HeaderSearch.h revision a30cfe5026b12c28b7b575b48176e0a3543ce939
1//===--- HeaderSearch.h - Resolve Header File Locations ---------*- C++ -*-===//
2//
3//                     The LLVM Compiler Infrastructure
4//
5// This file is distributed under the University of Illinois Open Source
6// License. See LICENSE.TXT for details.
7//
8//===----------------------------------------------------------------------===//
9//
10// This file defines the HeaderSearch interface.
11//
12//===----------------------------------------------------------------------===//
13
14#ifndef LLVM_CLANG_LEX_HEADERSEARCH_H
15#define LLVM_CLANG_LEX_HEADERSEARCH_H
16
17#include "clang/Lex/DirectoryLookup.h"
18#include "clang/Lex/ModuleMap.h"
19#include "llvm/ADT/StringMap.h"
20#include "llvm/ADT/StringSet.h"
21#include "llvm/Support/Allocator.h"
22#include <vector>
23
24namespace clang {
25
26class DiagnosticsEngine;
27class ExternalIdentifierLookup;
28class FileEntry;
29class FileManager;
30class IdentifierInfo;
31
32/// HeaderFileInfo - The preprocessor keeps track of this information for each
33/// file that is #included.
34struct HeaderFileInfo {
35  /// isImport - True if this is a #import'd or #pragma once file.
36  unsigned isImport : 1;
37
38  /// isPragmaOnce - True if this is  #pragma once file.
39  unsigned isPragmaOnce : 1;
40
41  /// DirInfo - Keep track of whether this is a system header, and if so,
42  /// whether it is C++ clean or not.  This can be set by the include paths or
43  /// by #pragma gcc system_header.  This is an instance of
44  /// SrcMgr::CharacteristicKind.
45  unsigned DirInfo : 2;
46
47  /// \brief Whether this header file info was supplied by an external source.
48  unsigned External : 1;
49
50  /// \brief Whether this structure is considered to already have been
51  /// "resolved", meaning that it was loaded from the external source.
52  unsigned Resolved : 1;
53
54  /// \brief Whether this is a header inside a framework that is currently
55  /// being built.
56  ///
57  /// When a framework is being built, the headers have not yet been placed
58  /// into the appropriate framework subdirectories, and therefore are
59  /// provided via a header map. This bit indicates when this is one of
60  /// those framework headers.
61  unsigned IndexHeaderMapHeader : 1;
62
63  /// NumIncludes - This is the number of times the file has been included
64  /// already.
65  unsigned short NumIncludes;
66
67  /// \brief The ID number of the controlling macro.
68  ///
69  /// This ID number will be non-zero when there is a controlling
70  /// macro whose IdentifierInfo may not yet have been loaded from
71  /// external storage.
72  unsigned ControllingMacroID;
73
74  /// ControllingMacro - If this file has a #ifndef XXX (or equivalent) guard
75  /// that protects the entire contents of the file, this is the identifier
76  /// for the macro that controls whether or not it has any effect.
77  ///
78  /// Note: Most clients should use getControllingMacro() to access
79  /// the controlling macro of this header, since
80  /// getControllingMacro() is able to load a controlling macro from
81  /// external storage.
82  const IdentifierInfo *ControllingMacro;
83
84  /// \brief If this header came from a framework include, this is the name
85  /// of the framework.
86  StringRef Framework;
87
88  HeaderFileInfo()
89    : isImport(false), isPragmaOnce(false), DirInfo(SrcMgr::C_User),
90      External(false), Resolved(false), IndexHeaderMapHeader(false),
91      NumIncludes(0), ControllingMacroID(0), ControllingMacro(0)  {}
92
93  /// \brief Retrieve the controlling macro for this header file, if
94  /// any.
95  const IdentifierInfo *getControllingMacro(ExternalIdentifierLookup *External);
96
97  /// \brief Determine whether this is a non-default header file info, e.g.,
98  /// it corresponds to an actual header we've included or tried to include.
99  bool isNonDefault() const {
100    return isImport || isPragmaOnce || NumIncludes || ControllingMacro ||
101      ControllingMacroID;
102  }
103};
104
105/// \brief An external source of header file information, which may supply
106/// information about header files already included.
107class ExternalHeaderFileInfoSource {
108public:
109  virtual ~ExternalHeaderFileInfoSource();
110
111  /// \brief Retrieve the header file information for the given file entry.
112  ///
113  /// \returns Header file information for the given file entry, with the
114  /// \c External bit set. If the file entry is not known, return a
115  /// default-constructed \c HeaderFileInfo.
