SourceManager.h revision e23ac65af568ffe611b0990818ac3a57c856a4d8
1//===--- SourceManager.h - Track and cache source files ---------*- 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 SourceManager interface. 11// 12//===----------------------------------------------------------------------===// 13 14#ifndef LLVM_CLANG_SOURCEMANAGER_H 15#define LLVM_CLANG_SOURCEMANAGER_H 16 17#include "clang/Basic/SourceLocation.h" 18#include "llvm/Support/Allocator.h" 19#include "llvm/Support/DataTypes.h" 20#include "llvm/ADT/PointerIntPair.h" 21#include "llvm/ADT/PointerUnion.h" 22#include "llvm/ADT/IntrusiveRefCntPtr.h" 23#include "llvm/ADT/DenseMap.h" 24#include <vector> 25#include <cassert> 26 27namespace llvm { 28class MemoryBuffer; 29class StringRef; 30} 31 32namespace clang { 33 34class Diagnostic; 35class SourceManager; 36class FileManager; 37class FileEntry; 38class LineTableInfo; 39 40/// SrcMgr - Public enums and private classes that are part of the 41/// SourceManager implementation. 42/// 43namespace SrcMgr { 44 /// CharacteristicKind - This is used to represent whether a file or directory 45 /// holds normal user code, system code, or system code which is implicitly 46 /// 'extern "C"' in C++ mode. Entire directories can be tagged with this 47 /// (this is maintained by DirectoryLookup and friends) as can specific 48 /// FileIDInfos when a #pragma system_header is seen or various other cases. 49 /// 50 enum CharacteristicKind { 51 C_User, C_System, C_ExternCSystem 52 }; 53 54 /// ContentCache - One instance of this struct is kept for every file 55 /// loaded or used. This object owns the MemoryBuffer object. 56 class ContentCache { 57 enum CCFlags { 58 /// \brief Whether the buffer is invalid. 59 InvalidFlag = 0x01, 60 /// \brief Whether the buffer should not be freed on destruction. 61 DoNotFreeFlag = 0x02 62 }; 63 64 /// Buffer - The actual buffer containing the characters from the input 65 /// file. This is owned by the ContentCache object. 66 /// The bits indicate indicates whether the buffer is invalid. 67 mutable llvm::PointerIntPair<const llvm::MemoryBuffer *, 2> Buffer; 68 69 public: 70 /// Reference to the file entry representing this ContentCache. 71 /// This reference does not own the FileEntry object. 72 /// It is possible for this to be NULL if 73 /// the ContentCache encapsulates an imaginary text buffer. 74 const FileEntry *OrigEntry; 75 76 /// \brief References the file which the contents were actually loaded from. 77 /// Can be different from 'Entry' if we overridden the contents of one file 78 /// with the contents of another file. 79 const FileEntry *ContentsEntry; 80 81 /// SourceLineCache - A bump pointer allocated array of offsets for each 82 /// source line. This is lazily computed. This is owned by the 83 /// SourceManager BumpPointerAllocator object. 84 unsigned *SourceLineCache; 85 86 /// NumLines - The number of lines in this ContentCache. This is only valid 87 /// if SourceLineCache is non-null. 88 unsigned NumLines; 89 90 /// getBuffer - Returns the memory buffer for the associated content. 91 /// 92 /// \param Diag Object through which diagnostics will be emitted it the 93 /// buffer cannot be retrieved. 94 /// 95 /// \param Loc If specified, is the location that invalid file diagnostics 96 /// will be emitted at. 97 /// 98 /// \param Invalid If non-NULL, will be set \c true if an error occurred. 99 const llvm::MemoryBuffer *getBuffer(Diagnostic &Diag, 100 const SourceManager &SM, 101 SourceLocation Loc = SourceLocation(), 102 bool *Invalid = 0) const; 103 104 /// getSize - Returns the size of the content encapsulated by this 105 /// ContentCache. This can be the size of the source file or the size of an 106 /// arbitrary scratch buffer. If the ContentCache encapsulates a source 107 /// file this size is retrieved from the file's FileEntry. 108 unsigned getSize() const; 109 110 /// getSizeBytesMapped - Returns the number of bytes actually mapped for 111 /// this ContentCache. This can be 0 if the MemBuffer was not actually 112 /// instantiated. 113 unsigned getSizeBytesMapped() const; 114 115 void setBuffer(const llvm::MemoryBuffer *B) { 116 assert(!Buffer.getPointer() && "MemoryBuffer already set."); 117 Buffer.setPointer(B); 118 Buffer.setInt(false); 119 } 120 121 /// \brief Get the underlying buffer, returning NULL if the buffer is not 122 /// yet available. 123 const llvm::MemoryBuffer *getRawBuffer() const { 124 return Buffer.getPointer(); 125 } 126 127 /// \brief Replace the existing buffer (which will be deleted) 128 /// with the given buffer. 129 void replaceBuffer(const llvm::MemoryBuffer *B, bool DoNotFree = false); 130 131 /// \brief Determine whether the buffer itself is invalid. 132 bool isBufferInvalid() const { 133 return Buffer.getInt() & InvalidFlag; 134 } 135 136 /// \brief Determine whether the buffer should be freed. 