Updating README.fb to have newest verison 2.4

Summary: Test Plan: visual
Change namespace from leveldb to rocksdb
2026-07-08 15:15:23 +08:00 · 2013-10-04 12:17:44 -07:00 · 2013-10-04 11:59:26 -07:00 · 2013-10-04 10:47:20 -07:00 · 2013-10-04 10:20:10 -07:00 · 2013-10-03 16:12:43 -07:00
251 changed files with 12260 additions and 4122 deletions
@@ -1,13 +1,16 @@
 build_config.mk
 *.a
-*.o
+*.arc
+*.d
 *.dylib*
+*.gcda
+*.gcno
+*.o
 *.so
 *.so.*
 *_test
-*.arc
-*.d
 db_bench
+db_repl_stress
 db_stress
 ldb
 leveldb_server
@@ -15,4 +18,7 @@ leveldb_shell
 manifest_dump
 sst_dump
 util/build_version.cc
-db_repl_stress
+build_tools/VALGRIND_LOGS/
+coverage/COVERAGE_REPORT
+util/build_version.cc.tmp
+.gdbhistory
@@ -14,13 +14,13 @@ OPT += -O2 -fno-omit-frame-pointer -momit-leaf-frame-pointer
 #-----------------------------------------------

 # detect what platform we're building on
-$(shell ./build_detect_platform build_config.mk)
+$(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

 WARNING_FLAGS = -Wall -Werror
 CFLAGS += -g $(WARNING_FLAGS) -I. -I./include $(PLATFORM_CCFLAGS) $(OPT)
-CXXFLAGS += -g $(WARNING_FLAGS) -I. -I./include $(PLATFORM_CXXFLAGS) $(OPT) -std=gnu++0x
+CXXFLAGS += -g $(WARNING_FLAGS) -I. -I./include $(PLATFORM_CXXFLAGS) $(OPT) -std=gnu++0x -Woverloaded-virtual

 LDFLAGS += $(PLATFORM_LDFLAGS)

@@ -31,39 +31,41 @@ MEMENVOBJECTS = $(MEMENV_SOURCES:.cc=.o)
 TESTUTIL = ./util/testutil.o
 TESTHARNESS = ./util/testharness.o $(TESTUTIL)
 VALGRIND_ERROR = 2
-VALGRIND_DIR = VALGRIND_LOGS
+VALGRIND_DIR = build_tools/VALGRIND_LOGS
 VALGRIND_VER := $(join $(VALGRIND_VER),valgrind)
 VALGRIND_OPTS = --error-exitcode=$(VALGRIND_ERROR) --leak-check=full

 TESTS = \
 	arena_test \
+	auto_roll_logger_test \
+	block_test \
 	bloom_test \
 	c_test \
 	cache_test \
 	coding_test \
-	histogram_test \
 	corruption_test \
 	crc32c_test \
 	db_test \
 	dbformat_test \
 	env_test \
+	filelock_test \
 	filename_test \
 	filter_block_test \
+	histogram_test \
 	log_test \
+	manual_compaction_test \
 	memenv_test \
-	skiplist_test \
-	table_test \
-	ttl_test \
-	block_test \
-	version_edit_test \
-	version_set_test \
-	reduce_levels_test \
-	write_batch_test \
-	auto_roll_logger_test \
-	filelock_test \
 	merge_test \
 	redis_test \
-	manual_compaction_test
+	reduce_levels_test \
+	skiplist_test \
+	stringappend_test \
+	table_test \
+	ttl_test \
+	version_edit_test \
+	version_set_test \
+	write_batch_test\
+	deletefile_test

 TOOLS = \
        sst_dump \
@@ -74,7 +76,10 @@ TOOLS = \
 PROGRAMS = db_bench signal_test $(TESTS) $(TOOLS)
 BENCHMARKS = db_bench_sqlite3 db_bench_tree_db

-LIBRARY = libleveldb.a
+# The library name is configurable since we are maintaining libraries of both
+# debug/release mode.
+LIBNAME = librocksdb
+LIBRARY = ${LIBNAME}.a
 MEMENVLIBRARY = libmemenv.a

 default: all
@@ -83,15 +88,15 @@ default: all
 ifneq ($(PLATFORM_SHARED_EXT),)

 ifneq ($(PLATFORM_SHARED_VERSIONED),true)
-SHARED1 = libleveldb.$(PLATFORM_SHARED_EXT)
+SHARED1 = ${LIBNAME}.$(PLATFORM_SHARED_EXT)
 SHARED2 = $(SHARED1)
 SHARED3 = $(SHARED1)
 SHARED = $(SHARED1)
 else
 # Update db.h if you change these.
-SHARED_MAJOR = 1
-SHARED_MINOR = 5
-SHARED1 = libleveldb.$(PLATFORM_SHARED_EXT)
+SHARED_MAJOR = 2
+SHARED_MINOR = 0
+SHARED1 = ${LIBNAME}.$(PLATFORM_SHARED_EXT)
 SHARED2 = $(SHARED1).$(SHARED_MAJOR)
 SHARED3 = $(SHARED1).$(SHARED_MAJOR).$(SHARED_MINOR)
 SHARED = $(SHARED1) $(SHARED2) $(SHARED3)
@@ -102,15 +107,25 @@ $(SHARED2): $(SHARED3)
 endif

 $(SHARED3):
-	$(CXX) $(LDFLAGS) $(PLATFORM_SHARED_LDFLAGS)$(SHARED2) $(CXXFLAGS) $(PLATFORM_SHARED_CFLAGS) $(SOURCES) -o $(SHARED3)
+	$(CXX) $(LDFLAGS) $(PLATFORM_SHARED_LDFLAGS)$(SHARED2) $(CXXFLAGS) $(COVERAGEFLAGS) $(PLATFORM_SHARED_CFLAGS) $(SOURCES) -o $(SHARED3)

 endif  # PLATFORM_SHARED_EXT

 all: $(SHARED) $(LIBRARY) $(PROGRAMS)

+.PHONY: blackbox_crash_test check clean coverage crash_test ldb_tests \
+	release tags valgrind_check whitebox_crash_test
+
 release:
-	make clean
-	OPT=-DNDEBUG make -j32
+	$(MAKE) clean
+	OPT=-DNDEBUG $(MAKE) -j32
+
+coverage:
+	$(MAKE) clean
+	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

 check: all $(PROGRAMS) $(TESTS) $(TOOLS) ldb_tests
 	for t in $(TESTS); do echo "***** Running $$t"; ./$$t || exit 1; done
@@ -127,6 +142,7 @@ whitebox_crash_test: db_stress
 	python -u tools/db_crashtest2.py

 valgrind_check: all $(PROGRAMS) $(TESTS)
+	mkdir -p $(VALGRIND_DIR)
 	echo TESTS THAT HAVE VALGRIND ERRORS > $(VALGRIND_DIR)/valgrind_failed_tests; \
 	echo TIMES in seconds TAKEN BY TESTS ON VALGRIND > $(VALGRIND_DIR)/valgrind_tests_times; \
 	for t in $(filter-out skiplist_test,$(TESTS)); do \
@@ -140,139 +156,151 @@ valgrind_check: all $(PROGRAMS) $(TESTS)
 	done

 clean:
-	-rm -f $(PROGRAMS) $(BENCHMARKS) $(LIBRARY) $(SHARED) $(MEMENVLIBRARY) $(THRIFTSERVER) */*.o */*/*.o ios-x86/*/*.o ios-arm/*/*.o build_config.mk */*.d */*/*.d
+	-rm -f $(PROGRAMS) $(BENCHMARKS) $(LIBRARY) $(SHARED) $(MEMENVLIBRARY) $(THRIFTSERVER) build_config.mk
 	-rm -rf ios-x86/* ios-arm/*
+	-find . -name "*.[od]" | xargs --no-run-if-empty rm
+	-find . -type f -regex ".*\.\(\(gcda\)\|\(gcno\)\)" | xargs --no-run-if-empty rm
+tags:
+	ctags * -R
+	cscope -b `find . -name '*.cc'` `find . -name '*.h'`

+# ---------------------------------------------------------------------------
+# 	Unit tests and tools
+# ---------------------------------------------------------------------------
 $(LIBRARY): $(LIBOBJECTS)
 	rm -f $@
 	$(AR) -rs $@ $(LIBOBJECTS)

-
 db_bench: db/db_bench.o $(LIBOBJECTS) $(TESTUTIL)
-	$(CXX) db/db_bench.o $(LIBOBJECTS) $(TESTUTIL) $(EXEC_LDFLAGS) -o $@  $(LDFLAGS)
+	$(CXX) db/db_bench.o $(LIBOBJECTS) $(TESTUTIL) $(EXEC_LDFLAGS) -o $@  $(LDFLAGS) $(COVERAGEFLAGS)

 db_stress: tools/db_stress.o $(LIBOBJECTS) $(TESTUTIL)
-	$(CXX) tools/db_stress.o $(LIBOBJECTS) $(TESTUTIL) $(EXEC_LDFLAGS) -o $@  $(LDFLAGS)
+	$(CXX) tools/db_stress.o $(LIBOBJECTS) $(TESTUTIL) $(EXEC_LDFLAGS) -o $@  $(LDFLAGS) $(COVERAGEFLAGS)

 db_repl_stress: tools/db_repl_stress.o $(LIBOBJECTS) $(TESTUTIL)
-	$(CXX) tools/db_repl_stress.o $(LIBOBJECTS) $(TESTUTIL) $(EXEC_LDFLAGS) -o $@  $(LDFLAGS)
+	$(CXX) tools/db_repl_stress.o $(LIBOBJECTS) $(TESTUTIL) $(EXEC_LDFLAGS) -o $@  $(LDFLAGS) $(COVERAGEFLAGS)

 db_bench_sqlite3: doc/bench/db_bench_sqlite3.o $(LIBOBJECTS) $(TESTUTIL)
-	$(CXX) doc/bench/db_bench_sqlite3.o $(LIBOBJECTS) $(TESTUTIL) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) -lsqlite3
+	$(CXX) doc/bench/db_bench_sqlite3.o $(LIBOBJECTS) $(TESTUTIL) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) -lsqlite3 $(COVERAGEFLAGS)

 db_bench_tree_db: doc/bench/db_bench_tree_db.o $(LIBOBJECTS) $(TESTUTIL)
-	$(CXX) doc/bench/db_bench_tree_db.o $(LIBOBJECTS) $(TESTUTIL) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) -lkyotocabinet
+	$(CXX) doc/bench/db_bench_tree_db.o $(LIBOBJECTS) $(TESTUTIL) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) -lkyotocabinet $(COVERAGEFLAGS)

 signal_test: util/signal_test.o $(LIBOBJECTS)
-	$(CXX) util/signal_test.o $(LIBOBJECTS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) util/signal_test.o $(LIBOBJECTS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 arena_test: util/arena_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) util/arena_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) util/arena_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 bloom_test: util/bloom_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) util/bloom_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) util/bloom_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 c_test: db/c_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/c_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) db/c_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 cache_test: util/cache_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) util/cache_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) util/cache_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 coding_test: util/coding_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) util/coding_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) util/coding_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 stringappend_test: utilities/merge_operators/string_append/stringappend_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) utilities/merge_operators/string_append/stringappend_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) utilities/merge_operators/string_append/stringappend_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 redis_test: utilities/redis/redis_lists_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) utilities/redis/redis_lists_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) utilities/redis/redis_lists_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 histogram_test: util/histogram_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) util/histogram_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o$@ $(LDFLAGS)
+	$(CXX) util/histogram_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o$@ $(LDFLAGS) $(COVERAGEFLAGS)

 corruption_test: db/corruption_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/corruption_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) db/corruption_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 crc32c_test: util/crc32c_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) util/crc32c_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) util/crc32c_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 db_test: db/db_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/db_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) db/db_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)
+
+perf_context_test: db/perf_context_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) db/perf_context_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)

 ttl_test: utilities/ttl/ttl_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) utilities/ttl/ttl_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@  $(LDFLAGS)
+	$(CXX) utilities/ttl/ttl_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@  $(LDFLAGS) $(COVERAGEFLAGS)

 dbformat_test: db/dbformat_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/dbformat_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) db/dbformat_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 env_test: util/env_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) util/env_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) util/env_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 filename_test: db/filename_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/filename_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) db/filename_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 filter_block_test: table/filter_block_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) table/filter_block_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) table/filter_block_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 log_test: db/log_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/log_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) db/log_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 table_test: table/table_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) table/table_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) table/table_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 block_test: table/block_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) table/block_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) table/block_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 skiplist_test: db/skiplist_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/skiplist_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) db/skiplist_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 version_edit_test: db/version_edit_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/version_edit_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) db/version_edit_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 version_set_test: db/version_set_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/version_set_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) db/version_set_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 reduce_levels_test: tools/reduce_levels_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) tools/reduce_levels_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) tools/reduce_levels_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 write_batch_test: db/write_batch_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/write_batch_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) db/write_batch_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

-merge_test: db/merge_test.o $(LIBOBJECTS)
-	$(CXX) db/merge_test.o $(LIBOBJECTS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+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)

 memenv_test : helpers/memenv/memenv_test.o $(MEMENVLIBRARY) $(LIBRARY) $(TESTHARNESS)
-	$(CXX) helpers/memenv/memenv_test.o $(MEMENVLIBRARY) $(LIBRARY) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) helpers/memenv/memenv_test.o $(MEMENVLIBRARY) $(LIBRARY) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 manual_compaction_test: util/manual_compaction_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) util/manual_compaction_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) util/manual_compaction_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

-leveldb_shell: tools/shell/ShellContext.o tools/shell/ShellState.o tools/shell/LeveldbShell.o tools/shell/DBClientProxy.o tools/shell/ShellContext.h tools/shell/ShellState.h tools/shell/DBClientProxy.h $(LIBOBJECTS)
-	$(CXX) tools/shell/ShellContext.o tools/shell/ShellState.o tools/shell/LeveldbShell.o tools/shell/DBClientProxy.o $(LIBOBJECTS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+rocksdb_shell: tools/shell/ShellContext.o tools/shell/ShellState.o tools/shell/LeveldbShell.o tools/shell/DBClientProxy.o tools/shell/ShellContext.h tools/shell/ShellState.h tools/shell/DBClientProxy.h $(LIBOBJECTS)
+	$(CXX) tools/shell/ShellContext.o tools/shell/ShellState.o tools/shell/LeveldbShell.o tools/shell/DBClientProxy.o $(LIBOBJECTS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 DBClientProxy_test: tools/shell/test/DBClientProxyTest.o tools/shell/DBClientProxy.o $(LIBRARY)
-	$(CXX) tools/shell/test/DBClientProxyTest.o tools/shell/DBClientProxy.o $(LIBRARY) $(EXEC_LDFLAGS) $(EXEC_LDFLAGS) -o $@  $(LDFLAGS)
+	$(CXX) tools/shell/test/DBClientProxyTest.o tools/shell/DBClientProxy.o $(LIBRARY) $(EXEC_LDFLAGS) $(EXEC_LDFLAGS) -o $@  $(LDFLAGS) $(COVERAGEFLAGS)

 filelock_test: util/filelock_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) util/filelock_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) util/filelock_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 auto_roll_logger_test: util/auto_roll_logger_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) util/auto_roll_logger_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) util/auto_roll_logger_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 sst_dump: tools/sst_dump.o $(LIBOBJECTS)
-	$(CXX) tools/sst_dump.o $(LIBOBJECTS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) tools/sst_dump.o $(LIBOBJECTS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

 ldb: tools/ldb.o $(LIBOBJECTS)
-	$(CXX) tools/ldb.o $(LIBOBJECTS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS)
+	$(CXX) tools/ldb.o $(LIBOBJECTS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

-.PHONY: tags
-tags:
-	ctags `find . -name '*.cc'` `find . -name '*.h'`
-	cscope -b `find . -name '*.cc'` `find . -name '*.h'`
+# ---------------------------------------------------------------------------
+#  	Platform-specific compilation
+# ---------------------------------------------------------------------------

 ifeq ($(PLATFORM), IOS)
 # For iOS, create universal object files to be used on both the simulator and
@@ -284,9 +312,9 @@ IOSVERSION=$(shell defaults read $(PLATFORMSROOT)/iPhoneOS.platform/versionCFBun

 .cc.o:
 	mkdir -p ios-x86/$(dir $@)
-	$(SIMULATORROOT)/usr/bin/$(CXX) $(CXXFLAGS) -isysroot $(SIMULATORROOT)/SDKs/iPhoneSimulator$(IOSVERSION).sdk -arch i686 -c $< -o ios-x86/$@
+	$(SIMULATORROOT)/usr/bin/$(CXX) $(CXXFLAGS) -isysroot $(SIMULATORROOT)/SDKs/iPhoneSimulator$(IOSVERSION).sdk -arch i686 -c $< -o ios-x86/$@ $(COVERAGEFLAGS)
 	mkdir -p ios-arm/$(dir $@)
-	$(DEVICEROOT)/usr/bin/$(CXX) $(CXXFLAGS) -isysroot $(DEVICEROOT)/SDKs/iPhoneOS$(IOSVERSION).sdk -arch armv6 -arch armv7 -c $< -o ios-arm/$@
+	$(DEVICEROOT)/usr/bin/$(CXX) $(CXXFLAGS) -isysroot $(DEVICEROOT)/SDKs/iPhoneOS$(IOSVERSION).sdk -arch armv6 -arch armv7 -c $< -o ios-arm/$@ $(COVERAGEFLAGS)
 	lipo ios-x86/$@ ios-arm/$@ -create -output $@

 .c.o:
@@ -298,17 +326,26 @@ IOSVERSION=$(shell defaults read $(PLATFORMSROOT)/iPhoneOS.platform/versionCFBun

 else
 .cc.o:
-	$(CXX) $(CXXFLAGS) $(PLATFORM_SHARED_CFLAGS) -c $< -o $@
+	$(CXX) $(CXXFLAGS) -c $< -o $@ $(COVERAGEFLAGS)

 .c.o:
-	$(CC) $(CFLAGS) $(PLATFORM_SHARED_CFLAGS) -c $< -o $@
+	$(CC) $(CFLAGS) -c $< -o $@
 endif

+# ---------------------------------------------------------------------------
+#  	Source files dependencies detection
+# ---------------------------------------------------------------------------
+
+# Add proper dependency support so changing a .h file forces a .cc file to
+# rebuild.
+
+# The .d file indicates .cc file's dependencies on .h files. We generate such
+# dependency by g++'s -MM option, whose output is a make dependency rule.
+# The sed command makes sure the "target" file in the generated .d file has
+# the correct path prefix.
 %.d: %.cc
-	$(CXX) $(CXXFLAGS) $(PLATFORM_SHARED_CFLAGS) -MM $< -o $*.d
-	@cp -f $*.d $*.d.tmp
-	@sed -e 's|.*:|$*.o:|' < $*.d.tmp > $*.d
-	@rm -f $*.d.tmp
+	$(CXX) $(CXXFLAGS) $(PLATFORM_SHARED_CFLAGS) -MM $< -o $@
+	@sed -i -e 's|.*:|$*.o:|' $@

 DEPFILES = $(filter-out util/build_version.d,$(SOURCES:.cc=.d))

