rn-leveldb 3.11.0

package/cpp/leveldb/db/skiplist_test.cc

@@ -0,0 +1,373 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "db/skiplist.h"
+
+#include <atomic>
+#include <set>
+
+#include "gtest/gtest.h"
+#include "leveldb/env.h"
+#include "port/port.h"
+#include "port/thread_annotations.h"
+#include "util/arena.h"
+#include "util/hash.h"
+#include "util/random.h"
+#include "util/testutil.h"
+
+namespace leveldb {
+
+typedef uint64_t Key;
+
+struct Comparator {
+  int operator()(const Key& a, const Key& b) const {
+    if (a < b) {
+      return -1;
+    } else if (a > b) {
+      return +1;
+    } else {
+      return 0;
+    }
+  }
+};
+
+TEST(SkipTest, Empty) {
+  Arena arena;
+  Comparator cmp;
+  SkipList<Key, Comparator> list(cmp, &arena);
+  ASSERT_TRUE(!list.Contains(10));
+
+  SkipList<Key, Comparator>::Iterator iter(&list);
+  ASSERT_TRUE(!iter.Valid());
+  iter.SeekToFirst();
+  ASSERT_TRUE(!iter.Valid());
+  iter.Seek(100);
+  ASSERT_TRUE(!iter.Valid());
+  iter.SeekToLast();
+  ASSERT_TRUE(!iter.Valid());
+}
+
+TEST(SkipTest, InsertAndLookup) {
+  const int N = 2000;
+  const int R = 5000;
+  Random rnd(1000);
+  std::set<Key> keys;
+  Arena arena;
+  Comparator cmp;
+  SkipList<Key, Comparator> list(cmp, &arena);
+  for (int i = 0; i < N; i++) {
+    Key key = rnd.Next() % R;
+    if (keys.insert(key).second) {
+      list.Insert(key);
+    }
+  }
+
+  for (int i = 0; i < R; i++) {
+    if (list.Contains(i)) {
+      ASSERT_EQ(keys.count(i), 1);
+    } else {
+      ASSERT_EQ(keys.count(i), 0);
+    }
+  }
+
+  // Simple iterator tests
+  {
+    SkipList<Key, Comparator>::Iterator iter(&list);
+    ASSERT_TRUE(!iter.Valid());
+
+    iter.Seek(0);
+    ASSERT_TRUE(iter.Valid());
+    ASSERT_EQ(*(keys.begin()), iter.key());
+
+    iter.SeekToFirst();
+    ASSERT_TRUE(iter.Valid());
+    ASSERT_EQ(*(keys.begin()), iter.key());
+
+    iter.SeekToLast();
+    ASSERT_TRUE(iter.Valid());
+    ASSERT_EQ(*(keys.rbegin()), iter.key());
+  }
+
+  // Forward iteration test
+  for (int i = 0; i < R; i++) {
+    SkipList<Key, Comparator>::Iterator iter(&list);
+    iter.Seek(i);
+
+    // Compare against model iterator
+    std::set<Key>::iterator model_iter = keys.lower_bound(i);
+    for (int j = 0; j < 3; j++) {
+      if (model_iter == keys.end()) {
+        ASSERT_TRUE(!iter.Valid());
+        break;
+      } else {
+        ASSERT_TRUE(iter.Valid());
+        ASSERT_EQ(*model_iter, iter.key());
+        ++model_iter;
+        iter.Next();
+      }
+    }
+  }
+
+  // Backward iteration test
+  {
+    SkipList<Key, Comparator>::Iterator iter(&list);
+    iter.SeekToLast();
+
+    // Compare against model iterator
+    for (std::set<Key>::reverse_iterator model_iter = keys.rbegin();
+         model_iter != keys.rend(); ++model_iter) {
+      ASSERT_TRUE(iter.Valid());
+      ASSERT_EQ(*model_iter, iter.key());
+      iter.Prev();
+    }
+    ASSERT_TRUE(!iter.Valid());
+  }
+}
+
+// We want to make sure that with a single writer and multiple
+// concurrent readers (with no synchronization other than when a
+// reader's iterator is created), the reader always observes all the
+// data that was present in the skip list when the iterator was
+// constructed.  Because insertions are happening concurrently, we may
+// also observe new values that were inserted since the iterator was
+// constructed, but we should never miss any values that were present
+// at iterator construction time.
