Preprocessor.h revision cae5095c116c29a4e52c91fa0ad88df09dd2b45f
1//===--- Preprocessor.h - C Language Family Preprocessor --------*- 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// This file defines the Preprocessor interface. 11// 12//===----------------------------------------------------------------------===// 13 14#ifndef LLVM_CLANG_LEX_PREPROCESSOR_H 15#define LLVM_CLANG_LEX_PREPROCESSOR_H 16 17#include "clang/Lex/MacroInfo.h" 18#include "clang/Lex/Lexer.h" 19#include "clang/Lex/PTHLexer.h" 20#include "clang/Lex/PPCallbacks.h" 21#include "clang/Lex/TokenLexer.h" 22#include "clang/Lex/PTHManager.h" 23#include "clang/Basic/Builtins.h" 24#include "clang/Basic/Diagnostic.h" 25#include "clang/Basic/IdentifierTable.h" 26#include "clang/Basic/SourceLocation.h" 27#include "llvm/ADT/DenseMap.h" 28#include "llvm/ADT/OwningPtr.h" 29#include "llvm/ADT/SmallVector.h" 30#include "llvm/Support/Allocator.h" 31#include <vector> 32 33namespace clang { 34 35class SourceManager; 36class ExternalPreprocessorSource; 37class FileManager; 38class FileEntry; 39class HeaderSearch; 40class PragmaNamespace; 41class PragmaHandler; 42class CommentHandler; 43class ScratchBuffer; 44class TargetInfo; 45class PPCallbacks; 46class CodeCompletionHandler; 47class DirectoryLookup; 48class PreprocessingRecord; 49 50/// Preprocessor - This object engages in a tight little dance with the lexer to 51/// efficiently preprocess tokens. Lexers know only about tokens within a 52/// single source file, and don't know anything about preprocessor-level issues 53/// like the #include stack, token expansion, etc. 54/// 55class Preprocessor { 56 Diagnostic *Diags; 57 LangOptions Features; 58 const TargetInfo &Target; 59 FileManager &FileMgr; 60 SourceManager &SourceMgr; 61 ScratchBuffer *ScratchBuf; 62 HeaderSearch &HeaderInfo; 63 64 /// \brief External source of macros. 65 ExternalPreprocessorSource *ExternalSource; 66 67 /// PTH - An optional PTHManager object used for getting tokens from 68 /// a token cache rather than lexing the original source file. 69 llvm::OwningPtr<PTHManager> PTH; 70 71 /// BP - A BumpPtrAllocator object used to quickly allocate and release 72 /// objects internal to the Preprocessor. 73 llvm::BumpPtrAllocator BP; 74 75 /// Identifiers for builtin macros and other builtins. 76 IdentifierInfo *Ident__LINE__, *Ident__FILE__; // __LINE__, __FILE__ 77 IdentifierInfo *Ident__DATE__, *Ident__TIME__; // __DATE__, __TIME__ 78 IdentifierInfo *Ident__INCLUDE_LEVEL__; // __INCLUDE_LEVEL__ 79 IdentifierInfo *Ident__BASE_FILE__; // __BASE_FILE__ 80 IdentifierInfo *Ident__TIMESTAMP__; // __TIMESTAMP__ 81 IdentifierInfo *Ident__COUNTER__; // __COUNTER__ 82 IdentifierInfo *Ident_Pragma, *Ident__pragma; // _Pragma, __pragma 83 IdentifierInfo *Ident__VA_ARGS__; // __VA_ARGS__ 84 IdentifierInfo *Ident__has_feature; // __has_feature 85 IdentifierInfo *Ident__has_builtin; // __has_builtin 86 IdentifierInfo *Ident__has_attribute; // __has_attribute 87 IdentifierInfo *Ident__has_include; // __has_include 88 IdentifierInfo *Ident__has_include_next; // __has_include_next 89 90 SourceLocation DATELoc, TIMELoc; 91 unsigned CounterValue; // Next __COUNTER__ value. 92 93 enum { 94 /// MaxIncludeStackDepth - Maximum depth of #includes. 95 MaxAllowedIncludeStackDepth = 200 96 }; 97 98 // State that is set before the preprocessor begins. 99 bool KeepComments : 1; 100 bool KeepMacroComments : 1; 101 102 // State that changes while the preprocessor runs: 103 bool InMacroArgs : 1; // True if parsing fn macro invocation args. 104 105 /// Whether the preprocessor owns the header search object. 106 bool OwnsHeaderSearch : 1; 107 108 /// DisableMacroExpansion - True if macro expansion is disabled. 109 bool DisableMacroExpansion : 1; 110 111 /// \brief Whether we have already loaded macros from the external source. 112 mutable bool ReadMacrosFromExternalSource : 1; 113 114 /// Identifiers - This is mapping/lookup information for all identifiers in 115 /// the program, including program keywords. 116 mutable IdentifierTable Identifiers; 117 118 /// Selectors - This table contains all the selectors in the program. Unlike 119 /// IdentifierTable above, this table *isn't* populated by the preprocessor. 120 /// It is declared/instantiated here because it's role/lifetime is 121 /// conceptually similar the IdentifierTable. In addition, the current control 122 /// flow (in clang::ParseAST()), make it convenient to put here. 123 /// FIXME: Make sure the lifetime of Identifiers/Selectors *isn't* tied to 124 /// the lifetime of the preprocessor. 125 SelectorTable Selectors; 126 127 /// BuiltinInfo - Information about builtins. 128 Builtin::Context BuiltinInfo; 129 130 /// PragmaHandlers - This tracks all of the pragmas that the client registered 131 /// with this preprocessor. 132 PragmaNamespace *PragmaHandlers; 133 134 /// \brief Tracks all of the comment handlers that the client registered 135 /// with this preprocessor. 136 std::vector<CommentHandler *> CommentHandlers; 137 138 /// \brief The code-completion handler. 139 CodeCompletionHandler *CodeComplete; 140 141 /// \brief The file that we're performing code-completion for, if any. 142 const FileEntry *CodeCompletionFile; 143 144 /// \brief The number of bytes that we will initially skip when entering the 145 /// main file, which is used when loading a precompiled preamble, along 146 /// with a flag that indicates whether skipping this number of bytes will 147 /// place the lexer at the start of a line. 148 std::pair<unsigned, bool> SkipMainFilePreamble; 149 150 /// CurLexer - This is the current top of the stack that we're lexing from if 151 /// not expanding a macro and we are lexing directly from source code. 152 /// Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null. 153 llvm::OwningPtr<Lexer> CurLexer; 154 155 /// CurPTHLexer - This is the current top of stack that we're lexing from if 156 /// not expanding from a macro and we are lexing from a PTH cache. 