| 1 | /* regcomp.h |
| 2 | * |
| 3 | * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, |
| 4 | * 2000, 2001, 2002, 2003, 2005, 2006, 2007, by Larry Wall and others |
| 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 | * |
| 9 | */ |
| 10 | #include "regcharclass.h" |
| 11 | |
| 12 | /* Convert branch sequences to more efficient trie ops? */ |
| 13 | #define PERL_ENABLE_TRIE_OPTIMISATION 1 |
| 14 | |
| 15 | /* Be really aggressive about optimising patterns with trie sequences? */ |
| 16 | #define PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION 1 |
| 17 | |
| 18 | /* Should the optimiser take positive assertions into account? */ |
| 19 | #define PERL_ENABLE_POSITIVE_ASSERTION_STUDY 0 |
| 20 | |
| 21 | /* Not for production use: */ |
| 22 | #define PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS 0 |
| 23 | |
| 24 | /* Activate offsets code - set to if 1 to enable */ |
| 25 | #ifdef DEBUGGING |
| 26 | #define RE_TRACK_PATTERN_OFFSETS |
| 27 | #endif |
| 28 | |
| 29 | /* Unless the next line is uncommented it is illegal to combine lazy |
| 30 | matching with possessive matching. Frankly it doesn't make much sense |
| 31 | to allow it as X*?+ matches nothing, X+?+ matches a single char only, |
| 32 | and X{min,max}?+ matches min times only. |
| 33 | */ |
| 34 | /* #define REG_ALLOW_MINMOD_SUSPEND */ |
| 35 | |
| 36 | /* |
| 37 | * The "internal use only" fields in regexp.h are present to pass info from |
| 38 | * compile to execute that permits the execute phase to run lots faster on |
| 39 | * simple cases. They are: |
| 40 | * |
| 41 | * regstart sv that must begin a match; NULL if none obvious |
| 42 | * reganch is the match anchored (at beginning-of-line only)? |
| 43 | * regmust string (pointer into program) that match must include, or NULL |
| 44 | * [regmust changed to SV* for bminstr()--law] |
| 45 | * regmlen length of regmust string |
| 46 | * [regmlen not used currently] |
| 47 | * |
| 48 | * Regstart and reganch permit very fast decisions on suitable starting points |
| 49 | * for a match, cutting down the work a lot. Regmust permits fast rejection |
| 50 | * of lines that cannot possibly match. The regmust tests are costly enough |
| 51 | * that pregcomp() supplies a regmust only if the r.e. contains something |
| 52 | * potentially expensive (at present, the only such thing detected is * or + |
| 53 | * at the start of the r.e., which can involve a lot of backup). Regmlen is |
| 54 | * supplied because the test in pregexec() needs it and pregcomp() is computing |
| 55 | * it anyway. |
| 56 | * [regmust is now supplied always. The tests that use regmust have a |
| 57 | * heuristic that disables the test if it usually matches.] |
| 58 | * |
| 59 | * [In fact, we now use regmust in many cases to locate where the search |
| 60 | * starts in the string, so if regback is >= 0, the regmust search is never |
| 61 | * wasted effort. The regback variable says how many characters back from |
| 62 | * where regmust matched is the earliest possible start of the match. |
| 63 | * For instance, /[a-z].foo/ has a regmust of 'foo' and a regback of 2.] |
| 64 | */ |
| 65 | |
| 66 | /* |
| 67 | * Structure for regexp "program". This is essentially a linear encoding |
| 68 | * of a nondeterministic finite-state machine (aka syntax charts or |
| 69 | * "railroad normal form" in parsing technology). Each node is an opcode |
| 70 | * plus a "next" pointer, possibly plus an operand. "Next" pointers of |
| 71 | * all nodes except BRANCH implement concatenation; a "next" pointer with |
| 72 | * a BRANCH on both ends of it is connecting two alternatives. (Here we |
| 73 | * have one of the subtle syntax dependencies: an individual BRANCH (as |
| 74 | * opposed to a collection of them) is never concatenated with anything |
| 75 | * because of operator precedence.) The operand of some types of node is |
| 76 | * a literal string; for others, it is a node leading into a sub-FSM. In |
| 77 | * particular, the operand of a BRANCH node is the first node of the branch. |
| 78 | * (NB this is *not* a tree structure: the tail of the branch connects |
| 79 | * to the thing following the set of BRANCHes.) The opcodes are defined |
| 80 | * in regnodes.h which is generated from regcomp.sym by regcomp.pl. |
| 81 | */ |
| 82 | |
| 83 | /* |
| 84 | * A node is one char of opcode followed by two chars of "next" pointer. |
| 85 | * "Next" pointers are stored as two 8-bit pieces, high order first. The |
| 86 | * value is a positive offset from the opcode of the node containing it. |
| 87 | * An operand, if any, simply follows the node. (Note that much of the |
| 88 | * code generation knows about this implicit relationship.) |
| 89 | * |
| 90 | * Using two bytes for the "next" pointer is vast overkill for most things, |
| 91 | * but allows patterns to get big without disasters. |
| 92 | * |
| 93 | * [The "next" pointer is always aligned on an even |
| 94 | * boundary, and reads the offset directly as a short. Also, there is no |
| 95 | * special test to reverse the sign of BACK pointers since the offset is |
| 96 | * stored negative.] |
