An odd fix for GCC 7

Disable warning as error
Add some include<functional>
2026-11-06 11:02:36 -05:00 · 2019-11-02 12:48:48 -07:00 · 2019-10-31 14:54:22 -07:00 · 2019-10-31 14:54:22 -07:00 · 2019-10-31 14:54:22 -07:00 · 2017-07-17 21:01:59 -07:00
339 changed files with 35017 additions and 10021 deletions
@@ -28,7 +28,13 @@ util/build_version.cc
 build_tools/VALGRIND_LOGS/
 coverage/COVERAGE_REPORT
 .gdbhistory
+package/
 .phutil_module_cache
 tags
 java/*.log
 java/include/org_rocksdb_*.h
+unity.cc
+java/crossbuild/.vagrant
+.vagrant/
+java/**.asc
+java/javadoc
@@ -14,7 +14,6 @@ before_install:
 - sudo dpkg -i libgflags-dev_2.0-1_amd64.deb
 # Lousy hack to disable use and testing of fallocate, which doesn't behave quite
 # as EnvPosixTest::AllocateTest expects within the Travis OpenVZ environment.
- - sed -i "s/fallocate(/HACK_NO_fallocate(/" build_tools/build_detect_platform
-script: make check -j8
+script: OPT=-DTRAVIS make unity  && OPT=-DTRAVIS make check -j8
 notifications:
    email: false
@@ -0,0 +1,11 @@
+Facebook Inc.
+Facebook Engineering Team
+
+Google Inc.
+# Initial version authors:
+Jeffrey Dean <jeff@google.com>
+Sanjay Ghemawat <sanjay@google.com>
+
+# Partial list of contributors:
+Kevin Regan <kevin.d.regan@gmail.com>
+Johan Bilien <jobi@litl.com>
@@ -10,9 +10,8 @@ the CLA and we can cross-check with your GitHub username.

 Complete your CLA here: <https://code.facebook.com/cla>

-If you don't have a Facebook account, we can send you a PDF that you can
-sign offline. Send us an e-mail or create a new github issue to
-request the CLA in PDF format.
+If you prefer to sign a paper copy, we can send you a PDF.  Send us an 
+e-mail or create a new github issue to request the CLA in PDF format.

 ## License

@@ -1,5 +1,74 @@
 # Rocksdb Change Log

+## 3.6.0 (10/7/2014)
+### Disk format changes
+* If you're using RocksDB on ARM platforms and you're using default bloom filter, there is a disk format change you need to be aware of. There are three steps you need to do when you convert to new release: 1. turn off filter policy, 2. compact the whole database, 3. turn on filter policy
+
+### Behavior changes
+* We have refactored our system of stalling writes.  Any stall-related statistics' meanings are changed. Instead of per-write stall counts, we now count stalls per-epoch, where epochs are periods between flushes and compactions. You'll find more information in our Tuning Perf Guide once we release RocksDB 3.6.
+* When disableDataSync=true, we no longer sync the MANIFEST file.
+* Add identity_as_first_hash property to CuckooTable. SST file needs to be rebuilt to be opened by reader properly.
+
+### Public API changes
+* Change target_file_size_base type to uint64_t from int.
+* Remove allow_thread_local. This feature was proved to be stable, so we are turning it always-on.
+
+## 3.5.0 (9/3/2014)
+### New Features
+* Add include/utilities/write_batch_with_index.h, providing a utilitiy class to query data out of WriteBatch when building it.
+* Move BlockBasedTable related options to BlockBasedTableOptions from Options. Change corresponding JNI interface. Options affected include:
+  no_block_cache, block_cache, block_cache_compressed, block_size, block_size_deviation, block_restart_interval, filter_policy, whole_key_filtering. filter_policy is changed to shared_ptr from a raw pointer.
+* Remove deprecated options: disable_seek_compaction and db_stats_log_interval
+* OptimizeForPointLookup() takes one parameter for block cache size. It now builds hash index, bloom filter, and block cache.
+
+### Public API changes
+* The Prefix Extractor used with V2 compaction filters is now passed user key to SliceTransform::Transform instead of unparsed RocksDB key.
+
+## 3.4.0 (8/18/2014)
+### New Features
+* Support Multiple DB paths in universal style compactions
+* Add feature of storing plain table index and bloom filter in SST file.
+* CompactRange() will never output compacted files to level 0. This used to be the case when all the compaction input files were at level 0.
+* Added iterate_upper_bound to define the extent upto which the forward iterator will return entries. This will prevent iterating over delete markers and overwritten entries for edge cases where you want to break out the iterator anyways. This may improve perfomance in case there are a large number of delete markers or overwritten entries.
+
+### Public API changes
+* DBOptions.db_paths now is a vector of a DBPath structure which indicates both of path and target size
+* NewPlainTableFactory instead of bunch of parameters now accepts PlainTableOptions, which is defined in include/rocksdb/table.h
+* Moved include/utilities/*.h to include/rocksdb/utilities/*.h
+* Statistics APIs now take uint32_t as type instead of Tickers. Also make two access functions getTickerCount and histogramData const
+* Add DB property rocksdb.estimate-num-keys, estimated number of live keys in DB.
+* Add DB::GetIntProperty(), which returns DB properties that are integer as uint64_t.
+* The Prefix Extractor used with V2 compaction filters is now passed user key to SliceTransform::Transform instead of unparsed RocksDB key.
+
+## 3.3.0 (7/10/2014)
+### New Features
+* Added JSON API prototype.
+* HashLinklist reduces performance outlier caused by skewed bucket by switching data in the bucket from linked list to skip list. Add parameter threshold_use_skiplist in NewHashLinkListRepFactory().
+* RocksDB is now able to reclaim storage space more effectively during the compaction process.  This is done by compensating the size of each deletion entry by the 2X average value size, which makes compaction to be triggerred by deletion entries more easily.
+* Add TimeOut API to write.  Now WriteOptions have a variable called timeout_hint_us.  With timeout_hint_us set to non-zero, any write associated with this timeout_hint_us may be aborted when it runs longer than the specified timeout_hint_us, and it is guaranteed that any write completes earlier than the specified time-out will not be aborted due to the time-out condition.
+* Add a rate_limiter option, which controls total throughput of flush and compaction. The throughput is specified in bytes/sec. Flush always has precedence over compaction when available bandwidth is constrained.
+
+### Public API changes
+* Removed NewTotalOrderPlainTableFactory because it is not used and implemented semantically incorrect.
+
+## 3.2.0 (06/20/2014)
+
+### Public API changes
+* We removed seek compaction as a concept from RocksDB because:
+1) It makes more sense for spinning disk workloads, while RocksDB is primarily designed for flash and memory,
+2) It added some complexity to the important code-paths,
+3) None of our internal customers were really using it.
+Because of that, Options::disable_seek_compaction is now obsolete. It is still a parameter in Options, so it does not break the build, but it does not have any effect. We plan to completely remove it at some point, so we ask users to please remove this option from your code base.
+* Add two paramters to NewHashLinkListRepFactory() for logging on too many entries in a hash bucket when flushing.
+* Added new option BlockBasedTableOptions::hash_index_allow_collision. When enabled, prefix hash index for block-based table will not store prefix and allow hash collision, reducing memory consumption.
+
+### New Features
+* PlainTable now supports a new key encoding: for keys of the same prefix, the prefix is only written once. It can be enabled through encoding_type paramter of NewPlainTableFactory()
+* Add AdaptiveTableFactory, which is used to convert from a DB of PlainTable to BlockBasedTabe, or vise versa. It can be created using NewAdaptiveTableFactory()
+
+### Performance Improvements
+* Tailing Iterator re-implemeted with ForwardIterator + Cascading Search Hint , see ~20% throughput improvement.
+
 ## 3.1.0 (05/21/2014)

 ### Public API changes
@@ -15,6 +15,10 @@ There are few options when compiling RocksDB:
 * `make all` will compile our static library, and all our tools and unit tests. Our tools
 depend on gflags. You will need to have gflags installed to run `make all`.

+* if Intel SSE instruction set is supported, set USE_SSE=" -msse -msse4.2 " to make sure
+SSE4.2 is used to speed up CRC32 when calculating data checksum.
+
+
 ## Dependencies

 * You can link RocksDB with following compression libraries:
@@ -74,11 +78,11 @@ depend on gflags. You will need to have gflags installed to run `make all`.
            * run `brew tap homebrew/dupes; brew install gcc47 --use-llvm` to install gcc 4.7 (or higher).
    * Install zlib, bzip2 and snappy libraries for compression.
    * Install gflags. We have included a script
-    `build_tools/mac-install-gflags.sh`, which should automatically install it.
+    `build_tools/mac-install-gflags.sh`, which should automatically install it (execute this file instead of runing using "source" command).
    If you installed gflags by other means (for example, `brew install gflags`),
    please set `LIBRARY_PATH` and `CPATH` accordingly.
    * Please note that some of the optimizations/features are disabled in OSX.
    We did not run any production workloads on it.

 * **iOS**:
-  * Run: `TARGET_OS=IOS make static_lib`
+  * Run: `TARGET_OS=IOS make static_lib`. When building the project which uses rocksdb iOS library, make sure to define two important pre-processing macros: `ROCKSDB_LITE` and `IOS_CROSS_COMPILE`.
@@ -3,7 +3,6 @@
 # found in the LICENSE file. See the AUTHORS file for names of contributors.

 # Inherit some settings from environment variables, if available
-INSTALL_PATH ?= $(CURDIR)

 #-----------------------------------------------

@@ -49,7 +48,27 @@ else
 	PLATFORM_CCFLAGS += $(JEMALLOC_INCLUDE) -DHAVE_JEMALLOC
 endif

-WARNING_FLAGS = -Wall -Werror
+#-------------------------------------------------
+# make install related stuff
+INSTALL_PATH ?= /usr/local
+
+uninstall:
+	@rm -rf $(INSTALL_PATH)/include/rocksdb
+	@rm -rf $(INSTALL_PATH)/lib/$(LIBRARY)
+	@rm -rf $(INSTALL_PATH)/lib/$(SHARED)
+
+install:
+	@install -d $(INSTALL_PATH)/lib
+	@for header_dir in `find "include/rocksdb" -type d`; do \
+		install -d $(INSTALL_PATH)/$$header_dir; \
+	done
+	@for header in `find "include/rocksdb" -type f -name *.h`; do \
+		install -C -m 644 $$header $(INSTALL_PATH)/$$header; \
+	done
+	@[ ! -e $(LIBRARY) ] || install -C -m 644 $(LIBRARY) $(INSTALL_PATH)/lib
+	@[ ! -e $(SHARED) ] || install -C -m 644 $(SHARED) $(INSTALL_PATH)/lib
+#-------------------------------------------------
+
 CFLAGS += $(WARNING_FLAGS) -I. -I./include $(PLATFORM_CCFLAGS) $(OPT)
 CXXFLAGS += $(WARNING_FLAGS) -I. -I./include $(PLATFORM_CXXFLAGS) $(OPT) -Woverloaded-virtual

@@ -69,6 +88,7 @@ VALGRIND_OPTS = --error-exitcode=$(VALGRIND_ERROR) --leak-check=full

 TESTS = \
 	db_test \
+	db_iter_test \
 	block_hash_index_test \
 	autovector_test \
 	column_family_test \
@@ -89,29 +109,40 @@ TESTS = \
 	blob_store_test \
 	filelock_test \
 	filename_test \
-	filter_block_test \
+	block_based_filter_block_test \
+	full_filter_block_test \
 	histogram_test \
 	log_test \
 	manual_compaction_test \
 	memenv_test \
 	merge_test \
+	merger_test \
 	redis_test \
 	reduce_levels_test \
 	plain_table_db_test \
 	prefix_test \
-	simple_table_db_test \
 	skiplist_test \
 	stringappend_test \
 	ttl_test \
 	backupable_db_test \
+	document_db_test \
+	json_document_test \
+	spatial_db_test \
 	version_edit_test \
 	version_set_test \
 	file_indexer_test \
-	write_batch_test\
+	write_batch_test \
+	write_controller_test\
 	deletefile_test \
 	table_test \
 	thread_local_test \
-        geodb_test
+	geodb_test \
+	rate_limiter_test \
+	options_test \
+	cuckoo_table_builder_test \
+	cuckoo_table_reader_test \
+	cuckoo_table_db_test \
+	write_batch_with_index_test

 TOOLS = \
        sst_dump \
@@ -119,9 +150,10 @@ TOOLS = \
        db_stress \
        ldb \
 	db_repl_stress \
+  options_test \
 	blob_store_bench

-PROGRAMS = db_bench signal_test table_reader_bench log_and_apply_bench $(TOOLS)
+PROGRAMS = db_bench signal_test table_reader_bench log_and_apply_bench cache_bench perf_context_test $(TOOLS)

 # The library name is configurable since we are maintaining libraries of both
 # debug/release mode.
@@ -131,6 +163,10 @@ endif
 LIBRARY = ${LIBNAME}.a
 MEMENVLIBRARY = libmemenv.a

+ROCKSDB_MAJOR = $(shell egrep "ROCKSDB_MAJOR.[0-9]" include/rocksdb/version.h | cut -d ' ' -f 3)
+ROCKSDB_MINOR = $(shell egrep "ROCKSDB_MINOR.[0-9]" include/rocksdb/version.h | cut -d ' ' -f 3)
+ROCKSDB_PATCH = $(shell egrep "ROCKSDB_PATCH.[0-9]" include/rocksdb/version.h | cut -d ' ' -f 3)
+
 default: all

 #-----------------------------------------------
@@ -142,29 +178,33 @@ ifneq ($(PLATFORM_SHARED_VERSIONED),true)
 SHARED1 = ${LIBNAME}.$(PLATFORM_SHARED_EXT)
 SHARED2 = $(SHARED1)
 SHARED3 = $(SHARED1)
+SHARED4 = $(SHARED1)
 SHARED = $(SHARED1)
 else
-# Update db.h if you change these.
-SHARED_MAJOR = 3
-SHARED_MINOR = 0
+SHARED_MAJOR = $(ROCKSDB_MAJOR)
+SHARED_MINOR = $(ROCKSDB_MINOR)
+SHARED_PATCH = $(ROCKSDB_PATCH)
 SHARED1 = ${LIBNAME}.$(PLATFORM_SHARED_EXT)
 SHARED2 = $(SHARED1).$(SHARED_MAJOR)
 SHARED3 = $(SHARED1).$(SHARED_MAJOR).$(SHARED_MINOR)
-SHARED = $(SHARED1) $(SHARED2) $(SHARED3)
-$(SHARED1): $(SHARED3)
-	ln -fs $(SHARED3) $(SHARED1)
-$(SHARED2): $(SHARED3)
-	ln -fs $(SHARED3) $(SHARED2)
+SHARED4 = $(SHARED1).$(SHARED_MAJOR).$(SHARED_MINOR).$(SHARED_PATCH)
+SHARED = $(SHARED1) $(SHARED2) $(SHARED3) $(SHARED4)
+$(SHARED1): $(SHARED4)
+	ln -fs $(SHARED4) $(SHARED1)
+$(SHARED2): $(SHARED4)
+	ln -fs $(SHARED4) $(SHARED2)
+$(SHARED3): $(SHARED4)
+	ln -fs $(SHARED4) $(SHARED3)
 endif

-$(SHARED3):
+$(SHARED4):
 	$(CXX) $(PLATFORM_SHARED_LDFLAGS)$(SHARED2) $(CXXFLAGS) $(PLATFORM_SHARED_CFLAGS) $(SOURCES) $(LDFLAGS) -o $@

 endif  # PLATFORM_SHARED_EXT

-.PHONY: blackbox_crash_test check clean coverage crash_test ldb_tests \
+.PHONY: blackbox_crash_test check clean coverage crash_test ldb_tests package \
 	release tags valgrind_check whitebox_crash_test format static_lib shared_lib all \
-	dbg
+	dbg rocksdbjavastatic rocksdbjava install uninstall

 all: $(LIBRARY) $(PROGRAMS) $(TESTS)

@@ -225,11 +265,18 @@ valgrind_check: all $(PROGRAMS) $(TESTS)
 		echo $$t $$((etime - stime)) >> $(VALGRIND_DIR)/valgrind_tests_times; \
 	done

+unity.cc:
+	$(shell (export ROCKSDB_ROOT="$(CURDIR)"; "$(CURDIR)/build_tools/unity" "$(CURDIR)/unity.cc"))
+
+unity: unity.cc unity.o
+	$(CXX) unity.o $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)
+
 clean:
-	-rm -f $(PROGRAMS) $(TESTS) $(LIBRARY) $(SHARED) $(MEMENVLIBRARY) build_config.mk
+	-rm -f $(PROGRAMS) $(TESTS) $(LIBRARY) $(SHARED) $(MEMENVLIBRARY) build_config.mk unity.cc
 	-rm -rf ios-x86/* ios-arm/*
-	-find . -name "*.[od]" -exec rm {} \;
+	-find . -name "*.[oda]" -exec rm {} \;
 	-find . -type f -regex ".*\.\(\(gcda\)\|\(gcno\)\)" -exec rm {} \;
+	-rm -rf bzip2* snappy* zlib*
 tags:
 	ctags * -R
 	cscope -b `find . -name '*.cc'` `find . -name '*.h'`
@@ -237,6 +284,9 @@ tags:
 format:
 	build_tools/format-diff.sh

+package:
+	bash build_tools/make_package.sh $(SHARED_MAJOR).$(SHARED_MINOR)
+
 # ---------------------------------------------------------------------------
 # 	Unit tests and tools
 # ---------------------------------------------------------------------------
@@ -247,6 +297,9 @@ $(LIBRARY): $(LIBOBJECTS)
 db_bench: db/db_bench.o $(LIBOBJECTS) $(TESTUTIL)
 	$(CXX) db/db_bench.o $(LIBOBJECTS) $(TESTUTIL) $(EXEC_LDFLAGS) -o $@  $(LDFLAGS) $(COVERAGEFLAGS)

+cache_bench: util/cache_bench.o $(LIBOBJECTS) $(TESTUTIL)
+	$(CXX) util/cache_bench.o $(LIBOBJECTS) $(TESTUTIL) $(EXEC_LDFLAGS) -o $@  $(LDFLAGS) $(COVERAGEFLAGS)
+
 block_hash_index_test: table/block_hash_index_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	 $(CXX) table/block_hash_index_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

@@ -319,15 +372,15 @@ crc32c_test: util/crc32c_test.o $(LIBOBJECTS) $(TESTHARNESS)
 db_test: db/db_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(CXX) db/db_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

+db_iter_test: db/db_iter_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) db/db_iter_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)
+
 log_write_bench: util/log_write_bench.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(CXX) util/log_write_bench.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS) -pg

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

-simple_table_db_test: db/simple_table_db_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/simple_table_db_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)
-
 table_reader_bench: table/table_reader_bench.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(CXX) table/table_reader_bench.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS) -pg

@@ -343,20 +396,38 @@ prefix_test: db/prefix_test.o $(LIBOBJECTS) $(TESTHARNESS)
 backupable_db_test: utilities/backupable/backupable_db_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(CXX) utilities/backupable/backupable_db_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@  $(LDFLAGS) $(COVERAGEFLAGS)

+document_db_test: utilities/document/document_db_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) utilities/document/document_db_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@  $(LDFLAGS) $(COVERAGEFLAGS)
+
+json_document_test: utilities/document/json_document_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) utilities/document/json_document_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@  $(LDFLAGS) $(COVERAGEFLAGS)
+
+spatial_db_test: utilities/spatialdb/spatial_db_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) utilities/spatialdb/spatial_db_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@  $(LDFLAGS) $(COVERAGEFLAGS)
+
 ttl_test: utilities/ttl/ttl_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(CXX) utilities/ttl/ttl_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@  $(LDFLAGS) $(COVERAGEFLAGS)

+write_batch_with_index_test: utilities/write_batch_with_index/write_batch_with_index_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) utilities/write_batch_with_index/write_batch_with_index_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) $(COVERAGEFLAGS)

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

+rate_limiter_test: util/rate_limiter_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) util/rate_limiter_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) $(COVERAGEFLAGS)

-filter_block_test: table/filter_block_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) table/filter_block_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)
+block_based_filter_block_test: table/block_based_filter_block_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) table/block_based_filter_block_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)
+
+full_filter_block_test: table/full_filter_block_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) table/full_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) $(COVERAGEFLAGS)
@@ -385,15 +456,33 @@ reduce_levels_test: tools/reduce_levels_test.o $(LIBOBJECTS) $(TESTHARNESS)
 write_batch_test: db/write_batch_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(CXX) db/write_batch_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

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

+merger_test: table/merger_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) table/merger_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)

 geodb_test: utilities/geodb/geodb_test.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(CXX) utilities/geodb/geodb_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)

+cuckoo_table_builder_test: table/cuckoo_table_builder_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) table/cuckoo_table_builder_test.o  $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)
+
+cuckoo_table_reader_test: table/cuckoo_table_reader_test.o $(LIBOBJECTS) $(TESTHARNESS) $(BENCHHARNESS)
+	$(CXX) table/cuckoo_table_reader_test.o  $(LIBOBJECTS) $(TESTHARNESS) $(BENCHHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)
+
+cuckoo_table_db_test: db/cuckoo_table_db_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) db/cuckoo_table_db_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)
+
+options_test: util/options_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) util/options_test.o $(LIBOBJECTS) $(TESTHARNESS) $(EXEC_LDFLAGS) -o $@ $(LDFLAGS) $(COVERAGEFLAGS)
+
 $(MEMENVLIBRARY) : $(MEMENVOBJECTS)
 	rm -f $@
 	$(AR) -rs $@ $(MEMENVOBJECTS)
@@ -428,18 +517,69 @@ ldb: tools/ldb.o $(LIBOBJECTS)

 JNI_NATIVE_SOURCES = ./java/rocksjni/*.cc
 JAVA_INCLUDE = -I$(JAVA_HOME)/include/ -I$(JAVA_HOME)/include/linux
-ROCKSDBJNILIB = ./java/librocksdbjni.so
+ARCH := $(shell getconf LONG_BIT)
+ROCKSDBJNILIB = librocksdbjni-linux$(ARCH).so
+ROCKSDB_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-linux$(ARCH).jar
+ROCKSDB_JAR_ALL = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH).jar
+ROCKSDB_JAVADOCS_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-javadoc.jar
+ROCKSDB_SOURCES_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-sources.jar

 ifeq ($(PLATFORM), OS_MACOSX)
-ROCKSDBJNILIB = ./java/librocksdbjni.jnilib
+ROCKSDBJNILIB = librocksdbjni-osx.jnilib
+ROCKSDB_JAR = rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-osx.jar
 JAVA_INCLUDE = -I/System/Library/Frameworks/JavaVM.framework/Headers/
 endif

-rocksdbjava: clean
+libz.a:
+	-rm -rf zlib-1.2.8
+	curl -O http://zlib.net/zlib-1.2.8.tar.gz
+	tar xvzf zlib-1.2.8.tar.gz
+	cd zlib-1.2.8 && CFLAGS='-fPIC' ./configure --static && make
+	cp zlib-1.2.8/libz.a .
+
+libbz2.a:
+	-rm -rf bzip2-1.0.6
+	curl -O  http://www.bzip.org/1.0.6/bzip2-1.0.6.tar.gz
+	tar xvzf bzip2-1.0.6.tar.gz
+	cd bzip2-1.0.6 && make CFLAGS='-fPIC -Wall -Winline -O2 -g -D_FILE_OFFSET_BITS=64'
+	cp bzip2-1.0.6/libbz2.a .
+
+libsnappy.a:
+	-rm -rf snappy-1.1.1
+	curl -O https://snappy.googlecode.com/files/snappy-1.1.1.tar.gz
+	tar xvzf snappy-1.1.1.tar.gz
+	cd snappy-1.1.1 && ./configure --with-pic --enable-static
+	cd snappy-1.1.1 && make
+	cp snappy-1.1.1/.libs/libsnappy.a .
+
+
+rocksdbjavastatic: libz.a libbz2.a libsnappy.a
+	OPT="-fPIC -DNDEBUG -O2" $(MAKE) $(LIBRARY) -j
+	cd java;$(MAKE) java;
+	rm -f ./java/$(ROCKSDBJNILIB)
+	$(CXX) $(CXXFLAGS) -I./java/. $(JAVA_INCLUDE) -shared -fPIC -o ./java/$(ROCKSDBJNILIB) $(JNI_NATIVE_SOURCES) $(LIBOBJECTS) $(COVERAGEFLAGS) libz.a libbz2.a libsnappy.a
+	cd java;jar -cf $(ROCKSDB_JAR) org/rocksdb/*.class org/rocksdb/util/*.class HISTORY*.md $(ROCKSDBJNILIB)
+	cd java/javadoc;jar -cf ../$(ROCKSDB_JAVADOCS_JAR) *
+	cd java;jar -cf $(ROCKSDB_SOURCES_JAR) org
+
+rocksdbjavastaticrelease: rocksdbjavastatic
+	cd java/crossbuild && vagrant destroy -f && vagrant up linux32 && vagrant halt linux32 && vagrant up linux64 && vagrant halt linux64
+	cd java;jar -cf $(ROCKSDB_JAR_ALL) org/rocksdb/*.class org/rocksdb/util/*.class HISTORY*.md librocksdbjni-*.so librocksdbjni-*.jnilib
+
+rocksdbjavastaticpublish: rocksdbjavastaticrelease
+	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-javadoc.jar -Dclassifier=javadoc
+	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-sources.jar -Dclassifier=sources
+	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-linux64.jar -Dclassifier=linux64
+	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-linux32.jar -Dclassifier=linux32
+	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH)-osx.jar -Dclassifier=osx
+	mvn gpg:sign-and-deploy-file -Durl=https://oss.sonatype.org/service/local/staging/deploy/maven2/ -DrepositoryId=sonatype-nexus-staging -DpomFile=java/rocksjni.pom -Dfile=java/rocksdbjni-$(ROCKSDB_MAJOR).$(ROCKSDB_MINOR).$(ROCKSDB_PATCH).jar
+
+rocksdbjava:
 	OPT="-fPIC -DNDEBUG -O2" $(MAKE) $(LIBRARY) -j32
 	cd java;$(MAKE) java;
-	rm -f $(ROCKSDBJNILIB)
-	$(CXX) $(CXXFLAGS) -I./java/. $(JAVA_INCLUDE) -shared -fPIC -o $(ROCKSDBJNILIB) $(JNI_NATIVE_SOURCES) $(LIBOBJECTS) $(LDFLAGS) $(COVERAGEFLAGS)
+	rm -f ./java/$(ROCKSDBJNILIB)
+	$(CXX) $(CXXFLAGS) -I./java/. $(JAVA_INCLUDE) -shared -fPIC -o ./java/$(ROCKSDBJNILIB) $(JNI_NATIVE_SOURCES) $(LIBOBJECTS) $(JAVA_LDFLAGS) $(COVERAGEFLAGS)
+	cd java;jar -cf $(ROCKSDB_JAR) org/rocksdb/*.class org/rocksdb/util/*.class HISTORY*.md $(ROCKSDBJNILIB)

 jclean:
 	cd java;$(MAKE) clean;
@@ -516,8 +656,10 @@ ifneq ($(MAKECMDGOALS),clean)
 ifneq ($(MAKECMDGOALS),format)
 ifneq ($(MAKECMDGOALS),jclean)
 ifneq ($(MAKECMDGOALS),jtest)
+ifneq ($(MAKECMDGOALS),package)
 -include $(DEPFILES)
 endif
 endif
 endif
 endif
+endif
@@ -3,12 +3,12 @@
 [![Build Status](https://travis-ci.org/facebook/rocksdb.svg?branch=master)](https://travis-ci.org/facebook/rocksdb)

 RocksDB is developed and maintained by Facebook Database Engineering Team.
-It is built on on earlier work on LevelDB by Sanjay Ghemawat (sanjay@google.com)
+It is built 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.
-It has an Log-Structured-Merge-Database (LSM) design with flexible tradeoffs
+It has a Log-Structured-Merge-Database (LSM) design with flexible tradeoffs
 between Write-Amplification-Factor (WAF), Read-Amplification-Factor (RAF)
 and Space-Amplification-Factor (SAF). It has multi-threaded compactions,
 making it specially suitable for storing multiple terabytes of data in a
@@ -16,8 +16,7 @@ single database.

 Start with example usage here: https://github.com/facebook/rocksdb/tree/master/examples

-See [doc/index.html](https://github.com/facebook/rocksdb/blob/master/doc/index.html) and
-[github wiki](https://github.com/facebook/rocksdb/wiki) for more explanation.
+See the [github wiki](https://github.com/facebook/rocksdb/wiki) for more explanation.

 The public interface is in `include/`.  Callers should not include or
 rely on the details of any other header files in this package.  Those
@@ -0,0 +1,16 @@
+Vagrant.configure("2") do |config|
+
+  config.vm.provider "virtualbox" do |v|
+    v.memory = 4096
+    v.cpus = 2
+  end
+
+  config.vm.define "ubuntu14" do |box|
+    box.vm.box = "ubuntu/trusty64"
+  end
+
+  config.vm.define "centos65" do |box|
+    box.vm.box = "chef/centos-6.5"
+  end
+
+end
@@ -7,6 +7,7 @@
 #   CC                          C Compiler path
 #   CXX                         C++ Compiler path
 #   PLATFORM_LDFLAGS            Linker flags
+#   JAVA_LDFLAGS                Linker flags for RocksDBJava
 #   PLATFORM_SHARED_EXT         Extension for shared libraries
 #   PLATFORM_SHARED_LDFLAGS     Flags for building shared library
 #   PLATFORM_SHARED_CFLAGS      Flags for compiling objects for shared library
@@ -21,6 +22,7 @@
 #       -DLEVELDB_PLATFORM_NOATOMIC if it is not
 #       -DSNAPPY                    if the Snappy library is present
 #       -DLZ4                       if the LZ4 library is present
+#       -DNUMA                      if the NUMA library is present
 #
 # Using gflags in rocksdb:
 # Our project depends on gflags, which requires users to take some extra steps
@@ -44,16 +46,10 @@ PLATFORM_CXXFLAGS="-std=c++11"
 COMMON_FLAGS="-DROCKSDB_PLATFORM_POSIX"

 # Default to fbcode gcc on internal fb machines
-if [ -d /mnt/gvfs/third-party -a -z "$CXX" ]; then
+if [ -z "$ROCKSDB_NO_FBCODE" -a -d /mnt/gvfs/third-party ]; then
    FBCODE_BUILD="true"
    if [ -z "$USE_CLANG" ]; then
-        CENTOS_VERSION=`rpm -q --qf "%{VERSION}" \
-          $(rpm -q --whatprovides redhat-release)`
-        if [ "$CENTOS_VERSION" = "6" ]; then
-          source "$PWD/build_tools/fbcode.gcc481.sh"
-        else
-          source "$PWD/build_tools/fbcode.gcc471.sh"
-        fi
+        source $PWD/build_tools/fbcode_config.sh
    else
        source "$PWD/build_tools/fbcode.clang31.sh"
    fi
@@ -156,6 +152,8 @@ case "$TARGET_OS" in
        exit 1
 esac

+JAVA_LDFLAGS="$PLATFORM_LDFLAGS"
+
 if test -z "$DO_NOT_RUN_BUILD_DETECT_VERSION"; then
  "$PWD/build_tools/build_detect_version"
 fi
@@ -214,6 +212,7 @@ EOF
    if [ "$?" = 0 ]; then
        COMMON_FLAGS="$COMMON_FLAGS -DSNAPPY"
        PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -lsnappy"
+        JAVA_LDFLAGS="$JAVA_LDFLAGS -lsnappy"
    fi


@@ -237,7 +236,7 @@ EOF
      int main() {}
 EOF
    if [ "$?" = 0 ]; then
-        COMMON_FLAGS="$COMMON_FLAGS -DGFLAGS=google"
+        COMMON_FLAGS="$COMMON_FLAGS -DGFLAGS=gflags"
        PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -lgflags"
    fi

@@ -249,6 +248,7 @@ EOF
    if [ "$?" = 0 ]; then
        COMMON_FLAGS="$COMMON_FLAGS -DZLIB"
        PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -lz"
+        JAVA_LDFLAGS="$JAVA_LDFLAGS -lz"
    fi

    # Test whether bzip library is installed
@@ -259,6 +259,7 @@ EOF
    if [ "$?" = 0 ]; then
        COMMON_FLAGS="$COMMON_FLAGS -DBZIP2"
        PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -lbz2"
+        JAVA_LDFLAGS="$JAVA_LDFLAGS -lbz2"
    fi

    # Test whether lz4 library is installed
@@ -270,6 +271,19 @@ EOF
    if [ "$?" = 0 ]; then
        COMMON_FLAGS="$COMMON_FLAGS -DLZ4"
        PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -llz4"
+        JAVA_LDFLAGS="$JAVA_LDFLAGS -llz4"
+    fi
+
+    # Test whether numa is available
+    $CXX $CFLAGS -x c++ - -o /dev/null -lnuma 2>/dev/null  <<EOF
+      #include <numa.h>
+      #inlcude <numaif.h>
+      int main() {}
+EOF
+    if [ "$?" = 0 ]; then
+        COMMON_FLAGS="$COMMON_FLAGS -DNUMA"
+        PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -lnuma"
+        JAVA_LDFLAGS="$JAVA_LDFLAGS -lnuma"
    fi

    # Test whether tcmalloc is available
@@ -278,6 +292,7 @@ EOF
 EOF
    if [ "$?" = 0 ]; then
        PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -ltcmalloc"
+        JAVA_LDFLAGS="$JAVA_LDFLAGS -ltcmalloc"
    fi
 fi

@@ -294,6 +309,7 @@ if test "$USE_HDFS"; then
  HDFS_LDFLAGS="$HDFS_LDFLAGS -ldl -lverify -ljava -ljvm"
  COMMON_FLAGS="$COMMON_FLAGS $HDFS_CCFLAGS"
  PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS $HDFS_LDFLAGS"
+  JAVA_LDFLAGS="$JAVA_LDFLAGS $HDFS_LDFLAGS"
 fi

 # if Intel SSE instruction set is supported, set USE_SSE=" -msse -msse4.2 "
@@ -304,10 +320,15 @@ PLATFORM_CXXFLAGS="$PLATFORM_CXXFLAGS $COMMON_FLAGS"

 VALGRIND_VER="$VALGRIND_VER"

