1 // Copyright 2006 The RE2 Authors. All Rights Reserved. 2 // Use of this source code is governed by a BSD-style 3 // license that can be found in the LICENSE file. 4 5 #ifndef RE2_REGEXP_H_ 6 #define RE2_REGEXP_H_ 7 8 // --- SPONSORED LINK -------------------------------------------------- 9 // If you want to use this library for regular expression matching, 10 // you should use re2/re2.h, which provides a class RE2 that 11 // mimics the PCRE interface provided by PCRE's C++ wrappers. 12 // This header describes the low-level interface used to implement RE2 13 // and may change in backwards-incompatible ways from time to time. 14 // In contrast, RE2's interface will not. 15 // --------------------------------------------------------------------- 16 17 // Regular expression library: parsing, execution, and manipulation 18 // of regular expressions. 19 // 20 // Any operation that traverses the Regexp structures should be written 21 // using Regexp::Walker (see walker-inl.h), not recursively, because deeply nested 22 // regular expressions such as x++++++++++++++++++++... might cause recursive 23 // traversals to overflow the stack. 24 // 25 // It is the caller's responsibility to provide appropriate mutual exclusion 26 // around manipulation of the regexps. RE2 does this. 27 // 28 // PARSING 29 // 30 // Regexp::Parse parses regular expressions encoded in UTF-8. 31 // The default syntax is POSIX extended regular expressions, 32 // with the following changes: 33 // 34 // 1. Backreferences (optional in POSIX EREs) are not supported. 35 // (Supporting them precludes the use of DFA-based 36 // matching engines.) 37 // 38 // 2. Collating elements and collation classes are not supported. 39 // (No one has needed or wanted them.) 40 // 41 // The exact syntax accepted can be modified by passing flags to 42 // Regexp::Parse. In particular, many of the basic Perl additions 43 // are available. The flags are documented below (search for LikePerl). 44 // 45 // If parsed with the flag Regexp::Latin1, both the regular expression 46 // and the input to the matching routines are assumed to be encoded in 47 // Latin-1, not UTF-8. 48 // 49 // EXECUTION 50 // 51 // Once Regexp has parsed a regular expression, it provides methods 52 // to search text using that regular expression. These methods are 53 // implemented via calling out to other regular expression libraries. 54 // (Let's call them the sublibraries.) 55 // 56 // To call a sublibrary, Regexp does not simply prepare a 57 // string version of the regular expression and hand it to the 58 // sublibrary. Instead, Regexp prepares, from its own parsed form, the 59 // corresponding internal representation used by the sublibrary. 60 // This has the drawback of needing to know the internal representation 61 // used by the sublibrary, but it has two important benefits: 62 // 63 // 1. The syntax and meaning of regular expressions is guaranteed 64 // to be that used by Regexp's parser, not the syntax expected 65 // by the sublibrary. Regexp might accept a restricted or 66 // expanded syntax for regular expressions as compared with 67 // the sublibrary. As long as Regexp can translate from its 68 // internal form into the sublibrary's, clients need not know 69 // exactly which sublibrary they are using. 