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 typedef OP OP_4tree; /* Will be redefined later. */ 13 14 15 /* Convert branch sequences to more efficient trie ops? */ 16 #define PERL_ENABLE_TRIE_OPTIMISATION 1 17 18 /* Be really agressive about optimising patterns with trie sequences? */ 19 #define PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION 1 20 21 /* Use old style unicode mappings for perl and posix character classes 22 * 23 * NOTE: Enabling this essentially breaks character class matching against unicode 24 * strings, so that POSIX char classes match when they shouldn't, and \d matches 25 * way more than 10 characters, and sometimes a charclass and its complement either 26 * both match or neither match. 27 * NOTE: Disabling this will cause various backwards compatibility issues to rear 28 * their head, and tests to fail. However it will make the charclass behaviour 29 * consistant regardless of internal string type, and make character class inversions 30 * consistant. The tests that fail in the regex engine are basically broken tests. 31 * 32 * Personally I think 5.12 should disable this for sure. Its a bit more debatable for 33 * 5.10, so for now im leaving it enabled. 34 * 35 * -demerphq 36 */ 37 #define PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS 1 38 39 /* Should the optimiser take positive assertions into account? */ 40 #define PERL_ENABLE_POSITIVE_ASSERTION_STUDY 0 41 42 /* Not for production use: */ 43 #define PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS 0 44 45 /* Activate offsets code - set to if 1 to enable */ 46 #ifdef DEBUGGING 47 #define RE_TRACK_PATTERN_OFFSETS 48 #endif 49 50 /* Unless the next line is uncommented it is illegal to combine lazy 51 matching with possessive matching. Frankly it doesn't make much sense 52 to allow it as X*?+ matches nothing, X+?+ matches a single char only, 53 and X{min,max}?+ matches min times only. 54 */ 55 /* #define REG_ALLOW_MINMOD_SUSPEND */ 56 57 /* 58 * The "internal use only" fields in regexp.h are present to pass info from 59 * compile to execute that permits the execute phase to run lots faster on 60 * simple cases. They are: 61 * 62 * regstart sv that must begin a match; NULL if none obvious 63 * reganch is the match anchored (at beginning-of-line only)? 64 * regmust string (pointer into program) that match must include, or NULL 65 * [regmust changed to SV* for bminstr()--law] 66 * regmlen length of regmust string 67 * [regmlen not used currently] 68 * 69 * Regstart and reganch permit very fast decisions on suitable starting points 70 * for a match, cutting down the work a lot. Regmust permits fast rejection 71 * of lines that cannot possibly match. The regmust tests are costly enough 72 * that pregcomp() supplies a regmust only if the r.e. contains something 73 * potentially expensive (at present, the only such thing detected is * or + 74 * at the start of the r.e., which can involve a lot of backup). Regmlen is 75 * supplied because the test in pregexec() needs it and pregcomp() is computing 76 * it anyway. 77 * [regmust is now supplied always. The tests that use regmust have a 78 * heuristic that disables the test if it usually matches.] 79 * 80 * [In fact, we now use regmust in many cases to locate where the search 81 * starts in the string, so if regback is >= 0, the regmust search is never 82 * wasted effort. The regback variable says how many characters back from 83 * where regmust matched is the earliest possible start of the match. 84 * For instance, /[a-z].foo/ has a regmust of 'foo' and a regback of 2.] 