SourceManager.h revision 62221b17c90457df9ca0ff20bb54d634e8951def
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/// \file
11/// \brief Defines the SourceManager interface.
12///
13/// There are three different types of locations in a file: a spelling
14/// location, an expansion location, and a presumed location.
15///
16/// Given an example of:
17/// \code
18/// #define min(x, y) x < y ? x : y
19/// \endcode
20///
21/// and then later on a use of min:
22/// \code
23/// #line 17
24/// return min(a, b);
25/// \endcode
26///
27/// The expansion location is the line in the source code where the macro
28/// was expanded (the return statement), the spelling location is the
29/// location in the source where the macro was originally defined,
30/// and the presumed location is where the line directive states that
31/// the line is 17, or any other line.
32///
33//===----------------------------------------------------------------------===//
34
35#ifndef LLVM_CLANG_SOURCEMANAGER_H
36#define LLVM_CLANG_SOURCEMANAGER_H
37
38#include "clang/Basic/LLVM.h"
39#include "clang/Basic/FileManager.h"
40#include "clang/Basic/SourceLocation.h"
41#include "llvm/Support/Allocator.h"
42#include "llvm/Support/DataTypes.h"
43#include "llvm/ADT/PointerIntPair.h"
44#include "llvm/ADT/PointerUnion.h"
45#include "llvm/ADT/IntrusiveRefCntPtr.h"
46#include "llvm/ADT/OwningPtr.h"
47#include "llvm/ADT/DenseMap.h"
48#include "llvm/ADT/DenseSet.h"
49#include "llvm/Support/MemoryBuffer.h"
50#include <map>
51#include <vector>
52#include <cassert>
53
54namespace clang {
55
56class DiagnosticsEngine;
57class SourceManager;
58class FileManager;
59class FileEntry;
60class LineTableInfo;
61class LangOptions;
62class ASTWriter;
63class ASTReader;
64
65/// \brief Public enums and private classes that are part of the
66/// SourceManager implementation.
67///
68namespace SrcMgr {
69  /// \brief Indicates whether a file or directory holds normal user code,
70  /// system code, or system code which is implicitly 'extern "C"' in C++ mode.
71  ///
72  /// Entire directories can be tagged with this (this is maintained by
73  /// DirectoryLookup and friends) as can specific FileInfos when a \#pragma
74  /// system_header is seen or in various other cases.
75  ///
76  enum CharacteristicKind {
77    C_User, C_System, C_ExternCSystem
78  };
79
80  /// \brief One instance of this struct is kept for every file loaded or used.
81  ////
82  /// This object owns the MemoryBuffer object.
83  class ContentCache {
84    enum CCFlags {
85      /// \brief Whether the buffer is invalid.
86      InvalidFlag = 0x01,
87      /// \brief Whether the buffer should not be freed on destruction.
88      DoNotFreeFlag = 0x02
89    };
90
91    /// \brief The actual buffer containing the characters from the input
92    /// file.
93    ///
94    /// This is owned by the ContentCache object.  The bits indicate
95    /// whether the buffer is invalid.
96    mutable llvm::PointerIntPair<const llvm::MemoryBuffer *, 2> Buffer;
97
98  public:
99    /// \brief Reference to the file entry representing this ContentCache.
100    ///
101    /// This reference does not own the FileEntry object.
102    ///
103    /// It is possible for this to be NULL if the ContentCache encapsulates
104    /// an imaginary text buffer.
105    const FileEntry *OrigEntry;
106
107    /// \brief References the file which the contents were actually loaded from.
108    ///
109    /// Can be different from 'Entry' if we overridden the contents of one file
110    /// with the contents of another file.
111    const FileEntry *ContentsEntry;
112
113    /// \brief A bump pointer allocated array of offsets for each source line.
114    ///
115    /// This is lazily computed.  This is owned by the SourceManager
116    /// BumpPointerAllocator object.
117    unsigned *SourceLineCache;
118
119    /// \brief The number of lines in this ContentCache.
120    ///
121    /// This is only valid if SourceLineCache is non-null.
122    unsigned NumLines : 31;
123
124    /// \brief Indicates whether the buffer itself was provided to override
125    /// the actual file contents.
126    ///
127    /// When true, the original entry may be a virtual file that does not
128    /// exist.
129    unsigned BufferOverridden : 1;
130
131    /// \brief True if this content cache was initially created for a source
132    /// file considered as a system one.
133    unsigned IsSystemFile : 1;
134
135    ContentCache(const FileEntry *Ent = 0)
136      : Buffer(0, false), OrigEntry(Ent), ContentsEntry(Ent),
137        SourceLineCache(0), NumLines(0), BufferOverridden(false),
138        IsSystemFile(false) {}
139
140    ContentCache(const FileEntry *Ent, const FileEntry *contentEnt)
141      : Buffer(0, false), OrigEntry(Ent), ContentsEntry(contentEnt),
142        SourceLineCache(0), NumLines(0), BufferOverridden(false),
143        IsSystemFile(false) {}
144
145    ~ContentCache();
146
147    /// The copy ctor does not allow copies where source object has either
148    /// a non-NULL Buffer or SourceLineCache.  Ownership of allocated memory
149    /// is not transferred, so this is a logical error.
150    ContentCache(const ContentCache &RHS)
151      : Buffer(0, false), SourceLineCache(0), BufferOverridden(false),
152        IsSystemFile(false)
153    {
154      OrigEntry = RHS.OrigEntry;
155      ContentsEntry = RHS.ContentsEntry;
156
157      assert (RHS.Buffer.getPointer() == 0 && RHS.SourceLineCache == 0 &&
158              "Passed ContentCache object cannot own a buffer.");
159
160      NumLines = RHS.NumLines;
161    }
162
163    /// \brief Returns the memory buffer for the associated content.
164    ///
165    /// \param Diag Object through which diagnostics will be emitted if the
166    ///   buffer cannot be retrieved.
167    ///
168    /// \param Loc If specified, is the location that invalid file diagnostics
169    ///   will be emitted at.
170    ///
171    /// \param Invalid If non-NULL, will be set \c true if an error occurred.
172    const llvm::MemoryBuffer *getBuffer(DiagnosticsEngine &Diag,
173                                        const SourceManager &SM,
174                                        SourceLocation Loc = SourceLocation(),
175                                        bool *Invalid = 0) const;
176
177    /// \brief Returns the size of the content encapsulated by this
178    /// ContentCache.
179    ///
180    /// This can be the size of the source file or the size of an
181    /// arbitrary scratch buffer.  If the ContentCache encapsulates a source
182    /// file this size is retrieved from the file's FileEntry.
183    unsigned getSize() const;
184
185    /// \brief Returns the number of bytes actually mapped for this
186    /// ContentCache.
187    ///
188    /// This can be 0 if the MemBuffer was not actually expanded.
189    unsigned getSizeBytesMapped() const;
190
191    /// Returns the kind of memory used to back the memory buffer for
192    /// this content cache.  This is used for performance analysis.
193    llvm::MemoryBuffer::BufferKind getMemoryBufferKind() const;
194
195    void setBuffer(const llvm::MemoryBuffer *B) {
196      assert(!Buffer.getPointer() && "MemoryBuffer already set.");
197      Buffer.setPointer(B);
198      Buffer.setInt(false);
199    }
200
201    /// \brief Get the underlying buffer, returning NULL if the buffer is not
202    /// yet available.
203    const llvm::MemoryBuffer *getRawBuffer() const {
204      return Buffer.getPointer();
205    }
206
207    /// \brief Replace the existing buffer (which will be deleted)
208    /// with the given buffer.
209    void replaceBuffer(const llvm::MemoryBuffer *B, bool DoNotFree = false);
210
211    /// \brief Determine whether the buffer itself is invalid.
212    bool isBufferInvalid() const {
213      return Buffer.getInt() & InvalidFlag;
214    }
215
216    /// \brief Determine whether the buffer should be freed.
217    bool shouldFreeBuffer() const {
218      return (Buffer.getInt() & DoNotFreeFlag) == 0;
219    }
220
221  private:
222    // Disable assignments.
223    ContentCache &operator=(const ContentCache& RHS) LLVM_DELETED_FUNCTION;
224  };
225
226  /// \brief Information about a FileID, basically just the logical file
227  /// that it represents and include stack information.
228  ///
229  /// Each FileInfo has include stack information, indicating where it came
230  /// from. This information encodes the \#include chain that a token was
231  /// expanded from. The main include file has an invalid IncludeLoc.
