1 //===--- IdentifierTable.h - Hash table for identifier lookup ---*- C++ -*-===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 /// 10 /// \file 11 /// \brief Defines the clang::IdentifierInfo, clang::IdentifierTable, and 12 /// clang::Selector interfaces. 13 /// 14 //===----------------------------------------------------------------------===// 15 16 #ifndef LLVM_CLANG_BASIC_IDENTIFIERTABLE_H 17 #define LLVM_CLANG_BASIC_IDENTIFIERTABLE_H 18 19 #include "clang/Basic/LLVM.h" 20 #include "clang/Basic/TokenKinds.h" 21 #include "llvm/ADT/StringMap.h" 22 #include "llvm/ADT/StringRef.h" 23 #include <cassert> 24 #include <string> 25 26 namespace llvm { 27 template <typename T> struct DenseMapInfo; 28 } 29 30 namespace clang { 31 class LangOptions; 32 class IdentifierInfo; 33 class IdentifierTable; 34 class SourceLocation; 35 class MultiKeywordSelector; // private class used by Selector 36 class DeclarationName; // AST class that stores declaration names 37 38 /// \brief A simple pair of identifier info and location. 39 typedef std::pair<IdentifierInfo*, SourceLocation> IdentifierLocPair; 40 41 42 /// One of these records is kept for each identifier that 43 /// is lexed. This contains information about whether the token was \#define'd, 44 /// is a language keyword, or if it is a front-end token of some sort (e.g. a 45 /// variable or function name). The preprocessor keeps this information in a 46 /// set, and all tok::identifier tokens have a pointer to one of these. 47 class IdentifierInfo { 48 unsigned TokenID : 9; // Front-end token ID or tok::identifier. 49 // Objective-C keyword ('protocol' in '@protocol') or builtin (__builtin_inf). 50 // First NUM_OBJC_KEYWORDS values are for Objective-C, the remaining values 51 // are for builtins. 52 unsigned ObjCOrBuiltinID :13; 53 bool HasMacro : 1; // True if there is a #define for this. 54 bool HadMacro : 1; // True if there was a #define for this. 55 bool IsExtension : 1; // True if identifier is a lang extension. 56 bool IsFutureCompatKeyword : 1; // True if identifier is a keyword in a 57 // newer Standard or proposed Standard. 58 bool IsPoisoned : 1; // True if identifier is poisoned. 59 bool IsCPPOperatorKeyword : 1; // True if ident is a C++ operator keyword. 60 bool NeedsHandleIdentifier : 1; // See "RecomputeNeedsHandleIdentifier". 61 bool IsFromAST : 1; // True if identifier was loaded (at least 62 // partially) from an AST file. 63 bool ChangedAfterLoad : 1; // True if identifier has changed from the 64 // definition loaded from an AST file. 65 bool RevertedTokenID : 1; // True if RevertTokenIDToIdentifier was 66 // called. 67 bool OutOfDate : 1; // True if there may be additional 68 // information about this identifier 69 // stored externally. 70 bool IsModulesImport : 1; // True if this is the 'import' contextual 71 // keyword. 72 // 30 bit left in 64-bit word. 73 74 void *FETokenInfo; // Managed by the language front-end. 75 llvm::StringMapEntry<IdentifierInfo*> *Entry; 76 77 IdentifierInfo(const IdentifierInfo&) = delete; 78 void operator=(const IdentifierInfo&) = delete; 79 80 friend class IdentifierTable; 81 82 public: 83 IdentifierInfo(); 84 85 86 /// \brief Return true if this is the identifier for the specified string. 87 /// 88 /// This is intended to be used for string literals only: II->isStr("foo"). 89 template <std::size_t StrLen> isStr(const char (& Str)[StrLen])90 bool isStr(const char (&Str)[StrLen]) const { 91 return getLength() == StrLen-1 && !memcmp(getNameStart(), Str, StrLen-1); 92 } 93 94 /// \brief Return the beginning of the actual null-terminated string for this 95 /// identifier. 