1 //===--- SourceManager.h - Track and cache source files ---------*- C++ -*-===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file defines the SourceManager interface.
12 //===----------------------------------------------------------------------===//
14 #ifndef LLVM_CLANG_SOURCEMANAGER_H
15 #define LLVM_CLANG_SOURCEMANAGER_H
17 #include "clang/Basic/LLVM.h"
18 #include "clang/Basic/SourceLocation.h"
19 #include "llvm/Support/Allocator.h"
20 #include "llvm/Support/DataTypes.h"
21 #include "llvm/ADT/PointerIntPair.h"
22 #include "llvm/ADT/PointerUnion.h"
23 #include "llvm/ADT/IntrusiveRefCntPtr.h"
24 #include "llvm/ADT/DenseMap.h"
25 #include "llvm/Support/MemoryBuffer.h"
41 /// SrcMgr - Public enums and private classes that are part of the
42 /// SourceManager implementation.
45 /// CharacteristicKind - This is used to represent whether a file or directory
46 /// holds normal user code, system code, or system code which is implicitly
47 /// 'extern "C"' in C++ mode. Entire directories can be tagged with this
48 /// (this is maintained by DirectoryLookup and friends) as can specific
49 /// FileInfos when a #pragma system_header is seen or various other cases.
51 enum CharacteristicKind {
52 C_User, C_System, C_ExternCSystem
55 /// ContentCache - One instance of this struct is kept for every file
56 /// loaded or used. This object owns the MemoryBuffer object.
59 /// \brief Whether the buffer is invalid.
61 /// \brief Whether the buffer should not be freed on destruction.
65 /// Buffer - The actual buffer containing the characters from the input
66 /// file. This is owned by the ContentCache object.
67 /// The bits indicate indicates whether the buffer is invalid.
68 mutable llvm::PointerIntPair<const llvm::MemoryBuffer *, 2> Buffer;
71 /// Reference to the file entry representing this ContentCache.
72 /// This reference does not own the FileEntry object.
73 /// It is possible for this to be NULL if
74 /// the ContentCache encapsulates an imaginary text buffer.
75 const FileEntry *OrigEntry;
77 /// \brief References the file which the contents were actually loaded from.
78 /// Can be different from 'Entry' if we overridden the contents of one file
79 /// with the contents of another file.
80 const FileEntry *ContentsEntry;
82 /// SourceLineCache - A bump pointer allocated array of offsets for each
83 /// source line. This is lazily computed. This is owned by the
84 /// SourceManager BumpPointerAllocator object.
85 unsigned *SourceLineCache;
87 /// NumLines - The number of lines in this ContentCache. This is only valid
88 /// if SourceLineCache is non-null.
91 /// \brief Lazily computed map of macro argument chunks to their expanded
93 typedef std::map<unsigned, SourceLocation> MacroArgsMap;
94 MacroArgsMap *MacroArgsCache;
96 /// getBuffer - Returns the memory buffer for the associated content.
98 /// \param Diag Object through which diagnostics will be emitted if the
99 /// buffer cannot be retrieved.
101 /// \param Loc If specified, is the location that invalid file diagnostics
102 /// will be emitted at.
104 /// \param Invalid If non-NULL, will be set \c true if an error occurred.
105 const llvm::MemoryBuffer *getBuffer(Diagnostic &Diag,
106 const SourceManager &SM,
107 SourceLocation Loc = SourceLocation(),
108 bool *Invalid = 0) const;
110 /// getSize - Returns the size of the content encapsulated by this
111 /// ContentCache. This can be the size of the source file or the size of an
112 /// arbitrary scratch buffer. If the ContentCache encapsulates a source
113 /// file this size is retrieved from the file's FileEntry.
114 unsigned getSize() const;
116 /// getSizeBytesMapped - Returns the number of bytes actually mapped for
117 /// this ContentCache. This can be 0 if the MemBuffer was not actually
119 unsigned getSizeBytesMapped() const;
121 /// Returns the kind of memory used to back the memory buffer for
122 /// this content cache. This is used for performance analysis.
123 llvm::MemoryBuffer::BufferKind getMemoryBufferKind() const;
125 void setBuffer(const llvm::MemoryBuffer *B) {
126 assert(!Buffer.getPointer() && "MemoryBuffer already set.");
127 Buffer.setPointer(B);
128 Buffer.setInt(false);
131 /// \brief Get the underlying buffer, returning NULL if the buffer is not
133 const llvm::MemoryBuffer *getRawBuffer() const {
134 return Buffer.getPointer();
137 /// \brief Replace the existing buffer (which will be deleted)
138 /// with the given buffer.
139 void replaceBuffer(const llvm::MemoryBuffer *B, bool DoNotFree = false);
141 /// \brief Determine whether the buffer itself is invalid.
142 bool isBufferInvalid() const {
143 return Buffer.getInt() & InvalidFlag;
146 /// \brief Determine whether the buffer should be freed.
147 bool shouldFreeBuffer() const {
148 return (Buffer.getInt() & DoNotFreeFlag) == 0;
151 ContentCache(const FileEntry *Ent = 0)
152 : Buffer(0, false), OrigEntry(Ent), ContentsEntry(Ent),
153 SourceLineCache(0), NumLines(0), MacroArgsCache(0) {}
155 ContentCache(const FileEntry *Ent, const FileEntry *contentEnt)
156 : Buffer(0, false), OrigEntry(Ent), ContentsEntry(contentEnt),
157 SourceLineCache(0), NumLines(0), MacroArgsCache(0) {}
161 /// The copy ctor does not allow copies where source object has either
162 /// a non-NULL Buffer or SourceLineCache. Ownership of allocated memory
163 /// is not transferred, so this is a logical error.
164 ContentCache(const ContentCache &RHS)
165 : Buffer(0, false), SourceLineCache(0), MacroArgsCache(0)
167 OrigEntry = RHS.OrigEntry;
168 ContentsEntry = RHS.ContentsEntry;
170 assert (RHS.Buffer.getPointer() == 0 && RHS.SourceLineCache == 0 &&
171 RHS.MacroArgsCache == 0
172 && "Passed ContentCache object cannot own a buffer.");
174 NumLines = RHS.NumLines;
178 // Disable assignments.
