Updating README.fb to have newest verison 2.4

Summary:

Test Plan: visual

.gitignore

@@ -21,3 +21,4 @@ util/build_version.cc
 build_tools/VALGRIND_LOGS/
 coverage/COVERAGE_REPORT
 util/build_version.cc.tmp
+.gdbhistory

Makefile

@@ -14,7 +14,7 @@ OPT += -O2 -fno-omit-frame-pointer -momit-leaf-frame-pointer
 #-----------------------------------------------
 
 # detect what platform we're building on
-$(shell (cd build_tools/; ROCKSDB_ROOT=.. ./build_detect_platform ../build_config.mk))
+$(shell (export ROCKSDB_ROOT=$(CURDIR); $(CURDIR)/build_tools/build_detect_platform $(CURDIR)/build_config.mk))
 # this file is generated by the previous line to set build flags and sources
 include build_config.mk
 
@@ -64,7 +64,8 @@ TESTS = \
 	ttl_test \
 	version_edit_test \
 	version_set_test \
-	write_batch_test
+	write_batch_test\
+	deletefile_test
 
 TOOLS = \
         sst_dump \
@@ -121,7 +122,7 @@ release:
 
 coverage:
 	$(MAKE) clean
-	COVERAGEFLAGS="-fprofile-arcs -ftest-coverage" $(MAKE) all check
+	COVERAGEFLAGS="-fprofile-arcs -ftest-coverage" LDFLAGS+="-lgcov" $(MAKE) all check
 	(cd coverage; ./coverage_test.sh)
 	# Delete intermediate files
 	find . -type f -regex ".*\.\(\(gcda\)\|\(gcno\)\)" | xargs --no-run-if-empty rm
@@ -266,6 +267,9 @@ write_batch_test: db/write_batch_test.o $(LIBOBJECTS) $(TESTHARNESS)
 merge_test: db/merge_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(CXX) db/merge_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)
 
+deletefile_test: db/deletefile_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) db/deletefile_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+
 $(MEMENVLIBRARY) : $(MEMENVOBJECTS)
 	rm -f $@
 	$(AR) -rs $@ $(MEMENVOBJECTS)

README

@@ -1,6 +1,6 @@
 rocksdb: A persistent key-value store for flash storage
 Authors: * The Facebook Database Engineering Team
-         * Build on earlier work on leveldb by Sanjay Ghemawat 
+         * Build on earlier work on leveldb by Sanjay Ghemawat
            (sanjay@google.com) and Jeff Dean (jeff@google.com)
 
 This code is a library that forms the core building block for a fast
@@ -19,64 +19,64 @@ persistent key/value store.
 See doc/index.html for more explanation.
 See doc/impl.html for a brief overview of the implementation.
 
-The public interface is in include/*.h.  Callers should not include or
+The public interface is in include/*.  Callers should not include or
 rely on the details of any other header files in this package.  Those
 internal APIs may be changed without warning.
 
 Guide to header files:
 
-include/db.h
+include/rocksdb/db.h
     Main interface to the DB: Start here
 
-include/options.h
+include/rocksdb/options.h
     Control over the behavior of an entire database, and also
     control over the behavior of individual reads and writes.
 
-include/comparator.h
+include/rocksdb/comparator.h
     Abstraction for user-specified comparison function.  If you want
     just bytewise comparison of keys, you can use the default comparator,
     but clients can write their own comparator implementations if they
     want custom ordering (e.g. to handle different character
     encodings, etc.)
 
-include/iterator.h
+include/rocksdb/iterator.h
     Interface for iterating over data. You can get an iterator
     from a DB object.
 
-include/write_batch.h
+include/rocksdb/write_batch.h
     Interface for atomically applying multiple updates to a database.
 
-include/slice.h
+include/rocksdb/slice.h
     A simple module for maintaining a pointer and a length into some
     other byte array.
 
-include/status.h
+include/rocksdb/status.h
     Status is returned from many of the public interfaces and is used
     to report success and various kinds of errors.
 
-include/env.h
+include/rocksdb/env.h
     Abstraction of the OS environment.  A posix implementation of
     this interface is in util/env_posix.cc
 
-include/table_builder.h
+include/rocksdb/table_builder.h
     Lower-level modules that most clients probably won't use directly
 
-include/cache.h
+include/rocksdb/cache.h
     An API for the block cache.
 
-include/compaction_filter.h
+include/rocksdb/compaction_filter.h
     An API for a application filter invoked on every compaction.
 
-include/filter_policy.h
+include/rocksdb/filter_policy.h
     An API for configuring a bloom filter.
 
-include/memtablerep.h
+include/rocksdb/memtablerep.h
     An API for implementing a memtable.
 
-include/statistics.h
+include/rocksdb/statistics.h
     An API to retrieve various database statistics.
 
-include/transaction_log_iterator.h
+include/rocksdb/transaction_log.h
     An API to retrieve transaction logs from a database.
 
 

README.fb

@@ -1,3 +1,3 @@
 * Detailed instructions on how to compile using fbcode and jemalloc
 
-* Latest release is 2.1.fb
+* Latest release is 2.4.fb

build_tools/build_detect_platform

@@ -31,9 +31,9 @@ fi
 # Default to fbcode gcc on internal fb machines
 if [ -d /mnt/gvfs/third-party -a -z "$CXX" ]; then
     if [ -z "$USE_CLANG" ]; then
-        source ./fbcode.gcc471.sh
+        source $PWD/build_tools/fbcode.gcc471.sh
     else
-        source ./fbcode.clang31.sh
+        source $PWD/build_tools/fbcode.clang31.sh
     fi
 fi
 
@@ -65,7 +65,7 @@ PLATFORM_SHARED_CFLAGS="-fPIC"
 PLATFORM_SHARED_VERSIONED=true
 
 # generic port files (working on all platform by #ifdef) go directly in /port
-GENERIC_PORT_FILES=`cd $ROCKSDB_ROOT; find port -name '*.cc' | tr "\n" " "`
+GENERIC_PORT_FILES=`find $ROCKSDB_ROOT/port -name '*.cc' | tr "\n" " "`
 
 # On GCC, we pick libc's memcmp over GCC's memcmp via -fno-builtin-memcmp
 case "$TARGET_OS" in
@@ -82,7 +82,7 @@ case "$TARGET_OS" in
         if [ -z "$USE_CLANG" ]; then
             COMMON_FLAGS="$COMMON_FLAGS -fno-builtin-memcmp"
         fi
-        PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -lpthread"
+        PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -lpthread -lrt"
         # PORT_FILES=port/linux/linux_specific.cc
         ;;
     SunOS)
@@ -127,7 +127,7 @@ case "$TARGET_OS" in
         exit 1
 esac
 
-./build_detect_version
+$PWD/build_tools/build_detect_version
 
 # We want to make a list of all cc files within util, db, table, and helpers
 # except for the test and benchmark files. By default, find will output a list

build_tools/build_detect_version

@@ -19,15 +19,15 @@ if [ "$?" = 0 ]; then
   awk '
   BEGIN {
     print "#include \"build_version.h\"\n"
-  } 
+  }
   { print "const char* leveldb_build_git_sha = \"leveldb_build_git_sha:" $0"\";" }
   ' >  ${VFILE}
 else
   echo "git not found" |
   awk '
     BEGIN {
-      print "#include \"build_version.h\"" 
-    } 
+      print "#include \"build_version.h\""
+    }
     { print "const char* leveldb_build_git_sha = \"leveldb_build_git_sha:git not found\";" }
     ' > ${VFILE}
 fi

coverage/coverage_test.sh

@@ -55,12 +55,19 @@ then
   exit 0
 fi
 
+LCOV_VERSION=$(lcov -v | grep 1.1 || true)
+if [ $LCOV_VERSION ]
+then
+  echo "Not supported lcov version. Expect lcov 1.1."
+  exit 0
+fi
+
 (cd $ROOT; lcov --no-external \
      --capture  \
      --directory $PWD \
      --gcov-tool $GCOV \
-     --output-file $COVERAGE_DIR/coverage.info &>/dev/null)
+     --output-file $COVERAGE_DIR/coverage.info)
 
-genhtml $COVERAGE_DIR/coverage.info -o $COVERAGE_DIR &>/dev/null
+genhtml $COVERAGE_DIR/coverage.info -o $COVERAGE_DIR
 
 echo "HTML Coverage report is generated in $COVERAGE_DIR"

db/builder.cc

@@ -9,12 +9,12 @@
 #include "db/merge_helper.h"
 #include "db/table_cache.h"
 #include "db/version_edit.h"
-#include "leveldb/db.h"
-#include "leveldb/env.h"
-#include "leveldb/iterator.h"
+#include "rocksdb/db.h"
+#include "rocksdb/env.h"
+#include "rocksdb/iterator.h"
 #include "util/stop_watch.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 Status BuildTable(const std::string& dbname,
                   Env* env,
@@ -204,4 +204,4 @@ Status BuildTable(const std::string& dbname,
   return s;
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb

db/builder.h

@@ -5,11 +5,11 @@
 #ifndef STORAGE_LEVELDB_DB_BUILDER_H_
 #define STORAGE_LEVELDB_DB_BUILDER_H_
 
-#include "leveldb/comparator.h"
-#include "leveldb/status.h"
-#include "leveldb/types.h"
+#include "rocksdb/comparator.h"
+#include "rocksdb/status.h"
+#include "rocksdb/types.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 struct Options;
 struct FileMetaData;
@@ -36,6 +36,6 @@ extern Status BuildTable(const std::string& dbname,
                          const SequenceNumber newest_snapshot,
                          const SequenceNumber earliest_seqno_in_memtable);
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 #endif  // STORAGE_LEVELDB_DB_BUILDER_H_

db/c.cc

@@ -2,42 +2,42 @@
 // 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 "leveldb/c.h"
+#include "rocksdb/c.h"
 
 #include <stdlib.h>
 #include <unistd.h>
-#include "leveldb/cache.h"
-#include "leveldb/comparator.h"
-#include "leveldb/db.h"
-#include "leveldb/env.h"
-#include "leveldb/filter_policy.h"
-#include "leveldb/iterator.h"
-#include "leveldb/options.h"
-#include "leveldb/status.h"
-#include "leveldb/write_batch.h"
+#include "rocksdb/cache.h"
+#include "rocksdb/comparator.h"
+#include "rocksdb/db.h"
+#include "rocksdb/env.h"
+#include "rocksdb/filter_policy.h"
+#include "rocksdb/iterator.h"
+#include "rocksdb/options.h"
+#include "rocksdb/status.h"
+#include "rocksdb/write_batch.h"
 
-using leveldb::Cache;
-using leveldb::Comparator;
-using leveldb::CompressionType;
-using leveldb::DB;
-using leveldb::Env;
-using leveldb::FileLock;
-using leveldb::FilterPolicy;
-using leveldb::Iterator;
-using leveldb::Logger;
-using leveldb::NewBloomFilterPolicy;
-using leveldb::NewLRUCache;
-using leveldb::Options;
-using leveldb::RandomAccessFile;
-using leveldb::Range;
-using leveldb::ReadOptions;
-using leveldb::SequentialFile;
-using leveldb::Slice;
-using leveldb::Snapshot;
-using leveldb::Status;
-using leveldb::WritableFile;
-using leveldb::WriteBatch;
-using leveldb::WriteOptions;
+using rocksdb::Cache;
+using rocksdb::Comparator;
+using rocksdb::CompressionType;
+using rocksdb::DB;
+using rocksdb::Env;
+using rocksdb::FileLock;
+using rocksdb::FilterPolicy;
+using rocksdb::Iterator;
+using rocksdb::Logger;
+using rocksdb::NewBloomFilterPolicy;
+using rocksdb::NewLRUCache;
+using rocksdb::Options;
+using rocksdb::RandomAccessFile;
+using rocksdb::Range;
+using rocksdb::ReadOptions;
+using rocksdb::SequentialFile;
+using rocksdb::Slice;
+using rocksdb::Snapshot;
+using rocksdb::Status;
+using rocksdb::WritableFile;
+using rocksdb::WriteBatch;
+using rocksdb::WriteOptions;
 
 using std::shared_ptr;
 

db/c_test.c

@@ -2,7 +2,7 @@
    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 "leveldb/c.h"
+#include "rocksdb/c.h"
 
 #include <stddef.h>
 #include <stdio.h>

db/corruption_test.cc

@@ -2,15 +2,15 @@
 // 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 "leveldb/db.h"
+#include "rocksdb/db.h"
 
 #include <errno.h>
 #include <fcntl.h>
 #include <sys/stat.h>
 #include <sys/types.h>
-#include "leveldb/cache.h"
-#include "leveldb/env.h"
-#include "leveldb/write_batch.h"
+#include "rocksdb/cache.h"
+#include "rocksdb/env.h"
+#include "rocksdb/write_batch.h"
 #include "db/db_impl.h"
 #include "db/filename.h"
 #include "db/log_format.h"
@@ -20,7 +20,7 @@
 #include "util/testharness.h"
 #include "util/testutil.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 static const int kValueSize = 1000;
 
@@ -68,7 +68,7 @@ class CorruptionTest {
   void RepairDB() {
     delete db_;
     db_ = nullptr;
-    ASSERT_OK(::leveldb::RepairDB(dbname_, options_));
+    ASSERT_OK(::rocksdb::RepairDB(dbname_, options_));
   }
 
   void Build(int n) {
@@ -354,8 +354,8 @@ TEST(CorruptionTest, UnrelatedKeys) {
   ASSERT_EQ(Value(1000, &tmp2).ToString(), v);
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 int main(int argc, char** argv) {
-  return leveldb::test::RunAllTests();
+  return rocksdb::test::RunAllTests();
 }

db/db_filesnapshot.cc

@@ -8,12 +8,12 @@
 #include "db/db_impl.h"
 #include "db/filename.h"
 #include "db/version_set.h"
-#include "leveldb/db.h"
-#include "leveldb/env.h"
+#include "rocksdb/db.h"
+#include "rocksdb/env.h"
 #include "port/port.h"
 #include "util/mutexlock.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 Status DBImpl::DisableFileDeletions() {
   MutexLock l(&mutex_);
@@ -30,16 +30,19 @@ Status DBImpl::EnableFileDeletions() {
 }
 
 Status DBImpl::GetLiveFiles(std::vector<std::string>& ret,
-                            uint64_t* manifest_file_size) {
+                            uint64_t* manifest_file_size,
+                            bool flush_memtable) {
 
   *manifest_file_size = 0;
 
-  // flush all dirty data to disk.
-  Status status =  Flush(FlushOptions());
-  if (!status.ok()) {
-    Log(options_.info_log, "Cannot Flush data %s\n",
-        status.ToString().c_str());
-    return status;
+  if (flush_memtable) {
+    // flush all dirty data to disk.
+    Status status =  Flush(FlushOptions());
+    if (!status.ok()) {
+      Log(options_.info_log, "Cannot Flush data %s\n",
+          status.ToString().c_str());
+      return status;
+    }
   }
 
   MutexLock l(&mutex_);
@@ -48,18 +51,17 @@ Status DBImpl::GetLiveFiles(std::vector<std::string>& ret,
   std::set<uint64_t> live;
   versions_->AddLiveFilesCurrentVersion(&live);
 
-  ret.resize(live.size() + 2); //*.sst + CURRENT + MANIFEST
+  ret.clear();
+  ret.reserve(live.size() + 2); //*.sst + CURRENT + MANIFEST
 
   // create names of the live files. The names are not absolute
   // paths, instead they are relative to dbname_;
-  std::set<uint64_t>::iterator it = live.begin();
-  for (unsigned int i = 0; i < live.size(); i++, it++) {
-    ret[i] = TableFileName("", *it);
+  for (auto live_file : live) {
+    ret.push_back(TableFileName("", live_file));
   }
 
-  ret[live.size()] = CurrentFileName("");
-  ret[live.size()+1] = DescriptorFileName("",
-                                          versions_->ManifestFileNumber());
+  ret.push_back(CurrentFileName(""));
+  ret.push_back(DescriptorFileName("", versions_->ManifestFileNumber()));
 
   // find length of manifest file while holding the mutex lock
   *manifest_file_size = versions_->ManifestFileSize();
@@ -71,7 +73,7 @@ Status DBImpl::GetSortedWalFiles(VectorLogPtr& files) {
   // First get sorted files in archive dir, then append sorted files from main
   // dir to maintain sorted order
 
-  //  list wal files in archive dir.
+  // list wal files in archive dir.
   Status s;
   std::string archivedir = ArchivalDirectory(dbname_);
   if (env_->FileExists(archivedir)) {
@@ -81,11 +83,7 @@ Status DBImpl::GetSortedWalFiles(VectorLogPtr& files) {
     }
   }
   // list wal files in main db dir.
-  s = AppendSortedWalsOfType(dbname_, files, kAliveLogFile);
-  if (!s.ok()) {
-    return s;
-  }
-  return s;
+  return AppendSortedWalsOfType(dbname_, files, kAliveLogFile);
 }
 
 Status DBImpl::DeleteWalFiles(const VectorLogPtr& files) {
@@ -93,14 +91,17 @@ Status DBImpl::DeleteWalFiles(const VectorLogPtr& files) {
   std::string archivedir = ArchivalDirectory(dbname_);
   std::string files_not_deleted;
   for (const auto& wal : files) {
-    /* Try deleting in archive dir. If fails, try deleting in main db dir.
-     * This is efficient because all except for very few wal files will be in
-     * archive. Checking for WalType is not much helpful because alive wal could
-       be archived now.
+    /* Try deleting in the dir that pathname points to for the logfile.
+       This may fail if we try to delete a log file which was live when captured
+       but is archived now. Try deleting it from archive also
      */
-    if (!env_->DeleteFile(archivedir + "/" + wal->Filename()).ok() &&
-        !env_->DeleteFile(dbname_ + "/" + wal->Filename()).ok()) {
-      files_not_deleted.append(wal->Filename());
+    Status st = env_->DeleteFile(dbname_ + "/" + wal->PathName());
+    if (!st.ok()) {
+      if (wal->Type() == kAliveLogFile &&
+          env_->DeleteFile(LogFileName(archivedir, wal->LogNumber())).ok()) {
+        continue;
+      }
+      files_not_deleted.append(wal->PathName());
     }
   }
   if (!files_not_deleted.empty()) {

db/db_impl.cc

@@ -28,13 +28,13 @@
 #include "db/version_set.h"
 #include "db/write_batch_internal.h"
 #include "db/transaction_log_impl.h"
-#include "leveldb/compaction_filter.h"
-#include "leveldb/db.h"
-#include "leveldb/env.h"
-#include "leveldb/merge_operator.h"
-#include "leveldb/statistics.h"
-#include "leveldb/status.h"
-#include "leveldb/table_builder.h"
+#include "rocksdb/compaction_filter.h"
+#include "rocksdb/db.h"
+#include "rocksdb/env.h"
+#include "rocksdb/merge_operator.h"
+#include "rocksdb/statistics.h"
+#include "rocksdb/status.h"
+#include "rocksdb/table_builder.h"
 #include "port/port.h"
 #include "table/block.h"
 #include "table/merger.h"
@@ -47,7 +47,7 @@
 #include "util/mutexlock.h"
 #include "util/stop_watch.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 void dumpLeveldbBuildVersion(Logger * log);
 
@@ -163,6 +163,22 @@ Options SanitizeOptions(const std::string& dbname,
       result.compaction_filter_factory->CreateCompactionFilter().get()) {
     Log(result.info_log, "Both filter and factory specified. Using filter");
   }
+  if (result.prefix_extractor) {
+    // If a prefix extractor has been supplied and a PrefixHashRepFactory is
+    // being used, make sure that the latter uses the former as its transform
+    // function.
+    auto factory = dynamic_cast<PrefixHashRepFactory*>(
+      result.memtable_factory.get());
+    if (factory &&
+        factory->GetTransform() != result.prefix_extractor) {
+      Log(result.info_log, "A prefix hash representation factory was supplied "
+          "whose prefix extractor does not match options.prefix_extractor. "
+          "Falling back to skip list representation factory");
+      result.memtable_factory = std::make_shared<SkipListFactory>();
+    } else if (factory) {
+      Log(result.info_log, "Prefix hash memtable rep is in use.");
+    }
+  }
   return result;
 }
 
@@ -184,6 +200,7 @@ DBImpl::DBImpl(const Options& options, const std::string& dbname)
       logfile_number_(0),
       tmp_batch_(),
       bg_compaction_scheduled_(0),
+      bg_flush_scheduled_(0),
       bg_logstats_scheduled_(false),
       manual_compaction_(nullptr),
       logger_(nullptr),
@@ -251,7 +268,9 @@ DBImpl::~DBImpl() {
   }
   mutex_.Lock();
   shutting_down_.Release_Store(this);  // Any non-nullptr value is ok
-  while (bg_compaction_scheduled_ || bg_logstats_scheduled_) {
+  while (bg_compaction_scheduled_ ||
+         bg_flush_scheduled_ ||
+         bg_logstats_scheduled_) {
     bg_cv_.Wait();
   }
   mutex_.Unlock();
@@ -271,13 +290,17 @@ void DBImpl::TEST_Destroy_DBImpl() {
 
   // wait till all background compactions are done.
   mutex_.Lock();
-  while (bg_compaction_scheduled_ || bg_logstats_scheduled_) {
+  while (bg_compaction_scheduled_ ||
+         bg_flush_scheduled_ ||
+         bg_logstats_scheduled_) {
     bg_cv_.Wait();
   }
 
   // Prevent new compactions from occuring.
+  bg_work_gate_closed_ = true;
   const int LargeNumber = 10000000;
   bg_compaction_scheduled_ += LargeNumber;
+
   mutex_.Unlock();
 
   // force release the lock file.
@@ -406,6 +429,26 @@ void DBImpl::FindObsoleteFiles(DeletionState& deletion_state) {
   deletion_state.prevlognumber = versions_->PrevLogNumber();
 }
 
+Status DBImpl::DeleteLogFile(uint64_t number) {
+  Status s;
+  auto filename = LogFileName(dbname_, number);
+  if (options_.WAL_ttl_seconds > 0) {
+    s = env_->RenameFile(filename,
+                         ArchivedLogFileName(dbname_, number));
+
+    if (!s.ok()) {
+      Log(options_.info_log, "RenameFile logfile #%lu FAILED", number);
+    }
+  } else {
+    s = env_->DeleteFile(filename);
+    if(!s.ok()) {
+      Log(options_.info_log, "Delete logfile #%lu FAILED", number);
+    }
+  }
+
+  return s;
+}
+
 // Diffs the files listed in filenames and those that do not
 // belong to live files are posibly removed. If the removed file
 // is a sst file, then it returns the file number in files_to_evict.
@@ -460,19 +503,9 @@ void DBImpl::PurgeObsoleteFiles(DeletionState& state) {
           state.files_to_evict.push_back(number);
         }
         Log(options_.info_log, "Delete type=%d #%lu", int(type), number);
-        if (type == kLogFile && options_.WAL_ttl_seconds > 0) {
-          Status st = env_->RenameFile(
-            LogFileName(dbname_, number),
-            ArchivedLogFileName(dbname_, number)
-          );
 
-          if (!st.ok()) {
-            Log(
-              options_.info_log, "RenameFile type=%d #%lu FAILED",
-              int(type),
-              number
-            );
-          }
+        if (type == kLogFile) {
+          DeleteLogFile(number);
         } else {
           Status st = env_->DeleteFile(dbname_ + "/" + state.allfiles[i]);
           if (!st.ok()) {
@@ -485,7 +518,7 @@ void DBImpl::PurgeObsoleteFiles(DeletionState& state) {
     }
   }
 
-  // Delete old log files.
+  // Delete old info log files.
   size_t old_log_file_count = old_log_files.size();
   // NOTE: Currently we only support log purge when options_.db_log_dir is
   // located in `dbname` directory.
@@ -890,7 +923,6 @@ Status DBImpl::CompactMemTable(bool* madeProgress) {
   }
 
   // Save the contents of the earliest memtable as a new Table
-  // This will release and re-acquire the mutex.
   uint64_t file_number;
   std::vector<MemTable*> mems;
   imm_.PickMemtablesToFlush(&mems);
@@ -904,8 +936,10 @@ Status DBImpl::CompactMemTable(bool* madeProgress) {
   MemTable* m = mems[0];
   VersionEdit* edit = m->GetEdits();
   edit->SetPrevLogNumber(0);
-  edit->SetLogNumber(m->GetLogNumber());  // Earlier logs no longer needed
+  edit->SetLogNumber(m->GetNextLogNumber());  // Earlier logs no longer needed
+  auto to_delete = m->GetLogNumber();
 
+  // This will release and re-acquire the mutex.
   Status s = WriteLevel0Table(mems, edit, &file_number);
 
   if (s.ok() && shutting_down_.Acquire_Load()) {
@@ -923,16 +957,25 @@ Status DBImpl::CompactMemTable(bool* madeProgress) {
     if (madeProgress) {
       *madeProgress = 1;
     }
+
     MaybeScheduleLogDBDeployStats();
-    // we could have deleted obsolete files here, but it is not
-    // absolutely necessary because it could be also done as part
-    // of other background compaction
+    // TODO: if log deletion failed for any reason, we probably
+    // should store the file number in the shared state, and retry
+    // However, for now, PurgeObsoleteFiles will take care of that
+    // anyways.
+    if (options_.purge_log_after_memtable_flush &&
+        !disable_delete_obsolete_files_ &&
+        to_delete > 0) {
+      mutex_.Unlock();
+      DeleteLogFile(to_delete);
+      mutex_.Lock();
+    }
   }
   return s;
 }
 
 void DBImpl::CompactRange(const Slice* begin, const Slice* end,
-                          bool reduce_level) {
+                          bool reduce_level, int target_level) {
   int max_level_with_files = 1;
   {
     MutexLock l(&mutex_);
@@ -949,7 +992,7 @@ void DBImpl::CompactRange(const Slice* begin, const Slice* end,
   }
 
   if (reduce_level) {
-    ReFitLevel(max_level_with_files);
+    ReFitLevel(max_level_with_files, target_level);
   }
 }
 
@@ -969,7 +1012,7 @@ int DBImpl::FindMinimumEmptyLevelFitting(int level) {
   return minimum_level;
 }
 
-void DBImpl::ReFitLevel(int level) {
+void DBImpl::ReFitLevel(int level, int target_level) {
   assert(level < NumberLevels());
 
   MutexLock l(&mutex_);
@@ -983,15 +1026,18 @@ void DBImpl::ReFitLevel(int level) {
 
   // wait for all background threads to stop
   bg_work_gate_closed_ = true;
-  while (bg_compaction_scheduled_ > 0) {
+  while (bg_compaction_scheduled_ > 0 || bg_flush_scheduled_) {
     Log(options_.info_log,
-        "RefitLevel: waiting for background threads to stop: %d",
-        bg_compaction_scheduled_);
+        "RefitLevel: waiting for background threads to stop: %d %d",
+        bg_compaction_scheduled_, bg_flush_scheduled_);
     bg_cv_.Wait();
   }
 
   // move to a smaller level
-  int to_level = FindMinimumEmptyLevelFitting(level);
+  int to_level = target_level;
+  if (target_level < 0) {
+    to_level = FindMinimumEmptyLevelFitting(level);
+  }
 
   assert(to_level <= level);
 
@@ -1046,6 +1092,7 @@ SequenceNumber DBImpl::GetLatestSequenceNumber() {
 Status DBImpl::GetUpdatesSince(SequenceNumber seq,
                                unique_ptr<TransactionLogIterator>* iter) {
 
+  RecordTick(options_.statistics, GET_UPDATES_SINCE_CALLS);
   if (seq > last_flushed_sequence_) {
     return Status::IOError("Requested sequence not yet written in the db");
   }
@@ -1196,6 +1243,12 @@ Status DBImpl::AppendSortedWalsOfType(const std::string& path,
     return status;
   }
   log_files.reserve(log_files.size() + all_files.size());
+  VectorLogPtr::iterator pos_start;
+  if (!log_files.empty()) {
+    pos_start = log_files.end() - 1;
+  } else {
+    pos_start = log_files.begin();
+  }
   for (const auto& f : all_files) {
     uint64_t number;
     FileType type;
@@ -1221,7 +1274,7 @@ Status DBImpl::AppendSortedWalsOfType(const std::string& path,
     }
   }
   CompareLogByPointer compare_log_files;
-  std::sort(log_files.begin(), log_files.end(), compare_log_files);
+  std::sort(pos_start, log_files.end(), compare_log_files);
   return status;
 }
 
@@ -1328,7 +1381,8 @@ Status DBImpl::TEST_WaitForCompactMemTable() {
 Status DBImpl::TEST_WaitForCompact() {
   // Wait until the compaction completes
   MutexLock l(&mutex_);
-  while (bg_compaction_scheduled_ && bg_error_.ok()) {
+  while ((bg_compaction_scheduled_ || bg_flush_scheduled_) &&
+         bg_error_.ok()) {
     bg_cv_.Wait();
   }
   return bg_error_;
@@ -1338,29 +1392,80 @@ void DBImpl::MaybeScheduleCompaction() {
   mutex_.AssertHeld();
   if (bg_work_gate_closed_) {
     // gate closed for backgrond work
-  } else if (bg_compaction_scheduled_ >= options_.max_background_compactions) {
-    // Already scheduled
   } else if (shutting_down_.Acquire_Load()) {
     // DB is being deleted; no more background compactions
-  } else if (!imm_.IsFlushPending(options_.min_write_buffer_number_to_merge) &&
-             manual_compaction_ == nullptr &&
-             !versions_->NeedsCompaction()) {
-    // No work to be done
   } else {
-    bg_compaction_scheduled_++;
-    env_->Schedule(&DBImpl::BGWork, this);
+    bool is_flush_pending =
+      imm_.IsFlushPending(options_.min_write_buffer_number_to_merge);
+    if (is_flush_pending &&
+        (bg_flush_scheduled_ < options_.max_background_flushes)) {
+      // memtable flush needed
+      bg_flush_scheduled_++;
+      env_->Schedule(&DBImpl::BGWorkFlush, this, Env::Priority::HIGH);
+    }
+
+    if ((manual_compaction_ ||
+         versions_->NeedsCompaction() ||
+         (is_flush_pending && (options_.max_background_flushes <= 0))) &&
+        bg_compaction_scheduled_ < options_.max_background_compactions) {
+      // compaction needed, or memtable flush needed but HIGH pool not enabled.
+      bg_compaction_scheduled_++;
+      env_->Schedule(&DBImpl::BGWorkCompaction, this, Env::Priority::LOW);
+    }
   }
 }
 
-void DBImpl::BGWork(void* db) {
-  reinterpret_cast<DBImpl*>(db)->BackgroundCall();
+void DBImpl::BGWorkFlush(void* db) {
+  reinterpret_cast<DBImpl*>(db)->BackgroundCallFlush();
 }
 
+void DBImpl::BGWorkCompaction(void* db) {
+  reinterpret_cast<DBImpl*>(db)->BackgroundCallCompaction();
+}
+
+Status DBImpl::BackgroundFlush() {
+  Status stat;
+  while (stat.ok() &&
+         imm_.IsFlushPending(options_.min_write_buffer_number_to_merge)) {
+    Log(options_.info_log,
+        "BackgroundCallFlush doing CompactMemTable, flush slots available %d",
+        options_.max_background_flushes - bg_flush_scheduled_);
+    stat = CompactMemTable();
+  }
+  return stat;
+}
+
+void DBImpl::BackgroundCallFlush() {
+  assert(bg_flush_scheduled_);
+  MutexLock l(&mutex_);
+
+  if (!shutting_down_.Acquire_Load()) {
+    Status s = BackgroundFlush();
+    if (!s.ok()) {
+      // Wait a little bit before retrying background compaction in
+      // case this is an environmental problem and we do not want to
+      // chew up resources for failed compactions for the duration of
+      // the problem.
+      bg_cv_.SignalAll();  // In case a waiter can proceed despite the error
+      Log(options_.info_log, "Waiting after background flush error: %s",
+          s.ToString().c_str());
+      mutex_.Unlock();
+      env_->SleepForMicroseconds(1000000);
+      mutex_.Lock();
+    }
+  }
+
+  bg_flush_scheduled_--;
+
+  bg_cv_.SignalAll();
+}
+
+
 void DBImpl::TEST_PurgeObsoleteteWAL() {
   PurgeObsoleteWALFiles();
 }
 