96 /// getNameStart()97 const char *getNameStart() const { 98 if (Entry) return Entry->getKeyData(); 99 // FIXME: This is gross. It would be best not to embed specific details 100 // of the PTH file format here. 101 // The 'this' pointer really points to a 102 // std::pair<IdentifierInfo, const char*>, where internal pointer 103 // points to the external string data. 104 typedef std::pair<IdentifierInfo, const char*> actualtype; 105 return ((const actualtype*) this)->second; 106 } 107 108 /// \brief Efficiently return the length of this identifier info. 109 /// getLength()110 unsigned getLength() const { 111 if (Entry) return Entry->getKeyLength(); 112 // FIXME: This is gross. It would be best not to embed specific details 113 // of the PTH file format here. 114 // The 'this' pointer really points to a 115 // std::pair<IdentifierInfo, const char*>, where internal pointer 116 // points to the external string data. 117 typedef std::pair<IdentifierInfo, const char*> actualtype; 118 const char* p = ((const actualtype*) this)->second - 2; 119 return (((unsigned) p[0]) | (((unsigned) p[1]) << 8)) - 1; 120 } 121 122 /// \brief Return the actual identifier string. getName()123 StringRef getName() const { 124 return StringRef(getNameStart(), getLength()); 125 } 126 127 /// \brief Return true if this identifier is \#defined to some other value. 128 /// \note The current definition may be in a module and not currently visible. hasMacroDefinition()129 bool hasMacroDefinition() const { 130 return HasMacro; 131 } setHasMacroDefinition(bool Val)132 void setHasMacroDefinition(bool Val) { 133 if (HasMacro == Val) return; 134 135 HasMacro = Val; 136 if (Val) { 137 NeedsHandleIdentifier = 1; 138 HadMacro = true; 139 } else { 140 RecomputeNeedsHandleIdentifier(); 141 } 142 } 143 /// \brief Returns true if this identifier was \#defined to some value at any 144 /// moment. In this case there should be an entry for the identifier in the 145 /// macro history table in Preprocessor. hadMacroDefinition()146 bool hadMacroDefinition() const { 147 return HadMacro; 148 } 149 150 /// If this is a source-language token (e.g. 'for'), this API 151 /// can be used to cause the lexer to map identifiers to source-language 152 /// tokens. getTokenID()153 tok::TokenKind getTokenID() const { return (tok::TokenKind)TokenID; } 154 155 /// \brief True if RevertTokenIDToIdentifier() was called. hasRevertedTokenIDToIdentifier()156 bool hasRevertedTokenIDToIdentifier() const { return RevertedTokenID; } 157 158 /// \brief Revert TokenID to tok::identifier; used for GNU libstdc++ 4.2 159 /// compatibility. 160 /// 161 /// TokenID is normally read-only but there are 2 instances where we revert it 162 /// to tok::identifier for libstdc++ 4.2. Keep track of when this happens 163 /// using this method so we can inform serialization about it. RevertTokenIDToIdentifier()164 void RevertTokenIDToIdentifier() { 165 assert(TokenID != tok::identifier && "Already at tok::identifier"); 166 TokenID = tok::identifier; 167 RevertedTokenID = true; 168 } 169 170 /// \brief Return the preprocessor keyword ID for this identifier. 171 /// 172 /// For example, "define" will return tok::pp_define. 173 tok::PPKeywordKind getPPKeywordID() const; 174 175 /// \brief Return the Objective-C keyword ID for the this identifier. 176 /// 177 /// For example, 'class' will return tok::objc_class if ObjC is enabled. getObjCKeywordID()178 tok::ObjCKeywordKind getObjCKeywordID() const { 179 if (ObjCOrBuiltinID < tok::NUM_OBJC_KEYWORDS) 180 return tok::ObjCKeywordKind(ObjCOrBuiltinID); 181 else 182 return tok::objc_not_keyword; 183 } setObjCKeywordID(tok::ObjCKeywordKind ID)184 void setObjCKeywordID(tok::ObjCKeywordKind ID) { ObjCOrBuiltinID = ID; } 185 186 /// \brief Return a value indicating whether this is a builtin function. 187 /// 188 /// 0 is not-built-in. 1 is builtin-for-some-nonprimary-target. 189 /// 2+ are specific builtin functions. getBuiltinID()190 unsigned getBuiltinID() const { 191 if (ObjCOrBuiltinID >= tok::NUM_OBJC_KEYWORDS) 192 return ObjCOrBuiltinID - tok::NUM_OBJC_KEYWORDS; 193 else 194 return 0; 195 } setBuiltinID(unsigned ID)196 void setBuiltinID(unsigned ID) { 197 ObjCOrBuiltinID = ID + tok::NUM_OBJC_KEYWORDS; 198 assert(ObjCOrBuiltinID - unsigned(tok::NUM_OBJC_KEYWORDS) == ID 199 && "ID too large for field!"); 200 } 201 getObjCOrBuiltinID()202 unsigned getObjCOrBuiltinID() const { return ObjCOrBuiltinID; } setObjCOrBuiltinID(unsigned ID)203 void setObjCOrBuiltinID(unsigned ID) { ObjCOrBuiltinID = ID; } 204 205 /// get/setExtension - Initialize information about whether or not this 206 /// language token is an extension. This controls extension warnings, and is 207 /// only valid if a custom token ID is set. isExtensionToken()208 bool isExtensionToken() const { return IsExtension; } setIsExtensionToken(bool Val)209 void setIsExtensionToken(bool Val) { 210 IsExtension = Val; 211 if (Val) 212 NeedsHandleIdentifier = 1; 213 else 214 RecomputeNeedsHandleIdentifier(); 215 } 216 217 /// is/setIsFutureCompatKeyword - Initialize information about whether or not 218 /// this language token is a keyword in a newer or proposed Standard. This 219 /// controls compatibility warnings, and is only true when not parsing the 220 /// corresponding Standard. Once a compatibility problem has been diagnosed 221 /// with this keyword, the flag will be cleared. isFutureCompatKeyword()222 bool isFutureCompatKeyword() const { return IsFutureCompatKeyword; } setIsFutureCompatKeyword(bool Val)223 void setIsFutureCompatKeyword(bool Val) { 224 IsFutureCompatKeyword = Val; 225 if (Val) 226 NeedsHandleIdentifier = 1; 227 else 228 RecomputeNeedsHandleIdentifier(); 229 } 230 231 /// setIsPoisoned - Mark this identifier as poisoned. After poisoning, the 232 /// Preprocessor will emit an error every time this token is used. 233 void setIsPoisoned(bool Value = true) { 234 IsPoisoned = Value; 235 if (Value) 236 NeedsHandleIdentifier = 1; 237 else 238 RecomputeNeedsHandleIdentifier(); 239 } 240 241 /// \brief Return true if this token has been poisoned. isPoisoned()242 bool isPoisoned() const { return IsPoisoned; } 243 244 /// isCPlusPlusOperatorKeyword/setIsCPlusPlusOperatorKeyword controls whether 245 /// this identifier is a C++ alternate representation of an operator. 246 void setIsCPlusPlusOperatorKeyword(bool Val = true) { 247 IsCPPOperatorKeyword = Val; 248 if (Val) 249 NeedsHandleIdentifier = 1; 250 else 251 RecomputeNeedsHandleIdentifier(); 252 } isCPlusPlusOperatorKeyword()253 bool isCPlusPlusOperatorKeyword() const { return IsCPPOperatorKeyword; } 254 255 /// \brief Return true if this token is a keyword in the specified language. 256 bool isKeyword(const LangOptions &LangOpts); 257 258 /// getFETokenInfo/setFETokenInfo - The language front-end is allowed to 259 /// associate arbitrary metadata with this token. 260 template<typename T> getFETokenInfo()261 T *getFETokenInfo() const { return static_cast<T*>(FETokenInfo); } setFETokenInfo(void * T)262 void setFETokenInfo(void *T) { FETokenInfo = T; } 263 264 /// \brief Return true if the Preprocessor::HandleIdentifier must be called 265 /// on a token of this identifier. 266 /// 267 /// If this returns false, we know that HandleIdentifier will not affect 268 /// the token. isHandleIdentifierCase()269 bool isHandleIdentifierCase() const { return NeedsHandleIdentifier; } 270 271 /// \brief Return true if the identifier in its current state was loaded 272 /// from an AST file. isFromAST()273 bool isFromAST() const { return IsFromAST; } 274 setIsFromAST()275 void setIsFromAST() { IsFromAST = true; } 276 277 /// \brief Determine whether this identifier has changed since it was loaded 278 /// from an AST file. hasChangedSinceDeserialization()279 bool hasChangedSinceDeserialization() const { 280 return ChangedAfterLoad; 281 } 282 283 /// \brief Note that this identifier has changed since it was loaded from 284 /// an AST file. setChangedSinceDeserialization()285 void setChangedSinceDeserialization() { 286 ChangedAfterLoad = true; 287 } 288 289 /// \brief Determine whether the information for this identifier is out of 290 /// date with respect to the external source. isOutOfDate()291 bool isOutOfDate() const { return OutOfDate; } 292 293 /// \brief Set whether the information for this identifier is out of 294 /// date with respect to the external source. setOutOfDate(bool OOD)295 void setOutOfDate(bool OOD) { 296 OutOfDate = OOD; 297 if (OOD) 298 NeedsHandleIdentifier = true; 299 else 300 RecomputeNeedsHandleIdentifier(); 301 } 302 303 /// \brief Determine whether this is the contextual keyword \c import. isModulesImport()304 bool isModulesImport() const { return IsModulesImport; } 305 306 /// \brief Set whether this identifier is the contextual keyword \c import. setModulesImport(bool I)307 void setModulesImport(bool I) { 308 IsModulesImport = I; 309 if (I) 310 NeedsHandleIdentifier = true; 311 else 312 RecomputeNeedsHandleIdentifier(); 313 } 314 315 /// \brief Provide less than operator for lexicographical sorting. 