157 /// Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null. 158 llvm::OwningPtr<PTHLexer> CurPTHLexer; 159 160 /// CurPPLexer - This is the current top of the stack what we're lexing from 161 /// if not expanding a macro. This is an alias for either CurLexer or 162 /// CurPTHLexer. 163 PreprocessorLexer *CurPPLexer; 164 165 /// CurLookup - The DirectoryLookup structure used to find the current 166 /// FileEntry, if CurLexer is non-null and if applicable. This allows us to 167 /// implement #include_next and find directory-specific properties. 168 const DirectoryLookup *CurDirLookup; 169 170 /// CurTokenLexer - This is the current macro we are expanding, if we are 171 /// expanding a macro. One of CurLexer and CurTokenLexer must be null. 172 llvm::OwningPtr<TokenLexer> CurTokenLexer; 173 174 /// IncludeMacroStack - This keeps track of the stack of files currently 175 /// #included, and macros currently being expanded from, not counting 176 /// CurLexer/CurTokenLexer. 177 struct IncludeStackInfo { 178 Lexer *TheLexer; 179 PTHLexer *ThePTHLexer; 180 PreprocessorLexer *ThePPLexer; 181 TokenLexer *TheTokenLexer; 182 const DirectoryLookup *TheDirLookup; 183 184 IncludeStackInfo(Lexer *L, PTHLexer* P, PreprocessorLexer* PPL, 185 TokenLexer* TL, const DirectoryLookup *D) 186 : TheLexer(L), ThePTHLexer(P), ThePPLexer(PPL), TheTokenLexer(TL), 187 TheDirLookup(D) {} 188 }; 189 std::vector<IncludeStackInfo> IncludeMacroStack; 190 191 /// Callbacks - These are actions invoked when some preprocessor activity is 192 /// encountered (e.g. a file is #included, etc). 193 PPCallbacks *Callbacks; 194 195 /// Macros - For each IdentifierInfo with 'HasMacro' set, we keep a mapping 196 /// to the actual definition of the macro. 197 llvm::DenseMap<IdentifierInfo*, MacroInfo*> Macros; 198 199 /// MacroArgCache - This is a "freelist" of MacroArg objects that can be 200 /// reused for quick allocation. 201 MacroArgs *MacroArgCache; 202 friend class MacroArgs; 203 204 /// PragmaPushMacroInfo - For each IdentifierInfo used in a #pragma 205 /// push_macro directive, we keep a MacroInfo stack used to restore 206 /// previous macro value. 207 llvm::DenseMap<IdentifierInfo*, std::vector<MacroInfo*> > PragmaPushMacroInfo; 208 209 // Various statistics we track for performance analysis. 210 unsigned NumDirectives, NumIncluded, NumDefined, NumUndefined, NumPragma; 211 unsigned NumIf, NumElse, NumEndif; 212 unsigned NumEnteredSourceFiles, MaxIncludeStackDepth; 213 unsigned NumMacroExpanded, NumFnMacroExpanded, NumBuiltinMacroExpanded; 214 unsigned NumFastMacroExpanded, NumTokenPaste, NumFastTokenPaste; 215 unsigned NumSkipped; 216 217 /// Predefines - This string is the predefined macros that preprocessor 218 /// should use from the command line etc. 219 std::string Predefines; 220 221 /// TokenLexerCache - Cache macro expanders to reduce malloc traffic. 222 enum { TokenLexerCacheSize = 8 }; 223 unsigned NumCachedTokenLexers; 224 TokenLexer *TokenLexerCache[TokenLexerCacheSize]; 225 226 /// \brief A record of the macro definitions and instantiations that 227 /// occurred during preprocessing. 228 /// 229 /// This is an optional side structure that can be enabled with 230 /// \c createPreprocessingRecord() prior to preprocessing. 231 PreprocessingRecord *Record; 232 233private: // Cached tokens state. 234 typedef llvm::SmallVector<Token, 1> CachedTokensTy; 235 236 /// CachedTokens - Cached tokens are stored here when we do backtracking or 237 /// lookahead. They are "lexed" by the CachingLex() method. 238 CachedTokensTy CachedTokens; 239 240 /// CachedLexPos - The position of the cached token that CachingLex() should 241 /// "lex" next. If it points beyond the CachedTokens vector, it means that 242 /// a normal Lex() should be invoked. 243 CachedTokensTy::size_type CachedLexPos; 244 245 /// BacktrackPositions - Stack of backtrack positions, allowing nested 246 /// backtracks. The EnableBacktrackAtThisPos() method pushes a position to 247 /// indicate where CachedLexPos should be set when the BackTrack() method is 248 /// invoked (at which point the last position is popped). 249 std::vector<CachedTokensTy::size_type> BacktrackPositions; 250 251 struct MacroInfoChain { 252 MacroInfo MI; 253 MacroInfoChain *Next; 254 MacroInfoChain *Prev; 255 }; 256 257 /// MacroInfos are managed as a chain for easy disposal. This is the head 258 /// of that list. 259 MacroInfoChain *MIChainHead; 260 261 /// MICache - A "freelist" of MacroInfo objects that can be reused for quick 262 /// allocation. 263 MacroInfoChain *MICache; 264 265public: 266 Preprocessor(Diagnostic &diags, const LangOptions &opts, 267 const TargetInfo &target, 268 SourceManager &SM, HeaderSearch &Headers, 269 IdentifierInfoLookup *IILookup = 0, 270 bool OwnsHeaderSearch = false); 271 272 ~Preprocessor(); 273 274 Diagnostic &getDiagnostics() const { return *Diags; } 275 void setDiagnostics(Diagnostic &D) { Diags = &D; } 276 277 const LangOptions &getLangOptions() const { return Features; } 278 const TargetInfo &getTargetInfo() const { return Target; } 279 FileManager &getFileManager() const { return FileMgr; } 280 SourceManager &getSourceManager() const { return SourceMgr; } 281 HeaderSearch &getHeaderSearchInfo() const { return HeaderInfo; } 282 283 IdentifierTable &getIdentifierTable() { return Identifiers; } 284 SelectorTable &getSelectorTable() { return Selectors; } 285 Builtin::Context &getBuiltinInfo() { return BuiltinInfo; } 286 llvm::BumpPtrAllocator &getPreprocessorAllocator() { return BP; } 287 288 void setPTHManager(PTHManager* pm); 289 290 PTHManager *getPTHManager() { return PTH.get(); } 291 292 void setExternalSource(ExternalPreprocessorSource *Source) { 293 ExternalSource = Source; 294 } 295 296 ExternalPreprocessorSource *getExternalSource() const { 297 return ExternalSource; 298 } 299 300 /// SetCommentRetentionState - Control whether or not the preprocessor retains 301 /// comments in output. 