-void DBImpl::BackgroundCall() {
+void DBImpl::BackgroundCallCompaction() {
   bool madeProgress = false;
   DeletionState deletion_state;
 
@@ -1413,6 +1518,7 @@ Status DBImpl::BackgroundCompaction(bool* madeProgress,
   *madeProgress = false;
   mutex_.AssertHeld();
 
+  // TODO: remove memtable flush from formal compaction
   while (imm_.IsFlushPending(options_.min_write_buffer_number_to_merge)) {
     Log(options_.info_log,
         "BackgroundCompaction doing CompactMemTable, compaction slots available %d",
@@ -1789,6 +1895,7 @@ Status DBImpl::DoCompactionWork(CompactionState* compact) {
   }
   for (; input->Valid() && !shutting_down_.Acquire_Load(); ) {
     // Prioritize immutable compaction work
+    // TODO: remove memtable flush from normal compaction work
     if (imm_.imm_flush_needed.NoBarrier_Load() != nullptr) {
       const uint64_t imm_start = env_->NowMicros();
       mutex_.Lock();
@@ -2085,7 +2192,7 @@ Status DBImpl::DoCompactionWork(CompactionState* compact) {
   }
 
   mutex_.Lock();
-  stats_[compact->compaction->level() + 1].Add(stats);
+  stats_[compact->compaction->output_level()].Add(stats);
 
   // if there were any unused file number (mostly in case of
   // compaction error), free up the entry from pending_putputs
@@ -2097,7 +2204,8 @@ Status DBImpl::DoCompactionWork(CompactionState* compact) {
   VersionSet::LevelSummaryStorage tmp;
   Log(options_.info_log,
       "compacted to: %s, %.1f MB/sec, level %d, files in(%d, %d) out(%d) "
-      "MB in(%.1f, %.1f) out(%.1f), amplify(%.1f) %s\n",
+      "MB in(%.1f, %.1f) out(%.1f), read-write-amplify(%.1f) "
+      "write-amplify(%.1f) %s\n",
       versions_->LevelSummary(&tmp),
       (stats.bytes_readn + stats.bytes_readnp1 + stats.bytes_written) /
           (double) stats.micros,
@@ -2106,8 +2214,9 @@ Status DBImpl::DoCompactionWork(CompactionState* compact) {
       stats.bytes_readn / 1048576.0,
       stats.bytes_readnp1 / 1048576.0,
       stats.bytes_written / 1048576.0,
-      (stats.bytes_written + stats.bytes_readnp1) /
+      (stats.bytes_written + stats.bytes_readnp1 + stats.bytes_readn) /
           (double) stats.bytes_readn,
+      stats.bytes_written / (double) stats.bytes_readn,
       status.ToString().c_str());
 
   return status;
@@ -2141,7 +2250,8 @@ Iterator* DBImpl::NewInternalIterator(const ReadOptions& options,
   // Collect together all needed child iterators for mem
   std::vector<Iterator*> list;
   mem_->Ref();
-  list.push_back(mem_->NewIterator());
+  list.push_back(mem_->NewIterator(options.prefix));
+
   cleanup->mem.push_back(mem_);
 
   // Collect together all needed child iterators for imm_
@@ -2150,7 +2260,7 @@ Iterator* DBImpl::NewInternalIterator(const ReadOptions& options,
   for (unsigned int i = 0; i < immutables.size(); i++) {
     MemTable* m = immutables[i];
     m->Ref();
-    list.push_back(m->NewIterator());
+    list.push_back(m->NewIterator(options.prefix));
     cleanup->mem.push_back(m);
   }
 
@@ -2187,13 +2297,12 @@ Status DBImpl::Get(const ReadOptions& options,
 Status DBImpl::GetImpl(const ReadOptions& options,
                        const Slice& key,
                        std::string* value,
-                       const bool no_io,
                        bool* value_found) {
   Status s;
 
   StopWatch sw(env_, options_.statistics, DB_GET);
   SequenceNumber snapshot;
-  MutexLock l(&mutex_);
+  mutex_.Lock();
   if (options.snapshot != nullptr) {
     snapshot = reinterpret_cast<const SnapshotImpl*>(options.snapshot)->number_;
   } else {
@@ -2238,6 +2347,9 @@ Status DBImpl::GetImpl(const ReadOptions& options,
   mem->Unref();
   imm.UnrefAll();
   current->Unref();
+  mutex_.Unlock();
+
+  // Note, tickers are atomic now - no lock protection needed any more.
   RecordTick(options_.statistics, NUMBER_KEYS_READ);
   RecordTick(options_.statistics, BYTES_READ, value->size());
   return s;
@@ -2249,7 +2361,7 @@ std::vector<Status> DBImpl::MultiGet(const ReadOptions& options,
 
   StopWatch sw(env_, options_.statistics, DB_MULTIGET);
   SequenceNumber snapshot;
-  MutexLock l(&mutex_);
+  mutex_.Lock();
   if (options.snapshot != nullptr) {
     snapshot = reinterpret_cast<const SnapshotImpl*>(options.snapshot)->number_;
   } else {
@@ -2313,6 +2425,8 @@ std::vector<Status> DBImpl::MultiGet(const ReadOptions& options,
   mem->Unref();
   imm.UnrefAll();
   current->Unref();
+  mutex_.Unlock();
+
   RecordTick(options_.statistics, NUMBER_MULTIGET_CALLS);
   RecordTick(options_.statistics, NUMBER_MULTIGET_KEYS_READ, numKeys);
   RecordTick(options_.statistics, NUMBER_MULTIGET_BYTES_READ, bytesRead);
@@ -2327,7 +2441,9 @@ bool DBImpl::KeyMayExist(const ReadOptions& options,
   if (value_found != nullptr) {
     *value_found = true; // falsify later if key-may-exist but can't fetch value
   }
-  return GetImpl(options, key, value, true, value_found).ok();
+  ReadOptions roptions = options;
+  roptions.read_tier = kBlockCacheTier; // read from block cache only
+  return GetImpl(roptions, key, value, value_found).ok();
 }
 
 Iterator* DBImpl::NewIterator(const ReadOptions& options) {
@@ -2397,22 +2513,28 @@ Status DBImpl::Write(const WriteOptions& options, WriteBatch* my_batch) {
   uint64_t last_sequence = versions_->LastSequence();
   Writer* last_writer = &w;
   if (status.ok() && my_batch != nullptr) {  // nullptr batch is for compactions
+    // TODO: BuildBatchGroup physically concatenate/copy all write batches into
+    // a new one. Mem copy is done with the lock held. Ideally, we only need
+    // the lock to obtain the last_writer and the references to all batches.
+    // Creation (copy) of the merged batch could have been done outside of the
+    // lock protected region.
     WriteBatch* updates = BuildBatchGroup(&last_writer);
-    const SequenceNumber current_sequence = last_sequence + 1;
-    WriteBatchInternal::SetSequence(updates, current_sequence);
-    int my_batch_count = WriteBatchInternal::Count(updates);
-    last_sequence += my_batch_count;
-    // Record statistics
-    RecordTick(options_.statistics, NUMBER_KEYS_WRITTEN, my_batch_count);
-    RecordTick(options_.statistics,
-               BYTES_WRITTEN,
-               WriteBatchInternal::ByteSize(updates));
+
     // Add to log and apply to memtable.  We can release the lock
     // during this phase since &w is currently responsible for logging
     // and protects against concurrent loggers and concurrent writes
     // into mem_.
     {
       mutex_.Unlock();
+      const SequenceNumber current_sequence = last_sequence + 1;
+      WriteBatchInternal::SetSequence(updates, current_sequence);
+      int my_batch_count = WriteBatchInternal::Count(updates);
+      last_sequence += my_batch_count;
+      // Record statistics
+      RecordTick(options_.statistics, NUMBER_KEYS_WRITTEN, my_batch_count);
+      RecordTick(options_.statistics,
+                 BYTES_WRITTEN,
+                 WriteBatchInternal::ByteSize(updates));
       if (options.disableWAL) {
         flush_on_destroy_ = true;
       }
@@ -2439,11 +2561,13 @@ Status DBImpl::Write(const WriteOptions& options, WriteBatch* my_batch) {
           // have succeeded in memtable but Status reports error for all writes.
           throw std::runtime_error("In memory WriteBatch corruption!");
         }
-        RecordTick(options_.statistics, SEQUENCE_NUMBER, my_batch_count);
+        SetTickerCount(options_.statistics, SEQUENCE_NUMBER, last_sequence);
+      }
+      mutex_.Lock();
+      if (status.ok()) {
         versions_->SetLastSequence(last_sequence);
         last_flushed_sequence_ = current_sequence;
       }
-      mutex_.Lock();
     }
     if (updates == &tmp_batch_) tmp_batch_.Clear();
   }
@@ -2705,11 +2829,15 @@ Status DBImpl::MakeRoomForWrite(bool force) {
       lfile->SetPreallocationBlockSize(1.1 * options_.write_buffer_size);
       logfile_number_ = new_log_number;
       log_.reset(new log::Writer(std::move(lfile)));
-      mem_->SetLogNumber(logfile_number_);
+      mem_->SetNextLogNumber(logfile_number_);
       imm_.Add(mem_);
+      if (force) {
+        imm_.FlushRequested();
+      }
       mem_ = new MemTable(internal_comparator_, mem_rep_factory_,
         NumberLevels(), options_);
       mem_->Ref();
+      mem_->SetLogNumber(logfile_number_);
       force = false;   // Do not force another compaction if have room
       MaybeScheduleCompaction();
     }
@@ -2773,7 +2901,7 @@ bool DBImpl::GetProperty(const Slice& property, std::string* value) {
     // Pardon the long line but I think it is easier to read this way.
     snprintf(buf, sizeof(buf),
              "                               Compactions\n"
-             "Level  Files Size(MB) Score Time(sec)  Read(MB) Write(MB)    Rn(MB)  Rnp1(MB)  Wnew(MB) Amplify Read(MB/s) Write(MB/s)      Rn     Rnp1     Wnp1     NewW    Count  Ln-stall Stall-cnt\n"
+             "Level  Files Size(MB) Score Time(sec)  Read(MB) Write(MB)    Rn(MB)  Rnp1(MB)  Wnew(MB) RW-Amplify Read(MB/s) Write(MB/s)      Rn     Rnp1     Wnp1     NewW    Count  Ln-stall Stall-cnt\n"
              "--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------\n"
              );
     value->append(buf);
@@ -2786,7 +2914,9 @@ bool DBImpl::GetProperty(const Slice& property, std::string* value) {
                             stats_[level].bytes_readnp1;
         double amplify = (stats_[level].bytes_readn == 0)
             ? 0.0
-            : (stats_[level].bytes_written + stats_[level].bytes_readnp1) /
+            : (stats_[level].bytes_written +
+               stats_[level].bytes_readnp1 +
+               stats_[level].bytes_readn) /
                 (double) stats_[level].bytes_readn;
 
         total_bytes_read += bytes_read;
@@ -2794,12 +2924,12 @@ bool DBImpl::GetProperty(const Slice& property, std::string* value) {
 
         snprintf(
             buf, sizeof(buf),
-            "%3d %8d %8.0f %5.1f %9.0f %9.0f %9.0f %9.0f %9.0f %9.0f %7.1f %9.1f %11.1f %8d %8d %8d %8d %8d %9.1f %9lu\n",
+            "%3d %8d %8.0f %5.1f %9.0f %9.0f %9.0f %9.0f %9.0f %9.0f %10.1f %9.1f %11.1f %8d %8d %8d %8d %8d %9.1f %9lu\n",
             level,
             files,
             versions_->NumLevelBytes(level) / 1048576.0,
             versions_->NumLevelBytes(level) /
-            versions_->MaxBytesForLevel(level),
+                versions_->MaxBytesForLevel(level),
             stats_[level].micros / 1e6,
             bytes_read / 1048576.0,
             stats_[level].bytes_written / 1048576.0,
@@ -2951,6 +3081,71 @@ inline void DBImpl::DelayLoggingAndReset() {
   }
 }
 
+Status DBImpl::DeleteFile(std::string name) {
+  uint64_t number;
+  FileType type;
+  if (!ParseFileName(name, &number, &type) ||
+      (type != kTableFile)) {
+    Log(options_.info_log, "DeleteFile #%lld FAILED. Invalid file name\n",
+        static_cast<unsigned long long>(number));
+    return Status::InvalidArgument("Invalid file name");
+  }
+
+  int level;
+  FileMetaData metadata;
+  int maxlevel = NumberLevels();
+  VersionEdit edit(maxlevel);
+  DeletionState deletion_state;
+  Status status;
+  {
+    MutexLock l(&mutex_);
+    status = versions_->GetMetadataForFile(number, &level, &metadata);
+    if (!status.ok()) {
+      Log(options_.info_log, "DeleteFile #%lld FAILED. File not found\n",
+          static_cast<unsigned long long>(number));
+      return Status::InvalidArgument("File not found");
+    }
+    assert((level > 0) && (level < maxlevel));
+
+    // If the file is being compacted no need to delete.
+    if (metadata.being_compacted) {
+      Log(options_.info_log,
+          "DeleteFile #%lld Skipped. File about to be compacted\n",
+          static_cast<unsigned long long>(number));
+      return Status::OK();
+    }
+
+    // Only the files in the last level can be deleted externally.
+    // This is to make sure that any deletion tombstones are not
+    // lost. Check that the level passed is the last level.
+    for (int i = level + 1; i < maxlevel; i++) {
+      if (versions_->NumLevelFiles(i) != 0) {
+        Log(options_.info_log,
+            "DeleteFile #%lld FAILED. File not in last level\n",
+            static_cast<unsigned long long>(number));
+        return Status::InvalidArgument("File not in last level");
+      }
+    }
+    edit.DeleteFile(level, number);
+    status = versions_->LogAndApply(&edit, &mutex_);
+    if (status.ok()) {
+      FindObsoleteFiles(deletion_state);
+    }
+  } // lock released here
+
+  if (status.ok()) {
+    // remove files outside the db-lock
+    PurgeObsoleteFiles(deletion_state);
+    EvictObsoleteFiles(deletion_state);
+  }
+  return status;
+}
+
+void DBImpl::GetLiveFilesMetaData(std::vector<LiveFileMetaData> *metadata) {
+  MutexLock l(&mutex_);
+  return versions_->GetLiveFilesMetaData(metadata);
+}
+
 // Default implementations of convenience methods that subclasses of DB
 // can call if they wish
 Status DB::Put(const WriteOptions& opt, const Slice& key, const Slice& value) {
@@ -3005,12 +3200,26 @@ Status DB::Open(const Options& options, const std::string& dbname, DB** dbptr) {
       s = impl->versions_->LogAndApply(&edit, &impl->mutex_);
     }
     if (s.ok()) {
+      impl->mem_->SetLogNumber(impl->logfile_number_);
       impl->DeleteObsoleteFiles();
       impl->MaybeScheduleCompaction();
       impl->MaybeScheduleLogDBDeployStats();
     }
   }
   impl->mutex_.Unlock();
+
+  if (options.compaction_style == kCompactionStyleUniversal) {
+    int num_files;
+    for (int i = 1; i < impl->NumberLevels(); i++) {
+      num_files = impl->versions_->NumLevelFiles(i);
+      if (num_files > 0) {
+        s = Status::InvalidArgument("Not all files are at level 0. Cannot "
+          "open with universal compaction style.");
+        break;
+      }
+    }
+  }
+
   if (s.ok()) {
     *dbptr = impl;
   } else {
@@ -3085,4 +3294,4 @@ void dumpLeveldbBuildVersion(Logger * log) {
       leveldb_build_compile_time, leveldb_build_compile_date);
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb

db/db_impl.h

@@ -12,10 +12,10 @@
 #include "db/dbformat.h"
 #include "db/log_writer.h"
 #include "db/snapshot.h"
-#include "leveldb/db.h"
-#include "leveldb/env.h"
-#include "leveldb/memtablerep.h"
-#include "leveldb/transaction_log.h"
+#include "rocksdb/db.h"
+#include "rocksdb/env.h"
+#include "rocksdb/memtablerep.h"
+#include "rocksdb/transaction_log.h"
 #include "port/port.h"
 #include "util/stats_logger.h"
 #include "memtablelist.h"
@@ -24,7 +24,7 @@
 #include "scribe/scribe_logger.h"
 #endif
 
-namespace leveldb {
+namespace rocksdb {
 
 class MemTable;
 class TableCache;
@@ -64,7 +64,7 @@ class DBImpl : public DB {
   virtual bool GetProperty(const Slice& property, std::string* value);
   virtual void GetApproximateSizes(const Range* range, int n, uint64_t* sizes);
   virtual void CompactRange(const Slice* begin, const Slice* end,
-                            bool reduce_level = false);
+                            bool reduce_level = false, int target_level = -1);
   virtual int NumberLevels();
   virtual int MaxMemCompactionLevel();
   virtual int Level0StopWriteTrigger();
@@ -72,12 +72,17 @@ class DBImpl : public DB {
   virtual Status DisableFileDeletions();
   virtual Status EnableFileDeletions();
   virtual Status GetLiveFiles(std::vector<std::string>&,
-                              uint64_t* manifest_file_size);
+                              uint64_t* manifest_file_size,
+                              bool flush_memtable = true);
   virtual Status GetSortedWalFiles(VectorLogPtr& files);
   virtual Status DeleteWalFiles(const VectorLogPtr& files);
   virtual SequenceNumber GetLatestSequenceNumber();
   virtual Status GetUpdatesSince(SequenceNumber seq_number,
                                  unique_ptr<TransactionLogIterator>* iter);
+  virtual Status DeleteFile(std::string name);
+
+  virtual void GetLiveFilesMetaData(
+    std::vector<LiveFileMetaData> *metadata);
 
   // Extra methods (for testing) that are not in the public DB interface
 
@@ -183,9 +188,12 @@ class DBImpl : public DB {
   void LogDBDeployStats();
 
   void MaybeScheduleCompaction();
-  static void BGWork(void* db);
-  void BackgroundCall();
+  static void BGWorkCompaction(void* db);
+  static void BGWorkFlush(void* db);
+  void BackgroundCallCompaction();
+  void BackgroundCallFlush();
   Status BackgroundCompaction(bool* madeProgress,DeletionState& deletion_state);
+  Status BackgroundFlush();
   void CleanupCompaction(CompactionState* compact);
   Status DoCompactionWork(CompactionState* compact);
 
@@ -208,6 +216,8 @@ class DBImpl : public DB {
   // Removes the file listed in files_to_evict from the table_cache
   void EvictObsoleteFiles(DeletionState& deletion_state);
 
+  Status DeleteLogFile(uint64_t number);
+
   void PurgeObsoleteWALFiles();
 
   Status AppendSortedWalsOfType(const std::string& path,
@@ -237,9 +247,10 @@ class DBImpl : public DB {
   // input level. Return the input level, if such level could not be found.
   int FindMinimumEmptyLevelFitting(int level);
 
-  // Move the files in the input level to the minimum level that could hold
-  // the data set.
-  void ReFitLevel(int level);
+  // Move the files in the input level to the target level.
+  // If target_level < 0, automatically calculate the minimum level that could
+  // hold the data set.
+  void ReFitLevel(int level, int target_level = -1);
 
   // Constant after construction
   const InternalFilterPolicy internal_filter_policy_;
@@ -276,6 +287,9 @@ class DBImpl : public DB {
   // count how many background compaction been scheduled or is running?
   int bg_compaction_scheduled_;
 
+  // number of background memtable flush jobs, submitted to the HIGH pool
+  int bg_flush_scheduled_;
+
   // Has a background stats log thread scheduled?
   bool bg_logstats_scheduled_;
 
@@ -420,7 +434,6 @@ class DBImpl : public DB {
   Status GetImpl(const ReadOptions& options,
                  const Slice& key,
                  std::string* value,
-                 const bool no_io = false,
                  bool* value_found = nullptr);
 };
 
@@ -431,6 +444,6 @@ extern Options SanitizeOptions(const std::string& db,
                                const InternalFilterPolicy* ipolicy,
                                const Options& src);
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 #endif  // STORAGE_LEVELDB_DB_DB_IMPL_H_

db/db_impl_readonly.cc

@@ -21,10 +21,10 @@
 #include "db/table_cache.h"
 #include "db/version_set.h"
 #include "db/write_batch_internal.h"
-#include "leveldb/db.h"
-#include "leveldb/env.h"
-#include "leveldb/status.h"
-#include "leveldb/table_builder.h"
+#include "rocksdb/db.h"
+#include "rocksdb/env.h"
+#include "rocksdb/status.h"
+#include "rocksdb/table_builder.h"
 #include "port/port.h"
 #include "table/block.h"
 #include "table/merger.h"
@@ -34,7 +34,7 @@
 #include "util/logging.h"
 #include "util/build_version.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 DBImplReadOnly::DBImplReadOnly(const Options& options,
     const std::string& dbname)

db/db_impl_readonly.h

@@ -12,8 +12,8 @@
 #include "db/dbformat.h"
 #include "db/log_writer.h"
 #include "db/snapshot.h"
-#include "leveldb/db.h"
-#include "leveldb/env.h"
+#include "rocksdb/db.h"
+#include "rocksdb/env.h"
 #include "port/port.h"
 #include "util/stats_logger.h"
 
@@ -21,7 +21,7 @@
 #include "scribe/scribe_logger.h"
 #endif
 
-namespace leveldb {
+namespace rocksdb {
 
 class DBImplReadOnly : public DBImpl {
 public:
@@ -51,7 +51,7 @@ public:
    return Status::NotSupported("Not supported operation in read only mode.");
  }
  virtual void CompactRange(const Slice* begin, const Slice* end,
-                           bool reduce_level = false) {
+                           bool reduce_level = false, int target_level = -1) {
  }
  virtual Status DisableFileDeletions() {
    return Status::NotSupported("Not supported operation in read only mode.");
@@ -60,7 +60,8 @@ public:
    return Status::NotSupported("Not supported operation in read only mode.");
  }
  virtual Status GetLiveFiles(std::vector<std::string>&,
-                             uint64_t* manifest_file_size) {
+                             uint64_t* manifest_file_size,
+                             bool flush_memtable = true) {
    return Status::NotSupported("Not supported operation in read only mode.");
  }
  virtual Status Flush(const FlushOptions& options) {

db/db_iter.cc

@@ -8,15 +8,16 @@
 
 #include "db/filename.h"
 #include "db/dbformat.h"
-#include "leveldb/env.h"
-#include "leveldb/options.h"
-#include "leveldb/iterator.h"
-#include "leveldb/merge_operator.h"
+#include "rocksdb/env.h"
+#include "rocksdb/options.h"
+#include "rocksdb/iterator.h"
+#include "rocksdb/merge_operator.h"
 #include "port/port.h"
 #include "util/logging.h"
 #include "util/mutexlock.h"
+#include "util/perf_context_imp.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 #if 0
 static void DumpInternalIter(Iterator* iter) {
@@ -65,6 +66,7 @@ class DBIter: public Iterator {
         current_entry_is_merged_(false),
         statistics_(options.statistics) {
     RecordTick(statistics_, NO_ITERATORS, 1);
+    max_skip_ = options.max_sequential_skip_in_iterations;
   }
   virtual ~DBIter() {
     RecordTick(statistics_, NO_ITERATORS, -1);
@@ -129,6 +131,7 @@ class DBIter: public Iterator {
   bool valid_;
   bool current_entry_is_merged_;
   std::shared_ptr<Statistics> statistics_;
+  uint64_t max_skip_;
 
   // No copying allowed
   DBIter(const DBIter&);
@@ -188,12 +191,14 @@ void DBIter::FindNextUserEntry(bool skipping) {
   assert(iter_->Valid());
   assert(direction_ == kForward);
   current_entry_is_merged_ = false;
+  uint64_t num_skipped = 0;
   do {
     ParsedInternalKey ikey;
     if (ParseKey(&ikey) && ikey.sequence <= sequence_) {
       if (skipping &&
           user_comparator_->Compare(ikey.user_key, saved_key_) <= 0) {
-        // skip this entry
+        num_skipped++; // skip this entry
+        BumpPerfCount(&perf_context.internal_key_skipped_count);
       } else {
         skipping = false;
         switch (ikey.type) {
@@ -202,6 +207,8 @@ void DBIter::FindNextUserEntry(bool skipping) {
             // they are hidden by this deletion.
             SaveKey(ikey.user_key, &saved_key_);
             skipping = true;
+            num_skipped = 0;
+            BumpPerfCount(&perf_context.internal_delete_skipped_count);
             break;
           case kTypeValue:
             valid_ = true;
@@ -220,7 +227,20 @@ void DBIter::FindNextUserEntry(bool skipping) {
         }
       }
     }
-    iter_->Next();
+    // If we have sequentially iterated via numerous keys and still not
+    // found the next user-key, then it is better to seek so that we can
+    // avoid too many key comparisons. We seek to the last occurence of
+    // our current key by looking for sequence number 0.
+    if (skipping && num_skipped > max_skip_) {
+      num_skipped = 0;
+      std::string last_key;
+      AppendInternalKey(&last_key,
+        ParsedInternalKey(Slice(saved_key_), 0, kValueTypeForSeek));
+      iter_->Seek(last_key);
+      RecordTick(statistics_, NUMBER_OF_RESEEKS_IN_ITERATION);
+    } else {
+      iter_->Next();
+    }
   } while (iter_->Valid());
   valid_ = false;
 }
@@ -342,6 +362,7 @@ void DBIter::Prev() {
 
 void DBIter::FindPrevUserEntry() {
   assert(direction_ == kReverse);
+  uint64_t num_skipped = 0;
 
   ValueType value_type = kTypeDeletion;
   if (iter_->Valid()) {
@@ -367,7 +388,22 @@ void DBIter::FindPrevUserEntry() {
           saved_value_.assign(raw_value.data(), raw_value.size());
         }
       }
-      iter_->Prev();
+      num_skipped++;
+      // If we have sequentially iterated via numerous keys and still not
+      // found the prev user-key, then it is better to seek so that we can
+      // avoid too many key comparisons. We seek to the first occurence of
+      // our current key by looking for max sequence number.
+      if (num_skipped > max_skip_) {
+        num_skipped = 0;
+        std::string last_key;
+        AppendInternalKey(&last_key,
+          ParsedInternalKey(Slice(saved_key_), kMaxSequenceNumber,
+                            kValueTypeForSeek));
+        iter_->Seek(last_key);
+        RecordTick(statistics_, NUMBER_OF_RESEEKS_IN_ITERATION);
+      } else {
+        iter_->Prev();
+      }
     } while (iter_->Valid());
   }
 
@@ -435,4 +471,4 @@ Iterator* NewDBIterator(
                     internal_iter, sequence);
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb

db/db_iter.h

@@ -6,10 +6,10 @@
 #define STORAGE_LEVELDB_DB_DB_ITER_H_
 
 #include <stdint.h>
-#include "leveldb/db.h"
+#include "rocksdb/db.h"
 #include "db/dbformat.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 // Return a new iterator that converts internal keys (yielded by
 // "*internal_iter") that were live at the specified "sequence" number
@@ -22,6 +22,6 @@ extern Iterator* NewDBIterator(
     Iterator* internal_iter,
     const SequenceNumber& sequence);
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 #endif  // STORAGE_LEVELDB_DB_DB_ITER_H_

db/db_statistics.h

@@ -10,13 +10,13 @@
 #include <vector>
 #include <memory>
 
-#include "leveldb/statistics.h"
+#include "rocksdb/statistics.h"
 #include "util/histogram.h"
 #include "port/port.h"
 #include "util/mutexlock.h"
 
 
-namespace leveldb {
+namespace rocksdb {
 
 class DBStatistics: public Statistics {
  public:
@@ -59,8 +59,6 @@ std::shared_ptr<Statistics> CreateDBStatistics() {
   return std::make_shared<DBStatistics>();
 }
 
-} // namespace leveldb
+} // namespace rocksdb
 
 #endif // LEVELDB_STORAGE_DB_DB_STATISTICS_H_
-
-

db/db_stats_logger.cc

@@ -7,12 +7,12 @@
 #include <stdint.h>
 #include <stdio.h>
 #include "db/version_set.h"
-#include "leveldb/db.h"
-#include "leveldb/env.h"
+#include "rocksdb/db.h"
+#include "rocksdb/env.h"
 #include "port/port.h"
 #include "util/mutexlock.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 void DBImpl::MaybeScheduleLogDBDeployStats() {
 

db/db_test.cc

@@ -5,15 +5,16 @@
 #include <algorithm>
 #include <set>
 
-#include "leveldb/db.h"
-#include "leveldb/filter_policy.h"
+#include "rocksdb/db.h"
+#include "rocksdb/filter_policy.h"
 #include "db/db_impl.h"
 #include "db/filename.h"
 #include "db/version_set.h"
 #include "db/write_batch_internal.h"
-#include "leveldb/cache.h"
-#include "leveldb/compaction_filter.h"
-#include "leveldb/env.h"
+#include "db/db_statistics.h"
+#include "rocksdb/cache.h"
+#include "rocksdb/compaction_filter.h"
+#include "rocksdb/env.h"
 #include "table/table.h"
 #include "util/hash.h"
 #include "util/logging.h"
@@ -22,7 +23,7 @@
 #include "util/testutil.h"
 #include "utilities/merge_operators.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 static bool SnappyCompressionSupported(const CompressionOptions& options) {
   std::string out;
@@ -68,6 +69,7 @@ class AtomicCounter {
     count_ = 0;
   }
 };
+
 }
 
 // Special Env used to delay background operations
@@ -207,9 +209,12 @@ class DBTest {
  private:
   const FilterPolicy* filter_policy_;
 
+ protected:
   // Sequence of option configurations to try
   enum OptionConfig {
     kDefault,
+    kVectorRep,
+    kUnsortedRep,
     kMergePut,
     kFilter,
     kUncompressed,
@@ -219,6 +224,7 @@ class DBTest {
     kCompactOnFlush,
     kPerfOptions,
     kDeletesFilterFirst,
+    kPrefixHashRep,
     kUniversalCompaction,
     kEnd
   };
@@ -293,6 +299,10 @@ class DBTest {
   Options CurrentOptions() {
     Options options;
     switch (option_config_) {
+      case kPrefixHashRep:
+        options.memtable_factory.reset(new
+          PrefixHashRepFactory(NewFixedPrefixTransform(1)));
+        break;
       case kMergePut:
         options.merge_operator = MergeOperators::CreatePutOperator();
         break;
@@ -321,6 +331,12 @@ class DBTest {
       case kDeletesFilterFirst:
         options.filter_deletes = true;
         break;
+      case kUnsortedRep:
+        options.memtable_factory.reset(new UnsortedRepFactory);
+        break;
+      case kVectorRep:
+        options.memtable_factory.reset(new VectorRepFactory(100));
+        break;
       case kUniversalCompaction:
         options.compaction_style = kCompactionStyleUniversal;
         break;
@@ -815,6 +831,7 @@ TEST(DBTest, KeyMayExist) {
     std::string value;
     Options options = CurrentOptions();
     options.filter_policy = NewBloomFilterPolicy(20);
+    options.statistics = rocksdb::CreateDBStatistics();
     Reopen(&options);
 
     ASSERT_TRUE(!db_->KeyMayExist(ropts, "a", &value));
@@ -827,25 +844,114 @@ TEST(DBTest, KeyMayExist) {
 
     dbfull()->Flush(FlushOptions());
     value.clear();
-    value_found = false;
+
+    long numopen = options.statistics.get()->getTickerCount(NO_FILE_OPENS);
+    long cache_miss =
+      options.statistics.get()->getTickerCount(BLOCK_CACHE_MISS);
     ASSERT_TRUE(db_->KeyMayExist(ropts, "a", &value, &value_found));
-    ASSERT_TRUE(value_found);
-    ASSERT_EQ("b", value);
+    ASSERT_TRUE(!value_found);
+    // assert that no new files were opened and no new blocks were
+    // read into block cache.
+    ASSERT_EQ(numopen, options.statistics.get()->getTickerCount(NO_FILE_OPENS));
+    ASSERT_EQ(cache_miss,
+              options.statistics.get()->getTickerCount(BLOCK_CACHE_MISS));
 