+ROCKSDB_MAJOR=`build_tools/version.sh major`
+ROCKSDB_MINOR=`build_tools/version.sh minor`
+ROCKSDB_PATCH=`build_tools/version.sh patch`
+
 echo "CC=$CC" >> "$OUTPUT"
 echo "CXX=$CXX" >> "$OUTPUT"
 echo "PLATFORM=$PLATFORM" >> "$OUTPUT"
 echo "PLATFORM_LDFLAGS=$PLATFORM_LDFLAGS" >> "$OUTPUT"
+echo "JAVA_LDFLAGS=$JAVA_LDFLAGS" >> "$OUTPUT"
 echo "VALGRIND_VER=$VALGRIND_VER" >> "$OUTPUT"
 echo "PLATFORM_CCFLAGS=$PLATFORM_CCFLAGS" >> "$OUTPUT"
 echo "PLATFORM_CXXFLAGS=$PLATFORM_CXXFLAGS" >> "$OUTPUT"
@@ -318,3 +339,6 @@ echo "PLATFORM_SHARED_VERSIONED=$PLATFORM_SHARED_VERSIONED" >> "$OUTPUT"
 echo "EXEC_LDFLAGS=$EXEC_LDFLAGS" >> "$OUTPUT"
 echo "JEMALLOC_INCLUDE=$JEMALLOC_INCLUDE" >> "$OUTPUT"
 echo "JEMALLOC_LIB=$JEMALLOC_LIB" >> "$OUTPUT"
+echo "ROCKSDB_MAJOR=$ROCKSDB_MAJOR" >> "$OUTPUT"
+echo "ROCKSDB_MINOR=$ROCKSDB_MINOR" >> "$OUTPUT"
+echo "ROCKSDB_PATCH=$ROCKSDB_PATCH" >> "$OUTPUT"
@@ -0,0 +1,19 @@
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
+GCC_BASE=/mnt/gvfs/third-party2/gcc/7331085db891a2ef4a88a48a751d834e8d68f4cb/7.x/centos7-native/b2ef2b6
+CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/963d9aeda70cc4779885b1277484fe7544a04e3e/9.0.0/platform007/9e92d53/
+LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/6ace84e956873d53638c738b6f65f3f469cca74c/7.x/platform007/5620abc
+GLIBC_BASE=/mnt/gvfs/third-party2/glibc/192b0f42d63dcf6210d6ceae387b49af049e6e0c/2.26/platform007/f259413
+SNAPPY_BASE=/mnt/gvfs/third-party2/snappy/7f9bdaada18f59bc27ec2b0871eb8a6144343aef/1.1.3/platform007/ca4da3d
+ZLIB_BASE=/mnt/gvfs/third-party2/zlib/2d9f0b9a4274cc21f61272a9e89bdb859bce8f1f/1.2.8/platform007/ca4da3d
+BZIP2_BASE=/mnt/gvfs/third-party2/bzip2/dc49a21c5fceec6456a7a28a94dcd16690af1337/1.0.6/platform007/ca4da3d
+LZ4_BASE=/mnt/gvfs/third-party2/lz4/0f607f8fc442ea7d6b876931b1898bb573d5e5da/1.9.1/platform007/ca4da3d
+ZSTD_BASE=/mnt/gvfs/third-party2/zstd/ca22bc441a4eb709e9e0b1f9fec9750fed7b31c5/1.4.x/platform007/15a3614
+GFLAGS_BASE=/mnt/gvfs/third-party2/gflags/0b9929d2588991c65a57168bf88aff2db87c5d48/2.2.0/platform007/ca4da3d
+JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/c26f08f47ac35fc31da2633b7da92d6b863246eb/master/platform007/c26c002
+NUMA_BASE=/mnt/gvfs/third-party2/numa/3f3fb57a5ccc5fd21c66416c0b83e0aa76a05376/2.0.11/platform007/ca4da3d
+LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/40c73d874898b386a71847f1b99115d93822d11f/1.4/platform007/6f3e0a9
+TBB_BASE=/mnt/gvfs/third-party2/tbb/4ce8e8dba77cdbd81b75d6f0c32fd7a1b76a11ec/2018_U5/platform007/ca4da3d
+KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/fb251ecd2f5ae16f8671f7014c246e52a748fe0b/fb/platform007/da39a3e
+BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/ab9f09bba370e7066cafd4eb59752db93f2e8312/2.29.1/platform007/15a3614
+VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/d42d152a15636529b0861ec493927200ebebca8e/3.15.0/platform007/ca4da3d
+LUA_BASE=/mnt/gvfs/third-party2/lua/f0cd714433206d5139df61659eb7b28b1dea6683/5.3.4/platform007/5007832
@@ -55,22 +55,27 @@ GFLAGS_LIBS=" $TOOLCHAIN_LIB_BASE/gflags/gflags-1.6/c3f970a/lib/libgflags.a"
 JEMALLOC_INCLUDE=" -I $TOOLCHAIN_LIB_BASE/jemalloc/jemalloc-3.4.1/4d53c6f/include/"
 JEMALLOC_LIB=" -Wl,--whole-archive $TOOLCHAIN_LIB_BASE/jemalloc/jemalloc-3.4.1/4d53c6f/lib/libjemalloc.a"

+# location of numa
+NUMA_REV=829d10dac0230f99cd7e1778869d2adf3da24b65
+NUMA_INCLUDE=" -I /mnt/gvfs/third-party2/numa/$NUMA_REV/2.0.8/gcc-4.8.1-glibc-2.17/c3f970a/include/"
+NUMA_LIB=" /mnt/gvfs/third-party2/numa/$NUMA_REV/2.0.8/gcc-4.8.1-glibc-2.17/c3f970a/lib/libnuma.a"
+
 # use Intel SSE support for checksum calculations
 export USE_SSE=" -msse -msse4.2 "

 CC="$TOOLCHAIN_EXECUTABLES/gcc/gcc-4.8.1/cc6c9dc/bin/gcc"
-CXX="$TOOLCHAIN_EXECUTABLES/gcc/gcc-4.8.1/cc6c9dc/bin/g++ $JINCLUDE $SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $GFLAGS_INCLUDE"
+CXX="$TOOLCHAIN_EXECUTABLES/gcc/gcc-4.8.1/cc6c9dc/bin/g++ $JINCLUDE $SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE"
 AR=$TOOLCHAIN_EXECUTABLES/binutils/binutils-2.21.1/da39a3e/bin/ar
 RANLIB=$TOOLCHAIN_EXECUTABLES/binutils/binutils-2.21.1/da39a3e/bin/ranlib

 CFLAGS="-B$TOOLCHAIN_EXECUTABLES/binutils/binutils-2.21.1/da39a3e/bin/gold -m64 -mtune=generic"
 CFLAGS+=" $LIBGCC_INCLUDE $GLIBC_INCLUDE"
 CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_ATOMIC_PRESENT -DROCKSDB_FALLOCATE_PRESENT"
-CFLAGS+=" -DSNAPPY -DGFLAGS=google -DZLIB -DBZIP2 -DLZ4"
+CFLAGS+=" -DSNAPPY -DGFLAGS=google -DZLIB -DBZIP2 -DLZ4 -DNUMA"

 EXEC_LDFLAGS="-Wl,--dynamic-linker,/usr/local/fbcode/gcc-4.8.1-glibc-2.17/lib/ld.so"
 EXEC_LDFLAGS+=" -Wl,--no-whole-archive $TOOLCHAIN_LIB_BASE/libunwind/libunwind-1.0.1/675d945/lib/libunwind.a"
-EXEC_LDFLAGS+=" $HDFSLIB $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $GFLAGS_LIBS"
+EXEC_LDFLAGS+=" $HDFSLIB $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $GFLAGS_LIBS $NUMA_LIB"

 PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS "

@@ -0,0 +1,134 @@
+#!/bin/sh
+#
+# Set environment variables so that we can compile rocksdb using
+# fbcode settings.  It uses the latest g++ and clang compilers and also
+# uses jemalloc
+# Environment variables that change the behavior of this script:
+# PIC_BUILD -- if true, it will only take pic versions of libraries from fbcode. libraries that don't have pic variant will not be included
+
+
+BASEDIR=`dirname $BASH_SOURCE`
+source "$BASEDIR/dependencies.sh"
+
+CFLAGS=""
+
+# libgcc
+LIBGCC_INCLUDE="$LIBGCC_BASE/include/c++/7.3.0"
+LIBGCC_LIBS=" -L $LIBGCC_BASE/lib"
+
+# glibc
+GLIBC_INCLUDE="$GLIBC_BASE/include"
+GLIBC_LIBS=" -L $GLIBC_BASE/lib"
+
+# snappy
+SNAPPY_INCLUDE=" -I $SNAPPY_BASE/include/"
+if test -z $PIC_BUILD; then
+  SNAPPY_LIBS=" $SNAPPY_BASE/lib/libsnappy.a"
+else
+  SNAPPY_LIBS=" $SNAPPY_BASE/lib/libsnappy_pic.a"
+fi
+CFLAGS+=" -DSNAPPY"
+
+if test -z $PIC_BUILD; then
+  # location of zlib headers and libraries
+  ZLIB_INCLUDE=" -I $ZLIB_BASE/include/"
+  ZLIB_LIBS=" $ZLIB_BASE/lib/libz.a"
+  CFLAGS+=" -DZLIB"
+
+  LZ4_INCLUDE=" -I $LZ4_BASE/include/"
+  LZ4_LIBS=" $LZ4_BASE/lib/liblz4.a"
+  CFLAGS+=" -DLZ4"
+fi
+
+# location of gflags headers and libraries
+GFLAGS_INCLUDE=" -I $GFLAGS_BASE/include/"
+if test -z $PIC_BUILD; then
+  GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags.a"
+else
+  GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags_pic.a"
+fi
+CFLAGS+=" -DGFLAGS=gflags"
+
+# location of jemalloc
+JEMALLOC_INCLUDE=" -I $JEMALLOC_BASE/include/"
+JEMALLOC_LIB=" $JEMALLOC_BASE/lib/libjemalloc.a"
+
+if test -z $PIC_BUILD; then
+  # location of libunwind
+  LIBUNWIND="$LIBUNWIND_BASE/lib/libunwind.a"
+fi
+
+# location of TBB
+TBB_INCLUDE=" -isystem $TBB_BASE/include/"
+if test -z $PIC_BUILD; then
+  TBB_LIBS="$TBB_BASE/lib/libtbb.a"
+else
+  TBB_LIBS="$TBB_BASE/lib/libtbb_pic.a"
+fi
+CFLAGS+=" -DTBB"
+
+# use Intel SSE support for checksum calculations
+export USE_SSE=" -msse -msse4.2 "
+
+BINUTILS="$BINUTILS_BASE/bin"
+AR="$BINUTILS/ar"
+
+DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $LZ4_INCLUDE $GFLAGS_INCLUDE"
+
+STDLIBS="-L $GCC_BASE/lib64"
+
+CLANG_BIN="$CLANG_BASE/bin"
+CLANG_LIB="$CLANG_BASE/lib"
+CLANG_SRC="$CLANG_BASE/../../src"
+
+CLANG_ANALYZER="$CLANG_BIN/clang++"
+CLANG_SCAN_BUILD="$CLANG_SRC/llvm/tools/clang/tools/scan-build/bin/scan-build"
+
+if [ -z "$USE_CLANG" ]; then
+  # gcc
+  CC="$GCC_BASE/bin/gcc"
+  CXX="$GCC_BASE/bin/g++"
+
+  CFLAGS+=" -B$BINUTILS/gold"
+  CFLAGS+=" -isystem $LIBGCC_INCLUDE"
+  CFLAGS+=" -isystem $GLIBC_INCLUDE"
+  JEMALLOC=1
+else
+  # clang
+  CLANG_INCLUDE="$CLANG_LIB/clang/stable/include"
+  CC="$CLANG_BIN/clang"
+  CXX="$CLANG_BIN/clang++"
+
+  KERNEL_HEADERS_INCLUDE="$KERNEL_HEADERS_BASE/include"
+
+  CFLAGS+=" -B$BINUTILS/gold -nostdinc -nostdlib"
+  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/7.x "
+  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/7.x/x86_64-facebook-linux "
+  CFLAGS+=" -isystem $GLIBC_INCLUDE"
+  CFLAGS+=" -isystem $LIBGCC_INCLUDE"
+  CFLAGS+=" -isystem $CLANG_INCLUDE"
+  CFLAGS+=" -isystem $KERNEL_HEADERS_INCLUDE/linux "
+  CFLAGS+=" -isystem $KERNEL_HEADERS_INCLUDE "
+  CFLAGS+=" -Wno-expansion-to-defined "
+  CXXFLAGS="-nostdinc++"
+fi
+
+CFLAGS+=" $DEPS_INCLUDE"
+CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE -DROCKSDB_RANGESYNC_PRESENT -DROCKSDB_SCHED_GETCPU_PRESENT -DROCKSDB_SUPPORT_THREAD_LOCAL -DHAVE_SSE42"
+CXXFLAGS+=" $CFLAGS"
+
+EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB $TBB_LIBS"
+EXEC_LDFLAGS+=" -B$BINUTILS/gold"
+EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/platform007/lib/ld.so"
+EXEC_LDFLAGS+=" $LIBUNWIND"
+EXEC_LDFLAGS+=" -Wl,-rpath=/usr/local/fbcode/platform007/lib"
+# required by libtbb
+EXEC_LDFLAGS+=" -ldl"
+
+PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS $STDLIBS -lgcc -lstdc++"
+
+EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $TBB_LIBS"
+
+VALGRIND_VER="$VALGRIND_BASE/bin/"
+
+export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE CLANG_ANALYZER CLANG_SCAN_BUILD 
@@ -22,4 +22,4 @@ echo ""
 echo "-----------------------------------------------------------------------------"
 echo "|                         Installation Completed                            |"
 echo "-----------------------------------------------------------------------------"
-echo "Please run `. ~/bash_profile` to be able to compile with gflags"
+echo "Please run \`. ~/.bash_profile\` to be able to compile with gflags"
@@ -0,0 +1,116 @@
+#/usr/bin/env bash
+
+set -e
+
+function log() {
+  echo "[+] $1"
+}
+
+function fatal() {
+  echo "[!] $1"
+  exit 1
+}
+
+function platform() {
+  local  __resultvar=$1
+  if [[ -f "/etc/yum.conf" ]]; then
+    eval $__resultvar="centos"
+  elif [[ -f "/etc/dpkg/dpkg.cfg" ]]; then
+    eval $__resultvar="ubuntu"
+  else
+    fatal "Unknwon operating system"
+  fi
+}
+platform OS
+
+function package() {
+  if [[ $OS = "ubuntu" ]]; then
+    if dpkg --get-selections | grep --quiet $1; then
+      log "$1 is already installed. skipping."
+    else
+      apt-get install $@ -y
+    fi
+  elif [[ $OS = "centos" ]]; then
+    if rpm -qa | grep --quiet $1; then
+      log "$1 is already installed. skipping."
+    else
+      yum install $@ -y
+    fi
+  fi
+}
+
+function detect_fpm_output() {
+  if [[ $OS = "ubuntu" ]]; then
+    export FPM_OUTPUT=deb
+  elif [[ $OS = "centos" ]]; then
+    export FPM_OUTPUT=rpm
+  fi
+}
+detect_fpm_output
+
+function gem_install() {
+  if gem list | grep --quiet $1; then
+    log "$1 is already installed. skipping."
+  else
+    gem install $@
+  fi
+}
+
+function main() {
+  if [[ $# -ne 1 ]]; then
+    fatal "Usage: $0 <rocksdb_version>"
+  else
+    log "using rocksdb version: $1"
+  fi
+
+  if [[ -d /vagrant ]]; then
+    if [[ $OS = "ubuntu" ]]; then
+      package g++-4.7
+      export CXX=g++-4.7
+
+      # the deb would depend on libgflags2, but the static lib is the only thing
+      # installed by make install
+      package libgflags-dev
+
+      package ruby-all-dev
+    elif [[ $OS = "centos" ]]; then
+      pushd /etc/yum.repos.d
+      if [[ ! -f /etc/yum.repos.d/devtools-1.1.repo ]]; then
+        wget http://people.centos.org/tru/devtools-1.1/devtools-1.1.repo
+      fi
+      package devtoolset-1.1-gcc --enablerepo=testing-1.1-devtools-6
+      package devtoolset-1.1-gcc-c++ --enablerepo=testing-1.1-devtools-6
+      export CC=/opt/centos/devtoolset-1.1/root/usr/bin/gcc
+      export CPP=/opt/centos/devtoolset-1.1/root/usr/bin/cpp
+      export CXX=/opt/centos/devtoolset-1.1/root/usr/bin/c++
+      export PATH=$PATH:/opt/centos/devtoolset-1.1/root/usr/bin
+      popd
+      if ! rpm -qa | grep --quiet gflags; then
+        rpm -i https://github.com/schuhschuh/gflags/releases/download/v2.1.0/gflags-devel-2.1.0-1.amd64.rpm
+      fi
+
+      package ruby
+      package ruby-devel
+      package rubygems
+      package rpm-build
+    fi
+  fi
+  gem_install fpm
+
+  make static_lib
+  make install INSTALL_PATH=package
+  fpm \
+    -s dir \
+    -t $FPM_OUTPUT \
+    -n rocksdb \
+    -v $1 \
+    --prefix /usr \
+    --url http://rocksdb.org/ \
+    -m rocksdb@fb.com \
+    --license BSD \
+    --vendor Facebook \
+    --description "RocksDB is an embeddable persistent key-value store for fast storage." \
+    package
+}
+
+main $@
@@ -109,7 +109,6 @@ make release
    --cache_numshardbits=6 \
    --table_cache_numshardbits=4 \
    --open_files=55000 \
-    --disable_seek_compaction=1 \
    --statistics=1 \
    --histogram=1 \
    --disable_data_sync=1 \
@@ -129,7 +128,6 @@ make release
    --cache_numshardbits=6 \
    --table_cache_numshardbits=4 \
    --open_files=55000 \
-    --disable_seek_compaction=1 \
    --use_tailing_iterator=1 \
    --statistics=1 \
    --histogram=1 \
@@ -150,7 +148,6 @@ make release
    --cache_numshardbits=6 \
    --table_cache_numshardbits=4 \
    --open_files=55000 \
-    --disable_seek_compaction=1 \
    --statistics=1 \
    --histogram=1 \
    --disable_data_sync=1 \
@@ -172,7 +169,6 @@ make release
    --table_cache_numshardbits=4 \
    --write_buffer_size=1000000000 \
    --open_files=55000 \
-    --disable_seek_compaction=1 \
    --statistics=1 \
    --histogram=1 \
    --disable_data_sync=1 \
@@ -231,7 +227,6 @@ make release
    --cache_numshardbits=6 \
    --table_cache_numshardbits=4 \
    --open_files=55000 \
-    --disable_seek_compaction=1 \
    --disable_auto_compactions=1 \
    --statistics=1 \
    --histogram=1 \
@@ -254,7 +249,6 @@ make release
    --cache_numshardbits=6 \
    --table_cache_numshardbits=4 \
    --open_files=55000 \
-    --disable_seek_compaction=1 \
    --statistics=1 \
    --histogram=1 \
    --disable_data_sync=1 \
@@ -274,7 +268,6 @@ make release
    --table_cache_numshardbits=4 \
    --write_buffer_size=1000000000 \
    --open_files=55000 \
-    --disable_seek_compaction=1 \
    --statistics=1 \
    --histogram=1 \
    --disable_data_sync=1 \
@@ -283,6 +276,107 @@ make release
    --value_size=10 \
    --threads=16 > ${STAT_FILE}.memtablefillreadrandom

+common_in_mem_args="--db=/dev/shm/rocksdb \
+    --num_levels=6 \
+    --key_size=20 \
+    --prefix_size=12 \
+    --keys_per_prefix=10 \
+    --value_size=100 \
+    --compression_type=none \
+    --compression_ratio=1 \
+    --hard_rate_limit=2 \
+    --write_buffer_size=134217728 \
+    --max_write_buffer_number=4 \
+    --level0_file_num_compaction_trigger=8 \
+    --level0_slowdown_writes_trigger=16 \
+    --level0_stop_writes_trigger=24 \
+    --target_file_size_base=134217728 \
+    --max_bytes_for_level_base=1073741824 \
+    --disable_wal=0 \
+    --wal_dir=/dev/shm/rocksdb \
+    --sync=0 \
+    --disable_data_sync=1 \
+    --verify_checksum=1 \
+    --delete_obsolete_files_period_micros=314572800 \
+    --max_grandparent_overlap_factor=10 \
+    --use_plain_table=1 \
+    --open_files=-1 \
+    --mmap_read=1 \
+    --mmap_write=0 \
+    --memtablerep=prefix_hash \
+    --bloom_bits=10 \
+    --bloom_locality=1 \
+    --perf_level=0"
+
+# prepare a in-memory DB with 50M keys, total DB size is ~6G
+./db_bench \
+    $common_in_mem_args \
+    --statistics=0 \
+    --max_background_compactions=16 \
+    --max_background_flushes=16 \
+    --benchmarks=filluniquerandom \
+    --use_existing_db=0 \
+    --num=52428800 \
+    --threads=1 > /dev/null
+
+# Readwhilewriting
+./db_bench \
+    $common_in_mem_args \
+    --statistics=1 \
+    --max_background_compactions=4 \
+    --max_background_flushes=0 \
+    --benchmarks=readwhilewriting\
+    --use_existing_db=1 \
+    --duration=600 \
+    --threads=32 \
+    --writes_per_second=81920 > ${STAT_FILE}.readwhilewriting_in_ram
+
+# Seekrandomwhilewriting
+./db_bench \
+    $common_in_mem_args \
+    --statistics=1 \
+    --max_background_compactions=4 \
+    --max_background_flushes=0 \
+    --benchmarks=seekrandomwhilewriting \
+    --use_existing_db=1 \
+    --use_tailing_iterator=1 \
+    --duration=600 \
+    --threads=32 \
+    --writes_per_second=81920 > ${STAT_FILE}.seekwhilewriting_in_ram
+
+# measure fillseq with bunch of column families
+./db_bench \
+    --benchmarks=fillseq \
+    --num_column_families=500 \
+    --write_buffer_size=1048576 \
+    --db=$DATA_DIR \
+    --use_existing_db=0 \
+    --num=$NUM \
+    --writes=$NUM \
+    --open_files=55000 \
+    --statistics=1 \
+    --histogram=1 \
+    --disable_data_sync=1 \
+    --disable_wal=1 \
+    --sync=0  > ${STAT_FILE}.fillseq_lots_column_families
+
+# measure overwrite performance with bunch of column families
+./db_bench \
+    --benchmarks=overwrite \
+    --num_column_families=500 \
+    --write_buffer_size=1048576 \
+    --db=$DATA_DIR \
+    --use_existing_db=1 \
+    --num=$NUM \
+    --writes=$((NUM / 10)) \
+    --open_files=55000 \
+    --statistics=1 \
+    --histogram=1 \
+    --disable_data_sync=1 \
+    --disable_wal=1 \
+    --sync=0 \
+    --threads=8 > ${STAT_FILE}.overwrite_lots_column_families
+
 # send data to ods
 function send_to_ods {
  key="$1"
@@ -308,9 +402,9 @@ function send_benchmark_to_ods {
  file="$3"

  QPS=$(grep $bench $file | awk '{print $5}')
-  P50_MICROS=$(grep $bench $file -A 4 | tail -n1 | awk '{print $3}' )
-  P75_MICROS=$(grep $bench $file -A 4 | tail -n1 | awk '{print $5}' )
-  P99_MICROS=$(grep $bench $file -A 4 | tail -n1 | awk '{print $7}' )
+  P50_MICROS=$(grep $bench $file -A 6 | grep "Percentiles" | awk '{print $3}' )
+  P75_MICROS=$(grep $bench $file -A 6 | grep "Percentiles" | awk '{print $5}' )
+  P99_MICROS=$(grep $bench $file -A 6 | grep "Percentiles" | awk '{print $7}' )

  send_to_ods rocksdb.build.$bench_key.qps $QPS
  send_to_ods rocksdb.build.$bench_key.p50_micros $P50_MICROS
@@ -328,3 +422,7 @@ send_benchmark_to_ods readrandom readrandom_fillunique_random $STAT_FILE.readran
 send_benchmark_to_ods fillrandom memtablefillrandom $STAT_FILE.memtablefillreadrandom
 send_benchmark_to_ods readrandom memtablereadrandom $STAT_FILE.memtablefillreadrandom
 send_benchmark_to_ods readwhilewriting readwhilewriting $STAT_FILE.readwhilewriting
+send_benchmark_to_ods readwhilewriting readwhilewriting_in_ram ${STAT_FILE}.readwhilewriting_in_ram
+send_benchmark_to_ods seekrandomwhilewriting seekwhilewriting_in_ram ${STAT_FILE}.seekwhilewriting_in_ram
+send_benchmark_to_ods fillseq fillseq_lots_column_families ${STAT_FILE}.fillseq_lots_column_families
+send_benchmark_to_ods overwrite overwrite_lots_column_families ${STAT_FILE}.overwrite_lots_column_families
@@ -0,0 +1,78 @@
+#!/bin/sh
+#
+# Create the unity file
+#
+
+OUTPUT=$1
+if test -z "$OUTPUT"; then
+  echo "usage: $0 <output-filename>" >&2
+  exit 1
+fi
+
+# Delete existing file, if it exists
+rm -f "$OUTPUT"
+touch "$OUTPUT"
+
+# Detect OS
+if test -z "$TARGET_OS"; then
+    TARGET_OS=`uname -s`
+fi
+
+# generic port files (working on all platform by #ifdef) go directly in /port
+GENERIC_PORT_FILES=`cd "$ROCKSDB_ROOT"; find port -name '*.cc' | tr "\n" " "`
+
+# On GCC, we pick libc's memcmp over GCC's memcmp via -fno-builtin-memcmp
+case "$TARGET_OS" in
+    Darwin)
+        # PORT_FILES=port/darwin/darwin_specific.cc
+        ;;
+    IOS)
+        ;;
+    Linux)
+        # PORT_FILES=port/linux/linux_specific.cc
+        ;;
+    SunOS)
+        # PORT_FILES=port/sunos/sunos_specific.cc
+        ;;
+    FreeBSD)
+        # PORT_FILES=port/freebsd/freebsd_specific.cc
+        ;;
+    NetBSD)
+        # PORT_FILES=port/netbsd/netbsd_specific.cc
+        ;;
+    OpenBSD)
+        # PORT_FILES=port/openbsd/openbsd_specific.cc
+        ;;
+    DragonFly)
+        # PORT_FILES=port/dragonfly/dragonfly_specific.cc
+        ;;
+    OS_ANDROID_CROSSCOMPILE)
+        # PORT_FILES=port/android/android.cc
+        ;;
+    *)
+        echo "Unknown platform!" >&2
+        exit 1
+esac
+
+# 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
+# of all files matching either rule, so we need to append -print to make the
+# prune take effect.
+DIRS="util db table utilities"
+
+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=`cd "$ROCKSDB_ROOT"; find $DIRS $PRUNE_TEST -o $PRUNE_BENCH -o -name '*.cc' -print | sort`
+PORTABLE_CPP=`cd "$ROCKSDB_ROOT"; find $DIRS $PRUNE_TEST -o $PRUNE_BENCH -o -name '*.cpp' -print | sort`
+set +f # re-enable globbing
+
+# The sources consist of the portable files, plus the platform-specific port
+# file.
+for SOURCE_FILE in $PORTABLE_FILES $GENERIC_PORT_FILES $PORT_FILES $PORTABLE_CPP
+do
+	echo "#include <$SOURCE_FILE>" >> "$OUTPUT"
+done
+
+echo "int main(int argc, char** argv){ return 0; }" >> "$OUTPUT"
+
@@ -0,0 +1,14 @@
+#!/bin/sh
+if [ "$#" = "0" ]; then
+  echo "Usage: $0 major|minor|patch"
+  exit 1
+fi
+if [ "$1" = "major" ]; then
+  cat include/rocksdb/version.h  | grep MAJOR | head -n1 | awk '{print $3}'
+fi
+if [ "$1" = "minor" ]; then
+  cat include/rocksdb/version.h  | grep MINOR | head -n1 | awk '{print $3}'
+fi
+if [ "$1" = "patch" ]; then
+  cat include/rocksdb/version.h  | grep PATCH | head -n1 | awk '{print $3}'
+fi
@@ -26,44 +26,51 @@ namespace rocksdb {

 class TableFactory;

-TableBuilder* NewTableBuilder(const Options& options,
+TableBuilder* NewTableBuilder(const ImmutableCFOptions& ioptions,
                              const InternalKeyComparator& internal_comparator,
                              WritableFile* file,
-                              CompressionType compression_type) {
-  return options.table_factory->NewTableBuilder(options, internal_comparator,
-                                                file, compression_type);
+                              const CompressionType compression_type,
+                              const CompressionOptions& compression_opts) {
+  return ioptions.table_factory->NewTableBuilder(
+      ioptions, internal_comparator, file, compression_type, compression_opts);
 }

-Status BuildTable(const std::string& dbname, Env* env, const Options& options,
-                  const EnvOptions& soptions, TableCache* table_cache,
+Status BuildTable(const std::string& dbname, Env* env,
+                  const ImmutableCFOptions& ioptions,
+                  const EnvOptions& env_options, TableCache* table_cache,
                  Iterator* iter, FileMetaData* meta,
                  const InternalKeyComparator& internal_comparator,
                  const SequenceNumber newest_snapshot,
                  const SequenceNumber earliest_seqno_in_memtable,
-                  const CompressionType compression) {
+                  const CompressionType compression,
+                  const CompressionOptions& compression_opts,
+                  const Env::IOPriority io_priority) {
  Status s;
-  meta->file_size = 0;
+  meta->fd.file_size = 0;
  meta->smallest_seqno = meta->largest_seqno = 0;
  iter->SeekToFirst();

  // If the sequence number of the smallest entry in the memtable is
  // smaller than the most recent snapshot, then we do not trigger
  // removal of duplicate/deleted keys as part of this builder.
-  bool purge = options.purge_redundant_kvs_while_flush;
+  bool purge = ioptions.purge_redundant_kvs_while_flush;
  if (earliest_seqno_in_memtable <= newest_snapshot) {
    purge = false;
  }

-  std::string fname = TableFileName(dbname, meta->number);
+  std::string fname = TableFileName(ioptions.db_paths, meta->fd.GetNumber(),
+                                    meta->fd.GetPathId());
  if (iter->Valid()) {
    unique_ptr<WritableFile> file;
-    s = env->NewWritableFile(fname, &file, soptions);
+    s = env->NewWritableFile(fname, &file, env_options);
    if (!s.ok()) {
      return s;
    }
+    file->SetIOPriority(io_priority);

-    TableBuilder* builder =
-        NewTableBuilder(options, internal_comparator, file.get(), compression);
+    TableBuilder* builder = NewTableBuilder(
+        ioptions, internal_comparator, file.get(),
+        compression, compression_opts);

    // the first key is the smallest key
    Slice key = iter->key();
@@ -72,8 +79,8 @@ Status BuildTable(const std::string& dbname, Env* env, const Options& options,
    meta->largest_seqno = meta->smallest_seqno;

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

    if (purge) {
@@ -110,6 +117,13 @@ Status BuildTable(const std::string& dbname, Env* env, const Options& options,
          is_first_key = false;

          if (this_ikey.type == kTypeMerge) {
+            // TODO(tbd): Add a check here to prevent RocksDB from crash when
+            // reopening a DB w/o properly specifying the merge operator.  But
+            // currently we observed a memory leak on failing in RocksDB
+            // recovery, so we decide to let it crash instead of causing
+            // memory leak for now before we have identified the real cause
+            // of the memory leak.
+
            // Handle merge-type keys using the MergeHelper
            // TODO: pass statistics to MergeUntil
            merge.MergeUntil(iter, 0 /* don't worry about snapshot */);
@@ -177,8 +191,8 @@ Status BuildTable(const std::string& dbname, Env* env, const Options& options,
    if (s.ok()) {
      s = builder->Finish();
      if (s.ok()) {
-        meta->file_size = builder->FileSize();
-        assert(meta->file_size > 0);
+        meta->fd.file_size = builder->FileSize();
+        assert(meta->fd.GetFileSize() > 0);
      }
    } else {
      builder->Abandon();
@@ -186,12 +200,12 @@ Status BuildTable(const std::string& dbname, Env* env, const Options& options,
    delete builder;

    // Finish and check for file errors
-    if (s.ok() && !options.disableDataSync) {
-      if (options.use_fsync) {
-        StopWatch sw(env, options.statistics.get(), TABLE_SYNC_MICROS);
+    if (s.ok() && !ioptions.disable_data_sync) {
+      if (ioptions.use_fsync) {
+        StopWatch sw(env, ioptions.statistics, TABLE_SYNC_MICROS);
        s = file->Fsync();
      } else {
-        StopWatch sw(env, options.statistics.get(), TABLE_SYNC_MICROS);
+        StopWatch sw(env, ioptions.statistics, TABLE_SYNC_MICROS);
        s = file->Sync();
      }
    }
@@ -201,8 +215,8 @@ Status BuildTable(const std::string& dbname, Env* env, const Options& options,

    if (s.ok()) {
      // Verify that the table is usable
-      Iterator* it = table_cache->NewIterator(ReadOptions(), soptions,
-                                              internal_comparator, *meta);
+      Iterator* it = table_cache->NewIterator(ReadOptions(), env_options,
+                                              internal_comparator, meta->fd);
      s = it->status();
      delete it;
    }
@@ -213,7 +227,7 @@ Status BuildTable(const std::string& dbname, Env* env, const Options& options,
    s = iter->status();
  }

-  if (s.ok() && meta->file_size > 0) {
+  if (s.ok() && meta->fd.GetFileSize() > 0) {
    // Keep it
  } else {
    env->DeleteFile(fname);
@@ -7,9 +7,11 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 #pragma once
 #include "rocksdb/comparator.h"
+#include "rocksdb/env.h"
 #include "rocksdb/status.h"
 #include "rocksdb/types.h"
 #include "rocksdb/options.h"
+#include "rocksdb/immutable_options.h"

 namespace rocksdb {

@@ -25,21 +27,26 @@ class TableBuilder;
 class WritableFile;

 extern TableBuilder* NewTableBuilder(
-    const Options& options, const InternalKeyComparator& internal_comparator,
-    WritableFile* file, CompressionType compression_type);
+    const ImmutableCFOptions& options,
+    const InternalKeyComparator& internal_comparator,
+    WritableFile* file, const CompressionType compression_type,
+    const CompressionOptions& compression_opts);

 // Build a Table file from the contents of *iter.  The generated file
-// will be named according to meta->number.  On success, the rest of
+// will be named according to number specified in meta. On success, the rest of
 // *meta will be filled with metadata about the generated table.
 // If no data is present in *iter, meta->file_size will be set to
 // zero, and no Table file will be produced.
 extern Status BuildTable(const std::string& dbname, Env* env,
-                         const Options& options, const EnvOptions& soptions,
+                         const ImmutableCFOptions& options,
+                         const EnvOptions& env_options,
                         TableCache* table_cache, Iterator* iter,
                         FileMetaData* meta,
                         const InternalKeyComparator& internal_comparator,
                         const SequenceNumber newest_snapshot,
                         const SequenceNumber earliest_seqno_in_memtable,
-                         const CompressionType compression);
+                         const CompressionType compression,
+                         const CompressionOptions& compression_opts,
+                         const Env::IOPriority io_priority = Env::IO_HIGH);

