Preprocessor.h revision f47724bf78299c7a50f008e0443c5f9f9f279ddc
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/Lexer.h" 18#include "clang/Lex/PTHLexer.h" 19#include "clang/Lex/PPCallbacks.h" 20#include "clang/Lex/TokenLexer.h" 21#include "clang/Lex/PTHManager.h" 22#include "clang/Basic/Builtins.h" 23#include "clang/Basic/Diagnostic.h" 24#include "clang/Basic/IdentifierTable.h" 25#include "clang/Basic/SourceLocation.h" 26#include "llvm/ADT/DenseMap.h" 27#include "llvm/ADT/OwningPtr.h" 28#include "llvm/ADT/SmallVector.h" 29#include "llvm/Support/Allocator.h" 30#include <vector> 31 32namespace clang { 33 34class SourceManager; 35class ExternalPreprocessorSource; 36class FileManager; 37class FileEntry; 38class HeaderSearch; 39class PragmaNamespace; 40class PragmaHandler; 41class CommentHandler; 42class ScratchBuffer; 43class TargetInfo; 44class PPCallbacks; 45class DirectoryLookup; 46class PreprocessingRecord; 47 48/// Preprocessor - This object engages in a tight little dance with the lexer to 49/// efficiently preprocess tokens. Lexers know only about tokens within a 50/// single source file, and don't know anything about preprocessor-level issues 51/// like the #include stack, token expansion, etc. 52/// 53class Preprocessor { 54 Diagnostic *Diags; 55 LangOptions Features; 56 const TargetInfo &Target; 57 FileManager &FileMgr; 58 SourceManager &SourceMgr; 59 ScratchBuffer *ScratchBuf; 60 HeaderSearch &HeaderInfo; 61 62 /// \brief External source of macros. 63 ExternalPreprocessorSource *ExternalSource; 64 65 /// PTH - An optional PTHManager object used for getting tokens from 66 /// a token cache rather than lexing the original source file. 67 llvm::OwningPtr<PTHManager> PTH; 68 69 /// BP - A BumpPtrAllocator object used to quickly allocate and release 70 /// objects internal to the Preprocessor. 71 llvm::BumpPtrAllocator BP; 72 73 /// Identifiers for builtin macros and other builtins. 74 IdentifierInfo *Ident__LINE__, *Ident__FILE__; // __LINE__, __FILE__ 75 IdentifierInfo *Ident__DATE__, *Ident__TIME__; // __DATE__, __TIME__ 76 IdentifierInfo *Ident__INCLUDE_LEVEL__; // __INCLUDE_LEVEL__ 77 IdentifierInfo *Ident__BASE_FILE__; // __BASE_FILE__ 78 IdentifierInfo *Ident__TIMESTAMP__; // __TIMESTAMP__ 79 IdentifierInfo *Ident__COUNTER__; // __COUNTER__ 80 IdentifierInfo *Ident_Pragma, *Ident__VA_ARGS__; // _Pragma, __VA_ARGS__ 81 IdentifierInfo *Ident__has_feature; // __has_feature 82 IdentifierInfo *Ident__has_builtin; // __has_builtin 83 IdentifierInfo *Ident__has_include; // __has_include 84 IdentifierInfo *Ident__has_include_next; // __has_include_next 85 86 SourceLocation DATELoc, TIMELoc; 87 unsigned CounterValue; // Next __COUNTER__ value. 88 89 enum { 90 /// MaxIncludeStackDepth - Maximum depth of #includes. 91 MaxAllowedIncludeStackDepth = 200 92 }; 93 94 // State that is set before the preprocessor begins. 95 bool KeepComments : 1; 96 bool KeepMacroComments : 1; 97 98 // State that changes while the preprocessor runs: 99 bool InMacroArgs : 1; // True if parsing fn macro invocation args. 100 101 /// Whether the preprocessor owns the header search object. 102 bool OwnsHeaderSearch : 1; 103 104 /// DisableMacroExpansion - True if macro expansion is disabled. 105 bool DisableMacroExpansion : 1; 106 107 /// \brief Whether we have already loaded macros from the external source. 108 mutable bool ReadMacrosFromExternalSource : 1; 109 110 /// Identifiers - This is mapping/lookup information for all identifiers in 111 /// the program, including program keywords. 112 mutable IdentifierTable Identifiers; 113 114 /// Selectors - This table contains all the selectors in the program. Unlike 115 /// IdentifierTable above, this table *isn't* populated by the preprocessor. 116 /// It is declared/instantiated here because it's role/lifetime is 117 /// conceptually similar the IdentifierTable. In addition, the current control 118 /// flow (in clang::ParseAST()), make it convenient to put here. 119 /// FIXME: Make sure the lifetime of Identifiers/Selectors *isn't* tied to 120 /// the lifetime of the preprocessor. 121 SelectorTable Selectors; 122 123 /// BuiltinInfo - Information about builtins. 124 Builtin::Context BuiltinInfo; 125 126 /// PragmaHandlers - This tracks all of the pragmas that the client registered 127 /// with this preprocessor. 128 PragmaNamespace *PragmaHandlers; 129 130 /// \brief Tracks all of the comment handlers that the client registered 131 /// with this preprocessor. 132 std::vector<CommentHandler *> CommentHandlers; 133 134 /// \brief The file that we're performing code-completion for, if any. 135 const FileEntry *CodeCompletionFile; 136 137 /// \brief The number of bytes that we will initially skip when entering the 138 /// main file, which is used when loading a precompiled preamble, along 139 /// with a flag that indicates whether skipping this number of bytes will 140 /// place the lexer at the start of a line. 141 std::pair<unsigned, bool> SkipMainFilePreamble; 142 143 /// CurLexer - This is the current top of the stack that we're lexing from if 144 /// not expanding a macro and we are lexing directly from source code. 145 /// Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null. 146 llvm::OwningPtr<Lexer> CurLexer; 147 148 /// CurPTHLexer - This is the current top of stack that we're lexing from if 149 /// not expanding from a macro and we are lexing from a PTH cache. 