179 ContentCache &operator=(const ContentCache& RHS);
182 /// FileInfo - Information about a FileID, basically just the logical file
183 /// that it represents and include stack information.
185 /// Each FileInfo has include stack information, indicating where it came
186 /// from. This information encodes the #include chain that a token was
187 /// expanded from. The main include file has an invalid IncludeLoc.
189 /// FileInfos contain a "ContentCache *", with the contents of the file.
192 /// IncludeLoc - The location of the #include that brought in this file.
193 /// This is an invalid SLOC for the main file (top of the #include chain).
194 unsigned IncludeLoc; // Really a SourceLocation
196 /// \brief Number of FileIDs (files and macros) that were created during
197 /// preprocessing of this #include, including this SLocEntry.
198 /// Zero means the preprocessor didn't provide such info for this SLocEntry.
199 unsigned NumCreatedFIDs;
201 /// Data - This contains the ContentCache* and the bits indicating the
202 /// characteristic of the file and whether it has #line info, all bitmangled
206 friend class clang::SourceManager;
207 friend class clang::ASTWriter;
208 friend class clang::ASTReader;
210 /// get - Return a FileInfo object.
211 static FileInfo get(SourceLocation IL, const ContentCache *Con,
212 CharacteristicKind FileCharacter) {
214 X.IncludeLoc = IL.getRawEncoding();
215 X.NumCreatedFIDs = 0;
216 X.Data = (uintptr_t)Con;
217 assert((X.Data & 7) == 0 &&"ContentCache pointer insufficiently aligned");
218 assert((unsigned)FileCharacter < 4 && "invalid file character");
219 X.Data |= (unsigned)FileCharacter;
223 SourceLocation getIncludeLoc() const {
224 return SourceLocation::getFromRawEncoding(IncludeLoc);
226 const ContentCache* getContentCache() const {
227 return reinterpret_cast<const ContentCache*>(Data & ~7UL);
230 /// getCharacteristic - Return whether this is a system header or not.
231 CharacteristicKind getFileCharacteristic() const {
232 return (CharacteristicKind)(Data & 3);
235 /// hasLineDirectives - Return true if this FileID has #line directives in
237 bool hasLineDirectives() const { return (Data & 4) != 0; }
239 /// setHasLineDirectives - Set the flag that indicates that this FileID has
240 /// line table entries associated with it.
241 void setHasLineDirectives() {
246 /// ExpansionInfo - Each ExpansionInfo encodes the expansion location - where
247 /// the token was ultimately expanded, and the SpellingLoc - where the actual
248 /// character data for the token came from.
249 class ExpansionInfo {
250 // Really these are all SourceLocations.
252 /// SpellingLoc - Where the spelling for the token can be found.
253 unsigned SpellingLoc;
255 /// ExpansionLocStart/ExpansionLocEnd - In a macro expansion, these
256 /// indicate the start and end of the expansion. In object-like macros,
257 /// these will be the same. In a function-like macro expansion, the start
258 /// will be the identifier and the end will be the ')'. Finally, in
259 /// macro-argument instantitions, the end will be 'SourceLocation()', an
260 /// invalid location.
261 unsigned ExpansionLocStart, ExpansionLocEnd;
264 SourceLocation getSpellingLoc() const {
265 return SourceLocation::getFromRawEncoding(SpellingLoc);
267 SourceLocation getExpansionLocStart() const {
268 return SourceLocation::getFromRawEncoding(ExpansionLocStart);
270 SourceLocation getExpansionLocEnd() const {
271 SourceLocation EndLoc =
272 SourceLocation::getFromRawEncoding(ExpansionLocEnd);
273 return EndLoc.isInvalid() ? getExpansionLocStart() : EndLoc;
276 std::pair<SourceLocation,SourceLocation> getExpansionLocRange() const {
277 return std::make_pair(getExpansionLocStart(), getExpansionLocEnd());
280 bool isMacroArgExpansion() const {
281 // Note that this needs to return false for default constructed objects.
282 return getExpansionLocStart().isValid() &&
283 SourceLocation::getFromRawEncoding(ExpansionLocEnd).isInvalid();
286 /// create - Return a ExpansionInfo for an expansion. Start and End specify
287 /// the expansion range (where the macro is expanded), and SpellingLoc
288 /// specifies the spelling location (where the characters from the token
289 /// come from). All three can refer to normal File SLocs or expansion
291 static ExpansionInfo create(SourceLocation SpellingLoc,
292 SourceLocation Start, SourceLocation End) {
294 X.SpellingLoc = SpellingLoc.getRawEncoding();
295 X.ExpansionLocStart = Start.getRawEncoding();
296 X.ExpansionLocEnd = End.getRawEncoding();
300 /// createForMacroArg - Return a special ExpansionInfo for the expansion of
301 /// a macro argument into a function-like macro's body. ExpansionLoc
302 /// specifies the expansion location (where the macro is expanded). This
303 /// doesn't need to be a range because a macro is always expanded at
304 /// a macro parameter reference, and macro parameters are always exactly
305 /// one token. SpellingLoc specifies the spelling location (where the
306 /// characters from the token come from). ExpansionLoc and SpellingLoc can
307 /// both refer to normal File SLocs or expansion locations.
311 /// #define F(x) f(x)
315 /// When expanding '\c F(42)', the '\c x' would call this with an
316 /// SpellingLoc pointing at '\c 42' anad an ExpansionLoc pointing at its
317 /// location in the definition of '\c F'.
318 static ExpansionInfo createForMacroArg(SourceLocation SpellingLoc,
319 SourceLocation ExpansionLoc) {
320 // We store an intentionally invalid source location for the end of the
321 // expansion range to mark that this is a macro argument ion rather than
323 return create(SpellingLoc, ExpansionLoc, SourceLocation());
327 /// SLocEntry - This is a discriminated union of FileInfo and
328 /// ExpansionInfo. SourceManager keeps an array of these objects, and
329 /// they are uniquely identified by the FileID datatype.
331 unsigned Offset; // low bit is set for expansion info.