232  ///
233  /// FileInfos contain a "ContentCache *", with the contents of the file.
234  ///
235  class FileInfo {
236    /// \brief The location of the \#include that brought in this file.
237    ///
238    /// This is an invalid SLOC for the main file (top of the \#include chain).
239    unsigned IncludeLoc;  // Really a SourceLocation
240
241    /// \brief Number of FileIDs (files and macros) that were created during
242    /// preprocessing of this \#include, including this SLocEntry.
243    ///
244    /// Zero means the preprocessor didn't provide such info for this SLocEntry.
245    unsigned NumCreatedFIDs;
246
247    /// \brief Contains the ContentCache* and the bits indicating the
248    /// characteristic of the file and whether it has \#line info, all
249    /// bitmangled together.
250    uintptr_t Data;
251
252    friend class clang::SourceManager;
253    friend class clang::ASTWriter;
254    friend class clang::ASTReader;
255  public:
256    /// \brief Return a FileInfo object.
257    static FileInfo get(SourceLocation IL, const ContentCache *Con,
258                        CharacteristicKind FileCharacter) {
259      FileInfo X;
260      X.IncludeLoc = IL.getRawEncoding();
261      X.NumCreatedFIDs = 0;
262      X.Data = (uintptr_t)Con;
263      assert((X.Data & 7) == 0 &&"ContentCache pointer insufficiently aligned");
264      assert((unsigned)FileCharacter < 4 && "invalid file character");
265      X.Data |= (unsigned)FileCharacter;
266      return X;
267    }
268
269    SourceLocation getIncludeLoc() const {
270      return SourceLocation::getFromRawEncoding(IncludeLoc);
271    }
272    const ContentCache* getContentCache() const {
273      return reinterpret_cast<const ContentCache*>(Data & ~7UL);
274    }
275
276    /// \brief Return whether this is a system header or not.
277    CharacteristicKind getFileCharacteristic() const {
278      return (CharacteristicKind)(Data & 3);
279    }
280
281    /// \brief Return true if this FileID has \#line directives in it.
282    bool hasLineDirectives() const { return (Data & 4) != 0; }
283
284    /// \brief Set the flag that indicates that this FileID has
285    /// line table entries associated with it.
286    void setHasLineDirectives() {
287      Data |= 4;
288    }
289  };
290
291  /// \brief Each ExpansionInfo encodes the expansion location - where
292  /// the token was ultimately expanded, and the SpellingLoc - where the actual
293  /// character data for the token came from.
294  class ExpansionInfo {
295    // Really these are all SourceLocations.
296
297    /// \brief Where the spelling for the token can be found.
298    unsigned SpellingLoc;
299
300    /// In a macro expansion, ExpansionLocStart and ExpansionLocEnd
301    /// indicate the start and end of the expansion. In object-like macros,
302    /// they will be the same. In a function-like macro expansion, the start
303    /// will be the identifier and the end will be the ')'. Finally, in
304    /// macro-argument instantiations, the end will be 'SourceLocation()', an
305    /// invalid location.
306    unsigned ExpansionLocStart, ExpansionLocEnd;
307
308  public:
309    SourceLocation getSpellingLoc() const {
310      return SourceLocation::getFromRawEncoding(SpellingLoc);
311    }
312    SourceLocation getExpansionLocStart() const {
313      return SourceLocation::getFromRawEncoding(ExpansionLocStart);
314    }
315    SourceLocation getExpansionLocEnd() const {
316      SourceLocation EndLoc =
317        SourceLocation::getFromRawEncoding(ExpansionLocEnd);
318      return EndLoc.isInvalid() ? getExpansionLocStart() : EndLoc;
319    }
320
321    std::pair<SourceLocation,SourceLocation> getExpansionLocRange() const {
322      return std::make_pair(getExpansionLocStart(), getExpansionLocEnd());
323    }
324
325    bool isMacroArgExpansion() const {
326      // Note that this needs to return false for default constructed objects.
327      return getExpansionLocStart().isValid() &&
328        SourceLocation::getFromRawEncoding(ExpansionLocEnd).isInvalid();
329    }
330
331    bool isFunctionMacroExpansion() const {
332      return getExpansionLocStart().isValid() &&
333          getExpansionLocStart() != getExpansionLocEnd();
334    }
335
336    /// \brief Return a ExpansionInfo for an expansion.
337    ///
338    /// Start and End specify the expansion range (where the macro is
339    /// expanded), and SpellingLoc specifies the spelling location (where
340    /// the characters from the token come from). All three can refer to
341    /// normal File SLocs or expansion locations.
342    static ExpansionInfo create(SourceLocation SpellingLoc,
343                                SourceLocation Start, SourceLocation End) {
344      ExpansionInfo X;
345      X.SpellingLoc = SpellingLoc.getRawEncoding();
346      X.ExpansionLocStart = Start.getRawEncoding();
347      X.ExpansionLocEnd = End.getRawEncoding();
348      return X;
349    }
350
351    /// \brief Return a special ExpansionInfo for the expansion of
352    /// a macro argument into a function-like macro's body.
353    ///
354    /// ExpansionLoc specifies the expansion location (where the macro is
355    /// expanded). This doesn't need to be a range because a macro is always
356    /// expanded at a macro parameter reference, and macro parameters are
357    /// always exactly one token. SpellingLoc specifies the spelling location
358    /// (where the characters from the token come from). ExpansionLoc and
359    /// SpellingLoc can both refer to normal File SLocs or expansion locations.
360    ///
361    /// Given the code:
362    /// \code
363    ///   #define F(x) f(x)
364    ///   F(42);
365    /// \endcode
366    ///
367    /// When expanding '\c F(42)', the '\c x' would call this with an
368    /// SpellingLoc pointing at '\c 42' and an ExpansionLoc pointing at its
369    /// location in the definition of '\c F'.
370    static ExpansionInfo createForMacroArg(SourceLocation SpellingLoc,
371                                           SourceLocation ExpansionLoc) {
372      // We store an intentionally invalid source location for the end of the
373      // expansion range to mark that this is a macro argument ion rather than
374      // a normal one.
375      return create(SpellingLoc, ExpansionLoc, SourceLocation());
376    }
377  };
378
379  /// \brief This is a discriminated union of FileInfo and ExpansionInfo.
380  ///
381  /// SourceManager keeps an array of these objects, and they are uniquely
382  /// identified by the FileID datatype.
383  class SLocEntry {
384    unsigned Offset;   // low bit is set for expansion info.
385    union {
386      FileInfo File;
387      ExpansionInfo Expansion;
388    };
389  public:
390    unsigned getOffset() const { return Offset >> 1; }
391
392    bool isExpansion() const { return Offset & 1; }
393    bool isFile() const { return !isExpansion(); }
394
395    const FileInfo &getFile() const {
396      assert(isFile() && "Not a file SLocEntry!");
397      return File;
398    }
399
400    const ExpansionInfo &getExpansion() const {
401      assert(isExpansion() && "Not a macro expansion SLocEntry!");
402      return Expansion;
403    }
404
405    static SLocEntry get(unsigned Offset, const FileInfo &FI) {
406      SLocEntry E;
407      E.Offset = Offset << 1;
408      E.File = FI;
409      return E;
410    }
411
412    static SLocEntry get(unsigned Offset, const ExpansionInfo &Expansion) {
413      SLocEntry E;
414      E.Offset = (Offset << 1) | 1;
415      E.Expansion = Expansion;
416      return E;
417    }
418  };
419}  // end SrcMgr namespace.
420
421/// \brief External source of source location entries.
422class ExternalSLocEntrySource {
423public:
424  virtual ~ExternalSLocEntrySource();
425
426  /// \brief Read the source location entry with index ID, which will always be
427  /// less than -1.
428  ///
429  /// \returns true if an error occurred that prevented the source-location
430  /// entry from being loaded.
431  virtual bool ReadSLocEntry(int ID) = 0;
432};
433
434
435/// \brief Holds the cache used by isBeforeInTranslationUnit.
436///
437/// The cache structure is complex enough to be worth breaking out of
438/// SourceManager.
439class IsBeforeInTranslationUnitCache {
440  /// \brief The FileID's of the cached query.
441  ///
442  /// If these match up with a subsequent query, the result can be reused.
443  FileID LQueryFID, RQueryFID;
444
445  /// \brief True if LQueryFID was created before RQueryFID.
