1 //===- VirtualFileSystem.h - Virtual File System Layer ----------*- 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 /// \file
10 /// \brief Defines the virtual file system interface vfs::FileSystem.
11 //===----------------------------------------------------------------------===//
12 
13 #ifndef LLVM_CLANG_BASIC_VIRTUALFILESYSTEM_H
14 #define LLVM_CLANG_BASIC_VIRTUALFILESYSTEM_H
15 
16 #include "clang/Basic/LLVM.h"
17 #include "llvm/ADT/IntrusiveRefCntPtr.h"
18 #include "llvm/ADT/Optional.h"
19 #include "llvm/Support/ErrorOr.h"
20 #include "llvm/Support/FileSystem.h"
21 #include "llvm/Support/SourceMgr.h"
22 #include "llvm/Support/raw_ostream.h"
23 
24 namespace llvm {
25 class MemoryBuffer;
26 }
27 
28 namespace clang {
29 namespace vfs {
30 
31 /// \brief The result of a \p status operation.
32 class Status {
33   std::string Name;
34   llvm::sys::fs::UniqueID UID;
35   llvm::sys::TimeValue MTime;
36   uint32_t User;
37   uint32_t Group;
38   uint64_t Size;
39   llvm::sys::fs::file_type Type;
40   llvm::sys::fs::perms Perms;
41 
42 public:
43   bool IsVFSMapped; // FIXME: remove when files support multiple names
44 
45 public:
Status()46   Status() : Type(llvm::sys::fs::file_type::status_error) {}
47   Status(const llvm::sys::fs::file_status &Status);
48   Status(StringRef Name, llvm::sys::fs::UniqueID UID,
49          llvm::sys::TimeValue MTime, uint32_t User, uint32_t Group,
50          uint64_t Size, llvm::sys::fs::file_type Type,
51          llvm::sys::fs::perms Perms);
52 
53   /// Get a copy of a Status with a different name.
54   static Status copyWithNewName(const Status &In, StringRef NewName);
55   static Status copyWithNewName(const llvm::sys::fs::file_status &In,
56                                 StringRef NewName);
57 
58   /// \brief Returns the name that should be used for this file or directory.
getName()59   StringRef getName() const { return Name; }
60 
61   /// @name Status interface from llvm::sys::fs
62   /// @{
getType()63   llvm::sys::fs::file_type getType() const { return Type; }
getPermissions()64   llvm::sys::fs::perms getPermissions() const { return Perms; }
getLastModificationTime()65   llvm::sys::TimeValue getLastModificationTime() const { return MTime; }
getUniqueID()66   llvm::sys::fs::UniqueID getUniqueID() const { return UID; }
getUser()67   uint32_t getUser() const { return User; }
getGroup()68   uint32_t getGroup() const { return Group; }
getSize()69   uint64_t getSize() const { return Size; }
70   /// @}
71   /// @name Status queries
72   /// These are static queries in llvm::sys::fs.
73   /// @{
74   bool equivalent(const Status &Other) const;
75   bool isDirectory() const;
76   bool isRegularFile() const;
77   bool isOther() const;
78   bool isSymlink() const;
79   bool isStatusKnown() const;
80   bool exists() const;
81   /// @}
82 };
83 
84 /// \brief Represents an open file.
85 class File {
86 public:
87   /// \brief Destroy the file after closing it (if open).
88   /// Sub-classes should generally call close() inside their destructors.  We
89   /// cannot do that from the base class, since close is virtual.
90   virtual ~File();
91   /// \brief Get the status of the file.
92   virtual llvm::ErrorOr<Status> status() = 0;
93   /// \brief Get the contents of the file as a \p MemoryBuffer.
94   virtual llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>>
95   getBuffer(const Twine &Name, int64_t FileSize = -1,
96             bool RequiresNullTerminator = true, bool IsVolatile = false) = 0;
97   /// \brief Closes the file.
98   virtual std::error_code close() = 0;
99 };
100 
101 namespace detail {
102 /// \brief An interface for virtual file systems to provide an iterator over the
103 /// (non-recursive) contents of a directory.
104 struct DirIterImpl {
105   virtual ~DirIterImpl();
106   /// \brief Sets \c CurrentEntry to the next entry in the directory on success,
107   /// or returns a system-defined \c error_code.
108   virtual std::error_code increment() = 0;
109   Status CurrentEntry;
110 };
111 } // end namespace detail
112 
113 /// \brief An input iterator over the entries in a virtual path, similar to
114 /// llvm::sys::fs::directory_iterator.
