Commit | Line | Data |
---|---|---|
a0d0e21e | 1 | /* regcomp.h |
d6376244 | 2 | * |
4bb101f2 | 3 | * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, |
1d325971 | 4 | * 2000, 2001, 2002, 2003, 2005 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 LW |
9 | */ |
10 | ||
c277df42 IZ |
11 | typedef OP OP_4tree; /* Will be redefined later. */ |
12 | ||
07be1b83 YO |
13 | |
14 | #define PERL_ENABLE_TRIE_OPTIMISATION 1 | |
15 | #define PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION 1 | |
16 | #define PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS 0 | |
17 | ||
a687059c LW |
18 | /* |
19 | * The "internal use only" fields in regexp.h are present to pass info from | |
20 | * compile to execute that permits the execute phase to run lots faster on | |
21 | * simple cases. They are: | |
22 | * | |
bd61b366 | 23 | * regstart sv that must begin a match; NULL if none obvious |
a687059c LW |
24 | * reganch is the match anchored (at beginning-of-line only)? |
25 | * regmust string (pointer into program) that match must include, or NULL | |
79072805 | 26 | * [regmust changed to SV* for bminstr()--law] |
a687059c LW |
27 | * regmlen length of regmust string |
28 | * [regmlen not used currently] | |
29 | * | |
30 | * Regstart and reganch permit very fast decisions on suitable starting points | |
31 | * for a match, cutting down the work a lot. Regmust permits fast rejection | |
32 | * of lines that cannot possibly match. The regmust tests are costly enough | |
e50aee73 | 33 | * that pregcomp() supplies a regmust only if the r.e. contains something |
a687059c LW |
34 | * potentially expensive (at present, the only such thing detected is * or + |
35 | * at the start of the r.e., which can involve a lot of backup). Regmlen is | |
e50aee73 | 36 | * supplied because the test in pregexec() needs it and pregcomp() is computing |
a687059c LW |
37 | * it anyway. |
38 | * [regmust is now supplied always. The tests that use regmust have a | |
39 | * heuristic that disables the test if it usually matches.] | |
40 | * | |
41 | * [In fact, we now use regmust in many cases to locate where the search | |
42 | * starts in the string, so if regback is >= 0, the regmust search is never | |
43 | * wasted effort. The regback variable says how many characters back from | |
44 | * where regmust matched is the earliest possible start of the match. | |
45 | * For instance, /[a-z].foo/ has a regmust of 'foo' and a regback of 2.] | |
46 | */ | |
47 | ||
48 | /* | |
49 | * Structure for regexp "program". This is essentially a linear encoding | |
50 | * of a nondeterministic finite-state machine (aka syntax charts or | |
51 | * "railroad normal form" in parsing technology). Each node is an opcode | |
52 | * plus a "next" pointer, possibly plus an operand. "Next" pointers of | |
53 | * all nodes except BRANCH implement concatenation; a "next" pointer with | |
54 | * a BRANCH on both ends of it is connecting two alternatives. (Here we | |
55 | * have one of the subtle syntax dependencies: an individual BRANCH (as | |
56 | * opposed to a collection of them) is never concatenated with anything | |
57 | * because of operator precedence.) The operand of some types of node is | |