 }  // namespace rocksdb
@@ -14,6 +14,7 @@
 #include <stdlib.h>
 #include <unistd.h>
 #include "rocksdb/cache.h"
+#include "rocksdb/compaction_filter.h"
 #include "rocksdb/comparator.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
@@ -30,9 +31,19 @@
 #include "rocksdb/table.h"

 using rocksdb::Cache;
+using rocksdb::ColumnFamilyDescriptor;
+using rocksdb::ColumnFamilyHandle;
+using rocksdb::ColumnFamilyOptions;
+using rocksdb::CompactionFilter;
+using rocksdb::CompactionFilterFactory;
+using rocksdb::CompactionFilterV2;
+using rocksdb::CompactionFilterFactoryV2;
+using rocksdb::CompactionFilterContext;
+using rocksdb::CompactionOptionsFIFO;
 using rocksdb::Comparator;
 using rocksdb::CompressionType;
 using rocksdb::DB;
+using rocksdb::DBOptions;
 using rocksdb::Env;
 using rocksdb::InfoLogLevel;
 using rocksdb::FileLock;
@@ -44,6 +55,7 @@ using rocksdb::MergeOperator;
 using rocksdb::NewBloomFilterPolicy;
 using rocksdb::NewLRUCache;
 using rocksdb::Options;
+using rocksdb::BlockBasedTableOptions;
 using rocksdb::RandomAccessFile;
 using rocksdb::Range;
 using rocksdb::ReadOptions;
@@ -66,9 +78,11 @@ struct rocksdb_iterator_t        { Iterator*         rep; };
 struct rocksdb_writebatch_t      { WriteBatch        rep; };
 struct rocksdb_snapshot_t        { const Snapshot*   rep; };
 struct rocksdb_flushoptions_t    { FlushOptions      rep; };
+struct rocksdb_fifo_compaction_options_t { CompactionOptionsFIFO rep; };
 struct rocksdb_readoptions_t     { ReadOptions       rep; };
 struct rocksdb_writeoptions_t    { WriteOptions      rep; };
 struct rocksdb_options_t         { Options           rep; };
+struct rocksdb_block_based_table_options_t  { BlockBasedTableOptions rep; };
 struct rocksdb_seqfile_t         { SequentialFile*   rep; };
 struct rocksdb_randomfile_t      { RandomAccessFile* rep; };
 struct rocksdb_writablefile_t    { WritableFile*     rep; };
@@ -76,6 +90,172 @@ struct rocksdb_filelock_t        { FileLock*         rep; };
 struct rocksdb_logger_t          { shared_ptr<Logger>  rep; };
 struct rocksdb_cache_t           { shared_ptr<Cache>   rep; };
 struct rocksdb_livefiles_t       { std::vector<LiveFileMetaData> rep; };
+struct rocksdb_column_family_handle_t  { ColumnFamilyHandle* rep; };
+
+struct rocksdb_compactionfiltercontext_t {
+  CompactionFilter::Context rep;
+};
+
+struct rocksdb_compactionfilter_t : public CompactionFilter {
+  void* state_;
+  void (*destructor_)(void*);
+  unsigned char (*filter_)(
+      void*,
+      int level,
+      const char* key, size_t key_length,
+      const char* existing_value, size_t value_length,
+      char** new_value, size_t *new_value_length,
+      unsigned char* value_changed);
+  const char* (*name_)(void*);
+
+  virtual ~rocksdb_compactionfilter_t() {
+    (*destructor_)(state_);
+  }
+
+  virtual bool Filter(
+      int level,
+      const Slice& key,
+      const Slice& existing_value,
+      std::string* new_value,
+      bool* value_changed) const {
+    char* c_new_value = nullptr;
+    size_t new_value_length = 0;
+    unsigned char c_value_changed = 0;
+    unsigned char result = (*filter_)(
+        state_,
+        level,
+        key.data(), key.size(),
+        existing_value.data(), existing_value.size(),
+        &c_new_value, &new_value_length, &c_value_changed);
+    if (c_value_changed) {
+      new_value->assign(c_new_value, new_value_length);
+      *value_changed = true;
+    }
+    return result;
+  }
+
+  virtual const char* Name() const {
+    return (*name_)(state_);
+  }
+};
+
+struct rocksdb_compactionfilterfactory_t : public CompactionFilterFactory {
+  void* state_;
+  void (*destructor_)(void*);
+  rocksdb_compactionfilter_t* (*create_compaction_filter_)(
+      void*, rocksdb_compactionfiltercontext_t* context);
+  const char* (*name_)(void*);
+
+  virtual ~rocksdb_compactionfilterfactory_t() { (*destructor_)(state_); }
+
+  virtual std::unique_ptr<CompactionFilter> CreateCompactionFilter(
+      const CompactionFilter::Context& context) {
+    rocksdb_compactionfiltercontext_t ccontext;
+    ccontext.rep = context;
+    CompactionFilter* cf = (*create_compaction_filter_)(state_, &ccontext);
+    return std::unique_ptr<CompactionFilter>(cf);
+  }
+
+  virtual const char* Name() const { return (*name_)(state_); }
+};
+
+struct rocksdb_compactionfilterv2_t : public CompactionFilterV2 {
+  void* state_;
+  void (*destructor_)(void*);
+  const char* (*name_)(void*);
+  void (*filter_)(void*, int level, size_t num_keys,
+                  const char* const* keys_list, const size_t* keys_list_sizes,
+                  const char* const* existing_values_list, const size_t* existing_values_list_sizes,
+                  char** new_values_list, size_t* new_values_list_sizes,
+                  unsigned char* to_delete_list);
+
+  virtual ~rocksdb_compactionfilterv2_t() {
+    (*destructor_)(state_);
+  }
+
+  virtual const char* Name() const {
+    return (*name_)(state_);
+  }
+
+  virtual std::vector<bool> Filter(int level,
+                                   const SliceVector& keys,
+                                   const SliceVector& existing_values,
+                                   std::vector<std::string>* new_values,
+                                   std::vector<bool>* values_changed) const {
+    // Make a vector pointing to the underlying key data.
+    size_t num_keys = keys.size();
+    std::vector<const char*> keys_list(num_keys);
+    std::vector<size_t> keys_list_sizes(num_keys);
+    for (size_t i = 0; i < num_keys; ++i) {
+      keys_list[i] = keys[i].data();
+      keys_list_sizes[i] = keys[i].size();
+    }
+    // Make a vector pointing to the underlying value data.
+    std::vector<const char*> existing_values_list(num_keys);
+    std::vector<size_t> existing_values_list_sizes(num_keys);
+    for (size_t i = 0; i < num_keys; ++i) {
+      existing_values_list[i] = existing_values[i].data();
+      existing_values_list_sizes[i] = existing_values[i].size();
+    }
+    // Make a vector which will accept newly-allocated char* arrays
+    // which we will take ownership of and assign to strings in new_values.
+    new_values->clear();
+    std::vector<char*> new_values_list(num_keys);
+    std::vector<size_t> new_values_list_sizes(num_keys);
+    // Resize values_changed to hold all keys.
+    values_changed->resize(num_keys);
+    // Make a vector for bools indicating a value should be deleted
+    // on compaction (true) or maintained (false).
+    std::vector<unsigned char> to_delete_list(num_keys);
+
+    (*filter_)(
+        state_, level, num_keys, &keys_list[0], &keys_list_sizes[0],
+        &existing_values_list[0], &existing_values_list_sizes[0],
+        &new_values_list[0], &new_values_list_sizes[0], &to_delete_list[0]);
+
+    // Now, we transfer any changed values, setting values_changed and
+    // initializing new_values in the event a value changed.
+    std::vector<bool> to_delete(num_keys);
+    for (size_t i = 0; i < num_keys; ++i) {
+      to_delete[i] = to_delete_list[i];
+      (*values_changed)[i] = new_values_list[i] != nullptr;
+      if ((*values_changed)[i]) {
+        new_values->push_back(std::string(new_values_list[i], new_values_list_sizes[i]));
+        free(new_values_list[i]);
+      }
+    }
+    return to_delete;
+  }
+};
+
+struct rocksdb_compactionfilterfactoryv2_t : public CompactionFilterFactoryV2 {
+  void* state_;
+  void (*destructor_)(void*);
+  const char* (*name_)(void*);
+  rocksdb_compactionfilterv2_t* (*create_compaction_filter_v2_)(
+      void* state, const rocksdb_compactionfiltercontext_t* context);
+
+  rocksdb_compactionfilterfactoryv2_t(const SliceTransform* prefix_extractor)
+      : CompactionFilterFactoryV2(prefix_extractor) {
+  }
+
+  virtual ~rocksdb_compactionfilterfactoryv2_t() {
+    (*destructor_)(state_);
+  }
+
+  virtual const char* Name() const {
+    return (*name_)(state_);
+  }
+
+  virtual std::unique_ptr<CompactionFilterV2> CreateCompactionFilterV2(
+      const CompactionFilterContext& context) {
+    struct rocksdb_compactionfiltercontext_t c_context;
+    c_context.rep.is_full_compaction = context.is_full_compaction;
+    c_context.rep.is_manual_compaction = context.is_manual_compaction;
+    return std::unique_ptr<CompactionFilterV2>(
+        (*create_compaction_filter_v2_)(state_, &c_context));
+  }
+};

 struct rocksdb_comparator_t : public Comparator {
  void* state_;
@@ -349,6 +529,120 @@ void rocksdb_close(rocksdb_t* db) {
  delete db;
 }

+rocksdb_t* rocksdb_open_column_families(
+    const rocksdb_options_t* db_options,
+    const char* name,
+    int num_column_families,
+    const char** column_family_names,
+    const rocksdb_options_t** column_family_options,
+    rocksdb_column_family_handle_t** column_family_handles,
+    char** errptr) {
+  std::vector<ColumnFamilyDescriptor> column_families;
+  for (int i = 0; i < num_column_families; i++) {
+    column_families.push_back(ColumnFamilyDescriptor(
+        std::string(column_family_names[i]),
+        ColumnFamilyOptions(column_family_options[i]->rep)));
+  }
+
+  DB* db;
+  std::vector<ColumnFamilyHandle*> handles;
+  if (SaveError(errptr, DB::Open(DBOptions(db_options->rep),
+          std::string(name), column_families, &handles, &db))) {
+    return nullptr;
+  }
+
+  for (size_t i = 0; i < handles.size(); i++) {
+    rocksdb_column_family_handle_t* c_handle = new rocksdb_column_family_handle_t;
+    c_handle->rep = handles[i];
+    column_family_handles[i] = c_handle;
+  }
+  rocksdb_t* result = new rocksdb_t;
+  result->rep = db;
+  return result;
+}
+
+rocksdb_t* rocksdb_open_for_read_only_column_families(
+    const rocksdb_options_t* db_options,
+    const char* name,
+    int num_column_families,
+    const char** column_family_names,
+    const rocksdb_options_t** column_family_options,
+    rocksdb_column_family_handle_t** column_family_handles,
+    unsigned char error_if_log_file_exist,
+    char** errptr) {
+  std::vector<ColumnFamilyDescriptor> column_families;
+  for (int i = 0; i < num_column_families; i++) {
+    column_families.push_back(ColumnFamilyDescriptor(
+        std::string(column_family_names[i]),
+        ColumnFamilyOptions(column_family_options[i]->rep)));
+  }
+
+  DB* db;
+  std::vector<ColumnFamilyHandle*> handles;
+  if (SaveError(errptr, DB::OpenForReadOnly(DBOptions(db_options->rep),
+          std::string(name), column_families, &handles, &db, error_if_log_file_exist))) {
+    return nullptr;
+  }
+
+  for (size_t i = 0; i < handles.size(); i++) {
+    rocksdb_column_family_handle_t* c_handle = new rocksdb_column_family_handle_t;
+    c_handle->rep = handles[i];
+    column_family_handles[i] = c_handle;
+  }
+  rocksdb_t* result = new rocksdb_t;
+  result->rep = db;
+  return result;
+}
+
+char** rocksdb_list_column_families(
+    const rocksdb_options_t* options,
+    const char* name,
+    size_t* lencfs,
+    char** errptr) {
+  std::vector<std::string> fams;
+  SaveError(errptr,
+      DB::ListColumnFamilies(DBOptions(options->rep),
+        std::string(name), &fams));
+
+  *lencfs = fams.size();
+  char** column_families = static_cast<char**>(malloc(sizeof(char*) * fams.size()));
+  for (size_t i = 0; i < fams.size(); i++) {
+    column_families[i] = strdup(fams[i].c_str());
+  }
+  return column_families;
+}
+
+void rocksdb_list_column_families_destroy(char** list, size_t len) {
+  for (size_t i = 0; i < len; ++i) {
+    free(list[i]);
+  }
+  free(list);
+}
+
+rocksdb_column_family_handle_t* rocksdb_create_column_family(
+    rocksdb_t* db,
+    const rocksdb_options_t* column_family_options,
+    const char* column_family_name,
+    char** errptr) {
+  rocksdb_column_family_handle_t* handle = new rocksdb_column_family_handle_t;
+  SaveError(errptr,
+      db->rep->CreateColumnFamily(ColumnFamilyOptions(column_family_options->rep),
+        std::string(column_family_name), &(handle->rep)));
+  return handle;
+}
+
+void rocksdb_drop_column_family(
+    rocksdb_t* db,
+    rocksdb_column_family_handle_t* handle,
+    char** errptr) {
+  SaveError(errptr, db->rep->DropColumnFamily(handle->rep));
+}
+
+void rocksdb_column_family_handle_destroy(rocksdb_column_family_handle_t* handle) {
+  delete handle->rep;
+  delete handle;
+}
+
 void rocksdb_put(
    rocksdb_t* db,
    const rocksdb_writeoptions_t* options,
@@ -359,6 +653,18 @@ void rocksdb_put(
            db->rep->Put(options->rep, Slice(key, keylen), Slice(val, vallen)));
 }

+void rocksdb_put_cf(
+    rocksdb_t* db,
+    const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t keylen,
+    const char* val, size_t vallen,
+    char** errptr) {
+  SaveError(errptr,
+            db->rep->Put(options->rep, column_family->rep,
+              Slice(key, keylen), Slice(val, vallen)));
+}
+
 void rocksdb_delete(
    rocksdb_t* db,
    const rocksdb_writeoptions_t* options,
@@ -367,6 +673,16 @@ void rocksdb_delete(
  SaveError(errptr, db->rep->Delete(options->rep, Slice(key, keylen)));
 }

+void rocksdb_delete_cf(
+    rocksdb_t* db,
+    const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t keylen,
+    char** errptr) {
+  SaveError(errptr, db->rep->Delete(options->rep, column_family->rep,
+        Slice(key, keylen)));
+}
+
 void rocksdb_merge(
    rocksdb_t* db,
    const rocksdb_writeoptions_t* options,
@@ -377,6 +693,18 @@ void rocksdb_merge(
            db->rep->Merge(options->rep, Slice(key, keylen), Slice(val, vallen)));
 }

+void rocksdb_merge_cf(
+    rocksdb_t* db,
+    const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t keylen,
+    const char* val, size_t vallen,
+    char** errptr) {
+  SaveError(errptr,
+            db->rep->Merge(options->rep, column_family->rep,
+              Slice(key, keylen), Slice(val, vallen)));
+}
+
 void rocksdb_write(
    rocksdb_t* db,
    const rocksdb_writeoptions_t* options,
@@ -406,6 +734,29 @@ char* rocksdb_get(
  return result;
 }

+char* rocksdb_get_cf(
+    rocksdb_t* db,
+    const rocksdb_readoptions_t* options,
+    rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t keylen,
+    size_t* vallen,
+    char** errptr) {
+  char* result = nullptr;
+  std::string tmp;
+  Status s = db->rep->Get(options->rep, column_family->rep,
+      Slice(key, keylen), &tmp);
+  if (s.ok()) {
+    *vallen = tmp.size();
+    result = CopyString(tmp);
+  } else {
+    *vallen = 0;
+    if (!s.IsNotFound()) {
+      SaveError(errptr, s);
+    }
+  }
+  return result;
+}
+
 rocksdb_iterator_t* rocksdb_create_iterator(
    rocksdb_t* db,
    const rocksdb_readoptions_t* options) {
@@ -414,6 +765,15 @@ rocksdb_iterator_t* rocksdb_create_iterator(
  return result;
 }

+rocksdb_iterator_t* rocksdb_create_iterator_cf(
+    rocksdb_t* db,
+    const rocksdb_readoptions_t* options,
+    rocksdb_column_family_handle_t* column_family) {
+  rocksdb_iterator_t* result = new rocksdb_iterator_t;
+  result->rep = db->rep->NewIterator(options->rep, column_family->rep);
+  return result;
+}
+
 const rocksdb_snapshot_t* rocksdb_create_snapshot(
    rocksdb_t* db) {
  rocksdb_snapshot_t* result = new rocksdb_snapshot_t;
@@ -440,6 +800,19 @@ char* rocksdb_property_value(
  }
 }

+char* rocksdb_property_value_cf(
+    rocksdb_t* db,
+    rocksdb_column_family_handle_t* column_family,
+    const char* propname) {
+  std::string tmp;
+  if (db->rep->GetProperty(column_family->rep, Slice(propname), &tmp)) {
+    // We use strdup() since we expect human readable output.
+    return strdup(tmp.c_str());
+  } else {
+    return nullptr;
+  }
+}
+
 void rocksdb_approximate_sizes(
    rocksdb_t* db,
    int num_ranges,
@@ -455,6 +828,22 @@ void rocksdb_approximate_sizes(
  delete[] ranges;
 }

+void rocksdb_approximate_sizes_cf(
+    rocksdb_t* db,
+    rocksdb_column_family_handle_t* column_family,
+    int num_ranges,
+    const char* const* range_start_key, const size_t* range_start_key_len,
+    const char* const* range_limit_key, const size_t* range_limit_key_len,
+    uint64_t* sizes) {
+  Range* ranges = new Range[num_ranges];
+  for (int i = 0; i < num_ranges; i++) {
+    ranges[i].start = Slice(range_start_key[i], range_start_key_len[i]);
+    ranges[i].limit = Slice(range_limit_key[i], range_limit_key_len[i]);
+  }
+  db->rep->GetApproximateSizes(column_family->rep, ranges, num_ranges, sizes);
+  delete[] ranges;
+}
+
 void rocksdb_delete_file(
    rocksdb_t* db,
    const char* name) {
@@ -479,6 +868,19 @@ void rocksdb_compact_range(
      (limit_key ? (b = Slice(limit_key, limit_key_len), &b) : nullptr));
 }

+void rocksdb_compact_range_cf(
+    rocksdb_t* db,
+    rocksdb_column_family_handle_t* column_family,
+    const char* start_key, size_t start_key_len,
+    const char* limit_key, size_t limit_key_len) {
+  Slice a, b;
+  db->rep->CompactRange(
+      column_family->rep,
+      // Pass nullptr Slice if corresponding "const char*" is nullptr
+      (start_key ? (a = Slice(start_key, start_key_len), &a) : nullptr),
+      (limit_key ? (b = Slice(limit_key, limit_key_len), &b) : nullptr));
+}
+
 void rocksdb_flush(
    rocksdb_t* db,
    const rocksdb_flushoptions_t* options,
@@ -562,6 +964,13 @@ rocksdb_writebatch_t* rocksdb_writebatch_create() {
  return new rocksdb_writebatch_t;
 }

+rocksdb_writebatch_t* rocksdb_writebatch_create_from(const char* rep,
+                                                     size_t size) {
+  rocksdb_writebatch_t* b = new rocksdb_writebatch_t;
+  b->rep = WriteBatch(std::string(rep, size));
+  return b;
+}
+
 void rocksdb_writebatch_destroy(rocksdb_writebatch_t* b) {
  delete b;
 }
@@ -581,6 +990,14 @@ void rocksdb_writebatch_put(
  b->rep.Put(Slice(key, klen), Slice(val, vlen));
 }

+void rocksdb_writebatch_put_cf(
+    rocksdb_writebatch_t* b,
+    rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t klen,
+    const char* val, size_t vlen) {
+  b->rep.Put(column_family->rep, Slice(key, klen), Slice(val, vlen));
+}
+
 void rocksdb_writebatch_merge(
    rocksdb_writebatch_t* b,
    const char* key, size_t klen,
@@ -588,12 +1005,27 @@ void rocksdb_writebatch_merge(
  b->rep.Merge(Slice(key, klen), Slice(val, vlen));
 }

+void rocksdb_writebatch_merge_cf(
+    rocksdb_writebatch_t* b,
+    rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t klen,
+    const char* val, size_t vlen) {
+  b->rep.Merge(column_family->rep, Slice(key, klen), Slice(val, vlen));
+}
+
 void rocksdb_writebatch_delete(
    rocksdb_writebatch_t* b,
    const char* key, size_t klen) {
  b->rep.Delete(Slice(key, klen));
 }

+void rocksdb_writebatch_delete_cf(
+    rocksdb_writebatch_t* b,
+    rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t klen) {
+  b->rep.Delete(column_family->rep, Slice(key, klen));
+}
+
 void rocksdb_writebatch_iterate(
    rocksdb_writebatch_t* b,
    void* state,
@@ -623,6 +1055,74 @@ const char* rocksdb_writebatch_data(rocksdb_writebatch_t* b, size_t* size) {
  return b->rep.Data().c_str();
 }

+rocksdb_block_based_table_options_t*
+rocksdb_block_based_options_create() {
+  return new rocksdb_block_based_table_options_t;
+}
+
+void rocksdb_block_based_options_destroy(
+    rocksdb_block_based_table_options_t* options) {
+  delete options;
+}
+
+void rocksdb_block_based_options_set_block_size(
+    rocksdb_block_based_table_options_t* options, size_t block_size) {
+  options->rep.block_size = block_size;
+}
+
+void rocksdb_block_based_options_set_block_size_deviation(
+    rocksdb_block_based_table_options_t* options, int block_size_deviation) {
+  options->rep.block_size_deviation = block_size_deviation;
+}
+
+void rocksdb_block_based_options_set_block_restart_interval(
+    rocksdb_block_based_table_options_t* options, int block_restart_interval) {
+  options->rep.block_restart_interval = block_restart_interval;
+}
+
+void rocksdb_block_based_options_set_filter_policy(
+    rocksdb_block_based_table_options_t* options,
+    rocksdb_filterpolicy_t* filter_policy) {
+  options->rep.filter_policy.reset(filter_policy);
+}
+
+void rocksdb_block_based_options_set_no_block_cache(
+    rocksdb_block_based_table_options_t* options,
+    unsigned char no_block_cache) {
+  options->rep.no_block_cache = no_block_cache;
+}
+
+void rocksdb_block_based_options_set_block_cache(
+    rocksdb_block_based_table_options_t* options,
+    rocksdb_cache_t* block_cache) {
+  if (block_cache) {
+    options->rep.block_cache = block_cache->rep;
+  }
+}
+
+void rocksdb_block_based_options_set_block_cache_compressed(
+    rocksdb_block_based_table_options_t* options,
+    rocksdb_cache_t* block_cache_compressed) {
+  if (block_cache_compressed) {
+    options->rep.block_cache_compressed = block_cache_compressed->rep;
+  }
+}
+
+void rocksdb_block_based_options_set_whole_key_filtering(
+    rocksdb_block_based_table_options_t* options, unsigned char v) {
+  options->rep.whole_key_filtering = v;
+}
+
+void rocksdb_options_set_block_based_table_factory(
+    rocksdb_options_t *opt,
+    rocksdb_block_based_table_options_t* table_options) {
+  if (table_options) {
+    opt->rep.table_factory.reset(
+        rocksdb::NewBlockBasedTableFactory(table_options->rep));
+  }
+}
+
+
 rocksdb_options_t* rocksdb_options_create() {
  return new rocksdb_options_t;
 }
@@ -631,6 +1131,38 @@ void rocksdb_options_destroy(rocksdb_options_t* options) {
  delete options;
 }

+void rocksdb_options_increase_parallelism(
+    rocksdb_options_t* opt, int total_threads) {
+  opt->rep.IncreaseParallelism(total_threads);
+}
+
+void rocksdb_options_optimize_for_point_lookup(
+    rocksdb_options_t* opt, uint64_t block_cache_size_mb) {
+  opt->rep.OptimizeForPointLookup(block_cache_size_mb);
+}
+
+void rocksdb_options_optimize_level_style_compaction(
+    rocksdb_options_t* opt, uint64_t memtable_memory_budget) {
+  opt->rep.OptimizeLevelStyleCompaction(memtable_memory_budget);
+}
+
+void rocksdb_options_optimize_universal_style_compaction(
+    rocksdb_options_t* opt, uint64_t memtable_memory_budget) {
+  opt->rep.OptimizeUniversalStyleCompaction(memtable_memory_budget);
+}
+
+void rocksdb_options_set_compaction_filter(
+    rocksdb_options_t* opt,
+    rocksdb_compactionfilter_t* filter) {
+  opt->rep.compaction_filter = filter;
+}
+
+void rocksdb_options_set_compaction_filter_factory(
+    rocksdb_options_t* opt, rocksdb_compactionfilterfactory_t* factory) {
+  opt->rep.compaction_filter_factory =
+      std::shared_ptr<CompactionFilterFactory>(factory);
+}
+
 void rocksdb_options_set_comparator(
    rocksdb_options_t* opt,
    rocksdb_comparator_t* cmp) {
@@ -643,10 +1175,10 @@ void rocksdb_options_set_merge_operator(
  opt->rep.merge_operator = std::shared_ptr<MergeOperator>(merge_operator);
 }

-void rocksdb_options_set_filter_policy(
+void rocksdb_options_set_compaction_filter_factory_v2(
    rocksdb_options_t* opt,
-    rocksdb_filterpolicy_t* policy) {
-  opt->rep.filter_policy = policy;
+    rocksdb_compactionfilterfactoryv2_t* compaction_filter_factory_v2) {
+  opt->rep.compaction_filter_factory_v2 = std::shared_ptr<CompactionFilterFactoryV2>(compaction_filter_factory_v2);
 }

 void rocksdb_options_set_create_if_missing(
@@ -654,6 +1186,11 @@ void rocksdb_options_set_create_if_missing(
  opt->rep.create_if_missing = v;
 }

+void rocksdb_options_set_create_missing_column_families(
+    rocksdb_options_t* opt, unsigned char v) {
+  opt->rep.create_missing_column_families = v;
+}
+
 void rocksdb_options_set_error_if_exists(
    rocksdb_options_t* opt, unsigned char v) {
  opt->rep.error_if_exists = v;
@@ -687,26 +1224,6 @@ void rocksdb_options_set_max_open_files(rocksdb_options_t* opt, int n) {
  opt->rep.max_open_files = n;
 }

-void rocksdb_options_set_cache(rocksdb_options_t* opt, rocksdb_cache_t* c) {
-  if (c) {
-    opt->rep.block_cache = c->rep;
-  }
-}
-
-void rocksdb_options_set_cache_compressed(rocksdb_options_t* opt, rocksdb_cache_t* c) {
-  if (c) {
-    opt->rep.block_cache_compressed = c->rep;
-  }
-}
-
-void rocksdb_options_set_block_size(rocksdb_options_t* opt, size_t s) {
-  opt->rep.block_size = s;
-}
-
-void rocksdb_options_set_block_restart_interval(rocksdb_options_t* opt, int n) {
-  opt->rep.block_restart_interval = n;
-}
-
 void rocksdb_options_set_target_file_size_base(
    rocksdb_options_t* opt, uint64_t n) {
  opt->rep.target_file_size_base = n;
@@ -799,11 +1316,6 @@ void rocksdb_options_set_prefix_extractor(
  opt->rep.prefix_extractor.reset(prefix_extractor);
 }

-void rocksdb_options_set_whole_key_filtering(
-    rocksdb_options_t* opt, unsigned char v) {
-  opt->rep.whole_key_filtering = v;
-}
-
 void rocksdb_options_set_disable_data_sync(
    rocksdb_options_t* opt, int disable_data_sync) {
  opt->rep.disableDataSync = disable_data_sync;
@@ -814,11 +1326,6 @@ void rocksdb_options_set_use_fsync(
  opt->rep.use_fsync = use_fsync;
 }

-void rocksdb_options_set_db_stats_log_interval(
-    rocksdb_options_t* opt, int db_stats_log_interval) {
-  opt->rep.db_stats_log_interval = db_stats_log_interval;
-}
-
 void rocksdb_options_set_db_log_dir(
    rocksdb_options_t* opt, const char* db_log_dir) {
  opt->rep.db_log_dir = db_log_dir;
@@ -848,8 +1355,8 @@ void rocksdb_options_set_purge_redundant_kvs_while_flush(
  opt->rep.purge_redundant_kvs_while_flush = v;
 }

-void rocksdb_options_set_allow_os_buffer(
-    rocksdb_options_t* opt, unsigned char v) {
+void rocksdb_options_set_allow_os_buffer(rocksdb_options_t* opt,
+                                         unsigned char v) {
  opt->rep.allow_os_buffer = v;
 }

@@ -878,11 +1385,6 @@ void rocksdb_options_set_stats_dump_period_sec(
  opt->rep.stats_dump_period_sec = v;
 }

-void rocksdb_options_set_block_size_deviation(
-    rocksdb_options_t* opt, int v) {
-  opt->rep.block_size_deviation = v;
-}
-
 void rocksdb_options_set_advise_random_on_open(
    rocksdb_options_t* opt, unsigned char v) {
  opt->rep.advise_random_on_open = v;
@@ -977,11 +1479,6 @@ void rocksdb_options_set_max_manifest_file_size(
  opt->rep.max_manifest_file_size = v;
 }

-void rocksdb_options_set_no_block_cache(
-    rocksdb_options_t* opt, unsigned char v) {
-  opt->rep.no_block_cache = v;
-}
-
 void rocksdb_options_set_table_cache_numshardbits(
    rocksdb_options_t* opt, int v) {
  opt->rep.table_cache_numshardbits = v;
@@ -1001,10 +1498,6 @@ void rocksdb_options_set_disable_auto_compactions(rocksdb_options_t* opt, int di
  opt->rep.disable_auto_compactions = disable;
 }

-void rocksdb_options_set_disable_seek_compaction(rocksdb_options_t* opt, int disable) {
-  opt->rep.disable_seek_compaction = disable;
-}
-
 void rocksdb_options_set_delete_obsolete_files_period_micros(
    rocksdb_options_t* opt, uint64_t v) {
  opt->rep.delete_obsolete_files_period_micros = v;
@@ -1062,9 +1555,13 @@ void rocksdb_options_set_plain_table_factory(
    double hash_table_ratio, size_t index_sparseness) {
  static rocksdb::TableFactory* factory = 0;
  if (!factory) {
-    factory = rocksdb::NewPlainTableFactory(
-        user_key_len, bloom_bits_per_key,
-        hash_table_ratio, index_sparseness);
+    rocksdb::PlainTableOptions options;
+    options.user_key_len = user_key_len;
+    options.bloom_bits_per_key = bloom_bits_per_key;
+    options.hash_table_ratio = hash_table_ratio;
+    options.index_sparseness = index_sparseness;
+
+    factory = rocksdb::NewPlainTableFactory(options);
  }
  opt->rep.table_factory.reset(factory);
 }
@@ -1084,11 +1581,6 @@ void rocksdb_options_set_bloom_locality(
  opt->rep.bloom_locality = v;
 }

-void rocksdb_options_set_allow_thread_local(
-    rocksdb_options_t* opt, unsigned char v) {
-  opt->rep.allow_thread_local = v;
-}
-
 void rocksdb_options_set_inplace_update_support(
    rocksdb_options_t* opt, unsigned char v) {
  opt->rep.inplace_update_support = v;
@@ -1107,6 +1599,12 @@ void rocksdb_options_set_universal_compaction_options(rocksdb_options_t *opt, ro
  opt->rep.compaction_options_universal = *(uco->rep);
 }

+void rocksdb_options_set_fifo_compaction_options(
+    rocksdb_options_t* opt,
+    rocksdb_fifo_compaction_options_t* fifo) {
+  opt->rep.compaction_options_fifo = fifo->rep;
+}
+
 /*
 TODO:
 DB::OpenForReadOnly
@@ -1119,10 +1617,101 @@ DB::GetUpdatesSince
 DB::GetDbIdentity
 DB::RunManualCompaction
 custom cache
-compaction_filter
 table_properties_collectors
 */

+rocksdb_compactionfilter_t* rocksdb_compactionfilter_create(
+    void* state,
+    void (*destructor)(void*),
+    unsigned char (*filter)(
+        void*,
+        int level,
+        const char* key, size_t key_length,
+        const char* existing_value, size_t value_length,
+        char** new_value, size_t *new_value_length,
+        unsigned char* value_changed),
+    const char* (*name)(void*)) {
+  rocksdb_compactionfilter_t* result = new rocksdb_compactionfilter_t;
+  result->state_ = state;
+  result->destructor_ = destructor;
+  result->filter_ = filter;
+  result->name_ = name;
+  return result;
+}
+
+void rocksdb_compactionfilter_destroy(rocksdb_compactionfilter_t* filter) {
+  delete filter;
+}
+
+unsigned char rocksdb_compactionfiltercontext_is_full_compaction(
+    rocksdb_compactionfiltercontext_t* context) {
+  return context->rep.is_full_compaction;
+}
+
+unsigned char rocksdb_compactionfiltercontext_is_manual_compaction(
+    rocksdb_compactionfiltercontext_t* context) {
+  return context->rep.is_manual_compaction;
+}
+
+rocksdb_compactionfilterfactory_t* rocksdb_compactionfilterfactory_create(
+    void* state, void (*destructor)(void*),
+    rocksdb_compactionfilter_t* (*create_compaction_filter)(
+        void*, rocksdb_compactionfiltercontext_t* context),
+    const char* (*name)(void*)) {
+  rocksdb_compactionfilterfactory_t* result =
+      new rocksdb_compactionfilterfactory_t;
+  result->state_ = state;
+  result->destructor_ = destructor;
+  result->create_compaction_filter_ = create_compaction_filter;
+  result->name_ = name;
+  return result;
+}
+
+void rocksdb_compactionfilterfactory_destroy(
+    rocksdb_compactionfilterfactory_t* factory) {
+  delete factory;
+}
+
+rocksdb_compactionfilterv2_t* rocksdb_compactionfilterv2_create(
+    void* state,
+    void (*destructor)(void*),
+    void (*filter)(void*, int level, size_t num_keys,
+                   const char* const* keys_list, const size_t* keys_list_sizes,
+                   const char* const* existing_values_list, const size_t* existing_values_list_sizes,
+                   char** new_values_list, size_t* new_values_list_sizes,
+                   unsigned char* to_delete_list),
+    const char* (*name)(void*)) {
+  rocksdb_compactionfilterv2_t* result = new rocksdb_compactionfilterv2_t;
+  result->state_ = state;
+  result->destructor_ = destructor;
+  result->filter_ = filter;
+  result->name_ = name;
+  return result;
+}
+
+void rocksdb_compactionfilterv2_destroy(rocksdb_compactionfilterv2_t* filter) {
+  delete filter;
+}
+
+rocksdb_compactionfilterfactoryv2_t* rocksdb_compactionfilterfactoryv2_create(
+    void* state,
+    rocksdb_slicetransform_t* prefix_extractor,
+    void (*destructor)(void*),
+    rocksdb_compactionfilterv2_t* (*create_compaction_filter_v2)(
+        void* state, const rocksdb_compactionfiltercontext_t* context),
+    const char* (*name)(void*)) {
+  rocksdb_compactionfilterfactoryv2_t* result = new rocksdb_compactionfilterfactoryv2_t(prefix_extractor);
+  result->state_ = state;
+  result->destructor_ = destructor;
+  result->create_compaction_filter_v2_ = create_compaction_filter_v2;
+  result->name_ = name;
+  return result;
+}
+
+void rocksdb_compactionfilterfactoryv2_destroy(rocksdb_compactionfilterfactoryv2_t* factory) {
+  delete factory;
+}
+
 rocksdb_comparator_t* rocksdb_comparator_create(
    void* state,
    void (*destructor)(void*),
@@ -1250,6 +1839,13 @@ void rocksdb_readoptions_set_snapshot(
  opt->rep.snapshot = (snap ? snap->rep : nullptr);
 }

+void rocksdb_readoptions_set_iterate_upper_bound(
+    rocksdb_readoptions_t* opt,
+    const char* key, size_t keylen) {
+  Slice prefix = Slice(key, keylen);
+  opt->rep.iterate_upper_bound = &prefix;
+}
+
 void rocksdb_readoptions_set_read_tier(
    rocksdb_readoptions_t* opt, int v) {
  opt->rep.read_tier = static_cast<rocksdb::ReadTier>(v);
@@ -1414,6 +2010,22 @@ void rocksdb_universal_compaction_options_destroy(
  delete uco;
 }

+rocksdb_fifo_compaction_options_t* rocksdb_fifo_compaction_options_create() {
+  rocksdb_fifo_compaction_options_t* result = new rocksdb_fifo_compaction_options_t;
+  result->rep =  CompactionOptionsFIFO();
+  return result;
+}
+
+void rocksdb_fifo_compaction_options_set_max_table_files_size(
+    rocksdb_fifo_compaction_options_t* fifo_opts, uint64_t size) {
+  fifo_opts->rep.max_table_files_size = size;
+}
+
+void rocksdb_fifo_compaction_options_destroy(
+    rocksdb_fifo_compaction_options_t* fifo_opts) {
+  delete fifo_opts;
+}
+
 void rocksdb_options_set_min_level_to_compress(rocksdb_options_t* opt, int level) {
  if (level >= 0) {
    assert(level <= opt->rep.num_levels);
@@ -75,6 +75,22 @@ static void CheckGet(
  Free(&val);
 }