334 ExpansionInfo Expansion;
337 unsigned getOffset() const { return Offset >> 1; }
339 bool isExpansion() const { return Offset & 1; }
340 bool isFile() const { return !isExpansion(); }
342 const FileInfo &getFile() const {
343 assert(isFile() && "Not a file SLocEntry!");
347 const ExpansionInfo &getExpansion() const {
348 assert(isExpansion() && "Not a macro expansion SLocEntry!");
352 static SLocEntry get(unsigned Offset, const FileInfo &FI) {
354 E.Offset = Offset << 1;
359 static SLocEntry get(unsigned Offset, const ExpansionInfo &Expansion) {
361 E.Offset = (Offset << 1) | 1;
362 E.Expansion = Expansion;
366 } // end SrcMgr namespace.
368 /// \brief External source of source location entries.
369 class ExternalSLocEntrySource {
371 virtual ~ExternalSLocEntrySource();
373 /// \brief Read the source location entry with index ID, which will always be
376 /// \returns true if an error occurred that prevented the source-location
377 /// entry from being loaded.
378 virtual bool ReadSLocEntry(int ID) = 0;
382 /// IsBeforeInTranslationUnitCache - This class holds the cache used by
383 /// isBeforeInTranslationUnit. The cache structure is complex enough to be
384 /// worth breaking out of SourceManager.
385 class IsBeforeInTranslationUnitCache {
386 /// L/R QueryFID - These are the FID's of the cached query. If these match up
387 /// with a subsequent query, the result can be reused.
388 FileID LQueryFID, RQueryFID;
390 /// \brief True if LQueryFID was created before RQueryFID. This is used
391 /// to compare macro expansion locations.
392 bool IsLQFIDBeforeRQFID;
394 /// CommonFID - This is the file found in common between the two #include
395 /// traces. It is the nearest common ancestor of the #include tree.
398 /// L/R CommonOffset - This is the offset of the previous query in CommonFID.
399 /// Usually, this represents the location of the #include for QueryFID, but if
400 /// LQueryFID is a parent of RQueryFID (or vise versa) then these can be a
401 /// random token in the parent.
402 unsigned LCommonOffset, RCommonOffset;
405 /// isCacheValid - Return true if the currently cached values match up with
406 /// the specified LHS/RHS query. If not, we can't use the cache.
407 bool isCacheValid(FileID LHS, FileID RHS) const {
408 return LQueryFID == LHS && RQueryFID == RHS;
411 /// getCachedResult - If the cache is valid, compute the result given the
412 /// specified offsets in the LHS/RHS FID's.
413 bool getCachedResult(unsigned LOffset, unsigned ROffset) const {
414 // If one of the query files is the common file, use the offset. Otherwise,
415 // use the #include loc in the common file.
416 if (LQueryFID != CommonFID) LOffset = LCommonOffset;
417 if (RQueryFID != CommonFID) ROffset = RCommonOffset;
419 // It is common for multiple macro expansions to be "included" from the same
420 // location (expansion location), in which case use the order of the FileIDs
421 // to determine which came first.
422 if (LOffset == ROffset && LQueryFID != CommonFID && RQueryFID != CommonFID)
423 return IsLQFIDBeforeRQFID;
425 return LOffset < ROffset;
428 // Set up a new query.
429 void setQueryFIDs(FileID LHS, FileID RHS, bool isLFIDBeforeRFID) {
433 IsLQFIDBeforeRQFID = isLFIDBeforeRFID;
437 LQueryFID = RQueryFID = FileID();
438 IsLQFIDBeforeRQFID = false;
441 void setCommonLoc(FileID commonFID, unsigned lCommonOffset,
442 unsigned rCommonOffset) {
443 CommonFID = commonFID;
444 LCommonOffset = lCommonOffset;
445 RCommonOffset = rCommonOffset;
450 /// \brief This class handles loading and caching of source files into memory.
452 /// This object owns the MemoryBuffer objects for all of the loaded
453 /// files and assigns unique FileID's for each unique #include chain.
455 /// The SourceManager can be queried for information about SourceLocation
456 /// objects, turning them into either spelling or expansion locations. Spelling
457 /// locations represent where the bytes corresponding to a token came from and
458 /// expansion locations represent where the location is in the user's view. In
459 /// the case of a macro expansion, for example, the spelling location indicates
460 /// where the expanded token came from and the expansion location specifies
461 /// where it was expanded.
462 class SourceManager : public llvm::RefCountedBase<SourceManager> {
463 /// \brief Diagnostic object.
466 FileManager &FileMgr;
468 mutable llvm::BumpPtrAllocator ContentCacheAlloc;
470 /// FileInfos - Memoized information about all of the files tracked by this
471 /// SourceManager. This set allows us to merge ContentCache entries based
472 /// on their FileEntry*. All ContentCache objects will thus have unique,
473 /// non-null, FileEntry pointers.
474 llvm::DenseMap<const FileEntry*, SrcMgr::ContentCache*> FileInfos;
476 /// \brief True if the ContentCache for files that are overriden by other
477 /// files, should report the original file name. Defaults to true.
478 bool OverridenFilesKeepOriginalName;
480 /// \brief Files that have been overriden with the contents from another file.
481 llvm::DenseMap<const FileEntry *, const FileEntry *> OverriddenFiles;
483 /// MemBufferInfos - Information about various memory buffers that we have
484 /// read in. All FileEntry* within the stored ContentCache objects are NULL,
485 /// as they do not refer to a file.
486 std::vector<SrcMgr::ContentCache*> MemBufferInfos;
488 /// \brief The table of SLocEntries that are local to this module.
490 /// Positive FileIDs are indexes into this table. Entry 0 indicates an invalid
492 std::vector<SrcMgr::SLocEntry> LocalSLocEntryTable;
494 /// \brief The table of SLocEntries that are loaded from other modules.
496 /// Negative FileIDs are indexes into this table. To get from ID to an index,
498 std::vector<SrcMgr::SLocEntry> LoadedSLocEntryTable;
500 /// \brief The starting offset of the next local SLocEntry.
502 /// This is LocalSLocEntryTable.back().Offset + the size of that entry.
503 unsigned NextLocalOffset;
505 /// \brief The starting offset of the latest batch of loaded SLocEntries.
507 /// This is LoadedSLocEntryTable.back().Offset, except that that entry might
508 /// not have been loaded, so that value would be unknown.