302 void SetCommentRetentionState(bool KeepComments, bool KeepMacroComments) { 303 this->KeepComments = KeepComments | KeepMacroComments; 304 this->KeepMacroComments = KeepMacroComments; 305 } 306 307 bool getCommentRetentionState() const { return KeepComments; } 308 309 /// isCurrentLexer - Return true if we are lexing directly from the specified 310 /// lexer. 311 bool isCurrentLexer(const PreprocessorLexer *L) const { 312 return CurPPLexer == L; 313 } 314 315 /// getCurrentLexer - Return the current lexer being lexed from. Note 316 /// that this ignores any potentially active macro expansions and _Pragma 317 /// expansions going on at the time. 318 PreprocessorLexer *getCurrentLexer() const { return CurPPLexer; } 319 320 /// getCurrentFileLexer - Return the current file lexer being lexed from. 321 /// Note that this ignores any potentially active macro expansions and _Pragma 322 /// expansions going on at the time. 323 PreprocessorLexer *getCurrentFileLexer() const; 324 325 /// getPPCallbacks/addPPCallbacks - Accessors for preprocessor callbacks. 326 /// Note that this class takes ownership of any PPCallbacks object given to 327 /// it. 328 PPCallbacks *getPPCallbacks() const { return Callbacks; } 329 void addPPCallbacks(PPCallbacks *C) { 330 if (Callbacks) 331 C = new PPChainedCallbacks(C, Callbacks); 332 Callbacks = C; 333 } 334 335 /// getMacroInfo - Given an identifier, return the MacroInfo it is #defined to 336 /// or null if it isn't #define'd. 337 MacroInfo *getMacroInfo(IdentifierInfo *II) const { 338 return II->hasMacroDefinition() ? Macros.find(II)->second : 0; 339 } 340 341 /// setMacroInfo - Specify a macro for this identifier. 342 /// 343 void setMacroInfo(IdentifierInfo *II, MacroInfo *MI); 344 345 /// macro_iterator/macro_begin/macro_end - This allows you to walk the current 346 /// state of the macro table. This visits every currently-defined macro. 347 typedef llvm::DenseMap<IdentifierInfo*, 348 MacroInfo*>::const_iterator macro_iterator; 349 macro_iterator macro_begin(bool IncludeExternalMacros = true) const; 350 macro_iterator macro_end(bool IncludeExternalMacros = true) const; 351 352 const std::string &getPredefines() const { return Predefines; } 353 /// setPredefines - Set the predefines for this Preprocessor. These 354 /// predefines are automatically injected when parsing the main file. 355 void setPredefines(const char *P) { Predefines = P; } 356 void setPredefines(const std::string &P) { Predefines = P; } 357 358 /// getIdentifierInfo - Return information about the specified preprocessor 359 /// identifier token. The version of this method that takes two character 360 /// pointers is preferred unless the identifier is already available as a 361 /// string (this avoids allocation and copying of memory to construct an 362 /// std::string). 363 IdentifierInfo *getIdentifierInfo(llvm::StringRef Name) const { 364 return &Identifiers.get(Name); 365 } 366 367 /// AddPragmaHandler - Add the specified pragma handler to the preprocessor. 368 /// If 'Namespace' is non-null, then it is a token required to exist on the 369 /// pragma line before the pragma string starts, e.g. "STDC" or "GCC". 370 void AddPragmaHandler(llvm::StringRef Namespace, PragmaHandler *Handler); 371 void AddPragmaHandler(PragmaHandler *Handler) { 372 AddPragmaHandler(llvm::StringRef(), Handler); 373 } 374 375 /// RemovePragmaHandler - Remove the specific pragma handler from 376 /// the preprocessor. If \arg Namespace is non-null, then it should 377 /// be the namespace that \arg Handler was added to. It is an error 378 /// to remove a handler that has not been registered. 379 void RemovePragmaHandler(llvm::StringRef Namespace, PragmaHandler *Handler); 380 void RemovePragmaHandler(PragmaHandler *Handler) { 381 RemovePragmaHandler(llvm::StringRef(), Handler); 382 } 383 384 /// \brief Add the specified comment handler to the preprocessor. 385 void AddCommentHandler(CommentHandler *Handler); 386 387 /// \brief Remove the specified comment handler. 388 /// 389 /// It is an error to remove a handler that has not been registered. 390 void RemoveCommentHandler(CommentHandler *Handler); 391 392 /// \brief Set the code completion handler to the given object. 393 void setCodeCompletionHandler(CodeCompletionHandler &Handler) { 394 CodeComplete = &Handler; 395 } 396 397 /// \brief Retrieve the current code-completion handler. 398 CodeCompletionHandler *getCodeCompletionHandler() const { 399 return CodeComplete; 400 } 401 402 /// \brief Clear out the code completion handler. 403 void clearCodeCompletionHandler() { 404 CodeComplete = 0; 405 } 406 407 /// \brief Hook used by the lexer to invoke the "natural language" code 408 /// completion point. 409 void CodeCompleteNaturalLanguage(); 410 411 /// \brief Retrieve the preprocessing record, or NULL if there is no 412 /// preprocessing record. 