     ASSERT_OK(db_->Delete(WriteOptions(), "a"));
+
+    numopen = options.statistics.get()->getTickerCount(NO_FILE_OPENS);
+    cache_miss = options.statistics.get()->getTickerCount(BLOCK_CACHE_MISS);
     ASSERT_TRUE(!db_->KeyMayExist(ropts, "a", &value));
+    ASSERT_EQ(numopen, options.statistics.get()->getTickerCount(NO_FILE_OPENS));
+    ASSERT_EQ(cache_miss,
+              options.statistics.get()->getTickerCount(BLOCK_CACHE_MISS));
 
     dbfull()->Flush(FlushOptions());
     dbfull()->CompactRange(nullptr, nullptr);
+
+    numopen = options.statistics.get()->getTickerCount(NO_FILE_OPENS);
+    cache_miss = options.statistics.get()->getTickerCount(BLOCK_CACHE_MISS);
     ASSERT_TRUE(!db_->KeyMayExist(ropts, "a", &value));
+    ASSERT_EQ(numopen, options.statistics.get()->getTickerCount(NO_FILE_OPENS));
+    ASSERT_EQ(cache_miss,
+              options.statistics.get()->getTickerCount(BLOCK_CACHE_MISS));
 
     ASSERT_OK(db_->Delete(WriteOptions(), "c"));
+
+    numopen = options.statistics.get()->getTickerCount(NO_FILE_OPENS);
+    cache_miss = options.statistics.get()->getTickerCount(BLOCK_CACHE_MISS);
     ASSERT_TRUE(!db_->KeyMayExist(ropts, "c", &value));
+    ASSERT_EQ(numopen, options.statistics.get()->getTickerCount(NO_FILE_OPENS));
+    ASSERT_EQ(cache_miss,
+              options.statistics.get()->getTickerCount(BLOCK_CACHE_MISS));
 
     delete options.filter_policy;
   } while (ChangeOptions());
 }
 
+TEST(DBTest, NonBlockingIteration) {
+  do {
+    ReadOptions non_blocking_opts, regular_opts;
+    Options options = CurrentOptions();
+    options.statistics = rocksdb::CreateDBStatistics();
+    non_blocking_opts.read_tier = kBlockCacheTier;
+    Reopen(&options);
+
+    // write one kv to the database.
+    ASSERT_OK(db_->Put(WriteOptions(), "a", "b"));
+
+    // scan using non-blocking iterator. We should find it because
+    // it is in memtable.
+    Iterator* iter = db_->NewIterator(non_blocking_opts);
+    int count = 0;
+    for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
+      ASSERT_TRUE(iter->status().ok());
+      count++;
+    }
+    ASSERT_EQ(count, 1);
+    delete iter;
+
+    // flush memtable to storage. Now, the key should not be in the
+    // memtable neither in the block cache.
+    dbfull()->Flush(FlushOptions());
+
+    // verify that a non-blocking iterator does not find any
+    // kvs. Neither does it do any IOs to storage.
+    long numopen = options.statistics.get()->getTickerCount(NO_FILE_OPENS);
+    long cache_miss =
+      options.statistics.get()->getTickerCount(BLOCK_CACHE_MISS);
+    iter = db_->NewIterator(non_blocking_opts);
+    count = 0;
+    for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
+      count++;
+    }
+    ASSERT_EQ(count, 0);
+    ASSERT_TRUE(iter->status().IsIncomplete());
+    ASSERT_EQ(numopen, options.statistics.get()->getTickerCount(NO_FILE_OPENS));
+    ASSERT_EQ(cache_miss,
+              options.statistics.get()->getTickerCount(BLOCK_CACHE_MISS));
+    delete iter;
+
+    // read in the specified block via a regular get
+    ASSERT_EQ(Get("a"), "b");
+
+    // verify that we can find it via a non-blocking scan
+    numopen = options.statistics.get()->getTickerCount(NO_FILE_OPENS);
+    cache_miss = options.statistics.get()->getTickerCount(BLOCK_CACHE_MISS);
+    iter = db_->NewIterator(non_blocking_opts);
+    count = 0;
+    for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
+      ASSERT_TRUE(iter->status().ok());
+      count++;
+    }
+    ASSERT_EQ(count, 1);
+    ASSERT_EQ(numopen, options.statistics.get()->getTickerCount(NO_FILE_OPENS));
+    ASSERT_EQ(cache_miss,
+              options.statistics.get()->getTickerCount(BLOCK_CACHE_MISS));
+    delete iter;
+
+  } while (ChangeOptions());
+}
+
 // A delete is skipped for key if KeyMayExist(key) returns False
 // Tests Writebatch consistency and proper delete behaviour
 TEST(DBTest, FilterDeletes) {
@@ -1028,6 +1134,95 @@ TEST(DBTest, IterMulti) {
   } while (ChangeCompactOptions());
 }
 
+// Check that we can skip over a run of user keys
+// by using reseek rather than sequential scan
+TEST(DBTest, IterReseek) {
+  Options options = CurrentOptions();
+  options.max_sequential_skip_in_iterations = 3;
+  options.create_if_missing = true;
+  options.statistics = rocksdb::CreateDBStatistics();
+  DestroyAndReopen(&options);
+
+  // insert two keys with same userkey and verify that
+  // reseek is not invoked. For each of these test cases,
+  // verify that we can find the next key "b".
+  ASSERT_OK(Put("a",  "one"));
+  ASSERT_OK(Put("a",  "two"));
+  ASSERT_OK(Put("b",  "bone"));
+  Iterator* iter = db_->NewIterator(ReadOptions());
+  iter->SeekToFirst();
+  ASSERT_EQ(options.statistics.get()->getTickerCount(
+            NUMBER_OF_RESEEKS_IN_ITERATION), 0);
+  ASSERT_EQ(IterStatus(iter), "a->two");
+  iter->Next();
+  ASSERT_EQ(options.statistics.get()->getTickerCount(
+            NUMBER_OF_RESEEKS_IN_ITERATION), 0);
+  ASSERT_EQ(IterStatus(iter), "b->bone");
+  delete iter;
+
+  // insert a total of three keys with same userkey and verify
+  // that reseek is still not invoked.
+  ASSERT_OK(Put("a",  "three"));
+  iter = db_->NewIterator(ReadOptions());
+  iter->SeekToFirst();
+  ASSERT_EQ(IterStatus(iter), "a->three");
+  iter->Next();
+  ASSERT_EQ(options.statistics.get()->getTickerCount(
+            NUMBER_OF_RESEEKS_IN_ITERATION), 0);
+  ASSERT_EQ(IterStatus(iter), "b->bone");
+  delete iter;
+
+  // insert a total of four keys with same userkey and verify
+  // that reseek is invoked.
+  ASSERT_OK(Put("a",  "four"));
+  iter = db_->NewIterator(ReadOptions());
+  iter->SeekToFirst();
+  ASSERT_EQ(IterStatus(iter), "a->four");
+  ASSERT_EQ(options.statistics.get()->getTickerCount(
+            NUMBER_OF_RESEEKS_IN_ITERATION), 0);
+  iter->Next();
+  ASSERT_EQ(options.statistics.get()->getTickerCount(
+            NUMBER_OF_RESEEKS_IN_ITERATION), 1);
+  ASSERT_EQ(IterStatus(iter), "b->bone");
+  delete iter;
+
+  // Testing reverse iterator
+  // At this point, we have three versions of "a" and one version of "b".
+  // The reseek statistics is already at 1.
+  int num_reseeks = (int)options.statistics.get()->getTickerCount(
+                 NUMBER_OF_RESEEKS_IN_ITERATION);
+
+  // Insert another version of b and assert that reseek is not invoked
+  ASSERT_OK(Put("b",  "btwo"));
+  iter = db_->NewIterator(ReadOptions());
+  iter->SeekToLast();
+  ASSERT_EQ(IterStatus(iter), "b->btwo");
+  ASSERT_EQ(options.statistics.get()->getTickerCount(
+            NUMBER_OF_RESEEKS_IN_ITERATION), num_reseeks);
+  iter->Prev();
+  ASSERT_EQ(options.statistics.get()->getTickerCount(
+            NUMBER_OF_RESEEKS_IN_ITERATION), num_reseeks+1);
+  ASSERT_EQ(IterStatus(iter), "a->four");
+  delete iter;
+
+  // insert two more versions of b. This makes a total of 4 versions
+  // of b and 4 versions of a.
+  ASSERT_OK(Put("b",  "bthree"));
+  ASSERT_OK(Put("b",  "bfour"));
+  iter = db_->NewIterator(ReadOptions());
+  iter->SeekToLast();
+  ASSERT_EQ(IterStatus(iter), "b->bfour");
+  ASSERT_EQ(options.statistics.get()->getTickerCount(
+            NUMBER_OF_RESEEKS_IN_ITERATION), num_reseeks + 2);
+  iter->Prev();
+
+  // the previous Prev call should have invoked reseek
+  ASSERT_EQ(options.statistics.get()->getTickerCount(
+            NUMBER_OF_RESEEKS_IN_ITERATION), num_reseeks + 3);
+  ASSERT_EQ(IterStatus(iter), "a->four");
+  delete iter;
+}
+
 TEST(DBTest, IterSmallAndLargeMix) {
   do {
     ASSERT_OK(Put("a", "va"));
@@ -1088,6 +1283,44 @@ TEST(DBTest, IterMultiWithDelete) {
   } while (ChangeOptions());
 }
 
+TEST(DBTest, IterWithSnapshot) {
+  do {
+    ASSERT_OK(Put("key1", "val1"));
+    ASSERT_OK(Put("key2", "val2"));
+    ASSERT_OK(Put("key3", "val3"));
+    ASSERT_OK(Put("key4", "val4"));
+    ASSERT_OK(Put("key5", "val5"));
+
+    const Snapshot *snapshot = db_->GetSnapshot();
+    ReadOptions options;
+    options.snapshot = snapshot;
+    Iterator* iter = db_->NewIterator(options);
+
+    // Put more values after the snapshot
+    ASSERT_OK(Put("key100", "val100"));
+    ASSERT_OK(Put("key101", "val101"));
+
+    iter->Seek("key5");
+    ASSERT_EQ(IterStatus(iter), "key5->val5");
+    if (!CurrentOptions().merge_operator) {
+      // TODO: merge operator does not support backward iteration yet
+      iter->Prev();
+      ASSERT_EQ(IterStatus(iter), "key4->val4");
+      iter->Prev();
+      ASSERT_EQ(IterStatus(iter), "key3->val3");
+
+      iter->Next();
+      ASSERT_EQ(IterStatus(iter), "key4->val4");
+      iter->Next();
+      ASSERT_EQ(IterStatus(iter), "key5->val5");
+      iter->Next();
+      ASSERT_TRUE(!iter->Valid());
+    }
+    db_->ReleaseSnapshot(snapshot);
+    delete iter;
+  } while (ChangeOptions());
+}
+
 TEST(DBTest, Recover) {
   do {
     ASSERT_OK(Put("foo", "v1"));
@@ -1171,6 +1404,24 @@ TEST(DBTest, CheckLock) {
   } while (ChangeCompactOptions());
 }
 
+TEST(DBTest, FlushMultipleMemtable) {
+  do {
+    Options options = CurrentOptions();
+    WriteOptions writeOpt = WriteOptions();
+    writeOpt.disableWAL = true;
+    options.max_write_buffer_number = 4;
+    options.min_write_buffer_number_to_merge = 3;
+    Reopen(&options);
+    ASSERT_OK(dbfull()->Put(writeOpt, "foo", "v1"));
+    dbfull()->Flush(FlushOptions());
+    ASSERT_OK(dbfull()->Put(writeOpt, "bar", "v1"));
+
+    ASSERT_EQ("v1", Get("foo"));
+    ASSERT_EQ("v1", Get("bar"));
+    dbfull()->Flush(FlushOptions());
+  } while (ChangeCompactOptions());
+}
+
 TEST(DBTest, FLUSH) {
   do {
     Options options = CurrentOptions();
@@ -1386,8 +1637,8 @@ TEST(DBTest, UniversalCompactionTrigger) {
   Options options = CurrentOptions();
   options.compaction_style = kCompactionStyleUniversal;
   options.write_buffer_size = 100<<10; //100KB
-  // trigger compaction if there are > 3 files
-  options.level0_file_num_compaction_trigger = 3;
+  // trigger compaction if there are >= 4 files
+  options.level0_file_num_compaction_trigger = 4;
   Reopen(&options);
 
   Random rnd(301);
@@ -1397,7 +1648,7 @@ TEST(DBTest, UniversalCompactionTrigger) {
   //   Generate a set of files at level 0, but don't trigger level-0
   //   compaction.
   for (int num = 0;
-       num < options.level0_file_num_compaction_trigger;
+       num < options.level0_file_num_compaction_trigger-1;
        num++) {
     // Write 120KB (12 values, each 10K)
     for (int i = 0; i < 12; i++) {
@@ -1432,7 +1683,7 @@ TEST(DBTest, UniversalCompactionTrigger) {
   //   data amount).
   dbfull()->Flush(FlushOptions());
   for (int num = 0;
-       num < options.level0_file_num_compaction_trigger-2;
+       num < options.level0_file_num_compaction_trigger-3;
        num++) {
     // Write 120KB (12 values, each 10K)
     for (int i = 0; i < 12; i++) {
@@ -1461,7 +1712,7 @@ TEST(DBTest, UniversalCompactionTrigger) {
   //   Now we have 2 files at level 0, with size 4 and 2.4. Continue
   //   generating new files at level 0.
   for (int num = 0;
-       num < options.level0_file_num_compaction_trigger-2;
+       num < options.level0_file_num_compaction_trigger-3;
        num++) {
     // Write 120KB (12 values, each 10K)
     for (int i = 0; i < 12; i++) {
@@ -1515,6 +1766,170 @@ TEST(DBTest, UniversalCompactionTrigger) {
   }
 }
 
+TEST(DBTest, UniversalCompactionSizeAmplification) {
+  Options options = CurrentOptions();
+  options.compaction_style = kCompactionStyleUniversal;
+  options.write_buffer_size = 100<<10; //100KB
+  options.level0_file_num_compaction_trigger = 3;
+
+  // Trigger compaction if size amplification exceeds 110%
+  options.compaction_options_universal.
+    max_size_amplification_percent = 110;
+  Reopen(&options);
+
+  Random rnd(301);
+  int key_idx = 0;
+
+  //   Generate two files in Level 0. Both files are approx the same size.
+  for (int num = 0;
+       num < options.level0_file_num_compaction_trigger-1;
+       num++) {
+    // Write 120KB (12 values, each 10K)
+    for (int i = 0; i < 12; i++) {
+      ASSERT_OK(Put(Key(key_idx), RandomString(&rnd, 10000)));
+      key_idx++;
+    }
+    dbfull()->TEST_WaitForCompactMemTable();
+    ASSERT_EQ(NumTableFilesAtLevel(0), num + 1);
+  }
+  ASSERT_EQ(NumTableFilesAtLevel(0), 2);
+
+  // Flush whatever is remaining in memtable. This is typically
+  // small, which should not trigger size ratio based compaction
+  // but will instead trigger size amplification.
+  dbfull()->Flush(FlushOptions());
+
+  // Verify that size amplification did occur
+  ASSERT_EQ(NumTableFilesAtLevel(0), 1);
+}
+
+TEST(DBTest, UniversalCompactionOptions) {
+  Options options = CurrentOptions();
+  options.compaction_style = kCompactionStyleUniversal;
+  options.write_buffer_size = 100<<10; //100KB
+  options.level0_file_num_compaction_trigger = 4;
+  options.num_levels = 1;
+  Reopen(&options);
+
+  Random rnd(301);
+  int key_idx = 0;
+
+  for (int num = 0;
+       num < options.level0_file_num_compaction_trigger;
+       num++) {
+    // Write 120KB (12 values, each 10K)
+    for (int i = 0; i < 12; i++) {
+      ASSERT_OK(Put(Key(key_idx), RandomString(&rnd, 10000)));
+      key_idx++;
+    }
+    dbfull()->TEST_WaitForCompactMemTable();
+
+    if (num < options.level0_file_num_compaction_trigger - 1) {
+      ASSERT_EQ(NumTableFilesAtLevel(0), num + 1);
+    }
+  }
+
+  dbfull()->TEST_WaitForCompact();
+  ASSERT_EQ(NumTableFilesAtLevel(0), 1);
+  for (int i = 1; i < options.num_levels ; i++) {
+    ASSERT_EQ(NumTableFilesAtLevel(i), 0);
+  }
+}
+
+TEST(DBTest, ConvertCompactionStyle) {
+  Random rnd(301);
+  int max_key_level_insert = 200;
+  int max_key_universal_insert = 600;
+
+  // Stage 1: generate a db with level compaction
+  Options options = CurrentOptions();
+  options.write_buffer_size = 100<<10; //100KB
+  options.num_levels = 4;
+  options.level0_file_num_compaction_trigger = 3;
+  options.max_bytes_for_level_base = 500<<10; // 500KB
+  options.max_bytes_for_level_multiplier = 1;
+  options.target_file_size_base = 200<<10; // 200KB
+  options.target_file_size_multiplier = 1;
+  Reopen(&options);
+
+  for (int i = 0; i <= max_key_level_insert; i++) {
+    // each value is 10K
+    ASSERT_OK(Put(Key(i), RandomString(&rnd, 10000)));
+  }
+  dbfull()->Flush(FlushOptions());
+  dbfull()->TEST_WaitForCompact();
+
+  ASSERT_GT(TotalTableFiles(), 1);
+  int non_level0_num_files = 0;
+  for (int i = 1; i < dbfull()->NumberLevels(); i++) {
+    non_level0_num_files += NumTableFilesAtLevel(i);
+  }
+  ASSERT_GT(non_level0_num_files, 0);
+
+  // Stage 2: reopen with universal compaction - should fail
+  options = CurrentOptions();
+  options.compaction_style = kCompactionStyleUniversal;
+  Status s = TryReopen(&options);
+  ASSERT_TRUE(s.IsInvalidArgument());
+
+  // Stage 3: compact into a single file and move the file to level 0
+  options = CurrentOptions();
+  options.disable_auto_compactions = true;
+  options.target_file_size_base = INT_MAX;
+  options.target_file_size_multiplier = 1;
+  options.max_bytes_for_level_base = INT_MAX;
+  options.max_bytes_for_level_multiplier = 1;
+  Reopen(&options);
+
+  dbfull()->CompactRange(nullptr, nullptr,
+                         true /* reduce level */,
+                         0    /* reduce to level 0 */);
+
+  for (int i = 0; i < dbfull()->NumberLevels(); i++) {
+    int num = NumTableFilesAtLevel(i);
+    if (i == 0) {
+      ASSERT_EQ(num, 1);
+    } else {
+      ASSERT_EQ(num, 0);
+    }
+  }
+
+  // Stage 4: re-open in universal compaction style and do some db operations
+  options = CurrentOptions();
+  options.compaction_style = kCompactionStyleUniversal;
+  options.write_buffer_size = 100<<10; //100KB
+  options.level0_file_num_compaction_trigger = 3;
+  Reopen(&options);
+
+  for (int i = max_key_level_insert / 2; i <= max_key_universal_insert; i++) {
+    ASSERT_OK(Put(Key(i), RandomString(&rnd, 10000)));
+  }
+  dbfull()->Flush(FlushOptions());
+  dbfull()->TEST_WaitForCompact();
+
+  for (int i = 1; i < dbfull()->NumberLevels(); i++) {
+    ASSERT_EQ(NumTableFilesAtLevel(i), 0);
+  }
+
+  // verify keys inserted in both level compaction style and universal
+  // compaction style
+  std::string keys_in_db;
+  Iterator* iter = dbfull()->NewIterator(ReadOptions());
+  for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
+    keys_in_db.append(iter->key().ToString());
+    keys_in_db.push_back(',');
+  }
+  delete iter;
+
+  std::string expected_keys;
+  for (int i = 0; i <= max_key_universal_insert; i++) {
+    expected_keys.append(Key(i));
+    expected_keys.push_back(',');
+  }
+
+  ASSERT_EQ(keys_in_db, expected_keys);
+}
+
 void MinLevelHelper(DBTest* self, Options& options) {
   Random rnd(301);
 
@@ -3425,7 +3840,7 @@ class ModelDB: public DB {
     }
   }
   virtual void CompactRange(const Slice* start, const Slice* end,
-                            bool reduce_level ) {
+                            bool reduce_level, int target_level) {
   }
 
   virtual int NumberLevels()
@@ -3443,7 +3858,7 @@ class ModelDB: public DB {
   return -1;
   }
 
-  virtual Status Flush(const leveldb::FlushOptions& options) {
+  virtual Status Flush(const rocksdb::FlushOptions& options) {
     Status ret;
     return ret;
   }
@@ -3454,7 +3869,8 @@ class ModelDB: public DB {
   virtual Status EnableFileDeletions() {
     return Status::OK();
   }
-  virtual Status GetLiveFiles(std::vector<std::string>&, uint64_t* size) {
+  virtual Status GetLiveFiles(std::vector<std::string>&, uint64_t* size,
+                              bool flush_memtable = true) {
     return Status::OK();
   }
 
@@ -3469,8 +3885,8 @@ class ModelDB: public DB {
   virtual SequenceNumber GetLatestSequenceNumber() {
     return 0;
   }
-  virtual Status GetUpdatesSince(leveldb::SequenceNumber,
-                                 unique_ptr<leveldb::TransactionLogIterator>*) {
+  virtual Status GetUpdatesSince(rocksdb::SequenceNumber,
+                                 unique_ptr<rocksdb::TransactionLogIterator>*) {
     return Status::NotSupported("Not supported in Model DB");
   }
 
@@ -3509,10 +3925,13 @@ class ModelDB: public DB {
   KVMap map_;
 };
 
-static std::string RandomKey(Random* rnd) {
-  int len = (rnd->OneIn(3)
-             ? 1                // Short sometimes to encourage collisions
-             : (rnd->OneIn(100) ? rnd->Skewed(10) : rnd->Uniform(10)));
+static std::string RandomKey(Random* rnd, int minimum = 0) {
+  int len;
+  do {
+    len = (rnd->OneIn(3)
+           ? 1                // Short sometimes to encourage collisions
+           : (rnd->OneIn(100) ? rnd->Skewed(10) : rnd->Uniform(10)));
+  } while (len < minimum);
   return test::RandomKey(rnd, len);
 }
 
@@ -3574,8 +3993,12 @@ TEST(DBTest, Randomized) {
     for (int step = 0; step < N; step++) {
       // TODO(sanjay): Test Get() works
       int p = rnd.Uniform(100);
+      int minimum = 0;
+      if (option_config_ == kPrefixHashRep) {
+        minimum = 1;
+      }
       if (p < 45) {                               // Put
-        k = RandomKey(&rnd);
+        k = RandomKey(&rnd, minimum);
         v = RandomString(&rnd,
                          rnd.OneIn(20)
                          ? 100 + rnd.Uniform(100)
@@ -3584,7 +4007,7 @@ TEST(DBTest, Randomized) {
         ASSERT_OK(db_->Put(WriteOptions(), k, v));
 
       } else if (p < 90) {                        // Delete
-        k = RandomKey(&rnd);
+        k = RandomKey(&rnd, minimum);
         ASSERT_OK(model.Delete(WriteOptions(), k));
         ASSERT_OK(db_->Delete(WriteOptions(), k));
 
@@ -3594,7 +4017,7 @@ TEST(DBTest, Randomized) {
         const int num = rnd.Uniform(8);
         for (int i = 0; i < num; i++) {
           if (i == 0 || !rnd.OneIn(10)) {
-            k = RandomKey(&rnd);
+            k = RandomKey(&rnd, minimum);
           } else {
             // Periodically re-use the same key from the previous iter, so
             // we have multiple entries in the write batch for the same key
@@ -3750,6 +4173,9 @@ TEST(DBTest, PrefixScan) {
   snprintf(buf, sizeof(buf), "03______:");
   prefix = Slice(buf, 8);
   key = Slice(buf, 9);
+  auto prefix_extractor = NewFixedPrefixTransform(8);
+  auto memtable_factory =
+    std::make_shared<PrefixHashRepFactory>(prefix_extractor);
 
   // db configs
   env_->count_random_reads_ = true;
@@ -3757,12 +4183,13 @@ TEST(DBTest, PrefixScan) {
   options.env = env_;
   options.block_cache = NewLRUCache(0);  // Prevent cache hits
   options.filter_policy =  NewBloomFilterPolicy(10);
-  options.prefix_extractor = NewFixedPrefixTransform(8);
+  options.prefix_extractor = prefix_extractor;
   options.whole_key_filtering = false;
   options.disable_auto_compactions = true;
   options.max_background_compactions = 2;
   options.create_if_missing = true;
   options.disable_seek_compaction = true;
+  options.memtable_factory = memtable_factory;
 
   // prefix specified, with blooms: 2 RAND I/Os
   // SeekToFirst
@@ -3816,7 +4243,6 @@ TEST(DBTest, PrefixScan) {
   ASSERT_EQ(env_->random_read_counter_.Read(), 11);
   Close();
   delete options.filter_policy;
-  delete options.prefix_extractor;
 }
 
 std::string MakeKey(unsigned int num) {
@@ -3877,16 +4303,16 @@ void BM_LogAndApply(int iters, int num_base_files) {
           buf, iters, us, ((float)us) / iters);
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 int main(int argc, char** argv) {
   if (argc > 1 && std::string(argv[1]) == "--benchmark") {
-    leveldb::BM_LogAndApply(1000, 1);
-    leveldb::BM_LogAndApply(1000, 100);
-    leveldb::BM_LogAndApply(1000, 10000);
-    leveldb::BM_LogAndApply(100, 100000);
+    rocksdb::BM_LogAndApply(1000, 1);
+    rocksdb::BM_LogAndApply(1000, 100);
+    rocksdb::BM_LogAndApply(1000, 10000);
+    rocksdb::BM_LogAndApply(100, 100000);
     return 0;
   }
 
-  return leveldb::test::RunAllTests();
+  return rocksdb::test::RunAllTests();
 }

db/dbformat.cc

@@ -6,9 +6,9 @@
 #include "db/dbformat.h"
 #include "port/port.h"
 #include "util/coding.h"
-#include "include/leveldb/perf_context.h"
+#include "util/perf_context_imp.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 static uint64_t PackSequenceAndType(uint64_t seq, ValueType t) {
   assert(seq <= kMaxSequenceNumber);
@@ -54,7 +54,7 @@ int InternalKeyComparator::Compare(const Slice& akey, const Slice& bkey) const {
   //    decreasing sequence number
   //    decreasing type (though sequence# should be enough to disambiguate)
   int r = user_comparator_->Compare(ExtractUserKey(akey), ExtractUserKey(bkey));
-  perf_context.user_key_comparison_count++;
+  BumpPerfCount(&perf_context.user_key_comparison_count);
   if (r == 0) {
     const uint64_t anum = DecodeFixed64(akey.data() + akey.size() - 8);
     const uint64_t bnum = DecodeFixed64(bkey.data() + bkey.size() - 8);
@@ -139,4 +139,4 @@ LookupKey::LookupKey(const Slice& user_key, SequenceNumber s) {
   end_ = dst;
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb

db/dbformat.h

@@ -6,16 +6,16 @@
 #define STORAGE_LEVELDB_DB_FORMAT_H_
 
 #include <stdio.h>
-#include "leveldb/comparator.h"
-#include "leveldb/db.h"
-#include "leveldb/filter_policy.h"
-#include "leveldb/slice.h"
-#include "leveldb/table_builder.h"
-#include "leveldb/types.h"
+#include "rocksdb/comparator.h"
+#include "rocksdb/db.h"
+#include "rocksdb/filter_policy.h"
+#include "rocksdb/slice.h"
+#include "rocksdb/table_builder.h"
+#include "rocksdb/types.h"
 #include "util/coding.h"
 #include "util/logging.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class InternalKey;
 