316 bool operator<(const IdentifierInfo &RHS) const { 317 return getName() < RHS.getName(); 318 } 319 320 private: 321 /// The Preprocessor::HandleIdentifier does several special (but rare) 322 /// things to identifiers of various sorts. For example, it changes the 323 /// \c for keyword token from tok::identifier to tok::for. 324 /// 325 /// This method is very tied to the definition of HandleIdentifier. Any 326 /// change to it should be reflected here. RecomputeNeedsHandleIdentifier()327 void RecomputeNeedsHandleIdentifier() { 328 NeedsHandleIdentifier = 329 (isPoisoned() | hasMacroDefinition() | isCPlusPlusOperatorKeyword() | 330 isExtensionToken() | isFutureCompatKeyword() || isOutOfDate() || 331 isModulesImport()); 332 } 333 }; 334 335 /// \brief An RAII object for [un]poisoning an identifier within a scope. 336 /// 337 /// \p II is allowed to be null, in which case objects of this type have 338 /// no effect. 339 class PoisonIdentifierRAIIObject { 340 IdentifierInfo *const II; 341 const bool OldValue; 342 public: PoisonIdentifierRAIIObject(IdentifierInfo * II,bool NewValue)343 PoisonIdentifierRAIIObject(IdentifierInfo *II, bool NewValue) 344 : II(II), OldValue(II ? II->isPoisoned() : false) { 345 if(II) 346 II->setIsPoisoned(NewValue); 347 } 348 ~PoisonIdentifierRAIIObject()349 ~PoisonIdentifierRAIIObject() { 350 if(II) 351 II->setIsPoisoned(OldValue); 352 } 353 }; 354 355 /// \brief An iterator that walks over all of the known identifiers 356 /// in the lookup table. 357 /// 358 /// Since this iterator uses an abstract interface via virtual 359 /// functions, it uses an object-oriented interface rather than the 360 /// more standard C++ STL iterator interface. In this OO-style 361 /// iteration, the single function \c Next() provides dereference, 362 /// advance, and end-of-sequence checking in a single 363 /// operation. Subclasses of this iterator type will provide the 364 /// actual functionality. 365 class IdentifierIterator { 366 private: 367 IdentifierIterator(const IdentifierIterator &) = delete; 368 void operator=(const IdentifierIterator &) = delete; 369 370 protected: IdentifierIterator()371 IdentifierIterator() { } 372 373 public: 374 virtual ~IdentifierIterator(); 375 376 /// \brief Retrieve the next string in the identifier table and 377 /// advances the iterator for the following string. 378 /// 379 /// \returns The next string in the identifier table. If there is 380 /// no such string, returns an empty \c StringRef. 381 virtual StringRef Next() = 0; 382 }; 383 384 /// \brief Provides lookups to, and iteration over, IdentiferInfo objects. 385 class IdentifierInfoLookup { 386 public: 387 virtual ~IdentifierInfoLookup(); 388 389 /// \brief Return the IdentifierInfo for the specified named identifier. 390 /// 391 /// Unlike the version in IdentifierTable, this returns a pointer instead 392 /// of a reference. If the pointer is null then the IdentifierInfo cannot 393 /// be found. 394 virtual IdentifierInfo* get(StringRef Name) = 0; 395 396 /// \brief Retrieve an iterator into the set of all identifiers 397 /// known to this identifier lookup source. 398 /// 399 /// This routine provides access to all of the identifiers known to 400 /// the identifier lookup, allowing access to the contents of the 401 /// identifiers without introducing the overhead of constructing 402 /// IdentifierInfo objects for each. 403 /// 404 /// \returns A new iterator into the set of known identifiers. The 405 /// caller is responsible for deleting this iterator. 406 virtual IdentifierIterator *getIdentifiers(); 407 }; 408 409 /// \brief Implements an efficient mapping from strings to IdentifierInfo nodes. 