150 /// Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null. 151 llvm::OwningPtr<PTHLexer> CurPTHLexer; 152 153 /// CurPPLexer - This is the current top of the stack what we're lexing from 154 /// if not expanding a macro. This is an alias for either CurLexer or 155 /// CurPTHLexer. 156 PreprocessorLexer *CurPPLexer; 157 158 /// CurLookup - The DirectoryLookup structure used to find the current 159 /// FileEntry, if CurLexer is non-null and if applicable. This allows us to 160 /// implement #include_next and find directory-specific properties. 161 const DirectoryLookup *CurDirLookup; 162 163 /// CurTokenLexer - This is the current macro we are expanding, if we are 164 /// expanding a macro. One of CurLexer and CurTokenLexer must be null. 165 llvm::OwningPtr<TokenLexer> CurTokenLexer; 166 167 /// IncludeMacroStack - This keeps track of the stack of files currently 168 /// #included, and macros currently being expanded from, not counting 169 /// CurLexer/CurTokenLexer. 170 struct IncludeStackInfo { 171 Lexer *TheLexer; 172 PTHLexer *ThePTHLexer; 173 PreprocessorLexer *ThePPLexer; 174 TokenLexer *TheTokenLexer; 175 const DirectoryLookup *TheDirLookup; 176 177 IncludeStackInfo(Lexer *L, PTHLexer* P, PreprocessorLexer* PPL, 178 TokenLexer* TL, const DirectoryLookup *D) 179 : TheLexer(L), ThePTHLexer(P), ThePPLexer(PPL), TheTokenLexer(TL), 180 TheDirLookup(D) {} 181 }; 182 std::vector<IncludeStackInfo> IncludeMacroStack; 183 184 /// Callbacks - These are actions invoked when some preprocessor activity is 185 /// encountered (e.g. a file is #included, etc). 186 PPCallbacks *Callbacks; 187 188 /// Macros - For each IdentifierInfo with 'HasMacro' set, we keep a mapping 189 /// to the actual definition of the macro. 190 llvm::DenseMap<IdentifierInfo*, MacroInfo*> Macros; 191 192 /// MICache - A "freelist" of MacroInfo objects that can be reused for quick 193 /// allocation. 194 /// FIXME: why not use a singly linked list? 195 std::vector<MacroInfo*> MICache; 196 197 /// MacroArgCache - This is a "freelist" of MacroArg objects that can be 198 /// reused for quick allocation. 199 MacroArgs *MacroArgCache; 200 friend class MacroArgs; 201 202 /// PragmaPushMacroInfo - For each IdentifierInfo used in a #pragma 203 /// push_macro directive, we keep a MacroInfo stack used to restore 204 /// previous macro value. 205 llvm::DenseMap<IdentifierInfo*, std::vector<MacroInfo*> > PragmaPushMacroInfo; 206 207 // Various statistics we track for performance analysis. 208 unsigned NumDirectives, NumIncluded, NumDefined, NumUndefined, NumPragma; 209 unsigned NumIf, NumElse, NumEndif; 210 unsigned NumEnteredSourceFiles, MaxIncludeStackDepth; 211 unsigned NumMacroExpanded, NumFnMacroExpanded, NumBuiltinMacroExpanded; 212 unsigned NumFastMacroExpanded, NumTokenPaste, NumFastTokenPaste; 213 unsigned NumSkipped; 214 215 /// Predefines - This string is the predefined macros that preprocessor 216 /// should use from the command line etc. 217 std::string Predefines; 218 219 /// TokenLexerCache - Cache macro expanders to reduce malloc traffic. 220 enum { TokenLexerCacheSize = 8 }; 221 unsigned NumCachedTokenLexers; 222 TokenLexer *TokenLexerCache[TokenLexerCacheSize]; 223 224 /// \brief A record of the macro definitions and instantiations that 225 /// occurred during preprocessing. 226 /// 227 /// This is an optional side structure that can be enabled with 228 /// \c createPreprocessingRecord() prior to preprocessing. 229 PreprocessingRecord *Record; 230 231private: // Cached tokens state. 232 typedef llvm::SmallVector<Token, 1> CachedTokensTy; 233 234 /// CachedTokens - Cached tokens are stored here when we do backtracking or 235 /// lookahead. They are "lexed" by the CachingLex() method. 236 CachedTokensTy CachedTokens; 237 238 /// CachedLexPos - The position of the cached token that CachingLex() should 239 /// "lex" next. If it points beyond the CachedTokens vector, it means that 240 /// a normal Lex() should be invoked. 241 CachedTokensTy::size_type CachedLexPos; 242 243 /// BacktrackPositions - Stack of backtrack positions, allowing nested 244 /// backtracks. The EnableBacktrackAtThisPos() method pushes a position to 245 /// indicate where CachedLexPos should be set when the BackTrack() method is 246 /// invoked (at which point the last position is popped). 247 std::vector<CachedTokensTy::size_type> BacktrackPositions; 248 249public: 250 Preprocessor(Diagnostic &diags, const LangOptions &opts, 251 const TargetInfo &target, 252 SourceManager &SM, HeaderSearch &Headers, 253 IdentifierInfoLookup *IILookup = 0, 254 bool OwnsHeaderSearch = false); 255 256 ~Preprocessor(); 257 258 Diagnostic &getDiagnostics() const { return *Diags; } 259 void setDiagnostics(Diagnostic &D) { Diags = &D; } 260 261 const LangOptions &getLangOptions() const { return Features; } 262 const TargetInfo &getTargetInfo() const { return Target; } 263 FileManager &getFileManager() const { return FileMgr; } 264 SourceManager &getSourceManager() const { return SourceMgr; } 265 HeaderSearch &getHeaderSearchInfo() const { return HeaderInfo; } 266 267 IdentifierTable &getIdentifierTable() { return Identifiers; } 268 SelectorTable &getSelectorTable() { return Selectors; } 269 Builtin::Context &getBuiltinInfo() { return BuiltinInfo; } 270 llvm::BumpPtrAllocator &getPreprocessorAllocator() { return BP; } 271 272 void setPTHManager(PTHManager* pm); 273 274 PTHManager *getPTHManager() { return PTH.get(); } 275 276 void setExternalSource(ExternalPreprocessorSource *Source) { 277 ExternalSource = Source; 278 } 279 280 ExternalPreprocessorSource *getExternalSource() const { 281 return ExternalSource; 282 } 283 284 /// SetCommentRetentionState - Control whether or not the preprocessor retains 285 /// comments in output. 