| 97 | */ |
| 98 | |
| 99 | /* This is the stuff that used to live in regexp.h that was truly |
| 100 | private to the engine itself. It now lives here. */ |
| 101 | |
| 102 | |
| 103 | |
| 104 | typedef struct regexp_internal { |
| 105 | int name_list_idx; /* Optional data index of an array of paren names */ |
| 106 | union { |
| 107 | U32 *offsets; /* offset annotations 20001228 MJD |
| 108 | data about mapping the program to the |
| 109 | string - |
| 110 | offsets[0] is proglen when this is used |
| 111 | */ |
| 112 | U32 proglen; |
| 113 | } u; |
| 114 | |
| 115 | regnode *regstclass; /* Optional startclass as identified or constructed |
| 116 | by the optimiser */ |
| 117 | struct reg_data *data; /* Additional miscellaneous data used by the program. |
| 118 | Used to make it easier to clone and free arbitrary |
| 119 | data that the regops need. Often the ARG field of |
| 120 | a regop is an index into this structure */ |
| 121 | struct reg_code_block *code_blocks;/* positions of literal (?{}) */ |
| 122 | int num_code_blocks; /* size of code_blocks[] */ |
| 123 | regnode program[1]; /* Unwarranted chumminess with compiler. */ |
| 124 | } regexp_internal; |
| 125 | |
| 126 | #define RXi_SET(x,y) (x)->pprivate = (void*)(y) |
| 127 | #define RXi_GET(x) ((regexp_internal *)((x)->pprivate)) |
| 128 | #define RXi_GET_DECL(r,ri) regexp_internal *ri = RXi_GET(r) |
| 129 | /* |
| 130 | * Flags stored in regexp->intflags |
| 131 | * These are used only internally to the regexp engine |
| 132 | * |
| 133 | * See regexp.h for flags used externally to the regexp engine |
| 134 | */ |
| 135 | #define PREGf_SKIP 0x00000001 |
| 136 | #define PREGf_IMPLICIT 0x00000002 /* Converted .* to ^.* */ |
| 137 | #define PREGf_NAUGHTY 0x00000004 /* how exponential is this pattern? */ |
| 138 | #define PREGf_VERBARG_SEEN 0x00000008 |
| 139 | #define PREGf_CUTGROUP_SEEN 0x00000010 |
| 140 | #define PREGf_USE_RE_EVAL 0x00000020 /* compiled with "use re 'eval'" */ |
| 141 | |
| 142 | |
| 143 | /* this is where the old regcomp.h started */ |
| 144 | |
| 145 | struct regnode_string { |
| 146 | U8 str_len; |
| 147 | U8 type; |
| 148 | U16 next_off; |
| 149 | char string[1]; |
| 150 | }; |
| 151 | |
| 152 | /* Argument bearing node - workhorse, |
| 153 | arg1 is often for the data field */ |
| 154 | struct regnode_1 { |
| 155 | U8 flags; |
| 156 | U8 type; |
| 157 | U16 next_off; |
| 158 | U32 arg1; |
| 159 | }; |
| 160 | |
| 161 | /* Similar to a regnode_1 but with an extra signed argument */ |
| 162 | struct regnode_2L { |
| 163 | U8 flags; |
| 164 | U8 type; |
| 165 | U16 next_off; |
| 166 | U32 arg1; |
| 167 | I32 arg2; |
| 168 | }; |
| 169 | |
| 170 | /* 'Two field' -- Two 16 bit unsigned args */ |
| 171 | struct regnode_2 { |
| 172 | U8 flags; |
| 173 | U8 type; |
| 174 | U16 next_off; |
| 175 | U16 arg1; |
| 176 | U16 arg2; |
| 177 | }; |
| 178 | |
| 179 | |
| 180 | #define ANYOF_BITMAP_SIZE 32 /* 256 b/(8 b/B) */ |
| 181 | #define ANYOF_CLASSBITMAP_SIZE 4 /* up to 32 (8*4) named classes */ |
| 182 | |
| 183 | /* also used by trie */ |
| 184 | struct regnode_charclass { |
| 185 | U8 flags; |
| 186 | U8 type; |
| 187 | U16 next_off; |
| 188 | U32 arg1; /* used as ptr in S_regclass */ |
| 189 | char bitmap[ANYOF_BITMAP_SIZE]; /* only compile-time */ |
| 190 | }; |
| 191 | |
| 192 | /* has runtime (locale) \d, \w, ..., [:posix:] classes */ |
| 193 | struct regnode_charclass_class { |
| 194 | U8 flags; /* ANYOF_CLASS bit must go here */ |
| 195 | U8 type; |
| 196 | U16 next_off; |
| 197 | U32 arg1; /* used as ptr in S_regclass */ |
| 198 | char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time */ |
| 199 | char classflags[ANYOF_CLASSBITMAP_SIZE]; /* and run-time */ |
| 200 | }; |
| 201 | |
| 202 | /* XXX fix this description. |
| 203 | Impose a limit of REG_INFTY on various pattern matching operations |
| 204 | to limit stack growth and to avoid "infinite" recursions. |
| 205 | */ |
| 206 | /* The default size for REG_INFTY is I16_MAX, which is the same as |
| 207 | SHORT_MAX (see perl.h). Unfortunately I16 isn't necessarily 16 bits |
| 208 | (see handy.h). On the Cray C90, sizeof(short)==4 and hence I16_MAX is |
| 209 | ((1<<31)-1), while on the Cray T90, sizeof(short)==8 and I16_MAX is |
| 210 | ((1<<63)-1). To limit stack growth to reasonable sizes, supply a |
| 211 | smaller default. |
| 212 | --Andy Dougherty 11 June 1998 |
| 213 | */ |
| 214 | #if SHORTSIZE > 2 |
| 215 | # ifndef REG_INFTY |
| 216 | # define REG_INFTY ((1<<15)-1) |
| 217 | # endif |
| 218 | #endif |
| 219 | |
| 220 | #ifndef REG_INFTY |
| 221 | # define REG_INFTY I16_MAX |
| 222 | #endif |
| 223 | |
| 224 | #define ARG_VALUE(arg) (arg) |
| 225 | #define ARG__SET(arg,val) ((arg) = (val)) |
| 226 | |
| 227 | #undef ARG |
| 228 | #undef ARG1 |
| 229 | #undef ARG2 |
| 230 | |
| 231 | #define ARG(p) ARG_VALUE(ARG_LOC(p)) |
| 232 | #define ARG1(p) ARG_VALUE(ARG1_LOC(p)) |
| 233 | #define ARG2(p) ARG_VALUE(ARG2_LOC(p)) |
| 234 | #define ARG2L(p) ARG_VALUE(ARG2L_LOC(p)) |
| 235 | |
| 236 | #define ARG_SET(p, val) ARG__SET(ARG_LOC(p), (val)) |
