SourceManager.h revision b8c879a5363f36bdae8831112b563333e3c05acb
1bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant//===--- SourceManager.h - Track and cache source files ---------*- C++ -*-===//
2bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant//
3f5256e16dfc425c1d466f6308d4026d529ce9e0bHoward Hinnant//                     The LLVM Compiler Infrastructure
4bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant//
5b64f8b07c104c6cc986570ac8ee0ed16a9f23976Howard Hinnant// This file is distributed under the University of Illinois Open Source
6b64f8b07c104c6cc986570ac8ee0ed16a9f23976Howard Hinnant// License. See LICENSE.TXT for details.
7bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant//
8bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant//===----------------------------------------------------------------------===//
9bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant//
10bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant//  This file defines the SourceManager interface.
11bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant//
12bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant//===----------------------------------------------------------------------===//
13bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant
14bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant#ifndef LLVM_CLANG_SOURCEMANAGER_H
15bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant#define LLVM_CLANG_SOURCEMANAGER_H
16bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant
17061d0cc4db18d17bf01ed14c5db0be098205bd47Marshall Clow#include "clang/Basic/LLVM.h"
18fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant#include "clang/Basic/SourceLocation.h"
19fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant#include "llvm/Support/Allocator.h"
20fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant#include "llvm/Support/DataTypes.h"
21bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant#include "llvm/ADT/PointerIntPair.h"
22bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant#include "llvm/ADT/PointerUnion.h"
23bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant#include "llvm/ADT/IntrusiveRefCntPtr.h"
24bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant#include "llvm/ADT/DenseMap.h"
25bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant#include "llvm/Support/MemoryBuffer.h"
26bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant#include <map>
27bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant#include <vector>
28bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant#include <cassert>
29bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant
30bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnantnamespace clang {
31fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant
32bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnantclass DiagnosticsEngine;
33bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnantclass SourceManager;
34bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnantclass FileManager;
35bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnantclass FileEntry;
36bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnantclass LineTableInfo;
37bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnantclass LangOptions;
38bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnantclass ASTWriter;
39bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnantclass ASTReader;
40bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant
41fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant/// There are three different types of locations in a file: a spelling
42bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant/// location, an expansion location, and a presumed location.
43fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant///
44bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant/// Given an example of:
45fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant/// #define min(x, y) x < y ? x : y
46bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant///
47bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant/// and then later on a use of min:
48bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant/// #line 17
49bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant/// return min(a, b);
50bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant///
51bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant/// The expansion location is the line in the source code where the macro
52bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant/// was expanded (the return statement), the spelling location is the
53bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant/// location in the source where the macro was originally defined,
54bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant/// and the presumed location is where the line directive states that
55bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant/// the line is 17, or any other line.
56bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant
57bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant/// SrcMgr - Public enums and private classes that are part of the
58bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant/// SourceManager implementation.
59bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant///
60fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnantnamespace SrcMgr {
61bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant  /// CharacteristicKind - This is used to represent whether a file or directory
62bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant  /// holds normal user code, system code, or system code which is implicitly
63bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant  /// 'extern "C"' in C++ mode.  Entire directories can be tagged with this
64bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant  /// (this is maintained by DirectoryLookup and friends) as can specific
65bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant  /// FileInfos when a #pragma system_header is seen or various other cases.
66bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant  ///
67fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant  enum CharacteristicKind {
68bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant    C_User, C_System, C_ExternCSystem
69bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant  };
70bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant
71bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant  /// ContentCache - One instance of this struct is kept for every file
72bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant  /// loaded or used.  This object owns the MemoryBuffer object.
73bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant  class ContentCache {
74fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant    enum CCFlags {
75bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant      /// \brief Whether the buffer is invalid.
76bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant      InvalidFlag = 0x01,
77bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant      /// \brief Whether the buffer should not be freed on destruction.
78bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant      DoNotFreeFlag = 0x02
79fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant    };
80fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant
81fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant    /// Buffer - The actual buffer containing the characters from the input
82fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant    /// file.  This is owned by the ContentCache object.
83fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant    /// The bits indicate indicates whether the buffer is invalid.
84fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant    mutable llvm::PointerIntPair<const llvm::MemoryBuffer *, 2> Buffer;
85fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant
86fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant  public:
87fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant    /// Reference to the file entry representing this ContentCache.
88fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant    /// This reference does not own the FileEntry object.
89fcd8db7133c56a5a627f3922ce4a180c12287dd9Howard Hinnant    /// It is possible for this to be NULL if
90bc8d3f97eb5c958007f2713238472e0c1c8fe02Howard Hinnant    /// the ContentCache encapsulates an imaginary text buffer.
91    const FileEntry *OrigEntry;
92
93    /// \brief References the file which the contents were actually loaded from.
94    /// Can be different from 'Entry' if we overridden the contents of one file
95    /// with the contents of another file.
96    const FileEntry *ContentsEntry;
97
98    /// SourceLineCache - A bump pointer allocated array of offsets for each
99    /// source line.  This is lazily computed.  This is owned by the
100    /// SourceManager BumpPointerAllocator object.
101    unsigned *SourceLineCache;
102
103    /// NumLines - The number of lines in this ContentCache.  This is only valid
104    /// if SourceLineCache is non-null.
105    unsigned NumLines : 31;
106
107    /// \brief Indicates whether the buffer itself was provided to override
108    /// the actual file contents.
109    ///
110    /// When true, the original entry may be a virtual file that does not
111    /// exist.
112    unsigned BufferOverridden : 1;
113
114    ContentCache(const FileEntry *Ent = 0)
115      : Buffer(0, false), OrigEntry(Ent), ContentsEntry(Ent),
116        SourceLineCache(0), NumLines(0), BufferOverridden(false) {}
117
118    ContentCache(const FileEntry *Ent, const FileEntry *contentEnt)
119      : Buffer(0, false), OrigEntry(Ent), ContentsEntry(contentEnt),
120        SourceLineCache(0), NumLines(0), BufferOverridden(false) {}
121
122    ~ContentCache();
123
124    /// The copy ctor does not allow copies where source object has either
125    ///  a non-NULL Buffer or SourceLineCache.  Ownership of allocated memory
126    ///  is not transferred, so this is a logical error.
127    ContentCache(const ContentCache &RHS)
128      : Buffer(0, false), SourceLineCache(0), BufferOverridden(false)
129    {
130      OrigEntry = RHS.OrigEntry;
131      ContentsEntry = RHS.ContentsEntry;
132
133      assert (RHS.Buffer.getPointer() == 0 && RHS.SourceLineCache == 0 &&
134              "Passed ContentCache object cannot own a buffer.");
135
136      NumLines = RHS.NumLines;
137    }
138
139    /// getBuffer - Returns the memory buffer for the associated content.
140    ///
141    /// \param Diag Object through which diagnostics will be emitted if the
142    /// buffer cannot be retrieved.
143    ///
144    /// \param Loc If specified, is the location that invalid file diagnostics
145    ///     will be emitted at.
146    ///
147    /// \param Invalid If non-NULL, will be set \c true if an error occurred.
148    const llvm::MemoryBuffer *getBuffer(DiagnosticsEngine &Diag,
149                                        const SourceManager &SM,
150                                        SourceLocation Loc = SourceLocation(),
151                                        bool *Invalid = 0) const;
152
153    /// getSize - Returns the size of the content encapsulated by this
154    ///  ContentCache. This can be the size of the source file or the size of an
155    ///  arbitrary scratch buffer.  If the ContentCache encapsulates a source
156    ///  file this size is retrieved from the file's FileEntry.
