SourceManager.h revision 414cb64f09ce48a36377458ce5e5a90c3ad41d00
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/DenseMap.h" 23#include <vector> 24#include <cassert> 25 26namespace llvm { 27class MemoryBuffer; 28class StringRef; 29} 30 31namespace clang { 32 33class Diagnostic; 34class SourceManager; 35class FileManager; 36class FileEntry; 37class LineTableInfo; 38 39/// SrcMgr - Public enums and private classes that are part of the 40/// SourceManager implementation. 41/// 42namespace SrcMgr { 43 /// CharacteristicKind - This is used to represent whether a file or directory 44 /// holds normal user code, system code, or system code which is implicitly 45 /// 'extern "C"' in C++ mode. Entire directories can be tagged with this 46 /// (this is maintained by DirectoryLookup and friends) as can specific 47 /// FileIDInfos when a #pragma system_header is seen or various other cases. 48 /// 49 enum CharacteristicKind { 50 C_User, C_System, C_ExternCSystem 51 }; 52 53 /// ContentCache - One instance of this struct is kept for every file 54 /// loaded or used. This object owns the MemoryBuffer object. 55 class ContentCache { 56 enum CCFlags { 57 /// \brief Whether the buffer is invalid. 58 InvalidFlag = 0x01, 59 /// \brief Whether the buffer should not be freed on destruction. 60 DoNotFreeFlag = 0x02 61 }; 62 63 /// Buffer - The actual buffer containing the characters from the input 64 /// file. This is owned by the ContentCache object. 65 /// The bits indicate indicates whether the buffer is invalid. 66 mutable llvm::PointerIntPair<const llvm::MemoryBuffer *, 2> Buffer; 67 68 public: 69 /// Reference to the file entry. This reference does not own 70 /// the FileEntry object. It is possible for this to be NULL if 71 /// the ContentCache encapsulates an imaginary text buffer. 72 const FileEntry *Entry; 73 74 /// SourceLineCache - A bump pointer allocated array of offsets for each 75 /// source line. This is lazily computed. This is owned by the 76 /// SourceManager BumpPointerAllocator object. 77 unsigned *SourceLineCache; 78 79 /// NumLines - The number of lines in this ContentCache. This is only valid 80 /// if SourceLineCache is non-null. 81 unsigned NumLines; 82 83 /// getBuffer - Returns the memory buffer for the associated content. 84 /// 85 /// \param Diag Object through which diagnostics will be emitted it the 86 /// buffer cannot be retrieved. 87 /// 88 /// \param Loc If specified, is the location that invalid file diagnostics 89 /// will be emitted at. 90 /// 91 /// \param Invalid If non-NULL, will be set \c true if an error occurred. 92 const llvm::MemoryBuffer *getBuffer(Diagnostic &Diag, 93 const SourceManager &SM, 94 SourceLocation Loc = SourceLocation(), 95 bool *Invalid = 0) const; 96 97 /// getSize - Returns the size of the content encapsulated by this 98 /// ContentCache. This can be the size of the source file or the size of an 99 /// arbitrary scratch buffer. If the ContentCache encapsulates a source 100 /// file this size is retrieved from the file's FileEntry. 101 unsigned getSize() const; 102 103 /// getSizeBytesMapped - Returns the number of bytes actually mapped for 104 /// this ContentCache. This can be 0 if the MemBuffer was not actually 105 /// instantiated. 106 unsigned getSizeBytesMapped() const; 107 108 void setBuffer(const llvm::MemoryBuffer *B) { 109 assert(!Buffer.getPointer() && "MemoryBuffer already set."); 110 Buffer.setPointer(B); 111 Buffer.setInt(false); 112 } 113 114 /// \brief Get the underlying buffer, returning NULL if the buffer is not 115 /// yet available. 116 const llvm::MemoryBuffer *getRawBuffer() const { 117 return Buffer.getPointer(); 118 } 119 120 /// \brief Replace the existing buffer (which will be deleted) 121 /// with the given buffer. 122 void replaceBuffer(const llvm::MemoryBuffer *B, bool DoNotFree = false); 123 124 /// \brief Determine whether the buffer itself is invalid. 125 bool isBufferInvalid() const { 126 return Buffer.getInt() & InvalidFlag; 127 } 128 129 /// \brief Determine whether the buffer should be freed. 130 bool shouldFreeBuffer() const { 131 return (Buffer.getInt() & DoNotFreeFlag) == 0; 132 } 133 134 ContentCache(const FileEntry *Ent = 0) 135 : Buffer(0, false), Entry(Ent), SourceLineCache(0), NumLines(0) {} 136 137 ~ContentCache(); 138 139 /// The copy ctor does not allow copies where source object has either 140 /// a non-NULL Buffer or SourceLineCache. Ownership of allocated memory 141 /// is not transfered, so this is a logical error. 