410 /// 411 /// This has no other purpose, but this is an extremely performance-critical 412 /// piece of the code, as each occurrence of every identifier goes through 413 /// here when lexed. 414 class IdentifierTable { 415 // Shark shows that using MallocAllocator is *much* slower than using this 416 // BumpPtrAllocator! 417 typedef llvm::StringMap<IdentifierInfo*, llvm::BumpPtrAllocator> HashTableTy; 418 HashTableTy HashTable; 419 420 IdentifierInfoLookup* ExternalLookup; 421 422 public: 423 /// \brief Create the identifier table, populating it with info about the 424 /// language keywords for the language specified by \p LangOpts. 425 IdentifierTable(const LangOptions &LangOpts, 426 IdentifierInfoLookup* externalLookup = nullptr); 427 428 /// \brief Set the external identifier lookup mechanism. setExternalIdentifierLookup(IdentifierInfoLookup * IILookup)429 void setExternalIdentifierLookup(IdentifierInfoLookup *IILookup) { 430 ExternalLookup = IILookup; 431 } 432 433 /// \brief Retrieve the external identifier lookup object, if any. getExternalIdentifierLookup()434 IdentifierInfoLookup *getExternalIdentifierLookup() const { 435 return ExternalLookup; 436 } 437 getAllocator()438 llvm::BumpPtrAllocator& getAllocator() { 439 return HashTable.getAllocator(); 440 } 441 442 /// \brief Return the identifier token info for the specified named 443 /// identifier. get(StringRef Name)444 IdentifierInfo &get(StringRef Name) { 445 auto &Entry = *HashTable.insert(std::make_pair(Name, nullptr)).first; 446 447 IdentifierInfo *&II = Entry.second; 448 if (II) return *II; 449 450 // No entry; if we have an external lookup, look there first. 451 if (ExternalLookup) { 452 II = ExternalLookup->get(Name); 453 if (II) 454 return *II; 455 } 456 457 // Lookups failed, make a new IdentifierInfo. 458 void *Mem = getAllocator().Allocate<IdentifierInfo>(); 459 II = new (Mem) IdentifierInfo(); 460 461 // Make sure getName() knows how to find the IdentifierInfo 462 // contents. 463 II->Entry = &Entry; 464 465 return *II; 466 } 467 get(StringRef Name,tok::TokenKind TokenCode)468 IdentifierInfo &get(StringRef Name, tok::TokenKind TokenCode) { 469 IdentifierInfo &II = get(Name); 470 II.TokenID = TokenCode; 471 assert(II.TokenID == (unsigned) TokenCode && "TokenCode too large"); 472 return II; 473 } 474 475 /// \brief Gets an IdentifierInfo for the given name without consulting 476 /// external sources. 477 /// 478 /// This is a version of get() meant for external sources that want to 479 /// introduce or modify an identifier. If they called get(), they would 480 /// likely end up in a recursion. getOwn(StringRef Name)481 IdentifierInfo &getOwn(StringRef Name) { 482 auto &Entry = *HashTable.insert(std::make_pair(Name, nullptr)).first; 483 484 IdentifierInfo *&II = Entry.second; 485 if (II) 486 return *II; 487 488 // Lookups failed, make a new IdentifierInfo. 489 void *Mem = getAllocator().Allocate<IdentifierInfo>(); 490 II = new (Mem) IdentifierInfo(); 491 492 // Make sure getName() knows how to find the IdentifierInfo 493 // contents. 494 II->Entry = &Entry; 495 496 // If this is the 'import' contextual keyword, mark it as such. 497 if (Name.equals("import")) 498 II->setModulesImport(true); 499 500 return *II; 501 } 502 503 typedef HashTableTy::const_iterator iterator; 504 typedef HashTableTy::const_iterator const_iterator; 505 begin()506 iterator begin() const { return HashTable.begin(); } end()507 iterator end() const { return HashTable.end(); } size()508 unsigned size() const { return HashTable.size(); } 509 510 /// \brief Print some statistics to stderr that indicate how well the 511 /// hashing is doing. 512 void PrintStats() const; 513 514 void AddKeywords(const LangOptions &LangOpts); 515 }; 516 517 /// \brief A family of Objective-C methods. 518 /// 519 /// These families have no inherent meaning in the language, but are 520 /// nonetheless central enough in the existing implementations to 521 /// merit direct AST support. While, in theory, arbitrary methods can 522 /// be considered to form families, we focus here on the methods 523 /// involving allocation and retain-count management, as these are the 524 /// most "core" and the most likely to be useful to diverse clients 525 /// without extra information. 