446  ///
447  /// This is used to compare macro expansion locations.
448  bool IsLQFIDBeforeRQFID;
449
450  /// \brief The file found in common between the two \#include traces, i.e.,
451  /// the nearest common ancestor of the \#include tree.
452  FileID CommonFID;
453
454  /// \brief The offset of the previous query in CommonFID.
455  ///
456  /// Usually, this represents the location of the \#include for QueryFID, but
457  /// if LQueryFID is a parent of RQueryFID (or vice versa) then these can be a
458  /// random token in the parent.
459  unsigned LCommonOffset, RCommonOffset;
460public:
461
462  /// \brief Return true if the currently cached values match up with
463  /// the specified LHS/RHS query.
464  ///
465  /// If not, we can't use the cache.
466  bool isCacheValid(FileID LHS, FileID RHS) const {
467    return LQueryFID == LHS && RQueryFID == RHS;
468  }
469
470  /// \brief If the cache is valid, compute the result given the
471  /// specified offsets in the LHS/RHS FileID's.
472  bool getCachedResult(unsigned LOffset, unsigned ROffset) const {
473    // If one of the query files is the common file, use the offset.  Otherwise,
474    // use the #include loc in the common file.
475    if (LQueryFID != CommonFID) LOffset = LCommonOffset;
476    if (RQueryFID != CommonFID) ROffset = RCommonOffset;
477
478    // It is common for multiple macro expansions to be "included" from the same
479    // location (expansion location), in which case use the order of the FileIDs
480    // to determine which came first. This will also take care the case where
481    // one of the locations points at the inclusion/expansion point of the other
482    // in which case its FileID will come before the other.
483    if (LOffset == ROffset)
484      return IsLQFIDBeforeRQFID;
485
486    return LOffset < ROffset;
487  }
488
489  /// \brief Set up a new query.
490  void setQueryFIDs(FileID LHS, FileID RHS, bool isLFIDBeforeRFID) {
491    assert(LHS != RHS);
492    LQueryFID = LHS;
493    RQueryFID = RHS;
494    IsLQFIDBeforeRQFID = isLFIDBeforeRFID;
495  }
496
497  void clear() {
498    LQueryFID = RQueryFID = FileID();
499    IsLQFIDBeforeRQFID = false;
500  }
501
502  void setCommonLoc(FileID commonFID, unsigned lCommonOffset,
503                    unsigned rCommonOffset) {
504    CommonFID = commonFID;
505    LCommonOffset = lCommonOffset;
506    RCommonOffset = rCommonOffset;
507  }
508
509};
510
511/// \brief This class handles loading and caching of source files into memory.
512///
513/// This object owns the MemoryBuffer objects for all of the loaded
514/// files and assigns unique FileID's for each unique \#include chain.
515///
516/// The SourceManager can be queried for information about SourceLocation
517/// objects, turning them into either spelling or expansion locations. Spelling
518/// locations represent where the bytes corresponding to a token came from and
519/// expansion locations represent where the location is in the user's view. In
520/// the case of a macro expansion, for example, the spelling location indicates
521/// where the expanded token came from and the expansion location specifies
522/// where it was expanded.
523class SourceManager : public RefCountedBase<SourceManager> {
524  /// \brief DiagnosticsEngine object.
525  DiagnosticsEngine &Diag;
526
527  FileManager &FileMgr;
528
529  mutable llvm::BumpPtrAllocator ContentCacheAlloc;
530
531  /// \brief Memoized information about all of the files tracked by this
532  /// SourceManager.
533  ///
534  /// This map allows us to merge ContentCache entries based
535  /// on their FileEntry*.  All ContentCache objects will thus have unique,
536  /// non-null, FileEntry pointers.
537  llvm::DenseMap<const FileEntry*, SrcMgr::ContentCache*> FileInfos;
538
539  /// \brief True if the ContentCache for files that are overriden by other
540  /// files, should report the original file name. Defaults to true.
541  bool OverridenFilesKeepOriginalName;
542
543  /// \brief True if non-system source files should be treated as volatile
544  /// (likely to change while trying to use them). Defaults to false.
545  bool UserFilesAreVolatile;
546
547  struct OverriddenFilesInfoTy {
548    /// \brief Files that have been overriden with the contents from another
549    /// file.
550    llvm::DenseMap<const FileEntry *, const FileEntry *> OverriddenFiles;
551    /// \brief Files that were overridden with a memory buffer.
552    llvm::DenseSet<const FileEntry *> OverriddenFilesWithBuffer;
553  };
554
555  /// \brief Lazily create the object keeping overridden files info, since
556  /// it is uncommonly used.
557  OwningPtr<OverriddenFilesInfoTy> OverriddenFilesInfo;
558
559  OverriddenFilesInfoTy &getOverriddenFilesInfo() {
560    if (!OverriddenFilesInfo)
561      OverriddenFilesInfo.reset(new OverriddenFilesInfoTy);
562    return *OverriddenFilesInfo;
563  }
564
565  /// \brief Information about various memory buffers that we have read in.
566  ///
567  /// All FileEntry* within the stored ContentCache objects are NULL,
568  /// as they do not refer to a file.
569  std::vector<SrcMgr::ContentCache*> MemBufferInfos;
570
571  /// \brief The table of SLocEntries that are local to this module.
572  ///
573  /// Positive FileIDs are indexes into this table. Entry 0 indicates an invalid
574  /// expansion.
575  std::vector<SrcMgr::SLocEntry> LocalSLocEntryTable;
576
577  /// \brief The table of SLocEntries that are loaded from other modules.
578  ///
579  /// Negative FileIDs are indexes into this table. To get from ID to an index,
580  /// use (-ID - 2).
581  mutable std::vector<SrcMgr::SLocEntry> LoadedSLocEntryTable;
582
583  /// \brief The starting offset of the next local SLocEntry.
584  ///
585  /// This is LocalSLocEntryTable.back().Offset + the size of that entry.
586  unsigned NextLocalOffset;
587
588  /// \brief The starting offset of the latest batch of loaded SLocEntries.
589  ///
590  /// This is LoadedSLocEntryTable.back().Offset, except that that entry might
591  /// not have been loaded, so that value would be unknown.
592  unsigned CurrentLoadedOffset;
593
594  /// \brief The highest possible offset is 2^31-1, so CurrentLoadedOffset
595  /// starts at 2^31.
596  static const unsigned MaxLoadedOffset = 1U << 31U;
597
598  /// \brief A bitmap that indicates whether the entries of LoadedSLocEntryTable
599  /// have already been loaded from the external source.
600  ///
601  /// Same indexing as LoadedSLocEntryTable.
602  std::vector<bool> SLocEntryLoaded;
603
604  /// \brief An external source for source location entries.
605  ExternalSLocEntrySource *ExternalSLocEntries;
606
607  /// \brief A one-entry cache to speed up getFileID.
608  ///
609  /// LastFileIDLookup records the last FileID looked up or created, because it
610  /// is very common to look up many tokens from the same file.
611  mutable FileID LastFileIDLookup;
612
613  /// \brief Holds information for \#line directives.
614  ///
615  /// This is referenced by indices from SLocEntryTable.
616  LineTableInfo *LineTable;
617
618  /// \brief These ivars serve as a cache used in the getLineNumber
619  /// method which is used to speedup getLineNumber calls to nearby locations.
620  mutable FileID LastLineNoFileIDQuery;
621  mutable SrcMgr::ContentCache *LastLineNoContentCache;
622  mutable unsigned LastLineNoFilePos;
623  mutable unsigned LastLineNoResult;
624
625  /// \brief The file ID for the main source file of the translation unit.
626  FileID MainFileID;
627
628  /// \brief The file ID for the precompiled preamble there is one.
629  FileID PreambleFileID;
630
631  // Statistics for -print-stats.
632  mutable unsigned NumLinearScans, NumBinaryProbes;
633
634  // Cache results for the isBeforeInTranslationUnit method.
635  mutable IsBeforeInTranslationUnitCache IsBeforeInTUCache;
636
637  // Cache for the "fake" buffer used for error-recovery purposes.
638  mutable llvm::MemoryBuffer *FakeBufferForRecovery;
639
640  mutable SrcMgr::ContentCache *FakeContentCacheForRecovery;
641
642  /// \brief Lazily computed map of macro argument chunks to their expanded
643  /// source location.