413 PreprocessingRecord *getPreprocessingRecord() const { return Record; } 414 415 /// \brief Create a new preprocessing record, which will keep track of 416 /// all macro expansions, macro definitions, etc. 417 void createPreprocessingRecord(); 418 419 /// EnterMainSourceFile - Enter the specified FileID as the main source file, 420 /// which implicitly adds the builtin defines etc. 421 void EnterMainSourceFile(); 422 423 /// EndSourceFile - Inform the preprocessor callbacks that processing is 424 /// complete. 425 void EndSourceFile(); 426 427 /// EnterSourceFile - Add a source file to the top of the include stack and 428 /// start lexing tokens from it instead of the current buffer. Emit an error 429 /// and don't enter the file on error. 430 void EnterSourceFile(FileID CurFileID, const DirectoryLookup *Dir, 431 SourceLocation Loc); 432 433 /// EnterMacro - Add a Macro to the top of the include stack and start lexing 434 /// tokens from it instead of the current buffer. Args specifies the 435 /// tokens input to a function-like macro. 436 /// 437 /// ILEnd specifies the location of the ')' for a function-like macro or the 438 /// identifier for an object-like macro. 439 void EnterMacro(Token &Identifier, SourceLocation ILEnd, MacroArgs *Args); 440 441 /// EnterTokenStream - Add a "macro" context to the top of the include stack, 442 /// which will cause the lexer to start returning the specified tokens. 443 /// 444 /// If DisableMacroExpansion is true, tokens lexed from the token stream will 445 /// not be subject to further macro expansion. Otherwise, these tokens will 446 /// be re-macro-expanded when/if expansion is enabled. 447 /// 448 /// If OwnsTokens is false, this method assumes that the specified stream of 449 /// tokens has a permanent owner somewhere, so they do not need to be copied. 450 /// If it is true, it assumes the array of tokens is allocated with new[] and 451 /// must be freed. 452 /// 453 void EnterTokenStream(const Token *Toks, unsigned NumToks, 454 bool DisableMacroExpansion, bool OwnsTokens); 455 456 /// RemoveTopOfLexerStack - Pop the current lexer/macro exp off the top of the 457 /// lexer stack. This should only be used in situations where the current 458 /// state of the top-of-stack lexer is known. 459 void RemoveTopOfLexerStack(); 460 461 /// EnableBacktrackAtThisPos - From the point that this method is called, and 462 /// until CommitBacktrackedTokens() or Backtrack() is called, the Preprocessor 463 /// keeps track of the lexed tokens so that a subsequent Backtrack() call will 464 /// make the Preprocessor re-lex the same tokens. 465 /// 466 /// Nested backtracks are allowed, meaning that EnableBacktrackAtThisPos can 467 /// be called multiple times and CommitBacktrackedTokens/Backtrack calls will 468 /// be combined with the EnableBacktrackAtThisPos calls in reverse order. 469 /// 470 /// NOTE: *DO NOT* forget to call either CommitBacktrackedTokens or Backtrack 471 /// at some point after EnableBacktrackAtThisPos. If you don't, caching of 472 /// tokens will continue indefinitely. 473 /// 474 void EnableBacktrackAtThisPos(); 475 476 /// CommitBacktrackedTokens - Disable the last EnableBacktrackAtThisPos call. 477 void CommitBacktrackedTokens(); 478 479 /// Backtrack - Make Preprocessor re-lex the tokens that were lexed since 480 /// EnableBacktrackAtThisPos() was previously called. 481 void Backtrack(); 482 483 /// isBacktrackEnabled - True if EnableBacktrackAtThisPos() was called and 484 /// caching of tokens is on. 485 bool isBacktrackEnabled() const { return !BacktrackPositions.empty(); } 486 487 /// Lex - To lex a token from the preprocessor, just pull a token from the 488 /// current lexer or macro object. 489 void Lex(Token &Result) { 490 if (CurLexer) 491 CurLexer->Lex(Result); 492 else if (CurPTHLexer) 493 CurPTHLexer->Lex(Result); 494 else if (CurTokenLexer) 495 CurTokenLexer->Lex(Result); 496 else 497 CachingLex(Result); 498 } 499 500 /// LexNonComment - Lex a token. If it's a comment, keep lexing until we get 501 /// something not a comment. This is useful in -E -C mode where comments 502 /// would foul up preprocessor directive handling. 503 void LexNonComment(Token &Result) { 504 do 505 Lex(Result); 506 while (Result.getKind() == tok::comment); 507 } 508 509 /// LexUnexpandedToken - This is just like Lex, but this disables macro 510 /// expansion of identifier tokens. 511 void LexUnexpandedToken(Token &Result) { 512 // Disable macro expansion. 513 bool OldVal = DisableMacroExpansion; 514 DisableMacroExpansion = true; 515 // Lex the token. 516 Lex(Result); 517 518 // Reenable it. 519 DisableMacroExpansion = OldVal; 520 } 521 522 /// LookAhead - This peeks ahead N tokens and returns that token without 523 /// consuming any tokens. LookAhead(0) returns the next token that would be 524 /// returned by Lex(), LookAhead(1) returns the token after it, etc. This 525 /// returns normal tokens after phase 5. As such, it is equivalent to using 526 /// 'Lex', not 'LexUnexpandedToken'. 527 const Token &LookAhead(unsigned N) { 528 if (CachedLexPos + N < CachedTokens.size()) 529 return CachedTokens[CachedLexPos+N]; 530 else 531 return PeekAhead(N+1); 532 } 533 534 /// RevertCachedTokens - When backtracking is enabled and tokens are cached, 535 /// this allows to revert a specific number of tokens. 