142 ContentCache(const ContentCache &RHS) 143 : Buffer(0, false), SourceLineCache(0) 144 { 145 Entry = RHS.Entry; 146 147 assert (RHS.Buffer.getPointer() == 0 && RHS.SourceLineCache == 0 148 && "Passed ContentCache object cannot own a buffer."); 149 150 NumLines = RHS.NumLines; 151 } 152 153 private: 154 // Disable assignments. 155 ContentCache &operator=(const ContentCache& RHS); 156 }; 157 158 /// FileInfo - Information about a FileID, basically just the logical file 159 /// that it represents and include stack information. 160 /// 161 /// Each FileInfo has include stack information, indicating where it came 162 /// from. This information encodes the #include chain that a token was 163 /// instantiated from. The main include file has an invalid IncludeLoc. 164 /// 165 /// FileInfos contain a "ContentCache *", with the contents of the file. 166 /// 167 class FileInfo { 168 /// IncludeLoc - The location of the #include that brought in this file. 169 /// This is an invalid SLOC for the main file (top of the #include chain). 170 unsigned IncludeLoc; // Really a SourceLocation 171 172 /// Data - This contains the ContentCache* and the bits indicating the 173 /// characteristic of the file and whether it has #line info, all bitmangled 174 /// together. 175 uintptr_t Data; 176 public: 177 /// get - Return a FileInfo object. 178 static FileInfo get(SourceLocation IL, const ContentCache *Con, 179 CharacteristicKind FileCharacter) { 180 FileInfo X; 181 X.IncludeLoc = IL.getRawEncoding(); 182 X.Data = (uintptr_t)Con; 183 assert((X.Data & 7) == 0 &&"ContentCache pointer insufficiently aligned"); 184 assert((unsigned)FileCharacter < 4 && "invalid file character"); 185 X.Data |= (unsigned)FileCharacter; 186 return X; 187 } 188 189 SourceLocation getIncludeLoc() const { 190 return SourceLocation::getFromRawEncoding(IncludeLoc); 191 } 192 const ContentCache* getContentCache() const { 193 return reinterpret_cast<const ContentCache*>(Data & ~7UL); 194 } 195 196 /// getCharacteristic - Return whether this is a system header or not. 197 CharacteristicKind getFileCharacteristic() const { 198 return (CharacteristicKind)(Data & 3); 199 } 200 201 /// hasLineDirectives - Return true if this FileID has #line directives in 202 /// it. 203 bool hasLineDirectives() const { return (Data & 4) != 0; } 204 205 /// setHasLineDirectives - Set the flag that indicates that this FileID has 206 /// line table entries associated with it. 207 void setHasLineDirectives() { 208 Data |= 4; 209 } 210 }; 211 212 /// InstantiationInfo - Each InstantiationInfo encodes the Instantiation 213 /// location - where the token was ultimately instantiated, and the 214 /// SpellingLoc - where the actual character data for the token came from. 215 class InstantiationInfo { 216 // Really these are all SourceLocations. 217 218 /// SpellingLoc - Where the spelling for the token can be found. 219 unsigned SpellingLoc; 220 221 /// InstantiationLocStart/InstantiationLocEnd - In a macro expansion, these 222 /// indicate the start and end of the instantiation. In object-like macros, 223 /// these will be the same. In a function-like macro instantiation, the 224 /// start will be the identifier and the end will be the ')'. 225 unsigned InstantiationLocStart, InstantiationLocEnd; 226 public: 227 SourceLocation getSpellingLoc() const { 228 return SourceLocation::getFromRawEncoding(SpellingLoc); 229 } 230 SourceLocation getInstantiationLocStart() const { 231 return SourceLocation::getFromRawEncoding(InstantiationLocStart); 232 } 233 SourceLocation getInstantiationLocEnd() const { 234 return SourceLocation::getFromRawEncoding(InstantiationLocEnd); 235 } 236 237 std::pair<SourceLocation,SourceLocation> getInstantiationLocRange() const { 238 return std::make_pair(getInstantiationLocStart(), 239 getInstantiationLocEnd()); 240 } 241 242 /// get - Return a InstantiationInfo for an expansion. IL specifies 243 /// the instantiation location (where the macro is expanded), and SL 244 /// specifies the spelling location (where the characters from the token 245 /// come from). IL and PL can both refer to normal File SLocs or 246 /// instantiation locations. 247 static InstantiationInfo get(SourceLocation ILStart, SourceLocation ILEnd, 248 SourceLocation SL) { 249 InstantiationInfo X; 250 X.SpellingLoc = SL.getRawEncoding(); 251 X.InstantiationLocStart = ILStart.getRawEncoding(); 252 X.InstantiationLocEnd = ILEnd.getRawEncoding(); 253 return X; 254 } 255 }; 256 257 /// SLocEntry - This is a discriminated union of FileInfo and 258 /// InstantiationInfo. SourceManager keeps an array of these objects, and 259 /// they are uniquely identified by the FileID datatype. 