137 bool shouldFreeBuffer() const { 138 return (Buffer.getInt() & DoNotFreeFlag) == 0; 139 } 140 141 ContentCache(const FileEntry *Ent = 0) 142 : Buffer(0, false), OrigEntry(Ent), ContentsEntry(Ent), 143 SourceLineCache(0), NumLines(0) {} 144 145 ContentCache(const FileEntry *Ent, const FileEntry *contentEnt) 146 : Buffer(0, false), OrigEntry(Ent), ContentsEntry(contentEnt), 147 SourceLineCache(0), NumLines(0) {} 148 149 ~ContentCache(); 150 151 /// The copy ctor does not allow copies where source object has either 152 /// a non-NULL Buffer or SourceLineCache. Ownership of allocated memory 153 /// is not transferred, so this is a logical error. 154 ContentCache(const ContentCache &RHS) 155 : Buffer(0, false), SourceLineCache(0) 156 { 157 OrigEntry = RHS.OrigEntry; 158 ContentsEntry = RHS.ContentsEntry; 159 160 assert (RHS.Buffer.getPointer() == 0 && RHS.SourceLineCache == 0 161 && "Passed ContentCache object cannot own a buffer."); 162 163 NumLines = RHS.NumLines; 164 } 165 166 private: 167 // Disable assignments. 168 ContentCache &operator=(const ContentCache& RHS); 169 }; 170 171 /// FileInfo - Information about a FileID, basically just the logical file 172 /// that it represents and include stack information. 173 /// 174 /// Each FileInfo has include stack information, indicating where it came 175 /// from. This information encodes the #include chain that a token was 176 /// instantiated from. The main include file has an invalid IncludeLoc. 177 /// 178 /// FileInfos contain a "ContentCache *", with the contents of the file. 179 /// 180 class FileInfo { 181 /// IncludeLoc - The location of the #include that brought in this file. 182 /// This is an invalid SLOC for the main file (top of the #include chain). 183 unsigned IncludeLoc; // Really a SourceLocation 184 185 /// Data - This contains the ContentCache* and the bits indicating the 186 /// characteristic of the file and whether it has #line info, all bitmangled 187 /// together. 188 uintptr_t Data; 189 public: 190 /// get - Return a FileInfo object. 191 static FileInfo get(SourceLocation IL, const ContentCache *Con, 192 CharacteristicKind FileCharacter) { 193 FileInfo X; 194 X.IncludeLoc = IL.getRawEncoding(); 195 X.Data = (uintptr_t)Con; 196 assert((X.Data & 7) == 0 &&"ContentCache pointer insufficiently aligned"); 197 assert((unsigned)FileCharacter < 4 && "invalid file character"); 198 X.Data |= (unsigned)FileCharacter; 199 return X; 200 } 201 202 SourceLocation getIncludeLoc() const { 203 return SourceLocation::getFromRawEncoding(IncludeLoc); 204 } 205 const ContentCache* getContentCache() const { 206 return reinterpret_cast<const ContentCache*>(Data & ~7UL); 207 } 208 209 /// getCharacteristic - Return whether this is a system header or not. 210 CharacteristicKind getFileCharacteristic() const { 211 return (CharacteristicKind)(Data & 3); 212 } 213 214 /// hasLineDirectives - Return true if this FileID has #line directives in 215 /// it. 216 bool hasLineDirectives() const { return (Data & 4) != 0; } 217 218 /// setHasLineDirectives - Set the flag that indicates that this FileID has 219 /// line table entries associated with it. 220 void setHasLineDirectives() { 221 Data |= 4; 222 } 223 }; 224 225 /// InstantiationInfo - Each InstantiationInfo encodes the Instantiation 226 /// location - where the token was ultimately instantiated, and the 227 /// SpellingLoc - where the actual character data for the token came from. 228 class InstantiationInfo { 229 // Really these are all SourceLocations. 230 231 /// SpellingLoc - Where the spelling for the token can be found. 232 unsigned SpellingLoc; 233 234 /// InstantiationLocStart/InstantiationLocEnd - In a macro expansion, these 235 /// indicate the start and end of the instantiation. In object-like macros, 236 /// these will be the same. In a function-like macro instantiation, the 237 /// start will be the identifier and the end will be the ')'. 238 unsigned InstantiationLocStart, InstantiationLocEnd; 239 public: 240 SourceLocation getSpellingLoc() const { 241 return SourceLocation::getFromRawEncoding(SpellingLoc); 242 } 243 SourceLocation getInstantiationLocStart() const { 244 return SourceLocation::getFromRawEncoding(InstantiationLocStart); 245 } 246 SourceLocation getInstantiationLocEnd() const { 247 return SourceLocation::getFromRawEncoding(InstantiationLocEnd); 248 } 249 250 std::pair<SourceLocation,SourceLocation> getInstantiationLocRange() const { 251 return std::make_pair(getInstantiationLocStart(), 252 getInstantiationLocEnd()); 253 } 254 255 /// get - Return a InstantiationInfo for an expansion. IL specifies 256 /// the instantiation location (where the macro is expanded), and SL 257 /// specifies the spelling location (where the characters from the token 258 /// come from). IL and PL can both refer to normal File SLocs or 259 /// instantiation locations. 