| 237 | #define ARG1_SET(p, val) ARG__SET(ARG1_LOC(p), (val)) |
| 238 | #define ARG2_SET(p, val) ARG__SET(ARG2_LOC(p), (val)) |
| 239 | #define ARG2L_SET(p, val) ARG__SET(ARG2L_LOC(p), (val)) |
| 240 | |
| 241 | #undef NEXT_OFF |
| 242 | #undef NODE_ALIGN |
| 243 | |
| 244 | #define NEXT_OFF(p) ((p)->next_off) |
| 245 | #define NODE_ALIGN(node) |
| 246 | #define NODE_ALIGN_FILL(node) ((node)->flags = 0xde) /* deadbeef */ |
| 247 | |
| 248 | #define SIZE_ALIGN NODE_ALIGN |
| 249 | |
| 250 | #undef OP |
| 251 | #undef OPERAND |
| 252 | #undef MASK |
| 253 | #undef STRING |
| 254 | |
| 255 | #define OP(p) ((p)->type) |
| 256 | #define FLAGS(p) ((p)->flags) /* Caution: Doesn't apply to all \ |
| 257 | regnode types. For some, it's the \ |
| 258 | character set of the regnode */ |
| 259 | #define OPERAND(p) (((struct regnode_string *)p)->string) |
| 260 | #define MASK(p) ((char*)OPERAND(p)) |
| 261 | #define STR_LEN(p) (((struct regnode_string *)p)->str_len) |
| 262 | #define STRING(p) (((struct regnode_string *)p)->string) |
| 263 | #define STR_SZ(l) ((l + sizeof(regnode) - 1) / sizeof(regnode)) |
| 264 | #define NODE_SZ_STR(p) (STR_SZ(STR_LEN(p))+1) |
| 265 | |
| 266 | #undef NODE_ALIGN |
| 267 | #undef ARG_LOC |
| 268 | #undef NEXTOPER |
| 269 | #undef PREVOPER |
| 270 | |
| 271 | #define NODE_ALIGN(node) |
| 272 | #define ARG_LOC(p) (((struct regnode_1 *)p)->arg1) |
| 273 | #define ARG1_LOC(p) (((struct regnode_2 *)p)->arg1) |
| 274 | #define ARG2_LOC(p) (((struct regnode_2 *)p)->arg2) |
| 275 | #define ARG2L_LOC(p) (((struct regnode_2L *)p)->arg2) |
| 276 | |
| 277 | #define NODE_STEP_REGNODE 1 /* sizeof(regnode)/sizeof(regnode) */ |
| 278 | #define EXTRA_STEP_2ARGS EXTRA_SIZE(struct regnode_2) |
| 279 | |
| 280 | #define NODE_STEP_B 4 |
| 281 | |
| 282 | #define NEXTOPER(p) ((p) + NODE_STEP_REGNODE) |
| 283 | #define PREVOPER(p) ((p) - NODE_STEP_REGNODE) |
| 284 | |
| 285 | #define FILL_ADVANCE_NODE(ptr, op) STMT_START { \ |
| 286 | (ptr)->type = op; (ptr)->next_off = 0; (ptr)++; } STMT_END |
| 287 | #define FILL_ADVANCE_NODE_ARG(ptr, op, arg) STMT_START { \ |
| 288 | ARG_SET(ptr, arg); FILL_ADVANCE_NODE(ptr, op); (ptr) += 1; } STMT_END |
| 289 | |
| 290 | #define REG_MAGIC 0234 |
| 291 | |
| 292 | #define SIZE_ONLY (RExC_emit == &PL_regdummy) |
| 293 | |
| 294 | /* If the bitmap doesn't fully represent what this ANYOF node can match, the |
| 295 | * ARG is set to this special value (since 0, 1, ... are legal, but will never |
| 296 | * reach this high). */ |
| 297 | #define ANYOF_NONBITMAP_EMPTY ((U32) -1) |
| 298 | |
| 299 | /* The information used to be stored as as combination of the ANYOF_UTF8 and |
| 300 | * ANYOF_NONBITMAP_NON_UTF8 bits in the flags field, but was moved out of there |
| 301 | * to free up a bit for other uses. This tries to hide the change from |
| 302 | * existing code as much as possible. Now, the data structure that goes in ARG |
| 303 | * is not allocated unless it is needed, and that is what is used to determine |
| 304 | * if there is something outside the bitmap. The code now assumes that if |
| 305 | * that structure exists, that any UTF-8 encoded string should be tried against |
| 306 | * it, but a non-UTF8-encoded string will be tried only if the |
| 307 | * ANYOF_NONBITMAP_NON_UTF8 bit is also set. */ |
| 308 | #define ANYOF_NONBITMAP(node) (ARG(node) != ANYOF_NONBITMAP_EMPTY) |
| 309 | |
| 310 | /* Flags for node->flags of ANYOF. These are in short supply, so some games |
| 311 | * are done to share them, as described below. Already, the ANYOF_LOCALE and |
| 312 | * ANYOF_CLASS bits are shared, making a space penalty for all locale nodes. |
| 313 | * An option would be to push them into new nodes. E.g. there could be an |
| 314 | * ANYOF_LOCALE node that would be in place of the flag of the same name. But |
| 315 | * there are better options. The UNICODE_ALL bit could be freed up by |
| 316 | * resorting to creating a swash containing everything above 255. This |
| 317 | * introduces a performance penalty. Better would be to split it off into a |
| 318 | * separate node, which actually would improve performance by allowing adding a |
| 319 | * case statement to regexec.c use the bit map for code points under 256, and |
| 320 | * to match everything above. If flags need to be added that are applicable to |
| 321 | * the synthetic start class only, with some work, they could be put in the |
| 322 | * next-node field, or in an unused bit of the classflags field. This could be |
| 323 | * done with the current EOS flag, and a new node type created that is just for |
| 324 | * the scc, freeing up that bit */ |
| 325 | |
| 326 | #define ANYOF_LOCALE 0x01 /* /l modifier */ |
| 327 | |
| 328 | /* The fold is calculated and stored in the bitmap where possible at compile |
| 329 | * time. However under locale, the actual folding varies depending on |
| 330 | * what the locale is at the time of execution, so it has to be deferred until |
| 331 | * then */ |
| 332 | #define ANYOF_LOC_FOLD 0x02 |
| 333 | |
| 334 | #define ANYOF_INVERT 0x04 |
| 335 | |
| 336 | /* Set if this is a struct regnode_charclass_class vs a regnode_charclass. This |