+//
+// We generate multi-part keys:
+//     <key,gen,hash>
+// where:
+//     key is in range [0..K-1]
+//     gen is a generation number for key
+//     hash is hash(key,gen)
+//
+// The insertion code picks a random key, sets gen to be 1 + the last
+// generation number inserted for that key, and sets hash to Hash(key,gen).
+//
+// At the beginning of a read, we snapshot the last inserted
+// generation number for each key.  We then iterate, including random
+// calls to Next() and Seek().  For every key we encounter, we
+// check that it is either expected given the initial snapshot or has
+// been concurrently added since the iterator started.
+class ConcurrentTest {
+ private:
+  static constexpr uint32_t K = 4;
+
+  static uint64_t key(Key key) { return (key >> 40); }
+  static uint64_t gen(Key key) { return (key >> 8) & 0xffffffffu; }
+  static uint64_t hash(Key key) { return key & 0xff; }
+
+  static uint64_t HashNumbers(uint64_t k, uint64_t g) {
+    uint64_t data[2] = {k, g};
+    return Hash(reinterpret_cast<char*>(data), sizeof(data), 0);
+  }
+
+  static Key MakeKey(uint64_t k, uint64_t g) {
+    static_assert(sizeof(Key) == sizeof(uint64_t), "");
+    assert(k <= K);  // We sometimes pass K to seek to the end of the skiplist
+    assert(g <= 0xffffffffu);
+    return ((k << 40) | (g << 8) | (HashNumbers(k, g) & 0xff));
+  }
+
+  static bool IsValidKey(Key k) {
+    return hash(k) == (HashNumbers(key(k), gen(k)) & 0xff);
+  }
+
+  static Key RandomTarget(Random* rnd) {
+    switch (rnd->Next() % 10) {
+      case 0:
+        // Seek to beginning
+        return MakeKey(0, 0);
+      case 1:
+        // Seek to end
+        return MakeKey(K, 0);
+      default:
+        // Seek to middle
+        return MakeKey(rnd->Next() % K, 0);
+    }
+  }
+
+  // Per-key generation
+  struct State {
+    std::atomic<int> generation[K];
+    void Set(int k, int v) {
+      generation[k].store(v, std::memory_order_release);
+    }
+    int Get(int k) { return generation[k].load(std::memory_order_acquire); }
+
+    State() {
+      for (int k = 0; k < K; k++) {
+        Set(k, 0);
+      }
+    }
+  };
+
+  // Current state of the test
+  State current_;
+
+  Arena arena_;
+
+  // SkipList is not protected by mu_.  We just use a single writer
+  // thread to modify it.
+  SkipList<Key, Comparator> list_;
+
+ public:
+  ConcurrentTest() : list_(Comparator(), &arena_) {}
+
+  // REQUIRES: External synchronization
+  void WriteStep(Random* rnd) {
+    const uint32_t k = rnd->Next() % K;
+    const intptr_t g = current_.Get(k) + 1;
+    const Key key = MakeKey(k, g);
+    list_.Insert(key);
+    current_.Set(k, g);
+  }
+
+  void ReadStep(Random* rnd) {
+    // Remember the initial committed state of the skiplist.
+    State initial_state;
+    for (int k = 0; k < K; k++) {
+      initial_state.Set(k, current_.Get(k));
+    }
+
+    Key pos = RandomTarget(rnd);
+    SkipList<Key, Comparator>::Iterator iter(&list_);
+    iter.Seek(pos);
+    while (true) {
+      Key current;
+      if (!iter.Valid()) {
+        current = MakeKey(K, 0);
+      } else {
+        current = iter.key();
+        ASSERT_TRUE(IsValidKey(current)) << current;
+      }
+      ASSERT_LE(pos, current) << "should not go backwards";
+
+      // Verify that everything in [pos,current) was not present in
+      // initial_state.
+      while (pos < current) {
+        ASSERT_LT(key(pos), K) << pos;
+
+        // Note that generation 0 is never inserted, so it is ok if
+        // <*,0,*> is missing.
