SourceManager.h revision 4f32786ac45210143654390177105eb749b614e9
1//===--- SourceManager.h - Track and cache source files ---------*- C++ -*-===//
2//
3//                     The LLVM Compiler Infrastructure
4//
5// This file is distributed under the University of Illinois Open Source
6// License. See LICENSE.TXT for details.
7//
8//===----------------------------------------------------------------------===//
9//
10//  This file defines the SourceManager interface.
11//
12//===----------------------------------------------------------------------===//
13
14#ifndef LLVM_CLANG_SOURCEMANAGER_H
15#define LLVM_CLANG_SOURCEMANAGER_H
16
17#include "clang/Basic/SourceLocation.h"
18#include "llvm/Support/Allocator.h"
19#include "llvm/Support/DataTypes.h"
20#include "llvm/ADT/PointerIntPair.h"
21#include "llvm/ADT/PointerUnion.h"
22#include "llvm/ADT/IntrusiveRefCntPtr.h"
23#include "llvm/ADT/DenseMap.h"
24#include <vector>
25#include <cassert>
26
27namespace llvm {
28class MemoryBuffer;
29class StringRef;
30}
31
32namespace clang {
33
34class Diagnostic;
35class SourceManager;
36class FileManager;
37class FileEntry;
38class LineTableInfo;
39
40/// SrcMgr - Public enums and private classes that are part of the
41/// SourceManager implementation.
42///
43namespace SrcMgr {
44  /// CharacteristicKind - This is used to represent whether a file or directory
45  /// holds normal user code, system code, or system code which is implicitly
46  /// 'extern "C"' in C++ mode.  Entire directories can be tagged with this
47  /// (this is maintained by DirectoryLookup and friends) as can specific
48  /// FileIDInfos when a #pragma system_header is seen or various other cases.
49  ///
50  enum CharacteristicKind {
51    C_User, C_System, C_ExternCSystem
52  };
53
54  /// ContentCache - One instance of this struct is kept for every file
55  /// loaded or used.  This object owns the MemoryBuffer object.
56  class ContentCache {
57    enum CCFlags {
58      /// \brief Whether the buffer is invalid.
59      InvalidFlag = 0x01,
60      /// \brief Whether the buffer should not be freed on destruction.
61      DoNotFreeFlag = 0x02
62    };
63
64    /// Buffer - The actual buffer containing the characters from the input
65    /// file.  This is owned by the ContentCache object.
66    /// The bits indicate indicates whether the buffer is invalid.
67    mutable llvm::PointerIntPair<const llvm::MemoryBuffer *, 2> Buffer;
68
69  public:
70    /// Reference to the file entry representing this ContentCache.
71    /// This reference does not own the FileEntry object.
72    /// It is possible for this to be NULL if
73    /// the ContentCache encapsulates an imaginary text buffer.
74    const FileEntry *OrigEntry;
75
76    /// \brief References the file which the contents were actually loaded from.
77    /// Can be different from 'Entry' if we overridden the contents of one file
78    /// with the contents of another file.
79    const FileEntry *ContentsEntry;
80
81    /// SourceLineCache - A bump pointer allocated array of offsets for each
82    /// source line.  This is lazily computed.  This is owned by the
83    /// SourceManager BumpPointerAllocator object.
84    unsigned *SourceLineCache;
85
86    /// NumLines - The number of lines in this ContentCache.  This is only valid
87    /// if SourceLineCache is non-null.
88    unsigned NumLines;
89
90    /// getBuffer - Returns the memory buffer for the associated content.
91    ///
92    /// \param Diag Object through which diagnostics will be emitted it the
93    /// buffer cannot be retrieved.
94    ///
95    /// \param Loc If specified, is the location that invalid file diagnostics
96    ///     will be emitted at.
97    ///
98    /// \param Invalid If non-NULL, will be set \c true if an error occurred.
99    const llvm::MemoryBuffer *getBuffer(Diagnostic &Diag,
100                                        const SourceManager &SM,
101                                        SourceLocation Loc = SourceLocation(),
102                                        bool *Invalid = 0) const;
103
104    /// getSize - Returns the size of the content encapsulated by this
105    ///  ContentCache. This can be the size of the source file or the size of an
106    ///  arbitrary scratch buffer.  If the ContentCache encapsulates a source
107    ///  file this size is retrieved from the file's FileEntry.
108    unsigned getSize() const;
109
110    /// getSizeBytesMapped - Returns the number of bytes actually mapped for
111    ///  this ContentCache.  This can be 0 if the MemBuffer was not actually
112    ///  instantiated.
113    unsigned getSizeBytesMapped() const;
114
115    void setBuffer(const llvm::MemoryBuffer *B) {
116      assert(!Buffer.getPointer() && "MemoryBuffer already set.");
117      Buffer.setPointer(B);
118      Buffer.setInt(false);
119    }
120
121    /// \brief Get the underlying buffer, returning NULL if the buffer is not
122    /// yet available.
123    const llvm::MemoryBuffer *getRawBuffer() const {
124      return Buffer.getPointer();
125    }
126
127    /// \brief Replace the existing buffer (which will be deleted)
128    /// with the given buffer.