286 void SetCommentRetentionState(bool KeepComments, bool KeepMacroComments) { 287 this->KeepComments = KeepComments | KeepMacroComments; 288 this->KeepMacroComments = KeepMacroComments; 289 } 290 291 bool getCommentRetentionState() const { return KeepComments; } 292 293 /// isCurrentLexer - Return true if we are lexing directly from the specified 294 /// lexer. 295 bool isCurrentLexer(const PreprocessorLexer *L) const { 296 return CurPPLexer == L; 297 } 298 299 /// getCurrentLexer - Return the current lexer being lexed from. Note 300 /// that this ignores any potentially active macro expansions and _Pragma 301 /// expansions going on at the time. 302 PreprocessorLexer *getCurrentLexer() const { return CurPPLexer; } 303 304 /// getCurrentFileLexer - Return the current file lexer being lexed from. 305 /// Note that this ignores any potentially active macro expansions and _Pragma 306 /// expansions going on at the time. 307 PreprocessorLexer *getCurrentFileLexer() const; 308 309 /// getPPCallbacks/addPPCallbacks - Accessors for preprocessor callbacks. 310 /// Note that this class takes ownership of any PPCallbacks object given to 311 /// it. 312 PPCallbacks *getPPCallbacks() const { return Callbacks; } 313 void addPPCallbacks(PPCallbacks *C) { 314 if (Callbacks) 315 C = new PPChainedCallbacks(C, Callbacks); 316 Callbacks = C; 317 } 318 319 /// getMacroInfo - Given an identifier, return the MacroInfo it is #defined to 320 /// or null if it isn't #define'd. 321 MacroInfo *getMacroInfo(IdentifierInfo *II) const { 322 return II->hasMacroDefinition() ? Macros.find(II)->second : 0; 323 } 324 325 /// setMacroInfo - Specify a macro for this identifier. 326 /// 327 void setMacroInfo(IdentifierInfo *II, MacroInfo *MI); 328 329 /// macro_iterator/macro_begin/macro_end - This allows you to walk the current 330 /// state of the macro table. This visits every currently-defined macro. 331 typedef llvm::DenseMap<IdentifierInfo*, 332 MacroInfo*>::const_iterator macro_iterator; 333 macro_iterator macro_begin(bool IncludeExternalMacros = true) const; 334 macro_iterator macro_end(bool IncludeExternalMacros = true) const; 335 336 const std::string &getPredefines() const { return Predefines; } 337 /// setPredefines - Set the predefines for this Preprocessor. These 338 /// predefines are automatically injected when parsing the main file. 339 void setPredefines(const char *P) { Predefines = P; } 340 void setPredefines(const std::string &P) { Predefines = P; } 341 342 /// getIdentifierInfo - Return information about the specified preprocessor 343 /// identifier token. The version of this method that takes two character 344 /// pointers is preferred unless the identifier is already available as a 345 /// string (this avoids allocation and copying of memory to construct an 346 /// std::string). 347 IdentifierInfo *getIdentifierInfo(llvm::StringRef Name) const { 348 return &Identifiers.get(Name); 349 } 350 351 /// AddPragmaHandler - Add the specified pragma handler to the preprocessor. 352 /// If 'Namespace' is non-null, then it is a token required to exist on the 353 /// pragma line before the pragma string starts, e.g. "STDC" or "GCC". 354 void AddPragmaHandler(llvm::StringRef Namespace, PragmaHandler *Handler); 355 void AddPragmaHandler(PragmaHandler *Handler) { 356 AddPragmaHandler(llvm::StringRef(), Handler); 357 } 358 359 /// RemovePragmaHandler - Remove the specific pragma handler from 360 /// the preprocessor. If \arg Namespace is non-null, then it should 361 /// be the namespace that \arg Handler was added to. It is an error 362 /// to remove a handler that has not been registered. 363 void RemovePragmaHandler(llvm::StringRef Namespace, PragmaHandler *Handler); 364 void RemovePragmaHandler(PragmaHandler *Handler) { 365 RemovePragmaHandler(llvm::StringRef(), Handler); 366 } 367 368 /// \brief Add the specified comment handler to the preprocessor. 369 void AddCommentHandler(CommentHandler *Handler); 370 371 /// \brief Remove the specified comment handler. 372 /// 373 /// It is an error to remove a handler that has not been registered. 374 void RemoveCommentHandler(CommentHandler *Handler); 375 376 /// \brief Retrieve the preprocessing record, or NULL if there is no 377 /// preprocessing record. 378 PreprocessingRecord *getPreprocessingRecord() const { return Record; } 379 380 /// \brief Create a new preprocessing record, which will keep track of 381 /// all macro expansions, macro definitions, etc. 382 void createPreprocessingRecord(); 383 384 /// EnterMainSourceFile - Enter the specified FileID as the main source file, 385 /// which implicitly adds the builtin defines etc. 386 void EnterMainSourceFile(); 387 388 /// EndSourceFile - Inform the preprocessor callbacks that processing is 389 /// complete. 390 void EndSourceFile(); 391 392 /// EnterSourceFile - Add a source file to the top of the include stack and 393 /// start lexing tokens from it instead of the current buffer. Emit an error 394 /// and don't enter the file on error. 