1//===- SourceManager.h - Track and cache source files -----------*- C++ -*-===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8//
9/// \file
10/// Defines the SourceManager interface.
11///
12/// There are three different types of locations in a %file: a spelling
13/// location, an expansion location, and a presumed location.
14///
15/// Given an example of:
16/// \code
17/// #define min(x, y) x < y ? x : y
18/// \endcode
19///
20/// and then later on a use of min:
21/// \code
22/// #line 17
23/// return min(a, b);
24/// \endcode
25///
26/// The expansion location is the line in the source code where the macro
27/// was expanded (the return statement), the spelling location is the
28/// location in the source where the macro was originally defined,
29/// and the presumed location is where the line directive states that
30/// the line is 17, or any other line.
31//
32//===----------------------------------------------------------------------===//
33
34#ifndef LLVM_CLANG_BASIC_SOURCEMANAGER_H
35#define LLVM_CLANG_BASIC_SOURCEMANAGER_H
36
37#include "clang/Basic/Diagnostic.h"
38#include "clang/Basic/FileEntry.h"
39#include "clang/Basic/FileManager.h"
40#include "clang/Basic/SourceLocation.h"
41#include "llvm/ADT/ArrayRef.h"
42#include "llvm/ADT/BitVector.h"
43#include "llvm/ADT/DenseMap.h"
44#include "llvm/ADT/DenseSet.h"
45#include "llvm/ADT/IntrusiveRefCntPtr.h"
46#include "llvm/ADT/PagedVector.h"
47#include "llvm/ADT/PointerIntPair.h"
48#include "llvm/ADT/SmallVector.h"
49#include "llvm/ADT/StringRef.h"
50#include "llvm/Support/Allocator.h"
51#include "llvm/Support/Compiler.h"
52#include "llvm/Support/MemoryBuffer.h"
53#include <cassert>
54#include <cstddef>
55#include <map>
56#include <memory>
57#include <optional>
58#include <string>
59#include <utility>
60#include <vector>
61
62namespace clang {
63
64class ASTReader;
65class ASTWriter;
66class FileManager;
67class LineTableInfo;
68class SourceManager;
69
70/// Public enums and private classes that are part of the
71/// SourceManager implementation.
72namespace SrcMgr {
73
74/// Indicates whether a file or directory holds normal user code,
75/// system code, or system code which is implicitly 'extern "C"' in C++ mode.
76///
77/// Entire directories can be tagged with this (this is maintained by
78/// DirectoryLookup and friends) as can specific FileInfos when a \#pragma
79/// system_header is seen or in various other cases.
80///
81enum CharacteristicKind {
82  C_User,
83  C_System,
84  C_ExternCSystem,
85  C_User_ModuleMap,
86  C_System_ModuleMap
87};
88
89/// Determine whether a file / directory characteristic is for system code.
90inline bool isSystem(CharacteristicKind CK) {
91  return CK != C_User && CK != C_User_ModuleMap;
92}
93
94/// Determine whether a file characteristic is for a module map.
95inline bool isModuleMap(CharacteristicKind CK) {
96  return CK == C_User_ModuleMap || CK == C_System_ModuleMap;
97}
98
99/// Mapping of line offsets into a source file. This does not own the storage
100/// for the line numbers.
101class LineOffsetMapping {
102public:
103  explicit operator bool() const { return Storage; }
104  unsigned size() const {
105    assert(Storage);
106    return Storage[0];
107  }
108  ArrayRef<unsigned> getLines() const {
109    assert(Storage);
110    return ArrayRef<unsigned>(Storage + 1, Storage + 1 + size());
111  }
112  const unsigned *begin() const { return getLines().begin(); }
113  const unsigned *end() const { return getLines().end(); }
114  const unsigned &operator[](int I) const { return getLines()[I]; }
115
116  static LineOffsetMapping get(llvm::MemoryBufferRef Buffer,
117                               llvm::BumpPtrAllocator &Alloc);
118
119  LineOffsetMapping() = default;
120  LineOffsetMapping(ArrayRef<unsigned> LineOffsets,
121                    llvm::BumpPtrAllocator &Alloc);
122
123private:
124  /// First element is the size, followed by elements at off-by-one indexes.
125  unsigned *Storage = nullptr;
126};
127
128/// One instance of this struct is kept for every file loaded or used.
129///
130/// This object owns the MemoryBuffer object.
131class alignas(8) ContentCache {
132  /// The actual buffer containing the characters from the input
133  /// file.
134  mutable std::unique_ptr<llvm::MemoryBuffer> Buffer;
135
136public:
137  /// Reference to the file entry representing this ContentCache.
138  ///
139  /// This reference does not own the FileEntry object.
140  ///
141  /// It is possible for this to be NULL if the ContentCache encapsulates
142  /// an imaginary text buffer.
143  ///
144  /// FIXME: Make non-optional using a virtual file as needed, remove \c
145  /// Filename and use \c OrigEntry.getNameAsRequested() instead.
146  OptionalFileEntryRef OrigEntry;
147
148  /// References the file which the contents were actually loaded from.
149  ///
150  /// Can be different from 'Entry' if we overridden the contents of one file
151  /// with the contents of another file.
152  OptionalFileEntryRef ContentsEntry;
153
154  /// The filename that is used to access OrigEntry.
155  ///
156  /// FIXME: Remove this once OrigEntry is a FileEntryRef with a stable name.
157  StringRef Filename;
158
159  /// A bump pointer allocated array of offsets for each source line.
160  ///
161  /// This is lazily computed.  The lines are owned by the SourceManager
162  /// BumpPointerAllocator object.
163  mutable LineOffsetMapping SourceLineCache;
164
165  /// Indicates whether the buffer itself was provided to override
166  /// the actual file contents.
167  ///
168  /// When true, the original entry may be a virtual file that does not
169  /// exist.
170  LLVM_PREFERRED_TYPE(bool)
171  unsigned BufferOverridden : 1;
172
173  /// True if this content cache was initially created for a source file
174  /// considered to be volatile (likely to change between stat and open).
175  LLVM_PREFERRED_TYPE(bool)
176  unsigned IsFileVolatile : 1;
177
178  /// True if this file may be transient, that is, if it might not
179  /// exist at some later point in time when this content entry is used,
180  /// after serialization and deserialization.
181  LLVM_PREFERRED_TYPE(bool)
182  unsigned IsTransient : 1;
183
184  LLVM_PREFERRED_TYPE(bool)
185  mutable unsigned IsBufferInvalid : 1;
186
187  ContentCache()
188      : OrigEntry(std::nullopt), ContentsEntry(std::nullopt),
189        BufferOverridden(false), IsFileVolatile(false), IsTransient(false),
190        IsBufferInvalid(false) {}
191
192  ContentCache(FileEntryRef Ent) : ContentCache(Ent, Ent) {}
193
194  ContentCache(FileEntryRef Ent, FileEntryRef contentEnt)
195      : OrigEntry(Ent), ContentsEntry(contentEnt), BufferOverridden(false),
196        IsFileVolatile(false), IsTransient(false), IsBufferInvalid(false) {}
197
198  /// The copy ctor does not allow copies where source object has either
199  /// a non-NULL Buffer or SourceLineCache.  Ownership of allocated memory
200  /// is not transferred, so this is a logical error.
201  ContentCache(const ContentCache &RHS)
202      : BufferOverridden(false), IsFileVolatile(false), IsTransient(false),
203        IsBufferInvalid(false) {
204    OrigEntry = RHS.OrigEntry;
205    ContentsEntry = RHS.ContentsEntry;
206
207    assert(!RHS.Buffer && !RHS.SourceLineCache &&
208           "Passed ContentCache object cannot own a buffer.");
209  }
210
211  ContentCache &operator=(const ContentCache &RHS) = delete;
212
213  /// Returns the memory buffer for the associated content.
214  ///
215  /// \param Diag Object through which diagnostics will be emitted if the
216  ///   buffer cannot be retrieved.
217  ///
218  /// \param Loc If specified, is the location that invalid file diagnostics
219  ///   will be emitted at.
220  std::optional<llvm::MemoryBufferRef>
221  getBufferOrNone(DiagnosticsEngine &Diag, FileManager &FM,
222                  SourceLocation Loc = SourceLocation()) const;
223
224  /// Returns the size of the content encapsulated by this
225  /// ContentCache.
226  ///
227  /// This can be the size of the source file or the size of an
228  /// arbitrary scratch buffer.  If the ContentCache encapsulates a source
229  /// file this size is retrieved from the file's FileEntry.
230  unsigned getSize() const;
231
232  /// Returns the number of bytes actually mapped for this
233  /// ContentCache.
234  ///
235  /// This can be 0 if the MemBuffer was not actually expanded.
236  unsigned getSizeBytesMapped() const;
237
238  /// Returns the kind of memory used to back the memory buffer for
239  /// this content cache.  This is used for performance analysis.
240  llvm::MemoryBuffer::BufferKind getMemoryBufferKind() const;
241
242  /// Return the buffer, only if it has been loaded.
243  std::optional<llvm::MemoryBufferRef> getBufferIfLoaded() const {
244    if (Buffer)
245      return Buffer->getMemBufferRef();
246    return std::nullopt;
247  }
248
249  /// Return a StringRef to the source buffer data, only if it has already
250  /// been loaded.
251  std::optional<StringRef> getBufferDataIfLoaded() const {
252    if (Buffer)
253      return Buffer->getBuffer();
254    return std::nullopt;
255  }
256
257  /// Set the buffer.
258  void setBuffer(std::unique_ptr<llvm::MemoryBuffer> B) {
259    IsBufferInvalid = false;
260    Buffer = std::move(B);
261  }
262
263  /// Set the buffer to one that's not owned (or to nullptr).
264  ///
265  /// \pre Buffer cannot already be set.
266  void setUnownedBuffer(std::optional<llvm::MemoryBufferRef> B) {
267    assert(!Buffer && "Expected to be called right after construction");
268    if (B)
269      setBuffer(llvm::MemoryBuffer::getMemBuffer(*B));
270  }
271
272  // If BufStr has an invalid BOM, returns the BOM name; otherwise, returns
273  // nullptr
274  static const char *getInvalidBOM(StringRef BufStr);
275};
276
277// Assert that the \c ContentCache objects will always be 8-byte aligned so
278// that we can pack 3 bits of integer into pointers to such objects.
279static_assert(alignof(ContentCache) >= 8,
280              "ContentCache must be 8-byte aligned.");
281
282/// Information about a FileID, basically just the logical file
283/// that it represents and include stack information.
284///
285/// Each FileInfo has include stack information, indicating where it came
286/// from. This information encodes the \#include chain that a token was
287/// expanded from. The main include file has an invalid IncludeLoc.