260 static InstantiationInfo get(SourceLocation ILStart, SourceLocation ILEnd, 261 SourceLocation SL) { 262 InstantiationInfo X; 263 X.SpellingLoc = SL.getRawEncoding(); 264 X.InstantiationLocStart = ILStart.getRawEncoding(); 265 X.InstantiationLocEnd = ILEnd.getRawEncoding(); 266 return X; 267 } 268 }; 269 270 /// SLocEntry - This is a discriminated union of FileInfo and 271 /// InstantiationInfo. SourceManager keeps an array of these objects, and 272 /// they are uniquely identified by the FileID datatype. 273 class SLocEntry { 274 unsigned Offset; // low bit is set for instantiation info. 275 union { 276 FileInfo File; 277 InstantiationInfo Instantiation; 278 }; 279 public: 280 unsigned getOffset() const { return Offset >> 1; } 281 282 bool isInstantiation() const { return Offset & 1; } 283 bool isFile() const { return !isInstantiation(); } 284 285 const FileInfo &getFile() const { 286 assert(isFile() && "Not a file SLocEntry!"); 287 return File; 288 } 289 290 const InstantiationInfo &getInstantiation() const { 291 assert(isInstantiation() && "Not an instantiation SLocEntry!"); 292 return Instantiation; 293 } 294 295 static SLocEntry get(unsigned Offset, const FileInfo &FI) { 296 SLocEntry E; 297 E.Offset = Offset << 1; 298 E.File = FI; 299 return E; 300 } 301 302 static SLocEntry get(unsigned Offset, const InstantiationInfo &II) { 303 SLocEntry E; 304 E.Offset = (Offset << 1) | 1; 305 E.Instantiation = II; 306 return E; 307 } 308 }; 309} // end SrcMgr namespace. 310 311/// \brief External source of source location entries. 312class ExternalSLocEntrySource { 313public: 314 virtual ~ExternalSLocEntrySource(); 315 316 /// \brief Read the source location entry with index ID. 317 /// 318 /// \returns true if an error occurred that prevented the source-location 319 /// entry from being loaded. 320 virtual bool ReadSLocEntry(unsigned ID) = 0; 321}; 322 323 324/// IsBeforeInTranslationUnitCache - This class holds the cache used by 325/// isBeforeInTranslationUnit. The cache structure is complex enough to be 326/// worth breaking out of SourceManager. 327class IsBeforeInTranslationUnitCache { 328 /// L/R QueryFID - These are the FID's of the cached query. If these match up 329 /// with a subsequent query, the result can be reused. 330 FileID LQueryFID, RQueryFID; 331 332 /// CommonFID - This is the file found in common between the two #include 333 /// traces. It is the nearest common ancestor of the #include tree. 334 FileID CommonFID; 335 336 /// L/R CommonOffset - This is the offset of the previous query in CommonFID. 337 /// Usually, this represents the location of the #include for QueryFID, but if 338 /// LQueryFID is a parent of RQueryFID (or vise versa) then these can be a 339 /// random token in the parent. 340 unsigned LCommonOffset, RCommonOffset; 341public: 342 343 /// isCacheValid - Return true if the currently cached values match up with 344 /// the specified LHS/RHS query. If not, we can't use the cache. 345 bool isCacheValid(FileID LHS, FileID RHS) const { 346 return LQueryFID == LHS && RQueryFID == RHS; 347 } 348 349 /// getCachedResult - If the cache is valid, compute the result given the 350 /// specified offsets in the LHS/RHS FID's. 351 bool getCachedResult(unsigned LOffset, unsigned ROffset) const { 352 // If one of the query files is the common file, use the offset. Otherwise, 353 // use the #include loc in the common file. 354 if (LQueryFID != CommonFID) LOffset = LCommonOffset; 355 if (RQueryFID != CommonFID) ROffset = RCommonOffset; 356 return LOffset < ROffset; 357 } 358 359 // Set up a new query. 360 void setQueryFIDs(FileID LHS, FileID RHS) { 361 LQueryFID = LHS; 362 RQueryFID = RHS; 363 } 364 365 void setCommonLoc(FileID commonFID, unsigned lCommonOffset, 366 unsigned rCommonOffset) { 367 CommonFID = commonFID; 368 LCommonOffset = lCommonOffset; 369 RCommonOffset = rCommonOffset; 370 } 371 372}; 373 374/// SourceManager - This file handles loading and caching of source files into 375/// memory. This object owns the MemoryBuffer objects for all of the loaded 376/// files and assigns unique FileID's for each unique #include chain. 377/// 378/// The SourceManager can be queried for information about SourceLocation 379/// objects, turning them into either spelling or instantiation locations. 380/// Spelling locations represent where the bytes corresponding to a token came 381/// from and instantiation locations represent where the location is in the 382/// user's view. In the case of a macro expansion, for example, the spelling 383/// location indicates where the expanded token came from and the instantiation 384/// location specifies where it was expanded. 385class SourceManager : public llvm::RefCountedBase<SourceManager> { 386 /// \brief Diagnostic object. 387 Diagnostic &Diag; 388 389 FileManager &FileMgr; 390 391 mutable llvm::BumpPtrAllocator ContentCacheAlloc; 392 393 /// FileInfos - Memoized information about all of the files tracked by this 394 /// SourceManager. This set allows us to merge ContentCache entries based 395 /// on their FileEntry*. All ContentCache objects will thus have unique, 396 /// non-null, FileEntry pointers. 