+static void CheckGetCF(
+    rocksdb_t* db,
+    const rocksdb_readoptions_t* options,
+    rocksdb_column_family_handle_t* handle,
+    const char* key,
+    const char* expected) {
+  char* err = NULL;
+  size_t val_len;
+  char* val;
+  val = rocksdb_get_cf(db, options, handle, key, strlen(key), &val_len, &err);
+  CheckNoError(err);
+  CheckEqual(expected, val, val_len);
+  Free(&val);
+}
+
+
 static void CheckIter(rocksdb_iterator_t* iter,
                      const char* key, const char* val) {
  size_t len;
@@ -154,6 +170,134 @@ static unsigned char FilterKeyMatch(
  return fake_filter_result;
 }

+// Custom compaction filter
+static void CFilterDestroy(void* arg) {}
+static const char* CFilterName(void* arg) { return "foo"; }
+static unsigned char CFilterFilter(void* arg, int level, const char* key,
+                                   size_t key_length,
+                                   const char* existing_value,
+                                   size_t value_length, char** new_value,
+                                   size_t* new_value_length,
+                                   unsigned char* value_changed) {
+  if (key_length == 3) {
+    if (memcmp(key, "bar", key_length) == 0) {
+      return 1;
+    } else if (memcmp(key, "baz", key_length) == 0) {
+      *value_changed = 1;
+      *new_value = "newbazvalue";
+      *new_value_length = 11;
+      return 0;
+    }
+  }
+  return 0;
+}
+
+static void CFilterFactoryDestroy(void* arg) {}
+static const char* CFilterFactoryName(void* arg) { return "foo"; }
+static rocksdb_compactionfilter_t* CFilterCreate(
+    void* arg, rocksdb_compactionfiltercontext_t* context) {
+  return rocksdb_compactionfilter_create(NULL, CFilterDestroy, CFilterFilter,
+                                         CFilterName);
+}
+
+static rocksdb_t* CheckCompaction(rocksdb_t* db, rocksdb_options_t* options,
+                                  rocksdb_readoptions_t* roptions,
+                                  rocksdb_writeoptions_t* woptions) {
+  char* err = NULL;
+  db = rocksdb_open(options, dbname, &err);
+  CheckNoError(err);
+  rocksdb_put(db, woptions, "foo", 3, "foovalue", 8, &err);
+  CheckNoError(err);
+  CheckGet(db, roptions, "foo", "foovalue");
+  rocksdb_put(db, woptions, "bar", 3, "barvalue", 8, &err);
+  CheckNoError(err);
+  CheckGet(db, roptions, "bar", "barvalue");
+  rocksdb_put(db, woptions, "baz", 3, "bazvalue", 8, &err);
+  CheckNoError(err);
+  CheckGet(db, roptions, "baz", "bazvalue");
+
+  // Force compaction
+  rocksdb_compact_range(db, NULL, 0, NULL, 0);
+  // should have filtered bar, but not foo
+  CheckGet(db, roptions, "foo", "foovalue");
+  CheckGet(db, roptions, "bar", NULL);
+  CheckGet(db, roptions, "baz", "newbazvalue");
+  return db;
+}
+
+// Custom compaction filter V2.
+static void CompactionFilterV2Destroy(void* arg) { }
+static const char* CompactionFilterV2Name(void* arg) {
+  return "TestCompactionFilterV2";
+}
+static void CompactionFilterV2Filter(
+    void* arg, int level, size_t num_keys,
+    const char* const* keys_list, const size_t* keys_list_sizes,
+    const char* const* existing_values_list, const size_t* existing_values_list_sizes,
+    char** new_values_list, size_t* new_values_list_sizes,
+    unsigned char* to_delete_list) {
+  size_t i;
+  for (i = 0; i < num_keys; i++) {
+    // If any value is "gc", it's removed.
+    if (existing_values_list_sizes[i] == 2 && memcmp(existing_values_list[i], "gc", 2) == 0) {
+      to_delete_list[i] = 1;
+    } else if (existing_values_list_sizes[i] == 6 && memcmp(existing_values_list[i], "gc all", 6) == 0) {
+      // If any value is "gc all", all keys are removed.
+      size_t j;
+      for (j = 0; j < num_keys; j++) {
+        to_delete_list[j] = 1;
+      }
+      return;
+    } else if (existing_values_list_sizes[i] == 6 && memcmp(existing_values_list[i], "change", 6) == 0) {
+      // If value is "change", set changed value to "changed".
+      size_t len;
+      len = strlen("changed");
+      new_values_list[i] = malloc(len);
+      memcpy(new_values_list[i], "changed", len);
+      new_values_list_sizes[i] = len;
+    } else {
+      // Otherwise, no keys are removed.
+    }
+  }
+}
+
+// Custom prefix extractor for compaction filter V2 which extracts first 3 characters.
+static void CFV2PrefixExtractorDestroy(void* arg) { }
+static char* CFV2PrefixExtractorTransform(void* arg, const char* key, size_t length, size_t* dst_length) {
+  // Verify keys are maximum length 4; this verifies fix for a
+  // prior bug which was passing the RocksDB-encoded key with
+  // logical timestamp suffix instead of parsed user key.
+  if (length > 4) {
+    fprintf(stderr, "%s:%d: %s: key %s is not user key\n", __FILE__, __LINE__, phase, key);
+    abort();
+  }
+  *dst_length = length < 3 ? length : 3;
+  return (char*)key;
+}
+static unsigned char CFV2PrefixExtractorInDomain(void* state, const char* key, size_t length) {
+  return 1;
+}
+static unsigned char CFV2PrefixExtractorInRange(void* state, const char* key, size_t length) {
+  return 1;
+}
+static const char* CFV2PrefixExtractorName(void* state) {
+  return "TestCFV2PrefixExtractor";
+}
+
+// Custom compaction filter factory V2.
+static void CompactionFilterFactoryV2Destroy(void* arg) {
+  rocksdb_slicetransform_destroy((rocksdb_slicetransform_t*)arg);
+}
+static const char* CompactionFilterFactoryV2Name(void* arg) {
+  return "TestCompactionFilterFactoryV2";
+}
+static rocksdb_compactionfilterv2_t* CompactionFilterFactoryV2Create(
+    void* state, const rocksdb_compactionfiltercontext_t* context) {
+  return rocksdb_compactionfilterv2_create(state, CompactionFilterV2Destroy,
+                                           CompactionFilterV2Filter,
+                                           CompactionFilterV2Name);
+}
+
 // Custom merge operator
 static void MergeOperatorDestroy(void* arg) { }
 static const char* MergeOperatorName(void* arg) {
@@ -191,6 +335,7 @@ int main(int argc, char** argv) {
  rocksdb_cache_t* cache;
  rocksdb_env_t* env;
  rocksdb_options_t* options;
+  rocksdb_block_based_table_options_t* table_options;
  rocksdb_readoptions_t* roptions;
  rocksdb_writeoptions_t* woptions;
  char* err = NULL;
@@ -209,14 +354,15 @@ int main(int argc, char** argv) {
  options = rocksdb_options_create();
  rocksdb_options_set_comparator(options, cmp);
  rocksdb_options_set_error_if_exists(options, 1);
-  rocksdb_options_set_cache(options, cache);
  rocksdb_options_set_env(options, env);
  rocksdb_options_set_info_log(options, NULL);
  rocksdb_options_set_write_buffer_size(options, 100000);
  rocksdb_options_set_paranoid_checks(options, 1);
  rocksdb_options_set_max_open_files(options, 10);
-  rocksdb_options_set_block_size(options, 1024);
-  rocksdb_options_set_block_restart_interval(options, 8);
+  table_options = rocksdb_block_based_options_create();
+  rocksdb_block_based_options_set_block_cache(table_options, cache);
+  rocksdb_options_set_block_based_table_factory(options, table_options);
+
  rocksdb_options_set_compression(options, rocksdb_no_compression);
  rocksdb_options_set_compression_options(options, -14, -1, 0);
  int compression_levels[] = {rocksdb_no_compression, rocksdb_no_compression,
@@ -277,6 +423,24 @@ int main(int argc, char** argv) {
    rocksdb_writebatch_destroy(wb);
  }

+  StartPhase("writebatch_rep");
+  {
+    rocksdb_writebatch_t* wb1 = rocksdb_writebatch_create();
+    rocksdb_writebatch_put(wb1, "baz", 3, "d", 1);
+    rocksdb_writebatch_put(wb1, "quux", 4, "e", 1);
+    rocksdb_writebatch_delete(wb1, "quux", 4);
+    size_t repsize1 = 0;
+    const char* rep = rocksdb_writebatch_data(wb1, &repsize1);
+    rocksdb_writebatch_t* wb2 = rocksdb_writebatch_create_from(rep, repsize1);
+    CheckCondition(rocksdb_writebatch_count(wb1) ==
+                   rocksdb_writebatch_count(wb2));
+    size_t repsize2 = 0;
+    CheckCondition(
+        memcmp(rep, rocksdb_writebatch_data(wb2, &repsize2), repsize1) == 0);
+    rocksdb_writebatch_destroy(wb1);
+    rocksdb_writebatch_destroy(wb2);
+  }
+
  StartPhase("iter");
  {
    rocksdb_iterator_t* iter = rocksdb_create_iterator(db, roptions);
@@ -378,10 +542,12 @@ int main(int argc, char** argv) {
      policy = rocksdb_filterpolicy_create_bloom(10);
    }

+    rocksdb_block_based_options_set_filter_policy(table_options, policy);
+
    // Create new database
    rocksdb_close(db);
    rocksdb_destroy_db(options, dbname, &err);
-    rocksdb_options_set_filter_policy(options, policy);
+    rocksdb_options_set_block_based_table_factory(options, table_options);
    db = rocksdb_open(options, dbname, &err);
    CheckNoError(err);
    rocksdb_put(db, woptions, "foo", 3, "foovalue", 8, &err);
@@ -403,8 +569,88 @@ int main(int argc, char** argv) {
      CheckGet(db, roptions, "foo", "foovalue");
      CheckGet(db, roptions, "bar", "barvalue");
    }
-    rocksdb_options_set_filter_policy(options, NULL);
-    rocksdb_filterpolicy_destroy(policy);
+    // Reset the policy
+    rocksdb_block_based_options_set_filter_policy(table_options, NULL);
+    rocksdb_options_set_block_based_table_factory(options, table_options);
+  }
+
+  StartPhase("compaction_filter");
+  {
+    rocksdb_options_t* options = rocksdb_options_create();
+    rocksdb_options_set_create_if_missing(options, 1);
+    rocksdb_compactionfilter_t* cfilter;
+    cfilter = rocksdb_compactionfilter_create(NULL, CFilterDestroy,
+                                              CFilterFilter, CFilterName);
+    // Create new database
+    rocksdb_close(db);
+    rocksdb_destroy_db(options, dbname, &err);
+    rocksdb_options_set_compaction_filter(options, cfilter);
+    db = CheckCompaction(db, options, roptions, woptions);
+
+    rocksdb_options_set_compaction_filter(options, NULL);
+    rocksdb_compactionfilter_destroy(cfilter);
+    rocksdb_options_destroy(options);
+  }
+
+  StartPhase("compaction_filter_factory");
+  {
+    rocksdb_options_t* options = rocksdb_options_create();
+    rocksdb_options_set_create_if_missing(options, 1);
+    rocksdb_compactionfilterfactory_t* factory;
+    factory = rocksdb_compactionfilterfactory_create(
+        NULL, CFilterFactoryDestroy, CFilterCreate, CFilterFactoryName);
+    // Create new database
+    rocksdb_close(db);
+    rocksdb_destroy_db(options, dbname, &err);
+    rocksdb_options_set_compaction_filter_factory(options, factory);
+    db = CheckCompaction(db, options, roptions, woptions);
+
+    rocksdb_options_set_compaction_filter_factory(options, NULL);
+    rocksdb_options_destroy(options);
+  }
+
+  StartPhase("compaction_filter_v2");
+  {
+    rocksdb_compactionfilterfactoryv2_t* factory;
+    rocksdb_slicetransform_t* prefix_extractor;
+    prefix_extractor = rocksdb_slicetransform_create(
+        NULL, CFV2PrefixExtractorDestroy, CFV2PrefixExtractorTransform,
+        CFV2PrefixExtractorInDomain, CFV2PrefixExtractorInRange,
+        CFV2PrefixExtractorName);
+    factory = rocksdb_compactionfilterfactoryv2_create(
+        prefix_extractor, prefix_extractor, CompactionFilterFactoryV2Destroy,
+        CompactionFilterFactoryV2Create, CompactionFilterFactoryV2Name);
+    // Create new database
+    rocksdb_close(db);
+    rocksdb_destroy_db(options, dbname, &err);
+    rocksdb_options_set_compaction_filter_factory_v2(options, factory);
+    db = rocksdb_open(options, dbname, &err);
+    CheckNoError(err);
+    // Only foo2 is GC'd, foo3 is changed.
+    rocksdb_put(db, woptions, "foo1", 4, "no gc", 5, &err);
+    CheckNoError(err);
+    rocksdb_put(db, woptions, "foo2", 4, "gc", 2, &err);
+    CheckNoError(err);
+    rocksdb_put(db, woptions, "foo3", 4, "change", 6, &err);
+    CheckNoError(err);
+    // All bars are GC'd.
+    rocksdb_put(db, woptions, "bar1", 4, "no gc", 5, &err);
+    CheckNoError(err);
+    rocksdb_put(db, woptions, "bar2", 4, "gc all", 6, &err);
+    CheckNoError(err);
+    rocksdb_put(db, woptions, "bar3", 4, "no gc", 5, &err);
+    CheckNoError(err);
+    // Compact the DB to garbage collect.
+    rocksdb_compact_range(db, NULL, 0, NULL, 0);
+
+    // Verify foo entries.
+    CheckGet(db, roptions, "foo1", "no gc");
+    CheckGet(db, roptions, "foo2", NULL);
+    CheckGet(db, roptions, "foo3", "changed");
+    // Verify bar entries were all deleted.
+    CheckGet(db, roptions, "bar1", NULL);
+    CheckGet(db, roptions, "bar2", NULL);
+    CheckGet(db, roptions, "bar3", NULL);
  }

  StartPhase("merge_operator");
@@ -433,16 +679,92 @@ int main(int argc, char** argv) {

  }

+  StartPhase("columnfamilies");
+  {
+    rocksdb_close(db);
+    rocksdb_destroy_db(options, dbname, &err);
+    CheckNoError(err)
+
+    rocksdb_options_t* db_options = rocksdb_options_create();
+    rocksdb_options_set_create_if_missing(db_options, 1);
+    db = rocksdb_open(db_options, dbname, &err);
+    CheckNoError(err)
+    rocksdb_column_family_handle_t* cfh;
+    cfh = rocksdb_create_column_family(db, db_options, "cf1", &err);
+    rocksdb_column_family_handle_destroy(cfh);
+    CheckNoError(err);
+    rocksdb_close(db);
+
+    size_t cflen;
+    char** column_fams = rocksdb_list_column_families(db_options, dbname, &cflen, &err);
+    CheckNoError(err);
+    CheckEqual("default", column_fams[0], 7);
+    CheckEqual("cf1", column_fams[1], 3);
+    CheckCondition(cflen == 2);
+    rocksdb_list_column_families_destroy(column_fams, cflen);
+
+    rocksdb_options_t* cf_options = rocksdb_options_create();
+
+    const char* cf_names[2] = {"default", "cf1"};
+    const rocksdb_options_t* cf_opts[2] = {cf_options, cf_options};
+    rocksdb_column_family_handle_t* handles[2];
+    db = rocksdb_open_column_families(db_options, dbname, 2, cf_names, cf_opts, handles, &err);
+    CheckNoError(err);
+
+    rocksdb_put_cf(db, woptions, handles[1], "foo", 3, "hello", 5, &err);
+    CheckNoError(err);
+
+    CheckGetCF(db, roptions, handles[1], "foo", "hello");
+
+    rocksdb_delete_cf(db, woptions, handles[1], "foo", 3, &err);
+    CheckNoError(err);
+
+    CheckGetCF(db, roptions, handles[1], "foo", NULL);
+
+    rocksdb_writebatch_t* wb = rocksdb_writebatch_create();
+    rocksdb_writebatch_put_cf(wb, handles[1], "baz", 3, "a", 1);
+    rocksdb_writebatch_clear(wb);
+    rocksdb_writebatch_put_cf(wb, handles[1], "bar", 3, "b", 1);
+    rocksdb_writebatch_put_cf(wb, handles[1], "box", 3, "c", 1);
+    rocksdb_writebatch_delete_cf(wb, handles[1], "bar", 3);
+    rocksdb_write(db, woptions, wb, &err);
+    CheckNoError(err);
+    CheckGetCF(db, roptions, handles[1], "baz", NULL);
+    CheckGetCF(db, roptions, handles[1], "bar", NULL);
+    CheckGetCF(db, roptions, handles[1], "box", "c");
+    rocksdb_writebatch_destroy(wb);
+
+    rocksdb_iterator_t* iter = rocksdb_create_iterator_cf(db, roptions, handles[1]);
+    CheckCondition(!rocksdb_iter_valid(iter));
+    rocksdb_iter_seek_to_first(iter);
+    CheckCondition(rocksdb_iter_valid(iter));
+
+    int i;
+    for (i = 0; rocksdb_iter_valid(iter) != 0; rocksdb_iter_next(iter)) {
+      i++;
+    }
+    CheckCondition(i == 1);
+    rocksdb_iter_get_error(iter, &err);
+    CheckNoError(err);
+    rocksdb_iter_destroy(iter);
+
+    rocksdb_drop_column_family(db, handles[1], &err);
+    CheckNoError(err);
+    for (i = 0; i < 2; i++) {
+      rocksdb_column_family_handle_destroy(handles[i]);
+    }
+    rocksdb_close(db);
+    rocksdb_destroy_db(options, dbname, &err);
+    rocksdb_options_destroy(db_options);
+    rocksdb_options_destroy(cf_options);
+  }
+
  StartPhase("prefix");
  {
    // Create new database
-    rocksdb_close(db);
-    rocksdb_destroy_db(options, dbname, &err);
-
-    rocksdb_filterpolicy_t* policy = rocksdb_filterpolicy_create_bloom(10);
-    rocksdb_options_set_filter_policy(options, policy);
+    rocksdb_options_set_allow_mmap_reads(options, 1);
    rocksdb_options_set_prefix_extractor(options, rocksdb_slicetransform_create_fixed_prefix(3));
-    rocksdb_options_set_hash_skip_list_rep(options, 50000, 4, 4);
+    rocksdb_options_set_hash_skip_list_rep(options, 5000, 4, 4);
    rocksdb_options_set_plain_table_factory(options, 4, 10, 0.75, 16);

    db = rocksdb_open(options, dbname, &err);
@@ -477,12 +799,13 @@ int main(int argc, char** argv) {
    rocksdb_iter_get_error(iter, &err);
    CheckNoError(err);
    rocksdb_iter_destroy(iter);
-    rocksdb_filterpolicy_destroy(policy);
  }

+
  StartPhase("cleanup");
  rocksdb_close(db);
  rocksdb_options_destroy(options);
+  rocksdb_block_based_options_destroy(table_options);
  rocksdb_readoptions_destroy(roptions);
  rocksdb_writeoptions_destroy(woptions);
  rocksdb_cache_destroy(cache);
@@ -9,6 +9,11 @@

 #include "db/column_family.h"

+#ifndef __STDC_FORMAT_MACROS
+#define __STDC_FORMAT_MACROS
+#endif
+
+#include <inttypes.h>
 #include <vector>
 #include <string>
 #include <algorithm>
@@ -19,11 +24,43 @@
 #include "db/internal_stats.h"
 #include "db/compaction_picker.h"
 #include "db/table_properties_collector.h"
+#include "db/write_controller.h"
 #include "util/autovector.h"
 #include "util/hash_skiplist_rep.h"
+#include "util/options_helper.h"

 namespace rocksdb {

+namespace {
+// This function computes the amount of time in microseconds by which a write
+// should be delayed based on the number of level-0 files according to the
+// following formula:
+// if n < bottom, return 0;
+// if n >= top, return 1000;
+// otherwise, let r = (n - bottom) /
+//                    (top - bottom)
+//  and return r^2 * 1000.
+// The goal of this formula is to gradually increase the rate at which writes
+// are slowed. We also tried linear delay (r * 1000), but it seemed to do
+// slightly worse. There is no other particular reason for choosing quadratic.
+uint64_t SlowdownAmount(int n, double bottom, double top) {
+  uint64_t delay;
+  if (n >= top) {
+    delay = 1000;
+  } else if (n < bottom) {
+    delay = 0;
+  } else {
+    // If we are here, we know that:
+    //   level0_start_slowdown <= n < level0_slowdown
+    // since the previous two conditions are false.
+    double how_much = static_cast<double>(n - bottom) / (top - bottom);
+    delay = std::max(how_much * how_much * 1000, 100.0);
+  }
+  assert(delay <= 1000);
+  return delay;
+}
+}  // namespace
+
 ColumnFamilyHandleImpl::ColumnFamilyHandleImpl(ColumnFamilyData* cfd,
                                               DBImpl* db, port::Mutex* mutex)
    : cfd_(cfd), db_(db), mutex_(mutex) {
@@ -49,21 +86,14 @@ ColumnFamilyHandleImpl::~ColumnFamilyHandleImpl() {

 uint32_t ColumnFamilyHandleImpl::GetID() const { return cfd()->GetID(); }

-namespace {
-// Fix user-supplied options to be reasonable
-template <class T, class V>
-static void ClipToRange(T* ptr, V minvalue, V maxvalue) {
-  if (static_cast<V>(*ptr) > maxvalue) *ptr = maxvalue;
-  if (static_cast<V>(*ptr) < minvalue) *ptr = minvalue;
+const Comparator* ColumnFamilyHandleImpl::user_comparator() const {
+  return cfd()->user_comparator();
 }
-}  // anonymous namespace

 ColumnFamilyOptions SanitizeOptions(const InternalKeyComparator* icmp,
-                                    const InternalFilterPolicy* ipolicy,
                                    const ColumnFamilyOptions& src) {
  ColumnFamilyOptions result = src;
  result.comparator = icmp;
-  result.filter_policy = (src.filter_policy != nullptr) ? ipolicy : nullptr;
 #ifdef OS_MACOSX
  // TODO(icanadi) make write_buffer_size uint64_t instead of size_t
  ClipToRange(&result.write_buffer_size, ((size_t)64) << 10, ((size_t)1) << 30);
@@ -79,19 +109,16 @@ ColumnFamilyOptions SanitizeOptions(const InternalKeyComparator* icmp,
  result.min_write_buffer_number_to_merge =
      std::min(result.min_write_buffer_number_to_merge,
               result.max_write_buffer_number - 1);
-  if (result.block_cache == nullptr && !result.no_block_cache) {
-    result.block_cache = NewLRUCache(8 << 20);
-  }
  result.compression_per_level = src.compression_per_level;
-  if (result.block_size_deviation < 0 || result.block_size_deviation > 100) {
-    result.block_size_deviation = 0;
-  }
  if (result.max_mem_compaction_level >= result.num_levels) {
    result.max_mem_compaction_level = result.num_levels - 1;
  }
  if (result.soft_rate_limit > result.hard_rate_limit) {
    result.soft_rate_limit = result.hard_rate_limit;
  }
+  if (result.max_write_buffer_number < 2) {
+    result.max_write_buffer_number = 2;
+  }
  if (!result.prefix_extractor) {
    assert(result.memtable_factory);
    Slice name = result.memtable_factory->Name();
@@ -188,12 +215,11 @@ void SuperVersionUnrefHandle(void* ptr) {
 }
 }  // anonymous namespace

-ColumnFamilyData::ColumnFamilyData(const std::string& dbname, uint32_t id,
-                                   const std::string& name,
+ColumnFamilyData::ColumnFamilyData(uint32_t id, const std::string& name,
                                   Version* dummy_versions, Cache* table_cache,
-                                   const ColumnFamilyOptions& options,
+                                   const ColumnFamilyOptions& cf_options,
                                   const DBOptions* db_options,
-                                   const EnvOptions& storage_options,
+                                   const EnvOptions& env_options,
                                   ColumnFamilySet* column_family_set)
    : id_(id),
      name_(name),
@@ -201,10 +227,10 @@ ColumnFamilyData::ColumnFamilyData(const std::string& dbname, uint32_t id,
      current_(nullptr),
      refs_(0),
      dropped_(false),
-      internal_comparator_(options.comparator),
-      internal_filter_policy_(options.filter_policy),
-      options_(*db_options, SanitizeOptions(&internal_comparator_,
-                                            &internal_filter_policy_, options)),
+      internal_comparator_(cf_options.comparator),
+      options_(*db_options, SanitizeOptions(&internal_comparator_, cf_options)),
+      ioptions_(options_),
+      mutable_cf_options_(options_, ioptions_),
      mem_(nullptr),
      imm_(options_.min_write_buffer_number_to_merge),
      super_version_(nullptr),
@@ -213,33 +239,33 @@ ColumnFamilyData::ColumnFamilyData(const std::string& dbname, uint32_t id,
      next_(nullptr),
      prev_(nullptr),
      log_number_(0),
-      need_slowdown_for_num_level0_files_(false),
      column_family_set_(column_family_set) {
  Ref();

  // if dummy_versions is nullptr, then this is a dummy column family.
  if (dummy_versions != nullptr) {
-    internal_stats_.reset(new InternalStats(
-        options_.num_levels, db_options->env, db_options->statistics.get()));
-    table_cache_.reset(
-        new TableCache(dbname, &options_, storage_options, table_cache));
-    if (options_.compaction_style == kCompactionStyleUniversal) {
+    internal_stats_.reset(
+        new InternalStats(ioptions_.num_levels, db_options->env, this));
+    table_cache_.reset(new TableCache(ioptions_, env_options, table_cache));
+    if (ioptions_.compaction_style == kCompactionStyleUniversal) {
      compaction_picker_.reset(
-          new UniversalCompactionPicker(&options_, &internal_comparator_));
-    } else if (options_.compaction_style == kCompactionStyleLevel) {
+          new UniversalCompactionPicker(ioptions_, &internal_comparator_));
+    } else if (ioptions_.compaction_style == kCompactionStyleLevel) {
      compaction_picker_.reset(
-          new LevelCompactionPicker(&options_, &internal_comparator_));
+          new LevelCompactionPicker(ioptions_, &internal_comparator_));
    } else {
-      assert(options_.compaction_style == kCompactionStyleFIFO);
+      assert(ioptions_.compaction_style == kCompactionStyleFIFO);
      compaction_picker_.reset(
-          new FIFOCompactionPicker(&options_, &internal_comparator_));
+          new FIFOCompactionPicker(ioptions_, &internal_comparator_));
    }

-    Log(options_.info_log, "Options for column family \"%s\":\n",
+    Log(ioptions_.info_log, "Options for column family \"%s\":\n",
        name.c_str());
    const ColumnFamilyOptions* cf_options = &options_;
-    cf_options->Dump(options_.info_log.get());
+    cf_options->Dump(ioptions_.info_log);
  }
+
+  RecalculateWriteStallConditions(mutable_cf_options_);
 }

 // DB mutex held
@@ -292,51 +318,107 @@ ColumnFamilyData::~ColumnFamilyData() {
  }
 }

+void ColumnFamilyData::RecalculateWriteStallConditions(
+      const MutableCFOptions& mutable_cf_options) {
+  if (current_ != nullptr) {
+    const double score = current_->MaxCompactionScore();
+    const int max_level = current_->MaxCompactionScoreLevel();
+
+    auto write_controller = column_family_set_->write_controller_;
+
+    if (imm()->size() == options_.max_write_buffer_number) {
+      write_controller_token_ = write_controller->GetStopToken();
+      internal_stats_->AddCFStats(InternalStats::MEMTABLE_COMPACTION, 1);
+      Log(ioptions_.info_log,
+          "[%s] Stopping writes because we have %d immutable memtables "
+          "(waiting for flush)",
+          name_.c_str(), imm()->size());
+    } else if (current_->NumLevelFiles(0) >=
+               mutable_cf_options.level0_stop_writes_trigger) {
+      write_controller_token_ = write_controller->GetStopToken();
+      internal_stats_->AddCFStats(InternalStats::LEVEL0_NUM_FILES, 1);
+      Log(ioptions_.info_log,
+          "[%s] Stopping writes because we have %d level-0 files",
+          name_.c_str(), current_->NumLevelFiles(0));
+    } else if (mutable_cf_options.level0_slowdown_writes_trigger >= 0 &&
+               current_->NumLevelFiles(0) >=
+                   mutable_cf_options.level0_slowdown_writes_trigger) {
+      uint64_t slowdown = SlowdownAmount(
+          current_->NumLevelFiles(0),
+          mutable_cf_options.level0_slowdown_writes_trigger,
+          mutable_cf_options.level0_stop_writes_trigger);
+      write_controller_token_ = write_controller->GetDelayToken(slowdown);
+      internal_stats_->AddCFStats(InternalStats::LEVEL0_SLOWDOWN, slowdown);
+      Log(ioptions_.info_log,
+          "[%s] Stalling writes because we have %d level-0 files (%" PRIu64
+          "us)",
+          name_.c_str(), current_->NumLevelFiles(0), slowdown);
+    } else if (options_.hard_rate_limit > 1.0 &&
+               score > options_.hard_rate_limit) {
+      uint64_t kHardLimitSlowdown = 1000;
+      write_controller_token_ =
+          write_controller->GetDelayToken(kHardLimitSlowdown);
+      internal_stats_->RecordLevelNSlowdown(max_level, kHardLimitSlowdown,
+                                            false);
+      Log(ioptions_.info_log,
+          "[%s] Stalling writes because we hit hard limit on level %d. "
+          "(%" PRIu64 "us)",
+          name_.c_str(), max_level, kHardLimitSlowdown);
+    } else if (options_.soft_rate_limit > 0.0 &&
+               score > options_.soft_rate_limit) {
+      uint64_t slowdown = SlowdownAmount(score, options_.soft_rate_limit,
+                                         options_.hard_rate_limit);
+      write_controller_token_ = write_controller->GetDelayToken(slowdown);
+      internal_stats_->RecordLevelNSlowdown(max_level, slowdown, true);
+      Log(ioptions_.info_log,
+          "[%s] Stalling writes because we hit soft limit on level %d (%" PRIu64
+          "us)",
+          name_.c_str(), max_level, slowdown);
+    } else {
+      write_controller_token_.reset();
+    }
+  }
+}
+
 const EnvOptions* ColumnFamilyData::soptions() const {
-  return &(column_family_set_->storage_options_);
+  return &(column_family_set_->env_options_);
 }

-void ColumnFamilyData::SetCurrent(Version* current) {
-  current_ = current;
-  need_slowdown_for_num_level0_files_ =
-      (options_.level0_slowdown_writes_trigger >= 0 &&
-       current_->NumLevelFiles(0) >= options_.level0_slowdown_writes_trigger);
-}
+void ColumnFamilyData::SetCurrent(Version* current) { current_ = current; }

-void ColumnFamilyData::CreateNewMemtable() {
+void ColumnFamilyData::CreateNewMemtable(const MemTableOptions& moptions) {
  assert(current_ != nullptr);
  if (mem_ != nullptr) {
    delete mem_->Unref();
  }
-  mem_ = new MemTable(internal_comparator_, options_);
+  mem_ = new MemTable(internal_comparator_, ioptions_, moptions);
  mem_->Ref();
 }

-Compaction* ColumnFamilyData::PickCompaction(LogBuffer* log_buffer) {
-  return compaction_picker_->PickCompaction(current_, log_buffer);
+Compaction* ColumnFamilyData::PickCompaction(
+    const MutableCFOptions& mutable_options, LogBuffer* log_buffer) {
+  auto result = compaction_picker_->PickCompaction(
+      mutable_options, current_, log_buffer);
+  return result;
 }

-Compaction* ColumnFamilyData::CompactRange(int input_level, int output_level,
-                                           const InternalKey* begin,
-                                           const InternalKey* end,
-                                           InternalKey** compaction_end) {
-  return compaction_picker_->CompactRange(current_, input_level, output_level,
-                                          begin, end, compaction_end);
+Compaction* ColumnFamilyData::CompactRange(
+    const MutableCFOptions& mutable_cf_options,
+    int input_level, int output_level, uint32_t output_path_id,
+    const InternalKey* begin, const InternalKey* end,
+    InternalKey** compaction_end) {
+  return compaction_picker_->CompactRange(
+      mutable_cf_options, current_, input_level, output_level,
+      output_path_id, begin, end, compaction_end);
 }

 SuperVersion* ColumnFamilyData::GetReferencedSuperVersion(
    port::Mutex* db_mutex) {
  SuperVersion* sv = nullptr;
-  if (LIKELY(column_family_set_->db_options_->allow_thread_local)) {
-    sv = GetThreadLocalSuperVersion(db_mutex);
-    sv->Ref();
-    if (!ReturnThreadLocalSuperVersion(sv)) {
-      sv->Unref();
-    }
-  } else {
-    db_mutex->Lock();
-    sv = super_version_->Ref();
-    db_mutex->Unlock();
+  sv = GetThreadLocalSuperVersion(db_mutex);
+  sv->Ref();
+  if (!ReturnThreadLocalSuperVersion(sv)) {
+    sv->Unref();
  }
  return sv;
 }
@@ -365,11 +447,11 @@ SuperVersion* ColumnFamilyData::GetThreadLocalSuperVersion(
  sv = static_cast<SuperVersion*>(ptr);
  if (sv == SuperVersion::kSVObsolete ||
      sv->version_number != super_version_number_.load()) {
-    RecordTick(options_.statistics.get(), NUMBER_SUPERVERSION_ACQUIRES);
+    RecordTick(ioptions_.statistics, NUMBER_SUPERVERSION_ACQUIRES);
    SuperVersion* sv_to_delete = nullptr;

    if (sv && sv->Unref()) {
-      RecordTick(options_.statistics.get(), NUMBER_SUPERVERSION_CLEANUPS);
+      RecordTick(ioptions_.statistics, NUMBER_SUPERVERSION_CLEANUPS);
      db_mutex->Lock();
      // NOTE: underlying resources held by superversion (sst files) might
      // not be released until the next background job.
@@ -407,16 +489,25 @@ bool ColumnFamilyData::ReturnThreadLocalSuperVersion(SuperVersion* sv) {

 SuperVersion* ColumnFamilyData::InstallSuperVersion(
    SuperVersion* new_superversion, port::Mutex* db_mutex) {
+  db_mutex->AssertHeld();
+  return InstallSuperVersion(new_superversion, db_mutex, mutable_cf_options_);
+}
+
+SuperVersion* ColumnFamilyData::InstallSuperVersion(
+    SuperVersion* new_superversion, port::Mutex* db_mutex,
+    const MutableCFOptions& mutable_cf_options) {
  new_superversion->db_mutex = db_mutex;
+  new_superversion->mutable_cf_options = mutable_cf_options;
  new_superversion->Init(mem_, imm_.current(), current_);
  SuperVersion* old_superversion = super_version_;
  super_version_ = new_superversion;
  ++super_version_number_;
  super_version_->version_number = super_version_number_;
  // Reset SuperVersions cached in thread local storage
-  if (column_family_set_->db_options_->allow_thread_local) {
-    ResetThreadLocalSuperVersions();
-  }
+  ResetThreadLocalSuperVersions();
+
+  RecalculateWriteStallConditions(mutable_cf_options);
+
  if (old_superversion != nullptr && old_superversion->Unref()) {
    old_superversion->Cleanup();
    return old_superversion;  // will let caller delete outside of mutex
@@ -440,19 +531,33 @@ void ColumnFamilyData::ResetThreadLocalSuperVersions() {
  }
 }