129    void replaceBuffer(const llvm::MemoryBuffer *B, bool DoNotFree = false);
130
131    /// \brief Determine whether the buffer itself is invalid.
132    bool isBufferInvalid() const {
133      return Buffer.getInt() & InvalidFlag;
134    }
135
136    /// \brief Determine whether the buffer should be freed.
137    bool shouldFreeBuffer() const {
138      return (Buffer.getInt() & DoNotFreeFlag) == 0;
139    }
140
141    ContentCache(const FileEntry *Ent = 0)
142      : Buffer(0, false), OrigEntry(Ent), ContentsEntry(Ent),
143        SourceLineCache(0), NumLines(0) {}
144
145    ContentCache(const FileEntry *Ent, const FileEntry *contentEnt)
146      : Buffer(0, false), OrigEntry(Ent), ContentsEntry(contentEnt),
147        SourceLineCache(0), NumLines(0) {}
148
149    ~ContentCache();
150
151    /// The copy ctor does not allow copies where source object has either
152    ///  a non-NULL Buffer or SourceLineCache.  Ownership of allocated memory
153    ///  is not transfered, so this is a logical error.
154    ContentCache(const ContentCache &RHS)
155      : Buffer(0, false), SourceLineCache(0)
156    {
157      OrigEntry = RHS.OrigEntry;
158      ContentsEntry = RHS.ContentsEntry;
159
160      assert (RHS.Buffer.getPointer() == 0 && RHS.SourceLineCache == 0
161              && "Passed ContentCache object cannot own a buffer.");
162
163      NumLines = RHS.NumLines;
164    }
165
166  private:
167    // Disable assignments.
168    ContentCache &operator=(const ContentCache& RHS);
169  };
170
171  /// FileInfo - Information about a FileID, basically just the logical file
172  /// that it represents and include stack information.
173  ///
174  /// Each FileInfo has include stack information, indicating where it came
175  /// from.  This information encodes the #include chain that a token was
176  /// instantiated from.  The main include file has an invalid IncludeLoc.
177  ///
178  /// FileInfos contain a "ContentCache *", with the contents of the file.
179  ///
180  class FileInfo {
181    /// IncludeLoc - The location of the #include that brought in this file.
182    /// This is an invalid SLOC for the main file (top of the #include chain).
183    unsigned IncludeLoc;  // Really a SourceLocation
184
185    /// Data - This contains the ContentCache* and the bits indicating the
186    /// characteristic of the file and whether it has #line info, all bitmangled
187    /// together.
188    uintptr_t Data;
189  public:
190    /// get - Return a FileInfo object.
191    static FileInfo get(SourceLocation IL, const ContentCache *Con,
192                        CharacteristicKind FileCharacter) {
193      FileInfo X;
194      X.IncludeLoc = IL.getRawEncoding();
195      X.Data = (uintptr_t)Con;
196      assert((X.Data & 7) == 0 &&"ContentCache pointer insufficiently aligned");
197      assert((unsigned)FileCharacter < 4 && "invalid file character");
198      X.Data |= (unsigned)FileCharacter;
199      return X;
200    }
201
202    SourceLocation getIncludeLoc() const {
203      return SourceLocation::getFromRawEncoding(IncludeLoc);
204    }
205    const ContentCache* getContentCache() const {
206      return reinterpret_cast<const ContentCache*>(Data & ~7UL);
207    }
208
209    /// getCharacteristic - Return whether this is a system header or not.
210    CharacteristicKind getFileCharacteristic() const {
211      return (CharacteristicKind)(Data & 3);
212    }
213
214    /// hasLineDirectives - Return true if this FileID has #line directives in
215    /// it.
216    bool hasLineDirectives() const { return (Data & 4) != 0; }
217
218    /// setHasLineDirectives - Set the flag that indicates that this FileID has
219    /// line table entries associated with it.
220    void setHasLineDirectives() {
221      Data |= 4;
222    }
223  };
224
225  /// InstantiationInfo - Each InstantiationInfo encodes the Instantiation
226  /// location - where the token was ultimately instantiated, and the
227  /// SpellingLoc - where the actual character data for the token came from.
228  class InstantiationInfo {
229     // Really these are all SourceLocations.
230
231    /// SpellingLoc - Where the spelling for the token can be found.
232    unsigned SpellingLoc;
233
234    /// InstantiationLocStart/InstantiationLocEnd - In a macro expansion, these
235    /// indicate the start and end of the instantiation.  In object-like macros,
236    /// these will be the same.  In a function-like macro instantiation, the
237    /// start will be the identifier and the end will be the ')'.