260 class SLocEntry { 261 unsigned Offset; // low bit is set for instantiation info. 262 union { 263 FileInfo File; 264 InstantiationInfo Instantiation; 265 }; 266 public: 267 unsigned getOffset() const { return Offset >> 1; } 268 269 bool isInstantiation() const { return Offset & 1; } 270 bool isFile() const { return !isInstantiation(); } 271 272 const FileInfo &getFile() const { 273 assert(isFile() && "Not a file SLocEntry!"); 274 return File; 275 } 276 277 const InstantiationInfo &getInstantiation() const { 278 assert(isInstantiation() && "Not an instantiation SLocEntry!"); 279 return Instantiation; 280 } 281 282 static SLocEntry get(unsigned Offset, const FileInfo &FI) { 283 SLocEntry E; 284 E.Offset = Offset << 1; 285 E.File = FI; 286 return E; 287 } 288 289 static SLocEntry get(unsigned Offset, const InstantiationInfo &II) { 290 SLocEntry E; 291 E.Offset = (Offset << 1) | 1; 292 E.Instantiation = II; 293 return E; 294 } 295 }; 296} // end SrcMgr namespace. 297 298/// \brief External source of source location entries. 299class ExternalSLocEntrySource { 300public: 301 virtual ~ExternalSLocEntrySource(); 302 303 /// \brief Read the source location entry with index ID. 304 virtual void ReadSLocEntry(unsigned ID) = 0; 305}; 306 307 308/// IsBeforeInTranslationUnitCache - This class holds the cache used by 309/// isBeforeInTranslationUnit. The cache structure is complex enough to be 310/// worth breaking out of SourceManager. 311class IsBeforeInTranslationUnitCache { 312 /// L/R QueryFID - These are the FID's of the cached query. If these match up 313 /// with a subsequent query, the result can be reused. 314 FileID LQueryFID, RQueryFID; 315 316 /// CommonFID - This is the file found in common between the two #include 317 /// traces. It is the nearest common ancestor of the #include tree. 318 FileID CommonFID; 319 320 /// L/R CommonOffset - This is the offset of the previous query in CommonFID. 321 /// Usually, this represents the location of the #include for QueryFID, but if 322 /// LQueryFID is a parent of RQueryFID (or vise versa) then these can be a 323 /// random token in the parent. 324 unsigned LCommonOffset, RCommonOffset; 325public: 326 327 /// isCacheValid - Return true if the currently cached values match up with 328 /// the specified LHS/RHS query. If not, we can't use the cache. 329 bool isCacheValid(FileID LHS, FileID RHS) const { 330 return LQueryFID == LHS && RQueryFID == RHS; 331 } 332 333 /// getCachedResult - If the cache is valid, compute the result given the 334 /// specified offsets in the LHS/RHS FID's. 335 bool getCachedResult(unsigned LOffset, unsigned ROffset) const { 336 // If one of the query files is the common file, use the offset. Otherwise, 337 // use the #include loc in the common file. 338 if (LQueryFID != CommonFID) LOffset = LCommonOffset; 339 if (RQueryFID != CommonFID) ROffset = RCommonOffset; 340 return LOffset < ROffset; 341 } 342 343 // Set up a new query. 344 void setQueryFIDs(FileID LHS, FileID RHS) { 345 LQueryFID = LHS; 346 RQueryFID = RHS; 347 } 348 349 void setCommonLoc(FileID commonFID, unsigned lCommonOffset, 350 unsigned rCommonOffset) { 351 CommonFID = commonFID; 352 LCommonOffset = lCommonOffset; 353 RCommonOffset = rCommonOffset; 354 } 355 356}; 357 358/// SourceManager - This file handles loading and caching of source files into 359/// memory. This object owns the MemoryBuffer objects for all of the loaded 360/// files and assigns unique FileID's for each unique #include chain. 361/// 362/// The SourceManager can be queried for information about SourceLocation 363/// objects, turning them into either spelling or instantiation locations. 364/// Spelling locations represent where the bytes corresponding to a token came 365/// from and instantiation locations represent where the location is in the 366/// user's view. In the case of a macro expansion, for example, the spelling 367/// location indicates where the expanded token came from and the instantiation 368/// location specifies where it was expanded. 369class SourceManager { 370 /// \brief Diagnostic object. 371 Diagnostic &Diag; 372 373 FileManager &FileMgr; 374 375 mutable llvm::BumpPtrAllocator ContentCacheAlloc; 376 377 /// FileInfos - Memoized information about all of the files tracked by this 378 /// SourceManager. This set allows us to merge ContentCache entries based 379 /// on their FileEntry*. All ContentCache objects will thus have unique, 380 /// non-null, FileEntry pointers. 