SourceManager.h revision a1f1fad8b60e1cb9d21a40a37f2e03150bcbeb6f
1d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch//===--- SourceManager.h - Track and cache source files ---------*- C++ -*-===//
2d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch//
3d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch//                     The LLVM Compiler Infrastructure
4d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch//
5d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch// This file is distributed under the University of Illinois Open Source
6d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch// License. See LICENSE.TXT for details.
7d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch//
85d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)//===----------------------------------------------------------------------===//
95d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)//
105d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)//  This file defines the SourceManager interface.
115d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)//
12d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch//===----------------------------------------------------------------------===//
135d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)
145d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)#ifndef LLVM_CLANG_SOURCEMANAGER_H
151320f92c476a1ad9d19dba2a48c72b75566198e9Primiano Tucci#define LLVM_CLANG_SOURCEMANAGER_H
16d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch
175d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)#include "clang/Basic/LLVM.h"
18d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch#include "clang/Basic/SourceLocation.h"
19d0247b1b59f9c528cb6df88b4f2b9afaf80d181eTorne (Richard Coles)#include "llvm/Support/Allocator.h"
20d0247b1b59f9c528cb6df88b4f2b9afaf80d181eTorne (Richard Coles)#include "llvm/Support/DataTypes.h"
21d0247b1b59f9c528cb6df88b4f2b9afaf80d181eTorne (Richard Coles)#include "llvm/ADT/PointerIntPair.h"
22d0247b1b59f9c528cb6df88b4f2b9afaf80d181eTorne (Richard Coles)#include "llvm/ADT/PointerUnion.h"
23d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch#include "llvm/ADT/IntrusiveRefCntPtr.h"
24d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch#include "llvm/ADT/DenseMap.h"
25d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch#include "llvm/Support/MemoryBuffer.h"
26d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch#include <map>
27d0247b1b59f9c528cb6df88b4f2b9afaf80d181eTorne (Richard Coles)#include <vector>
28d0247b1b59f9c528cb6df88b4f2b9afaf80d181eTorne (Richard Coles)#include <cassert>
291320f92c476a1ad9d19dba2a48c72b75566198e9Primiano Tucci
30d3868032626d59662ff73b372b5d584c1d144c53Ben Murdochnamespace clang {
31d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch
324e180b6a0b4720a9b8e9e959a882386f690f08ffTorne (Richard Coles)class DiagnosticsEngine;
334e180b6a0b4720a9b8e9e959a882386f690f08ffTorne (Richard Coles)class SourceManager;
344e180b6a0b4720a9b8e9e959a882386f690f08ffTorne (Richard Coles)class FileManager;
354e180b6a0b4720a9b8e9e959a882386f690f08ffTorne (Richard Coles)class FileEntry;
365d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)class LineTableInfo;
374e180b6a0b4720a9b8e9e959a882386f690f08ffTorne (Richard Coles)class LangOptions;
385d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)class ASTWriter;
39d3868032626d59662ff73b372b5d584c1d144c53Ben Murdochclass ASTReader;
40d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch
41d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch/// There are three different types of locations in a file: a spelling
42d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch/// location, an expansion location, and a presumed location.
43d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch///
44d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch/// Given an example of:
45d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch/// #define min(x, y) x < y ? x : y
46d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch///
47d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch/// and then later on a use of min:
48d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch/// #line 17
49d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch/// return min(a, b);
50d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch///
51d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch/// The expansion location is the line in the source code where the macro
52d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch/// was expanded (the return statement), the spelling location is the
536e8cce623b6e4fe0c9e4af605d675dd9d0338c38Torne (Richard Coles)/// location in the source where the macro was originally defined,
546e8cce623b6e4fe0c9e4af605d675dd9d0338c38Torne (Richard Coles)/// and the presumed location is where the line directive states that
55d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch/// the line is 17, or any other line.
56d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch
57d0247b1b59f9c528cb6df88b4f2b9afaf80d181eTorne (Richard Coles)/// SrcMgr - Public enums and private classes that are part of the
58ba5b9a6411cb1792fd21f0a078d7a25cd1ceec16Ben Murdoch/// SourceManager implementation.
59ba5b9a6411cb1792fd21f0a078d7a25cd1ceec16Ben Murdoch///
60ba5b9a6411cb1792fd21f0a078d7a25cd1ceec16Ben Murdochnamespace SrcMgr {
615d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)  /// CharacteristicKind - This is used to represent whether a file or directory
625d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)  /// holds normal user code, system code, or system code which is implicitly
635d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)  /// 'extern "C"' in C++ mode.  Entire directories can be tagged with this
645d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)  /// (this is maintained by DirectoryLookup and friends) as can specific
655d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)  /// FileInfos when a #pragma system_header is seen or various other cases.
665d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)  ///
675d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)  enum CharacteristicKind {
685d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    C_User, C_System, C_ExternCSystem
69ba5b9a6411cb1792fd21f0a078d7a25cd1ceec16Ben Murdoch  };
70ba5b9a6411cb1792fd21f0a078d7a25cd1ceec16Ben Murdoch
71ba5b9a6411cb1792fd21f0a078d7a25cd1ceec16Ben Murdoch  /// ContentCache - One instance of this struct is kept for every file
72d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch  /// loaded or used.  This object owns the MemoryBuffer object.
73d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch  class ContentCache {
74d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch    enum CCFlags {
75d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch      /// \brief Whether the buffer is invalid.
76d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch      InvalidFlag = 0x01,
77d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch      /// \brief Whether the buffer should not be freed on destruction.