238    unsigned InstantiationLocStart, InstantiationLocEnd;
239  public:
240    SourceLocation getSpellingLoc() const {
241      return SourceLocation::getFromRawEncoding(SpellingLoc);
242    }
243    SourceLocation getInstantiationLocStart() const {
244      return SourceLocation::getFromRawEncoding(InstantiationLocStart);
245    }
246    SourceLocation getInstantiationLocEnd() const {
247      return SourceLocation::getFromRawEncoding(InstantiationLocEnd);
248    }
249
250    std::pair<SourceLocation,SourceLocation> getInstantiationLocRange() const {
251      return std::make_pair(getInstantiationLocStart(),
252                            getInstantiationLocEnd());
253    }
254
255    /// get - Return a InstantiationInfo for an expansion.  IL specifies
256    /// the instantiation location (where the macro is expanded), and SL
257    /// specifies the spelling location (where the characters from the token
258    /// come from).  IL and PL can both refer to normal File SLocs or
259    /// instantiation locations.
260    static InstantiationInfo get(SourceLocation ILStart, SourceLocation ILEnd,
261                                 SourceLocation SL) {
262      InstantiationInfo X;
263      X.SpellingLoc = SL.getRawEncoding();
264      X.InstantiationLocStart = ILStart.getRawEncoding();
265      X.InstantiationLocEnd = ILEnd.getRawEncoding();
266      return X;
267    }
268  };
269
270  /// SLocEntry - This is a discriminated union of FileInfo and
271  /// InstantiationInfo.  SourceManager keeps an array of these objects, and
272  /// they are uniquely identified by the FileID datatype.
273  class SLocEntry {
274    unsigned Offset;   // low bit is set for instantiation info.
275    union {
276      FileInfo File;
277      InstantiationInfo Instantiation;
278    };
279  public:
280    unsigned getOffset() const { return Offset >> 1; }
281
282    bool isInstantiation() const { return Offset & 1; }
283    bool isFile() const { return !isInstantiation(); }
284
285    const FileInfo &getFile() const {
286      assert(isFile() && "Not a file SLocEntry!");
287      return File;
288    }
289
290    const InstantiationInfo &getInstantiation() const {
291      assert(isInstantiation() && "Not an instantiation SLocEntry!");
292      return Instantiation;
293    }
294
295    static SLocEntry get(unsigned Offset, const FileInfo &FI) {
296      SLocEntry E;
297      E.Offset = Offset << 1;
298      E.File = FI;
299      return E;
300    }
301
302    static SLocEntry get(unsigned Offset, const InstantiationInfo &II) {
303      SLocEntry E;
304      E.Offset = (Offset << 1) | 1;
305      E.Instantiation = II;
306      return E;
307    }
308  };
309}  // end SrcMgr namespace.
310
311/// \brief External source of source location entries.
312class ExternalSLocEntrySource {
313public:
314  virtual ~ExternalSLocEntrySource();
315
316  /// \brief Read the source location entry with index ID.
317  virtual void ReadSLocEntry(unsigned ID) = 0;
318};
319
320
321/// IsBeforeInTranslationUnitCache - This class holds the cache used by
322/// isBeforeInTranslationUnit.  The cache structure is complex enough to be
323/// worth breaking out of SourceManager.
324class IsBeforeInTranslationUnitCache {
325  /// L/R QueryFID - These are the FID's of the cached query.  If these match up
326  /// with a subsequent query, the result can be reused.
327  FileID LQueryFID, RQueryFID;
328
329  /// CommonFID - This is the file found in common between the two #include
330  /// traces.  It is the nearest common ancestor of the #include tree.
331  FileID CommonFID;
332
333  /// L/R CommonOffset - This is the offset of the previous query in CommonFID.
334  /// Usually, this represents the location of the #include for QueryFID, but if
335  /// LQueryFID is a parent of RQueryFID (or vise versa) then these can be a
336  /// random token in the parent.
337  unsigned LCommonOffset, RCommonOffset;
338public:
339
340  /// isCacheValid - Return true if the currently cached values match up with
341  /// the specified LHS/RHS query.  If not, we can't use the cache.
342  bool isCacheValid(FileID LHS, FileID RHS) const {
343    return LQueryFID == LHS && RQueryFID == RHS;
344  }
345
346  /// getCachedResult - If the cache is valid, compute the result given the
347  /// specified offsets in the LHS/RHS FID's.
348  bool getCachedResult(unsigned LOffset, unsigned ROffset) const {
349    // If one of the query files is the common file, use the offset.  Otherwise,
350    // use the #include loc in the common file.
351    if (LQueryFID != CommonFID) LOffset = LCommonOffset;
352    if (RQueryFID != CommonFID) ROffset = RCommonOffset;
353    return LOffset < ROffset;
354  }
355
356  // Set up a new query.
357  void setQueryFIDs(FileID LHS, FileID RHS) {
358    LQueryFID = LHS;
359    RQueryFID = RHS;
360  }
361
362  void setCommonLoc(FileID commonFID, unsigned lCommonOffset,
363                    unsigned rCommonOffset) {
364    CommonFID = commonFID;
365    LCommonOffset = lCommonOffset;
366    RCommonOffset = rCommonOffset;
367  }
368
369};
370
371/// SourceManager - This file handles loading and caching of source files into
372/// memory.  This object owns the MemoryBuffer objects for all of the loaded
373/// files and assigns unique FileID's for each unique #include chain.
374///
375/// The SourceManager can be queried for information about SourceLocation
376/// objects, turning them into either spelling or instantiation locations.