397 llvm::DenseMap<const FileEntry*, SrcMgr::ContentCache*> FileInfos; 398 399 /// \brief True if the ContentCache for files that are overriden by other 400 /// files, should report the original file name. Defaults to true. 401 bool OverridenFilesKeepOriginalName; 402 403 /// \brief Files that have been overriden with the contents from another file. 404 llvm::DenseMap<const FileEntry *, const FileEntry *> OverriddenFiles; 405 406 /// MemBufferInfos - Information about various memory buffers that we have 407 /// read in. All FileEntry* within the stored ContentCache objects are NULL, 408 /// as they do not refer to a file. 409 std::vector<SrcMgr::ContentCache*> MemBufferInfos; 410 411 /// SLocEntryTable - This is an array of SLocEntry's that we have created. 412 /// FileID is an index into this vector. This array is sorted by the offset. 413 std::vector<SrcMgr::SLocEntry> SLocEntryTable; 414 /// NextOffset - This is the next available offset that a new SLocEntry can 415 /// start at. It is SLocEntryTable.back().getOffset()+size of back() entry. 416 unsigned NextOffset; 417 418 /// \brief If source location entries are being lazily loaded from 419 /// an external source, this vector indicates whether the Ith source 420 /// location entry has already been loaded from the external storage. 421 std::vector<bool> SLocEntryLoaded; 422 423 /// \brief An external source for source location entries. 424 ExternalSLocEntrySource *ExternalSLocEntries; 425 426 /// LastFileIDLookup - This is a one-entry cache to speed up getFileID. 427 /// LastFileIDLookup records the last FileID looked up or created, because it 428 /// is very common to look up many tokens from the same file. 429 mutable FileID LastFileIDLookup; 430 431 /// LineTable - This holds information for #line directives. It is referenced 432 /// by indices from SLocEntryTable. 433 LineTableInfo *LineTable; 434 435 /// LastLineNo - These ivars serve as a cache used in the getLineNumber 436 /// method which is used to speedup getLineNumber calls to nearby locations. 437 mutable FileID LastLineNoFileIDQuery; 438 mutable SrcMgr::ContentCache *LastLineNoContentCache; 439 mutable unsigned LastLineNoFilePos; 440 mutable unsigned LastLineNoResult; 441 442 /// MainFileID - The file ID for the main source file of the translation unit. 443 FileID MainFileID; 444 445 // Statistics for -print-stats. 446 mutable unsigned NumLinearScans, NumBinaryProbes; 447 448 // Cache results for the isBeforeInTranslationUnit method. 449 mutable IsBeforeInTranslationUnitCache IsBeforeInTUCache; 450 451 // Cache for the "fake" buffer used for error-recovery purposes. 452 mutable llvm::MemoryBuffer *FakeBufferForRecovery; 453 454 // SourceManager doesn't support copy construction. 455 explicit SourceManager(const SourceManager&); 456 void operator=(const SourceManager&); 457public: 458 SourceManager(Diagnostic &Diag, FileManager &FileMgr); 459 ~SourceManager(); 460 461 void clearIDTables(); 462 463 Diagnostic &getDiagnostics() const { return Diag; } 464 465 FileManager &getFileManager() const { return FileMgr; } 466 467 /// \brief Set true if the SourceManager should report the original file name 468 /// for contents of files that were overriden by other files.Defaults to true. 469 void setOverridenFilesKeepOriginalName(bool value) { 470 OverridenFilesKeepOriginalName = value; 471 } 472 473 //===--------------------------------------------------------------------===// 474 // MainFileID creation and querying methods. 475 //===--------------------------------------------------------------------===// 476 477 /// getMainFileID - Returns the FileID of the main source file. 478 FileID getMainFileID() const { return MainFileID; } 479 480 /// createMainFileID - Create the FileID for the main source file. 481 FileID createMainFileID(const FileEntry *SourceFile) { 482 assert(MainFileID.isInvalid() && "MainFileID already set!"); 483 MainFileID = createFileID(SourceFile, SourceLocation(), SrcMgr::C_User); 484 return MainFileID; 485 } 486 487 /// \brief Set the file ID for the precompiled preamble, which is also the 488 /// main file. 489 void SetPreambleFileID(FileID Preamble) { 490 assert(MainFileID.isInvalid() && "MainFileID already set!"); 491 MainFileID = Preamble; 492 } 493 494 //===--------------------------------------------------------------------===// 495 // Methods to create new FileID's and instantiations. 496 //===--------------------------------------------------------------------===// 497 498 /// createFileID - Create a new FileID that represents the specified file 499 /// being #included from the specified IncludePosition. This returns 0 on 500 /// error and translates NULL into standard input. 501 /// PreallocateID should be non-zero to specify which pre-allocated, 502 /// lazily computed source location is being filled in by this operation. 