78d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch      DoNotFreeFlag = 0x02
79d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch    };
805d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)
815d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    /// Buffer - The actual buffer containing the characters from the input
825d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    /// file.  This is owned by the ContentCache object.
835d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    /// The bits indicate indicates whether the buffer is invalid.
84d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch    mutable llvm::PointerIntPair<const llvm::MemoryBuffer *, 2> Buffer;
855d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)
865d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)  public:
875d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    /// Reference to the file entry representing this ContentCache.
885d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    /// This reference does not own the FileEntry object.
895d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    /// It is possible for this to be NULL if
90d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch    /// the ContentCache encapsulates an imaginary text buffer.
91d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch    const FileEntry *OrigEntry;
92d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch
93d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch    /// \brief References the file which the contents were actually loaded from.
94d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch    /// Can be different from 'Entry' if we overridden the contents of one file
956e8cce623b6e4fe0c9e4af605d675dd9d0338c38Torne (Richard Coles)    /// with the contents of another file.
96d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch    const FileEntry *ContentsEntry;
97d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch
98d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch    /// SourceLineCache - A bump pointer allocated array of offsets for each
99d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch    /// source line.  This is lazily computed.  This is owned by the
100d0247b1b59f9c528cb6df88b4f2b9afaf80d181eTorne (Richard Coles)    /// SourceManager BumpPointerAllocator object.
101d0247b1b59f9c528cb6df88b4f2b9afaf80d181eTorne (Richard Coles)    unsigned *SourceLineCache;
102ba5b9a6411cb1792fd21f0a078d7a25cd1ceec16Ben Murdoch
1035d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    /// NumLines - The number of lines in this ContentCache.  This is only valid
1045d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    /// if SourceLineCache is non-null.
1055d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    unsigned NumLines : 31;
1065d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)
1075d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    /// \brief Indicates whether the buffer itself was provided to override
1085d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    /// the actual file contents.
1095d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    ///
1105d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    /// When true, the original entry may be a virtual file that does not
1115d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    /// exist.
1125d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    unsigned BufferOverridden : 1;
1135d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)
1145d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    ContentCache(const FileEntry *Ent = 0)
115d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch      : Buffer(0, false), OrigEntry(Ent), ContentsEntry(Ent),
116d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch        SourceLineCache(0), NumLines(0), BufferOverridden(false) {}
117d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch
1181320f92c476a1ad9d19dba2a48c72b75566198e9Primiano Tucci    ContentCache(const FileEntry *Ent, const FileEntry *contentEnt)
1191320f92c476a1ad9d19dba2a48c72b75566198e9Primiano Tucci      : Buffer(0, false), OrigEntry(Ent), ContentsEntry(contentEnt),
1201320f92c476a1ad9d19dba2a48c72b75566198e9Primiano Tucci        SourceLineCache(0), NumLines(0), BufferOverridden(false) {}
1211320f92c476a1ad9d19dba2a48c72b75566198e9Primiano Tucci
1221320f92c476a1ad9d19dba2a48c72b75566198e9Primiano Tucci    ~ContentCache();
1231320f92c476a1ad9d19dba2a48c72b75566198e9Primiano Tucci
124d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch    /// The copy ctor does not allow copies where source object has either
125d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch    ///  a non-NULL Buffer or SourceLineCache.  Ownership of allocated memory
126d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch    ///  is not transferred, so this is a logical error.
1270f1bc08d4cfcc34181b0b5cbf065c40f687bf740Torne (Richard Coles)    ContentCache(const ContentCache &RHS)
128d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch      : Buffer(0, false), SourceLineCache(0), BufferOverridden(false)
129d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch    {
130d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch      OrigEntry = RHS.OrigEntry;
1310f1bc08d4cfcc34181b0b5cbf065c40f687bf740Torne (Richard Coles)      ContentsEntry = RHS.ContentsEntry;
1321320f92c476a1ad9d19dba2a48c72b75566198e9Primiano Tucci
1331320f92c476a1ad9d19dba2a48c72b75566198e9Primiano Tucci      assert (RHS.Buffer.getPointer() == 0 && RHS.SourceLineCache == 0 &&
134d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch              "Passed ContentCache object cannot own a buffer.");
1355d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)
1365d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)      NumLines = RHS.NumLines;
1375d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    }
1385d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)
1395d1f7b1de12d16ceb2c938c56701a3e8bfa558f7Torne (Richard Coles)    /// getBuffer - Returns the memory buffer for the associated content.
140d0247b1b59f9c528cb6df88b4f2b9afaf80d181eTorne (Richard Coles)    ///
141d3868032626d59662ff73b372b5d584c1d144c53Ben Murdoch    /// \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                          SrcMgr::CharacteristicKind Kind = SrcMgr::C_User) {
623    assert(MainFileID.isInvalid() && "MainFileID already set!");
624    MainFileID = createFileID(SourceFile, SourceLocation(), Kind);
625    return MainFileID;
626  }
627
628  /// \brief Set the file ID for the main source file.
629  void setMainFileID(FileID FID) {
630    assert(MainFileID.isInvalid() && "MainFileID already set!");
631    MainFileID = FID;
632  }
633
634  /// \brief Set the file ID for the precompiled preamble.
635  void setPreambleFileID(FileID Preamble) {
636    assert(PreambleFileID.isInvalid() && "PreambleFileID already set!");
637    PreambleFileID = Preamble;
638  }
639
640  /// \brief Get the file ID for the precompiled preamble if there is one.
641  FileID getPreambleFileID() const { return PreambleFileID; }
642
643  //===--------------------------------------------------------------------===//
644  // Methods to create new FileID's and macro expansions.
645  //===--------------------------------------------------------------------===//
646
647  /// createFileID - Create a new FileID that represents the specified file
648  /// being #included from the specified IncludePosition.  This translates NULL
649  /// into standard input.