@@ -223,6 +223,6 @@ inline LookupKey::~LookupKey() {
   if (start_ != space_) delete[] start_;
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 #endif  // STORAGE_LEVELDB_DB_FORMAT_H_

db/dbformat_test.cc

@@ -6,7 +6,7 @@
 #include "util/logging.h"
 #include "util/testharness.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 static std::string IKey(const std::string& user_key,
                         uint64_t seq,
@@ -105,8 +105,8 @@ TEST(FormatTest, InternalKeyShortestSuccessor) {
             ShortSuccessor(IKey("\xff\xff", 100, kTypeValue)));
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 int main(int argc, char** argv) {
-  return leveldb::test::RunAllTests();
+  return rocksdb::test::RunAllTests();
 }

db/deletefile_test.cc

@@ -0,0 +1,188 @@
+// 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 "rocksdb/db.h"
+#include "db/db_impl.h"
+#include "db/filename.h"
+#include "db/version_set.h"
+#include "db/write_batch_internal.h"
+#include "util/testharness.h"
+#include "util/testutil.h"
+#include "boost/lexical_cast.hpp"
+#include "rocksdb/env.h"
+#include <vector>
+#include <boost/algorithm/string.hpp>
+#include <stdlib.h>
+#include <map>
+
+namespace rocksdb {
+
+class DeleteFileTest {
+ public:
+  std::string dbname_;
+  Options options_;
+  DB* db_;
+  Env* env_;
+  int numlevels_;
+
+  DeleteFileTest() {
+    db_ = nullptr;
+    env_ = Env::Default();
+    options_.write_buffer_size = 1024*1024*1000;
+    options_.target_file_size_base = 1024*1024*1000;
+    options_.max_bytes_for_level_base = 1024*1024*1000;
+    dbname_ = test::TmpDir() + "/deletefile_test";
+    DestroyDB(dbname_, options_);
+    numlevels_ = 7;
+    ASSERT_OK(ReopenDB(true));
+  }
+
+  Status ReopenDB(bool create) {
+    delete db_;
+    if (create) {
+      DestroyDB(dbname_, options_);
+    }
+    db_ = nullptr;
+    options_.create_if_missing = create;
+    return DB::Open(options_, dbname_, &db_);
+  }
+
+  void CloseDB() {
+    delete db_;
+  }
+
+  void AddKeys(int numkeys, int startkey = 0) {
+    WriteOptions options;
+    options.sync = false;
+    ReadOptions roptions;
+    for (int i = startkey; i < (numkeys + startkey) ; i++) {
+      std::string temp = boost::lexical_cast<std::string>(i);
+      Slice key(temp);
+      Slice value(temp);
+      ASSERT_OK(db_->Put(options, key, value));
+    }
+  }
+
+  int numKeysInLevels(
+    std::vector<LiveFileMetaData> &metadata,
+    std::vector<int> *keysperlevel = nullptr) {
+
+    if (keysperlevel != nullptr) {
+      keysperlevel->resize(numlevels_);
+    }
+
+    int numKeys = 0;
+    for (size_t i = 0; i < metadata.size(); i++) {
+      int startkey = atoi(metadata[i].smallestkey.c_str());
+      int endkey = atoi(metadata[i].largestkey.c_str());
+      int numkeysinfile = (endkey - startkey + 1);
+      numKeys += numkeysinfile;
+      if (keysperlevel != nullptr) {
+        (*keysperlevel)[(int)metadata[i].level] += numkeysinfile;
+      }
+      fprintf(stderr, "level %d name %s smallest %s largest %s\n",
+              metadata[i].level, metadata[i].name.c_str(),
+              metadata[i].smallestkey.c_str(),
+              metadata[i].largestkey.c_str());
+    }
+    return numKeys;
+  }
+
+  void CreateTwoLevels() {
+    AddKeys(50000, 10000);
+    DBImpl* dbi = reinterpret_cast<DBImpl*>(db_);
+    ASSERT_OK(dbi->TEST_CompactMemTable());
+    ASSERT_OK(dbi->TEST_WaitForCompactMemTable());
+
+    AddKeys(50000, 10000);
+    ASSERT_OK(dbi->TEST_CompactMemTable());
+    ASSERT_OK(dbi->TEST_WaitForCompactMemTable());
+  }
+
+};
+
+TEST(DeleteFileTest, AddKeysAndQueryLevels) {
+  CreateTwoLevels();
+  std::vector<LiveFileMetaData> metadata;
+  std::vector<int> keysinlevel;
+  db_->GetLiveFilesMetaData(&metadata);
+
+  std::string level1file = "";
+  int level1keycount = 0;
+  std::string level2file = "";
+  int level2keycount = 0;
+  int level1index = 0;
+  int level2index = 1;
+
+  ASSERT_EQ((int)metadata.size(), 2);
+  if (metadata[0].level == 2) {
+    level1index = 1;
+    level2index = 0;
+  }
+
+  level1file = metadata[level1index].name;
+  int startkey = atoi(metadata[level1index].smallestkey.c_str());
+  int endkey = atoi(metadata[level1index].largestkey.c_str());
+  level1keycount = (endkey - startkey + 1);
+  level2file = metadata[level2index].name;
+  startkey = atoi(metadata[level2index].smallestkey.c_str());
+  endkey = atoi(metadata[level2index].largestkey.c_str());
+  level2keycount = (endkey - startkey + 1);
+
+  // COntrolled setup. Levels 1 and 2 should both have 50K files.
+  // This is a little fragile as it depends on the current
+  // compaction heuristics.
+  ASSERT_EQ(level1keycount, 50000);
+  ASSERT_EQ(level2keycount, 50000);
+
+  Status status = db_->DeleteFile("0.sst");
+  ASSERT_TRUE(status.IsInvalidArgument());
+
+  // intermediate level files cannot be deleted.
+  status = db_->DeleteFile(level1file);
+  ASSERT_TRUE(status.IsInvalidArgument());
+
+  // Lowest level file deletion should succeed.
+  ASSERT_OK(db_->DeleteFile(level2file));
+
+  CloseDB();
+}
+
+
+TEST(DeleteFileTest, DeleteFileWithIterator) {
+  CreateTwoLevels();
+  ReadOptions options;
+  Iterator* it = db_->NewIterator(options);
+  std::vector<LiveFileMetaData> metadata;
+  db_->GetLiveFilesMetaData(&metadata);
+
+  std::string level2file = "";
+
+  ASSERT_EQ((int)metadata.size(), 2);
+  if (metadata[0].level == 1) {
+    level2file = metadata[1].name;
+  } else {
+    level2file = metadata[0].name;
+  }
+
+  Status status = db_->DeleteFile(level2file);
+  fprintf(stdout, "Deletion status %s: %s\n",
+          level2file.c_str(), status.ToString().c_str());
+  ASSERT_TRUE(status.ok());
+  it->SeekToFirst();
+  int numKeysIterated = 0;
+  while(it->Valid()) {
+    numKeysIterated++;
+    it->Next();
+  }
+  ASSERT_EQ(numKeysIterated, 50000);
+  delete it;
+  CloseDB();
+}
+} //namespace rocksdb
+
+int main(int argc, char** argv) {
+  return rocksdb::test::RunAllTests();
+}
+

db/filename.cc

@@ -7,10 +7,10 @@
 #include <ctype.h>
 #include <stdio.h>
 #include "db/dbformat.h"
-#include "leveldb/env.h"
+#include "rocksdb/env.h"
 #include "util/logging.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 // Given a path, flatten the path name by replacing all chars not in
 // {[0-9,a-z,A-Z,-,_,.]} with _. And append '\0' at the end.
@@ -46,13 +46,10 @@ extern Status WriteStringToFileSync(Env* env, const Slice& data,
 static std::string MakeFileName(const std::string& name, uint64_t number,
                                 const char* suffix) {
   char buf[100];
-  snprintf(buf, sizeof(buf), "%06llu.%s",
+  snprintf(buf, sizeof(buf), "/%06llu.%s",
            static_cast<unsigned long long>(number),
            suffix);
-  if (name.empty()) {
-    return buf;
-  }
-  return name + "/" + buf;
+  return name + buf;
 }
 
 std::string LogFileName(const std::string& name, uint64_t number) {
@@ -65,7 +62,7 @@ std::string ArchivalDirectory(const std::string& dbname) {
 }
 std::string ArchivedLogFileName(const std::string& name, uint64_t number) {
   assert(number > 0);
-  return MakeFileName(name + "/archive", number, "log");
+  return MakeFileName(name + "/" + ARCHIVAL_DIR, number, "log");
 }
 
 std::string TableFileName(const std::string& name, uint64_t number) {
@@ -133,10 +130,14 @@ std::string MetaDatabaseName(const std::string& dbname, uint64_t number) {
 //    dbname/MANIFEST-[0-9]+
 //    dbname/[0-9]+.(log|sst)
 //    dbname/METADB-[0-9]+
+//    Disregards / at the beginning
 bool ParseFileName(const std::string& fname,
                    uint64_t* number,
                    FileType* type) {
   Slice rest(fname);
+  if (fname.length() > 1 && fname[0] == '/') {
+    rest.remove_prefix(1);
+  }
   if (rest == "CURRENT") {
     *number = 0;
     *type = kCurrentFile;
@@ -217,4 +218,4 @@ Status SetCurrentFile(Env* env, const std::string& dbname,
   return s;
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb

db/filename.h

@@ -9,11 +9,11 @@
 
 #include <stdint.h>
 #include <string>
-#include "leveldb/slice.h"
-#include "leveldb/status.h"
+#include "rocksdb/slice.h"
+#include "rocksdb/status.h"
 #include "port/port.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class Env;
 
@@ -92,6 +92,6 @@ extern Status SetCurrentFile(Env* env, const std::string& dbname,
                              uint64_t descriptor_number);
 
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 #endif  // STORAGE_LEVELDB_DB_FILENAME_H_

db/filename_test.cc

@@ -9,7 +9,7 @@
 #include "util/logging.h"
 #include "util/testharness.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class FileNameTest { };
 
@@ -128,8 +128,8 @@ TEST(FileNameTest, Construction) {
   ASSERT_EQ(kMetaDatabase, type);
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 int main(int argc, char** argv) {
-  return leveldb::test::RunAllTests();
+  return rocksdb::test::RunAllTests();
 }

db/log_format.h

@@ -8,7 +8,7 @@
 #ifndef STORAGE_LEVELDB_DB_LOG_FORMAT_H_
 #define STORAGE_LEVELDB_DB_LOG_FORMAT_H_
 
-namespace leveldb {
+namespace rocksdb {
 namespace log {
 
 enum RecordType {
@@ -30,6 +30,6 @@ static const unsigned int kBlockSize = 32768;
 static const int kHeaderSize = 4 + 1 + 2;
 
 }  // namespace log
-}  // namespace leveldb
+}  // namespace rocksdb
 
 #endif  // STORAGE_LEVELDB_DB_LOG_FORMAT_H_

db/log_reader.cc

@@ -5,11 +5,11 @@
 #include "db/log_reader.h"
 
 #include <stdio.h>
-#include "leveldb/env.h"
+#include "rocksdb/env.h"
 #include "util/coding.h"
 #include "util/crc32c.h"
 
-namespace leveldb {
+namespace rocksdb {
 namespace log {
 
 Reader::Reporter::~Reporter() {
@@ -256,4 +256,4 @@ unsigned int Reader::ReadPhysicalRecord(Slice* result) {
 }
 
 }  // namespace log
-}  // namespace leveldb
+}  // namespace rocksdb

db/log_reader.h

@@ -9,10 +9,10 @@
 #include <stdint.h>
 
 #include "db/log_format.h"
-#include "leveldb/slice.h"
-#include "leveldb/status.h"
+#include "rocksdb/slice.h"
+#include "rocksdb/status.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class SequentialFile;
 using std::unique_ptr;
@@ -118,6 +118,6 @@ class Reader {
 };
 
 }  // namespace log
-}  // namespace leveldb
+}  // namespace rocksdb
 
 #endif  // STORAGE_LEVELDB_DB_LOG_READER_H_

db/log_test.cc

@@ -4,13 +4,13 @@
 
 #include "db/log_reader.h"
 #include "db/log_writer.h"
-#include "leveldb/env.h"
+#include "rocksdb/env.h"
 #include "util/coding.h"
 #include "util/crc32c.h"
 #include "util/random.h"
 #include "util/testharness.h"
 
-namespace leveldb {
+namespace rocksdb {
 namespace log {
 
 // Construct a string of the specified length made out of the supplied
@@ -516,8 +516,8 @@ TEST(LogTest, ReadPastEnd) {
 }
 
 }  // namespace log
-}  // namespace leveldb
+}  // namespace rocksdb
 
 int main(int argc, char** argv) {
-  return leveldb::test::RunAllTests();
+  return rocksdb::test::RunAllTests();
 }

db/log_writer.cc

@@ -5,11 +5,11 @@
 #include "db/log_writer.h"
 
 #include <stdint.h>
-#include "leveldb/env.h"
+#include "rocksdb/env.h"
 #include "util/coding.h"
 #include "util/crc32c.h"
 
-namespace leveldb {
+namespace rocksdb {
 namespace log {
 
 Writer::Writer(unique_ptr<WritableFile>&& dest)
@@ -100,4 +100,4 @@ Status Writer::EmitPhysicalRecord(RecordType t, const char* ptr, size_t n) {
 }
 
 }  // namespace log
-}  // namespace leveldb
+}  // namespace rocksdb

db/log_writer.h

@@ -8,10 +8,10 @@
 #include <memory>
 #include <stdint.h>
 #include "db/log_format.h"
-#include "leveldb/slice.h"
-#include "leveldb/status.h"
+#include "rocksdb/slice.h"
+#include "rocksdb/status.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class WritableFile;
 
@@ -49,6 +49,6 @@ class Writer {
 };
 
 }  // namespace log
-}  // namespace leveldb
+}  // namespace rocksdb
 
 #endif  // STORAGE_LEVELDB_DB_LOG_WRITER_H_

db/memtable.cc

@@ -7,20 +7,14 @@
 #include <memory>
 
 #include "db/dbformat.h"
-#include "leveldb/comparator.h"
-#include "leveldb/env.h"
-#include "leveldb/iterator.h"
-#include "leveldb/merge_operator.h"
+#include "rocksdb/comparator.h"
+#include "rocksdb/env.h"
+#include "rocksdb/iterator.h"
+#include "rocksdb/merge_operator.h"
 #include "util/coding.h"
+#include "util/murmurhash.h"
 
-namespace leveldb {
-
-static Slice GetLengthPrefixedSlice(const char* data) {
-  uint32_t len;
-  const char* p = data;
-  p = GetVarint32Ptr(p, p + 5, &len);  // +5: we assume "p" is not corrupted
-  return Slice(p, len);
-}
+namespace rocksdb {
 
 MemTable::MemTable(const InternalKeyComparator& cmp,
                    std::shared_ptr<MemTableRepFactory> table_factory,
@@ -35,6 +29,7 @@ MemTable::MemTable(const InternalKeyComparator& cmp,
       file_number_(0),
       edit_(numlevel),
       first_seqno_(0),
+      mem_next_logfile_number_(0),
       mem_logfile_number_(0) { }
 
 MemTable::~MemTable() {
@@ -42,7 +37,8 @@ MemTable::~MemTable() {
 }
 
 size_t MemTable::ApproximateMemoryUsage() {
-  return arena_impl_.ApproximateMemoryUsage();
+  return arena_impl_.ApproximateMemoryUsage() +
+    table_->ApproximateMemoryUsage();
 }
 
 int MemTable::KeyComparator::operator()(const char* aptr, const char* bptr)
@@ -53,6 +49,11 @@ int MemTable::KeyComparator::operator()(const char* aptr, const char* bptr)
   return comparator.Compare(a, b);
 }
 
+Slice MemTableRep::UserKey(const char* key) const {
+  Slice slice = GetLengthPrefixedSlice(key);
+  return Slice(slice.data(), slice.size() - 8);
+}
+
 // Encode a suitable internal key target for "target" and return it.
 // Uses *scratch as scratch space, and the returned pointer will point
 // into this scratch space.
@@ -68,6 +69,9 @@ class MemTableIterator: public Iterator {
   explicit MemTableIterator(MemTableRep* table)
     : iter_(table->GetIterator()) { }
 
+  MemTableIterator(MemTableRep* table, const Slice* prefix)
+    : iter_(table->GetPrefixIterator(*prefix)) { }
+
   virtual bool Valid() const { return iter_->Valid(); }
   virtual void Seek(const Slice& k) { iter_->Seek(EncodeKey(&tmp_, k)); }
   virtual void SeekToFirst() { iter_->SeekToFirst(); }
@@ -93,8 +97,12 @@ class MemTableIterator: public Iterator {
   void operator=(const MemTableIterator&);
 };
 
-Iterator* MemTable::NewIterator() {
-  return new MemTableIterator(table_.get());
+Iterator* MemTable::NewIterator(const Slice* prefix) {
+  if (prefix) {
+    return new MemTableIterator(table_.get(), prefix);
+  } else {
+    return new MemTableIterator(table_.get());
+  }
 }
 
 void MemTable::Add(SequenceNumber s, ValueType type,
@@ -132,7 +140,8 @@ void MemTable::Add(SequenceNumber s, ValueType type,
 bool MemTable::Get(const LookupKey& key, std::string* value, Status* s,
                    std::deque<std::string>* operands, const Options& options) {
   Slice memkey = key.memtable_key();
-  std::shared_ptr<MemTableRep::Iterator> iter(table_.get()->GetIterator());
+  std::shared_ptr<MemTableRep::Iterator> iter(
+    table_->GetIterator(key.user_key()));
   iter->Seek(memkey.data());
 
   // It is the caller's responsibility to allocate/delete operands list
@@ -229,4 +238,4 @@ bool MemTable::Get(const LookupKey& key, std::string* value, Status* s,
   return false;
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb

db/memtable.h

@@ -11,11 +11,11 @@
 #include "db/dbformat.h"
 #include "db/skiplist.h"
 #include "db/version_set.h"
-#include "leveldb/db.h"
-#include "leveldb/memtablerep.h"
+#include "rocksdb/db.h"
+#include "rocksdb/memtablerep.h"
 #include "util/arena_impl.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class Mutex;
 class MemTableIterator;
@@ -61,7 +61,11 @@ class MemTable {
   // while the returned iterator is live.  The keys returned by this
   // iterator are internal keys encoded by AppendInternalKey in the
   // db/dbformat.{h,cc} module.
-  Iterator* NewIterator();
+  //
+  // If a prefix is supplied, it is passed to the underlying MemTableRep as a
+  // hint that the iterator only need to support access to keys with that
+  // prefix.
+  Iterator* NewIterator(const Slice* prefix = nullptr);
 
   // Add an entry into memtable that maps key to value at the
   // specified sequence number and with the specified type.
@@ -88,6 +92,14 @@ class MemTable {
   // into the memtable
   SequenceNumber GetFirstSequenceNumber() { return first_seqno_; }
 
+  // Returns the next active logfile number when this memtable is about to
+  // be flushed to storage
+  uint64_t GetNextLogNumber() { return mem_next_logfile_number_; }
+
+  // Sets the next active logfile number when this memtable is about to
+  // be flushed to storage
+  void SetNextLogNumber(uint64_t num) { mem_next_logfile_number_ = num; }
+
   // Returns the logfile number that can be safely deleted when this
   // memstore is flushed to storage
   uint64_t GetLogNumber() { return mem_logfile_number_; }
@@ -96,6 +108,9 @@ class MemTable {
   // memstore is flushed to storage
   void SetLogNumber(uint64_t num) { mem_logfile_number_ = num; }
 
+  // Notify the underlying storage that no more items will be added
+  void MarkImmutable() { table_->MarkReadOnly(); }
+
  private:
   ~MemTable();  // Private since only Unref() should be used to delete it
   friend class MemTableIterator;
@@ -120,6 +135,10 @@ class MemTable {
   SequenceNumber first_seqno_;
 
   // The log files earlier than this number can be deleted.
+  uint64_t mem_next_logfile_number_;
+
+  // The log file that backs this memtable (to be deleted when
+  // memtable flush is done)
   uint64_t mem_logfile_number_;
 
   // No copying allowed
@@ -127,6 +146,6 @@ class MemTable {
   void operator=(const MemTable&);
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 #endif  // STORAGE_LEVELDB_DB_MEMTABLE_H_

db/memtablelist.cc

@@ -4,13 +4,13 @@
 #include "db/memtablelist.h"
 
 #include <string>
-#include "leveldb/db.h"
+#include "rocksdb/db.h"
 #include "db/memtable.h"
-#include "leveldb/env.h"
-#include "leveldb/iterator.h"
+#include "rocksdb/env.h"
+#include "rocksdb/iterator.h"
 #include "util/coding.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class InternalKeyComparator;
 class Mutex;
@@ -42,7 +42,8 @@ int MemTableList::size() {
 // Returns true if there is at least one memtable on which flush has
 // not yet started.
 bool MemTableList::IsFlushPending(int min_write_buffer_number_to_merge) {
-  if (num_flush_not_started_ >= min_write_buffer_number_to_merge) {
+  if ((flush_requested_ && num_flush_not_started_ >= 1) ||
+      (num_flush_not_started_ >= min_write_buffer_number_to_merge)) {
     assert(imm_flush_needed.NoBarrier_Load() != nullptr);
     return true;
   }
@@ -63,6 +64,7 @@ void MemTableList::PickMemtablesToFlush(std::vector<MemTable*>* ret) {
       ret->push_back(m);
     }
   }
+  flush_requested_ = false; // start-flush request is complete
 }
 
 // Record a successful flush in the manifest file
@@ -172,6 +174,7 @@ void MemTableList::Add(MemTable* m) {
   assert(size_ >= num_flush_not_started_);
   size_++;
   memlist_.push_front(m);
+  m->MarkImmutable();
   num_flush_not_started_++;
   if (num_flush_not_started_ == 1) {
     imm_flush_needed.Release_Store((void *)1);
@@ -207,4 +210,4 @@ void MemTableList::GetMemTables(std::vector<MemTable*>* output) {
   }
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb

db/memtablelist.h

@@ -6,12 +6,12 @@
 #include <string>
 #include <list>
 #include <deque>
-#include "leveldb/db.h"
+#include "rocksdb/db.h"
 #include "db/dbformat.h"
 #include "db/skiplist.h"
 #include "memtable.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class InternalKeyComparator;
 class Mutex;
@@ -29,7 +29,8 @@ class MemTableList {
  public:
   // A list of memtables.
   MemTableList() : size_(0), num_flush_not_started_(0),
-    commit_in_progress_(false) {
+    commit_in_progress_(false),
+    flush_requested_(false) {
     imm_flush_needed.Release_Store(nullptr);
   }
   ~MemTableList() {};
@@ -76,6 +77,9 @@ class MemTableList {
   // Returns the list of underlying memtables.
   void GetMemTables(std::vector<MemTable*>* list);
 
+  // Request a flush of all existing memtables to storage
+  void FlushRequested() { flush_requested_ = true; }
+
   // Copying allowed
   // MemTableList(const MemTableList&);
   // void operator=(const MemTableList&);
@@ -90,8 +94,11 @@ class MemTableList {
   // committing in progress
   bool commit_in_progress_;
 
+  // Requested a flush of all memtables to storage
+  bool flush_requested_;
+
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 #endif  // STORAGE_LEVELDB_DB_MEMTABLELIST_H_

db/merge_helper.cc

@@ -1,12 +1,12 @@
 #include "merge_helper.h"
 #include "db/dbformat.h"
-#include "leveldb/comparator.h"
-#include "leveldb/db.h"
-#include "leveldb/merge_operator.h"
+#include "rocksdb/comparator.h"
+#include "rocksdb/db.h"
+#include "rocksdb/merge_operator.h"
 #include <string>
 #include <stdio.h>
 
-namespace leveldb {
+namespace rocksdb {
 
 // PRE:  iter points to the first merge type entry
 // POST: iter points to the first entry beyond the merge process (or the end)
@@ -189,4 +189,4 @@ void MergeHelper::MergeUntil(Iterator* iter, SequenceNumber stop_before,
   }
 }
 
-} // namespace leveldb
+} // namespace rocksdb

db/merge_helper.h

@@ -2,12 +2,12 @@
 #define MERGE_HELPER_H
 
 #include "db/dbformat.h"
-#include "leveldb/slice.h"
-#include "leveldb/statistics.h"
+#include "rocksdb/slice.h"
+#include "rocksdb/statistics.h"
 #include <string>
 #include <deque>
 
-namespace leveldb {
+namespace rocksdb {
 
 class Comparator;
 class Iterator;
@@ -93,6 +93,6 @@ class MergeHelper {
   bool success_;
 };
 
-} // namespace leveldb
+} // namespace rocksdb
 
 #endif

db/merge_operator.cc

@@ -5,9 +5,9 @@
  * Copyright 2013 Facebook
  */
 
-#include "leveldb/merge_operator.h"
+#include "rocksdb/merge_operator.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 // Given a "real" merge from the library, call the user's
 // associative merge function one-by-one on each of the operands.
@@ -48,4 +48,4 @@ bool AssociativeMergeOperator::PartialMerge(
   return Merge(key, &left_operand, right_operand, new_value, logger);
 }
 
-} // namespace leveldb
+} // namespace rocksdb

db/merge_test.cc

@@ -2,11 +2,11 @@
 #include <memory>
 #include <iostream>
 
-#include "leveldb/cache.h"
-#include "leveldb/comparator.h"
-#include "leveldb/db.h"
-#include "leveldb/env.h"
-#include "leveldb/merge_operator.h"
+#include "rocksdb/cache.h"
+#include "rocksdb/comparator.h"
+#include "rocksdb/db.h"
+#include "rocksdb/env.h"
+#include "rocksdb/merge_operator.h"
 #include "db/dbformat.h"
 #include "db/db_impl.h"
 #include "utilities/merge_operators.h"
@@ -14,7 +14,7 @@
 #include "utilities/utility_db.h"
 
 using namespace std;
-using namespace leveldb;
+using namespace rocksdb;
 
 
 std::shared_ptr<DB> OpenDb(const string& dbname, const bool ttl = false) {

db/perf_context.cc

@@ -1,12 +0,0 @@
-#include "include/leveldb/perf_context.h"
-
-
-namespace leveldb {
-
-void PerfContext::Reset() {
-  user_key_comparison_count = 0;
-}
-
-__thread PerfContext perf_context;
-
-}

db/perf_context_test.cc

@@ -1,20 +1,52 @@
+#include <algorithm>
 #include <iostream>
+#include <vector>
+#include "/usr/include/valgrind/callgrind.h"
 
-#include "leveldb/db.h"
-#include "leveldb/perf_context.h"
+#include "rocksdb/db.h"
+#include "rocksdb/perf_context.h"
+#include "util/histogram.h"
+#include "util/stop_watch.h"
 #include "util/testharness.h"
 
 
-namespace leveldb {
+bool FLAGS_random_key = false;
+bool FLAGS_use_set_based_memetable = false;
+int FLAGS_total_keys = 100;
+int FLAGS_write_buffer_size = 1000000000;
+int FLAGS_max_write_buffer_number = 8;
+int FLAGS_min_write_buffer_number_to_merge = 7;
 
 // Path to the database on file system
-const std::string kDbName = test::TmpDir() + "/perf_context_test";
+const std::string kDbName = rocksdb::test::TmpDir() + "/perf_context_test";
+
+void SeekToFirst(rocksdb::Iterator* iter) {
+//  std::cout << "Press a key to continue:";
+//  std::string s;
+//  std::cin >> s;
+  iter->SeekToFirst();
+//  std::cout << "Press a key to continue:";
+//  std::string s2;
+//  std::cin >> s2;
+}
+
+namespace rocksdb {
 
 std::shared_ptr<DB> OpenDb() {
     DB* db;
     Options options;
     options.create_if_missing = true;
-    options.write_buffer_size = 1000000000;     // give it a big memtable
+    options.write_buffer_size = FLAGS_write_buffer_size;
+    options.max_write_buffer_number = FLAGS_max_write_buffer_number;
+    options.min_write_buffer_number_to_merge =
+      FLAGS_min_write_buffer_number_to_merge;
+
+    if (FLAGS_use_set_based_memetable) {
+      auto prefix_extractor = rocksdb::NewFixedPrefixTransform(0);
+      options.memtable_factory =
+        std::make_shared<rocksdb::PrefixHashRepFactory>(prefix_extractor);
+    }
+
     Status s = DB::Open(options, kDbName,  &db);
     ASSERT_OK(s);
     return std::shared_ptr<DB>(db);
@@ -22,10 +54,125 @@ std::shared_ptr<DB> OpenDb() {
 
 class PerfContextTest { };
 
-int kTotalKeys = 100;
+TEST(PerfContextTest, SeekIntoDeletion) {
+  DestroyDB(kDbName, Options());
+  auto db = OpenDb();
+  WriteOptions write_options;
+  ReadOptions read_options;
 
-TEST(PerfContextTest, KeyComparisonCount) {
+  for (int i = 0; i < FLAGS_total_keys; ++i) {
+    std::string key = "k" + std::to_string(i);
+    std::string value = "v" + std::to_string(i);
 
+    db->Put(write_options, key, value);
+  }
+
+  for (int i = 0; i < FLAGS_total_keys -1 ; ++i) {
+    std::string key = "k" + std::to_string(i);
+    db->Delete(write_options, key);
+  }
+
+  HistogramImpl hist_get;
+  HistogramImpl hist_get_time;
+  for (int i = 0; i < FLAGS_total_keys - 1; ++i) {
+    std::string key = "k" + std::to_string(i);
+    std::string value;
+
+    perf_context.Reset();
+    StopWatchNano timer(Env::Default(), true);
+    auto status = db->Get(read_options, key, &value);
+    auto elapsed_nanos = timer.ElapsedNanos();
+    ASSERT_TRUE(status.IsNotFound());
+    hist_get.Add(perf_context.user_key_comparison_count);
+    hist_get_time.Add(elapsed_nanos);
+  }
+
+  std::cout << "Get uesr key comparison: \n" << hist_get.ToString()
+            << "Get time: \n" << hist_get_time.ToString();
+
+  HistogramImpl hist_seek_to_first;
+  std::unique_ptr<Iterator> iter(db->NewIterator(read_options));
+
+  perf_context.Reset();
+  StopWatchNano timer(Env::Default(), true);
+  //CALLGRIND_ZERO_STATS;
+  SeekToFirst(iter.get());
+  //iter->SeekToFirst();
+  //CALLGRIND_DUMP_STATS;
+  hist_seek_to_first.Add(perf_context.user_key_comparison_count);
+  auto elapsed_nanos = timer.ElapsedNanos();
+
+  std::cout << "SeekToFirst uesr key comparison: \n" << hist_seek_to_first.ToString()
+            << "ikey skipped: " << perf_context.internal_key_skipped_count << "\n"
+            << "idelete skipped: " << perf_context.internal_delete_skipped_count << "\n"
+            << "elapsed: " << elapsed_nanos << "\n";
+
+  HistogramImpl hist_seek;
+  for (int i = 0; i < FLAGS_total_keys; ++i) {
+    std::unique_ptr<Iterator> iter(db->NewIterator(read_options));
+    std::string key = "k" + std::to_string(i);
+
+    perf_context.Reset();
+    StopWatchNano timer(Env::Default(), true);
+    iter->Seek(key);
+    auto elapsed_nanos = timer.ElapsedNanos();
+    hist_seek.Add(perf_context.user_key_comparison_count);
+    std::cout << "seek cmp: " << perf_context.user_key_comparison_count
+              << " ikey skipped " << perf_context.internal_key_skipped_count
+              << " idelete skipped " << perf_context.internal_delete_skipped_count
+              << " elapsed: " << elapsed_nanos << "ns\n";
+
+    perf_context.Reset();
+    ASSERT_TRUE(iter->Valid());
+    StopWatchNano timer2(Env::Default(), true);
+    iter->Next();
+    auto elapsed_nanos2 = timer2.ElapsedNanos();
+    std::cout << "next cmp: " << perf_context.user_key_comparison_count
+              << "elapsed: " << elapsed_nanos2 << "ns\n";
+  }
+
+  std::cout << "Seek uesr key comparison: \n" << hist_seek.ToString();
+}
+
+TEST(PerfContextTest, StopWatchNanoOverhead) {
+  // profile the timer cost by itself!
+  const int kTotalIterations = 1000000;
+  std::vector<uint64_t> timings(kTotalIterations);
+
+  StopWatchNano timer(Env::Default(), true);
+  for (auto& timing : timings) {
+    timing = timer.ElapsedNanos(true /* reset */);
+  }
+
+  HistogramImpl histogram;
+  for (const auto timing : timings) {
+    histogram.Add(timing);
+  }
+
+  std::cout << histogram.ToString();
+}
+
+TEST(PerfContextTest, StopWatchOverhead) {
+  // profile the timer cost by itself!
+  const int kTotalIterations = 1000000;
+  std::vector<uint64_t> timings(kTotalIterations);
+
+  StopWatch timer(Env::Default());
+  for (auto& timing : timings) {
+    timing = timer.ElapsedMicros();
+  }
+
+  HistogramImpl histogram;
+  uint64_t prev_timing = 0;
+  for (const auto timing : timings) {
+    histogram.Add(timing - prev_timing);
+    prev_timing = timing;
+  }
+
+  std::cout << histogram.ToString();
+}
+
+void ProfileKeyComparison() {
   DestroyDB(kDbName, Options());    // Start this test with a fresh DB
 
   auto db = OpenDb();
@@ -33,49 +180,141 @@ TEST(PerfContextTest, KeyComparisonCount) {
   WriteOptions write_options;
   ReadOptions read_options;
 
-  uint64_t total_user_key_comparison_get = 0;
-  uint64_t total_user_key_comparison_put = 0;
-  uint64_t max_user_key_comparison_get = 0;
+  HistogramImpl hist_put;
+  HistogramImpl hist_get;
 
-  std::cout << "Inserting " << kTotalKeys << " key/value pairs\n...\n";
+  std::cout << "Inserting " << FLAGS_total_keys << " key/value pairs\n...\n";
 
-  for (int i = 0; i < kTotalKeys; ++i) {
+  std::vector<int> keys;
+  for (int i = 0; i < FLAGS_total_keys; ++i) {
+    keys.push_back(i);
+  }
+
+  if (FLAGS_random_key) {
+    std::random_shuffle(keys.begin(), keys.end());
+  }
+
+  for (const int i : keys) {
     std::string key = "k" + std::to_string(i);
     std::string value = "v" + std::to_string(i);
 
     perf_context.Reset();
     db->Put(write_options, key, value);
-    total_user_key_comparison_put += perf_context.user_key_comparison_count;
+    hist_put.Add(perf_context.user_key_comparison_count);
 
     perf_context.Reset();
     db->Get(read_options, key, &value);
-    total_user_key_comparison_get += perf_context.user_key_comparison_count;
-    max_user_key_comparison_get =
-      std::max(max_user_key_comparison_get,
-               perf_context.user_key_comparison_count);
+    hist_get.Add(perf_context.user_key_comparison_count);
   }
 
-  std::cout << "total user key comparison get: "
-            << total_user_key_comparison_get << "\n"
-            << "total user key comparison put: "
-            << total_user_key_comparison_put << "\n"
-            << "max user key comparison get: "
-            << max_user_key_comparison_get << "\n"
-            << "avg user key comparison get:"
-            << total_user_key_comparison_get/kTotalKeys << "\n";
+  std::cout << "Put uesr key comparison: \n" << hist_put.ToString()
+            << "Get uesr key comparison: \n" << hist_get.ToString();
 
 }
 
+TEST(PerfContextTest, KeyComparisonCount) {
+  SetPerfLevel(kEnableCount);
+  ProfileKeyComparison();
 
+  SetPerfLevel(kDisable);
+  ProfileKeyComparison();
+
+  SetPerfLevel(kEnableTime);
+  ProfileKeyComparison();
 }
 