+bool ColumnFamilyData::SetOptions(
+      const std::unordered_map<std::string, std::string>& options_map) {
+  MutableCFOptions new_mutable_cf_options;
+  if (GetMutableOptionsFromStrings(mutable_cf_options_, options_map,
+                                   &new_mutable_cf_options)) {
+    mutable_cf_options_ = new_mutable_cf_options;
+    mutable_cf_options_.RefreshDerivedOptions(ioptions_);
+    return true;
+  }
+  return false;
+}
+
 ColumnFamilySet::ColumnFamilySet(const std::string& dbname,
                                 const DBOptions* db_options,
-                                 const EnvOptions& storage_options,
-                                 Cache* table_cache)
+                                 const EnvOptions& env_options,
+                                 Cache* table_cache,
+                                 WriteController* write_controller)
    : max_column_family_(0),
-      dummy_cfd_(new ColumnFamilyData(dbname, 0, "", nullptr, nullptr,
+      dummy_cfd_(new ColumnFamilyData(0, "", nullptr, nullptr,
                                      ColumnFamilyOptions(), db_options,
-                                      storage_options_, nullptr)),
+                                      env_options, nullptr)),
      default_cfd_cache_(nullptr),
      db_name_(dbname),
      db_options_(db_options),
-      storage_options_(storage_options),
+      env_options_(env_options),
      table_cache_(table_cache),
+      write_controller_(write_controller),
      spin_lock_(ATOMIC_FLAG_INIT) {
  // initialize linked list
  dummy_cfd_->prev_ = dummy_cfd_;
@@ -506,14 +611,18 @@ void ColumnFamilySet::UpdateMaxColumnFamily(uint32_t new_max_column_family) {
  max_column_family_ = std::max(new_max_column_family, max_column_family_);
 }

+size_t ColumnFamilySet::NumberOfColumnFamilies() const {
+  return column_families_.size();
+}
+
 // under a DB mutex
 ColumnFamilyData* ColumnFamilySet::CreateColumnFamily(
    const std::string& name, uint32_t id, Version* dummy_versions,
    const ColumnFamilyOptions& options) {
  assert(column_families_.find(name) == column_families_.end());
  ColumnFamilyData* new_cfd =
-      new ColumnFamilyData(db_name_, id, name, dummy_versions, table_cache_,
-                           options, db_options_, storage_options_, this);
+      new ColumnFamilyData(id, name, dummy_versions, table_cache_, options,
+                           db_options_, env_options_, this);
  Lock();
  column_families_.insert({name, id});
  column_family_data_.insert({id, new_cfd});
@@ -572,6 +681,11 @@ bool ColumnFamilyMemTablesImpl::Seek(uint32_t column_family_id) {
    column_family_set_->Lock();
    current_ = column_family_set_->GetColumnFamily(column_family_id);
    column_family_set_->Unlock();
+    // TODO(icanadi) Maybe remove column family from the hash table when it's
+    // dropped?
+    if (current_ != nullptr && current_->IsDropped()) {
+      current_ = nullptr;
+    }
  }
  handle_.SetCFD(current_);
  return current_ != nullptr;
@@ -597,4 +711,29 @@ ColumnFamilyHandle* ColumnFamilyMemTablesImpl::GetColumnFamilyHandle() {
  return &handle_;
 }

+void ColumnFamilyMemTablesImpl::CheckMemtableFull() {
+  if (current_ != nullptr && current_->mem()->ShouldScheduleFlush()) {
+    flush_scheduler_->ScheduleFlush(current_);
+    current_->mem()->MarkFlushScheduled();
+  }
+}
+
+uint32_t GetColumnFamilyID(ColumnFamilyHandle* column_family) {
+  uint32_t column_family_id = 0;
+  if (column_family != nullptr) {
+    auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
+    column_family_id = cfh->GetID();
+  }
+  return column_family_id;
+}
+
+const Comparator* GetColumnFamilyUserComparator(
+    ColumnFamilyHandle* column_family) {
+  if (column_family != nullptr) {
+    auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
+    return cfh->user_comparator();
+  }
+  return nullptr;
+}
+
 }  // namespace rocksdb
@@ -19,8 +19,11 @@
 #include "rocksdb/env.h"
 #include "db/memtable_list.h"
 #include "db/write_batch_internal.h"
+#include "db/write_controller.h"
 #include "db/table_cache.h"
 #include "util/thread_local.h"
+#include "db/flush_scheduler.h"
+#include "util/mutable_cf_options.h"

 namespace rocksdb {

@@ -46,6 +49,7 @@ class ColumnFamilyHandleImpl : public ColumnFamilyHandle {
  // destroy without mutex
  virtual ~ColumnFamilyHandleImpl();
  virtual ColumnFamilyData* cfd() const { return cfd_; }
+  virtual const Comparator* user_comparator() const;

  virtual uint32_t GetID() const;

@@ -78,6 +82,7 @@ struct SuperVersion {
  MemTable* mem;
  MemTableListVersion* imm;
  Version* current;
+  MutableCFOptions mutable_cf_options;
  std::atomic<uint32_t> refs;
  // We need to_delete because during Cleanup(), imm->Unref() returns
  // all memtables that we need to free through this vector. We then
@@ -113,7 +118,6 @@ struct SuperVersion {
 };

 extern ColumnFamilyOptions SanitizeOptions(const InternalKeyComparator* icmp,
-                                           const InternalFilterPolicy* ipolicy,
                                           const ColumnFamilyOptions& src);

 class ColumnFamilySet;
@@ -133,7 +137,7 @@ class ColumnFamilyData {
  void Ref() { ++refs_; }
  // will just decrease reference count to 0, but will not delete it. returns
  // true if the ref count was decreased to zero. in that case, it can be
-  // deleted by the caller immediatelly, or later, by calling
+  // deleted by the caller immediately, or later, by calling
  // FreeDeadColumnFamilies()
  bool Unref() {
    assert(refs_ > 0);
@@ -157,6 +161,7 @@ class ColumnFamilyData {
    // can't drop default CF
    assert(id_ != 0);
    dropped_ = true;
+    write_controller_token_.reset();
  }
  bool IsDropped() const { return dropped_; }

@@ -169,6 +174,21 @@ class ColumnFamilyData {
  // thread-safe
  const Options* options() const { return &options_; }
  const EnvOptions* soptions() const;
+  const ImmutableCFOptions* ioptions() const { return &ioptions_; }
+  // REQUIRES: DB mutex held
+  // This returns the MutableCFOptions used by current SuperVersion
+  // You shoul use this API to reference MutableCFOptions most of the time.
+  const MutableCFOptions* mutable_cf_options() const {
+    return &(super_version_->mutable_cf_options);
+  }
+  // REQUIRES: DB mutex held
+  // This returns the latest MutableCFOptions, which may be not in effect yet.
+  const MutableCFOptions* GetLatestMutableCFOptions() const {
+    return &mutable_cf_options_;
+  }
+  // REQUIRES: DB mutex held
+  bool SetOptions(
+      const std::unordered_map<std::string, std::string>& options_map);

  InternalStats* internal_stats() { return internal_stats_.get(); }

@@ -178,15 +198,19 @@ class ColumnFamilyData {
  Version* dummy_versions() { return dummy_versions_; }
  void SetMemtable(MemTable* new_mem) { mem_ = new_mem; }
  void SetCurrent(Version* current);
-  void CreateNewMemtable();
+  void CreateNewMemtable(const MemTableOptions& moptions);

-  TableCache* table_cache() { return table_cache_.get(); }
+  TableCache* table_cache() const { return table_cache_.get(); }

  // See documentation in compaction_picker.h
-  Compaction* PickCompaction(LogBuffer* log_buffer);
-  Compaction* CompactRange(int input_level, int output_level,
-                           const InternalKey* begin, const InternalKey* end,
-                           InternalKey** compaction_end);
+  // REQUIRES: DB mutex held
+  Compaction* PickCompaction(const MutableCFOptions& mutable_options,
+                             LogBuffer* log_buffer);
+  Compaction* CompactRange(
+      const MutableCFOptions& mutable_cf_options,
+      int input_level, int output_level, uint32_t output_path_id,
+      const InternalKey* begin, const InternalKey* end,
+      InternalKey** compaction_end);

  CompactionPicker* compaction_picker() { return compaction_picker_.get(); }
  // thread-safe
@@ -218,26 +242,30 @@ class ColumnFamilyData {
  // if its reference count is zero and needs deletion or nullptr if not
  // As argument takes a pointer to allocated SuperVersion to enable
  // the clients to allocate SuperVersion outside of mutex.
+  SuperVersion* InstallSuperVersion(SuperVersion* new_superversion,
+                                    port::Mutex* db_mutex,
+                                    const MutableCFOptions& mutable_cf_options);
  SuperVersion* InstallSuperVersion(SuperVersion* new_superversion,
                                    port::Mutex* db_mutex);

  void ResetThreadLocalSuperVersions();

-  // A Flag indicating whether write needs to slowdown because of there are
-  // too many number of level0 files.
-  bool NeedSlowdownForNumLevel0Files() const {
-    return need_slowdown_for_num_level0_files_;
-  }
-
 private:
  friend class ColumnFamilySet;
-  ColumnFamilyData(const std::string& dbname, uint32_t id,
-                   const std::string& name, Version* dummy_versions,
-                   Cache* table_cache, const ColumnFamilyOptions& options,
-                   const DBOptions* db_options,
-                   const EnvOptions& storage_options,
+  ColumnFamilyData(uint32_t id, const std::string& name,
+                   Version* dummy_versions, Cache* table_cache,
+                   const ColumnFamilyOptions& options,
+                   const DBOptions* db_options, const EnvOptions& env_options,
                   ColumnFamilySet* column_family_set);

+  // Recalculate some small conditions, which are changed only during
+  // compaction, adding new memtable and/or
+  // recalculation of compaction score. These values are used in
+  // DBImpl::MakeRoomForWrite function to decide, if it need to make
+  // a write stall
+  void RecalculateWriteStallConditions(
+      const MutableCFOptions& mutable_cf_options);
+
  uint32_t id_;
  const std::string name_;
  Version* dummy_versions_;  // Head of circular doubly-linked list of versions.
@@ -247,9 +275,10 @@ class ColumnFamilyData {
  bool dropped_;               // true if client dropped it

  const InternalKeyComparator internal_comparator_;
-  const InternalFilterPolicy internal_filter_policy_;

-  Options const options_;
+  const Options options_;
+  const ImmutableCFOptions ioptions_;
+  MutableCFOptions mutable_cf_options_;

  std::unique_ptr<TableCache> table_cache_;

@@ -278,15 +307,13 @@ class ColumnFamilyData {
  // recovered from
  uint64_t log_number_;

-  // A flag indicating whether we should delay writes because
-  // we have too many level 0 files
-  bool need_slowdown_for_num_level0_files_;
-
  // An object that keeps all the compaction stats
  // and picks the next compaction
  std::unique_ptr<CompactionPicker> compaction_picker_;

  ColumnFamilySet* column_family_set_;
+
+  std::unique_ptr<WriteControllerToken> write_controller_token_;
 };

 // ColumnFamilySet has interesting thread-safety requirements
@@ -302,8 +329,8 @@ class ColumnFamilyData {
 // family might get dropped when the DB mutex is released
 // * GetDefault() -- thread safe
 // * GetColumnFamily() -- either inside of DB mutex or call Lock() <-> Unlock()
-// * GetNextColumnFamilyID(), GetMaxColumnFamily(), UpdateMaxColumnFamily() --
-// inside of DB mutex
+// * GetNextColumnFamilyID(), GetMaxColumnFamily(), UpdateMaxColumnFamily(),
+// NumberOfColumnFamilies -- inside of DB mutex
 class ColumnFamilySet {
 public:
  // ColumnFamilySet supports iteration
@@ -328,7 +355,8 @@ class ColumnFamilySet {
  };

  ColumnFamilySet(const std::string& dbname, const DBOptions* db_options,
-                  const EnvOptions& storage_options, Cache* table_cache);
+                  const EnvOptions& env_options, Cache* table_cache,
+                  WriteController* write_controller);
  ~ColumnFamilySet();

  ColumnFamilyData* GetDefault() const;
@@ -342,6 +370,7 @@ class ColumnFamilySet {
  uint32_t GetNextColumnFamilyID();
  uint32_t GetMaxColumnFamily();
  void UpdateMaxColumnFamily(uint32_t new_max_column_family);
+  size_t NumberOfColumnFamilies() const;

  ColumnFamilyData* CreateColumnFamily(const std::string& name, uint32_t id,
                                       Version* dummy_version,
@@ -380,8 +409,9 @@ class ColumnFamilySet {

  const std::string db_name_;
  const DBOptions* const db_options_;
-  const EnvOptions storage_options_;
+  const EnvOptions env_options_;
  Cache* table_cache_;
+  WriteController* write_controller_;
  std::atomic_flag spin_lock_;
 };

@@ -389,8 +419,11 @@ class ColumnFamilySet {
 // memtables of different column families (specified by ID in the write batch)
 class ColumnFamilyMemTablesImpl : public ColumnFamilyMemTables {
 public:
-  explicit ColumnFamilyMemTablesImpl(ColumnFamilySet* column_family_set)
-      : column_family_set_(column_family_set), current_(nullptr) {}
+  explicit ColumnFamilyMemTablesImpl(ColumnFamilySet* column_family_set,
+                                     FlushScheduler* flush_scheduler)
+      : column_family_set_(column_family_set),
+        current_(nullptr),
+        flush_scheduler_(flush_scheduler) {}

  // sets current_ to ColumnFamilyData with column_family_id
  // returns false if column family doesn't exist
@@ -409,10 +442,18 @@ class ColumnFamilyMemTablesImpl : public ColumnFamilyMemTables {
  // Returns column family handle for the selected column family
  virtual ColumnFamilyHandle* GetColumnFamilyHandle() override;

+  virtual void CheckMemtableFull() override;
+
 private:
  ColumnFamilySet* column_family_set_;
  ColumnFamilyData* current_;
+  FlushScheduler* flush_scheduler_;
  ColumnFamilyHandleInternal handle_;
 };

+extern uint32_t GetColumnFamilyID(ColumnFamilyHandle* column_family);
+
+extern const Comparator* GetColumnFamilyUserComparator(
+    ColumnFamilyHandle* column_family);
+
 }  // namespace rocksdb
@@ -408,9 +408,15 @@ TEST(ColumnFamilyTest, WriteBatchFailure) {
  Open();
  CreateColumnFamiliesAndReopen({"one", "two"});
  WriteBatch batch;
+  batch.Put(handles_[0], Slice("existing"), Slice("column-family"));
  batch.Put(handles_[1], Slice("non-existing"), Slice("column-family"));
  ASSERT_OK(db_->Write(WriteOptions(), &batch));
  DropColumnFamilies({1});
+  WriteOptions woptions_ignore_missing_cf;
+  woptions_ignore_missing_cf.ignore_missing_column_families = true;
+  batch.Put(handles_[0], Slice("still here"), Slice("column-family"));
+  ASSERT_OK(db_->Write(woptions_ignore_missing_cf, &batch));
+  ASSERT_EQ("column-family", Get(0, "still here"));
  Status s = db_->Write(WriteOptions(), &batch);
  ASSERT_TRUE(s.IsInvalidArgument());
  Close();
@@ -746,10 +752,10 @@ TEST(ColumnFamilyTest, DifferentCompactionStyles) {
  default_cf.num_levels = 3;
  default_cf.write_buffer_size = 64 << 10;  // 64KB
  default_cf.target_file_size_base = 30 << 10;
-  default_cf.filter_policy = nullptr;
-  default_cf.no_block_cache = true;
  default_cf.source_compaction_factor = 100;
-  default_cf.disable_seek_compaction = false;
+  BlockBasedTableOptions table_options;
+  table_options.no_block_cache = true;
+  default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));

  one.compaction_style = kCompactionStyleUniversal;
  // trigger compaction if there are >= 4 files
@@ -764,33 +770,6 @@ TEST(ColumnFamilyTest, DifferentCompactionStyles) {

  Reopen({default_cf, one, two});

-  // SETUP column family "default" - test read compaction
-  ASSERT_EQ("", FilesPerLevel(0));
-  PutRandomData(0, 1, 4096);
-  ASSERT_OK(Flush(0));
-  ASSERT_EQ("0,0,1", FilesPerLevel(0));
-  // write 8MB
-  PutRandomData(0, 2000, 4096);
-  ASSERT_OK(Flush(0));
-  // clear levels 0 and 1
-  dbfull()->TEST_CompactRange(0, nullptr, nullptr, handles_[0]);
-  dbfull()->TEST_CompactRange(1, nullptr, nullptr, handles_[0]);
-  ASSERT_EQ(NumTableFilesAtLevel(0, 0), 0);
-  ASSERT_EQ(NumTableFilesAtLevel(1, 0), 0);
-  // write some new keys into level 0 and 1
-  PutRandomData(0, 1024, 512);
-  ASSERT_OK(Flush(0));
-  WaitForCompaction();
-  PutRandomData(0, 10, 512);
-  ASSERT_OK(Flush(0));
-  // remember number of files in each level
-  int l1 = NumTableFilesAtLevel(0, 0);
-  int l2 = NumTableFilesAtLevel(1, 0);
-  int l3 = NumTableFilesAtLevel(2, 0);
-  ASSERT_NE(l1, 0);
-  ASSERT_NE(l2, 0);
-  ASSERT_NE(l3, 0);
-
  // SETUP column family "one" -- universal style
  for (int i = 0; i < one.level0_file_num_compaction_trigger - 1; ++i) {
    PutRandomData(1, 11, 10000);
@@ -805,12 +784,6 @@ TEST(ColumnFamilyTest, DifferentCompactionStyles) {
    ASSERT_EQ(std::to_string(i + 1), FilesPerLevel(2));
  }

-  // TRIGGER compaction "default"
-  // read a bunch of times, trigger read compaction
-  for (int i = 0; i < 200000; ++i) {
-    Get(0, std::to_string(i));
-  }
-
  // TRIGGER compaction "one"
  PutRandomData(1, 12, 10000);

@@ -820,13 +793,6 @@ TEST(ColumnFamilyTest, DifferentCompactionStyles) {
  // WAIT for compactions
  WaitForCompaction();

-  // VERIFY compaction "default"
-  // verify that the number of files have decreased
-  // in some level, indicating that there was a compaction
-  ASSERT_TRUE(NumTableFilesAtLevel(0, 0) < l1 ||
-              NumTableFilesAtLevel(1, 0) < l2 ||
-              NumTableFilesAtLevel(2, 0) < l3);
-
  // VERIFY compaction "one"
  ASSERT_EQ("1", FilesPerLevel(1));

@@ -898,6 +864,7 @@ TEST(ColumnFamilyTest, NewIteratorsTest) {
 TEST(ColumnFamilyTest, ReadOnlyDBTest) {
  Open();
  CreateColumnFamiliesAndReopen({"one", "two", "three", "four"});
+  ASSERT_OK(Put(0, "a", "b"));
  ASSERT_OK(Put(1, "foo", "bla"));
  ASSERT_OK(Put(2, "foo", "blabla"));
  ASSERT_OK(Put(3, "foo", "blablabla"));
@@ -911,6 +878,29 @@ TEST(ColumnFamilyTest, ReadOnlyDBTest) {
  ASSERT_EQ("bla", Get(1, "foo"));
  ASSERT_EQ("blablablabla", Get(2, "foo"));

+
+  // test newiterators
+  {
+    std::vector<Iterator*> iterators;
+    ASSERT_OK(db_->NewIterators(ReadOptions(), handles_, &iterators));
+    for (auto it : iterators) {
+      it->SeekToFirst();
+    }
+    ASSERT_EQ(IterStatus(iterators[0]), "a->b");
+    ASSERT_EQ(IterStatus(iterators[1]), "foo->bla");
+    ASSERT_EQ(IterStatus(iterators[2]), "foo->blablablabla");
+    for (auto it : iterators) {
+      it->Next();
+    }
+    ASSERT_EQ(IterStatus(iterators[0]), "(invalid)");
+    ASSERT_EQ(IterStatus(iterators[1]), "(invalid)");
+    ASSERT_EQ(IterStatus(iterators[2]), "(invalid)");
+
+    for (auto it : iterators) {
+      delete it;
+    }
+  }
+
  Close();
  // can't open dropped column family
  Status s = OpenReadOnly({"default", "one", "two"});
@@ -930,7 +920,9 @@ TEST(ColumnFamilyTest, DontRollEmptyLogs) {
  }
  int num_writable_file_start = env_->GetNumberOfNewWritableFileCalls();
  // this will trigger the flushes
-  ASSERT_OK(db_->Write(WriteOptions(), nullptr));
+  for (size_t i = 0; i <= 4; ++i) {
+    ASSERT_OK(Flush(i));
+  }

  for (int i = 0; i < 4; ++i) {
    dbfull()->TEST_WaitForFlushMemTable(handles_[i]);
@@ -970,6 +962,15 @@ TEST(ColumnFamilyTest, FlushStaleColumnFamilies) {
  Close();
 }

+TEST(ColumnFamilyTest, CreateMissingColumnFamilies) {
+  Status s = TryOpen({"one", "two"});
+  ASSERT_TRUE(!s.ok());
+  db_options_.create_missing_column_families = true;
+  s = TryOpen({"default", "one", "two"});
+  ASSERT_TRUE(s.ok());
+  Close();
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -9,7 +9,10 @@

 #include "db/compaction.h"

+#ifndef __STDC_FORMAT_MACROS
 #define __STDC_FORMAT_MACROS
+#endif
+
 #include <inttypes.h>
 #include <vector>

@@ -18,28 +21,30 @@

 namespace rocksdb {

-static uint64_t TotalFileSize(const std::vector<FileMetaData*>& files) {
+uint64_t TotalFileSize(const std::vector<FileMetaData*>& files) {
  uint64_t sum = 0;
  for (size_t i = 0; i < files.size() && files[i]; i++) {
-    sum += files[i]->file_size;
+    sum += files[i]->fd.GetFileSize();
  }
  return sum;
 }

-Compaction::Compaction(Version* input_version, int level, int out_level,
+Compaction::Compaction(Version* input_version, int start_level, int out_level,
                       uint64_t target_file_size,
                       uint64_t max_grandparent_overlap_bytes,
-                       bool seek_compaction, bool enable_compression,
+                       uint32_t output_path_id,
+                       CompressionType output_compression, bool seek_compaction,
                       bool deletion_compaction)
-    : level_(level),
-      out_level_(out_level),
+    : start_level_(start_level),
+      output_level_(out_level),
      max_output_file_size_(target_file_size),
      max_grandparent_overlap_bytes_(max_grandparent_overlap_bytes),
      input_version_(input_version),
      number_levels_(input_version_->NumberLevels()),
      cfd_(input_version_->cfd_),
+      output_path_id_(output_path_id),
+      output_compression_(output_compression),
      seek_compaction_(seek_compaction),
-      enable_compression_(enable_compression),
      deletion_compaction_(deletion_compaction),
      grandparent_index_(0),
      seen_key_(false),
@@ -51,7 +56,6 @@ Compaction::Compaction(Version* input_version, int level, int out_level,
      is_full_compaction_(false),
      is_manual_compaction_(false),
      level_ptrs_(std::vector<size_t>(number_levels_)) {
-
  cfd_->Ref();
  input_version_->Ref();
  edit_ = new VersionEdit();
@@ -59,6 +63,12 @@ Compaction::Compaction(Version* input_version, int level, int out_level,
  for (int i = 0; i < number_levels_; i++) {
    level_ptrs_[i] = 0;
  }
+  int num_levels = output_level_ - start_level_ + 1;
+  input_levels_.resize(num_levels);
+  inputs_.resize(num_levels);
+  for (int i = 0; i < num_levels; ++i) {
+    inputs_[i].level = start_level_ + i;
+  }
 }

 Compaction::~Compaction() {
@@ -73,43 +83,50 @@ Compaction::~Compaction() {
  }
 }

+void Compaction::GenerateFileLevels() {
+  input_levels_.resize(num_input_levels());
+  for (int which = 0; which < num_input_levels(); which++) {
+    DoGenerateFileLevel(&input_levels_[which], inputs_[which].files, &arena_);
+  }
+}
+
 bool Compaction::IsTrivialMove() const {
  // Avoid a move if there is lots of overlapping grandparent data.
  // Otherwise, the move could create a parent file that will require
  // a very expensive merge later on.
-  // If level_== out_level_, the purpose is to force compaction filter to be
-  // applied to that level, and thus cannot be a trivia move.
-  return (level_ != out_level_ &&
+  // If start_level_== output_level_, the purpose is to force compaction
+  // filter to be applied to that level, and thus cannot be a trivia move.
+  return (start_level_ != output_level_ &&
+          num_input_levels() == 2 &&
          num_input_files(0) == 1 &&
          num_input_files(1) == 0 &&
          TotalFileSize(grandparents_) <= max_grandparent_overlap_bytes_);
 }

-bool Compaction::IsDeletionCompaction() const { return deletion_compaction_; }
-
 void Compaction::AddInputDeletions(VersionEdit* edit) {
-  for (int which = 0; which < 2; which++) {
+  for (int which = 0; which < num_input_levels(); which++) {
    for (size_t i = 0; i < inputs_[which].size(); i++) {
-      edit->DeleteFile(level_ + which, inputs_[which][i]->number);
+      edit->DeleteFile(level(which), inputs_[which][i]->fd.GetNumber());
    }
  }
 }

-bool Compaction::IsBaseLevelForKey(const Slice& user_key) {
-  assert(cfd_->options()->compaction_style != kCompactionStyleFIFO);
-  if (cfd_->options()->compaction_style == kCompactionStyleUniversal) {
+bool Compaction::KeyNotExistsBeyondOutputLevel(const Slice& user_key) {
+  assert(cfd_->ioptions()->compaction_style != kCompactionStyleFIFO);
+  if (cfd_->ioptions()->compaction_style == kCompactionStyleUniversal) {
    return bottommost_level_;
  }
  // Maybe use binary search to find right entry instead of linear search?
  const Comparator* user_cmp = cfd_->user_comparator();
-  for (int lvl = level_ + 2; lvl < number_levels_; lvl++) {
+  for (int lvl = output_level_ + 1; lvl < number_levels_; lvl++) {
    const std::vector<FileMetaData*>& files = input_version_->files_[lvl];
    for (; level_ptrs_[lvl] < files.size(); ) {
      FileMetaData* f = files[level_ptrs_[lvl]];
      if (user_cmp->Compare(user_key, f->largest.user_key()) <= 0) {
        // We've advanced far enough
        if (user_cmp->Compare(user_key, f->smallest.user_key()) >= 0) {
-          // Key falls in this file's range, so definitely not base level
+          // Key falls in this file's range, so definitely
+          // exists beyond output level
          return false;
        }
        break;
@@ -127,7 +144,7 @@ bool Compaction::ShouldStopBefore(const Slice& internal_key) {
      icmp->Compare(internal_key,
                    grandparents_[grandparent_index_]->largest.Encode()) > 0) {
    if (seen_key_) {
-      overlapped_bytes_ += grandparents_[grandparent_index_]->file_size;
+      overlapped_bytes_ += grandparents_[grandparent_index_]->fd.GetFileSize();
    }
    assert(grandparent_index_ + 1 >= grandparents_.size() ||
           icmp->Compare(grandparents_[grandparent_index_]->largest.Encode(),
@@ -147,34 +164,34 @@ bool Compaction::ShouldStopBefore(const Slice& internal_key) {
 }

 // Mark (or clear) each file that is being compacted
-void Compaction::MarkFilesBeingCompacted(bool value) {
-  for (int i = 0; i < 2; i++) {
-    std::vector<FileMetaData*> v = inputs_[i];
+void Compaction::MarkFilesBeingCompacted(bool mark_as_compacted) {
+  for (int i = 0; i < num_input_levels(); i++) {
    for (unsigned int j = 0; j < inputs_[i].size(); j++) {
-      assert(value ? !inputs_[i][j]->being_compacted :
-                      inputs_[i][j]->being_compacted);
-      inputs_[i][j]->being_compacted = value;
+      assert(mark_as_compacted ? !inputs_[i][j]->being_compacted :
+                                  inputs_[i][j]->being_compacted);
+      inputs_[i][j]->being_compacted = mark_as_compacted;
    }
  }
 }

 // Is this compaction producing files at the bottommost level?
-void Compaction::SetupBottomMostLevel(bool isManual) {
-  assert(cfd_->options()->compaction_style != kCompactionStyleFIFO);
-  if (cfd_->options()->compaction_style == kCompactionStyleUniversal) {
+void Compaction::SetupBottomMostLevel(bool is_manual) {
+  assert(cfd_->ioptions()->compaction_style != kCompactionStyleFIFO);
+  if (cfd_->ioptions()->compaction_style == kCompactionStyleUniversal) {
    // If universal compaction style is used and manual
    // compaction is occuring, then we are guaranteed that
    // all files will be picked in a single compaction
    // run. We can safely set bottommost_level_ = true.
    // If it is not manual compaction, then bottommost_level_
    // is already set when the Compaction was created.
-    if (isManual) {
+    if (is_manual) {
      bottommost_level_ = true;
    }
    return;
  }
  bottommost_level_ = true;
-  for (int i = output_level() + 1; i < number_levels_; i++) {
+  // checks whether there are files living beyond the output_level.
+  for (int i = output_level_ + 1; i < number_levels_; i++) {
    if (input_version_->NumLevelFiles(i) > 0) {
      bottommost_level_ = false;
      break;
@@ -200,7 +217,7 @@ void Compaction::ReleaseCompactionFiles(Status status) {
 }

 void Compaction::ResetNextCompactionIndex() {
-  input_version_->ResetNextCompactionIndex(level_);
+  input_version_->ResetNextCompactionIndex(start_level_);
 }

 namespace {
@@ -212,9 +229,9 @@ int InputSummary(const std::vector<FileMetaData*>& files, char* output,
    int sz = len - write;
    int ret;
    char sztxt[16];
-    AppendHumanBytes(files.at(i)->file_size, sztxt, 16);
-    ret = snprintf(output + write, sz, "%" PRIu64 "(%s) ", files.at(i)->number,
-                   sztxt);
+    AppendHumanBytes(files.at(i)->fd.GetFileSize(), sztxt, 16);
+    ret = snprintf(output + write, sz, "%" PRIu64 "(%s) ",
+                   files.at(i)->fd.GetNumber(), sztxt);
    if (ret < 0 || ret >= sz) break;
    write += ret;
  }
@@ -227,27 +244,44 @@ void Compaction::Summary(char* output, int len) {
  int write =
      snprintf(output, len, "Base version %" PRIu64
                            " Base level %d, seek compaction:%d, inputs: [",
-               input_version_->GetVersionNumber(), level_, seek_compaction_);
+               input_version_->GetVersionNumber(),
+               start_level_, seek_compaction_);
  if (write < 0 || write >= len) {
    return;
  }

-  write += InputSummary(inputs_[0], output + write, len - write);
-  if (write < 0 || write >= len) {
-    return;
-  }
-
-  write += snprintf(output + write, len - write, "], [");
-  if (write < 0 || write >= len) {
-    return;
-  }
-
-  write += InputSummary(inputs_[1], output + write, len - write);
-  if (write < 0 || write >= len) {
-    return;
+  for (int level = 0; level < num_input_levels(); ++level) {
+    if (level > 0) {
+      write += snprintf(output + write, len - write, "], [");
+      if (write < 0 || write >= len) {
+        return;
+      }
+    }
+    write += InputSummary(inputs_[level].files, output + write, len - write);
+    if (write < 0 || write >= len) {
+      return;
+    }
  }

  snprintf(output + write, len - write, "]");
 }

+uint64_t Compaction::OutputFilePreallocationSize(
+    const MutableCFOptions& mutable_options) {
+  uint64_t preallocation_size = 0;
+
+  if (cfd_->ioptions()->compaction_style == kCompactionStyleLevel) {
+    preallocation_size = mutable_options.MaxFileSizeForLevel(output_level());
+  } else {
+    for (int level = 0; level < num_input_levels(); ++level) {
+      for (const auto& f : inputs_[level].files) {
+        preallocation_size += f->fd.GetFileSize();
+      }
+    }
+  }
+  // Over-estimate slightly so we don't end up just barely crossing
+  // the threshold
+  return preallocation_size * 1.1;
+}
+
 }  // namespace rocksdb
@@ -8,62 +8,120 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.