58 | * a literal string; for others, it is a node leading into a sub-FSM. In | |
59 | * particular, the operand of a BRANCH node is the first node of the branch. | |
60 | * (NB this is *not* a tree structure: the tail of the branch connects | |
a3621e74 YO |
61 | * to the thing following the set of BRANCHes.) The opcodes are defined |
62 | * in regnodes.h which is generated from regcomp.sym by regcomp.pl. | |
a687059c LW |
63 | */ |
64 | ||
a687059c LW |
65 | /* |
66 | * A node is one char of opcode followed by two chars of "next" pointer. | |
67 | * "Next" pointers are stored as two 8-bit pieces, high order first. The | |
68 | * value is a positive offset from the opcode of the node containing it. | |
69 | * An operand, if any, simply follows the node. (Note that much of the | |
70 | * code generation knows about this implicit relationship.) | |
71 | * | |
72 | * Using two bytes for the "next" pointer is vast overkill for most things, | |
73 | * but allows patterns to get big without disasters. | |
74 | * | |
e91177ed | 75 | * [The "next" pointer is always aligned on an even |
a687059c LW |
76 | * boundary, and reads the offset directly as a short. Also, there is no |
77 | * special test to reverse the sign of BACK pointers since the offset is | |
78 | * stored negative.] | |
79 | */ | |
80 | ||
c277df42 | 81 | struct regnode_string { |
cd439c50 | 82 | U8 str_len; |
c277df42 IZ |
83 | U8 type; |
84 | U16 next_off; | |
cd439c50 | 85 | char string[1]; |
c277df42 | 86 | }; |
a687059c | 87 | |
c277df42 IZ |
88 | struct regnode_1 { |
89 | U8 flags; | |
90 | U8 type; | |
91 | U16 next_off; | |
92 | U32 arg1; | |
93 | }; | |
94 | ||
95 | struct regnode_2 { | |
96 | U8 flags; | |
97 | U8 type; | |
98 | U16 next_off; | |
99 | U16 arg1; | |
100 | U16 arg2; | |
101 | }; | |
102 | ||
936ed897 | 103 | #define ANYOF_BITMAP_SIZE 32 /* 256 b/(8 b/B) */ |
ffc61ed2 | 104 | #define ANYOF_CLASSBITMAP_SIZE 4 /* up to 32 (8*4) named classes */ |
936ed897 IZ |
105 | |
106 | struct regnode_charclass { | |
107 | U8 flags; | |
108 | U8 type; | |
109 | U16 next_off; | |
ffc61ed2 | 110 | U32 arg1; |
19103b91 | 111 | char bitmap[ANYOF_BITMAP_SIZE]; /* only compile-time */ |
936ed897 IZ |
112 | }; |
113 | ||
19103b91 JH |
114 | struct regnode_charclass_class { /* has [[:blah:]] classes */ |
115 | U8 flags; /* should have ANYOF_CLASS here */ | |
936ed897 IZ |
116 | U8 type; |
117 | U16 next_off; | |
ffc61ed2 | 118 | U32 arg1; |
19103b91 JH |
119 | char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time */ |
120 | char classflags[ANYOF_CLASSBITMAP_SIZE]; /* and run-time */ | |
936ed897 IZ |
121 | }; |
122 | ||
9c7e81e8 AD |
123 | /* XXX fix this description. |
124 | Impose a limit of REG_INFTY on various pattern matching operations | |
125 | to limit stack growth and to avoid "infinite" recursions. | |
126 | */ | |
127 | /* The default size for REG_INFTY is I16_MAX, which is the same as | |
128 | SHORT_MAX (see perl.h). Unfortunately I16 isn't necessarily 16 bits | |
129 | (see handy.h). On the Cray C90, sizeof(short)==4 and hence I16_MAX is | |
130 | ((1<<31)-1), while on the Cray T90, sizeof(short)==8 and I16_MAX is | |