509 unsigned CurrentLoadedOffset;
511 /// \brief The highest possible offset is 2^31-1, so CurrentLoadedOffset
513 static const unsigned MaxLoadedOffset = 1U << 31U;
515 /// \brief A bitmap that indicates whether the entries of LoadedSLocEntryTable
516 /// have already been loaded from the external source.
518 /// Same indexing as LoadedSLocEntryTable.
519 std::vector<bool> SLocEntryLoaded;
521 /// \brief An external source for source location entries.
522 ExternalSLocEntrySource *ExternalSLocEntries;
524 /// LastFileIDLookup - This is a one-entry cache to speed up getFileID.
525 /// LastFileIDLookup records the last FileID looked up or created, because it
526 /// is very common to look up many tokens from the same file.
527 mutable FileID LastFileIDLookup;
529 /// LineTable - This holds information for #line directives. It is referenced
530 /// by indices from SLocEntryTable.
531 LineTableInfo *LineTable;
533 /// LastLineNo - These ivars serve as a cache used in the getLineNumber
534 /// method which is used to speedup getLineNumber calls to nearby locations.
535 mutable FileID LastLineNoFileIDQuery;
536 mutable SrcMgr::ContentCache *LastLineNoContentCache;
537 mutable unsigned LastLineNoFilePos;
538 mutable unsigned LastLineNoResult;
540 /// MainFileID - The file ID for the main source file of the translation unit.
543 // Statistics for -print-stats.
544 mutable unsigned NumLinearScans, NumBinaryProbes;
546 // Cache results for the isBeforeInTranslationUnit method.
547 mutable IsBeforeInTranslationUnitCache IsBeforeInTUCache;
549 // Cache for the "fake" buffer used for error-recovery purposes.
550 mutable llvm::MemoryBuffer *FakeBufferForRecovery;
552 // SourceManager doesn't support copy construction.
553 explicit SourceManager(const SourceManager&);
554 void operator=(const SourceManager&);
556 SourceManager(Diagnostic &Diag, FileManager &FileMgr);
559 void clearIDTables();
561 Diagnostic &getDiagnostics() const { return Diag; }
563 FileManager &getFileManager() const { return FileMgr; }
565 /// \brief Set true if the SourceManager should report the original file name
566 /// for contents of files that were overriden by other files.Defaults to true.
567 void setOverridenFilesKeepOriginalName(bool value) {
568 OverridenFilesKeepOriginalName = value;
571 /// createMainFileIDForMembuffer - Create the FileID for a memory buffer
572 /// that will represent the FileID for the main source. One example
573 /// of when this would be used is when the main source is read from STDIN.
574 FileID createMainFileIDForMemBuffer(const llvm::MemoryBuffer *Buffer) {
575 assert(MainFileID.isInvalid() && "MainFileID already set!");
576 MainFileID = createFileIDForMemBuffer(Buffer);
580 //===--------------------------------------------------------------------===//
581 // MainFileID creation and querying methods.
582 //===--------------------------------------------------------------------===//
584 /// getMainFileID - Returns the FileID of the main source file.
585 FileID getMainFileID() const { return MainFileID; }
587 /// createMainFileID - Create the FileID for the main source file.
588 FileID createMainFileID(const FileEntry *SourceFile) {
589 assert(MainFileID.isInvalid() && "MainFileID already set!");
590 MainFileID = createFileID(SourceFile, SourceLocation(), SrcMgr::C_User);
594 /// \brief Set the file ID for the precompiled preamble, which is also the
596 void SetPreambleFileID(FileID Preamble) {
597 assert(MainFileID.isInvalid() && "MainFileID already set!");
598 MainFileID = Preamble;
601 //===--------------------------------------------------------------------===//
602 // Methods to create new FileID's and macro expansions.
603 //===--------------------------------------------------------------------===//
605 /// createFileID - Create a new FileID that represents the specified file
606 /// being #included from the specified IncludePosition. This translates NULL
607 /// into standard input.
608 FileID createFileID(const FileEntry *SourceFile, SourceLocation IncludePos,
609 SrcMgr::CharacteristicKind FileCharacter,
610 int LoadedID = 0, unsigned LoadedOffset = 0) {
611 const SrcMgr::ContentCache *IR = getOrCreateContentCache(SourceFile);
612 assert(IR && "getOrCreateContentCache() cannot return NULL");
613 return createFileID(IR, IncludePos, FileCharacter, LoadedID, LoadedOffset);
616 /// createFileIDForMemBuffer - Create a new FileID that represents the
617 /// specified memory buffer. This does no caching of the buffer and takes
618 /// ownership of the MemoryBuffer, so only pass a MemoryBuffer to this once.
619 FileID createFileIDForMemBuffer(const llvm::MemoryBuffer *Buffer,
620 int LoadedID = 0, unsigned LoadedOffset = 0) {
621 return createFileID(createMemBufferContentCache(Buffer), SourceLocation(),
622 SrcMgr::C_User, LoadedID, LoadedOffset);
625 /// createMacroArgExpansionLoc - Return a new SourceLocation that encodes the
626 /// fact that a token from SpellingLoc should actually be referenced from
627 /// ExpansionLoc, and that it represents the expansion of a macro argument
628 /// into the function-like macro body.
629 SourceLocation createMacroArgExpansionLoc(SourceLocation Loc,
630 SourceLocation ExpansionLoc,
633 /// createExpansionLoc - Return a new SourceLocation that encodes the fact
634 /// that a token from SpellingLoc should actually be referenced from
636 SourceLocation createExpansionLoc(SourceLocation Loc,
637 SourceLocation ExpansionLocStart,
638 SourceLocation ExpansionLocEnd,
641 unsigned LoadedOffset = 0);
643 /// \brief Retrieve the memory buffer associated with the given file.
645 /// \param Invalid If non-NULL, will be set \c true if an error
646 /// occurs while retrieving the memory buffer.
647 const llvm::MemoryBuffer *getMemoryBufferForFile(const FileEntry *File,
650 /// \brief Override the contents of the given source file by providing an
651 /// already-allocated buffer.
653 /// \param SourceFile the source file whose contents will be overriden.
655 /// \param Buffer the memory buffer whose contents will be used as the
656 /// data in the given source file.
658 /// \param DoNotFree If true, then the buffer will not be freed when the
659 /// source manager is destroyed.