503 FileID createFileID(const FileEntry *SourceFile, SourceLocation IncludePos, 504 SrcMgr::CharacteristicKind FileCharacter, 505 unsigned PreallocatedID = 0, 506 unsigned Offset = 0) { 507 const SrcMgr::ContentCache *IR = getOrCreateContentCache(SourceFile); 508 assert(IR && "getOrCreateContentCache() cannot return NULL"); 509 return createFileID(IR, IncludePos, FileCharacter, PreallocatedID, Offset); 510 } 511 512 /// createFileIDForMemBuffer - Create a new FileID that represents the 513 /// specified memory buffer. This does no caching of the buffer and takes 514 /// ownership of the MemoryBuffer, so only pass a MemoryBuffer to this once. 515 FileID createFileIDForMemBuffer(const llvm::MemoryBuffer *Buffer, 516 unsigned PreallocatedID = 0, 517 unsigned Offset = 0) { 518 return createFileID(createMemBufferContentCache(Buffer), SourceLocation(), 519 SrcMgr::C_User, PreallocatedID, Offset); 520 } 521 522 /// createMainFileIDForMembuffer - Create the FileID for a memory buffer 523 /// that will represent the FileID for the main source. One example 524 /// of when this would be used is when the main source is read from STDIN. 525 FileID createMainFileIDForMemBuffer(const llvm::MemoryBuffer *Buffer) { 526 assert(MainFileID.isInvalid() && "MainFileID already set!"); 527 MainFileID = createFileIDForMemBuffer(Buffer); 528 return MainFileID; 529 } 530 531 /// createInstantiationLoc - Return a new SourceLocation that encodes the fact 532 /// that a token at Loc should actually be referenced from InstantiationLoc. 533 /// TokLength is the length of the token being instantiated. 534 SourceLocation createInstantiationLoc(SourceLocation Loc, 535 SourceLocation InstantiationLocStart, 536 SourceLocation InstantiationLocEnd, 537 unsigned TokLength, 538 unsigned PreallocatedID = 0, 539 unsigned Offset = 0); 540 541 /// \brief Retrieve the memory buffer associated with the given file. 542 /// 543 /// \param Invalid If non-NULL, will be set \c true if an error 544 /// occurs while retrieving the memory buffer. 545 const llvm::MemoryBuffer *getMemoryBufferForFile(const FileEntry *File, 546 bool *Invalid = 0); 547 548 /// \brief Override the contents of the given source file by providing an 549 /// already-allocated buffer. 550 /// 551 /// \param SourceFile the source file whose contents will be overriden. 552 /// 553 /// \param Buffer the memory buffer whose contents will be used as the 554 /// data in the given source file. 555 /// 556 /// \param DoNotFree If true, then the buffer will not be freed when the 557 /// source manager is destroyed. 558 void overrideFileContents(const FileEntry *SourceFile, 559 const llvm::MemoryBuffer *Buffer, 560 bool DoNotFree = false); 561 562 /// \brief Override the the given source file with another one. 563 /// 564 /// \param SourceFile the source file which will be overriden. 565 /// 566 /// \param NewFile the file whose contents will be used as the 567 /// data instead of the contents of the given source file. 568 void overrideFileContents(const FileEntry *SourceFile, 569 const FileEntry *NewFile); 570 571 //===--------------------------------------------------------------------===// 572 // FileID manipulation methods. 573 //===--------------------------------------------------------------------===// 574 575 /// getBuffer - Return the buffer for the specified FileID. If there is an 576 /// error opening this buffer the first time, this manufactures a temporary 577 /// buffer and returns a non-empty error string. 578 const llvm::MemoryBuffer *getBuffer(FileID FID, SourceLocation Loc, 579 bool *Invalid = 0) const { 580 bool MyInvalid = false; 581 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid); 582 if (MyInvalid || !Entry.isFile()) { 583 if (Invalid) 584 *Invalid = true; 585 586 return getFakeBufferForRecovery(); 587 } 588 589 return Entry.getFile().getContentCache()->getBuffer(Diag, *this, Loc, 590 Invalid); 591 } 592 593 const llvm::MemoryBuffer *getBuffer(FileID FID, bool *Invalid = 0) const { 594 bool MyInvalid = false; 595 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid); 596 if (MyInvalid || !Entry.isFile()) { 597 if (Invalid) 598 *Invalid = true; 599 600 return getFakeBufferForRecovery(); 601 } 602 603 return Entry.getFile().getContentCache()->getBuffer(Diag, *this, 604 SourceLocation(), 605 Invalid); 606 } 607 608 /// getFileEntryForID - Returns the FileEntry record for the provided FileID. 609 const FileEntry *getFileEntryForID(FileID FID) const { 610 bool MyInvalid = false; 611 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid); 612 if (MyInvalid || !Entry.isFile()) 613 return 0; 614 615 return Entry.getFile().getContentCache()->OrigEntry; 616 } 617 618 /// Returns the FileEntry record for the provided SLocEntry. 619 const FileEntry *getFileEntryForSLocEntry(const SrcMgr::SLocEntry &sloc) const 620 { 621 return sloc.getFile().getContentCache()->OrigEntry; 622 } 623 624 /// getBufferData - Return a StringRef to the source buffer data for the 625 /// specified FileID. 626 /// 627 /// \param FID The file ID whose contents will be returned. 628 /// \param Invalid If non-NULL, will be set true if an error occurred. 