23f37 | 1065 | #define DEBUG_STACK_r(x) DEBUG_r( \ |
55a0b497 | 1066 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXTRA_STACK)) x ) |
e7707071 | 1067 | #define DEBUG_BUFFERS_r(x) DEBUG_r( \ |
55a0b497 | 1068 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXTRA_BUFFERS)) x ) |
e7707071 | 1069 | |
a5ca303d | 1070 | #define DEBUG_OPTIMISE_MORE_r(x) DEBUG_r( \ |
dafb2544 | 1071 | if (DEBUG_v_TEST || ((RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE) == \ |
55a0b497 | 1072 | RE_DEBUG_FLAG(RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE))) x ) |
be8e71aa | 1073 | #define DEBUG_TRIE_COMPILE_MORE_r(x) DEBUG_TRIE_COMPILE_r( \ |
55a0b497 | 1074 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXTRA_TRIE)) x ) |
be8e71aa | 1075 | #define DEBUG_TRIE_EXECUTE_MORE_r(x) DEBUG_TRIE_EXECUTE_r( \ |
55a0b497 | 1076 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXTRA_TRIE)) x ) |
be8e71aa YO |
1077 | |
1078 | #define DEBUG_TRIE_r(x) DEBUG_r( \ | |
55a0b497 KW |
1079 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_COMPILE_TRIE \ |
1080 | | RE_DEBUG_EXECUTE_TRIE )) x ) | |
2c296965 | 1081 | #define DEBUG_GPOS_r(x) DEBUG_r( \ |
55a0b497 | 1082 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXTRA_GPOS)) x ) |
be8e71aa | 1083 | |
fd41b2d1 | 1084 | #define DEBUG_DUMP_PRE_OPTIMIZE_r(x) DEBUG_r( \ |
55a0b497 | 1085 | if (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXTRA_DUMP_PRE_OPTIMIZE)) x ) |
fd41b2d1 | 1086 | |
be8e71aa | 1087 | /* initialization */ |
1fe1d354 KW |
1088 | /* Get the debug flags for code not in regcomp.c nor regexec.c. This doesn't |
1089 | * initialize the variable if it isn't already there, instead it just assumes | |
1090 | * the flags are 0 */ | |
1091 | #define DECLARE_AND_GET_RE_DEBUG_FLAGS_NON_REGEX \ | |
1092 | volatile IV re_debug_flags = 0; PERL_UNUSED_VAR(re_debug_flags); \ | |
1093 | STMT_START { \ | |
1094 | SV * re_debug_flags_sv = NULL; \ | |
1095 | /* get_sv() can return NULL during global destruction. */ \ | |
1096 | re_debug_flags_sv = PL_curcop ? get_sv(RE_DEBUG_FLAGS, GV_ADD) : NULL; \ | |
1097 | if (re_debug_flags_sv && SvIOK(re_debug_flags_sv)) \ | |
1098 | re_debug_flags=SvIV(re_debug_flags_sv); \ | |
1099 | } STMT_END | |
1100 | ||
1101 | ||
a3621e74 | 1102 | #ifdef DEBUGGING |
ab3bbdeb | 1103 | |
1fe1d354 KW |
1104 | /* For use in regcomp.c and regexec.c, Get the debug flags, and initialize to |
1105 | * the defaults if not done already */ | |
619d7873 KW |
1106 | #define DECLARE_AND_GET_RE_DEBUG_FLAGS \ |
1107 | volatile IV re_debug_flags = 0; PERL_UNUSED_VAR(re_debug_flags); \ | |
64901762 | 1108 | DEBUG_r({ \ |
619d7873 KW |
1109 | SV * re_debug_flags_sv = NULL; \ |
1110 | /* get_sv() can return NULL during global destruction. */ \ | |
1111 | re_debug_flags_sv = PL_curcop ? get_sv(RE_DEBUG_FLAGS, GV_ADD) : NULL; \ | |
1112 | if (re_debug_flags_sv) { \ | |
1113 | if (!SvIOK(re_debug_flags_sv)) /* If doesnt exist set to default */\ | |
1114 | sv_setuv(re_debug_flags_sv, \ | |
1115 | /* These defaults should be kept in sync with re.pm */ \ | |
1116 | RE_DEBUG_COMPILE_DUMP | RE_DEBUG_EXECUTE_MASK ); \ | |
1117 | re_debug_flags=SvIV(re_debug_flags_sv); \ | |
1118 | } \ | |
64901762 | 1119 | }) |
ab3bbdeb | 1120 | |
51bd1941 KW |
1121 | #define isDEBUG_WILDCARD (DEBUG_v_TEST || RE_DEBUG_FLAG(RE_DEBUG_EXTRA_WILDCARD)) |
1122 | ||
eecd4d11 KW |
1123 | #define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2) \ |
1124 | const char * const rpv = \ | |
1125 | pv_pretty((dsv), (pv), (l), (m), \ | |
1126 | PL_colors[(c1)],PL_colors[(c2)], \ | |