131 | ((1<<63)-1). To limit stack growth to reasonable sizes, supply a | |
132 | smaller default. | |
133 | --Andy Dougherty 11 June 1998 | |
134 | */ | |
135 | #if SHORTSIZE > 2 | |
136 | # ifndef REG_INFTY | |
137 | # define REG_INFTY ((1<<15)-1) | |
138 | # endif | |
139 | #endif | |
140 | ||
141 | #ifndef REG_INFTY | |
83cfe48c IZ |
142 | # define REG_INFTY I16_MAX |
143 | #endif | |
a687059c | 144 | |
e91177ed IZ |
145 | #define ARG_VALUE(arg) (arg) |
146 | #define ARG__SET(arg,val) ((arg) = (val)) | |
c277df42 | 147 | |
e8e3b0fb JH |
148 | #undef ARG |
149 | #undef ARG1 | |
150 | #undef ARG2 | |
151 | ||
c277df42 IZ |
152 | #define ARG(p) ARG_VALUE(ARG_LOC(p)) |
153 | #define ARG1(p) ARG_VALUE(ARG1_LOC(p)) | |
154 | #define ARG2(p) ARG_VALUE(ARG2_LOC(p)) | |
155 | #define ARG_SET(p, val) ARG__SET(ARG_LOC(p), (val)) | |
156 | #define ARG1_SET(p, val) ARG__SET(ARG1_LOC(p), (val)) | |
157 | #define ARG2_SET(p, val) ARG__SET(ARG2_LOC(p), (val)) | |
158 | ||
e8e3b0fb JH |
159 | #undef NEXT_OFF |
160 | #undef NODE_ALIGN | |
161 | ||
075abff3 AL |
162 | #define NEXT_OFF(p) ((p)->next_off) |
163 | #define NODE_ALIGN(node) | |
164 | #define NODE_ALIGN_FILL(node) ((node)->flags = 0xde) /* deadbeef */ | |
a687059c | 165 | |
c277df42 IZ |
166 | #define SIZE_ALIGN NODE_ALIGN |
167 | ||
e8e3b0fb JH |
168 | #undef OP |
169 | #undef OPERAND | |
170 | #undef MASK | |
171 | #undef STRING | |
172 | ||
e91177ed IZ |
173 | #define OP(p) ((p)->type) |
174 | #define OPERAND(p) (((struct regnode_string *)p)->string) | |
cd439c50 IZ |
175 | #define MASK(p) ((char*)OPERAND(p)) |
176 | #define STR_LEN(p) (((struct regnode_string *)p)->str_len) | |
177 | #define STRING(p) (((struct regnode_string *)p)->string) | |
178 | #define STR_SZ(l) ((l + sizeof(regnode) - 1) / sizeof(regnode)) | |
179 | #define NODE_SZ_STR(p) (STR_SZ(STR_LEN(p))+1) | |
180 | ||
e8e3b0fb JH |
181 | #undef NODE_ALIGN |
182 | #undef ARG_LOC | |
183 | #undef NEXTOPER | |
184 | #undef PREVOPER | |
185 | ||
e91177ed IZ |
186 | #define NODE_ALIGN(node) |
187 | #define ARG_LOC(p) (((struct regnode_1 *)p)->arg1) | |
188 | #define ARG1_LOC(p) (((struct regnode_2 *)p)->arg1) | |
189 | #define ARG2_LOC(p) (((struct regnode_2 *)p)->arg2) | |
190 | #define NODE_STEP_REGNODE 1 /* sizeof(regnode)/sizeof(regnode) */ | |
191 | #define EXTRA_STEP_2ARGS EXTRA_SIZE(struct regnode_2) | |
192 | ||
193 | #define NODE_STEP_B 4 | |
c277df42 IZ |
194 | |
195 | #define NEXTOPER(p) ((p) + NODE_STEP_REGNODE) | |
196 | #define PREVOPER(p) ((p) - NODE_STEP_REGNODE) | |
197 | ||
e91177ed | 198 | #define FILL_ADVANCE_NODE(ptr, op) STMT_START { \ |
c277df42 | 199 | (ptr)->type = op; (ptr)->next_off = 0; (ptr)++; } STMT_END |
e91177ed | 200 | #define FILL_ADVANCE_NODE_ARG(ptr, op, arg) STMT_START { \ |
c277df42 | 201 | ARG_SET(ptr, arg); FILL_ADVANCE_NODE(ptr, op); (ptr) += 1; } STMT_END |
a687059c | 202 | |
22c35a8c | 203 | #define REG_MAGIC 0234 |
a687059c | 204 | |
830247a4 | 205 | #define SIZE_ONLY (RExC_emit == &PL_regdummy) |
c277df42 | 206 | |
936ed897 | 207 | /* Flags for node->flags of ANYOF */ |
b8c5462f | 208 | |
19103b91 | 209 | #define ANYOF_CLASS 0x08 /* has [[:blah:]] classes */ |
ffc61ed2 JH |
210 | #define ANYOF_INVERT 0x04 |