| 337 | * is used for runtime \d, \w, [:posix:], ..., which are used only in locale |
| 338 | * and the optimizer's synthetic start class. Non-locale \d, etc are resolved |
| 339 | * at compile-time. Now shared with ANYOF_LOCALE, forcing all locale nodes to |
| 340 | * be large */ |
| 341 | #define ANYOF_CLASS ANYOF_LOCALE |
| 342 | #define ANYOF_LARGE ANYOF_CLASS /* Same; name retained for back compat */ |
| 343 | |
| 344 | /* Should this character class warn if matched against a character above |
| 345 | * Unicode */ |
| 346 | #define ANYOF_WARN_SUPER 0x08 |
| 347 | |
| 348 | /* EOS, meaning that it can match an empty string too, is used for the |
| 349 | * synthetic start class only. */ |
| 350 | #define ANYOF_EOS 0x10 |
| 351 | |
| 352 | /* ? Is this node the synthetic start class (ssc). This bit is shared with |
| 353 | * ANYOF_EOS, as the latter is used only for the ssc, and then not used by |
| 354 | * regexec.c. And, the code is structured so that if it is set, the ssc is |
| 355 | * not used, so it is guaranteed to be 0 for the ssc by the time regexec.c |
| 356 | * gets executed, and 0 for a non-ssc ANYOF node, as it only ever gets set for |
| 357 | * a potential ssc candidate. Thus setting it to 1 after it has been |
| 358 | * determined that the ssc will be used is not ambiguous */ |
| 359 | #define ANYOF_IS_SYNTHETIC ANYOF_EOS |
| 360 | |
| 361 | /* Can match something outside the bitmap that isn't in utf8 */ |
| 362 | #define ANYOF_NONBITMAP_NON_UTF8 0x20 |
| 363 | |
| 364 | /* Matches every code point 0x100 and above*/ |
| 365 | #define ANYOF_UNICODE_ALL 0x40 |
| 366 | |
| 367 | /* Match all Latin1 characters that aren't ASCII when the target string is not |
| 368 | * in utf8. */ |
| 369 | #define ANYOF_NON_UTF8_LATIN1_ALL 0x80 |
| 370 | |
| 371 | #define ANYOF_FLAGS_ALL 0xff |
| 372 | |
| 373 | /* These are the flags that ANYOF_INVERT being set or not doesn't affect |
| 374 | * whether they are operative or not. e.g., the node still has LOCALE |
| 375 | * regardless of being inverted; whereas ANYOF_UNICODE_ALL means something |
| 376 | * different if inverted */ |
| 377 | #define INVERSION_UNAFFECTED_FLAGS (ANYOF_LOCALE \ |
| 378 | |ANYOF_LOC_FOLD \ |
| 379 | |ANYOF_CLASS \ |
| 380 | |ANYOF_EOS \ |
| 381 | |ANYOF_NONBITMAP_NON_UTF8) |
| 382 | |
| 383 | /* Character classes for node->classflags of ANYOF */ |
| 384 | /* Should be synchronized with a table in regprop() */ |
| 385 | /* 2n should be the normal one, paired with its complement at 2n+1 */ |
| 386 | |
| 387 | #define ANYOF_WORDCHAR ((_CC_WORDCHAR) * 2) /* \w, PL_utf8_alnum, utf8::IsWord, ALNUM */ |
| 388 | #define ANYOF_NWORDCHAR ((ANYOF_WORDCHAR) + 1) |
| 389 | #define ANYOF_SPACE ((_CC_SPACE) * 2) /* \s */ |
| 390 | #define ANYOF_NSPACE ((ANYOF_SPACE) + 1) |
| 391 | #define ANYOF_DIGIT ((_CC_DIGIT) * 2) /* \d */ |
| 392 | #define ANYOF_NDIGIT ((ANYOF_DIGIT) + 1) |
| 393 | #define ANYOF_ALNUMC ((_CC_ALNUMC) * 2) /* [[:alnum:]] isalnum(3), utf8::IsAlnum, ALNUMC */ |
| 394 | #define ANYOF_NALNUMC ((ANYOF_ALNUMC) + 1) |
| 395 | #define ANYOF_ALPHA ((_CC_ALPHA) * 2) |
| 396 | #define ANYOF_NALPHA ((ANYOF_ALPHA) + 1) |
| 397 | #define ANYOF_ASCII ((_CC_ASCII) * 2) |
| 398 | #define ANYOF_NASCII ((ANYOF_ASCII) + 1) |
| 399 | #define ANYOF_CNTRL ((_CC_CNTRL) * 2) |
| 400 | #define ANYOF_NCNTRL ((ANYOF_CNTRL) + 1) |
| 401 | #define ANYOF_GRAPH ((_CC_GRAPH) * 2) |
| 402 | #define ANYOF_NGRAPH ((ANYOF_GRAPH) + 1) |
| 403 | #define ANYOF_LOWER ((_CC_LOWER) * 2) |
| 404 | #define ANYOF_NLOWER ((ANYOF_LOWER) + 1) |
| 405 | #define ANYOF_PRINT ((_CC_PRINT) * 2) |
| 406 | #define ANYOF_NPRINT ((ANYOF_PRINT) + 1) |
| 407 | #define ANYOF_PUNCT ((_CC_PUNCT) * 2) |
| 408 | #define ANYOF_NPUNCT ((ANYOF_PUNCT) + 1) |
| 409 | #define ANYOF_UPPER ((_CC_UPPER) * 2) |
| 410 | #define ANYOF_NUPPER ((ANYOF_UPPER) + 1) |
| 411 | #define ANYOF_XDIGIT ((_CC_XDIGIT) * 2) |
| 412 | #define ANYOF_NXDIGIT ((ANYOF_XDIGIT) + 1) |
| 413 | #define ANYOF_PSXSPC ((_CC_PSXSPC) * 2) /* POSIX space: \s plus the vertical tab */ |
| 414 | #define ANYOF_NPSXSPC ((ANYOF_PSXSPC) + 1) |
| 415 | #define ANYOF_BLANK ((_CC_BLANK) * 2) /* GNU extension: space and tab: non-vertical space */ |
| 416 | #define ANYOF_NBLANK ((ANYOF_BLANK) + 1) |
| 417 | |
| 418 | #define ANYOF_MAX ((ANYOF_NBLANK) + 1) /* So upper loop limit is written: |
| 419 | '< ANYOF_MAX' */ |
| 420 | #if (ANYOF_MAX > 32) /* Must fit in 32-bit word */ |
| 421 | # error Problem with handy.h _CC_foo #defines |
| 422 | #endif |
| 423 | |
| 424 | /* pseudo classes below this, not stored in the class bitmap, but used as flags |
| 425 | during compilation of char classes */ |
| 426 | |
| 427 | #define ANYOF_VERTWS ((ANYOF_MAX)+0) |
| 428 | #define ANYOF_NVERTWS ((ANYOF_MAX)+1) |
| 429 | |
| 430 | #if (ANYOF_VERTWS != (_CC_VERTSPACE) * 2) \ |
| 431 | || (_CC_VERTSPACE != _HIGHEST_REGCOMP_DOT_H_SYNC) |
| 432 | # error Problem with handy.h _CC_VERTSPACE #define |
| 433 | #endif |
| 434 | |
| 435 | #define ANYOF_HORIZWS ((ANYOF_MAX)+2) |
| 436 | #define ANYOF_NHORIZWS ((ANYOF_MAX)+3) |
| 437 | |
