1 //===--- FileManager.h - File System Probing and Caching --------*- 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 clang::FileManager interface and associated types. 11 /// 12 //===----------------------------------------------------------------------===// 13 14 #ifndef LLVM_CLANG_BASIC_FILEMANAGER_H 15 #define LLVM_CLANG_BASIC_FILEMANAGER_H 16 17 #include "clang/Basic/DirectoryEntry.h" 18 #include "clang/Basic/FileEntry.h" 19 #include "clang/Basic/FileSystemOptions.h" 20 #include "clang/Basic/LLVM.h" 21 #include "llvm/ADT/DenseMap.h" 22 #include "llvm/ADT/IntrusiveRefCntPtr.h" 23 #include "llvm/ADT/PointerUnion.h" 24 #include "llvm/ADT/SmallVector.h" 25 #include "llvm/ADT/StringMap.h" 26 #include "llvm/ADT/StringRef.h" 27 #include "llvm/Support/Allocator.h" 28 #include "llvm/Support/ErrorOr.h" 29 #include "llvm/Support/FileSystem.h" 30 #include "llvm/Support/VirtualFileSystem.h" 31 #include <ctime> 32 #include <map> 33 #include <memory> 34 #include <string> 35 36 namespace llvm { 37 38 class MemoryBuffer; 39 40 } // end namespace llvm 41 42 namespace clang { 43 44 class FileSystemStatCache; 45 46 /// Implements support for file system lookup, file system caching, 47 /// and directory search management. 48 /// 49 /// This also handles more advanced properties, such as uniquing files based 50 /// on "inode", so that a file with two names (e.g. symlinked) will be treated 51 /// as a single file. 52 /// 53 class FileManager : public RefCountedBase<FileManager> { 54 IntrusiveRefCntPtr<llvm::vfs::FileSystem> FS; 55 FileSystemOptions FileSystemOpts; 56 57 /// Cache for existing real directories. 58 std::map<llvm::sys::fs::UniqueID, DirectoryEntry> UniqueRealDirs; 59 60 /// Cache for existing real files. 61 std::map<llvm::sys::fs::UniqueID, FileEntry> UniqueRealFiles; 62 63 /// The virtual directories that we have allocated. 64 /// 65 /// For each virtual file (e.g. foo/bar/baz.cpp), we add all of its parent 66 /// directories (foo/ and foo/bar/) here. 67 SmallVector<std::unique_ptr<DirectoryEntry>, 4> VirtualDirectoryEntries; 68 /// The virtual files that we have allocated. 69 SmallVector<std::unique_ptr<FileEntry>, 4> VirtualFileEntries; 70 71 /// A set of files that bypass the maps and uniquing. They can have 72 /// conflicting filenames. 73 SmallVector<std::unique_ptr<FileEntry>, 0> BypassFileEntries; 74 75 /// A cache that maps paths to directory entries (either real or 76 /// virtual) we have looked up, or an error that occurred when we looked up 77 /// the directory. 78 /// 79 /// The actual Entries for real directories/files are 80 /// owned by UniqueRealDirs/UniqueRealFiles above, while the Entries 81 /// for virtual directories/files are owned by 82 /// VirtualDirectoryEntries/VirtualFileEntries above. 83 /// 84 llvm::StringMap<llvm::ErrorOr<DirectoryEntry &>, llvm::BumpPtrAllocator> 85 SeenDirEntries; 86 87 /// A cache that maps paths to file entries (either real or 88 /// virtual) we have looked up, or an error that occurred when we looked up 89 /// the file. 90 /// 91 /// \see SeenDirEntries 92 llvm::StringMap<llvm::ErrorOr<FileEntryRef::MapValue>, llvm::BumpPtrAllocator> 93 SeenFileEntries; 94 95 /// A mirror of SeenFileEntries to give fake answers for getBypassFile(). 96 /// 97 /// Don't bother hooking up a BumpPtrAllocator. This should be rarely used, 98 /// and only on error paths. 99 std::unique_ptr<llvm::StringMap<llvm::ErrorOr<FileEntryRef::MapValue>>> 100 SeenBypassFileEntries; 101 102 /// The canonical names of files and directories . 103 llvm::DenseMap<const void *, llvm::StringRef> CanonicalNames; 104 105 /// Storage for canonical names that we have computed. 106 llvm::BumpPtrAllocator CanonicalNameStorage; 107 108 /// Each FileEntry we create is assigned a unique ID #. 109 /// 110 unsigned NextFileUID; 111 112 // Caching. 113 std::unique_ptr<FileSystemStatCache> StatCache; 114 115 std::error_code getStatValue(StringRef Path, llvm::vfs::Status &Status, 116 bool isFile, 117 std::unique_ptr<llvm::vfs::File> *F); 118 119 /// Add all ancestors of the given path (pointing to either a file 120 /// or a directory) as virtual directories. 121 void addAncestorsAsVirtualDirs(StringRef Path); 122 123 /// Fills the RealPathName in file entry. 