+// make perf_context_test
+// export LEVELDB_TESTS=PerfContextTest.SeekKeyComparison
+// For one memtable:
+// ./perf_context_test --write_buffer_size=500000 --total_keys=10000
+// For two memtables:
+// ./perf_context_test --write_buffer_size=250000 --total_keys=10000
+// Specify --random_key=1 to shuffle the key before insertion
+// Results show that, for sequential insertion, worst-case Seek Key comparison
+// is close to the total number of keys (linear), when there is only one
+// memtable. When there are two memtables, even the avg Seek Key comparison
+// starts to become linear to the input size.
+
+TEST(PerfContextTest, SeekKeyComparison) {
+  DestroyDB(kDbName, Options());
+  auto db = OpenDb();
+  WriteOptions write_options;
+  ReadOptions read_options;
+
+  std::cout << "Inserting " << FLAGS_total_keys << " key/value pairs\n...\n";
+
+  std::vector<int> keys;
+  for (int i = 0; i < FLAGS_total_keys; ++i) {
+    keys.push_back(i);
+  }
+
+  if (FLAGS_random_key) {
+    std::random_shuffle(keys.begin(), keys.end());
+  }
+
+  for (const int i : keys) {
+    std::string key = "k" + std::to_string(i);
+    std::string value = "v" + std::to_string(i);
+
+    db->Put(write_options, key, value);
+  }
+
+  HistogramImpl hist_seek;
+  HistogramImpl hist_next;
+
+  for (int i = 0; i < FLAGS_total_keys; ++i) {
+    std::string key = "k" + std::to_string(i);
+    std::string value = "v" + std::to_string(i);
+
+    std::unique_ptr<Iterator> iter(db->NewIterator(read_options));
+    perf_context.Reset();
+    iter->Seek(key);
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ(iter->value().ToString(), value);
+    hist_seek.Add(perf_context.user_key_comparison_count);
+  }
+
+  std::unique_ptr<Iterator> iter(db->NewIterator(read_options));
+  for (iter->SeekToFirst(); iter->Valid();) {
+    perf_context.Reset();
+    iter->Next();
+    hist_next.Add(perf_context.user_key_comparison_count);
+  }
+
+  std::cout << "Seek:\n" << hist_seek.ToString()
+            << "Next:\n" << hist_next.ToString();
+}
+
+}
 
 int main(int argc, char** argv) {
 
-  if (argc > 1) {
-    leveldb::kTotalKeys = std::stoi(argv[1]);
+  for (int i = 1; i < argc; i++) {
+    int n;
+    char junk;
+
+    if (sscanf(argv[i], "--write_buffer_size=%d%c", &n, &junk) == 1) {
+      FLAGS_write_buffer_size = n;
+    }
+
+    if (sscanf(argv[i], "--total_keys=%d%c", &n, &junk) == 1) {
+      FLAGS_total_keys = n;
+    }
+
+    if (sscanf(argv[i], "--random_key=%d%c", &n, &junk) == 1 &&
+        (n == 0 || n == 1)) {
+      FLAGS_random_key = n;
+    }
+
+    if (sscanf(argv[i], "--use_set_based_memetable=%d%c", &n, &junk) == 1 &&
+        (n == 0 || n == 1)) {
+      FLAGS_use_set_based_memetable = n;
+    }
+
   }
 
-  leveldb::test::RunAllTests();
+  std::cout << kDbName << "\n";
+
+  rocksdb::test::RunAllTests();
   return 0;
 }

db/prefix_filter_iterator.h

@@ -12,9 +12,9 @@
 #ifndef STORAGE_LEVELDB_DB_PREFIX_FILTER_ITERATOR_H_
 #define STORAGE_LEVELDB_DB_PREFIX_FILTER_ITERATOR_H_
 
-#include "leveldb/iterator.h"
+#include "rocksdb/iterator.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class PrefixFilterIterator : public Iterator {
  private:
@@ -71,6 +71,6 @@ class PrefixFilterIterator : public Iterator {
   }
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 #endif

db/repair.cc

@@ -34,11 +34,11 @@
 #include "db/table_cache.h"
 #include "db/version_edit.h"
 #include "db/write_batch_internal.h"
-#include "leveldb/comparator.h"
-#include "leveldb/db.h"
-#include "leveldb/env.h"
+#include "rocksdb/comparator.h"
+#include "rocksdb/db.h"
+#include "rocksdb/env.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 namespace {
 
@@ -383,4 +383,4 @@ Status RepairDB(const std::string& dbname, const Options& options) {
   return repairer.Run();
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb

db/skiplist.h

@@ -33,7 +33,7 @@
 #include "port/port.h"
 #include "util/random.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 template<typename Key, class Comparator>
 class SkipList {
@@ -107,6 +107,7 @@ class SkipList {
 
   // Used for optimizing sequential insert patterns
   Node* prev_[kMaxHeight];
+  int   prev_height_;
 
   inline int GetMaxHeight() const {
     return static_cast<int>(
@@ -267,6 +268,10 @@ typename SkipList<Key,Comparator>::Node* SkipList<Key,Comparator>::FindGreaterOr
     Node* x = prev[0];
     Node* next = x->Next(0);
     if ((x == head_) || KeyIsAfterNode(key, x)) {
+      // Adjust all relevant insertion points to the previous entry
+      for (int i = 1; i < prev_height_; i++) {
+        prev[i] = x;
+      }
       return next;
     }
   }
@@ -275,6 +280,9 @@ typename SkipList<Key,Comparator>::Node* SkipList<Key,Comparator>::FindGreaterOr
   int level = GetMaxHeight() - 1;
   while (true) {
     Node* next = x->Next(level);
+    // Make sure the lists are sorted.
+    // If x points to head_ or next points nullptr, it is trivially satisfied.
+    assert((x == head_) || (next == nullptr) || KeyIsAfterNode(next->key, x));
     if (KeyIsAfterNode(key, next)) {
       // Keep searching in this list
       x = next;
@@ -337,6 +345,7 @@ SkipList<Key,Comparator>::SkipList(Comparator cmp, Arena* arena)
       arena_(arena),
       head_(NewNode(0 /* any key will do */, kMaxHeight)),
       max_height_(reinterpret_cast<void*>(1)),
+      prev_height_(1),
       rnd_(0xdeadbeef) {
   for (int i = 0; i < kMaxHeight; i++) {
     head_->SetNext(i, nullptr);
@@ -378,6 +387,7 @@ void SkipList<Key,Comparator>::Insert(const Key& key) {
     prev_[i]->SetNext(i, x);
   }
   prev_[0] = x;
+  prev_height_ = height;
 }
 
 template<typename Key, class Comparator>
@@ -390,6 +400,6 @@ bool SkipList<Key,Comparator>::Contains(const Key& key) const {
   }
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 #endif // STORAGE_LEVELDB_DB_SKIPLIST_H_

db/skiplist_test.cc

@@ -4,13 +4,13 @@
 
 #include "db/skiplist.h"
 #include <set>
-#include "leveldb/env.h"
+#include "rocksdb/env.h"
 #include "util/arena_impl.h"
 #include "util/hash.h"
 #include "util/random.h"
 #include "util/testharness.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 typedef uint64_t Key;
 
@@ -371,8 +371,8 @@ TEST(SkipTest, Concurrent3) { RunConcurrent(3); }
 TEST(SkipTest, Concurrent4) { RunConcurrent(4); }
 TEST(SkipTest, Concurrent5) { RunConcurrent(5); }
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 int main(int argc, char** argv) {
-  return leveldb::test::RunAllTests();
+  return rocksdb::test::RunAllTests();
 }

db/snapshot.h

@@ -5,9 +5,9 @@
 #ifndef STORAGE_LEVELDB_DB_SNAPSHOT_H_
 #define STORAGE_LEVELDB_DB_SNAPSHOT_H_
 
-#include "leveldb/db.h"
+#include "rocksdb/db.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class SnapshotList;
 
@@ -80,6 +80,6 @@ class SnapshotList {
   SnapshotImpl list_;
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 #endif  // STORAGE_LEVELDB_DB_SNAPSHOT_H_

db/table_cache.cc

@@ -6,12 +6,12 @@
 
 #include "db/filename.h"
 
-#include "leveldb/statistics.h"
+#include "rocksdb/statistics.h"
 #include "table/table.h"
 #include "util/coding.h"
 #include "util/stop_watch.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 static void DeleteEntry(const Slice& key, void* value) {
   Table* table = reinterpret_cast<Table*>(value);
@@ -48,7 +48,7 @@ Status TableCache::FindTable(const EnvOptions& toptions,
   *handle = cache_->Lookup(key);
   if (*handle == nullptr) {
     if (no_io) { // Dont do IO and return a not-found status
-      return Status::NotFound("Table not found in table_cache, no_io is set");
+      return Status::Incomplete("Table not found in table_cache, no_io is set");
     }
     if (table_io != nullptr) {
       *table_io = true;    // we had to do IO from storage
@@ -90,7 +90,8 @@ Iterator* TableCache::NewIterator(const ReadOptions& options,
   }
 
   Cache::Handle* handle = nullptr;
-  Status s = FindTable(toptions, file_number, file_size, &handle);
+  Status s = FindTable(toptions, file_number, file_size, &handle,
+                       nullptr, options.read_tier == kBlockCacheTier);
   if (!s.ok()) {
     return NewErrorIterator(s);
   }
@@ -117,17 +118,17 @@ Status TableCache::Get(const ReadOptions& options,
                        void* arg,
                        bool (*saver)(void*, const Slice&, const Slice&, bool),
                        bool* table_io,
-                       void (*mark_key_may_exist)(void*),
-                       const bool no_io) {
+                       void (*mark_key_may_exist)(void*)) {
   Cache::Handle* handle = nullptr;
   Status s = FindTable(storage_options_, file_number, file_size,
-                       &handle, table_io, no_io);
+                       &handle, table_io,
+                       options.read_tier == kBlockCacheTier);
   if (s.ok()) {
     Table* t =
       reinterpret_cast<Table*>(cache_->Value(handle));
-    s = t->InternalGet(options, k, arg, saver, mark_key_may_exist, no_io);
+    s = t->InternalGet(options, k, arg, saver, mark_key_may_exist);
     cache_->Release(handle);
-  } else if (no_io && s.IsNotFound()) {
+  } else if (options.read_tier && s.IsIncomplete()) {
     // Couldnt find Table in cache but treat as kFound if no_io set
     (*mark_key_may_exist)(arg);
     return Status::OK();
@@ -135,10 +136,28 @@ Status TableCache::Get(const ReadOptions& options,
   return s;
 }
 
+bool TableCache::PrefixMayMatch(const ReadOptions& options,
+                                uint64_t file_number,
+                                uint64_t file_size,
+                                const Slice& internal_prefix,
+                                bool* table_io) {
+  Cache::Handle* handle = nullptr;
+  Status s = FindTable(storage_options_, file_number,
+                       file_size, &handle, table_io);
+  bool may_match = true;
+  if (s.ok()) {
+    Table* t =
+      reinterpret_cast<Table*>(cache_->Value(handle));
+    may_match = t->PrefixMayMatch(internal_prefix);
+    cache_->Release(handle);
+  }
+  return may_match;
+}
+
 void TableCache::Evict(uint64_t file_number) {
   char buf[sizeof(file_number)];
   EncodeFixed64(buf, file_number);
   cache_->Erase(Slice(buf, sizeof(buf)));
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb

db/table_cache.h

@@ -10,12 +10,12 @@
 #include <string>
 #include <stdint.h>
 #include "db/dbformat.h"
-#include "leveldb/env.h"
-#include "leveldb/cache.h"
+#include "rocksdb/env.h"
+#include "rocksdb/cache.h"
 #include "port/port.h"
 #include "table/table.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class Env;
 
@@ -49,8 +49,13 @@ class TableCache {
              void* arg,
              bool (*handle_result)(void*, const Slice&, const Slice&, bool),
              bool* table_io,
-             void (*mark_key_may_exist)(void*) = nullptr,
-             const bool no_io = false);
+             void (*mark_key_may_exist)(void*) = nullptr);
+
+  // Determine whether the table may contain the specified prefix.  If
+  // the table index of blooms are not in memory, this may cause an I/O
+  bool PrefixMayMatch(const ReadOptions& options, uint64_t file_number,
+                      uint64_t file_size, const Slice& internal_prefix,
+                      bool* table_io);
 
   // Evict any entry for the specified file number
   void Evict(uint64_t file_number);
@@ -67,6 +72,6 @@ class TableCache {
                    const bool no_io = false);
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 #endif  // STORAGE_LEVELDB_DB_TABLE_CACHE_H_

db/transaction_log_impl.cc

@@ -1,7 +1,7 @@
 #include "db/transaction_log_impl.h"
 #include "db/write_batch_internal.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 TransactionLogIteratorImpl::TransactionLogIteratorImpl(
                            const std::string& dbname,
@@ -168,4 +168,4 @@ Status TransactionLogIteratorImpl::OpenLogReader(const LogFile* logFile) {
   );
   return Status::OK();
 }
-}  //  namespace leveldb
+}  //  namespace rocksdb

db/transaction_log_impl.h

@@ -4,14 +4,14 @@
 
 #include <vector>
 
-#include "leveldb/env.h"
-#include "leveldb/options.h"
-#include "leveldb/types.h"
-#include "leveldb/transaction_log.h"
+#include "rocksdb/env.h"
+#include "rocksdb/options.h"
+#include "rocksdb/types.h"
+#include "rocksdb/transaction_log.h"
 #include "db/log_reader.h"
 #include "db/filename.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 struct LogReporter : public log::Reader::Reporter {
   Env* env;
@@ -31,7 +31,12 @@ class LogFileImpl : public LogFile {
     sizeFileBytes_(sizeBytes) {
   }
 
-  std::string Filename() const { return LogFileName("", logNumber_); }
+  std::string PathName() const {
+    if (type_ == kArchivedLogFile) {
+      return ArchivedLogFileName("", logNumber_);
+    }
+    return LogFileName("", logNumber_);
+  }
 
   uint64_t LogNumber() const { return logNumber_; }
 
@@ -94,5 +99,5 @@ class TransactionLogIteratorImpl : public TransactionLogIterator {
 
 
 
-}  //  namespace leveldb
+}  //  namespace rocksdb
 #endif  //  STORAGE_LEVELDB_INCLUDE_WRITES_ITERATOR_IMPL_H_

db/version_edit.cc

@@ -7,7 +7,7 @@
 #include "db/version_set.h"
 #include "util/coding.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 // Tag numbers for serialized VersionEdit.  These numbers are written to
 // disk and should not be changed.
@@ -293,4 +293,4 @@ std::string VersionEdit::DebugString(bool hex_key) const {
   return r;
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb

db/version_edit.h

@@ -10,7 +10,7 @@
 #include <vector>
 #include "db/dbformat.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class VersionSet;
 
@@ -122,6 +122,6 @@ class VersionEdit {
   std::vector< std::pair<int, FileMetaData> > new_files_;
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 #endif  // STORAGE_LEVELDB_DB_VERSION_EDIT_H_

db/version_edit_test.cc

@@ -5,7 +5,7 @@
 #include "db/version_edit.h"
 #include "util/testharness.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 static void TestEncodeDecode(const VersionEdit& edit) {
   std::string encoded, encoded2;
@@ -41,8 +41,8 @@ TEST(VersionEditTest, EncodeDecode) {
   TestEncodeDecode(edit);
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 int main(int argc, char** argv) {
-  return leveldb::test::RunAllTests();
+  return rocksdb::test::RunAllTests();
 }

db/version_set.cc

@@ -12,16 +12,16 @@
 #include "db/log_writer.h"
 #include "db/memtable.h"
 #include "db/table_cache.h"
-#include "leveldb/env.h"
-#include "leveldb/merge_operator.h"
-#include "leveldb/table_builder.h"
+#include "rocksdb/env.h"
+#include "rocksdb/merge_operator.h"
+#include "rocksdb/table_builder.h"
 #include "table/merger.h"
 #include "table/two_level_iterator.h"
 #include "util/coding.h"
 #include "util/logging.h"
 #include "util/stop_watch.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 static uint64_t TotalFileSize(const std::vector<FileMetaData*>& files) {
   uint64_t sum = 0;
@@ -189,7 +189,14 @@ static Iterator* GetFileIterator(void* arg,
     return NewErrorIterator(
         Status::Corruption("FileReader invoked with unexpected value"));
   } else {
-    return cache->NewIterator(options,
+    ReadOptions options_copy;
+    if (options.prefix) {
+      // suppress prefix filtering since we have already checked the
+      // filters once at this point
+      options_copy = options;
+      options_copy.prefix = nullptr;
+    }
+    return cache->NewIterator(options.prefix ? options_copy : options,
                               soptions,
                               DecodeFixed64(file_value.data()),
                               DecodeFixed64(file_value.data() + 8),
@@ -198,22 +205,55 @@ static Iterator* GetFileIterator(void* arg,
   }
 }
 
+bool Version::PrefixMayMatch(const ReadOptions& options,
+                             const EnvOptions& soptions,
+                             const Slice& internal_prefix,
+                             Iterator* level_iter) const {
+  bool may_match = true;
+  level_iter->Seek(internal_prefix);
+  if (!level_iter->Valid()) {
+    // we're past end of level
+    may_match = false;
+  } else if (ExtractUserKey(level_iter->key()).starts_with(
+                                             ExtractUserKey(internal_prefix))) {
+    // TODO(tylerharter): do we need this case?  Or are we guaranteed
+    // key() will always be the biggest value for this SST?
+    may_match = true;
+  } else {
+    may_match = vset_->table_cache_->PrefixMayMatch(
+                           options,
+                           DecodeFixed64(level_iter->value().data()),
+                           DecodeFixed64(level_iter->value().data() + 8),
+                           internal_prefix, nullptr);
+  }
+  return may_match;
+}
+
 Iterator* Version::NewConcatenatingIterator(const ReadOptions& options,
                                             const EnvOptions& soptions,
                                             int level) const {
-  return NewTwoLevelIterator(
-      new LevelFileNumIterator(vset_->icmp_, &files_[level]),
-      &GetFileIterator, vset_->table_cache_, options, soptions);
+  Iterator* level_iter = new LevelFileNumIterator(vset_->icmp_, &files_[level]);
+  if (options.prefix) {
+    InternalKey internal_prefix(*options.prefix, 0, kTypeValue);
+    if (!PrefixMayMatch(options, soptions,
+                        internal_prefix.Encode(), level_iter)) {
+      delete level_iter;
+      // nothing in this level can match the prefix
+      return NewEmptyIterator();
+    }
+  }
+  return NewTwoLevelIterator(level_iter, &GetFileIterator,
+                             vset_->table_cache_, options, soptions);
 }
 
 void Version::AddIterators(const ReadOptions& options,
                            const EnvOptions& soptions,
                            std::vector<Iterator*>* iters) {
   // Merge all level zero files together since they may overlap
-  for (size_t i = 0; i < files_[0].size(); i++) {
+  for (const FileMetaData* file : files_[0]) {
     iters->push_back(
         vset_->table_cache_->NewIterator(
-            options, soptions, files_[0][i]->number, files_[0][i]->file_size));
+            options, soptions, file->number, file->file_size));
   }
 
   // For levels > 0, we can use a concatenating iterator that sequentially
@@ -357,6 +397,7 @@ static bool NewestFirstBySeqNo(FileMetaData* a, FileMetaData* b) {
 
 Version::Version(VersionSet* vset, uint64_t version_number)
     : vset_(vset), next_(this), prev_(this), refs_(0),
+      files_(new std::vector<FileMetaData*>[vset->NumberLevels()]),
       files_by_size_(vset->NumberLevels()),
       next_file_to_compact_by_size_(vset->NumberLevels()),
       file_to_compact_(nullptr),
@@ -365,7 +406,6 @@ Version::Version(VersionSet* vset, uint64_t version_number)
       compaction_level_(vset->NumberLevels()),
       offset_manifest_file_(0),
       version_number_(version_number) {
-  files_ = new std::vector<FileMetaData*>[vset->NumberLevels()];
 }
 
 void Version::Get(const ReadOptions& options,
@@ -375,7 +415,6 @@ void Version::Get(const ReadOptions& options,
                   std::deque<std::string>* operands,
                   GetStats* stats,
                   const Options& db_options,
-                  const bool no_io,
                   bool* value_found) {
   Slice ikey = k.internal_key();
   Slice user_key = k.user_key();
@@ -385,9 +424,6 @@ void Version::Get(const ReadOptions& options,
   auto logger = db_options.info_log;
 
   assert(status->ok() || status->IsMergeInProgress());
-  if (no_io) {
-    assert(status->ok());
-  }
   Saver saver;
   saver.state = status->ok()? kNotFound : kMerge;
   saver.ucmp = ucmp;
@@ -476,7 +512,7 @@ void Version::Get(const ReadOptions& options,
       bool tableIO = false;
       *status = vset_->table_cache_->Get(options, f->number, f->file_size,
                                          ikey, &saver, SaveValue, &tableIO,
-                                         MarkKeyMayExist, no_io);
+                                         MarkKeyMayExist);
       // TODO: examine the behavior for corrupted key
       if (!status->ok()) {
         return;
@@ -2123,17 +2159,215 @@ void VersionSet::SizeBeingCompacted(std::vector<uint64_t>& sizes) {
   }
 }
 
-Compaction* VersionSet::PickCompactionUniversal(int level, double score) {
+//
+// Look at overall size amplification. If size amplification
+// exceeeds the configured value, then do a compaction
+// of the candidate files all the way upto the earliest
+// base file (overrides configured values of file-size ratios,
+// min_merge_width and max_merge_width).
+//
+Compaction* VersionSet::PickCompactionUniversalSizeAmp(
+    int level, double score) {
   assert (level == 0);
 
-  // percentage flexibilty while comparing file sizes
-  uint64_t ratio = options_->compaction_options_universal.size_ratio;
+  // percentage flexibilty while reducing size amplification
+  uint64_t ratio = options_->compaction_options_universal.
+                     max_size_amplification_percent;
+
+  // The files are sorted from newest first to oldest last.
+  std::vector<int>& file_by_time = current_->files_by_size_[level];
+  assert(file_by_time.size() == current_->files_[level].size());
+
+  unsigned int candidate_count = 0;
+  uint64_t candidate_size = 0;
+  unsigned int start_index = 0;
+  FileMetaData* f = nullptr;
+
+  // Skip files that are already being compacted
+  for (unsigned int loop = 0; loop < file_by_time.size() - 1; loop++) {
+    int index = file_by_time[loop];
+    f = current_->files_[level][index];
+    if (!f->being_compacted) {
+      start_index = loop;         // Consider this as the first candidate.
+      break;
+    }
+    Log(options_->info_log, "Universal: skipping file %ld[%d] compacted %s",
+        f->number, loop, " cannot be a candidate to reduce size amp.\n");
+    f = nullptr;
+  }
+  if (f == nullptr) {
+    return nullptr;             // no candidate files
+  }
+
+  Log(options_->info_log, "Universal: First candidate file %ld[%d] %s",
+      f->number, start_index, " to reduce size amp.\n");
+
+  // keep adding up all the remaining files
+  for (unsigned int loop = start_index; loop < file_by_time.size() - 1;
+       loop++) {
+    int index = file_by_time[loop];
+    f = current_->files_[level][index];
+    if (f->being_compacted) {
+      Log(options_->info_log,
+          "Universal: Possible candidate file %ld[%d] %s.", f->number, loop,
+          " is already being compacted. No size amp reduction possible.\n");
+      return nullptr;
+    }
+    candidate_size += f->file_size;
+    candidate_count++;
+  }
+  if (candidate_count == 0) {
+    return nullptr;
+  }
+
+  // size of earliest file
+  int index = file_by_time[file_by_time.size() - 1];
+  uint64_t earliest_file_size = current_->files_[level][index]->file_size;
+
+  // size amplification = percentage of additional size
+  if (candidate_size * 100 < ratio * earliest_file_size) {
+    Log(options_->info_log,
+        "Universal: size amp not needed. newer-files-total-size %ld "
+        "earliest-file-size %ld",
+        candidate_size, earliest_file_size);
+    return nullptr;
+  } else {
+    Log(options_->info_log,
+        "Universal: size amp needed. newer-files-total-size %ld "
+        "earliest-file-size %ld",
+        candidate_size, earliest_file_size);
+  }
+  assert(start_index >= 0 && start_index < file_by_time.size() - 1);
+
+  // create a compaction request
+  Compaction* c = new Compaction(level, level, MaxFileSizeForLevel(level),
+                                 LLONG_MAX, NumberLevels());
+  c->score_ = score;
+  for (unsigned int loop = start_index; loop < file_by_time.size(); loop++) {
+    int index = file_by_time[loop];
+    f = current_->files_[level][index];
+    c->inputs_[0].push_back(f);
+    Log(options_->info_log,
+        "Universal: size amp picking file %ld[%d] with size %ld",
+        f->number, index, f->file_size);
+  }
+  return c;
+}
+
+//
+// Consider compaction files based on their size differences with
+// the next file in time order.
+//
+Compaction* VersionSet::PickCompactionUniversalReadAmp(
+    int level, double score, unsigned int ratio,
+    unsigned int max_number_of_files_to_compact) {
+
   unsigned int min_merge_width =
     options_->compaction_options_universal.min_merge_width;
   unsigned int max_merge_width =
     options_->compaction_options_universal.max_merge_width;
 
-  if ((current_->files_[level].size() <=
+  // The files are sorted from newest first to oldest last.
+  std::vector<int>& file_by_time = current_->files_by_size_[level];
+  FileMetaData* f = nullptr;
+  bool done = false;
+  int start_index = 0;
+  unsigned int candidate_count;
+  assert(file_by_time.size() == current_->files_[level].size());
+
+  unsigned int max_files_to_compact = std::min(max_merge_width,
+                                       max_number_of_files_to_compact);
+  min_merge_width = std::max(min_merge_width, 2U);
+
+  // Considers a candidate file only if it is smaller than the
+  // total size accumulated so far.
+  for (unsigned int loop = 0; loop < file_by_time.size(); loop++) {
+
+    candidate_count = 0;
+
+    // Skip files that are already being compacted
+    for (f = nullptr; loop < file_by_time.size(); loop++) {
+      int index = file_by_time[loop];
+      f = current_->files_[level][index];
+
+      if (!f->being_compacted) {
+        candidate_count = 1;
+        break;
+      }
+      Log(options_->info_log,
+          "Universal: file %ld[%d] being compacted, skipping",
+          f->number, loop);
+      f = nullptr;
+    }
+
+    // This file is not being compacted. Consider it as the
+    // first candidate to be compacted.
+    uint64_t candidate_size =  f != nullptr? f->file_size : 0;
+    if (f != nullptr) {
+      Log(options_->info_log, "Universal: Possible candidate file %ld[%d].",
+          f->number, loop);
+    }
+
+    // Check if the suceeding files need compaction.
+    for (unsigned int i = loop+1;
+         candidate_count < max_files_to_compact && i < file_by_time.size();
+         i++) {
+      int index = file_by_time[i];
+      FileMetaData* f = current_->files_[level][index];
+      if (f->being_compacted) {
+        break;
+      }
+      // pick files if the total candidate file size (increased by the
+      // specified ratio) is still larger than the next candidate file.
+      uint64_t sz = (candidate_size * (100L + ratio)) /100;
+      if (sz < f->file_size) {
+        break;
+      }
+      candidate_count++;
+      candidate_size += f->file_size;
+    }
+
+    // Found a series of consecutive files that need compaction.
+    if (candidate_count >= (unsigned int)min_merge_width) {
+      start_index = loop;
+      done = true;
+      break;
+    } else {
+      for (unsigned int i = loop;
+           i < loop + candidate_count && i < file_by_time.size(); i++) {
+       int index = file_by_time[i];
+       FileMetaData* f = current_->files_[level][index];
+       Log(options_->info_log,
+           "Universal: Skipping file %ld[%d] with size %ld %d\n",
+           f->number, i, f->file_size, f->being_compacted);
+      }
+    }
+  }
+  if (!done || candidate_count <= 1) {
+    return nullptr;
+  }
+  Compaction* c = new Compaction(level, level, MaxFileSizeForLevel(level),
+                                 LLONG_MAX, NumberLevels());
+  c->score_ = score;
+
+  for (unsigned int i = start_index; i < start_index + candidate_count; i++) {
+    int index = file_by_time[i];
+    FileMetaData* f = current_->files_[level][index];
+    c->inputs_[0].push_back(f);
+    Log(options_->info_log, "Universal: Picking file %ld[%d] with size %ld\n",
+        f->number, i, f->file_size);
+  }
+  return c;
+}
+
+//
+// Universal style of compaction. Pick files that are contiguous in
+// time-range to compact.
+//
+Compaction* VersionSet::PickCompactionUniversal(int level, double score) {
+  assert (level == 0);
+
+  if ((current_->files_[level].size() <
       (unsigned int)options_->level0_file_num_compaction_trigger)) {
     Log(options_->info_log, "Universal: nothing to do\n");
     return nullptr;
@@ -2143,127 +2377,29 @@ Compaction* VersionSet::PickCompactionUniversal(int level, double score) {
       current_->files_[level].size(),
       LevelFileSummary(&tmp, 0));
 
-  Compaction* c = nullptr;
-  c = new Compaction(level, level, MaxFileSizeForLevel(level),
-                     LLONG_MAX, NumberLevels());
-  c->score_ = score;
+  // Check for size amplification first.
+  Compaction* c = PickCompactionUniversalSizeAmp(level, score);
+  if (c == nullptr) {
 
-  // The files are sorted from newest first to oldest last.
-  std::vector<int>& file_by_time = current_->files_by_size_[level];
-  FileMetaData* f = nullptr;
-  bool done = false;
-  assert(file_by_time.size() == current_->files_[level].size());
+    // Size amplification is within limits. Try reducing read
+    // amplification while maintaining file size ratios.
+    unsigned int ratio = options_->compaction_options_universal.size_ratio;
+    c = PickCompactionUniversalReadAmp(level, score, ratio, UINT_MAX);
 
-  unsigned int max_files_to_compact = std::min(max_merge_width, UINT_MAX);
-
-  // Make two pass. The first pass considers a candidate file
-  // only if it is smaller than the total size accumulated so far.
-  // The second pass does not look at the slope of the
-  // file-size  curve to decide what to pick for compaction.
-  for (int iter = 0; !done && iter < 2; iter++) {
-
-    for (unsigned int loop = 0; loop < file_by_time.size(); ) {
-
-      // Skip files that are already being compacted
-      for (f = nullptr; loop < file_by_time.size(); loop++) {
-        int index = file_by_time[loop];
-        f = current_->files_[level][index];
-
-        if (!f->being_compacted) {
-          break;
-        }
-        Log(options_->info_log, "Universal: file %ld[%d] being compacted, skipping",
-            f->number, loop);
-        f = nullptr;
-      }
-
-      // This file is not being compacted. Consider it as the
-      // first candidate to be compacted.
-      unsigned int candidate_count = 1;
-      uint64_t candidate_size =  f != nullptr? f->file_size : 0;
-      if (f != nullptr) {
-        Log(options_->info_log, "Universal: Possible candidate file %ld[%d] %s.",
-            f->number, loop, iter == 0? "" : "forced ");
-      }
-
-      // Check if the suceeding files need compaction.
-      for (unsigned int i = loop+1;
-           candidate_count < max_files_to_compact && i < file_by_time.size();
-           i++) {
-        int index = file_by_time[i];
-        FileMetaData* f = current_->files_[level][index];
-        if (f->being_compacted) {
-          break;
-        }
-        // If this is the first iteration, then we pick files if the
-        // total candidate file size (increased by the specified ratio)
-        // is still larger than the next candidate file.
-        if (iter == 0) {
-          uint64_t sz = (candidate_size * (100 + ratio)) /100;
-          if (sz < f->file_size) {
-            break;
-          }
-        }
-        candidate_count++;
-        candidate_size += f->file_size;
-      }
-
-      // Found a series of consecutive files that need compaction.
-      if (candidate_count >= (unsigned int)min_merge_width) {
-        for (unsigned int i = loop; i < loop + candidate_count; i++) {
-          int index = file_by_time[i];
-          FileMetaData* f = current_->files_[level][index];
-          c->inputs_[0].push_back(f);
-          Log(options_->info_log, "Universal: Picking file %ld[%d] with size %ld %s",
-              f->number, i, f->file_size,
-              (iter == 0 ? "" : "forced"));
-        }
-        done = true;
-        break;
-      } else {
-        for (unsigned int i = loop;
-             i < loop + candidate_count && i < file_by_time.size(); i++) {
-         int index = file_by_time[i];
-         FileMetaData* f = current_->files_[level][index];
-         Log(options_->info_log, "Universal: Skipping file %ld[%d] with size %ld %d %s",
-             f->number, i, f->file_size, f->being_compacted,
-              (iter == 0 ? "" : "forced"));
-        }
-      }
-      loop += candidate_count;
-    }
-    assert(done || c->inputs_[0].size() == 0);
-
-    // If we are unable to find a normal compaction run and we are still
-    // above the compaction threshold, iterate again to pick compaction
-    // candidates, this time without considering their size differences.
-    if (!done) {
-      int files_not_in_compaction = 0;
-      for (unsigned int i = 0; i < current_->files_[level].size(); i++) {
-        f = current_->files_[level][i];
-        if (!f->being_compacted) {
-          files_not_in_compaction++;
-        }
-      }
-      int expected_num_files = files_not_in_compaction +
-                compactions_in_progress_[level].size();
-      if (expected_num_files <=
-          options_->level0_file_num_compaction_trigger + 1) {
-        done = true;     // nothing more to do
-      } else {
-        max_files_to_compact = std::min((int)max_merge_width,
-          expected_num_files - options_->level0_file_num_compaction_trigger);
-        Log(options_->info_log, "Universal: second loop with maxfiles %d",
-            max_files_to_compact);
-      }
+    // Size amplification and file size ratios are within configured limits.
+    // If max read amplification is exceeding configured limits, then force
+    // compaction without looking at filesize ratios and try to reduce
+    // the number of files to fewer than level0_file_num_compaction_trigger.
+    if (c == nullptr) {
+      unsigned int num_files = current_->files_[level].size() -
+                               options_->level0_file_num_compaction_trigger;
+      c = PickCompactionUniversalReadAmp(level, score, UINT_MAX, num_files);
     }
   }
-  if (c->inputs_[0].size() <= 1) {
-    Log(options_->info_log, "Universal: only %ld files, nothing to do.\n",
-        c->inputs_[0].size());
-    delete c;
+  if (c == nullptr) {
     return nullptr;
   }
+  assert(c->inputs_[0].size() > 1);
 