395 void EnterSourceFile(FileID CurFileID, const DirectoryLookup *Dir, 396 SourceLocation Loc); 397 398 /// EnterMacro - Add a Macro to the top of the include stack and start lexing 399 /// tokens from it instead of the current buffer. Args specifies the 400 /// tokens input to a function-like macro. 401 /// 402 /// ILEnd specifies the location of the ')' for a function-like macro or the 403 /// identifier for an object-like macro. 404 void EnterMacro(Token &Identifier, SourceLocation ILEnd, MacroArgs *Args); 405 406 /// EnterTokenStream - Add a "macro" context to the top of the include stack, 407 /// which will cause the lexer to start returning the specified tokens. 408 /// 409 /// If DisableMacroExpansion is true, tokens lexed from the token stream will 410 /// not be subject to further macro expansion. Otherwise, these tokens will 411 /// be re-macro-expanded when/if expansion is enabled. 412 /// 413 /// If OwnsTokens is false, this method assumes that the specified stream of 414 /// tokens has a permanent owner somewhere, so they do not need to be copied. 415 /// If it is true, it assumes the array of tokens is allocated with new[] and 416 /// must be freed. 417 /// 418 void EnterTokenStream(const Token *Toks, unsigned NumToks, 419 bool DisableMacroExpansion, bool OwnsTokens); 420 421 /// RemoveTopOfLexerStack - Pop the current lexer/macro exp off the top of the 422 /// lexer stack. This should only be used in situations where the current 423 /// state of the top-of-stack lexer is known. 424 void RemoveTopOfLexerStack(); 425 426 /// EnableBacktrackAtThisPos - From the point that this method is called, and 427 /// until CommitBacktrackedTokens() or Backtrack() is called, the Preprocessor 428 /// keeps track of the lexed tokens so that a subsequent Backtrack() call will 429 /// make the Preprocessor re-lex the same tokens. 430 /// 431 /// Nested backtracks are allowed, meaning that EnableBacktrackAtThisPos can 432 /// be called multiple times and CommitBacktrackedTokens/Backtrack calls will 433 /// be combined with the EnableBacktrackAtThisPos calls in reverse order. 434 /// 435 /// NOTE: *DO NOT* forget to call either CommitBacktrackedTokens or Backtrack 436 /// at some point after EnableBacktrackAtThisPos. If you don't, caching of 437 /// tokens will continue indefinitely. 438 /// 439 void EnableBacktrackAtThisPos(); 440 441 /// CommitBacktrackedTokens - Disable the last EnableBacktrackAtThisPos call. 442 void CommitBacktrackedTokens(); 443 444 /// Backtrack - Make Preprocessor re-lex the tokens that were lexed since 445 /// EnableBacktrackAtThisPos() was previously called. 446 void Backtrack(); 447 448 /// isBacktrackEnabled - True if EnableBacktrackAtThisPos() was called and 449 /// caching of tokens is on. 450 bool isBacktrackEnabled() const { return !BacktrackPositions.empty(); } 451 452 /// Lex - To lex a token from the preprocessor, just pull a token from the 453 /// current lexer or macro object. 454 void Lex(Token &Result) { 455 if (CurLexer) 456 CurLexer->Lex(Result); 457 else if (CurPTHLexer) 458 CurPTHLexer->Lex(Result); 459 else if (CurTokenLexer) 460 CurTokenLexer->Lex(Result); 461 else 462 CachingLex(Result); 463 } 464 465 /// LexNonComment - Lex a token. If it's a comment, keep lexing until we get 466 /// something not a comment. This is useful in -E -C mode where comments 467 /// would foul up preprocessor directive handling. 468 void LexNonComment(Token &Result) { 469 do 470 Lex(Result); 471 while (Result.getKind() == tok::comment); 472 } 473 474 /// LexUnexpandedToken - This is just like Lex, but this disables macro 475 /// expansion of identifier tokens. 476 void LexUnexpandedToken(Token &Result) { 477 // Disable macro expansion. 478 bool OldVal = DisableMacroExpansion; 479 DisableMacroExpansion = true; 480 // Lex the token. 481 Lex(Result); 482 483 // Reenable it. 484 DisableMacroExpansion = OldVal; 485 } 486 487 /// LookAhead - This peeks ahead N tokens and returns that token without 488 /// consuming any tokens. LookAhead(0) returns the next token that would be 489 /// returned by Lex(), LookAhead(1) returns the token after it, etc. This 490 /// returns normal tokens after phase 5. As such, it is equivalent to using 491 /// 'Lex', not 'LexUnexpandedToken'. 492 const Token &LookAhead(unsigned N) { 493 if (CachedLexPos + N < CachedTokens.size()) 494 return CachedTokens[CachedLexPos+N]; 495 else 496 return PeekAhead(N+1); 497 } 498 499 /// RevertCachedTokens - When backtracking is enabled and tokens are cached, 500 /// this allows to revert a specific number of tokens. 501 /// Note that the number of tokens being reverted should be up to the last 502 /// backtrack position, not more. 503 void RevertCachedTokens(unsigned N) { 504 assert(isBacktrackEnabled() && 505 "Should only be called when tokens are cached for backtracking"); 506 assert(signed(CachedLexPos) - signed(N) >= signed(BacktrackPositions.back()) 507 && "Should revert tokens up to the last backtrack position, not more"); 508 assert(signed(CachedLexPos) - signed(N) >= 0 && 509 "Corrupted backtrack positions ?"); 510 CachedLexPos -= N; 511 } 512 513 /// EnterToken - Enters a token in the token stream to be lexed next. If 514 /// BackTrack() is called afterwards, the token will remain at the insertion 515 /// point. 