288///
289/// FileInfo should not grow larger than ExpansionInfo. Doing so will
290/// cause memory to bloat in compilations with many unloaded macro
291/// expansions, since the two data structurs are stored in a union in
292/// SLocEntry. Extra fields should instead go in "ContentCache *", which
293/// stores file contents and other bits on the side.
294///
295class FileInfo {
296  friend class clang::SourceManager;
297  friend class clang::ASTWriter;
298  friend class clang::ASTReader;
299
300  /// The location of the \#include that brought in this file.
301  ///
302  /// This is an invalid SLOC for the main file (top of the \#include chain).
303  SourceLocation IncludeLoc;
304
305  /// Number of FileIDs (files and macros) that were created during
306  /// preprocessing of this \#include, including this SLocEntry.
307  ///
308  /// Zero means the preprocessor didn't provide such info for this SLocEntry.
309  unsigned NumCreatedFIDs : 31;
310
311  /// Whether this FileInfo has any \#line directives.
312  LLVM_PREFERRED_TYPE(bool)
313  unsigned HasLineDirectives : 1;
314
315  /// The content cache and the characteristic of the file.
316  llvm::PointerIntPair<const ContentCache *, 3, CharacteristicKind>
317      ContentAndKind;
318
319public:
320  /// Return a FileInfo object.
321  static FileInfo get(SourceLocation IL, ContentCache &Con,
322                      CharacteristicKind FileCharacter, StringRef Filename) {
323    FileInfo X;
324    X.IncludeLoc = IL;
325    X.NumCreatedFIDs = 0;
326    X.HasLineDirectives = false;
327    X.ContentAndKind.setPointer(&Con);
328    X.ContentAndKind.setInt(FileCharacter);
329    Con.Filename = Filename;
330    return X;
331  }
332
333  SourceLocation getIncludeLoc() const {
334    return IncludeLoc;
335  }
336
337  const ContentCache &getContentCache() const {
338    return *ContentAndKind.getPointer();
339  }
340
341  /// Return whether this is a system header or not.
342  CharacteristicKind getFileCharacteristic() const {
343    return ContentAndKind.getInt();
344  }
345
346  /// Return true if this FileID has \#line directives in it.
347  bool hasLineDirectives() const { return HasLineDirectives; }
348
349  /// Set the flag that indicates that this FileID has
350  /// line table entries associated with it.
351  void setHasLineDirectives() { HasLineDirectives = true; }
352
353  /// Returns the name of the file that was used when the file was loaded from
354  /// the underlying file system.
355  StringRef getName() const { return getContentCache().Filename; }
356};
357
358/// Each ExpansionInfo encodes the expansion location - where
359/// the token was ultimately expanded, and the SpellingLoc - where the actual
360/// character data for the token came from.
361class ExpansionInfo {
362  // Really these are all SourceLocations.
363
364  /// Where the spelling for the token can be found.
365  SourceLocation SpellingLoc;
366
367  /// In a macro expansion, ExpansionLocStart and ExpansionLocEnd
368  /// indicate the start and end of the expansion. In object-like macros,
369  /// they will be the same. In a function-like macro expansion, the start
370  /// will be the identifier and the end will be the ')'. Finally, in
371  /// macro-argument instantiations, the end will be 'SourceLocation()', an
372  /// invalid location.
373  SourceLocation ExpansionLocStart, ExpansionLocEnd;
374
375  /// Whether the expansion range is a token range.
376  bool ExpansionIsTokenRange;
377
378public:
379  SourceLocation getSpellingLoc() const {
380    return SpellingLoc.isInvalid() ? getExpansionLocStart() : SpellingLoc;
381  }
382
383  SourceLocation getExpansionLocStart() const {
384    return ExpansionLocStart;
385  }
386
387  SourceLocation getExpansionLocEnd() const {
388    return ExpansionLocEnd.isInvalid() ? getExpansionLocStart()
389                                       : ExpansionLocEnd;
390  }
391
392  bool isExpansionTokenRange() const { return ExpansionIsTokenRange; }
393
394  CharSourceRange getExpansionLocRange() const {
395    return CharSourceRange(
396        SourceRange(getExpansionLocStart(), getExpansionLocEnd()),
397        isExpansionTokenRange());
398  }
399
400  bool isMacroArgExpansion() const {
401    // Note that this needs to return false for default constructed objects.
402    return getExpansionLocStart().isValid() && ExpansionLocEnd.isInvalid();
403  }
404
405  bool isMacroBodyExpansion() const {
406    return getExpansionLocStart().isValid() && ExpansionLocEnd.isValid();
407  }
408
409  bool isFunctionMacroExpansion() const {
410    return getExpansionLocStart().isValid() &&
411           getExpansionLocStart() != getExpansionLocEnd();
412  }
413
414  /// Return a ExpansionInfo for an expansion.
415  ///
416  /// Start and End specify the expansion range (where the macro is
417  /// expanded), and SpellingLoc specifies the spelling location (where
418  /// the characters from the token come from). All three can refer to
419  /// normal File SLocs or expansion locations.
420  static ExpansionInfo create(SourceLocation SpellingLoc, SourceLocation Start,
421                              SourceLocation End,
422                              bool ExpansionIsTokenRange = true) {
423    ExpansionInfo X;
424    X.SpellingLoc = SpellingLoc;
425    X.ExpansionLocStart = Start;
426    X.ExpansionLocEnd = End;
427    X.ExpansionIsTokenRange = ExpansionIsTokenRange;
428    return X;
429  }
430
431  /// Return a special ExpansionInfo for the expansion of
432  /// a macro argument into a function-like macro's body.
433  ///
434  /// ExpansionLoc specifies the expansion location (where the macro is
435  /// expanded). This doesn't need to be a range because a macro is always
436  /// expanded at a macro parameter reference, and macro parameters are
437  /// always exactly one token. SpellingLoc specifies the spelling location
438  /// (where the characters from the token come from). ExpansionLoc and
439  /// SpellingLoc can both refer to normal File SLocs or expansion locations.
440  ///
441  /// Given the code:
442  /// \code
443  ///   #define F(x) f(x)
444  ///   F(42);
445  /// \endcode
446  ///
447  /// When expanding '\c F(42)', the '\c x' would call this with an
448  /// SpellingLoc pointing at '\c 42' and an ExpansionLoc pointing at its
449  /// location in the definition of '\c F'.
450  static ExpansionInfo createForMacroArg(SourceLocation SpellingLoc,
451                                         SourceLocation ExpansionLoc) {
452    // We store an intentionally invalid source location for the end of the
453    // expansion range to mark that this is a macro argument location rather
454    // than a normal one.
455    return create(SpellingLoc, ExpansionLoc, SourceLocation());
456  }
457
458  /// Return a special ExpansionInfo representing a token that ends
459  /// prematurely. This is used to model a '>>' token that has been split
460  /// into '>' tokens and similar cases. Unlike for the other forms of
461  /// expansion, the expansion range in this case is a character range, not
462  /// a token range.
463  static ExpansionInfo createForTokenSplit(SourceLocation SpellingLoc,
464                                           SourceLocation Start,
465                                           SourceLocation End) {
466    return create(SpellingLoc, Start, End, false);
467  }
468};
469
470// Assert that the \c FileInfo objects are no bigger than \c ExpansionInfo
471// objects. This controls the size of \c SLocEntry, of which we have one for
472// each macro expansion. The number of (unloaded) macro expansions can be
473// very large. Any other fields needed in FileInfo should go in ContentCache.
474static_assert(sizeof(FileInfo) <= sizeof(ExpansionInfo),
475              "FileInfo must be no larger than ExpansionInfo.");
476
477/// This is a discriminated union of FileInfo and ExpansionInfo.
478///
479/// SourceManager keeps an array of these objects, and they are uniquely
480/// identified by the FileID datatype.
481class SLocEntry {
482  static constexpr int OffsetBits = 8 * sizeof(SourceLocation::UIntTy) - 1;
483  SourceLocation::UIntTy Offset : OffsetBits;
484  LLVM_PREFERRED_TYPE(bool)
485  SourceLocation::UIntTy IsExpansion : 1;
486  union {
487    FileInfo File;
488    ExpansionInfo Expansion;
489  };
490
491public:
492  SLocEntry() : Offset(), IsExpansion(), File() {}
493
494  SourceLocation::UIntTy getOffset() const { return Offset; }
495
496  bool isExpansion() const { return IsExpansion; }
497  bool isFile() const { return !isExpansion(); }
498
499  const FileInfo &getFile() const {
500    assert(isFile() && "Not a file SLocEntry!");
501    return File;
502  }
503
504  const ExpansionInfo &getExpansion() const {
505    assert(isExpansion() && "Not a macro expansion SLocEntry!");
506    return Expansion;
507  }
508
509  /// Creates an incomplete SLocEntry that is only able to report its offset.
510  static SLocEntry getOffsetOnly(SourceLocation::UIntTy Offset) {
511    assert(!(Offset & (1ULL << OffsetBits)) && "Offset is too large");
512    SLocEntry E;
513    E.Offset = Offset;
514    return E;
515  }
516
517  static SLocEntry get(SourceLocation::UIntTy Offset, const FileInfo &FI) {
518    assert(!(Offset & (1ULL << OffsetBits)) && "Offset is too large");
519    SLocEntry E;
520    E.Offset = Offset;
521    E.IsExpansion = false;
522    E.File = FI;
523    return E;
524  }
525
526  static SLocEntry get(SourceLocation::UIntTy Offset,
527                       const ExpansionInfo &Expansion) {
528    assert(!(Offset & (1ULL << OffsetBits)) && "Offset is too large");
529    SLocEntry E;
530    E.Offset = Offset;
531    E.IsExpansion = true;
532    new (&E.Expansion) ExpansionInfo(Expansion);
533    return E;
534  }
535};
536
537} // namespace SrcMgr
538
539/// External source of source location entries.
540class ExternalSLocEntrySource {
541public:
542  virtual ~ExternalSLocEntrySource();
543
544  /// Read the source location entry with index ID, which will always be
545  /// less than -1.
546  ///
547  /// \returns true if an error occurred that prevented the source-location
548  /// entry from being loaded.
549  virtual bool ReadSLocEntry(int ID) = 0;
550
551  /// Get the index ID for the loaded SourceLocation offset.
552  ///
553  /// \returns Invalid index ID (0) if an error occurred that prevented the
554  /// SLocEntry  from being loaded.
555  virtual int getSLocEntryID(SourceLocation::UIntTy SLocOffset) = 0;
556
557  /// Retrieve the module import location and name for the given ID, if
558  /// in fact it was loaded from a module (rather than, say, a precompiled
559  /// header).