157    unsigned getSize() const;
158
159    /// getSizeBytesMapped - Returns the number of bytes actually mapped for
160    /// this ContentCache. This can be 0 if the MemBuffer was not actually
161    /// expanded.
162    unsigned getSizeBytesMapped() const;
163
164    /// Returns the kind of memory used to back the memory buffer for
165    /// this content cache.  This is used for performance analysis.
166    llvm::MemoryBuffer::BufferKind getMemoryBufferKind() const;
167
168    void setBuffer(const llvm::MemoryBuffer *B) {
169      assert(!Buffer.getPointer() && "MemoryBuffer already set.");
170      Buffer.setPointer(B);
171      Buffer.setInt(false);
172    }
173
174    /// \brief Get the underlying buffer, returning NULL if the buffer is not
175    /// yet available.
176    const llvm::MemoryBuffer *getRawBuffer() const {
177      return Buffer.getPointer();
178    }
179
180    /// \brief Replace the existing buffer (which will be deleted)
181    /// with the given buffer.
182    void replaceBuffer(const llvm::MemoryBuffer *B, bool DoNotFree = false);
183
184    /// \brief Determine whether the buffer itself is invalid.
185    bool isBufferInvalid() const {
186      return Buffer.getInt() & InvalidFlag;
187    }
188
189    /// \brief Determine whether the buffer should be freed.
190    bool shouldFreeBuffer() const {
191      return (Buffer.getInt() & DoNotFreeFlag) == 0;
192    }
193
194  private:
195    // Disable assignments.
196    ContentCache &operator=(const ContentCache& RHS);
197  };
198
199  /// FileInfo - Information about a FileID, basically just the logical file
200  /// that it represents and include stack information.
201  ///
202  /// Each FileInfo has include stack information, indicating where it came
203  /// from. This information encodes the #include chain that a token was
204  /// expanded from. The main include file has an invalid IncludeLoc.
205  ///
206  /// FileInfos contain a "ContentCache *", with the contents of the file.
207  ///
208  class FileInfo {
209    /// IncludeLoc - The location of the #include that brought in this file.
210    /// This is an invalid SLOC for the main file (top of the #include chain).
211    unsigned IncludeLoc;  // Really a SourceLocation
212
213    /// \brief Number of FileIDs (files and macros) that were created during
214    /// preprocessing of this #include, including this SLocEntry.
215    /// Zero means the preprocessor didn't provide such info for this SLocEntry.
216    unsigned NumCreatedFIDs;
217
218    /// Data - This contains the ContentCache* and the bits indicating the
219    /// characteristic of the file and whether it has #line info, all bitmangled
220    /// together.
221    uintptr_t Data;
222
223    friend class clang::SourceManager;
224    friend class clang::ASTWriter;
225    friend class clang::ASTReader;
226  public:
227    /// get - Return a FileInfo object.
228    static FileInfo get(SourceLocation IL, const ContentCache *Con,
229                        CharacteristicKind FileCharacter) {
230      FileInfo X;
231      X.IncludeLoc = IL.getRawEncoding();
232      X.NumCreatedFIDs = 0;
233      X.Data = (uintptr_t)Con;
234      assert((X.Data & 7) == 0 &&"ContentCache pointer insufficiently aligned");
235      assert((unsigned)FileCharacter < 4 && "invalid file character");
236      X.Data |= (unsigned)FileCharacter;
237      return X;
238    }
239
240    SourceLocation getIncludeLoc() const {
241      return SourceLocation::getFromRawEncoding(IncludeLoc);
242    }
243    const ContentCache* getContentCache() const {
244      return reinterpret_cast<const ContentCache*>(Data & ~7UL);
245    }
246
247    /// getCharacteristic - Return whether this is a system header or not.
248    CharacteristicKind getFileCharacteristic() const {
249      return (CharacteristicKind)(Data & 3);
250    }
251
252    /// hasLineDirectives - Return true if this FileID has #line directives in
253    /// it.
254    bool hasLineDirectives() const { return (Data & 4) != 0; }
255
256    /// setHasLineDirectives - Set the flag that indicates that this FileID has
257    /// line table entries associated with it.
258    void setHasLineDirectives() {
259      Data |= 4;
260    }
261  };
262
263  /// ExpansionInfo - Each ExpansionInfo encodes the expansion location - where
264  /// the token was ultimately expanded, and the SpellingLoc - where the actual
265  /// character data for the token came from.
266  class ExpansionInfo {
267    // Really these are all SourceLocations.
268
269    /// SpellingLoc - Where the spelling for the token can be found.
270    unsigned SpellingLoc;
271
272    /// ExpansionLocStart/ExpansionLocEnd - In a macro expansion, these
273    /// indicate the start and end of the expansion. In object-like macros,
274    /// these will be the same. In a function-like macro expansion, the start
275    /// will be the identifier and the end will be the ')'. Finally, in
276    /// macro-argument instantitions, the end will be 'SourceLocation()', an
277    /// invalid location.
278    unsigned ExpansionLocStart, ExpansionLocEnd;
279
280  public:
281    SourceLocation getSpellingLoc() const {
282      return SourceLocation::getFromRawEncoding(SpellingLoc);
283    }
284    SourceLocation getExpansionLocStart() const {
285      return SourceLocation::getFromRawEncoding(ExpansionLocStart);
286    }
287    SourceLocation getExpansionLocEnd() const {
288      SourceLocation EndLoc =
289        SourceLocation::getFromRawEncoding(ExpansionLocEnd);
290      return EndLoc.isInvalid() ? getExpansionLocStart() : EndLoc;
291    }
292
293    std::pair<SourceLocation,SourceLocation> getExpansionLocRange() const {
294      return std::make_pair(getExpansionLocStart(), getExpansionLocEnd());
295    }
296
297    bool isMacroArgExpansion() const {
298      // Note that this needs to return false for default constructed objects.
299      return getExpansionLocStart().isValid() &&
300        SourceLocation::getFromRawEncoding(ExpansionLocEnd).isInvalid();
301    }
302
303    bool isFunctionMacroExpansion() const {
304      return getExpansionLocStart().isValid() &&
305          getExpansionLocStart() != getExpansionLocEnd();
306    }
307
308    /// create - Return a ExpansionInfo for an expansion. Start and End specify
309    /// the expansion range (where the macro is expanded), and SpellingLoc
310    /// specifies the spelling location (where the characters from the token
311    /// come from). All three can refer to normal File SLocs or expansion
312    /// locations.
313    static ExpansionInfo create(SourceLocation SpellingLoc,
314                                SourceLocation Start, SourceLocation End) {
315      ExpansionInfo X;
316      X.SpellingLoc = SpellingLoc.getRawEncoding();
317      X.ExpansionLocStart = Start.getRawEncoding();
318      X.ExpansionLocEnd = End.getRawEncoding();
319      return X;
320    }
321
322    /// createForMacroArg - Return a special ExpansionInfo for the expansion of
323    /// a macro argument into a function-like macro's body. ExpansionLoc
324    /// specifies the expansion location (where the macro is expanded). This
325    /// doesn't need to be a range because a macro is always expanded at
326    /// a macro parameter reference, and macro parameters are always exactly
327    /// one token. SpellingLoc specifies the spelling location (where the
328    /// characters from the token come from). ExpansionLoc and SpellingLoc can
329    /// both refer to normal File SLocs or expansion locations.
330    ///
331    /// Given the code:
332    /// \code
333    ///   #define F(x) f(x)
334    ///   F(42);
335    /// \endcode
336    ///
337    /// When expanding '\c F(42)', the '\c x' would call this with an
338    /// SpellingLoc pointing at '\c 42' anad an ExpansionLoc pointing at its
339    /// location in the definition of '\c F'.