@@ -317,4 +354,3 @@ depend: $(DEPFILES)
 ifneq ($(MAKECMDGOALS),clean)
 -include $(DEPFILES)
 endif
-
@@ -1,5 +1,7 @@
 rocksdb: A persistent key-value store for flash storage
-Authors: The Facebook Database Engineering Team
+Authors: * The Facebook Database Engineering Team
+         * 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
 key value server, especially suited for storing data on flash drives.
@@ -17,45 +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.h
-include/table_builder.h
+include/rocksdb/table_builder.h
    Lower-level modules that most clients probably won't use directly
+
+include/rocksdb/cache.h
+    An API for the block cache.
+
+include/rocksdb/compaction_filter.h
+    An API for a application filter invoked on every compaction.
+
+include/rocksdb/filter_policy.h
+    An API for configuring a bloom filter.
+
+include/rocksdb/memtablerep.h
+    An API for implementing a memtable.
+
+include/rocksdb/statistics.h
+    An API to retrieve various database statistics.
+
+include/rocksdb/transaction_log.h
+    An API to retrieve transaction logs from a database.
+
+
@@ -1,3 +1,3 @@
 * Detailed instructions on how to compile using fbcode and jemalloc

-* Latest release is 1.5.9.fb
+* Latest release is 2.4.fb
@@ -1,4 +0,0 @@
-This directory stores the tests that failed valgrind and the times associated
-with the failed runs.
-"make valgrind_check" can be invoked to call valgrind on the rocksdb tests and
-generate files in this directory
@@ -1,30 +0,0 @@
-#!/bin/sh
-#
-# Record the version of the source that we are compiling.
-# We keep a record of the git revision in util/version.cc. This source file
-# is then built as a regular source file as part of the compilation process.
-# One can run "strings executable_filename | grep _build_" to find the version of
-# the source that we used to build the executable file.
-#
-
-# create git version file
-VFILE=$(mktemp)
-trap "rm $VFILE" EXIT
-
-# check to see if git is in the path
-which git > /dev/null
-
-if [ "$?" = 0 ]; then
-  env -i git rev-parse HEAD | awk ' BEGIN {print "#include \"build_version.h\""} {print "const char * leveldb_build_git_sha = \"leveldb_build_git_sha:" $0"\";"} END {}' >  ${VFILE}
-else
-  echo "git not found"| awk ' BEGIN {print "#include \"build_version.h\""} {print "const char * leveldb_build_git_sha = \"leveldb_build_git_sha:git not found\";"} END {}' > ${VFILE}
-fi
-
-date | awk 'BEGIN {} {print "const char * leveldb_build_git_datetime = \"leveldb_build_git_datetime:"$0"\";"} END {} ' >> ${VFILE}
-echo "const char * leveldb_build_compile_date = __DATE__;" >> ${VFILE}
-echo "const char * leveldb_build_compile_time = __TIME__;" >> ${VFILE}
-
-OUTFILE=util/build_version.cc
-if [ ! -e $OUTFILE ] || ! cmp -s $VFILE $OUTFILE; then
-    cp $VFILE $OUTFILE
-fi
@@ -31,9 +31,9 @@ fi
 # Default to fbcode gcc on internal fb machines
 if [ -d /mnt/gvfs/third-party -a -z "$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=`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 "$TARGET_OS" in
@@ -78,11 +78,11 @@ case "$TARGET_OS" in
        ;;
    Linux)
        PLATFORM=OS_LINUX
-        COMMON_FLAGS="$COMMON_FLAGS -DOS_LINUX -fPIC"
+        COMMON_FLAGS="$COMMON_FLAGS -DOS_LINUX"
        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
@@ -146,8 +146,8 @@ fi
 set -f # temporarily disable globbing so that our patterns arent expanded
 PRUNE_TEST="-name *test*.cc -prune"
 PRUNE_BENCH="-name *_bench.cc -prune"
-PORTABLE_FILES=`find $DIRS $PRUNE_TEST -o $PRUNE_BENCH -o -name '*.cc' -print | sort | tr "\n" " "`
-PORTABLE_CPP=`find $DIRS $PRUNE_TEST -o $PRUNE_BENCH -o -name '*.cpp' -print | sort | tr "\n" " "`
+PORTABLE_FILES=`cd $ROCKSDB_ROOT; find $DIRS $PRUNE_TEST -o $PRUNE_BENCH -o -name '*.cc' -print | sort | tr "\n" " "`
+PORTABLE_CPP=`cd $ROCKSDB_ROOT; find $DIRS $PRUNE_TEST -o $PRUNE_BENCH -o -name '*.cpp' -print | sort | tr "\n" " "`
 set +f # re-enable globbing

 # The sources consist of the portable files, plus the platform-specific port
@@ -0,0 +1,42 @@
+#!/bin/sh
+#
+# Record the version of the source that we are compiling.
+# We keep a record of the git revision in util/version.cc. This source file
+# is then built as a regular source file as part of the compilation process.
+# One can run "strings executable_filename | grep _build_" to find the version of
+# the source that we used to build the executable file.
+#
+
+# create git version file
+VFILE=$ROCKSDB_ROOT/util/build_version.cc.tmp
+trap "rm $VFILE" EXIT
+
+# check to see if git is in the path
+which git > /dev/null
+
+if [ "$?" = 0 ]; then
+  env -i git rev-parse HEAD |
+  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 "const char* leveldb_build_git_sha = \"leveldb_build_git_sha:git not found\";" }
+    ' > ${VFILE}
+fi
+
+echo "const char* leveldb_build_git_datetime = \"leveldb_build_git_datetime:$(date)\";" >> ${VFILE}
+echo "const char* leveldb_build_compile_date = __DATE__;" >> ${VFILE}
+echo "const char* leveldb_build_compile_time = __TIME__;" >> ${VFILE}
+
+OUTFILE=$ROCKSDB_ROOT/util/build_version.cc
+if [ ! -e $OUTFILE ] || ! cmp -s $VFILE $OUTFILE; then
+    cp $VFILE $OUTFILE
+fi
@@ -14,8 +14,8 @@ fi
 # On the production build servers, set data and stat
 # files/directories not in /tmp or else the tempdir cleaning
 # scripts will make you very unhappy.
-DATA_DIR=${DATA_DIR:-$(mktemp --tmpdir -d leveldb_XXXX)}
-STAT_FILE=${STAT_FILE:-$(mktemp --tmpdir -u leveldb_test_stats_XXXX)}
+DATA_DIR=${DATA_DIR:-$(mktemp -t -d rocksdb_XXXX)}
+STAT_FILE=${STAT_FILE:-$(mktemp -t -u rocksdb_test_stats_XXXX)}

 function cleanup {
  rm -rf $DATA_DIR
@@ -34,7 +34,8 @@ function send_to_ods {
    echo >&2 "ERROR: Key $key doesn't have a value."
    return
  fi
-  curl -s "https://www.intern.facebook.com/intern/agent/ods_set.php?entity=rocksdb_build&key=$key&value=$value"
+  curl -s "https://www.intern.facebook.com/intern/agent/ods_set.php?entity=rocksdb_build&key=$key&value=$value" \
+    --connect-timeout 60
 }

 make clean
@@ -2,7 +2,7 @@
 #A shell script for Jenknis to run valgrind on rocksdb tests
 #Returns 0 on success when there are no failed tests 

-VALGRIND_DIR=VALGRIND_LOGS
+VALGRIND_DIR=build_tools/VALGRIND_LOGS
 make clean
 make -j$(nproc) valgrind_check
 NUM_FAILED_TESTS=$((`wc -l $VALGRIND_DIR/valgrind_failed_tests | awk '{print $1}'` - 1))
@@ -0,0 +1,73 @@
+#!/bin/bash
+
+# Exit on error.
+set -e
+
+if [ -n "$USE_CLANG" ]; then
+  echo "Error: Coverage test is supported only for gcc."
+  exit 1
+fi
+
+ROOT=".."
+# Fetch right version of gcov
+if [ -d /mnt/gvfs/third-party -a -z "$CXX" ]; then
+  source $ROOT/build_tools/fbcode.gcc471.sh
+  GCOV=$TOOLCHAIN_EXECUTABLES/gcc/gcc-4.7.1/cc6c9dc/bin/gcov
+else
+  GCOV=$(which gcov)
+fi
+
+COVERAGE_DIR="$PWD/COVERAGE_REPORT"
+mkdir -p $COVERAGE_DIR
+
+# Find all gcno files to generate the coverage report
+
+GCNO_FILES=`find $ROOT -name "*.gcno"`
+$GCOV --preserve-paths --relative-only --no-output $GCNO_FILES 2>/dev/null |
+  # Parse the raw gcov report to more human readable form.
+  python $ROOT/coverage/parse_gcov_output.py |
+  # Write the output to both stdout and report file.
+  tee $COVERAGE_DIR/coverage_report_all.txt &&
+echo -e "Generated coverage report for all files: $COVERAGE_DIR/coverage_report_all.txt\n"
+
+# TODO: we also need to get the files of the latest commits.
+# Get the most recently committed files.
+LATEST_FILES=`
+  git show --pretty="format:" --name-only HEAD |
+  grep -v "^$" |
+  paste -s -d,`
+RECENT_REPORT=$COVERAGE_DIR/coverage_report_recent.txt
+
+echo -e "Recently updated files: $LATEST_FILES\n" > $RECENT_REPORT
+$GCOV --preserve-paths --relative-only --no-output $GCNO_FILES 2>/dev/null |
+  python $ROOT/coverage/parse_gcov_output.py -interested-files $LATEST_FILES |
+  tee -a $RECENT_REPORT &&
+echo -e "Generated coverage report for recently updated files: $RECENT_REPORT\n"
+
+# Generate the html report. If we cannot find lcov in this machine, we'll simply
+# skip this step.
+echo "Generating the html coverage report..."
+
+LCOV=$(which lcov || true 2>/dev/null)
+if [ -z $LCOV ]
+then
+  echo "Skip: Cannot find lcov to generate the html report."
+  exit 0
+fi
+
+LCOV_VERSION=$(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)
+
+genhtml $COVERAGE_DIR/coverage.info -o $COVERAGE_DIR
+
+echo "HTML Coverage report is generated in $COVERAGE_DIR"
@@ -0,0 +1,118 @@
+import optparse
+import re
+import sys
+
+from optparse import OptionParser
+
+# the gcov report follows certain pattern. Each file will have two lines
+# of report, from which we can extract the file name, total lines and coverage
+# percentage.
+def parse_gcov_report(gcov_input):
+    per_file_coverage = {}
+    total_coverage = None
+
+    for line in sys.stdin:
+        line = line.strip()
+
+        # --First line of the coverage report (with file name in it)?
+        match_obj = re.match("^File '(.*)'$", line)
+        if match_obj:
+            # fetch the file name from the first line of the report.
+            current_file = match_obj.group(1)
+            continue
+
+        # -- Second line of the file report (with coverage percentage)
+        match_obj = re.match("^Lines executed:(.*)% of (.*)", line)
+
+        if match_obj:
+            coverage = float(match_obj.group(1))
+            lines = int(match_obj.group(2))
+
+            if current_file is not None:
+                per_file_coverage[current_file] = (coverage, lines)
+                current_file = None
+            else:
+                # If current_file is not set, we reach the last line of report,
+                # which contains the summarized coverage percentage.
+                total_coverage = (coverage, lines)
+            continue
+
+        # If the line's pattern doesn't fall into the above categories. We
+        # can simply ignore them since they're either empty line or doesn't
+        # find executable lines of the given file.
+        current_file = None
+
+    return per_file_coverage, total_coverage
+
+def get_option_parser():
+    usage = "Parse the gcov output and generate more human-readable code " +\
+            "coverage report."
+    parser = OptionParser(usage)
+
+    parser.add_option(
+        "--interested-files", "-i",
+        dest="filenames",
+        help="Comma separated files names. if specified, we will display " +
+             "the coverage report only for interested source files. " +
+             "Otherwise we will display the coverage report for all " +
+             "source files."
+    )
+    return parser
+
+def display_file_coverage(per_file_coverage, total_coverage):
+    # To print out auto-adjustable column, we need to know the longest
+    # length of file names.
+    max_file_name_length = max(
+        len(fname) for fname in per_file_coverage.keys()
+    )
+
+    # -- Print header
+    # size of separator is determined by 3 column sizes:
+    # file name, coverage percentage and lines.
+    header_template = \
+        "%" + str(max_file_name_length) + "s\t%s\t%s"
+    separator = "-" * (max_file_name_length + 10 + 20)
+    print header_template % ("Filename", "Coverage", "Lines")
+    print separator
+
+    # -- Print body
+    # template for printing coverage report for each file.
+    record_template = "%" + str(max_file_name_length) + "s\t%5.2f%%\t%10d"
+
+    for fname, coverage_info in per_file_coverage.items():
+        coverage, lines = coverage_info
+        print record_template % (fname, coverage, lines)
+
+    # -- Print footer
+    if total_coverage:
+        print separator
+        print record_template % ("Total", total_coverage[0], total_coverage[1])
+
+def report_coverage():
+    parser = get_option_parser()
+    (options, args) = parser.parse_args()
+
+    interested_files = set()
+    if options.filenames is not None:
+        interested_files = set(f.strip() for f in options.filenames.split(','))
+
+    # To make things simple, right now we only read gcov report from the input
+    per_file_coverage, total_coverage = parse_gcov_report(sys.stdin)
+
+    # Check if we need to display coverage info for interested files.
+    if len(interested_files):
+        per_file_coverage = dict(
+            (fname, per_file_coverage[fname]) for fname in interested_files
+            if fname in per_file_coverage
+        )
+        # If we only interested in several files, it makes no sense to report
+        # the total_coverage
+        total_coverage = None
+
+    if not len(per_file_coverage):
+        print >> sys.stderr, "Cannot find coverage info for the given files."
+        return
+    display_file_coverage(per_file_coverage, total_coverage)
+
+if __name__ == "__main__":
+    report_coverage()
@@ -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,
@@ -28,6 +28,7 @@ Status BuildTable(const std::string& dbname,
                  const SequenceNumber earliest_seqno_in_memtable) {
  Status s;
  meta->file_size = 0;
+  meta->smallest_seqno = meta->largest_seqno = 0;
  iter->SeekToFirst();

  // If the sequence number of the smallest entry in the memtable is
@@ -50,86 +51,106 @@ Status BuildTable(const std::string& dbname,
    // the first key is the smallest key
    Slice key = iter->key();
    meta->smallest.DecodeFrom(key);
+    meta->smallest_seqno = GetInternalKeySeqno(key);
+    meta->largest_seqno = meta->smallest_seqno;

-    MergeHelper merge(user_comparator, options.merge_operator,
+    MergeHelper merge(user_comparator, options.merge_operator.get(),
                      options.info_log.get(),
                      true /* internal key corruption is not ok */);

    if (purge) {
-      ParsedInternalKey ikey;
      // Ugly walkaround to avoid compiler error for release build
-      // TODO: find a clean way to treat in memory key corruption
-      ikey.type = kTypeValue;
+      bool ok __attribute__((unused)) = true;
+
+      // Will write to builder if current key != prev key
      ParsedInternalKey prev_ikey;
-      std::string prev_value;
      std::string prev_key;
-
-      // Ugly walkaround to avoid compiler error for release build
-      // TODO: find a clean way to treat in memory key corruption
-      auto ok __attribute__((unused)) = ParseInternalKey(key, &ikey);
-      // in-memory key corruption is not ok;
-      assert(ok);
-
-      if (ikey.type == kTypeMerge) {
-        // merge values if the first entry is of merge type
-        merge.MergeUntil(iter,  0 /* don't worry about snapshot */);
-        prev_key.assign(merge.key().data(), merge.key().size());
-        ok = ParseInternalKey(Slice(prev_key), &prev_ikey);
-        assert(ok);
-        prev_value.assign(merge.value().data(), merge.value().size());
-      } else {
-        // store first key-value
-        prev_key.assign(key.data(), key.size());
-        prev_value.assign(iter->value().data(), iter->value().size());
-        ok = ParseInternalKey(Slice(prev_key), &prev_ikey);
-        assert(ok);
-        assert(prev_ikey.sequence >= earliest_seqno_in_memtable);
-        iter->Next();
-      }
+      bool is_first_key = true;    // Also write if this is the very first key

      while (iter->Valid()) {
        bool iterator_at_next = false;
+
+        // Get current key
        ParsedInternalKey this_ikey;
        Slice key = iter->key();
+        Slice value = iter->value();
+
+        // In-memory key corruption is not ok;
+        // TODO: find a clean way to treat in memory key corruption
        ok = ParseInternalKey(key, &this_ikey);
        assert(ok);
        assert(this_ikey.sequence >= earliest_seqno_in_memtable);

-        if (user_comparator->Compare(prev_ikey.user_key, this_ikey.user_key)) {
-          // This key is different from previous key.
-          // Output prev key and remember current key
-          builder->Add(Slice(prev_key), Slice(prev_value));
+        // If the key is the same as the previous key (and it is not the
+        // first key), then we skip it, since it is an older version.
+        // Otherwise we output the key and mark it as the "new" previous key.
+        if (!is_first_key && !user_comparator->Compare(prev_ikey.user_key,
+                                                       this_ikey.user_key)) {
+          // seqno within the same key are in decreasing order
+          assert(this_ikey.sequence < prev_ikey.sequence);
+        } else {
+          is_first_key = false;
+
          if (this_ikey.type == kTypeMerge) {
+            // Handle merge-type keys using the MergeHelper
            merge.MergeUntil(iter, 0 /* don't worry about snapshot */);
            iterator_at_next = true;
-            prev_key.assign(merge.key().data(), merge.key().size());
-            ok = ParseInternalKey(Slice(prev_key), &prev_ikey);
-            assert(ok);
-            prev_value.assign(merge.value().data(), merge.value().size());
+            if (merge.IsSuccess()) {
+              // Merge completed correctly.
+              // Add the resulting merge key/value and continue to next
+              builder->Add(merge.key(), merge.value());
+              prev_key.assign(merge.key().data(), merge.key().size());
+              ok = ParseInternalKey(Slice(prev_key), &prev_ikey);
+              assert(ok);
+            } else {
+              // Merge did not find a Put/Delete.
+              // Can not compact these merges into a kValueType.
+              // Write them out one-by-one. (Proceed back() to front())
+              const std::deque<std::string>& keys = merge.keys();
+              const std::deque<std::string>& values = merge.values();
+              assert(keys.size() == values.size() && keys.size() >= 1);
+              std::deque<std::string>::const_reverse_iterator key_iter;
+              std::deque<std::string>::const_reverse_iterator value_iter;
+              for (key_iter=keys.rbegin(), value_iter = values.rbegin();
+                   key_iter != keys.rend() && value_iter != values.rend();
+                   ++key_iter, ++value_iter) {
+
+                builder->Add(Slice(*key_iter), Slice(*value_iter));
+              }
+
+              // Sanity check. Both iterators should end at the same time
+              assert(key_iter == keys.rend() && value_iter == values.rend());
+
+              prev_key.assign(keys.front());
+              ok = ParseInternalKey(Slice(prev_key), &prev_ikey);
+              assert(ok);
+            }
          } else {
+            // Handle Put/Delete-type keys by simply writing them
+            builder->Add(key, value);
            prev_key.assign(key.data(), key.size());
-            prev_value.assign(iter->value().data(), iter->value().size());
            ok = ParseInternalKey(Slice(prev_key), &prev_ikey);
            assert(ok);
          }
-        } else {
-          // seqno within the same key are in decreasing order
-          assert(this_ikey.sequence < prev_ikey.sequence);
-          // This key is an earlier version of the same key in prev_key.
-          // Skip current key.
        }

        if (!iterator_at_next) iter->Next();
      }
-      // output last key
-      builder->Add(Slice(prev_key), Slice(prev_value));
+
+      // The last key is the largest key
      meta->largest.DecodeFrom(Slice(prev_key));
+      SequenceNumber seqno = GetInternalKeySeqno(Slice(prev_key));
+      meta->smallest_seqno = std::min(meta->smallest_seqno, seqno);
+      meta->largest_seqno = std::max(meta->largest_seqno, seqno);

    } else {
      for (; iter->Valid(); iter->Next()) {
        Slice key = iter->key();
        meta->largest.DecodeFrom(key);
        builder->Add(key, iter->value());
+        SequenceNumber seqno = GetInternalKeySeqno(key);
+        meta->smallest_seqno = std::min(meta->smallest_seqno, seqno);
+        meta->largest_seqno = std::max(meta->largest_seqno, seqno);
      }
    }

@@ -183,4 +204,4 @@ Status BuildTable(const std::string& dbname,
  return s;
 }

-}  // namespace leveldb
+}  // namespace rocksdb
@@ -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_
@@ -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;

@@ -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>
@@ -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;

@@ -57,6 +57,7 @@ class CorruptionTest {
    opt.env = &env_;
    opt.block_cache = tiny_cache_;
    opt.block_size_deviation = 0;
+    opt.arena_block_size = 4096;
    return DB::Open(opt, dbname_, &db_);
  }

@@ -67,7 +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) {
@@ -353,8 +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();
 }
@@ -2,17 +2,18 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.