377/// Spelling locations represent where the bytes corresponding to a token came
378/// from and instantiation locations represent where the location is in the
379/// user's view.  In the case of a macro expansion, for example, the spelling
380/// location indicates where the expanded token came from and the instantiation
381/// location specifies where it was expanded.
382class SourceManager : public llvm::RefCountedBase<SourceManager> {
383  /// \brief Diagnostic object.
384  Diagnostic &Diag;
385
386  FileManager &FileMgr;
387
388  mutable llvm::BumpPtrAllocator ContentCacheAlloc;
389
390  /// FileInfos - Memoized information about all of the files tracked by this
391  /// SourceManager.  This set allows us to merge ContentCache entries based
392  /// on their FileEntry*.  All ContentCache objects will thus have unique,
393  /// non-null, FileEntry pointers.
394  llvm::DenseMap<const FileEntry*, SrcMgr::ContentCache*> FileInfos;
395
396  /// \brief True if the ContentCache for files that are overriden by other
397  /// files, should report the original file name. Defaults to true.
398  bool OverridenFilesKeepOriginalName;
399
400  /// \brief Files that have been overriden with the contents from another file.
401  llvm::DenseMap<const FileEntry *, const FileEntry *> OverriddenFiles;
402
403  /// MemBufferInfos - Information about various memory buffers that we have
404  /// read in.  All FileEntry* within the stored ContentCache objects are NULL,
405  /// as they do not refer to a file.
406  std::vector<SrcMgr::ContentCache*> MemBufferInfos;
407
408  /// SLocEntryTable - This is an array of SLocEntry's that we have created.
409  /// FileID is an index into this vector.  This array is sorted by the offset.
410  std::vector<SrcMgr::SLocEntry> SLocEntryTable;
411  /// NextOffset - This is the next available offset that a new SLocEntry can
412  /// start at.  It is SLocEntryTable.back().getOffset()+size of back() entry.
413  unsigned NextOffset;
414
415  /// \brief If source location entries are being lazily loaded from
416  /// an external source, this vector indicates whether the Ith source
417  /// location entry has already been loaded from the external storage.
418  std::vector<bool> SLocEntryLoaded;
419
420  /// \brief An external source for source location entries.
421  ExternalSLocEntrySource *ExternalSLocEntries;
422
423  /// LastFileIDLookup - This is a one-entry cache to speed up getFileID.
424  /// LastFileIDLookup records the last FileID looked up or created, because it
425  /// is very common to look up many tokens from the same file.
426  mutable FileID LastFileIDLookup;
427
428  /// LineTable - This holds information for #line directives.  It is referenced
429  /// by indices from SLocEntryTable.
430  LineTableInfo *LineTable;
431
432  /// LastLineNo - These ivars serve as a cache used in the getLineNumber
433  /// method which is used to speedup getLineNumber calls to nearby locations.
434  mutable FileID LastLineNoFileIDQuery;
435  mutable SrcMgr::ContentCache *LastLineNoContentCache;
436  mutable unsigned LastLineNoFilePos;
437  mutable unsigned LastLineNoResult;
438
439  /// MainFileID - The file ID for the main source file of the translation unit.
440  FileID MainFileID;
441
442  // Statistics for -print-stats.
443  mutable unsigned NumLinearScans, NumBinaryProbes;
444
445  // Cache results for the isBeforeInTranslationUnit method.
446  mutable IsBeforeInTranslationUnitCache IsBeforeInTUCache;
447
448  // SourceManager doesn't support copy construction.
449  explicit SourceManager(const SourceManager&);
450  void operator=(const SourceManager&);
451public:
452  SourceManager(Diagnostic &Diag, FileManager &FileMgr);
453  ~SourceManager();
454
455  void clearIDTables();
456
457  Diagnostic &getDiagnostics() const { return Diag; }
458
459  FileManager &getFileManager() const { return FileMgr; }
460
461  /// \brief Set true if the SourceManager should report the original file name
462  /// for contents of files that were overriden by other files.Defaults to true.
463  void setOverridenFilesKeepOriginalName(bool value) {
464    OverridenFilesKeepOriginalName = value;
465  }
466
467  //===--------------------------------------------------------------------===//
468  // MainFileID creation and querying methods.
469  //===--------------------------------------------------------------------===//
470
471  /// getMainFileID - Returns the FileID of the main source file.
472  FileID getMainFileID() const { return MainFileID; }
473
474  /// createMainFileID - Create the FileID for the main source file.
475  FileID createMainFileID(const FileEntry *SourceFile) {
476    assert(MainFileID.isInvalid() && "MainFileID already set!");
477    MainFileID = createFileID(SourceFile, SourceLocation(), SrcMgr::C_User);
478    return MainFileID;
479  }
480
481  /// \brief Set the file ID for the precompiled preamble, which is also the
482  /// main file.
483  void SetPreambleFileID(FileID Preamble) {
484    assert(MainFileID.isInvalid() && "MainFileID already set!");
485    MainFileID = Preamble;
486  }
487
488  //===--------------------------------------------------------------------===//
489  // Methods to create new FileID's and instantiations.
