[clangd] Store paths as requested in PreambleStatCache

[kadircet]

Summary

Underlying FS can store different file names inside the stat response
(e.g. symlinks resolved, absolute paths, dots removed). But we store path names
as requested inside the preamble,
https://github.com/llvm/llvm-project/blob/main/clang/lib/Serialization/ASTWriter.cpp#L1635.

This improves cache hit rates from ~30% to 90% in a build system that uses
symlinks.

[hokein]

good catch!

clang-tools-extra/clangd/FS.h

 ///
 /// Note that the cache is only valid when reusing preamble.
 class PreambleFileStatusCache {
 public:
   /// \p MainFilePath is the absolute path of the main source file this preamble
   /// corresponds to. The stat for the main file will not be cached.
   PreambleFileStatusCache(llvm::StringRef MainFilePath);
 
-  void update(const llvm::vfs::FileSystem &FS, llvm::vfs::Status S);
+  void update(const llvm::vfs::FileSystem &FS, llvm::vfs::Status S,
+              llvm::StringRef File);
 
   /// \p Path is a path stored in preamble.
   std::optional<llvm::vfs::Status> lookup(llvm::StringRef Path) const;
 
   /// Returns a VFS that collects file status.
   /// Only cache stats for files that exist because
   ///   1) we only care about existing files when reusing preamble, unlike
   ///   building preamble.

clang-tools-extra/clangd/FS.cpp

 //===--- FS.cpp - File system related utils ----------------------*- C++-*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #include "FS.h"
 #include "clang/Basic/LLVM.h"
 #include "llvm/Support/Path.h"
 #include "llvm/Support/VirtualFileSystem.h"
 #include <optional>
+#include <utility>
 
 namespace clang {
 namespace clangd {
 
 PreambleFileStatusCache::PreambleFileStatusCache(llvm::StringRef MainFilePath){
   assert(llvm::sys::path::is_absolute(MainFilePath));
   llvm::SmallString<256> MainFileCanonical(MainFilePath);
   llvm::sys::path::remove_dots(MainFileCanonical, /*remove_dot_dot=*/true);
   this->MainFilePath = std::string(MainFileCanonical.str());
 }
 
 void PreambleFileStatusCache::update(const llvm::vfs::FileSystem &FS,
-                                     llvm::vfs::Status S) {
+                                     llvm::vfs::Status S,
+                                     llvm::StringRef File) {
   // Canonicalize path for later lookup, which is usually by absolute path.
-  llvm::SmallString<32> PathStore(S.getName());
+  llvm::SmallString<32> PathStore(File);
   if (FS.makeAbsolute(PathStore))
     return;
   llvm::sys::path::remove_dots(PathStore, /*remove_dot_dot=*/true);
   // Do not cache status for the main file.
   if (PathStore == MainFilePath)
     return;
   // Stores the latest status in cache as it can change in a preamble build.
   StatCache.insert({PathStore, std::move(S)});
openFileForRead(const llvm::Twine &Path) override {
       if (!File || !*File)
         return File;
       // Eagerly stat opened file, as the followup `status` call on the file
       // doesn't necessarily go through this FS. This puts some extra work on
       // preamble build, but it should be worth it as preamble can be reused
       // many times (e.g. code completion) and the repeated status call is
       // likely to be cached in the underlying file system anyway.
       if (auto S = File->get()->status())
-        StatCache.update(getUnderlyingFS(), std::move(*S));
+        StatCache.update(getUnderlyingFS(), std::move(*S), Path.str());
       return File;
     }
 
     llvm::ErrorOr<llvm::vfs::Status> status(const llvm::Twine &Path) override {
       auto S = getUnderlyingFS().status(Path);
       if (S)
-        StatCache.update(getUnderlyingFS(), *S);
+        StatCache.update(getUnderlyingFS(), *S, Path.str());
       return S;
     }
 
   private:
     PreambleFileStatusCache &StatCache;
   };
   return llvm::IntrusiveRefCntPtr<CollectFS>(
       new CollectFS(std::move(FS), *this));

clang-tools-extra/clangd/unittests/FSTests.cpp

TEST(FSTests, PreambleStatusCache) {
   EXPECT_TRUE(StatCache.lookup(testPath("y")).has_value());
   // Main file is not cached.
   EXPECT_FALSE(StatCache.lookup(testPath("main")).has_value());
 
   llvm::vfs::Status S("fake", llvm::sys::fs::UniqueID(123, 456),
                       std::chrono::system_clock::now(), 0, 0, 1024,
                       llvm::sys::fs::file_type::regular_file,
                       llvm::sys::fs::all_all);
-  StatCache.update(*FS, S);
+  StatCache.update(*FS, S, "real");
   auto ConsumeFS = StatCache.getConsumingFS(FS);
-  auto Cached = ConsumeFS->status(testPath("fake"));
+  EXPECT_FALSE(ConsumeFS->status(testPath("fake")));
+  auto Cached = ConsumeFS->status(testPath("real"));
   EXPECT_TRUE(Cached);
-  EXPECT_EQ(Cached->getName(), testPath("fake"));
+  EXPECT_EQ(Cached->getName(), testPath("real"));
   EXPECT_EQ(Cached->getUniqueID(), S.getUniqueID());
 
-  // fake and temp/../fake should hit the same cache entry.
+  // real and temp/../real should hit the same cache entry.
   // However, the Status returned reflects the actual path requested.
-  auto CachedDotDot = ConsumeFS->status(testPath("temp/../fake"));
+  auto CachedDotDot = ConsumeFS->status(testPath("temp/../real"));
   EXPECT_TRUE(CachedDotDot);
-  EXPECT_EQ(CachedDotDot->getName(), testPath("temp/../fake"));
+  EXPECT_EQ(CachedDotDot->getName(), testPath("temp/../real"));
   EXPECT_EQ(CachedDotDot->getUniqueID(), S.getUniqueID());
 }
 
 } // namespace
 } // namespace clangd
 } // namespace clang