   // validate that all the chosen files are non overlapping in time
   FileMetaData* newerfile __attribute__((unused)) = nullptr;
@@ -2275,6 +2411,9 @@ Compaction* VersionSet::PickCompactionUniversal(int level, double score) {
     newerfile = f;
   }
 
+  // The files are sorted from newest first to oldest last.
+  std::vector<int>& file_by_time = current_->files_by_size_[level];
+
   // Is the earliest file part of this compaction?
   int last_index = file_by_time[file_by_time.size()-1];
   FileMetaData* last_file = current_->files_[level][last_index];
@@ -2637,6 +2776,41 @@ void VersionSet::SetupOtherInputs(Compaction* c) {
   c->edit_->SetCompactPointer(level, largest);
 }
 
+Status VersionSet::GetMetadataForFile(
+    uint64_t number,
+    int *filelevel,
+    FileMetaData *meta) {
+  for (int level = 0; level < NumberLevels(); level++) {
+    const std::vector<FileMetaData*>& files = current_->files_[level];
+    for (size_t i = 0; i < files.size(); i++) {
+      if (files[i]->number == number) {
+        *meta = *files[i];
+        *filelevel = level;
+        return Status::OK();
+      }
+    }
+  }
+  return Status::NotFound("File not present in any level");
+}
+
+void VersionSet::GetLiveFilesMetaData(
+    std::vector<LiveFileMetaData> * metadata) {
+  for (int level = 0; level < NumberLevels(); level++) {
+    const std::vector<FileMetaData*>& files = current_->files_[level];
+    for (size_t i = 0; i < files.size(); i++) {
+      LiveFileMetaData filemetadata;
+      filemetadata.name = TableFileName("", files[i]->number);
+      filemetadata.level = level;
+      filemetadata.size = files[i]->file_size;
+      filemetadata.smallestkey = files[i]->smallest.user_key().ToString();
+      filemetadata.largestkey = files[i]->largest.user_key().ToString();
+      filemetadata.smallest_seqno = files[i]->smallest_seqno;
+      filemetadata.largest_seqno = files[i]->largest_seqno;
+      metadata->push_back(filemetadata);
+    }
+  }
+}
+
 Compaction* VersionSet::CompactRange(
     int level,
     const InternalKey* begin,
@@ -2881,4 +3055,4 @@ void Compaction::Summary(char* output, int len) {
       level_low_summary, level_up_summary);
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb

db/version_set.h

@@ -25,7 +25,7 @@
 #include "port/port.h"
 #include "db/table_cache.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 namespace log { class Writer; }
 
@@ -76,7 +76,7 @@ class Version {
   };
   void Get(const ReadOptions&, const LookupKey& key, std::string* val,
            Status* status, std::deque<std::string>* operands, GetStats* stats,
-           const Options& db_option, const bool no_io = false,
+           const Options& db_option,
            bool* value_found = nullptr);
 
   // Adds "stats" into the current state.  Returns true if a new
@@ -152,6 +152,8 @@ class Version {
   Iterator* NewConcatenatingIterator(const ReadOptions&,
                                      const EnvOptions& soptions,
                                      int level) const;
+  bool PrefixMayMatch(const ReadOptions& options, const EnvOptions& soptions,
+                      const Slice& internal_prefix, Iterator* level_iter) const;
 
   VersionSet* vset_;            // VersionSet to which this Version belongs
   Version* next_;               // Next version in linked list
@@ -381,6 +383,13 @@ class VersionSet {
   // Pick files to compact in Universal mode
   Compaction* PickCompactionUniversal(int level, double score);
 
+  // Pick Universal compaction to limit read amplification
+  Compaction* PickCompactionUniversalReadAmp(int level, double score,
+                unsigned int ratio, unsigned int num_files);
+
+  // Pick Universal compaction to limit space amplification.
+  Compaction* PickCompactionUniversalSizeAmp(int level, double score);
+
   // Free up the files that were participated in a compaction
   void ReleaseCompactionFiles(Compaction* c, Status status);
 
@@ -405,6 +414,12 @@ class VersionSet {
 
   double MaxBytesForLevel(int level);
 
+  Status GetMetadataForFile(
+    uint64_t number, int *filelevel, FileMetaData *metadata);
+
+  void GetLiveFilesMetaData(
+    std::vector<LiveFileMetaData> *metadata);
+
  private:
   class Builder;
   struct ManifestWriter;
@@ -612,6 +627,6 @@ class Compaction {
   void ResetNextCompactionIndex();
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 #endif  // STORAGE_LEVELDB_DB_VERSION_SET_H_

db/version_set_reduce_num_levels.cc

@@ -10,7 +10,7 @@
 #include "db/log_writer.h"
 #include "util/logging.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 Status VersionSet::ReduceNumberOfLevels(int new_levels, port::Mutex* mu) {
 

db/version_set_test.cc

@@ -7,7 +7,7 @@
 #include "util/testharness.h"
 #include "util/testutil.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class FindFileTest {
  public:
@@ -172,8 +172,8 @@ TEST(FindFileTest, OverlappingFiles) {
   ASSERT_TRUE(Overlaps("600", "700"));
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 int main(int argc, char** argv) {
-  return leveldb::test::RunAllTests();
+  return rocksdb::test::RunAllTests();
 }

db/write_batch.cc

@@ -14,10 +14,10 @@
 //    len: varint32
 //    data: uint8[len]
 
-#include "leveldb/write_batch.h"
+#include "rocksdb/write_batch.h"
 
-#include "leveldb/options.h"
-#include "leveldb/statistics.h"
+#include "rocksdb/options.h"
+#include "rocksdb/statistics.h"
 #include "db/dbformat.h"
 #include "db/db_impl.h"
 #include "db/memtable.h"
@@ -26,7 +26,7 @@
 #include "util/coding.h"
 #include <stdexcept>
 
-namespace leveldb {
+namespace rocksdb {
 
 // WriteBatch header has an 8-byte sequence number followed by a 4-byte count.
 static const size_t kHeader = 12;
@@ -48,6 +48,10 @@ void WriteBatch::Handler::LogData(const Slice& blob) {
   // them.
 }
 
+bool WriteBatch::Handler::Continue() {
+  return true;
+}
+
 void WriteBatch::Clear() {
   rep_.clear();
   rep_.resize(kHeader);
@@ -66,7 +70,7 @@ Status WriteBatch::Iterate(Handler* handler) const {
   input.remove_prefix(kHeader);
   Slice key, value, blob;
   int found = 0;
-  while (!input.empty()) {
+  while (!input.empty() && handler->Continue()) {
     char tag = input[0];
     input.remove_prefix(1);
     switch (tag) {
@@ -223,4 +227,4 @@ void WriteBatchInternal::Append(WriteBatch* dst, const WriteBatch* src) {
   dst->rep_.append(src->rep_.data() + kHeader, src->rep_.size() - kHeader);
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb

db/write_batch_internal.h

@@ -5,12 +5,12 @@
 #ifndef STORAGE_LEVELDB_DB_WRITE_BATCH_INTERNAL_H_
 #define STORAGE_LEVELDB_DB_WRITE_BATCH_INTERNAL_H_
 
-#include "leveldb/types.h"
-#include "leveldb/write_batch.h"
-#include "leveldb/db.h"
-#include "leveldb/options.h"
+#include "rocksdb/types.h"
+#include "rocksdb/write_batch.h"
+#include "rocksdb/db.h"
+#include "rocksdb/options.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class MemTable;
 
@@ -51,7 +51,7 @@ class WriteBatchInternal {
   static void Append(WriteBatch* dst, const WriteBatch* src);
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 
 #endif  // STORAGE_LEVELDB_DB_WRITE_BATCH_INTERNAL_H_

db/write_batch_test.cc

@@ -2,17 +2,17 @@
 // 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 "leveldb/db.h"
+#include "rocksdb/db.h"
 
 #include <memory>
-#include "db/skiplistrep.h"
 #include "db/memtable.h"
 #include "db/write_batch_internal.h"
-#include "leveldb/env.h"
+#include "rocksdb/env.h"
+#include "rocksdb/memtablerep.h"
 #include "util/logging.h"
 #include "util/testharness.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 static std::string PrintContents(WriteBatch* b) {
   InternalKeyComparator cmp(BytewiseComparator());
@@ -134,6 +134,24 @@ TEST(WriteBatchTest, Append) {
   ASSERT_EQ(4, b1.Count());
 }
 
+namespace {
+  struct TestHandler : public WriteBatch::Handler {
+    std::string seen;
+    virtual void Put(const Slice& key, const Slice& value) {
+      seen += "Put(" + key.ToString() + ", " + value.ToString() + ")";
+    }
+    virtual void Merge(const Slice& key, const Slice& value) {
+      seen += "Merge(" + key.ToString() + ", " + value.ToString() + ")";
+    }
+    virtual void LogData(const Slice& blob) {
+      seen += "LogData(" + blob.ToString() + ")";
+    }
+    virtual void Delete(const Slice& key) {
+      seen += "Delete(" + key.ToString() + ")";
+    }
+  };
+}
+
 TEST(WriteBatchTest, Blob) {
   WriteBatch batch;
   batch.Put(Slice("k1"), Slice("v1"));
@@ -151,21 +169,7 @@ TEST(WriteBatchTest, Blob) {
             "Put(k3, v3)@2",
             PrintContents(&batch));
 
-  struct Handler : public WriteBatch::Handler {
-    std::string seen;
-    virtual void Put(const Slice& key, const Slice& value) {
-      seen += "Put(" + key.ToString() + ", " + value.ToString() + ")";
-    }
-    virtual void Merge(const Slice& key, const Slice& value) {
-      seen += "Merge(" + key.ToString() + ", " + value.ToString() + ")";
-    }
-    virtual void LogData(const Slice& blob) {
-      seen += "LogData(" + blob.ToString() + ")";
-    }
-    virtual void Delete(const Slice& key) {
-      seen += "Delete(" + key.ToString() + ")";
-    }
-  } handler;
+  TestHandler handler;
   batch.Iterate(&handler);
   ASSERT_EQ(
             "Put(k1, v1)"
@@ -178,8 +182,47 @@ TEST(WriteBatchTest, Blob) {
             handler.seen);
 }
 
-}  // namespace leveldb
+TEST(WriteBatchTest, Continue) {
+  WriteBatch batch;
+
+  struct Handler : public TestHandler {
+    int num_seen = 0;
+    virtual void Put(const Slice& key, const Slice& value) {
+      ++num_seen;
+      TestHandler::Put(key, value);
+    }
+    virtual void Merge(const Slice& key, const Slice& value) {
+      ++num_seen;
+      TestHandler::Merge(key, value);
+    }
+    virtual void LogData(const Slice& blob) {
+      ++num_seen;
+      TestHandler::LogData(blob);
+    }
+    virtual void Delete(const Slice& key) {
+      ++num_seen;
+      TestHandler::Delete(key);
+    }
+    virtual bool Continue() override {
+      return num_seen < 3;
+    }
+  } handler;
+
+  batch.Put(Slice("k1"), Slice("v1"));
+  batch.PutLogData(Slice("blob1"));
+  batch.Delete(Slice("k1"));
+  batch.PutLogData(Slice("blob2"));
+  batch.Merge(Slice("foo"), Slice("bar"));
+  batch.Iterate(&handler);
+  ASSERT_EQ(
+            "Put(k1, v1)"
+            "LogData(blob1)"
+            "Delete(k1)",
+            handler.seen);
+}
+
+}  // namespace rocksdb
 
 int main(int argc, char** argv) {
-  return leveldb::test::RunAllTests();
+  return rocksdb::test::RunAllTests();
 }

doc/bench/db_bench_sqlite3.cc

@@ -111,7 +111,7 @@ static void WalCheckpoint(sqlite3* db_) {
   }
 }
 
-namespace leveldb {
+namespace rocksdb {
 
 // Helper for quickly generating random data.
 namespace {
@@ -664,7 +664,7 @@ class Benchmark {
 
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 int main(int argc, char** argv) {
   std::string default_db_path;
@@ -672,7 +672,7 @@ int main(int argc, char** argv) {
     double d;
     int n;
     char junk;
-    if (leveldb::Slice(argv[i]).starts_with("--benchmarks=")) {
+    if (rocksdb::Slice(argv[i]).starts_with("--benchmarks=")) {
       FLAGS_benchmarks = argv[i] + strlen("--benchmarks=");
     } else if (sscanf(argv[i], "--histogram=%d%c", &n, &junk) == 1 &&
                (n == 0 || n == 1)) {
@@ -688,7 +688,7 @@ int main(int argc, char** argv) {
       FLAGS_reads = n;
     } else if (sscanf(argv[i], "--value_size=%d%c", &n, &junk) == 1) {
       FLAGS_value_size = n;
-    } else if (leveldb::Slice(argv[i]) == leveldb::Slice("--no_transaction")) {
+    } else if (rocksdb::Slice(argv[i]) == rocksdb::Slice("--no_transaction")) {
       FLAGS_transaction = false;
     } else if (sscanf(argv[i], "--page_size=%d%c", &n, &junk) == 1) {
       FLAGS_page_size = n;
@@ -707,12 +707,12 @@ int main(int argc, char** argv) {
 
   // Choose a location for the test database if none given with --db=<path>
   if (FLAGS_db == NULL) {
-      leveldb::Env::Default()->GetTestDirectory(&default_db_path);
+      rocksdb::Env::Default()->GetTestDirectory(&default_db_path);
       default_db_path += "/dbbench";
       FLAGS_db = default_db_path.c_str();
   }
 
-  leveldb::Benchmark benchmark;
+  rocksdb::Benchmark benchmark;
   benchmark.Run();
   return 0;
 }

doc/bench/db_bench_tree_db.cc

@@ -80,7 +80,7 @@ static void DBSynchronize(kyotocabinet::TreeDB* db_)
   }
 }
 
-namespace leveldb {
+namespace rocksdb {
 
 // Helper for quickly generating random data.
 namespace {
@@ -479,7 +479,7 @@ class Benchmark {
   }
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 int main(int argc, char** argv) {
   std::string default_db_path;
@@ -487,7 +487,7 @@ int main(int argc, char** argv) {
     double d;
     int n;
     char junk;
-    if (leveldb::Slice(argv[i]).starts_with("--benchmarks=")) {
+    if (rocksdb::Slice(argv[i]).starts_with("--benchmarks=")) {
       FLAGS_benchmarks = argv[i] + strlen("--benchmarks=");
     } else if (sscanf(argv[i], "--compression_ratio=%lf%c", &d, &junk) == 1) {
       FLAGS_compression_ratio = d;
@@ -517,12 +517,12 @@ int main(int argc, char** argv) {
 
   // Choose a location for the test database if none given with --db=<path>
   if (FLAGS_db == NULL) {
-      leveldb::Env::Default()->GetTestDirectory(&default_db_path);
+      rocksdb::Env::Default()->GetTestDirectory(&default_db_path);
       default_db_path += "/dbbench";
       FLAGS_db = default_db_path.c_str();
   }
 
-  leveldb::Benchmark benchmark;
+  rocksdb::Benchmark benchmark;
   benchmark.Run();
   return 0;
 }

doc/index.html

@@ -25,27 +25,27 @@ creating it if necessary:
   #include <assert>
   #include "leveldb/db.h"
 
-  leveldb::DB* db;
-  leveldb::Options options;
+  rocksdb::DB* db;
+  rocksdb::Options options;
   options.create_if_missing = true;
-  leveldb::Status status = leveldb::DB::Open(options, "/tmp/testdb", &db);
+  rocksdb::Status status = rocksdb::DB::Open(options, "/tmp/testdb", &db);
   assert(status.ok());
   ...
 </pre>
 If you want to raise an error if the database already exists, add
-the following line before the <code>leveldb::DB::Open</code> call:
+the following line before the <code>rocksdb::DB::Open</code> call:
 <pre>
   options.error_if_exists = true;
 </pre>
 <h1>Status</h1>
 <p>
-You may have noticed the <code>leveldb::Status</code> type above.  Values of this
+You may have noticed the <code>rocksdb::Status</code> type above.  Values of this
 type are returned by most functions in <code>leveldb</code> that may encounter an
 error.  You can check if such a result is ok, and also print an
 associated error message:
 <p>
 <pre>
-   leveldb::Status s = ...;
+   rocksdb::Status s = ...;
    if (!s.ok()) cerr &lt;&lt; s.ToString() &lt;&lt; endl;
 </pre>
 <h1>Closing A Database</h1>
@@ -65,9 +65,9 @@ modify/query the database.  For example, the following code
 moves the value stored under key1 to key2.
 <pre>
   std::string value;
-  leveldb::Status s = db-&gt;Get(leveldb::ReadOptions(), key1, &amp;value);
-  if (s.ok()) s = db-&gt;Put(leveldb::WriteOptions(), key2, value);
-  if (s.ok()) s = db-&gt;Delete(leveldb::WriteOptions(), key1);
+  rocksdb::Status s = db-&gt;Get(rocksdb::ReadOptions(), key1, &amp;value);
+  if (s.ok()) s = db-&gt;Put(rocksdb::WriteOptions(), key2, value);
+  if (s.ok()) s = db-&gt;Delete(rocksdb::WriteOptions(), key1);
 </pre>
 
 <h1>Atomic Updates</h1>
@@ -81,12 +81,12 @@ atomically apply a set of updates:
   #include "leveldb/write_batch.h"
   ...
   std::string value;
-  leveldb::Status s = db-&gt;Get(leveldb::ReadOptions(), key1, &amp;value);
+  rocksdb::Status s = db-&gt;Get(rocksdb::ReadOptions(), key1, &amp;value);
   if (s.ok()) {
-    leveldb::WriteBatch batch;
+    rocksdb::WriteBatch batch;
     batch.Delete(key1);
     batch.Put(key2, value);
-    s = db-&gt;Write(leveldb::WriteOptions(), &amp;batch);
+    s = db-&gt;Write(rocksdb::WriteOptions(), &amp;batch);
   }
 </pre>
 The <code>WriteBatch</code> holds a sequence of edits to be made to the database,
@@ -109,7 +109,7 @@ persistent storage.  (On Posix systems, this is implemented by calling
 either <code>fsync(...)</code> or <code>fdatasync(...)</code> or
 <code>msync(..., MS_SYNC)</code> before the write operation returns.)
 <pre>
-  leveldb::WriteOptions write_options;
+  rocksdb::WriteOptions write_options;
   write_options.sync = true;
   db-&gt;Put(write_options, ...);
 </pre>
@@ -144,7 +144,7 @@ the batch.
 A database may only be opened by one process at a time.
 The <code>leveldb</code> implementation acquires a lock from the
 operating system to prevent misuse.  Within a single process, the
-same <code>leveldb::DB</code> object may be safely shared by multiple
+same <code>rocksdb::DB</code> object may be safely shared by multiple
 concurrent threads.  I.e., different threads may write into or fetch
 iterators or call <code>Get</code> on the same database without any
 external synchronization (the leveldb implementation will
@@ -160,7 +160,7 @@ The following example demonstrates how to print all key,value pairs
 in a database.
 <p>
 <pre>
-  leveldb::Iterator* it = db-&gt;NewIterator(leveldb::ReadOptions());
+  rocksdb::Iterator* it = db-&gt;NewIterator(rocksdb::ReadOptions());
   for (it-&gt;SeekToFirst(); it-&gt;Valid(); it-&gt;Next()) {
     cout &lt;&lt; it-&gt;key().ToString() &lt;&lt; ": "  &lt;&lt; it-&gt;value().ToString() &lt;&lt; endl;
   }
@@ -196,10 +196,10 @@ implicit snapshot of the current state.
 Snapshots are created by the DB::GetSnapshot() method:
 <p>
 <pre>
-  leveldb::ReadOptions options;
+  rocksdb::ReadOptions options;
   options.snapshot = db-&gt;GetSnapshot();
   ... apply some updates to db ...
-  leveldb::Iterator* iter = db-&gt;NewIterator(options);
+  rocksdb::Iterator* iter = db-&gt;NewIterator(options);
   ... read using iter to view the state when the snapshot was created ...
   delete iter;
   db-&gt;ReleaseSnapshot(options.snapshot);
@@ -211,7 +211,7 @@ support reading as of that snapshot.
 <h1>Slice</h1>
 <p>
 The return value of the <code>it->key()</code> and <code>it->value()</code> calls above
-are instances of the <code>leveldb::Slice</code> type.  <code>Slice</code> is a simple
+are instances of the <code>rocksdb::Slice</code> type.  <code>Slice</code> is a simple
 structure that contains a length and a pointer to an external byte
 array.  Returning a <code>Slice</code> is a cheaper alternative to returning a
 <code>std::string</code> since we do not need to copy potentially large keys and
@@ -223,10 +223,10 @@ C++ strings and null-terminated C-style strings can be easily converted
 to a Slice:
 <p>
 <pre>
-   leveldb::Slice s1 = "hello";
+   rocksdb::Slice s1 = "hello";
 
    std::string str("world");
-   leveldb::Slice s2 = str;
+   rocksdb::Slice s2 = str;
 </pre>
 A Slice can be easily converted back to a C++ string:
 <pre>
@@ -238,7 +238,7 @@ the external byte array into which the Slice points remains live while
 the Slice is in use.  For example, the following is buggy:
 <p>
 <pre>
-   leveldb::Slice slice;
+   rocksdb::Slice slice;
    if (...) {
      std::string str = ...;
      slice = str;
@@ -255,16 +255,16 @@ which orders bytes lexicographically.  You can however supply a custom
 comparator when opening a database.  For example, suppose each
 database key consists of two numbers and we should sort by the first
 number, breaking ties by the second number.  First, define a proper
-subclass of <code>leveldb::Comparator</code> that expresses these rules:
+subclass of <code>rocksdb::Comparator</code> that expresses these rules:
 <p>
 <pre>
-  class TwoPartComparator : public leveldb::Comparator {
+  class TwoPartComparator : public rocksdb::Comparator {
    public:
     // Three-way comparison function:
     //   if a &lt; b: negative result
     //   if a &gt; b: positive result
     //   else: zero result
-    int Compare(const leveldb::Slice&amp; a, const leveldb::Slice&amp; b) const {
+    int Compare(const rocksdb::Slice&amp; a, const rocksdb::Slice&amp; b) const {
       int a1, a2, b1, b2;
       ParseKey(a, &amp;a1, &amp;a2);
       ParseKey(b, &amp;b1, &amp;b2);
@@ -277,7 +277,7 @@ subclass of <code>leveldb::Comparator</code> that expresses these rules:
 
     // Ignore the following methods for now:
     const char* Name() const { return "TwoPartComparator"; }
-    void FindShortestSeparator(std::string*, const leveldb::Slice&amp;) const { }
+    void FindShortestSeparator(std::string*, const rocksdb::Slice&amp;) const { }
     void FindShortSuccessor(std::string*) const { }
   };
 </pre>
@@ -285,18 +285,18 @@ Now create a database using this custom comparator:
 <p>
 <pre>
   TwoPartComparator cmp;
-  leveldb::DB* db;
-  leveldb::Options options;
+  rocksdb::DB* db;
+  rocksdb::Options options;
   options.create_if_missing = true;
   options.comparator = &amp;cmp;
-  leveldb::Status status = leveldb::DB::Open(options, "/tmp/testdb", &amp;db);
+  rocksdb::Status status = rocksdb::DB::Open(options, "/tmp/testdb", &amp;db);
   ...
 </pre>
 <h2>Backwards compatibility</h2>
 <p>
 The result of the comparator's <code>Name</code> method is attached to the
 database when it is created, and is checked on every subsequent
-database open.  If the name changes, the <code>leveldb::DB::Open</code> call will
+database open.  If the name changes, the <code>rocksdb::DB::Open</code> call will
 fail.  Therefore, change the name if and only if the new key format
 and comparison function are incompatible with existing databases, and
 it is ok to discard the contents of all existing databases.
@@ -339,9 +339,9 @@ compression entirely, but should only do so if benchmarks show a
 performance improvement:
 <p>
 <pre>
-  leveldb::Options options;
-  options.compression = leveldb::kNoCompression;
-  ... leveldb::DB::Open(options, name, ...) ....
+  rocksdb::Options options;
+  options.compression = rocksdb::kNoCompression;
+  ... rocksdb::DB::Open(options, name, ...) ....
 </pre>
 <h2>Cache</h2>
 <p>
@@ -353,10 +353,10 @@ uncompressed block contents.
 <pre>
   #include "leveldb/cache.h"
 
-  leveldb::Options options;
-  options.cache = leveldb::NewLRUCache(100 * 1048576);  // 100MB cache
-  leveldb::DB* db;
-  leveldb::DB::Open(options, name, &db);
+  rocksdb::Options options;
+  options.cache = rocksdb::NewLRUCache(100 * 1048576);  // 100MB cache
+  rocksdb::DB* db;
+  rocksdb::DB::Open(options, name, &db);
   ... use the db ...
   delete db
   delete options.cache;
@@ -373,9 +373,9 @@ displacing most of the cached contents.  A per-iterator option can be
 used to achieve this:
 <p>
 <pre>
-  leveldb::ReadOptions options;
+  rocksdb::ReadOptions options;
   options.fill_cache = false;
-  leveldb::Iterator* it = db-&gt;NewIterator(options);
+  rocksdb::Iterator* it = db-&gt;NewIterator(options);
   for (it-&gt;SeekToFirst(); it-&gt;Valid(); it-&gt;Next()) {
     ...
   }
@@ -407,10 +407,10 @@ a single <code>Get()</code> call may involve multiple reads from disk.
 The optional <code>FilterPolicy</code> mechanism can be used to reduce
 the number of disk reads substantially.
 <pre>
-   leveldb::Options options;
+   rocksdb::Options options;
    options.filter_policy = NewBloomFilter(10);
-   leveldb::DB* db;
-   leveldb::DB::Open(options, "/tmp/testdb", &amp;db);
+   rocksdb::DB* db;
+   rocksdb::DB::Open(options, "/tmp/testdb", &amp;db);
    ... use the database ...
    delete db;
    delete options.filter_policy;
@@ -434,7 +434,7 @@ consider a comparator that ignores trailing spaces when comparing keys.
 Instead, the application should provide a custom filter policy that
 also ignores trailing spaces.  For example:
 <pre>
-  class CustomFilterPolicy : public leveldb::FilterPolicy {
+  class CustomFilterPolicy : public rocksdb::FilterPolicy {
    private:
     FilterPolicy* builtin_policy_;
    public:
@@ -484,7 +484,7 @@ checksums are verified:
   parts of its persistent storage have been corrupted.
 <p>
   If a database is corrupted (perhaps it cannot be opened when
-  paranoid checking is turned on), the <code>leveldb::RepairDB</code> function
+  paranoid checking is turned on), the <code>rocksdb::RepairDB</code> function
   may be used to recover as much of the data as possible
 <p>
 </ul>
@@ -494,11 +494,11 @@ The <code>GetApproximateSizes</code> method can used to get the approximate
 number of bytes of file system space used by one or more key ranges.
 <p>
 <pre>
-   leveldb::Range ranges[2];
-   ranges[0] = leveldb::Range("a", "c");
-   ranges[1] = leveldb::Range("x", "z");
+   rocksdb::Range ranges[2];
+   ranges[0] = rocksdb::Range("a", "c");
+   ranges[1] = rocksdb::Range("x", "z");
    uint64_t sizes[2];
-   leveldb::Status s = db-&gt;GetApproximateSizes(ranges, 2, sizes);
+   rocksdb::Status s = db-&gt;GetApproximateSizes(ranges, 2, sizes);
 </pre>
 The preceding call will set <code>sizes[0]</code> to the approximate number of
 bytes of file system space used by the key range <code>[a..c)</code> and
@@ -508,21 +508,21 @@ bytes of file system space used by the key range <code>[a..c)</code> and
 <h1>Environment</h1>
 <p>
 All file operations (and other operating system calls) issued by the
-<code>leveldb</code> implementation are routed through a <code>leveldb::Env</code> object.
+<code>leveldb</code> implementation are routed through a <code>rocksdb::Env</code> object.
 Sophisticated clients may wish to provide their own <code>Env</code>
 implementation to get better control.  For example, an application may
 introduce artificial delays in the file IO paths to limit the impact
 of <code>leveldb</code> on other activities in the system.
 <p>
 <pre>
-  class SlowEnv : public leveldb::Env {
+  class SlowEnv : public rocksdb::Env {
     .. implementation of the Env interface ...
   };
 
   SlowEnv env;
-  leveldb::Options options;
+  rocksdb::Options options;
   options.env = &amp;env;
-  Status s = leveldb::DB::Open(options, ...);
+  Status s = rocksdb::DB::Open(options, ...);
 </pre>
 <h1>Porting</h1>
 <p>
@@ -531,7 +531,7 @@ specific implementations of the types/methods/functions exported by
 <code>leveldb/port/port.h</code>.  See <code>leveldb/port/port_example.h</code> for more
 details.
 <p>
-In addition, the new platform may need a new default <code>leveldb::Env</code>
+In addition, the new platform may need a new default <code>rocksdb::Env</code>
 implementation.  See <code>leveldb/util/env_posix.h</code> for an example.
 