644  typedef std::map<unsigned, SourceLocation> MacroArgsMap;
645
646  mutable llvm::DenseMap<FileID, MacroArgsMap *> MacroArgsCacheMap;
647
648  // SourceManager doesn't support copy construction.
649  explicit SourceManager(const SourceManager&) LLVM_DELETED_FUNCTION;
650  void operator=(const SourceManager&) LLVM_DELETED_FUNCTION;
651public:
652  SourceManager(DiagnosticsEngine &Diag, FileManager &FileMgr,
653                bool UserFilesAreVolatile = false);
654  ~SourceManager();
655
656  void clearIDTables();
657
658  DiagnosticsEngine &getDiagnostics() const { return Diag; }
659
660  FileManager &getFileManager() const { return FileMgr; }
661
662  /// \brief Set true if the SourceManager should report the original file name
663  /// for contents of files that were overriden by other files.Defaults to true.
664  void setOverridenFilesKeepOriginalName(bool value) {
665    OverridenFilesKeepOriginalName = value;
666  }
667
668  /// \brief True if non-system source files should be treated as volatile
669  /// (likely to change while trying to use them).
670  bool userFilesAreVolatile() const { return UserFilesAreVolatile; }
671
672  /// \brief Create the FileID for a memory buffer that will represent the
673  /// FileID for the main source.
674  ///
675  /// One example of when this would be used is when the main source is read
676  /// from STDIN.
677  FileID createMainFileIDForMemBuffer(const llvm::MemoryBuffer *Buffer,
678                             SrcMgr::CharacteristicKind Kind = SrcMgr::C_User) {
679    assert(MainFileID.isInvalid() && "MainFileID already set!");
680    MainFileID = createFileIDForMemBuffer(Buffer, Kind);
681    return MainFileID;
682  }
683
684  //===--------------------------------------------------------------------===//
685  // MainFileID creation and querying methods.
686  //===--------------------------------------------------------------------===//
687
688  /// \brief Returns the FileID of the main source file.
689  FileID getMainFileID() const { return MainFileID; }
690
691  /// \brief Create the FileID for the main source file.
692  FileID createMainFileID(const FileEntry *SourceFile,
693                          SrcMgr::CharacteristicKind Kind = SrcMgr::C_User) {
694    assert(MainFileID.isInvalid() && "MainFileID already set!");
695    MainFileID = createFileID(SourceFile, SourceLocation(), Kind);
696    return MainFileID;
697  }
698
699  /// \brief Set the file ID for the main source file.
700  void setMainFileID(FileID FID) {
701    assert(MainFileID.isInvalid() && "MainFileID already set!");
702    MainFileID = FID;
703  }
704
705  /// \brief Set the file ID for the precompiled preamble.
706  void setPreambleFileID(FileID Preamble) {
707    assert(PreambleFileID.isInvalid() && "PreambleFileID already set!");
708    PreambleFileID = Preamble;
709  }
710
711  /// \brief Get the file ID for the precompiled preamble if there is one.
712  FileID getPreambleFileID() const { return PreambleFileID; }
713
714  //===--------------------------------------------------------------------===//
715  // Methods to create new FileID's and macro expansions.
716  //===--------------------------------------------------------------------===//
717
718  /// \brief Create a new FileID that represents the specified file
719  /// being \#included from the specified IncludePosition.
720  ///
721  /// This translates NULL into standard input.
722  FileID createFileID(const FileEntry *SourceFile, SourceLocation IncludePos,
723                      SrcMgr::CharacteristicKind FileCharacter,
724                      int LoadedID = 0, unsigned LoadedOffset = 0) {
725    const SrcMgr::ContentCache *
726      IR = getOrCreateContentCache(SourceFile,
727                              /*isSystemFile=*/FileCharacter != SrcMgr::C_User);
728    assert(IR && "getOrCreateContentCache() cannot return NULL");
729    return createFileID(IR, IncludePos, FileCharacter, LoadedID, LoadedOffset);
730  }
731
732  /// \brief Create a new FileID that represents the specified memory buffer.
733  ///
734  /// This does no caching of the buffer and takes ownership of the
735  /// MemoryBuffer, so only pass a MemoryBuffer to this once.
736  FileID createFileIDForMemBuffer(const llvm::MemoryBuffer *Buffer,
737                      SrcMgr::CharacteristicKind FileCharacter = SrcMgr::C_User,
738                                  int LoadedID = 0, unsigned LoadedOffset = 0,
739                                 SourceLocation IncludeLoc = SourceLocation()) {
740    return createFileID(createMemBufferContentCache(Buffer), IncludeLoc,
741                        FileCharacter, LoadedID, LoadedOffset);
742  }
743
744  /// \brief Return a new SourceLocation that encodes the
745  /// fact that a token from SpellingLoc should actually be referenced from
746  /// ExpansionLoc, and that it represents the expansion of a macro argument
747  /// into the function-like macro body.
748  SourceLocation createMacroArgExpansionLoc(SourceLocation Loc,
749                                            SourceLocation ExpansionLoc,
750                                            unsigned TokLength);
751
752  /// \brief Return a new SourceLocation that encodes the fact
753  /// that a token from SpellingLoc should actually be referenced from
754  /// ExpansionLoc.
755  SourceLocation createExpansionLoc(SourceLocation Loc,
756                                    SourceLocation ExpansionLocStart,
757                                    SourceLocation ExpansionLocEnd,
758                                    unsigned TokLength,
759                                    int LoadedID = 0,
760                                    unsigned LoadedOffset = 0);
761
762  /// \brief Retrieve the memory buffer associated with the given file.
763  ///
764  /// \param Invalid If non-NULL, will be set \c true if an error
765  /// occurs while retrieving the memory buffer.
766  const llvm::MemoryBuffer *getMemoryBufferForFile(const FileEntry *File,
767                                                   bool *Invalid = 0);
768
769  /// \brief Override the contents of the given source file by providing an
770  /// already-allocated buffer.
771  ///
772  /// \param SourceFile the source file whose contents will be overriden.
773  ///
774  /// \param Buffer the memory buffer whose contents will be used as the
775  /// data in the given source file.
776  ///
777  /// \param DoNotFree If true, then the buffer will not be freed when the
778  /// source manager is destroyed.
779  void overrideFileContents(const FileEntry *SourceFile,
780                            const llvm::MemoryBuffer *Buffer,
781                            bool DoNotFree = false);
782
783  /// \brief Override the given source file with another one.
784  ///
785  /// \param SourceFile the source file which will be overriden.
786  ///
787  /// \param NewFile the file whose contents will be used as the
788  /// data instead of the contents of the given source file.
789  void overrideFileContents(const FileEntry *SourceFile,
790                            const FileEntry *NewFile);
791
792  /// \brief Returns true if the file contents have been overridden.
793  bool isFileOverridden(const FileEntry *File) {
794    if (OverriddenFilesInfo) {
795      if (OverriddenFilesInfo->OverriddenFilesWithBuffer.count(File))
796        return true;
797      if (OverriddenFilesInfo->OverriddenFiles.find(File) !=
798          OverriddenFilesInfo->OverriddenFiles.end())
799        return true;
800    }
801    return false;
802  }
803
804  /// \brief Disable overridding the contents of a file, previously enabled
805  /// with #overrideFileContents.
806  ///
807  /// This should be called before parsing has begun.
808  void disableFileContentsOverride(const FileEntry *File);
809
810  //===--------------------------------------------------------------------===//
811  // FileID manipulation methods.
812  //===--------------------------------------------------------------------===//
813
814  /// \brief Return the buffer for the specified FileID.
815  ///
816  /// If there is an error opening this buffer the first time, this
817  /// manufactures a temporary buffer and returns a non-empty error string.
818  const llvm::MemoryBuffer *getBuffer(FileID FID, SourceLocation Loc,
819                                      bool *Invalid = 0) const {
820    bool MyInvalid = false;
821    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid);
822    if (MyInvalid || !Entry.isFile()) {
823      if (Invalid)
824        *Invalid = true;
825
826      return getFakeBufferForRecovery();
827    }
828
829    return Entry.getFile().getContentCache()->getBuffer(Diag, *this, Loc,
830                                                        Invalid);
831  }
832
833  const llvm::MemoryBuffer *getBuffer(FileID FID, bool *Invalid = 0) const {
834    bool MyInvalid = false;
835    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid);
836    if (MyInvalid || !Entry.isFile()) {
837      if (Invalid)
838        *Invalid = true;
839
840      return getFakeBufferForRecovery();
841    }
842
843    return Entry.getFile().getContentCache()->getBuffer(Diag, *this,
844                                                        SourceLocation(),
845                                                        Invalid);
846  }
847
848  /// \brief Returns the FileEntry record for the provided FileID.