490  //===--------------------------------------------------------------------===//
491
492  /// createFileID - Create a new FileID that represents the specified file
493  /// being #included from the specified IncludePosition.  This returns 0 on
494  /// error and translates NULL into standard input.
495  /// PreallocateID should be non-zero to specify which pre-allocated,
496  /// lazily computed source location is being filled in by this operation.
497  FileID createFileID(const FileEntry *SourceFile, SourceLocation IncludePos,
498                      SrcMgr::CharacteristicKind FileCharacter,
499                      unsigned PreallocatedID = 0,
500                      unsigned Offset = 0) {
501    const SrcMgr::ContentCache *IR = getOrCreateContentCache(SourceFile);
502    assert(IR && "getOrCreateContentCache() cannot return NULL");
503    return createFileID(IR, IncludePos, FileCharacter, PreallocatedID, Offset);
504  }
505
506  /// createFileIDForMemBuffer - Create a new FileID that represents the
507  /// specified memory buffer.  This does no caching of the buffer and takes
508  /// ownership of the MemoryBuffer, so only pass a MemoryBuffer to this once.
509  FileID createFileIDForMemBuffer(const llvm::MemoryBuffer *Buffer,
510                                  unsigned PreallocatedID = 0,
511                                  unsigned Offset = 0) {
512    return createFileID(createMemBufferContentCache(Buffer), SourceLocation(),
513                        SrcMgr::C_User, PreallocatedID, Offset);
514  }
515
516  /// createMainFileIDForMembuffer - Create the FileID for a memory buffer
517  ///  that will represent the FileID for the main source.  One example
518  ///  of when this would be used is when the main source is read from STDIN.
519  FileID createMainFileIDForMemBuffer(const llvm::MemoryBuffer *Buffer) {
520    assert(MainFileID.isInvalid() && "MainFileID already set!");
521    MainFileID = createFileIDForMemBuffer(Buffer);
522    return MainFileID;
523  }
524
525  /// createInstantiationLoc - Return a new SourceLocation that encodes the fact
526  /// that a token at Loc should actually be referenced from InstantiationLoc.
527  /// TokLength is the length of the token being instantiated.
528  SourceLocation createInstantiationLoc(SourceLocation Loc,
529                                        SourceLocation InstantiationLocStart,
530                                        SourceLocation InstantiationLocEnd,
531                                        unsigned TokLength,
532                                        unsigned PreallocatedID = 0,
533                                        unsigned Offset = 0);
534
535  /// \brief Retrieve the memory buffer associated with the given file.
536  ///
537  /// \param Invalid If non-NULL, will be set \c true if an error
538  /// occurs while retrieving the memory buffer.
539  const llvm::MemoryBuffer *getMemoryBufferForFile(const FileEntry *File,
540                                                   bool *Invalid = 0);
541
542  /// \brief Override the contents of the given source file by providing an
543  /// already-allocated buffer.
544  ///
545  /// \param SourceFile the source file whose contents will be overriden.
546  ///
547  /// \param Buffer the memory buffer whose contents will be used as the
548  /// data in the given source file.
549  ///
550  /// \param DoNotFree If true, then the buffer will not be freed when the
551  /// source manager is destroyed.
552  void overrideFileContents(const FileEntry *SourceFile,
553                            const llvm::MemoryBuffer *Buffer,
554                            bool DoNotFree = false);
555
556  /// \brief Override the the given source file with another one.
557  ///
558  /// \param SourceFile the source file which will be overriden.
559  ///
560  /// \param NewFile the file whose contents will be used as the
561  /// data instead of the contents of the given source file.
562  void overrideFileContents(const FileEntry *SourceFile,
563                            const FileEntry *NewFile);
564
565  //===--------------------------------------------------------------------===//
566  // FileID manipulation methods.
567  //===--------------------------------------------------------------------===//
568
569  /// getBuffer - Return the buffer for the specified FileID. If there is an
570  /// error opening this buffer the first time, this manufactures a temporary
571  /// buffer and returns a non-empty error string.
572  const llvm::MemoryBuffer *getBuffer(FileID FID, SourceLocation Loc,
573                                      bool *Invalid = 0) const {
574    return getSLocEntry(FID).getFile().getContentCache()
575       ->getBuffer(Diag, *this, Loc, Invalid);
576  }
577
578  const llvm::MemoryBuffer *getBuffer(FileID FID, bool *Invalid = 0) const {
579    return getSLocEntry(FID).getFile().getContentCache()
580       ->getBuffer(Diag, *this, SourceLocation(), Invalid);
581  }
582
583  /// getFileEntryForID - Returns the FileEntry record for the provided FileID.
584  const FileEntry *getFileEntryForID(FileID FID) const {
585    return getSLocEntry(FID).getFile().getContentCache()->OrigEntry;
586  }
587
588  /// getBufferData - Return a StringRef to the source buffer data for the
589  /// specified FileID.
590  ///
591  /// \param FID The file ID whose contents will be returned.
592  /// \param Invalid If non-NULL, will be set true if an error occurred.