660 void overrideFileContents(const FileEntry *SourceFile,
661 const llvm::MemoryBuffer *Buffer,
662 bool DoNotFree = false);
664 /// \brief Override the the given source file with another one.
666 /// \param SourceFile the source file which will be overriden.
668 /// \param NewFile the file whose contents will be used as the
669 /// data instead of the contents of the given source file.
670 void overrideFileContents(const FileEntry *SourceFile,
671 const FileEntry *NewFile);
673 //===--------------------------------------------------------------------===//
674 // FileID manipulation methods.
675 //===--------------------------------------------------------------------===//
677 /// getBuffer - Return the buffer for the specified FileID. If there is an
678 /// error opening this buffer the first time, this manufactures a temporary
679 /// buffer and returns a non-empty error string.
680 const llvm::MemoryBuffer *getBuffer(FileID FID, SourceLocation Loc,
681 bool *Invalid = 0) const {
682 bool MyInvalid = false;
683 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid);
684 if (MyInvalid || !Entry.isFile()) {
688 return getFakeBufferForRecovery();
691 return Entry.getFile().getContentCache()->getBuffer(Diag, *this, Loc,
695 const llvm::MemoryBuffer *getBuffer(FileID FID, bool *Invalid = 0) const {
696 bool MyInvalid = false;
697 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid);
698 if (MyInvalid || !Entry.isFile()) {
702 return getFakeBufferForRecovery();
705 return Entry.getFile().getContentCache()->getBuffer(Diag, *this,
710 /// getFileEntryForID - Returns the FileEntry record for the provided FileID.
711 const FileEntry *getFileEntryForID(FileID FID) const {
712 bool MyInvalid = false;
713 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid);
714 if (MyInvalid || !Entry.isFile())
717 return Entry.getFile().getContentCache()->OrigEntry;
720 /// Returns the FileEntry record for the provided SLocEntry.
721 const FileEntry *getFileEntryForSLocEntry(const SrcMgr::SLocEntry &sloc) const
723 return sloc.getFile().getContentCache()->OrigEntry;
726 /// getBufferData - Return a StringRef to the source buffer data for the
727 /// specified FileID.
729 /// \param FID The file ID whose contents will be returned.
730 /// \param Invalid If non-NULL, will be set true if an error occurred.
731 StringRef getBufferData(FileID FID, bool *Invalid = 0) const;
733 /// \brief Get the number of FileIDs (files and macros) that were created
734 /// during preprocessing of \arg FID, including it.
735 unsigned getNumCreatedFIDsForFileID(FileID FID) const {
736 bool Invalid = false;
737 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
738 if (Invalid || !Entry.isFile())
741 return Entry.getFile().NumCreatedFIDs;
744 /// \brief Set the number of FileIDs (files and macros) that were created
745 /// during preprocessing of \arg FID, including it.
746 void setNumCreatedFIDsForFileID(FileID FID, unsigned NumFIDs) const {
747 bool Invalid = false;
748 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
749 if (Invalid || !Entry.isFile())
752 assert(Entry.getFile().NumCreatedFIDs == 0 && "Already set!");
753 const_cast<SrcMgr::FileInfo &>(Entry.getFile()).NumCreatedFIDs = NumFIDs;
756 //===--------------------------------------------------------------------===//
757 // SourceLocation manipulation methods.
758 //===--------------------------------------------------------------------===//
760 /// getFileID - Return the FileID for a SourceLocation. This is a very
761 /// hot method that is used for all SourceManager queries that start with a
762 /// SourceLocation object. It is responsible for finding the entry in
763 /// SLocEntryTable which contains the specified location.
765 FileID getFileID(SourceLocation SpellingLoc) const {
766 unsigned SLocOffset = SpellingLoc.getOffset();
768 // If our one-entry cache covers this offset, just return it.
769 if (isOffsetInFileID(LastFileIDLookup, SLocOffset))
770 return LastFileIDLookup;
772 return getFileIDSlow(SLocOffset);
775 /// getLocForStartOfFile - Return the source location corresponding to the
776 /// first byte of the specified file.
777 SourceLocation getLocForStartOfFile(FileID FID) const {
778 bool Invalid = false;
779 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
780 if (Invalid || !Entry.isFile())
781 return SourceLocation();
783 unsigned FileOffset = Entry.getOffset();
784 return SourceLocation::getFileLoc(FileOffset);
787 /// \brief Returns the include location if \arg FID is a #include'd file
788 /// otherwise it returns an invalid location.
789 SourceLocation getIncludeLoc(FileID FID) const {
790 bool Invalid = false;
791 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
792 if (Invalid || !Entry.isFile())
793 return SourceLocation();
795 return Entry.getFile().getIncludeLoc();
798 /// getExpansionLoc - Given a SourceLocation object, return the expansion
799 /// location referenced by the ID.
800 SourceLocation getExpansionLoc(SourceLocation Loc) const {
801 // Handle the non-mapped case inline, defer to out of line code to handle
803 if (Loc.isFileID()) return Loc;
804 return getExpansionLocSlowCase(Loc);
807 /// getImmediateExpansionRange - Loc is required to be an expansion location.
808 /// Return the start/end of the expansion information.
809 std::pair<SourceLocation,SourceLocation>
810 getImmediateExpansionRange(SourceLocation Loc) const;
812 /// getExpansionRange - Given a SourceLocation object, return the range of
813 /// tokens covered by the expansion the ultimate file.
814 std::pair<SourceLocation,SourceLocation>
815 getExpansionRange(SourceLocation Loc) const;
818 /// getSpellingLoc - Given a SourceLocation object, return the spelling
819 /// location referenced by the ID. This is the place where the characters
820 /// that make up the lexed token can be found.
821 SourceLocation getSpellingLoc(SourceLocation Loc) const {
822 // Handle the non-mapped case inline, defer to out of line code to handle
824 if (Loc.isFileID()) return Loc;
825 return getSpellingLocSlowCase(Loc);
828 /// getImmediateSpellingLoc - Given a SourceLocation object, return the
829 /// spelling location referenced by the ID. This is the first level down
830 /// towards the place where the characters that make up the lexed token can be
831 /// found. This should not generally be used by clients.