70 // 71 // 2. The sublibrary parsers are bypassed. For whatever reason, 72 // sublibrary regular expression parsers often have security 73 // problems. For example, plan9grep's regular expression parser 74 // has a buffer overflow in its handling of large character 75 // classes, and PCRE's parser has had buffer overflow problems 76 // in the past. Security-team requires sandboxing of sublibrary 77 // regular expression parsers. Avoiding the sublibrary parsers 78 // avoids the sandbox. 79 // 80 // The execution methods we use now are provided by the compiled form, 81 // Prog, described in prog.h 82 // 83 // MANIPULATION 84 // 85 // Unlike other regular expression libraries, Regexp makes its parsed 86 // form accessible to clients, so that client code can analyze the 87 // parsed regular expressions. 88 89 #include <stdint.h> 90 #include <map> 91 #include <set> 92 #include <string> 93 94 #include "util/util.h" 95 #include "util/logging.h" 96 #include "util/utf.h" 97 #include "re2/stringpiece.h" 98 99 namespace re2 { 100 101 // Keep in sync with string list kOpcodeNames[] in testing/dump.cc 102 enum RegexpOp { 103 // Matches no strings. 104 kRegexpNoMatch = 1, 105 106 // Matches empty string. 107 kRegexpEmptyMatch, 108 109 // Matches rune_. 110 kRegexpLiteral, 111 112 // Matches runes_. 113 kRegexpLiteralString, 114 115 // Matches concatenation of sub_[0..nsub-1]. 116 kRegexpConcat, 117 // Matches union of sub_[0..nsub-1]. 118 kRegexpAlternate, 119 120 // Matches sub_[0] zero or more times. 121 kRegexpStar, 122 // Matches sub_[0] one or more times. 123 kRegexpPlus, 124 // Matches sub_[0] zero or one times. 125 kRegexpQuest, 126 127 // Matches sub_[0] at least min_ times, at most max_ times. 128 // max_ == -1 means no upper limit. 129 kRegexpRepeat, 130 131 // Parenthesized (capturing) subexpression. Index is cap_. 132 // Optionally, capturing name is name_. 133 kRegexpCapture, 134 135 // Matches any character. 136 kRegexpAnyChar, 137 138 // Matches any byte [sic]. 139 kRegexpAnyByte, 140 141 // Matches empty string at beginning of line. 142 kRegexpBeginLine, 143 // Matches empty string at end of line. 144 kRegexpEndLine, 145 146 // Matches word boundary "\b". 147 kRegexpWordBoundary, 148 // Matches not-a-word boundary "\B". 149 kRegexpNoWordBoundary, 150 151 // Matches empty string at beginning of text. 152 kRegexpBeginText, 153 // Matches empty string at end of text. 154 kRegexpEndText, 155 156 // Matches character class given by cc_. 157 kRegexpCharClass, 158 159 // Forces match of entire expression right now, 160 // with match ID match_id_ (used by RE2::Set). 161 kRegexpHaveMatch, 162 163 kMaxRegexpOp = kRegexpHaveMatch, 164 }; 165 166 // Keep in sync with string list in regexp.cc 167 enum RegexpStatusCode { 168 // No error 169 kRegexpSuccess = 0, 170 171 // Unexpected error 172 kRegexpInternalError, 173 174 // Parse errors 175 kRegexpBadEscape, // bad escape sequence 176 kRegexpBadCharClass, // bad character class 177 kRegexpBadCharRange, // bad character class range 178 kRegexpMissingBracket, // missing closing ] 179 kRegexpMissingParen, // missing closing ) 180 kRegexpTrailingBackslash, // at end of regexp 181 kRegexpRepeatArgument, // repeat argument missing, e.g. "*" 182 kRegexpRepeatSize, // bad repetition argument 183 kRegexpRepeatOp, // bad repetition operator 184 kRegexpBadPerlOp, // bad perl operator 185 kRegexpBadUTF8, // invalid UTF-8 in regexp 186 kRegexpBadNamedCapture, // bad named capture 187 }; 188 189 // Error status for certain operations. 