-#include "db/db_impl.h"
-#include "db/filename.h"
+#include <algorithm>
 #include <string>
 #include <stdint.h>
+#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_);
@@ -29,16 +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_);
@@ -47,18 +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();
@@ -66,4 +69,46 @@ Status DBImpl::GetLiveFiles(std::vector<std::string>& ret,
  return Status::OK();
 }

+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.
+  Status s;
+  std::string archivedir = ArchivalDirectory(dbname_);
+  if (env_->FileExists(archivedir)) {
+    s = AppendSortedWalsOfType(archivedir, files, kArchivedLogFile);
+    if (!s.ok()) {
+      return s;
+    }
+  }
+  // list wal files in main db dir.
+  return AppendSortedWalsOfType(dbname_, files, kAliveLogFile);
+}
+
+Status DBImpl::DeleteWalFiles(const VectorLogPtr& files) {
+  Status s;
+  std::string archivedir = ArchivalDirectory(dbname_);
+  std::string files_not_deleted;
+  for (const auto& wal : files) {
+    /* 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
+     */
+    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()) {
+    return Status::IOError("Deleted all requested files except: " +
+                           files_not_deleted);
+  }
+  return Status::OK();
+}
+
 }
@@ -10,11 +10,12 @@
 #include <set>
 #include <vector>
 #include "db/dbformat.h"
-#include "db/log_file.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 "rocksdb/memtablerep.h"
+#include "rocksdb/transaction_log.h"
 #include "port/port.h"
 #include "util/stats_logger.h"
 #include "memtablelist.h"
@@ -23,7 +24,7 @@
 #include "scribe/scribe_logger.h"
 #endif

-namespace leveldb {
+namespace rocksdb {

 class MemTable;
 class TableCache;
@@ -48,12 +49,22 @@ class DBImpl : public DB {
  virtual std::vector<Status> MultiGet(const ReadOptions& options,
                                       const std::vector<Slice>& keys,
                                       std::vector<std::string>* values);
+
+  // Returns false if key doesn't exist in the database and true if it may.
+  // If value_found is not passed in as null, then return the value if found in
+  // memory. On return, if value was found, then value_found will be set to true
+  // , otherwise false.
+  virtual bool KeyMayExist(const ReadOptions& options,
+                           const Slice& key,
+                           std::string* value,
+                           bool* value_found = nullptr);
  virtual Iterator* NewIterator(const ReadOptions&);
  virtual const Snapshot* GetSnapshot();
  virtual void ReleaseSnapshot(const Snapshot* snapshot);
  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);
+  virtual void CompactRange(const Slice* begin, const Slice* end,
+                            bool reduce_level = false, int target_level = -1);
  virtual int NumberLevels();
  virtual int MaxMemCompactionLevel();
  virtual int Level0StopWriteTrigger();
@@ -61,10 +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

@@ -155,6 +173,7 @@ class DBImpl : public DB {
  Status WriteLevel0Table(std::vector<MemTable*> &mems, VersionEdit* edit,
                                uint64_t* filenumber);

+  uint64_t SlowdownAmount(int n, int top, int bottom);
  Status MakeRoomForWrite(bool force /* compact even if there is room? */);
  WriteBatch* BuildBatchGroup(Writer** last_writer);

@@ -169,9 +188,12 @@ class DBImpl : public DB {
  void LogDBDeployStats();

  void MaybeScheduleCompaction();
-  static void BGWork(void* db);
-  void BackgroundCall();
-  Status BackgroundCompaction(bool* madeProgress, DeletionState& deletion_state);
+  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);

@@ -194,21 +216,25 @@ 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 ListAllWALFiles(const std::string& path,
-                         std::vector<LogFile>* logFiles,
-                         WalFileType type);
+  Status AppendSortedWalsOfType(const std::string& path,
+                                VectorLogPtr& log_files,
+                                WalFileType type);

-  //  Find's all the log files which contain updates with seq no.
-  //  Greater Than or Equal to the requested SequenceNumber
-  Status FindProbableWALFiles(std::vector<LogFile>* const allLogs,
-                              std::vector<LogFile>* const result,
-                              const SequenceNumber target);
+  // Requires: all_logs should be sorted with earliest log file first
+  // Retains all log files in all_logs which contain updates with seq no.
+  // Greater Than or Equal to the requested SequenceNumber.
+  Status RetainProbableWalFiles(VectorLogPtr& all_logs,
+                                const SequenceNumber target);
  //  return true if
-  bool CheckFileExistsAndEmpty(const LogFile& file);
+  bool CheckWalFileExistsAndEmpty(const WalFileType type,
+                                  const uint64_t number);

-  Status ReadFirstRecord(const LogFile& file, WriteBatch* const result);
+  Status ReadFirstRecord(const WalFileType type, const uint64_t number,
+                         WriteBatch* const result);

  Status ReadFirstLine(const std::string& fname, WriteBatch* const batch);

@@ -217,6 +243,15 @@ class DBImpl : public DB {
  // dump leveldb.stats to LOG
  void MaybeDumpStats();

+  // Return the minimum empty level that could hold the total data in the
+  // 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 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_;
  bool owns_info_log_;
@@ -231,6 +266,7 @@ class DBImpl : public DB {
  port::Mutex mutex_;
  port::AtomicPointer shutting_down_;
  port::CondVar bg_cv_;          // Signalled when background work finishes
+  std::shared_ptr<MemTableRepFactory> mem_rep_factory_;
  MemTable* mem_;
  MemTableList imm_;             // Memtable that are not changing
  uint64_t logfile_number_;
@@ -251,6 +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_;

@@ -289,6 +328,10 @@ class DBImpl : public DB {
  uint64_t stall_memtable_compaction_;
  uint64_t stall_level0_num_files_;
  std::vector<uint64_t> stall_leveln_slowdown_;
+  uint64_t stall_level0_slowdown_count_;
+  uint64_t stall_memtable_compaction_count_;
+  uint64_t stall_level0_num_files_count_;
+  std::vector<uint64_t> stall_leveln_slowdown_count_;

  // Time at which this instance was started.
  const uint64_t started_at_;
@@ -366,6 +409,12 @@ class DBImpl : public DB {
  // The options to access storage files
  const EnvOptions storage_options_;

+  // A value of true temporarily disables scheduling of background work
+  bool bg_work_gate_closed_;
+
+  // Guard against multiple concurrent refitting
+  bool refitting_level_;
+
  // No copying allowed
  DBImpl(const DBImpl&);
  void operator=(const DBImpl&);
@@ -379,6 +428,13 @@ class DBImpl : public DB {
    SequenceNumber in,
    std::vector<SequenceNumber>& snapshots,
    SequenceNumber* prev_snapshot);
+
+  // Function that Get and KeyMayExist call with no_io true or false
+  // Note: 'value_found' from KeyMayExist propagates here
+  Status GetImpl(const ReadOptions& options,
+                 const Slice& key,
+                 std::string* value,
+                 bool* value_found = nullptr);
 };

 // Sanitize db options.  The caller should delete result.info_log if
@@ -388,6 +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_
@@ -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)
@@ -53,11 +53,12 @@ Status DBImplReadOnly::Get(const ReadOptions& options,
  MemTable* mem = GetMemTable();
  Version* current = versions_->current();
  SequenceNumber snapshot = versions_->LastSequence();
+  std::deque<std::string> merge_operands;
  LookupKey lkey(key, snapshot);
-  if (mem->Get(lkey, value, &s, options_)) {
+  if (mem->Get(lkey, value, &s, &merge_operands, options_)) {
  } else {
    Version::GetStats stats;
-    current->Get(options, lkey, value, &s, &stats, options_);
+    current->Get(options, lkey, value, &s, &merge_operands, &stats, options_);
  }
  return s;
 }
@@ -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:
@@ -32,6 +32,9 @@ public:
 virtual Status Get(const ReadOptions& options,
                    const Slice& key,
                    std::string* value);
+
+ // TODO: Implement ReadOnly MultiGet?
+
 virtual Iterator* NewIterator(const ReadOptions&);

 virtual Status Put(const WriteOptions&, const Slice& key, const Slice& value) {
@@ -47,7 +50,8 @@ public:
 virtual Status Write(const WriteOptions& options, WriteBatch* updates) {
   return Status::NotSupported("Not supported operation in read only mode.");
 }
- virtual void CompactRange(const Slice* begin, const Slice* end) {
+ virtual void CompactRange(const Slice* begin, const Slice* end,
+                           bool reduce_level = false, int target_level = -1) {
 }
 virtual Status DisableFileDeletions() {
   return Status::NotSupported("Not supported operation in read only mode.");
@@ -56,7 +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) {
@@ -4,18 +4,20 @@

 #include "db/db_iter.h"
 #include <stdexcept>
+#include <deque>

 #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) {
@@ -56,7 +58,7 @@ class DBIter: public Iterator {
        env_(env),
        logger_(options.info_log),
        user_comparator_(cmp),
-        user_merge_operator_(options.merge_operator),
+        user_merge_operator_(options.merge_operator.get()),
        iter_(iter),
        sequence_(s),
        direction_(kForward),
@@ -64,6 +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);
@@ -128,6 +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&);
@@ -187,12 +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) {
@@ -201,6 +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;
@@ -211,15 +219,28 @@ void DBIter::FindNextUserEntry(bool skipping) {
            SaveKey(ikey.user_key, &saved_key_);
            current_entry_is_merged_ = true;
            valid_ = true;
-            // Go to a different state machine
-            MergeValuesNewToOld();
-            // TODO: what if !iter_->Valid()
+            MergeValuesNewToOld();  // Go to a different state machine
            return;
+          case kTypeLogData:
+            assert(false);
            break;
        }
      }
    }
-    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;
 }
@@ -231,10 +252,17 @@ void DBIter::FindNextUserEntry(bool skipping) {
 // POST: saved_value_ has the merged value for the user key
 //       iter_ points to the next entry (or invalid)
 void DBIter::MergeValuesNewToOld() {
+  if (!user_merge_operator_) {
+    Log(logger_, "Options::merge_operator is null.");
+    throw std::logic_error("DBIter::MergeValuesNewToOld() with"
+                           " Options::merge_operator null");
+  }

-  const Slice value = iter_->value();
-  std::string operand(value.data(), value.size());
+  // Start the merge process by pushing the first operand
+  std::deque<std::string> operands;
+  operands.push_front(iter_->value().ToString());

+  std::string merge_result;   // Temporary string to hold merge result later
  ParsedInternalKey ikey;
  for (iter_->Next(); iter_->Valid(); iter_->Next()) {
    if (!ParseKey(&ikey)) {
@@ -255,40 +283,53 @@ void DBIter::MergeValuesNewToOld() {
    }

    if (kTypeValue == ikey.type) {
-      // hit a put, merge the put value with operand and store it in the
-      // final result saved_value_. We are done!
+      // hit a put, merge the put value with operands and store the
+      // final result in saved_value_. We are done!
+      // ignore corruption if there is any.
      const Slice value = iter_->value();
-      user_merge_operator_->Merge(ikey.user_key, &value, Slice(operand),
-                                  &saved_value_, logger_.get());
+      user_merge_operator_->FullMerge(ikey.user_key, &value, operands,
+                                      &saved_value_, logger_.get());
      // iter_ is positioned after put
      iter_->Next();
      return;
    }

    if (kTypeMerge == ikey.type) {
-      // hit a merge, merge the value with operand and continue.
-      // saved_value_ is used as a scratch area. The result is put
-      // back in operand
-      const Slice value = iter_->value();
-      user_merge_operator_->Merge(ikey.user_key, &value, operand,
-                                  &saved_value_, logger_.get());
-      swap(saved_value_, operand);
+      // hit a merge, add the value as an operand and run associative merge.
+      // when complete, add result to operands and continue.
+      const Slice& value = iter_->value();
+      operands.push_front(value.ToString());
+      while(operands.size() >= 2) {
+        // Call user associative-merge until it returns false
+        if (user_merge_operator_->PartialMerge(ikey.user_key,
+                                               Slice(operands[0]),
+                                               Slice(operands[1]),
+                                               &merge_result,
+                                               logger_.get())) {
+          operands.pop_front();
+          swap(operands.front(), merge_result);
+        } else {
+          // Associative merge returns false ==> stack the operands
+          break;
+        }
+      }
+
    }
  }

  // we either exhausted all internal keys under this user key, or hit
  // a deletion marker.
-  // feed null as the existing value to the merge opexrator, such that
+  // feed null as the existing value to the merge operator, such that
  // client can differentiate this scenario and do things accordingly.
-  user_merge_operator_->Merge(ikey.user_key, nullptr, operand,
-                              &saved_value_, logger_.get());
+  user_merge_operator_->FullMerge(saved_key_, nullptr, operands,
+                                  &saved_value_, logger_.get());
 }

 void DBIter::Prev() {
  assert(valid_);

-  // TODO: support backward iteration
  // Throw an exception now if merge_operator is provided
+  // TODO: support backward iteration
  if (user_merge_operator_) {
    Log(logger_, "Prev not supported yet if merge_operator is provided");
    throw std::logic_error("DBIter::Prev backward iteration not supported"
@@ -321,6 +362,7 @@ void DBIter::Prev() {

 void DBIter::FindPrevUserEntry() {
  assert(direction_ == kReverse);
+  uint64_t num_skipped = 0;

  ValueType value_type = kTypeDeletion;
  if (iter_->Valid()) {
@@ -346,7 +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());
  }

@@ -387,8 +444,8 @@ void DBIter::SeekToFirst() {
 }

 void DBIter::SeekToLast() {
+  // Throw an exception for now if merge_operator is provided
  // TODO: support backward iteration
-  // throw an exception for now if merge_operator is provided
  if (user_merge_operator_) {
    Log(logger_, "SeekToLast not supported yet if merge_operator is provided");
    throw std::logic_error("DBIter::SeekToLast: backward iteration not"
@@ -414,4 +471,4 @@ Iterator* NewDBIterator(
                    internal_iter, sequence);
 }

-}  // namespace leveldb
+}  // namespace rocksdb
@@ -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_
@@ -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:
@@ -30,6 +30,11 @@ class DBStatistics: public Statistics {
    return allTickers_[tickerType].getCount();
  }

+  virtual void setTickerCount(Tickers tickerType, uint64_t count) {
+    assert(tickerType < TICKER_ENUM_MAX);
+    allTickers_[tickerType].setTickerCount(count);
+  }
+
  virtual void recordTick(Tickers tickerType, uint64_t count) {
    assert(tickerType < TICKER_ENUM_MAX);
    allTickers_[tickerType].recordTick(count);
@@ -54,8 +59,6 @@ std::shared_ptr<Statistics> CreateDBStatistics() {
  return std::make_shared<DBStatistics>();
 }

-} // namespace leveldb
+} // namespace rocksdb

 #endif // LEVELDB_STORAGE_DB_DB_STATISTICS_H_
-
-
@@ -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() {

@@ -6,8 +6,9 @@
 #include "db/dbformat.h"
 #include "port/port.h"
 #include "util/coding.h"
+#include "util/perf_context_imp.h"

-namespace leveldb {
+namespace rocksdb {

 static uint64_t PackSequenceAndType(uint64_t seq, ValueType t) {
  assert(seq <= kMaxSequenceNumber);
@@ -53,6 +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));
+  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);
@@ -137,4 +139,4 @@ LookupKey::LookupKey(const Slice& user_key, SequenceNumber s) {
  end_ = dst;
 }

-}  // namespace leveldb
+}  // namespace rocksdb
@@ -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;

@@ -25,7 +25,8 @@ class InternalKey;
 enum ValueType {
  kTypeDeletion = 0x0,
  kTypeValue = 0x1,
-  kTypeMerge = 0x2
+  kTypeMerge = 0x2,
+  kTypeLogData = 0x3
 };
 // kValueTypeForSeek defines the ValueType that should be passed when
 // constructing a ParsedInternalKey object for seeking to a particular
@@ -173,6 +174,15 @@ inline void UpdateInternalKey(char* internal_key,
  EncodeFixed64(seqtype, newval);
 }

+// Get the sequence number from the internal key
+inline uint64_t GetInternalKeySeqno(const Slice& internal_key) {
+  const size_t n = internal_key.size();
+  assert(n >= 8);
+  uint64_t num = DecodeFixed64(internal_key.data() + n - 8);
+  return num >> 8;
+}
+
+
 // A helper class useful for DBImpl::Get()
 class LookupKey {
 public:
@@ -213,6 +223,6 @@ inline LookupKey::~LookupKey() {
  if (start_ != space_) delete[] start_;
 }

-}  // namespace leveldb
+}  // namespace rocksdb

 #endif  // STORAGE_LEVELDB_DB_FORMAT_H_
@@ -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();
 }
@@ -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();
+}
+
@@ -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.
@@ -62,7 +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) {
@@ -130,10 +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;
@@ -214,4 +218,4 @@ Status SetCurrentFile(Env* env, const std::string& dbname,
  return s;
 }

-}  // namespace leveldb
+}  // namespace rocksdb
@@ -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_
@@ -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();
 }
@@ -1,48 +0,0 @@
-// Copyright 2008-present Facebook. All Rights Reserved.
-
-#ifndef STORAGE_LEVELDB_DB_LOG_FILE_H_
-#define STORAGE_LEVELDB_DB_LOG_FILE_H_
-
-namespace leveldb {
-
-enum  WalFileType {
-  kArchivedLogFile = 0,
-  kAliveLogFile = 1
-} ;
-
-class LogFile {
-
- public:
-  uint64_t logNumber;
-  WalFileType type;
-
-  LogFile(uint64_t logNum,WalFileType logType) :
-    logNumber(logNum),
-    type(logType) {}
-
-  LogFile(const LogFile& that) {
-   logNumber = that.logNumber;
-   type = that.type;
-  }
-
-  bool operator < (const LogFile& that) const {
-    return logNumber < that.logNumber;
-  }
-
-  std::string ToString() const {
-    char response[100];
-    const char* typeOfLog;
-    if (type == kAliveLogFile) {
-      typeOfLog = "Alive Log";
-    } else {
-      typeOfLog = "Archived Log";
-    }
-    sprintf(response,
-            "LogNumber  : %ld LogType : %s",
-            logNumber,
-            typeOfLog);
-    return std::string(response);
-  }
-};
-}  //  namespace leveldb
-#endif  // STORAGE_LEVELDB_DB_LOG_FILE_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_
@@ -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
@@ -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_
@@ -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();
 }
@@ -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
@@ -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_
@@ -3,39 +3,43 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.