| 438 | #define ANYOF_UNIPROP ((ANYOF_MAX)+4) /* Used to indicate a Unicode |
| 439 | property: \p{} or \P{} */ |
| 440 | |
| 441 | /* Backward source code compatibility. */ |
| 442 | |
| 443 | #define ANYOF_ALNUML ANYOF_ALNUM |
| 444 | #define ANYOF_NALNUML ANYOF_NALNUM |
| 445 | #define ANYOF_SPACEL ANYOF_SPACE |
| 446 | #define ANYOF_NSPACEL ANYOF_NSPACE |
| 447 | #define ANYOF_ALNUM ANYOF_WORDCHAR |
| 448 | #define ANYOF_NALNUM ANYOF_NWORDCHAR |
| 449 | |
| 450 | /* Utility macros for the bitmap and classes of ANYOF */ |
| 451 | |
| 452 | #define ANYOF_SIZE (sizeof(struct regnode_charclass)) |
| 453 | #define ANYOF_CLASS_SIZE (sizeof(struct regnode_charclass_class)) |
| 454 | |
| 455 | #define ANYOF_FLAGS(p) ((p)->flags) |
| 456 | |
| 457 | #define ANYOF_BIT(c) (1 << ((c) & 7)) |
| 458 | |
| 459 | #define ANYOF_CLASS_BYTE(p, c) (((struct regnode_charclass_class*)(p))->classflags[((c) >> 3) & 3]) |
| 460 | #define ANYOF_CLASS_SET(p, c) (ANYOF_CLASS_BYTE(p, c) |= ANYOF_BIT(c)) |
| 461 | #define ANYOF_CLASS_CLEAR(p, c) (ANYOF_CLASS_BYTE(p, c) &= ~ANYOF_BIT(c)) |
| 462 | #define ANYOF_CLASS_TEST(p, c) (ANYOF_CLASS_BYTE(p, c) & ANYOF_BIT(c)) |
| 463 | |
| 464 | #define ANYOF_CLASS_ZERO(ret) Zero(((struct regnode_charclass_class*)(ret))->classflags, ANYOF_CLASSBITMAP_SIZE, char) |
| 465 | #define ANYOF_CLASS_SETALL(ret) \ |
| 466 | memset (((struct regnode_charclass_class*)(ret))->classflags, 255, ANYOF_CLASSBITMAP_SIZE) |
| 467 | #define ANYOF_BITMAP_ZERO(ret) Zero(((struct regnode_charclass*)(ret))->bitmap, ANYOF_BITMAP_SIZE, char) |
| 468 | |
| 469 | #define ANYOF_BITMAP(p) (((struct regnode_charclass*)(p))->bitmap) |
| 470 | #define ANYOF_BITMAP_BYTE(p, c) (ANYOF_BITMAP(p)[(((U8)(c)) >> 3) & 31]) |
| 471 | #define ANYOF_BITMAP_SET(p, c) (ANYOF_BITMAP_BYTE(p, c) |= ANYOF_BIT(c)) |
| 472 | #define ANYOF_BITMAP_CLEAR(p,c) (ANYOF_BITMAP_BYTE(p, c) &= ~ANYOF_BIT(c)) |
| 473 | #define ANYOF_BITMAP_TEST(p, c) (ANYOF_BITMAP_BYTE(p, c) & ANYOF_BIT(c)) |
| 474 | |
| 475 | #define ANYOF_BITMAP_SETALL(p) \ |
| 476 | memset (ANYOF_BITMAP(p), 255, ANYOF_BITMAP_SIZE) |
| 477 | #define ANYOF_BITMAP_CLEARALL(p) \ |
| 478 | Zero (ANYOF_BITMAP(p), ANYOF_BITMAP_SIZE) |
| 479 | /* Check that all 256 bits are all set. Used in S_cl_is_anything() */ |
| 480 | #define ANYOF_BITMAP_TESTALLSET(p) /* Assumes sizeof(p) == 32 */ \ |
| 481 | 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) |
| 482 | |
| 483 | #define ANYOF_SKIP ((ANYOF_SIZE - 1)/sizeof(regnode)) |
| 484 | #define ANYOF_CLASS_SKIP ((ANYOF_CLASS_SIZE - 1)/sizeof(regnode)) |
| 485 | |
| 486 | #if ANYOF_CLASSBITMAP_SIZE != 4 |
| 487 | # error ANYOF_CLASSBITMAP_SIZE is expected to be 4 |
| 488 | #endif |
| 489 | #define ANYOF_CLASS_TEST_ANY_SET(p) ((ANYOF_FLAGS(p) & ANYOF_CLASS) \ |
| 490 | && memNE (((struct regnode_charclass_class*)(p))->classflags, \ |
| 491 | "\0\0\0\0", ANYOF_CLASSBITMAP_SIZE)) |
| 492 | /*#define ANYOF_CLASS_ADD_SKIP (ANYOF_CLASS_SKIP - ANYOF_SKIP) |
| 493 | * */ |
| 494 | |
| 495 | |
| 496 | /* |
| 497 | * Utility definitions. |
| 498 | */ |
| 499 | #ifndef CHARMASK |
| 500 | # define UCHARAT(p) ((int)*(const U8*)(p)) |
| 501 | #else |
| 502 | # define UCHARAT(p) ((int)*(p)&CHARMASK) |
| 503 | #endif |
| 504 | |
| 505 | #define EXTRA_SIZE(guy) ((sizeof(guy)-1)/sizeof(struct regnode)) |
| 506 | |
| 507 | #define REG_SEEN_ZERO_LEN 0x00000001 |
| 508 | #define REG_SEEN_LOOKBEHIND 0x00000002 |
| 509 | #define REG_SEEN_GPOS 0x00000004 |
| 510 | /* spare */ |
| 511 | #define REG_SEEN_CANY 0x00000010 |
| 512 | #define REG_SEEN_SANY REG_SEEN_CANY /* src bckwrd cmpt */ |
| 513 | #define REG_SEEN_RECURSE 0x00000020 |
| 514 | #define REG_TOP_LEVEL_BRANCHES 0x00000040 |
| 515 | #define REG_SEEN_VERBARG 0x00000080 |
| 516 | #define REG_SEEN_CUTGROUP 0x00000100 |
| 517 | #define REG_SEEN_RUN_ON_COMMENT 0x00000200 |
| 518 | #define REG_SEEN_EXACTF_SHARP_S 0x00000400 |
| 519 | |
| 520 | START_EXTERN_C |
| 521 | |
| 522 | #ifdef PLUGGABLE_RE_EXTENSION |
| 523 | #include "re_nodes.h" |
| 524 | #else |
| 525 | #include "regnodes.h" |
| 526 | #endif |
| 527 | |
| 528 | #ifndef PLUGGABLE_RE_EXTENSION |
| 529 | #ifndef DOINIT |
| 530 | EXTCONST regexp_engine PL_core_reg_engine; |
| 531 | #else /* DOINIT */ |
| 532 | EXTCONST regexp_engine PL_core_reg_engine = { |
| 533 | Perl_re_compile, |
| 534 | Perl_regexec_flags, |
| 535 | Perl_re_intuit_start, |
| 536 | Perl_re_intuit_string, |
| 537 | Perl_regfree_internal, |
| 538 | Perl_reg_numbered_buff_fetch, |
| 539 | Perl_reg_numbered_buff_store, |
| 540 | Perl_reg_numbered_buff_length, |
| 541 | Perl_reg_named_buff, |
| 542 | Perl_reg_named_buff_iter, |
| 543 | Perl_reg_qr_package, |
| 544 | #if defined(USE_ITHREADS) |
| 545 | Perl_regdupe_internal, |
| 546 | #endif |
| 547 | Perl_re_op_compile |
| 548 | }; |
| 549 | #endif /* DOINIT */ |
| 550 | #endif /* PLUGGABLE_RE_EXTENSION */ |
| 551 | |
| 552 | |
| 553 | END_EXTERN_C |
| 554 | |
| 555 | |
| 556 | /* .what is a character array with one character for each member of .data |
| 557 | * The character describes the function of the corresponding .data item: |
| 558 | * a - AV for paren_name_list under DEBUGGING |