190 class RegexpStatus { 191 public: RegexpStatus()192 RegexpStatus() : code_(kRegexpSuccess), tmp_(NULL) {} ~RegexpStatus()193 ~RegexpStatus() { delete tmp_; } 194 set_code(RegexpStatusCode code)195 void set_code(RegexpStatusCode code) { code_ = code; } set_error_arg(const StringPiece & error_arg)196 void set_error_arg(const StringPiece& error_arg) { error_arg_ = error_arg; } set_tmp(std::string * tmp)197 void set_tmp(std::string* tmp) { delete tmp_; tmp_ = tmp; } code()198 RegexpStatusCode code() const { return code_; } error_arg()199 const StringPiece& error_arg() const { return error_arg_; } ok()200 bool ok() const { return code() == kRegexpSuccess; } 201 202 // Copies state from status. 203 void Copy(const RegexpStatus& status); 204 205 // Returns text equivalent of code, e.g.: 206 // "Bad character class" 207 static std::string CodeText(RegexpStatusCode code); 208 209 // Returns text describing error, e.g.: 210 // "Bad character class: [z-a]" 211 std::string Text() const; 212 213 private: 214 RegexpStatusCode code_; // Kind of error 215 StringPiece error_arg_; // Piece of regexp containing syntax error. 216 std::string* tmp_; // Temporary storage, possibly where error_arg_ is. 217 218 RegexpStatus(const RegexpStatus&) = delete; 219 RegexpStatus& operator=(const RegexpStatus&) = delete; 220 }; 221 222 // Compiled form; see prog.h 223 class Prog; 224 225 struct RuneRange { RuneRangeRuneRange226 RuneRange() : lo(0), hi(0) { } RuneRangeRuneRange227 RuneRange(int l, int h) : lo(l), hi(h) { } 228 Rune lo; 229 Rune hi; 230 }; 231 232 // Less-than on RuneRanges treats a == b if they overlap at all. 233 // This lets us look in a set to find the range covering a particular Rune. 234 struct RuneRangeLess { operatorRuneRangeLess235 bool operator()(const RuneRange& a, const RuneRange& b) const { 236 return a.hi < b.lo; 237 } 238 }; 239 240 class CharClassBuilder; 241 242 class CharClass { 243 public: 244 void Delete(); 245 246 typedef RuneRange* iterator; begin()247 iterator begin() { return ranges_; } end()248 iterator end() { return ranges_ + nranges_; } 249 size()250 int size() { return nrunes_; } empty()251 bool empty() { return nrunes_ == 0; } full()252 bool full() { return nrunes_ == Runemax+1; } FoldsASCII()253 bool FoldsASCII() { return folds_ascii_; } 254 255 bool Contains(Rune r); 256 CharClass* Negate(); 257 258 private: 259 CharClass(); // not implemented 260 ~CharClass(); // not implemented 261 static CharClass* New(int maxranges); 262 263 friend class CharClassBuilder; 264 265 bool folds_ascii_; 266 int nrunes_; 267 RuneRange *ranges_; 268 int nranges_; 269 270 CharClass(const CharClass&) = delete; 271 CharClass& operator=(const CharClass&) = delete; 272 }; 273 274 class Regexp { 275 public: 276 277 // Flags for parsing. Can be ORed together. 278 enum ParseFlags { 279 NoParseFlags = 0, 280 FoldCase = 1<<0, // Fold case during matching (case-insensitive). 281 Literal = 1<<1, // Treat s as literal string instead of a regexp. 282 ClassNL = 1<<2, // Allow char classes like [^a-z] and \D and \s 283 // and [[:space:]] to match newline. 284 DotNL = 1<<3, // Allow . to match newline. 285 MatchNL = ClassNL | DotNL, 286 OneLine = 1<<4, // Treat ^ and $ as only matching at beginning and 287 // end of text, not around embedded newlines. 288 // (Perl's default) 289 Latin1 = 1<<5, // Regexp and text are in Latin1, not UTF-8. 290 NonGreedy = 1<<6, // Repetition operators are non-greedy by default. 291 PerlClasses = 1<<7, // Allow Perl character classes like \d. 292 PerlB = 1<<8, // Allow Perl's \b and \B. 293 PerlX = 1<<9, // Perl extensions: 294 // non-capturing parens - (?: ) 295 // non-greedy operators - *? +? ?? {}? 296 // flag edits - (?i) (?