381 llvm::DenseMap<const FileEntry*, SrcMgr::ContentCache*> FileInfos; 382 383 /// MemBufferInfos - Information about various memory buffers that we have 384 /// read in. All FileEntry* within the stored ContentCache objects are NULL, 385 /// as they do not refer to a file. 386 std::vector<SrcMgr::ContentCache*> MemBufferInfos; 387 388 /// SLocEntryTable - This is an array of SLocEntry's that we have created. 389 /// FileID is an index into this vector. This array is sorted by the offset. 390 std::vector<SrcMgr::SLocEntry> SLocEntryTable; 391 /// NextOffset - This is the next available offset that a new SLocEntry can 392 /// start at. It is SLocEntryTable.back().getOffset()+size of back() entry. 393 unsigned NextOffset; 394 395 /// \brief If source location entries are being lazily loaded from 396 /// an external source, this vector indicates whether the Ith source 397 /// location entry has already been loaded from the external storage. 398 std::vector<bool> SLocEntryLoaded; 399 400 /// \brief An external source for source location entries. 401 ExternalSLocEntrySource *ExternalSLocEntries; 402 403 /// LastFileIDLookup - This is a one-entry cache to speed up getFileID. 404 /// LastFileIDLookup records the last FileID looked up or created, because it 405 /// is very common to look up many tokens from the same file. 406 mutable FileID LastFileIDLookup; 407 408 /// LineTable - This holds information for #line directives. It is referenced 409 /// by indices from SLocEntryTable. 410 LineTableInfo *LineTable; 411 412 /// LastLineNo - These ivars serve as a cache used in the getLineNumber 413 /// method which is used to speedup getLineNumber calls to nearby locations. 414 mutable FileID LastLineNoFileIDQuery; 415 mutable SrcMgr::ContentCache *LastLineNoContentCache; 416 mutable unsigned LastLineNoFilePos; 417 mutable unsigned LastLineNoResult; 418 419 /// MainFileID - The file ID for the main source file of the translation unit. 420 FileID MainFileID; 421 422 // Statistics for -print-stats. 423 mutable unsigned NumLinearScans, NumBinaryProbes; 424 425 // Cache results for the isBeforeInTranslationUnit method. 426 mutable IsBeforeInTranslationUnitCache IsBeforeInTUCache; 427 428 // SourceManager doesn't support copy construction. 429 explicit SourceManager(const SourceManager&); 430 void operator=(const SourceManager&); 431public: 432 SourceManager(Diagnostic &Diag, FileManager &FileMgr); 433 ~SourceManager(); 434 435 void clearIDTables(); 436 437 Diagnostic &getDiagnostics() const { return Diag; } 438 439 FileManager &getFileManager() const { return FileMgr; } 440 441 //===--------------------------------------------------------------------===// 442 // MainFileID creation and querying methods. 443 //===--------------------------------------------------------------------===// 444 445 /// getMainFileID - Returns the FileID of the main source file. 446 FileID getMainFileID() const { return MainFileID; } 447 448 /// createMainFileID - Create the FileID for the main source file. 449 FileID createMainFileID(const FileEntry *SourceFile) { 450 assert(MainFileID.isInvalid() && "MainFileID already set!"); 451 MainFileID = createFileID(SourceFile, SourceLocation(), SrcMgr::C_User); 452 return MainFileID; 453 } 454 455 /// \brief Set the file ID for the precompiled preamble, which is also the 456 /// main file. 457 void SetPreambleFileID(FileID Preamble) { 458 assert(MainFileID.isInvalid() && "MainFileID already set!"); 459 MainFileID = Preamble; 460 } 461 462 //===--------------------------------------------------------------------===// 463 // Methods to create new FileID's and instantiations. 464 //===--------------------------------------------------------------------===// 465 466 /// createFileID - Create a new FileID that represents the specified file 467 /// being #included from the specified IncludePosition. This returns 0 on 468 /// error and translates NULL into standard input. 469 /// PreallocateID should be non-zero to specify which pre-allocated, 470 /// lazily computed source location is being filled in by this operation. 