 #pragma once
+#include "util/arena.h"
+#include "util/autovector.h"
+#include "util/mutable_cf_options.h"
 #include "db/version_set.h"

 namespace rocksdb {

+// The structure that manages compaction input files associated
+// with the same physical level.
+struct CompactionInputFiles {
+  int level;
+  std::vector<FileMetaData*> files;
+  inline bool empty() const { return files.empty(); }
+  inline size_t size() const { return files.size(); }
+  inline void clear() { files.clear(); }
+  inline FileMetaData* operator[](int i) const { return files[i]; }
+};
+
 class Version;
 class ColumnFamilyData;

 // A Compaction encapsulates information about a compaction.
 class Compaction {
 public:
+  // No copying allowed
+  Compaction(const Compaction&) = delete;
+  void operator=(const Compaction&) = delete;
+
  ~Compaction();

-  // Return the level that is being compacted.  Inputs from "level"
-  // will be merged.
-  int level() const { return level_; }
+  // Returns the level associated to the specified compaction input level.
+  // If compaction_input_level is not specified, then input_level is set to 0.
+  int level(int compaction_input_level = 0) const {
+    return inputs_[compaction_input_level].level;
+  }

  // Outputs will go to this level
-  int output_level() const { return out_level_; }
+  int output_level() const { return output_level_; }
+
+  // Returns the number of input levels in this compaction.
+  int num_input_levels() const { return inputs_.size(); }

  // Return the object that holds the edits to the descriptor done
  // by this compaction.
-  VersionEdit* edit() { return edit_; }
+  VersionEdit* edit() const { return edit_; }

-  // "which" must be either 0 or 1
-  int num_input_files(int which) const { return inputs_[which].size(); }
+  // Returns the number of input files associated to the specified
+  // compaction input level.
+  // The function will return 0 if when "compaction_input_level" < 0
+  // or "compaction_input_level" >= "num_input_levels()".
+  int num_input_files(size_t compaction_input_level) const {
+    if (compaction_input_level < inputs_.size()) {
+      return inputs_[compaction_input_level].size();
+    }
+    return 0;
+  }

  // Returns input version of the compaction
  Version* input_version() const { return input_version_; }

+  // Returns the ColumnFamilyData associated with the compaction.
  ColumnFamilyData* column_family_data() const { return cfd_; }

-  // Return the ith input file at "level()+which" ("which" must be 0 or 1).
-  FileMetaData* input(int which, int i) const { return inputs_[which][i]; }
+  // Returns the file meta data of the 'i'th input file at the
+  // specified compaction input level.
+  // REQUIREMENT: "compaction_input_level" must be >= 0 and
+  //              < "input_levels()"
+  FileMetaData* input(size_t compaction_input_level, int i) const {
+    assert(compaction_input_level < inputs_.size());
+    return inputs_[compaction_input_level][i];
+  }

-  std::vector<FileMetaData*>* inputs(int which) { return &inputs_[which]; }
+  // Returns the list of file meta data of the specified compaction
+  // input level.
+  // REQUIREMENT: "compaction_input_level" must be >= 0 and
+  //              < "input_levels()"
+  std::vector<FileMetaData*>* const inputs(size_t compaction_input_level) {
+    assert(compaction_input_level < inputs_.size());
+    return &inputs_[compaction_input_level].files;
+  }
+
+  // Returns the FileLevel of the specified compaction input level.
+  FileLevel* input_levels(int compaction_input_level) {
+    return &input_levels_[compaction_input_level];
+  }

  // Maximum size of files to build during this compaction.
  uint64_t MaxOutputFileSize() const { return max_output_file_size_; }

-  // Whether compression will be enabled for compaction outputs
-  bool enable_compression() const { return enable_compression_; }
+  // What compression for output
+  CompressionType OutputCompressionType() const { return output_compression_; }
+
+  // Whether need to write output file to second DB path.
+  uint32_t GetOutputPathId() const { return output_path_id_; }
+
+  // Generate input_levels_ from inputs_
+  // Should be called when inputs_ is stable
+  void GenerateFileLevels();

  // Is this a trivial compaction that can be implemented by just
  // moving a single input file to the next level (no merging or splitting)
  bool IsTrivialMove() const;

-  // If true, just delete all files in inputs_[0]
-  bool IsDeletionCompaction() const;
+  // If true, then the comaction can be done by simply deleting input files.
+  bool IsDeletionCompaction() const {
+    return deletion_compaction_;
+  }

  // Add all inputs to this compaction as delete operations to *edit.
  void AddInputDeletions(VersionEdit* edit);

-  // Returns true if the information we have available guarantees that
-  // the compaction is producing data in "level+1" for which no data exists
-  // in levels greater than "level+1".
-  bool IsBaseLevelForKey(const Slice& user_key);
+  // Returns true if the available information we have guarantees that
+  // the input "user_key" does not exist in any level beyond "output_level()".
+  bool KeyNotExistsBeyondOutputLevel(const Slice& user_key);

  // Returns true iff we should stop building the current output
  // before processing "internal_key".
@@ -77,6 +135,9 @@ class Compaction {
  // Delete this compaction from the list of running compactions.
  void ReleaseCompactionFiles(Status status);

+  // Returns the summary of the compaction in "output" with maximum "len"
+  // in bytes.  The caller is responsible for the memory management of
+  // "output".
  void Summary(char* output, int len);

  // Return the score that was used to pick this compaction run.
@@ -91,44 +152,61 @@ class Compaction {
  // Was this compaction triggered manually by the client?
  bool IsManualCompaction() { return is_manual_compaction_; }

+  // Return the MutableCFOptions that should be used throughout the compaction
+  // procedure
+  const MutableCFOptions* mutable_cf_options() { return &mutable_cf_options_; }
+
+  // Returns the size in bytes that the output file should be preallocated to.
+  // In level compaction, that is max_file_size_. In universal compaction, that
+  // is the sum of all input file sizes.
+  uint64_t OutputFilePreallocationSize(const MutableCFOptions& mutable_options);
+
 private:
  friend class CompactionPicker;
  friend class UniversalCompactionPicker;
  friend class FIFOCompactionPicker;
  friend class LevelCompactionPicker;

-  Compaction(Version* input_version, int level, int out_level,
+  Compaction(Version* input_version, int start_level, int out_level,
             uint64_t target_file_size, uint64_t max_grandparent_overlap_bytes,
-             bool seek_compaction = false, bool enable_compression = true,
-             bool deletion_compaction = false);
+             uint32_t output_path_id, CompressionType output_compression,
+             bool seek_compaction = false, bool deletion_compaction = false);

-  int level_;
-  int out_level_; // levels to which output files are stored
+  const int start_level_;    // the lowest level to be compacted
+  const int output_level_;  // levels to which output files are stored
  uint64_t max_output_file_size_;
  uint64_t max_grandparent_overlap_bytes_;
+  MutableCFOptions mutable_cf_options_;
  Version* input_version_;
  VersionEdit* edit_;
  int number_levels_;
  ColumnFamilyData* cfd_;
+  Arena arena_;          // Arena used to allocate space for file_levels_

+  uint32_t output_path_id_;
+  CompressionType output_compression_;
  bool seek_compaction_;
-  bool enable_compression_;
-  // if true, just delete files in inputs_[0]
+  // If true, then the comaction can be done by simply deleting input files.
  bool deletion_compaction_;

-  // Each compaction reads inputs from "level_" and "level_+1"
-  std::vector<FileMetaData*> inputs_[2];      // The two sets of inputs
+  // Compaction input files organized by level.
+  autovector<CompactionInputFiles> inputs_;
+
+  // A copy of inputs_, organized more closely in memory
+  autovector<FileLevel, 2> input_levels_;

  // State used to check for number of of overlapping grandparent files
-  // (parent == level_ + 1, grandparent == level_ + 2)
+  // (grandparent == "output_level_ + 1")
+  // This vector is updated by Version::GetOverlappingInputs().
  std::vector<FileMetaData*> grandparents_;
-  size_t grandparent_index_;  // Index in grandparent_starts_
-  bool seen_key_;             // Some output key has been seen
+  size_t grandparent_index_;   // Index in grandparent_starts_
+  bool seen_key_;              // Some output key has been seen
  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.
+                               // and grandparent files
+  int base_index_;    // index of the file in files_[start_level_]
+  int parent_index_;  // index of some file with same range in
+                      // files_[start_level_+1]
+  double score_;      // score that was used to pick this compaction.

  // Is this compaction creating a file in the bottom most level?
  bool bottommost_level_;
@@ -138,21 +216,28 @@ class Compaction {
  // Is this compaction requested by the client?
  bool is_manual_compaction_;

-  // level_ptrs_ holds indices into input_version_->levels_: our state
-  // is that we are positioned at one of the file ranges for each
-  // higher level than the ones involved in this compaction (i.e. for
-  // all L >= level_ + 2).
+  // "level_ptrs_" holds indices into "input_version_->levels_", where each
+  // index remembers which file of an associated level we are currently used
+  // to check KeyNotExistsBeyondOutputLevel() for deletion operation.
+  // As it is for checking KeyNotExistsBeyondOutputLevel(), it only
+  // records indices for all levels beyond "output_level_".
  std::vector<size_t> level_ptrs_;

  // mark (or clear) all files that are being compacted
-  void MarkFilesBeingCompacted(bool);
+  void MarkFilesBeingCompacted(bool mark_as_compacted);

-  // Initialize whether compaction producing files at the bottommost level
-  void SetupBottomMostLevel(bool isManual);
+  // Initialize whether the compaction is producing files at the
+  // bottommost level.
+  //
+  // @see BottomMostLevel()
+  void SetupBottomMostLevel(bool is_manual);

  // 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();
 };

+// Utility function
+extern uint64_t TotalFileSize(const std::vector<FileMetaData*>& files);
+
 }  // namespace rocksdb
@@ -13,6 +13,7 @@
 #include "rocksdb/status.h"
 #include "rocksdb/options.h"
 #include "rocksdb/env.h"
+#include "util/mutable_cf_options.h"

 #include <vector>
 #include <memory>
@@ -26,15 +27,17 @@ class Version;

 class CompactionPicker {
 public:
-  CompactionPicker(const Options* options, const InternalKeyComparator* icmp);
+  CompactionPicker(const ImmutableCFOptions& ioptions,
+                   const InternalKeyComparator* icmp);
  virtual ~CompactionPicker();

  // Pick level and inputs for a new compaction.
  // Returns nullptr if there is no compaction to be done.
  // Otherwise returns a pointer to a heap-allocated object that
  // describes the compaction.  Caller should delete the result.
-  virtual Compaction* PickCompaction(Version* version,
-                                     LogBuffer* log_buffer) = 0;
+  virtual Compaction* PickCompaction(
+      const MutableCFOptions& mutable_cf_options,
+      Version* version, LogBuffer* log_buffer) = 0;

  // Return a compaction object for compacting the range [begin,end] in
  // the specified level.  Returns nullptr if there is nothing in that
@@ -47,10 +50,15 @@ class CompactionPicker {
  // compaction_end will be set to nullptr.
  // Client is responsible for compaction_end storage -- when called,
  // *compaction_end should point to valid InternalKey!
-  virtual Compaction* CompactRange(Version* version, int input_level,
-                                   int output_level, const InternalKey* begin,
-                                   const InternalKey* end,
-                                   InternalKey** compaction_end);
+  virtual Compaction* CompactRange(
+      const MutableCFOptions& mutable_cf_options, Version* version,
+      int input_level, int output_level, uint32_t output_path_id,
+      const InternalKey* begin, const InternalKey* end,
+      InternalKey** compaction_end);
+
+  // Given the current number of levels, returns the lowest allowed level
+  // for compaction input.
+  virtual int MaxInputLevel(int current_num_levels) const = 0;

  // Free up the files that participated in a compaction
  void ReleaseCompactionFiles(Compaction* c, Status status);
@@ -59,19 +67,8 @@ class CompactionPicker {
  // compactions per level
  void SizeBeingCompacted(std::vector<uint64_t>& sizes);

-  // Returns maximum total overlap bytes with grandparent
-  // level (i.e., level+2) before we stop building a single
-  // file in level->level+1 compaction.
-  uint64_t MaxGrandParentOverlapBytes(int level);
-
-  // Returns maximum total bytes of data on a given level.
-  double MaxBytesForLevel(int level);
-
-  // Get the max file size in a given level.
-  uint64_t MaxFileSizeForLevel(int level) const;
-
 protected:
-  int NumberLevels() const { return num_levels_; }
+  int NumberLevels() const { return ioptions_.num_levels; }

  // Stores the minimal range that covers all entries in inputs in
  // *smallest, *largest.
@@ -98,8 +95,6 @@ class CompactionPicker {
  // Will return false if it is impossible to apply this compaction.
  bool ExpandWhileOverlapping(Compaction* c);

-  uint64_t ExpandedCompactionByteSizeLimit(int level);
-
  // Returns true if any one of the specified files are being compacted
  bool FilesInCompaction(std::vector<FileMetaData*>& files);

@@ -108,74 +103,98 @@ class CompactionPicker {
                               const InternalKey* largest, int level,
                               int* index);

-  void SetupOtherInputs(Compaction* c);
+  void SetupOtherInputs(const MutableCFOptions& mutable_cf_options,
+                        Compaction* c);
+
+  const ImmutableCFOptions& ioptions_;

  // record all the ongoing compactions for all levels
  std::vector<std::set<Compaction*>> compactions_in_progress_;

-  // Per-level target file size.
-  std::unique_ptr<uint64_t[]> max_file_size_;
-
-  // Per-level max bytes
-  std::unique_ptr<uint64_t[]> level_max_bytes_;
-
-  const Options* const options_;

 private:
-  int num_levels_;
-
  const InternalKeyComparator* const icmp_;
 };

 class UniversalCompactionPicker : public CompactionPicker {
 public:
-  UniversalCompactionPicker(const Options* options,
+  UniversalCompactionPicker(const ImmutableCFOptions& ioptions,
                            const InternalKeyComparator* icmp)
-      : CompactionPicker(options, icmp) {}
-  virtual Compaction* PickCompaction(Version* version,
-                                     LogBuffer* log_buffer) override;
+      : CompactionPicker(ioptions, icmp) {}
+  virtual Compaction* PickCompaction(
+      const MutableCFOptions& mutable_cf_options,
+      Version* version, LogBuffer* log_buffer) override;
+
+  // The maxinum allowed input level.  Always return 0.
+  virtual int MaxInputLevel(int current_num_levels) const override {
+    return 0;
+  }

 private:
  // Pick Universal compaction to limit read amplification
-  Compaction* PickCompactionUniversalReadAmp(Version* version, double score,
-                                             unsigned int ratio,
-                                             unsigned int num_files,
-                                             LogBuffer* log_buffer);
+  Compaction* PickCompactionUniversalReadAmp(
+      const MutableCFOptions& mutable_cf_options,
+      Version* version, double score, unsigned int ratio,
+      unsigned int num_files, LogBuffer* log_buffer);

  // Pick Universal compaction to limit space amplification.
-  Compaction* PickCompactionUniversalSizeAmp(Version* version, double score,
-                                             LogBuffer* log_buffer);
+  Compaction* PickCompactionUniversalSizeAmp(
+      const MutableCFOptions& mutable_cf_options,
+      Version* version, double score, LogBuffer* log_buffer);
+
+  // Pick a path ID to place a newly generated file, with its estimated file
+  // size.
+  static uint32_t GetPathId(const ImmutableCFOptions& ioptions,
+                            uint64_t file_size);
 };

 class LevelCompactionPicker : public CompactionPicker {
 public:
-  LevelCompactionPicker(const Options* options,
+  LevelCompactionPicker(const ImmutableCFOptions& ioptions,
                        const InternalKeyComparator* icmp)
-      : CompactionPicker(options, icmp) {}
-  virtual Compaction* PickCompaction(Version* version,
-                                     LogBuffer* log_buffer) override;
+      : CompactionPicker(ioptions, icmp) {}
+  virtual Compaction* PickCompaction(
+      const MutableCFOptions& mutable_cf_options,
+      Version* version, LogBuffer* log_buffer) override;
+
+  // Returns current_num_levels - 2, meaning the last level cannot be
+  // compaction input level.
+  virtual int MaxInputLevel(int current_num_levels) const override {
+    return current_num_levels - 2;
+  }

 private:
  // For the specfied level, pick a compaction.
  // Returns nullptr if there is no compaction to be done.
  // If level is 0 and there is already a compaction on that level, this
  // function will return nullptr.
-  Compaction* PickCompactionBySize(Version* version, int level, double score);
+  Compaction* PickCompactionBySize(const MutableCFOptions& mutable_cf_options,
+      Version* version, int level, double score);
 };

 class FIFOCompactionPicker : public CompactionPicker {
 public:
-  FIFOCompactionPicker(const Options* options,
+  FIFOCompactionPicker(const ImmutableCFOptions& ioptions,
                       const InternalKeyComparator* icmp)
-      : CompactionPicker(options, icmp) {}
+      : CompactionPicker(ioptions, icmp) {}

-  virtual Compaction* PickCompaction(Version* version,
-                                     LogBuffer* log_buffer) override;
+  virtual Compaction* PickCompaction(
+      const MutableCFOptions& mutable_cf_options,
+      Version* version, LogBuffer* log_buffer) override;

-  virtual Compaction* CompactRange(Version* version, int input_level,
-                                   int output_level, const InternalKey* begin,
-                                   const InternalKey* end,
-                                   InternalKey** compaction_end) override;
+  virtual Compaction* CompactRange(
+      const MutableCFOptions& mutable_cf_options, Version* version,
+      int input_level, int output_level, uint32_t output_path_id,
+      const InternalKey* begin, const InternalKey* end,
+      InternalKey** compaction_end) override;
+
+  // The maxinum allowed input level.  Always return 0.
+  virtual int MaxInputLevel(int current_num_levels) const override {
+    return 0;
+  }
 };

+// Utility function
+extern uint64_t TotalCompensatedFileSize(const std::vector<FileMetaData*>& files);
+
 }  // namespace rocksdb
@@ -45,7 +45,9 @@ class CorruptionTest {

    db_ = nullptr;
    options_.create_if_missing = true;
-    options_.block_size_deviation = 0; // make unit test pass for now
+    BlockBasedTableOptions table_options;
+    table_options.block_size_deviation = 0;  // make unit test pass for now
+    options_.table_factory.reset(NewBlockBasedTableFactory(table_options));
    Reopen();
    options_.create_if_missing = false;
  }
@@ -60,9 +62,11 @@ class CorruptionTest {
    db_ = nullptr;
    Options opt = (options ? *options : options_);
    opt.env = &env_;
-    opt.block_cache = tiny_cache_;
-    opt.block_size_deviation = 0;
    opt.arena_block_size = 4096;
+    BlockBasedTableOptions table_options;
+    table_options.block_cache = tiny_cache_;
+    table_options.block_size_deviation = 0;
+    opt.table_factory.reset(NewBlockBasedTableFactory(table_options));
    return DB::Open(opt, dbname_, &db_);
  }

@@ -127,7 +131,7 @@ class CorruptionTest {
    ASSERT_GE(max_expected, correct);
  }

-  void CorruptFile(const std::string fname, int offset, int bytes_to_corrupt) {
+  void CorruptFile(const std::string& fname, int offset, int bytes_to_corrupt) {
    struct stat sbuf;
    if (stat(fname.c_str(), &sbuf) != 0) {
      const char* msg = strerror(errno);
@@ -328,6 +332,9 @@ TEST(CorruptionTest, CorruptedDescriptor) {
 }

 TEST(CorruptionTest, CompactionInputError) {
+  Options options;
+  options.max_background_flushes = 0;
+  Reopen(&options);
  Build(10);
  DBImpl* dbi = reinterpret_cast<DBImpl*>(db_);
  dbi->TEST_FlushMemTable();
@@ -347,6 +354,7 @@ TEST(CorruptionTest, CompactionInputErrorParanoid) {
  options.paranoid_checks = true;
  options.write_buffer_size = 131072;
  options.max_write_buffer_number = 2;
+  options.max_background_flushes = 0;
  Reopen(&options);
  DBImpl* dbi = reinterpret_cast<DBImpl*>(db_);

@@ -0,0 +1,346 @@
+//  Copyright (c) 2014, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+
+#include "db/db_impl.h"
+#include "rocksdb/db.h"
+#include "rocksdb/env.h"
+#include "table/meta_blocks.h"
+#include "table/cuckoo_table_factory.h"
+#include "table/cuckoo_table_reader.h"
+#include "util/testharness.h"
+#include "util/testutil.h"
+
+namespace rocksdb {
+
+class CuckooTableDBTest {
+ private:
+  std::string dbname_;
+  Env* env_;
+  DB* db_;
+
+ public:
+  CuckooTableDBTest() : env_(Env::Default()) {
+    dbname_ = test::TmpDir() + "/cuckoo_table_db_test";
+    ASSERT_OK(DestroyDB(dbname_, Options()));
+    db_ = nullptr;
+    Reopen();
+  }
+
+  ~CuckooTableDBTest() {
+    delete db_;
+    ASSERT_OK(DestroyDB(dbname_, Options()));
+  }
+
+  Options CurrentOptions() {
+    Options options;
+    options.table_factory.reset(NewCuckooTableFactory());
+    options.memtable_factory.reset(NewHashLinkListRepFactory(4, 0, 3, true));
+    options.allow_mmap_reads = true;
+    options.create_if_missing = true;
+    options.max_mem_compaction_level = 0;
+    return options;
+  }
+
+  DBImpl* dbfull() {
+    return reinterpret_cast<DBImpl*>(db_);
+  }
+
+  // The following util methods are copied from plain_table_db_test.
+  void Reopen(Options* options = nullptr) {
+    delete db_;
+    db_ = nullptr;
+    Options opts;
+    if (options != nullptr) {
+      opts = *options;
+    } else {
+      opts = CurrentOptions();
+      opts.create_if_missing = true;
+    }
+    ASSERT_OK(DB::Open(opts, dbname_, &db_));
+  }
+
+  Status Put(const Slice& k, const Slice& v) {
+    return db_->Put(WriteOptions(), k, v);
+  }
+
+  Status Delete(const std::string& k) {
+    return db_->Delete(WriteOptions(), k);
+  }
+
+  std::string Get(const std::string& k) {
+    ReadOptions options;
+    std::string result;
+    Status s = db_->Get(options, k, &result);
+    if (s.IsNotFound()) {
+      result = "NOT_FOUND";
+    } else if (!s.ok()) {
+      result = s.ToString();
+    }
+    return result;
+  }
+
+  int NumTableFilesAtLevel(int level) {
+    std::string property;
+    ASSERT_TRUE(
+        db_->GetProperty("rocksdb.num-files-at-level" + NumberToString(level),
+                         &property));
+    return atoi(property.c_str());
+  }
+
+  // Return spread of files per level
+  std::string FilesPerLevel() {
+    std::string result;
+    int last_non_zero_offset = 0;
+    for (int level = 0; level < db_->NumberLevels(); level++) {
+      int f = NumTableFilesAtLevel(level);
+      char buf[100];
+      snprintf(buf, sizeof(buf), "%s%d", (level ? "," : ""), f);
+      result += buf;
+      if (f > 0) {
+        last_non_zero_offset = result.size();
+      }
+    }
+    result.resize(last_non_zero_offset);
+    return result;
+  }
+};
+
+TEST(CuckooTableDBTest, Flush) {
+  // Try with empty DB first.
+  ASSERT_TRUE(dbfull() != nullptr);
+  ASSERT_EQ("NOT_FOUND", Get("key2"));
+
+  // Add some values to db.
+  Options options = CurrentOptions();
+  Reopen(&options);
+
+  ASSERT_OK(Put("key1", "v1"));
+  ASSERT_OK(Put("key2", "v2"));
+  ASSERT_OK(Put("key3", "v3"));
+  dbfull()->TEST_FlushMemTable();
+
+  TablePropertiesCollection ptc;
+  reinterpret_cast<DB*>(dbfull())->GetPropertiesOfAllTables(&ptc);
+  ASSERT_EQ(1U, ptc.size());
+  ASSERT_EQ(3U, ptc.begin()->second->num_entries);
+  ASSERT_EQ("1", FilesPerLevel());
+
+  ASSERT_EQ("v1", Get("key1"));
+  ASSERT_EQ("v2", Get("key2"));
+  ASSERT_EQ("v3", Get("key3"));
+  ASSERT_EQ("NOT_FOUND", Get("key4"));
+
+  // Now add more keys and flush.
+  ASSERT_OK(Put("key4", "v4"));
+  ASSERT_OK(Put("key5", "v5"));
+  ASSERT_OK(Put("key6", "v6"));
+  dbfull()->TEST_FlushMemTable();
+
+  reinterpret_cast<DB*>(dbfull())->GetPropertiesOfAllTables(&ptc);
+  ASSERT_EQ(2U, ptc.size());
+  auto row = ptc.begin();
+  ASSERT_EQ(3U, row->second->num_entries);
+  ASSERT_EQ(3U, (++row)->second->num_entries);
+  ASSERT_EQ("2", FilesPerLevel());
+  ASSERT_EQ("v1", Get("key1"));
+  ASSERT_EQ("v2", Get("key2"));
+  ASSERT_EQ("v3", Get("key3"));
+  ASSERT_EQ("v4", Get("key4"));
+  ASSERT_EQ("v5", Get("key5"));
+  ASSERT_EQ("v6", Get("key6"));
+
+  ASSERT_OK(Delete("key6"));
+  ASSERT_OK(Delete("key5"));
+  ASSERT_OK(Delete("key4"));
+  dbfull()->TEST_FlushMemTable();
+  reinterpret_cast<DB*>(dbfull())->GetPropertiesOfAllTables(&ptc);
+  ASSERT_EQ(3U, ptc.size());
+  row = ptc.begin();
+  ASSERT_EQ(3U, row->second->num_entries);
+  ASSERT_EQ(3U, (++row)->second->num_entries);
+  ASSERT_EQ(3U, (++row)->second->num_entries);
+  ASSERT_EQ("3", FilesPerLevel());
+  ASSERT_EQ("v1", Get("key1"));
+  ASSERT_EQ("v2", Get("key2"));
+  ASSERT_EQ("v3", Get("key3"));
+  ASSERT_EQ("NOT_FOUND", Get("key4"));
+  ASSERT_EQ("NOT_FOUND", Get("key5"));
+  ASSERT_EQ("NOT_FOUND", Get("key6"));
+}
+
+TEST(CuckooTableDBTest, FlushWithDuplicateKeys) {
+  Options options = CurrentOptions();
+  Reopen(&options);
+  ASSERT_OK(Put("key1", "v1"));
+  ASSERT_OK(Put("key2", "v2"));
+  ASSERT_OK(Put("key1", "v3"));  // Duplicate
+  dbfull()->TEST_FlushMemTable();
+
+  TablePropertiesCollection ptc;
+  reinterpret_cast<DB*>(dbfull())->GetPropertiesOfAllTables(&ptc);
+  ASSERT_EQ(1U, ptc.size());
+  ASSERT_EQ(2U, ptc.begin()->second->num_entries);
+  ASSERT_EQ("1", FilesPerLevel());
+  ASSERT_EQ("v3", Get("key1"));
+  ASSERT_EQ("v2", Get("key2"));
+}
+
+namespace {
+static std::string Key(int i) {
+  char buf[100];
+  snprintf(buf, sizeof(buf), "key_______%06d", i);
+  return std::string(buf);
+}
+static std::string Uint64Key(uint64_t i) {
+  std::string str;
+  str.resize(8);
+  memcpy(&str[0], static_cast<void*>(&i), 8);
+  return str;
+}
+}  // namespace.
+
+TEST(CuckooTableDBTest, Uint64Comparator) {
+  Options options = CurrentOptions();
+  options.comparator = test::Uint64Comparator();
+  Reopen(&options);
+
+  ASSERT_OK(Put(Uint64Key(1), "v1"));
+  ASSERT_OK(Put(Uint64Key(2), "v2"));
+  ASSERT_OK(Put(Uint64Key(3), "v3"));
+  dbfull()->TEST_FlushMemTable();
+
+  ASSERT_EQ("v1", Get(Uint64Key(1)));
+  ASSERT_EQ("v2", Get(Uint64Key(2)));
+  ASSERT_EQ("v3", Get(Uint64Key(3)));
+  ASSERT_EQ("NOT_FOUND", Get(Uint64Key(4)));
+
+  // Add more keys.
+  ASSERT_OK(Delete(Uint64Key(2)));  // Delete.
+  dbfull()->TEST_FlushMemTable();
+  ASSERT_OK(Put(Uint64Key(3), "v0"));  // Update.
+  ASSERT_OK(Put(Uint64Key(4), "v4"));
+  dbfull()->TEST_FlushMemTable();
+  ASSERT_EQ("v1", Get(Uint64Key(1)));
+  ASSERT_EQ("NOT_FOUND", Get(Uint64Key(2)));
+  ASSERT_EQ("v0", Get(Uint64Key(3)));
+  ASSERT_EQ("v4", Get(Uint64Key(4)));
+}
+
+TEST(CuckooTableDBTest, CompactionTrigger) {
+  Options options = CurrentOptions();
+  options.write_buffer_size = 100 << 10;  // 100KB
+  options.level0_file_num_compaction_trigger = 2;
+  Reopen(&options);
+
+  // Write 11 values, each 10016 B
+  for (int idx = 0; idx < 11; ++idx) {
+    ASSERT_OK(Put(Key(idx), std::string(10000, 'a' + idx)));
+  }
+  dbfull()->TEST_WaitForFlushMemTable();
+  ASSERT_EQ("1", FilesPerLevel());
+
+  // Generate one more file in level-0, and should trigger level-0 compaction
+  for (int idx = 11; idx < 22; ++idx) {
+    ASSERT_OK(Put(Key(idx), std::string(10000, 'a' + idx)));
+  }
+  dbfull()->TEST_WaitForFlushMemTable();
+  ASSERT_EQ("2", FilesPerLevel());
+
+  dbfull()->TEST_CompactRange(0, nullptr, nullptr);
+  ASSERT_EQ("0,2", FilesPerLevel());
+  for (int idx = 0; idx < 22; ++idx) {
+    ASSERT_EQ(std::string(10000, 'a' + idx), Get(Key(idx)));
+  }
+}
+
+TEST(CuckooTableDBTest, CompactionIntoMultipleFiles) {
+  // Create a big L0 file and check it compacts into multiple files in L1.
+  Options options = CurrentOptions();
+  options.write_buffer_size = 270 << 10;
+  // Two SST files should be created, each containing 14 keys.
+  // Number of buckets will be 16. Total size ~156 KB.
+  options.target_file_size_base = 160 << 10;
+  Reopen(&options);
+
+  // Write 28 values, each 10016 B ~ 10KB
+  for (int idx = 0; idx < 28; ++idx) {
+    ASSERT_OK(Put(Key(idx), std::string(10000, 'a' + idx)));
+  }
+  dbfull()->TEST_WaitForFlushMemTable();
+  ASSERT_EQ("1", FilesPerLevel());
+
+  dbfull()->TEST_CompactRange(0, nullptr, nullptr);
+  ASSERT_EQ("0,2", FilesPerLevel());
+  for (int idx = 0; idx < 28; ++idx) {
+    ASSERT_EQ(std::string(10000, 'a' + idx), Get(Key(idx)));
+  }
+}
+
+TEST(CuckooTableDBTest, SameKeyInsertedInTwoDifferentFilesAndCompacted) {
+  // Insert same key twice so that they go to different SST files. Then wait for
+  // compaction and check if the latest value is stored and old value removed.
+  Options options = CurrentOptions();
+  options.write_buffer_size = 100 << 10;  // 100KB
+  options.level0_file_num_compaction_trigger = 2;
+  Reopen(&options);
+
+  // Write 11 values, each 10016 B
+  for (int idx = 0; idx < 11; ++idx) {
+    ASSERT_OK(Put(Key(idx), std::string(10000, 'a')));
+  }
+  dbfull()->TEST_WaitForFlushMemTable();
+  ASSERT_EQ("1", FilesPerLevel());
+
+  // Generate one more file in level-0, and should trigger level-0 compaction
+  for (int idx = 0; idx < 11; ++idx) {
+    ASSERT_OK(Put(Key(idx), std::string(10000, 'a' + idx)));
+  }
+  dbfull()->TEST_WaitForFlushMemTable();
+  dbfull()->TEST_CompactRange(0, nullptr, nullptr);
+
+  ASSERT_EQ("0,1", FilesPerLevel());
+  for (int idx = 0; idx < 11; ++idx) {
+    ASSERT_EQ(std::string(10000, 'a' + idx), Get(Key(idx)));
+  }
+}
+
+TEST(CuckooTableDBTest, AdaptiveTable) {
+  Options options = CurrentOptions();
+
+  // Write some keys using cuckoo table.
+  options.table_factory.reset(NewCuckooTableFactory());
+  Reopen(&options);
+
+  ASSERT_OK(Put("key1", "v1"));
+  ASSERT_OK(Put("key2", "v2"));
+  ASSERT_OK(Put("key3", "v3"));
+  dbfull()->TEST_FlushMemTable();
+
+  // Write some keys using plain table.
+  options.create_if_missing = false;
+  options.table_factory.reset(NewPlainTableFactory());
+  Reopen(&options);
+  ASSERT_OK(Put("key4", "v4"));
+  ASSERT_OK(Put("key1", "v5"));
+  dbfull()->TEST_FlushMemTable();
+
+  // Write some keys using block based table.
+  std::shared_ptr<TableFactory> block_based_factory(
+      NewBlockBasedTableFactory());
+  options.table_factory.reset(NewAdaptiveTableFactory(block_based_factory));
+  Reopen(&options);
+  ASSERT_OK(Put("key5", "v6"));
+  ASSERT_OK(Put("key2", "v7"));
+  dbfull()->TEST_FlushMemTable();
+
+  ASSERT_EQ("v5", Get("key1"));
+  ASSERT_EQ("v7", Get("key2"));
+  ASSERT_EQ("v3", Get("key3"));
+  ASSERT_EQ("v4", Get("key4"));
+  ASSERT_EQ("v6", Get("key5"));
+}
+}  // namespace rocksdb
+
+int main(int argc, char** argv) { return rocksdb::test::RunAllTests(); }
@@ -7,7 +7,9 @@
 // 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 __STDC_FORMAT_MACROS
 #define __STDC_FORMAT_MACROS
+#endif

 #ifndef GFLAGS
 #include <cstdio>
@@ -17,6 +19,11 @@ int main() {
 }
 #else

+#ifdef NUMA
+#include <numa.h>
+#include <numaif.h>
+#endif
+
 #include <inttypes.h>
 #include <cstddef>
 #include <sys/types.h>
@@ -25,7 +32,6 @@ int main() {
 #include <gflags/gflags.h>
 #include "db/db_impl.h"
 #include "db/version_set.h"
-#include "rocksdb/statistics.h"
 #include "rocksdb/options.h"
 #include "rocksdb/cache.h"
 #include "rocksdb/db.h"
@@ -33,8 +39,8 @@ int main() {
 #include "rocksdb/memtablerep.h"
 #include "rocksdb/write_batch.h"
 #include "rocksdb/slice.h"
+#include "rocksdb/filter_policy.h"
 #include "rocksdb/slice_transform.h"
-#include "rocksdb/statistics.h"
 #include "rocksdb/perf_context.h"
 #include "port/port.h"
 #include "port/stack_trace.h"
@@ -80,7 +86,8 @@ DEFINE_string(benchmarks,
              "xxhash,"
              "compress,"
              "uncompress,"
-              "acquireload,",
+              "acquireload,"
+              "fillseekseq,",

              "Comma-separated list of operations to run in the specified order"
              "Actual benchmarks:\n"
@@ -123,6 +130,8 @@ DEFINE_string(benchmarks,
              "\tcrc32c        -- repeated crc32c of 4K of data\n"
              "\txxhash        -- repeated xxHash of 4K of data\n"
              "\tacquireload   -- load N*1000 times\n"
+              "\tfillseekseq   -- write N values in sequential key, then read "
+              "them by seeking to each key\n"
              "Meta operations:\n"
              "\tcompact     -- Compact the entire DB\n"
              "\tstats       -- Print DB stats\n"
@@ -142,6 +151,7 @@ DEFINE_int64(merge_keys, -1,
             "Number of distinct keys to use for MergeRandom and "
             "ReadRandomMergeRandom. "
             "If negative, there will be FLAGS_num keys.");
+DEFINE_int32(num_column_families, 1, "Number of Column Families to use.");

 DEFINE_int64(reads, -1, "Number of read operations to do.  "
             "If negative, do FLAGS_num reads.");
@@ -158,6 +168,12 @@ DEFINE_int32(duration, 0, "Time in seconds for the random-ops tests to run."