85 */ 86 87 /* 88 * Structure for regexp "program". This is essentially a linear encoding 89 * of a nondeterministic finite-state machine (aka syntax charts or 90 * "railroad normal form" in parsing technology). Each node is an opcode 91 * plus a "next" pointer, possibly plus an operand. "Next" pointers of 92 * all nodes except BRANCH implement concatenation; a "next" pointer with 93 * a BRANCH on both ends of it is connecting two alternatives. (Here we 94 * have one of the subtle syntax dependencies: an individual BRANCH (as 95 * opposed to a collection of them) is never concatenated with anything 96 * because of operator precedence.) The operand of some types of node is 97 * a literal string; for others, it is a node leading into a sub-FSM. In 98 * particular, the operand of a BRANCH node is the first node of the branch. 99 * (NB this is *not* a tree structure: the tail of the branch connects 100 * to the thing following the set of BRANCHes.) The opcodes are defined 101 * in regnodes.h which is generated from regcomp.sym by regcomp.pl. 102 */ 103 104 /* 105 * A node is one char of opcode followed by two chars of "next" pointer. 106 * "Next" pointers are stored as two 8-bit pieces, high order first. The 107 * value is a positive offset from the opcode of the node containing it. 108 * An operand, if any, simply follows the node. (Note that much of the 109 * code generation knows about this implicit relationship.) 110 * 111 * Using two bytes for the "next" pointer is vast overkill for most things, 112 * but allows patterns to get big without disasters. 113 * 114 * [The "next" pointer is always aligned on an even 115 * boundary, and reads the offset directly as a short. Also, there is no 116 * special test to reverse the sign of BACK pointers since the offset is 117 * stored negative.] 118 */ 119 120 /* This is the stuff that used to live in regexp.h that was truly 121 private to the engine itself. It now lives here. */ 122 123 124 125 typedef struct regexp_internal { 126 int name_list_idx; /* Optional data index of an array of paren names */ 127 union { 128 U32 *offsets; /* offset annotations 20001228 MJD 129 data about mapping the program to the 130 string - 131 offsets[0] is proglen when this is used 132 */ 133 U32 proglen; 134 } u; 135 136 regnode *regstclass; /* Optional startclass as identified or constructed 137 by the optimiser */ 138 struct reg_data *data; /* Additional miscellaneous data used by the program. 139 Used to make it easier to clone and free arbitrary 140 data that the regops need. Often the ARG field of 141 a regop is an index into this structure */ 142 regnode program[1]; /* Unwarranted chumminess with compiler. */ 143 } regexp_internal; 144 145 #define RXi_SET(x,y) (x)->pprivate = (void*)(y) 146 #define RXi_GET(x) ((regexp_internal *)((x)->pprivate)) 147 #define RXi_GET_DECL(r,ri) regexp_internal *ri = RXi_GET(r) 148 /* 149 * Flags stored in regexp->intflags 150 * These are used only internally to the regexp engine 151 * 152 * See regexp.h for flags used externally to the regexp engine 153 */ 154 #define PREGf_SKIP 0x00000001 155 #define PREGf_IMPLICIT 0x00000002 /* Converted .* to ^.* */ 156 #define PREGf_NAUGHTY 0x00000004 /* how exponential is this pattern? */ 157 #define PREGf_VERBARG_SEEN 0x00000008 158 #define PREGf_CUTGROUP_SEEN 0x00000010 159 160 161 /* this is where the old regcomp.h started */ 162 163 struct regnode_string { 164 U8 str_len; 165 U8 type; 166 U16 next_off; 167 char string[1]; 168 }; 169 170 /* Argument bearing node - workhorse, 171 arg1 is often for the data field */ 172 struct regnode_1 { 173 U8 flags; 174 U8 type; 175 U16 next_off; 176 U32 arg1; 177 }; 178 179 /* Similar to a regnode_1 but with an extra signed argument */ 180 struct regnode_2L { 181 U8 flags; 182 U8 type; 183 U16 next_off; 184 U32 arg1; 185 I32 arg2; 186 }; 187 188 /* 'Two field' -- Two 16 bit unsigned args */ 189 struct regnode_2 { 190 U8 flags; 191 U8 type; 192 U16 next_off; 193 U16 arg1; 194 U16 