471 FileID createFileID(const FileEntry *SourceFile, SourceLocation IncludePos, 472 SrcMgr::CharacteristicKind FileCharacter, 473 unsigned PreallocatedID = 0, 474 unsigned Offset = 0) { 475 const SrcMgr::ContentCache *IR = getOrCreateContentCache(SourceFile); 476 assert(IR && "getOrCreateContentCache() cannot return NULL"); 477 return createFileID(IR, IncludePos, FileCharacter, PreallocatedID, Offset); 478 } 479 480 /// createFileIDForMemBuffer - Create a new FileID that represents the 481 /// specified memory buffer. This does no caching of the buffer and takes 482 /// ownership of the MemoryBuffer, so only pass a MemoryBuffer to this once. 483 FileID createFileIDForMemBuffer(const llvm::MemoryBuffer *Buffer, 484 unsigned PreallocatedID = 0, 485 unsigned Offset = 0) { 486 return createFileID(createMemBufferContentCache(Buffer), SourceLocation(), 487 SrcMgr::C_User, PreallocatedID, Offset); 488 } 489 490 /// createMainFileIDForMembuffer - Create the FileID for a memory buffer 491 /// that will represent the FileID for the main source. One example 492 /// of when this would be used is when the main source is read from STDIN. 493 FileID createMainFileIDForMemBuffer(const llvm::MemoryBuffer *Buffer) { 494 assert(MainFileID.isInvalid() && "MainFileID already set!"); 495 MainFileID = createFileIDForMemBuffer(Buffer); 496 return MainFileID; 497 } 498 499 /// createInstantiationLoc - Return a new SourceLocation that encodes the fact 500 /// that a token at Loc should actually be referenced from InstantiationLoc. 501 /// TokLength is the length of the token being instantiated. 502 SourceLocation createInstantiationLoc(SourceLocation Loc, 503 SourceLocation InstantiationLocStart, 504 SourceLocation InstantiationLocEnd, 505 unsigned TokLength, 506 unsigned PreallocatedID = 0, 507 unsigned Offset = 0); 508 509 /// \brief Retrieve the memory buffer associated with the given file. 510 /// 511 /// \param Invalid If non-NULL, will be set \c true if an error 512 /// occurs while retrieving the memory buffer. 513 const llvm::MemoryBuffer *getMemoryBufferForFile(const FileEntry *File, 514 bool *Invalid = 0); 515 516 /// \brief Override the contents of the given source file by providing an 517 /// already-allocated buffer. 518 /// 519 /// \param SourceFile the source file whose contents will be overriden. 520 /// 521 /// \param Buffer the memory buffer whose contents will be used as the 522 /// data in the given source file. 523 /// 524 /// \param DoNotFree If true, then the buffer will not be freed when the 525 /// source manager is destroyed. 526 void overrideFileContents(const FileEntry *SourceFile, 527 const llvm::MemoryBuffer *Buffer, 528 bool DoNotFree = false); 529 530 //===--------------------------------------------------------------------===// 531 // FileID manipulation methods. 532 //===--------------------------------------------------------------------===// 533 534 /// getBuffer - Return the buffer for the specified FileID. If there is an 535 /// error opening this buffer the first time, this manufactures a temporary 536 /// buffer and returns a non-empty error string. 537 const llvm::MemoryBuffer *getBuffer(FileID FID, SourceLocation Loc, 538 bool *Invalid = 0) const { 539 return getSLocEntry(FID).getFile().getContentCache() 540 ->getBuffer(Diag, *this, Loc, Invalid); 541 } 542 543 const llvm::MemoryBuffer *getBuffer(FileID FID, bool *Invalid = 0) const { 544 return getSLocEntry(FID).getFile().getContentCache() 545 ->getBuffer(Diag, *this, SourceLocation(), Invalid); 546 } 547 548 /// getFileEntryForID - Returns the FileEntry record for the provided FileID. 549 const FileEntry *getFileEntryForID(FileID FID) const { 550 return getSLocEntry(FID).getFile().getContentCache()->Entry; 551 } 552 553 /// getBufferData - Return a StringRef to the source buffer data for the 554 /// specified FileID. 555 /// 556 /// \param FID The file ID whose contents will be returned. 557 /// \param Invalid If non-NULL, will be set true if an error occurred. 558 llvm::StringRef getBufferData(FileID FID, bool *Invalid = 0) const; 559 560 561 //===--------------------------------------------------------------------===// 562 // SourceLocation manipulation methods. 563 //===--------------------------------------------------------------------===// 564 565 /// getFileID - Return the FileID for a SourceLocation. This is a very 566 /// hot method that is used for all SourceManager queries that start with a 567 /// SourceLocation object. It is responsible for finding the entry in 568 /// SLocEntryTable which contains the specified location. 