 #include "db/memtable.h"
+
+#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 {
+namespace rocksdb {

-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);
-}
-
-MemTable::MemTable(const InternalKeyComparator& cmp, int numlevel)
+MemTable::MemTable(const InternalKeyComparator& cmp,
+                   std::shared_ptr<MemTableRepFactory> table_factory,
+                   int numlevel,
+                   const Options& options)
    : comparator_(cmp),
      refs_(0),
-      table_(comparator_, &arena_),
+      arena_impl_(options.arena_block_size),
+      table_(table_factory->CreateMemTableRep(comparator_, &arena_impl_)),
      flush_in_progress_(false),
      flush_completed_(false),
      file_number_(0),
      edit_(numlevel),
      first_seqno_(0),
-      mem_logfile_number_(0) {
-}
+      mem_next_logfile_number_(0),
+      mem_logfile_number_(0) { }

 MemTable::~MemTable() {
  assert(refs_ == 0);
 }

-size_t MemTable::ApproximateMemoryUsage() { return arena_.MemoryUsage(); }
+size_t MemTable::ApproximateMemoryUsage() {
+  return arena_impl_.ApproximateMemoryUsage() +
+    table_->ApproximateMemoryUsage();
+}

 int MemTable::KeyComparator::operator()(const char* aptr, const char* bptr)
    const {
@@ -45,6 +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.
@@ -57,24 +66,30 @@ static const char* EncodeKey(std::string* scratch, const Slice& target) {

 class MemTableIterator: public Iterator {
 public:
-  explicit MemTableIterator(MemTable::Table* table) : iter_(table) { }
+  explicit MemTableIterator(MemTableRep* table)
+    : iter_(table->GetIterator()) { }

-  virtual bool Valid() const { return iter_.Valid(); }
-  virtual void Seek(const Slice& k) { iter_.Seek(EncodeKey(&tmp_, k)); }
-  virtual void SeekToFirst() { iter_.SeekToFirst(); }
-  virtual void SeekToLast() { iter_.SeekToLast(); }
-  virtual void Next() { iter_.Next(); }
-  virtual void Prev() { iter_.Prev(); }
-  virtual Slice key() const { return GetLengthPrefixedSlice(iter_.key()); }
+  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(); }
+  virtual void SeekToLast() { iter_->SeekToLast(); }
+  virtual void Next() { iter_->Next(); }
+  virtual void Prev() { iter_->Prev(); }
+  virtual Slice key() const {
+    return GetLengthPrefixedSlice(iter_->key());
+  }
  virtual Slice value() const {
-    Slice key_slice = GetLengthPrefixedSlice(iter_.key());
+    Slice key_slice = GetLengthPrefixedSlice(iter_->key());
    return GetLengthPrefixedSlice(key_slice.data() + key_slice.size());
  }

  virtual Status status() const { return Status::OK(); }

 private:
-  MemTable::Table::Iterator iter_;
+  std::shared_ptr<MemTableRep::Iterator> iter_;
  std::string tmp_;       // For passing to EncodeKey

  // No copying allowed
@@ -82,8 +97,12 @@ class MemTableIterator: public Iterator {
  void operator=(const MemTableIterator&);
 };

-Iterator* MemTable::NewIterator() {
-  return new MemTableIterator(&table_);
+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,
@@ -100,7 +119,7 @@ void MemTable::Add(SequenceNumber s, ValueType type,
  const size_t encoded_len =
      VarintLength(internal_key_size) + internal_key_size +
      VarintLength(val_size) + val_size;
-  char* buf = arena_.Allocate(encoded_len);
+  char* buf = arena_impl_.Allocate(encoded_len);
  char* p = EncodeVarint32(buf, internal_key_size);
  memcpy(p, key.data(), key_size);
  p += key_size;
@@ -109,7 +128,7 @@ void MemTable::Add(SequenceNumber s, ValueType type,
  p = EncodeVarint32(p, val_size);
  memcpy(p, value.data(), val_size);
  assert((p + val_size) - buf == (unsigned)encoded_len);
-  table_.Insert(buf);
+  table_->Insert(buf);

  // The first sequence number inserted into the memtable
  assert(first_seqno_ == 0 || s > first_seqno_);
@@ -119,22 +138,21 @@ void MemTable::Add(SequenceNumber s, ValueType type,
 }

 bool MemTable::Get(const LookupKey& key, std::string* value, Status* s,
-                  const Options& options) {
+                   std::deque<std::string>* operands, const Options& options) {
  Slice memkey = key.memtable_key();
-  Table::Iterator iter(&table_);
-  iter.Seek(memkey.data());
+  std::shared_ptr<MemTableRep::Iterator> iter(
+    table_->GetIterator(key.user_key()));
+  iter->Seek(memkey.data());

-  bool merge_in_progress = false;
-  std::string operand;
-  if (s->IsMergeInProgress()) {
-    swap(*value, operand);
-    merge_in_progress = true;
-  }
+  // It is the caller's responsibility to allocate/delete operands list
+  assert(operands != nullptr);

-
-  auto merge_operator = options.merge_operator;
+  bool merge_in_progress = s->IsMergeInProgress();
+  auto merge_operator = options.merge_operator.get();
  auto logger = options.info_log;
-  for (; iter.Valid(); iter.Next()) {
+  std::string merge_result;
+
+  for (; iter->Valid(); iter->Next()) {
    // entry format is:
    //    klength  varint32
    //    userkey  char[klength-8]
@@ -144,7 +162,7 @@ bool MemTable::Get(const LookupKey& key, std::string* value, Status* s,
    // Check that it belongs to same user key.  We do not check the
    // sequence number since the Seek() call above should have skipped
    // all entries with overly large sequence numbers.
-    const char* entry = iter.key();
+    const char* entry = iter->key();
    uint32_t key_length;
    const char* key_ptr = GetVarint32Ptr(entry, entry+5, &key_length);
    if (comparator_.comparator.user_comparator()->Compare(
@@ -155,36 +173,56 @@ bool MemTable::Get(const LookupKey& key, std::string* value, Status* s,
      switch (static_cast<ValueType>(tag & 0xff)) {
        case kTypeValue: {
          Slice v = GetLengthPrefixedSlice(key_ptr + key_length);
+          *s = Status::OK();
          if (merge_in_progress) {
-            merge_operator->Merge(key.user_key(), &v, operand,
-                                   value, logger.get());
+            assert(merge_operator);
+            if (!merge_operator->FullMerge(key.user_key(), &v, *operands,
+                                           value, logger.get())) {
+              RecordTick(options.statistics, NUMBER_MERGE_FAILURES);
+              *s = Status::Corruption("Error: Could not perform merge.");
+            }
          } else {
            value->assign(v.data(), v.size());
          }
          return true;
        }
-        case kTypeMerge: {
-          Slice v = GetLengthPrefixedSlice(key_ptr + key_length);
-          if (merge_in_progress) {
-            merge_operator->Merge(key.user_key(), &v, operand,
-                                  value, logger.get());
-            swap(*value, operand);
-          } else {
-            assert(merge_operator);
-            merge_in_progress = true;
-            operand.assign(v.data(), v.size());
-          }
-          break;
-        }
        case kTypeDeletion: {
          if (merge_in_progress) {
-            merge_operator->Merge(key.user_key(), nullptr, operand,
-                                   value, logger.get());
+            assert(merge_operator);
+            *s = Status::OK();
+            if (!merge_operator->FullMerge(key.user_key(), nullptr, *operands,
+                                           value, logger.get())) {
+              RecordTick(options.statistics, NUMBER_MERGE_FAILURES);
+              *s = Status::Corruption("Error: Could not perform merge.");
+            }
          } else {
            *s = Status::NotFound(Slice());
          }
          return true;
        }
+        case kTypeMerge: {
+          Slice v = GetLengthPrefixedSlice(key_ptr + key_length);
+          merge_in_progress = true;
+          operands->push_front(v.ToString());
+          while(operands->size() >= 2) {
+            // Attempt to associative merge. (Returns true if successful)
+            if (merge_operator->PartialMerge(key.user_key(),
+                                             Slice((*operands)[0]),
+                                             Slice((*operands)[1]),
+                                             &merge_result,
+                                             logger.get())) {
+              operands->pop_front();
+              swap(operands->front(), merge_result);
+            } else {
+              // Stack them because user can't associative merge
+              break;
+            }
+          }
+          break;
+        }
+        case kTypeLogData:
+          assert(false);
+          break;
      }
    } else {
      // exit loop if user key does not match
@@ -192,11 +230,12 @@ bool MemTable::Get(const LookupKey& key, std::string* value, Status* s,
    }
  }

+  // No change to value, since we have not yet found a Put/Delete
+
  if (merge_in_progress) {
-    swap(*value, operand);
    *s = Status::MergeInProgress("");
  }
  return false;
 }

-}  // namespace leveldb
+}  // namespace rocksdb
@@ -6,24 +6,35 @@
 #define STORAGE_LEVELDB_DB_MEMTABLE_H_

 #include <string>
-#include "leveldb/db.h"
+#include <memory>
+#include <deque>
 #include "db/dbformat.h"
 #include "db/skiplist.h"
 #include "db/version_set.h"
-#include "util/arena.h"
+#include "rocksdb/db.h"
+#include "rocksdb/memtablerep.h"
+#include "util/arena_impl.h"

-namespace leveldb {
+namespace rocksdb {

-class InternalKeyComparator;
 class Mutex;
 class MemTableIterator;

 class MemTable {
 public:
+  struct KeyComparator : public MemTableRep::KeyComparator {
+    const InternalKeyComparator comparator;
+    explicit KeyComparator(const InternalKeyComparator& c) : comparator(c) { }
+    virtual int operator()(const char* a, const char* b) const;
+  };
+
  // MemTables are reference counted.  The initial reference count
  // is zero and the caller must call Ref() at least once.
-  explicit MemTable(const InternalKeyComparator& comparator,
-                    int numlevel = 7);
+  explicit MemTable(
+    const InternalKeyComparator& comparator,
+    std::shared_ptr<MemTableRepFactory> table_factory,
+    int numlevel = 7,
+    const Options& options = Options());

  // Increase reference count.
  void Ref() { ++refs_; }
@@ -50,7 +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.
@@ -64,11 +79,11 @@ class MemTable {
  // in *status and return true.
  // If memtable contains Merge operation as the most recent entry for a key,
  //   and the merge process does not stop (not reaching a value or delete),
-  //   store the current merged result in value and MergeInProgress in s.
-  //   return false
+  //   prepend the current merge operand to *operands.
+  //   store MergeInProgress in s, and return false.
  // Else, return false.
  bool Get(const LookupKey& key, std::string* value, Status* s,
-          const Options& options);
+           std::deque<std::string>* operands, const Options& options);

  // Returns the edits area that is needed for flushing the memtable
  VersionEdit* GetEdits() { return &edit_; }
@@ -77,6 +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_; }
@@ -85,24 +108,19 @@ 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
-
-  struct KeyComparator {
-    const InternalKeyComparator comparator;
-    explicit KeyComparator(const InternalKeyComparator& c) : comparator(c) { }
-    int operator()(const char* a, const char* b) const;
-  };
  friend class MemTableIterator;
  friend class MemTableBackwardIterator;
  friend class MemTableList;

-  typedef SkipList<const char*, KeyComparator> Table;
-
  KeyComparator comparator_;
  int refs_;
-  Arena arena_;
-  Table table_;
+  ArenaImpl arena_impl_;
+  shared_ptr<MemTableRep> table_;

  // These are used to manage memtable flushes to storage
  bool flush_in_progress_; // started the flush
@@ -117,6 +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
@@ -124,6 +146,6 @@ class MemTable {
  void operator=(const MemTable&);
 };

-}  // namespace leveldb
+}  // namespace rocksdb

 #endif  // STORAGE_LEVELDB_DB_MEMTABLE_H_
@@ -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;
@@ -21,17 +21,15 @@ using std::list;

 // Increase reference count on all underling memtables
 void MemTableList::RefAll() {
-  for (list<MemTable*>::iterator it = memlist_.begin();
-       it != memlist_.end() ; ++it) {
-    (*it)->Ref();
+  for (auto &memtable : memlist_) {
+    memtable->Ref();
  }
 }

 // Drop reference count on all underling memtables
 void MemTableList::UnrefAll() {
-  for (list<MemTable*>::iterator it = memlist_.begin();
-       it != memlist_.end() ; ++it) {
-    (*it)->Unref();
+  for (auto &memtable : memlist_) {
+    memtable->Unref();
  }
 }

@@ -44,7 +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;
  }
@@ -53,8 +52,7 @@ bool MemTableList::IsFlushPending(int min_write_buffer_number_to_merge) {

 // Returns the memtables that need to be flushed.
 void MemTableList::PickMemtablesToFlush(std::vector<MemTable*>* ret) {
-  for (list<MemTable*>::reverse_iterator it = memlist_.rbegin();
-       it != memlist_.rend(); it++) {
+  for (auto it = memlist_.rbegin(); it != memlist_.rend(); it++) {
    MemTable* m = *it;
    if (!m->flush_in_progress_) {
      assert(!m->flush_completed_);
@@ -66,6 +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
@@ -175,6 +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);
@@ -184,20 +184,20 @@ void MemTableList::Add(MemTable* m) {
 // Returns an estimate of the number of bytes of data in use.
 size_t MemTableList::ApproximateMemoryUsage() {
  size_t size = 0;
-  for (list<MemTable*>::iterator it = memlist_.begin();
-       it != memlist_.end(); ++it) {
-    size += (*it)->ApproximateMemoryUsage();
+  for (auto &memtable : memlist_) {
+    size += memtable->ApproximateMemoryUsage();
  }
  return size;
 }

 // Search all the memtables starting from the most recent one.
 // Return the most recent value found, if any.
+// Operands stores the list of merge operations to apply, so far.
 bool MemTableList::Get(const LookupKey& key, std::string* value, Status* s,
-                      const Options& options ) {
-  for (list<MemTable*>::iterator it = memlist_.begin();
-       it != memlist_.end(); ++it) {
-    if ((*it)->Get(key, value, s, options)) {
+                       std::deque<std::string>* operands,
+                       const Options& options) {
+  for (auto &memtable : memlist_) {
+    if (memtable->Get(key, value, s, operands, options)) {
      return true;
    }
  }
@@ -205,10 +205,9 @@ bool MemTableList::Get(const LookupKey& key, std::string* value, Status* s,
 }

 void MemTableList::GetMemTables(std::vector<MemTable*>* output) {
-  for (list<MemTable*>::iterator it = memlist_.begin();
-       it != memlist_.end(); ++it) {
-    output->push_back(*it);
+  for (auto &memtable : memlist_) {
+    output->push_back(memtable);
  }
 }

-}  // namespace leveldb
+}  // namespace rocksdb
@@ -5,13 +5,13 @@

 #include <string>
 #include <list>
-#include "leveldb/db.h"
+#include <deque>
+#include "rocksdb/db.h"
 #include "db/dbformat.h"
 #include "db/skiplist.h"
-#include "util/arena.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() {};
@@ -71,11 +72,14 @@ class MemTableList {
  // Search all the memtables starting from the most recent one.
  // Return the most recent value found, if any.
  bool Get(const LookupKey& key, std::string* value, Status* s,
-           const Options& options);
+           std::deque<std::string>* operands, const Options& options);

  // 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_
@@ -1,32 +1,45 @@
 #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)
-//       key_, value_ are updated to reflect the merge result
+//       keys_, operands_ are updated to reflect the merge result.
+//       keys_ stores the list of keys encountered while merging.
+//       operands_ stores the list of merge operands encountered while merging.
+//       keys_[i] corresponds to operands_[i] for each i.
 void MergeHelper::MergeUntil(Iterator* iter, SequenceNumber stop_before,
-                             bool at_bottom) {
-  // get a copy of the internal key, before it's invalidated by iter->Next()
-  key_.assign(iter->key().data(), iter->key().size());
-  // we need to parse the internal key again as the parsed key is
+                             bool at_bottom, shared_ptr<Statistics> stats) {
+  // Get a copy of the internal key, before it's invalidated by iter->Next()
+  // Also maintain the list of merge operands seen.
+  keys_.clear();
+  operands_.clear();
+  keys_.push_front(iter->key().ToString());
+  operands_.push_front(iter->value().ToString());
+
+  success_ = false;   // Will become true if we hit Put/Delete or bottom
+
+  // We need to parse the internal key again as the parsed key is
  // backed by the internal key!
-  ParsedInternalKey orig_ikey;
  // Assume no internal key corruption as it has been successfully parsed
  // by the caller.
-  // TODO: determine a good alternative of assert (exception?)
-  ParseInternalKey(key_, &orig_ikey);
-  std::string operand(iter->value().data(), iter->value().size());
+  // Invariant: keys_.back() will not change. Hence, orig_ikey is always valid.
+  ParsedInternalKey orig_ikey;
+  ParseInternalKey(keys_.back(), &orig_ikey);

  bool hit_the_next_user_key = false;
  ParsedInternalKey ikey;
+  std::string merge_result;  // Temporary value for merge results
  for (iter->Next(); iter->Valid(); iter->Next()) {
+    assert(operands_.size() >= 1);        // Should be invariants!
+    assert(keys_.size() == operands_.size());
+
    if (!ParseInternalKey(iter->key(), &ikey)) {
      // stop at corrupted key
      if (assert_valid_internal_key_) {
@@ -46,32 +59,58 @@ void MergeHelper::MergeUntil(Iterator* iter, SequenceNumber stop_before,
      break;
    }