arg2; 195 }; 196 197 198 #define ANYOF_BITMAP_SIZE 32 /* 256 b/(8 b/B) */ 199 #define ANYOF_CLASSBITMAP_SIZE 4 /* up to 40 (8*5) named classes */ 200 201 /* also used by trie */ 202 struct regnode_charclass { 203 U8 flags; 204 U8 type; 205 U16 next_off; 206 U32 arg1; 207 char bitmap[ANYOF_BITMAP_SIZE]; /* only compile-time */ 208 }; 209 210 struct regnode_charclass_class { /* has [[:blah:]] classes */ 211 U8 flags; /* should have ANYOF_CLASS here */ 212 U8 type; 213 U16 next_off; 214 U32 arg1; 215 char bitmap[ANYOF_BITMAP_SIZE]; /* both compile-time */ 216 char classflags[ANYOF_CLASSBITMAP_SIZE]; /* and run-time */ 217 }; 218 219 /* XXX fix this description. 220 Impose a limit of REG_INFTY on various pattern matching operations 221 to limit stack growth and to avoid "infinite" recursions. 222 */ 223 /* The default size for REG_INFTY is I16_MAX, which is the same as 224 SHORT_MAX (see perl.h). Unfortunately I16 isn't necessarily 16 bits 225 (see handy.h). On the Cray C90, sizeof(short)==4 and hence I16_MAX is 226 ((1<<31)-1), while on the Cray T90, sizeof(short)==8 and I16_MAX is 227 ((1<<63)-1). To limit stack growth to reasonable sizes, supply a 228 smaller default. 229 --Andy Dougherty 11 June 1998 230 */ 231 #if SHORTSIZE > 2 232 # ifndef REG_INFTY 233 # define REG_INFTY ((1<<15)-1) 234 # endif 235 #endif 236 237 #ifndef REG_INFTY 238 # define REG_INFTY I16_MAX 239 #endif 240 241 #define ARG_VALUE(arg) (arg) 242 #define ARG__SET(arg,val) ((arg) = (val)) 243 244 #undef ARG 245 #undef ARG1 246 #undef ARG2 247 248 #define ARG(p) ARG_VALUE(ARG_LOC(p)) 249 #define ARG1(p) ARG_VALUE(ARG1_LOC(p)) 250 #define ARG2(p) ARG_VALUE(ARG2_LOC(p)) 251 #define ARG2L(p) ARG_VALUE(ARG2L_LOC(p)) 252 253 #define ARG_SET(p, val) ARG__SET(ARG_LOC(p), (val)) 254 #define ARG1_SET(p, val) ARG__SET(ARG1_LOC(p), (val)) 255 #define ARG2_SET(p, val) ARG__SET(ARG2_LOC(p), (val)) 256 #define ARG2L_SET(p, val) ARG__SET(ARG2L_LOC(p), (val)) 257 258 #undef NEXT_OFF 259 #undef NODE_ALIGN 260 261 #define NEXT_OFF(p) ((p)->next_off) 262 #define NODE_ALIGN(node) 263 #define NODE_ALIGN_FILL(node) ((node)->flags = 0xde) /* deadbeef */ 264 265 #define SIZE_ALIGN NODE_ALIGN 266 267 #undef OP 268 #undef OPERAND 269 #undef MASK 270 #undef STRING 271 272 #define OP(p) ((p)->type) 273 #define OPERAND(p) (((struct regnode_string *)p)->string) 274 #define MASK(p) ((char*)OPERAND(p)) 275 #define STR_LEN(p) (((struct regnode_string *)p)->str_len) 276 #define STRING(p) (((struct regnode_string *)p)->string) 277 #define STR_SZ(l) ((l + sizeof(regnode) - 1) / sizeof(regnode)) 278 #define NODE_SZ_STR(p) (STR_SZ(STR_LEN(p))+1) 279 280 #undef NODE_ALIGN 281 #undef ARG_LOC 282 #undef NEXTOPER 283 #undef PREVOPER 284 285 #define NODE_ALIGN(node) 286 #define ARG_LOC(p) (((struct regnode_1 *)p)->arg1) 287 #define ARG1_LOC(p) (((struct regnode_2 *)p)->arg1) 288 #define ARG2_LOC(p) (((struct regnode_2 *)p)->arg2) 289 #define ARG2L_LOC(p) (((struct regnode_2L *)p)->arg2) 290 291 #define NODE_STEP_REGNODE 1 /* sizeof(regnode)/sizeof(regnode) */ 292 #define EXTRA_STEP_2ARGS EXTRA_SIZE(struct regnode_2) 293 294 #define NODE_STEP_B 4 295 296 #define NEXTOPER(p) ((p) + NODE_STEP_REGNODE) 297 #define PREVOPER(p) ((p) - NODE_STEP_REGNODE) 298 299 #define FILL_ADVANCE_NODE(ptr, op) STMT_START { \ 300 (ptr)->type = op; (ptr)->next_off = 0; (ptr)++; } STMT_END 301 #define FILL_ADVANCE_NODE_ARG(ptr, op, arg) STMT_START { \ 302 ARG_SET(ptr, arg); FILL_ADVANCE_NODE(ptr, op); (ptr) += 1; } STMT_END 303 304 #define REG_MAGIC 0234 305 306 #define SIZE_ONLY (RExC_emit == &PL_regdummy) 307 308 /* Flags for node->flags of ANYOF */ 309 310 #define ANYOF_CLASS 0x08 /* has [[:blah:]] classes */ 311 #define ANYOF_INVERT 0x04 312 #define ANYOF_FOLD 0x02 313 #define ANYOF_LOCALE 0x01 314 315 /* Used for