124 void fillRealPathName(FileEntry *UFE, llvm::StringRef FileName); 125 126 public: 127 /// Construct a file manager, optionally with a custom VFS. 128 /// 129 /// \param FS if non-null, the VFS to use. Otherwise uses 130 /// llvm::vfs::getRealFileSystem(). 131 FileManager(const FileSystemOptions &FileSystemOpts, 132 IntrusiveRefCntPtr<llvm::vfs::FileSystem> FS = nullptr); 133 ~FileManager(); 134 135 /// Installs the provided FileSystemStatCache object within 136 /// the FileManager. 137 /// 138 /// Ownership of this object is transferred to the FileManager. 139 /// 140 /// \param statCache the new stat cache to install. Ownership of this 141 /// object is transferred to the FileManager. 142 void setStatCache(std::unique_ptr<FileSystemStatCache> statCache); 143 144 /// Removes the FileSystemStatCache object from the manager. 145 void clearStatCache(); 146 147 /// Returns the number of unique real file entries cached by the file manager. getNumUniqueRealFiles()148 size_t getNumUniqueRealFiles() const { return UniqueRealFiles.size(); } 149 150 /// Lookup, cache, and verify the specified directory (real or 151 /// virtual). 152 /// 153 /// This returns a \c std::error_code if there was an error reading the 154 /// directory. On success, returns the reference to the directory entry 155 /// together with the exact path that was used to access a file by a 156 /// particular call to getDirectoryRef. 157 /// 158 /// \param CacheFailure If true and the file does not exist, we'll cache 159 /// the failure to find this file. 160 llvm::Expected<DirectoryEntryRef> getDirectoryRef(StringRef DirName, 161 bool CacheFailure = true); 162 163 /// Get a \c DirectoryEntryRef if it exists, without doing anything on error. 164 llvm::Optional<DirectoryEntryRef> 165 getOptionalDirectoryRef(StringRef DirName, bool CacheFailure = true) { 166 return llvm::expectedToOptional(getDirectoryRef(DirName, CacheFailure)); 167 } 168 169 /// Lookup, cache, and verify the specified directory (real or 170 /// virtual). 171 /// 172 /// This function is deprecated and will be removed at some point in the 173 /// future, new clients should use 174 /// \c getDirectoryRef. 175 /// 176 /// This returns a \c std::error_code if there was an error reading the 177 /// directory. If there is no error, the DirectoryEntry is guaranteed to be 178 /// non-NULL. 179 /// 180 /// \param CacheFailure If true and the file does not exist, we'll cache 181 /// the failure to find this file. 182 llvm::ErrorOr<const DirectoryEntry *> 183 getDirectory(StringRef DirName, bool CacheFailure = true); 184 185 /// Lookup, cache, and verify the specified file (real or 186 /// virtual). 187 /// 188 /// This function is deprecated and will be removed at some point in the 189 /// future, new clients should use 190 /// \c getFileRef. 191 /// 192 /// This returns a \c std::error_code if there was an error loading the file. 193 /// If there is no error, the FileEntry is guaranteed to be non-NULL. 194 /// 195 /// \param OpenFile if true and the file exists, it will be opened. 196 /// 197 /// \param CacheFailure If true and the file does not exist, we'll cache 198 /// the failure to find this file. 199 llvm::ErrorOr<const FileEntry *> 200 getFile(StringRef Filename, bool OpenFile = false, bool CacheFailure = true); 201 202 /// Lookup, cache, and verify the specified file (real or virtual). Return the 203 /// reference to the file entry together with the exact path that was used to 204 /// access a file by a particular call to getFileRef. If the underlying VFS is 205 /// a redirecting VFS that uses external file names, the returned FileEntryRef 206 /// will use the external name instead of the filename that was passed to this 207 /// method. 208 /// 209 /// This returns a \c std::error_code if there was an error loading the file, 210 /// or a \c FileEntryRef otherwise. 211 /// 212 /// \param OpenFile if true and the file exists, it will be opened. 213 /// 214 /// \param CacheFailure If true and the file does not exist, we'll cache 215 /// the failure to find this file. 216 llvm::Expected<FileEntryRef> getFileRef(StringRef Filename, 217 bool OpenFile = false, 218 bool CacheFailure = true); 219 220 /// Get a FileEntryRef if it exists, without doing anything on error. 