ASTUnit.h revision 03013fa9a0bf1ef4b907f5fec006c8f4000fdd21
1//===--- ASTUnit.h - ASTUnit utility ----------------------------*- 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// ASTUnit utility class. 11// 12//===----------------------------------------------------------------------===// 13 14#ifndef LLVM_CLANG_FRONTEND_ASTUNIT_H 15#define LLVM_CLANG_FRONTEND_ASTUNIT_H 16 17#include "clang/Index/ASTLocation.h" 18#include "clang/Serialization/ASTBitCodes.h" 19#include "clang/Sema/Sema.h" 20#include "clang/Sema/CodeCompleteConsumer.h" 21#include "clang/Lex/PreprocessingRecord.h" 22#include "clang/Basic/SourceManager.h" 23#include "clang/Basic/FileManager.h" 24#include "clang/Basic/FileSystemOptions.h" 25#include "clang-c/Index.h" 26#include "llvm/ADT/IntrusiveRefCntPtr.h" 27#include "llvm/ADT/OwningPtr.h" 28#include "llvm/ADT/SmallVector.h" 29#include "llvm/ADT/StringMap.h" 30#include "llvm/Support/Path.h" 31#include <map> 32#include <string> 33#include <vector> 34#include <cassert> 35#include <utility> 36#include <sys/types.h> 37 38namespace llvm { 39 class MemoryBuffer; 40} 41 42namespace clang { 43class ASTContext; 44class CodeCompleteConsumer; 45class CompilerInvocation; 46class Decl; 47class Diagnostic; 48class FileEntry; 49class FileManager; 50class HeaderSearch; 51class Preprocessor; 52class SourceManager; 53class TargetInfo; 54 55using namespace idx; 56 57/// \brief Utility class for loading a ASTContext from an AST file. 58/// 59class ASTUnit { 60public: 61 typedef std::map<FileID, std::vector<PreprocessedEntity *> > 62 PreprocessedEntitiesByFileMap; 63 64private: 65 llvm::IntrusiveRefCntPtr<Diagnostic> Diagnostics; 66 llvm::OwningPtr<FileManager> FileMgr; 67 llvm::OwningPtr<SourceManager> SourceMgr; 68 llvm::OwningPtr<HeaderSearch> HeaderInfo; 69 llvm::OwningPtr<TargetInfo> Target; 70 llvm::OwningPtr<Preprocessor> PP; 71 llvm::OwningPtr<ASTContext> Ctx; 72 73 FileSystemOptions FileSystemOpts; 74 75 /// \brief The AST consumer that received information about the translation 76 /// unit as it was parsed or loaded. 77 llvm::OwningPtr<ASTConsumer> Consumer; 78 79 /// \brief The semantic analysis object used to type-check the translation 80 /// unit. 81 llvm::OwningPtr<Sema> TheSema; 82 83 /// Optional owned invocation, just used to make the invocation used in 84 /// LoadFromCommandLine available. 85 llvm::OwningPtr<CompilerInvocation> Invocation; 86 87 // OnlyLocalDecls - when true, walking this AST should only visit declarations 88 // that come from the AST itself, not from included precompiled headers. 89 // FIXME: This is temporary; eventually, CIndex will always do this. 90 bool OnlyLocalDecls; 91 92 /// \brief Whether to capture any diagnostics produced. 93 bool CaptureDiagnostics; 94 95 /// \brief Track whether the main file was loaded from an AST or not. 96 bool MainFileIsAST; 97 98 /// \brief Whether this AST represents a complete translation unit. 99 bool CompleteTranslationUnit; 100 101 /// \brief Whether we should time each operation. 102 bool WantTiming; 103 104 /// Track the top-level decls which appeared in an ASTUnit which was loaded 105 /// from a source file. 106 // 107 // FIXME: This is just an optimization hack to avoid deserializing large parts 108 // of a PCH file when using the Index library on an ASTUnit loaded from 109 // source. In the long term we should make the Index library use efficient and 110 // more scalable search mechanisms. 111 std::vector<Decl*> TopLevelDecls; 112 113 /// The name of the original source file used to generate this ASTUnit. 114 std::string OriginalSourceFile; 115 116 // Critical optimization when using clang_getCursor(). 