560  virtual std::pair<SourceLocation, StringRef> getModuleImportLoc(int ID) = 0;
561};
562
563/// Holds the cache used by isBeforeInTranslationUnit.
564///
565/// The cache structure is complex enough to be worth breaking out of
566/// SourceManager.
567class InBeforeInTUCacheEntry {
568  /// The FileID's of the cached query.
569  ///
570  /// If these match up with a subsequent query, the result can be reused.
571  FileID LQueryFID, RQueryFID;
572
573  /// The relative order of FileIDs that the CommonFID *immediately* includes.
574  ///
575  /// This is used to compare macro expansion locations.
576  bool LChildBeforeRChild;
577
578  /// The file found in common between the two \#include traces, i.e.,
579  /// the nearest common ancestor of the \#include tree.
580  FileID CommonFID;
581
582  /// The offset of the previous query in CommonFID.
583  ///
584  /// Usually, this represents the location of the \#include for QueryFID, but
585  /// if LQueryFID is a parent of RQueryFID (or vice versa) then these can be a
586  /// random token in the parent.
587  unsigned LCommonOffset, RCommonOffset;
588
589public:
590  InBeforeInTUCacheEntry() = default;
591  InBeforeInTUCacheEntry(FileID L, FileID R) : LQueryFID(L), RQueryFID(R) {
592    assert(L != R);
593  }
594
595  /// Return true if the currently cached values match up with
596  /// the specified LHS/RHS query.
597  ///
598  /// If not, we can't use the cache.
599  bool isCacheValid() const {
600    return CommonFID.isValid();
601  }
602
603  /// If the cache is valid, compute the result given the
604  /// specified offsets in the LHS/RHS FileID's.
605  bool getCachedResult(unsigned LOffset, unsigned ROffset) const {
606    // If one of the query files is the common file, use the offset.  Otherwise,
607    // use the #include loc in the common file.
608    if (LQueryFID != CommonFID) LOffset = LCommonOffset;
609    if (RQueryFID != CommonFID) ROffset = RCommonOffset;
610
611    // It is common for multiple macro expansions to be "included" from the same
612    // location (expansion location), in which case use the order of the FileIDs
613    // to determine which came first. This will also take care the case where
614    // one of the locations points at the inclusion/expansion point of the other
615    // in which case its FileID will come before the other.
616    if (LOffset == ROffset)
617      return LChildBeforeRChild;
618
619    return LOffset < ROffset;
620  }
621
622  /// Set up a new query.
623  /// If it matches the old query, we can keep the cached answer.
624  void setQueryFIDs(FileID LHS, FileID RHS) {
625    assert(LHS != RHS);
626    if (LQueryFID != LHS || RQueryFID != RHS) {
627      LQueryFID = LHS;
628      RQueryFID = RHS;
629      CommonFID = FileID();
630    }
631  }
632
633  void setCommonLoc(FileID commonFID, unsigned lCommonOffset,
634                    unsigned rCommonOffset, bool LParentBeforeRParent) {
635    CommonFID = commonFID;
636    LCommonOffset = lCommonOffset;
637    RCommonOffset = rCommonOffset;
638    LChildBeforeRChild = LParentBeforeRParent;
639  }
640};
641
642/// The stack used when building modules on demand, which is used
643/// to provide a link between the source managers of the different compiler
644/// instances.
645using ModuleBuildStack = ArrayRef<std::pair<std::string, FullSourceLoc>>;
646
647/// This class handles loading and caching of source files into memory.
648///
649/// This object owns the MemoryBuffer objects for all of the loaded
650/// files and assigns unique FileID's for each unique \#include chain.
651///
652/// The SourceManager can be queried for information about SourceLocation
653/// objects, turning them into either spelling or expansion locations. Spelling
654/// locations represent where the bytes corresponding to a token came from and
655/// expansion locations represent where the location is in the user's view. In
656/// the case of a macro expansion, for example, the spelling location indicates
657/// where the expanded token came from and the expansion location specifies
658/// where it was expanded.
659class SourceManager : public RefCountedBase<SourceManager> {
660  /// DiagnosticsEngine object.
661  DiagnosticsEngine &Diag;
662
663  FileManager &FileMgr;
664
665  mutable llvm::BumpPtrAllocator ContentCacheAlloc;
666
667  /// Memoized information about all of the files tracked by this
668  /// SourceManager.
669  ///
670  /// This map allows us to merge ContentCache entries based
671  /// on their FileEntry*.  All ContentCache objects will thus have unique,
672  /// non-null, FileEntry pointers.
673  llvm::DenseMap<FileEntryRef, SrcMgr::ContentCache*> FileInfos;
674
675  /// True if the ContentCache for files that are overridden by other
676  /// files, should report the original file name. Defaults to true.
677  bool OverridenFilesKeepOriginalName = true;
678
679  /// True if non-system source files should be treated as volatile
680  /// (likely to change while trying to use them). Defaults to false.
681  bool UserFilesAreVolatile;
682
683  /// True if all files read during this compilation should be treated
684  /// as transient (may not be present in later compilations using a module
685  /// file created from this compilation). Defaults to false.
686  bool FilesAreTransient = false;
687
688  struct OverriddenFilesInfoTy {
689    /// Files that have been overridden with the contents from another
690    /// file.
691    llvm::DenseMap<const FileEntry *, FileEntryRef> OverriddenFiles;
692
693    /// Files that were overridden with a memory buffer.
694    llvm::DenseSet<const FileEntry *> OverriddenFilesWithBuffer;
695  };
696
697  /// Lazily create the object keeping overridden files info, since
698  /// it is uncommonly used.
699  std::unique_ptr<OverriddenFilesInfoTy> OverriddenFilesInfo;
700
701  OverriddenFilesInfoTy &getOverriddenFilesInfo() {
702    if (!OverriddenFilesInfo)
703      OverriddenFilesInfo.reset(new OverriddenFilesInfoTy);
704    return *OverriddenFilesInfo;
705  }
706
707  /// Information about various memory buffers that we have read in.
708  ///
709  /// All FileEntry* within the stored ContentCache objects are NULL,
710  /// as they do not refer to a file.
711  std::vector<SrcMgr::ContentCache*> MemBufferInfos;
712
713  /// The table of SLocEntries that are local to this module.
714  ///
715  /// Positive FileIDs are indexes into this table. Entry 0 indicates an invalid
716  /// expansion.
717  SmallVector<SrcMgr::SLocEntry, 0> LocalSLocEntryTable;
718
719  /// The table of SLocEntries that are loaded from other modules.
720  ///
721  /// Negative FileIDs are indexes into this table. To get from ID to an index,
722  /// use (-ID - 2).
723  llvm::PagedVector<SrcMgr::SLocEntry> LoadedSLocEntryTable;
724
725  /// For each allocation in LoadedSLocEntryTable, we keep the first FileID.
726  /// We assume exactly one allocation per AST file, and use that to determine
727  /// whether two FileIDs come from the same AST file.
728  SmallVector<FileID, 0> LoadedSLocEntryAllocBegin;
729
730  /// The starting offset of the next local SLocEntry.
731  ///
732  /// This is LocalSLocEntryTable.back().Offset + the size of that entry.
733  SourceLocation::UIntTy NextLocalOffset;
734
735  /// The starting offset of the latest batch of loaded SLocEntries.
736  ///
737  /// This is LoadedSLocEntryTable.back().Offset, except that that entry might
738  /// not have been loaded, so that value would be unknown.
739  SourceLocation::UIntTy CurrentLoadedOffset;
740
741  /// The highest possible offset is 2^31-1 (2^63-1 for 64-bit source
742  /// locations), so CurrentLoadedOffset starts at 2^31 (2^63 resp.).
743  static const SourceLocation::UIntTy MaxLoadedOffset =
744      1ULL << (8 * sizeof(SourceLocation::UIntTy) - 1);
745
746  /// A bitmap that indicates whether the entries of LoadedSLocEntryTable
747  /// have already been loaded from the external source.
748  ///
749  /// Same indexing as LoadedSLocEntryTable.
750  llvm::BitVector SLocEntryLoaded;
751
752  /// A bitmap that indicates whether the entries of LoadedSLocEntryTable
753  /// have already had their offset loaded from the external source.
754  ///
755  /// Superset of SLocEntryLoaded. Same indexing as SLocEntryLoaded.
756  llvm::BitVector SLocEntryOffsetLoaded;
757
758  /// An external source for source location entries.
759  ExternalSLocEntrySource *ExternalSLocEntries = nullptr;
760
761  /// A one-entry cache to speed up getFileID.
762  ///
763  /// LastFileIDLookup records the last FileID looked up or created, because it
764  /// is very common to look up many tokens from the same file.
765  mutable FileID LastFileIDLookup;
766
767  /// Holds information for \#line directives.
768  ///
769  /// This is referenced by indices from SLocEntryTable.
770  std::unique_ptr<LineTableInfo> LineTable;
771
772  /// These ivars serve as a cache used in the getLineNumber
773  /// method which is used to speedup getLineNumber calls to nearby locations.
774  mutable FileID LastLineNoFileIDQuery;
775  mutable const SrcMgr::ContentCache *LastLineNoContentCache;
776  mutable unsigned LastLineNoFilePos;
777  mutable unsigned LastLineNoResult;
778
779  /// The file ID for the main source file of the translation unit.
780  FileID MainFileID;
781
782  /// The file ID for the precompiled preamble there is one.
783  FileID PreambleFileID;
784
785  // Statistics for -print-stats.
786  mutable unsigned NumLinearScans = 0;
787  mutable unsigned NumBinaryProbes = 0;
788
789  /// Associates a FileID with its "included/expanded in" decomposed
790  /// location.
791  ///
792  /// Used to cache results from and speed-up \c getDecomposedIncludedLoc
793  /// function.
794  mutable llvm::DenseMap<FileID, std::pair<FileID, unsigned>> IncludedLocMap;
795
796  /// The key value into the IsBeforeInTUCache table.
797  using IsBeforeInTUCacheKey = std::pair<FileID, FileID>;
798
799  /// The IsBeforeInTranslationUnitCache is a mapping from FileID pairs
800  /// to cache results.
801  using InBeforeInTUCache =
802      llvm::DenseMap<IsBeforeInTUCacheKey, InBeforeInTUCacheEntry>;
803
804  /// Cache results for the isBeforeInTranslationUnit method.
805  mutable InBeforeInTUCache IBTUCache;
806  mutable InBeforeInTUCacheEntry IBTUCacheOverflow;
807
808  /// Return the cache entry for comparing the given file IDs
809  /// for isBeforeInTranslationUnit.