-i) (?i: ) 297 // i - FoldCase 298 // m - !OneLine 299 // s - DotNL 300 // U - NonGreedy 301 // line ends: \A \z 302 // \Q and \E to disable/enable metacharacters 303 // (?P<name>expr) for named captures 304 // \C to match any single byte 305 UnicodeGroups = 1<<10, // Allow \p{Han} for Unicode Han group 306 // and \P{Han} for its negation. 307 NeverNL = 1<<11, // Never match NL, even if the regexp mentions 308 // it explicitly. 309 NeverCapture = 1<<12, // Parse all parens as non-capturing. 310 311 // As close to Perl as we can get. 312 LikePerl = ClassNL | OneLine | PerlClasses | PerlB | PerlX | 313 UnicodeGroups, 314 315 // Internal use only. 316 WasDollar = 1<<13, // on kRegexpEndText: was $ in regexp text 317 AllParseFlags = (1<<14)-1, 318 }; 319 320 // Get. No set, Regexps are logically immutable once created. op()321 RegexpOp op() { return static_cast<RegexpOp>(op_); } nsub()322 int nsub() { return nsub_; } simple()323 bool simple() { return simple_ != 0; } parse_flags()324 ParseFlags parse_flags() { return static_cast<ParseFlags>(parse_flags_); } 325 int Ref(); // For testing. 326 sub()327 Regexp** sub() { 328 if(nsub_ <= 1) 329 return &subone_; 330 else 331 return submany_; 332 } 333 min()334 int min() { DCHECK_EQ(op_, kRegexpRepeat); return min_; } max()335 int max() { DCHECK_EQ(op_, kRegexpRepeat); return max_; } rune()336 Rune rune() { DCHECK_EQ(op_, kRegexpLiteral); return rune_; } cc()337 CharClass* cc() { DCHECK_EQ(op_, kRegexpCharClass); return cc_; } cap()338 int cap() { DCHECK_EQ(op_, kRegexpCapture); return cap_; } name()339 const std::string* name() { DCHECK_EQ(op_, kRegexpCapture); return name_; } runes()340 Rune* runes() { DCHECK_EQ(op_, kRegexpLiteralString); return runes_; } nrunes()341 int nrunes() { DCHECK_EQ(op_, kRegexpLiteralString); return nrunes_; } match_id()342 int match_id() { DCHECK_EQ(op_, kRegexpHaveMatch); return match_id_; } 343 344 // Increments reference count, returns object as convenience. 345 Regexp* Incref(); 346 347 // Decrements reference count and deletes this object if count reaches 0. 348 void Decref(); 349 350 // Parses string s to produce regular expression, returned. 351 // Caller must release return value with re->Decref(). 352 // On failure, sets *status (if status != NULL) and returns NULL. 353 static Regexp* Parse(const StringPiece& s, ParseFlags flags, 354 RegexpStatus* status); 355 356 // Returns a _new_ simplified version of the current regexp. 357 // Does not edit the current regexp. 358 // Caller must release return value with re->Decref(). 359 // Simplified means that counted repetition has been rewritten 360 // into simpler terms and all Perl/POSIX features have been 361 // removed. The result will capture exactly the same 362 // subexpressions the original did, unless formatted with ToString. 363 Regexp* Simplify(); 364 friend class CoalesceWalker; 365 friend class SimplifyWalker; 366 367 // Parses the regexp src and then simplifies it and sets *dst to the 368 // string representation of the simplified form. Returns true on success. 369 // Returns false and sets *status (if status != NULL) on parse error. 370 static bool SimplifyRegexp(const StringPiece& src, ParseFlags flags, 371 std::string* dst, RegexpStatus* status); 372 373 // Returns the number of capturing groups in the regexp. 374 int NumCaptures(); 375 friend class NumCapturesWalker; 376 377 // Returns a map from names to capturing group indices, 378 // or NULL if the regexp contains no named capture groups. 