569 /// 570 FileID getFileID(SourceLocation SpellingLoc) const { 571 unsigned SLocOffset = SpellingLoc.getOffset(); 572 573 // If our one-entry cache covers this offset, just return it. 574 if (isOffsetInFileID(LastFileIDLookup, SLocOffset)) 575 return LastFileIDLookup; 576 577 return getFileIDSlow(SLocOffset); 578 } 579 580 /// getLocForStartOfFile - Return the source location corresponding to the 581 /// first byte of the specified file. 582 SourceLocation getLocForStartOfFile(FileID FID) const { 583 assert(FID.ID < SLocEntryTable.size() && "FileID out of range"); 584 assert(getSLocEntry(FID).isFile() && "FileID is not a file"); 585 unsigned FileOffset = getSLocEntry(FID).getOffset(); 586 return SourceLocation::getFileLoc(FileOffset); 587 } 588 589 /// getInstantiationLoc - Given a SourceLocation object, return the 590 /// instantiation location referenced by the ID. 591 SourceLocation getInstantiationLoc(SourceLocation Loc) const { 592 // Handle the non-mapped case inline, defer to out of line code to handle 593 // instantiations. 594 if (Loc.isFileID()) return Loc; 595 return getInstantiationLocSlowCase(Loc); 596 } 597 598 /// getImmediateInstantiationRange - Loc is required to be an instantiation 599 /// location. Return the start/end of the instantiation information. 600 std::pair<SourceLocation,SourceLocation> 601 getImmediateInstantiationRange(SourceLocation Loc) const; 602 603 /// getInstantiationRange - Given a SourceLocation object, return the 604 /// range of tokens covered by the instantiation in the ultimate file. 605 std::pair<SourceLocation,SourceLocation> 606 getInstantiationRange(SourceLocation Loc) const; 607 608 609 /// getSpellingLoc - Given a SourceLocation object, return the spelling 610 /// location referenced by the ID. This is the place where the characters 611 /// that make up the lexed token can be found. 612 SourceLocation getSpellingLoc(SourceLocation Loc) const { 613 // Handle the non-mapped case inline, defer to out of line code to handle 614 // instantiations. 615 if (Loc.isFileID()) return Loc; 616 return getSpellingLocSlowCase(Loc); 617 } 618 619 /// getImmediateSpellingLoc - Given a SourceLocation object, return the 620 /// spelling location referenced by the ID. This is the first level down 621 /// towards the place where the characters that make up the lexed token can be 622 /// found. This should not generally be used by clients. 623 SourceLocation getImmediateSpellingLoc(SourceLocation Loc) const; 624 625 /// getDecomposedLoc - Decompose the specified location into a raw FileID + 626 /// Offset pair. The first element is the FileID, the second is the 627 /// offset from the start of the buffer of the location. 628 std::pair<FileID, unsigned> getDecomposedLoc(SourceLocation Loc) const { 629 FileID FID = getFileID(Loc); 630 return std::make_pair(FID, Loc.getOffset()-getSLocEntry(FID).getOffset()); 631 } 632 633 /// getDecomposedInstantiationLoc - Decompose the specified location into a 634 /// raw FileID + Offset pair. If the location is an instantiation record, 635 /// walk through it until we find the final location instantiated. 636 std::pair<FileID, unsigned> 637 getDecomposedInstantiationLoc(SourceLocation Loc) const { 638 FileID FID = getFileID(Loc); 639 const SrcMgr::SLocEntry *E = &getSLocEntry(FID); 640 641 unsigned Offset = Loc.getOffset()-E->getOffset(); 642 if (Loc.isFileID()) 643 return std::make_pair(FID, Offset); 644 645 return getDecomposedInstantiationLocSlowCase(E, Offset); 646 } 647 648 /// getDecomposedSpellingLoc - Decompose the specified location into a raw 649 /// FileID + Offset pair. If the location is an instantiation record, walk 650 /// through it until we find its spelling record. 651 std::pair<FileID, unsigned> 652 getDecomposedSpellingLoc(SourceLocation Loc) const { 653 FileID FID = getFileID(Loc); 654 const SrcMgr::SLocEntry *E = &getSLocEntry(FID); 655 656 unsigned Offset = Loc.getOffset()-E->getOffset(); 657 if (Loc.isFileID()) 658 return std::make_pair(FID, Offset); 659 return getDecomposedSpellingLocSlowCase(E, Offset); 660 } 661 662 /// getFileOffset - This method returns the offset from the start 663 /// of the file that the specified SourceLocation represents. This is not very 664 /// meaningful for a macro ID. 665 unsigned getFileOffset(SourceLocation SpellingLoc) const { 666 return getDecomposedLoc(SpellingLoc).second; 667 } 668 669 670 //===--------------------------------------------------------------------===// 671 // Queries about the code at a SourceLocation. 672 //===--------------------------------------------------------------------===// 673 674 /// getCharacterData - Return a pointer to the start of the specified location 675 /// in the appropriate spelling MemoryBuffer. 676 /// 677 /// \param Invalid If non-NULL, will be set \c true if an error occurs. 