810  InBeforeInTUCacheEntry &getInBeforeInTUCache(FileID LFID, FileID RFID) const;
811
812  // Cache for the "fake" buffer used for error-recovery purposes.
813  mutable std::unique_ptr<llvm::MemoryBuffer> FakeBufferForRecovery;
814
815  mutable std::unique_ptr<SrcMgr::ContentCache> FakeContentCacheForRecovery;
816
817  mutable std::unique_ptr<SrcMgr::SLocEntry> FakeSLocEntryForRecovery;
818
819  /// Lazily computed map of macro argument chunks to their expanded
820  /// source location.
821  using MacroArgsMap = std::map<unsigned, SourceLocation>;
822
823  mutable llvm::DenseMap<FileID, std::unique_ptr<MacroArgsMap>>
824      MacroArgsCacheMap;
825
826  /// The stack of modules being built, which is used to detect
827  /// cycles in the module dependency graph as modules are being built, as
828  /// well as to describe why we're rebuilding a particular module.
829  ///
830  /// There is no way to set this value from the command line. If we ever need
831  /// to do so (e.g., if on-demand module construction moves out-of-process),
832  /// we can add a cc1-level option to do so.
833  SmallVector<std::pair<std::string, FullSourceLoc>, 2> StoredModuleBuildStack;
834
835public:
836  SourceManager(DiagnosticsEngine &Diag, FileManager &FileMgr,
837                bool UserFilesAreVolatile = false);
838  explicit SourceManager(const SourceManager &) = delete;
839  SourceManager &operator=(const SourceManager &) = delete;
840  ~SourceManager();
841
842  void clearIDTables();
843
844  /// Initialize this source manager suitably to replay the compilation
845  /// described by \p Old. Requires that \p Old outlive \p *this.
846  void initializeForReplay(const SourceManager &Old);
847
848  DiagnosticsEngine &getDiagnostics() const { return Diag; }
849
850  FileManager &getFileManager() const { return FileMgr; }
851
852  /// Set true if the SourceManager should report the original file name
853  /// for contents of files that were overridden by other files. Defaults to
854  /// true.
855  void setOverridenFilesKeepOriginalName(bool value) {
856    OverridenFilesKeepOriginalName = value;
857  }
858
859  /// True if non-system source files should be treated as volatile
860  /// (likely to change while trying to use them).
861  bool userFilesAreVolatile() const { return UserFilesAreVolatile; }
862
863  /// Retrieve the module build stack.
864  ModuleBuildStack getModuleBuildStack() const {
865    return StoredModuleBuildStack;
866  }
867
868  /// Set the module build stack.
869  void setModuleBuildStack(ModuleBuildStack stack) {
870    StoredModuleBuildStack.clear();
871    StoredModuleBuildStack.append(stack.begin(), stack.end());
872  }
873
874  /// Push an entry to the module build stack.
875  void pushModuleBuildStack(StringRef moduleName, FullSourceLoc importLoc) {
876    StoredModuleBuildStack.push_back(std::make_pair(moduleName.str(),importLoc));
877  }
878
879  //===--------------------------------------------------------------------===//
880  // MainFileID creation and querying methods.
881  //===--------------------------------------------------------------------===//
882
883  /// Returns the FileID of the main source file.
884  FileID getMainFileID() const { return MainFileID; }
885
886  /// Set the file ID for the main source file.
887  void setMainFileID(FileID FID) {
888    MainFileID = FID;
889  }
890
891  /// Returns true when the given FileEntry corresponds to the main file.
892  ///
893  /// The main file should be set prior to calling this function.
894  bool isMainFile(const FileEntry &SourceFile);
895
896  /// Set the file ID for the precompiled preamble.
897  void setPreambleFileID(FileID Preamble) {
898    assert(PreambleFileID.isInvalid() && "PreambleFileID already set!");
899    PreambleFileID = Preamble;
900  }
901
902  /// Get the file ID for the precompiled preamble if there is one.
903  FileID getPreambleFileID() const { return PreambleFileID; }
904
905  //===--------------------------------------------------------------------===//
906  // Methods to create new FileID's and macro expansions.
907  //===--------------------------------------------------------------------===//
908
909  /// Create a new FileID that represents the specified file
910  /// being \#included from the specified IncludePosition.
911  FileID createFileID(FileEntryRef SourceFile, SourceLocation IncludePos,
912                      SrcMgr::CharacteristicKind FileCharacter,
913                      int LoadedID = 0,
914                      SourceLocation::UIntTy LoadedOffset = 0);
915
916  /// Create a new FileID that represents the specified memory buffer.
917  ///
918  /// This does no caching of the buffer and takes ownership of the
919  /// MemoryBuffer, so only pass a MemoryBuffer to this once.
920  FileID createFileID(std::unique_ptr<llvm::MemoryBuffer> Buffer,
921                      SrcMgr::CharacteristicKind FileCharacter = SrcMgr::C_User,
922                      int LoadedID = 0, SourceLocation::UIntTy LoadedOffset = 0,
923                      SourceLocation IncludeLoc = SourceLocation());
924
925  /// Create a new FileID that represents the specified memory buffer.
926  ///
927  /// This does not take ownership of the MemoryBuffer. The memory buffer must
928  /// outlive the SourceManager.
929  FileID createFileID(const llvm::MemoryBufferRef &Buffer,
930                      SrcMgr::CharacteristicKind FileCharacter = SrcMgr::C_User,
931                      int LoadedID = 0, SourceLocation::UIntTy LoadedOffset = 0,
932                      SourceLocation IncludeLoc = SourceLocation());
933
934  /// Get the FileID for \p SourceFile if it exists. Otherwise, create a
935  /// new FileID for the \p SourceFile.
936  FileID getOrCreateFileID(FileEntryRef SourceFile,
937                           SrcMgr::CharacteristicKind FileCharacter);
938
939  /// Creates an expansion SLocEntry for the substitution of an argument into a
940  /// function-like macro's body. Returns the start of the expansion.
941  ///
942  /// The macro argument was written at \p SpellingLoc with length \p Length.
943  /// \p ExpansionLoc is the parameter name in the (expanded) macro body.
944  SourceLocation createMacroArgExpansionLoc(SourceLocation SpellingLoc,
945                                            SourceLocation ExpansionLoc,
946                                            unsigned Length);
947
948  /// Creates an expansion SLocEntry for a macro use. Returns its start.
949  ///
950  /// The macro body begins at \p SpellingLoc with length \p Length.
951  /// The macro use spans [ExpansionLocStart, ExpansionLocEnd].
952  SourceLocation createExpansionLoc(SourceLocation SpellingLoc,
953                                    SourceLocation ExpansionLocStart,
954                                    SourceLocation ExpansionLocEnd,
955                                    unsigned Length,
956                                    bool ExpansionIsTokenRange = true,
957                                    int LoadedID = 0,
958                                    SourceLocation::UIntTy LoadedOffset = 0);
959
960  /// Return a new SourceLocation that encodes that the token starting
961  /// at \p TokenStart ends prematurely at \p TokenEnd.
962  SourceLocation createTokenSplitLoc(SourceLocation SpellingLoc,
963                                     SourceLocation TokenStart,
964                                     SourceLocation TokenEnd);
965
966  /// Retrieve the memory buffer associated with the given file.
967  ///
968  /// Returns std::nullopt if the buffer is not valid.
969  std::optional<llvm::MemoryBufferRef>
970  getMemoryBufferForFileOrNone(FileEntryRef File);
971
972  /// Retrieve the memory buffer associated with the given file.
973  ///
974  /// Returns a fake buffer if there isn't a real one.
975  llvm::MemoryBufferRef getMemoryBufferForFileOrFake(FileEntryRef File) {
976    if (auto B = getMemoryBufferForFileOrNone(File))
977      return *B;
978    return getFakeBufferForRecovery();
979  }
980
981  /// Override the contents of the given source file by providing an
982  /// already-allocated buffer.
983  ///
984  /// \param SourceFile the source file whose contents will be overridden.
985  ///
986  /// \param Buffer the memory buffer whose contents will be used as the
987  /// data in the given source file.
988  void overrideFileContents(FileEntryRef SourceFile,
989                            const llvm::MemoryBufferRef &Buffer) {
990    overrideFileContents(SourceFile, llvm::MemoryBuffer::getMemBuffer(Buffer));
991  }
992
993  /// Override the contents of the given source file by providing an
994  /// already-allocated buffer.
995  ///
996  /// \param SourceFile the source file whose contents will be overridden.
997  ///
998  /// \param Buffer the memory buffer whose contents will be used as the
999  /// data in the given source file.
1000  void overrideFileContents(FileEntryRef SourceFile,
1001                            std::unique_ptr<llvm::MemoryBuffer> Buffer);
1002
1003  /// Override the given source file with another one.
1004  ///
1005  /// \param SourceFile the source file which will be overridden.
1006  ///
1007  /// \param NewFile the file whose contents will be used as the
1008  /// data instead of the contents of the given source file.
1009  void overrideFileContents(const FileEntry *SourceFile, FileEntryRef NewFile);
1010
1011  /// Returns true if the file contents have been overridden.
1012  bool isFileOverridden(const FileEntry *File) const {
1013    if (OverriddenFilesInfo) {
1014      if (OverriddenFilesInfo->OverriddenFilesWithBuffer.count(File))
1015        return true;
1016      if (OverriddenFilesInfo->OverriddenFiles.contains(File))
1017        return true;
1018    }
1019    return false;
1020  }
1021
1022  /// Bypass the overridden contents of a file.  This creates a new FileEntry
1023  /// and initializes the content cache for it.  Returns std::nullopt if there
1024  /// is no such file in the filesystem.
1025  ///
1026  /// This should be called before parsing has begun.
1027  OptionalFileEntryRef bypassFileContentsOverride(FileEntryRef File);
1028
1029  /// Specify that a file is transient.
1030  void setFileIsTransient(FileEntryRef SourceFile);
1031
1032  /// Specify that all files that are read during this compilation are
1033  /// transient.
1034  void setAllFilesAreTransient(bool Transient) {
1035    FilesAreTransient = Transient;
1036  }
1037
1038  //===--------------------------------------------------------------------===//
1039  // FileID manipulation methods.
1040  //===--------------------------------------------------------------------===//
1041
1042  /// Return the buffer for the specified FileID.
1043  ///
1044  /// If there is an error opening this buffer the first time, return
1045  /// std::nullopt.
1046  std::optional<llvm::MemoryBufferRef>
1047  getBufferOrNone(FileID FID, SourceLocation Loc = SourceLocation()) const {
1048    if (auto *Entry = getSLocEntryForFile(FID))
1049      return Entry->getFile().getContentCache().getBufferOrNone(
1050          Diag, getFileManager(), Loc);
1051    return std::nullopt;
1052  }
1053
1054  /// Return the buffer for the specified FileID.