regstclass only */ 316 #define ANYOF_EOS 0x10 /* Can match an empty string too */ 317 318 /* There is a character or a range past 0xff */ 319 #define ANYOF_UNICODE 0x20 320 #define ANYOF_UNICODE_ALL 0x40 /* Can match any char past 0xff */ 321 322 /* size of node is large (includes class pointer) */ 323 #define ANYOF_LARGE 0x80 324 325 /* Are there any runtime flags on in this node? */ 326 #define ANYOF_RUNTIME(s) (ANYOF_FLAGS(s) & 0x0f) 327 328 #define ANYOF_FLAGS_ALL 0xff 329 330 /* Character classes for node->classflags of ANYOF */ 331 /* Should be synchronized with a table in regprop() */ 332 /* 2n should pair with 2n+1 */ 333 334 #define ANYOF_ALNUM 0 /* \w, PL_utf8_alnum, utf8::IsWord, ALNUM */ 335 #define ANYOF_NALNUM 1 336 #define ANYOF_SPACE 2 /* \s */ 337 #define ANYOF_NSPACE 3 338 #define ANYOF_DIGIT 4 /* \d */ 339 #define ANYOF_NDIGIT 5 340 #define ANYOF_ALNUMC 6 /* [[:alnum:]] isalnum(3), utf8::IsAlnum, ALNUMC */ 341 #define ANYOF_NALNUMC 7 342 #define ANYOF_ALPHA 8 343 #define ANYOF_NALPHA 9 344 #define ANYOF_ASCII 10 345 #define ANYOF_NASCII 11 346 #define ANYOF_CNTRL 12 347 #define ANYOF_NCNTRL 13 348 #define ANYOF_GRAPH 14 349 #define ANYOF_NGRAPH 15 350 #define ANYOF_LOWER 16 351 #define ANYOF_NLOWER 17 352 #define ANYOF_PRINT 18 353 #define ANYOF_NPRINT 19 354 #define ANYOF_PUNCT 20 355 #define ANYOF_NPUNCT 21 356 #define ANYOF_UPPER 22 357 #define ANYOF_NUPPER 23 358 #define ANYOF_XDIGIT 24 359 #define ANYOF_NXDIGIT 25 360 #define ANYOF_PSXSPC 26 /* POSIX space: \s plus the vertical tab */ 361 #define ANYOF_NPSXSPC 27 362 #define ANYOF_BLANK 28 /* GNU extension: space and tab: non-vertical space */ 363 #define ANYOF_NBLANK 29 364 365 #define ANYOF_MAX 32 366 367 /* pseudo classes, not stored in the class bitmap, but used as flags 368 during compilation of char classes */ 369 370 #define ANYOF_VERTWS (ANYOF_MAX+1) 371 #define ANYOF_NVERTWS (ANYOF_MAX+2) 372 #define ANYOF_HORIZWS (ANYOF_MAX+3) 373 #define ANYOF_NHORIZWS (ANYOF_MAX+4) 374 375 /* Backward source code compatibility. */ 376 377 #define ANYOF_ALNUML ANYOF_ALNUM 378 #define ANYOF_NALNUML ANYOF_NALNUM 379 #define ANYOF_SPACEL ANYOF_SPACE 380 #define ANYOF_NSPACEL ANYOF_NSPACE 381 382 /* Utility macros for the bitmap and classes of ANYOF */ 383 384 #define ANYOF_SIZE (sizeof(struct regnode_charclass)) 385 #define ANYOF_CLASS_SIZE (sizeof(struct regnode_charclass_class)) 386 387 #define ANYOF_FLAGS(p) ((p)->flags) 388 389 #define ANYOF_BIT(c) (1 << ((c) & 7)) 390 391 #define ANYOF_CLASS_BYTE(p, c) (((struct regnode_charclass_class*)(p))->classflags[((c) >> 3) & 3]) 392 #define ANYOF_CLASS_SET(p, c) (ANYOF_CLASS_BYTE(p, c) |= ANYOF_BIT(c)) 393 #define ANYOF_CLASS_CLEAR(p, c) (ANYOF_CLASS_BYTE(p, c) &= ~ANYOF_BIT(c)) 394 #define ANYOF_CLASS_TEST(p, c) (ANYOF_CLASS_BYTE(p, c) & ANYOF_BIT(c)) 395 396 #define ANYOF_CLASS_ZERO(ret) Zero(((struct regnode_charclass_class*)(ret))->classflags, ANYOF_CLASSBITMAP_SIZE, char) 397 #define ANYOF_BITMAP_ZERO(ret) Zero(((struct regnode_charclass*)(ret))->bitmap, ANYOF_BITMAP_SIZE, char) 398 399 #define ANYOF_BITMAP(p) (((struct regnode_charclass*)(p))->bitmap) 400 #define ANYOF_BITMAP_BYTE(p, c) (ANYOF_BITMAP(p)[(((U8)(c)) >> 3) & 31]) 401 #define ANYOF_BITMAP_SET(p, c) (ANYOF_BITMAP_BYTE(p, c) |= ANYOF_BIT(c)) 402 #define ANYOF_BITMAP_CLEAR(p,c) (ANYOF_BITMAP_BYTE(p, c) &= ~ANYOF_BIT(c)) 403 #define ANYOF_BITMAP_TEST(p, c) (ANYOF_BITMAP_BYTE(p, c) & ANYOF_BIT(c)) 404 405 #define ANYOF_BITMAP_SETALL(p) \ 406 memset (ANYOF_BITMAP(p), 255, ANYOF_BITMAP_SIZE) 407 #define ANYOF_BITMAP_CLEARALL(p) \ 408 Zero (ANYOF_BITMAP(p), ANYOF_BITMAP_SIZE) 409 /* Check that all 256 bits are all set. Used in S_cl_is_anything() */ 410 #define ANYOF_BITMAP_TESTALLSET(p) \ 411 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) 412 413 #define ANYOF_SKIP ((ANYOF_SIZE - 1)/sizeof(regnode)) 414 #define ANYOF_CLASS_SKIP ((ANYOF_CLASS_SIZE - 1)/sizeof(regnode)) 415 #define ANYOF_CLASS_ADD_SKIP (ANYOF_CLASS_SKIP - ANYOF_SKIP) 416 417 418 /* 419 * Utility definitions. 