211 | #define ANYOF_FOLD 0x02 | |
212 | #define ANYOF_LOCALE 0x01 | |
b8c5462f | 213 | |
653099ff | 214 | /* Used for regstclass only */ |
ffc61ed2 JH |
215 | #define ANYOF_EOS 0x10 /* Can match an empty string too */ |
216 | ||
217 | /* There is a character or a range past 0xff */ | |
218 | #define ANYOF_UNICODE 0x20 | |
1aa99e6b | 219 | #define ANYOF_UNICODE_ALL 0x40 /* Can match any char past 0xff */ |
ffc61ed2 | 220 | |
4f66b38d HS |
221 | /* size of node is large (includes class pointer) */ |
222 | #define ANYOF_LARGE 0x80 | |
223 | ||
ffc61ed2 JH |
224 | /* Are there any runtime flags on in this node? */ |
225 | #define ANYOF_RUNTIME(s) (ANYOF_FLAGS(s) & 0x0f) | |
226 | ||
227 | #define ANYOF_FLAGS_ALL 0xff | |
653099ff | 228 | |
936ed897 | 229 | /* Character classes for node->classflags of ANYOF */ |
653099ff GS |
230 | /* Should be synchronized with a table in regprop() */ |
231 | /* 2n should pair with 2n+1 */ | |
b8c5462f | 232 | |
289d4f09 | 233 | #define ANYOF_ALNUM 0 /* \w, PL_utf8_alnum, utf8::IsWord, ALNUM */ |
b8c5462f | 234 | #define ANYOF_NALNUM 1 |
aaa51d5e | 235 | #define ANYOF_SPACE 2 /* \s */ |
b8c5462f JH |
236 | #define ANYOF_NSPACE 3 |
237 | #define ANYOF_DIGIT 4 | |
238 | #define ANYOF_NDIGIT 5 | |
289d4f09 | 239 | #define ANYOF_ALNUMC 6 /* isalnum(3), utf8::IsAlnum, ALNUMC */ |
b8c5462f JH |
240 | #define ANYOF_NALNUMC 7 |
241 | #define ANYOF_ALPHA 8 | |
242 | #define ANYOF_NALPHA 9 | |
243 | #define ANYOF_ASCII 10 | |
244 | #define ANYOF_NASCII 11 | |
245 | #define ANYOF_CNTRL 12 | |
246 | #define ANYOF_NCNTRL 13 | |
247 | #define ANYOF_GRAPH 14 | |
248 | #define ANYOF_NGRAPH 15 | |
249 | #define ANYOF_LOWER 16 | |
250 | #define ANYOF_NLOWER 17 | |
251 | #define ANYOF_PRINT 18 | |
252 | #define ANYOF_NPRINT 19 | |
253 | #define ANYOF_PUNCT 20 | |
254 | #define ANYOF_NPUNCT 21 | |
255 | #define ANYOF_UPPER 22 | |
256 | #define ANYOF_NUPPER 23 | |
257 | #define ANYOF_XDIGIT 24 | |
258 | #define ANYOF_NXDIGIT 25 | |
aaa51d5e JF |
259 | #define ANYOF_PSXSPC 26 /* POSIX space: \s plus the vertical tab */ |
260 | #define ANYOF_NPSXSPC 27 | |
ffc61ed2 | 261 | #define ANYOF_BLANK 28 /* GNU extension: space and tab: non-vertical space */ |
289d4f09 | 262 | #define ANYOF_NBLANK 29 |
b8c5462f | 263 | |
aaa51d5e | 264 | #define ANYOF_MAX 32 |
b8c5462f JH |
265 | |
266 | /* Backward source code compatibility. */ | |
267 | ||
268 | #define ANYOF_ALNUML ANYOF_ALNUM | |
269 | #define ANYOF_NALNUML ANYOF_NALNUM | |
270 | #define ANYOF_SPACEL ANYOF_SPACE | |
271 | #define ANYOF_NSPACEL ANYOF_NSPACE | |
272 | ||
273 | /* Utility macros for the bitmap and classes of ANYOF */ | |
274 | ||
936ed897 IZ |
275 | #define ANYOF_SIZE (sizeof(struct regnode_charclass)) |
276 | #define ANYOF_CLASS_SIZE (sizeof(struct regnode_charclass_class)) | |
b8c5462f | 277 | |
936ed897 | 278 | #define ANYOF_FLAGS(p) ((p)->flags) |
b8c5462f JH |
279 | |
280 | #define ANYOF_BIT(c) (1 << ((c) & 7)) | |
281 | ||
936ed897 | 282 | #define ANYOF_CLASS_BYTE(p, c) (((struct regnode_charclass_class*)(p))->classflags[((c) >> 3) & 3]) |
b8c5462f JH |
283 | #define ANYOF_CLASS_SET(p, c) (ANYOF_CLASS_BYTE(p, c) |= ANYOF_BIT(c)) |