849  const FileEntry *getFileEntryForID(FileID FID) const {
850    bool MyInvalid = false;
851    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid);
852    if (MyInvalid || !Entry.isFile())
853      return 0;
854
855    const SrcMgr::ContentCache *Content = Entry.getFile().getContentCache();
856    if (!Content)
857      return 0;
858    return Content->OrigEntry;
859  }
860
861  /// \brief Returns the FileEntry record for the provided SLocEntry.
862  const FileEntry *getFileEntryForSLocEntry(const SrcMgr::SLocEntry &sloc) const
863  {
864    const SrcMgr::ContentCache *Content = sloc.getFile().getContentCache();
865    if (!Content)
866      return 0;
867    return Content->OrigEntry;
868  }
869
870  /// \brief Return a StringRef to the source buffer data for the
871  /// specified FileID.
872  ///
873  /// \param FID The file ID whose contents will be returned.
874  /// \param Invalid If non-NULL, will be set true if an error occurred.
875  StringRef getBufferData(FileID FID, bool *Invalid = 0) const;
876
877  /// \brief Get the number of FileIDs (files and macros) that were created
878  /// during preprocessing of \p FID, including it.
879  unsigned getNumCreatedFIDsForFileID(FileID FID) const {
880    bool Invalid = false;
881    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
882    if (Invalid || !Entry.isFile())
883      return 0;
884
885    return Entry.getFile().NumCreatedFIDs;
886  }
887
888  /// \brief Set the number of FileIDs (files and macros) that were created
889  /// during preprocessing of \p FID, including it.
890  void setNumCreatedFIDsForFileID(FileID FID, unsigned NumFIDs) const {
891    bool Invalid = false;
892    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
893    if (Invalid || !Entry.isFile())
894      return;
895
896    assert(Entry.getFile().NumCreatedFIDs == 0 && "Already set!");
897    const_cast<SrcMgr::FileInfo &>(Entry.getFile()).NumCreatedFIDs = NumFIDs;
898  }
899
900  //===--------------------------------------------------------------------===//
901  // SourceLocation manipulation methods.
902  //===--------------------------------------------------------------------===//
903
904  /// \brief Return the FileID for a SourceLocation.
905  ///
906  /// This is a very hot method that is used for all SourceManager queries
907  /// that start with a SourceLocation object.  It is responsible for finding
908  /// the entry in SLocEntryTable which contains the specified location.
909  ///
910  FileID getFileID(SourceLocation SpellingLoc) const {
911    unsigned SLocOffset = SpellingLoc.getOffset();
912
913    // If our one-entry cache covers this offset, just return it.
914    if (isOffsetInFileID(LastFileIDLookup, SLocOffset))
915      return LastFileIDLookup;
916
917    return getFileIDSlow(SLocOffset);
918  }
919
920  /// \brief Return the filename of the file containing a SourceLocation.
921  StringRef getFilename(SourceLocation SpellingLoc) const {
922    if (const FileEntry *F = getFileEntryForID(getFileID(SpellingLoc)))
923      return F->getName();
924    return StringRef();
925  }
926
927  /// \brief Return the source location corresponding to the first byte of
928  /// the specified file.
929  SourceLocation getLocForStartOfFile(FileID FID) const {
930    bool Invalid = false;
931    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
932    if (Invalid || !Entry.isFile())
933      return SourceLocation();
934
935    unsigned FileOffset = Entry.getOffset();
936    return SourceLocation::getFileLoc(FileOffset);
937  }
938
939  /// \brief Return the source location corresponding to the last byte of the
940  /// specified file.
941  SourceLocation getLocForEndOfFile(FileID FID) const {
942    bool Invalid = false;
943    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
944    if (Invalid || !Entry.isFile())
945      return SourceLocation();
946
947    unsigned FileOffset = Entry.getOffset();
948    return SourceLocation::getFileLoc(FileOffset + getFileIDSize(FID) - 1);
949  }
950
951  /// \brief Returns the include location if \p FID is a \#include'd file
952  /// otherwise it returns an invalid location.
953  SourceLocation getIncludeLoc(FileID FID) const {
954    bool Invalid = false;
955    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
956    if (Invalid || !Entry.isFile())
957      return SourceLocation();
958
959    return Entry.getFile().getIncludeLoc();
960  }
961
962  /// \brief Given a SourceLocation object \p Loc, return the expansion
963  /// location referenced by the ID.
964  SourceLocation getExpansionLoc(SourceLocation Loc) const {
965    // Handle the non-mapped case inline, defer to out of line code to handle
966    // expansions.
967    if (Loc.isFileID()) return Loc;
968    return getExpansionLocSlowCase(Loc);
969  }
970
971  /// \brief Given \p Loc, if it is a macro location return the expansion
972  /// location or the spelling location, depending on if it comes from a
973  /// macro argument or not.
974  SourceLocation getFileLoc(SourceLocation Loc) const {
975    if (Loc.isFileID()) return Loc;
976    return getFileLocSlowCase(Loc);
977  }
978
979  /// \brief Return the start/end of the expansion information for an
980  /// expansion location.
981  ///
982  /// \pre \p Loc is required to be an expansion location.
983  std::pair<SourceLocation,SourceLocation>
984  getImmediateExpansionRange(SourceLocation Loc) const;
985
986  /// \brief Given a SourceLocation object, return the range of
987  /// tokens covered by the expansion the ultimate file.
988  std::pair<SourceLocation,SourceLocation>
989  getExpansionRange(SourceLocation Loc) const;
990
991
992  /// \brief Given a SourceLocation object, return the spelling
993  /// location referenced by the ID.
994  ///
995  /// This is the place where the characters that make up the lexed token
996  /// can be found.
997  SourceLocation getSpellingLoc(SourceLocation Loc) const {
998    // Handle the non-mapped case inline, defer to out of line code to handle
999    // expansions.
1000    if (Loc.isFileID()) return Loc;
1001    return getSpellingLocSlowCase(Loc);
1002  }
1003
1004  /// \brief Given a SourceLocation object, return the spelling location
1005  /// referenced by the ID.
1006  ///
1007  /// This is the first level down towards the place where the characters
1008  /// that make up the lexed token can be found.  This should not generally
1009  /// be used by clients.
1010  SourceLocation getImmediateSpellingLoc(SourceLocation Loc) const;
1011
1012  /// \brief Decompose the specified location into a raw FileID + Offset pair.
1013  ///
1014  /// The first element is the FileID, the second is the offset from the
1015  /// start of the buffer of the location.
1016  std::pair<FileID, unsigned> getDecomposedLoc(SourceLocation Loc) const {
1017    FileID FID = getFileID(Loc);
1018    bool Invalid = false;
1019    const SrcMgr::SLocEntry &E = getSLocEntry(FID, &Invalid);
1020    if (Invalid)
1021      return std::make_pair(FileID(), 0);
1022    return std::make_pair(FID, Loc.getOffset()-E.getOffset());
1023  }
1024
1025  /// \brief Decompose the specified location into a raw FileID + Offset pair.
1026  ///
1027  /// If the location is an expansion record, walk through it until we find
1028  /// the final location expanded.
1029  std::pair<FileID, unsigned>
1030  getDecomposedExpansionLoc(SourceLocation Loc) const {
1031    FileID FID = getFileID(Loc);
1032    bool Invalid = false;
1033    const SrcMgr::SLocEntry *E = &getSLocEntry(FID, &Invalid);
1034    if (Invalid)
1035      return std::make_pair(FileID(), 0);
1036
1037    unsigned Offset = Loc.getOffset()-E->getOffset();
1038    if (Loc.isFileID())
1039      return std::make_pair(FID, Offset);
1040
1041    return getDecomposedExpansionLocSlowCase(E);
1042  }
1043
1044  /// \brief Decompose the specified location into a raw FileID + Offset pair.
1045  ///
1046  /// If the location is an expansion record, walk through it until we find
1047  /// its spelling record.