+        ASSERT_TRUE((gen(pos) == 0) ||
+                    (gen(pos) > static_cast<Key>(initial_state.Get(key(pos)))))
+            << "key: " << key(pos) << "; gen: " << gen(pos)
+            << "; initgen: " << initial_state.Get(key(pos));
+
+        // Advance to next key in the valid key space
+        if (key(pos) < key(current)) {
+          pos = MakeKey(key(pos) + 1, 0);
+        } else {
+          pos = MakeKey(key(pos), gen(pos) + 1);
+        }
+      }
+
+      if (!iter.Valid()) {
+        break;
+      }
+
+      if (rnd->Next() % 2) {
+        iter.Next();
+        pos = MakeKey(key(pos), gen(pos) + 1);
+      } else {
+        Key new_target = RandomTarget(rnd);
+        if (new_target > pos) {
+          pos = new_target;
+          iter.Seek(new_target);
+        }
+      }
+    }
+  }
+};
+
+// Needed when building in C++11 mode.
+constexpr uint32_t ConcurrentTest::K;
+
+// Simple test that does single-threaded testing of the ConcurrentTest
+// scaffolding.
+TEST(SkipTest, ConcurrentWithoutThreads) {
+  ConcurrentTest test;
+  Random rnd(test::RandomSeed());
+  for (int i = 0; i < 10000; i++) {
+    test.ReadStep(&rnd);
+    test.WriteStep(&rnd);
+  }
+}
+
+class TestState {
+ public:
+  ConcurrentTest t_;
+  int seed_;
+  std::atomic<bool> quit_flag_;
+
+  enum ReaderState { STARTING, RUNNING, DONE };
+
+  explicit TestState(int s)
+      : seed_(s), quit_flag_(false), state_(STARTING), state_cv_(&mu_) {}
+
+  void Wait(ReaderState s) LOCKS_EXCLUDED(mu_) {
+    mu_.Lock();
+    while (state_ != s) {
+      state_cv_.Wait();
+    }
+    mu_.Unlock();
+  }
+
+  void Change(ReaderState s) LOCKS_EXCLUDED(mu_) {
+    mu_.Lock();
+    state_ = s;
+    state_cv_.Signal();
+    mu_.Unlock();
+  }
+
+ private:
+  port::Mutex mu_;
+  ReaderState state_ GUARDED_BY(mu_);
+  port::CondVar state_cv_ GUARDED_BY(mu_);
+};
+
+static void ConcurrentReader(void* arg) {
+  TestState* state = reinterpret_cast<TestState*>(arg);
+  Random rnd(state->seed_);
+  int64_t reads = 0;
+  state->Change(TestState::RUNNING);
+  while (!state->quit_flag_.load(std::memory_order_acquire)) {
+    state->t_.ReadStep(&rnd);
+    ++reads;
+  }
+  state->Change(TestState::DONE);
+}
+
+static void RunConcurrent(int run) {
+  const int seed = test::RandomSeed() + (run * 100);
+  Random rnd(seed);
+  const int N = 1000;
+  const int kSize = 1000;
+  for (int i = 0; i < N; i++) {
+    if ((i % 100) == 0) {
+      std::fprintf(stderr, "Run %d of %d\n", i, N);
+    }
+    TestState state(seed + 1);
+    Env::Default()->Schedule(ConcurrentReader, &state);
+    state.Wait(TestState::RUNNING);
+    for (int i = 0; i < kSize; i++) {
+      state.t_.WriteStep(&rnd);
+    }
+    state.quit_flag_.store(true, std::memory_order_release);
+    state.Wait(TestState::DONE);
+  }
+}
+
+TEST(SkipTest, Concurrent1) { RunConcurrent(1); }
+TEST(SkipTest, Concurrent2) { RunConcurrent(2); }
+TEST(SkipTest, Concurrent3) { RunConcurrent(3); }
+TEST(SkipTest, Concurrent4) { RunConcurrent(4); }
+TEST(SkipTest, Concurrent5) { RunConcurrent(5); }
+
+}  // namespace leveldb
+
+int main(int argc, char** argv) {
+  testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}

package/cpp/leveldb/db/snapshot.h

@@ -0,0 +1,95 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#ifndef STORAGE_LEVELDB_DB_SNAPSHOT_H_
+#define STORAGE_LEVELDB_DB_SNAPSHOT_H_
+
+#include "db/dbformat.h"
+#include "leveldb/db.h"
+
+namespace leveldb {
+
+class SnapshotList;
+
+// Snapshots are kept in a doubly-linked list in the DB.
+// Each SnapshotImpl corresponds to a particular sequence number.