284 | #define ANYOF_CLASS_CLEAR(p, c) (ANYOF_CLASS_BYTE(p, c) &= ~ANYOF_BIT(c)) | |
285 | #define ANYOF_CLASS_TEST(p, c) (ANYOF_CLASS_BYTE(p, c) & ANYOF_BIT(c)) | |
286 | ||
936ed897 IZ |
287 | #define ANYOF_CLASS_ZERO(ret) Zero(((struct regnode_charclass_class*)(ret))->classflags, ANYOF_CLASSBITMAP_SIZE, char) |
288 | #define ANYOF_BITMAP_ZERO(ret) Zero(((struct regnode_charclass*)(ret))->bitmap, ANYOF_BITMAP_SIZE, char) | |
289 | ||
290 | #define ANYOF_BITMAP(p) (((struct regnode_charclass*)(p))->bitmap) | |
291 | #define ANYOF_BITMAP_BYTE(p, c) (ANYOF_BITMAP(p)[((c) >> 3) & 31]) | |
b8c5462f JH |
292 | #define ANYOF_BITMAP_SET(p, c) (ANYOF_BITMAP_BYTE(p, c) |= ANYOF_BIT(c)) |
293 | #define ANYOF_BITMAP_CLEAR(p,c) (ANYOF_BITMAP_BYTE(p, c) &= ~ANYOF_BIT(c)) | |
294 | #define ANYOF_BITMAP_TEST(p, c) (ANYOF_BITMAP_BYTE(p, c) & ANYOF_BIT(c)) | |
295 | ||
f8bef550 NC |
296 | #define ANYOF_BITMAP_SETALL(p) \ |
297 | memset (ANYOF_BITMAP(p), 255, ANYOF_BITMAP_SIZE) | |
298 | #define ANYOF_BITMAP_CLEARALL(p) \ | |
299 | Zero (ANYOF_BITMAP(p), ANYOF_BITMAP_SIZE) | |
300 | /* Check that all 256 bits are all set. Used in S_cl_is_anything() */ | |
301 | #define ANYOF_BITMAP_TESTALLSET(p) \ | |
302 | 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) | |
303 | ||
936ed897 IZ |
304 | #define ANYOF_SKIP ((ANYOF_SIZE - 1)/sizeof(regnode)) |
305 | #define ANYOF_CLASS_SKIP ((ANYOF_CLASS_SIZE - 1)/sizeof(regnode)) | |
306 | #define ANYOF_CLASS_ADD_SKIP (ANYOF_CLASS_SKIP - ANYOF_SKIP) | |
c277df42 | 307 | |
a687059c LW |
308 | /* |
309 | * Utility definitions. | |
310 | */ | |
2304df62 | 311 | #ifndef CHARMASK |
075abff3 | 312 | # define UCHARAT(p) ((int)*(const U8*)(p)) |
a687059c | 313 | #else |
075abff3 | 314 | # define UCHARAT(p) ((int)*(p)&CHARMASK) |
a687059c | 315 | #endif |
a687059c | 316 | |
c277df42 IZ |
317 | #define EXTRA_SIZE(guy) ((sizeof(guy)-1)/sizeof(struct regnode)) |
318 | ||
3dab1dad YO |
319 | #define REG_SEEN_ZERO_LEN 0x00000001 |
320 | #define REG_SEEN_LOOKBEHIND 0x00000002 | |
321 | #define REG_SEEN_GPOS 0x00000004 | |
322 | #define REG_SEEN_EVAL 0x00000008 | |
323 | #define REG_SEEN_CANY 0x00000010 | |
cce850e4 | 324 | #define REG_SEEN_SANY REG_SEEN_CANY /* src bckwrd cmpt */ |
d09b2d29 | 325 | |
73c4f7a1 GS |
326 | START_EXTERN_C |
327 | ||
d09b2d29 IZ |
328 | #include "regnodes.h" |
329 | ||
4d61ec05 | 330 | /* The following have no fixed length. U8 so we can do strchr() on it. */ |
d09b2d29 | 331 | #ifndef DOINIT |
4d61ec05 | 332 | EXTCONST U8 PL_varies[]; |
d09b2d29 | 333 | #else |
4d61ec05 | 334 | EXTCONST U8 PL_varies[] = { |
0111c4fd | 335 | BRANCH, BACK, STAR, PLUS, CURLY, CURLYX, REF, REFF, REFFL, |
a0ed51b3 | 336 | WHILEM, CURLYM, CURLYN, BRANCHJ, IFTHEN, SUSPEND, CLUMP, 0 |
c277df42 | 337 | }; |
d09b2d29 | 338 | #endif |
c277df42 | 339 | |
4d61ec05 GS |
340 | /* The following always have a length of 1. U8 we can do strchr() on it. */ |
341 | /* (Note that length 1 means "one character" under UTF8, not "one octet".) */ | |
d09b2d29 | 342 | #ifndef DOINIT |
4d61ec05 | 343 | EXTCONST U8 PL_simple[]; |
d09b2d29 | 344 | #else |
4d61ec05 | 345 | EXTCONST U8 PL_simple[] = { |
f33976b4 | 346 | REG_ANY, SANY, CANY, |
ffc61ed2 JH |
347 | ANYOF, |
348 | ALNUM, ALNUML, | |