221 llvm::Optional<FileEntryRef> getOptionalFileRef(StringRef Filename, 222 bool OpenFile = false, 223 bool CacheFailure = true) { 224 return llvm::expectedToOptional( 225 getFileRef(Filename, OpenFile, CacheFailure)); 226 } 227 228 /// Returns the current file system options getFileSystemOpts()229 FileSystemOptions &getFileSystemOpts() { return FileSystemOpts; } getFileSystemOpts()230 const FileSystemOptions &getFileSystemOpts() const { return FileSystemOpts; } 231 getVirtualFileSystem()232 llvm::vfs::FileSystem &getVirtualFileSystem() const { return *FS; } 233 setVirtualFileSystem(IntrusiveRefCntPtr<llvm::vfs::FileSystem> FS)234 void setVirtualFileSystem(IntrusiveRefCntPtr<llvm::vfs::FileSystem> FS) { 235 this->FS = std::move(FS); 236 } 237 238 /// Retrieve a file entry for a "virtual" file that acts as 239 /// if there were a file with the given name on disk. 240 /// 241 /// The file itself is not accessed. 242 FileEntryRef getVirtualFileRef(StringRef Filename, off_t Size, 243 time_t ModificationTime); 244 245 const FileEntry *getVirtualFile(StringRef Filename, off_t Size, 246 time_t ModificationTime); 247 248 /// Retrieve a FileEntry that bypasses VFE, which is expected to be a virtual 249 /// file entry, to access the real file. The returned FileEntry will have 250 /// the same filename as FE but a different identity and its own stat. 251 /// 252 /// This should be used only for rare error recovery paths because it 253 /// bypasses all mapping and uniquing, blindly creating a new FileEntry. 254 /// There is no attempt to deduplicate these; if you bypass the same file 255 /// twice, you get two new file entries. 256 llvm::Optional<FileEntryRef> getBypassFile(FileEntryRef VFE); 257 258 /// Open the specified file as a MemoryBuffer, returning a new 259 /// MemoryBuffer if successful, otherwise returning null. 260 llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>> 261 getBufferForFile(const FileEntry *Entry, bool isVolatile = false, 262 bool RequiresNullTerminator = true); 263 llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>> 264 getBufferForFile(StringRef Filename, bool isVolatile = false, 265 bool RequiresNullTerminator = true) { 266 return getBufferForFileImpl(Filename, /*FileSize=*/-1, isVolatile, 267 RequiresNullTerminator); 268 } 269 270 private: 271 llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>> 272 getBufferForFileImpl(StringRef Filename, int64_t FileSize, bool isVolatile, 273 bool RequiresNullTerminator); 274 275 public: 276 /// Get the 'stat' information for the given \p Path. 277 /// 278 /// If the path is relative, it will be resolved against the WorkingDir of the 279 /// FileManager's FileSystemOptions. 280 /// 281 /// \returns a \c std::error_code describing an error, if there was one 282 std::error_code getNoncachedStatValue(StringRef Path, 283 llvm::vfs::Status &Result); 284 285 /// If path is not absolute and FileSystemOptions set the working 286 /// directory, the path is modified to be relative to the given 287 /// working directory. 288 /// \returns true if \c path changed. 289 bool FixupRelativePath(SmallVectorImpl<char> &path) const; 290 291 /// Makes \c Path absolute taking into account FileSystemOptions and the 292 /// working directory option. 293 /// \returns true if \c Path changed to absolute. 294 bool makeAbsolutePath(SmallVectorImpl<char> &Path) const; 295 296 /// Produce an array mapping from the unique IDs assigned to each 297 /// file to the corresponding FileEntry pointer. 298 void GetUniqueIDMapping( 299 SmallVectorImpl<const FileEntry *> &UIDToFiles) const; 300 301 /// Retrieve the canonical name for a given directory. 302 /// 303 /// This is a very expensive operation, despite its results being cached, 304 /// and should only be used when the physical layout of the file system is 305 /// required, which is (almost) never. 306 StringRef getCanonicalName(const DirectoryEntry *Dir); 307 308 /// Retrieve the canonical name for a given file. 309 /// 310 /// This is a very expensive operation, despite its results being cached, 311 /// and should only be used when the physical layout of the file system is 312 /// required, which is (almost) never. 313 StringRef getCanonicalName(const FileEntry *File); 314 315 void PrintStats() const; 316 }; 317 318 } // end namespace clang 319 320 #endif // LLVM_CLANG_BASIC_FILEMANAGER_H 321