526 /// 527 /// Both selectors and actual method declarations may be classified 528 /// into families. Method families may impose additional restrictions 529 /// beyond their selector name; for example, a method called '_init' 530 /// that returns void is not considered to be in the 'init' family 531 /// (but would be if it returned 'id'). It is also possible to 532 /// explicitly change or remove a method's family. Therefore the 533 /// method's family should be considered the single source of truth. 534 enum ObjCMethodFamily { 535 /// \brief No particular method family. 536 OMF_None, 537 538 // Selectors in these families may have arbitrary arity, may be 539 // written with arbitrary leading underscores, and may have 540 // additional CamelCase "words" in their first selector chunk 541 // following the family name. 542 OMF_alloc, 543 OMF_copy, 544 OMF_init, 545 OMF_mutableCopy, 546 OMF_new, 547 548 // These families are singletons consisting only of the nullary 549 // selector with the given name. 550 OMF_autorelease, 551 OMF_dealloc, 552 OMF_finalize, 553 OMF_release, 554 OMF_retain, 555 OMF_retainCount, 556 OMF_self, 557 OMF_initialize, 558 559 // performSelector families 560 OMF_performSelector 561 }; 562 563 /// Enough bits to store any enumerator in ObjCMethodFamily or 564 /// InvalidObjCMethodFamily. 565 enum { ObjCMethodFamilyBitWidth = 4 }; 566 567 /// \brief An invalid value of ObjCMethodFamily. 568 enum { InvalidObjCMethodFamily = (1 << ObjCMethodFamilyBitWidth) - 1 }; 569 570 /// \brief A family of Objective-C methods. 571 /// 572 /// These are family of methods whose result type is initially 'id', but 573 /// but are candidate for the result type to be changed to 'instancetype'. 574 enum ObjCInstanceTypeFamily { 575 OIT_None, 576 OIT_Array, 577 OIT_Dictionary, 578 OIT_Singleton, 579 OIT_Init, 580 OIT_ReturnsSelf 581 }; 582 583 enum ObjCStringFormatFamily { 584 SFF_None, 585 SFF_NSString, 586 SFF_CFString 587 }; 588 589 /// \brief Smart pointer class that efficiently represents Objective-C method 590 /// names. 591 /// 592 /// This class will either point to an IdentifierInfo or a 593 /// MultiKeywordSelector (which is private). This enables us to optimize 594 /// selectors that take no arguments and selectors that take 1 argument, which 595 /// accounts for 78% of all selectors in Cocoa.h. 596 class Selector { 597 friend class Diagnostic; 598 599 enum IdentifierInfoFlag { 600 // Empty selector = 0. 601 ZeroArg = 0x1, 602 OneArg = 0x2, 603 MultiArg = 0x3, 604 ArgFlags = ZeroArg|OneArg 605 }; 606 uintptr_t InfoPtr; // a pointer to the MultiKeywordSelector or IdentifierInfo. 607 Selector(IdentifierInfo * II,unsigned nArgs)608 Selector(IdentifierInfo *II, unsigned nArgs) { 609 InfoPtr = reinterpret_cast<uintptr_t>(II); 610 assert((InfoPtr & ArgFlags) == 0 &&"Insufficiently aligned IdentifierInfo"); 611 assert(nArgs < 2 && "nArgs not equal to 0/1"); 612 InfoPtr |= nArgs+1; 613 } Selector(MultiKeywordSelector * SI)614 Selector(MultiKeywordSelector *SI) { 615 InfoPtr = reinterpret_cast<uintptr_t>(SI); 616 assert((InfoPtr & ArgFlags) == 0 &&"Insufficiently aligned IdentifierInfo"); 617 InfoPtr |= MultiArg; 618 } 619 getAsIdentifierInfo()620 IdentifierInfo *getAsIdentifierInfo() const { 621 if (getIdentifierInfoFlag() < MultiArg) 622 return reinterpret_cast<IdentifierInfo *>(InfoPtr & ~ArgFlags); 623 return nullptr; 624 } getMultiKeywordSelector()625 MultiKeywordSelector *getMultiKeywordSelector() const { 626 return reinterpret_cast<MultiKeywordSelector *>(InfoPtr & ~ArgFlags); 627 } 628 getIdentifierInfoFlag()629 unsigned getIdentifierInfoFlag() const { 630 return InfoPtr & ArgFlags; 631 } 632 633 static ObjCMethodFamily getMethodFamilyImpl(Selector sel); 634 635 static ObjCStringFormatFamily getStringFormatFamilyImpl(Selector sel); 636 637 public: 638 friend class SelectorTable; // only the SelectorTable can create these 639 friend class DeclarationName; // and the AST's DeclarationName. 