516 void EnterToken(const Token &Tok) { 517 EnterCachingLexMode(); 518 CachedTokens.insert(CachedTokens.begin()+CachedLexPos, Tok); 519 } 520 521 /// AnnotateCachedTokens - We notify the Preprocessor that if it is caching 522 /// tokens (because backtrack is enabled) it should replace the most recent 523 /// cached tokens with the given annotation token. This function has no effect 524 /// if backtracking is not enabled. 525 /// 526 /// Note that the use of this function is just for optimization; so that the 527 /// cached tokens doesn't get re-parsed and re-resolved after a backtrack is 528 /// invoked. 529 void AnnotateCachedTokens(const Token &Tok) { 530 assert(Tok.isAnnotation() && "Expected annotation token"); 531 if (CachedLexPos != 0 && isBacktrackEnabled()) 532 AnnotatePreviousCachedTokens(Tok); 533 } 534 535 /// \brief Replace the last token with an annotation token. 536 /// 537 /// Like AnnotateCachedTokens(), this routine replaces an 538 /// already-parsed (and resolved) token with an annotation 539 /// token. However, this routine only replaces the last token with 540 /// the annotation token; it does not affect any other cached 541 /// tokens. This function has no effect if backtracking is not 542 /// enabled. 543 void ReplaceLastTokenWithAnnotation(const Token &Tok) { 544 assert(Tok.isAnnotation() && "Expected annotation token"); 545 if (CachedLexPos != 0 && isBacktrackEnabled()) 546 CachedTokens[CachedLexPos-1] = Tok; 547 } 548 549 /// \brief Specify the point at which code-completion will be performed. 550 /// 551 /// \param File the file in which code completion should occur. If 552 /// this file is included multiple times, code-completion will 553 /// perform completion the first time it is included. If NULL, this 554 /// function clears out the code-completion point. 555 /// 556 /// \param Line the line at which code completion should occur 557 /// (1-based). 558 /// 559 /// \param Column the column at which code completion should occur 560 /// (1-based). 561 /// 562 /// \returns true if an error occurred, false otherwise. 563 bool SetCodeCompletionPoint(const FileEntry *File, 564 unsigned Line, unsigned Column); 565 566 /// \brief Determine if this source location refers into the file 567 /// for which we are performing code completion. 568 bool isCodeCompletionFile(SourceLocation FileLoc) const; 569 570 /// \brief Instruct the preprocessor to skip part of the main 571 /// the main source file. 572 /// 573 /// \brief Bytes The number of bytes in the preamble to skip. 574 /// 575 /// \brief StartOfLine Whether skipping these bytes puts the lexer at the 576 /// start of a line. 577 void setSkipMainFilePreamble(unsigned Bytes, bool StartOfLine) { 578 SkipMainFilePreamble.first = Bytes; 579 SkipMainFilePreamble.second = StartOfLine; 580 } 581 582 /// Diag - Forwarding function for diagnostics. This emits a diagnostic at 583 /// the specified Token's location, translating the token's start 584 /// position in the current buffer into a SourcePosition object for rendering. 585 DiagnosticBuilder Diag(SourceLocation Loc, unsigned DiagID) { 586 return Diags->Report(FullSourceLoc(Loc, getSourceManager()), DiagID); 587 } 588 589 DiagnosticBuilder Diag(const Token &Tok, unsigned DiagID) { 590 return Diags->Report(FullSourceLoc(Tok.getLocation(), getSourceManager()), 591 DiagID); 592 } 593 594 /// getSpelling() - Return the 'spelling' of the Tok token. The spelling of a 595 /// token is the characters used to represent the token in the source file 596 /// after trigraph expansion and escaped-newline folding. In particular, this 597 /// wants to get the true, uncanonicalized, spelling of things like digraphs 598 /// UCNs, etc. 599 /// 600 /// \param Invalid If non-NULL, will be set \c true if an error occurs. 601 std::string getSpelling(const Token &Tok, bool *Invalid = 0) const; 602 603 /// getSpelling() - Return the 'spelling' of the Tok token. The spelling of a 604 /// token is the characters used to represent the token in the source file 605 /// after trigraph expansion and escaped-newline folding. In particular, this 606 /// wants to get the true, uncanonicalized, spelling of things like digraphs 607 /// UCNs, etc. 608 static std::string getSpelling(const Token &Tok, 609 const SourceManager &SourceMgr, 610 const LangOptions &Features, 611 bool *Invalid = 0); 612 613 /// getSpelling - This method is used to get the spelling of a token into a 614 /// preallocated buffer, instead of as an std::string. The caller is required 615 /// to allocate enough space for the token, which is guaranteed to be at least 616 /// Tok.getLength() bytes long. The length of the actual result is returned. 617 /// 618 /// Note that this method may do two possible things: it may either fill in 619 /// the buffer specified with characters, or it may *change the input pointer* 620 /// to point to a constant buffer with the data already in it (avoiding a 621 /// copy). The caller is not allowed to modify the returned buffer pointer 622 /// if an internal buffer is returned. 623 unsigned getSpelling(const Token &Tok, const char *&Buffer, 624 bool *Invalid = 0) const; 625 626 /// getSpelling - This method is used to get the spelling of a token into a 627 /// SmallVector. Note that the returned StringRef may not point to the 628 /// supplied buffer if a copy can be avoided. 629 llvm::StringRef getSpelling(const Token &Tok, 630 llvm::SmallVectorImpl<char> &Buffer, 631 bool *Invalid = 0) const; 632 633 /// getSpellingOfSingleCharacterNumericConstant - Tok is a numeric constant 634 /// with length 1, return the character. 