340    static ExpansionInfo createForMacroArg(SourceLocation SpellingLoc,
341                                           SourceLocation ExpansionLoc) {
342      // We store an intentionally invalid source location for the end of the
343      // expansion range to mark that this is a macro argument ion rather than
344      // a normal one.
345      return create(SpellingLoc, ExpansionLoc, SourceLocation());
346    }
347  };
348
349  /// SLocEntry - This is a discriminated union of FileInfo and
350  /// ExpansionInfo.  SourceManager keeps an array of these objects, and
351  /// they are uniquely identified by the FileID datatype.
352  class SLocEntry {
353    unsigned Offset;   // low bit is set for expansion info.
354    union {
355      FileInfo File;
356      ExpansionInfo Expansion;
357    };
358  public:
359    unsigned getOffset() const { return Offset >> 1; }
360
361    bool isExpansion() const { return Offset & 1; }
362    bool isFile() const { return !isExpansion(); }
363
364    const FileInfo &getFile() const {
365      assert(isFile() && "Not a file SLocEntry!");
366      return File;
367    }
368
369    const ExpansionInfo &getExpansion() const {
370      assert(isExpansion() && "Not a macro expansion SLocEntry!");
371      return Expansion;
372    }
373
374    static SLocEntry get(unsigned Offset, const FileInfo &FI) {
375      SLocEntry E;
376      E.Offset = Offset << 1;
377      E.File = FI;
378      return E;
379    }
380
381    static SLocEntry get(unsigned Offset, const ExpansionInfo &Expansion) {
382      SLocEntry E;
383      E.Offset = (Offset << 1) | 1;
384      E.Expansion = Expansion;
385      return E;
386    }
387  };
388}  // end SrcMgr namespace.
389
390/// \brief External source of source location entries.
391class ExternalSLocEntrySource {
392public:
393  virtual ~ExternalSLocEntrySource();
394
395  /// \brief Read the source location entry with index ID, which will always be
396  /// less than -1.
397  ///
398  /// \returns true if an error occurred that prevented the source-location
399  /// entry from being loaded.
400  virtual bool ReadSLocEntry(int ID) = 0;
401};
402
403
404/// IsBeforeInTranslationUnitCache - This class holds the cache used by
405/// isBeforeInTranslationUnit.  The cache structure is complex enough to be
406/// worth breaking out of SourceManager.
407class IsBeforeInTranslationUnitCache {
408  /// L/R QueryFID - These are the FID's of the cached query.  If these match up
409  /// with a subsequent query, the result can be reused.
410  FileID LQueryFID, RQueryFID;
411
412  /// \brief True if LQueryFID was created before RQueryFID. This is used
413  /// to compare macro expansion locations.
414  bool IsLQFIDBeforeRQFID;
415
416  /// CommonFID - This is the file found in common between the two #include
417  /// traces.  It is the nearest common ancestor of the #include tree.
418  FileID CommonFID;
419
420  /// L/R CommonOffset - This is the offset of the previous query in CommonFID.
421  /// Usually, this represents the location of the #include for QueryFID, but if
422  /// LQueryFID is a parent of RQueryFID (or vise versa) then these can be a
423  /// random token in the parent.
424  unsigned LCommonOffset, RCommonOffset;
425public:
426
427  /// isCacheValid - Return true if the currently cached values match up with
428  /// the specified LHS/RHS query.  If not, we can't use the cache.
429  bool isCacheValid(FileID LHS, FileID RHS) const {
430    return LQueryFID == LHS && RQueryFID == RHS;
431  }
432
433  /// getCachedResult - If the cache is valid, compute the result given the
434  /// specified offsets in the LHS/RHS FID's.
435  bool getCachedResult(unsigned LOffset, unsigned ROffset) const {
436    // If one of the query files is the common file, use the offset.  Otherwise,
437    // use the #include loc in the common file.
438    if (LQueryFID != CommonFID) LOffset = LCommonOffset;
439    if (RQueryFID != CommonFID) ROffset = RCommonOffset;
440
441    // It is common for multiple macro expansions to be "included" from the same
442    // location (expansion location), in which case use the order of the FileIDs
443    // to determine which came first. This will also take care the case where
444    // one of the locations points at the inclusion/expansion point of the other
445    // in which case its FileID will come before the other.
446    if (LOffset == ROffset)
447      return IsLQFIDBeforeRQFID;
448
449    return LOffset < ROffset;
450  }
451
452  // Set up a new query.
453  void setQueryFIDs(FileID LHS, FileID RHS, bool isLFIDBeforeRFID) {
454    assert(LHS != RHS);
455    LQueryFID = LHS;
456    RQueryFID = RHS;
457    IsLQFIDBeforeRQFID = isLFIDBeforeRFID;
458  }
459
460  void clear() {
461    LQueryFID = RQueryFID = FileID();
462    IsLQFIDBeforeRQFID = false;
463  }
464
465  void setCommonLoc(FileID commonFID, unsigned lCommonOffset,
466                    unsigned rCommonOffset) {
467    CommonFID = commonFID;
468    LCommonOffset = lCommonOffset;
469    RCommonOffset = rCommonOffset;
470  }
471
472};
473
474/// \brief This class handles loading and caching of source files into memory.
475///
476/// This object owns the MemoryBuffer objects for all of the loaded
477/// files and assigns unique FileID's for each unique #include chain.
478///
479/// The SourceManager can be queried for information about SourceLocation
480/// objects, turning them into either spelling or expansion locations. Spelling
481/// locations represent where the bytes corresponding to a token came from and
482/// expansion locations represent where the location is in the user's view. In
483/// the case of a macro expansion, for example, the spelling location indicates
484/// where the expanded token came from and the expansion location specifies
485/// where it was expanded.
486class SourceManager : public llvm::RefCountedBase<SourceManager> {
487  /// \brief DiagnosticsEngine object.
488  DiagnosticsEngine &Diag;
489
490  FileManager &FileMgr;
491
492  mutable llvm::BumpPtrAllocator ContentCacheAlloc;
493
494  /// FileInfos - Memoized information about all of the files tracked by this
495  /// SourceManager.  This set allows us to merge ContentCache entries based
496  /// on their FileEntry*.  All ContentCache objects will thus have unique,
497  /// non-null, FileEntry pointers.
498  llvm::DenseMap<const FileEntry*, SrcMgr::ContentCache*> FileInfos;
499
500  /// \brief True if the ContentCache for files that are overriden by other
501  /// files, should report the original file name. Defaults to true.
502  bool OverridenFilesKeepOriginalName;
503
504  /// \brief Files that have been overriden with the contents from another file.
505  llvm::DenseMap<const FileEntry *, const FileEntry *> OverriddenFiles;
506
507  /// MemBufferInfos - Information about various memory buffers that we have
508  /// read in.  All FileEntry* within the stored ContentCache objects are NULL,
509  /// as they do not refer to a file.
510  std::vector<SrcMgr::ContentCache*> MemBufferInfos;
511
512  /// \brief The table of SLocEntries that are local to this module.
513  ///
514  /// Positive FileIDs are indexes into this table. Entry 0 indicates an invalid
515  /// expansion.
516  std::vector<SrcMgr::SLocEntry> LocalSLocEntryTable;
517
518  /// \brief The table of SLocEntries that are loaded from other modules.
519  ///
520  /// Negative FileIDs are indexes into this table. To get from ID to an index,
521  /// use (-ID - 2).
522  std::vector<SrcMgr::SLocEntry> LoadedSLocEntryTable;
523
524  /// \brief The starting offset of the next local SLocEntry.
525  ///
526  /// This is LocalSLocEntryTable.back().Offset + the size of that entry.
527  unsigned NextLocalOffset;
528
529  /// \brief The starting offset of the latest batch of loaded SLocEntries.