1055  ///
1056  /// If there is an error opening this buffer the first time, this
1057  /// manufactures a temporary buffer and returns it.
1058  llvm::MemoryBufferRef
1059  getBufferOrFake(FileID FID, SourceLocation Loc = SourceLocation()) const {
1060    if (auto B = getBufferOrNone(FID, Loc))
1061      return *B;
1062    return getFakeBufferForRecovery();
1063  }
1064
1065  /// Returns the FileEntry record for the provided FileID.
1066  const FileEntry *getFileEntryForID(FileID FID) const {
1067    if (auto FE = getFileEntryRefForID(FID))
1068      return *FE;
1069    return nullptr;
1070  }
1071
1072  /// Returns the FileEntryRef for the provided FileID.
1073  OptionalFileEntryRef getFileEntryRefForID(FileID FID) const {
1074    if (auto *Entry = getSLocEntryForFile(FID))
1075      return Entry->getFile().getContentCache().OrigEntry;
1076    return std::nullopt;
1077  }
1078
1079  /// Returns the filename for the provided FileID, unless it's a built-in
1080  /// buffer that's not represented by a filename.
1081  ///
1082  /// Returns std::nullopt for non-files and built-in files.
1083  std::optional<StringRef> getNonBuiltinFilenameForID(FileID FID) const;
1084
1085  /// Returns the FileEntry record for the provided SLocEntry.
1086  const FileEntry *
1087  getFileEntryForSLocEntry(const SrcMgr::SLocEntry &SLocEntry) const {
1088    if (auto FE = SLocEntry.getFile().getContentCache().OrigEntry)
1089      return *FE;
1090    return nullptr;
1091  }
1092
1093  /// Return a StringRef to the source buffer data for the
1094  /// specified FileID.
1095  ///
1096  /// \param FID The file ID whose contents will be returned.
1097  /// \param Invalid If non-NULL, will be set true if an error occurred.
1098  StringRef getBufferData(FileID FID, bool *Invalid = nullptr) const;
1099
1100  /// Return a StringRef to the source buffer data for the
1101  /// specified FileID, returning std::nullopt if invalid.
1102  ///
1103  /// \param FID The file ID whose contents will be returned.
1104  std::optional<StringRef> getBufferDataOrNone(FileID FID) const;
1105
1106  /// Return a StringRef to the source buffer data for the
1107  /// specified FileID, returning std::nullopt if it's not yet loaded.
1108  ///
1109  /// \param FID The file ID whose contents will be returned.
1110  std::optional<StringRef> getBufferDataIfLoaded(FileID FID) const;
1111
1112  /// Get the number of FileIDs (files and macros) that were created
1113  /// during preprocessing of \p FID, including it.
1114  unsigned getNumCreatedFIDsForFileID(FileID FID) const {
1115    if (auto *Entry = getSLocEntryForFile(FID))
1116      return Entry->getFile().NumCreatedFIDs;
1117    return 0;
1118  }
1119
1120  /// Set the number of FileIDs (files and macros) that were created
1121  /// during preprocessing of \p FID, including it.
1122  void setNumCreatedFIDsForFileID(FileID FID, unsigned NumFIDs,
1123                                  bool Force = false) const {
1124    auto *Entry = getSLocEntryForFile(FID);
1125    if (!Entry)
1126      return;
1127    assert((Force || Entry->getFile().NumCreatedFIDs == 0) && "Already set!");
1128    const_cast<SrcMgr::FileInfo &>(Entry->getFile()).NumCreatedFIDs = NumFIDs;
1129  }
1130
1131  //===--------------------------------------------------------------------===//
1132  // SourceLocation manipulation methods.
1133  //===--------------------------------------------------------------------===//
1134
1135  /// Return the FileID for a SourceLocation.
1136  ///
1137  /// This is a very hot method that is used for all SourceManager queries
1138  /// that start with a SourceLocation object.  It is responsible for finding
1139  /// the entry in SLocEntryTable which contains the specified location.
1140  ///
1141  FileID getFileID(SourceLocation SpellingLoc) const {
1142    return getFileID(SpellingLoc.getOffset());
1143  }
1144
1145  /// Return the filename of the file containing a SourceLocation.
1146  StringRef getFilename(SourceLocation SpellingLoc) const;
1147
1148  /// Return the source location corresponding to the first byte of
1149  /// the specified file.
1150  SourceLocation getLocForStartOfFile(FileID FID) const {
1151    if (auto *Entry = getSLocEntryForFile(FID))
1152      return SourceLocation::getFileLoc(Entry->getOffset());
1153    return SourceLocation();
1154  }
1155
1156  /// Return the source location corresponding to the last byte of the
1157  /// specified file.
1158  SourceLocation getLocForEndOfFile(FileID FID) const {
1159    if (auto *Entry = getSLocEntryForFile(FID))
1160      return SourceLocation::getFileLoc(Entry->getOffset() +
1161                                        getFileIDSize(FID));
1162    return SourceLocation();
1163  }
1164
1165  /// Returns the include location if \p FID is a \#include'd file
1166  /// otherwise it returns an invalid location.
1167  SourceLocation getIncludeLoc(FileID FID) const {
1168    if (auto *Entry = getSLocEntryForFile(FID))
1169      return Entry->getFile().getIncludeLoc();
1170    return SourceLocation();
1171  }
1172
1173  // Returns the import location if the given source location is
1174  // located within a module, or an invalid location if the source location
1175  // is within the current translation unit.
1176  std::pair<SourceLocation, StringRef>
1177  getModuleImportLoc(SourceLocation Loc) const {
1178    FileID FID = getFileID(Loc);
1179
1180    // Positive file IDs are in the current translation unit, and -1 is a
1181    // placeholder.
1182    if (FID.ID >= -1)
1183      return std::make_pair(SourceLocation(), "");
1184
1185    return ExternalSLocEntries->getModuleImportLoc(FID.ID);
1186  }
1187
1188  /// Given a SourceLocation object \p Loc, return the expansion
1189  /// location referenced by the ID.
1190  SourceLocation getExpansionLoc(SourceLocation Loc) const {
1191    // Handle the non-mapped case inline, defer to out of line code to handle
1192    // expansions.
1193    if (Loc.isFileID()) return Loc;
1194    return getExpansionLocSlowCase(Loc);
1195  }
1196
1197  /// Given \p Loc, if it is a macro location return the expansion
1198  /// location or the spelling location, depending on if it comes from a
1199  /// macro argument or not.
1200  SourceLocation getFileLoc(SourceLocation Loc) const {
1201    if (Loc.isFileID()) return Loc;
1202    return getFileLocSlowCase(Loc);
1203  }
1204
1205  /// Return the start/end of the expansion information for an
1206  /// expansion location.
1207  ///
1208  /// \pre \p Loc is required to be an expansion location.
1209  CharSourceRange getImmediateExpansionRange(SourceLocation Loc) const;
1210
1211  /// Given a SourceLocation object, return the range of
1212  /// tokens covered by the expansion in the ultimate file.
1213  CharSourceRange getExpansionRange(SourceLocation Loc) const;
1214
1215  /// Given a SourceRange object, return the range of
1216  /// tokens or characters covered by the expansion in the ultimate file.
1217  CharSourceRange getExpansionRange(SourceRange Range) const {
1218    SourceLocation Begin = getExpansionRange(Range.getBegin()).getBegin();
1219    CharSourceRange End = getExpansionRange(Range.getEnd());
1220    return CharSourceRange(SourceRange(Begin, End.getEnd()),
1221                           End.isTokenRange());
1222  }
1223
1224  /// Given a CharSourceRange object, return the range of
1225  /// tokens or characters covered by the expansion in the ultimate file.
1226  CharSourceRange getExpansionRange(CharSourceRange Range) const {
1227    CharSourceRange Expansion = getExpansionRange(Range.getAsRange());
1228    if (Expansion.getEnd() == Range.getEnd())
1229      Expansion.setTokenRange(Range.isTokenRange());
1230    return Expansion;
1231  }
1232
1233  /// Given a SourceLocation object, return the spelling
1234  /// location referenced by the ID.
1235  ///
1236  /// This is the place where the characters that make up the lexed token
1237  /// can be found.
1238  SourceLocation getSpellingLoc(SourceLocation Loc) const {
1239    // Handle the non-mapped case inline, defer to out of line code to handle
1240    // expansions.
1241    if (Loc.isFileID()) return Loc;
1242    return getSpellingLocSlowCase(Loc);
1243  }
1244
1245  /// Given a SourceLocation object, return the spelling location
1246  /// referenced by the ID.
1247  ///
1248  /// This is the first level down towards the place where the characters
1249  /// that make up the lexed token can be found.  This should not generally
1250  /// be used by clients.
1251  SourceLocation getImmediateSpellingLoc(SourceLocation Loc) const;
1252
1253  /// Form a SourceLocation from a FileID and Offset pair.
1254  SourceLocation getComposedLoc(FileID FID, unsigned Offset) const {
1255    auto *Entry = getSLocEntryOrNull(FID);
1256    if (!Entry)
1257      return SourceLocation();
1258
1259    SourceLocation::UIntTy GlobalOffset = Entry->getOffset() + Offset;
1260    return Entry->isFile() ? SourceLocation::getFileLoc(GlobalOffset)
1261                           : SourceLocation::getMacroLoc(GlobalOffset);
1262  }
1263
1264  /// Decompose the specified location into a raw FileID + Offset pair.
1265  ///
1266  /// The first element is the FileID, the second is the offset from the
1267  /// start of the buffer of the location.
1268  std::pair<FileID, unsigned> getDecomposedLoc(SourceLocation Loc) const {
1269    FileID FID = getFileID(Loc);
1270    auto *Entry = getSLocEntryOrNull(FID);
1271    if (!Entry)
1272      return std::make_pair(FileID(), 0);
1273    return std::make_pair(FID, Loc.getOffset() - Entry->getOffset());
1274  }
1275
1276  /// Decompose the specified location into a raw FileID + Offset pair.
1277  ///
1278  /// If the location is an expansion record, walk through it until we find
1279  /// the final location expanded.
1280  std::pair<FileID, unsigned>
1281  getDecomposedExpansionLoc(SourceLocation Loc) const {
1282    FileID FID = getFileID(Loc);
1283    auto *E = getSLocEntryOrNull(FID);
1284    if (!E)
1285      return std::make_pair(FileID(), 0);
1286
1287    unsigned Offset = Loc.getOffset()-E->getOffset();
1288    if (Loc.isFileID())
1289      return std::make_pair(FID, Offset);
1290
1291    return getDecomposedExpansionLocSlowCase(E);
1292  }
1293
1294  /// Decompose the specified location into a raw FileID + Offset pair.