629 llvm::StringRef getBufferData(FileID FID, bool *Invalid = 0) const; 630 631 632 //===--------------------------------------------------------------------===// 633 // SourceLocation manipulation methods. 634 //===--------------------------------------------------------------------===// 635 636 /// getFileID - Return the FileID for a SourceLocation. This is a very 637 /// hot method that is used for all SourceManager queries that start with a 638 /// SourceLocation object. It is responsible for finding the entry in 639 /// SLocEntryTable which contains the specified location. 640 /// 641 FileID getFileID(SourceLocation SpellingLoc) const { 642 unsigned SLocOffset = SpellingLoc.getOffset(); 643 644 // If our one-entry cache covers this offset, just return it. 645 if (isOffsetInFileID(LastFileIDLookup, SLocOffset)) 646 return LastFileIDLookup; 647 648 return getFileIDSlow(SLocOffset); 649 } 650 651 /// getLocForStartOfFile - Return the source location corresponding to the 652 /// first byte of the specified file. 653 SourceLocation getLocForStartOfFile(FileID FID) const { 654 assert(FID.ID < SLocEntryTable.size() && "FileID out of range"); 655 bool Invalid = false; 656 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid); 657 if (Invalid || !Entry.isFile()) 658 return SourceLocation(); 659 660 unsigned FileOffset = Entry.getOffset(); 661 return SourceLocation::getFileLoc(FileOffset); 662 } 663 664 /// getInstantiationLoc - Given a SourceLocation object, return the 665 /// instantiation location referenced by the ID. 666 SourceLocation getInstantiationLoc(SourceLocation Loc) const { 667 // Handle the non-mapped case inline, defer to out of line code to handle 668 // instantiations. 669 if (Loc.isFileID()) return Loc; 670 return getInstantiationLocSlowCase(Loc); 671 } 672 673 /// getImmediateInstantiationRange - Loc is required to be an instantiation 674 /// location. Return the start/end of the instantiation information. 675 std::pair<SourceLocation,SourceLocation> 676 getImmediateInstantiationRange(SourceLocation Loc) const; 677 678 /// getInstantiationRange - Given a SourceLocation object, return the 679 /// range of tokens covered by the instantiation in the ultimate file. 680 std::pair<SourceLocation,SourceLocation> 681 getInstantiationRange(SourceLocation Loc) const; 682 683 684 /// getSpellingLoc - Given a SourceLocation object, return the spelling 685 /// location referenced by the ID. This is the place where the characters 686 /// that make up the lexed token can be found. 687 SourceLocation getSpellingLoc(SourceLocation Loc) const { 688 // Handle the non-mapped case inline, defer to out of line code to handle 689 // instantiations. 690 if (Loc.isFileID()) return Loc; 691 return getSpellingLocSlowCase(Loc); 692 } 693 694 /// getImmediateSpellingLoc - Given a SourceLocation object, return the 695 /// spelling location referenced by the ID. This is the first level down 696 /// towards the place where the characters that make up the lexed token can be 697 /// found. This should not generally be used by clients. 698 SourceLocation getImmediateSpellingLoc(SourceLocation Loc) const; 699 700 /// getDecomposedLoc - Decompose the specified location into a raw FileID + 701 /// Offset pair. The first element is the FileID, the second is the 702 /// offset from the start of the buffer of the location. 703 std::pair<FileID, unsigned> getDecomposedLoc(SourceLocation Loc) const { 704 FileID FID = getFileID(Loc); 705 return std::make_pair(FID, Loc.getOffset()-getSLocEntry(FID).getOffset()); 706 } 707 708 /// getDecomposedInstantiationLoc - Decompose the specified location into a 709 /// raw FileID + Offset pair. If the location is an instantiation record, 710 /// walk through it until we find the final location instantiated. 711 std::pair<FileID, unsigned> 712 getDecomposedInstantiationLoc(SourceLocation Loc) const { 713 FileID FID = getFileID(Loc); 714 const SrcMgr::SLocEntry *E = &getSLocEntry(FID); 715 716 unsigned Offset = Loc.getOffset()-E->getOffset(); 717 if (Loc.isFileID()) 718 return std::make_pair(FID, Offset); 719 720 return getDecomposedInstantiationLocSlowCase(E, Offset); 721 } 722 723 /// getDecomposedSpellingLoc - Decompose the specified location into a raw 724 /// FileID + Offset pair. If the location is an instantiation record, walk 725 /// through it until we find its spelling record. 726 std::pair<FileID, unsigned> 727 getDecomposedSpellingLoc(SourceLocation Loc) const { 728 FileID FID = getFileID(Loc); 729 const SrcMgr::SLocEntry *E = &getSLocEntry(FID); 730 731 unsigned Offset = Loc.getOffset()-E->getOffset(); 732 if (Loc.isFileID()) 733 return std::make_pair(FID, Offset); 734 return getDecomposedSpellingLocSlowCase(E, Offset); 735 } 736 737 /// getFileOffset - This method returns the offset from the start 738 /// of the file that the specified SourceLocation represents. This is not very 739 /// meaningful for a macro ID. 740 unsigned getFileOffset(SourceLocation SpellingLoc) const { 741 return getDecomposedLoc(SpellingLoc).second; 742 } 743 744 745 //===--------------------------------------------------------------------===// 746 // Queries about the code at a SourceLocation. 