117 ASTLocation LastLoc; 118 119 /// \brief The set of diagnostics produced when creating this 120 /// translation unit. 121 llvm::SmallVector<StoredDiagnostic, 4> StoredDiagnostics; 122 123 /// \brief The number of stored diagnostics that come from the driver 124 /// itself. 125 /// 126 /// Diagnostics that come from the driver are retained from one parse to 127 /// the next. 128 unsigned NumStoredDiagnosticsFromDriver; 129 130 /// \brief Temporary files that should be removed when the ASTUnit is 131 /// destroyed. 132 llvm::SmallVector<llvm::sys::Path, 4> TemporaryFiles; 133 134 /// \brief A mapping from file IDs to the set of preprocessed entities 135 /// stored in that file. 136 /// 137 /// FIXME: This is just an optimization hack to avoid searching through 138 /// many preprocessed entities during cursor traversal in the CIndex library. 139 /// Ideally, we would just be able to perform a binary search within the 140 /// list of preprocessed entities. 141 PreprocessedEntitiesByFileMap PreprocessedEntitiesByFile; 142 143 /// \brief Simple hack to allow us to assert that ASTUnit is not being 144 /// used concurrently, which is not supported. 145 /// 146 /// Clients should create instances of the ConcurrencyCheck class whenever 147 /// using the ASTUnit in a way that isn't intended to be concurrent, which is 148 /// just about any usage. 149 unsigned int ConcurrencyCheckValue; 150 static const unsigned int CheckLocked = 28573289; 151 static const unsigned int CheckUnlocked = 9803453; 152 153 /// \brief Counter that determines when we want to try building a 154 /// precompiled preamble. 155 /// 156 /// If zero, we will never build a precompiled preamble. Otherwise, 157 /// it's treated as a counter that decrements each time we reparse 158 /// without the benefit of a precompiled preamble. When it hits 1, 159 /// we'll attempt to rebuild the precompiled header. This way, if 160 /// building the precompiled preamble fails, we won't try again for 161 /// some number of calls. 162 unsigned PreambleRebuildCounter; 163 164 /// \brief The file in which the precompiled preamble is stored. 165 std::string PreambleFile; 166 167 /// \brief The contents of the preamble that has been precompiled to 168 /// \c PreambleFile. 169 std::vector<char> Preamble; 170 171 /// \brief Whether the preamble ends at the start of a new line. 172 /// 173 /// Used to inform the lexer as to whether it's starting at the beginning of 174 /// a line after skipping the preamble. 175 bool PreambleEndsAtStartOfLine; 176 177 /// \brief The size of the source buffer that we've reserved for the main 178 /// file within the precompiled preamble. 179 unsigned PreambleReservedSize; 180 181 /// \brief Keeps track of the files that were used when computing the 182 /// preamble, with both their buffer size and their modification time. 183 /// 184 /// If any of the files have changed from one compile to the next, 185 /// the preamble must be thrown away. 186 llvm::StringMap<std::pair<off_t, time_t> > FilesInPreamble; 187 188 /// \brief When non-NULL, this is the buffer used to store the contents of 189 /// the main file when it has been padded for use with the precompiled 190 /// preamble. 191 llvm::MemoryBuffer *SavedMainFileBuffer; 192 193 /// \brief When non-NULL, this is the buffer used to store the 194 /// contents of the preamble when it has been padded to build the 195 /// precompiled preamble. 196 llvm::MemoryBuffer *PreambleBuffer; 197 198 /// \brief The number of warnings that occurred while parsing the preamble. 199 /// 200 /// This value will be used to restore the state of the \c Diagnostic object 201 /// when re-using the precompiled preamble. Note that only the 202 /// number of warnings matters, since we will not save the preamble 203 /// when any errors are present. 