 <h1>Other Information</h1>

hdfs/env_hdfs.h

@@ -25,13 +25,13 @@
 #include <sys/time.h>
 #include <time.h>
 #include <iostream>
-#include "leveldb/env.h"
-#include "leveldb/status.h"
+#include "rocksdb/env.h"
+#include "rocksdb/status.h"
 
 #ifdef USE_HDFS
 #include "hdfs/hdfs.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 static const std::string kProto = "hdfs://";
 static const std::string pathsep = "/";
@@ -107,8 +107,9 @@ class HdfsEnv : public Env {
 
   virtual Status NewLogger(const std::string& fname, Logger** result);
 
-  virtual void Schedule( void (*function)(void* arg), void* arg) {
-    posixEnv->Schedule(function, arg);
+  virtual void Schedule(void (*function)(void* arg), void* arg,
+                        Priority pri = LOW) {
+    posixEnv->Schedule(function, arg, pri);
   }
 
   virtual void StartThread(void (*function)(void* arg), void* arg) {
@@ -140,8 +141,8 @@ class HdfsEnv : public Env {
     return posixEnv->GetAbsolutePath(db_path, output_path);
   }
 
-  virtual void SetBackgroundThreads(int number) {
-    posixEnv->SetBackgroundThreads(number);
+  virtual void SetBackgroundThreads(int number, Priority pri = LOW) {
+    posixEnv->SetBackgroundThreads(number, pri);
   }
 
   virtual std::string TimeToString(uint64_t number) {
@@ -213,12 +214,12 @@ class HdfsEnv : public Env {
   }
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 #else // USE_HDFS
 
 
-namespace leveldb {
+namespace rocksdb {
 
 static const Status notsup;
 
@@ -279,7 +280,8 @@ class HdfsEnv : public Env {
   virtual Status NewLogger(const std::string& fname,
                            shared_ptr<Logger>* result){return notsup;}
 
-  virtual void Schedule( void (*function)(void* arg), void* arg) {}
+  virtual void Schedule(void (*function)(void* arg), void* arg,
+                        Priority pri = LOW) {}
 
   virtual void StartThread(void (*function)(void* arg), void* arg) {}
 
@@ -296,7 +298,7 @@ class HdfsEnv : public Env {
   virtual Status GetAbsolutePath(const std::string& db_path,
       std::string* outputpath) {return notsup;}
 
-  virtual void SetBackgroundThreads(int number) {}
+  virtual void SetBackgroundThreads(int number, Priority pri = LOW) {}
 
   virtual std::string TimeToString(uint64_t number) { return "";}
 };

helpers/memenv/memenv.cc

@@ -4,8 +4,8 @@
 
 #include "helpers/memenv/memenv.h"
 
-#include "leveldb/env.h"
-#include "leveldb/status.h"
+#include "rocksdb/env.h"
+#include "rocksdb/status.h"
 #include "port/port.h"
 #include "util/mutexlock.h"
 #include <map>
@@ -13,7 +13,7 @@
 #include <string>
 #include <vector>
 
-namespace leveldb {
+namespace rocksdb {
 
 namespace {
 
@@ -383,4 +383,4 @@ Env* NewMemEnv(Env* base_env) {
   return new InMemoryEnv(base_env);
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb

helpers/memenv/memenv.h

@@ -2,10 +2,9 @@
 // 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_HELPERS_MEMENV_MEMENV_H_
-#define STORAGE_LEVELDB_HELPERS_MEMENV_MEMENV_H_
-
-namespace leveldb {
+#ifndef STORAGE_ROCKSDB_HELPERS_MEMENV_MEMENV_H_
+#define STORAGE_ROCKSDB_HELPERS_MEMENV_MEMENV_H_
+namespace rocksdb {
 
 class Env;
 
@@ -15,6 +14,6 @@ class Env;
 // *base_env must remain live while the result is in use.
 Env* NewMemEnv(Env* base_env);
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
-#endif  // STORAGE_LEVELDB_HELPERS_MEMENV_MEMENV_H_
+#endif  // STORAGE_ROCKSDB_HELPERS_MEMENV_MEMENV_H_

helpers/memenv/memenv_test.cc

@@ -5,14 +5,14 @@
 #include "helpers/memenv/memenv.h"
 
 #include "db/db_impl.h"
-#include "leveldb/db.h"
-#include "leveldb/env.h"
+#include "rocksdb/db.h"
+#include "rocksdb/env.h"
 #include "util/testharness.h"
 #include <memory>
 #include <string>
 #include <vector>
 
-namespace leveldb {
+namespace rocksdb {
 
 class MemEnvTest {
  public:
@@ -226,8 +226,8 @@ TEST(MemEnvTest, DBTest) {
   delete db;
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
 int main(int argc, char** argv) {
-  return leveldb::test::RunAllTests();
+  return rocksdb::test::RunAllTests();
 }

include/leveldb/ldb_tool.h

@@ -1,14 +0,0 @@
-// Copyright 2008-present Facebook. All Rights Reserved.
-#ifndef STORAGE_LEVELDB_INCLUDE_LDB_TOOL_H
-#define STORAGE_LEVELDB_INCLUDE_LDB_TOOL_H
-#include "leveldb/options.h"
-
-namespace leveldb {
-
-class LDBTool {
- public:
-  void Run(int argc, char** argv, Options = Options());
-};
-
-} // namespace leveldb
-#endif // STORAGE_LEVELDB_INCLUDE_LDB_TOOL_H

include/leveldb/memtablerep.h

@@ -1,88 +0,0 @@
-// This file contains the interface that must be implemented by any collection
-// to be used as the backing store for a MemTable. Such a collection must
-// satisfy the following properties:
-//  (1) It does not store duplicate items.
-//  (2) It uses MemTableRep::KeyComparator to compare items for iteration and
-//     equality.
-//  (3) It can be accessed concurrently by multiple readers but need not support
-//      concurrent writes.
-//  (4) Items are never deleted.
-// The liberal use of assertions is encouraged to enforce (1).
-
-#ifndef STORAGE_LEVELDB_DB_TABLE_H_
-#define STORAGE_LEVELDB_DB_TABLE_H_
-
-#include <memory>
-#include "leveldb/arena.h"
-
-namespace leveldb {
-
-class MemTableRep {
- public:
-  // KeyComparator(a, b) returns a negative value if a is less than b, 0 if they
-  // are equal, and a positive value if b is greater than a
-  class KeyComparator {
-  public:
-    virtual int operator()(const char* a, const char* b) const = 0;
-    virtual ~KeyComparator() { }
-  };
-
-  // Insert key into the collection. (The caller will pack key and value into a
-  // single buffer and pass that in as the parameter to Insert)
-  // REQUIRES: nothing that compares equal to key is currently in the
-  // collection.
-  virtual void Insert(const char* key) = 0;
-
-  // Returns true iff an entry that compares equal to key is in the collection.
-  virtual bool Contains(const char* key) const = 0;
-
-  virtual ~MemTableRep() { }
-
-  // Iteration over the contents of a skip collection
-  class Iterator {
-   public:
-    // Initialize an iterator over the specified collection.
-    // The returned iterator is not valid.
-    // explicit Iterator(const MemTableRep* collection);
-    virtual ~Iterator() { };
-
-    // Returns true iff the iterator is positioned at a valid node.
-    virtual bool Valid() const = 0;
-
-    // Returns the key at the current position.
-    // REQUIRES: Valid()
-    virtual const char* key() const = 0;
-
-    // Advances to the next position.
-    // REQUIRES: Valid()
-    virtual void Next() = 0;
-
-    // Advances to the previous position.
-    // REQUIRES: Valid()
-    virtual void Prev() = 0;
-
-    // Advance to the first entry with a key >= target
-    virtual void Seek(const char* target) = 0;
-
-    // Position at the first entry in collection.
-    // Final state of iterator is Valid() iff collection is not empty.
-    virtual void SeekToFirst() = 0;
-
-    // Position at the last entry in collection.
-    // Final state of iterator is Valid() iff collection is not empty.
-    virtual void SeekToLast() = 0;
-  };
-
-  virtual std::shared_ptr<Iterator> GetIterator() = 0;
-};
-
-class MemTableRepFactory {
- public:
-  virtual ~MemTableRepFactory() { };
-  virtual std::shared_ptr<MemTableRep> CreateMemTableRep(
-    MemTableRep::KeyComparator&, Arena* arena) = 0;
-};
-
-}
-
-#endif // STORAGE_LEVELDB_DB_TABLE_H_

include/leveldb/perf_context.h

@@ -1,24 +0,0 @@
-#ifndef STORAGE_LEVELDB_INCLUDE_PERF_CONTEXT_H
-#define STORAGE_LEVELDB_INCLUDE_PERF_CONTEXT_H
-
-#include <stdint.h>
-
-namespace leveldb {
-
-// A thread local context for gathering performance counter efficiently
-// and transparently.
-
-struct PerfContext {
-
-  void Reset(); // reset all performance counters to zero
-
-
-  uint64_t user_key_comparison_count; // total number of user key comparisons
-};
-
-extern __thread PerfContext perf_context;
-
-}
-
-
-#endif

include/leveldb/types.h

@@ -1,14 +0,0 @@
-#ifndef STORAGE_LEVELDB_INCLUDE_TYPES_H_
-#define STORAGE_LEVELDB_INCLUDE_TYPES_H_
-
-#include <stdint.h>
-
-namespace leveldb {
-
-// Define all public custom types here.
-
-// Represents a sequence number in a WAL file.
-typedef uint64_t SequenceNumber;
-
-}  //  namespace leveldb
-#endif //  STORAGE_LEVELDB_INCLUDE_TYPES_H_

include/rocksdb/arena.h

@@ -5,10 +5,13 @@
 // Arena class defines memory allocation methods. It's used by memtable and
 // skiplist.
 
-#ifndef STORAGE_LEVELDB_INCLUDE_ARENA_H_
-#define STORAGE_LEVELDB_INCLUDE_ARENA_H_
+#ifndef STORAGE_ROCKSDB_INCLUDE_ARENA_H_
+#define STORAGE_ROCKSDB_INCLUDE_ARENA_H_
 
-namespace leveldb {
+#include <limits>
+#include <memory>
+
+namespace rocksdb {
 
 class Arena {
  public:
@@ -33,6 +36,8 @@ class Arena {
   void operator=(const Arena&);
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
-#endif  // STORAGE_LEVELDB_INCLUDE_ARENA_H_
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif  // STORAGE_ROCKSDB_INCLUDE_ARENA_H_

include/rocksdb/c.h

@@ -37,8 +37,8 @@
   (5) All of the pointer arguments must be non-NULL.
 */
 
-#ifndef STORAGE_LEVELDB_INCLUDE_C_H_
-#define STORAGE_LEVELDB_INCLUDE_C_H_
+#ifndef STORAGE_ROCKSDB_INCLUDE_C_H_
+#define STORAGE_ROCKSDB_INCLUDE_C_H_
 
 #ifdef __cplusplus
 extern "C" {
@@ -278,4 +278,4 @@ extern void leveldb_env_destroy(leveldb_env_t*);
 }  /* end extern "C" */
 #endif
 
-#endif  /* STORAGE_LEVELDB_INCLUDE_C_H_ */
+#endif  /* STORAGE_ROCKSDB_INCLUDE_C_H_ */

include/rocksdb/cache.h

@@ -15,14 +15,14 @@
 // they want something more sophisticated (like scan-resistance, a
 // custom eviction policy, variable cache sizing, etc.)
 
-#ifndef STORAGE_LEVELDB_INCLUDE_CACHE_H_
-#define STORAGE_LEVELDB_INCLUDE_CACHE_H_
+#ifndef STORAGE_ROCKSDB_INCLUDE_CACHE_H_
+#define STORAGE_ROCKSDB_INCLUDE_CACHE_H_
 
 #include <memory>
 #include <stdint.h>
-#include "leveldb/slice.h"
+#include "rocksdb/slice.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 using std::shared_ptr;
 
@@ -101,6 +101,8 @@ class Cache {
   void operator=(const Cache&);
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
-#endif  // STORAGE_LEVELDB_UTIL_CACHE_H_
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif  // STORAGE_ROCKSDB_UTIL_CACHE_H_

include/rocksdb/compaction_filter.h

@@ -2,12 +2,12 @@
 // 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_INCLUDE_COMPACTION_FILTER_H_
-#define STORAGE_LEVELDB_INCLUDE_COMPACTION_FILTER_H_
+#ifndef STORAGE_ROCKSDB_INCLUDE_COMPACTION_FILTER_H_
+#define STORAGE_ROCKSDB_INCLUDE_COMPACTION_FILTER_H_
 
 #include <string>
 
-namespace leveldb {
+namespace rocksdb {
 
 class Slice;
 
@@ -66,6 +66,8 @@ class DefaultCompactionFilterFactory : public CompactionFilterFactory {
     }
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
-#endif  // STORAGE_LEVELDB_INCLUDE_COMPACTION_FILTER_H_
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif  // STORAGE_ROCKSDB_INCLUDE_COMPACTION_FILTER_H_

include/rocksdb/comparator.h

@@ -2,12 +2,12 @@
 // 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_INCLUDE_COMPARATOR_H_
-#define STORAGE_LEVELDB_INCLUDE_COMPARATOR_H_
+#ifndef STORAGE_ROCKSDB_INCLUDE_COMPARATOR_H_
+#define STORAGE_ROCKSDB_INCLUDE_COMPARATOR_H_
 
 #include <string>
 
-namespace leveldb {
+namespace rocksdb {
 
 class Slice;
 
@@ -58,6 +58,8 @@ class Comparator {
 // must not be deleted.
 extern const Comparator* BytewiseComparator();
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
-#endif  // STORAGE_LEVELDB_INCLUDE_COMPARATOR_H_
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif  // STORAGE_ROCKSDB_INCLUDE_COMPARATOR_H_

include/rocksdb/db.h

@@ -2,19 +2,19 @@
 // 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_INCLUDE_DB_H_
-#define STORAGE_LEVELDB_INCLUDE_DB_H_
+#ifndef STORAGE_ROCKSDB_INCLUDE_DB_H_
+#define STORAGE_ROCKSDB_INCLUDE_DB_H_
 
 #include <stdint.h>
 #include <stdio.h>
 #include <memory>
 #include <vector>
-#include "leveldb/iterator.h"
-#include "leveldb/options.h"
-#include "leveldb/types.h"
-#include "leveldb/transaction_log.h"
+#include "rocksdb/iterator.h"
+#include "rocksdb/options.h"
+#include "rocksdb/types.h"
+#include "rocksdb/transaction_log.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 using std::unique_ptr;
 
@@ -28,6 +28,17 @@ struct WriteOptions;
 struct FlushOptions;
 class WriteBatch;
 
+// Metadata associated with each SST file.
+struct LiveFileMetaData {
+  std::string name;        // Name of the file
+  int level;               // Level at which this file resides.
+  size_t size;             // File size in bytes.
+  std::string smallestkey; // Smallest user defined key in the file.
+  std::string largestkey;  // Largest user defined key in the file.
+  SequenceNumber smallest_seqno; // smallest seqno in file
+  SequenceNumber largest_seqno;  // largest seqno in file
+};
+
 // Abstract handle to particular state of a DB.
 // A Snapshot is an immutable object and can therefore be safely
 // accessed from multiple threads without any external synchronization.
@@ -198,9 +209,10 @@ class DB {
   // after compaction is reduced, that level might not be appropriate for
   // hosting all the files. In this case, client could set reduce_level
   // to true, to move the files back to the minimum level capable of holding
-  // the data set.
+  // the data set or a given level (specified by non-negative target_level).
   virtual void CompactRange(const Slice* begin, const Slice* end,
-                            bool reduce_level = false) = 0;
+                            bool reduce_level = false,
+                            int target_level = -1) = 0;
 
   // Number of levels used for this DB.
   virtual int NumberLevels() = 0;
@@ -223,21 +235,30 @@ class DB {
   // Allow compactions to delete obselete files.
   virtual Status EnableFileDeletions() = 0;
 
+  // GetLiveFiles followed by GetSortedWalFiles can generate a lossless backup
+
+  // THIS METHOD IS DEPRECATED. Use the GetTableMetaData to get more
+  // detailed information on the live files.
   // Retrieve the list of all files in the database. The files are
-  // relative to the dbname and are not absolute paths. This list
-  // can be used to generate a backup. The valid size of the manifest
-  // file is returned in manifest_file_size. The manifest file is
-  // an ever growing file, but only the portion specified
-  // by manifest_file_size is valid for this snapshot.
+  // relative to the dbname and are not absolute paths. The valid size of the
+  // manifest file is returned in manifest_file_size. The manifest file is an
+  // ever growing file, but only the portion specified by manifest_file_size is
+  // valid for this snapshot.
+  // Setting flush_memtable to true does Flush before recording the live files.
+  // Setting flush_memtable to false is useful when we don't want to wait for
+  // flush which may have to wait for compaction to complete taking an
+  // indeterminate time. But this will have to use GetSortedWalFiles after
+  // GetLiveFiles to compensate for memtables missed in this snapshot due to the
+  // absence of Flush, by WAL files to recover the database consistently later
   virtual Status GetLiveFiles(std::vector<std::string>&,
-                              uint64_t* manifest_file_size) = 0;
+                              uint64_t* manifest_file_size,
+                              bool flush_memtable = true) = 0;
 
   // Retrieve the sorted list of all wal files with earliest file first
   virtual Status GetSortedWalFiles(VectorLogPtr& files) = 0;
 
   // Delete wal files in files. These can be either live or archived.
   // Returns Status::OK if all files could be deleted, otherwise Status::IOError
-  // which contains information about files that could not be deleted.
   virtual Status DeleteWalFiles(const VectorLogPtr& files) = 0;
 
   // The sequence number of the most recent transaction.
@@ -256,6 +277,19 @@ class DB {
   // an update is read.
   virtual Status GetUpdatesSince(SequenceNumber seq_number,
                                  unique_ptr<TransactionLogIterator>* iter) = 0;
+
+  // Delete the file name from the db directory and update the internal
+  // state to reflect that.
+  virtual Status DeleteFile(std::string name) {
+    return Status::OK();
+  }
+
+  // Returns a list of all table files with their level, start key
+  // and end key
+  virtual void GetLiveFilesMetaData(
+    std::vector<LiveFileMetaData> *metadata) {
+  }
+
  private:
   // No copying allowed
   DB(const DB&);
@@ -272,6 +306,8 @@ Status DestroyDB(const std::string& name, const Options& options);
 // on a database that contains important information.
 Status RepairDB(const std::string& dbname, const Options& options);
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
-#endif  // STORAGE_LEVELDB_INCLUDE_DB_H_
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif  // STORAGE_ROCKSDB_INCLUDE_DB_H_

include/rocksdb/env.h

@@ -10,17 +10,17 @@
 // All Env implementations are safe for concurrent access from
 // multiple threads without any external synchronization.
 
-#ifndef STORAGE_LEVELDB_INCLUDE_ENV_H_
-#define STORAGE_LEVELDB_INCLUDE_ENV_H_
+#ifndef STORAGE_ROCKSDB_INCLUDE_ENV_H_
+#define STORAGE_ROCKSDB_INCLUDE_ENV_H_
 
 #include <cstdarg>
 #include <string>
 #include <memory>
 #include <vector>
 #include <stdint.h>
-#include "leveldb/status.h"
+#include "rocksdb/status.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class FileLock;
 class Logger;
@@ -41,7 +41,7 @@ struct EnvOptions {
   EnvOptions();
 
   // construct from Options
-  EnvOptions(const Options& options);
+  explicit EnvOptions(const Options& options);
 
   // If true, then allow caching of data in environment buffers
   bool use_os_buffer;
@@ -162,15 +162,20 @@ class Env {
   // REQUIRES: lock has not already been unlocked.
   virtual Status UnlockFile(FileLock* lock) = 0;
 
-  // Arrange to run "(*function)(arg)" once in a background thread.
-  //
+  enum Priority { LOW, HIGH, TOTAL };
+
+  // Arrange to run "(*function)(arg)" once in a background thread, in
+  // the thread pool specified by pri. By default, jobs go to the 'LOW'
+  // priority thread pool.
+
   // "function" may run in an unspecified thread.  Multiple functions
   // added to the same Env may run concurrently in different threads.
   // I.e., the caller may not assume that background work items are
   // serialized.
   virtual void Schedule(
       void (*function)(void* arg),
-      void* arg) = 0;
+      void* arg,
+      Priority pri = LOW) = 0;
 
   // Start a new thread, invoking "function(arg)" within the new thread.
   // When "function(arg)" returns, the thread will be destroyed.
@@ -190,6 +195,13 @@ class Env {
   // useful for computing deltas of time.
   virtual uint64_t NowMicros() = 0;
 
+  // Returns the number of nano-seconds since some fixed point in time. Only
+  // useful for computing deltas of time in one run.
+  // Default implementation simply relies on NowMicros
+  virtual uint64_t NowNanos() {
+    return NowMicros() * 1000;
+  }
+
   // Sleep/delay the thread for the perscribed number of micro-seconds.
   virtual void SleepForMicroseconds(int micros) = 0;
 
@@ -203,9 +215,10 @@ class Env {
   virtual Status GetAbsolutePath(const std::string& db_path,
       std::string* output_path) = 0;
 
-  // The number of background worker threads for this environment.
-  // default: 1
-  virtual void SetBackgroundThreads(int number) = 0;
+  // The number of background worker threads of a specific thread pool
+  // for this environment. 'LOW' is the default pool.
+  // default number: 1
+  virtual void SetBackgroundThreads(int number, Priority pri = LOW) = 0;
 
   // Converts seconds-since-Jan-01-1970 to a printable string
   virtual std::string TimeToString(uint64_t time) = 0;
@@ -240,6 +253,13 @@ class SequentialFile {
   //
   // REQUIRES: External synchronization
   virtual Status Skip(uint64_t n) = 0;
+
+  // Remove any kind of caching of data from the offset to offset+length
+  // of this file. If the length is 0, then it refers to the end of file.
+  // If the system is not caching the file contents, then this is a noop.
+  virtual Status InvalidateCache(size_t offset, size_t length) {
+    return Status::NotSupported("InvalidateCache not supported.");
+  }
 };
 
 // A file abstraction for randomly reading the contents of a file.
@@ -285,6 +305,12 @@ class RandomAccessFile {
 
   virtual void Hint(AccessPattern pattern) {}
 
+  // Remove any kind of caching of data from the offset to offset+length
+  // of this file. If the length is 0, then it refers to the end of file.
+  // If the system is not caching the file contents, then this is a noop.
+  virtual Status InvalidateCache(size_t offset, size_t length) {
+    return Status::NotSupported("InvalidateCache not supported.");
+  }
 };
 
 // A file abstraction for sequential writing.  The implementation
@@ -334,6 +360,14 @@ class WritableFile {
     *block_size = preallocation_block_size_;
   }
 
+  // Remove any kind of caching of data from the offset to offset+length
+  // of this file. If the length is 0, then it refers to the end of file.
+  // If the system is not caching the file contents, then this is a noop.
+  // This call has no effect on dirty pages in the cache.
+  virtual Status InvalidateCache(size_t offset, size_t length) {
+    return Status::NotSupported("InvalidateCache not supported.");
+  }
+
  protected:
   // PrepareWrite performs any necessary preparation for a write
   // before the write actually occurs.  This allows for pre-allocation
@@ -489,8 +523,8 @@ class EnvWrapper : public Env {
     return target_->LockFile(f, l);
   }
   Status UnlockFile(FileLock* l) { return target_->UnlockFile(l); }
-  void Schedule(void (*f)(void*), void* a) {
-    return target_->Schedule(f, a);
+  void Schedule(void (*f)(void*), void* a, Priority pri) {
+    return target_->Schedule(f, a, pri);
   }
   void StartThread(void (*f)(void*), void* a) {
     return target_->StartThread(f, a);
@@ -518,8 +552,8 @@ class EnvWrapper : public Env {
       std::string* output_path) {
     return target_->GetAbsolutePath(db_path, output_path);
   }
-  void SetBackgroundThreads(int num) {
-    return target_->SetBackgroundThreads(num);
+  void SetBackgroundThreads(int num, Priority pri) {
+    return target_->SetBackgroundThreads(num, pri);
   }
   std::string TimeToString(uint64_t time) {
     return target_->TimeToString(time);
@@ -529,6 +563,8 @@ class EnvWrapper : public Env {
   Env* target_;
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
-#endif  // STORAGE_LEVELDB_INCLUDE_ENV_H_
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif  // STORAGE_ROCKSDB_INCLUDE_ENV_H_

include/rocksdb/filter_policy.h

@@ -13,12 +13,12 @@
 // Most people will want to use the builtin bloom filter support (see
 // NewBloomFilterPolicy() below).
 
-#ifndef STORAGE_LEVELDB_INCLUDE_FILTER_POLICY_H_
-#define STORAGE_LEVELDB_INCLUDE_FILTER_POLICY_H_
+#ifndef STORAGE_ROCKSDB_INCLUDE_FILTER_POLICY_H_
+#define STORAGE_ROCKSDB_INCLUDE_FILTER_POLICY_H_
 
 #include <string>
 
-namespace leveldb {
+namespace rocksdb {
 
 class Slice;
 
@@ -66,5 +66,6 @@ class FilterPolicy {
 extern const FilterPolicy* NewBloomFilterPolicy(int bits_per_key);
 
 }
+#include "rocksdb/rocksdb_to_leveldb.h"
 
-#endif  // STORAGE_LEVELDB_INCLUDE_FILTER_POLICY_H_
+#endif  // STORAGE_ROCKSDB_INCLUDE_FILTER_POLICY_H_

include/rocksdb/iterator.h

@@ -12,13 +12,13 @@
 // non-const method, all threads accessing the same Iterator must use
 // external synchronization.
 
-#ifndef STORAGE_LEVELDB_INCLUDE_ITERATOR_H_
-#define STORAGE_LEVELDB_INCLUDE_ITERATOR_H_
+#ifndef STORAGE_ROCKSDB_INCLUDE_ITERATOR_H_
+#define STORAGE_ROCKSDB_INCLUDE_ITERATOR_H_
 
-#include "leveldb/slice.h"
-#include "leveldb/status.h"
+#include "rocksdb/slice.h"
+#include "rocksdb/status.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class Iterator {
  public:
@@ -65,6 +65,8 @@ class Iterator {
   virtual Slice value() const = 0;
 
   // If an error has occurred, return it.  Else return an ok status.
+  // If non-blocking IO is requested and this operation cannot be
+  // satisfied without doing some IO, then this returns Status::Incomplete().
   virtual Status status() const = 0;
 
   // Clients are allowed to register function/arg1/arg2 triples that
@@ -95,6 +97,8 @@ extern Iterator* NewEmptyIterator();
 // Return an empty iterator with the specified status.
 extern Iterator* NewErrorIterator(const Status& status);
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
-#endif  // STORAGE_LEVELDB_INCLUDE_ITERATOR_H_
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif  // STORAGE_ROCKSDB_INCLUDE_ITERATOR_H_

include/rocksdb/ldb_tool.h

@@ -0,0 +1,20 @@
+// Copyright 2008-present Facebook. All Rights Reserved.
+#ifndef STORAGE_ROCKSDB_INCLUDE_LDB_TOOL_H
+#define STORAGE_ROCKSDB_INCLUDE_LDB_TOOL_H
+#include "rocksdb/options.h"
+
+namespace rocksdb {
+
+class LDBTool {
+ public:
+  void Run(int argc, char** argv, Options = Options());
+};
+
+namespace leveldb = rocksdb;
+
+
+} // namespace rocksdb
+
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif // STORAGE_ROCKSDB_INCLUDE_LDB_TOOL_H

include/rocksdb/memtablerep.h

@@ -0,0 +1,261 @@
+// This file contains the interface that must be implemented by any collection
+// to be used as the backing store for a MemTable. Such a collection must
+// satisfy the following properties:
+//  (1) It does not store duplicate items.
+//  (2) It uses MemTableRep::KeyComparator to compare items for iteration and
+//     equality.
+//  (3) It can be accessed concurrently by multiple readers and can support
+//     during reads. However, it needn't support multiple concurrent writes.
+//  (4) Items are never deleted.
+// The liberal use of assertions is encouraged to enforce (1).
+//
+// The factory will be passed an Arena object when a new MemTableRep is
+// requested. The API for this object is in leveldb/arena.h.
+//
+// Users can implement their own memtable representations. We include four
+// types built in:
+//  - SkipListRep: This is the default; it is backed by a skip list.
+//  - TransformRep: This is backed by an std::unordered_map<Slice,
+// std::set>. On construction, they are given a SliceTransform object. This
+// object is applied to the user key of stored items which indexes into the
+// unordered map to yield a set containing all records that share the same user
+// key under the transform function.
+//  - UnsortedRep: A subclass of TransformRep where the transform function is
+// the identity function. Optimized for point lookups.
+//  - PrefixHashRep: A subclass of TransformRep where the transform function is
+// a fixed-size prefix extractor. If you use PrefixHashRepFactory, the transform
+// must be identical to options.prefix_extractor, otherwise it will be discarded
+// and the default will be used. It is optimized for ranged scans over a
+// prefix.
+//  - VectorRep: This is backed by an unordered std::vector. On iteration, the
+// vector is sorted. It is intelligent about sorting; once the MarkReadOnly()
+// has been called, the vector will only be sorted once. It is optimized for
+// random-write-heavy workloads.
+//
+// The last four implementations are designed for situations in which
+// iteration over the entire collection is rare since doing so requires all the
+// keys to be copied into a sorted data structure.
+
+#ifndef STORAGE_ROCKSDB_DB_MEMTABLEREP_H_
+#define STORAGE_ROCKSDB_DB_MEMTABLEREP_H_
+
+#include <memory>
+#include "rocksdb/arena.h"
+#include "rocksdb/slice.h"
+#include "rocksdb/slice_transform.h"
+
+namespace rocksdb {
+
+class MemTableRep {
+ public:
+  // KeyComparator provides a means to compare keys, which are internal keys
+  // concatenated with values.
+  class KeyComparator {
+   public:
+    // Compare a and b. Return a negative value if a is less than b, 0 if they
+    // are equal, and a positive value if a is greater than b
+    virtual int operator()(const char* a, const char* b) const = 0;
+
+    virtual ~KeyComparator() { }
+  };
+
+  // Insert key into the collection. (The caller will pack key and value into a
+  // single buffer and pass that in as the parameter to Insert)
+  // REQUIRES: nothing that compares equal to key is currently in the
+  // collection.
+  virtual void Insert(const char* key) = 0;
+
+  // Returns true iff an entry that compares equal to key is in the collection.
+  virtual bool Contains(const char* key) const = 0;
+
+  // Notify this table rep that it will no longer be added to. By default, does
+  // nothing.
+  virtual void MarkReadOnly() { }
+
+  // Report an approximation of how much memory has been used other than memory
+  // that was allocated through the arena.
+  virtual size_t ApproximateMemoryUsage() = 0;
+
+  virtual ~MemTableRep() { }
+
+  // Iteration over the contents of a skip collection
+  class Iterator {
+   public:
+    // Initialize an iterator over the specified collection.
+    // The returned iterator is not valid.
+    // explicit Iterator(const MemTableRep* collection);
+    virtual ~Iterator() { };
+
+    // Returns true iff the iterator is positioned at a valid node.
+    virtual bool Valid() const = 0;
+
+    // Returns the key at the current position.
+    // REQUIRES: Valid()
+    virtual const char* key() const = 0;
+
+    // Advances to the next position.
+    // REQUIRES: Valid()
+    virtual void Next() = 0;
+
+    // Advances to the previous position.
+    // REQUIRES: Valid()
+    virtual void Prev() = 0;
+
+    // Advance to the first entry with a key >= target
+    virtual void Seek(const char* target) = 0;
+
+    // Position at the first entry in collection.
+    // Final state of iterator is Valid() iff collection is not empty.
+    virtual void SeekToFirst() = 0;
+
+    // Position at the last entry in collection.
+    // Final state of iterator is Valid() iff collection is not empty.
+    virtual void SeekToLast() = 0;
+  };
+
+  // Return an iterator over the keys in this representation.
+  virtual std::shared_ptr<Iterator> GetIterator() = 0;
+
+  // Return an iterator over at least the keys with the specified user key. The
+  // iterator may also allow access to other keys, but doesn't have to. Default:
+  // GetIterator().
+  virtual std::shared_ptr<Iterator> GetIterator(const Slice& user_key) {
+    return GetIterator();
+  }
+
+  // Return an iterator over at least the keys with the specified prefix. The
+  // iterator may also allow access to other keys, but doesn't have to. Default:
+  // GetIterator().
+  virtual std::shared_ptr<Iterator> GetPrefixIterator(const Slice& prefix) {
+    return GetIterator();
+  }
+
+ protected:
+  // When *key is an internal key concatenated with the value, returns the
+  // user key.
+  virtual Slice UserKey(const char* key) const;
+};
+
+// This is the base class for all factories that are used by RocksDB to create
+// new MemTableRep objects
+class MemTableRepFactory {
+ public:
+  virtual ~MemTableRepFactory() { };
+  virtual std::shared_ptr<MemTableRep> CreateMemTableRep(
+    MemTableRep::KeyComparator&, Arena*) = 0;
+  virtual const char* Name() const = 0;
+};
+
+// This creates MemTableReps that are backed by an std::vector. On iteration,
+// the vector is sorted. This is useful for workloads where iteration is very
+// rare and writes are generally not issued after reads begin.
+//
+// Parameters:
+//   count: Passed to the constructor of the underlying std::vector of each
+//     VectorRep. On initialization, the underlying array will be at least count
+//     bytes reserved for usage.
+class VectorRepFactory : public MemTableRepFactory {
+  const size_t count_;
+public:
+  explicit VectorRepFactory(size_t count = 0) : count_(count) { }
+  virtual std::shared_ptr<MemTableRep> CreateMemTableRep(
+    MemTableRep::KeyComparator&, Arena*) override;
+  virtual const char* Name() const override {
+    return "VectorRepFactory";
+  }
+};
+
+// This uses a skip list to store keys. It is the default.
+class SkipListFactory : public MemTableRepFactory {
+public:
+  virtual std::shared_ptr<MemTableRep> CreateMemTableRep(
+    MemTableRep::KeyComparator&, Arena*) override;
+  virtual const char* Name() const override {
+    return "SkipListFactory";
+  }
+};
+
+// TransformReps are backed by an unordered map of buffers to buckets. When
+// looking up a key, the user key is extracted and a user-supplied transform
+// function (see leveldb/slice_transform.h) is applied to get the key into the
+// unordered map. This allows the user to bin user keys based on arbitrary
+// criteria. Two example implementations are UnsortedRepFactory and
+// PrefixHashRepFactory.
+//
+// Iteration over the entire collection is implemented by dumping all the keys
+// into an std::set. Thus, these data structures are best used when iteration
+// over the entire collection is rare.
+//
+// Parameters:
+//   transform: The SliceTransform to bucket user keys on. TransformRepFactory
+//     owns the pointer.
+//   bucket_count: Passed to the constructor of the underlying
+//     std::unordered_map of each TransformRep. On initialization, the
+//     underlying array will be at least bucket_count size.
+//   num_locks: Number of read-write locks to have for the rep. Each bucket is
+//     hashed onto a read-write lock which controls access to that lock. More
+//     locks means finer-grained concurrency but more memory overhead.
+class TransformRepFactory : public MemTableRepFactory {
+ public:
+  explicit TransformRepFactory(const SliceTransform* transform,
+    size_t bucket_count, size_t num_locks = 1000)
+    : transform_(transform),
+      bucket_count_(bucket_count),
+      num_locks_(num_locks) { }
+
+  virtual ~TransformRepFactory() { delete transform_; }
+
+  virtual std::shared_ptr<MemTableRep> CreateMemTableRep(
+    MemTableRep::KeyComparator&, Arena*) override;
+
+  virtual const char* Name() const override {
+    return "TransformRepFactory";
+  }
+
+  const SliceTransform* GetTransform() { return transform_; }
+
+ protected:
+  const SliceTransform* transform_;
+  const size_t bucket_count_;
+  const size_t num_locks_;
+};
+
+// UnsortedReps bin user keys based on an identity function transform -- that
+// is, transform(key) = key. This optimizes for point look-ups.
+//
+// Parameters: See TransformRepFactory.
+class UnsortedRepFactory : public TransformRepFactory {
+public:
+  explicit UnsortedRepFactory(size_t bucket_count = 0, size_t num_locks = 1000)
+    : TransformRepFactory(NewNoopTransform(),
+                          bucket_count,
+                          num_locks) { }
+  virtual const char* Name() const override {
+    return "UnsortedRepFactory";
+  }
+};
+
+// PrefixHashReps bin user keys based on a fixed-size prefix. This optimizes for
+// short ranged scans over a given prefix.
+//
+// Parameters: See TransformRepFactory.
+class PrefixHashRepFactory : public TransformRepFactory {
+public:
+  explicit PrefixHashRepFactory(const SliceTransform* prefix_extractor,
+    size_t bucket_count = 0, size_t num_locks = 1000)
+    : TransformRepFactory(prefix_extractor, bucket_count, num_locks)
+    { }
+
+  virtual std::shared_ptr<MemTableRep> CreateMemTableRep(
+    MemTableRep::KeyComparator&, Arena*) override;
+
+  virtual const char* Name() const override {
+    return "PrefixHashRepFactory";
+  }
+};
+
+}
+
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif // STORAGE_ROCKSDB_DB_MEMTABLEREP_H_

include/rocksdb/merge_operator.h

@@ -2,14 +2,14 @@
 // 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_INCLUDE_MERGE_OPERATOR_H_
-#define STORAGE_LEVELDB_INCLUDE_MERGE_OPERATOR_H_
+#ifndef STORAGE_ROCKSDB_INCLUDE_MERGE_OPERATOR_H_
+#define STORAGE_ROCKSDB_INCLUDE_MERGE_OPERATOR_H_
 