115 class directory_iterator {
116   std::shared_ptr<detail::DirIterImpl> Impl; // Input iterator semantics on copy
117 
118 public:
directory_iterator(std::shared_ptr<detail::DirIterImpl> I)119   directory_iterator(std::shared_ptr<detail::DirIterImpl> I) : Impl(I) {
120     assert(Impl.get() != nullptr && "requires non-null implementation");
121     if (!Impl->CurrentEntry.isStatusKnown())
122       Impl.reset(); // Normalize the end iterator to Impl == nullptr.
123   }
124 
125   /// \brief Construct an 'end' iterator.
directory_iterator()126   directory_iterator() { }
127 
128   /// \brief Equivalent to operator++, with an error code.
increment(std::error_code & EC)129   directory_iterator &increment(std::error_code &EC) {
130     assert(Impl && "attempting to increment past end");
131     EC = Impl->increment();
132     if (EC || !Impl->CurrentEntry.isStatusKnown())
133       Impl.reset(); // Normalize the end iterator to Impl == nullptr.
134     return *this;
135   }
136 
137   const Status &operator*() const { return Impl->CurrentEntry; }
138   const Status *operator->() const { return &Impl->CurrentEntry; }
139 
140   bool operator==(const directory_iterator &RHS) const {
141     if (Impl && RHS.Impl)
142       return Impl->CurrentEntry.equivalent(RHS.Impl->CurrentEntry);
143     return !Impl && !RHS.Impl;
144   }
145   bool operator!=(const directory_iterator &RHS) const {
146     return !(*this == RHS);
147   }
148 };
149 
150 class FileSystem;
151 
152 /// \brief An input iterator over the recursive contents of a virtual path,
153 /// similar to llvm::sys::fs::recursive_directory_iterator.
154 class recursive_directory_iterator {
155   typedef std::stack<directory_iterator, std::vector<directory_iterator>>
156       IterState;
157 
158   FileSystem *FS;
159   std::shared_ptr<IterState> State; // Input iterator semantics on copy.
160 
161 public:
162   recursive_directory_iterator(FileSystem &FS, const Twine &Path,
163                                std::error_code &EC);
164   /// \brief Construct an 'end' iterator.
recursive_directory_iterator()165   recursive_directory_iterator() { }
166 
167   /// \brief Equivalent to operator++, with an error code.
168   recursive_directory_iterator &increment(std::error_code &EC);
169 
170   const Status &operator*() const { return *State->top(); }
171   const Status *operator->() const { return &*State->top(); }
172 
173   bool operator==(const recursive_directory_iterator &Other) const {
174     return State == Other.State; // identity
175   }
176   bool operator!=(const recursive_directory_iterator &RHS) const {
177     return !(*this == RHS);
178   }
179 };
180 
181 /// \brief The virtual file system interface.
182 class FileSystem : public llvm::ThreadSafeRefCountedBase<FileSystem> {
183 public:
184   virtual ~FileSystem();
185 
186   /// \brief Get the status of the entry at \p Path, if one exists.
187   virtual llvm::ErrorOr<Status> status(const Twine &Path) = 0;
188   /// \brief Get a \p File object for the file at \p Path, if one exists.
189   virtual llvm::ErrorOr<std::unique_ptr<File>>
190   openFileForRead(const Twine &Path) = 0;
191 
192   /// This is a convenience method that opens a file, gets its content and then
193   /// closes the file.
194   llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>>
195   getBufferForFile(const Twine &Name, int64_t FileSize = -1,
196                    bool RequiresNullTerminator = true, bool IsVolatile = false);
197 
198   /// \brief Get a directory_iterator for \p Dir.
199   /// \note The 'end' iterator is directory_iterator().
200   virtual directory_iterator dir_begin(const Twine &Dir,
201                                        std::error_code &EC) = 0;
202 
203   /// Set the working directory. This will affect all following operations on
204   /// this file system and may propagate down for nested file systems.
205   virtual std::error_code setCurrentWorkingDirectory(const Twine &Path) = 0;
206   /// Get the working directory of this file system.
207   virtual llvm::ErrorOr<std::string> getCurrentWorkingDirectory() const = 0;
208 
209   /// Check whether a file exists. Provided for convenience.
210   bool exists(const Twine &Path);
211 
212   /// Make \a Path an absolute path.
213   ///
214   /// Makes \a Path absolute using the current directory if it is not already.
215   /// An empty \a Path will result in the current directory.
216   ///
217   /// /absolute/path   => /absolute/path
218   /// relative/../path => <current-directory>/relative/../path
219   ///
220   /// \param Path A path that is modified to be an absolute path.
221   /// \returns success if \a path has been made absolute, otherwise a
222   ///          platform-specific error_code.
223   std::error_code makeAbsolute(SmallVectorImpl<char> &Path) const;
224 };
225 
226 /// \brief Gets an \p vfs::FileSystem for the 'real' file system, as seen by
227 /// the operating system.
228 IntrusiveRefCntPtr<FileSystem> getRealFileSystem();
229 
230 /// \brief A file system that allows overlaying one \p AbstractFileSystem on top
231 /// of another.
232 ///
233 /// Consists of a stack of >=1 \p FileSystem objects, which are treated as being
234 /// one merged file system. When there is a directory that exists in more than
235 /// one file system, the \p OverlayFileSystem contains a directory containing
236 /// the union of their contents.  The attributes (permissions, etc.) of the
237 /// top-most (most recently added) directory are used.  When there is a file
238 /// that exists in more than one file system, the file in the top-most file
239 /// system overrides the other(s).