204 unsigned NumWarningsInPreamble; 205 206 /// \brief The number of diagnostics that were stored when parsing 207 /// the precompiled preamble. 208 /// 209 /// This value is used to determine how many of the stored 210 /// diagnostics should be retained when reparsing in the presence of 211 /// a precompiled preamble. 212 unsigned NumStoredDiagnosticsInPreamble; 213 214 /// \brief A list of the serialization ID numbers for each of the top-level 215 /// declarations parsed within the precompiled preamble. 216 std::vector<serialization::DeclID> TopLevelDeclsInPreamble; 217 218 /// \brief Whether we should be caching code-completion results. 219 bool ShouldCacheCodeCompletionResults; 220 221 static void ConfigureDiags(llvm::IntrusiveRefCntPtr<Diagnostic> &Diags, 222 ASTUnit &AST, bool CaptureDiagnostics); 223 224public: 225 /// \brief A cached code-completion result, which may be introduced in one of 226 /// many different contexts. 227 struct CachedCodeCompletionResult { 228 /// \brief The code-completion string corresponding to this completion 229 /// result. 230 CodeCompletionString *Completion; 231 232 /// \brief A bitmask that indicates which code-completion contexts should 233 /// contain this completion result. 234 /// 235 /// The bits in the bitmask correspond to the values of 236 /// CodeCompleteContext::Kind. To map from a completion context kind to a 237 /// bit, subtract one from the completion context kind and shift 1 by that 238 /// number of bits. Many completions can occur in several different 239 /// contexts. 240 unsigned ShowInContexts; 241 242 /// \brief The priority given to this code-completion result. 243 unsigned Priority; 244 245 /// \brief The libclang cursor kind corresponding to this code-completion 246 /// result. 247 CXCursorKind Kind; 248 249 /// \brief The availability of this code-completion result. 250 CXAvailabilityKind Availability; 251 252 /// \brief The simplified type class for a non-macro completion result. 253 SimplifiedTypeClass TypeClass; 254 255 /// \brief The type of a non-macro completion result, stored as a unique 256 /// integer used by the string map of cached completion types. 257 /// 258 /// This value will be zero if the type is not known, or a unique value 259 /// determined by the formatted type string. Se \c CachedCompletionTypes 260 /// for more information. 261 unsigned Type; 262 }; 263 264 /// \brief Retrieve the mapping from formatted type names to unique type 265 /// identifiers. 266 llvm::StringMap<unsigned> &getCachedCompletionTypes() { 267 return CachedCompletionTypes; 268 } 269 270private: 271 /// \brief The set of cached code-completion results. 272 std::vector<CachedCodeCompletionResult> CachedCompletionResults; 273 274 /// \brief A mapping from the formatted type name to a unique number for that 275 /// type, which is used for type equality comparisons. 276 llvm::StringMap<unsigned> CachedCompletionTypes; 277 278 /// \brief The number of top-level declarations present the last time we 279 /// cached code-completion results. 280 /// 281 /// The value is used to help detect when we should repopulate the global 282 /// completion cache. 283 unsigned NumTopLevelDeclsAtLastCompletionCache; 284 285 /// \brief The number of reparses left until we'll consider updating the 286 /// code-completion cache. 287 /// 288 /// This is meant to avoid thrashing during reparsing, by not allowing the 289 /// code-completion cache to be updated on every reparse. 290 unsigned CacheCodeCompletionCoolDown; 291 292 /// \brief Bit used by CIndex to mark when a translation unit may be in an 293 /// inconsistent state, and is not safe to free. 294 unsigned UnsafeToFree : 1; 295 296 /// \brief Cache any "global" code-completion results, so that we can avoid 297 /// recomputing them with each completion. 