593  llvm::StringRef getBufferData(FileID FID, bool *Invalid = 0) const;
594
595
596  //===--------------------------------------------------------------------===//
597  // SourceLocation manipulation methods.
598  //===--------------------------------------------------------------------===//
599
600  /// getFileID - Return the FileID for a SourceLocation.  This is a very
601  /// hot method that is used for all SourceManager queries that start with a
602  /// SourceLocation object.  It is responsible for finding the entry in
603  /// SLocEntryTable which contains the specified location.
604  ///
605  FileID getFileID(SourceLocation SpellingLoc) const {
606    unsigned SLocOffset = SpellingLoc.getOffset();
607
608    // If our one-entry cache covers this offset, just return it.
609    if (isOffsetInFileID(LastFileIDLookup, SLocOffset))
610      return LastFileIDLookup;
611
612    return getFileIDSlow(SLocOffset);
613  }
614
615  /// getLocForStartOfFile - Return the source location corresponding to the
616  /// first byte of the specified file.
617  SourceLocation getLocForStartOfFile(FileID FID) const {
618    assert(FID.ID < SLocEntryTable.size() && "FileID out of range");
619    assert(getSLocEntry(FID).isFile() && "FileID is not a file");
620    unsigned FileOffset = getSLocEntry(FID).getOffset();
621    return SourceLocation::getFileLoc(FileOffset);
622  }
623
624  /// getInstantiationLoc - Given a SourceLocation object, return the
625  /// instantiation location referenced by the ID.
626  SourceLocation getInstantiationLoc(SourceLocation Loc) const {
627    // Handle the non-mapped case inline, defer to out of line code to handle
628    // instantiations.
629    if (Loc.isFileID()) return Loc;
630    return getInstantiationLocSlowCase(Loc);
631  }
632
633  /// getImmediateInstantiationRange - Loc is required to be an instantiation
634  /// location.  Return the start/end of the instantiation information.
635  std::pair<SourceLocation,SourceLocation>
636  getImmediateInstantiationRange(SourceLocation Loc) const;
637
638  /// getInstantiationRange - Given a SourceLocation object, return the
639  /// range of tokens covered by the instantiation in the ultimate file.
640  std::pair<SourceLocation,SourceLocation>
641  getInstantiationRange(SourceLocation Loc) const;
642
643
644  /// getSpellingLoc - Given a SourceLocation object, return the spelling
645  /// location referenced by the ID.  This is the place where the characters
646  /// that make up the lexed token can be found.
647  SourceLocation getSpellingLoc(SourceLocation Loc) const {
648    // Handle the non-mapped case inline, defer to out of line code to handle
649    // instantiations.
650    if (Loc.isFileID()) return Loc;
651    return getSpellingLocSlowCase(Loc);
652  }
653
654  /// getImmediateSpellingLoc - Given a SourceLocation object, return the
655  /// spelling location referenced by the ID.  This is the first level down
656  /// towards the place where the characters that make up the lexed token can be
657  /// found.  This should not generally be used by clients.
658  SourceLocation getImmediateSpellingLoc(SourceLocation Loc) const;
659
660  /// getDecomposedLoc - Decompose the specified location into a raw FileID +
661  /// Offset pair.  The first element is the FileID, the second is the
662  /// offset from the start of the buffer of the location.
663  std::pair<FileID, unsigned> getDecomposedLoc(SourceLocation Loc) const {
664    FileID FID = getFileID(Loc);
665    return std::make_pair(FID, Loc.getOffset()-getSLocEntry(FID).getOffset());
666  }
667
668  /// getDecomposedInstantiationLoc - Decompose the specified location into a
669  /// raw FileID + Offset pair.  If the location is an instantiation record,
670  /// walk through it until we find the final location instantiated.
671  std::pair<FileID, unsigned>
672  getDecomposedInstantiationLoc(SourceLocation Loc) const {
673    FileID FID = getFileID(Loc);
674    const SrcMgr::SLocEntry *E = &getSLocEntry(FID);
675
676    unsigned Offset = Loc.getOffset()-E->getOffset();
677    if (Loc.isFileID())
678      return std::make_pair(FID, Offset);
679
680    return getDecomposedInstantiationLocSlowCase(E, Offset);
681  }
682
683  /// getDecomposedSpellingLoc - Decompose the specified location into a raw
684  /// FileID + Offset pair.  If the location is an instantiation record, walk
685  /// through it until we find its spelling record.
686  std::pair<FileID, unsigned>
687  getDecomposedSpellingLoc(SourceLocation Loc) const {
688    FileID FID = getFileID(Loc);
689    const SrcMgr::SLocEntry *E = &getSLocEntry(FID);
690
691    unsigned Offset = Loc.getOffset()-E->getOffset();
692    if (Loc.isFileID())
693      return std::make_pair(FID, Offset);
694    return getDecomposedSpellingLocSlowCase(E, Offset);
695  }
696
697  /// getFileOffset - This method returns the offset from the start
698  /// of the file that the specified SourceLocation represents. This is not very
699  /// meaningful for a macro ID.