1295  ///
1296  /// If the location is an expansion record, walk through it until we find
1297  /// its spelling record.
1298  std::pair<FileID, unsigned>
1299  getDecomposedSpellingLoc(SourceLocation Loc) const {
1300    FileID FID = getFileID(Loc);
1301    auto *E = getSLocEntryOrNull(FID);
1302    if (!E)
1303      return std::make_pair(FileID(), 0);
1304
1305    unsigned Offset = Loc.getOffset()-E->getOffset();
1306    if (Loc.isFileID())
1307      return std::make_pair(FID, Offset);
1308    return getDecomposedSpellingLocSlowCase(E, Offset);
1309  }
1310
1311  /// Returns the "included/expanded in" decomposed location of the given
1312  /// FileID.
1313  std::pair<FileID, unsigned> getDecomposedIncludedLoc(FileID FID) const;
1314
1315  /// Returns the offset from the start of the file that the
1316  /// specified SourceLocation represents.
1317  ///
1318  /// This is not very meaningful for a macro ID.
1319  unsigned getFileOffset(SourceLocation SpellingLoc) const {
1320    return getDecomposedLoc(SpellingLoc).second;
1321  }
1322
1323  /// Tests whether the given source location represents a macro
1324  /// argument's expansion into the function-like macro definition.
1325  ///
1326  /// \param StartLoc If non-null and function returns true, it is set to the
1327  /// start location of the macro argument expansion.
1328  ///
1329  /// Such source locations only appear inside of the expansion
1330  /// locations representing where a particular function-like macro was
1331  /// expanded.
1332  bool isMacroArgExpansion(SourceLocation Loc,
1333                           SourceLocation *StartLoc = nullptr) const;
1334
1335  /// Tests whether the given source location represents the expansion of
1336  /// a macro body.
1337  ///
1338  /// This is equivalent to testing whether the location is part of a macro
1339  /// expansion but not the expansion of an argument to a function-like macro.
1340  bool isMacroBodyExpansion(SourceLocation Loc) const;
1341
1342  /// Returns true if the given MacroID location points at the beginning
1343  /// of the immediate macro expansion.
1344  ///
1345  /// \param MacroBegin If non-null and function returns true, it is set to the
1346  /// begin location of the immediate macro expansion.
1347  bool isAtStartOfImmediateMacroExpansion(SourceLocation Loc,
1348                                    SourceLocation *MacroBegin = nullptr) const;
1349
1350  /// Returns true if the given MacroID location points at the character
1351  /// end of the immediate macro expansion.
1352  ///
1353  /// \param MacroEnd If non-null and function returns true, it is set to the
1354  /// character end location of the immediate macro expansion.
1355  bool
1356  isAtEndOfImmediateMacroExpansion(SourceLocation Loc,
1357                                   SourceLocation *MacroEnd = nullptr) const;
1358
1359  /// Returns true if \p Loc is inside the [\p Start, +\p Length)
1360  /// chunk of the source location address space.
1361  ///
1362  /// If it's true and \p RelativeOffset is non-null, it will be set to the
1363  /// relative offset of \p Loc inside the chunk.
1364  bool
1365  isInSLocAddrSpace(SourceLocation Loc, SourceLocation Start, unsigned Length,
1366                    SourceLocation::UIntTy *RelativeOffset = nullptr) const {
1367    assert(((Start.getOffset() < NextLocalOffset &&
1368               Start.getOffset()+Length <= NextLocalOffset) ||
1369            (Start.getOffset() >= CurrentLoadedOffset &&
1370                Start.getOffset()+Length < MaxLoadedOffset)) &&
1371           "Chunk is not valid SLoc address space");
1372    SourceLocation::UIntTy LocOffs = Loc.getOffset();
1373    SourceLocation::UIntTy BeginOffs = Start.getOffset();
1374    SourceLocation::UIntTy EndOffs = BeginOffs + Length;
1375    if (LocOffs >= BeginOffs && LocOffs < EndOffs) {
1376      if (RelativeOffset)
1377        *RelativeOffset = LocOffs - BeginOffs;
1378      return true;
1379    }
1380
1381    return false;
1382  }
1383
1384  /// Return true if both \p LHS and \p RHS are in the local source
1385  /// location address space or the loaded one.
1386  ///
1387  /// If it's true and \p RelativeOffset is non-null, it will be set to the
1388  /// offset of \p RHS relative to \p LHS.
1389  bool isInSameSLocAddrSpace(SourceLocation LHS, SourceLocation RHS,
1390                             SourceLocation::IntTy *RelativeOffset) const {
1391    SourceLocation::UIntTy LHSOffs = LHS.getOffset(), RHSOffs = RHS.getOffset();
1392    bool LHSLoaded = LHSOffs >= CurrentLoadedOffset;
1393    bool RHSLoaded = RHSOffs >= CurrentLoadedOffset;
1394
1395    if (LHSLoaded == RHSLoaded) {
1396      if (RelativeOffset)
1397        *RelativeOffset = RHSOffs - LHSOffs;
1398      return true;
1399    }
1400
1401    return false;
1402  }
1403
1404  //===--------------------------------------------------------------------===//
1405  // Queries about the code at a SourceLocation.
1406  //===--------------------------------------------------------------------===//
1407
1408  /// Return a pointer to the start of the specified location
1409  /// in the appropriate spelling MemoryBuffer.
1410  ///
1411  /// \param Invalid If non-NULL, will be set \c true if an error occurs.
1412  const char *getCharacterData(SourceLocation SL,
1413                               bool *Invalid = nullptr) const;
1414
1415  /// Return the column # for the specified file position.
1416  ///
1417  /// This is significantly cheaper to compute than the line number.  This
1418  /// returns zero if the column number isn't known.  This may only be called
1419  /// on a file sloc, so you must choose a spelling or expansion location
1420  /// before calling this method.
1421  unsigned getColumnNumber(FileID FID, unsigned FilePos,
1422                           bool *Invalid = nullptr) const;
1423  unsigned getSpellingColumnNumber(SourceLocation Loc,
1424                                   bool *Invalid = nullptr) const;
1425  unsigned getExpansionColumnNumber(SourceLocation Loc,
1426                                    bool *Invalid = nullptr) const;
1427  unsigned getPresumedColumnNumber(SourceLocation Loc,
1428                                   bool *Invalid = nullptr) const;
1429
1430  /// Given a SourceLocation, return the spelling line number
1431  /// for the position indicated.
1432  ///
1433  /// This requires building and caching a table of line offsets for the
1434  /// MemoryBuffer, so this is not cheap: use only when about to emit a
1435  /// diagnostic.
1436  unsigned getLineNumber(FileID FID, unsigned FilePos, bool *Invalid = nullptr) const;
1437  unsigned getSpellingLineNumber(SourceLocation Loc, bool *Invalid = nullptr) const;
1438  unsigned getExpansionLineNumber(SourceLocation Loc, bool *Invalid = nullptr) const;
1439  unsigned getPresumedLineNumber(SourceLocation Loc, bool *Invalid = nullptr) const;
1440
1441  /// Return the filename or buffer identifier of the buffer the
1442  /// location is in.
1443  ///
1444  /// Note that this name does not respect \#line directives.  Use
1445  /// getPresumedLoc for normal clients.
1446  StringRef getBufferName(SourceLocation Loc, bool *Invalid = nullptr) const;
1447
1448  /// Return the file characteristic of the specified source
1449  /// location, indicating whether this is a normal file, a system
1450  /// header, or an "implicit extern C" system header.
1451  ///
1452  /// This state can be modified with flags on GNU linemarker directives like:
1453  /// \code
1454  ///   # 4 "foo.h" 3
1455  /// \endcode
1456  /// which changes all source locations in the current file after that to be
1457  /// considered to be from a system header.
1458  SrcMgr::CharacteristicKind getFileCharacteristic(SourceLocation Loc) const;
1459
1460  /// Returns the "presumed" location of a SourceLocation specifies.
1461  ///
1462  /// A "presumed location" can be modified by \#line or GNU line marker
1463  /// directives.  This provides a view on the data that a user should see
1464  /// in diagnostics, for example.
1465  ///
1466  /// Note that a presumed location is always given as the expansion point of
1467  /// an expansion location, not at the spelling location.
1468  ///
1469  /// \returns The presumed location of the specified SourceLocation. If the
1470  /// presumed location cannot be calculated (e.g., because \p Loc is invalid
1471  /// or the file containing \p Loc has changed on disk), returns an invalid
1472  /// presumed location.
1473  PresumedLoc getPresumedLoc(SourceLocation Loc,
1474                             bool UseLineDirectives = true) const;
1475
1476  /// Returns whether the PresumedLoc for a given SourceLocation is
1477  /// in the main file.
1478  ///
1479  /// This computes the "presumed" location for a SourceLocation, then checks
1480  /// whether it came from a file other than the main file. This is different
1481  /// from isWrittenInMainFile() because it takes line marker directives into
1482  /// account.
1483  bool isInMainFile(SourceLocation Loc) const;
1484
1485  /// Returns true if the spelling locations for both SourceLocations
1486  /// are part of the same file buffer.
1487  ///
1488  /// This check ignores line marker directives.
1489  bool isWrittenInSameFile(SourceLocation Loc1, SourceLocation Loc2) const {
1490    return getFileID(Loc1) == getFileID(Loc2);
1491  }
1492
1493  /// Returns true if the spelling location for the given location
1494  /// is in the main file buffer.
1495  ///
1496  /// This check ignores line marker directives.
1497  bool isWrittenInMainFile(SourceLocation Loc) const {
1498    return getFileID(Loc) == getMainFileID();
1499  }
1500
1501  /// Returns whether \p Loc is located in a <built-in> file.
1502  bool isWrittenInBuiltinFile(SourceLocation Loc) const {
1503    PresumedLoc Presumed = getPresumedLoc(Loc);
1504    if (Presumed.isInvalid())
1505      return false;
1506    StringRef Filename(Presumed.getFilename());
1507    return Filename.equals("<built-in>");
1508  }
1509
1510  /// Returns whether \p Loc is located in a <command line> file.
1511  bool isWrittenInCommandLineFile(SourceLocation Loc) const {
1512    PresumedLoc Presumed = getPresumedLoc(Loc);
1513    if (Presumed.isInvalid())
1514      return false;
1515    StringRef Filename(Presumed.getFilename());
1516    return Filename.equals("<command line>");
1517  }
1518
1519  /// Returns whether \p Loc is located in a <scratch space> file.