298 void CacheCodeCompletionResults(); 299 300 /// \brief Clear out and deallocate 301 void ClearCachedCompletionResults(); 302 303 ASTUnit(const ASTUnit&); // DO NOT IMPLEMENT 304 ASTUnit &operator=(const ASTUnit &); // DO NOT IMPLEMENT 305 306 explicit ASTUnit(bool MainFileIsAST); 307 308 void CleanTemporaryFiles(); 309 bool Parse(llvm::MemoryBuffer *OverrideMainBuffer); 310 311 std::pair<llvm::MemoryBuffer *, std::pair<unsigned, bool> > 312 ComputePreamble(CompilerInvocation &Invocation, 313 unsigned MaxLines, bool &CreatedBuffer); 314 315 llvm::MemoryBuffer *getMainBufferWithPrecompiledPreamble( 316 CompilerInvocation PreambleInvocation, 317 bool AllowRebuild = true, 318 unsigned MaxLines = 0); 319 void RealizeTopLevelDeclsFromPreamble(); 320 321public: 322 class ConcurrencyCheck { 323 volatile ASTUnit &Self; 324 325 public: 326 explicit ConcurrencyCheck(ASTUnit &Self) 327 : Self(Self) 328 { 329 assert(Self.ConcurrencyCheckValue == CheckUnlocked && 330 "Concurrent access to ASTUnit!"); 331 Self.ConcurrencyCheckValue = CheckLocked; 332 } 333 334 ~ConcurrencyCheck() { 335 Self.ConcurrencyCheckValue = CheckUnlocked; 336 } 337 }; 338 friend class ConcurrencyCheck; 339 340 ~ASTUnit(); 341 342 bool isMainFileAST() const { return MainFileIsAST; } 343 344 bool isUnsafeToFree() const { return UnsafeToFree; } 345 void setUnsafeToFree(bool Value) { UnsafeToFree = Value; } 346 347 const Diagnostic &getDiagnostics() const { return *Diagnostics; } 348 Diagnostic &getDiagnostics() { return *Diagnostics; } 349 350 const SourceManager &getSourceManager() const { return *SourceMgr; } 351 SourceManager &getSourceManager() { return *SourceMgr; } 352 353 const Preprocessor &getPreprocessor() const { return *PP.get(); } 354 Preprocessor &getPreprocessor() { return *PP.get(); } 355 356 const ASTContext &getASTContext() const { return *Ctx.get(); } 357 ASTContext &getASTContext() { return *Ctx.get(); } 358 359 bool hasSema() const { return TheSema; } 360 Sema &getSema() const { 361 assert(TheSema && "ASTUnit does not have a Sema object!"); 362 return *TheSema; 363 } 364 365 const FileManager &getFileManager() const { return *FileMgr; } 366 FileManager &getFileManager() { return *FileMgr; } 367 368 const FileSystemOptions &getFileSystemOpts() const { return FileSystemOpts; } 369 370 const std::string &getOriginalSourceFileName(); 371 const std::string &getASTFileName(); 372 373 /// \brief Add a temporary file that the ASTUnit depends on. 374 /// 375 /// This file will be erased when the ASTUnit is destroyed. 376 void addTemporaryFile(const llvm::sys::Path &TempFile) { 377 TemporaryFiles.push_back(TempFile); 378 } 379 380 bool getOnlyLocalDecls() const { return OnlyLocalDecls; } 381 382 /// \brief Retrieve the maximum PCH level of declarations that a 383 /// traversal of the translation unit should consider. 384 unsigned getMaxPCHLevel() const; 385 386 void setLastASTLocation(ASTLocation ALoc) { LastLoc = ALoc; } 387 ASTLocation getLastASTLocation() const { return LastLoc; } 388 389 390 llvm::StringRef getMainFileName() const; 391 392 typedef std::vector<Decl *>::iterator top_level_iterator; 393 394 top_level_iterator top_level_begin() { 395 assert(!isMainFileAST() && "Invalid call for AST based ASTUnit!"); 396 if (!TopLevelDeclsInPreamble.empty()) 397 RealizeTopLevelDeclsFromPreamble(); 398 return TopLevelDecls.begin(); 399 } 400 401 top_level_iterator top_level_end() { 402 assert(!isMainFileAST() && "Invalid call for AST based ASTUnit!"); 403 if (!TopLevelDeclsInPreamble.empty()) 404 RealizeTopLevelDeclsFromPreamble(); 405 return TopLevelDecls.end(); 406 } 407 408 std::size_t top_level_size() const { 409 assert(!isMainFileAST() && "Invalid call for AST based ASTUnit!"); 410 return TopLevelDeclsInPreamble.size() + TopLevelDecls.size(); 411 } 412 413 bool top_level_empty() const { 414 assert(!isMainFileAST() && "Invalid call for AST based ASTUnit!"); 415 return TopLevelDeclsInPreamble.empty() && TopLevelDecls.empty(); 416 } 417 418 /// \brief Add a new top-level declaration. 