700  unsigned getFileOffset(SourceLocation SpellingLoc) const {
701    return getDecomposedLoc(SpellingLoc).second;
702  }
703
704
705  //===--------------------------------------------------------------------===//
706  // Queries about the code at a SourceLocation.
707  //===--------------------------------------------------------------------===//
708
709  /// getCharacterData - Return a pointer to the start of the specified location
710  /// in the appropriate spelling MemoryBuffer.
711  ///
712  /// \param Invalid If non-NULL, will be set \c true if an error occurs.
713  const char *getCharacterData(SourceLocation SL, bool *Invalid = 0) const;
714
715  /// getColumnNumber - Return the column # for the specified file position.
716  /// This is significantly cheaper to compute than the line number.  This
717  /// returns zero if the column number isn't known.  This may only be called on
718  /// a file sloc, so you must choose a spelling or instantiation location
719  /// before calling this method.
720  unsigned getColumnNumber(FileID FID, unsigned FilePos,
721                           bool *Invalid = 0) const;
722  unsigned getSpellingColumnNumber(SourceLocation Loc, bool *Invalid = 0) const;
723  unsigned getInstantiationColumnNumber(SourceLocation Loc,
724                                        bool *Invalid = 0) const;
725  unsigned getPresumedColumnNumber(SourceLocation Loc, bool *Invalid = 0) const;
726
727
728  /// getLineNumber - Given a SourceLocation, return the spelling line number
729  /// for the position indicated.  This requires building and caching a table of
730  /// line offsets for the MemoryBuffer, so this is not cheap: use only when
731  /// about to emit a diagnostic.
732  unsigned getLineNumber(FileID FID, unsigned FilePos, bool *Invalid = 0) const;
733  unsigned getSpellingLineNumber(SourceLocation Loc, bool *Invalid = 0) const;
734  unsigned getInstantiationLineNumber(SourceLocation Loc,
735                                      bool *Invalid = 0) const;
736  unsigned getPresumedLineNumber(SourceLocation Loc, bool *Invalid = 0) const;
737
738  /// Return the filename or buffer identifier of the buffer the location is in.
739  /// Note that this name does not respect #line directives.  Use getPresumedLoc
740  /// for normal clients.
741  const char *getBufferName(SourceLocation Loc, bool *Invalid = 0) const;
742
743  /// getFileCharacteristic - return the file characteristic of the specified
744  /// source location, indicating whether this is a normal file, a system
745  /// header, or an "implicit extern C" system header.
746  ///
747  /// This state can be modified with flags on GNU linemarker directives like:
748  ///   # 4 "foo.h" 3
749  /// which changes all source locations in the current file after that to be
750  /// considered to be from a system header.
751  SrcMgr::CharacteristicKind getFileCharacteristic(SourceLocation Loc) const;
752
753  /// getPresumedLoc - This method returns the "presumed" location of a
754  /// SourceLocation specifies.  A "presumed location" can be modified by #line
755  /// or GNU line marker directives.  This provides a view on the data that a
756  /// user should see in diagnostics, for example.
757  ///
758  /// Note that a presumed location is always given as the instantiation point
759  /// of an instantiation location, not at the spelling location.
760  ///
761  /// \returns The presumed location of the specified SourceLocation. If the
762  /// presumed location cannot be calculate (e.g., because \p Loc is invalid
763  /// or the file containing \p Loc has changed on disk), returns an invalid
764  /// presumed location.
765  PresumedLoc getPresumedLoc(SourceLocation Loc) const;
766
767  /// isFromSameFile - Returns true if both SourceLocations correspond to
768  ///  the same file.
769  bool isFromSameFile(SourceLocation Loc1, SourceLocation Loc2) const {
770    return getFileID(Loc1) == getFileID(Loc2);
771  }
772
773  /// isFromMainFile - Returns true if the file of provided SourceLocation is
774  ///   the main file.
775  bool isFromMainFile(SourceLocation Loc) const {
776    return getFileID(Loc) == getMainFileID();
777  }
778
779  /// isInSystemHeader - Returns if a SourceLocation is in a system header.
780  bool isInSystemHeader(SourceLocation Loc) const {
781    return getFileCharacteristic(Loc) != SrcMgr::C_User;
782  }
783
784  /// isInExternCSystemHeader - Returns if a SourceLocation is in an "extern C"
785  /// system header.
786  bool isInExternCSystemHeader(SourceLocation Loc) const {
787    return getFileCharacteristic(Loc) == SrcMgr::C_ExternCSystem;
788  }
789
790  //===--------------------------------------------------------------------===//
791  // Line Table Manipulation Routines
792  //===--------------------------------------------------------------------===//
793
794  /// getLineTableFilenameID - Return the uniqued ID for the specified filename.
795  ///
796  unsigned getLineTableFilenameID(const char *Ptr, unsigned Len);
797
798  /// AddLineNote - Add a line note to the line table for the FileID and offset
799  /// specified by Loc.  If FilenameID is -1, it is considered to be
800  /// unspecified.