349 | NALNUM, NALNUML, | |
350 | SPACE, SPACEL, | |
351 | NSPACE, NSPACEL, | |
352 | DIGIT, NDIGIT, | |
353 | 0 | |
c277df42 IZ |
354 | }; |
355 | #endif | |
356 | ||
73c4f7a1 | 357 | END_EXTERN_C |
cad2e5aa JH |
358 | |
359 | typedef struct re_scream_pos_data_s | |
360 | { | |
361 | char **scream_olds; /* match pos */ | |
362 | I32 *scream_pos; /* Internal iterator of scream. */ | |
363 | } re_scream_pos_data; | |
364 | ||
261faec3 MJD |
365 | /* .what is a character array with one character for each member of .data |
366 | * The character describes the function of the corresponding .data item: | |
0111c4fd | 367 | * f - start-class data for regstclass optimization |
261faec3 MJD |
368 | * n - Root of op tree for (?{EVAL}) item |
369 | * o - Start op for (?{EVAL}) item | |
3dab1dad | 370 | * p - Pad for (?{EVAL}) item |
9e55ce06 JH |
371 | * s - swash for unicode-style character class, and the multicharacter |
372 | * strings resulting from casefolding the single-character entries | |
373 | * in the character class | |
a3621e74 | 374 | * t - trie struct |
07be1b83 | 375 | * T - aho-trie struct |
261faec3 | 376 | * 20010712 mjd@plover.com |
c3613ed7 | 377 | * (Remember to update re_dup() and pregfree() if you add any items.) |
261faec3 | 378 | */ |
cad2e5aa JH |
379 | struct reg_data { |
380 | U32 count; | |
381 | U8 *what; | |
382 | void* data[1]; | |
383 | }; | |
384 | ||
385 | struct reg_substr_datum { | |
386 | I32 min_offset; | |
387 | I32 max_offset; | |
33b8afdf JH |
388 | SV *substr; /* non-utf8 variant */ |
389 | SV *utf8_substr; /* utf8 variant */ | |
cad2e5aa JH |
390 | }; |
391 | ||
392 | struct reg_substr_data { | |
393 | struct reg_substr_datum data[3]; /* Actual array */ | |
394 | }; | |
395 | ||
396 | #define anchored_substr substrs->data[0].substr | |
33b8afdf | 397 | #define anchored_utf8 substrs->data[0].utf8_substr |
cad2e5aa JH |
398 | #define anchored_offset substrs->data[0].min_offset |
399 | #define float_substr substrs->data[1].substr | |
33b8afdf | 400 | #define float_utf8 substrs->data[1].utf8_substr |
cad2e5aa JH |
401 | #define float_min_offset substrs->data[1].min_offset |
402 | #define float_max_offset substrs->data[1].max_offset | |
403 | #define check_substr substrs->data[2].substr | |
33b8afdf | 404 | #define check_utf8 substrs->data[2].utf8_substr |
cad2e5aa JH |
405 | #define check_offset_min substrs->data[2].min_offset |
406 | #define check_offset_max substrs->data[2].max_offset | |
a3621e74 YO |
407 | |
408 | ||
409 | ||
410 | /* trie related stuff */ | |
5d9a96ca | 411 | |
a3621e74 YO |
412 | /* a transition record for the state machine. the |
413 | check field determines which state "owns" the | |
414 | transition. the char the transition is for is | |
415 | determined by offset from the owning states base | |
416 | field. the next field determines which state | |
417 | is to be transitioned to if any. | |
418 | */ | |
419 | struct _reg_trie_trans { | |
420 | U32 next; | |
421 | U32 check; | |
422 | }; | |
423 | ||
424 | /* a transition list element for the list based representation */ | |
425 | struct _reg_trie_trans_list_elem { | |
426 | U16 forid; | |
427 | U32 newstate; | |
428 | }; | |