1520  bool isWrittenInScratchSpace(SourceLocation Loc) const {
1521    PresumedLoc Presumed = getPresumedLoc(Loc);
1522    if (Presumed.isInvalid())
1523      return false;
1524    StringRef Filename(Presumed.getFilename());
1525    return Filename.equals("<scratch space>");
1526  }
1527
1528  /// Returns if a SourceLocation is in a system header.
1529  bool isInSystemHeader(SourceLocation Loc) const {
1530    if (Loc.isInvalid())
1531      return false;
1532    return isSystem(getFileCharacteristic(Loc));
1533  }
1534
1535  /// Returns if a SourceLocation is in an "extern C" system header.
1536  bool isInExternCSystemHeader(SourceLocation Loc) const {
1537    return getFileCharacteristic(Loc) == SrcMgr::C_ExternCSystem;
1538  }
1539
1540  /// Returns whether \p Loc is expanded from a macro in a system header.
1541  bool isInSystemMacro(SourceLocation loc) const {
1542    if (!loc.isMacroID())
1543      return false;
1544
1545    // This happens when the macro is the result of a paste, in that case
1546    // its spelling is the scratch memory, so we take the parent context.
1547    // There can be several level of token pasting.
1548    if (isWrittenInScratchSpace(getSpellingLoc(loc))) {
1549      do {
1550        loc = getImmediateMacroCallerLoc(loc);
1551      } while (isWrittenInScratchSpace(getSpellingLoc(loc)));
1552      return isInSystemMacro(loc);
1553    }
1554
1555    return isInSystemHeader(getSpellingLoc(loc));
1556  }
1557
1558  /// The size of the SLocEntry that \p FID represents.
1559  unsigned getFileIDSize(FileID FID) const;
1560
1561  /// Given a specific FileID, returns true if \p Loc is inside that
1562  /// FileID chunk and sets relative offset (offset of \p Loc from beginning
1563  /// of FileID) to \p relativeOffset.
1564  bool isInFileID(SourceLocation Loc, FileID FID,
1565                  unsigned *RelativeOffset = nullptr) const {
1566    SourceLocation::UIntTy Offs = Loc.getOffset();
1567    if (isOffsetInFileID(FID, Offs)) {
1568      if (RelativeOffset)
1569        *RelativeOffset = Offs - getSLocEntry(FID).getOffset();
1570      return true;
1571    }
1572
1573    return false;
1574  }
1575
1576  //===--------------------------------------------------------------------===//
1577  // Line Table Manipulation Routines
1578  //===--------------------------------------------------------------------===//
1579
1580  /// Return the uniqued ID for the specified filename.
1581  unsigned getLineTableFilenameID(StringRef Str);
1582
1583  /// Add a line note to the line table for the FileID and offset
1584  /// specified by Loc.
1585  ///
1586  /// If FilenameID is -1, it is considered to be unspecified.
1587  void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID,
1588                   bool IsFileEntry, bool IsFileExit,
1589                   SrcMgr::CharacteristicKind FileKind);
1590
1591  /// Determine if the source manager has a line table.
1592  bool hasLineTable() const { return LineTable != nullptr; }
1593
1594  /// Retrieve the stored line table.
1595  LineTableInfo &getLineTable();
1596
1597  //===--------------------------------------------------------------------===//
1598  // Queries for performance analysis.
1599  //===--------------------------------------------------------------------===//
1600
1601  /// Return the total amount of physical memory allocated by the
1602  /// ContentCache allocator.
1603  size_t getContentCacheSize() const {
1604    return ContentCacheAlloc.getTotalMemory();
1605  }
1606
1607  struct MemoryBufferSizes {
1608    const size_t malloc_bytes;
1609    const size_t mmap_bytes;
1610
1611    MemoryBufferSizes(size_t malloc_bytes, size_t mmap_bytes)
1612      : malloc_bytes(malloc_bytes), mmap_bytes(mmap_bytes) {}
1613  };
1614
1615  /// Return the amount of memory used by memory buffers, breaking down
1616  /// by heap-backed versus mmap'ed memory.
1617  MemoryBufferSizes getMemoryBufferSizes() const;
1618
1619  /// Return the amount of memory used for various side tables and
1620  /// data structures in the SourceManager.
1621  size_t getDataStructureSizes() const;
1622
1623  //===--------------------------------------------------------------------===//
1624  // Other miscellaneous methods.
1625  //===--------------------------------------------------------------------===//
1626
1627  /// Get the source location for the given file:line:col triplet.
1628  ///
1629  /// If the source file is included multiple times, the source location will
1630  /// be based upon the first inclusion.
1631  SourceLocation translateFileLineCol(const FileEntry *SourceFile,
1632                                      unsigned Line, unsigned Col) const;
1633
1634  /// Get the FileID for the given file.
1635  ///
1636  /// If the source file is included multiple times, the FileID will be the
1637  /// first inclusion.
1638  FileID translateFile(const FileEntry *SourceFile) const;
1639  FileID translateFile(FileEntryRef SourceFile) const {
1640    return translateFile(&SourceFile.getFileEntry());
1641  }
1642
1643  /// Get the source location in \p FID for the given line:col.
1644  /// Returns null location if \p FID is not a file SLocEntry.
1645  SourceLocation translateLineCol(FileID FID,
1646                                  unsigned Line, unsigned Col) const;
1647
1648  /// If \p Loc points inside a function macro argument, the returned
1649  /// location will be the macro location in which the argument was expanded.
1650  /// If a macro argument is used multiple times, the expanded location will
1651  /// be at the first expansion of the argument.
1652  /// e.g.
1653  ///   MY_MACRO(foo);
1654  ///             ^
1655  /// Passing a file location pointing at 'foo', will yield a macro location
1656  /// where 'foo' was expanded into.
1657  SourceLocation getMacroArgExpandedLocation(SourceLocation Loc) const;
1658
1659  /// Determines the order of 2 source locations in the translation unit.
1660  ///
1661  /// \returns true if LHS source location comes before RHS, false otherwise.
1662  bool isBeforeInTranslationUnit(SourceLocation LHS, SourceLocation RHS) const;
1663
1664  /// Determines whether the two decomposed source location is in the
1665  ///        same translation unit. As a byproduct, it also calculates the order
1666  ///        of the source locations in case they are in the same TU.
1667  ///
1668  /// \returns Pair of bools the first component is true if the two locations
1669  ///          are in the same TU. The second bool is true if the first is true
1670  ///          and \p LOffs is before \p ROffs.
1671  std::pair<bool, bool>
1672  isInTheSameTranslationUnit(std::pair<FileID, unsigned> &LOffs,
1673                             std::pair<FileID, unsigned> &ROffs) const;
1674
1675  /// Determines whether the two decomposed source location is in the same TU.
1676  bool isInTheSameTranslationUnitImpl(
1677      const std::pair<FileID, unsigned> &LOffs,
1678      const std::pair<FileID, unsigned> &ROffs) const;
1679
1680  /// Determines the order of 2 source locations in the "source location
1681  /// address space".
1682  bool isBeforeInSLocAddrSpace(SourceLocation LHS, SourceLocation RHS) const {
1683    return isBeforeInSLocAddrSpace(LHS, RHS.getOffset());
1684  }
1685
1686  /// Determines the order of a source location and a source location
1687  /// offset in the "source location address space".
1688  ///
1689  /// Note that we always consider source locations loaded from
1690  bool isBeforeInSLocAddrSpace(SourceLocation LHS,
1691                               SourceLocation::UIntTy RHS) const {
1692    SourceLocation::UIntTy LHSOffset = LHS.getOffset();
1693    bool LHSLoaded = LHSOffset >= CurrentLoadedOffset;
1694    bool RHSLoaded = RHS >= CurrentLoadedOffset;
1695    if (LHSLoaded == RHSLoaded)
1696      return LHSOffset < RHS;
1697
1698    return LHSLoaded;
1699  }
1700
1701  /// Return true if the Point is within Start and End.
1702  bool isPointWithin(SourceLocation Location, SourceLocation Start,
1703                     SourceLocation End) const {
1704    return Location == Start || Location == End ||
1705           (isBeforeInTranslationUnit(Start, Location) &&
1706            isBeforeInTranslationUnit(Location, End));
1707  }
1708
1709  // Iterators over FileInfos.
1710  using fileinfo_iterator =
1711      llvm::DenseMap<FileEntryRef, SrcMgr::ContentCache *>::const_iterator;
1712
1713  fileinfo_iterator fileinfo_begin() const { return FileInfos.begin(); }
1714  fileinfo_iterator fileinfo_end() const { return FileInfos.end(); }
1715  bool hasFileInfo(const FileEntry *File) const {
1716    return FileInfos.find_as(File) != FileInfos.end();
1717  }
1718
1719  /// Print statistics to stderr.
1720  void PrintStats() const;
1721
1722  void dump() const;
1723
1724  // Produce notes describing the current source location address space usage.
1725  void noteSLocAddressSpaceUsage(DiagnosticsEngine &Diag,
1726                                 std::optional<unsigned> MaxNotes = 32) const;
1727
1728  /// Get the number of local SLocEntries we have.
1729  unsigned local_sloc_entry_size() const { return LocalSLocEntryTable.size(); }
1730
1731  /// Get a local SLocEntry. This is exposed for indexing.
1732  const SrcMgr::SLocEntry &getLocalSLocEntry(unsigned Index) const {
1733    assert(Index < LocalSLocEntryTable.size() && "Invalid index");
1734    return LocalSLocEntryTable[Index];
1735  }
1736
1737  /// Get the number of loaded SLocEntries we have.
1738  unsigned loaded_sloc_entry_size() const { return LoadedSLocEntryTable.size();}
1739
1740  /// Get a loaded SLocEntry. This is exposed for indexing.
1741  const SrcMgr::SLocEntry &getLoadedSLocEntry(unsigned Index,
1742                                              bool *Invalid = nullptr) const {
1743    assert(Index < LoadedSLocEntryTable.size() && "Invalid index");
1744    if (SLocEntryLoaded[Index])
1745      return LoadedSLocEntryTable[Index];
1746    return loadSLocEntry(Index, Invalid);
1747  }
1748
1749  const SrcMgr::SLocEntry &getSLocEntry(FileID FID,
1750                                        bool *Invalid = nullptr) const {
1751    if (FID.ID == 0 || FID.ID == -1) {
1752      if (Invalid) *Invalid = true;
1753      return LocalSLocEntryTable[0];
1754    }
1755    return getSLocEntryByID(FID.ID, Invalid);
1756  }
1757
1758  SourceLocation::UIntTy getNextLocalOffset() const { return NextLocalOffset; }
1759
1760  void setExternalSLocEntrySource(ExternalSLocEntrySource *Source) {
1761    assert(LoadedSLocEntryTable.empty() &&
1762           "Invalidating existing loaded entries");
1763    ExternalSLocEntries = Source;
1764  }
1765
1766  /// Allocate a number of loaded SLocEntries, which will be actually
1767  /// loaded on demand from the external source.