+    // At this point we are guaranteed that we need to process this key.
+
    if (kTypeDeletion == ikey.type) {
      // hit a delete
-      //   => merge nullptr with operand
-      //   => change the entry type to kTypeValue
-      // We are done!
-      user_merge_operator_->Merge(ikey.user_key, nullptr, operand,
-                                  &value_, logger_);
-      orig_ikey.type = kTypeValue;
-      UpdateInternalKey(&key_[0], key_.size(),
-                        orig_ikey.sequence, orig_ikey.type);
-      // move iter to the next entry
+      //   => merge nullptr with operands_
+      //   => store result in operands_.back() (and update keys_.back())
+      //   => change the entry type to kTypeValue for keys_.back()
+      // We are done! Return a success if the merge passes.
+      success_ = user_merge_operator_->FullMerge(ikey.user_key, nullptr,
+                                                 operands_, &merge_result,
+                                                 logger_);
+
+      // We store the result in keys_.back() and operands_.back()
+      // if nothing went wrong (i.e.: no operand corruption on disk)
+      if (success_) {
+        std::string& key = keys_.back();  // The original key encountered
+        orig_ikey.type = kTypeValue;
+        UpdateInternalKey(&key[0], key.size(),
+                          orig_ikey.sequence, orig_ikey.type);
+        swap(operands_.back(), merge_result);
+      } else {
+        RecordTick(stats, NUMBER_MERGE_FAILURES);
+      }
+
+      // move iter to the next entry (before doing anything else)
      iter->Next();
      return;
    }

    if (kTypeValue == ikey.type) {
      // hit a put
-      //   => merge the put value with operand
-      //   => change the entry type to kTypeValue
-      // We are done!
+      //   => merge the put value with operands_
+      //   => store result in operands_.back() (and update keys_.back())
+      //   => change the entry type to kTypeValue for keys_.back()
+      // We are done! Success!
      const Slice value = iter->value();
-      user_merge_operator_->Merge(ikey.user_key, &value, Slice(operand),
-                                  &value_, logger_);
-      orig_ikey.type = kTypeValue;
-      UpdateInternalKey(&key_[0], key_.size(),
-                        orig_ikey.sequence, orig_ikey.type);
+      success_ = user_merge_operator_->FullMerge(ikey.user_key, &value,
+                                                 operands_, &merge_result,
+                                                 logger_);
+
+      // We store the result in keys_.back() and operands_.back()
+      // if nothing went wrong (i.e.: no operand corruption on disk)
+      if (success_) {
+        std::string& key = keys_.back();  // The original key encountered
+        orig_ikey.type = kTypeValue;
+        UpdateInternalKey(&key[0], key.size(),
+                          orig_ikey.sequence, orig_ikey.type);
+        swap(operands_.back(), merge_result);
+      } else {
+        RecordTick(stats, NUMBER_MERGE_FAILURES);
+      }
+
      // move iter to the next entry
      iter->Next();
      return;
@@ -79,36 +118,75 @@ void MergeHelper::MergeUntil(Iterator* iter, SequenceNumber stop_before,

    if (kTypeMerge == ikey.type) {
      // hit a merge
-      //   => merge the value with operand.
-      //   => put the result back to operand and continue
-      const Slice value = iter->value();
-      user_merge_operator_->Merge(ikey.user_key, &value, operand,
-                                  &value_, logger_);
-      swap(value_, operand);
+      //   => merge the operand into the front of the operands_ list
+      //   => use the user's associative merge function to determine how.
+      //   => then continue because we haven't yet seen a Put/Delete.
+      assert(!operands_.empty()); // Should have at least one element in it
+
+      keys_.push_front(iter->key().ToString());
+      operands_.push_front(iter->value().ToString());
+      while (operands_.size() >= 2) {
+        // Returns false when the merge_operator can no longer process it
+        if (user_merge_operator_->PartialMerge(ikey.user_key,
+                                               Slice(operands_[0]),
+                                               Slice(operands_[1]),
+                                               &merge_result,
+                                               logger_)) {
+          // Merging of operands (associative merge) was successful.
+          // Replace these frontmost two operands with the merge result
+          keys_.pop_front();
+          operands_.pop_front();
+          swap(operands_.front(), merge_result);
+        } else {
+          // Merging of operands (associative merge) returned false.
+          // The user merge_operator does not know how to merge these operands.
+          // So we just stack them up until we find a Put/Delete or end of key.
+          break;
+        }
+      }
      continue;
    }
  }

-  // We have seen the root history of this key if we are at the
-  // bottem level and exhausted all internal keys of this user key
+  // We are sure we have seen this key's entire history if we are at the
+  // last level and exhausted all internal keys of this user key.
  // NOTE: !iter->Valid() does not necessarily mean we hit the
  // beginning of a user key, as versions of a user key might be
-  // split into multiple files and some files might not be included
-  // in the merge.
-  bool seen_the_beginning = hit_the_next_user_key && at_bottom;
-
-  if (seen_the_beginning) {
+  // split into multiple files (even files on the same level)
+  // and some files might not be included in the compaction/merge.
+  //
+  // There are also cases where we have seen the root of history of this
+  // key without being sure of it. Then, we simply miss the opportunity
+  // to combine the keys. Since VersionSet::SetupOtherInputs() always makes
+  // sure that all merge-operands on the same level get compacted together,
+  // this will simply lead to these merge operands moving to the next level.
+  //
+  // So, we only perform the following logic (to merge all operands together
+  // without a Put/Delete) if we are certain that we have seen the end of key.
+  bool surely_seen_the_beginning = hit_the_next_user_key && at_bottom;
+  if (surely_seen_the_beginning) {
    // do a final merge with nullptr as the existing value and say
    // bye to the merge type (it's now converted to a Put)
    assert(kTypeMerge == orig_ikey.type);
-    user_merge_operator_->Merge(orig_ikey.user_key, nullptr, operand,
-                                &value_, logger_);
-    orig_ikey.type = kTypeValue;
-    UpdateInternalKey(&key_[0], key_.size(),
-                      orig_ikey.sequence, orig_ikey.type);
-  } else {
-    swap(value_, operand);
+    assert(operands_.size() >= 1);
+    assert(operands_.size() == keys_.size());
+    success_ = user_merge_operator_->FullMerge(ikey.user_key, nullptr,
+                                               operands_, &merge_result,
+                                               logger_);
+
+    if (success_) {
+      std::string& key = keys_.back();  // The original key encountered
+      orig_ikey.type = kTypeValue;
+      UpdateInternalKey(&key[0], key.size(),
+                        orig_ikey.sequence, orig_ikey.type);
+
+      // The final value() is always stored in operands_.back()
+      swap(operands_.back(),merge_result);
+    } else {
+      RecordTick(stats, NUMBER_MERGE_FAILURES);
+      // Do nothing if not success_. Leave keys() and operands() as they are.
+    }
  }
 }

-} // namespace leveldb
+} // namespace rocksdb
@@ -2,10 +2,12 @@
 #define MERGE_HELPER_H

 #include "db/dbformat.h"
-#include "leveldb/slice.h"
+#include "rocksdb/slice.h"
+#include "rocksdb/statistics.h"
 #include <string>
+#include <deque>

-namespace leveldb {
+namespace rocksdb {

 class Comparator;
 class Iterator;
@@ -21,7 +23,10 @@ class MergeHelper {
      : user_comparator_(user_comparator),
        user_merge_operator_(user_merge_operator),
        logger_(logger),
-        assert_valid_internal_key_(assert_valid_internal_key) {}
+        assert_valid_internal_key_(assert_valid_internal_key),
+        keys_(),
+        operands_(),
+        success_(false) {}

  // Merge entries until we hit
  //     - a corrupted key
@@ -36,15 +41,43 @@ class MergeHelper {
  // at_bottom:   (IN) true if the iterator covers the bottem level, which means
  //                   we could reach the start of the history of this user key.
  void MergeUntil(Iterator* iter, SequenceNumber stop_before = 0,
-                  bool at_bottom = false);
+                  bool at_bottom = false, shared_ptr<Statistics> stats=nullptr);

  // Query the merge result
-  // These are valid until the next MergeUtil call
-  // IMPORTANT: the key type could change after the MergeUntil call.
-  //            Put/Delete + Merge + ... + Merge => Put
-  //            Merge + ... + Merge => Merge
-  Slice key() { return Slice(key_); }
-  Slice value() { return Slice(value_); }
+  // These are valid until the next MergeUntil call
+  // If the merging was successful:
+  //   - IsSuccess() will be true
+  //   - key() will have the latest sequence number of the merges.
+  //           The type will be Put or Merge. See IMPORTANT 1 note, below.
+  //   - value() will be the result of merging all the operands together
+  //   - The user should ignore keys() and values().
+  //
+  //   IMPORTANT 1: the key type could change after the MergeUntil call.
+  //        Put/Delete + Merge + ... + Merge => Put
+  //        Merge + ... + Merge => Merge
+  //
+  // If the merge operator is not associative, and if a Put/Delete is not found
+  // then the merging will be unsuccessful. In this case:
+  //   - IsSuccess() will be false
+  //   - keys() contains the list of internal keys seen in order of iteration.
+  //   - values() contains the list of values (merges) seen in the same order.
+  //              values() is parallel to keys() so that the first entry in
+  //              keys() is the key associated with the first entry in values()
+  //              and so on. These lists will be the same length.
+  //              All of these pairs will be merges over the same user key.
+  //              See IMPORTANT 2 note below.
+  //   - The user should ignore key() and value().
+  //
+  //   IMPORTANT 2: The entries were traversed in order from BACK to FRONT.
+  //                So keys().back() was the first key seen by iterator.
+  // TODO: Re-style this comment to be like the first one
+  bool IsSuccess() { return success_; }
+  Slice key() { assert(success_); return Slice(keys_.back()); }
+  Slice value() { assert(success_); return Slice(operands_.back()); }
+  const std::deque<std::string>& keys() { assert(!success_); return keys_; }
+  const std::deque<std::string>& values() {
+    assert(!success_); return operands_;
+  }

 private:
  const Comparator* user_comparator_;
@@ -55,10 +88,11 @@ class MergeHelper {

  // the scratch area that holds the result of MergeUntil
  // valid up to the next MergeUntil call
-  std::string key_;
-  std::string value_;
+  std::deque<std::string> keys_;    // Keeps track of the sequence of keys seen
+  std::deque<std::string> operands_;  // Parallel with keys_; stores the values
+  bool success_;
 };

-} // namespace leveldb
+} // namespace rocksdb

 #endif
@@ -0,0 +1,51 @@
+/**
+ * Back-end implementation details specific to the Merge Operator.
+ *
+ * @author Deon Nicholas (dnicholas@fb.com)
+ * Copyright 2013 Facebook
+ */
+
+#include "rocksdb/merge_operator.h"
+
+namespace rocksdb {
+
+// Given a "real" merge from the library, call the user's
+// associative merge function one-by-one on each of the operands.
+// NOTE: It is assumed that the client's merge-operator will handle any errors.
+bool AssociativeMergeOperator::FullMerge(
+    const Slice& key,
+    const Slice* existing_value,
+    const std::deque<std::string>& operand_list,
+    std::string* new_value,
+    Logger* logger) const {
+
+  // Simply loop through the operands
+  Slice temp_existing;
+  std::string temp_value;
+  for (const auto& operand : operand_list) {
+    Slice value(operand);
+    if (!Merge(key, existing_value, value, &temp_value, logger)) {
+      return false;
+    }
+    swap(temp_value, *new_value);
+    temp_existing = Slice(*new_value);
+    existing_value = &temp_existing;
+  }
+
+  // The result will be in *new_value. All merges succeeded.
+  return true;
+}
+
+// Call the user defined simple merge on the operands;
+// NOTE: It is assumed that the client's merge-operator will handle any errors.
+bool AssociativeMergeOperator::PartialMerge(
+    const Slice& key,
+    const Slice& left_operand,
+    const Slice& right_operand,
+    std::string* new_value,
+    Logger* logger) const {
+
+  return Merge(key, &left_operand, right_operand, new_value, logger);
+}
+
+} // namespace rocksdb
@@ -2,25 +2,36 @@
 #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"
+#include "util/testharness.h"
+#include "utilities/utility_db.h"

 using namespace std;
-using namespace leveldb;
+using namespace rocksdb;

-auto mergeOperator = MergeOperators::CreateUInt64AddOperator();

-std::shared_ptr<DB> OpenDb() {
+std::shared_ptr<DB> OpenDb(const string& dbname, const bool ttl = false) {
  DB* db;
+  StackableDB* sdb;
  Options options;
  options.create_if_missing = true;
-  options.merge_operator = mergeOperator.get();
-  Status s = DB::Open(options, "/tmp/testdb", &db);
+  options.merge_operator = MergeOperators::CreateUInt64AddOperator();
+  Status s;
+  DestroyDB(dbname, Options());
+  if (ttl) {
+    cout << "Opening database with TTL\n";
+    s = UtilityDB::OpenTtlDB(options, dbname, &sdb);
+    db = sdb;
+  } else {
+    s = DB::Open(options, dbname, &db);
+  }
  if (!s.ok()) {
    cerr << s.ToString() << endl;
    assert(false);
@@ -45,7 +56,7 @@ class Counters {
  uint64_t default_;

 public:
-  Counters(std::shared_ptr<DB> db, uint64_t defaultCount = 0)
+  explicit Counters(std::shared_ptr<DB> db, uint64_t defaultCount = 0)
      : db_(db),
        put_option_(),
        get_option_(),
@@ -143,7 +154,7 @@ class MergeBasedCounters : public Counters {
  WriteOptions merge_option_; // for merge

 public:
-  MergeBasedCounters(std::shared_ptr<DB> db, uint64_t defaultCount = 0)
+  explicit MergeBasedCounters(std::shared_ptr<DB> db, uint64_t defaultCount = 0)
      : Counters(db, defaultCount),
        merge_option_() {
  }
@@ -227,9 +238,8 @@ void testCounters(Counters& counters, DB* db, bool test_compaction) {
  }
 }

-int main(int argc, char *argv[]) {
-
-  auto db = OpenDb();
+void runTest(int argc, const string& dbname, const bool use_ttl = false) {
+  auto db = OpenDb(dbname, use_ttl);

  {
    cout << "Test read-modify-write counters... \n";
@@ -249,5 +259,12 @@ int main(int argc, char *argv[]) {
    testCounters(counters, db.get(), compact);
  }

+  DestroyDB(dbname, Options());
+}
+
+int main(int argc, char *argv[]) {
+  //TODO: Make this test like a general rocksdb unit-test
+  runTest(argc, test::TmpDir() + "/merge_testdb");
+  runTest(argc, test::TmpDir() + "/merge_testdbttl", true); // Run test on TTL database
  return 0;
 }
@@ -0,0 +1,320 @@
+#include <algorithm>
+#include <iostream>
+#include <vector>
+#include "/usr/include/valgrind/callgrind.h"
+
+#include "rocksdb/db.h"
+#include "rocksdb/perf_context.h"
+#include "util/histogram.h"
+#include "util/stop_watch.h"
+#include "util/testharness.h"
+
+
+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 = 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 = 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);
+}
+
+class PerfContextTest { };
+
+TEST(PerfContextTest, SeekIntoDeletion) {
+  DestroyDB(kDbName, Options());
+  auto db = OpenDb();
+  WriteOptions write_options;
+  ReadOptions read_options;
+
+  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();
+
+  WriteOptions write_options;
+  ReadOptions read_options;
+
+  HistogramImpl hist_put;
+  HistogramImpl hist_get;
+
+  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);
+
+    perf_context.Reset();
+    db->Put(write_options, key, value);
+    hist_put.Add(perf_context.user_key_comparison_count);
+
+    perf_context.Reset();
+    db->Get(read_options, key, &value);
+    hist_get.Add(perf_context.user_key_comparison_count);
+  }
+
+  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) {
+
+  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;
+    }
+
+  }
+
+  std::cout << kDbName << "\n";
+
+  rocksdb::test::RunAllTests();
+  return 0;
+}
@@ -0,0 +1,76 @@
+// 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.
+//
+// Wrap an underlying iterator, but exclude any results not starting
+// with a given prefix.  Seeking to keys not beginning with the prefix
+// is invalid, and SeekToLast is not implemented (that would be
+// non-trivial), but otherwise this iterator will behave just like the
+// underlying iterator would if there happened to be no non-matching
+// keys in the dataset.
+
+#ifndef STORAGE_LEVELDB_DB_PREFIX_FILTER_ITERATOR_H_
+#define STORAGE_LEVELDB_DB_PREFIX_FILTER_ITERATOR_H_
+
+#include "rocksdb/iterator.h"
+
+namespace rocksdb {
+
+class PrefixFilterIterator : public Iterator {
+ private:
+  Iterator* iter_;
+  const Slice &prefix_;
+  const SliceTransform *prefix_extractor_;
+  Status status_;
+
+ public:
+  PrefixFilterIterator(Iterator* iter,
+                       const Slice &prefix,
+                       const SliceTransform* prefix_extractor)
+                             : iter_(iter), prefix_(prefix),
+                               prefix_extractor_(prefix_extractor),
+                               status_(Status::OK()) {
+    if (prefix_extractor == nullptr) {
+      status_ = Status::InvalidArgument("A prefix filter may not be used "
+                                        "unless a function is also defined "
+                                        "for extracting prefixes");
+    } else if (!prefix_extractor_->InRange(prefix)) {
+      status_ = Status::InvalidArgument("Must provide a slice for prefix which"
+                                        "is a prefix for some key");
+    }
+  }
+  ~PrefixFilterIterator() {
+    delete iter_;
+  }
+  Slice key() const { return iter_->key(); }
+  Slice value() const { return iter_->value(); }
+  Status status() const {
+    if (!status_.ok()) {
+      return status_;
+    }
+    return iter_->status();
+  }
+  void Next() { iter_->Next(); }
+  void Prev() { iter_->Prev(); }
+  void Seek(const Slice& k) {
+    if (prefix_extractor_->Transform(k) == prefix_) {
+      iter_->Seek(k);
+    } else {
+      status_ = Status::InvalidArgument("Seek must begin with target prefix");
+    }
+  }
+  void SeekToFirst() {
+    Seek(prefix_);
+  }
+  void SeekToLast() {
+    status_ = Status::NotSupported("SeekToLast is incompatible with prefixes");
+  }
+  bool Valid() const {
+    return (status_.ok() && iter_->Valid() &&
+            prefix_extractor_->Transform(iter_->key()) == prefix_);
+  }
+};
+
+}  // namespace rocksdb
+
+#endif
@@ -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 {

@@ -88,6 +88,7 @@ class Repairer {
 private:
  struct TableInfo {
    FileMetaData meta;
+    SequenceNumber min_sequence;
    SequenceNumber max_sequence;
  };

@@ -191,7 +192,8 @@ class Repairer {
    std::string scratch;
    Slice record;
    WriteBatch batch;
-    MemTable* mem = new MemTable(icmp_, options_.num_levels);
+    MemTable* mem = new MemTable(icmp_, options_.memtable_factory,
+      options_.num_levels);
    mem->Ref();
    int counter = 0;
    while (reader.ReadRecord(&record, &scratch)) {
@@ -263,6 +265,7 @@ class Repairer {
          ReadOptions(), storage_options_, t->meta.number, t->meta.file_size);
      bool empty = true;
      ParsedInternalKey parsed;
+      t->min_sequence = 0;
      t->max_sequence = 0;
      for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
        Slice key = iter->key();
@@ -279,6 +282,9 @@ class Repairer {
          t->meta.smallest.DecodeFrom(key);
        }
        t->meta.largest.DecodeFrom(key);
+        if (parsed.sequence < t->min_sequence) {
+          t->min_sequence = parsed.sequence;
+        }
        if (parsed.sequence > t->max_sequence) {
          t->max_sequence = parsed.sequence;
        }
@@ -319,7 +325,8 @@ class Repairer {
      // TODO(opt): separate out into multiple levels
      const TableInfo& t = tables_[i];
      edit_->AddFile(0, t.meta.number, t.meta.file_size,
-                    t.meta.smallest, t.meta.largest);
+                    t.meta.smallest, t.meta.largest,
+                    t.min_sequence, t.max_sequence);
    }

    //fprintf(stderr, "NewDescriptor:\n%s\n", edit_.DebugString().c_str());
@@ -376,4 +383,4 @@ Status RepairDB(const std::string& dbname, const Options& options) {
  return repairer.Run();
 }

-}  // namespace leveldb
+}  // namespace rocksdb
@@ -31,12 +31,9 @@
 #include <assert.h>
 #include <stdlib.h>
 #include "port/port.h"
-#include "util/arena.h"
 #include "util/random.h"

-namespace leveldb {
-
-class Arena;
+namespace rocksdb {

 template<typename Key, class Comparator>
 class SkipList {
@@ -110,6 +107,7 @@ class SkipList {

  // Used for optimizing sequential insert patterns
  Node* prev_[kMaxHeight];
+  int   prev_height_;

  inline int GetMaxHeight() const {
    return static_cast<int>(
@@ -270,6 +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;
    }
  }
@@ -278,6 +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;
@@ -340,6 +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);
@@ -381,6 +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>
@@ -393,6 +400,6 @@ bool SkipList<Key,Comparator>::Contains(const Key& key) const {
  }
 }