635 char getSpellingOfSingleCharacterNumericConstant(const Token &Tok, 636 bool *Invalid = 0) const { 637 assert(Tok.is(tok::numeric_constant) && 638 Tok.getLength() == 1 && "Called on unsupported token"); 639 assert(!Tok.needsCleaning() && "Token can't need cleaning with length 1"); 640 641 // If the token is carrying a literal data pointer, just use it. 642 if (const char *D = Tok.getLiteralData()) 643 return *D; 644 645 // Otherwise, fall back on getCharacterData, which is slower, but always 646 // works. 647 return *SourceMgr.getCharacterData(Tok.getLocation(), Invalid); 648 } 649 650 /// CreateString - Plop the specified string into a scratch buffer and set the 651 /// specified token's location and length to it. If specified, the source 652 /// location provides a location of the instantiation point of the token. 653 void CreateString(const char *Buf, unsigned Len, 654 Token &Tok, SourceLocation SourceLoc = SourceLocation()); 655 656 /// \brief Computes the source location just past the end of the 657 /// token at this source location. 658 /// 659 /// This routine can be used to produce a source location that 660 /// points just past the end of the token referenced by \p Loc, and 661 /// is generally used when a diagnostic needs to point just after a 662 /// token where it expected something different that it received. If 663 /// the returned source location would not be meaningful (e.g., if 664 /// it points into a macro), this routine returns an invalid 665 /// source location. 666 /// 667 /// \param Offset an offset from the end of the token, where the source 668 /// location should refer to. The default offset (0) produces a source 669 /// location pointing just past the end of the token; an offset of 1 produces 670 /// a source location pointing to the last character in the token, etc. 671 SourceLocation getLocForEndOfToken(SourceLocation Loc, unsigned Offset = 0); 672 673 /// DumpToken - Print the token to stderr, used for debugging. 674 /// 675 void DumpToken(const Token &Tok, bool DumpFlags = false) const; 676 void DumpLocation(SourceLocation Loc) const; 677 void DumpMacro(const MacroInfo &MI) const; 678 679 /// AdvanceToTokenCharacter - Given a location that specifies the start of a 680 /// token, return a new location that specifies a character within the token. 681 SourceLocation AdvanceToTokenCharacter(SourceLocation TokStart,unsigned Char); 682 683 /// IncrementPasteCounter - Increment the counters for the number of token 684 /// paste operations performed. If fast was specified, this is a 'fast paste' 685 /// case we handled. 686 /// 687 void IncrementPasteCounter(bool isFast) { 688 if (isFast) 689 ++NumFastTokenPaste; 690 else 691 ++NumTokenPaste; 692 } 693 694 void PrintStats(); 695 696 /// HandleMicrosoftCommentPaste - When the macro expander pastes together a 697 /// comment (/##/) in microsoft mode, this method handles updating the current 698 /// state, returning the token on the next source line. 699 void HandleMicrosoftCommentPaste(Token &Tok); 700 701 //===--------------------------------------------------------------------===// 702 // Preprocessor callback methods. These are invoked by a lexer as various 703 // directives and events are found. 704 705 /// LookUpIdentifierInfo - Given a tok::identifier token, look up the 706 /// identifier information for the token and install it into the token. 707 IdentifierInfo *LookUpIdentifierInfo(Token &Identifier, 708 const char *BufPtr = 0) const; 709 710 /// HandleIdentifier - This callback is invoked when the lexer reads an 711 /// identifier and has filled in the tokens IdentifierInfo member. This 712 /// callback potentially macro expands it or turns it into a named token (like 713 /// 'for'). 714 void HandleIdentifier(Token &Identifier); 715 716 717 /// HandleEndOfFile - This callback is invoked when the lexer hits the end of 718 /// the current file. This either returns the EOF token and returns true, or 719 /// pops a level off the include stack and returns false, at which point the 720 /// client should call lex again. 721 bool HandleEndOfFile(Token &Result, bool isEndOfMacro = false); 722 723 /// HandleEndOfTokenLexer - This callback is invoked when the current 724 /// TokenLexer hits the end of its token stream. 725 bool HandleEndOfTokenLexer(Token &Result); 726 727 /// HandleDirective - This callback is invoked when the lexer sees a # token 728 /// at the start of a line. This consumes the directive, modifies the 729 /// lexer/preprocessor state, and advances the lexer(s) so that the next token 730 /// read is the correct one. 731 void HandleDirective(Token &Result); 732 733 /// CheckEndOfDirective - Ensure that the next token is a tok::eom token. If 734 /// not, emit a diagnostic and consume up until the eom. If EnableMacros is 735 /// true, then we consider macros that expand to zero tokens as being ok. 736 void CheckEndOfDirective(const char *Directive, bool EnableMacros = false); 737 738 /// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the 739 /// current line until the tok::eom token is found. 