1768  ///
1769  /// NumSLocEntries will be allocated, which occupy a total of TotalSize space
1770  /// in the global source view. The lowest ID and the base offset of the
1771  /// entries will be returned.
1772  std::pair<int, SourceLocation::UIntTy>
1773  AllocateLoadedSLocEntries(unsigned NumSLocEntries,
1774                            SourceLocation::UIntTy TotalSize);
1775
1776  /// Returns true if \p Loc came from a PCH/Module.
1777  bool isLoadedSourceLocation(SourceLocation Loc) const {
1778    return isLoadedOffset(Loc.getOffset());
1779  }
1780
1781  /// Returns true if \p Loc did not come from a PCH/Module.
1782  bool isLocalSourceLocation(SourceLocation Loc) const {
1783    return isLocalOffset(Loc.getOffset());
1784  }
1785
1786  /// Returns true if \p FID came from a PCH/Module.
1787  bool isLoadedFileID(FileID FID) const {
1788    assert(FID.ID != -1 && "Using FileID sentinel value");
1789    return FID.ID < 0;
1790  }
1791
1792  /// Returns true if \p FID did not come from a PCH/Module.
1793  bool isLocalFileID(FileID FID) const {
1794    return !isLoadedFileID(FID);
1795  }
1796
1797  /// Gets the location of the immediate macro caller, one level up the stack
1798  /// toward the initial macro typed into the source.
1799  SourceLocation getImmediateMacroCallerLoc(SourceLocation Loc) const {
1800    if (!Loc.isMacroID()) return Loc;
1801
1802    // When we have the location of (part of) an expanded parameter, its
1803    // spelling location points to the argument as expanded in the macro call,
1804    // and therefore is used to locate the macro caller.
1805    if (isMacroArgExpansion(Loc))
1806      return getImmediateSpellingLoc(Loc);
1807
1808    // Otherwise, the caller of the macro is located where this macro is
1809    // expanded (while the spelling is part of the macro definition).
1810    return getImmediateExpansionRange(Loc).getBegin();
1811  }
1812
1813  /// \return Location of the top-level macro caller.
1814  SourceLocation getTopMacroCallerLoc(SourceLocation Loc) const;
1815
1816private:
1817  friend class ASTReader;
1818  friend class ASTWriter;
1819
1820  llvm::MemoryBufferRef getFakeBufferForRecovery() const;
1821  SrcMgr::ContentCache &getFakeContentCacheForRecovery() const;
1822
1823  const SrcMgr::SLocEntry &loadSLocEntry(unsigned Index, bool *Invalid) const;
1824
1825  const SrcMgr::SLocEntry *getSLocEntryOrNull(FileID FID) const {
1826    bool Invalid = false;
1827    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
1828    return Invalid ? nullptr : &Entry;
1829  }
1830
1831  const SrcMgr::SLocEntry *getSLocEntryForFile(FileID FID) const {
1832    if (auto *Entry = getSLocEntryOrNull(FID))
1833      if (Entry->isFile())
1834        return Entry;
1835    return nullptr;
1836  }
1837
1838  /// Get the entry with the given unwrapped FileID.
1839  /// Invalid will not be modified for Local IDs.
1840  const SrcMgr::SLocEntry &getSLocEntryByID(int ID,
1841                                            bool *Invalid = nullptr) const {
1842    assert(ID != -1 && "Using FileID sentinel value");
1843    if (ID < 0)
1844      return getLoadedSLocEntryByID(ID, Invalid);
1845    return getLocalSLocEntry(static_cast<unsigned>(ID));
1846  }
1847
1848  const SrcMgr::SLocEntry &
1849  getLoadedSLocEntryByID(int ID, bool *Invalid = nullptr) const {
1850    return getLoadedSLocEntry(static_cast<unsigned>(-ID - 2), Invalid);
1851  }
1852
1853  FileID getFileID(SourceLocation::UIntTy SLocOffset) const {
1854    // If our one-entry cache covers this offset, just return it.
1855    if (isOffsetInFileID(LastFileIDLookup, SLocOffset))
1856      return LastFileIDLookup;
1857
1858    return getFileIDSlow(SLocOffset);
1859  }
1860
1861  bool isLocalOffset(SourceLocation::UIntTy SLocOffset) const {
1862    return SLocOffset < CurrentLoadedOffset;
1863  }
1864
1865  bool isLoadedOffset(SourceLocation::UIntTy SLocOffset) const {
1866    return SLocOffset >= CurrentLoadedOffset;
1867  }
1868
1869  /// Implements the common elements of storing an expansion info struct into
1870  /// the SLocEntry table and producing a source location that refers to it.
1871  SourceLocation
1872  createExpansionLocImpl(const SrcMgr::ExpansionInfo &Expansion,
1873                         unsigned Length, int LoadedID = 0,
1874                         SourceLocation::UIntTy LoadedOffset = 0);
1875
1876  /// Return true if the specified FileID contains the
1877  /// specified SourceLocation offset.  This is a very hot method.
1878  inline bool isOffsetInFileID(FileID FID,
1879                               SourceLocation::UIntTy SLocOffset) const {
1880    const SrcMgr::SLocEntry &Entry = getSLocEntry(FID);
1881    // If the entry is after the offset, it can't contain it.
1882    if (SLocOffset < Entry.getOffset()) return false;
1883
1884    // If this is the very last entry then it does.
1885    if (FID.ID == -2)
1886      return true;
1887
1888    // If it is the last local entry, then it does if the location is local.
1889    if (FID.ID+1 == static_cast<int>(LocalSLocEntryTable.size()))
1890      return SLocOffset < NextLocalOffset;
1891
1892    // Otherwise, the entry after it has to not include it. This works for both
1893    // local and loaded entries.
1894    return SLocOffset < getSLocEntryByID(FID.ID+1).getOffset();
1895  }
1896
1897  /// Returns the previous in-order FileID or an invalid FileID if there
1898  /// is no previous one.
1899  FileID getPreviousFileID(FileID FID) const;
1900
1901  /// Returns the next in-order FileID or an invalid FileID if there is
1902  /// no next one.
1903  FileID getNextFileID(FileID FID) const;
1904
1905  /// Create a new fileID for the specified ContentCache and
1906  /// include position.
1907  ///
1908  /// This works regardless of whether the ContentCache corresponds to a
1909  /// file or some other input source.
1910  FileID createFileIDImpl(SrcMgr::ContentCache &File, StringRef Filename,
1911                          SourceLocation IncludePos,
1912                          SrcMgr::CharacteristicKind DirCharacter, int LoadedID,
1913                          SourceLocation::UIntTy LoadedOffset);
1914
1915  SrcMgr::ContentCache &getOrCreateContentCache(FileEntryRef SourceFile,
1916                                                bool isSystemFile = false);
1917
1918  /// Create a new ContentCache for the specified  memory buffer.
1919  SrcMgr::ContentCache &
1920  createMemBufferContentCache(std::unique_ptr<llvm::MemoryBuffer> Buf);
1921
1922  FileID getFileIDSlow(SourceLocation::UIntTy SLocOffset) const;
1923  FileID getFileIDLocal(SourceLocation::UIntTy SLocOffset) const;
1924  FileID getFileIDLoaded(SourceLocation::UIntTy SLocOffset) const;
1925
1926  SourceLocation getExpansionLocSlowCase(SourceLocation Loc) const;
1927  SourceLocation getSpellingLocSlowCase(SourceLocation Loc) const;
1928  SourceLocation getFileLocSlowCase(SourceLocation Loc) const;
1929
1930  std::pair<FileID, unsigned>
1931  getDecomposedExpansionLocSlowCase(const SrcMgr::SLocEntry *E) const;
1932  std::pair<FileID, unsigned>
1933  getDecomposedSpellingLocSlowCase(const SrcMgr::SLocEntry *E,
1934                                   unsigned Offset) const;
1935  void computeMacroArgsCache(MacroArgsMap &MacroArgsCache, FileID FID) const;
1936  void associateFileChunkWithMacroArgExp(MacroArgsMap &MacroArgsCache,
1937                                         FileID FID,
1938                                         SourceLocation SpellLoc,
1939                                         SourceLocation ExpansionLoc,
1940                                         unsigned ExpansionLength) const;
1941};
1942
1943/// Comparison function object.
1944template<typename T>
1945class BeforeThanCompare;
1946
1947/// Compare two source locations.
1948template<>
1949class BeforeThanCompare<SourceLocation> {
1950  SourceManager &SM;
1951
1952public:
1953  explicit BeforeThanCompare(SourceManager &SM) : SM(SM) {}
1954
1955  bool operator()(SourceLocation LHS, SourceLocation RHS) const {
1956    return SM.isBeforeInTranslationUnit(LHS, RHS);
1957  }
1958};
1959
1960/// Compare two non-overlapping source ranges.
1961template<>
1962class BeforeThanCompare<SourceRange> {
1963  SourceManager &SM;
1964
1965public:
1966  explicit BeforeThanCompare(SourceManager &SM) : SM(SM) {}
1967
1968  bool operator()(SourceRange LHS, SourceRange RHS) const {
1969    return SM.isBeforeInTranslationUnit(LHS.getBegin(), RHS.getBegin());
1970  }
1971};
1972
1973/// SourceManager and necessary dependencies (e.g. VFS, FileManager) for a
1974/// single in-memorty file.
1975class SourceManagerForFile {
1976public:
1977  /// Creates SourceManager and necessary dependencies (e.g. VFS, FileManager).
1978  /// The main file in the SourceManager will be \p FileName with \p Content.
1979  SourceManagerForFile(StringRef FileName, StringRef Content);
1980
1981  SourceManager &get() {
1982    assert(SourceMgr);
1983    return *SourceMgr;
1984  }
1985
1986private:
1987  // The order of these fields are important - they should be in the same order
1988  // as they are created in `createSourceManagerForFile` so that they can be
1989  // deleted in the reverse order as they are created.
1990  std::unique_ptr<FileManager> FileMgr;
1991  std::unique_ptr<DiagnosticsEngine> Diagnostics;
1992  std::unique_ptr<SourceManager> SourceMgr;
1993};
1994
1995} // namespace clang
1996
1997#endif // LLVM_CLANG_BASIC_SOURCEMANAGER_H
1998