116  virtual HeaderFileInfo GetHeaderFileInfo(const FileEntry *FE) = 0;
117};
118
119/// HeaderSearch - This class encapsulates the information needed to find the
120/// file referenced by a #include or #include_next, (sub-)framework lookup, etc.
121class HeaderSearch {
122  FileManager &FileMgr;
123  DiagnosticsEngine &Diags;
124  /// #include search path information.  Requests for #include "x" search the
125  /// directory of the #including file first, then each directory in SearchDirs
126  /// consecutively. Requests for <x> search the current dir first, then each
127  /// directory in SearchDirs, starting at AngledDirIdx, consecutively.  If
128  /// NoCurDirSearch is true, then the check for the file in the current
129  /// directory is suppressed.
130  std::vector<DirectoryLookup> SearchDirs;
131  unsigned AngledDirIdx;
132  unsigned SystemDirIdx;
133  bool NoCurDirSearch;
134
135  /// \brief The path to the module cache.
136  std::string ModuleCachePath;
137
138  /// \brief The name of the module we're building.
139  std::string BuildingModule;
140
141  /// FileInfo - This contains all of the preprocessor-specific data about files
142  /// that are included.  The vector is indexed by the FileEntry's UID.
143  ///
144  std::vector<HeaderFileInfo> FileInfo;
145
146  /// LookupFileCache - This is keeps track of each lookup performed by
147  /// LookupFile.  The first part of the value is the starting index in
148  /// SearchDirs that the cached search was performed from.  If there is a hit
149  /// and this value doesn't match the current query, the cache has to be
150  /// ignored.  The second value is the entry in SearchDirs that satisfied the
151  /// query.
152  llvm::StringMap<std::pair<unsigned, unsigned>, llvm::BumpPtrAllocator>
153    LookupFileCache;
154
155  /// FrameworkMap - This is a collection mapping a framework or subframework
156  /// name like "Carbon" to the Carbon.framework directory.
157  llvm::StringMap<const DirectoryEntry *, llvm::BumpPtrAllocator>
158    FrameworkMap;
159
160  /// HeaderMaps - This is a mapping from FileEntry -> HeaderMap, uniquing
161  /// headermaps.  This vector owns the headermap.
162  std::vector<std::pair<const FileEntry*, const HeaderMap*> > HeaderMaps;
163
164  /// \brief The mapping between modules and headers.
165  ModuleMap ModMap;
166
167  /// \brief Describes whether a given directory has a module map in it.
168  llvm::DenseMap<const DirectoryEntry *, bool> DirectoryHasModuleMap;
169
170  /// \brief Uniqued set of framework names, which is used to track which
171  /// headers were included as framework headers.
172  llvm::StringSet<llvm::BumpPtrAllocator> FrameworkNames;
173
174  /// \brief Entity used to resolve the identifier IDs of controlling
175  /// macros into IdentifierInfo pointers, as needed.
176  ExternalIdentifierLookup *ExternalLookup;
177
178  /// \brief Entity used to look up stored header file information.
179  ExternalHeaderFileInfoSource *ExternalSource;
180
181  // Various statistics we track for performance analysis.
182  unsigned NumIncluded;
183  unsigned NumMultiIncludeFileOptzn;
184  unsigned NumFrameworkLookups, NumSubFrameworkLookups;
185
186  // HeaderSearch doesn't support default or copy construction.
187  explicit HeaderSearch();
188  explicit HeaderSearch(const HeaderSearch&);
189  void operator=(const HeaderSearch&);
190public:
191  HeaderSearch(FileManager &FM, DiagnosticsEngine &Diags);
192  ~HeaderSearch();
193
194  FileManager &getFileMgr() const { return FileMgr; }
195
196  /// SetSearchPaths - Interface for setting the file search paths.
197  ///
198  void SetSearchPaths(const std::vector<DirectoryLookup> &dirs,
199                      unsigned angledDirIdx, unsigned systemDirIdx,
200                      bool noCurDirSearch) {
201    assert(angledDirIdx <= systemDirIdx && systemDirIdx <= dirs.size() &&
202        "Directory indicies are unordered");
203    SearchDirs = dirs;
204    AngledDirIdx = angledDirIdx;
205    SystemDirIdx = systemDirIdx;
206    NoCurDirSearch = noCurDirSearch;
207    //LookupFileCache.clear();
208  }
209
210  /// \brief Set the path to the module cache and the name of the module
211  /// we're building
212  void configureModules(StringRef CachePath, StringRef BuildingModule) {
213    ModuleCachePath = CachePath;
214    this->BuildingModule = BuildingModule;
215  }
216
217  /// ClearFileInfo - Forget everything we know about headers so far.