530  ///
531  /// This is LoadedSLocEntryTable.back().Offset, except that that entry might
532  /// not have been loaded, so that value would be unknown.
533  unsigned CurrentLoadedOffset;
534
535  /// \brief The highest possible offset is 2^31-1, so CurrentLoadedOffset
536  /// starts at 2^31.
537  static const unsigned MaxLoadedOffset = 1U << 31U;
538
539  /// \brief A bitmap that indicates whether the entries of LoadedSLocEntryTable
540  /// have already been loaded from the external source.
541  ///
542  /// Same indexing as LoadedSLocEntryTable.
543  std::vector<bool> SLocEntryLoaded;
544
545  /// \brief An external source for source location entries.
546  ExternalSLocEntrySource *ExternalSLocEntries;
547
548  /// LastFileIDLookup - This is a one-entry cache to speed up getFileID.
549  /// LastFileIDLookup records the last FileID looked up or created, because it
550  /// is very common to look up many tokens from the same file.
551  mutable FileID LastFileIDLookup;
552
553  /// LineTable - This holds information for #line directives.  It is referenced
554  /// by indices from SLocEntryTable.
555  LineTableInfo *LineTable;
556
557  /// LastLineNo - These ivars serve as a cache used in the getLineNumber
558  /// method which is used to speedup getLineNumber calls to nearby locations.
559  mutable FileID LastLineNoFileIDQuery;
560  mutable SrcMgr::ContentCache *LastLineNoContentCache;
561  mutable unsigned LastLineNoFilePos;
562  mutable unsigned LastLineNoResult;
563
564  /// MainFileID - The file ID for the main source file of the translation unit.
565  FileID MainFileID;
566
567  /// \brief The file ID for the precompiled preamble there is one.
568  FileID PreambleFileID;
569
570  // Statistics for -print-stats.
571  mutable unsigned NumLinearScans, NumBinaryProbes;
572
573  // Cache results for the isBeforeInTranslationUnit method.
574  mutable IsBeforeInTranslationUnitCache IsBeforeInTUCache;
575
576  // Cache for the "fake" buffer used for error-recovery purposes.
577  mutable llvm::MemoryBuffer *FakeBufferForRecovery;
578
579  /// \brief Lazily computed map of macro argument chunks to their expanded
580  /// source location.
581  typedef std::map<unsigned, SourceLocation> MacroArgsMap;
582
583  mutable llvm::DenseMap<FileID, MacroArgsMap *> MacroArgsCacheMap;
584
585  // SourceManager doesn't support copy construction.
586  explicit SourceManager(const SourceManager&);
587  void operator=(const SourceManager&);
588public:
589  SourceManager(DiagnosticsEngine &Diag, FileManager &FileMgr);
590  ~SourceManager();
591
592  void clearIDTables();
593
594  DiagnosticsEngine &getDiagnostics() const { return Diag; }
595
596  FileManager &getFileManager() const { return FileMgr; }
597
598  /// \brief Set true if the SourceManager should report the original file name
599  /// for contents of files that were overriden by other files.Defaults to true.
600  void setOverridenFilesKeepOriginalName(bool value) {
601    OverridenFilesKeepOriginalName = value;
602  }
603
604  /// createMainFileIDForMembuffer - Create the FileID for a memory buffer
605  ///  that will represent the FileID for the main source.  One example
606  ///  of when this would be used is when the main source is read from STDIN.
607  FileID createMainFileIDForMemBuffer(const llvm::MemoryBuffer *Buffer) {
608    assert(MainFileID.isInvalid() && "MainFileID already set!");
609    MainFileID = createFileIDForMemBuffer(Buffer);
610    return MainFileID;
611  }
612
613  //===--------------------------------------------------------------------===//
614  // MainFileID creation and querying methods.
615  //===--------------------------------------------------------------------===//
616
617  /// getMainFileID - Returns the FileID of the main source file.
618  FileID getMainFileID() const { return MainFileID; }
619
620  /// createMainFileID - Create the FileID for the main source file.
621  FileID createMainFileID(const FileEntry *SourceFile) {
622    assert(MainFileID.isInvalid() && "MainFileID already set!");
623    MainFileID = createFileID(SourceFile, SourceLocation(), SrcMgr::C_User);
624    return MainFileID;
625  }
626
627  /// \brief Set the file ID for the main source file.
628  void setMainFileID(FileID FID) {
629    assert(MainFileID.isInvalid() && "MainFileID already set!");
630    MainFileID = FID;
631  }
632
633  /// \brief Set the file ID for the precompiled preamble.
634  void setPreambleFileID(FileID Preamble) {
635    assert(PreambleFileID.isInvalid() && "PreambleFileID already set!");
636    PreambleFileID = Preamble;
637  }
638
639  /// \brief Get the file ID for the precompiled preamble if there is one.
640  FileID getPreambleFileID() const { return PreambleFileID; }
641
642  //===--------------------------------------------------------------------===//
643  // Methods to create new FileID's and macro expansions.
644  //===--------------------------------------------------------------------===//
645
646  /// createFileID - Create a new FileID that represents the specified file
647  /// being #included from the specified IncludePosition.  This translates NULL
648  /// into standard input.
649  FileID createFileID(const FileEntry *SourceFile, SourceLocation IncludePos,
650                      SrcMgr::CharacteristicKind FileCharacter,
651                      int LoadedID = 0, unsigned LoadedOffset = 0) {
652    const SrcMgr::ContentCache *IR = getOrCreateContentCache(SourceFile);
653    assert(IR && "getOrCreateContentCache() cannot return NULL");
654    return createFileID(IR, IncludePos, FileCharacter, LoadedID, LoadedOffset);
655  }
656
657  /// createFileIDForMemBuffer - Create a new FileID that represents the
658  /// specified memory buffer.  This does no caching of the buffer and takes
659  /// ownership of the MemoryBuffer, so only pass a MemoryBuffer to this once.
660  FileID createFileIDForMemBuffer(const llvm::MemoryBuffer *Buffer,
661                                  int LoadedID = 0, unsigned LoadedOffset = 0,
662                                 SourceLocation IncludeLoc = SourceLocation()) {
663    return createFileID(createMemBufferContentCache(Buffer), IncludeLoc,
664                        SrcMgr::C_User, LoadedID, LoadedOffset);
665  }
666
667  /// createMacroArgExpansionLoc - Return a new SourceLocation that encodes the
668  /// fact that a token from SpellingLoc should actually be referenced from
669  /// ExpansionLoc, and that it represents the expansion of a macro argument
670  /// into the function-like macro body.
671  SourceLocation createMacroArgExpansionLoc(SourceLocation Loc,
672                                            SourceLocation ExpansionLoc,
673                                            unsigned TokLength);
674
675  /// createExpansionLoc - Return a new SourceLocation that encodes the fact
676  /// that a token from SpellingLoc should actually be referenced from
677  /// ExpansionLoc.
678  SourceLocation createExpansionLoc(SourceLocation Loc,
679                                    SourceLocation ExpansionLocStart,
680                                    SourceLocation ExpansionLocEnd,
681                                    unsigned TokLength,
682                                    int LoadedID = 0,
683                                    unsigned LoadedOffset = 0);
684
685  /// \brief Retrieve the memory buffer associated with the given file.
686  ///
687  /// \param Invalid If non-NULL, will be set \c true if an error
688  /// occurs while retrieving the memory buffer.
689  const llvm::MemoryBuffer *getMemoryBufferForFile(const FileEntry *File,
690                                                   bool *Invalid = 0);
691
692  /// \brief Override the contents of the given source file by providing an
693  /// already-allocated buffer.
694  ///
695  /// \param SourceFile the source file whose contents will be overriden.
696  ///
697  /// \param Buffer the memory buffer whose contents will be used as the
698  /// data in the given source file.