379 // The caller is responsible for deleting the map. 380 std::map<std::string, int>* NamedCaptures(); 381 382 // Returns a map from capturing group indices to capturing group 383 // names or NULL if the regexp contains no named capture groups. The 384 // caller is responsible for deleting the map. 385 std::map<int, std::string>* CaptureNames(); 386 387 // Returns a string representation of the current regexp, 388 // using as few parentheses as possible. 389 std::string ToString(); 390 391 // Convenience functions. They consume the passed reference, 392 // so in many cases you should use, e.g., Plus(re->Incref(), flags). 393 // They do not consume allocated arrays like subs or runes. 394 static Regexp* Plus(Regexp* sub, ParseFlags flags); 395 static Regexp* Star(Regexp* sub, ParseFlags flags); 396 static Regexp* Quest(Regexp* sub, ParseFlags flags); 397 static Regexp* Concat(Regexp** subs, int nsubs, ParseFlags flags); 398 static Regexp* Alternate(Regexp** subs, int nsubs, ParseFlags flags); 399 static Regexp* Capture(Regexp* sub, ParseFlags flags, int cap); 400 static Regexp* Repeat(Regexp* sub, ParseFlags flags, int min, int max); 401 static Regexp* NewLiteral(Rune rune, ParseFlags flags); 402 static Regexp* NewCharClass(CharClass* cc, ParseFlags flags); 403 static Regexp* LiteralString(Rune* runes, int nrunes, ParseFlags flags); 404 static Regexp* HaveMatch(int match_id, ParseFlags flags); 405 406 // Like Alternate but does not factor out common prefixes. 407 static Regexp* AlternateNoFactor(Regexp** subs, int nsubs, ParseFlags flags); 408 409 // Debugging function. Returns string format for regexp 410 // that makes structure clear. Does NOT use regexp syntax. 411 std::string Dump(); 412 413 // Helper traversal class, defined fully in walker-inl.h. 414 template<typename T> class Walker; 415 416 // Compile to Prog. See prog.h 417 // Reverse prog expects to be run over text backward. 418 // Construction and execution of prog will 419 // stay within approximately max_mem bytes of memory. 420 // If max_mem <= 0, a reasonable default is used. 421 Prog* CompileToProg(int64_t max_mem); 422 Prog* CompileToReverseProg(int64_t max_mem); 423 424 // Whether to expect this library to find exactly the same answer as PCRE 425 // when running this regexp. Most regexps do mimic PCRE exactly, but a few 426 // obscure cases behave differently. Technically this is more a property 427 // of the Prog than the Regexp, but the computation is much easier to do 428 // on the Regexp. See mimics_pcre.cc for the exact conditions. 429 bool MimicsPCRE(); 430 431 // Benchmarking function. 432 void NullWalk(); 433 434 // Whether every match of this regexp must be anchored and 435 // begin with a non-empty fixed string (perhaps after ASCII 436 // case-folding). If so, returns the prefix and the sub-regexp that 437 // follows it. 438 // Callers should expect *prefix, *foldcase and *suffix to be "zeroed" 439 // regardless of the return value. 440 bool RequiredPrefix(std::string* prefix, bool* foldcase, 441 Regexp** suffix); 442 443 // Whether every match of this regexp must be unanchored and 444 // begin with a non-empty fixed string (perhaps after ASCII 445 // case-folding). If so, returns the prefix. 446 // Callers should expect *prefix and *foldcase to be "zeroed" 447 // regardless of the return value. 448 bool RequiredPrefixForAccel(std::string* prefix, bool* foldcase); 449 450 private: 451 // Constructor allocates vectors as appropriate for operator. 452 explicit Regexp(RegexpOp op, ParseFlags parse_flags); 453 454 // Use Decref() instead of delete to release Regexps. 