832 SourceLocation getImmediateSpellingLoc(SourceLocation Loc) const;
834 /// getDecomposedLoc - Decompose the specified location into a raw FileID +
835 /// Offset pair. The first element is the FileID, the second is the
836 /// offset from the start of the buffer of the location.
837 std::pair<FileID, unsigned> getDecomposedLoc(SourceLocation Loc) const {
838 FileID FID = getFileID(Loc);
839 return std::make_pair(FID, Loc.getOffset()-getSLocEntry(FID).getOffset());
842 /// getDecomposedExpansionLoc - Decompose the specified location into a raw
843 /// FileID + Offset pair. If the location is an expansion record, walk
844 /// through it until we find the final location expanded.
845 std::pair<FileID, unsigned>
846 getDecomposedExpansionLoc(SourceLocation Loc) const {
847 FileID FID = getFileID(Loc);
848 const SrcMgr::SLocEntry *E = &getSLocEntry(FID);
850 unsigned Offset = Loc.getOffset()-E->getOffset();
852 return std::make_pair(FID, Offset);
854 return getDecomposedExpansionLocSlowCase(E);
857 /// getDecomposedSpellingLoc - Decompose the specified location into a raw
858 /// FileID + Offset pair. If the location is an expansion record, walk
859 /// through it until we find its spelling record.
860 std::pair<FileID, unsigned>
861 getDecomposedSpellingLoc(SourceLocation Loc) const {
862 FileID FID = getFileID(Loc);
863 const SrcMgr::SLocEntry *E = &getSLocEntry(FID);
865 unsigned Offset = Loc.getOffset()-E->getOffset();
867 return std::make_pair(FID, Offset);
868 return getDecomposedSpellingLocSlowCase(E, Offset);
871 /// getFileOffset - This method returns the offset from the start
872 /// of the file that the specified SourceLocation represents. This is not very
873 /// meaningful for a macro ID.
874 unsigned getFileOffset(SourceLocation SpellingLoc) const {
875 return getDecomposedLoc(SpellingLoc).second;
878 /// isMacroArgExpansion - This method tests whether the given source location
879 /// represents a macro argument's expansion into the function-like macro
880 /// definition. Such source locations only appear inside of the expansion
881 /// locations representing where a particular function-like macro was
883 bool isMacroArgExpansion(SourceLocation Loc) const;
885 /// \brief Returns true if \arg Loc is inside the [\arg Start, +\arg Length)
886 /// chunk of the source location address space.
887 /// If it's true and \arg RelativeOffset is non-null, it will be set to the
888 /// relative offset of \arg Loc inside the chunk.
889 bool isInSLocAddrSpace(SourceLocation Loc,
890 SourceLocation Start, unsigned Length,
891 unsigned *RelativeOffset = 0) const {
892 assert(((Start.getOffset() < NextLocalOffset &&
893 Start.getOffset()+Length <= NextLocalOffset) ||
894 (Start.getOffset() >= CurrentLoadedOffset &&
895 Start.getOffset()+Length < MaxLoadedOffset)) &&
896 "Chunk is not valid SLoc address space");
897 unsigned LocOffs = Loc.getOffset();
898 unsigned BeginOffs = Start.getOffset();
899 unsigned EndOffs = BeginOffs + Length;
900 if (LocOffs >= BeginOffs && LocOffs < EndOffs) {
902 *RelativeOffset = LocOffs - BeginOffs;
909 //===--------------------------------------------------------------------===//
910 // Queries about the code at a SourceLocation.
911 //===--------------------------------------------------------------------===//
913 /// getCharacterData - Return a pointer to the start of the specified location
914 /// in the appropriate spelling MemoryBuffer.
916 /// \param Invalid If non-NULL, will be set \c true if an error occurs.
917 const char *getCharacterData(SourceLocation SL, bool *Invalid = 0) const;
919 /// getColumnNumber - Return the column # for the specified file position.
920 /// This is significantly cheaper to compute than the line number. This
921 /// returns zero if the column number isn't known. This may only be called
922 /// on a file sloc, so you must choose a spelling or expansion location
923 /// before calling this method.
924 unsigned getColumnNumber(FileID FID, unsigned FilePos,
925 bool *Invalid = 0) const;
926 unsigned getSpellingColumnNumber(SourceLocation Loc, bool *Invalid = 0) const;
927 unsigned getExpansionColumnNumber(SourceLocation Loc,
928 bool *Invalid = 0) const;
929 unsigned getPresumedColumnNumber(SourceLocation Loc, bool *Invalid = 0) const;
932 /// getLineNumber - Given a SourceLocation, return the spelling line number
933 /// for the position indicated. This requires building and caching a table of
934 /// line offsets for the MemoryBuffer, so this is not cheap: use only when
935 /// about to emit a diagnostic.
936 unsigned getLineNumber(FileID FID, unsigned FilePos, bool *Invalid = 0) const;
937 unsigned getSpellingLineNumber(SourceLocation Loc, bool *Invalid = 0) const;
938 unsigned getExpansionLineNumber(SourceLocation Loc, bool *Invalid = 0) const;
939 unsigned getPresumedLineNumber(SourceLocation Loc, bool *Invalid = 0) const;
941 /// Return the filename or buffer identifier of the buffer the location is in.
942 /// Note that this name does not respect #line directives. Use getPresumedLoc
943 /// for normal clients.
944 const char *getBufferName(SourceLocation Loc, bool *Invalid = 0) const;
946 /// getFileCharacteristic - return the file characteristic of the specified
947 /// source location, indicating whether this is a normal file, a system
948 /// header, or an "implicit extern C" system header.
950 /// This state can be modified with flags on GNU linemarker directives like:
952 /// which changes all source locations in the current file after that to be
953 /// considered to be from a system header.
954 SrcMgr::CharacteristicKind getFileCharacteristic(SourceLocation Loc) const;
956 /// getPresumedLoc - This method returns the "presumed" location of a
957 /// SourceLocation specifies. A "presumed location" can be modified by #line
958 /// or GNU line marker directives. This provides a view on the data that a
959 /// user should see in diagnostics, for example.
961 /// Note that a presumed location is always given as the expansion point of
962 /// an expansion location, not at the spelling location.