699  ///
700  /// \param DoNotFree If true, then the buffer will not be freed when the
701  /// source manager is destroyed.
702  void overrideFileContents(const FileEntry *SourceFile,
703                            const llvm::MemoryBuffer *Buffer,
704                            bool DoNotFree = false);
705
706  /// \brief Override the the given source file with another one.
707  ///
708  /// \param SourceFile the source file which will be overriden.
709  ///
710  /// \param NewFile the file whose contents will be used as the
711  /// data instead of the contents of the given source file.
712  void overrideFileContents(const FileEntry *SourceFile,
713                            const FileEntry *NewFile);
714
715  //===--------------------------------------------------------------------===//
716  // FileID manipulation methods.
717  //===--------------------------------------------------------------------===//
718
719  /// getBuffer - Return the buffer for the specified FileID. If there is an
720  /// error opening this buffer the first time, this manufactures a temporary
721  /// buffer and returns a non-empty error string.
722  const llvm::MemoryBuffer *getBuffer(FileID FID, SourceLocation Loc,
723                                      bool *Invalid = 0) const {
724    bool MyInvalid = false;
725    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid);
726    if (MyInvalid || !Entry.isFile()) {
727      if (Invalid)
728        *Invalid = true;
729
730      return getFakeBufferForRecovery();
731    }
732
733    return Entry.getFile().getContentCache()->getBuffer(Diag, *this, Loc,
734                                                        Invalid);
735  }
736
737  const llvm::MemoryBuffer *getBuffer(FileID FID, bool *Invalid = 0) const {
738    bool MyInvalid = false;
739    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid);
740    if (MyInvalid || !Entry.isFile()) {
741      if (Invalid)
742        *Invalid = true;
743
744      return getFakeBufferForRecovery();
745    }
746
747    return Entry.getFile().getContentCache()->getBuffer(Diag, *this,
748                                                        SourceLocation(),
749                                                        Invalid);
750  }
751
752  /// getFileEntryForID - Returns the FileEntry record for the provided FileID.
753  const FileEntry *getFileEntryForID(FileID FID) const {
754    bool MyInvalid = false;
755    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid);
756    if (MyInvalid || !Entry.isFile())
757      return 0;
758
759    const SrcMgr::ContentCache *Content = Entry.getFile().getContentCache();
760    if (!Content)
761      return 0;
762    return Content->OrigEntry;
763  }
764
765  /// Returns the FileEntry record for the provided SLocEntry.
766  const FileEntry *getFileEntryForSLocEntry(const SrcMgr::SLocEntry &sloc) const
767  {
768    const SrcMgr::ContentCache *Content = sloc.getFile().getContentCache();
769    if (!Content)
770      return 0;
771    return Content->OrigEntry;
772  }
773
774  /// getBufferData - Return a StringRef to the source buffer data for the
775  /// specified FileID.
776  ///
777  /// \param FID The file ID whose contents will be returned.
778  /// \param Invalid If non-NULL, will be set true if an error occurred.
779  StringRef getBufferData(FileID FID, bool *Invalid = 0) const;
780
781  /// \brief Get the number of FileIDs (files and macros) that were created
782  /// during preprocessing of \p FID, including it.
783  unsigned getNumCreatedFIDsForFileID(FileID FID) const {
784    bool Invalid = false;
785    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
786    if (Invalid || !Entry.isFile())
787      return 0;
788
789    return Entry.getFile().NumCreatedFIDs;
790  }
791
792  /// \brief Set the number of FileIDs (files and macros) that were created
793  /// during preprocessing of \p FID, including it.
794  void setNumCreatedFIDsForFileID(FileID FID, unsigned NumFIDs) const {
795    bool Invalid = false;
796    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
797    if (Invalid || !Entry.isFile())
798      return;
799
800    assert(Entry.getFile().NumCreatedFIDs == 0 && "Already set!");
801    const_cast<SrcMgr::FileInfo &>(Entry.getFile()).NumCreatedFIDs = NumFIDs;
802  }
803
804  //===--------------------------------------------------------------------===//
805  // SourceLocation manipulation methods.
806  //===--------------------------------------------------------------------===//
807
808  /// getFileID - Return the FileID for a SourceLocation.  This is a very
809  /// hot method that is used for all SourceManager queries that start with a
810  /// SourceLocation object.  It is responsible for finding the entry in
811  /// SLocEntryTable which contains the specified location.
812  ///
813  FileID getFileID(SourceLocation SpellingLoc) const {
814    unsigned SLocOffset = SpellingLoc.getOffset();
815
816    // If our one-entry cache covers this offset, just return it.
817    if (isOffsetInFileID(LastFileIDLookup, SLocOffset))
818      return LastFileIDLookup;
819
820    return getFileIDSlow(SLocOffset);
821  }
822
823  /// getLocForStartOfFile - Return the source location corresponding to the
824  /// first byte of the specified file.
825  SourceLocation getLocForStartOfFile(FileID FID) const {
826    bool Invalid = false;
827    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
828    if (Invalid || !Entry.isFile())
829      return SourceLocation();
830
831    unsigned FileOffset = Entry.getOffset();
832    return SourceLocation::getFileLoc(FileOffset);
833  }
834
835  /// \brief Return the source location corresponding to the last byte of the
836  /// specified file.
837  SourceLocation getLocForEndOfFile(FileID FID) const {
838    bool Invalid = false;
839    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
840    if (Invalid || !Entry.isFile())
841      return SourceLocation();
842
843    unsigned FileOffset = Entry.getOffset();
844    return SourceLocation::getFileLoc(FileOffset + getFileIDSize(FID) - 1);
845  }
846
847  /// \brief Returns the include location if \p FID is a #include'd file
848  /// otherwise it returns an invalid location.
849  SourceLocation getIncludeLoc(FileID FID) const {
850    bool Invalid = false;
851    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
852    if (Invalid || !Entry.isFile())
853      return SourceLocation();
854
855    return Entry.getFile().getIncludeLoc();
856  }
857
858  /// getExpansionLoc - Given a SourceLocation object, return the expansion
859  /// location referenced by the ID.
860  SourceLocation getExpansionLoc(SourceLocation Loc) const {
861    // Handle the non-mapped case inline, defer to out of line code to handle
862    // expansions.
863    if (Loc.isFileID()) return Loc;
864    return getExpansionLocSlowCase(Loc);
865  }
866
867  /// \brief Given \p Loc, if it is a macro location return the expansion
868  /// location or the spelling location, depending on if it comes from a
869  /// macro argument or not.
870  SourceLocation getFileLoc(SourceLocation Loc) const {
871    if (Loc.isFileID()) return Loc;
872    return getFileLocSlowCase(Loc);
873  }
874
875  /// getImmediateExpansionRange - Loc is required to be an expansion location.
876  /// Return the start/end of the expansion information.
877  std::pair<SourceLocation,SourceLocation>
878  getImmediateExpansionRange(SourceLocation Loc) const;
879
880  /// getExpansionRange - Given a SourceLocation object, return the range of
881  /// tokens covered by the expansion the ultimate file.
882  std::pair<SourceLocation,SourceLocation>
883  getExpansionRange(SourceLocation Loc) const;
884
885
886  /// getSpellingLoc - Given a SourceLocation object, return the spelling
887  /// location referenced by the ID.  This is the place where the characters
888  /// that make up the lexed token can be found.
889  SourceLocation getSpellingLoc(SourceLocation Loc) const {
890    // Handle the non-mapped case inline, defer to out of line code to handle
891    // expansions.
892    if (Loc.isFileID()) return Loc;
893    return getSpellingLocSlowCase(Loc);
894  }
895
896  /// getImmediateSpellingLoc - Given a SourceLocation object, return the
897  /// spelling location referenced by the ID.  This is the first level down
898  /// towards the place where the characters that make up the lexed token can be
899  /// found.  This should not generally be used by clients.