678 const char *getCharacterData(SourceLocation SL, bool *Invalid = 0) const; 679 680 /// getColumnNumber - Return the column # for the specified file position. 681 /// This is significantly cheaper to compute than the line number. This 682 /// returns zero if the column number isn't known. This may only be called on 683 /// a file sloc, so you must choose a spelling or instantiation location 684 /// before calling this method. 685 unsigned getColumnNumber(FileID FID, unsigned FilePos, 686 bool *Invalid = 0) const; 687 unsigned getSpellingColumnNumber(SourceLocation Loc, 688 bool *Invalid = 0) const; 689 unsigned getInstantiationColumnNumber(SourceLocation Loc, 690 bool *Invalid = 0) const; 691 692 693 /// getLineNumber - Given a SourceLocation, return the spelling line number 694 /// for the position indicated. This requires building and caching a table of 695 /// line offsets for the MemoryBuffer, so this is not cheap: use only when 696 /// about to emit a diagnostic. 697 unsigned getLineNumber(FileID FID, unsigned FilePos, bool *Invalid = 0) const; 698 699 unsigned getInstantiationLineNumber(SourceLocation Loc, 700 bool *Invalid = 0) const; 701 unsigned getSpellingLineNumber(SourceLocation Loc, bool *Invalid = 0) const; 702 703 /// Return the filename or buffer identifier of the buffer the location is in. 704 /// Note that this name does not respect #line directives. Use getPresumedLoc 705 /// for normal clients. 706 const char *getBufferName(SourceLocation Loc, bool *Invalid = 0) const; 707 708 /// getFileCharacteristic - return the file characteristic of the specified 709 /// source location, indicating whether this is a normal file, a system 710 /// header, or an "implicit extern C" system header. 711 /// 712 /// This state can be modified with flags on GNU linemarker directives like: 713 /// # 4 "foo.h" 3 714 /// which changes all source locations in the current file after that to be 715 /// considered to be from a system header. 716 SrcMgr::CharacteristicKind getFileCharacteristic(SourceLocation Loc) const; 717 718 /// getPresumedLoc - This method returns the "presumed" location of a 719 /// SourceLocation specifies. A "presumed location" can be modified by #line 720 /// or GNU line marker directives. This provides a view on the data that a 721 /// user should see in diagnostics, for example. 722 /// 723 /// Note that a presumed location is always given as the instantiation point 724 /// of an instantiation location, not at the spelling location. 725 /// 726 /// \returns The presumed location of the specified SourceLocation. If the 727 /// presumed location cannot be calculate (e.g., because \p Loc is invalid 728 /// or the file containing \p Loc has changed on disk), returns an invalid 729 /// presumed location. 730 PresumedLoc getPresumedLoc(SourceLocation Loc) const; 731 732 /// isFromSameFile - Returns true if both SourceLocations correspond to 733 /// the same file. 734 bool isFromSameFile(SourceLocation Loc1, SourceLocation Loc2) const { 735 return getFileID(Loc1) == getFileID(Loc2); 736 } 737 738 /// isFromMainFile - Returns true if the file of provided SourceLocation is 739 /// the main file. 740 bool isFromMainFile(SourceLocation Loc) const { 741 return getFileID(Loc) == getMainFileID(); 742 } 743 744 /// isInSystemHeader - Returns if a SourceLocation is in a system header. 745 bool isInSystemHeader(SourceLocation Loc) const { 746 return getFileCharacteristic(Loc) != SrcMgr::C_User; 747 } 748 749 /// isInExternCSystemHeader - Returns if a SourceLocation is in an "extern C" 750 /// system header. 751 bool isInExternCSystemHeader(SourceLocation Loc) const { 752 return getFileCharacteristic(Loc) == SrcMgr::C_ExternCSystem; 753 } 754 755 //===--------------------------------------------------------------------===// 756 // Line Table Manipulation Routines 757 //===--------------------------------------------------------------------===// 758 759 /// getLineTableFilenameID - Return the uniqued ID for the specified filename. 760 /// 761 unsigned getLineTableFilenameID(const char *Ptr, unsigned Len); 762 763 /// AddLineNote - Add a line note to the line table for the FileID and offset 764 /// specified by Loc. If FilenameID is -1, it is considered to be 765 /// unspecified. 766 void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID); 767 void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID, 768 bool IsFileEntry, bool IsFileExit, 769 bool IsSystemHeader, bool IsExternCHeader); 770 771 /// \brief Determine if the source manager has a line table. 