419 void addTopLevelDecl(Decl *D) { 420 TopLevelDecls.push_back(D); 421 } 422 423 /// \brief Add a new top-level declaration, identified by its ID in 424 /// the precompiled preamble. 425 void addTopLevelDeclFromPreamble(serialization::DeclID D) { 426 TopLevelDeclsInPreamble.push_back(D); 427 } 428 429 /// \brief Retrieve the mapping from File IDs to the preprocessed entities 430 /// within that file. 431 PreprocessedEntitiesByFileMap &getPreprocessedEntitiesByFile() { 432 return PreprocessedEntitiesByFile; 433 } 434 435 // Retrieve the diagnostics associated with this AST 436 typedef const StoredDiagnostic *stored_diag_iterator; 437 stored_diag_iterator stored_diag_begin() const { 438 return StoredDiagnostics.begin(); 439 } 440 stored_diag_iterator stored_diag_end() const { 441 return StoredDiagnostics.end(); 442 } 443 unsigned stored_diag_size() const { return StoredDiagnostics.size(); } 444 445 llvm::SmallVector<StoredDiagnostic, 4> &getStoredDiagnostics() { 446 return StoredDiagnostics; 447 } 448 449 typedef std::vector<CachedCodeCompletionResult>::iterator 450 cached_completion_iterator; 451 452 cached_completion_iterator cached_completion_begin() { 453 return CachedCompletionResults.begin(); 454 } 455 456 cached_completion_iterator cached_completion_end() { 457 return CachedCompletionResults.end(); 458 } 459 460 unsigned cached_completion_size() const { 461 return CachedCompletionResults.size(); 462 } 463 464 llvm::MemoryBuffer *getBufferForFile(llvm::StringRef Filename, 465 std::string *ErrorStr = 0); 466 467 /// \brief Whether this AST represents a complete translation unit. 468 /// 469 /// If false, this AST is only a partial translation unit, e.g., one 470 /// that might still be used as a precompiled header or preamble. 471 bool isCompleteTranslationUnit() const { return CompleteTranslationUnit; } 472 473 /// \brief A mapping from a file name to the memory buffer that stores the 474 /// remapped contents of that file. 475 typedef std::pair<std::string, const llvm::MemoryBuffer *> RemappedFile; 476 477 /// \brief Create a ASTUnit from an AST file. 478 /// 479 /// \param Filename - The AST file to load. 480 /// 481 /// \param Diags - The diagnostics engine to use for reporting errors; its 482 /// lifetime is expected to extend past that of the returned ASTUnit. 483 /// 484 /// \returns - The initialized ASTUnit or null if the AST failed to load. 485 static ASTUnit *LoadFromASTFile(const std::string &Filename, 486 llvm::IntrusiveRefCntPtr<Diagnostic> Diags, 487 const FileSystemOptions &FileSystemOpts, 488 bool OnlyLocalDecls = false, 489 RemappedFile *RemappedFiles = 0, 490 unsigned NumRemappedFiles = 0, 491 bool CaptureDiagnostics = false); 492 493private: 494 /// \brief Helper function for \c LoadFromCompilerInvocation() and 495 /// \c LoadFromCommandLine(), which loads an AST from a compiler invocation. 496 /// 497 /// \param PrecompilePreamble Whether to precompile the preamble of this 498 /// translation unit, to improve the performance of reparsing. 499 /// 500 /// \returns \c true if a catastrophic failure occurred (which means that the 501 /// \c ASTUnit itself is invalid), or \c false otherwise. 502 bool LoadFromCompilerInvocation(bool PrecompilePreamble); 503 504public: 505 506 /// LoadFromCompilerInvocation - Create an ASTUnit from a source file, via a 507 /// CompilerInvocation object. 