900  SourceLocation getImmediateSpellingLoc(SourceLocation Loc) const;
901
902  /// getDecomposedLoc - Decompose the specified location into a raw FileID +
903  /// Offset pair.  The first element is the FileID, the second is the
904  /// offset from the start of the buffer of the location.
905  std::pair<FileID, unsigned> getDecomposedLoc(SourceLocation Loc) const {
906    FileID FID = getFileID(Loc);
907    bool Invalid = false;
908    const SrcMgr::SLocEntry &E = getSLocEntry(FID, &Invalid);
909    if (Invalid)
910      return std::make_pair(FileID(), 0);
911    return std::make_pair(FID, Loc.getOffset()-E.getOffset());
912  }
913
914  /// getDecomposedExpansionLoc - Decompose the specified location into a raw
915  /// FileID + Offset pair. If the location is an expansion record, walk
916  /// through it until we find the final location expanded.
917  std::pair<FileID, unsigned>
918  getDecomposedExpansionLoc(SourceLocation Loc) const {
919    FileID FID = getFileID(Loc);
920    bool Invalid = false;
921    const SrcMgr::SLocEntry *E = &getSLocEntry(FID, &Invalid);
922    if (Invalid)
923      return std::make_pair(FileID(), 0);
924
925    unsigned Offset = Loc.getOffset()-E->getOffset();
926    if (Loc.isFileID())
927      return std::make_pair(FID, Offset);
928
929    return getDecomposedExpansionLocSlowCase(E);
930  }
931
932  /// getDecomposedSpellingLoc - Decompose the specified location into a raw
933  /// FileID + Offset pair.  If the location is an expansion record, walk
934  /// through it until we find its spelling record.
935  std::pair<FileID, unsigned>
936  getDecomposedSpellingLoc(SourceLocation Loc) const {
937    FileID FID = getFileID(Loc);
938    bool Invalid = false;
939    const SrcMgr::SLocEntry *E = &getSLocEntry(FID, &Invalid);
940    if (Invalid)
941      return std::make_pair(FileID(), 0);
942
943    unsigned Offset = Loc.getOffset()-E->getOffset();
944    if (Loc.isFileID())
945      return std::make_pair(FID, Offset);
946    return getDecomposedSpellingLocSlowCase(E, Offset);
947  }
948
949  /// getFileOffset - This method returns the offset from the start
950  /// of the file that the specified SourceLocation represents. This is not very
951  /// meaningful for a macro ID.
952  unsigned getFileOffset(SourceLocation SpellingLoc) const {
953    return getDecomposedLoc(SpellingLoc).second;
954  }
955
956  /// isMacroArgExpansion - This method tests whether the given source location
957  /// represents a macro argument's expansion into the function-like macro
958  /// definition. Such source locations only appear inside of the expansion
959  /// locations representing where a particular function-like macro was
960  /// expanded.
961  bool isMacroArgExpansion(SourceLocation Loc) const;
962
963  /// \brief Returns true if \p Loc is inside the [\p Start, +\p Length)
964  /// chunk of the source location address space.
965  /// If it's true and \p RelativeOffset is non-null, it will be set to the
966  /// relative offset of \p Loc inside the chunk.
967  bool isInSLocAddrSpace(SourceLocation Loc,
968                         SourceLocation Start, unsigned Length,
969                         unsigned *RelativeOffset = 0) const {
970    assert(((Start.getOffset() < NextLocalOffset &&
971               Start.getOffset()+Length <= NextLocalOffset) ||
972            (Start.getOffset() >= CurrentLoadedOffset &&
973                Start.getOffset()+Length < MaxLoadedOffset)) &&
974           "Chunk is not valid SLoc address space");
975    unsigned LocOffs = Loc.getOffset();
976    unsigned BeginOffs = Start.getOffset();
977    unsigned EndOffs = BeginOffs + Length;
978    if (LocOffs >= BeginOffs && LocOffs < EndOffs) {
979      if (RelativeOffset)
980        *RelativeOffset = LocOffs - BeginOffs;
981      return true;
982    }
983
984    return false;
985  }
986
987  /// \brief Return true if both \p LHS and \p RHS are in the local source
988  /// location address space or the loaded one. If it's true and \p
989  /// RelativeOffset is non-null, it will be set to the offset of \p RHS
990  /// relative to \p LHS.
991  bool isInSameSLocAddrSpace(SourceLocation LHS, SourceLocation RHS,
992                             int *RelativeOffset) const {
993    unsigned LHSOffs = LHS.getOffset(), RHSOffs = RHS.getOffset();
994    bool LHSLoaded = LHSOffs >= CurrentLoadedOffset;
995    bool RHSLoaded = RHSOffs >= CurrentLoadedOffset;
996
997    if (LHSLoaded == RHSLoaded) {
998      if (RelativeOffset)
999        *RelativeOffset = RHSOffs - LHSOffs;
1000      return true;
1001    }
1002
1003    return false;
1004  }
1005
1006  //===--------------------------------------------------------------------===//
1007  // Queries about the code at a SourceLocation.
1008  //===--------------------------------------------------------------------===//
1009
1010  /// getCharacterData - Return a pointer to the start of the specified location
1011  /// in the appropriate spelling MemoryBuffer.
1012  ///
1013  /// \param Invalid If non-NULL, will be set \c true if an error occurs.
1014  const char *getCharacterData(SourceLocation SL, bool *Invalid = 0) const;
1015
1016  /// getColumnNumber - Return the column # for the specified file position.
1017  /// This is significantly cheaper to compute than the line number.  This
1018  /// returns zero if the column number isn't known.  This may only be called
1019  /// on a file sloc, so you must choose a spelling or expansion location
1020  /// before calling this method.
1021  unsigned getColumnNumber(FileID FID, unsigned FilePos,
1022                           bool *Invalid = 0) const;
1023  unsigned getSpellingColumnNumber(SourceLocation Loc, bool *Invalid = 0) const;
1024  unsigned getExpansionColumnNumber(SourceLocation Loc,
1025                                    bool *Invalid = 0) const;
1026  unsigned getPresumedColumnNumber(SourceLocation Loc, bool *Invalid = 0) const;
1027
1028
1029  /// getLineNumber - Given a SourceLocation, return the spelling line number
1030  /// for the position indicated.  This requires building and caching a table of
1031  /// line offsets for the MemoryBuffer, so this is not cheap: use only when
1032  /// about to emit a diagnostic.
1033  unsigned getLineNumber(FileID FID, unsigned FilePos, bool *Invalid = 0) const;
1034  unsigned getSpellingLineNumber(SourceLocation Loc, bool *Invalid = 0) const;
1035  unsigned getExpansionLineNumber(SourceLocation Loc, bool *Invalid = 0) const;
1036  unsigned getPresumedLineNumber(SourceLocation Loc, bool *Invalid = 0) const;
1037
1038  /// Return the filename or buffer identifier of the buffer the location is in.
1039  /// Note that this name does not respect #line directives.  Use getPresumedLoc
1040  /// for normal clients.
1041  const char *getBufferName(SourceLocation Loc, bool *Invalid = 0) const;
1042
1043  /// getFileCharacteristic - return the file characteristic of the specified
1044  /// source location, indicating whether this is a normal file, a system
1045  /// header, or an "implicit extern C" system header.
1046  ///
1047  /// This state can be modified with flags on GNU linemarker directives like:
1048  ///   # 4 "foo.h" 3
1049  /// which changes all source locations in the current file after that to be
1050  /// considered to be from a system header.