640 641 /// The default ctor should only be used when creating data structures that 642 /// will contain selectors. Selector()643 Selector() : InfoPtr(0) {} Selector(uintptr_t V)644 Selector(uintptr_t V) : InfoPtr(V) {} 645 646 /// operator==/!= - Indicate whether the specified selectors are identical. 647 bool operator==(Selector RHS) const { 648 return InfoPtr == RHS.InfoPtr; 649 } 650 bool operator!=(Selector RHS) const { 651 return InfoPtr != RHS.InfoPtr; 652 } getAsOpaquePtr()653 void *getAsOpaquePtr() const { 654 return reinterpret_cast<void*>(InfoPtr); 655 } 656 657 /// \brief Determine whether this is the empty selector. isNull()658 bool isNull() const { return InfoPtr == 0; } 659 660 // Predicates to identify the selector type. isKeywordSelector()661 bool isKeywordSelector() const { 662 return getIdentifierInfoFlag() != ZeroArg; 663 } isUnarySelector()664 bool isUnarySelector() const { 665 return getIdentifierInfoFlag() == ZeroArg; 666 } 667 unsigned getNumArgs() const; 668 669 670 /// \brief Retrieve the identifier at a given position in the selector. 671 /// 672 /// Note that the identifier pointer returned may be NULL. Clients that only 673 /// care about the text of the identifier string, and not the specific, 674 /// uniqued identifier pointer, should use \c getNameForSlot(), which returns 675 /// an empty string when the identifier pointer would be NULL. 676 /// 677 /// \param argIndex The index for which we want to retrieve the identifier. 678 /// This index shall be less than \c getNumArgs() unless this is a keyword 679 /// selector, in which case 0 is the only permissible value. 680 /// 681 /// \returns the uniqued identifier for this slot, or NULL if this slot has 682 /// no corresponding identifier. 683 IdentifierInfo *getIdentifierInfoForSlot(unsigned argIndex) const; 684 685 /// \brief Retrieve the name at a given position in the selector. 686 /// 687 /// \param argIndex The index for which we want to retrieve the name. 688 /// This index shall be less than \c getNumArgs() unless this is a keyword 689 /// selector, in which case 0 is the only permissible value. 690 /// 691 /// \returns the name for this slot, which may be the empty string if no 692 /// name was supplied. 693 StringRef getNameForSlot(unsigned argIndex) const; 694 695 /// \brief Derive the full selector name (e.g. "foo:bar:") and return 696 /// it as an std::string. 697 std::string getAsString() const; 698 699 /// \brief Prints the full selector name (e.g. "foo:bar:"). 700 void print(llvm::raw_ostream &OS) const; 701 702 /// \brief Derive the conventional family of this method. getMethodFamily()703 ObjCMethodFamily getMethodFamily() const { 704 return getMethodFamilyImpl(*this); 705 } 706 getStringFormatFamily()707 ObjCStringFormatFamily getStringFormatFamily() const { 708 return getStringFormatFamilyImpl(*this); 709 } 710 getEmptyMarker()711 static Selector getEmptyMarker() { 712 return Selector(uintptr_t(-1)); 713 } getTombstoneMarker()714 static Selector getTombstoneMarker() { 715 return Selector(uintptr_t(-2)); 716 } 717 718 static ObjCInstanceTypeFamily getInstTypeMethodFamily(Selector sel); 719 }; 720 721 /// \brief This table allows us to fully hide how we implement 722 /// multi-keyword caching. 723 class SelectorTable { 724 void *Impl; // Actually a SelectorTableImpl 725 SelectorTable(const SelectorTable &) = delete; 726 void operator=(const SelectorTable &) = delete; 727 public: 728 SelectorTable(); 729 ~SelectorTable(); 730 731 /// \brief Can create any sort of selector. 732 /// 733 /// \p NumArgs indicates whether this is a no argument selector "foo", a 734 /// single argument selector "foo:" or multi-argument "foo:bar:". 735 Selector getSelector(unsigned NumArgs, IdentifierInfo **IIV); 736 getUnarySelector(IdentifierInfo * ID)737 Selector getUnarySelector(IdentifierInfo *ID) { 738 return Selector(ID, 1); 739 } getNullarySelector(IdentifierInfo * ID)740 Selector getNullarySelector(IdentifierInfo *ID) { 741 return Selector(ID, 0); 742 } 743 744 /// \brief Return the total amount of memory allocated for managing selectors. 