c89df6cf | 1127 | PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) ); \ |
0df25f3d | 1128 | const int rlen = SvCUR(dsv) |
ab3bbdeb | 1129 | |
8ec77c3c | 1130 | /* This is currently unsed in the core */ |
eecd4d11 KW |
1131 | #define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m) \ |
1132 | const char * const rpv = \ | |
1133 | pv_pretty((dsv), (SvPV_nolen_const(sv)), (SvCUR(sv)), (m), \ | |
1134 | PL_colors[(c1)],PL_colors[(c2)], \ | |
c89df6cf | 1135 | PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) ) |
ab3bbdeb | 1136 | |
eecd4d11 KW |
1137 | #define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m) \ |
1138 | const char * const rpv = \ | |
1139 | pv_pretty((dsv), (pv), (l), (m), \ | |
1140 | PL_colors[0], PL_colors[1], \ | |
c89df6cf | 1141 | ( PERL_PV_PRETTY_QUOTE | PERL_PV_ESCAPE_RE | PERL_PV_ESCAPE_NONASCII | PERL_PV_PRETTY_ELLIPSES | \ |
ab3bbdeb | 1142 | ((isuni) ? PERL_PV_ESCAPE_UNI : 0)) \ |
0a0e9afc | 1143 | ) |
ab3bbdeb YO |
1144 | |
1145 | #define RE_SV_DUMPLEN(ItEm) (SvCUR(ItEm) - (SvTAIL(ItEm)!=0)) | |
1146 | #define RE_SV_TAIL(ItEm) (SvTAIL(ItEm) ? "$" : "") | |
1147 | ||
1148 | #else /* if not DEBUGGING */ | |
1149 | ||
82dc0782 KW |
1150 | #define DECLARE_AND_GET_RE_DEBUG_FLAGS dNOOP |
1151 | #define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2) dNOOP | |
ab3bbdeb | 1152 | #define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m) |
82dc0782 | 1153 | #define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m) dNOOP |
ab3bbdeb YO |
1154 | #define RE_SV_DUMPLEN(ItEm) |
1155 | #define RE_SV_TAIL(ItEm) | |
51bd1941 | 1156 | #define isDEBUG_WILDCARD 0 |
ab3bbdeb YO |
1157 | |
1158 | #endif /* DEBUG RELATED DEFINES */ | |
a3621e74 | 1159 | |
0bc0e44f | 1160 | #define FIRST_NON_ASCII_DECIMAL_DIGIT 0x660 /* ARABIC_INDIC_DIGIT_ZERO */ |
2ca74f27 | 1161 | |
64935bc6 | 1162 | typedef enum { |
1604cfb0 MS |
1163 | TRADITIONAL_BOUND = _CC_WORDCHAR, |
1164 | GCB_BOUND, | |
1165 | LB_BOUND, | |
1166 | SB_BOUND, | |
1167 | WB_BOUND | |
64935bc6 KW |
1168 | } bound_type; |
1169 | ||
13fcf652 | 1170 | /* This unpacks the FLAGS field of ANYOF[HR]x nodes. The value it contains |
3146c00a KW |
1171 | * gives the strict lower bound for the UTF-8 start byte of any code point |
1172 | * matchable by the node, and a loose upper bound as well. | |
1173 | * | |
1a737818 | 1174 | * The low bound is stored as 0xC0 + ((the upper 6 bits) >> 2) |
3146c00a KW |
1175 | * The loose upper bound is determined from the lowest 2 bits and the low bound |
1176 | * (called x) as follows: | |
1177 | * | |
1178 | * 11 The upper limit of the range can be as much as (EF - x) / 8 | |
1179 | * 10 The upper limit of the range can be as much as (EF - x) / 4 | |
1180 | * 01 The upper limit of the range can be as much as (EF - x) / 2 | |
1181 | * 00 The upper limit of the range can be as much as EF | |
1182 | * | |
0bf4efd7 KW |
1183 | * For motivation of this design, see commit message in |
1184 | * 3146c00a633e9cbed741e10146662fbcedfdb8d3 */ | |
3146c00a KW |
1185 | #ifdef EBCDIC |
1186 | # define MAX_ANYOF_HRx_BYTE 0xF4 | |
1187 | #else | |
1188 | # define MAX_ANYOF_HRx_BYTE 0xEF | |
1189 | #endif | |
1190 | #define LOWEST_ANYOF_HRx_BYTE(b) (((b) >> 2) + 0xC0) | |
1191 | #define HIGHEST_ANYOF_HRx_BYTE(b) \ | |
1192 | (LOWEST_ANYOF_HRx_BYTE(b) \ | |
1193 | + ((MAX_ANYOF_HRx_BYTE - LOWEST_ANYOF_HRx_BYTE(b)) >> ((b) & 3))) | |
1194 | ||
3dd7db29 JK |
1195 | #endif /* PERL_REGCOMP_H_ */ |
1196 | ||
e9a8c099 | 1197 | /* |
14d04a33 | 1198 | * ex: set ts=8 sts=4 sw=4 et: |
e9a8c099 | 1199 | */ |