1051  SrcMgr::CharacteristicKind getFileCharacteristic(SourceLocation Loc) const;
1052
1053  /// getPresumedLoc - This method returns the "presumed" location of a
1054  /// SourceLocation specifies.  A "presumed location" can be modified by #line
1055  /// or GNU line marker directives.  This provides a view on the data that a
1056  /// user should see in diagnostics, for example.
1057  ///
1058  /// Note that a presumed location is always given as the expansion point of
1059  /// an expansion location, not at the spelling location.
1060  ///
1061  /// \returns The presumed location of the specified SourceLocation. If the
1062  /// presumed location cannot be calculate (e.g., because \p Loc is invalid
1063  /// or the file containing \p Loc has changed on disk), returns an invalid
1064  /// presumed location.
1065  PresumedLoc getPresumedLoc(SourceLocation Loc) const;
1066
1067  /// isFromSameFile - Returns true if both SourceLocations correspond to
1068  ///  the same file.
1069  bool isFromSameFile(SourceLocation Loc1, SourceLocation Loc2) const {
1070    return getFileID(Loc1) == getFileID(Loc2);
1071  }
1072
1073  /// isFromMainFile - Returns true if the file of provided SourceLocation is
1074  ///   the main file.
1075  bool isFromMainFile(SourceLocation Loc) const {
1076    return getFileID(Loc) == getMainFileID();
1077  }
1078
1079  /// isInSystemHeader - Returns if a SourceLocation is in a system header.
1080  bool isInSystemHeader(SourceLocation Loc) const {
1081    return getFileCharacteristic(Loc) != SrcMgr::C_User;
1082  }
1083
1084  /// isInExternCSystemHeader - Returns if a SourceLocation is in an "extern C"
1085  /// system header.
1086  bool isInExternCSystemHeader(SourceLocation Loc) const {
1087    return getFileCharacteristic(Loc) == SrcMgr::C_ExternCSystem;
1088  }
1089
1090  /// \brief The size of the SLocEnty that \p FID represents.
1091  unsigned getFileIDSize(FileID FID) const;
1092
1093  /// \brief Given a specific FileID, returns true if \p Loc is inside that
1094  /// FileID chunk and sets relative offset (offset of \p Loc from beginning
1095  /// of FileID) to \p relativeOffset.
1096  bool isInFileID(SourceLocation Loc, FileID FID,
1097                  unsigned *RelativeOffset = 0) const {
1098    unsigned Offs = Loc.getOffset();
1099    if (isOffsetInFileID(FID, Offs)) {
1100      if (RelativeOffset)
1101        *RelativeOffset = Offs - getSLocEntry(FID).getOffset();
1102      return true;
1103    }
1104
1105    return false;
1106  }
1107
1108  //===--------------------------------------------------------------------===//
1109  // Line Table Manipulation Routines
1110  //===--------------------------------------------------------------------===//
1111
1112  /// getLineTableFilenameID - Return the uniqued ID for the specified filename.
1113  ///
1114  unsigned getLineTableFilenameID(StringRef Str);
1115
1116  /// AddLineNote - Add a line note to the line table for the FileID and offset
1117  /// specified by Loc.  If FilenameID is -1, it is considered to be
1118  /// unspecified.
1119  void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID);
1120  void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID,
1121                   bool IsFileEntry, bool IsFileExit,
1122                   bool IsSystemHeader, bool IsExternCHeader);
1123
1124  /// \brief Determine if the source manager has a line table.
1125  bool hasLineTable() const { return LineTable != 0; }
1126
1127  /// \brief Retrieve the stored line table.
1128  LineTableInfo &getLineTable();
1129
1130  //===--------------------------------------------------------------------===//
1131  // Queries for performance analysis.
1132  //===--------------------------------------------------------------------===//
1133
1134  /// Return the total amount of physical memory allocated by the
1135  /// ContentCache allocator.
1136  size_t getContentCacheSize() const {
1137    return ContentCacheAlloc.getTotalMemory();
1138  }
1139
1140  struct MemoryBufferSizes {
1141    const size_t malloc_bytes;
1142    const size_t mmap_bytes;
1143
1144    MemoryBufferSizes(size_t malloc_bytes, size_t mmap_bytes)
1145      : malloc_bytes(malloc_bytes), mmap_bytes(mmap_bytes) {}
1146  };
1147
1148  /// Return the amount of memory used by memory buffers, breaking down
1149  /// by heap-backed versus mmap'ed memory.
1150  MemoryBufferSizes getMemoryBufferSizes() const;
1151
1152  // Return the amount of memory used for various side tables and
1153  // data structures in the SourceManager.
1154  size_t getDataStructureSizes() const;
1155
1156  //===--------------------------------------------------------------------===//
1157  // Other miscellaneous methods.
1158  //===--------------------------------------------------------------------===//
1159
1160  /// \brief Get the source location for the given file:line:col triplet.
1161  ///
1162  /// If the source file is included multiple times, the source location will
1163  /// be based upon the first inclusion.
1164  SourceLocation translateFileLineCol(const FileEntry *SourceFile,
1165                                      unsigned Line, unsigned Col) const;
1166
1167  /// \brief Get the FileID for the given file.
1168  ///
1169  /// If the source file is included multiple times, the FileID will be the
1170  /// first inclusion.
1171  FileID translateFile(const FileEntry *SourceFile) const;
1172
1173  /// \brief Get the source location in \p FID for the given line:col.
1174  /// Returns null location if \p FID is not a file SLocEntry.
1175  SourceLocation translateLineCol(FileID FID,
1176                                  unsigned Line, unsigned Col) const;
1177
1178  /// \brief If \p Loc points inside a function macro argument, the returned
1179  /// location will be the macro location in which the argument was expanded.
1180  /// If a macro argument is used multiple times, the expanded location will
1181  /// be at the first expansion of the argument.
1182  /// e.g.
1183  ///   MY_MACRO(foo);
1184  ///             ^
1185  /// Passing a file location pointing at 'foo', will yield a macro location
1186  /// where 'foo' was expanded into.
1187  SourceLocation getMacroArgExpandedLocation(SourceLocation Loc) const;
1188
1189  /// \brief Determines the order of 2 source locations in the translation unit.
1190  ///
1191  /// \returns true if LHS source location comes before RHS, false otherwise.
1192  bool isBeforeInTranslationUnit(SourceLocation LHS, SourceLocation RHS) const;
1193
1194  /// \brief Comparison function class.
1195  class LocBeforeThanCompare : public std::binary_function<SourceLocation,
1196                                                         SourceLocation, bool> {
1197    SourceManager &SM;
1198
1199  public:
1200    explicit LocBeforeThanCompare(SourceManager &SM) : SM(SM) { }
1201
1202    bool operator()(SourceLocation LHS, SourceLocation RHS) const {
1203      return SM.isBeforeInTranslationUnit(LHS, RHS);
1204    }
1205  };
1206
1207  /// \brief Determines the order of 2 source locations in the "source location
1208  /// address space".
1209  bool isBeforeInSLocAddrSpace(SourceLocation LHS, SourceLocation RHS) const {
1210    return isBeforeInSLocAddrSpace(LHS, RHS.getOffset());
1211  }
1212
1213  /// \brief Determines the order of a source location and a source location
1214  /// offset in the "source location address space".
1215  ///
1216  /// Note that we always consider source locations loaded from
1217  bool isBeforeInSLocAddrSpace(SourceLocation LHS, unsigned RHS) const {
1218    unsigned LHSOffset = LHS.getOffset();
1219    bool LHSLoaded = LHSOffset >= CurrentLoadedOffset;
1220    bool RHSLoaded = RHS >= CurrentLoadedOffset;
1221    if (LHSLoaded == RHSLoaded)
1222      return LHSOffset < RHS;
1223
1224    return LHSLoaded;
1225  }
1226
1227  // Iterators over FileInfos.