964 /// \returns The presumed location of the specified SourceLocation. If the
965 /// presumed location cannot be calculate (e.g., because \p Loc is invalid
966 /// or the file containing \p Loc has changed on disk), returns an invalid
967 /// presumed location.
968 PresumedLoc getPresumedLoc(SourceLocation Loc) const;
970 /// isFromSameFile - Returns true if both SourceLocations correspond to
972 bool isFromSameFile(SourceLocation Loc1, SourceLocation Loc2) const {
973 return getFileID(Loc1) == getFileID(Loc2);
976 /// isFromMainFile - Returns true if the file of provided SourceLocation is
978 bool isFromMainFile(SourceLocation Loc) const {
979 return getFileID(Loc) == getMainFileID();
982 /// isInSystemHeader - Returns if a SourceLocation is in a system header.
983 bool isInSystemHeader(SourceLocation Loc) const {
984 return getFileCharacteristic(Loc) != SrcMgr::C_User;
987 /// isInExternCSystemHeader - Returns if a SourceLocation is in an "extern C"
989 bool isInExternCSystemHeader(SourceLocation Loc) const {
990 return getFileCharacteristic(Loc) == SrcMgr::C_ExternCSystem;
993 /// \brief The size of the SLocEnty that \arg FID represents.
994 unsigned getFileIDSize(FileID FID) const;
996 /// \brief Given a specific FileID, returns true if \arg Loc is inside that
997 /// FileID chunk and sets relative offset (offset of \arg Loc from beginning
998 /// of FileID) to \arg relativeOffset.
999 bool isInFileID(SourceLocation Loc, FileID FID,
1000 unsigned *RelativeOffset = 0) const {
1001 return isInFileID(Loc, FID, 0, getFileIDSize(FID), RelativeOffset);
1004 /// \brief Given a specific chunk of a FileID (FileID with offset+length),
1005 /// returns true if \arg Loc is inside that chunk and sets relative offset
1006 /// (offset of \arg Loc from beginning of chunk) to \arg relativeOffset.
1007 bool isInFileID(SourceLocation Loc,
1008 FileID FID, unsigned offset, unsigned length,
1009 unsigned *relativeOffset = 0) const;
1011 //===--------------------------------------------------------------------===//
1012 // Line Table Manipulation Routines
1013 //===--------------------------------------------------------------------===//
1015 /// getLineTableFilenameID - Return the uniqued ID for the specified filename.
1017 unsigned getLineTableFilenameID(StringRef Str);
1019 /// AddLineNote - Add a line note to the line table for the FileID and offset
1020 /// specified by Loc. If FilenameID is -1, it is considered to be
1022 void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID);
1023 void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID,
1024 bool IsFileEntry, bool IsFileExit,
1025 bool IsSystemHeader, bool IsExternCHeader);
1027 /// \brief Determine if the source manager has a line table.
1028 bool hasLineTable() const { return LineTable != 0; }
1030 /// \brief Retrieve the stored line table.
1031 LineTableInfo &getLineTable();
1033 //===--------------------------------------------------------------------===//
1034 // Queries for performance analysis.
1035 //===--------------------------------------------------------------------===//
1037 /// Return the total amount of physical memory allocated by the
1038 /// ContentCache allocator.
1039 size_t getContentCacheSize() const {
1040 return ContentCacheAlloc.getTotalMemory();
1043 struct MemoryBufferSizes {
1044 const size_t malloc_bytes;
1045 const size_t mmap_bytes;
1047 MemoryBufferSizes(size_t malloc_bytes, size_t mmap_bytes)
1048 : malloc_bytes(malloc_bytes), mmap_bytes(mmap_bytes) {}
1051 /// Return the amount of memory used by memory buffers, breaking down
1052 /// by heap-backed versus mmap'ed memory.
1053 MemoryBufferSizes getMemoryBufferSizes() const;
1055 // Return the amount of memory used for various side tables and
1056 // data structures in the SourceManager.
1057 size_t getDataStructureSizes() const;
1059 //===--------------------------------------------------------------------===//
1060 // Other miscellaneous methods.
1061 //===--------------------------------------------------------------------===//
1063 /// \brief Get the source location for the given file:line:col triplet.
1065 /// If the source file is included multiple times, the source location will
1066 /// be based upon the first inclusion.
1068 /// If the location points inside a function macro argument, the returned
1069 /// location will be the macro location in which the argument was expanded.
1070 /// \sa getMacroArgExpandedLocation
1071 SourceLocation getLocation(const FileEntry *SourceFile,
1072 unsigned Line, unsigned Col) {
1073 SourceLocation Loc = translateFileLineCol(SourceFile, Line, Col);
1074 return getMacroArgExpandedLocation(Loc);
1077 /// \brief Get the source location for the given file:line:col triplet.
1079 /// If the source file is included multiple times, the source location will
1080 /// be based upon the first inclusion.
1081 SourceLocation translateFileLineCol(const FileEntry *SourceFile,
1082 unsigned Line, unsigned Col);
1084 /// \brief If \arg Loc points inside a function macro argument, the returned
1085 /// location will be the macro location in which the argument was expanded.
1086 /// If a macro argument is used multiple times, the expanded location will
1087 /// be at the first expansion of the argument.
1091 /// Passing a file location pointing at 'foo', will yield a macro location
1092 /// where 'foo' was expanded into.
1093 SourceLocation getMacroArgExpandedLocation(SourceLocation Loc);
1095 /// \brief Determines the order of 2 source locations in the translation unit.
1097 /// \returns true if LHS source location comes before RHS, false otherwise.
1098 bool isBeforeInTranslationUnit(SourceLocation LHS, SourceLocation RHS) const;
1100 /// \brief Determines the order of 2 source locations in the "source location
1102 bool isBeforeInSLocAddrSpace(SourceLocation LHS, SourceLocation RHS) const {
1103 return isBeforeInSLocAddrSpace(LHS, RHS.getOffset());
1106 /// \brief Determines the order of a source location and a source location
1107 /// offset in the "source location address space".