 DEFINE_int32(value_size, 100, "Size of each value");

+DEFINE_int32(seekseq_next, 0, "How many times to call Next() after Seek() in "
+             "fillseekseq");
+
+DEFINE_bool(use_uint64_comparator, false, "use Uint64 user comparator");
+
+DEFINE_int64(batch_size, 1, "Batch size");

 static bool ValidateKeySize(const char* flagname, int32_t value) {
  return true;
@@ -173,6 +189,14 @@ DEFINE_double(compression_ratio, 0.5, "Arrange to generate values that shrink"

 DEFINE_bool(histogram, false, "Print histogram of operation timings");

+DEFINE_bool(enable_numa, false,
+            "Make operations aware of NUMA architecture and bind memory "
+            "and cpus corresponding to nodes together. In NUMA, memory "
+            "in same node as CPUs are closer when compared to memory in "
+            "other nodes. Reads can be faster when the process is bound to "
+            "CPU and memory of same node. Use \"$numactl --hardware\" command "
+            "to see NUMA memory architecture.");
+
 DEFINE_int64(write_buffer_size, rocksdb::Options().write_buffer_size,
             "Number of bytes to buffer in memtable before compacting");

@@ -226,9 +250,14 @@ DEFINE_int32(universal_compression_size_percent, -1,
 DEFINE_int64(cache_size, -1, "Number of bytes to use as a cache of uncompressed"
             "data. Negative means use default settings.");

-DEFINE_int32(block_size, rocksdb::Options().block_size,
+DEFINE_int32(block_size, rocksdb::BlockBasedTableOptions().block_size,
             "Number of bytes in a block.");

+DEFINE_int32(block_restart_interval,
+             rocksdb::BlockBasedTableOptions().block_restart_interval,
+             "Number of keys between restart points "
+             "for delta encoding of keys.");
+
 DEFINE_int64(compressed_cache_size, -1,
             "Number of bytes to use as a cache of compressed data.");

@@ -286,7 +315,7 @@ DEFINE_string(wal_dir, "", "If not empty, use the given dir for WAL");

 DEFINE_int32(num_levels, 7, "The total number of levels");

-DEFINE_int32(target_file_size_base, 2 * 1048576, "Target file size at level-1");
+DEFINE_int64(target_file_size_base, 2 * 1048576, "Target file size at level-1");

 DEFINE_int32(target_file_size_multiplier, 1,
             "A multiplier to compute target level-N file size (N >= 2)");
@@ -333,9 +362,6 @@ DEFINE_int32(deletepercent, 2, "Percentage of deletes out of reads/writes/"
             "deletepercent), so deletepercent must be smaller than (100 - "
             "FLAGS_readwritepercent)");

-DEFINE_int32(disable_seek_compaction, false, "Option to disable compaction"
-             " triggered by read.");
-
 DEFINE_uint64(delete_obsolete_files_period_micros, 0, "Option to delete "
              "obsolete files periodically. 0 means that obsolete files are"
              " deleted after every compaction run.");
@@ -465,6 +491,8 @@ static auto FLAGS_compaction_fadvice_e =

 DEFINE_bool(use_tailing_iterator, false,
            "Use tailing iterator to access a series of keys instead of get");
+DEFINE_int64(iter_refresh_interval_us, -1,
+             "How often to refresh iterators. Disable refresh when -1");

 DEFINE_bool(use_adaptive_mutex, rocksdb::Options().use_adaptive_mutex,
            "Use adaptive mutex");
@@ -492,6 +520,11 @@ DEFINE_int32(prefix_size, 0, "control the prefix size for HashSkipList and "
 DEFINE_int64(keys_per_prefix, 0, "control average number of keys generated "
             "per prefix, 0 means no special handling of the prefix, "
             "i.e. use the prefix comes with the generated random number.");
+DEFINE_bool(enable_io_prio, false, "Lower the background flush/compaction "
+            "threads' IO priority");
+DEFINE_bool(identity_as_first_hash, false, "the first hash function of cuckoo "
+            "table becomes an identity function. This is only valid when key "
+            "is 8 bytes");

 enum RepFactory {
  kSkipList,
@@ -526,11 +559,21 @@ DEFINE_string(memtablerep, "skip_list", "");
 DEFINE_int64(hash_bucket_count, 1024 * 1024, "hash bucket count");
 DEFINE_bool(use_plain_table, false, "if use plain table "
            "instead of block-based table format");
-
+DEFINE_bool(use_cuckoo_table, false, "if use cuckoo table format");
+DEFINE_double(cuckoo_hash_ratio, 0.9, "Hash ratio for Cuckoo SST table.");
+DEFINE_bool(use_hash_search, false, "if use kHashSearch "
+            "instead of kBinarySearch. "
+            "This is valid if only we use BlockTable");
+DEFINE_bool(use_block_based_filter, false, "if use kBlockBasedFilter "
+            "instead of kFullFilter for filter block. "
+            "This is valid if only we use BlockTable");
 DEFINE_string(merge_operator, "", "The merge operator to use with the database."
              "If a new merge operator is specified, be sure to use fresh"
              " database The possible merge operators are defined in"
              " utilities/merge_operators.h");
+DEFINE_int32(skip_list_lookahead, 0, "Used with skip_list memtablerep; try "
+             "linear search first for this many steps from the previous "
+             "position");

 static const bool FLAGS_soft_rate_limit_dummy __attribute__((unused)) =
    RegisterFlagValidator(&FLAGS_soft_rate_limit, &ValidateRateLimit);
@@ -551,6 +594,9 @@ static const bool FLAGS_cache_numshardbits_dummy __attribute__((unused)) =
 static const bool FLAGS_readwritepercent_dummy __attribute__((unused)) =
    RegisterFlagValidator(&FLAGS_readwritepercent, &ValidateInt32Percent);

+DEFINE_int32(disable_seek_compaction, false,
+             "Not used, left here for backwards compatibility");
+
 static const bool FLAGS_deletepercent_dummy __attribute__((unused)) =
    RegisterFlagValidator(&FLAGS_deletepercent, &ValidateInt32Percent);
 static const bool FLAGS_table_cache_numshardbits_dummy __attribute__((unused)) =
@@ -582,9 +628,9 @@ class RandomGenerator {
  }

  Slice Generate(unsigned int len) {
+    assert(len <= data_.size());
    if (pos_ + len > data_.size()) {
      pos_ = 0;
-      assert(len < data_.size());
    }
    pos_ += len;
    return Slice(data_.data() + pos_ - len, len);
@@ -599,6 +645,14 @@ static void AppendWithSpace(std::string* str, Slice msg) {
  str->append(msg.data(), msg.size());
 }

+struct DBWithColumnFamilies {
+  std::vector<ColumnFamilyHandle*> cfh;
+  DB* db;
+  DBWithColumnFamilies() : db(nullptr) {
+    cfh.clear();
+  }
+};
+
 class Stats {
 private:
  int id_;
@@ -662,7 +716,7 @@ class Stats {
  void SetId(int id) { id_ = id; }
  void SetExcludeFromMerge() { exclude_from_merge_ = true; }

-  void FinishedSingleOp(DB* db) {
+  void FinishedOps(DBWithColumnFamilies* db_with_cfh, DB* db, int64_t num_ops) {
    if (FLAGS_histogram) {
      double now = FLAGS_env->NowMicros();
      double micros = now - last_op_finish_;
@@ -674,7 +728,7 @@ class Stats {
      last_op_finish_ = now;
    }

-    done_++;
+    done_ += num_ops;
    if (done_ >= next_report_) {
      if (!FLAGS_stats_interval) {
        if      (next_report_ < 1000)   next_report_ += 100;
@@ -702,8 +756,17 @@ class Stats {

        if (FLAGS_stats_per_interval) {
          std::string stats;
-          if (db && db->GetProperty("rocksdb.stats", &stats))
+
+          if (db_with_cfh && db_with_cfh->cfh.size()) {
+            for (size_t i = 0; i < db_with_cfh->cfh.size(); ++i) {
+              if (db->GetProperty(db_with_cfh->cfh[i], "rocksdb.cfstats",
+                                  &stats))
+                fprintf(stderr, "%s\n", stats.c_str());
+            }
+
+          } else if (db && db->GetProperty("rocksdb.stats", &stats)) {
            fprintf(stderr, "%s\n", stats.c_str());
+          }
        }

        fflush(stderr);
@@ -720,7 +783,7 @@ class Stats {

  void Report(const Slice& name) {
    // Pretend at least one op was done in case we are running a benchmark
-    // that does not call FinishedSingleOp().
+    // that does not call FinishedOps().
    if (done_ < 1) done_ = 1;

    std::string extra;
@@ -818,12 +881,12 @@ class Duration {

 class Benchmark {
 private:
-  shared_ptr<Cache> cache_;
-  shared_ptr<Cache> compressed_cache_;
-  const FilterPolicy* filter_policy_;
+  std::shared_ptr<Cache> cache_;
+  std::shared_ptr<Cache> compressed_cache_;
+  std::shared_ptr<const FilterPolicy> filter_policy_;
  const SliceTransform* prefix_extractor_;
-  DB* db_;
-  std::vector<DB*> multi_dbs_;
+  DBWithColumnFamilies db_;
+  std::vector<DBWithColumnFamilies> multi_dbs_;
  int64_t num_;
  int value_size_;
  int key_size_;
@@ -835,6 +898,15 @@ class Benchmark {
  int64_t writes_;
  int64_t readwrites_;
  int64_t merge_keys_;
+
+  bool SanityCheck() {
+    if (FLAGS_compression_ratio > 1) {
+      fprintf(stderr, "compression_ratio should be between 0 and 1\n");
+      return false;
+    }
+    return true;
+  }
+
  void PrintHeader() {
    PrintEnvironment();
    fprintf(stdout, "Keys:       %d bytes each\n", FLAGS_key_size);
@@ -852,6 +924,18 @@ class Benchmark {
              * num_)
             / 1048576.0));
    fprintf(stdout, "Write rate limit: %d\n", FLAGS_writes_per_second);
+    if (FLAGS_enable_numa) {
+      fprintf(stderr, "Running in NUMA enabled mode.\n");
+#ifndef NUMA
+      fprintf(stderr, "NUMA is not defined in the system.\n");
+      exit(1);
+#else
+      if (numa_available() == -1) {
+        fprintf(stderr, "NUMA is not supported by the system.\n");
+        exit(1);
+      }
+#endif
+    }
    switch (FLAGS_compression_type_e) {
      case rocksdb::kNoCompression:
        fprintf(stdout, "Compression: none\n");
@@ -1018,11 +1102,10 @@ class Benchmark {
           (FLAGS_cache_numshardbits >= 1 ?
            NewLRUCache(FLAGS_compressed_cache_size, FLAGS_cache_numshardbits) :
            NewLRUCache(FLAGS_compressed_cache_size)) : nullptr),
-    filter_policy_(FLAGS_bloom_bits >= 0
-                   ? NewBloomFilterPolicy(FLAGS_bloom_bits)
-                   : nullptr),
+    filter_policy_(FLAGS_bloom_bits >= 0 ?
+        NewBloomFilterPolicy(FLAGS_bloom_bits, FLAGS_use_block_based_filter)
+        : nullptr),
    prefix_extractor_(NewFixedPrefixTransform(FLAGS_prefix_size)),
-    db_(nullptr),
    num_(FLAGS_num),
    value_size_(FLAGS_value_size),
    key_size_(FLAGS_key_size),
@@ -1053,8 +1136,9 @@ class Benchmark {
  }

  ~Benchmark() {
-    delete db_;
-    delete filter_policy_;
+    std::for_each(db_.cfh.begin(), db_.cfh.end(),
+                  [](ColumnFamilyHandle* cfh) { delete cfh; });
+    delete db_.db;
    delete prefix_extractor_;
  }

@@ -1113,7 +1197,20 @@ class Benchmark {
    return base_name + std::to_string(id);
  }

+  std::string ColumnFamilyName(int i) {
+    if (i == 0) {
+      return kDefaultColumnFamilyName;
+    } else {
+      char name[100];
+      snprintf(name, sizeof(name), "column_family_name_%06d", i);
+      return std::string(name);
+    }
+  }
+
  void Run() {
+    if (!SanityCheck()) {
+      exit(1);
+    }
    PrintHeader();
    Open();
    const char* benchmarks = FLAGS_benchmarks.c_str();
@@ -1186,7 +1283,12 @@ class Benchmark {
        method = &Benchmark::ReadReverse;
      } else if (name == Slice("readrandom")) {
        method = &Benchmark::ReadRandom;
+      } else if (name == Slice("readrandomfast")) {
+        method = &Benchmark::ReadRandomFast;
      } else if (name == Slice("multireadrandom")) {
+        entries_per_batch_ = FLAGS_batch_size;
+        fprintf(stderr, "entries_per_batch = %" PRIi64 "\n",
+                entries_per_batch_);
        method = &Benchmark::MultiReadRandom;
      } else if (name == Slice("readmissing")) {
        ++key_size_;
@@ -1233,6 +1335,8 @@ class Benchmark {
        method = &Benchmark::MergeRandom;
      } else if (name == Slice("randomwithverify")) {
        method = &Benchmark::RandomWithVerify;
+      } else if (name == Slice("fillseekseq")) {
+        method = &Benchmark::WriteSeqSeekSeq;
      } else if (name == Slice("compact")) {
        method = &Benchmark::Compact;
      } else if (name == Slice("crc32c")) {
@@ -1264,13 +1368,16 @@ class Benchmark {
                  name.ToString().c_str());
          method = nullptr;
        } else {
-          if (db_ != nullptr) {
-            delete db_;
-            db_ = nullptr;
+          if (db_.db != nullptr) {
+            std::for_each(db_.cfh.begin(), db_.cfh.end(),
+                          [](ColumnFamilyHandle* cfh) { delete cfh; });
+            delete db_.db;
+            db_.db = nullptr;
+            db_.cfh.clear();
            DestroyDB(FLAGS_db, Options());
          }
          for (size_t i = 0; i < multi_dbs_.size(); i++) {
-            delete multi_dbs_[i];
+            delete multi_dbs_[i].db;
            DestroyDB(GetDbNameForMultiple(FLAGS_db, i), Options());
          }
          multi_dbs_.clear();
@@ -1334,7 +1441,25 @@ class Benchmark {
    shared.start = false;

    ThreadArg* arg = new ThreadArg[n];
+
    for (int i = 0; i < n; i++) {
+#ifdef NUMA
+      if (FLAGS_enable_numa) {
+        // Performs a local allocation of memory to threads in numa node.
+        int n_nodes = numa_num_task_nodes();  // Number of nodes in NUMA.
+        numa_exit_on_error = 1;
+        int numa_node = i % n_nodes;
+        bitmask* nodes = numa_allocate_nodemask();
+        numa_bitmask_clearall(nodes);
+        numa_bitmask_setbit(nodes, numa_node);
+        // numa_bind() call binds the process to the node and these
+        // properties are passed on to the thread that is created in
+        // StartThread method called later in the loop.
+        numa_bind(nodes);
+        numa_set_strict(1);
+        numa_free_nodemask(nodes);
+      }
+#endif
      arg[i].bm = this;
      arg[i].method = method;
      arg[i].shared = &shared;
@@ -1377,7 +1502,7 @@ class Benchmark {
    uint32_t crc = 0;
    while (bytes < 500 * 1048576) {
      crc = crc32c::Value(data.data(), size);
-      thread->stats.FinishedSingleOp(nullptr);
+      thread->stats.FinishedOps(nullptr, nullptr, 1);
      bytes += size;
    }
    // Print so result is not dead
@@ -1396,7 +1521,7 @@ class Benchmark {
    unsigned int xxh32 = 0;
    while (bytes < 500 * 1048576) {
      xxh32 = XXH32(data.data(), size, 0);
-      thread->stats.FinishedSingleOp(nullptr);
+      thread->stats.FinishedOps(nullptr, nullptr, 1);
      bytes += size;
    }
    // Print so result is not dead
@@ -1417,14 +1542,14 @@ class Benchmark {
        ptr = ap.Acquire_Load();
      }
      count++;
-      thread->stats.FinishedSingleOp(nullptr);
+      thread->stats.FinishedOps(nullptr, nullptr, 1);
    }
    if (ptr == nullptr) exit(1); // Disable unused variable warning.
  }

  void Compress(ThreadState *thread) {
    RandomGenerator gen;
-    Slice input = gen.Generate(Options().block_size);
+    Slice input = gen.Generate(FLAGS_block_size);
    int64_t bytes = 0;
    int64_t produced = 0;
    bool ok = true;
@@ -1458,7 +1583,7 @@ class Benchmark {
      }
      produced += compressed.size();
      bytes += input.size();
-      thread->stats.FinishedSingleOp(nullptr);
+      thread->stats.FinishedOps(nullptr, nullptr, 1);
    }

    if (!ok) {
@@ -1474,7 +1599,7 @@ class Benchmark {

  void Uncompress(ThreadState *thread) {
    RandomGenerator gen;
-    Slice input = gen.Generate(Options().block_size);
+    Slice input = gen.Generate(FLAGS_block_size);
    std::string compressed;

    bool ok;
@@ -1539,7 +1664,7 @@ class Benchmark {
      }
      delete[] uncompressed;
      bytes += input.size();
-      thread->stats.FinishedSingleOp(nullptr);
+      thread->stats.FinishedOps(nullptr, nullptr, 1);
    }

    if (!ok) {
@@ -1550,14 +1675,10 @@ class Benchmark {
  }

  void Open() {
-    assert(db_ == nullptr);
+    assert(db_.db == nullptr);
    Options options;
    options.create_if_missing = !FLAGS_use_existing_db;
-    options.block_cache = cache_;
-    options.block_cache_compressed = compressed_cache_;
-    if (cache_ == nullptr) {
-      options.no_block_cache = true;
-    }
+    options.create_missing_column_families = FLAGS_num_column_families > 1;
    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 =
@@ -1565,16 +1686,25 @@ class Benchmark {
    options.max_background_compactions = FLAGS_max_background_compactions;
    options.max_background_flushes = FLAGS_max_background_flushes;
    options.compaction_style = FLAGS_compaction_style_e;
-    options.block_size = FLAGS_block_size;
-    options.filter_policy = filter_policy_;
-    if (FLAGS_use_plain_table) {
+    if (FLAGS_prefix_size != 0) {
      options.prefix_extractor.reset(
          NewFixedPrefixTransform(FLAGS_prefix_size));
    }
+    if (FLAGS_use_uint64_comparator) {
+      options.comparator = test::Uint64Comparator();
+      if (FLAGS_key_size != 8) {
+        fprintf(stderr, "Using Uint64 comparator but key size is not 8.\n");
+        exit(1);
+      }
+    }
    options.memtable_prefix_bloom_bits = FLAGS_memtable_bloom_bits;
    options.bloom_locality = FLAGS_bloom_locality;
    options.max_open_files = FLAGS_open_files;
    options.statistics = dbstats;
+    if (FLAGS_enable_io_prio) {
+      FLAGS_env->LowerThreadPoolIOPriority(Env::LOW);
+      FLAGS_env->LowerThreadPoolIOPriority(Env::HIGH);
+    }
    options.env = FLAGS_env;
    options.disableDataSync = FLAGS_disable_data_sync;
    options.use_fsync = FLAGS_use_fsync;
@@ -1598,7 +1728,8 @@ class Benchmark {
            FLAGS_hash_bucket_count));
        break;
      case kSkipList:
-        // no need to do anything
+        options.memtable_factory.reset(new SkipListFactory(
+            FLAGS_skip_list_lookahead));
        break;
      case kHashLinkedList:
        options.memtable_factory.reset(NewHashLinkListRepFactory(
@@ -1628,8 +1759,45 @@ class Benchmark {
      if (bloom_bits_per_key < 0) {
        bloom_bits_per_key = 0;
      }
+
+      PlainTableOptions plain_table_options;
+      plain_table_options.user_key_len = FLAGS_key_size;
+      plain_table_options.bloom_bits_per_key = bloom_bits_per_key;
+      plain_table_options.hash_table_ratio = 0.75;
      options.table_factory = std::shared_ptr<TableFactory>(
-          NewPlainTableFactory(FLAGS_key_size, bloom_bits_per_key, 0.75));
+          NewPlainTableFactory(plain_table_options));
+    } else if (FLAGS_use_cuckoo_table) {
+      if (FLAGS_cuckoo_hash_ratio > 1 || FLAGS_cuckoo_hash_ratio < 0) {
+        fprintf(stderr, "Invalid cuckoo_hash_ratio\n");
+        exit(1);
+      }
+      rocksdb::CuckooTableOptions table_options;
+      table_options.hash_table_ratio = FLAGS_cuckoo_hash_ratio;
+      table_options.identity_as_first_hash = FLAGS_identity_as_first_hash;
+      options.table_factory = std::shared_ptr<TableFactory>(
+          NewCuckooTableFactory(table_options));
+    } else {
+      BlockBasedTableOptions block_based_options;
+      if (FLAGS_use_hash_search) {
+        if (FLAGS_prefix_size == 0) {
+          fprintf(stderr,
+              "prefix_size not assigned when enable use_hash_search \n");
+          exit(1);
+        }
+        block_based_options.index_type = BlockBasedTableOptions::kHashSearch;
+      } else {
+        block_based_options.index_type = BlockBasedTableOptions::kBinarySearch;
+      }
+      if (cache_ == nullptr) {
+        block_based_options.no_block_cache = true;
+      }
+      block_based_options.block_cache = cache_;
+      block_based_options.block_cache_compressed = compressed_cache_;
+      block_based_options.block_size = FLAGS_block_size;
+      block_based_options.block_restart_interval = FLAGS_block_restart_interval;
+      block_based_options.filter_policy = filter_policy_;
+      options.table_factory.reset(
+          NewBlockBasedTableFactory(block_based_options));
    }
    if (FLAGS_max_bytes_for_level_multiplier_additional_v.size() > 0) {
      if (FLAGS_max_bytes_for_level_multiplier_additional_v.size() !=
@@ -1661,7 +1829,6 @@ class Benchmark {
        options.compression_per_level[i] = FLAGS_compression_type_e;
      }
    }
-    options.disable_seek_compaction = FLAGS_disable_seek_compaction;
    options.delete_obsolete_files_period_micros =
      FLAGS_delete_obsolete_files_period_micros;
    options.soft_rate_limit = FLAGS_soft_rate_limit;
@@ -1719,10 +1886,9 @@ class Benchmark {
      OpenDb(options, FLAGS_db, &db_);
    } else {
      multi_dbs_.clear();
+      multi_dbs_.resize(FLAGS_num_multi_db);
      for (int i = 0; i < FLAGS_num_multi_db; i++) {
-        DB* db;
-        OpenDb(options, GetDbNameForMultiple(FLAGS_db, i), &db);
-        multi_dbs_.push_back(db);
+        OpenDb(options, GetDbNameForMultiple(FLAGS_db, i), &multi_dbs_[i]);
      }
    }
    if (FLAGS_min_level_to_compress >= 0) {
@@ -1730,12 +1896,27 @@ class Benchmark {
    }
  }

-  void OpenDb(Options options, std::string db_name, DB** db) {
+  void OpenDb(const Options& options, const std::string& db_name,
+      DBWithColumnFamilies* db) {
    Status s;
-    if(FLAGS_readonly) {
-      s = DB::OpenForReadOnly(options, db_name, db);
+    // Open with column families if necessary.
+    if (FLAGS_num_column_families > 1) {
+      db->cfh.resize(FLAGS_num_column_families);
+      std::vector<ColumnFamilyDescriptor> column_families;
+      for (int i = 0; i < FLAGS_num_column_families; i++) {
+        column_families.push_back(ColumnFamilyDescriptor(
+              ColumnFamilyName(i), ColumnFamilyOptions(options)));
+      }
+      if (FLAGS_readonly) {
+        s = DB::OpenForReadOnly(options, db_name, column_families,
+            &db->cfh, &db->db);
+      } else {
+        s = DB::Open(options, db_name, column_families, &db->cfh, &db->db);
+      }
+    } else if (FLAGS_readonly) {
+      s = DB::OpenForReadOnly(options, db_name, &db->db);
    } else {
-      s = DB::Open(options, db_name, db);
+      s = DB::Open(options, db_name, &db->db);
    }
    if (!s.ok()) {
      fprintf(stderr, "open error: %s\n", s.ToString().c_str());
@@ -1803,10 +1984,18 @@ class Benchmark {
  };

  DB* SelectDB(ThreadState* thread) {
-    if (db_ != nullptr) {
-      return db_;
-    } else {
-      return multi_dbs_[thread->rand.Next() % multi_dbs_.size()];
+    return SelectDBWithCfh(thread)->db;
+  }
+
+  DBWithColumnFamilies* SelectDBWithCfh(ThreadState* thread) {
+    return SelectDBWithCfh(thread->rand.Next());
+  }
+
+  DBWithColumnFamilies* SelectDBWithCfh(uint64_t rand_int) {
+    if (db_.db != nullptr) {
+      return &db_;
+    } else  {
+      return &multi_dbs_[rand_int % multi_dbs_.size()];
    }
  }

@@ -1815,7 +2004,7 @@ class Benchmark {
    const int64_t num_ops = writes_ == 0 ? num_ : writes_;

    size_t num_key_gens = 1;
-    if (db_ == nullptr) {
+    if (db_.db == nullptr) {
      num_key_gens = multi_dbs_.size();
    }
    std::vector<std::unique_ptr<KeyGenerator>> key_gens(num_key_gens);
@@ -1838,20 +2027,26 @@ class Benchmark {
    Slice key = AllocateKey();
    std::unique_ptr<const char[]> key_guard(key.data());
    while (!duration.Done(entries_per_batch_)) {
-      size_t id = 0;
-      DB* db_to_write = db_;
-      if (db_to_write == nullptr) {
-        id = thread->rand.Next() % num_key_gens;
-        db_to_write = multi_dbs_[id];
-      }
+      size_t id = thread->rand.Next() % num_key_gens;
+      DBWithColumnFamilies* db_with_cfh = SelectDBWithCfh(id);
      batch.Clear();
      for (int64_t j = 0; j < entries_per_batch_; j++) {
-        GenerateKeyFromInt(key_gens[id]->Next(), FLAGS_num, &key);
-        batch.Put(key, gen.Generate(value_size_));
+        int64_t rand_num = key_gens[id]->Next();
+        GenerateKeyFromInt(rand_num, FLAGS_num, &key);
+        if (FLAGS_num_column_families <= 1) {
+          batch.Put(key, gen.Generate(value_size_));
+        } else {
+          // We use same rand_num as seed for key and column family so that we
+          // can deterministically find the cfh corresponding to a particular
+          // key while reading the key.
+          batch.Put(db_with_cfh->cfh[rand_num % db_with_cfh->cfh.size()],
+              key, gen.Generate(value_size_));
+        }
        bytes += value_size_ + key_size_;
-        thread->stats.FinishedSingleOp(db_to_write);
      }
-      s = db_to_write->Write(write_options_, &batch);
+      s = db_with_cfh->db->Write(write_options_, &batch);
+      thread->stats.FinishedOps(db_with_cfh, db_with_cfh->db,
+                                entries_per_batch_);
      if (!s.ok()) {
        fprintf(stderr, "put error: %s\n", s.ToString().c_str());
        exit(1);
@@ -1861,11 +2056,11 @@ class Benchmark {
  }

  void ReadSequential(ThreadState* thread) {
-    if (db_ != nullptr) {
-      ReadSequential(thread, db_);
+    if (db_.db != nullptr) {
+      ReadSequential(thread, db_.db);
    } else {
-      for (DB* db : multi_dbs_) {
-        ReadSequential(thread, db);
+      for (const auto& db_with_cfh : multi_dbs_) {
+        ReadSequential(thread, db_with_cfh.db);
      }
    }
  }
@@ -1876,7 +2071,7 @@ class Benchmark {
    int64_t bytes = 0;
    for (iter->SeekToFirst(); i < reads_ && iter->Valid(); iter->Next()) {
      bytes += iter->key().size() + iter->value().size();
-      thread->stats.FinishedSingleOp(db);
+      thread->stats.FinishedOps(nullptr, db, 1);
      ++i;
    }
    delete iter;
@@ -1884,11 +2079,11 @@ class Benchmark {
  }

  void ReadReverse(ThreadState* thread) {
-    if (db_ != nullptr) {
-      ReadReverse(thread, db_);
+    if (db_.db != nullptr) {
+      ReadReverse(thread, db_.db);
    } else {
-      for (DB* db : multi_dbs_) {
-        ReadReverse(thread, db);
+      for (const auto& db_with_cfh : multi_dbs_) {
+        ReadReverse(thread, db_with_cfh.db);
      }
    }
  }
@@ -1899,13 +2094,56 @@ class Benchmark {
    int64_t bytes = 0;
    for (iter->SeekToLast(); i < reads_ && iter->Valid(); iter->Prev()) {
      bytes += iter->key().size() + iter->value().size();
-      thread->stats.FinishedSingleOp(db_);
+      thread->stats.FinishedOps(nullptr, db, 1);
      ++i;
    }
    delete iter;
    thread->stats.AddBytes(bytes);
  }

+  void ReadRandomFast(ThreadState* thread) {
+    int64_t read = 0;
+    int64_t found = 0;
+    int64_t nonexist = 0;
+    ReadOptions options(FLAGS_verify_checksum, true);
+    Slice key = AllocateKey();
+    std::unique_ptr<const char[]> key_guard(key.data());
+    std::string value;
+    DB* db = SelectDBWithCfh(thread)->db;
+
+    int64_t pot = 1;
+    while (pot < FLAGS_num) {
+      pot <<= 1;
+    }
+
+    Duration duration(FLAGS_duration, reads_);
+    do {
+      for (int i = 0; i < 100; ++i) {
+        int64_t key_rand = thread->rand.Next() & (pot - 1);
+        GenerateKeyFromInt(key_rand, FLAGS_num, &key);
+        ++read;
+        if (db->Get(options, key, &value).ok()) {
+          ++found;
+        }
+        if (key_rand >= FLAGS_num) {
+          ++nonexist;
+        }
+      }
+      thread->stats.FinishedOps(nullptr, db, 100);
+    } while (!duration.Done(100));
+
+    char msg[100];
+    snprintf(msg, sizeof(msg), "(%" PRIu64 " of %" PRIu64 " found, "
+             "issued %" PRIu64 " non-exist keys)\n",
+             found, read, nonexist);
+
+    thread->stats.AddMessage(msg);
+
+    if (FLAGS_perf_level > 0) {
+      thread->stats.AddMessage(perf_context.ToString());
+    }
+  }
+
  void ReadRandom(ThreadState* thread) {
    int64_t read = 0;
    int64_t found = 0;
@@ -1916,17 +2154,28 @@ class Benchmark {

    Duration duration(FLAGS_duration, reads_);
    while (!duration.Done(1)) {
-      DB* db = SelectDB(thread);
-      GenerateKeyFromInt(thread->rand.Next() % FLAGS_num, FLAGS_num, &key);
+      DBWithColumnFamilies* db_with_cfh = SelectDBWithCfh(thread);
+      // We use same key_rand as seed for key and column family so that we can
+      // deterministically find the cfh corresponding to a particular key, as it
+      // is done in DoWrite method.
+      int64_t key_rand = thread->rand.Next() % FLAGS_num;
+      GenerateKeyFromInt(key_rand, FLAGS_num, &key);
      read++;
-      if (db->Get(options, key, &value).ok()) {
+      Status s;
+      if (FLAGS_num_column_families > 1) {
+        s = db_with_cfh->db->Get(options,
+            db_with_cfh->cfh[key_rand % db_with_cfh->cfh.size()], key, &value);
+      } else {
+        s = db_with_cfh->db->Get(options, key, &value);
+      }
+      if (s.ok()) {
        found++;
      }
-      thread->stats.FinishedSingleOp(db_);
+      thread->stats.FinishedOps(db_with_cfh, db_with_cfh->db, 1);
    }

    char msg[100];
-    snprintf(msg, sizeof(msg), "(%" PRIu64 " of %" PRIu64 " found)",
+    snprintf(msg, sizeof(msg), "(%" PRIu64 " of %" PRIu64 " found)\n",
             found, read);

    thread->stats.AddMessage(msg);
@@ -1964,6 +2213,7 @@ class Benchmark {
          ++found;
        }
      }
+      thread->stats.FinishedOps(nullptr, db, entries_per_batch_);
    }
    for (auto& k : keys) {
      delete k.data();
@@ -1982,7 +2232,7 @@ class Benchmark {
      DB* db = SelectDB(thread);
      Iterator* iter = db->NewIterator(options);
      delete iter;
-      thread->stats.FinishedSingleOp(db);
+      thread->stats.FinishedOps(nullptr, db, 1);
    }
  }

@@ -2002,19 +2252,38 @@ class Benchmark {

    Iterator* single_iter = nullptr;
    std::vector<Iterator*> multi_iters;
-    if (db_ != nullptr) {
-      single_iter = db_->NewIterator(options);
+    if (db_.db != nullptr) {
+      single_iter = db_.db->NewIterator(options);
    } else {
-      for (DB* db : multi_dbs_) {
-        multi_iters.push_back(db->NewIterator(options));
+      for (const auto& db_with_cfh : multi_dbs_) {
+        multi_iters.push_back(db_with_cfh.db->NewIterator(options));
      }
    }
+    uint64_t last_refresh = FLAGS_env->NowMicros();

    Slice key = AllocateKey();
    std::unique_ptr<const char[]> key_guard(key.data());

    Duration duration(FLAGS_duration, reads_);
    while (!duration.Done(1)) {
+      if (!FLAGS_use_tailing_iterator && FLAGS_iter_refresh_interval_us >= 0) {
+        uint64_t now = FLAGS_env->NowMicros();
+        if (now - last_refresh > (uint64_t)FLAGS_iter_refresh_interval_us) {
+          if (db_.db != nullptr) {
+            delete single_iter;
+            single_iter = db_.db->NewIterator(options);
+          } else {
+            for (auto iter : multi_iters) {
+              delete iter;
+            }
+            multi_iters.clear();
+            for (const auto& db_with_cfh : multi_dbs_) {
+              multi_iters.push_back(db_with_cfh.db->NewIterator(options));
+            }
+          }
+        }
+        last_refresh = now;
+      }
      // Pick a Iterator to use
      Iterator* iter_to_use = single_iter;
      if (single_iter == nullptr) {
@@ -2027,7 +2296,7 @@ class Benchmark {
      if (iter_to_use->Valid() && iter_to_use->key().compare(key) == 0) {
        found++;
      }
-      thread->stats.FinishedSingleOp(db_);
+      thread->stats.FinishedOps(&db_, db_.db, 1);
    }
    delete single_iter;
    for (auto iter : multi_iters) {
@@ -2035,9 +2304,12 @@ class Benchmark {
    }

    char msg[100];
-    snprintf(msg, sizeof(msg), "(%" PRIu64 " of %" PRIu64 " found)",
+    snprintf(msg, sizeof(msg), "(%" PRIu64 " of %" PRIu64 " found)\n",
             found, read);
    thread->stats.AddMessage(msg);
+    if (FLAGS_perf_level > 0) {
+      thread->stats.AddMessage(perf_context.ToString());
+    }
  }

  void SeekRandomWhileWriting(ThreadState* thread) {
@@ -2062,9 +2334,9 @@ class Benchmark {
        const int64_t k = seq ? i + j : (thread->rand.Next() % FLAGS_num);
        GenerateKeyFromInt(k, FLAGS_num, &key);
        batch.Delete(key);
-        thread->stats.FinishedSingleOp(db);
      }
      auto s = db->Write(write_options_, &batch);
+      thread->stats.FinishedOps(nullptr, db, entries_per_batch_);
      if (!s.ok()) {
        fprintf(stderr, "del error: %s\n", s.ToString().c_str());
        exit(1);
@@ -2124,7 +2396,7 @@ class Benchmark {
        fprintf(stderr, "put error: %s\n", s.ToString().c_str());
        exit(1);
      }
-      thread->stats.FinishedSingleOp(db_);
+      thread->stats.FinishedOps(&db_, db_.db, 1);

      ++num_writes;
      if (writes_per_second_by_10 && num_writes >= writes_per_second_by_10) {
@@ -2284,7 +2556,7 @@ class Benchmark {
        deletes_done++;
      }

-      thread->stats.FinishedSingleOp(db_);
+      thread->stats.FinishedOps(&db_, db_.db, 1);
    }
    char msg[100];
    snprintf(msg, sizeof(msg),
@@ -2342,7 +2614,7 @@ class Benchmark {
        put_weight--;
        writes_done++;
      }
-      thread->stats.FinishedSingleOp(db);
+      thread->stats.FinishedOps(nullptr, db, 1);
    }
    char msg[100];
    snprintf(msg, sizeof(msg), "( reads:%" PRIu64 " writes:%" PRIu64 \
@@ -2376,7 +2648,7 @@ class Benchmark {
        fprintf(stderr, "put error: %s\n", s.ToString().c_str());
        exit(1);
      }
-      thread->stats.FinishedSingleOp(db);
+      thread->stats.FinishedOps(nullptr, db, 1);
    }
    char msg[100];
    snprintf(msg, sizeof(msg),
@@ -2423,7 +2695,7 @@ class Benchmark {
        fprintf(stderr, "put error: %s\n", s.ToString().c_str());
        exit(1);
      }
-      thread->stats.FinishedSingleOp(db_);
+      thread->stats.FinishedOps(nullptr, db, 1);
    }

    char msg[100];
@@ -2459,7 +2731,7 @@ class Benchmark {
        fprintf(stderr, "merge error: %s\n", s.ToString().c_str());
        exit(1);
      }
-      thread->stats.FinishedSingleOp(db_);
+      thread->stats.FinishedOps(nullptr, db, 1);
    }

    // Print some statistics
@@ -2520,7 +2792,7 @@ class Benchmark {

      }

-      thread->stats.FinishedSingleOp(db_);
+      thread->stats.FinishedOps(nullptr, db, 1);
    }

    char msg[100];
@@ -2531,17 +2803,47 @@ class Benchmark {
    thread->stats.AddMessage(msg);
  }