1228  typedef llvm::DenseMap<const FileEntry*, SrcMgr::ContentCache*>
1229      ::const_iterator fileinfo_iterator;
1230  fileinfo_iterator fileinfo_begin() const { return FileInfos.begin(); }
1231  fileinfo_iterator fileinfo_end() const { return FileInfos.end(); }
1232  bool hasFileInfo(const FileEntry *File) const {
1233    return FileInfos.find(File) != FileInfos.end();
1234  }
1235
1236  /// PrintStats - Print statistics to stderr.
1237  ///
1238  void PrintStats() const;
1239
1240  /// \brief Get the number of local SLocEntries we have.
1241  unsigned local_sloc_entry_size() const { return LocalSLocEntryTable.size(); }
1242
1243  /// \brief Get a local SLocEntry. This is exposed for indexing.
1244  const SrcMgr::SLocEntry &getLocalSLocEntry(unsigned Index,
1245                                             bool *Invalid = 0) const {
1246    assert(Index < LocalSLocEntryTable.size() && "Invalid index");
1247    return LocalSLocEntryTable[Index];
1248  }
1249
1250  /// \brief Get the number of loaded SLocEntries we have.
1251  unsigned loaded_sloc_entry_size() const { return LoadedSLocEntryTable.size();}
1252
1253  /// \brief Get a loaded SLocEntry. This is exposed for indexing.
1254  const SrcMgr::SLocEntry &getLoadedSLocEntry(unsigned Index,
1255                                              bool *Invalid = 0) const {
1256    assert(Index < LoadedSLocEntryTable.size() && "Invalid index");
1257    if (!SLocEntryLoaded[Index])
1258      ExternalSLocEntries->ReadSLocEntry(-(static_cast<int>(Index) + 2));
1259    return LoadedSLocEntryTable[Index];
1260  }
1261
1262  const SrcMgr::SLocEntry &getSLocEntry(FileID FID, bool *Invalid = 0) const {
1263    if (FID.ID == 0 || FID.ID == -1) {
1264      if (Invalid) *Invalid = true;
1265      return LocalSLocEntryTable[0];
1266    }
1267    return getSLocEntryByID(FID.ID);
1268  }
1269
1270  unsigned getNextLocalOffset() const { return NextLocalOffset; }
1271
1272  void setExternalSLocEntrySource(ExternalSLocEntrySource *Source) {
1273    assert(LoadedSLocEntryTable.empty() &&
1274           "Invalidating existing loaded entries");
1275    ExternalSLocEntries = Source;
1276  }
1277
1278  /// \brief Allocate a number of loaded SLocEntries, which will be actually
1279  /// loaded on demand from the external source.
1280  ///
1281  /// NumSLocEntries will be allocated, which occupy a total of TotalSize space
1282  /// in the global source view. The lowest ID and the base offset of the
1283  /// entries will be returned.
1284  std::pair<int, unsigned>
1285  AllocateLoadedSLocEntries(unsigned NumSLocEntries, unsigned TotalSize);
1286
1287  /// \brief Returns true if \p Loc came from a PCH/Module.
1288  bool isLoadedSourceLocation(SourceLocation Loc) const {
1289    return Loc.getOffset() >= CurrentLoadedOffset;
1290  }
1291
1292  /// \brief Returns true if \p Loc did not come from a PCH/Module.
1293  bool isLocalSourceLocation(SourceLocation Loc) const {
1294    return Loc.getOffset() < NextLocalOffset;
1295  }
1296
1297  /// \brief Returns true if \p FID came from a PCH/Module.
1298  bool isLoadedFileID(FileID FID) const {
1299    assert(FID.ID != -1 && "Using FileID sentinel value");
1300    return FID.ID < 0;
1301  }
1302
1303  /// \brief Returns true if \p FID did not come from a PCH/Module.
1304  bool isLocalFileID(FileID FID) const {
1305    return !isLoadedFileID(FID);
1306  }
1307
1308private:
1309  const llvm::MemoryBuffer *getFakeBufferForRecovery() const;
1310
1311  /// \brief Get the entry with the given unwrapped FileID.
1312  const SrcMgr::SLocEntry &getSLocEntryByID(int ID) const {
1313    assert(ID != -1 && "Using FileID sentinel value");
1314    if (ID < 0)
1315      return getLoadedSLocEntryByID(ID);
1316    return getLocalSLocEntry(static_cast<unsigned>(ID));
1317  }
1318
1319  const SrcMgr::SLocEntry &getLoadedSLocEntryByID(int ID) const {
1320    return getLoadedSLocEntry(static_cast<unsigned>(-ID - 2));
1321  }
1322
1323  /// createExpansionLoc - Implements the common elements of storing an
1324  /// expansion info struct into the SLocEntry table and producing a source
1325  /// location that refers to it.
1326  SourceLocation createExpansionLocImpl(const SrcMgr::ExpansionInfo &Expansion,
1327                                        unsigned TokLength,
1328                                        int LoadedID = 0,
1329                                        unsigned LoadedOffset = 0);
1330
1331  /// isOffsetInFileID - Return true if the specified FileID contains the
1332  /// specified SourceLocation offset.  This is a very hot method.
1333  inline bool isOffsetInFileID(FileID FID, unsigned SLocOffset) const {
1334    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID);
1335    // If the entry is after the offset, it can't contain it.
1336    if (SLocOffset < Entry.getOffset()) return false;
1337
1338    // If this is the very last entry then it does.
1339    if (FID.ID == -2)
1340      return true;
1341
1342    // If it is the last local entry, then it does if the location is local.
1343    if (static_cast<unsigned>(FID.ID+1) == LocalSLocEntryTable.size()) {
1344      return SLocOffset < NextLocalOffset;
1345    }
1346
1347    // Otherwise, the entry after it has to not include it. This works for both
1348    // local and loaded entries.
1349    return SLocOffset < getSLocEntry(FileID::get(FID.ID+1)).getOffset();
1350  }
1351
1352  /// createFileID - Create a new fileID for the specified ContentCache and
1353  ///  include position.  This works regardless of whether the ContentCache
1354  ///  corresponds to a file or some other input source.
1355  FileID createFileID(const SrcMgr::ContentCache* File,
1356                      SourceLocation IncludePos,
1357                      SrcMgr::CharacteristicKind DirCharacter,
1358                      int LoadedID, unsigned LoadedOffset);
1359
1360  const SrcMgr::ContentCache *
1361    getOrCreateContentCache(const FileEntry *SourceFile);
1362
1363  /// createMemBufferContentCache - Create a new ContentCache for the specified
1364  ///  memory buffer.
1365  const SrcMgr::ContentCache*
1366  createMemBufferContentCache(const llvm::MemoryBuffer *Buf);
1367
1368  FileID getFileIDSlow(unsigned SLocOffset) const;
1369  FileID getFileIDLocal(unsigned SLocOffset) const;
1370  FileID getFileIDLoaded(unsigned SLocOffset) const;
1371
1372  SourceLocation getExpansionLocSlowCase(SourceLocation Loc) const;
1373  SourceLocation getSpellingLocSlowCase(SourceLocation Loc) const;
1374  SourceLocation getFileLocSlowCase(SourceLocation Loc) const;
1375
1376  std::pair<FileID, unsigned>
1377  getDecomposedExpansionLocSlowCase(const SrcMgr::SLocEntry *E) const;
1378  std::pair<FileID, unsigned>
1379  getDecomposedSpellingLocSlowCase(const SrcMgr::SLocEntry *E,
1380                                   unsigned Offset) const;
1381  void computeMacroArgsCache(MacroArgsMap *&MacroArgsCache, FileID FID) const;
1382
1383  friend class ASTReader;
1384  friend class ASTWriter;
1385};
1386
1387
1388}  // end namespace clang
1389
1390#endif
1391