536 /// Note that the number of tokens being reverted should be up to the last 537 /// backtrack position, not more. 538 void RevertCachedTokens(unsigned N) { 539 assert(isBacktrackEnabled() && 540 "Should only be called when tokens are cached for backtracking"); 541 assert(signed(CachedLexPos) - signed(N) >= signed(BacktrackPositions.back()) 542 && "Should revert tokens up to the last backtrack position, not more"); 543 assert(signed(CachedLexPos) - signed(N) >= 0 && 544 "Corrupted backtrack positions ?"); 545 CachedLexPos -= N; 546 } 547 548 /// EnterToken - Enters a token in the token stream to be lexed next. If 549 /// BackTrack() is called afterwards, the token will remain at the insertion 550 /// point. 551 void EnterToken(const Token &Tok) { 552 EnterCachingLexMode(); 553 CachedTokens.insert(CachedTokens.begin()+CachedLexPos, Tok); 554 } 555 556 /// AnnotateCachedTokens - We notify the Preprocessor that if it is caching 557 /// tokens (because backtrack is enabled) it should replace the most recent 558 /// cached tokens with the given annotation token. This function has no effect 559 /// if backtracking is not enabled. 560 /// 561 /// Note that the use of this function is just for optimization; so that the 562 /// cached tokens doesn't get re-parsed and re-resolved after a backtrack is 563 /// invoked. 564 void AnnotateCachedTokens(const Token &Tok) { 565 assert(Tok.isAnnotation() && "Expected annotation token"); 566 if (CachedLexPos != 0 && isBacktrackEnabled()) 567 AnnotatePreviousCachedTokens(Tok); 568 } 569 570 /// \brief Replace the last token with an annotation token. 571 /// 572 /// Like AnnotateCachedTokens(), this routine replaces an 573 /// already-parsed (and resolved) token with an annotation 574 /// token. However, this routine only replaces the last token with 575 /// the annotation token; it does not affect any other cached 576 /// tokens. This function has no effect if backtracking is not 577 /// enabled. 578 void ReplaceLastTokenWithAnnotation(const Token &Tok) { 579 assert(Tok.isAnnotation() && "Expected annotation token"); 580 if (CachedLexPos != 0 && isBacktrackEnabled()) 581 CachedTokens[CachedLexPos-1] = Tok; 582 } 583 584 /// \brief Specify the point at which code-completion will be performed. 585 /// 586 /// \param File the file in which code completion should occur. If 587 /// this file is included multiple times, code-completion will 588 /// perform completion the first time it is included. If NULL, this 589 /// function clears out the code-completion point. 590 /// 591 /// \param Line the line at which code completion should occur 592 /// (1-based). 593 /// 594 /// \param Column the column at which code completion should occur 595 /// (1-based). 596 /// 597 /// \returns true if an error occurred, false otherwise. 598 bool SetCodeCompletionPoint(const FileEntry *File, 599 unsigned Line, unsigned Column); 600 601 /// \brief Determine if this source location refers into the file 602 /// for which we are performing code completion. 603 bool isCodeCompletionFile(SourceLocation FileLoc) const; 604 605 /// \brief Instruct the preprocessor to skip part of the main 606 /// the main source file. 607 /// 608 /// \brief Bytes The number of bytes in the preamble to skip. 609 /// 610 /// \brief StartOfLine Whether skipping these bytes puts the lexer at the 611 /// start of a line. 612 void setSkipMainFilePreamble(unsigned Bytes, bool StartOfLine) { 613 SkipMainFilePreamble.first = Bytes; 614 SkipMainFilePreamble.second = StartOfLine; 615 } 616 617 /// Diag - Forwarding function for diagnostics. This emits a diagnostic at 618 /// the specified Token's location, translating the token's start 619 /// position in the current buffer into a SourcePosition object for rendering. 620 DiagnosticBuilder Diag(SourceLocation Loc, unsigned DiagID) { 621 return Diags->Report(FullSourceLoc(Loc, getSourceManager()), DiagID); 622 } 623 624 DiagnosticBuilder Diag(const Token &Tok, unsigned DiagID) { 625 return Diags->Report(FullSourceLoc(Tok.getLocation(), getSourceManager()), 626 DiagID); 627 } 628 629 /// getSpelling() - Return the 'spelling' of the Tok token. The spelling of a 630 /// token is the characters used to represent the token in the source file 631 /// after trigraph expansion and escaped-newline folding. In particular, this 632 /// wants to get the true, uncanonicalized, spelling of things like digraphs 633 /// UCNs, etc. 634 /// 635 /// \param Invalid If non-NULL, will be set \c true if an error occurs. 636 std::string getSpelling(const Token &Tok, bool *Invalid = 0) const; 637 638 /// getSpelling() - Return the 'spelling' of the Tok token. The spelling of a 639 /// token is the characters used to represent the token in the source file 640 /// after trigraph expansion and escaped-newline folding. In particular, this 641 /// wants to get the true, uncanonicalized, spelling of things like digraphs 642 /// UCNs, etc. 643 static std::string getSpelling(const Token &Tok, 644 const SourceManager &SourceMgr, 645 const LangOptions &Features, 646 bool *Invalid = 0); 647 648 /// getSpelling - This method is used to get the spelling of a token into a 649 /// preallocated buffer, instead of as an std::string. The caller is required 650 /// to allocate enough space for the token, which is guaranteed to be at least 651 /// Tok.getLength() bytes long. The length of the actual result is returned. 652 /// 653 /// Note that this method may do two possible things: it may either fill in 654 /// the buffer specified with characters, or it may *change the input pointer* 655 /// to point to a constant buffer with the data already in it (avoiding a 656 /// copy). The caller is not allowed to modify the returned buffer pointer 657 /// if an internal buffer is returned. 658 unsigned getSpelling(const Token &Tok, const char *&Buffer, 659 bool *Invalid = 0) const; 660 661 /// getSpelling - This method is used to get the spelling of a token into a 662 /// SmallVector. Note that the returned StringRef may not point to the 663 /// supplied buffer if a copy can be avoided. 