429 | typedef struct _reg_trie_trans_list_elem reg_trie_trans_le; | |
430 | ||
431 | /* a state for compressed nodes. base is an offset | |
432 | into an array of reg_trie_trans array. If wordnum is | |
433 | nonzero the state is accepting. if base is zero then | |
434 | the state has no children (and will be accepting) | |
435 | */ | |
436 | struct _reg_trie_state { | |
437 | U16 wordnum; | |
438 | union { | |
439 | U32 base; | |
440 | reg_trie_trans_le* list; | |
441 | } trans; | |
442 | }; | |
443 | ||
444 | ||
445 | ||
a3621e74 YO |
446 | typedef struct _reg_trie_state reg_trie_state; |
447 | typedef struct _reg_trie_trans reg_trie_trans; | |
448 | ||
449 | ||
450 | /* anything in here that needs to be freed later | |
8e11feef | 451 | should be dealt with in pregfree */ |
a3621e74 YO |
452 | struct _reg_trie_data { |
453 | U16 uniquecharcount; | |
cc601c31 | 454 | U32 lasttrans; |
a3621e74 YO |
455 | U16 *charmap; |
456 | HV *widecharmap; | |
457 | reg_trie_state *states; | |
458 | reg_trie_trans *trans; | |
3dab1dad | 459 | char *bitmap; |
a3621e74 | 460 | U32 refcount; |
3dab1dad | 461 | U32 startstate; |
8e11feef RGS |
462 | STRLEN minlen; |
463 | STRLEN maxlen; | |
07be1b83 YO |
464 | U32 *wordlen; |
465 | U32 laststate; /* Build only */ | |
a3621e74 | 466 | #ifdef DEBUGGING |
3dab1dad YO |
467 | U16 wordcount; /* Build only */ |
468 | STRLEN charcount; /* Build only */ | |
a3621e74 YO |
469 | AV *words; |
470 | AV *revcharmap; | |
471 | #endif | |
472 | }; | |
a3621e74 YO |
473 | typedef struct _reg_trie_data reg_trie_data; |
474 | ||
07be1b83 YO |
475 | struct _reg_ac_data { |
476 | U32 *fail; | |
477 | reg_trie_state *states; | |
478 | reg_trie_data *trie; | |
479 | U32 refcount; | |
480 | }; | |
481 | typedef struct _reg_ac_data reg_ac_data; | |
482 | ||
3dab1dad YO |
483 | /* ANY_BIT doesnt use the structure, so we can borrow it here. |
484 | This is simpler than refactoring all of it as wed end up with | |
485 | three different sets... */ | |
486 | ||
487 | #define TRIE_BITMAP(p) (((reg_trie_data *)(p))->bitmap) | |
488 | #define TRIE_BITMAP_BYTE(p, c) (TRIE_BITMAP(p)[((c) >> 3) & 31]) | |
489 | #define TRIE_BITMAP_SET(p, c) (TRIE_BITMAP_BYTE(p, c) |= ANYOF_BIT(c)) | |
490 | #define TRIE_BITMAP_CLEAR(p,c) (TRIE_BITMAP_BYTE(p, c) &= ~ANYOF_BIT(c)) | |
491 | #define TRIE_BITMAP_TEST(p, c) (TRIE_BITMAP_BYTE(p, c) & ANYOF_BIT(c)) | |
492 | ||
493 | ||
a3621e74 YO |
494 | /* these defines assume uniquecharcount is the correct variable, and state may be evaluated twice */ |
495 | #define TRIE_NODENUM(state) (((state)-1)/(trie->uniquecharcount)+1) | |
496 | #define SAFE_TRIE_NODENUM(state) ((state) ? (((state)-1)/(trie->uniquecharcount)+1) : (state)) | |
497 | #define TRIE_NODEIDX(state) ((state) ? (((state)-1)*(trie->uniquecharcount)+1) : (state)) | |
498 | ||
3dab1dad YO |
499 | #ifdef DEBUGGING |
500 | #define TRIE_WORDCOUNT(trie) ((trie)->wordcount) | |
501 | #define TRIE_CHARCOUNT(trie) ((trie)->charcount) | |
502 | #define TRIE_LASTSTATE(trie) ((trie)->laststate) | |
503 | #define TRIE_REVCHARMAP(trie) ((trie)->revcharmap) | |
504 | #else | |
505 | #define TRIE_WORDCOUNT(trie) (trie_wordcount) | |
506 | #define TRIE_CHARCOUNT(trie) (trie_charcount) | |