1109 /// Note that we always consider source locations loaded from
1110 bool isBeforeInSLocAddrSpace(SourceLocation LHS, unsigned RHS) const {
1111 unsigned LHSOffset = LHS.getOffset();
1112 bool LHSLoaded = LHSOffset >= CurrentLoadedOffset;
1113 bool RHSLoaded = RHS >= CurrentLoadedOffset;
1114 if (LHSLoaded == RHSLoaded)
1115 return LHSOffset < RHS;
1120 // Iterators over FileInfos.
1121 typedef llvm::DenseMap<const FileEntry*, SrcMgr::ContentCache*>
1122 ::const_iterator fileinfo_iterator;
1123 fileinfo_iterator fileinfo_begin() const { return FileInfos.begin(); }
1124 fileinfo_iterator fileinfo_end() const { return FileInfos.end(); }
1125 bool hasFileInfo(const FileEntry *File) const {
1126 return FileInfos.find(File) != FileInfos.end();
1129 /// PrintStats - Print statistics to stderr.
1131 void PrintStats() const;
1133 /// \brief Get the number of local SLocEntries we have.
1134 unsigned local_sloc_entry_size() const { return LocalSLocEntryTable.size(); }
1136 /// \brief Get a local SLocEntry. This is exposed for indexing.
1137 const SrcMgr::SLocEntry &getLocalSLocEntry(unsigned Index,
1138 bool *Invalid = 0) const {
1139 assert(Index < LocalSLocEntryTable.size() && "Invalid index");
1140 return LocalSLocEntryTable[Index];
1143 /// \brief Get the number of loaded SLocEntries we have.
1144 unsigned loaded_sloc_entry_size() const { return LoadedSLocEntryTable.size();}
1146 /// \brief Get a loaded SLocEntry. This is exposed for indexing.
1147 const SrcMgr::SLocEntry &getLoadedSLocEntry(unsigned Index, bool *Invalid=0) const {
1148 assert(Index < LoadedSLocEntryTable.size() && "Invalid index");
1149 if (!SLocEntryLoaded[Index])
1150 ExternalSLocEntries->ReadSLocEntry(-(static_cast<int>(Index) + 2));
1151 return LoadedSLocEntryTable[Index];
1154 const SrcMgr::SLocEntry &getSLocEntry(FileID FID, bool *Invalid = 0) const {
1155 return getSLocEntryByID(FID.ID);
1158 unsigned getNextLocalOffset() const { return NextLocalOffset; }
1160 void setExternalSLocEntrySource(ExternalSLocEntrySource *Source) {
1161 assert(LoadedSLocEntryTable.empty() &&
1162 "Invalidating existing loaded entries");
1163 ExternalSLocEntries = Source;
1166 /// \brief Allocate a number of loaded SLocEntries, which will be actually
1167 /// loaded on demand from the external source.
1169 /// NumSLocEntries will be allocated, which occupy a total of TotalSize space
1170 /// in the global source view. The lowest ID and the base offset of the
1171 /// entries will be returned.
1172 std::pair<int, unsigned>
1173 AllocateLoadedSLocEntries(unsigned NumSLocEntries, unsigned TotalSize);
1176 const llvm::MemoryBuffer *getFakeBufferForRecovery() const;
1178 /// \brief Get the entry with the given unwrapped FileID.
1179 const SrcMgr::SLocEntry &getSLocEntryByID(int ID) const {
1180 assert(ID != -1 && "Using FileID sentinel value");
1182 return getLoadedSLocEntryByID(ID);
1183 return getLocalSLocEntry(static_cast<unsigned>(ID));
1186 const SrcMgr::SLocEntry &getLoadedSLocEntryByID(int ID) const {
1187 return getLoadedSLocEntry(static_cast<unsigned>(-ID - 2));
1190 /// createExpansionLoc - Implements the common elements of storing an
1191 /// expansion info struct into the SLocEntry table and producing a source
1192 /// location that refers to it.
1193 SourceLocation createExpansionLocImpl(const SrcMgr::ExpansionInfo &Expansion,
1196 unsigned LoadedOffset = 0);
1198 /// isOffsetInFileID - Return true if the specified FileID contains the
1199 /// specified SourceLocation offset. This is a very hot method.
1200 inline bool isOffsetInFileID(FileID FID, unsigned SLocOffset) const {
1201 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID);
1202 // If the entry is after the offset, it can't contain it.
1203 if (SLocOffset < Entry.getOffset()) return false;
1205 // If this is the very last entry then it does.
1209 // If it is the last local entry, then it does if the location is local.
1210 if (static_cast<unsigned>(FID.ID+1) == LocalSLocEntryTable.size()) {
1211 return SLocOffset < NextLocalOffset;
1214 // Otherwise, the entry after it has to not include it. This works for both
1215 // local and loaded entries.
1216 return SLocOffset < getSLocEntry(FileID::get(FID.ID+1)).getOffset();
1219 /// createFileID - Create a new fileID for the specified ContentCache and
1220 /// include position. This works regardless of whether the ContentCache
1221 /// corresponds to a file or some other input source.
1222 FileID createFileID(const SrcMgr::ContentCache* File,
1223 SourceLocation IncludePos,
1224 SrcMgr::CharacteristicKind DirCharacter,
1225 int LoadedID, unsigned LoadedOffset);
1227 const SrcMgr::ContentCache *
1228 getOrCreateContentCache(const FileEntry *SourceFile);
1230 /// createMemBufferContentCache - Create a new ContentCache for the specified
1232 const SrcMgr::ContentCache*
1233 createMemBufferContentCache(const llvm::MemoryBuffer *Buf);
1235 FileID getFileIDSlow(unsigned SLocOffset) const;
1236 FileID getFileIDLocal(unsigned SLocOffset) const;
1237 FileID getFileIDLoaded(unsigned SLocOffset) const;
1239 SourceLocation getExpansionLocSlowCase(SourceLocation Loc) const;
1240 SourceLocation getSpellingLocSlowCase(SourceLocation Loc) const;
1242 std::pair<FileID, unsigned>
1243 getDecomposedExpansionLocSlowCase(const SrcMgr::SLocEntry *E) const;
1244 std::pair<FileID, unsigned>
1245 getDecomposedSpellingLocSlowCase(const SrcMgr::SLocEntry *E,
1246 unsigned Offset) const;
1247 void computeMacroArgsCache(SrcMgr::ContentCache *Content, FileID FID);
1251 } // end namespace clang