740 void DiscardUntilEndOfDirective(); 741 742 /// SawDateOrTime - This returns true if the preprocessor has seen a use of 743 /// __DATE__ or __TIME__ in the file so far. 744 bool SawDateOrTime() const { 745 return DATELoc != SourceLocation() || TIMELoc != SourceLocation(); 746 } 747 unsigned getCounterValue() const { return CounterValue; } 748 void setCounterValue(unsigned V) { CounterValue = V; } 749 750 /// AllocateMacroInfo - Allocate a new MacroInfo object with the provide 751 /// SourceLocation. 752 MacroInfo *AllocateMacroInfo(SourceLocation L); 753 754 /// CloneMacroInfo - Allocate a new MacroInfo object which is clone of MI. 755 MacroInfo *CloneMacroInfo(const MacroInfo &MI); 756 757 /// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully 758 /// checked and spelled filename, e.g. as an operand of #include. This returns 759 /// true if the input filename was in <>'s or false if it were in ""'s. The 760 /// caller is expected to provide a buffer that is large enough to hold the 761 /// spelling of the filename, but is also expected to handle the case when 762 /// this method decides to use a different buffer. 763 bool GetIncludeFilenameSpelling(SourceLocation Loc,llvm::StringRef &Filename); 764 765 /// LookupFile - Given a "foo" or <foo> reference, look up the indicated file, 766 /// return null on failure. isAngled indicates whether the file reference is 767 /// for system #include's or not (i.e. using <> instead of ""). 768 const FileEntry *LookupFile(llvm::StringRef Filename, 769 bool isAngled, const DirectoryLookup *FromDir, 770 const DirectoryLookup *&CurDir); 771 772 /// GetCurLookup - The DirectoryLookup structure used to find the current 773 /// FileEntry, if CurLexer is non-null and if applicable. This allows us to 774 /// implement #include_next and find directory-specific properties. 775 const DirectoryLookup *GetCurDirLookup() { return CurDirLookup; } 776 777 /// isInPrimaryFile - Return true if we're in the top-level file, not in a 778 /// #include. 779 bool isInPrimaryFile() const; 780 781 /// ConcatenateIncludeName - Handle cases where the #include name is expanded 782 /// from a macro as multiple tokens, which need to be glued together. This 783 /// occurs for code like: 784 /// #define FOO <a/b.h> 785 /// #include FOO 786 /// because in this case, "<a/b.h>" is returned as 7 tokens, not one. 787 /// 788 /// This code concatenates and consumes tokens up to the '>' token. It 789 /// returns false if the > was found, otherwise it returns true if it finds 790 /// and consumes the EOM marker. 791 bool ConcatenateIncludeName(llvm::SmallString<128> &FilenameBuffer); 792 793private: 794 795 void PushIncludeMacroStack() { 796 IncludeMacroStack.push_back(IncludeStackInfo(CurLexer.take(), 797 CurPTHLexer.take(), 798 CurPPLexer, 799 CurTokenLexer.take(), 800 CurDirLookup)); 801 CurPPLexer = 0; 802 } 803 804 void PopIncludeMacroStack() { 805 CurLexer.reset(IncludeMacroStack.back().TheLexer); 806 CurPTHLexer.reset(IncludeMacroStack.back().ThePTHLexer); 807 CurPPLexer = IncludeMacroStack.back().ThePPLexer; 808 CurTokenLexer.reset(IncludeMacroStack.back().TheTokenLexer); 809 CurDirLookup = IncludeMacroStack.back().TheDirLookup; 810 IncludeMacroStack.pop_back(); 811 } 812 813 /// AllocateMacroInfo - Allocate a new MacroInfo object. 814 MacroInfo *AllocateMacroInfo(); 815 816 /// ReleaseMacroInfo - Release the specified MacroInfo. This memory will 817 /// be reused for allocating new MacroInfo objects. 818 void ReleaseMacroInfo(MacroInfo* MI); 819 820 /// ReadMacroName - Lex and validate a macro name, which occurs after a 821 /// #define or #undef. This emits a diagnostic, sets the token kind to eom, 822 /// and discards the rest of the macro line if the macro name is invalid. 823 void ReadMacroName(Token &MacroNameTok, char isDefineUndef = 0); 824 825 /// ReadMacroDefinitionArgList - The ( starting an argument list of a macro 826 /// definition has just been read. Lex the rest of the arguments and the 827 /// closing ), updating MI with what we learn. Return true if an error occurs 828 /// parsing the arg list. 829 bool ReadMacroDefinitionArgList(MacroInfo *MI); 830 831 /// SkipExcludedConditionalBlock - We just read a #if or related directive and 832 /// decided that the subsequent tokens are in the #if'd out portion of the 833 /// file. Lex the rest of the file, until we see an #endif. If 834 /// FoundNonSkipPortion is true, then we have already emitted code for part of 835 /// this #if directive, so #else/#elif blocks should never be entered. If 836 /// FoundElse is false, then #else directives are ok, if not, then we have 837 /// already seen one so a #else directive is a duplicate. When this returns, 838 /// the caller can lex the first valid token. 839 void SkipExcludedConditionalBlock(SourceLocation IfTokenLoc, 840 bool FoundNonSkipPortion, bool FoundElse); 841 842 /// PTHSkipExcludedConditionalBlock - A fast PTH version of 843 /// SkipExcludedConditionalBlock. 844 void PTHSkipExcludedConditionalBlock(); 845 846 /// EvaluateDirectiveExpression - Evaluate an integer constant expression that 847 /// may occur after a #if or #elif directive and return it as a bool. If the 848 /// expression is equivalent to "!defined(X)" return X in IfNDefMacro. 849 bool EvaluateDirectiveExpression(IdentifierInfo *&IfNDefMacro); 850 851 /// RegisterBuiltinPragmas - Install the standard preprocessor pragmas: 852 /// #pragma GCC poison/system_header/dependency and #pragma once. 853 void RegisterBuiltinPragmas(); 854 855 /// RegisterBuiltinMacros - Register builtin macros, such as __LINE__ with the 856 /// identifier table. 