747 //===--------------------------------------------------------------------===// 748 749 /// getCharacterData - Return a pointer to the start of the specified location 750 /// in the appropriate spelling MemoryBuffer. 751 /// 752 /// \param Invalid If non-NULL, will be set \c true if an error occurs. 753 const char *getCharacterData(SourceLocation SL, bool *Invalid = 0) const; 754 755 /// getColumnNumber - Return the column # for the specified file position. 756 /// This is significantly cheaper to compute than the line number. This 757 /// returns zero if the column number isn't known. This may only be called on 758 /// a file sloc, so you must choose a spelling or instantiation location 759 /// before calling this method. 760 unsigned getColumnNumber(FileID FID, unsigned FilePos, 761 bool *Invalid = 0) const; 762 unsigned getSpellingColumnNumber(SourceLocation Loc, bool *Invalid = 0) const; 763 unsigned getInstantiationColumnNumber(SourceLocation Loc, 764 bool *Invalid = 0) const; 765 unsigned getPresumedColumnNumber(SourceLocation Loc, bool *Invalid = 0) const; 766 767 768 /// getLineNumber - Given a SourceLocation, return the spelling line number 769 /// for the position indicated. This requires building and caching a table of 770 /// line offsets for the MemoryBuffer, so this is not cheap: use only when 771 /// about to emit a diagnostic. 772 unsigned getLineNumber(FileID FID, unsigned FilePos, bool *Invalid = 0) const; 773 unsigned getSpellingLineNumber(SourceLocation Loc, bool *Invalid = 0) const; 774 unsigned getInstantiationLineNumber(SourceLocation Loc, 775 bool *Invalid = 0) const; 776 unsigned getPresumedLineNumber(SourceLocation Loc, bool *Invalid = 0) const; 777 778 /// Return the filename or buffer identifier of the buffer the location is in. 779 /// Note that this name does not respect #line directives. Use getPresumedLoc 780 /// for normal clients. 781 const char *getBufferName(SourceLocation Loc, bool *Invalid = 0) const; 782 783 /// getFileCharacteristic - return the file characteristic of the specified 784 /// source location, indicating whether this is a normal file, a system 785 /// header, or an "implicit extern C" system header. 786 /// 787 /// This state can be modified with flags on GNU linemarker directives like: 788 /// # 4 "foo.h" 3 789 /// which changes all source locations in the current file after that to be 790 /// considered to be from a system header. 791 SrcMgr::CharacteristicKind getFileCharacteristic(SourceLocation Loc) const; 792 793 /// getPresumedLoc - This method returns the "presumed" location of a 794 /// SourceLocation specifies. A "presumed location" can be modified by #line 795 /// or GNU line marker directives. This provides a view on the data that a 796 /// user should see in diagnostics, for example. 797 /// 798 /// Note that a presumed location is always given as the instantiation point 799 /// of an instantiation location, not at the spelling location. 800 /// 801 /// \returns The presumed location of the specified SourceLocation. If the 802 /// presumed location cannot be calculate (e.g., because \p Loc is invalid 803 /// or the file containing \p Loc has changed on disk), returns an invalid 804 /// presumed location. 805 PresumedLoc getPresumedLoc(SourceLocation Loc) const; 806 807 /// isFromSameFile - Returns true if both SourceLocations correspond to 808 /// the same file. 809 bool isFromSameFile(SourceLocation Loc1, SourceLocation Loc2) const { 810 return getFileID(Loc1) == getFileID(Loc2); 811 } 812 813 /// isFromMainFile - Returns true if the file of provided SourceLocation is 814 /// the main file. 815 bool isFromMainFile(SourceLocation Loc) const { 816 return getFileID(Loc) == getMainFileID(); 817 } 818 819 /// isInSystemHeader - Returns if a SourceLocation is in a system header. 820 bool isInSystemHeader(SourceLocation Loc) const { 821 return getFileCharacteristic(Loc) != SrcMgr::C_User; 822 } 823 824 /// isInExternCSystemHeader - Returns if a SourceLocation is in an "extern C" 825 /// system header. 826 bool isInExternCSystemHeader(SourceLocation Loc) const { 827 return getFileCharacteristic(Loc) == SrcMgr::C_ExternCSystem; 828 } 829 830 //===--------------------------------------------------------------------===// 831 // Line Table Manipulation Routines 832 //===--------------------------------------------------------------------===// 833 834 /// getLineTableFilenameID - Return the uniqued ID for the specified filename. 835 /// 836 unsigned getLineTableFilenameID(const char *Ptr, unsigned Len); 837 838 /// AddLineNote - Add a line note to the line table for the FileID and offset 839 /// specified by Loc. If FilenameID is -1, it is considered to be 840 /// unspecified. 