07be1b83 YO |
507 | /*#define TRIE_LASTSTATE(trie) (trie_laststate)*/ |
508 | #define TRIE_LASTSTATE(trie) ((trie)->laststate) | |
3dab1dad YO |
509 | #define TRIE_REVCHARMAP(trie) (trie_revcharmap) |
510 | #endif | |
a3621e74 | 511 | |
0111c4fd RGS |
512 | #define RE_TRIE_MAXBUF_INIT 65536 |
513 | #define RE_TRIE_MAXBUF_NAME "\022E_TRIE_MAXBUF" | |
a3621e74 YO |
514 | #define RE_DEBUG_FLAGS "\022E_DEBUG_FLAGS" |
515 | ||
516 | /* If you change these be sure to update ext/re/re.pm as well */ | |
3dab1dad YO |
517 | #define RE_DEBUG_COMPILE 0x0001 |
518 | #define RE_DEBUG_EXECUTE 0x0002 | |
519 | #define RE_DEBUG_TRIE_COMPILE 0x0004 | |
520 | #define RE_DEBUG_TRIE_EXECUTE 0x0008 | |
521 | #define RE_DEBUG_TRIE_MORE 0x0010 | |
522 | #define RE_DEBUG_OPTIMISE 0x0020 | |
523 | #define RE_DEBUG_OFFSETS 0x0040 | |
524 | #define RE_DEBUG_PARSE 0x0080 | |
07be1b83 YO |
525 | #define RE_DEBUG_OFFSETS_DEBUG 0x0100 |
526 | #define RE_DEBUG_OLD_OFFSETS 0x0200 | |
3dab1dad | 527 | |
a3621e74 | 528 | |
3dab1dad | 529 | #define DEBUG_PARSE_r(x) DEBUG_r( if (SvIV(re_debug_flags) & RE_DEBUG_PARSE) x ) |
a3621e74 YO |
530 | #define DEBUG_OPTIMISE_r(x) DEBUG_r( if (SvIV(re_debug_flags) & RE_DEBUG_OPTIMISE) x ) |
531 | #define DEBUG_EXECUTE_r(x) DEBUG_r( if (SvIV(re_debug_flags) & RE_DEBUG_EXECUTE) x ) | |
532 | #define DEBUG_COMPILE_r(x) DEBUG_r( if (SvIV(re_debug_flags) & RE_DEBUG_COMPILE) x ) | |
533 | #define DEBUG_OFFSETS_r(x) DEBUG_r( if (SvIV(re_debug_flags) & RE_DEBUG_OFFSETS) x ) | |
07be1b83 YO |
534 | #define DEBUG_OLD_OFFSETS_r(x) DEBUG_r( if (SvIV(re_debug_flags) & RE_DEBUG_OLD_OFFSETS) x ) |
535 | ||
a3621e74 YO |
536 | #define DEBUG_TRIE_r(x) DEBUG_r( \ |
537 | if (SvIV(re_debug_flags) & RE_DEBUG_TRIE_COMPILE \ | |
538 | || SvIV(re_debug_flags) & RE_DEBUG_TRIE_EXECUTE ) \ | |
539 | x \ | |
540 | ) | |
541 | #define DEBUG_TRIE_EXECUTE_r(x) \ | |
542 | DEBUG_r( if (SvIV(re_debug_flags) & RE_DEBUG_TRIE_EXECUTE) x ) | |
543 | ||
544 | #define DEBUG_TRIE_COMPILE_r(x) \ | |
545 | DEBUG_r( if (SvIV(re_debug_flags) & RE_DEBUG_TRIE_COMPILE) x ) | |
546 | ||
547 | #define DEBUG_TRIE_EXECUTE_MORE_r(x) \ | |
548 | DEBUG_TRIE_EXECUTE_r( if (SvIV(re_debug_flags) & RE_DEBUG_TRIE_MORE) x ) | |
549 | ||
550 | #define DEBUG_TRIE_COMPILE_MORE_r(x) \ | |
551 | DEBUG_TRIE_COMPILE_r( if (SvIV(re_debug_flags) & RE_DEBUG_TRIE_MORE) x ) | |
552 | ||
e7f343b6 | 553 | /* get_sv() can return NULL during global destruction. */ |
a3621e74 YO |
554 | #define GET_RE_DEBUG_FLAGS DEBUG_r( \ |
555 | re_debug_flags=get_sv(RE_DEBUG_FLAGS, 1); \ | |
e7f343b6 | 556 | if (re_debug_flags && !SvIOK(re_debug_flags)) { \ |
3dab1dad | 557 | sv_setiv(re_debug_flags, RE_DEBUG_COMPILE | RE_DEBUG_EXECUTE ); \ |
a3621e74 YO |
558 | } \ |
559 | ) | |
560 | ||
07be1b83 YO |
561 | #define MJD_OFFSET_DEBUG(x) DEBUG_r( \ |
562 | if (SvIV(re_debug_flags) & RE_DEBUG_OFFSETS_DEBUG) \ | |
563 | Perl_warn_nocontext x \ | |
564 | ) | |
a3621e74 YO |
565 | |
566 | #ifdef DEBUGGING | |
e1ec3a88 | 567 | #define GET_RE_DEBUG_FLAGS_DECL SV *re_debug_flags = NULL; GET_RE_DEBUG_FLAGS; |
a3621e74 YO |
568 | #else |
569 | #define GET_RE_DEBUG_FLAGS_DECL | |
570 | #endif | |
571 | ||
572 |