745 size_t getTotalMemory() const; 746 747 /// \brief Return the default setter name for the given identifier. 748 /// 749 /// This is "set" + \p Name where the initial character of \p Name 750 /// has been capitalized. 751 static SmallString<64> constructSetterName(StringRef Name); 752 753 /// \brief Return the default setter selector for the given identifier. 754 /// 755 /// This is "set" + \p Name where the initial character of \p Name 756 /// has been capitalized. 757 static Selector constructSetterSelector(IdentifierTable &Idents, 758 SelectorTable &SelTable, 759 const IdentifierInfo *Name); 760 }; 761 762 /// DeclarationNameExtra - Common base of the MultiKeywordSelector, 763 /// CXXSpecialName, and CXXOperatorIdName classes, all of which are 764 /// private classes that describe different kinds of names. 765 class DeclarationNameExtra { 766 public: 767 /// ExtraKind - The kind of "extra" information stored in the 768 /// DeclarationName. See @c ExtraKindOrNumArgs for an explanation of 769 /// how these enumerator values are used. 770 enum ExtraKind { 771 CXXConstructor = 0, 772 CXXDestructor, 773 CXXConversionFunction, 774 #define OVERLOADED_OPERATOR(Name,Spelling,Token,Unary,Binary,MemberOnly) \ 775 CXXOperator##Name, 776 #include "clang/Basic/OperatorKinds.def" 777 CXXLiteralOperator, 778 CXXUsingDirective, 779 NUM_EXTRA_KINDS 780 }; 781 782 /// ExtraKindOrNumArgs - Either the kind of C++ special name or 783 /// operator-id (if the value is one of the CXX* enumerators of 784 /// ExtraKind), in which case the DeclarationNameExtra is also a 785 /// CXXSpecialName, (for CXXConstructor, CXXDestructor, or 786 /// CXXConversionFunction) CXXOperatorIdName, or CXXLiteralOperatorName, 787 /// it may be also name common to C++ using-directives (CXXUsingDirective), 788 /// otherwise it is NUM_EXTRA_KINDS+NumArgs, where NumArgs is the number of 789 /// arguments in the Objective-C selector, in which case the 790 /// DeclarationNameExtra is also a MultiKeywordSelector. 791 unsigned ExtraKindOrNumArgs; 792 }; 793 794 } // end namespace clang 795 796 namespace llvm { 797 /// Define DenseMapInfo so that Selectors can be used as keys in DenseMap and 798 /// DenseSets. 799 template <> 800 struct DenseMapInfo<clang::Selector> { 801 static inline clang::Selector getEmptyKey() { 802 return clang::Selector::getEmptyMarker(); 803 } 804 static inline clang::Selector getTombstoneKey() { 805 return clang::Selector::getTombstoneMarker(); 806 } 807 808 static unsigned getHashValue(clang::Selector S); 809 810 static bool isEqual(clang::Selector LHS, clang::Selector RHS) { 811 return LHS == RHS; 812 } 813 }; 814 815 template <> 816 struct isPodLike<clang::Selector> { static const bool value = true; }; 817 818 template <typename T> class PointerLikeTypeTraits; 819 820 template<> 821 class PointerLikeTypeTraits<clang::Selector> { 822 public: 823 static inline const void *getAsVoidPointer(clang::Selector P) { 824 return P.getAsOpaquePtr(); 825 } 826 static inline clang::Selector getFromVoidPointer(const void *P) { 827 return clang::Selector(reinterpret_cast<uintptr_t>(P)); 828 } 829 enum { NumLowBitsAvailable = 0 }; 830 }; 831 832 // Provide PointerLikeTypeTraits for IdentifierInfo pointers, which 833 // are not guaranteed to be 8-byte aligned. 834 template<> 835 class PointerLikeTypeTraits<clang::IdentifierInfo*> { 836 public: 837 static inline void *getAsVoidPointer(clang::IdentifierInfo* P) { 838 return P; 839 } 840 static inline clang::IdentifierInfo *getFromVoidPointer(void *P) { 841 return static_cast<clang::IdentifierInfo*>(P); 842 } 843 enum { NumLowBitsAvailable = 1 }; 844 }; 845 846 template<> 847 class PointerLikeTypeTraits<const clang::IdentifierInfo*> { 848 public: 849 static inline const void *getAsVoidPointer(const clang::IdentifierInfo* P) { 850 return P; 851 } 852 static inline const clang::IdentifierInfo *getFromVoidPointer(const void *P) { 853 return static_cast<const clang::IdentifierInfo*>(P); 854 } 855 enum { NumLowBitsAvailable = 1 }; 856 }; 857 858 } // end namespace llvm 859 #endif 860