420 */ 421 #ifndef CHARMASK 422 # define UCHARAT(p) ((int)*(const U8*)(p)) 423 #else 424 # define UCHARAT(p) ((int)*(p)&CHARMASK) 425 #endif 426 427 #define EXTRA_SIZE(guy) ((sizeof(guy)-1)/sizeof(struct regnode)) 428 429 #define REG_SEEN_ZERO_LEN 0x00000001 430 #define REG_SEEN_LOOKBEHIND 0x00000002 431 #define REG_SEEN_GPOS 0x00000004 432 #define REG_SEEN_EVAL 0x00000008 433 #define REG_SEEN_CANY 0x00000010 434 #define REG_SEEN_SANY REG_SEEN_CANY /* src bckwrd cmpt */ 435 #define REG_SEEN_RECURSE 0x00000020 436 #define REG_TOP_LEVEL_BRANCHES 0x00000040 437 #define REG_SEEN_VERBARG 0x00000080 438 #define REG_SEEN_CUTGROUP 0x00000100 439 #define REG_SEEN_RUN_ON_COMMENT 0x00000200 440 441 START_EXTERN_C 442 443 #ifdef PLUGGABLE_RE_EXTENSION 444 #include "re_nodes.h" 445 #else 446 #include "regnodes.h" 447 #endif 448 449 /* The following have no fixed length. U8 so we can do strchr() on it. */ 450 #ifndef DOINIT 451 EXTCONST U8 PL_varies[]; 452 #else 453 EXTCONST U8 PL_varies[] = { 454 BRANCH, BACK, STAR, PLUS, CURLY, CURLYX, REF, REFF, REFFL, 455 WHILEM, CURLYM, CURLYN, BRANCHJ, IFTHEN, SUSPEND, CLUMP, 456 NREF, NREFF, NREFFL, 457 0 458 }; 459 #endif 460 461 /* The following always have a length of 1. U8 we can do strchr() on it. */ 462 /* (Note that length 1 means "one character" under UTF8, not "one octet".) */ 463 #ifndef DOINIT 464 EXTCONST U8 PL_simple[]; 465 #else 466 EXTCONST U8 PL_simple[] = { 467 REG_ANY, SANY, CANY, 468 ANYOF, 469 ALNUM, ALNUML, 470 NALNUM, NALNUML, 471 SPACE, SPACEL, 472 NSPACE, NSPACEL, 473 DIGIT, NDIGIT, 474 VERTWS, NVERTWS, 475 HORIZWS, NHORIZWS, 476 0 477 }; 478 #endif 479 480 #ifndef PLUGGABLE_RE_EXTENSION 481 #ifndef DOINIT 482 EXTCONST regexp_engine PL_core_reg_engine; 483 #else /* DOINIT */ 484 EXTCONST regexp_engine PL_core_reg_engine = { 485 Perl_re_compile, 486 Perl_regexec_flags, 487 Perl_re_intuit_start, 488 Perl_re_intuit_string, 489 Perl_regfree_internal, 490 Perl_reg_numbered_buff_fetch, 491 Perl_reg_numbered_buff_store, 492 Perl_reg_numbered_buff_length, 493 Perl_reg_named_buff, 494 Perl_reg_named_buff_iter, 495 Perl_reg_qr_package, 496 #if defined(USE_ITHREADS) 497 Perl_regdupe_internal 498 #endif 499 }; 500 #endif /* DOINIT */ 501 #endif /* PLUGGABLE_RE_EXTENSION */ 502 503 504 END_EXTERN_C 505 506 507 /* .what is a character array with one character for each member of .data 508 * The character describes the function of the corresponding .data item: 509 * f - start-class data for regstclass optimization 510 * n - Root of op tree for (?{EVAL}) item 511 * o - Start op for (?{EVAL}) item 512 * p - Pad for (?{EVAL}) item 513 * s - swash for Unicode-style character class, and the multicharacter 514 * strings resulting from casefolding the single-character entries 515 * in the character class 516 * t - trie struct 517 * u - trie struct's widecharmap (a HV, so can't share, must dup) 518 * also used for revcharmap and words under DEBUGGING 519 * T - aho-trie struct 520 * S - sv for named capture lookup 521 * 20010712 mjd@plover.com 522 * (Remember to update re_dup() and pregfree() if you add any items.) 