772 bool hasLineTable() const { return LineTable != 0; } 773 774 /// \brief Retrieve the stored line table. 775 LineTableInfo &getLineTable(); 776 777 //===--------------------------------------------------------------------===// 778 // Other miscellaneous methods. 779 //===--------------------------------------------------------------------===// 780 781 /// \brief Get the source location for the given file:line:col triplet. 782 /// 783 /// If the source file is included multiple times, the source location will 784 /// be based upon the first inclusion. 785 SourceLocation getLocation(const FileEntry *SourceFile, 786 unsigned Line, unsigned Col) const; 787 788 /// \brief Determines the order of 2 source locations in the translation unit. 789 /// 790 /// \returns true if LHS source location comes before RHS, false otherwise. 791 bool isBeforeInTranslationUnit(SourceLocation LHS, SourceLocation RHS) const; 792 793 // Iterators over FileInfos. 794 typedef llvm::DenseMap<const FileEntry*, SrcMgr::ContentCache*> 795 ::const_iterator fileinfo_iterator; 796 fileinfo_iterator fileinfo_begin() const { return FileInfos.begin(); } 797 fileinfo_iterator fileinfo_end() const { return FileInfos.end(); } 798 bool hasFileInfo(const FileEntry *File) const { 799 return FileInfos.find(File) != FileInfos.end(); 800 } 801 802 /// PrintStats - Print statistics to stderr. 803 /// 804 void PrintStats() const; 805 806 unsigned sloc_entry_size() const { return SLocEntryTable.size(); } 807 808 // FIXME: Exposing this is a little gross; what we want is a good way 809 // to iterate the entries that were not defined in an AST file (or 810 // any other external source). 811 unsigned sloc_loaded_entry_size() const { return SLocEntryLoaded.size(); } 812 813 const SrcMgr::SLocEntry &getSLocEntry(unsigned ID) const { 814 assert(ID < SLocEntryTable.size() && "Invalid id"); 815 if (ExternalSLocEntries && 816 ID < SLocEntryLoaded.size() && 817 !SLocEntryLoaded[ID]) 818 ExternalSLocEntries->ReadSLocEntry(ID); 819 return SLocEntryTable[ID]; 820 } 821 822 const SrcMgr::SLocEntry &getSLocEntry(FileID FID) const { 823 return getSLocEntry(FID.ID); 824 } 825 826 unsigned getNextOffset() const { return NextOffset; } 827 828 /// \brief Preallocate some number of source location entries, which 829 /// will be loaded as needed from the given external source. 830 void PreallocateSLocEntries(ExternalSLocEntrySource *Source, 831 unsigned NumSLocEntries, 832 unsigned NextOffset); 833 834 /// \brief Clear out any preallocated source location entries that 835 /// haven't already been loaded. 836 void ClearPreallocatedSLocEntries(); 837 838private: 839 /// isOffsetInFileID - Return true if the specified FileID contains the 840 /// specified SourceLocation offset. This is a very hot method. 841 inline bool isOffsetInFileID(FileID FID, unsigned SLocOffset) const { 842 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID); 843 // If the entry is after the offset, it can't contain it. 844 if (SLocOffset < Entry.getOffset()) return false; 845 846 // If this is the last entry than it does. Otherwise, the entry after it 847 // has to not include it. 848 if (FID.ID+1 == SLocEntryTable.size()) return true; 849 850 return SLocOffset < getSLocEntry(FileID::get(FID.ID+1)).getOffset(); 851 } 852 853 /// createFileID - Create a new fileID for the specified ContentCache and 854 /// include position. This works regardless of whether the ContentCache 855 /// corresponds to a file or some other input source. 856 FileID createFileID(const SrcMgr::ContentCache* File, 857 SourceLocation IncludePos, 858 SrcMgr::CharacteristicKind DirCharacter, 859 unsigned PreallocatedID = 0, 860 unsigned Offset = 0); 861 862 const SrcMgr::ContentCache * 863 getOrCreateContentCache(const FileEntry *SourceFile); 864 865 /// createMemBufferContentCache - Create a new ContentCache for the specified 866 /// memory buffer. 867 const SrcMgr::ContentCache* 868 createMemBufferContentCache(const llvm::MemoryBuffer *Buf); 869 870 FileID getFileIDSlow(unsigned SLocOffset) const; 871 872 SourceLocation getInstantiationLocSlowCase(SourceLocation Loc) const; 873 SourceLocation getSpellingLocSlowCase(SourceLocation Loc) const; 874 875 std::pair<FileID, unsigned> 876 getDecomposedInstantiationLocSlowCase(const SrcMgr::SLocEntry *E, 877 unsigned Offset) const; 878 std::pair<FileID, unsigned> 879 getDecomposedSpellingLocSlowCase(const SrcMgr::SLocEntry *E, 880 unsigned Offset) const; 881}; 882 883 884} // end namespace clang 885 886#endif 887