801  void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID);
802  void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID,
803                   bool IsFileEntry, bool IsFileExit,
804                   bool IsSystemHeader, bool IsExternCHeader);
805
806  /// \brief Determine if the source manager has a line table.
807  bool hasLineTable() const { return LineTable != 0; }
808
809  /// \brief Retrieve the stored line table.
810  LineTableInfo &getLineTable();
811
812  //===--------------------------------------------------------------------===//
813  // Other miscellaneous methods.
814  //===--------------------------------------------------------------------===//
815
816  /// \brief Get the source location for the given file:line:col triplet.
817  ///
818  /// If the source file is included multiple times, the source location will
819  /// be based upon the first inclusion.
820  SourceLocation getLocation(const FileEntry *SourceFile,
821                             unsigned Line, unsigned Col);
822
823  /// \brief Determines the order of 2 source locations in the translation unit.
824  ///
825  /// \returns true if LHS source location comes before RHS, false otherwise.
826  bool isBeforeInTranslationUnit(SourceLocation LHS, SourceLocation RHS) const;
827
828  // Iterators over FileInfos.
829  typedef llvm::DenseMap<const FileEntry*, SrcMgr::ContentCache*>
830      ::const_iterator fileinfo_iterator;
831  fileinfo_iterator fileinfo_begin() const { return FileInfos.begin(); }
832  fileinfo_iterator fileinfo_end() const { return FileInfos.end(); }
833  bool hasFileInfo(const FileEntry *File) const {
834    return FileInfos.find(File) != FileInfos.end();
835  }
836
837  /// PrintStats - Print statistics to stderr.
838  ///
839  void PrintStats() const;
840
841  unsigned sloc_entry_size() const { return SLocEntryTable.size(); }
842
843  // FIXME: Exposing this is a little gross; what we want is a good way
844  //  to iterate the entries that were not defined in an AST file (or
845  //  any other external source).
846  unsigned sloc_loaded_entry_size() const { return SLocEntryLoaded.size(); }
847
848  const SrcMgr::SLocEntry &getSLocEntry(unsigned ID) const {
849    assert(ID < SLocEntryTable.size() && "Invalid id");
850    if (ExternalSLocEntries &&
851        ID < SLocEntryLoaded.size() &&
852        !SLocEntryLoaded[ID])
853      ExternalSLocEntries->ReadSLocEntry(ID);
854    return SLocEntryTable[ID];
855  }
856
857  const SrcMgr::SLocEntry &getSLocEntry(FileID FID) const {
858    return getSLocEntry(FID.ID);
859  }
860
861  unsigned getNextOffset() const { return NextOffset; }
862
863  /// \brief Preallocate some number of source location entries, which
864  /// will be loaded as needed from the given external source.
865  void PreallocateSLocEntries(ExternalSLocEntrySource *Source,
866                              unsigned NumSLocEntries,
867                              unsigned NextOffset);
868
869  /// \brief Clear out any preallocated source location entries that
870  /// haven't already been loaded.
871  void ClearPreallocatedSLocEntries();
872
873private:
874  /// isOffsetInFileID - Return true if the specified FileID contains the
875  /// specified SourceLocation offset.  This is a very hot method.
876  inline bool isOffsetInFileID(FileID FID, unsigned SLocOffset) const {
877    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID);
878    // If the entry is after the offset, it can't contain it.
879    if (SLocOffset < Entry.getOffset()) return false;
880
881    // If this is the last entry than it does.  Otherwise, the entry after it
882    // has to not include it.
883    if (FID.ID+1 == SLocEntryTable.size()) return true;
884
885    return SLocOffset < getSLocEntry(FileID::get(FID.ID+1)).getOffset();
886  }
887
888  /// createFileID - Create a new fileID for the specified ContentCache and
889  ///  include position.  This works regardless of whether the ContentCache
890  ///  corresponds to a file or some other input source.
891  FileID createFileID(const SrcMgr::ContentCache* File,
892                      SourceLocation IncludePos,
893                      SrcMgr::CharacteristicKind DirCharacter,
894                      unsigned PreallocatedID = 0,
895                      unsigned Offset = 0);
896
897  const SrcMgr::ContentCache *
898    getOrCreateContentCache(const FileEntry *SourceFile);
899
900  /// createMemBufferContentCache - Create a new ContentCache for the specified
901  ///  memory buffer.
902  const SrcMgr::ContentCache*
903  createMemBufferContentCache(const llvm::MemoryBuffer *Buf);
904
905  FileID getFileIDSlow(unsigned SLocOffset) const;
906
907  SourceLocation getInstantiationLocSlowCase(SourceLocation Loc) const;
908  SourceLocation getSpellingLocSlowCase(SourceLocation Loc) const;
909
910  std::pair<FileID, unsigned>
911  getDecomposedInstantiationLocSlowCase(const SrcMgr::SLocEntry *E,
912                                        unsigned Offset) const;
913  std::pair<FileID, unsigned>
914  getDecomposedSpellingLocSlowCase(const SrcMgr::SLocEntry *E,
915                                   unsigned Offset) const;
916};
917
918
919}  // end namespace clang
920
921#endif
922