650  FileID createFileID(const FileEntry *SourceFile, SourceLocation IncludePos,
651                      SrcMgr::CharacteristicKind FileCharacter,
652                      int LoadedID = 0, unsigned LoadedOffset = 0) {
653    const SrcMgr::ContentCache *IR = getOrCreateContentCache(SourceFile);
654    assert(IR && "getOrCreateContentCache() cannot return NULL");
655    return createFileID(IR, IncludePos, FileCharacter, LoadedID, LoadedOffset);
656  }
657
658  /// createFileIDForMemBuffer - Create a new FileID that represents the
659  /// specified memory buffer.  This does no caching of the buffer and takes
660  /// ownership of the MemoryBuffer, so only pass a MemoryBuffer to this once.
661  FileID createFileIDForMemBuffer(const llvm::MemoryBuffer *Buffer,
662                                  int LoadedID = 0, unsigned LoadedOffset = 0,
663                                 SourceLocation IncludeLoc = SourceLocation()) {
664    return createFileID(createMemBufferContentCache(Buffer), IncludeLoc,
665                        SrcMgr::C_User, LoadedID, LoadedOffset);
666  }
667
668  /// createMacroArgExpansionLoc - Return a new SourceLocation that encodes the
669  /// fact that a token from SpellingLoc should actually be referenced from
670  /// ExpansionLoc, and that it represents the expansion of a macro argument
671  /// into the function-like macro body.
672  SourceLocation createMacroArgExpansionLoc(SourceLocation Loc,
673                                            SourceLocation ExpansionLoc,
674                                            unsigned TokLength);
675
676  /// createExpansionLoc - Return a new SourceLocation that encodes the fact
677  /// that a token from SpellingLoc should actually be referenced from
678  /// ExpansionLoc.
679  SourceLocation createExpansionLoc(SourceLocation Loc,
680                                    SourceLocation ExpansionLocStart,
681                                    SourceLocation ExpansionLocEnd,
682                                    unsigned TokLength,
683                                    int LoadedID = 0,
684                                    unsigned LoadedOffset = 0);
685
686  /// \brief Retrieve the memory buffer associated with the given file.
687  ///
688  /// \param Invalid If non-NULL, will be set \c true if an error
689  /// occurs while retrieving the memory buffer.
690  const llvm::MemoryBuffer *getMemoryBufferForFile(const FileEntry *File,
691                                                   bool *Invalid = 0);
692
693  /// \brief Override the contents of the given source file by providing an
694  /// already-allocated buffer.
695  ///
696  /// \param SourceFile the source file whose contents will be overriden.
697  ///
698  /// \param Buffer the memory buffer whose contents will be used as the
699  /// data in the given source file.
700  ///
701  /// \param DoNotFree If true, then the buffer will not be freed when the
702  /// source manager is destroyed.
703  void overrideFileContents(const FileEntry *SourceFile,
704                            const llvm::MemoryBuffer *Buffer,
705                            bool DoNotFree = false);
706
707  /// \brief Override the the given source file with another one.
708  ///
709  /// \param SourceFile the source file which will be overriden.
710  ///
711  /// \param NewFile the file whose contents will be used as the
712  /// data instead of the contents of the given source file.
713  void overrideFileContents(const FileEntry *SourceFile,
714                            const FileEntry *NewFile);
715
716  //===--------------------------------------------------------------------===//
717  // FileID manipulation methods.
718  //===--------------------------------------------------------------------===//
719
720  /// getBuffer - Return the buffer for the specified FileID. If there is an
721  /// error opening this buffer the first time, this manufactures a temporary
722  /// buffer and returns a non-empty error string.
723  const llvm::MemoryBuffer *getBuffer(FileID FID, SourceLocation Loc,
724                                      bool *Invalid = 0) const {
725    bool MyInvalid = false;
726    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid);
727    if (MyInvalid || !Entry.isFile()) {
728      if (Invalid)
729        *Invalid = true;
730
731      return getFakeBufferForRecovery();
732    }
733
734    return Entry.getFile().getContentCache()->getBuffer(Diag, *this, Loc,
735                                                        Invalid);
736  }
737
738  const llvm::MemoryBuffer *getBuffer(FileID FID, bool *Invalid = 0) const {
739    bool MyInvalid = false;
740    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid);
741    if (MyInvalid || !Entry.isFile()) {
742      if (Invalid)
743        *Invalid = true;
744
745      return getFakeBufferForRecovery();
746    }
747
748    return Entry.getFile().getContentCache()->getBuffer(Diag, *this,
749                                                        SourceLocation(),
750                                                        Invalid);
751  }
752
753  /// getFileEntryForID - Returns the FileEntry record for the provided FileID.
754  const FileEntry *getFileEntryForID(FileID FID) const {
755    bool MyInvalid = false;
756    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid);
757    if (MyInvalid || !Entry.isFile())
758      return 0;
759
760    const SrcMgr::ContentCache *Content = Entry.getFile().getContentCache();
761    if (!Content)
762      return 0;
763    return Content->OrigEntry;
764  }
765
766  /// Returns the FileEntry record for the provided SLocEntry.
767  const FileEntry *getFileEntryForSLocEntry(const SrcMgr::SLocEntry &sloc) const
768  {
769    const SrcMgr::ContentCache *Content = sloc.getFile().getContentCache();
770    if (!Content)
771      return 0;
772    return Content->OrigEntry;
773  }
774
775  /// getBufferData - Return a StringRef to the source buffer data for the
776  /// specified FileID.
777  ///
778  /// \param FID The file ID whose contents will be returned.
779  /// \param Invalid If non-NULL, will be set true if an error occurred.