841 void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID); 842 void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID, 843 bool IsFileEntry, bool IsFileExit, 844 bool IsSystemHeader, bool IsExternCHeader); 845 846 /// \brief Determine if the source manager has a line table. 847 bool hasLineTable() const { return LineTable != 0; } 848 849 /// \brief Retrieve the stored line table. 850 LineTableInfo &getLineTable(); 851 852 //===--------------------------------------------------------------------===// 853 // Other miscellaneous methods. 854 //===--------------------------------------------------------------------===// 855 856 /// \brief Get the source location for the given file:line:col triplet. 857 /// 858 /// If the source file is included multiple times, the source location will 859 /// be based upon the first inclusion. 860 SourceLocation getLocation(const FileEntry *SourceFile, 861 unsigned Line, unsigned Col); 862 863 /// \brief Determines the order of 2 source locations in the translation unit. 864 /// 865 /// \returns true if LHS source location comes before RHS, false otherwise. 866 bool isBeforeInTranslationUnit(SourceLocation LHS, SourceLocation RHS) const; 867 868 // Iterators over FileInfos. 869 typedef llvm::DenseMap<const FileEntry*, SrcMgr::ContentCache*> 870 ::const_iterator fileinfo_iterator; 871 fileinfo_iterator fileinfo_begin() const { return FileInfos.begin(); } 872 fileinfo_iterator fileinfo_end() const { return FileInfos.end(); } 873 bool hasFileInfo(const FileEntry *File) const { 874 return FileInfos.find(File) != FileInfos.end(); 875 } 876 877 /// PrintStats - Print statistics to stderr. 878 /// 879 void PrintStats() const; 880 881 unsigned sloc_entry_size() const { return SLocEntryTable.size(); } 882 883 // FIXME: Exposing this is a little gross; what we want is a good way 884 // to iterate the entries that were not defined in an AST file (or 885 // any other external source). 886 unsigned sloc_loaded_entry_size() const { return SLocEntryLoaded.size(); } 887 888 const SrcMgr::SLocEntry &getSLocEntry(unsigned ID, bool *Invalid = 0) const { 889 assert(ID < SLocEntryTable.size() && "Invalid id"); 890 // If we haven't loaded this source-location entry from the external source 891 // yet, do so now. 892 if (ExternalSLocEntries && 893 ID < SLocEntryLoaded.size() && 894 !SLocEntryLoaded[ID] && 895 ExternalSLocEntries->ReadSLocEntry(ID) && 896 Invalid) 897 *Invalid = true; 898 899 return SLocEntryTable[ID]; 900 } 901 902 const SrcMgr::SLocEntry &getSLocEntry(FileID FID, bool *Invalid = 0) const { 903 return getSLocEntry(FID.ID, Invalid); 904 } 905 906 unsigned getNextOffset() const { return NextOffset; } 907 908 /// \brief Preallocate some number of source location entries, which 909 /// will be loaded as needed from the given external source. 910 void PreallocateSLocEntries(ExternalSLocEntrySource *Source, 911 unsigned NumSLocEntries, 912 unsigned NextOffset); 913 914 /// \brief Clear out any preallocated source location entries that 915 /// haven't already been loaded. 916 void ClearPreallocatedSLocEntries(); 917 918private: 919 const llvm::MemoryBuffer *getFakeBufferForRecovery() const; 920 921 /// isOffsetInFileID - Return true if the specified FileID contains the 922 /// specified SourceLocation offset. This is a very hot method. 923 inline bool isOffsetInFileID(FileID FID, unsigned SLocOffset) const { 924 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID); 925 // If the entry is after the offset, it can't contain it. 926 if (SLocOffset < Entry.getOffset()) return false; 927 928 // If this is the last entry than it does. Otherwise, the entry after it 929 // has to not include it. 930 if (FID.ID+1 == SLocEntryTable.size()) return true; 931 932 return SLocOffset < getSLocEntry(FileID::get(FID.ID+1)).getOffset(); 933 } 934 935 /// createFileID - Create a new fileID for the specified ContentCache and 936 /// include position. This works regardless of whether the ContentCache 937 /// corresponds to a file or some other input source. 938 FileID createFileID(const SrcMgr::ContentCache* File, 939 SourceLocation IncludePos, 940 SrcMgr::CharacteristicKind DirCharacter, 941 unsigned PreallocatedID = 0, 942 unsigned Offset = 0); 943 944 const SrcMgr::ContentCache * 945 getOrCreateContentCache(const FileEntry *SourceFile); 946 947 /// createMemBufferContentCache - Create a new ContentCache for the specified 948 /// memory buffer. 949 const SrcMgr::ContentCache* 950 createMemBufferContentCache(const llvm::MemoryBuffer *Buf); 951 952 FileID getFileIDSlow(unsigned SLocOffset) const; 953 954 SourceLocation getInstantiationLocSlowCase(SourceLocation Loc) const; 955 SourceLocation getSpellingLocSlowCase(SourceLocation Loc) const; 956 957 std::pair<FileID, unsigned> 958 getDecomposedInstantiationLocSlowCase(const SrcMgr::SLocEntry *E, 959 unsigned Offset) const; 960 std::pair<FileID, unsigned> 961 getDecomposedSpellingLocSlowCase(const SrcMgr::SLocEntry *E, 962 unsigned Offset) const; 963}; 964 965 966} // end namespace clang 967 968#endif 969