523 */ 524 struct reg_data { 525 U32 count; 526 U8 *what; 527 void* data[1]; 528 }; 529 530 /* Code in S_to_utf8_substr() and S_to_byte_substr() in regexec.c accesses 531 anchored* and float* via array indexes 0 and 1. */ 532 #define anchored_substr substrs->data[0].substr 533 #define anchored_utf8 substrs->data[0].utf8_substr 534 #define anchored_offset substrs->data[0].min_offset 535 #define anchored_end_shift substrs->data[0].end_shift 536 537 #define float_substr substrs->data[1].substr 538 #define float_utf8 substrs->data[1].utf8_substr 539 #define float_min_offset substrs->data[1].min_offset 540 #define float_max_offset substrs->data[1].max_offset 541 #define float_end_shift substrs->data[1].end_shift 542 543 #define check_substr substrs->data[2].substr 544 #define check_utf8 substrs->data[2].utf8_substr 545 #define check_offset_min substrs->data[2].min_offset 546 #define check_offset_max substrs->data[2].max_offset 547 #define check_end_shift substrs->data[2].end_shift 548 549 #define RX_ANCHORED_SUBSTR(rx) ((rx)->anchored_substr) 550 #define RX_ANCHORED_UTF8(rx) ((rx)->anchored_utf8) 551 #define RX_FLOAT_SUBSTR(rx) ((rx)->float_substr) 552 #define RX_FLOAT_UTF8(rx) ((rx)->float_utf8) 553 554 /* trie related stuff */ 555 556 /* a transition record for the state machine. the 557 check field determines which state "owns" the 558 transition. the char the transition is for is 559 determined by offset from the owning states base 560 field. the next field determines which state 561 is to be transitioned to if any. 562 */ 563 struct _reg_trie_trans { 564 U32 next; 565 U32 check; 566 }; 567 568 /* a transition list element for the list based representation */ 569 struct _reg_trie_trans_list_elem { 570 U16 forid; 571 U32 newstate; 572 }; 573 typedef struct _reg_trie_trans_list_elem reg_trie_trans_le; 574 575 /* a state for compressed nodes. base is an offset 576 into an array of reg_trie_trans array. If wordnum is 577 nonzero the state is accepting. if base is zero then 578 the state has no children (and will be accepting) 579 */ 580 struct _reg_trie_state { 581 U16 wordnum; 582 union { 583 U32 base; 584 reg_trie_trans_le* list; 585 } trans; 586 }; 587 588 589 590 typedef struct _reg_trie_state reg_trie_state; 591 typedef struct _reg_trie_trans reg_trie_trans; 592 593 594 /* anything in here that needs to be freed later 595 should be dealt with in pregfree. 596 refcount is first in both this and _reg_ac_data to allow a space 597 optimisation in Perl_regdupe. */ 598 struct _reg_trie_data { 599 U32 refcount; /* number of times this trie is referenced */ 600 U32 lasttrans; /* last valid transition element */ 601 U16 *charmap; /* byte to charid lookup array */ 602 reg_trie_state *states; /* state data */ 603 reg_trie_trans *trans; /* array of transition elements */ 604 char *bitmap; /* stclass bitmap */ 605 U32 *wordlen; /* array of lengths of words */ 606 U16 *jump; /* optional 1 indexed array of offsets before tail 607 for the node following a given word. */ 608 U16 *nextword; /* optional 1 indexed array to support linked list 609 of duplicate wordnums */ 610 U16 uniquecharcount; /* unique chars in trie (width of trans table) */ 611 U32 startstate; /* initial state - used for common prefix optimisation */ 612 STRLEN minlen; /* minimum length of words in trie - build/opt only? */ 613 STRLEN maxlen; /* maximum length of words in trie - build/opt only? */ 614 U32 statecount; /* Build only - number of states in the states array 615 (including the unused zero state) */ 616 U32 wordcount; /* Build only */ 617 #ifdef DEBUGGING 618 STRLEN charcount; /* Build only */ 619 #endif 620 }; 621 /* There is one (3 under DEBUGGING) pointers that logically belong in this 622 structure, but are held outside as they need duplication on thread cloning, 623 whereas the rest of the structure can be read only: 624 HV *widecharmap; code points > 255 to charid 625 #ifdef DEBUGGING 626 AV *words; Array of words contained in trie, for dumping 627 AV *revcharmap; Map of each charid back to its character representation 628 #endif 629 */ 630 631 #define TRIE_WORDS_OFFSET 2 632 633 typedef struct _reg_trie_data reg_trie_data; 634 635 /* refcount is first in both this and _reg_trie_data to