780  StringRef getBufferData(FileID FID, bool *Invalid = 0) const;
781
782  /// \brief Get the number of FileIDs (files and macros) that were created
783  /// during preprocessing of \p FID, including it.
784  unsigned getNumCreatedFIDsForFileID(FileID FID) const {
785    bool Invalid = false;
786    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
787    if (Invalid || !Entry.isFile())
788      return 0;
789
790    return Entry.getFile().NumCreatedFIDs;
791  }
792
793  /// \brief Set the number of FileIDs (files and macros) that were created
794  /// during preprocessing of \p FID, including it.
795  void setNumCreatedFIDsForFileID(FileID FID, unsigned NumFIDs) const {
796    bool Invalid = false;
797    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
798    if (Invalid || !Entry.isFile())
799      return;
800
801    assert(Entry.getFile().NumCreatedFIDs == 0 && "Already set!");
802    const_cast<SrcMgr::FileInfo &>(Entry.getFile()).NumCreatedFIDs = NumFIDs;
803  }
804
805  //===--------------------------------------------------------------------===//
806  // SourceLocation manipulation methods.
807  //===--------------------------------------------------------------------===//
808
809  /// getFileID - Return the FileID for a SourceLocation.  This is a very
810  /// hot method that is used for all SourceManager queries that start with a
811  /// SourceLocation object.  It is responsible for finding the entry in
812  /// SLocEntryTable which contains the specified location.
813  ///
814  FileID getFileID(SourceLocation SpellingLoc) const {
815    unsigned SLocOffset = SpellingLoc.getOffset();
816
817    // If our one-entry cache covers this offset, just return it.
818    if (isOffsetInFileID(LastFileIDLookup, SLocOffset))
819      return LastFileIDLookup;
820
821    return getFileIDSlow(SLocOffset);
822  }
823
824  /// getLocForStartOfFile - Return the source location corresponding to the
825  /// first byte of the specified file.
826  SourceLocation getLocForStartOfFile(FileID FID) const {
827    bool Invalid = false;
828    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
829    if (Invalid || !Entry.isFile())
830      return SourceLocation();
831
832    unsigned FileOffset = Entry.getOffset();
833    return SourceLocation::getFileLoc(FileOffset);
834  }
835
836  /// \brief Return the source location corresponding to the last byte of the
837  /// specified file.
838  SourceLocation getLocForEndOfFile(FileID FID) const {
839    bool Invalid = false;
840    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
841    if (Invalid || !Entry.isFile())
842      return SourceLocation();
843
844    unsigned FileOffset = Entry.getOffset();
845    return SourceLocation::getFileLoc(FileOffset + getFileIDSize(FID) - 1);
846  }
847
848  /// \brief Returns the include location if \p FID is a #include'd file
849  /// otherwise it returns an invalid location.
850  SourceLocation getIncludeLoc(FileID FID) const {
851    bool Invalid = false;
852    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
853    if (Invalid || !Entry.isFile())
854      return SourceLocation();
855
856    return Entry.getFile().getIncludeLoc();
857  }
858
859  /// getExpansionLoc - Given a SourceLocation object, return the expansion
860  /// location referenced by the ID.
861  SourceLocation getExpansionLoc(SourceLocation Loc) const {
862    // Handle the non-mapped case inline, defer to out of line code to handle
863    // expansions.
864    if (Loc.isFileID()) return Loc;
865    return getExpansionLocSlowCase(Loc);
866  }
867
868  /// \brief Given \p Loc, if it is a macro location return the expansion
869  /// location or the spelling location, depending on if it comes from a
870  /// macro argument or not.
871  SourceLocation getFileLoc(SourceLocation Loc) const {
872    if (Loc.isFileID()) return Loc;
873    return getFileLocSlowCase(Loc);
874  }
875
876  /// getImmediateExpansionRange - Loc is required to be an expansion location.
877  /// Return the start/end of the expansion information.
878  std::pair<SourceLocation,SourceLocation>
879  getImmediateExpansionRange(SourceLocation Loc) const;
880
881  /// getExpansionRange - Given a SourceLocation object, return the range of
882  /// tokens covered by the expansion the ultimate file.
883  std::pair<SourceLocation,SourceLocation>
884  getExpansionRange(SourceLocation Loc) const;
885
886
887  /// getSpellingLoc - Given a SourceLocation object, return the spelling
888  /// location referenced by the ID.  This is the place where the characters
889  /// that make up the lexed token can be found.
890  SourceLocation getSpellingLoc(SourceLocation Loc) const {
891    // Handle the non-mapped case inline, defer to out of line code to handle
892    // expansions.
893    if (Loc.isFileID()) return Loc;
894    return getSpellingLocSlowCase(Loc);
895  }
896
897  /// getImmediateSpellingLoc - Given a SourceLocation object, return the
898  /// spelling location referenced by the ID.  This is the first level down
899  /// towards the place where the characters that make up the lexed token can be
900  /// found.  This should not generally be used by clients.
901  SourceLocation getImmediateSpellingLoc(SourceLocation Loc) const;
902
903  /// getDecomposedLoc - Decompose the specified location into a raw FileID +
904  /// Offset pair.  The first element is the FileID, the second is the
905  /// offset from the start of the buffer of the location.
906  std::pair<FileID, unsigned> getDecomposedLoc(SourceLocation Loc) const {
907    FileID FID = getFileID(Loc);
908    bool Invalid = false;
909    const SrcMgr::SLocEntry &E = getSLocEntry(FID, &Invalid);
910    if (Invalid)
911      return std::make_pair(FileID(), 0);
912    return std::make_pair(FID, Loc.getOffset()-E.getOffset());
913  }
914
915  /// getDecomposedExpansionLoc - Decompose the specified location into a raw
916  /// FileID + Offset pair. If the location is an expansion record, walk
917  /// through it until we find the final location expanded.