664 llvm::StringRef getSpelling(const Token &Tok, 665 llvm::SmallVectorImpl<char> &Buffer, 666 bool *Invalid = 0) const; 667 668 /// getSpellingOfSingleCharacterNumericConstant - Tok is a numeric constant 669 /// with length 1, return the character. 670 char getSpellingOfSingleCharacterNumericConstant(const Token &Tok, 671 bool *Invalid = 0) const { 672 assert(Tok.is(tok::numeric_constant) && 673 Tok.getLength() == 1 && "Called on unsupported token"); 674 assert(!Tok.needsCleaning() && "Token can't need cleaning with length 1"); 675 676 // If the token is carrying a literal data pointer, just use it. 677 if (const char *D = Tok.getLiteralData()) 678 return *D; 679 680 // Otherwise, fall back on getCharacterData, which is slower, but always 681 // works. 682 return *SourceMgr.getCharacterData(Tok.getLocation(), Invalid); 683 } 684 685 /// CreateString - Plop the specified string into a scratch buffer and set the 686 /// specified token's location and length to it. If specified, the source 687 /// location provides a location of the instantiation point of the token. 688 void CreateString(const char *Buf, unsigned Len, 689 Token &Tok, SourceLocation SourceLoc = SourceLocation()); 690 691 /// \brief Computes the source location just past the end of the 692 /// token at this source location. 693 /// 694 /// This routine can be used to produce a source location that 695 /// points just past the end of the token referenced by \p Loc, and 696 /// is generally used when a diagnostic needs to point just after a 697 /// token where it expected something different that it received. If 698 /// the returned source location would not be meaningful (e.g., if 699 /// it points into a macro), this routine returns an invalid 700 /// source location. 701 /// 702 /// \param Offset an offset from the end of the token, where the source 703 /// location should refer to. The default offset (0) produces a source 704 /// location pointing just past the end of the token; an offset of 1 produces 705 /// a source location pointing to the last character in the token, etc. 706 SourceLocation getLocForEndOfToken(SourceLocation Loc, unsigned Offset = 0); 707 708 /// DumpToken - Print the token to stderr, used for debugging. 709 /// 710 void DumpToken(const Token &Tok, bool DumpFlags = false) const; 711 void DumpLocation(SourceLocation Loc) const; 712 void DumpMacro(const MacroInfo &MI) const; 713 714 /// AdvanceToTokenCharacter - Given a location that specifies the start of a 715 /// token, return a new location that specifies a character within the token. 716 SourceLocation AdvanceToTokenCharacter(SourceLocation TokStart,unsigned Char); 717 718 /// IncrementPasteCounter - Increment the counters for the number of token 719 /// paste operations performed. If fast was specified, this is a 'fast paste' 720 /// case we handled. 721 /// 722 void IncrementPasteCounter(bool isFast) { 723 if (isFast) 724 ++NumFastTokenPaste; 725 else 726 ++NumTokenPaste; 727 } 728 729 void PrintStats(); 730 731 /// HandleMicrosoftCommentPaste - When the macro expander pastes together a 732 /// comment (/##/) in microsoft mode, this method handles updating the current 733 /// state, returning the token on the next source line. 734 void HandleMicrosoftCommentPaste(Token &Tok); 735 736 //===--------------------------------------------------------------------===// 737 // Preprocessor callback methods. These are invoked by a lexer as various 738 // directives and events are found. 739 740 /// LookUpIdentifierInfo - Given a tok::identifier token, look up the 741 /// identifier information for the token and install it into the token. 742 IdentifierInfo *LookUpIdentifierInfo(Token &Identifier, 743 const char *BufPtr = 0) const; 744 745 /// HandleIdentifier - This callback is invoked when the lexer reads an 746 /// identifier and has filled in the tokens IdentifierInfo member. This 747 /// callback potentially macro expands it or turns it into a named token (like 748 /// 'for'). 749 void HandleIdentifier(Token &Identifier); 750 751 752 /// HandleEndOfFile - This callback is invoked when the lexer hits the end of 753 /// the current file. This either returns the EOF token and returns true, or 754 /// pops a level off the include stack and returns false, at which point the 755 /// client should call lex again. 756 bool HandleEndOfFile(Token &Result, bool isEndOfMacro = false); 757 758 /// HandleEndOfTokenLexer - This callback is invoked when the current 759 /// TokenLexer hits the end of its token stream. 760 bool HandleEndOfTokenLexer(Token &Result); 761 762 /// HandleDirective - This callback is invoked when the lexer sees a # token 763 /// at the start of a line. This consumes the directive, modifies the 764 /// lexer/preprocessor state, and advances the lexer(s) so that the next token 765 /// read is the correct one. 766 void HandleDirective(Token &Result); 767 768 /// CheckEndOfDirective - Ensure that the next token is a tok::eom token. If 769 /// not, emit a diagnostic and consume up until the eom. If EnableMacros is 770 /// true, then we consider macros that expand to zero tokens as being ok. 771 void CheckEndOfDirective(const char *Directive, bool EnableMacros = false); 772 773 /// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the 774 /// current line until the tok::eom token is found. 