-}  // namespace leveldb
+}  // namespace rocksdb

 #endif // STORAGE_LEVELDB_DB_SKIPLIST_H_
@@ -4,13 +4,13 @@

 #include "db/skiplist.h"
 #include <set>
-#include "leveldb/env.h"
-#include "util/arena.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;

@@ -29,9 +29,9 @@ struct TestComparator {
 class SkipTest { };

 TEST(SkipTest, Empty) {
-  Arena arena;
+  ArenaImpl arena_impl;
  TestComparator cmp;
-  SkipList<Key, TestComparator> list(cmp, &arena);
+  SkipList<Key, TestComparator> list(cmp, &arena_impl);
  ASSERT_TRUE(!list.Contains(10));

  SkipList<Key, TestComparator>::Iterator iter(&list);
@@ -49,9 +49,9 @@ TEST(SkipTest, InsertAndLookup) {
  const int R = 5000;
  Random rnd(1000);
  std::set<Key> keys;
-  Arena arena;
+  ArenaImpl arena_impl;
  TestComparator cmp;
-  SkipList<Key, TestComparator> list(cmp, &arena);
+  SkipList<Key, TestComparator> list(cmp, &arena_impl);
  for (int i = 0; i < N; i++) {
    Key key = rnd.Next() % R;
    if (keys.insert(key).second) {
@@ -204,14 +204,14 @@ class ConcurrentTest {
  // Current state of the test
  State current_;

-  Arena arena_;
+  ArenaImpl arena_impl_;

  // SkipList is not protected by mu_.  We just use a single writer
  // thread to modify it.
  SkipList<Key, TestComparator> list_;

 public:
-  ConcurrentTest() : list_(TestComparator(), &arena_) { }
+  ConcurrentTest() : list_(TestComparator(), &arena_impl_) { }

  // REQUIRES: External synchronization
  void WriteStep(Random* rnd) {
@@ -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();
 }
@@ -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_
@@ -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);
@@ -39,15 +39,19 @@ TableCache::~TableCache() {

 Status TableCache::FindTable(const EnvOptions& toptions,
                             uint64_t file_number, uint64_t file_size,
-                             Cache::Handle** handle, bool* tableIO) {
+                             Cache::Handle** handle, bool* table_io,
+                             const bool no_io) {
  Status s;
  char buf[sizeof(file_number)];
  EncodeFixed64(buf, file_number);
  Slice key(buf, sizeof(buf));
  *handle = cache_->Lookup(key);
  if (*handle == nullptr) {
-    if (tableIO != nullptr) {
-      *tableIO = true;    // we had to do IO from storage
+    if (no_io) { // Dont do IO and return a not-found status
+      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
    }
    std::string fname = TableFileName(dbname_, file_number);
    unique_ptr<RandomAccessFile> file;
@@ -86,7 +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);
  }
@@ -112,23 +117,47 @@ Status TableCache::Get(const ReadOptions& options,
                       const Slice& k,
                       void* arg,
                       bool (*saver)(void*, const Slice&, const Slice&, bool),
-                       bool* tableIO) {
+                       bool* table_io,
+                       void (*mark_key_may_exist)(void*)) {
  Cache::Handle* handle = nullptr;
  Status s = FindTable(storage_options_, file_number, file_size,
-                       &handle, tableIO);
+                       &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);
+    s = t->InternalGet(options, k, arg, saver, mark_key_may_exist);
    cache_->Release(handle);
+  } 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();
  }
  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
@@ -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;

@@ -48,7 +48,14 @@ class TableCache {
             const Slice& k,
             void* arg,
             bool (*handle_result)(void*, const Slice&, const Slice&, bool),
-             bool* tableIO);
+             bool* table_io,
+             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);
@@ -60,11 +67,11 @@ class TableCache {
  const EnvOptions& storage_options_;
  std::shared_ptr<Cache> cache_;

-  Status FindTable(const EnvOptions& toptions,
-                   uint64_t file_number, uint64_t file_size, Cache::Handle**,
-                   bool* tableIO = nullptr);
+  Status FindTable(const EnvOptions& toptions, uint64_t file_number,
+                   uint64_t file_size, Cache::Handle**, bool* table_io=nullptr,
+                   const bool no_io = false);
 };

-}  // namespace leveldb
+}  // namespace rocksdb

 #endif  // STORAGE_LEVELDB_DB_TABLE_CACHE_H_
@@ -1,46 +1,45 @@
-#include "db/transaction_log_iterator_impl.h"
+#include "db/transaction_log_impl.h"
 #include "db/write_batch_internal.h"
-#include "db/filename.h"

-namespace leveldb {
+namespace rocksdb {

 TransactionLogIteratorImpl::TransactionLogIteratorImpl(
                           const std::string& dbname,
                           const Options* options,
                           const EnvOptions& soptions,
-                           SequenceNumber& seq,
-                           std::unique_ptr<std::vector<LogFile>> files,
+                           const SequenceNumber seq,
+                           std::unique_ptr<VectorLogPtr> files,
                           SequenceNumber const * const lastFlushedSequence) :
-  dbname_(dbname),
-  options_(options),
-  soptions_(soptions),
-  startingSequenceNumber_(seq),
-  files_(std::move(files)),
-  started_(false),
-  isValid_(false),
-  currentFileIndex_(0),
-  lastFlushedSequence_(lastFlushedSequence) {
+    dbname_(dbname),
+    options_(options),
+    soptions_(soptions),
+    startingSequenceNumber_(seq),
+    files_(std::move(files)),
+    started_(false),
+    isValid_(false),
+    currentFileIndex_(0),
+    lastFlushedSequence_(lastFlushedSequence) {
+  assert(startingSequenceNumber_ <= *lastFlushedSequence_);
  assert(files_.get() != nullptr);
-  assert(lastFlushedSequence_);

  reporter_.env = options_->env;
  reporter_.info_log = options_->info_log.get();
 }

 Status TransactionLogIteratorImpl::OpenLogFile(
-  const LogFile& logFile,
+  const LogFile* logFile,
  unique_ptr<SequentialFile>* file) {
  Env* env = options_->env;
-  if (logFile.type == kArchivedLogFile) {
-    std::string fname = ArchivedLogFileName(dbname_, logFile.logNumber);
+  if (logFile->Type() == kArchivedLogFile) {
+    std::string fname = ArchivedLogFileName(dbname_, logFile->LogNumber());
    return env->NewSequentialFile(fname, file, soptions_);
  } else {
-    std::string fname = LogFileName(dbname_, logFile.logNumber);
+    std::string fname = LogFileName(dbname_, logFile->LogNumber());
    Status status = env->NewSequentialFile(fname, file, soptions_);
    if (!status.ok()) {
      //  If cannot open file in DB directory.
      //  Try the archive dir, as it could have moved in the meanwhile.
-      fname = ArchivedLogFileName(dbname_, logFile.logNumber);
+      fname = ArchivedLogFileName(dbname_, logFile->LogNumber());
      status = env->NewSequentialFile(fname, file, soptions_);
      if (!status.ok()) {
        return Status::IOError(" Requested file not present in the dir");
@@ -67,7 +66,7 @@ bool TransactionLogIteratorImpl::Valid() {
 }

 void TransactionLogIteratorImpl::Next() {
-  LogFile currentLogFile = files_.get()->at(currentFileIndex_);
+  LogFile* currentLogFile = files_.get()->at(currentFileIndex_).get();

 //  First seek to the given seqNo. in the current file.
  std::string scratch;
@@ -129,7 +128,7 @@ void TransactionLogIteratorImpl::Next() {
    if (openNextFile) {
      if (currentFileIndex_ < files_.get()->size() - 1) {
        ++currentFileIndex_;
-        Status status = OpenLogReader(files_.get()->at(currentFileIndex_));
+        Status status =OpenLogReader(files_.get()->at(currentFileIndex_).get());
        if (!status.ok()) {
          isValid_ = false;
          currentStatus_ = status;
@@ -157,7 +156,7 @@ void TransactionLogIteratorImpl::UpdateCurrentWriteBatch(const Slice& record) {
  currentStatus_ = Status::OK();
 }

-Status TransactionLogIteratorImpl::OpenLogReader(const LogFile& logFile) {
+Status TransactionLogIteratorImpl::OpenLogReader(const LogFile* logFile) {
  unique_ptr<SequentialFile> file;
  Status status = OpenLogFile(logFile, &file);
  if (!status.ok()) {
@@ -169,4 +168,4 @@ Status TransactionLogIteratorImpl::OpenLogReader(const LogFile& logFile) {
  );
  return Status::OK();
 }
-}  //  namespace leveldb
+}  //  namespace rocksdb
@@ -4,14 +4,14 @@

 #include <vector>

-#include "leveldb/env.h"
-#include "leveldb/options.h"
-#include "leveldb/types.h"
-#include "leveldb/transaction_log_iterator.h"
-#include "db/log_file.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;
@@ -21,13 +21,50 @@ struct LogReporter : public log::Reader::Reporter {
  }
 };

+class LogFileImpl : public LogFile {
+ public:
+  LogFileImpl(uint64_t logNum, WalFileType logType, SequenceNumber startSeq,
+              uint64_t sizeBytes) :
+    logNumber_(logNum),
+    type_(logType),
+    startSequence_(startSeq),
+    sizeFileBytes_(sizeBytes) {
+  }
+
+  std::string PathName() const {
+    if (type_ == kArchivedLogFile) {
+      return ArchivedLogFileName("", logNumber_);
+    }
+    return LogFileName("", logNumber_);
+  }
+
+  uint64_t LogNumber() const { return logNumber_; }
+
+  WalFileType Type() const { return type_; }
+
+  SequenceNumber StartSequence() const { return startSequence_; }
+
+  uint64_t SizeFileBytes() const { return sizeFileBytes_; }
+
+  bool operator < (const LogFile& that) const {
+    return LogNumber() < that.LogNumber();
+  }
+
+ private:
+  uint64_t logNumber_;
+  WalFileType type_;
+  SequenceNumber startSequence_;
+  uint64_t sizeFileBytes_;
+
+};
+
 class TransactionLogIteratorImpl : public TransactionLogIterator {
 public:
  TransactionLogIteratorImpl(const std::string& dbname,
                             const Options* options,
                             const EnvOptions& soptions,
-                             SequenceNumber& seqNum,
-                             std::unique_ptr<std::vector<LogFile>> files,
+                             const SequenceNumber seqNum,
+                             std::unique_ptr<VectorLogPtr> files,
                             SequenceNumber const * const lastFlushedSequence);

  virtual bool Valid();
@@ -42,25 +79,25 @@ class TransactionLogIteratorImpl : public TransactionLogIterator {
  const std::string& dbname_;
  const Options* options_;
  const EnvOptions& soptions_;
-  const uint64_t startingSequenceNumber_;
-  std::unique_ptr<std::vector<LogFile>> files_;
+  const SequenceNumber startingSequenceNumber_;
+  std::unique_ptr<VectorLogPtr> files_;
  bool started_;
  bool isValid_;  // not valid when it starts of.
  Status currentStatus_;
  size_t currentFileIndex_;
  std::unique_ptr<WriteBatch> currentBatch_;
  unique_ptr<log::Reader> currentLogReader_;
-  Status OpenLogFile(const LogFile& logFile, unique_ptr<SequentialFile>* file);
+  Status OpenLogFile(const LogFile* logFile, unique_ptr<SequentialFile>* file);
  LogReporter reporter_;
  SequenceNumber const * const lastFlushedSequence_;
  // represents the sequence number being read currently.
  SequenceNumber currentSequence_;

  void UpdateCurrentWriteBatch(const Slice& record);
-  Status OpenLogReader(const LogFile& file);
+  Status OpenLogReader(const LogFile* file);
 };



-}  //  namespace leveldb
+}  //  namespace rocksdb
 #endif  //  STORAGE_LEVELDB_INCLUDE_WRITES_ITERATOR_IMPL_H_
@@ -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.
@@ -20,7 +20,10 @@ enum Tag {
  kDeletedFile          = 6,
  kNewFile              = 7,
  // 8 was used for large value refs
-  kPrevLogNumber        = 9
+  kPrevLogNumber        = 9,
+
+  // these are new formats divergent from open source leveldb
+  kNewFile2             = 100  // store smallest & largest seqno
 };

 void VersionEdit::Clear() {
@@ -76,12 +79,14 @@ void VersionEdit::EncodeTo(std::string* dst) const {

  for (size_t i = 0; i < new_files_.size(); i++) {
    const FileMetaData& f = new_files_[i].second;
-    PutVarint32(dst, kNewFile);
+    PutVarint32(dst, kNewFile2);
    PutVarint32(dst, new_files_[i].first);  // level
    PutVarint64(dst, f.number);
    PutVarint64(dst, f.file_size);
    PutLengthPrefixedSlice(dst, f.smallest.Encode());
    PutLengthPrefixedSlice(dst, f.largest.Encode());
+    PutVarint64(dst, f.smallest_seqno);
+    PutVarint64(dst, f.largest_seqno);
  }
 }

@@ -201,6 +206,22 @@ Status VersionEdit::DecodeFrom(const Slice& src) {
        }
        break;

+      case kNewFile2:
+        if (GetLevel(&input, &level, &msg) &&
+            GetVarint64(&input, &f.number) &&
+            GetVarint64(&input, &f.file_size) &&
+            GetInternalKey(&input, &f.smallest) &&
+            GetInternalKey(&input, &f.largest) &&
+            GetVarint64(&input, &f.smallest_seqno) &&
+            GetVarint64(&input, &f.largest_seqno) ) {
+          new_files_.push_back(std::make_pair(level, f));
+        } else {
+          if (!msg) {
+            msg = "new-file2 entry";
+          }
+        }
+        break;
+
      default:
        msg = "unknown tag";
        break;
@@ -218,7 +239,7 @@ Status VersionEdit::DecodeFrom(const Slice& src) {
  return result;
 }

-std::string VersionEdit::DebugString() const {
+std::string VersionEdit::DebugString(bool hex_key) const {
  std::string r;
  r.append("VersionEdit {");
  if (has_comparator_) {
@@ -245,7 +266,7 @@ std::string VersionEdit::DebugString() const {
    r.append("\n  CompactPointer: ");
    AppendNumberTo(&r, compact_pointers_[i].first);
    r.append(" ");
-    r.append(compact_pointers_[i].second.DebugString());
+    r.append(compact_pointers_[i].second.DebugString(hex_key));
  }
  for (DeletedFileSet::const_iterator iter = deleted_files_.begin();
       iter != deleted_files_.end();
@@ -264,12 +285,12 @@ std::string VersionEdit::DebugString() const {
    r.append(" ");
    AppendNumberTo(&r, f.file_size);
    r.append(" ");
-    r.append(f.smallest.DebugString());
+    r.append(f.smallest.DebugString(hex_key));
    r.append(" .. ");
-    r.append(f.largest.DebugString());
+    r.append(f.largest.DebugString(hex_key));
  }
  r.append("\n}\n");
  return r;
 }

-}  // namespace leveldb
+}  // namespace rocksdb
@@ -10,7 +10,7 @@
 #include <vector>
 #include "db/dbformat.h"

-namespace leveldb {
+namespace rocksdb {

 class VersionSet;

@@ -22,6 +22,8 @@ struct FileMetaData {
  InternalKey smallest;       // Smallest internal key served by table
  InternalKey largest;        // Largest internal key served by table
  bool being_compacted;       // Is this file undergoing compaction?
+  SequenceNumber smallest_seqno;// The smallest seqno in this file
+  SequenceNumber largest_seqno; // The largest seqno in this file

  FileMetaData() : refs(0), allowed_seeks(1 << 30), file_size(0),
                   being_compacted(false) { }
@@ -67,12 +69,17 @@ class VersionEdit {
  void AddFile(int level, uint64_t file,
               uint64_t file_size,
               const InternalKey& smallest,
-               const InternalKey& largest) {
+               const InternalKey& largest,
+               const SequenceNumber& smallest_seqno,
+               const SequenceNumber& largest_seqno) {
    FileMetaData f;
    f.number = file;
    f.file_size = file_size;
    f.smallest = smallest;
    f.largest = largest;
+    f.smallest_seqno = smallest_seqno;
+    f.largest_seqno = largest_seqno;
+    assert(smallest_seqno <= largest_seqno);
    new_files_.push_back(std::make_pair(level, f));
  }

@@ -89,7 +96,7 @@ class VersionEdit {
  void EncodeTo(std::string* dst) const;
  Status DecodeFrom(const Slice& src);

-  std::string DebugString() const;
+  std::string DebugString(bool hex_key = false) const;

 private:
  friend class VersionSet;
@@ -115,6 +122,6 @@ class VersionEdit {
  std::vector< std::pair<int, FileMetaData> > new_files_;
 };

-}  // namespace leveldb
+}  // namespace rocksdb

 #endif  // STORAGE_LEVELDB_DB_VERSION_EDIT_H_
@@ -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;
@@ -27,7 +27,9 @@ TEST(VersionEditTest, EncodeDecode) {
    TestEncodeDecode(edit);
    edit.AddFile(3, kBig + 300 + i, kBig + 400 + i,
                 InternalKey("foo", kBig + 500 + i, kTypeValue),
-                 InternalKey("zoo", kBig + 600 + i, kTypeDeletion));
+                 InternalKey("zoo", kBig + 600 + i, kTypeDeletion),
+                 kBig + 500 + i,
+                 kBig + 600 + i);
    edit.DeleteFile(4, kBig + 700 + i);
    edit.SetCompactPointer(i, InternalKey("x", kBig + 900 + i, kTypeValue));
  }
@@ -39,8 +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();
 }
@@ -25,7 +25,7 @@
 #include "port/port.h"
 #include "db/table_cache.h"

-namespace leveldb {
+namespace rocksdb {

 namespace log { class Writer; }

@@ -68,13 +68,16 @@ class Version {

  // Lookup the value for key.  If found, store it in *val and
  // return OK.  Else return a non-OK status.  Fills *stats.
+  // Uses *operands to store merge_operator operations to apply later
  // REQUIRES: lock is not held
  struct GetStats {
    FileMetaData* seek_file;
    int seek_file_level;
  };
  void Get(const ReadOptions&, const LookupKey& key, std::string* val,
-           Status* status, GetStats* stats,  const Options& db_option);
+           Status* status, std::deque<std::string>* operands, GetStats* stats,
+           const Options& db_option,
+           bool* value_found = nullptr);

  // Adds "stats" into the current state.  Returns true if a new
  // compaction may need to be triggered, false otherwise.
@@ -117,6 +120,14 @@ class Version {
                      const Slice* smallest_user_key,
                      const Slice* largest_user_key);

+  // Returns true iff the first or last file in inputs contains
+  // an overlapping user key to the file "just outside" of it (i.e.
+  // just after the last file, or just before the first file)
+  // REQUIRES: "*inputs" is a sorted list of non-overlapping files
+  bool HasOverlappingUserKey(const std::vector<FileMetaData*>* inputs,
+                             int level);
+
+
  // Return the level at which we should place a new memtable compaction
  // result that covers the range [smallest_user_key,largest_user_key].
  int PickLevelForMemTableOutput(const Slice& smallest_user_key,
@@ -135,11 +146,14 @@ class Version {
 private:
  friend class Compaction;
  friend class VersionSet;
+  friend class DBImpl;

  class LevelFileNumIterator;
  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
@@ -340,6 +354,9 @@ class VersionSet {
  struct LevelSummaryStorage {
    char buffer[100];
  };
+  struct FileSummaryStorage {
+    char buffer[1000];
+  };
  const char* LevelSummary(LevelSummaryStorage* scratch) const;