| 559 | * f - start-class data for regstclass optimization |
| 560 | * l - start op for literal (?{EVAL}) item |
| 561 | * L - start op for literal (?{EVAL}) item, with separate CV (qr//) |
| 562 | * r - pointer to an embedded code-containing qr, e.g. /ab$qr/ |
| 563 | * s - swash for Unicode-style character class, and the multicharacter |
| 564 | * strings resulting from casefolding the single-character entries |
| 565 | * in the character class |
| 566 | * t - trie struct |
| 567 | * u - trie struct's widecharmap (a HV, so can't share, must dup) |
| 568 | * also used for revcharmap and words under DEBUGGING |
| 569 | * T - aho-trie struct |
| 570 | * S - sv for named capture lookup |
| 571 | * 20010712 mjd@plover.com |
| 572 | * (Remember to update re_dup() and pregfree() if you add any items.) |
| 573 | */ |
| 574 | struct reg_data { |
| 575 | U32 count; |
| 576 | U8 *what; |
| 577 | void* data[1]; |
| 578 | }; |
| 579 | |
| 580 | /* Code in S_to_utf8_substr() and S_to_byte_substr() in regexec.c accesses |
| 581 | anchored* and float* via array indexes 0 and 1. */ |
| 582 | #define anchored_substr substrs->data[0].substr |
| 583 | #define anchored_utf8 substrs->data[0].utf8_substr |
| 584 | #define anchored_offset substrs->data[0].min_offset |
| 585 | #define anchored_end_shift substrs->data[0].end_shift |
| 586 | |
| 587 | #define float_substr substrs->data[1].substr |
| 588 | #define float_utf8 substrs->data[1].utf8_substr |
| 589 | #define float_min_offset substrs->data[1].min_offset |
| 590 | #define float_max_offset substrs->data[1].max_offset |
| 591 | #define float_end_shift substrs->data[1].end_shift |
| 592 | |
| 593 | #define check_substr substrs->data[2].substr |
| 594 | #define check_utf8 substrs->data[2].utf8_substr |
| 595 | #define check_offset_min substrs->data[2].min_offset |
| 596 | #define check_offset_max substrs->data[2].max_offset |
| 597 | #define check_end_shift substrs->data[2].end_shift |
| 598 | |
| 599 | #define RX_ANCHORED_SUBSTR(rx) (ReANY(rx)->anchored_substr) |
| 600 | #define RX_ANCHORED_UTF8(rx) (ReANY(rx)->anchored_utf8) |
| 601 | #define RX_FLOAT_SUBSTR(rx) (ReANY(rx)->float_substr) |
| 602 | #define RX_FLOAT_UTF8(rx) (ReANY(rx)->float_utf8) |
| 603 | |
| 604 | /* trie related stuff */ |
| 605 | |
| 606 | /* a transition record for the state machine. the |
| 607 | check field determines which state "owns" the |
| 608 | transition. the char the transition is for is |
| 609 | determined by offset from the owning states base |
| 610 | field. the next field determines which state |
| 611 | is to be transitioned to if any. |
| 612 | */ |
| 613 | struct _reg_trie_trans { |
| 614 | U32 next; |
| 615 | U32 check; |
| 616 | }; |
| 617 | |
| 618 | /* a transition list element for the list based representation */ |
| 619 | struct _reg_trie_trans_list_elem { |
| 620 | U16 forid; |
| 621 | U32 newstate; |
| 622 | }; |
| 623 | typedef struct _reg_trie_trans_list_elem reg_trie_trans_le; |
| 624 | |
| 625 | /* a state for compressed nodes. base is an offset |
| 626 | into an array of reg_trie_trans array. If wordnum is |
| 627 | nonzero the state is accepting. if base is zero then |
| 628 | the state has no children (and will be accepting) |
| 629 | */ |
| 630 | struct _reg_trie_state { |
| 631 | U16 wordnum; |
| 632 | union { |
| 633 | U32 base; |
| 634 | reg_trie_trans_le* list; |
| 635 | } trans; |
| 636 | }; |
| 637 | |
| 638 | /* info per word; indexed by wordnum */ |
| 639 | typedef struct { |
| 640 | U16 prev; /* previous word in acceptance chain; eg in |
| 641 | * zzz|abc|ab/ after matching the chars abc, the |
| 642 | * accepted word is #2, and the previous accepted |
| 643 | * word is #3 */ |
| 644 | U32 len; /* how many chars long is this word? */ |
| 645 | U32 accept; /* accept state for this word */ |
| 646 | } reg_trie_wordinfo; |
| 647 | |
| 648 | |
| 649 | typedef struct _reg_trie_state reg_trie_state; |
| 650 | typedef struct _reg_trie_trans reg_trie_trans; |
| 651 | |
| 652 | |
| 653 | /* anything in here that needs to be freed later |
| 654 | should be dealt with in pregfree. |
| 655 | refcount is first in both this and _reg_ac_data to allow a space |
| 656 | optimisation in Perl_regdupe. */ |
| 657 | struct _reg_trie_data { |
| 658 | U32 refcount; /* number of times this trie is referenced */ |
| 659 | U32 lasttrans; /* last valid transition element */ |
| 660 | U16 *charmap; /* byte to charid lookup array */ |
| 661 | reg_trie_state *states; /* state data */ |
| 662 | reg_trie_trans *trans; /* array of transition elements */ |
| 663 | char *bitmap; /* stclass bitmap */ |
| 664 | U16 *jump; /* optional 1 indexed array of offsets before tail |
| 665 | for the node following a given word. */ |
| 666 | reg_trie_wordinfo *wordinfo; /* array of info per word */ |
| 667 | U16 uniquecharcount; /* unique chars in trie (width of trans table) */ |
| 668 | U32 startstate; /* initial state - used for common prefix optimisation */ |
| 669 | STRLEN minlen; /* minimum length of words in trie - build/opt only? */ |
| 670 | STRLEN maxlen; /* maximum length of words in trie - build/opt only? */ |