1048  std::pair<FileID, unsigned>
1049  getDecomposedSpellingLoc(SourceLocation Loc) const {
1050    FileID FID = getFileID(Loc);
1051    bool Invalid = false;
1052    const SrcMgr::SLocEntry *E = &getSLocEntry(FID, &Invalid);
1053    if (Invalid)
1054      return std::make_pair(FileID(), 0);
1055
1056    unsigned Offset = Loc.getOffset()-E->getOffset();
1057    if (Loc.isFileID())
1058      return std::make_pair(FID, Offset);
1059    return getDecomposedSpellingLocSlowCase(E, Offset);
1060  }
1061
1062  /// \brief Returns the offset from the start of the file that the
1063  /// specified SourceLocation represents.
1064  ///
1065  /// This is not very meaningful for a macro ID.
1066  unsigned getFileOffset(SourceLocation SpellingLoc) const {
1067    return getDecomposedLoc(SpellingLoc).second;
1068  }
1069
1070  /// \brief Tests whether the given source location represents a macro
1071  /// argument's expansion into the function-like macro definition.
1072  ///
1073  /// Such source locations only appear inside of the expansion
1074  /// locations representing where a particular function-like macro was
1075  /// expanded.
1076  bool isMacroArgExpansion(SourceLocation Loc) const;
1077
1078  /// \brief Returns true if \p Loc is inside the [\p Start, +\p Length)
1079  /// chunk of the source location address space.
1080  ///
1081  /// If it's true and \p RelativeOffset is non-null, it will be set to the
1082  /// relative offset of \p Loc inside the chunk.
1083  bool isInSLocAddrSpace(SourceLocation Loc,
1084                         SourceLocation Start, unsigned Length,
1085                         unsigned *RelativeOffset = 0) const {
1086    assert(((Start.getOffset() < NextLocalOffset &&
1087               Start.getOffset()+Length <= NextLocalOffset) ||
1088            (Start.getOffset() >= CurrentLoadedOffset &&
1089                Start.getOffset()+Length < MaxLoadedOffset)) &&
1090           "Chunk is not valid SLoc address space");
1091    unsigned LocOffs = Loc.getOffset();
1092    unsigned BeginOffs = Start.getOffset();
1093    unsigned EndOffs = BeginOffs + Length;
1094    if (LocOffs >= BeginOffs && LocOffs < EndOffs) {
1095      if (RelativeOffset)
1096        *RelativeOffset = LocOffs - BeginOffs;
1097      return true;
1098    }
1099
1100    return false;
1101  }
1102
1103  /// \brief Return true if both \p LHS and \p RHS are in the local source
1104  /// location address space or the loaded one.
1105  ///
1106  /// If it's true and \p RelativeOffset is non-null, it will be set to the
1107  /// offset of \p RHS relative to \p LHS.
1108  bool isInSameSLocAddrSpace(SourceLocation LHS, SourceLocation RHS,
1109                             int *RelativeOffset) const {
1110    unsigned LHSOffs = LHS.getOffset(), RHSOffs = RHS.getOffset();
1111    bool LHSLoaded = LHSOffs >= CurrentLoadedOffset;
1112    bool RHSLoaded = RHSOffs >= CurrentLoadedOffset;
1113
1114    if (LHSLoaded == RHSLoaded) {
1115      if (RelativeOffset)
1116        *RelativeOffset = RHSOffs - LHSOffs;
1117      return true;
1118    }
1119
1120    return false;
1121  }
1122
1123  //===--------------------------------------------------------------------===//
1124  // Queries about the code at a SourceLocation.
1125  //===--------------------------------------------------------------------===//
1126
1127  /// \brief Return a pointer to the start of the specified location
1128  /// in the appropriate spelling MemoryBuffer.
1129  ///
1130  /// \param Invalid If non-NULL, will be set \c true if an error occurs.
1131  const char *getCharacterData(SourceLocation SL, bool *Invalid = 0) const;
1132
1133  /// \brief Return the column # for the specified file position.
1134  ///
1135  /// This is significantly cheaper to compute than the line number.  This
1136  /// returns zero if the column number isn't known.  This may only be called
1137  /// on a file sloc, so you must choose a spelling or expansion location
1138  /// before calling this method.
1139  unsigned getColumnNumber(FileID FID, unsigned FilePos,
1140                           bool *Invalid = 0) const;
1141  unsigned getSpellingColumnNumber(SourceLocation Loc, bool *Invalid = 0) const;
1142  unsigned getExpansionColumnNumber(SourceLocation Loc,
1143                                    bool *Invalid = 0) const;
1144  unsigned getPresumedColumnNumber(SourceLocation Loc, bool *Invalid = 0) const;
1145
1146
1147  /// \brief Given a SourceLocation, return the spelling line number
1148  /// for the position indicated.
1149  ///
1150  /// This requires building and caching a table of line offsets for the
1151  /// MemoryBuffer, so this is not cheap: use only when about to emit a
1152  /// diagnostic.
1153  unsigned getLineNumber(FileID FID, unsigned FilePos, bool *Invalid = 0) const;
1154  unsigned getSpellingLineNumber(SourceLocation Loc, bool *Invalid = 0) const;
1155  unsigned getExpansionLineNumber(SourceLocation Loc, bool *Invalid = 0) const;
1156  unsigned getPresumedLineNumber(SourceLocation Loc, bool *Invalid = 0) const;
1157
1158  /// \brief Return the filename or buffer identifier of the buffer the
1159  /// location is in.
1160  ///
1161  /// Note that this name does not respect \#line directives.  Use
1162  /// getPresumedLoc for normal clients.
1163  const char *getBufferName(SourceLocation Loc, bool *Invalid = 0) const;
1164
1165  /// \brief Return the file characteristic of the specified source
1166  /// location, indicating whether this is a normal file, a system
1167  /// header, or an "implicit extern C" system header.
1168  ///
1169  /// This state can be modified with flags on GNU linemarker directives like:
1170  /// \code
1171  ///   # 4 "foo.h" 3
1172  /// \endcode
1173  /// which changes all source locations in the current file after that to be
1174  /// considered to be from a system header.
1175  SrcMgr::CharacteristicKind getFileCharacteristic(SourceLocation Loc) const;
1176
1177  /// \brief Returns the "presumed" location of a SourceLocation specifies.
1178  ///
1179  /// A "presumed location" can be modified by \#line or GNU line marker
1180  /// directives.  This provides a view on the data that a user should see
1181  /// in diagnostics, for example.
1182  ///
1183  /// Note that a presumed location is always given as the expansion point of
1184  /// an expansion location, not at the spelling location.
1185  ///
1186  /// \returns The presumed location of the specified SourceLocation. If the
1187  /// presumed location cannot be calculate (e.g., because \p Loc is invalid
1188  /// or the file containing \p Loc has changed on disk), returns an invalid
1189  /// presumed location.
1190  PresumedLoc getPresumedLoc(SourceLocation Loc,
1191                             bool UseLineDirectives = true) const;
1192
1193  /// \brief Returns true if both SourceLocations correspond to the same file.
1194  bool isFromSameFile(SourceLocation Loc1, SourceLocation Loc2) const {
1195    return getFileID(Loc1) == getFileID(Loc2);
1196  }
1197
1198  /// \brief Returns true if the file of provided SourceLocation is the main
1199  /// file.
1200  bool isFromMainFile(SourceLocation Loc) const {
1201    return getFileID(Loc) == getMainFileID();
1202  }
1203
1204  /// \brief Returns if a SourceLocation is in a system header.
1205  bool isInSystemHeader(SourceLocation Loc) const {
1206    return getFileCharacteristic(Loc) != SrcMgr::C_User;
1207  }
1208
1209  /// \brief Returns if a SourceLocation is in an "extern C" system header.
1210  bool isInExternCSystemHeader(SourceLocation Loc) const {
1211    return getFileCharacteristic(Loc) == SrcMgr::C_ExternCSystem;
1212  }
1213
1214  /// \brief Returns whether \p Loc is expanded from a macro in a system header.
1215  bool isInSystemMacro(SourceLocation loc) {
1216    return loc.isMacroID() && isInSystemHeader(getSpellingLoc(loc));
1217  }
1218
1219  /// \brief The size of the SLocEnty that \p FID represents.
1220  unsigned getFileIDSize(FileID FID) const;
1221
1222  /// \brief Given a specific FileID, returns true if \p Loc is inside that
1223  /// FileID chunk and sets relative offset (offset of \p Loc from beginning
1224  /// of FileID) to \p relativeOffset.