508 /// 509 /// \param CI - The compiler invocation to use; it must have exactly one input 510 /// source file. The ASTUnit takes ownership of the CompilerInvocation object. 511 /// 512 /// \param Diags - The diagnostics engine to use for reporting errors; its 513 /// lifetime is expected to extend past that of the returned ASTUnit. 514 // 515 // FIXME: Move OnlyLocalDecls, UseBumpAllocator to setters on the ASTUnit, we 516 // shouldn't need to specify them at construction time. 517 static ASTUnit *LoadFromCompilerInvocation(CompilerInvocation *CI, 518 llvm::IntrusiveRefCntPtr<Diagnostic> Diags, 519 bool OnlyLocalDecls = false, 520 bool CaptureDiagnostics = false, 521 bool PrecompilePreamble = false, 522 bool CompleteTranslationUnit = true, 523 bool CacheCodeCompletionResults = false); 524 525 /// LoadFromCommandLine - Create an ASTUnit from a vector of command line 526 /// arguments, which must specify exactly one source file. 527 /// 528 /// \param ArgBegin - The beginning of the argument vector. 529 /// 530 /// \param ArgEnd - The end of the argument vector. 531 /// 532 /// \param Diags - The diagnostics engine to use for reporting errors; its 533 /// lifetime is expected to extend past that of the returned ASTUnit. 534 /// 535 /// \param ResourceFilesPath - The path to the compiler resource files. 536 // 537 // FIXME: Move OnlyLocalDecls, UseBumpAllocator to setters on the ASTUnit, we 538 // shouldn't need to specify them at construction time. 539 static ASTUnit *LoadFromCommandLine(const char **ArgBegin, 540 const char **ArgEnd, 541 llvm::IntrusiveRefCntPtr<Diagnostic> Diags, 542 llvm::StringRef ResourceFilesPath, 543 bool OnlyLocalDecls = false, 544 bool CaptureDiagnostics = false, 545 RemappedFile *RemappedFiles = 0, 546 unsigned NumRemappedFiles = 0, 547 bool PrecompilePreamble = false, 548 bool CompleteTranslationUnit = true, 549 bool CacheCodeCompletionResults = false, 550 bool CXXPrecompilePreamble = false, 551 bool CXXChainedPCH = false); 552 553 /// \brief Reparse the source files using the same command-line options that 554 /// were originally used to produce this translation unit. 555 /// 556 /// \returns True if a failure occurred that causes the ASTUnit not to 557 /// contain any translation-unit information, false otherwise. 558 bool Reparse(RemappedFile *RemappedFiles = 0, 559 unsigned NumRemappedFiles = 0); 560 561 /// \brief Perform code completion at the given file, line, and 562 /// column within this translation unit. 563 /// 564 /// \param File The file in which code completion will occur. 565 /// 566 /// \param Line The line at which code completion will occur. 567 /// 568 /// \param Column The column at which code completion will occur. 569 /// 570 /// \param IncludeMacros Whether to include macros in the code-completion 571 /// results. 572 /// 573 /// \param IncludeCodePatterns Whether to include code patterns (such as a 574 /// for loop) in the code-completion results. 575 /// 576 /// FIXME: The Diag, LangOpts, SourceMgr, FileMgr, StoredDiagnostics, and 577 /// OwnedBuffers parameters are all disgusting hacks. They will go away. 578 void CodeComplete(llvm::StringRef File, unsigned Line, unsigned Column, 579 RemappedFile *RemappedFiles, unsigned NumRemappedFiles, 580 bool IncludeMacros, bool IncludeCodePatterns, 581 CodeCompleteConsumer &Consumer, 582 Diagnostic &Diag, LangOptions &LangOpts, 583 SourceManager &SourceMgr, FileManager &FileMgr, 584 llvm::SmallVectorImpl<StoredDiagnostic> &StoredDiagnostics, 585 llvm::SmallVectorImpl<const llvm::MemoryBuffer *> &OwnedBuffers); 586 587 /// \brief Save this translation unit to a file with the given name. 588 /// 589 /// \returns True if an error occurred, false otherwise. 590 bool Save(llvm::StringRef File); 591}; 592 593} // namespace clang 594 595#endif 596