918  std::pair<FileID, unsigned>
919  getDecomposedExpansionLoc(SourceLocation Loc) const {
920    FileID FID = getFileID(Loc);
921    bool Invalid = false;
922    const SrcMgr::SLocEntry *E = &getSLocEntry(FID, &Invalid);
923    if (Invalid)
924      return std::make_pair(FileID(), 0);
925
926    unsigned Offset = Loc.getOffset()-E->getOffset();
927    if (Loc.isFileID())
928      return std::make_pair(FID, Offset);
929
930    return getDecomposedExpansionLocSlowCase(E);
931  }
932
933  /// getDecomposedSpellingLoc - Decompose the specified location into a raw
934  /// FileID + Offset pair.  If the location is an expansion record, walk
935  /// through it until we find its spelling record.
936  std::pair<FileID, unsigned>
937  getDecomposedSpellingLoc(SourceLocation Loc) const {
938    FileID FID = getFileID(Loc);
939    bool Invalid = false;
940    const SrcMgr::SLocEntry *E = &getSLocEntry(FID, &Invalid);
941    if (Invalid)
942      return std::make_pair(FileID(), 0);
943
944    unsigned Offset = Loc.getOffset()-E->getOffset();
945    if (Loc.isFileID())
946      return std::make_pair(FID, Offset);
947    return getDecomposedSpellingLocSlowCase(E, Offset);
948  }
949
950  /// getFileOffset - This method returns the offset from the start
951  /// of the file that the specified SourceLocation represents. This is not very
952  /// meaningful for a macro ID.
953  unsigned getFileOffset(SourceLocation SpellingLoc) const {
954    return getDecomposedLoc(SpellingLoc).second;
955  }
956
957  /// isMacroArgExpansion - This method tests whether the given source location
958  /// represents a macro argument's expansion into the function-like macro
959  /// definition. Such source locations only appear inside of the expansion
960  /// locations representing where a particular function-like macro was
961  /// expanded.
962  bool isMacroArgExpansion(SourceLocation Loc) const;
963
964  /// \brief Returns true if \p Loc is inside the [\p Start, +\p Length)
965  /// chunk of the source location address space.
966  /// If it's true and \p RelativeOffset is non-null, it will be set to the
967  /// relative offset of \p Loc inside the chunk.
968  bool isInSLocAddrSpace(SourceLocation Loc,
969                         SourceLocation Start, unsigned Length,
970                         unsigned *RelativeOffset = 0) const {
971    assert(((Start.getOffset() < NextLocalOffset &&
972               Start.getOffset()+Length <= NextLocalOffset) ||
973            (Start.getOffset() >= CurrentLoadedOffset &&
974                Start.getOffset()+Length < MaxLoadedOffset)) &&
975           "Chunk is not valid SLoc address space");
976    unsigned LocOffs = Loc.getOffset();
977    unsigned BeginOffs = Start.getOffset();
978    unsigned EndOffs = BeginOffs + Length;
979    if (LocOffs >= BeginOffs && LocOffs < EndOffs) {
980      if (RelativeOffset)
981        *RelativeOffset = LocOffs - BeginOffs;
982      return true;
983    }
984
985    return false;
986  }
987
988  /// \brief Return true if both \p LHS and \p RHS are in the local source
989  /// location address space or the loaded one. If it's true and \p
990  /// RelativeOffset is non-null, it will be set to the offset of \p RHS
991  /// relative to \p LHS.
992  bool isInSameSLocAddrSpace(SourceLocation LHS, SourceLocation RHS,
993                             int *RelativeOffset) const {
994    unsigned LHSOffs = LHS.getOffset(), RHSOffs = RHS.getOffset();
995    bool LHSLoaded = LHSOffs >= CurrentLoadedOffset;
996    bool RHSLoaded = RHSOffs >= CurrentLoadedOffset;
997
998    if (LHSLoaded == RHSLoaded) {
999      if (RelativeOffset)
1000        *RelativeOffset = RHSOffs - LHSOffs;
1001      return true;
1002    }
1003
1004    return false;
1005  }
1006
1007  //===--------------------------------------------------------------------===//
1008  // Queries about the code at a SourceLocation.
1009  //===--------------------------------------------------------------------===//
1010
1011  /// getCharacterData - Return a pointer to the start of the specified location
1012  /// in the appropriate spelling MemoryBuffer.
1013  ///
1014  /// \param Invalid If non-NULL, will be set \c true if an error occurs.
1015  const char *getCharacterData(SourceLocation SL, bool *Invalid = 0) const;
1016
1017  /// getColumnNumber - Return the column # for the specified file position.
1018  /// This is significantly cheaper to compute than the line number.  This
1019  /// returns zero if the column number isn't known.  This may only be called
1020  /// on a file sloc, so you must choose a spelling or expansion location
1021  /// before calling this method.
1022  unsigned getColumnNumber(FileID FID, unsigned FilePos,
1023                           bool *Invalid = 0) const;
1024  unsigned getSpellingColumnNumber(SourceLocation Loc, bool *Invalid = 0) const;
1025  unsigned getExpansionColumnNumber(SourceLocation Loc,
1026                                    bool *Invalid = 0) const;
1027  unsigned getPresumedColumnNumber(SourceLocation Loc, bool *Invalid = 0) const;
1028
1029
1030  /// getLineNumber - Given a SourceLocation, return the spelling line number
1031  /// for the position indicated.  This requires building and caching a table of
1032  /// line offsets for the MemoryBuffer, so this is not cheap: use only when
1033  /// about to emit a diagnostic.