455 // This is private to catch deletes at compile time. 456 ~Regexp(); 457 void Destroy(); 458 bool QuickDestroy(); 459 460 // Helpers for Parse. Listed here so they can edit Regexps. 461 class ParseState; 462 463 friend class ParseState; 464 friend bool ParseCharClass(StringPiece* s, Regexp** out_re, 465 RegexpStatus* status); 466 467 // Helper for testing [sic]. 468 friend bool RegexpEqualTestingOnly(Regexp*, Regexp*); 469 470 // Computes whether Regexp is already simple. 471 bool ComputeSimple(); 472 473 // Constructor that generates a Star, Plus or Quest, 474 // squashing the pair if sub is also a Star, Plus or Quest. 475 static Regexp* StarPlusOrQuest(RegexpOp op, Regexp* sub, ParseFlags flags); 476 477 // Constructor that generates a concatenation or alternation, 478 // enforcing the limit on the number of subexpressions for 479 // a particular Regexp. 480 static Regexp* ConcatOrAlternate(RegexpOp op, Regexp** subs, int nsubs, 481 ParseFlags flags, bool can_factor); 482 483 // Returns the leading string that re starts with. 484 // The returned Rune* points into a piece of re, 485 // so it must not be used after the caller calls re->Decref(). 486 static Rune* LeadingString(Regexp* re, int* nrune, ParseFlags* flags); 487 488 // Removes the first n leading runes from the beginning of re. 489 // Edits re in place. 490 static void RemoveLeadingString(Regexp* re, int n); 491 492 // Returns the leading regexp in re's top-level concatenation. 493 // The returned Regexp* points at re or a sub-expression of re, 494 // so it must not be used after the caller calls re->Decref(). 495 static Regexp* LeadingRegexp(Regexp* re); 496 497 // Removes LeadingRegexp(re) from re and returns the remainder. 498 // Might edit re in place. 499 static Regexp* RemoveLeadingRegexp(Regexp* re); 500 501 // Simplifies an alternation of literal strings by factoring out 502 // common prefixes. 503 static int FactorAlternation(Regexp** sub, int nsub, ParseFlags flags); 504 friend class FactorAlternationImpl; 505 506 // Is a == b? Only efficient on regexps that have not been through 507 // Simplify yet - the expansion of a kRegexpRepeat will make this 508 // take a long time. Do not call on such regexps, hence private. 509 static bool Equal(Regexp* a, Regexp* b); 510 511 // Allocate space for n sub-regexps. AllocSub(int n)512 void AllocSub(int n) { 513 DCHECK(n >= 0 && static_cast<uint16_t>(n) == n); 514 if (n > 1) 515 submany_ = new Regexp*[n]; 516 nsub_ = static_cast<uint16_t>(n); 517 } 518 519 // Add Rune to LiteralString 520 void AddRuneToString(Rune r); 521 522 // Swaps this with that, in place. 523 void Swap(Regexp *that); 524 525 // Operator. See description of operators above. 526 // uint8_t instead of RegexpOp to control space usage. 527 uint8_t op_; 528 529 // Is this regexp structure already simple 530 // (has it been returned by Simplify)? 531 // uint8_t instead of bool to control space usage. 532 uint8_t simple_; 533 534 // Flags saved from parsing and used during execution. 535 // (Only FoldCase is used.) 536 // uint16_t instead of ParseFlags to control space usage. 537 uint16_t parse_flags_; 538 539 // Reference count. Exists so that SimplifyRegexp can build 540 // regexp structures that are dags rather than trees to avoid 541 // exponential blowup in space requirements. 542 // uint16_t to control space usage. 543 // The standard regexp routines will never generate a 544 // ref greater than the maximum repeat count (kMaxRepeat), 545 // but even so, Incref and Decref consult an overflow map 546 // when ref_ reaches kMaxRef. 