| 671 | U32 prefixlen; /* #chars in common prefix */ |
| 672 | U32 statecount; /* Build only - number of states in the states array |
| 673 | (including the unused zero state) */ |
| 674 | U32 wordcount; /* Build only */ |
| 675 | #ifdef DEBUGGING |
| 676 | STRLEN charcount; /* Build only */ |
| 677 | #endif |
| 678 | }; |
| 679 | /* There is one (3 under DEBUGGING) pointers that logically belong in this |
| 680 | structure, but are held outside as they need duplication on thread cloning, |
| 681 | whereas the rest of the structure can be read only: |
| 682 | HV *widecharmap; code points > 255 to charid |
| 683 | #ifdef DEBUGGING |
| 684 | AV *words; Array of words contained in trie, for dumping |
| 685 | AV *revcharmap; Map of each charid back to its character representation |
| 686 | #endif |
| 687 | */ |
| 688 | |
| 689 | #define TRIE_WORDS_OFFSET 2 |
| 690 | |
| 691 | typedef struct _reg_trie_data reg_trie_data; |
| 692 | |
| 693 | /* refcount is first in both this and _reg_trie_data to allow a space |
| 694 | optimisation in Perl_regdupe. */ |
| 695 | struct _reg_ac_data { |
| 696 | U32 refcount; |
| 697 | U32 trie; |
| 698 | U32 *fail; |
| 699 | reg_trie_state *states; |
| 700 | }; |
| 701 | typedef struct _reg_ac_data reg_ac_data; |
| 702 | |
| 703 | /* ANY_BIT doesn't use the structure, so we can borrow it here. |
| 704 | This is simpler than refactoring all of it as wed end up with |
| 705 | three different sets... */ |
| 706 | |
| 707 | #define TRIE_BITMAP(p) (((reg_trie_data *)(p))->bitmap) |
| 708 | #define TRIE_BITMAP_BYTE(p, c) (TRIE_BITMAP(p)[(((U8)(c)) >> 3) & 31]) |
| 709 | #define TRIE_BITMAP_SET(p, c) (TRIE_BITMAP_BYTE(p, c) |= ANYOF_BIT((U8)c)) |
| 710 | #define TRIE_BITMAP_CLEAR(p,c) (TRIE_BITMAP_BYTE(p, c) &= ~ANYOF_BIT((U8)c)) |
| 711 | #define TRIE_BITMAP_TEST(p, c) (TRIE_BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c)) |
| 712 | |
| 713 | #define IS_ANYOF_TRIE(op) ((op)==TRIEC || (op)==AHOCORASICKC) |
| 714 | #define IS_TRIE_AC(op) ((op)>=AHOCORASICK) |
| 715 | |
| 716 | |
| 717 | #define BITMAP_BYTE(p, c) (((U8*)p)[(((U8)(c)) >> 3) & 31]) |
| 718 | #define BITMAP_TEST(p, c) (BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c)) |
| 719 | |
| 720 | /* these defines assume uniquecharcount is the correct variable, and state may be evaluated twice */ |
| 721 | #define TRIE_NODENUM(state) (((state)-1)/(trie->uniquecharcount)+1) |
| 722 | #define SAFE_TRIE_NODENUM(state) ((state) ? (((state)-1)/(trie->uniquecharcount)+1) : (state)) |
| 723 | #define TRIE_NODEIDX(state) ((state) ? (((state)-1)*(trie->uniquecharcount)+1) : (state)) |
| 724 | |
| 725 | #ifdef DEBUGGING |
| 726 | #define TRIE_CHARCOUNT(trie) ((trie)->charcount) |
| 727 | #else |
| 728 | #define TRIE_CHARCOUNT(trie) (trie_charcount) |
| 729 | #endif |
| 730 | |
| 731 | #define RE_TRIE_MAXBUF_INIT 65536 |
| 732 | #define RE_TRIE_MAXBUF_NAME "\022E_TRIE_MAXBUF" |
| 733 | #define RE_DEBUG_FLAGS "\022E_DEBUG_FLAGS" |
| 734 | |
| 735 | /* |
| 736 | |
| 737 | RE_DEBUG_FLAGS is used to control what debug output is emitted |
| 738 | its divided into three groups of options, some of which interact. |
| 739 | The three groups are: Compile, Execute, Extra. There is room for a |
| 740 | further group, as currently only the low three bytes are used. |
| 741 | |
| 742 | Compile Options: |
| 743 | |
| 744 | PARSE |
| 745 | PEEP |
| 746 | TRIE |
| 747 | PROGRAM |
| 748 | OFFSETS |
| 749 | |
| 750 | Execute Options: |
| 751 | |
| 752 | INTUIT |
| 753 | MATCH |
| 754 | TRIE |
| 755 | |
| 756 | Extra Options |
| 757 | |
| 758 | TRIE |
| 759 | OFFSETS |
| 760 | |
| 761 | If you modify any of these make sure you make corresponding changes to |
| 762 | re.pm, especially to the documentation. |
| 763 | |
| 764 | */ |
| 765 | |
| 766 | |
| 767 | /* Compile */ |
| 768 | #define RE_DEBUG_COMPILE_MASK 0x0000FF |
| 769 | #define RE_DEBUG_COMPILE_PARSE 0x000001 |
| 770 | #define RE_DEBUG_COMPILE_OPTIMISE 0x000002 |
| 771 | #define RE_DEBUG_COMPILE_TRIE 0x000004 |
| 772 | #define RE_DEBUG_COMPILE_DUMP 0x000008 |
| 773 | #define RE_DEBUG_COMPILE_FLAGS 0x000010 |
| 774 | |
| 775 | /* Execute */ |
| 776 | #define RE_DEBUG_EXECUTE_MASK 0x00FF00 |
| 777 | #define RE_DEBUG_EXECUTE_INTUIT 0x000100 |
| 778 | #define RE_DEBUG_EXECUTE_MATCH 0x000200 |
| 779 | #define RE_DEBUG_EXECUTE_TRIE 0x000400 |
| 780 | |
| 781 | /* Extra */ |
| 782 | #define RE_DEBUG_EXTRA_MASK 0xFF0000 |
| 783 | #define RE_DEBUG_EXTRA_TRIE 0x010000 |
| 784 | #define RE_DEBUG_EXTRA_OFFSETS 0x020000 |
| 785 | #define RE_DEBUG_EXTRA_OFFDEBUG 0x040000 |
| 786 | #define RE_DEBUG_EXTRA_STATE 0x080000 |
| 787 | #define RE_DEBUG_EXTRA_OPTIMISE 0x100000 |
| 788 | #define RE_DEBUG_EXTRA_BUFFERS 0x400000 |
| 789 | #define RE_DEBUG_EXTRA_GPOS 0x800000 |
| 790 | /* combined */ |
| 791 | #define RE_DEBUG_EXTRA_STACK 0x280000 |
| 792 | |
| 793 | #define RE_DEBUG_FLAG(x) (re_debug_flags & x) |
| 794 | /* Compile */ |
| 795 | #define DEBUG_COMPILE_r(x) DEBUG_r( \ |
| 796 | if (re_debug_flags & RE_DEBUG_COMPILE_MASK) x ) |
| 797 | #define DEBUG_PARSE_r(x) DEBUG_r( \ |