1034  unsigned getLineNumber(FileID FID, unsigned FilePos, bool *Invalid = 0) const;
1035  unsigned getSpellingLineNumber(SourceLocation Loc, bool *Invalid = 0) const;
1036  unsigned getExpansionLineNumber(SourceLocation Loc, bool *Invalid = 0) const;
1037  unsigned getPresumedLineNumber(SourceLocation Loc, bool *Invalid = 0) const;
1038
1039  /// Return the filename or buffer identifier of the buffer the location is in.
1040  /// Note that this name does not respect #line directives.  Use getPresumedLoc
1041  /// for normal clients.
1042  const char *getBufferName(SourceLocation Loc, bool *Invalid = 0) const;
1043
1044  /// getFileCharacteristic - return the file characteristic of the specified
1045  /// source location, indicating whether this is a normal file, a system
1046  /// header, or an "implicit extern C" system header.
1047  ///
1048  /// This state can be modified with flags on GNU linemarker directives like:
1049  ///   # 4 "foo.h" 3
1050  /// which changes all source locations in the current file after that to be
1051  /// considered to be from a system header.
1052  SrcMgr::CharacteristicKind getFileCharacteristic(SourceLocation Loc) const;
1053
1054  /// getPresumedLoc - This method returns the "presumed" location of a
1055  /// SourceLocation specifies.  A "presumed location" can be modified by #line
1056  /// or GNU line marker directives.  This provides a view on the data that a
1057  /// user should see in diagnostics, for example.
1058  ///
1059  /// Note that a presumed location is always given as the expansion point of
1060  /// an expansion location, not at the spelling location.
1061  ///
1062  /// \returns The presumed location of the specified SourceLocation. If the
1063  /// presumed location cannot be calculate (e.g., because \p Loc is invalid
1064  /// or the file containing \p Loc has changed on disk), returns an invalid
1065  /// presumed location.
1066  PresumedLoc getPresumedLoc(SourceLocation Loc) const;
1067
1068  /// isFromSameFile - Returns true if both SourceLocations correspond to
1069  ///  the same file.
1070  bool isFromSameFile(SourceLocation Loc1, SourceLocation Loc2) const {
1071    return getFileID(Loc1) == getFileID(Loc2);
1072  }
1073
1074  /// isFromMainFile - Returns true if the file of provided SourceLocation is
1075  ///   the main file.
1076  bool isFromMainFile(SourceLocation Loc) const {
1077    return getFileID(Loc) == getMainFileID();
1078  }
1079
1080  /// isInSystemHeader - Returns if a SourceLocation is in a system header.
1081  bool isInSystemHeader(SourceLocation Loc) const {
1082    return getFileCharacteristic(Loc) != SrcMgr::C_User;
1083  }
1084
1085  /// isInExternCSystemHeader - Returns if a SourceLocation is in an "extern C"
1086  /// system header.
1087  bool isInExternCSystemHeader(SourceLocation Loc) const {
1088    return getFileCharacteristic(Loc) == SrcMgr::C_ExternCSystem;
1089  }
1090
1091  /// \brief Returns whether \p Loc is expanded from a macro in a system header.
1092  bool isInSystemMacro(SourceLocation loc) {
1093    return loc.isMacroID() && isInSystemHeader(getSpellingLoc(loc));
1094  }
1095
1096  /// \brief The size of the SLocEnty that \p FID represents.
1097  unsigned getFileIDSize(FileID FID) const;
1098
1099  /// \brief Given a specific FileID, returns true if \p Loc is inside that
1100  /// FileID chunk and sets relative offset (offset of \p Loc from beginning
1101  /// of FileID) to \p relativeOffset.
1102  bool isInFileID(SourceLocation Loc, FileID FID,
1103                  unsigned *RelativeOffset = 0) const {
1104    unsigned Offs = Loc.getOffset();
1105    if (isOffsetInFileID(FID, Offs)) {
1106      if (RelativeOffset)
1107        *RelativeOffset = Offs - getSLocEntry(FID).getOffset();
1108      return true;
1109    }
1110
1111    return false;
1112  }
1113
1114  //===--------------------------------------------------------------------===//
1115  // Line Table Manipulation Routines
1116  //===--------------------------------------------------------------------===//
1117
1118  /// getLineTableFilenameID - Return the uniqued ID for the specified filename.
1119  ///
1120  unsigned getLineTableFilenameID(StringRef Str);
1121
1122  /// AddLineNote - Add a line note to the line table for the FileID and offset
1123  /// specified by Loc.  If FilenameID is -1, it is considered to be
1124  /// unspecified.
1125  void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID);
1126  void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID,
1127                   bool IsFileEntry, bool IsFileExit,
1128                   bool IsSystemHeader, bool IsExternCHeader);
1129
1130  /// \brief Determine if the source manager has a line table.
1131  bool hasLineTable() const { return LineTable != 0; }
1132
1133  /// \brief Retrieve the stored line table.
1134  LineTableInfo &getLineTable();
1135
1136  //===--------------------------------------------------------------------===//
1137  // Queries for performance analysis.
1138  //===--------------------------------------------------------------------===//
1139
1140  /// Return the total amount of physical memory allocated by the
1141  /// ContentCache allocator.
1142  size_t getContentCacheSize() const {
1143    return ContentCacheAlloc.getTotalMemory();
1144  }
1145
1146  struct MemoryBufferSizes {
1147    const size_t malloc_bytes;
1148    const size_t mmap_bytes;
1149
1150    MemoryBufferSizes(size_t malloc_bytes, size_t mmap_bytes)
1151      : malloc_bytes(malloc_bytes), mmap_bytes(mmap_bytes) {}
1152  };
1153
1154  /// Return the amount of memory used by memory buffers, breaking down
1155  /// by heap-backed versus mmap'ed memory.