218  void ClearFileInfo() {
219    FileInfo.clear();
220  }
221
222  void SetExternalLookup(ExternalIdentifierLookup *EIL) {
223    ExternalLookup = EIL;
224  }
225
226  ExternalIdentifierLookup *getExternalLookup() const {
227    return ExternalLookup;
228  }
229
230  /// \brief Set the external source of header information.
231  void SetExternalSource(ExternalHeaderFileInfoSource *ES) {
232    ExternalSource = ES;
233  }
234
235  /// LookupFile - Given a "foo" or <foo> reference, look up the indicated file,
236  /// return null on failure.
237  ///
238  /// \returns If successful, this returns 'UsedDir', the DirectoryLookup member
239  /// the file was found in, or null if not applicable.
240  ///
241  /// \param isAngled indicates whether the file reference is a <> reference.
242  ///
243  /// \param CurDir If non-null, the file was found in the specified directory
244  /// search location.  This is used to implement #include_next.
245  ///
246  /// \param CurFileEnt If non-null, indicates where the #including file is, in
247  /// case a relative search is needed.
248  ///
249  /// \param SearchPath If non-null, will be set to the search path relative
250  /// to which the file was found. If the include path is absolute, SearchPath
251  /// will be set to an empty string.
252  ///
253  /// \param RelativePath If non-null, will be set to the path relative to
254  /// SearchPath at which the file was found. This only differs from the
255  /// Filename for framework includes.
256  ///
257  /// \param SuggestedModule If non-null, and the file found is semantically
258  /// part of a known module, this will be set to the name of the module that
259  /// could be imported instead of preprocessing/parsing the file found.
260  const FileEntry *LookupFile(StringRef Filename, bool isAngled,
261                              const DirectoryLookup *FromDir,
262                              const DirectoryLookup *&CurDir,
263                              const FileEntry *CurFileEnt,
264                              SmallVectorImpl<char> *SearchPath,
265                              SmallVectorImpl<char> *RelativePath,
266                              StringRef *SuggestedModule);
267
268  /// LookupSubframeworkHeader - Look up a subframework for the specified
269  /// #include file.  For example, if #include'ing <HIToolbox/HIToolbox.h> from
270  /// within ".../Carbon.framework/Headers/Carbon.h", check to see if HIToolbox
271  /// is a subframework within Carbon.framework.  If so, return the FileEntry
272  /// for the designated file, otherwise return null.
273  const FileEntry *LookupSubframeworkHeader(
274      StringRef Filename,
275      const FileEntry *RelativeFileEnt,
276      SmallVectorImpl<char> *SearchPath,
277      SmallVectorImpl<char> *RelativePath);
278
279  /// LookupFrameworkCache - Look up the specified framework name in our
280  /// framework cache, returning the DirectoryEntry it is in if we know,
281  /// otherwise, return null.
282  const DirectoryEntry *&LookupFrameworkCache(StringRef FWName) {
283    return FrameworkMap.GetOrCreateValue(FWName).getValue();
284  }
285
286  /// ShouldEnterIncludeFile - Mark the specified file as a target of of a
287  /// #include, #include_next, or #import directive.  Return false if #including
288  /// the file will have no effect or true if we should include it.
289  bool ShouldEnterIncludeFile(const FileEntry *File, bool isImport);
290
291
292  /// getFileDirFlavor - Return whether the specified file is a normal header,
293  /// a system header, or a C++ friendly system header.
294  SrcMgr::CharacteristicKind getFileDirFlavor(const FileEntry *File) {
295    return (SrcMgr::CharacteristicKind)getFileInfo(File).DirInfo;
296  }
297
298  /// MarkFileIncludeOnce - Mark the specified file as a "once only" file, e.g.
299  /// due to #pragma once.
300  void MarkFileIncludeOnce(const FileEntry *File) {
301    HeaderFileInfo &FI = getFileInfo(File);
302    FI.isImport = true;
303    FI.isPragmaOnce = true;
304  }
305
306  /// MarkFileSystemHeader - Mark the specified file as a system header, e.g.
307  /// due to #pragma GCC system_header.
308  void MarkFileSystemHeader(const FileEntry *File) {
309    getFileInfo(File).DirInfo = SrcMgr::C_System;
310  }
311
312  /// IncrementIncludeCount - Increment the count for the number of times the
313  /// specified FileEntry has been entered.