| 798 | if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x ) |
| 799 | #define DEBUG_OPTIMISE_r(x) DEBUG_r( \ |
| 800 | if (re_debug_flags & RE_DEBUG_COMPILE_OPTIMISE) x ) |
| 801 | #define DEBUG_PARSE_r(x) DEBUG_r( \ |
| 802 | if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x ) |
| 803 | #define DEBUG_DUMP_r(x) DEBUG_r( \ |
| 804 | if (re_debug_flags & RE_DEBUG_COMPILE_DUMP) x ) |
| 805 | #define DEBUG_TRIE_COMPILE_r(x) DEBUG_r( \ |
| 806 | if (re_debug_flags & RE_DEBUG_COMPILE_TRIE) x ) |
| 807 | #define DEBUG_FLAGS_r(x) DEBUG_r( \ |
| 808 | if (re_debug_flags & RE_DEBUG_COMPILE_FLAGS) x ) |
| 809 | /* Execute */ |
| 810 | #define DEBUG_EXECUTE_r(x) DEBUG_r( \ |
| 811 | if (re_debug_flags & RE_DEBUG_EXECUTE_MASK) x ) |
| 812 | #define DEBUG_INTUIT_r(x) DEBUG_r( \ |
| 813 | if (re_debug_flags & RE_DEBUG_EXECUTE_INTUIT) x ) |
| 814 | #define DEBUG_MATCH_r(x) DEBUG_r( \ |
| 815 | if (re_debug_flags & RE_DEBUG_EXECUTE_MATCH) x ) |
| 816 | #define DEBUG_TRIE_EXECUTE_r(x) DEBUG_r( \ |
| 817 | if (re_debug_flags & RE_DEBUG_EXECUTE_TRIE) x ) |
| 818 | |
| 819 | /* Extra */ |
| 820 | #define DEBUG_EXTRA_r(x) DEBUG_r( \ |
| 821 | if (re_debug_flags & RE_DEBUG_EXTRA_MASK) x ) |
| 822 | #define DEBUG_OFFSETS_r(x) DEBUG_r( \ |
| 823 | if (re_debug_flags & RE_DEBUG_EXTRA_OFFSETS) x ) |
| 824 | #define DEBUG_STATE_r(x) DEBUG_r( \ |
| 825 | if (re_debug_flags & RE_DEBUG_EXTRA_STATE) x ) |
| 826 | #define DEBUG_STACK_r(x) DEBUG_r( \ |
| 827 | if (re_debug_flags & RE_DEBUG_EXTRA_STACK) x ) |
| 828 | #define DEBUG_BUFFERS_r(x) DEBUG_r( \ |
| 829 | if (re_debug_flags & RE_DEBUG_EXTRA_BUFFERS) x ) |
| 830 | |
| 831 | #define DEBUG_OPTIMISE_MORE_r(x) DEBUG_r( \ |
| 832 | if ((RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE) == \ |
| 833 | (re_debug_flags & (RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE)) ) x ) |
| 834 | #define MJD_OFFSET_DEBUG(x) DEBUG_r( \ |
| 835 | if (re_debug_flags & RE_DEBUG_EXTRA_OFFDEBUG) \ |
| 836 | Perl_warn_nocontext x ) |
| 837 | #define DEBUG_TRIE_COMPILE_MORE_r(x) DEBUG_TRIE_COMPILE_r( \ |
| 838 | if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x ) |
| 839 | #define DEBUG_TRIE_EXECUTE_MORE_r(x) DEBUG_TRIE_EXECUTE_r( \ |
| 840 | if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x ) |
| 841 | |
| 842 | #define DEBUG_TRIE_r(x) DEBUG_r( \ |
| 843 | if (re_debug_flags & (RE_DEBUG_COMPILE_TRIE \ |
| 844 | | RE_DEBUG_EXECUTE_TRIE )) x ) |
| 845 | #define DEBUG_GPOS_r(x) DEBUG_r( \ |
| 846 | if (re_debug_flags & RE_DEBUG_EXTRA_GPOS) x ) |
| 847 | |
| 848 | /* initialization */ |
| 849 | /* get_sv() can return NULL during global destruction. re_debug_flags can get |
| 850 | * clobbered by a longjmp, so must be initialized */ |
| 851 | #define GET_RE_DEBUG_FLAGS DEBUG_r({ \ |
| 852 | SV * re_debug_flags_sv = NULL; \ |
| 853 | re_debug_flags = 0; \ |
| 854 | re_debug_flags_sv = get_sv(RE_DEBUG_FLAGS, 1); \ |
| 855 | if (re_debug_flags_sv) { \ |
| 856 | if (!SvIOK(re_debug_flags_sv)) \ |
| 857 | sv_setuv(re_debug_flags_sv, RE_DEBUG_COMPILE_DUMP | RE_DEBUG_EXECUTE_MASK ); \ |
| 858 | re_debug_flags=SvIV(re_debug_flags_sv); \ |
| 859 | }\ |
| 860 | }) |
| 861 | |
| 862 | #ifdef DEBUGGING |
| 863 | |
| 864 | #define GET_RE_DEBUG_FLAGS_DECL VOL IV re_debug_flags \ |
| 865 | PERL_UNUSED_DECL = 0; GET_RE_DEBUG_FLAGS; |
| 866 | |
| 867 | #define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2) \ |
| 868 | const char * const rpv = \ |
| 869 | pv_pretty((dsv), (pv), (l), (m), \ |
| 870 | PL_colors[(c1)],PL_colors[(c2)], \ |
| 871 | PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) ); \ |
| 872 | const int rlen = SvCUR(dsv) |
| 873 | |
| 874 | #define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m) \ |
| 875 | const char * const rpv = \ |
| 876 | pv_pretty((dsv), (SvPV_nolen_const(sv)), (SvCUR(sv)), (m), \ |
| 877 | PL_colors[(c1)],PL_colors[(c2)], \ |
| 878 | PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) ) |
| 879 | |
| 880 | #define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m) \ |
| 881 | const char * const rpv = \ |
| 882 | pv_pretty((dsv), (pv), (l), (m), \ |
| 883 | PL_colors[0], PL_colors[1], \ |
| 884 | ( PERL_PV_PRETTY_QUOTE | PERL_PV_ESCAPE_RE | PERL_PV_ESCAPE_NONASCII | PERL_PV_PRETTY_ELLIPSES | \ |
| 885 | ((isuni) ? PERL_PV_ESCAPE_UNI : 0)) \ |
| 886 | ) |
| 887 | |
| 888 | #define RE_SV_DUMPLEN(ItEm) (SvCUR(ItEm) - (SvTAIL(ItEm)!=0)) |
| 889 | #define RE_SV_TAIL(ItEm) (SvTAIL(ItEm) ? "$" : "") |
| 890 | |
| 891 | #else /* if not DEBUGGING */ |
| 892 | |
| 893 | #define GET_RE_DEBUG_FLAGS_DECL |
| 894 | #define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2) |
| 895 | #define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m) |
| 896 | #define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m) |
| 897 | #define RE_SV_DUMPLEN(ItEm) |
| 898 | #define RE_SV_TAIL(ItEm) |
| 899 | |
| 900 | #endif /* DEBUG RELATED DEFINES */ |
| 901 | |
| 902 | /* |
| 903 | * Local variables: |
| 904 | * c-indentation-style: bsd |
| 905 | * c-basic-offset: 4 |
| 906 | * indent-tabs-mode: nil |
| 907 | * End: |
| 908 | * |
| 909 | * ex: set ts=8 sts=4 sw=4 et: |
| 910 | */ |