+class SnapshotImpl : public Snapshot {
+ public:
+  SnapshotImpl(SequenceNumber sequence_number)
+      : sequence_number_(sequence_number) {}
+
+  SequenceNumber sequence_number() const { return sequence_number_; }
+
+ private:
+  friend class SnapshotList;
+
+  // SnapshotImpl is kept in a doubly-linked circular list. The SnapshotList
+  // implementation operates on the next/previous fields direcly.
+  SnapshotImpl* prev_;
+  SnapshotImpl* next_;
+
+  const SequenceNumber sequence_number_;
+
+#if !defined(NDEBUG)
+  SnapshotList* list_ = nullptr;
+#endif  // !defined(NDEBUG)
+};
+
+class SnapshotList {
+ public:
+  SnapshotList() : head_(0) {
+    head_.prev_ = &head_;
+    head_.next_ = &head_;
+  }
+
+  bool empty() const { return head_.next_ == &head_; }
+  SnapshotImpl* oldest() const {
+    assert(!empty());
+    return head_.next_;
+  }
+  SnapshotImpl* newest() const {
+    assert(!empty());
+    return head_.prev_;
+  }
+
+  // Creates a SnapshotImpl and appends it to the end of the list.
+  SnapshotImpl* New(SequenceNumber sequence_number) {
+    assert(empty() || newest()->sequence_number_ <= sequence_number);
+
+    SnapshotImpl* snapshot = new SnapshotImpl(sequence_number);
+
+#if !defined(NDEBUG)
+    snapshot->list_ = this;
+#endif  // !defined(NDEBUG)
+    snapshot->next_ = &head_;
+    snapshot->prev_ = head_.prev_;
+    snapshot->prev_->next_ = snapshot;
+    snapshot->next_->prev_ = snapshot;
+    return snapshot;
+  }
+
+  // Removes a SnapshotImpl from this list.
+  //
+  // The snapshot must have been created by calling New() on this list.
+  //
+  // The snapshot pointer should not be const, because its memory is
+  // deallocated. However, that would force us to change DB::ReleaseSnapshot(),
+  // which is in the API, and currently takes a const Snapshot.
+  void Delete(const SnapshotImpl* snapshot) {
+#if !defined(NDEBUG)
+    assert(snapshot->list_ == this);
+#endif  // !defined(NDEBUG)
+    snapshot->prev_->next_ = snapshot->next_;
+    snapshot->next_->prev_ = snapshot->prev_;
+    delete snapshot;
+  }
+
+ private:
+  // Dummy head of doubly-linked list of snapshots
+  SnapshotImpl head_;
+};
+
+}  // namespace leveldb
+
+#endif  // STORAGE_LEVELDB_DB_SNAPSHOT_H_

package/cpp/leveldb/db/table_cache.cc

@@ -0,0 +1,120 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "db/table_cache.h"
+
+#include "db/filename.h"
+#include "leveldb/env.h"
+#include "leveldb/table.h"
+#include "util/coding.h"
+
+namespace leveldb {
+
+struct TableAndFile {
+  RandomAccessFile* file;
+  Table* table;
+};
+
+static void DeleteEntry(const Slice& key, void* value) {
+  TableAndFile* tf = reinterpret_cast<TableAndFile*>(value);
+  delete tf->table;
+  delete tf->file;
+  delete tf;
+}
+
+static void UnrefEntry(void* arg1, void* arg2) {
+  Cache* cache = reinterpret_cast<Cache*>(arg1);
+  Cache::Handle* h = reinterpret_cast<Cache::Handle*>(arg2);
+  cache->Release(h);
+}
+
+TableCache::TableCache(const std::string& dbname, const Options& options,
+                       int entries)
+    : env_(options.env),
+      dbname_(dbname),
+      options_(options),
+      cache_(NewLRUCache(entries)) {}
+
+TableCache::~TableCache() { delete cache_; }
+
+Status TableCache::FindTable(uint64_t file_number, uint64_t file_size,
+                             Cache::Handle** handle) {
+  Status s;
+  char buf[sizeof(file_number)];
+  EncodeFixed64(buf, file_number);
+  Slice key(buf, sizeof(buf));
+  *handle = cache_->Lookup(key);
+  if (*handle == nullptr) {
+    std::string fname = TableFileName(dbname_, file_number);
+    RandomAccessFile* file = nullptr;
+    Table* table = nullptr;
+    s = env_->NewRandomAccessFile(fname, &file);
+    if (!s.ok()) {
+      std::string old_fname = SSTTableFileName(dbname_, file_number);
+      if (env_->NewRandomAccessFile(old_fname, &file).ok()) {
+        s = Status::OK();
+      }
+    }
+    if (s.ok()) {
+      s = Table::Open(options_, file, file_size, &table);
+    }
+
+    if (!s.ok()) {
+      assert(table == nullptr);
+      delete file;
+      // We do not cache error results so that if the error is transient,
+      // or somebody repairs the file, we recover automatically.