  // printf contents (for debugging)
@@ -350,6 +367,10 @@ class VersionSet {
  // of files per level.  Uses *scratch as backing store.
  const char* LevelDataSizeSummary(LevelSummaryStorage* scratch) const;

+  // Return a human-readable short (single-line) summary of files
+  // in a specified level.  Uses *scratch as backing store.
+  const char* LevelFileSummary(FileSummaryStorage* scratch, int level) const;
+
  // Return the size of the current manifest file
  const uint64_t ManifestFileSize() { return current_->offset_manifest_file_; }

@@ -359,6 +380,16 @@ class VersionSet {
  // function will return nullptr.
  Compaction* PickCompactionBySize(int level, double score);

+  // 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);

@@ -383,6 +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;
@@ -403,6 +440,8 @@ class VersionSet {
                 InternalKey* smallest,
                 InternalKey* largest);

+  void ExpandWhileOverlapping(Compaction* c);
+
  void SetupOtherInputs(Compaction* c);

  // Save current contents to *log
@@ -412,9 +451,9 @@ class VersionSet {

  bool ManifestContains(const std::string& record) const;

-  int64_t ExpandedCompactionByteSizeLimit(int level);
+  uint64_t ExpandedCompactionByteSizeLimit(int level);

-  int64_t MaxGrandParentOverlapBytes(int level);
+  uint64_t MaxGrandParentOverlapBytes(int level);

  Env* const env_;
  const std::string dbname_;
@@ -489,9 +528,12 @@ class Compaction {
  ~Compaction();

  // Return the level that is being compacted.  Inputs from "level"
-  // and "level+1" will be merged to produce a set of "level+1" files.
+  // will be merged.
  int level() const { return level_; }

+  // Outputs will go to this level
+  int output_level() const { return out_level_; }
+
  // Return the object that holds the edits to the descriptor done
  // by this compaction.
  VersionEdit* edit() { return edit_; }
@@ -530,17 +572,21 @@ class Compaction {
  // Return the score that was used to pick this compaction run.
  double score() const { return score_; }

+  // Is this compaction creating a file in the bottom most level?
+  bool BottomMostLevel() { return bottommost_level_; }
+
 private:
  friend class Version;
  friend class VersionSet;

-  explicit Compaction(int level, uint64_t target_file_size,
+  explicit Compaction(int level, int out_level, uint64_t target_file_size,
    uint64_t max_grandparent_overlap_bytes, int number_levels,
    bool seek_compaction = false);

  int level_;
+  int out_level_; // levels to which output files are stored
  uint64_t max_output_file_size_;
-  int64_t maxGrandParentOverlapBytes_;
+  uint64_t maxGrandParentOverlapBytes_;
  Version* input_version_;
  VersionEdit* edit_;
  int number_levels_;
@@ -555,13 +601,14 @@ class Compaction {
  std::vector<FileMetaData*> grandparents_;
  size_t grandparent_index_;  // Index in grandparent_starts_
  bool seen_key_;             // Some output key has been seen
-  int64_t overlapped_bytes_;  // Bytes of overlap between current output
+  uint64_t overlapped_bytes_;  // Bytes of overlap between current output
                              // and grandparent files
  int base_index_;   // index of the file in files_[level_]
  int parent_index_; // index of some file with same range in files_[level_+1]
  double score_;     // score that was used to pick this compaction.

-  // State for implementing IsBaseLevelForKey
+  // Is this compaction creating a file in the bottom most level?
+  bool bottommost_level_;

  // level_ptrs_ holds indices into input_version_->levels_: our state
  // is that we are positioned at one of the file ranges for each
@@ -572,11 +619,14 @@ class Compaction {
  // mark (or clear) all files that are being compacted
  void MarkFilesBeingCompacted(bool);

+  // Initialize whether compaction producing files at the bottommost level
+  void SetupBottomMostLevel(bool isManual);
+
  // In case of compaction error, reset the nextIndex that is used
  // to pick up the next file to be compacted from files_by_size_
  void ResetNextCompactionIndex();
 };

-}  // namespace leveldb
+}  // namespace rocksdb

 #endif  // STORAGE_LEVELDB_DB_VERSION_SET_H_
@@ -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) {

@@ -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();
 }
@@ -14,16 +14,19 @@
 //    len: varint32
 //    data: uint8[len]

-#include "leveldb/write_batch.h"
+#include "rocksdb/write_batch.h"

-#include "leveldb/db.h"
+#include "rocksdb/options.h"
+#include "rocksdb/statistics.h"
 #include "db/dbformat.h"
+#include "db/db_impl.h"
 #include "db/memtable.h"
+#include "db/snapshot.h"
 #include "db/write_batch_internal.h"
 #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;
@@ -40,11 +43,24 @@ void WriteBatch::Handler::Merge(const Slice& key, const Slice& value) {
  throw std::runtime_error("Handler::Merge not implemented!");
 }

+void WriteBatch::Handler::LogData(const Slice& blob) {
+  // If the user has not specified something to do with blobs, then we ignore
+  // them.
+}
+
+bool WriteBatch::Handler::Continue() {
+  return true;
+}
+
 void WriteBatch::Clear() {
  rep_.clear();
  rep_.resize(kHeader);
 }

+int WriteBatch::Count() const {
+  return WriteBatchInternal::Count(this);
+}
+
 Status WriteBatch::Iterate(Handler* handler) const {
  Slice input(rep_);
  if (input.size() < kHeader) {
@@ -52,10 +68,9 @@ Status WriteBatch::Iterate(Handler* handler) const {
  }

  input.remove_prefix(kHeader);
-  Slice key, value;
+  Slice key, value, blob;
  int found = 0;
-  while (!input.empty()) {
-    found++;
+  while (!input.empty() && handler->Continue()) {
    char tag = input[0];
    input.remove_prefix(1);
    switch (tag) {
@@ -63,6 +78,7 @@ Status WriteBatch::Iterate(Handler* handler) const {
        if (GetLengthPrefixedSlice(&input, &key) &&
            GetLengthPrefixedSlice(&input, &value)) {
          handler->Put(key, value);
+          found++;
        } else {
          return Status::Corruption("bad WriteBatch Put");
        }
@@ -70,6 +86,7 @@ Status WriteBatch::Iterate(Handler* handler) const {
      case kTypeDeletion:
        if (GetLengthPrefixedSlice(&input, &key)) {
          handler->Delete(key);
+          found++;
        } else {
          return Status::Corruption("bad WriteBatch Delete");
        }
@@ -78,10 +95,18 @@ Status WriteBatch::Iterate(Handler* handler) const {
        if (GetLengthPrefixedSlice(&input, &key) &&
            GetLengthPrefixedSlice(&input, &value)) {
          handler->Merge(key, value);
+          found++;
        } else {
          return Status::Corruption("bad WriteBatch Merge");
        }
        break;
+      case kTypeLogData:
+        if (GetLengthPrefixedSlice(&input, &blob)) {
+          handler->LogData(blob);
+        } else {
+          return Status::Corruption("bad WriteBatch Blob");
+        }
+        break;
      default:
        return Status::Corruption("unknown WriteBatch tag");
    }
@@ -129,12 +154,33 @@ void WriteBatch::Merge(const Slice& key, const Slice& value) {
  PutLengthPrefixedSlice(&rep_, value);
 }

+void WriteBatch::PutLogData(const Slice& blob) {
+  rep_.push_back(static_cast<char>(kTypeLogData));
+  PutLengthPrefixedSlice(&rep_, blob);
+}

 namespace {
 class MemTableInserter : public WriteBatch::Handler {
 public:
  SequenceNumber sequence_;
  MemTable* mem_;
+  const Options* options_;
+  DBImpl* db_;
+  const bool filter_deletes_;
+
+  MemTableInserter(SequenceNumber sequence, MemTable* mem, const Options* opts,
+                   DB* db, const bool filter_deletes)
+    : sequence_(sequence),
+      mem_(mem),
+      options_(opts),
+      db_(reinterpret_cast<DBImpl*>(db)),
+      filter_deletes_(filter_deletes) {
+    assert(mem_);
+    if (filter_deletes_) {
+      assert(options_);
+      assert(db_);
+    }
+  }

  virtual void Put(const Slice& key, const Slice& value) {
    mem_->Add(sequence_, kTypeValue, key, value);
@@ -145,17 +191,28 @@ class MemTableInserter : public WriteBatch::Handler {
    sequence_++;
  }
  virtual void Delete(const Slice& key) {
+    if (filter_deletes_) {
+      SnapshotImpl read_from_snapshot;
+      read_from_snapshot.number_ = sequence_;
+      ReadOptions ropts;
+      ropts.snapshot = &read_from_snapshot;
+      std::string value;
+      if (!db_->KeyMayExist(ropts, key, &value)) {
+        RecordTick(options_->statistics, NUMBER_FILTERED_DELETES);
+        return;
+      }
+    }
    mem_->Add(sequence_, kTypeDeletion, key, Slice());
    sequence_++;
  }
 };
 }  // namespace

-Status WriteBatchInternal::InsertInto(const WriteBatch* b,
-                                      MemTable* memtable) {
-  MemTableInserter inserter;
-  inserter.sequence_ = WriteBatchInternal::Sequence(b);
-  inserter.mem_ = memtable;
+Status WriteBatchInternal::InsertInto(const WriteBatch* b, MemTable* mem,
+                                      const Options* opts, DB* db,
+                                      const bool filter_deletes) {
+  MemTableInserter inserter(WriteBatchInternal::Sequence(b), mem, opts, db,
+                            filter_deletes);
  return b->Iterate(&inserter);
 }

@@ -170,4 +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
@@ -5,10 +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 "rocksdb/types.h"
+#include "rocksdb/write_batch.h"
+#include "rocksdb/db.h"
+#include "rocksdb/options.h"

-namespace leveldb {
+namespace rocksdb {

 class MemTable;

@@ -39,12 +41,17 @@ class WriteBatchInternal {

  static void SetContents(WriteBatch* batch, const Slice& contents);

-  static Status InsertInto(const WriteBatch* batch, MemTable* memtable);
+  // Inserts batch entries into memtable
+  // Drops deletes in batch if filter_del is set to true and
+  // db->KeyMayExist returns false
+  static Status InsertInto(const WriteBatch* batch, MemTable* memtable,
+                           const Options* opts = nullptr, DB* db = nullptr,
+                           const bool filter_del = false);

  static void Append(WriteBatch* dst, const WriteBatch* src);
 };

-}  // namespace leveldb
+}  // namespace rocksdb


 #endif  // STORAGE_LEVELDB_DB_WRITE_BATCH_INTERNAL_H_
@@ -2,19 +2,22 @@
 // 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/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());
-  MemTable* mem = new MemTable(cmp);
+  auto factory = std::make_shared<SkipListFactory>();
+  MemTable* mem = new MemTable(cmp, factory);
  mem->Ref();
  std::string state;
  Status s = WriteBatchInternal::InsertInto(b, mem);
@@ -47,13 +50,16 @@ static std::string PrintContents(WriteBatch* b) {
        state.append(")");
        count++;
        break;
+      case kTypeLogData:
+        assert(false);
+        break;
    }
    state.append("@");
    state.append(NumberToString(ikey.sequence));
  }
  delete iter;
  if (!s.ok()) {
-    state.append("ParseError()");
+    state.append(s.ToString());
  } else if (count != WriteBatchInternal::Count(b)) {
    state.append("CountMismatch()");
  }
@@ -67,6 +73,7 @@ TEST(WriteBatchTest, Empty) {
  WriteBatch batch;
  ASSERT_EQ("", PrintContents(&batch));
  ASSERT_EQ(0, WriteBatchInternal::Count(&batch));
+  ASSERT_EQ(0, batch.Count());
 }

 TEST(WriteBatchTest, Multiple) {
@@ -81,6 +88,7 @@ TEST(WriteBatchTest, Multiple) {
            "Delete(box)@101"
            "Put(foo, bar)@100",
            PrintContents(&batch));
+  ASSERT_EQ(3, batch.Count());
 }

 TEST(WriteBatchTest, Corruption) {
@@ -92,7 +100,7 @@ TEST(WriteBatchTest, Corruption) {
  WriteBatchInternal::SetContents(&batch,
                                  Slice(contents.data(),contents.size()-1));
  ASSERT_EQ("Put(foo, bar)@200"
-            "ParseError()",
+            "Corruption: bad WriteBatch Delete",
            PrintContents(&batch));
 }

@@ -103,16 +111,19 @@ TEST(WriteBatchTest, Append) {
  WriteBatchInternal::Append(&b1, &b2);
  ASSERT_EQ("",
            PrintContents(&b1));
+  ASSERT_EQ(0, b1.Count());
  b2.Put("a", "va");
  WriteBatchInternal::Append(&b1, &b2);
  ASSERT_EQ("Put(a, va)@200",
            PrintContents(&b1));
+  ASSERT_EQ(1, b1.Count());
  b2.Clear();
  b2.Put("b", "vb");
  WriteBatchInternal::Append(&b1, &b2);
  ASSERT_EQ("Put(a, va)@200"
            "Put(b, vb)@201",
            PrintContents(&b1));
+  ASSERT_EQ(2, b1.Count());
  b2.Delete("foo");
  WriteBatchInternal::Append(&b1, &b2);
  ASSERT_EQ("Put(a, va)@200"
@@ -120,10 +131,98 @@ TEST(WriteBatchTest, Append) {
            "Put(b, vb)@201"
            "Delete(foo)@203",
            PrintContents(&b1));
+  ASSERT_EQ(4, b1.Count());
 }

-}  // namespace leveldb
+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"));
+  batch.Put(Slice("k2"), Slice("v2"));
+  batch.Put(Slice("k3"), Slice("v3"));
+  batch.PutLogData(Slice("blob1"));
+  batch.Delete(Slice("k2"));
+  batch.PutLogData(Slice("blob2"));
+  batch.Merge(Slice("foo"), Slice("bar"));
+  ASSERT_EQ(5, batch.Count());
+  ASSERT_EQ("Merge(foo, bar)@4"
+            "Put(k1, v1)@0"
+            "Delete(k2)@3"
+            "Put(k2, v2)@1"
+            "Put(k3, v3)@2",
+            PrintContents(&batch));
+
+  TestHandler handler;
+  batch.Iterate(&handler);
+  ASSERT_EQ(
+            "Put(k1, v1)"
+            "Put(k2, v2)"
+            "Put(k3, v3)"
+            "LogData(blob1)"
+            "Delete(k2)"
+            "LogData(blob2)"
+            "Merge(foo, bar)",
+            handler.seen);
+}
+
+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();
 }
@@ -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;
 }
@@ -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;
 }
@@ -25,27 +25,27 @@ creating it if necessary:
  #include &lt;assert&gt;
  #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", &amp;db);
+  rocksdb::Status status = rocksdb::DB::Open(options, "/tmp/testdb", &amp;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>
@@ -1,2 +0,0 @@
-ar: creating libmemenv.a
-ar: creating libleveldb.a
@@ -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 "";}
 };
@@ -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
@@ -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_
@@ -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();
 }
@@ -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
@@ -1,74 +0,0 @@
-// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file. See the AUTHORS file for names of contributors.
-
-#ifndef STORAGE_LEVELDB_INCLUDE_MERGE_OPERATOR_H_
-#define STORAGE_LEVELDB_INCLUDE_MERGE_OPERATOR_H_
-
-#include <string>
-
-namespace leveldb {
-
-class Slice;
-class Logger;
-
-// The Merge Operator interface.
-// Client needs to provide an object implementing this interface if Merge
-// operation is accessed.
-// Essentially, MergeOperator specifies the SEMANTICS of a merge, which only
-// client knows. It could be numeric addition, list append, string
-// concatenation, ... , anything.
-// The library, on the other hand, is concerned with the exercise of this
-// interface, at the right time (during get, iteration, compaction...)
-// Note that, even though in principle we don't require any special property
-// of the merge operator, the current rocksdb compaction order does imply that
-// an associative operator could be exercised more naturally (and more
-// efficiently).
-//
-// Refer to my_test.cc for an example of implementation
-//
-class MergeOperator {
- public:
-  virtual ~MergeOperator() {}
-
-  // Gives the client a way to express the read -> modify -> write semantics
-  // key:      (IN)    The key that's associated with this merge operation.
-  //                   Client could multiplex the merge operator based on it
-  //                   if the key space is partitioned and different subspaces
-  //                   refer to different types of data which have different
-  //                   merge operation semantics
-  // existing: (IN)    null indicates that the key does not exist before this op
-  // value:    (IN)    The passed-in merge operand value (when Merge is issued)
-  // new_value:(OUT)   Client is responsible for filling the merge result here
-  // logger:   (IN)    Client could use this to log errors during merge.
-  //
-  // Note: Merge does not return anything to indicate if a merge is successful
-  //       or not.
-  // Rationale: If a merge failed due to, say de-serialization error, we still
-  //            need to define a consistent merge result. Should we throw away
-  //            the existing value? the merge operand? Or reset the merged value
-  //            to sth? The rocksdb library is not in a position to make the
-  //            right choice. On the other hand, client knows exactly what
-  //            happened during Merge, thus is able to make the best decision.
-  //            Just save the final decision in new_value. logger is passed in,
-  //            in case client wants to leave a trace of what went wrong.
-  virtual void Merge(const Slice& key,
-                     const Slice* existing_value,
-                     const Slice& value,
-                     std::string* new_value,
-                     Logger* logger) const = 0;
-
-
-  // The name of the MergeOperator.  Used to check for MergeOperator
-  // mismatches (i.e., a DB created with one MergeOperator is
-  // accessed using a different MergeOperator)
-  // TODO: the name is currently not stored persistently and thus
-  //       no checking is enforced. Client is responsible for providing
-  //       consistent MergeOperator between DB opens.
-  virtual const char* Name() const = 0;
-
-};
-
-}  // namespace leveldb
-
-#endif  // STORAGE_LEVELDB_INCLUDE_MERGE_OPERATOR_H_
@@ -1,42 +0,0 @@
-// Copyright 2008-present Facebook. All Rights Reserved.
-#ifndef STORAGE_LEVELDB_INCLUDE_TRANSACTION_LOG_ITERATOR_H_
-#define STORAGE_LEVELDB_INCLUDE_TRANSACTION_LOG_ITERATOR_H_
-
-#include "leveldb/status.h"
-#include "leveldb/write_batch.h"
-
-namespace leveldb {
-
-
-struct BatchResult {
-  SequenceNumber sequence;
-  std::unique_ptr<WriteBatch> writeBatchPtr;
-};
-
-//  A TransactionLogIterator is used to iterate over the Transaction's in a db.
-class TransactionLogIterator {
- public:
-  TransactionLogIterator() {}
-  virtual ~TransactionLogIterator() {}
-
-  // An iterator is either positioned at a WriteBatch or not valid.
-  // This method returns true if the iterator is valid.
-  // Can read data from a valid iterator.
-  virtual bool Valid() = 0;
-
-  // Moves the iterator to the next WriteBatch.
-  // REQUIRES: Valid() to be true.
-  virtual void Next() = 0;
-
-  // Return's ok if the iterator is valid.
-  // Return the Error when something has gone wrong.
-  virtual Status status() = 0;
-
-  // If valid return's the current write_batch and the sequence number of the
-  // latest transaction contained in the batch.
-  // ONLY use if Valid() is true and status() is OK.
-  virtual BatchResult GetBatch() = 0;
-};
-} //  namespace leveldb
-
-#endif  // STORAGE_LEVELDB_INCLUDE_TRANSACTION_LOG_ITERATOR_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_
@@ -0,0 +1,43 @@
+// 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.
+//
+// Arena class defines memory allocation methods. It's used by memtable and
+// skiplist.
+
+#ifndef STORAGE_ROCKSDB_INCLUDE_ARENA_H_
+#define STORAGE_ROCKSDB_INCLUDE_ARENA_H_
+
+#include <limits>
+#include <memory>
+
+namespace rocksdb {
+
+class Arena {
+ public:
+  Arena() {};
+  virtual ~Arena() {};
+
+  // Return a pointer to a newly allocated memory block of "bytes" bytes.
+  virtual char* Allocate(size_t bytes) = 0;
+
+  // Allocate memory with the normal alignment guarantees provided by malloc.
+  virtual char* AllocateAligned(size_t bytes) = 0;
+
+  // Returns an estimate of the total memory used by arena.
+  virtual const size_t ApproximateMemoryUsage() = 0;
+
+  // Returns the total number of bytes in all blocks allocated so far.
+  virtual const size_t MemoryAllocatedBytes() = 0;
+
+ private:
+  // No copying allowed
+  Arena(const Arena&);
+  void operator=(const Arena&);
+};
+
+}  // namespace rocksdb
+
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif  // STORAGE_ROCKSDB_INCLUDE_ARENA_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_ */
@@ -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_
@@ -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;