allow a space 636 optimisation in Perl_regdupe. */ 637 struct _reg_ac_data { 638 U32 refcount; 639 U32 trie; 640 U32 *fail; 641 reg_trie_state *states; 642 }; 643 typedef struct _reg_ac_data reg_ac_data; 644 645 /* ANY_BIT doesnt use the structure, so we can borrow it here. 646 This is simpler than refactoring all of it as wed end up with 647 three different sets... */ 648 649 #define TRIE_BITMAP(p) (((reg_trie_data *)(p))->bitmap) 650 #define TRIE_BITMAP_BYTE(p, c) (TRIE_BITMAP(p)[(((U8)(c)) >> 3) & 31]) 651 #define TRIE_BITMAP_SET(p, c) (TRIE_BITMAP_BYTE(p, c) |= ANYOF_BIT((U8)c)) 652 #define TRIE_BITMAP_CLEAR(p,c) (TRIE_BITMAP_BYTE(p, c) &= ~ANYOF_BIT((U8)c)) 653 #define TRIE_BITMAP_TEST(p, c) (TRIE_BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c)) 654 655 #define IS_ANYOF_TRIE(op) ((op)==TRIEC || (op)==AHOCORASICKC) 656 #define IS_TRIE_AC(op) ((op)>=AHOCORASICK) 657 658 659 #define BITMAP_BYTE(p, c) (((U8*)p)[(((U8)(c)) >> 3) & 31]) 660 #define BITMAP_TEST(p, c) (BITMAP_BYTE(p, c) & ANYOF_BIT((U8)c)) 661 662 /* these defines assume uniquecharcount is the correct variable, and state may be evaluated twice */ 663 #define TRIE_NODENUM(state) (((state)-1)/(trie->uniquecharcount)+1) 664 #define SAFE_TRIE_NODENUM(state) ((state) ? (((state)-1)/(trie->uniquecharcount)+1) : (state)) 665 #define TRIE_NODEIDX(state) ((state) ? (((state)-1)*(trie->uniquecharcount)+1) : (state)) 666 667 #ifdef DEBUGGING 668 #define TRIE_CHARCOUNT(trie) ((trie)->charcount) 669 #else 670 #define TRIE_CHARCOUNT(trie) (trie_charcount) 671 #endif 672 673 #define RE_TRIE_MAXBUF_INIT 65536 674 #define RE_TRIE_MAXBUF_NAME "\022E_TRIE_MAXBUF" 675 #define RE_DEBUG_FLAGS "\022E_DEBUG_FLAGS" 676 677 /* 678 679 RE_DEBUG_FLAGS is used to control what debug output is emitted 680 its divided into three groups of options, some of which interact. 681 The three groups are: Compile, Execute, Extra. There is room for a 682 further group, as currently only the low three bytes are used. 683 684 Compile Options: 685 686 PARSE 687 PEEP 688 TRIE 689 PROGRAM 690 OFFSETS 691 692 Execute Options: 693 694 INTUIT 695 MATCH 696 TRIE 697 698 Extra Options 699 700 TRIE 701 OFFSETS 702 703 If you modify any of these make sure you make corresponding changes to 704 re.pm, especially to the documentation. 705 706 */ 707 708 709 /* Compile */ 710 #define RE_DEBUG_COMPILE_MASK 0x0000FF 711 #define RE_DEBUG_COMPILE_PARSE 0x000001 712 #define RE_DEBUG_COMPILE_OPTIMISE 0x000002 713 #define RE_DEBUG_COMPILE_TRIE 0x000004 714 #define RE_DEBUG_COMPILE_DUMP 0x000008 715 #define RE_DEBUG_COMPILE_FLAGS 0x000010 716 717 /* Execute */ 718 #define RE_DEBUG_EXECUTE_MASK 0x00FF00 719 #define RE_DEBUG_EXECUTE_INTUIT 0x000100 720 #define RE_DEBUG_EXECUTE_MATCH 0x000200 721 #define RE_DEBUG_EXECUTE_TRIE 0x000400 722 723 /* Extra */ 724 #define RE_DEBUG_EXTRA_MASK 0xFF0000 725 #define RE_DEBUG_EXTRA_TRIE 0x010000 726 #define RE_DEBUG_EXTRA_OFFSETS 0x020000 727 #define RE_DEBUG_EXTRA_OFFDEBUG 0x040000 728 #define RE_DEBUG_EXTRA_STATE 0x080000 729 #define RE_DEBUG_EXTRA_OPTIMISE 0x100000 730 #define RE_DEBUG_EXTRA_BUFFERS 0x400000 731 /* combined */ 732 #define RE_DEBUG_EXTRA_STACK 0x280000 733 734 #define RE_DEBUG_FLAG(x) (re_debug_flags & x) 735 /* Compile */ 736 #define DEBUG_COMPILE_r(x) DEBUG_r( \ 737 if (re_debug_flags & RE_DEBUG_COMPILE_MASK) x ) 738 #define DEBUG_PARSE_r(x) DEBUG_r( \ 739 if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x ) 740 #define DEBUG_OPTIMISE_r(x) DEBUG_r( \ 741 if (re_debug_flags & RE_DEBUG_COMPILE_OPTIMISE) x ) 742 #define DEBUG_PARSE_r(x) DEBUG_r( \ 743 if (re_debug_flags & RE_DEBUG_COMPILE_PARSE) x ) 744 #define DEBUG_DUMP_r(x) DEBUG_r( \ 745 if (re_debug_flags & RE_DEBUG_COMPILE_DUMP) x ) 746 #define DEBUG_TRIE_COMPILE_r(x) DEBUG_r( \ 747 if (re_debug_flags & RE_DEBUG_COMPILE_TRIE) x ) 748 #define DEBUG_FLAGS_r(x) DEBUG_r( \ 749 if (re_debug_flags & RE_DEBUG_COMPILE_FLAGS) x ) 750 /* Execute */ 751 #define DEBUG_EXECUTE_r(x) DEBUG_r( \ 752 if (re_debug_flags & RE_DEBUG_EXECUTE_MASK) x ) 753 #define DEBUG_INTUIT_r(x) DEBUG_r( \ 754 if (re_debug_flags & RE_DEBUG_EXECUTE_INTUIT) x ) 755 #define DEBUG_MATCH_r(x) DEBUG_r( \ 756 if (re_debug_flags & RE_DEBUG_EXECUTE_MATCH) x ) 757 #define DEBUG_TRIE_EXECUTE_r(x) DEBUG_r( \ 758 if (re_debug_flags & RE_DEBUG_EXECUTE_TRIE) x ) 759 760 /* Extra */ 761 #define DEBUG_EXTRA_r(x) DEBUG_r( \ 762 if (re_debug_flags & RE_DEBUG_EXTRA_MASK) x ) 763 #define DEBUG_OFFSETS_r(x) DEBUG_r( \ 764 if (re_debug_flags & RE_DEBUG_EXTRA_OFFSETS) x ) 765 #define DEBUG_STATE_r(x) DEBUG_r( \ 766 if (re_debug_flags & RE_DEBUG_EXTRA_STATE) x ) 767 #define DEBUG_STACK_r(x) DEBUG_r( \ 768 if (re_debug_flags & RE_DEBUG_EXTRA_STACK) x ) 769 #define DEBUG_BUFFERS_r(x) DEBUG_r( \ 770 if (re_debug_flags & RE_DEBUG_EXTRA_BUFFERS) x ) 771 772 #define DEBUG_OPTIMISE_MORE_r(x) DEBUG_r( \ 773 if ((RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE) == \ 774 (re_debug_flags & (RE_DEBUG_EXTRA_OPTIMISE|RE_DEBUG_COMPILE_OPTIMISE)) ) x ) 775 #define MJD_OFFSET_DEBUG(x) DEBUG_r( \ 776 if (re_debug_flags & RE_DEBUG_EXTRA_OFFDEBUG) \ 777 Perl_warn_nocontext x ) 778 #define DEBUG_TRIE_COMPILE_MORE_r(x) DEBUG_TRIE_COMPILE_r( \ 779 if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x ) 780 #define DEBUG_TRIE_EXECUTE_MORE_r(x) DEBUG_TRIE_EXECUTE_r( \ 781 if (re_debug_flags & RE_DEBUG_EXTRA_TRIE) x ) 782 783 #define DEBUG_TRIE_r(x) DEBUG_r( \ 784 if (re_debug_flags & (RE_DEBUG_COMPILE_TRIE \ 785 | RE_DEBUG_EXECUTE_TRIE )) x ) 786 787 /* initialization */ 788 /* get_sv() can return NULL during global destruction. */ 789 #define GET_RE_DEBUG_FLAGS DEBUG_r({ \ 790 SV * re_debug_flags_sv = NULL; \ 791 re_debug_flags_sv = get_sv(RE_DEBUG_FLAGS, 1); \ 792 if (re_debug_flags_sv) { \ 793 if (!SvIOK(re_debug_flags_sv)) \ 794 sv_setuv(re_debug_flags_sv, RE_DEBUG_COMPILE_DUMP | RE_DEBUG_EXECUTE_MASK ); \ 795 re_debug_flags=SvIV(re_debug_flags_sv); \ 796 }\ 797 }) 798 799 #ifdef DEBUGGING 800 801 #define GET_RE_DEBUG_FLAGS_DECL IV re_debug_flags = 0; GET_RE_DEBUG_FLAGS; 802 803 #define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2) \ 804 const char * const rpv = \ 805 pv_pretty((dsv), (pv), (l), (m), \ 806 PL_colors[(c1)],PL_colors[(c2)], \ 807 PERL_PV_ESCAPE_RE |((isuni) ? PERL_PV_ESCAPE_UNI : 0) ); \ 808 const int rlen = SvCUR(dsv) 809 810 #define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m) \ 811 const char * const rpv = \ 812 pv_pretty((dsv), (SvPV_nolen_const(sv)), (SvCUR(sv)), (m), \ 813 PL_colors[(c1)],PL_colors[(c2)], \ 814 PERL_PV_ESCAPE_RE |((isuni) ? PERL_PV_ESCAPE_UNI : 0) ) 815 816 #define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m) \ 817 const char * const rpv = \ 818 pv_pretty((dsv), (pv), (l), (m), \ 819 PL_colors[0], PL_colors[1], \ 820 ( PERL_PV_PRETTY_QUOTE | PERL_PV_ESCAPE_RE | PERL_PV_PRETTY_ELLIPSES | \ 821 ((isuni) ? PERL_PV_ESCAPE_UNI : 0)) \ 822 ) 823 824 #define RE_SV_DUMPLEN(ItEm) (SvCUR(ItEm) - (SvTAIL(ItEm)!=0)) 825 #define RE_SV_TAIL(ItEm) (SvTAIL(ItEm) ? "$" : "") 826 827 #else /* if not DEBUGGING */ 828 829 #define GET_RE_DEBUG_FLAGS_DECL 830 #define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2) 831 #define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m) 832 #define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m) 833 #define RE_SV_DUMPLEN(ItEm) 834 #define RE_SV_TAIL(ItEm) 835 836 #endif /* DEBUG RELATED DEFINES */ 837 838 /* 839 * Local variables: 840 * c-indentation-style: bsd 841 * c-basic-offset: 4 842 * indent-tabs-mode: t 843 * End: 844 * 845 * ex: set ts=8 sts=4 sw=4 noet: 846 */ 847