240 class OverlayFileSystem : public FileSystem {
241   typedef SmallVector<IntrusiveRefCntPtr<FileSystem>, 1> FileSystemList;
242   /// \brief The stack of file systems, implemented as a list in order of
243   /// their addition.
244   FileSystemList FSList;
245 
246 public:
247   OverlayFileSystem(IntrusiveRefCntPtr<FileSystem> Base);
248   /// \brief Pushes a file system on top of the stack.
249   void pushOverlay(IntrusiveRefCntPtr<FileSystem> FS);
250 
251   llvm::ErrorOr<Status> status(const Twine &Path) override;
252   llvm::ErrorOr<std::unique_ptr<File>>
253   openFileForRead(const Twine &Path) override;
254   directory_iterator dir_begin(const Twine &Dir, std::error_code &EC) override;
255   llvm::ErrorOr<std::string> getCurrentWorkingDirectory() const override;
256   std::error_code setCurrentWorkingDirectory(const Twine &Path) override;
257 
258   typedef FileSystemList::reverse_iterator iterator;
259 
260   /// \brief Get an iterator pointing to the most recently added file system.
overlays_begin()261   iterator overlays_begin() { return FSList.rbegin(); }
262 
263   /// \brief Get an iterator pointing one-past the least recently added file
264   /// system.
overlays_end()265   iterator overlays_end() { return FSList.rend(); }
266 };
267 
268 namespace detail {
269 class InMemoryDirectory;
270 } // end namespace detail
271 
272 /// An in-memory file system.
273 class InMemoryFileSystem : public FileSystem {
274   std::unique_ptr<detail::InMemoryDirectory> Root;
275   std::string WorkingDirectory;
276   bool UseNormalizedPaths = true;
277 
278 public:
279   explicit InMemoryFileSystem(bool UseNormalizedPaths = true);
280   ~InMemoryFileSystem() override;
281   /// Add a buffer to the VFS with a path. The VFS owns the buffer.
282   /// \return true if the file was successfully added, false if the file already
283   /// exists in the file system with different contents.
284   bool addFile(const Twine &Path, time_t ModificationTime,
285                std::unique_ptr<llvm::MemoryBuffer> Buffer);
286   /// Add a buffer to the VFS with a path. The VFS does not own the buffer.
287   /// \return true if the file was successfully added, false if the file already
288   /// exists in the file system with different contents.
289   bool addFileNoOwn(const Twine &Path, time_t ModificationTime,
290                     llvm::MemoryBuffer *Buffer);
291   std::string toString() const;
292   /// Return true if this file system normalizes . and .. in paths.
useNormalizedPaths()293   bool useNormalizedPaths() const { return UseNormalizedPaths; }
294 
295   llvm::ErrorOr<Status> status(const Twine &Path) override;
296   llvm::ErrorOr<std::unique_ptr<File>>
297   openFileForRead(const Twine &Path) override;
298   directory_iterator dir_begin(const Twine &Dir, std::error_code &EC) override;
getCurrentWorkingDirectory()299   llvm::ErrorOr<std::string> getCurrentWorkingDirectory() const override {
300     return WorkingDirectory;
301   }
setCurrentWorkingDirectory(const Twine & Path)302   std::error_code setCurrentWorkingDirectory(const Twine &Path) override {
303     WorkingDirectory = Path.str();
304     return std::error_code();
305   }
306 };
307 
308 /// \brief Get a globally unique ID for a virtual file or directory.
309 llvm::sys::fs::UniqueID getNextVirtualUniqueID();
310 
311 /// \brief Gets a \p FileSystem for a virtual file system described in YAML
312 /// format.
313 IntrusiveRefCntPtr<FileSystem>
314 getVFSFromYAML(std::unique_ptr<llvm::MemoryBuffer> Buffer,
315                llvm::SourceMgr::DiagHandlerTy DiagHandler,
316                void *DiagContext = nullptr,
317                IntrusiveRefCntPtr<FileSystem> ExternalFS = getRealFileSystem());
318 
319 struct YAMLVFSEntry {
YAMLVFSEntryYAMLVFSEntry320   template <typename T1, typename T2> YAMLVFSEntry(T1 &&VPath, T2 &&RPath)
321       : VPath(std::forward<T1>(VPath)), RPath(std::forward<T2>(RPath)) {}
322   std::string VPath;
323   std::string RPath;
324 };
325 
326 class YAMLVFSWriter {
327   std::vector<YAMLVFSEntry> Mappings;
328   Optional<bool> IsCaseSensitive;
329 
330 public:
YAMLVFSWriter()331   YAMLVFSWriter() {}
332   void addFileMapping(StringRef VirtualPath, StringRef RealPath);
setCaseSensitivity(bool CaseSensitive)333   void setCaseSensitivity(bool CaseSensitive) {
334     IsCaseSensitive = CaseSensitive;
335   }
336   void write(llvm::raw_ostream &OS);
337 };
338 
339 } // end namespace vfs
340 } // end namespace clang
341 #endif
342