1225  bool isInFileID(SourceLocation Loc, FileID FID,
1226                  unsigned *RelativeOffset = 0) const {
1227    unsigned Offs = Loc.getOffset();
1228    if (isOffsetInFileID(FID, Offs)) {
1229      if (RelativeOffset)
1230        *RelativeOffset = Offs - getSLocEntry(FID).getOffset();
1231      return true;
1232    }
1233
1234    return false;
1235  }
1236
1237  //===--------------------------------------------------------------------===//
1238  // Line Table Manipulation Routines
1239  //===--------------------------------------------------------------------===//
1240
1241  /// \brief Return the uniqued ID for the specified filename.
1242  ///
1243  unsigned getLineTableFilenameID(StringRef Str);
1244
1245  /// \brief Add a line note to the line table for the FileID and offset
1246  /// specified by Loc.
1247  ///
1248  /// If FilenameID is -1, it is considered to be unspecified.
1249  void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID);
1250  void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID,
1251                   bool IsFileEntry, bool IsFileExit,
1252                   bool IsSystemHeader, bool IsExternCHeader);
1253
1254  /// \brief Determine if the source manager has a line table.
1255  bool hasLineTable() const { return LineTable != 0; }
1256
1257  /// \brief Retrieve the stored line table.
1258  LineTableInfo &getLineTable();
1259
1260  //===--------------------------------------------------------------------===//
1261  // Queries for performance analysis.
1262  //===--------------------------------------------------------------------===//
1263
1264  /// \brief Return the total amount of physical memory allocated by the
1265  /// ContentCache allocator.
1266  size_t getContentCacheSize() const {
1267    return ContentCacheAlloc.getTotalMemory();
1268  }
1269
1270  struct MemoryBufferSizes {
1271    const size_t malloc_bytes;
1272    const size_t mmap_bytes;
1273
1274    MemoryBufferSizes(size_t malloc_bytes, size_t mmap_bytes)
1275      : malloc_bytes(malloc_bytes), mmap_bytes(mmap_bytes) {}
1276  };
1277
1278  /// \brief Return the amount of memory used by memory buffers, breaking down
1279  /// by heap-backed versus mmap'ed memory.
1280  MemoryBufferSizes getMemoryBufferSizes() const;
1281
1282  /// \brief Return the amount of memory used for various side tables and
1283  /// data structures in the SourceManager.
1284  size_t getDataStructureSizes() const;
1285
1286  //===--------------------------------------------------------------------===//
1287  // Other miscellaneous methods.
1288  //===--------------------------------------------------------------------===//
1289
1290  /// \brief Get the source location for the given file:line:col triplet.
1291  ///
1292  /// If the source file is included multiple times, the source location will
1293  /// be based upon the first inclusion.
1294  SourceLocation translateFileLineCol(const FileEntry *SourceFile,
1295                                      unsigned Line, unsigned Col) const;
1296
1297  /// \brief Get the FileID for the given file.
1298  ///
1299  /// If the source file is included multiple times, the FileID will be the
1300  /// first inclusion.
1301  FileID translateFile(const FileEntry *SourceFile) const;
1302
1303  /// \brief Get the source location in \p FID for the given line:col.
1304  /// Returns null location if \p FID is not a file SLocEntry.
1305  SourceLocation translateLineCol(FileID FID,
1306                                  unsigned Line, unsigned Col) const;
1307
1308  /// \brief If \p Loc points inside a function macro argument, the returned
1309  /// location will be the macro location in which the argument was expanded.
1310  /// If a macro argument is used multiple times, the expanded location will
1311  /// be at the first expansion of the argument.
1312  /// e.g.
1313  ///   MY_MACRO(foo);
1314  ///             ^
1315  /// Passing a file location pointing at 'foo', will yield a macro location
1316  /// where 'foo' was expanded into.
1317  SourceLocation getMacroArgExpandedLocation(SourceLocation Loc) const;
1318
1319  /// \brief Determines the order of 2 source locations in the translation unit.
1320  ///
1321  /// \returns true if LHS source location comes before RHS, false otherwise.
1322  bool isBeforeInTranslationUnit(SourceLocation LHS, SourceLocation RHS) const;
1323
1324  /// \brief Determines the order of 2 source locations in the "source location
1325  /// address space".
1326  bool isBeforeInSLocAddrSpace(SourceLocation LHS, SourceLocation RHS) const {
1327    return isBeforeInSLocAddrSpace(LHS, RHS.getOffset());
1328  }
1329
1330  /// \brief Determines the order of a source location and a source location
1331  /// offset in the "source location address space".
1332  ///
1333  /// Note that we always consider source locations loaded from
1334  bool isBeforeInSLocAddrSpace(SourceLocation LHS, unsigned RHS) const {
1335    unsigned LHSOffset = LHS.getOffset();
1336    bool LHSLoaded = LHSOffset >= CurrentLoadedOffset;
1337    bool RHSLoaded = RHS >= CurrentLoadedOffset;
1338    if (LHSLoaded == RHSLoaded)
1339      return LHSOffset < RHS;
1340
1341    return LHSLoaded;
1342  }
1343
1344  // Iterators over FileInfos.
1345  typedef llvm::DenseMap<const FileEntry*, SrcMgr::ContentCache*>
1346      ::const_iterator fileinfo_iterator;
1347  fileinfo_iterator fileinfo_begin() const { return FileInfos.begin(); }
1348  fileinfo_iterator fileinfo_end() const { return FileInfos.end(); }
1349  bool hasFileInfo(const FileEntry *File) const {
1350    return FileInfos.find(File) != FileInfos.end();
1351  }
1352
1353  /// \brief Print statistics to stderr.
1354  ///
1355  void PrintStats() const;
1356
1357  /// \brief Get the number of local SLocEntries we have.
1358  unsigned local_sloc_entry_size() const { return LocalSLocEntryTable.size(); }
1359
1360  /// \brief Get a local SLocEntry. This is exposed for indexing.
1361  const SrcMgr::SLocEntry &getLocalSLocEntry(unsigned Index,
1362                                             bool *Invalid = 0) const {
1363    assert(Index < LocalSLocEntryTable.size() && "Invalid index");
1364    return LocalSLocEntryTable[Index];
1365  }
1366
1367  /// \brief Get the number of loaded SLocEntries we have.
1368  unsigned loaded_sloc_entry_size() const { return LoadedSLocEntryTable.size();}
1369
1370  /// \brief Get a loaded SLocEntry. This is exposed for indexing.
1371  const SrcMgr::SLocEntry &getLoadedSLocEntry(unsigned Index,
1372                                              bool *Invalid = 0) const {
1373    assert(Index < LoadedSLocEntryTable.size() && "Invalid index");
1374    if (SLocEntryLoaded[Index])
1375      return LoadedSLocEntryTable[Index];
1376    return loadSLocEntry(Index, Invalid);
1377  }
1378
1379  const SrcMgr::SLocEntry &getSLocEntry(FileID FID, bool *Invalid = 0) const {
1380    if (FID.ID == 0 || FID.ID == -1) {
1381      if (Invalid) *Invalid = true;
1382      return LocalSLocEntryTable[0];
1383    }
1384    return getSLocEntryByID(FID.ID);
1385  }
1386
1387  unsigned getNextLocalOffset() const { return NextLocalOffset; }
1388
1389  void setExternalSLocEntrySource(ExternalSLocEntrySource *Source) {
1390    assert(LoadedSLocEntryTable.empty() &&
1391           "Invalidating existing loaded entries");
1392    ExternalSLocEntries = Source;
1393  }
1394
1395  /// \brief Allocate a number of loaded SLocEntries, which will be actually
1396  /// loaded on demand from the external source.
1397  ///
1398  /// NumSLocEntries will be allocated, which occupy a total of TotalSize space
1399  /// in the global source view. The lowest ID and the base offset of the
1400  /// entries will be returned.
1401  std::pair<int, unsigned>
1402  AllocateLoadedSLocEntries(unsigned NumSLocEntries, unsigned TotalSize);
1403
1404  /// \brief Returns true if \p Loc came from a PCH/Module.
1405  bool isLoadedSourceLocation(SourceLocation Loc) const {
1406    return Loc.getOffset() >= CurrentLoadedOffset;
1407  }
1408
1409  /// \brief Returns true if \p Loc did not come from a PCH/Module.
1410  bool isLocalSourceLocation(SourceLocation Loc) const {
1411    return Loc.getOffset() < NextLocalOffset;
1412  }
1413
1414  /// \brief Returns true if \p FID came from a PCH/Module.