857 void RegisterBuiltinMacros(); 858 859 /// HandleMacroExpandedIdentifier - If an identifier token is read that is to 860 /// be expanded as a macro, handle it and return the next token as 'Tok'. If 861 /// the macro should not be expanded return true, otherwise return false. 862 bool HandleMacroExpandedIdentifier(Token &Tok, MacroInfo *MI); 863 864 /// isNextPPTokenLParen - Determine whether the next preprocessor token to be 865 /// lexed is a '('. If so, consume the token and return true, if not, this 866 /// method should have no observable side-effect on the lexed tokens. 867 bool isNextPPTokenLParen(); 868 869 /// ReadFunctionLikeMacroArgs - After reading "MACRO(", this method is 870 /// invoked to read all of the formal arguments specified for the macro 871 /// invocation. This returns null on error. 872 MacroArgs *ReadFunctionLikeMacroArgs(Token &MacroName, MacroInfo *MI, 873 SourceLocation &InstantiationEnd); 874 875 /// ExpandBuiltinMacro - If an identifier token is read that is to be expanded 876 /// as a builtin macro, handle it and return the next token as 'Tok'. 877 void ExpandBuiltinMacro(Token &Tok); 878 879 /// Handle_Pragma - Read a _Pragma directive, slice it up, process it, then 880 /// return the first token after the directive. The _Pragma token has just 881 /// been read into 'Tok'. 882 void Handle_Pragma(Token &Tok); 883 884 /// EnterSourceFileWithLexer - Add a lexer to the top of the include stack and 885 /// start lexing tokens from it instead of the current buffer. 886 void EnterSourceFileWithLexer(Lexer *TheLexer, const DirectoryLookup *Dir); 887 888 /// EnterSourceFileWithPTH - Add a lexer to the top of the include stack and 889 /// start getting tokens from it using the PTH cache. 890 void EnterSourceFileWithPTH(PTHLexer *PL, const DirectoryLookup *Dir); 891 892 /// IsFileLexer - Returns true if we are lexing from a file and not a 893 /// pragma or a macro. 894 static bool IsFileLexer(const Lexer* L, const PreprocessorLexer* P) { 895 return L ? !L->isPragmaLexer() : P != 0; 896 } 897 898 static bool IsFileLexer(const IncludeStackInfo& I) { 899 return IsFileLexer(I.TheLexer, I.ThePPLexer); 900 } 901 902 bool IsFileLexer() const { 903 return IsFileLexer(CurLexer.get(), CurPPLexer); 904 } 905 906 //===--------------------------------------------------------------------===// 907 // Caching stuff. 908 void CachingLex(Token &Result); 909 bool InCachingLexMode() const { 910 // If the Lexer pointers are 0 and IncludeMacroStack is empty, it means 911 // that we are past EOF, not that we are in CachingLex mode. 912 return CurPPLexer == 0 && CurTokenLexer == 0 && CurPTHLexer == 0 && 913 !IncludeMacroStack.empty(); 914 } 915 void EnterCachingLexMode(); 916 void ExitCachingLexMode() { 917 if (InCachingLexMode()) 918 RemoveTopOfLexerStack(); 919 } 920 const Token &PeekAhead(unsigned N); 921 void AnnotatePreviousCachedTokens(const Token &Tok); 922 923 //===--------------------------------------------------------------------===// 924 /// Handle*Directive - implement the various preprocessor directives. These 925 /// should side-effect the current preprocessor object so that the next call 926 /// to Lex() will return the appropriate token next. 927 void HandleLineDirective(Token &Tok); 928 void HandleDigitDirective(Token &Tok); 929 void HandleUserDiagnosticDirective(Token &Tok, bool isWarning); 930 void HandleIdentSCCSDirective(Token &Tok); 931 932 // File inclusion. 933 void HandleIncludeDirective(Token &Tok, 934 const DirectoryLookup *LookupFrom = 0, 935 bool isImport = false); 936 void HandleIncludeNextDirective(Token &Tok); 937 void HandleIncludeMacrosDirective(Token &Tok); 938 void HandleImportDirective(Token &Tok); 939 940 // Macro handling. 941 void HandleDefineDirective(Token &Tok); 942 void HandleUndefDirective(Token &Tok); 943 944 // Conditional Inclusion. 945 void HandleIfdefDirective(Token &Tok, bool isIfndef, 946 bool ReadAnyTokensBeforeDirective); 947 void HandleIfDirective(Token &Tok, bool ReadAnyTokensBeforeDirective); 948 void HandleEndifDirective(Token &Tok); 949 void HandleElseDirective(Token &Tok); 950 void HandleElifDirective(Token &Tok); 951 952 // Pragmas. 953 void HandlePragmaDirective(); 954public: 955 void HandlePragmaOnce(Token &OnceTok); 956 void HandlePragmaMark(); 957 void HandlePragmaPoison(Token &PoisonTok); 958 void HandlePragmaSystemHeader(Token &SysHeaderTok); 959 void HandlePragmaDependency(Token &DependencyTok); 960 void HandlePragmaComment(Token &CommentTok); 961 void HandlePragmaMessage(Token &MessageTok); 962 void HandlePragmaPushMacro(Token &Tok); 963 void HandlePragmaPopMacro(Token &Tok); 964 IdentifierInfo *ParsePragmaPushOrPopMacro(Token &Tok); 965 966 // Return true and store the first token only if any CommentHandler 967 // has inserted some tokens and getCommentRetentionState() is false. 968 bool HandleComment(Token &Token, SourceRange Comment); 969}; 970 971/// \brief Abstract base class that describes a handler that will receive 972/// source ranges for each of the comments encountered in the source file. 973class CommentHandler { 974public: 975 virtual ~CommentHandler(); 976 977 // The handler shall return true if it has pushed any tokens 978 // to be read using e.g. EnterToken or EnterTokenStream. 979 virtual bool HandleComment(Preprocessor &PP, SourceRange Comment) = 0; 980}; 981 982} // end namespace clang 983 984#endif 985