@@ -41,6 +41,33 @@ class CompactionFilter {
  virtual const char* Name() const = 0;
 };

-}  // namespace leveldb
+// Each compaction will create a new CompactionFilter allowing the
+// application to know about different campactions
+class CompactionFilterFactory {
+  public:
+    virtual ~CompactionFilterFactory() { };
+    virtual std::unique_ptr<CompactionFilter> CreateCompactionFilter() = 0;

-#endif  // STORAGE_LEVELDB_INCLUDE_COMPACTION_FILTER_H_
+    // Returns a name that identifies this compaction filter factory.
+    virtual const char* Name() const = 0;
+};
+
+// Default implementaion of CompactionFilterFactory which does not
+// return any filter
+class DefaultCompactionFilterFactory : public CompactionFilterFactory {
+  public:
+    virtual std::unique_ptr<CompactionFilter>
+    CreateCompactionFilter() override {
+      return std::unique_ptr<CompactionFilter>(nullptr);
+    }
+
+    virtual const char* Name() const override {
+      return "DefaultCompactionFilterFactory";
+    }
+};
+
+}  // namespace rocksdb
+
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif  // STORAGE_ROCKSDB_INCLUDE_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_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_
@@ -2,25 +2,25 @@
 // 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_iterator.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;

 // Update Makefile if you change these
-static const int kMajorVersion = 1;
-static const int kMinorVersion = 5;
+static const int kMajorVersion = 2;
+static const int kMinorVersion = 0;

 struct Options;
 struct ReadOptions;
@@ -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.
@@ -104,7 +115,8 @@ class DB {
  //
  // May return some other Status on an error.
  virtual Status Get(const ReadOptions& options,
-                     const Slice& key, std::string* value) = 0;
+                     const Slice& key,
+                     std::string* value) = 0;

  // If keys[i] does not exist in the database, then the i'th returned
  // status will be one for which Status::IsNotFound() is true, and
@@ -120,6 +132,23 @@ class DB {
                                       const std::vector<Slice>& keys,
                                       std::vector<std::string>* values) = 0;

+  // If the key definitely does not exist in the database, then this method
+  // returns false, else true. If the caller wants to obtain value when the key
+  // is found in memory, a bool for 'value_found' must be passed. 'value_found'
+  // will be true on return if value has been set properly.
+  // This check is potentially lighter-weight than invoking DB::Get(). One way
+  // to make this lighter weight is to avoid doing any IOs.
+  // Default implementation here returns true and sets 'value_found' to false
+  virtual bool KeyMayExist(const ReadOptions& options,
+                           const Slice& key,
+                           std::string* value,
+                           bool* value_found = nullptr) {
+    if (value_found != nullptr) {
+      *value_found = false;
+    }
+    return true;
+  }
+
  // Return a heap-allocated iterator over the contents of the database.
  // The result of NewIterator() is initially invalid (caller must
  // call one of the Seek methods on the iterator before using it).
@@ -175,7 +204,15 @@ class DB {
  // end==nullptr is treated as a key after all keys in the database.
  // Therefore the following call will compact the entire database:
  //    db->CompactRange(nullptr, nullptr);
-  virtual void CompactRange(const Slice* begin, const Slice* end) = 0;
+  // Note that after the entire database is compacted, all data are pushed
+  // down to the last level containing any data. If the total data size
+  // 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 or a given level (specified by non-negative target_level).
+  virtual void CompactRange(const Slice* begin, const Slice* end,
+                            bool reduce_level = false,
+                            int target_level = -1) = 0;

  // Number of levels used for this DB.
  virtual int NumberLevels() = 0;
@@ -198,14 +235,31 @@ 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
+  virtual Status DeleteWalFiles(const VectorLogPtr& files) = 0;

  // The sequence number of the most recent transaction.
  virtual SequenceNumber GetLatestSequenceNumber() = 0;
@@ -223,6 +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&);
@@ -239,6 +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_
@@ -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_
@@ -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_
@@ -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_
@@ -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
@@ -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_
@@ -0,0 +1,149 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#ifndef STORAGE_ROCKSDB_INCLUDE_MERGE_OPERATOR_H_
+#define STORAGE_ROCKSDB_INCLUDE_MERGE_OPERATOR_H_
+
+#include <string>
+#include <deque>
+#include "rocksdb/slice.h"  // TODO: Remove this when migration is done;
+
+namespace rocksdb {
+
+class Slice;
+class Logger;
+
+// The Merge Operator
+//
+// Essentially, a MergeOperator specifies the SEMANTICS of a merge, which only
+// client knows. It could be numeric addition, list append, string
+// concatenation, edit data structure, ... , anything.
+// The library, on the other hand, is concerned with the exercise of this
+// interface, at the right time (during get, iteration, compaction...)
+//
+// To use merge, the client needs to provide an object implementing one of
+// the following interfaces:
+//  a) AssociativeMergeOperator - for most simple semantics (always take
+//    two values, and merge them into one value, which is then put back
+//    into rocksdb); numeric addition and string concatenation are examples;
+//
+//  b) MergeOperator - the generic class for all the more abstract / complex
+//    operations; one method (FullMerge) to merge a Put/Delete value with a
+//    merge operand; and another method (PartialMerge) that merges two
+//    operands together. this is especially useful if your key values have a
+//    complex structure but you would still like to support client-specific
+//    incremental updates.
+//
+// AssociativeMergeOperator is simpler to implement. MergeOperator is simply
+// more powerful.
+//
+// Refer to rocksdb-merge wiki for more details and example implementations.
+//
+class MergeOperator {
+ public:
+  virtual ~MergeOperator() {}
+
+  // Gives the client a way to express the read -> modify -> write semantics
+  // key:      (IN)    The key that's associated with this merge operation.
+  //                   Client could multiplex the merge operator based on it
+  //                   if the key space is partitioned and different subspaces
+  //                   refer to different types of data which have different
+  //                   merge operation semantics
+  // existing: (IN)    null indicates that the key does not exist before this op
+  // operand_list:(IN) the sequence of merge operations to apply, front() first.
+  // new_value:(OUT)   Client is responsible for filling the merge result here
+  // logger:   (IN)    Client could use this to log errors during merge.
+  //
+  // Return true on success.
+  // All values passed in will be client-specific values. So if this method
+  // returns false, it is because client specified bad data or there was
+  // internal corruption. This will be treated as an error by the library.
+  //
+  // Also make use of the *logger for error messages.
+  virtual bool FullMerge(const Slice& key,
+                         const Slice* existing_value,
+                         const std::deque<std::string>& operand_list,
+                         std::string* new_value,
+                         Logger* logger) const = 0;
+
+  // This function performs merge(left_op, right_op)
+  // when both the operands are themselves merge operation types
+  // that you would have passed to a DB::Merge() call in the same order
+  // (i.e.: DB::Merge(key,left_op), followed by DB::Merge(key,right_op)).
+  //
+  // PartialMerge should combine them into a single merge operation that is
+  // saved into *new_value, and then it should return true.
+  // *new_value should be constructed such that a call to
+  // DB::Merge(key, *new_value) would yield the same result as a call
+  // to DB::Merge(key, left_op) followed by DB::Merge(key, right_op).
+  //
+  // If it is impossible or infeasible to combine the two operations,
+  // leave new_value unchanged and return false. The library will
+  // internally keep track of the operations, and apply them in the
+  // correct order once a base-value (a Put/Delete/End-of-Database) is seen.
+  //
+  // TODO: Presently there is no way to differentiate between error/corruption
+  // and simply "return false". For now, the client should simply return
+  // false in any case it cannot perform partial-merge, regardless of reason.
+  // If there is corruption in the data, handle it in the FullMerge() function,
+  // and return false there.
+  virtual bool PartialMerge(const Slice& key,
+                            const Slice& left_operand,
+                            const Slice& right_operand,
+                            std::string* new_value,
+                            Logger* logger) const = 0;
+
+  // The name of the MergeOperator. Used to check for MergeOperator
+  // mismatches (i.e., a DB created with one MergeOperator is
+  // accessed using a different MergeOperator)
+  // TODO: the name is currently not stored persistently and thus
+  //       no checking is enforced. Client is responsible for providing
+  //       consistent MergeOperator between DB opens.
+  virtual const char* Name() const = 0;
+};
+
+// The simpler, associative merge operator.
+class AssociativeMergeOperator : public MergeOperator {
+ public:
+  virtual ~AssociativeMergeOperator() {}
+
+  // Gives the client a way to express the read -> modify -> write semantics
+  // key:           (IN) The key that's associated with this merge operation.
+  // existing_value:(IN) null indicates the key does not exist before this op
+  // value:         (IN) the value to update/merge the existing_value with
+  // new_value:    (OUT) Client is responsible for filling the merge result here
+  // logger:        (IN) Client could use this to log errors during merge.
+  //
+  // Return true on success.
+  // All values passed in will be client-specific values. So if this method
+  // returns false, it is because client specified bad data or there was
+  // internal corruption. The client should assume that this will be treated
+  // as an error by the library.
+  virtual bool Merge(const Slice& key,
+                     const Slice* existing_value,
+                     const Slice& value,
+                     std::string* new_value,
+                     Logger* logger) const = 0;
+
+
+ private:
+  // Default implementations of the MergeOperator functions
+  virtual bool FullMerge(const Slice& key,
+                         const Slice* existing_value,
+                         const std::deque<std::string>& operand_list,
+                         std::string* new_value,
+                         Logger* logger) const override;
+
+  virtual bool PartialMerge(const Slice& key,
+                            const Slice& left_operand,
+                            const Slice& right_operand,
+                            std::string* new_value,
+                            Logger* logger) const override;
+};
+
+}  // namespace rocksdb
+
+#include "rocksdb/rocksdb_to_leveldb.h"
+
+#endif  // STORAGE_ROCKSDB_INCLUDE_MERGE_OPERATOR_H_
@@ -2,18 +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 "rocksdb/slice.h"
+#include "rocksdb/statistics.h"
+#include "rocksdb/universal_compaction.h"
+#include "rocksdb/memtablerep.h"
+#include "rocksdb/slice_transform.h"

-namespace leveldb {
+namespace rocksdb {

 class Cache;
 class Comparator;
@@ -23,6 +26,7 @@ class Logger;
 class MergeOperator;
 class Snapshot;
 class CompactionFilter;
+class CompactionFilterFactory;

 using std::shared_ptr;

@@ -39,6 +43,11 @@ enum CompressionType {
  kBZip2Compression = 0x3
 };

+enum CompactionStyle {
+  kCompactionStyleLevel       = 0x0, // level based compaction style
+  kCompactionStyleUniversal   = 0x1  // Universal compaction style
+};
+
 // Compression options for different compression algorithms like Zlib
 struct CompressionOptions {
  int window_bits;
@@ -75,10 +84,15 @@ struct Options {
  // for the first time. It's necessary to specify a merge operator when
  // openning the DB in this case.
  // Default: nullptr
-  const MergeOperator* merge_operator;
+  shared_ptr<MergeOperator> merge_operator;

+  // The client must provide compaction_filter_factory if it requires a new
+  // compaction filter to be used for different compaction processes
  // Allows an application to modify/delete a key-value during background
  // compaction.
+  // Ideally, client should specify only one of filter or factory.
+  // compaction_filter takes precedence over compaction_filter_factory if
+  // client specifies both.
  // Default: nullptr
  const CompactionFilter* compaction_filter;

@@ -132,7 +146,7 @@ struct Options {
  int max_write_buffer_number;

  // The minimum number of write buffers that will be merged together
-  // before writing to storage.  If set to 1, then 
+  // before writing to storage.  If set to 1, then
  // all write buffers are fushed to L0 as individual files and this increases
  // read amplification because a get request has to check in all of these
  // files. Also, an in-memory merge may result in writing lesser
@@ -211,6 +225,28 @@ struct Options {
  // Default: nullptr
  const FilterPolicy* filter_policy;

+  // If non-nullptr, use the specified function to determine the
+  // prefixes for keys.  These prefixes will be placed in the filter.
+  // Depending on the workload, this can reduce the number of read-IOP
+  // cost for scans when a prefix is passed via ReadOptions to
+  // db.NewIterator().  For prefix filtering to work properly,
+  // "prefix_extractor" and "comparator" must be such that the following
+  // properties hold:
+  //
+  // 1) key.starts_with(prefix(key))
+  // 2) Compare(prefix(key), key) <= 0.
+  // 3) If Compare(k1, k2) <= 0, then Compare(prefix(k1), prefix(k2)) <= 0
+  // 4) prefix(prefix(key)) == prefix(key)
+  //
+  // Default: nullptr
+  const SliceTransform* prefix_extractor;
+
+  // If true, place whole keys in the filter (not just prefixes).
+  // This must generally be true for gets to be efficient.
+  //
+  // Default: true
+  bool whole_key_filtering;
+
  // Number of levels for this database
  int num_levels;

@@ -218,9 +254,9 @@ struct Options {
  // level-0 compaction will not be triggered by number of files at all.
  int level0_file_num_compaction_trigger;

-  // Soft limit on number of level-0 files. We slow down writes at this point.
-  // A value <0 means that no writing slow down will be triggered by number
-  // of files in level-0.
+  // Soft limit on number of level-0 files. We start slowing down writes at this
+  // point. A value <0 means that no writing slow down will be triggered by
+  // number of files in level-0.
  int level0_slowdown_writes_trigger;

  // Maximum number of level-0 files.  We stop writes at this point.
@@ -333,10 +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.
@@ -354,26 +403,44 @@ struct Options {
  // Default: 1000
  size_t keep_log_file_num;

-  // Puts are delayed when any level has a compaction score that
-  // exceeds rate_limit. This is ignored when <= 1.0.
-  double rate_limit;
+  // Puts are delayed 0-1 ms when any level has a compaction score that exceeds
+  // soft_rate_limit. This is ignored when == 0.0.
+  // CONSTRAINT: soft_rate_limit <= hard_rate_limit. If this constraint does not
+  // hold, RocksDB will set soft_rate_limit = hard_rate_limit
+  // Default: 0 (disabled)
+  double soft_rate_limit;

-  // Max time a put will be stalled when rate_limit is enforced
-  unsigned int rate_limit_delay_milliseconds;
+  // Puts are delayed 1ms at a time when any level has a compaction score that
+  // exceeds hard_rate_limit. This is ignored when <= 1.0.
+  // Default: 0 (disabled)
+  double hard_rate_limit;
+
+  // Max time a put will be stalled when hard_rate_limit is enforced. If 0, then
+  // there is no limit.
+  // Default: 1000
+  unsigned int rate_limit_delay_max_milliseconds;

  // manifest file is rolled over on reaching this limit.
  // The older manifest file be deleted.
  // The default value is MAX_INT so that roll-over does not take place.
  uint64_t max_manifest_file_size;

-  // Disable block cache. If this is set to false,
+  // Disable block cache. If this is set to true,
  // then no block cache should be used, and the block_cache should
  // point to a nullptr object.
+  // Default: false
  bool no_block_cache;

  // Number of shards used for table cache.
  int table_cache_numshardbits;

+  // size of one block in arena memory allocation.
+  // If <= 0, a proper value is automatically calculated (usually 1/10 of
+  // writer_buffer_size).
+  //
+  // Default: 0
+  size_t arena_block_size;
+
  // Create an Options object with default values for all fields.
  Options();

@@ -416,17 +483,6 @@ struct Options {
  // Default: true
  bool allow_os_buffer;

-  // Reading a single block from a file can cause the OS/FS to start
-  // readaheads of other blocks from the file. Default: true
-  // Note: Deprecated
-  bool allow_readahead;
-
-  // The reads triggered by compaction allows data to be readahead
-  // by the OS/FS. This overrides the setting of 'allow_readahead'
-  // for compaction-reads. Default: true
-  // Note: Deprecated
-  bool allow_readahead_compactions;
-
  // Allow the OS to mmap file for reading sst tables. Default: false
  bool allow_mmap_reads;

@@ -476,6 +532,52 @@ struct Options {
  // Default: 0
  uint64_t bytes_per_sync;

+  // The compaction style. Default: kCompactionStyleLevel
+  CompactionStyle compaction_style;
+
+  // The options needed to support Universal Style compactions
+  CompactionOptionsUniversal compaction_options_universal;
+
+  // Use KeyMayExist API to filter deletes when this is true.
+  // If KeyMayExist returns false, i.e. the key definitely does not exist, then
+  // the delete is a noop. KeyMayExist only incurs in-memory look up.
+  // This optimization avoids writing the delete to storage when appropriate.
+  // 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.
+  std::shared_ptr<MemTableRepFactory> memtable_factory;
+
+  // This is a factory that provides compaction filter objects which allow
+  // 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
@@ -497,14 +599,36 @@ struct ReadOptions {
  // Default: nullptr
  const Snapshot* snapshot;

+  // If "prefix" is non-nullptr, and ReadOptions is being passed to
+  // db.NewIterator, only return results when the key begins with this
+  // prefix.  This field is ignored by other calls (e.g., Get).
+  // Options.prefix_extractor must also be set, and
+  // prefix_extractor.InRange(prefix) must be true.  The iterator
+  // returned by NewIterator when this option is set will behave just
+  // as if the underlying store did not contain any non-matching keys,
+  // with two exceptions.  Seek() only accepts keys starting with the
+  // prefix, and SeekToLast() is not supported.  prefix filter with this
+  // option will sometimes reduce the number of read IOPs.
+  // 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) {
+        snapshot(nullptr),
+        prefix(nullptr),
+        read_tier(kReadAllTier) {
  }
  ReadOptions(bool cksum, bool cache) :
              verify_checksums(cksum), fill_cache(cache),
-              snapshot(nullptr) {
+              snapshot(nullptr), prefix(nullptr),
+              read_tier(kReadAllTier) {
  }
 };

@@ -549,6 +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_
@@ -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
@@ -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;
--- a/Show More
+++ b/Show More