314  void IncrementIncludeCount(const FileEntry *File) {
315    ++getFileInfo(File).NumIncludes;
316  }
317
318  /// SetFileControllingMacro - Mark the specified file as having a controlling
319  /// macro.  This is used by the multiple-include optimization to eliminate
320  /// no-op #includes.
321  void SetFileControllingMacro(const FileEntry *File,
322                               const IdentifierInfo *ControllingMacro) {
323    getFileInfo(File).ControllingMacro = ControllingMacro;
324  }
325
326  /// \brief Determine whether this file is intended to be safe from
327  /// multiple inclusions, e.g., it has #pragma once or a controlling
328  /// macro.
329  ///
330  /// This routine does not consider the effect of #import
331  bool isFileMultipleIncludeGuarded(const FileEntry *File);
332
333  /// CreateHeaderMap - This method returns a HeaderMap for the specified
334  /// FileEntry, uniquing them through the the 'HeaderMaps' datastructure.
335  const HeaderMap *CreateHeaderMap(const FileEntry *FE);
336
337  /// \brief Search in the module cache path for a module with the given
338  /// name.
339  ///
340  /// \param If non-NULL, will be set to the module file name we expected to
341  /// find (regardless of whether it was actually found or not).
342  ///
343  /// \param UmbrellaHeader If non-NULL, and no module was found in the module
344  /// cache, this routine will search in the framework paths to determine
345  /// whether a module can be built from an umbrella header. If so, the pointee
346  /// will be set to the path of the umbrella header.
347  ///
348  /// \returns A file describing the named module, if available, or NULL to
349  /// indicate that the module could not be found.
350  const FileEntry *lookupModule(StringRef ModuleName,
351                                std::string *ModuleFileName = 0,
352                                std::string *UmbrellaHeader = 0);
353
354  void IncrementFrameworkLookupCount() { ++NumFrameworkLookups; }
355
356  /// \brief Determine whether there is a module map that may map the header
357  /// with the given file name to a (sub)module.
358  ///
359  /// \param Filename The name of the file.
360  ///
361  /// \param Root The "root" directory, at which we should stop looking for
362  /// module maps.
363  bool hasModuleMap(StringRef Filename, const DirectoryEntry *Root);
364
365  /// \brief Retrieve the module that corresponds to the given file, if any.
366  ///
367  /// FIXME: This will need to be generalized for submodules.
368  StringRef getModuleForHeader(const FileEntry *File);
369
370  typedef std::vector<HeaderFileInfo>::const_iterator header_file_iterator;
371  header_file_iterator header_file_begin() const { return FileInfo.begin(); }
372  header_file_iterator header_file_end() const { return FileInfo.end(); }
373  unsigned header_file_size() const { return FileInfo.size(); }
374
375  // Used by ASTReader.
376  void setHeaderFileInfoForUID(HeaderFileInfo HFI, unsigned UID);
377
378  // Used by external tools
379  typedef std::vector<DirectoryLookup>::const_iterator search_dir_iterator;
380  search_dir_iterator search_dir_begin() const { return SearchDirs.begin(); }
381  search_dir_iterator search_dir_end() const { return SearchDirs.end(); }
382  unsigned search_dir_size() const { return SearchDirs.size(); }
383
384  search_dir_iterator quoted_dir_begin() const {
385    return SearchDirs.begin();
386  }
387  search_dir_iterator quoted_dir_end() const {
388    return SearchDirs.begin() + AngledDirIdx;
389  }
390
391  search_dir_iterator angled_dir_begin() const {
392    return SearchDirs.begin() + AngledDirIdx;
393  }
394  search_dir_iterator angled_dir_end() const {
395    return SearchDirs.begin() + SystemDirIdx;
396  }
397
398  search_dir_iterator system_dir_begin() const {
399    return SearchDirs.begin() + SystemDirIdx;
400  }
401  search_dir_iterator system_dir_end() const { return SearchDirs.end(); }
402
403  /// \brief Retrieve a uniqued framework name.
404  StringRef getUniqueFrameworkName(StringRef Framework);
405
406  void PrintStats();
407
408  size_t getTotalMemory() const;
409
410private:
411
412  /// getFileInfo - Return the HeaderFileInfo structure for the specified
413  /// FileEntry.
414  HeaderFileInfo &getFileInfo(const FileEntry *FE);
415};
416
417}  // end namespace clang
418
419#endif
420