+    } else {
+      TableAndFile* tf = new TableAndFile;
+      tf->file = file;
+      tf->table = table;
+      *handle = cache_->Insert(key, tf, 1, &DeleteEntry);
+    }
+  }
+  return s;
+}
+
+Iterator* TableCache::NewIterator(const ReadOptions& options,
+                                  uint64_t file_number, uint64_t file_size,
+                                  Table** tableptr) {
+  if (tableptr != nullptr) {
+    *tableptr = nullptr;
+  }
+
+  Cache::Handle* handle = nullptr;
+  Status s = FindTable(file_number, file_size, &handle);
+  if (!s.ok()) {
+    return NewErrorIterator(s);
+  }
+
+  Table* table = reinterpret_cast<TableAndFile*>(cache_->Value(handle))->table;
+  Iterator* result = table->NewIterator(options);
+  result->RegisterCleanup(&UnrefEntry, cache_, handle);
+  if (tableptr != nullptr) {
+    *tableptr = table;
+  }
+  return result;
+}
+
+Status TableCache::Get(const ReadOptions& options, uint64_t file_number,
+                       uint64_t file_size, const Slice& k, void* arg,
+                       void (*handle_result)(void*, const Slice&,
+                                             const Slice&)) {
+  Cache::Handle* handle = nullptr;
+  Status s = FindTable(file_number, file_size, &handle);
+  if (s.ok()) {
+    Table* t = reinterpret_cast<TableAndFile*>(cache_->Value(handle))->table;
+    s = t->InternalGet(options, k, arg, handle_result);
+    cache_->Release(handle);
+  }
+  return s;
+}
+
+void TableCache::Evict(uint64_t file_number) {
+  char buf[sizeof(file_number)];
+  EncodeFixed64(buf, file_number);
+  cache_->Erase(Slice(buf, sizeof(buf)));
+}
+
+}  // namespace leveldb

package/cpp/leveldb/db/table_cache.h

@@ -0,0 +1,57 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+//
+// Thread-safe (provides internal synchronization)
+
+#ifndef STORAGE_LEVELDB_DB_TABLE_CACHE_H_
+#define STORAGE_LEVELDB_DB_TABLE_CACHE_H_
+
+#include <cstdint>
+#include <string>
+
+#include "db/dbformat.h"
+#include "leveldb/cache.h"
+#include "leveldb/table.h"
+#include "port/port.h"
+
+namespace leveldb {
+
+class Env;
+
+class TableCache {
+ public:
+  TableCache(const std::string& dbname, const Options& options, int entries);
+  ~TableCache();
+
+  // Return an iterator for the specified file number (the corresponding
+  // file length must be exactly "file_size" bytes).  If "tableptr" is
+  // non-null, also sets "*tableptr" to point to the Table object
+  // underlying the returned iterator, or to nullptr if no Table object
+  // underlies the returned iterator.  The returned "*tableptr" object is owned
+  // by the cache and should not be deleted, and is valid for as long as the
+  // returned iterator is live.
+  Iterator* NewIterator(const ReadOptions& options, uint64_t file_number,
+                        uint64_t file_size, Table** tableptr = nullptr);
+
+  // If a seek to internal key "k" in specified file finds an entry,
+  // call (*handle_result)(arg, found_key, found_value).
+  Status Get(const ReadOptions& options, uint64_t file_number,
+             uint64_t file_size, const Slice& k, void* arg,
+             void (*handle_result)(void*, const Slice&, const Slice&));
+
+  // Evict any entry for the specified file number
+  void Evict(uint64_t file_number);
+
+ private:
+  Status FindTable(uint64_t file_number, uint64_t file_size, Cache::Handle**);
+
+  Env* const env_;
+  const std::string dbname_;
+  const Options& options_;
+  Cache* cache_;
+};
+
+}  // namespace leveldb
+
+#endif  // STORAGE_LEVELDB_DB_TABLE_CACHE_H_