1156  MemoryBufferSizes getMemoryBufferSizes() const;
1157
1158  // Return the amount of memory used for various side tables and
1159  // data structures in the SourceManager.
1160  size_t getDataStructureSizes() const;
1161
1162  //===--------------------------------------------------------------------===//
1163  // Other miscellaneous methods.
1164  //===--------------------------------------------------------------------===//
1165
1166  /// \brief Get the source location for the given file:line:col triplet.
1167  ///
1168  /// If the source file is included multiple times, the source location will
1169  /// be based upon the first inclusion.
1170  SourceLocation translateFileLineCol(const FileEntry *SourceFile,
1171                                      unsigned Line, unsigned Col) const;
1172
1173  /// \brief Get the FileID for the given file.
1174  ///
1175  /// If the source file is included multiple times, the FileID will be the
1176  /// first inclusion.
1177  FileID translateFile(const FileEntry *SourceFile) const;
1178
1179  /// \brief Get the source location in \p FID for the given line:col.
1180  /// Returns null location if \p FID is not a file SLocEntry.
1181  SourceLocation translateLineCol(FileID FID,
1182                                  unsigned Line, unsigned Col) const;
1183
1184  /// \brief If \p Loc points inside a function macro argument, the returned
1185  /// location will be the macro location in which the argument was expanded.
1186  /// If a macro argument is used multiple times, the expanded location will
1187  /// be at the first expansion of the argument.
1188  /// e.g.
1189  ///   MY_MACRO(foo);
1190  ///             ^
1191  /// Passing a file location pointing at 'foo', will yield a macro location
1192  /// where 'foo' was expanded into.
1193  SourceLocation getMacroArgExpandedLocation(SourceLocation Loc) const;
1194
1195  /// \brief Determines the order of 2 source locations in the translation unit.
1196  ///
1197  /// \returns true if LHS source location comes before RHS, false otherwise.
1198  bool isBeforeInTranslationUnit(SourceLocation LHS, SourceLocation RHS) const;
1199
1200  /// \brief Comparison function class.
1201  class LocBeforeThanCompare : public std::binary_function<SourceLocation,
1202                                                         SourceLocation, bool> {
1203    SourceManager &SM;
1204
1205  public:
1206    explicit LocBeforeThanCompare(SourceManager &SM) : SM(SM) { }
1207
1208    bool operator()(SourceLocation LHS, SourceLocation RHS) const {
1209      return SM.isBeforeInTranslationUnit(LHS, RHS);
1210    }
1211  };
1212
1213  /// \brief Determines the order of 2 source locations in the "source location
1214  /// address space".
1215  bool isBeforeInSLocAddrSpace(SourceLocation LHS, SourceLocation RHS) const {
1216    return isBeforeInSLocAddrSpace(LHS, RHS.getOffset());
1217  }
1218
1219  /// \brief Determines the order of a source location and a source location
1220  /// offset in the "source location address space".
1221  ///
1222  /// Note that we always consider source locations loaded from
1223  bool isBeforeInSLocAddrSpace(SourceLocation LHS, unsigned RHS) const {
1224    unsigned LHSOffset = LHS.getOffset();
1225    bool LHSLoaded = LHSOffset >= CurrentLoadedOffset;
1226    bool RHSLoaded = RHS >= CurrentLoadedOffset;
1227    if (LHSLoaded == RHSLoaded)
1228      return LHSOffset < RHS;
1229
1230    return LHSLoaded;
1231  }
1232
1233  // Iterators over FileInfos.
1234  typedef llvm::DenseMap<const FileEntry*, SrcMgr::ContentCache*>
1235      ::const_iterator fileinfo_iterator;
1236  fileinfo_iterator fileinfo_begin() const { return FileInfos.begin(); }
1237  fileinfo_iterator fileinfo_end() const { return FileInfos.end(); }
1238  bool hasFileInfo(const FileEntry *File) const {
1239    return FileInfos.find(File) != FileInfos.end();
1240  }
1241
1242  /// PrintStats - Print statistics to stderr.
1243  ///
1244  void PrintStats() const;
1245
1246  /// \brief Get the number of local SLocEntries we have.
1247  unsigned local_sloc_entry_size() const { return LocalSLocEntryTable.size(); }
1248
1249  /// \brief Get a local SLocEntry. This is exposed for indexing.
1250  const SrcMgr::SLocEntry &getLocalSLocEntry(unsigned Index,
1251                                             bool *Invalid = 0) const {
1252    assert(Index < LocalSLocEntryTable.size() && "Invalid index");
1253    return LocalSLocEntryTable[Index];
1254  }
1255
1256  /// \brief Get the number of loaded SLocEntries we have.
1257  unsigned loaded_sloc_entry_size() const { return LoadedSLocEntryTable.size();}
1258
1259  /// \brief Get a loaded SLocEntry. This is exposed for indexing.
1260  const SrcMgr::SLocEntry &getLoadedSLocEntry(unsigned Index,
1261                                              bool *Invalid = 0) const {
1262    assert(Index < LoadedSLocEntryTable.size() && "Invalid index");
1263    if (!SLocEntryLoaded[Index])
1264      ExternalSLocEntries->ReadSLocEntry(-(static_cast<int>(Index) + 2));
1265    return LoadedSLocEntryTable[Index];
1266  }
1267
1268  const SrcMgr::SLocEntry &getSLocEntry(FileID FID, bool *Invalid = 0) const {
1269    if (FID.ID == 0 || FID.ID == -1) {
1270      if (Invalid) *Invalid = true;
1271      return LocalSLocEntryTable[0];
1272    }
1273    return getSLocEntryByID(FID.ID);
1274  }
1275
1276  unsigned getNextLocalOffset() const { return NextLocalOffset; }
1277
1278  void setExternalSLocEntrySource(ExternalSLocEntrySource *Source) {
1279    assert(LoadedSLocEntryTable.empty() &&
1280           "Invalidating existing loaded entries");
1281    ExternalSLocEntries = Source;
1282  }
1283
1284  /// \brief Allocate a number of loaded SLocEntries, which will be actually
1285  /// loaded on demand from the external source.