775 void DiscardUntilEndOfDirective(); 776 777 /// SawDateOrTime - This returns true if the preprocessor has seen a use of 778 /// __DATE__ or __TIME__ in the file so far. 779 bool SawDateOrTime() const { 780 return DATELoc != SourceLocation() || TIMELoc != SourceLocation(); 781 } 782 unsigned getCounterValue() const { return CounterValue; } 783 void setCounterValue(unsigned V) { CounterValue = V; } 784 785 /// AllocateMacroInfo - Allocate a new MacroInfo object with the provide 786 /// SourceLocation. 787 MacroInfo *AllocateMacroInfo(SourceLocation L); 788 789 /// CloneMacroInfo - Allocate a new MacroInfo object which is clone of MI. 790 MacroInfo *CloneMacroInfo(const MacroInfo &MI); 791 792 /// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully 793 /// checked and spelled filename, e.g. as an operand of #include. This returns 794 /// true if the input filename was in <>'s or false if it were in ""'s. The 795 /// caller is expected to provide a buffer that is large enough to hold the 796 /// spelling of the filename, but is also expected to handle the case when 797 /// this method decides to use a different buffer. 798 bool GetIncludeFilenameSpelling(SourceLocation Loc,llvm::StringRef &Filename); 799 800 /// LookupFile - Given a "foo" or <foo> reference, look up the indicated file, 801 /// return null on failure. isAngled indicates whether the file reference is 802 /// for system #include's or not (i.e. using <> instead of ""). 803 const FileEntry *LookupFile(llvm::StringRef Filename, 804 bool isAngled, const DirectoryLookup *FromDir, 805 const DirectoryLookup *&CurDir); 806 807 /// GetCurLookup - The DirectoryLookup structure used to find the current 808 /// FileEntry, if CurLexer is non-null and if applicable. This allows us to 809 /// implement #include_next and find directory-specific properties. 810 const DirectoryLookup *GetCurDirLookup() { return CurDirLookup; } 811 812 /// isInPrimaryFile - Return true if we're in the top-level file, not in a 813 /// #include. 814 bool isInPrimaryFile() const; 815 816 /// ConcatenateIncludeName - Handle cases where the #include name is expanded 817 /// from a macro as multiple tokens, which need to be glued together. This 818 /// occurs for code like: 819 /// #define FOO <a/b.h> 820 /// #include FOO 821 /// because in this case, "<a/b.h>" is returned as 7 tokens, not one. 822 /// 823 /// This code concatenates and consumes tokens up to the '>' token. It 824 /// returns false if the > was found, otherwise it returns true if it finds 825 /// and consumes the EOM marker. 826 bool ConcatenateIncludeName(llvm::SmallString<128> &FilenameBuffer); 827 828private: 829 830 void PushIncludeMacroStack() { 831 IncludeMacroStack.push_back(IncludeStackInfo(CurLexer.take(), 832 CurPTHLexer.take(), 833 CurPPLexer, 834 CurTokenLexer.take(), 835 CurDirLookup)); 836 CurPPLexer = 0; 837 } 838 839 void PopIncludeMacroStack() { 840 CurLexer.reset(IncludeMacroStack.back().TheLexer); 841 CurPTHLexer.reset(IncludeMacroStack.back().ThePTHLexer); 842 CurPPLexer = IncludeMacroStack.back().ThePPLexer; 843 CurTokenLexer.reset(IncludeMacroStack.back().TheTokenLexer); 844 CurDirLookup = IncludeMacroStack.back().TheDirLookup; 845 IncludeMacroStack.pop_back(); 846 } 847 848 /// AllocateMacroInfo - Allocate a new MacroInfo object. 849 MacroInfo *AllocateMacroInfo(); 850 851 /// ReleaseMacroInfo - Release the specified MacroInfo. This memory will 852 /// be reused for allocating new MacroInfo objects. 853 void ReleaseMacroInfo(MacroInfo* MI); 854 855 /// ReadMacroName - Lex and validate a macro name, which occurs after a 856 /// #define or #undef. This emits a diagnostic, sets the token kind to eom, 857 /// and discards the rest of the macro line if the macro name is invalid. 858 void ReadMacroName(Token &MacroNameTok, char isDefineUndef = 0); 859 860 /// ReadMacroDefinitionArgList - The ( starting an argument list of a macro 861 /// definition has just been read. Lex the rest of the arguments and the 862 /// closing ), updating MI with what we learn. Return true if an error occurs 863 /// parsing the arg list. 864 bool ReadMacroDefinitionArgList(MacroInfo *MI); 865 866 /// SkipExcludedConditionalBlock - We just read a #if or related directive and 867 /// decided that the subsequent tokens are in the #if'd out portion of the 868 /// file. Lex the rest of the file, until we see an #endif. If 869 /// FoundNonSkipPortion is true, then we have already emitted code for part of 870 /// this #if directive, so #else/#elif blocks should never be entered. If 871 /// FoundElse is false, then #else directives are ok, if not, then we have 872 /// already seen one so a #else directive is a duplicate. When this returns, 873 /// the caller can lex the first valid token. 874 void SkipExcludedConditionalBlock(SourceLocation IfTokenLoc, 875 bool FoundNonSkipPortion, bool FoundElse); 876 877 /// PTHSkipExcludedConditionalBlock - A fast PTH version of 878 /// SkipExcludedConditionalBlock. 879 void PTHSkipExcludedConditionalBlock(); 880 881 /// EvaluateDirectiveExpression - Evaluate an integer constant expression that 882 /// may occur after a #if or #elif directive and return it as a bool. If the 883 /// expression is equivalent to "!defined(X)" return X in IfNDefMacro. 884 bool EvaluateDirectiveExpression(IdentifierInfo *&IfNDefMacro); 885 886 /// RegisterBuiltinPragmas - Install the standard preprocessor pragmas: 887 /// #pragma GCC poison/system_header/dependency and #pragma once. 888 void RegisterBuiltinPragmas(); 889 890 /// RegisterBuiltinMacros - Register builtin macros, such as __LINE__ with the 891 /// identifier table. 