 #include <string>
 #include <deque>
-#include "leveldb/slice.h"  // TODO: Remove this when migration is done;
+#include "rocksdb/slice.h"  // TODO: Remove this when migration is done;
 
-namespace leveldb {
+namespace rocksdb {
 
 class Slice;
 class Logger;
@@ -142,6 +142,8 @@ class AssociativeMergeOperator : public MergeOperator {
                             Logger* logger) const override;
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
-#endif  // STORAGE_LEVELDB_INCLUDE_MERGE_OPERATOR_H_
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif  // STORAGE_ROCKSDB_INCLUDE_MERGE_OPERATOR_H_

include/rocksdb/options.h

@@ -2,22 +2,21 @@
 // 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_INCLUDE_OPTIONS_H_
-#define STORAGE_LEVELDB_INCLUDE_OPTIONS_H_
+#ifndef STORAGE_ROCKSDB_INCLUDE_OPTIONS_H_
+#define STORAGE_ROCKSDB_INCLUDE_OPTIONS_H_
 
 #include <stddef.h>
 #include <string>
 #include <memory>
 #include <vector>
 #include <stdint.h>
-#include "leveldb/slice.h"
-#include "leveldb/statistics.h"
-#include "leveldb/universal_compaction.h"
-#include "leveldb/memtablerep.h"
+#include "rocksdb/slice.h"
+#include "rocksdb/statistics.h"
+#include "rocksdb/universal_compaction.h"
+#include "rocksdb/memtablerep.h"
+#include "rocksdb/slice_transform.h"
 
-#include "leveldb/slice_transform.h"
-
-namespace leveldb {
+namespace rocksdb {
 
 class Cache;
 class Comparator;
@@ -370,10 +369,23 @@ struct Options {
   // every compaction run.
   uint64_t delete_obsolete_files_period_micros;
 
-  // Maximum number of concurrent background compactions.
+  // Maximum number of concurrent background jobs, submitted to
+  // the default LOW priority thread pool
   // Default: 1
   int max_background_compactions;
 
+  // Maximum number of concurrent background memtable flush jobs, submitted to
+  // the HIGH priority thread pool.
+  // By default, all background jobs (major compaction and memtable flush) go
+  // to the LOW priority pool. If this option is set to a positive number,
+  // memtable flush jobs will be submitted to the HIGH priority pool.
+  // It is important when the same Env is shared by multiple db instances.
+  // Without a separate pool, long running major compaction jobs could
+  // potentially block memtable flush jobs of other db instances, leading to
+  // unnecessary Put stalls.
+  // Default: 0
+  int max_background_flushes;
+
   // Specify the maximal size of the info log file. If the log file
   // is larger than `max_log_file_size`, a new info log file will
   // be created.
@@ -533,6 +545,13 @@ struct Options {
   // Default: false
   bool filter_deletes;
 
+  // An iteration->Next() sequentially skips over keys with the same
+  // user-key unless this option is set. This number specifies the number
+  // of keys (with the same userkey) that will be sequentially
+  // skipped before a reseek is issued.
+  // Default: 8
+  uint64_t max_sequential_skip_in_iterations;
+
   // This is a factory that provides MemTableRep objects.
   // Default: a factory that provides a skip-list-based implementation of
   // MemTableRep.
@@ -542,6 +561,23 @@ struct Options {
   // an application to modify/delete a key-value during background compaction.
   // Default: a factory that doesn't provide any object
   std::shared_ptr<CompactionFilterFactory> compaction_filter_factory;
+
+  // Remove the log file immediately after the corresponding memtable is flushed
+  // to data file.
+  // Default: true
+  bool purge_log_after_memtable_flush;
+};
+
+//
+// An application can issue a read request (via Get/Iterators) and specify
+// if that read should process data that ALREADY resides on a specified cache
+// level. For example, if an application specifies kBlockCacheTier then the
+// Get call will process data that is already processed in the memtable or
+// the block cache. It will not page in data from the OS cache or data that
+// resides in storage.
+enum ReadTier {
+  kReadAllTier    = 0x0, // data in memtable, block cache, OS cache or storage
+  kBlockCacheTier = 0x1  // data in memtable or block cache
 };
 
 // Options that control read operations
@@ -576,15 +612,23 @@ struct ReadOptions {
   // Default: nullptr
   const Slice* prefix;
 
+  // Specify if this read request should process data that ALREADY
+  // resides on a particular cache. If the required data is not
+  // found at the specified cache, then Status::Incomplete is returned.
+  // Default: kReadAllTier
+  ReadTier read_tier;
+
   ReadOptions()
       : verify_checksums(false),
         fill_cache(true),
         snapshot(nullptr),
-        prefix(nullptr) {
+        prefix(nullptr),
+        read_tier(kReadAllTier) {
   }
   ReadOptions(bool cksum, bool cache) :
               verify_checksums(cksum), fill_cache(cache),
-              snapshot(nullptr), prefix(nullptr) {
+              snapshot(nullptr), prefix(nullptr),
+              read_tier(kReadAllTier) {
   }
 };
 
@@ -629,6 +673,8 @@ struct FlushOptions {
   }
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
-#endif  // STORAGE_LEVELDB_INCLUDE_OPTIONS_H_
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif  // STORAGE_ROCKSDB_INCLUDE_OPTIONS_H_

include/rocksdb/perf_context.h

@@ -0,0 +1,41 @@
+#ifndef STORAGE_ROCKSDB_INCLUDE_PERF_CONTEXT_H
+#define STORAGE_ROCKSDB_INCLUDE_PERF_CONTEXT_H
+
+#include <stdint.h>
+
+namespace rocksdb {
+
+enum PerfLevel {
+  kDisable        = 0,  // disable perf stats
+  kEnableCount    = 1,  // enable only count stats
+  kEnableTime     = 2   // enable time stats too
+};
+
+// set the perf stats level
+void SetPerfLevel(PerfLevel level);
+
+// A thread local context for gathering performance counter efficiently
+// and transparently.
+
+struct PerfContext {
+
+  void Reset(); // reset all performance counters to zero
+
+  uint64_t user_key_comparison_count; // total number of user key comparisons
+  uint64_t block_cache_hit_count;
+  uint64_t block_read_count;
+  uint64_t block_read_byte;
+  uint64_t block_read_time;
+  uint64_t block_checksum_time;
+  uint64_t block_decompress_time;
+  uint64_t internal_key_skipped_count;
+  uint64_t internal_delete_skipped_count;
+};
+
+extern __thread PerfContext perf_context;
+
+}
+
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif

include/rocksdb/rocksdb_to_leveldb.h

@@ -0,0 +1,9 @@
+// Copyright (c) 2013 Facebook.
+
+#pragma once
+
+//
+// This is for backward compatibility with applications that use the
+// 'leveldb' namespace. This file will be deleted in a future release.
+//
+namespace leveldb = rocksdb;

include/rocksdb/slice.h

@@ -12,15 +12,15 @@
 // non-const method, all threads accessing the same Slice must use
 // external synchronization.
 
-#ifndef STORAGE_LEVELDB_INCLUDE_SLICE_H_
-#define STORAGE_LEVELDB_INCLUDE_SLICE_H_
+#ifndef STORAGE_ROCKSDB_INCLUDE_SLICE_H_
+#define STORAGE_ROCKSDB_INCLUDE_SLICE_H_
 
 #include <assert.h>
 #include <stddef.h>
 #include <string.h>
 #include <string>
 
-namespace leveldb {
+namespace rocksdb {
 
 class Slice {
  public:
@@ -31,9 +31,11 @@ class Slice {
   Slice(const char* d, size_t n) : data_(d), size_(n) { }
 
   // Create a slice that refers to the contents of "s"
+  /* implicit */
   Slice(const std::string& s) : data_(s.data()), size_(s.size()) { }
 
   // Create a slice that refers to s[0,strlen(s)-1]
+  /* implicit */
   Slice(const char* s) : data_(s), size_(strlen(s)) { }
 
   // Return a pointer to the beginning of the referenced data
@@ -115,7 +117,8 @@ inline int Slice::compare(const Slice& b) const {
   return r;
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
+#include "rocksdb/rocksdb_to_leveldb.h"
 
-#endif  // STORAGE_LEVELDB_INCLUDE_SLICE_H_
+#endif  // STORAGE_ROCKSDB_INCLUDE_SLICE_H_

include/rocksdb/slice_transform.h

@@ -8,12 +8,12 @@
 // define InDomain and InRange to determine which slices are in either
 // of these sets respectively.
 
-#ifndef STORAGE_LEVELDB_INCLUDE_SLICE_TRANSFORM_H_
-#define STORAGE_LEVELDB_INCLUDE_SLICE_TRANSFORM_H_
+#ifndef STORAGE_ROCKSDB_INCLUDE_SLICE_TRANSFORM_H_
+#define STORAGE_ROCKSDB_INCLUDE_SLICE_TRANSFORM_H_
 
 #include <string>
 
-namespace leveldb {
+namespace rocksdb {
 
 class Slice;
 
@@ -36,6 +36,10 @@ class SliceTransform {
 
 extern const SliceTransform* NewFixedPrefixTransform(size_t prefix_len);
 
+extern const SliceTransform* NewNoopTransform();
+
 }
 
-#endif  // STORAGE_LEVELDB_INCLUDE_SLICE_TRANSFORM_H_
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif  // STORAGE_ROCKSDB_INCLUDE_SLICE_TRANSFORM_H_

include/rocksdb/statistics.h

@@ -2,8 +2,8 @@
 // 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_INCLUDE_STATISTICS_H_
-#define STORAGE_LEVELDB_INCLUDE_STATISTICS_H_
+#ifndef STORAGE_ROCKSDB_INCLUDE_STATISTICS_H_
+#define STORAGE_ROCKSDB_INCLUDE_STATISTICS_H_
 
 #include <atomic>
 #include <cassert>
@@ -13,7 +13,7 @@
 #include <memory>
 #include <vector>
 
-namespace leveldb {
+namespace rocksdb {
 
 /**
  * Keep adding ticker's here.
@@ -23,48 +23,68 @@ namespace leveldb {
  * And incrementing TICKER_ENUM_MAX.
  */
 enum Tickers {
-  BLOCK_CACHE_MISS = 0,
-  BLOCK_CACHE_HIT = 1,
-  BLOOM_FILTER_USEFUL = 2, // no. of times bloom filter has avoided file reads.
+  BLOCK_CACHE_MISS,
+  BLOCK_CACHE_HIT,
+  BLOOM_FILTER_USEFUL, // no. of times bloom filter has avoided file reads.
+
   /**
    * COMPACTION_KEY_DROP_* count the reasons for key drop during compaction
    * There are 3 reasons currently.
    */
-  COMPACTION_KEY_DROP_NEWER_ENTRY = 3, // key was written with a newer value.
-  COMPACTION_KEY_DROP_OBSOLETE = 4, // The key is obsolete.
-  COMPACTION_KEY_DROP_USER = 5, // user compaction function has dropped the key.
-  // Number of keys written to the database via the Put and Write call's
-  NUMBER_KEYS_WRITTEN = 6,
-  // Number of Keys read,
-  NUMBER_KEYS_READ = 7,
-  // Bytes written / read
-  BYTES_WRITTEN = 8,
-  BYTES_READ = 9,
-  NO_FILE_CLOSES = 10,
-  NO_FILE_OPENS = 11,
-  NO_FILE_ERRORS = 12,
-  // Time system had to wait to do LO-L1 compactions
-  STALL_L0_SLOWDOWN_MICROS = 13,
-  // Time system had to wait to move memtable to L1.
-  STALL_MEMTABLE_COMPACTION_MICROS = 14,
-  // write throttle because of too many files in L0
-  STALL_L0_NUM_FILES_MICROS = 15,
-  RATE_LIMIT_DELAY_MILLIS = 16,
+  COMPACTION_KEY_DROP_NEWER_ENTRY, // key was written with a newer value.
+  COMPACTION_KEY_DROP_OBSOLETE, // The key is obsolete.
+  COMPACTION_KEY_DROP_USER, // user compaction function has dropped the key.
 
-  NO_ITERATORS = 17, // number of iterators currently open
+  // Number of keys written to the database via the Put and Write call's
+  NUMBER_KEYS_WRITTEN,
+  // Number of Keys read,
+  NUMBER_KEYS_READ,
+  // Bytes written / read
+  BYTES_WRITTEN,
+  BYTES_READ,
+  NO_FILE_CLOSES,
+  NO_FILE_OPENS,
+  NO_FILE_ERRORS,
+  // Time system had to wait to do LO-L1 compactions
+  STALL_L0_SLOWDOWN_MICROS,
+  // Time system had to wait to move memtable to L1.
+  STALL_MEMTABLE_COMPACTION_MICROS,
+  // write throttle because of too many files in L0
+  STALL_L0_NUM_FILES_MICROS,
+  RATE_LIMIT_DELAY_MILLIS,
+
+  NO_ITERATORS, // number of iterators currently open
 
   // Number of MultiGet calls, keys read, and bytes read
-  NUMBER_MULTIGET_CALLS = 18,
-  NUMBER_MULTIGET_KEYS_READ = 19,
-  NUMBER_MULTIGET_BYTES_READ = 20,
+  NUMBER_MULTIGET_CALLS,
+  NUMBER_MULTIGET_KEYS_READ,
+  NUMBER_MULTIGET_BYTES_READ,
 
-  NUMBER_FILTERED_DELETES = 21,
-  NUMBER_MERGE_FAILURES = 22,
-  SEQUENCE_NUMBER = 23,
+  // Number of deletes records that were not required to be
+  // written to storage because key does not exist
+  NUMBER_FILTERED_DELETES,
+  NUMBER_MERGE_FAILURES,
+  SEQUENCE_NUMBER,
 
-  TICKER_ENUM_MAX = 24
+  // number of times bloom was checked before creating iterator on a
+  // file, and the number of times the check was useful in avoiding
+  // iterator creation (and thus likely IOPs).
+  BLOOM_FILTER_PREFIX_CHECKED,
+  BLOOM_FILTER_PREFIX_USEFUL,
+
+  // Number of times we had to reseek inside an iteration to skip
+  // over large number of keys with same userkey.
+  NUMBER_OF_RESEEKS_IN_ITERATION,
+
+  // Record the number of calls to GetUpadtesSince. Useful to keep track of
+  // transaction log iterator refreshes
+  GET_UPDATES_SINCE_CALLS,
+
+  TICKER_ENUM_MAX
 };
 
+// The order of items listed in  Tickers should be the same as
+// the order listed in TickersNameMap
 const std::vector<std::pair<Tickers, std::string>> TickersNameMap = {
   { BLOCK_CACHE_MISS, "rocksdb.block.cache.miss" },
   { BLOCK_CACHE_HIT, "rocksdb.block.cache.hit" },
@@ -89,7 +109,11 @@ const std::vector<std::pair<Tickers, std::string>> TickersNameMap = {
   { NUMBER_MULTIGET_BYTES_READ, "rocksdb.number.multiget.bytes.read" },
   { NUMBER_FILTERED_DELETES, "rocksdb.number.deletes.filtered" },
   { NUMBER_MERGE_FAILURES, "rocksdb.number.merge.failures" },
-  { SEQUENCE_NUMBER, "rocksdb.sequence.number" }
+  { SEQUENCE_NUMBER, "rocksdb.sequence.number" },
+  { BLOOM_FILTER_PREFIX_CHECKED, "rocksdb.bloom.filter.prefix.checked" },
+  { BLOOM_FILTER_PREFIX_USEFUL, "rocksdb.bloom.filter.prefix.useful" },
+  { NUMBER_OF_RESEEKS_IN_ITERATION, "rocksdb.number.reseeks.iteration" },
+  { GET_UPDATES_SINCE_CALLS, "rocksdb.getupdatessince.calls" }
 };
 
 /**
@@ -100,27 +124,27 @@ const std::vector<std::pair<Tickers, std::string>> TickersNameMap = {
  * And increment HISTOGRAM_ENUM_MAX
  */
 enum Histograms {
-  DB_GET = 0,
-  DB_WRITE = 1,
-  COMPACTION_TIME = 2,
-  TABLE_SYNC_MICROS = 3,
-  COMPACTION_OUTFILE_SYNC_MICROS = 4,
-  WAL_FILE_SYNC_MICROS = 5,
-  MANIFEST_FILE_SYNC_MICROS = 6,
+  DB_GET,
+  DB_WRITE,
+  COMPACTION_TIME,
+  TABLE_SYNC_MICROS,
+  COMPACTION_OUTFILE_SYNC_MICROS,
+  WAL_FILE_SYNC_MICROS,
+  MANIFEST_FILE_SYNC_MICROS,
   // TIME SPENT IN IO DURING TABLE OPEN
-  TABLE_OPEN_IO_MICROS = 7,
-  DB_MULTIGET = 8,
-  READ_BLOCK_COMPACTION_MICROS = 9,
-  READ_BLOCK_GET_MICROS = 10,
-  WRITE_RAW_BLOCK_MICROS = 11,
+  TABLE_OPEN_IO_MICROS,
+  DB_MULTIGET,
+  READ_BLOCK_COMPACTION_MICROS,
+  READ_BLOCK_GET_MICROS,
+  WRITE_RAW_BLOCK_MICROS,
 
-  STALL_L0_SLOWDOWN_COUNT = 12,
-  STALL_MEMTABLE_COMPACTION_COUNT = 13,
-  STALL_L0_NUM_FILES_COUNT = 14,
-  HARD_RATE_LIMIT_DELAY_COUNT = 15,
-  SOFT_RATE_LIMIT_DELAY_COUNT = 16,
-  NUM_FILES_IN_SINGLE_COMPACTION = 17,
-  HISTOGRAM_ENUM_MAX = 18
+  STALL_L0_SLOWDOWN_COUNT,
+  STALL_MEMTABLE_COMPACTION_COUNT,
+  STALL_L0_NUM_FILES_COUNT,
+  HARD_RATE_LIMIT_DELAY_COUNT,
+  SOFT_RATE_LIMIT_DELAY_COUNT,
+  NUM_FILES_IN_SINGLE_COMPACTION,
+  HISTOGRAM_ENUM_MAX
 };
 
 const std::vector<std::pair<Histograms, std::string>> HistogramsNameMap = {
@@ -234,6 +258,8 @@ inline void SetTickerCount(std::shared_ptr<Statistics> statistics,
   }
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
-#endif  // STORAGE_LEVELDB_INCLUDE_STATISTICS_H_
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif  // STORAGE_ROCKSDB_INCLUDE_STATISTICS_H_

include/rocksdb/status.h

@@ -10,13 +10,13 @@
 // non-const method, all threads accessing the same Status must use
 // external synchronization.
 
-#ifndef STORAGE_LEVELDB_INCLUDE_STATUS_H_
-#define STORAGE_LEVELDB_INCLUDE_STATUS_H_
+#ifndef STORAGE_ROCKSDB_INCLUDE_STATUS_H_
+#define STORAGE_ROCKSDB_INCLUDE_STATUS_H_
 
 #include <string>
-#include "leveldb/slice.h"
+#include "rocksdb/slice.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class Status {
  public:
@@ -50,6 +50,9 @@ class Status {
   static Status MergeInProgress(const Slice& msg, const Slice& msg2 = Slice()) {
     return Status(kMergeInProgress, msg, msg2);
   }
+  static Status Incomplete(const Slice& msg, const Slice& msg2 = Slice()) {
+    return Status(kIncomplete, msg, msg2);
+  }
 
   // Returns true iff the status indicates success.
   bool ok() const { return (state_ == nullptr); }
@@ -72,6 +75,9 @@ class Status {
   // Returns true iff the status indicates an MergeInProgress.
   bool IsMergeInProgress() const { return code() == kMergeInProgress; }
 
+  // Returns true iff the status indicates Incomplete
+  bool IsIncomplete() const { return code() == kIncomplete; }
+
   // Return a string representation of this status suitable for printing.
   // Returns the string "OK" for success.
   std::string ToString() const;
@@ -91,7 +97,8 @@ class Status {
     kNotSupported = 3,
     kInvalidArgument = 4,
     kIOError = 5,
-    kMergeInProgress = 6
+    kMergeInProgress = 6,
+    kIncomplete = 7
   };
 
   Code code() const {
@@ -114,6 +121,8 @@ inline void Status::operator=(const Status& s) {
   }
 }
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
-#endif  // STORAGE_LEVELDB_INCLUDE_STATUS_H_
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif  // STORAGE_ROCKSDB_INCLUDE_STATUS_H_

include/rocksdb/table_builder.h

@@ -10,14 +10,14 @@
 // non-const method, all threads accessing the same TableBuilder must use
 // external synchronization.
 
-#ifndef STORAGE_LEVELDB_INCLUDE_TABLE_BUILDER_H_
-#define STORAGE_LEVELDB_INCLUDE_TABLE_BUILDER_H_
+#ifndef STORAGE_ROCKSDB_INCLUDE_TABLE_BUILDER_H_
+#define STORAGE_ROCKSDB_INCLUDE_TABLE_BUILDER_H_
 
 #include <stdint.h>
-#include "leveldb/options.h"
-#include "leveldb/status.h"
+#include "rocksdb/options.h"
+#include "rocksdb/status.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class BlockBuilder;
 class BlockHandle;
@@ -90,6 +90,8 @@ class TableBuilder {
   void operator=(const TableBuilder&);
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
-#endif  // STORAGE_LEVELDB_INCLUDE_TABLE_BUILDER_H_
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif  // STORAGE_ROCKSDB_INCLUDE_TABLE_BUILDER_H_

include/rocksdb/transaction_log.h

@@ -1,12 +1,12 @@
 // Copyright 2008-present Facebook. All Rights Reserved.
-#ifndef STORAGE_LEVELDB_INCLUDE_TRANSACTION_LOG_ITERATOR_H_
-#define STORAGE_LEVELDB_INCLUDE_TRANSACTION_LOG_ITERATOR_H_
+#ifndef STORAGE_ROCKSDB_INCLUDE_TRANSACTION_LOG_ITERATOR_H_
+#define STORAGE_ROCKSDB_INCLUDE_TRANSACTION_LOG_ITERATOR_H_
 
-#include "leveldb/status.h"
-#include "leveldb/types.h"
-#include "leveldb/write_batch.h"
+#include "rocksdb/status.h"
+#include "rocksdb/types.h"
+#include "rocksdb/write_batch.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class LogFile;
 typedef std::vector<std::unique_ptr<LogFile>> VectorLogPtr;
@@ -27,8 +27,11 @@ class LogFile {
   LogFile() {}
   virtual ~LogFile() {}
 
-  // Returns log file's name excluding the db path
-  virtual std::string Filename() const = 0;
+  // Returns log file's pathname relative to the main db dir
+  // Eg. For a live-log-file = /000003.log
+  //     For an archived-log-file = /archive/000003.log
+  virtual std::string PathName() const = 0;
+
 
   // Primary identifier for log file.
   // This is directly proportional to creation time of the log file
@@ -73,6 +76,8 @@ class TransactionLogIterator {
   // ONLY use if Valid() is true and status() is OK.
   virtual BatchResult GetBatch() = 0;
 };
-} //  namespace leveldb
+} //  namespace rocksdb
 
-#endif  // STORAGE_LEVELDB_INCLUDE_TRANSACTION_LOG_ITERATOR_H_
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif  // STORAGE_ROCKSDB_INCLUDE_TRANSACTION_LOG_ITERATOR_H_

include/rocksdb/types.h

@@ -0,0 +1,17 @@
+#ifndef STORAGE_ROCKSDB_INCLUDE_TYPES_H_
+#define STORAGE_ROCKSDB_INCLUDE_TYPES_H_
+
+#include <stdint.h>
+
+namespace rocksdb {
+
+// Define all public custom types here.
+
+// Represents a sequence number in a WAL file.
+typedef uint64_t SequenceNumber;
+
+}  //  namespace rocksdb
+
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif //  STORAGE_ROCKSDB_INCLUDE_TYPES_H_

include/rocksdb/universal_compaction.h

@@ -11,10 +11,10 @@
 #include <vector>
 #include <stdint.h>
 #include <climits>
-#include "leveldb/slice.h"
-#include "leveldb/statistics.h"
+#include "rocksdb/slice.h"
+#include "rocksdb/statistics.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 //
 // Algorithm used to make a compaction request stop picking new files
@@ -36,9 +36,21 @@ class CompactionOptionsUniversal {
   // The minimum number of files in a single compaction run. Default: 2
   unsigned int min_merge_width;
 
-  // The maximum number of files in a single compaction run. Default: INT_MAX
+  // The maximum number of files in a single compaction run. Default: UINT_MAX
   unsigned int max_merge_width;
 
+  // The size amplification is defined as the amount (in percentage) of
+  // additional storage needed to store a single byte of data in the database.
+  // For example, a size amplification of 2% means that a database that
+  // contains 100 bytes of user-data may occupy upto 102 bytes of
+  // physical storage. By this definition, a fully compacted database has
+  // a size amplification of 0%. Rocksdb uses the following heuristic
+  // to calculate size amplification: it assumes that all files excluding
+  // the earliest file contribute to the size amplification.
+  // Default: 200, which means that a 100 byte database could require upto
+  // 300 bytes of storage.
+  unsigned int max_size_amplification_percent;
+
   // The algorithm used to stop picking files into a single compaction run
   // Default: kCompactionStopStyleTotalSize
   CompactionStopStyle stop_style;
@@ -48,10 +60,13 @@ class CompactionOptionsUniversal {
     size_ratio(1),
     min_merge_width(2),
     max_merge_width(UINT_MAX),
+    max_size_amplification_percent(200),
     stop_style(kCompactionStopStyleTotalSize) {
   }
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
+
+#include "rocksdb/rocksdb_to_leveldb.h"
 
 #endif  // STORAGE_ROCKSDB_UNIVERSAL_COMPACTION_OPTIONS_H

include/rocksdb/write_batch.h

@@ -18,13 +18,13 @@
 // non-const method, all threads accessing the same WriteBatch must use
 // external synchronization.
 
-#ifndef STORAGE_LEVELDB_INCLUDE_WRITE_BATCH_H_
-#define STORAGE_LEVELDB_INCLUDE_WRITE_BATCH_H_
+#ifndef STORAGE_ROCKSDB_INCLUDE_WRITE_BATCH_H_
+#define STORAGE_ROCKSDB_INCLUDE_WRITE_BATCH_H_
 
 #include <string>
-#include "leveldb/status.h"
+#include "rocksdb/status.h"
 
-namespace leveldb {
+namespace rocksdb {
 
 class Slice;
 
@@ -43,12 +43,13 @@ class WriteBatch {
   // If the database contains a mapping for "key", erase it.  Else do nothing.
   void Delete(const Slice& key);
 
-  // Append a blob of arbitrary to the records in this batch. The blob will be
-  // stored in the transaction log but not in any other file. In particular, it
-  // will not be persisted to the SST files. When iterating over this
+  // Append a blob of arbitrary size to the records in this batch. The blob will
+  // be stored in the transaction log but not in any other file. In particular,
+  // it will not be persisted to the SST files. When iterating over this
   // WriteBatch, WriteBatch::Handler::LogData will be called with the contents
   // of the blob as it is encountered. Blobs, puts, deletes, and merges will be
-  // encountered in the same order in thich they were inserted.
+  // encountered in the same order in thich they were inserted. The blob will
+  // NOT consume sequence number(s) and will NOT increase the count of the batch
   //
   // Example application: add timestamps to the transaction log for use in
   // replication.
@@ -69,6 +70,10 @@ class WriteBatch {
     // The default implementation of LogData does nothing.
     virtual void LogData(const Slice& blob);
     virtual void Delete(const Slice& key) = 0;
+    // Continue is called by WriteBatch::Iterate. If it returns false,
+    // iteration is halted. Otherwise, it continues iterating. The default
+    // implementation always returns true.
+    virtual bool Continue();
   };
   Status Iterate(Handler* handler) const;
 
@@ -89,6 +94,8 @@ class WriteBatch {
   // Intentionally copyable
 };
 
-}  // namespace leveldb
+}  // namespace rocksdb
 
-#endif  // STORAGE_LEVELDB_INCLUDE_WRITE_BATCH_H_
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif  // STORAGE_ROCKSDB_INCLUDE_WRITE_BATCH_H_