1286  ///
1287  /// NumSLocEntries will be allocated, which occupy a total of TotalSize space
1288  /// in the global source view. The lowest ID and the base offset of the
1289  /// entries will be returned.
1290  std::pair<int, unsigned>
1291  AllocateLoadedSLocEntries(unsigned NumSLocEntries, unsigned TotalSize);
1292
1293  /// \brief Returns true if \p Loc came from a PCH/Module.
1294  bool isLoadedSourceLocation(SourceLocation Loc) const {
1295    return Loc.getOffset() >= CurrentLoadedOffset;
1296  }
1297
1298  /// \brief Returns true if \p Loc did not come from a PCH/Module.
1299  bool isLocalSourceLocation(SourceLocation Loc) const {
1300    return Loc.getOffset() < NextLocalOffset;
1301  }
1302
1303  /// \brief Returns true if \p FID came from a PCH/Module.
1304  bool isLoadedFileID(FileID FID) const {
1305    assert(FID.ID != -1 && "Using FileID sentinel value");
1306    return FID.ID < 0;
1307  }
1308
1309  /// \brief Returns true if \p FID did not come from a PCH/Module.
1310  bool isLocalFileID(FileID FID) const {
1311    return !isLoadedFileID(FID);
1312  }
1313
1314private:
1315  const llvm::MemoryBuffer *getFakeBufferForRecovery() const;
1316
1317  /// \brief Get the entry with the given unwrapped FileID.
1318  const SrcMgr::SLocEntry &getSLocEntryByID(int ID) const {
1319    assert(ID != -1 && "Using FileID sentinel value");
1320    if (ID < 0)
1321      return getLoadedSLocEntryByID(ID);
1322    return getLocalSLocEntry(static_cast<unsigned>(ID));
1323  }
1324
1325  const SrcMgr::SLocEntry &getLoadedSLocEntryByID(int ID) const {
1326    return getLoadedSLocEntry(static_cast<unsigned>(-ID - 2));
1327  }
1328
1329  /// createExpansionLoc - Implements the common elements of storing an
1330  /// expansion info struct into the SLocEntry table and producing a source
1331  /// location that refers to it.
1332  SourceLocation createExpansionLocImpl(const SrcMgr::ExpansionInfo &Expansion,
1333                                        unsigned TokLength,
1334                                        int LoadedID = 0,
1335                                        unsigned LoadedOffset = 0);
1336
1337  /// isOffsetInFileID - Return true if the specified FileID contains the
1338  /// specified SourceLocation offset.  This is a very hot method.
1339  inline bool isOffsetInFileID(FileID FID, unsigned SLocOffset) const {
1340    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID);
1341    // If the entry is after the offset, it can't contain it.
1342    if (SLocOffset < Entry.getOffset()) return false;
1343
1344    // If this is the very last entry then it does.
1345    if (FID.ID == -2)
1346      return true;
1347
1348    // If it is the last local entry, then it does if the location is local.
1349    if (static_cast<unsigned>(FID.ID+1) == LocalSLocEntryTable.size()) {
1350      return SLocOffset < NextLocalOffset;
1351    }
1352
1353    // Otherwise, the entry after it has to not include it. This works for both
1354    // local and loaded entries.
1355    return SLocOffset < getSLocEntry(FileID::get(FID.ID+1)).getOffset();
1356  }
1357
1358  /// createFileID - Create a new fileID for the specified ContentCache and
1359  ///  include position.  This works regardless of whether the ContentCache
1360  ///  corresponds to a file or some other input source.
1361  FileID createFileID(const SrcMgr::ContentCache* File,
1362                      SourceLocation IncludePos,
1363                      SrcMgr::CharacteristicKind DirCharacter,
1364                      int LoadedID, unsigned LoadedOffset);
1365
1366  const SrcMgr::ContentCache *
1367    getOrCreateContentCache(const FileEntry *SourceFile);
1368
1369  /// createMemBufferContentCache - Create a new ContentCache for the specified
1370  ///  memory buffer.
1371  const SrcMgr::ContentCache*
1372  createMemBufferContentCache(const llvm::MemoryBuffer *Buf);
1373
1374  FileID getFileIDSlow(unsigned SLocOffset) const;
1375  FileID getFileIDLocal(unsigned SLocOffset) const;
1376  FileID getFileIDLoaded(unsigned SLocOffset) const;
1377
1378  SourceLocation getExpansionLocSlowCase(SourceLocation Loc) const;
1379  SourceLocation getSpellingLocSlowCase(SourceLocation Loc) const;
1380  SourceLocation getFileLocSlowCase(SourceLocation Loc) const;
1381
1382  std::pair<FileID, unsigned>
1383  getDecomposedExpansionLocSlowCase(const SrcMgr::SLocEntry *E) const;
1384  std::pair<FileID, unsigned>
1385  getDecomposedSpellingLocSlowCase(const SrcMgr::SLocEntry *E,
1386                                   unsigned Offset) const;
1387  void computeMacroArgsCache(MacroArgsMap *&MacroArgsCache, FileID FID) const;
1388
1389  friend class ASTReader;
1390  friend class ASTWriter;
1391};
1392
1393
1394}  // end namespace clang
1395
1396#endif
1397