547 uint16_t ref_; 548 static const uint16_t kMaxRef = 0xffff; 549 550 // Subexpressions. 551 // uint16_t to control space usage. 552 // Concat and Alternate handle larger numbers of subexpressions 553 // by building concatenation or alternation trees. 554 // Other routines should call Concat or Alternate instead of 555 // filling in sub() by hand. 556 uint16_t nsub_; 557 static const uint16_t kMaxNsub = 0xffff; 558 union { 559 Regexp** submany_; // if nsub_ > 1 560 Regexp* subone_; // if nsub_ == 1 561 }; 562 563 // Extra space for parse and teardown stacks. 564 Regexp* down_; 565 566 // Arguments to operator. See description of operators above. 567 union { 568 struct { // Repeat 569 int max_; 570 int min_; 571 }; 572 struct { // Capture 573 int cap_; 574 std::string* name_; 575 }; 576 struct { // LiteralString 577 int nrunes_; 578 Rune* runes_; 579 }; 580 struct { // CharClass 581 // These two could be in separate union members, 582 // but it wouldn't save any space (there are other two-word structs) 583 // and keeping them separate avoids confusion during parsing. 584 CharClass* cc_; 585 CharClassBuilder* ccb_; 586 }; 587 Rune rune_; // Literal 588 int match_id_; // HaveMatch 589 void *the_union_[2]; // as big as any other element, for memset 590 }; 591 592 Regexp(const Regexp&) = delete; 593 Regexp& operator=(const Regexp&) = delete; 594 }; 595 596 // Character class set: contains non-overlapping, non-abutting RuneRanges. 597 typedef std::set<RuneRange, RuneRangeLess> RuneRangeSet; 598 599 class CharClassBuilder { 600 public: 601 CharClassBuilder(); 602 603 typedef RuneRangeSet::iterator iterator; begin()604 iterator begin() { return ranges_.begin(); } end()605 iterator end() { return ranges_.end(); } 606 size()607 int size() { return nrunes_; } empty()608 bool empty() { return nrunes_ == 0; } full()609 bool full() { return nrunes_ == Runemax+1; } 610 611 bool Contains(Rune r); 612 bool FoldsASCII(); 613 bool AddRange(Rune lo, Rune hi); // returns whether class changed 614 CharClassBuilder* Copy(); 615 void AddCharClass(CharClassBuilder* cc); 616 void Negate(); 617 void RemoveAbove(Rune r); 618 CharClass* GetCharClass(); 619 void AddRangeFlags(Rune lo, Rune hi, Regexp::ParseFlags parse_flags); 620 621 private: 622 static const uint32_t AlphaMask = (1<<26) - 1; 623 uint32_t upper_; // bitmap of A-Z 624 uint32_t lower_; // bitmap of a-z 625 int nrunes_; 626 RuneRangeSet ranges_; 627 628 CharClassBuilder(const CharClassBuilder&) = delete; 629 CharClassBuilder& operator=(const CharClassBuilder&) = delete; 630 }; 631 632 // Bitwise ops on ParseFlags produce ParseFlags. 633 inline Regexp::ParseFlags operator|(Regexp::ParseFlags a, 634 Regexp::ParseFlags b) { 635 return static_cast<Regexp::ParseFlags>( 636 static_cast<int>(a) | static_cast<int>(b)); 637 } 638 639 inline Regexp::ParseFlags operator^(Regexp::ParseFlags a, 640 Regexp::ParseFlags b) { 641 return static_cast<Regexp::ParseFlags>( 642 static_cast<int>(a) ^ static_cast<int>(b)); 643 } 644 645 inline Regexp::ParseFlags operator&(Regexp::ParseFlags a, 646 Regexp::ParseFlags b) { 647 return static_cast<Regexp::ParseFlags>( 648 static_cast<int>(a) & static_cast<int>(b)); 649 } 650 651 inline Regexp::ParseFlags operator~(Regexp::ParseFlags a) { 652 // Attempting to produce a value out of enum's range has undefined behaviour. 653 return static_cast<Regexp::ParseFlags>( 654 ~static_cast<int>(a) & static_cast<int>(Regexp::AllParseFlags)); 655 } 656 657 } // namespace re2 658 659 #endif // RE2_REGEXP_H_ 660