1415  bool isLoadedFileID(FileID FID) const {
1416    assert(FID.ID != -1 && "Using FileID sentinel value");
1417    return FID.ID < 0;
1418  }
1419
1420  /// \brief Returns true if \p FID did not come from a PCH/Module.
1421  bool isLocalFileID(FileID FID) const {
1422    return !isLoadedFileID(FID);
1423  }
1424
1425  /// Get a presumed location suitable for displaying in a diagnostic message,
1426  /// taking into account macro arguments and expansions.
1427  PresumedLoc getPresumedLocForDisplay(SourceLocation Loc,
1428                                       bool UseLineDirectives = true) const{
1429    // This is a condensed form of the algorithm used by emitCaretDiagnostic to
1430    // walk to the top of the macro call stack.
1431    while (Loc.isMacroID()) {
1432      Loc = skipToMacroArgExpansion(Loc);
1433      Loc = getImmediateMacroCallerLoc(Loc);
1434    }
1435
1436    return getPresumedLoc(Loc, UseLineDirectives);
1437  }
1438
1439  /// Look through spelling locations for a macro argument expansion, and if
1440  /// found skip to it so that we can trace the argument rather than the macros
1441  /// in which that argument is used. If no macro argument expansion is found,
1442  /// don't skip anything and return the starting location.
1443  SourceLocation skipToMacroArgExpansion(SourceLocation StartLoc) const {
1444    for (SourceLocation L = StartLoc; L.isMacroID();
1445         L = getImmediateSpellingLoc(L)) {
1446      if (isMacroArgExpansion(L))
1447        return L;
1448    }
1449    // Otherwise just return initial location, there's nothing to skip.
1450    return StartLoc;
1451  }
1452
1453  /// Gets the location of the immediate macro caller, one level up the stack
1454  /// toward the initial macro typed into the source.
1455  SourceLocation getImmediateMacroCallerLoc(SourceLocation Loc) const {
1456    if (!Loc.isMacroID()) return Loc;
1457
1458    // When we have the location of (part of) an expanded parameter, its
1459    // spelling location points to the argument as typed into the macro call,
1460    // and therefore is used to locate the macro caller.
1461    if (isMacroArgExpansion(Loc))
1462      return getImmediateSpellingLoc(Loc);
1463
1464    // Otherwise, the caller of the macro is located where this macro is
1465    // expanded (while the spelling is part of the macro definition).
1466    return getImmediateExpansionRange(Loc).first;
1467  }
1468
1469  /// Gets the location of the immediate macro callee, one level down the stack
1470  /// toward the leaf macro.
1471  SourceLocation getImmediateMacroCalleeLoc(SourceLocation Loc) const {
1472    if (!Loc.isMacroID()) return Loc;
1473
1474    // When we have the location of (part of) an expanded parameter, its
1475    // expansion location points to the unexpanded parameter reference within
1476    // the macro definition (or callee).
1477    if (isMacroArgExpansion(Loc))
1478      return getImmediateExpansionRange(Loc).first;
1479
1480    // Otherwise, the callee of the macro is located where this location was
1481    // spelled inside the macro definition.
1482    return getImmediateSpellingLoc(Loc);
1483  }
1484
1485private:
1486  const llvm::MemoryBuffer *getFakeBufferForRecovery() const;
1487  const SrcMgr::ContentCache *getFakeContentCacheForRecovery() const;
1488
1489  const SrcMgr::SLocEntry &loadSLocEntry(unsigned Index, bool *Invalid) const;
1490
1491  /// \brief Get the entry with the given unwrapped FileID.
1492  const SrcMgr::SLocEntry &getSLocEntryByID(int ID) const {
1493    assert(ID != -1 && "Using FileID sentinel value");
1494    if (ID < 0)
1495      return getLoadedSLocEntryByID(ID);
1496    return getLocalSLocEntry(static_cast<unsigned>(ID));
1497  }
1498
1499  const SrcMgr::SLocEntry &getLoadedSLocEntryByID(int ID,
1500                                                  bool *Invalid = 0) const {
1501    return getLoadedSLocEntry(static_cast<unsigned>(-ID - 2), Invalid);
1502  }
1503
1504  /// Implements the common elements of storing an expansion info struct into
1505  /// the SLocEntry table and producing a source location that refers to it.
1506  SourceLocation createExpansionLocImpl(const SrcMgr::ExpansionInfo &Expansion,
1507                                        unsigned TokLength,
1508                                        int LoadedID = 0,
1509                                        unsigned LoadedOffset = 0);
1510
1511  /// \brief Return true if the specified FileID contains the
1512  /// specified SourceLocation offset.  This is a very hot method.
1513  inline bool isOffsetInFileID(FileID FID, unsigned SLocOffset) const {
1514    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID);
1515    // If the entry is after the offset, it can't contain it.
1516    if (SLocOffset < Entry.getOffset()) return false;
1517
1518    // If this is the very last entry then it does.
1519    if (FID.ID == -2)
1520      return true;
1521
1522    // If it is the last local entry, then it does if the location is local.
1523    if (FID.ID+1 == static_cast<int>(LocalSLocEntryTable.size()))
1524      return SLocOffset < NextLocalOffset;
1525
1526    // Otherwise, the entry after it has to not include it. This works for both
1527    // local and loaded entries.
1528    return SLocOffset < getSLocEntryByID(FID.ID+1).getOffset();
1529  }
1530
1531  /// \brief Create a new fileID for the specified ContentCache and
1532  /// include position.
1533  ///
1534  /// This works regardless of whether the ContentCache corresponds to a
1535  /// file or some other input source.
1536  FileID createFileID(const SrcMgr::ContentCache* File,
1537                      SourceLocation IncludePos,
1538                      SrcMgr::CharacteristicKind DirCharacter,
1539                      int LoadedID, unsigned LoadedOffset);
1540
1541  const SrcMgr::ContentCache *
1542    getOrCreateContentCache(const FileEntry *SourceFile,
1543                            bool isSystemFile = false);
1544
1545  /// \brief Create a new ContentCache for the specified  memory buffer.
1546  const SrcMgr::ContentCache*
1547  createMemBufferContentCache(const llvm::MemoryBuffer *Buf);
1548
1549  FileID getFileIDSlow(unsigned SLocOffset) const;
1550  FileID getFileIDLocal(unsigned SLocOffset) const;
1551  FileID getFileIDLoaded(unsigned SLocOffset) const;
1552
1553  SourceLocation getExpansionLocSlowCase(SourceLocation Loc) const;
1554  SourceLocation getSpellingLocSlowCase(SourceLocation Loc) const;
1555  SourceLocation getFileLocSlowCase(SourceLocation Loc) const;
1556
1557  std::pair<FileID, unsigned>
1558  getDecomposedExpansionLocSlowCase(const SrcMgr::SLocEntry *E) const;
1559  std::pair<FileID, unsigned>
1560  getDecomposedSpellingLocSlowCase(const SrcMgr::SLocEntry *E,
1561                                   unsigned Offset) const;
1562  void computeMacroArgsCache(MacroArgsMap *&MacroArgsCache, FileID FID) const;
1563  void associateFileChunkWithMacroArgExp(MacroArgsMap &MacroArgsCache,
1564                                         FileID FID,
1565                                         SourceLocation SpellLoc,
1566                                         SourceLocation ExpansionLoc,
1567                                         unsigned ExpansionLength) const;
1568  friend class ASTReader;
1569  friend class ASTWriter;
1570};
1571
1572/// \brief Comparison function object.
1573template<typename T>
1574class BeforeThanCompare;
1575
1576/// \brief Compare two source locations.
1577template<>
1578class BeforeThanCompare<SourceLocation> {
1579  SourceManager &SM;
1580
1581public:
1582  explicit BeforeThanCompare(SourceManager &SM) : SM(SM) { }
1583
1584  bool operator()(SourceLocation LHS, SourceLocation RHS) const {
1585    return SM.isBeforeInTranslationUnit(LHS, RHS);
1586  }
1587};
1588
1589/// \brief Compare two non-overlapping source ranges.
1590template<>
1591class BeforeThanCompare<SourceRange> {
1592  SourceManager &SM;
1593
1594public:
1595  explicit BeforeThanCompare(SourceManager &SM) : SM(SM) { }
1596
1597  bool operator()(SourceRange LHS, SourceRange RHS) {
1598    return SM.isBeforeInTranslationUnit(LHS.getBegin(), RHS.getBegin());
1599  }
1600};
1601
1602}  // end namespace clang
1603
1604#endif
1605