892 void RegisterBuiltinMacros(); 893 894 /// HandleMacroExpandedIdentifier - If an identifier token is read that is to 895 /// be expanded as a macro, handle it and return the next token as 'Tok'. If 896 /// the macro should not be expanded return true, otherwise return false. 897 bool HandleMacroExpandedIdentifier(Token &Tok, MacroInfo *MI); 898 899 /// isNextPPTokenLParen - Determine whether the next preprocessor token to be 900 /// lexed is a '('. If so, consume the token and return true, if not, this 901 /// method should have no observable side-effect on the lexed tokens. 902 bool isNextPPTokenLParen(); 903 904 /// ReadFunctionLikeMacroArgs - After reading "MACRO(", this method is 905 /// invoked to read all of the formal arguments specified for the macro 906 /// invocation. This returns null on error. 907 MacroArgs *ReadFunctionLikeMacroArgs(Token &MacroName, MacroInfo *MI, 908 SourceLocation &InstantiationEnd); 909 910 /// ExpandBuiltinMacro - If an identifier token is read that is to be expanded 911 /// as a builtin macro, handle it and return the next token as 'Tok'. 912 void ExpandBuiltinMacro(Token &Tok); 913 914 /// Handle_Pragma - Read a _Pragma directive, slice it up, process it, then 915 /// return the first token after the directive. The _Pragma token has just 916 /// been read into 'Tok'. 917 void Handle_Pragma(Token &Tok); 918 919 /// HandleMicrosoft__pragma - Like Handle_Pragma except the pragma text 920 /// is not enclosed within a string literal. 921 void HandleMicrosoft__pragma(Token &Tok); 922 923 void Handle_Pragma(unsigned Introducer, const std::string &StrVal, 924 SourceLocation PragmaLoc, SourceLocation RParenLoc); 925 926 /// EnterSourceFileWithLexer - Add a lexer to the top of the include stack and 927 /// start lexing tokens from it instead of the current buffer. 928 void EnterSourceFileWithLexer(Lexer *TheLexer, const DirectoryLookup *Dir); 929 930 /// EnterSourceFileWithPTH - Add a lexer to the top of the include stack and 931 /// start getting tokens from it using the PTH cache. 932 void EnterSourceFileWithPTH(PTHLexer *PL, const DirectoryLookup *Dir); 933 934 /// IsFileLexer - Returns true if we are lexing from a file and not a 935 /// pragma or a macro. 936 static bool IsFileLexer(const Lexer* L, const PreprocessorLexer* P) { 937 return L ? !L->isPragmaLexer() : P != 0; 938 } 939 940 static bool IsFileLexer(const IncludeStackInfo& I) { 941 return IsFileLexer(I.TheLexer, I.ThePPLexer); 942 } 943 944 bool IsFileLexer() const { 945 return IsFileLexer(CurLexer.get(), CurPPLexer); 946 } 947 948 //===--------------------------------------------------------------------===// 949 // Caching stuff. 950 void CachingLex(Token &Result); 951 bool InCachingLexMode() const { 952 // If the Lexer pointers are 0 and IncludeMacroStack is empty, it means 953 // that we are past EOF, not that we are in CachingLex mode. 954 return CurPPLexer == 0 && CurTokenLexer == 0 && CurPTHLexer == 0 && 955 !IncludeMacroStack.empty(); 956 } 957 void EnterCachingLexMode(); 958 void ExitCachingLexMode() { 959 if (InCachingLexMode()) 960 RemoveTopOfLexerStack(); 961 } 962 const Token &PeekAhead(unsigned N); 963 void AnnotatePreviousCachedTokens(const Token &Tok); 964 965 //===--------------------------------------------------------------------===// 966 /// Handle*Directive - implement the various preprocessor directives. These 967 /// should side-effect the current preprocessor object so that the next call 968 /// to Lex() will return the appropriate token next. 969 void HandleLineDirective(Token &Tok); 970 void HandleDigitDirective(Token &Tok); 971 void HandleUserDiagnosticDirective(Token &Tok, bool isWarning); 972 void HandleIdentSCCSDirective(Token &Tok); 973 974 // File inclusion. 975 void HandleIncludeDirective(Token &Tok, 976 const DirectoryLookup *LookupFrom = 0, 977 bool isImport = false); 978 void HandleIncludeNextDirective(Token &Tok); 979 void HandleIncludeMacrosDirective(Token &Tok); 980 void HandleImportDirective(Token &Tok); 981 982 // Macro handling. 983 void HandleDefineDirective(Token &Tok); 984 void HandleUndefDirective(Token &Tok); 985 986 // Conditional Inclusion. 987 void HandleIfdefDirective(Token &Tok, bool isIfndef, 988 bool ReadAnyTokensBeforeDirective); 989 void HandleIfDirective(Token &Tok, bool ReadAnyTokensBeforeDirective); 990 void HandleEndifDirective(Token &Tok); 991 void HandleElseDirective(Token &Tok); 992 void HandleElifDirective(Token &Tok); 993 994 // Pragmas. 995 void HandlePragmaDirective(unsigned Introducer); 996public: 997 void HandlePragmaOnce(Token &OnceTok); 998 void HandlePragmaMark(); 999 void HandlePragmaPoison(Token &PoisonTok); 1000 void HandlePragmaSystemHeader(Token &SysHeaderTok); 1001 void HandlePragmaDependency(Token &DependencyTok); 1002 void HandlePragmaComment(Token &CommentTok); 1003 void HandlePragmaMessage(Token &MessageTok); 1004 void HandlePragmaPushMacro(Token &Tok); 1005 void HandlePragmaPopMacro(Token &Tok); 1006 IdentifierInfo *ParsePragmaPushOrPopMacro(Token &Tok); 1007 1008 // Return true and store the first token only if any CommentHandler 1009 // has inserted some tokens and getCommentRetentionState() is false. 1010 bool HandleComment(Token &Token, SourceRange Comment); 1011}; 1012 1013/// \brief Abstract base class that describes a handler that will receive 1014/// source ranges for each of the comments encountered in the source file. 1015class CommentHandler { 1016public: 1017 virtual ~CommentHandler(); 1018 1019 // The handler shall return true if it has pushed any tokens 1020 // to be read using e.g. EnterToken or EnterTokenStream. 1021 virtual bool HandleComment(Preprocessor &PP, SourceRange Comment) = 0; 1022}; 1023 1024} // end namespace clang 1025 1026#endif 1027