+  void WriteSeqSeekSeq(ThreadState* thread) {
+    writes_ = FLAGS_num;
+    DoWrite(thread, SEQUENTIAL);
+    // exclude writes from the ops/sec calculation
+    thread->stats.Start(thread->tid);
+
+    DB* db = SelectDB(thread);
+    std::unique_ptr<Iterator> iter(
+      db->NewIterator(ReadOptions(FLAGS_verify_checksum, true)));
+
+    Slice key = AllocateKey();
+    for (int64_t i = 0; i < FLAGS_num; ++i) {
+      GenerateKeyFromInt(i, FLAGS_num, &key);
+      iter->Seek(key);
+      assert(iter->Valid() && iter->key() == key);
+      thread->stats.FinishedOps(nullptr, db, 1);
+
+      for (int j = 0; j < FLAGS_seekseq_next && i+1 < FLAGS_num; ++j) {
+        iter->Next();
+        GenerateKeyFromInt(++i, FLAGS_num, &key);
+        assert(iter->Valid() && iter->key() == key);
+        thread->stats.FinishedOps(nullptr, db, 1);
+      }
+
+      iter->Seek(key);
+      assert(iter->Valid() && iter->key() == key);
+      thread->stats.FinishedOps(nullptr, db, 1);
+    }
+  }
+
  void Compact(ThreadState* thread) {
    DB* db = SelectDB(thread);
    db->CompactRange(nullptr, nullptr);
  }

  void PrintStats(const char* key) {
-    if (db_ != nullptr) {
-      PrintStats(db_, key, false);
+    if (db_.db != nullptr) {
+      PrintStats(db_.db, key, false);
    }
-    for (DB* db : multi_dbs_) {
-      PrintStats(db, key, true);
+    for (const auto& db_with_cfh : multi_dbs_) {
+      PrintStats(db_with_cfh.db, key, true);
    }
  }

@@ -9,7 +9,10 @@

 #ifndef ROCKSDB_LITE

+#ifndef __STDC_FORMAT_MACROS
 #define __STDC_FORMAT_MACROS
+#endif
+
 #include <inttypes.h>
 #include <algorithm>
 #include <string>
@@ -29,9 +32,9 @@ Status DBImpl::DisableFileDeletions() {
  MutexLock l(&mutex_);
  ++disable_delete_obsolete_files_;
  if (disable_delete_obsolete_files_ == 1) {
-    Log(options_.info_log, "File Deletions Disabled");
+    Log(db_options_.info_log, "File Deletions Disabled");
  } else {
-    Log(options_.info_log,
+    Log(db_options_.info_log,
        "File Deletions Disabled, but already disabled. Counter: %d",
        disable_delete_obsolete_files_);
  }
@@ -50,11 +53,11 @@ Status DBImpl::EnableFileDeletions(bool force) {
      --disable_delete_obsolete_files_;
    }
    if (disable_delete_obsolete_files_ == 0)  {
-      Log(options_.info_log, "File Deletions Enabled");
+      Log(db_options_.info_log, "File Deletions Enabled");
      should_purge_files = true;
      FindObsoleteFiles(deletion_state, true);
    } else {
-      Log(options_.info_log,
+      Log(db_options_.info_log,
          "File Deletions Enable, but not really enabled. Counter: %d",
          disable_delete_obsolete_files_);
    }
@@ -62,10 +65,14 @@ Status DBImpl::EnableFileDeletions(bool force) {
  if (should_purge_files)  {
    PurgeObsoleteFiles(deletion_state);
  }
-  LogFlush(options_.info_log);
+  LogFlush(db_options_.info_log);
  return Status::OK();
 }

+int DBImpl::IsFileDeletionsEnabled() const {
+  return disable_delete_obsolete_files_;
+}
+
 Status DBImpl::GetLiveFiles(std::vector<std::string>& ret,
                            uint64_t* manifest_file_size,
                            bool flush_memtable) {
@@ -91,14 +98,14 @@ Status DBImpl::GetLiveFiles(std::vector<std::string>& ret,

    if (!status.ok()) {
      mutex_.Unlock();
-      Log(options_.info_log, "Cannot Flush data %s\n",
+      Log(db_options_.info_log, "Cannot Flush data %s\n",
          status.ToString().c_str());
      return status;
    }
  }

  // Make a set of all of the live *.sst files
-  std::set<uint64_t> live;
+  std::vector<FileDescriptor> live;
  for (auto cfd : *versions_->GetColumnFamilySet()) {
    cfd->current()->AddLiveFiles(&live);
  }
@@ -109,7 +116,7 @@ Status DBImpl::GetLiveFiles(std::vector<std::string>& ret,
  // create names of the live files. The names are not absolute
  // paths, instead they are relative to dbname_;
  for (auto live_file : live) {
-    ret.push_back(TableFileName("", live_file));
+    ret.push_back(MakeTableFileName("", live_file.GetNumber()));
  }

  ret.push_back(CurrentFileName(""));
@@ -129,7 +136,7 @@ Status DBImpl::GetSortedWalFiles(VectorLogPtr& files) {
  Status s;
  // list wal files in main db dir.
  VectorLogPtr logs;
-  s = GetSortedWalsOfType(options_.wal_dir, logs, kAliveLogFile);
+  s = GetSortedWalsOfType(db_options_.wal_dir, logs, kAliveLogFile);
  if (!s.ok()) {
    return s;
  }
@@ -142,7 +149,7 @@ Status DBImpl::GetSortedWalFiles(VectorLogPtr& files) {

  files.clear();
  // list wal files in archive dir.
-  std::string archivedir = ArchivalDirectory(options_.wal_dir);
+  std::string archivedir = ArchivalDirectory(db_options_.wal_dir);
  if (env_->FileExists(archivedir)) {
    s = GetSortedWalsOfType(archivedir, files, kArchivedLogFile);
    if (!s.ok()) {
@@ -153,7 +160,7 @@ Status DBImpl::GetSortedWalFiles(VectorLogPtr& files) {
  uint64_t latest_archived_log_number = 0;
  if (!files.empty()) {
    latest_archived_log_number = files.back()->LogNumber();
-    Log(options_.info_log, "Latest Archived log: %" PRIu64,
+    Log(db_options_.info_log, "Latest Archived log: %" PRIu64,
        latest_archived_log_number);
  }

@@ -166,7 +173,7 @@ Status DBImpl::GetSortedWalFiles(VectorLogPtr& files) {
      // same log in both db dir and archived dir. Simply
      // ignore the one in db dir. Note that, if we read
      // archived dir first, we would have missed the log file.
-      Log(options_.info_log, "%s already moved to archive",
+      Log(db_options_.info_log, "%s already moved to archive",
          log->PathName().c_str());
    }
  }
@@ -10,6 +10,7 @@

 #include <atomic>
 #include <deque>
+#include <limits>
 #include <set>
 #include <utility>
 #include <vector>
@@ -27,9 +28,13 @@
 #include "rocksdb/memtablerep.h"
 #include "rocksdb/transaction_log.h"
 #include "util/autovector.h"
-#include "util/stats_logger.h"
+#include "util/stop_watch.h"
 #include "util/thread_local.h"
+#include "util/scoped_arena_iterator.h"
 #include "db/internal_stats.h"
+#include "db/write_controller.h"
+#include "db/flush_scheduler.h"
+#include "db/write_thread.h"

 namespace rocksdb {

@@ -39,6 +44,7 @@ class Version;
 class VersionEdit;
 class VersionSet;
 class CompactionFilterV2;
+class Arena;

 class DBImpl : public DB {
 public:
@@ -94,13 +100,21 @@ class DBImpl : public DB {
  using DB::GetProperty;
  virtual bool GetProperty(ColumnFamilyHandle* column_family,
                           const Slice& property, std::string* value);
+  using DB::GetIntProperty;
+  virtual bool GetIntProperty(ColumnFamilyHandle* column_family,
+                              const Slice& property, uint64_t* value) override;
  using DB::GetApproximateSizes;
  virtual void GetApproximateSizes(ColumnFamilyHandle* column_family,
                                   const Range* range, int n, uint64_t* sizes);
  using DB::CompactRange;
  virtual Status CompactRange(ColumnFamilyHandle* column_family,
                              const Slice* begin, const Slice* end,
-                              bool reduce_level = false, int target_level = -1);
+                              bool reduce_level = false, int target_level = -1,
+                              uint32_t target_path_id = 0);
+
+  using DB::SetOptions;
+  bool SetOptions(ColumnFamilyHandle* column_family,
+      const std::unordered_map<std::string, std::string>& options_map);

  using DB::NumberLevels;
  virtual int NumberLevels(ColumnFamilyHandle* column_family);
@@ -121,6 +135,7 @@ class DBImpl : public DB {
 #ifndef ROCKSDB_LITE
  virtual Status DisableFileDeletions();
  virtual Status EnableFileDeletions(bool force);
+  virtual int IsFileDeletionsEnabled() const;
  // All the returned filenames start with "/"
  virtual Status GetLiveFiles(std::vector<std::string>&,
                              uint64_t* manifest_file_size,
@@ -143,8 +158,8 @@ class DBImpl : public DB {
  virtual Status GetDbIdentity(std::string& identity);

  Status RunManualCompaction(ColumnFamilyData* cfd, int input_level,
-                             int output_level, const Slice* begin,
-                             const Slice* end);
+                             int output_level, uint32_t output_path_id,
+                             const Slice* begin, const Slice* end);

 #ifndef ROCKSDB_LITE
  // Extra methods (for testing) that are not in the public DB interface
@@ -166,8 +181,8 @@ class DBImpl : public DB {
  // Return an internal iterator over the current state of the database.
  // The keys of this iterator are internal keys (see format.h).
  // The returned iterator should be deleted when no longer needed.
-  Iterator* TEST_NewInternalIterator(ColumnFamilyHandle* column_family =
-                                         nullptr);
+  Iterator* TEST_NewInternalIterator(
+      Arena* arena, ColumnFamilyHandle* column_family = nullptr);

  // Return the maximum overlapping data (in bytes) at next level for any
  // file at a level >= 1.
@@ -195,8 +210,30 @@ class DBImpl : public DB {
                              SequenceNumber* sequence);

  Status TEST_ReadFirstLine(const std::string& fname, SequenceNumber* sequence);
+
+  void TEST_LockMutex();
+
+  void TEST_UnlockMutex();
+
+  // REQUIRES: mutex locked
+  void* TEST_BeginWrite();
+
+  // REQUIRES: mutex locked
+  // pass the pointer that you got from TEST_BeginWrite()
+  void TEST_EndWrite(void* w);
 #endif  // NDEBUG

+  // Structure to store information for candidate files to delete.
+  struct CandidateFileInfo {
+    std::string file_name;
+    uint32_t path_id;
+    CandidateFileInfo(std::string name, uint32_t path)
+        : file_name(name), path_id(path) {}
+    bool operator==(const CandidateFileInfo& other) const {
+      return file_name == other.file_name && path_id == other.path_id;
+    }
+  };
+
  // needed for CleanupIteratorState
  struct DeletionState {
    inline bool HaveSomethingToDelete() const {
@@ -208,10 +245,10 @@ class DBImpl : public DB {
    // a list of all files that we'll consider deleting
    // (every once in a while this is filled up with all files
    // in the DB directory)
-    std::vector<std::string> candidate_files;
+    std::vector<CandidateFileInfo> candidate_files;

    // the list of all live sst files that cannot be deleted
-    std::vector<uint64_t> sst_live;
+    std::vector<FileDescriptor> sst_live;

    // a list of sst files that we need to delete
    std::vector<FileMetaData*> sst_delete_files;
@@ -257,7 +294,7 @@ class DBImpl : public DB {
  // Returns the list of live files in 'live' and the list
  // of all files in the filesystem in 'candidate_files'.
  // If force == false and the last call was less than
-  // options_.delete_obsolete_files_period_micros microseconds ago,
+  // db_options_.delete_obsolete_files_period_micros microseconds ago,
  // it will not fill up the deletion_state
  void FindObsoleteFiles(DeletionState& deletion_state,
                         bool force,
@@ -275,20 +312,23 @@ class DBImpl : public DB {
  Env* const env_;
  const std::string dbname_;
  unique_ptr<VersionSet> versions_;
-  const DBOptions options_;
+  const DBOptions db_options_;
+  Statistics* stats_;

  Iterator* NewInternalIterator(const ReadOptions&, ColumnFamilyData* cfd,
-                                SuperVersion* super_version);
+                                SuperVersion* super_version, Arena* arena);

 private:
  friend class DB;
  friend class InternalStats;
 #ifndef ROCKSDB_LITE
-  friend class TailingIterator;
+  friend class ForwardIterator;
 #endif
  friend struct SuperVersion;
+  friend class CompactedDBImpl;
  struct CompactionState;
-  struct Writer;
+
+  struct WriteContext;

  Status NewDB();

@@ -307,12 +347,13 @@ class DBImpl : public DB {

  // Flush the in-memory write buffer to storage.  Switches to a new
  // log-file/memtable and writes a new descriptor iff successful.
-  Status FlushMemTableToOutputFile(ColumnFamilyData* cfd, bool* madeProgress,
-                                   DeletionState& deletion_state,
-                                   LogBuffer* log_buffer);
+  Status FlushMemTableToOutputFile(
+      ColumnFamilyData* cfd, const MutableCFOptions& mutable_cf_options,
+      bool* madeProgress, DeletionState& deletion_state, LogBuffer* log_buffer);

-  Status RecoverLogFile(uint64_t log_number, SequenceNumber* max_sequence,
-                        bool read_only);
+  // REQUIRES: log_numbers are sorted in ascending order
+  Status RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
+                         SequenceNumber* max_sequence, bool read_only);

  // The following two methods are used to flush a memtable to
  // storage. The first one is used atdatabase RecoveryTime (when the
@@ -321,20 +362,17 @@ class DBImpl : public DB {
  // concurrent flush memtables to storage.
  Status WriteLevel0TableForRecovery(ColumnFamilyData* cfd, MemTable* mem,
                                     VersionEdit* edit);
-  Status WriteLevel0Table(ColumnFamilyData* cfd, autovector<MemTable*>& mems,
-                          VersionEdit* edit, uint64_t* filenumber,
-                          LogBuffer* log_buffer);
+  Status WriteLevel0Table(ColumnFamilyData* cfd,
+      const MutableCFOptions& mutable_cf_options,
+      const autovector<MemTable*>& mems,
+      VersionEdit* edit, uint64_t* filenumber, LogBuffer* log_buffer);

-  uint64_t SlowdownAmount(int n, double bottom, double top);
+  void DelayWrite(uint64_t expiration_time);

-  // TODO(icanadi) free superversion_to_free and old_log outside of mutex
-  Status MakeRoomForWrite(ColumnFamilyData* cfd,
-                          bool force /* flush even if there is room? */,
-                          autovector<SuperVersion*>* superversions_to_free,
-                          autovector<log::Writer*>* logs_to_free);
+  Status ScheduleFlushes(WriteContext* context);

-  void BuildBatchGroup(Writer** last_writer,
-                       autovector<WriteBatch*>* write_batch_group);
+  Status SetNewMemtableAndNewLogFile(ColumnFamilyData* cfd,
+                                     WriteContext* context);

  // Force current memtable contents to be flushed.
  Status FlushMemTable(ColumnFamilyData* cfd, const FlushOptions& options);
@@ -342,12 +380,8 @@ class DBImpl : public DB {
  // Wait for memtable flushed
  Status WaitForFlushMemTable(ColumnFamilyData* cfd);

-  void MaybeScheduleLogDBDeployStats();
-
-#ifndef ROCKSDB_LITE
-  static void BGLogDBDeployStats(void* db);
-  void LogDBDeployStats();
-#endif  // ROCKSDB_LITE
+  void RecordFlushIOStats();
+  void RecordCompactionIOStats();

  void MaybeScheduleFlushOrCompaction();
  static void BGWorkCompaction(void* db);
@@ -360,6 +394,7 @@ class DBImpl : public DB {
                         LogBuffer* log_buffer);
  void CleanupCompaction(CompactionState* compact, Status status);
  Status DoCompactionWork(CompactionState* compact,
+                          const MutableCFOptions& mutable_cf_options,
                          DeletionState& deletion_state,
                          LogBuffer* log_buffer);

@@ -367,12 +402,14 @@ class DBImpl : public DB {
  // preempt compaction, since it's higher prioirty
  // Returns: micros spent executing
  uint64_t CallFlushDuringCompaction(ColumnFamilyData* cfd,
-                                     DeletionState& deletion_state,
-                                     LogBuffer* log_buffer);
+      const MutableCFOptions& mutable_cf_options, DeletionState& deletion_state,
+      LogBuffer* log_buffer);

  // Call compaction filter if is_compaction_v2 is not true. Then iterate
  // through input and compact the kv-pairs
  Status ProcessKeyValueCompaction(
+    const MutableCFOptions& mutable_cf_options,
+    bool is_snapshot_supported,
    SequenceNumber visible_at_tip,
    SequenceNumber earliest_snapshot,
    SequenceNumber latest_snapshot,
@@ -382,16 +419,18 @@ class DBImpl : public DB {
    Iterator* input,
    CompactionState* compact,
    bool is_compaction_v2,
+    int* num_output_records,
    LogBuffer* log_buffer);

  // Call compaction_filter_v2->Filter() on kv-pairs in compact
  void CallCompactionFilterV2(CompactionState* compact,
    CompactionFilterV2* compaction_filter_v2);

-  Status OpenCompactionOutputFile(CompactionState* compact);
+  Status OpenCompactionOutputFile(CompactionState* compact,
+      const MutableCFOptions& mutable_cf_options);
  Status FinishCompactionOutputFile(CompactionState* compact, Iterator* input);
  Status InstallCompactionResults(CompactionState* compact,
-                                  LogBuffer* log_buffer);
+      const MutableCFOptions& mutable_cf_options, LogBuffer* log_buffer);
  void AllocateCompactionOutputFileNumbers(CompactionState* compact);
  void ReleaseCompactionUnusedFileNumbers(CompactionState* compact);

@@ -433,7 +472,8 @@ class DBImpl : public DB {

  // 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(ColumnFamilyData* cfd, int level);
+  int FindMinimumEmptyLevelFitting(ColumnFamilyData* cfd,
+      const MutableCFOptions& mutable_cf_options, int level);

  // Move the files in the input level to the target level.
  // If target_level < 0, automatically calculate the minimum level that could
@@ -449,11 +489,20 @@ class DBImpl : public DB {
  // State below is protected by mutex_
  port::Mutex mutex_;
  port::AtomicPointer shutting_down_;
-  port::CondVar bg_cv_;          // Signalled when background work finishes
+  // This condition variable is signaled on these conditions:
+  // * whenever bg_compaction_scheduled_ goes down to 0
+  // * if bg_manual_only_ > 0, whenever a compaction finishes, even if it hasn't
+  // made any progress
+  // * whenever a compaction made any progress
+  // * whenever bg_flush_scheduled_ value decreases (i.e. whenever a flush is
+  // done, even if it didn't make any progress)
+  // * whenever there is an error in background flush or compaction
+  port::CondVar bg_cv_;
  uint64_t logfile_number_;
  unique_ptr<log::Writer> log_;
  bool log_empty_;
  ColumnFamilyHandleImpl* default_cf_handle_;
+  InternalStats* default_cf_internal_stats_;
  unique_ptr<ColumnFamilyMemTablesImpl> column_family_memtables_;
  struct LogFileNumberSize {
    explicit LogFileNumberSize(uint64_t _number)
@@ -468,15 +517,19 @@ class DBImpl : public DB {
  // only used for dynamically adjusting max_total_wal_size. it is a sum of
  // [write_buffer_size * max_write_buffer_number] over all column families
  uint64_t max_total_in_memory_state_;
-
-  std::string host_name_;
+  // If true, we have only one (default) column family. We use this to optimize
+  // some code-paths
+  bool single_column_family_mode_;

  std::unique_ptr<Directory> db_directory_;

-  // Queue of writers.
-  std::deque<Writer*> writers_;
+  WriteThread write_thread_;
+
  WriteBatch tmp_batch_;

+  WriteController write_controller_;
+  FlushScheduler flush_scheduler_;
+
  SnapshotList snapshots_;

  // cache for ReadFirstRecord() calls
@@ -485,7 +538,8 @@ class DBImpl : public DB {

  // Set of table files to protect from deletion because they are
  // part of ongoing compactions.
-  std::set<uint64_t> pending_outputs_;
+  // map from pending file number ID to their path IDs.
+  FileNumToPathIdMap pending_outputs_;

  // At least one compaction or flush job is pending but not yet scheduled
  // because of the max background thread limit.
@@ -502,14 +556,12 @@ class DBImpl : public DB {
  // 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_;
-
  // Information for a manual compaction
  struct ManualCompaction {
    ColumnFamilyData* cfd;
    int input_level;
    int output_level;
+    uint32_t output_path_id;
    bool done;
    Status status;
    bool in_progress;           // compaction request being processed?
@@ -522,10 +574,6 @@ class DBImpl : public DB {
  // Have we encountered a background error in paranoid mode?
  Status bg_error_;

-  std::unique_ptr<StatsLogger> logger_;
-
-  int64_t volatile last_log_ts;
-
  // shall we disable deletion of obsolete files
  // if 0 the deletion is enabled.
  // if non-zero, files will not be getting deleted
@@ -552,11 +600,8 @@ class DBImpl : public DB {
  static const int KEEP_LOG_FILE_NUM = 1000;
  std::string db_absolute_path_;

-  // count of the number of contiguous delaying writes
-  int delayed_writes_;
-
  // The options to access storage files
-  const EnvOptions storage_options_;
+  const EnvOptions env_options_;

  // A value of true temporarily disables scheduling of background work
  bool bg_work_gate_closed_;
@@ -571,9 +616,6 @@ class DBImpl : public DB {
  DBImpl(const DBImpl&);
  void operator=(const DBImpl&);

-  // dump the delayed_writes_ to the log file and reset counter.
-  void DelayLoggingAndReset();
-
  // Return the earliest snapshot where seqno is visible.
  // Store the snapshot right before that, if any, in prev_snapshot
  inline SequenceNumber findEarliestVisibleSnapshot(
@@ -585,7 +627,18 @@ class DBImpl : public DB {
  // the cfd->InstallSuperVersion() function. Background threads carry
  // deletion_state which can have new_superversion already allocated.
  void InstallSuperVersion(ColumnFamilyData* cfd,
-                           DeletionState& deletion_state);
+                           DeletionState& deletion_state,
+                           const MutableCFOptions& mutable_cf_options);
+
+  // Find Super version and reference it. Based on options, it might return
+  // the thread local cached one.
+  inline SuperVersion* GetAndRefSuperVersion(ColumnFamilyData* cfd);
+
+  // Un-reference the super version and return it to thread local cache if
+  // needed. If it is the last reference of the super version. Clean it up
+  // after un-referencing it.
+  inline void ReturnAndCleanupSuperVersion(ColumnFamilyData* cfd,
+                                           SuperVersion* sv);

 #ifndef ROCKSDB_LITE
  using DB::GetPropertiesOfAllTables;
@@ -599,25 +652,28 @@ class DBImpl : public DB {
  Status GetImpl(const ReadOptions& options, ColumnFamilyHandle* column_family,
                 const Slice& key, std::string* value,
                 bool* value_found = nullptr);
+
+  bool GetIntPropertyInternal(ColumnFamilyHandle* column_family,
+                              DBPropertyType property_type,
+                              bool need_out_of_mutex, uint64_t* value);
 };

 // Sanitize db options.  The caller should delete result.info_log if
 // it is not equal to src.info_log.
 extern Options SanitizeOptions(const std::string& db,
                               const InternalKeyComparator* icmp,
-                               const InternalFilterPolicy* ipolicy,
                               const Options& src);
 extern DBOptions SanitizeOptions(const std::string& db, const DBOptions& src);

-// Determine compression type, based on user options, level of the output
-// file and whether compression is disabled.
-// If enable_compression is false, then compression is always disabled no
-// matter what the values of the other two parameters are.
-// Otherwise, the compression type is determined based on options and level.
-CompressionType GetCompressionType(const Options& options, int level,
-                                   const bool enable_compression);
+// Fix user-supplied options to be reasonable
+template <class T, class V>
+static void ClipToRange(T* ptr, V minvalue, V maxvalue) {
+  if (static_cast<V>(*ptr) > maxvalue) *ptr = maxvalue;
+  if (static_cast<V>(*ptr) < minvalue) *ptr = minvalue;
+}

-// Determine compression type for L0 file written by memtable flush.
-CompressionType GetCompressionFlush(const Options& options);
+// Dump db file summary, implemented in util/
+extern void DumpDBFileSummary(const DBOptions& options,
+                              const std::string& dbname);

 }  // namespace rocksdb
@@ -20,7 +20,8 @@ uint64_t DBImpl::TEST_GetLevel0TotalSize() {
  return default_cf_handle_->cfd()->current()->NumLevelBytes(0);
 }

-Iterator* DBImpl::TEST_NewInternalIterator(ColumnFamilyHandle* column_family) {
+Iterator* DBImpl::TEST_NewInternalIterator(Arena* arena,
+                                           ColumnFamilyHandle* column_family) {
  ColumnFamilyData* cfd;
  if (column_family == nullptr) {
    cfd = default_cf_handle_->cfd();
@@ -33,7 +34,7 @@ Iterator* DBImpl::TEST_NewInternalIterator(ColumnFamilyHandle* column_family) {
  SuperVersion* super_version = cfd->GetSuperVersion()->Ref();
  mutex_.Unlock();
  ReadOptions roptions;
-  return NewInternalIterator(roptions, cfd, super_version);
+  return NewInternalIterator(roptions, cfd, super_version, arena);
 }

 int64_t DBImpl::TEST_MaxNextLevelOverlappingBytes(
@@ -85,7 +86,7 @@ Status DBImpl::TEST_CompactRange(int level, const Slice* begin,
       cfd->options()->compaction_style == kCompactionStyleFIFO)
          ? level
          : level + 1;
-  return RunManualCompaction(cfd, level, output_level, begin, end);
+  return RunManualCompaction(cfd, level, output_level, 0, begin, end);
 }

 Status DBImpl::TEST_FlushMemTable(bool wait) {
@@ -129,5 +130,27 @@ Status DBImpl::TEST_ReadFirstLine(const std::string& fname,
                                  SequenceNumber* sequence) {
  return ReadFirstLine(fname, sequence);
 }
+
+void DBImpl::TEST_LockMutex() {
+  mutex_.Lock();
+}
+
+void DBImpl::TEST_UnlockMutex() {
+  mutex_.Unlock();
+}
+
+void* DBImpl::TEST_BeginWrite() {
+  auto w = new WriteThread::Writer(&mutex_);
+  Status s = write_thread_.EnterWriteThread(w, 0);
+  assert(s.ok() && !w->done);  // No timeout and nobody should do our job
+  return reinterpret_cast<void*>(w);
+}
+
+void DBImpl::TEST_EndWrite(void* w) {
+  auto writer = reinterpret_cast<WriteThread::Writer*>(w);
+  write_thread_.ExitWriteThread(writer, writer, Status::OK());
+  delete writer;
+}
+
 }  // namespace rocksdb
 #endif  // ROCKSDB_LITE
@@ -2,57 +2,28 @@
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
-//
-// Copyright (c) 2012 Facebook. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.

 #include "db/db_impl_readonly.h"
+#include "utilities/compacted_db/compacted_db_impl.h"
 #include "db/db_impl.h"
-
-#include <algorithm>
-#include <set>
-#include <string>
-#include <stdint.h>
-#include <stdio.h>
-#include <vector>
-#include <algorithm>
-#include "db/db_iter.h"
-#include "db/dbformat.h"
-#include "db/filename.h"
-#include "db/log_reader.h"
-#include "db/log_writer.h"
-#include "db/memtable.h"
 #include "db/merge_context.h"
-#include "db/table_cache.h"
-#include "db/version_set.h"
-#include "db/write_batch_internal.h"
-#include "rocksdb/db.h"
-#include "rocksdb/env.h"
-#include "rocksdb/status.h"
-#include "rocksdb/table.h"
-#include "rocksdb/merge_operator.h"
-#include "port/port.h"
-#include "table/block.h"
-#include "table/merger.h"
-#include "table/two_level_iterator.h"
-#include "util/coding.h"
-#include "util/logging.h"
-#include "util/build_version.h"
+#include "db/db_iter.h"
+#include "util/perf_context_imp.h"

 namespace rocksdb {

-DBImplReadOnly::DBImplReadOnly(const DBOptions& options,
+DBImplReadOnly::DBImplReadOnly(const DBOptions& db_options,
                               const std::string& dbname)
-    : DBImpl(options, dbname) {
-  Log(options_.info_log, "Opening the db in read only mode");
+    : DBImpl(db_options, dbname) {
+  Log(INFO_LEVEL, db_options_.info_log, "Opening the db in read only mode");
+  LogFlush(db_options_.info_log);
 }

 DBImplReadOnly::~DBImplReadOnly() {
 }

 // Implementations of the DB interface
-Status DBImplReadOnly::Get(const ReadOptions& options,
+Status DBImplReadOnly::Get(const ReadOptions& read_options,
                           ColumnFamilyHandle* column_family, const Slice& key,
                           std::string* value) {
  Status s;
@@ -62,34 +33,75 @@ Status DBImplReadOnly::Get(const ReadOptions& options,
  SuperVersion* super_version = cfd->GetSuperVersion();
  MergeContext merge_context;
  LookupKey lkey(key, snapshot);
-  if (super_version->mem->Get(lkey, value, &s, merge_context,
-                              *cfd->options())) {
+  if (super_version->mem->Get(lkey, value, &s, &merge_context)) {
  } else {
-    Version::GetStats stats;
-    super_version->current->Get(options, lkey, value, &s, &merge_context,
-                                &stats);
+    PERF_TIMER_GUARD(get_from_output_files_time);
+    super_version->current->Get(read_options, lkey, value, &s, &merge_context);
  }
  return s;
 }

-Iterator* DBImplReadOnly::NewIterator(const ReadOptions& options,
+Iterator* DBImplReadOnly::NewIterator(const ReadOptions& read_options,
                                      ColumnFamilyHandle* column_family) {
  auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
  auto cfd = cfh->cfd();
  SuperVersion* super_version = cfd->GetSuperVersion()->Ref();
  SequenceNumber latest_snapshot = versions_->LastSequence();
-  Iterator* internal_iter = NewInternalIterator(options, cfd, super_version);
-  return NewDBIterator(
-      env_, *cfd->options(), cfd->user_comparator(), internal_iter,
-      (options.snapshot != nullptr
-           ? reinterpret_cast<const SnapshotImpl*>(options.snapshot)->number_
-           : latest_snapshot));
+  auto db_iter = NewArenaWrappedDbIterator(
+      env_, *cfd->ioptions(), cfd->user_comparator(),
+      (read_options.snapshot != nullptr
+           ? reinterpret_cast<const SnapshotImpl*>(
+                read_options.snapshot)->number_
+           : latest_snapshot),
+      cfd->options()->max_sequential_skip_in_iterations);
+  auto internal_iter = NewInternalIterator(
+      read_options, cfd, super_version, db_iter->GetArena());
+  db_iter->SetIterUnderDBIter(internal_iter);
+  return db_iter;
+}
+
+Status DBImplReadOnly::NewIterators(
+    const ReadOptions& read_options,
+    const std::vector<ColumnFamilyHandle*>& column_families,
+    std::vector<Iterator*>* iterators) {
+  if (iterators == nullptr) {
+    return Status::InvalidArgument("iterators not allowed to be nullptr");
+  }
+  iterators->clear();
+  iterators->reserve(column_families.size());
+  SequenceNumber latest_snapshot = versions_->LastSequence();
+
+  for (auto cfh : column_families) {
+    auto cfd = reinterpret_cast<ColumnFamilyHandleImpl*>(cfh)->cfd();
+    auto db_iter = NewArenaWrappedDbIterator(
+        env_, *cfd->ioptions(), cfd->user_comparator(),
+        (read_options.snapshot != nullptr
+            ? reinterpret_cast<const SnapshotImpl*>(
+                  read_options.snapshot)->number_
+            : latest_snapshot),
+        cfd->options()->max_sequential_skip_in_iterations);
+    auto internal_iter = NewInternalIterator(
+        read_options, cfd, cfd->GetSuperVersion()->Ref(), db_iter->GetArena());
+    db_iter->SetIterUnderDBIter(internal_iter);
+    iterators->push_back(db_iter);
+  }
+
+  return Status::OK();
 }

 Status DB::OpenForReadOnly(const Options& options, const std::string& dbname,
                           DB** dbptr, bool error_if_log_file_exist) {
  *dbptr = nullptr;

+  // Try to first open DB as fully compacted DB
+  Status s;
+#ifndef ROCKSDB_LITE
+  s = CompactedDBImpl::Open(options, dbname, dbptr);
+  if (s.ok()) {
+    return s;
+  }
+#endif
+
  DBOptions db_options(options);
  ColumnFamilyOptions cf_options(options);
  std::vector<ColumnFamilyDescriptor> column_families;
@@ -97,8 +109,7 @@ Status DB::OpenForReadOnly(const Options& options, const std::string& dbname,
      ColumnFamilyDescriptor(kDefaultColumnFamilyName, cf_options));
  std::vector<ColumnFamilyHandle*> handles;

-  Status s =
-      DB::OpenForReadOnly(db_options, dbname, column_families, &handles, dbptr);
+  s = DB::OpenForReadOnly(db_options, dbname, column_families, &handles, dbptr);
  if (s.ok()) {
    assert(handles.size() == 1);
    // i can delete the handle since DBImpl is always holding a
@@ -2,25 +2,11 @@
 //  This source code is licensed under the BSD-style license found in the
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
-//
-// Copyright (c) 2012 Facebook. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.

 #pragma once
 #include "db/db_impl.h"
-
-#include <deque>
-#include <set>
 #include <vector>
 #include <string>
-#include "db/dbformat.h"
-#include "db/log_writer.h"
-#include "db/snapshot.h"
-#include "rocksdb/db.h"
-#include "rocksdb/env.h"
-#include "port/port.h"
-#include "util/stats_logger.h"

 namespace rocksdb {

@@ -33,63 +19,66 @@ class DBImplReadOnly : public DBImpl {
  using DB::Get;
  virtual Status Get(const ReadOptions& options,
                     ColumnFamilyHandle* column_family, const Slice& key,
-                     std::string* value);
+                     std::string* value) override;

  // TODO: Implement ReadOnly MultiGet?

  using DBImpl::NewIterator;
  virtual Iterator* NewIterator(const ReadOptions&,
-                                ColumnFamilyHandle* column_family);
+                                ColumnFamilyHandle* column_family) override;

  virtual Status NewIterators(
      const ReadOptions& options,
-      const std::vector<ColumnFamilyHandle*>& column_family,
-      std::vector<Iterator*>* iterators) {
-   // TODO
-    return Status::NotSupported("Not supported yet.");
-  }
+      const std::vector<ColumnFamilyHandle*>& column_families,
+      std::vector<Iterator*>* iterators) override;

  using DBImpl::Put;
  virtual Status Put(const WriteOptions& options,
                     ColumnFamilyHandle* column_family, const Slice& key,
-                     const Slice& value) {
+                     const Slice& value) override {
    return Status::NotSupported("Not supported operation in read only mode.");
  }
  using DBImpl::Merge;
  virtual Status Merge(const WriteOptions& options,
                       ColumnFamilyHandle* column_family, const Slice& key,
-                       const Slice& value) {
+                       const Slice& value) override {
    return Status::NotSupported("Not supported operation in read only mode.");
  }
  using DBImpl::Delete;
  virtual Status Delete(const WriteOptions& options,
-                        ColumnFamilyHandle* column_family, const Slice& key) {
+                        ColumnFamilyHandle* column_family,
+                        const Slice& key) override {
    return Status::NotSupported("Not supported operation in read only mode.");
  }
-  virtual Status Write(const WriteOptions& options, WriteBatch* updates) {
+  virtual Status Write(const WriteOptions& options,
+                       WriteBatch* updates) override {
    return Status::NotSupported("Not supported operation in read only mode.");
  }
  using DBImpl::CompactRange;
  virtual Status CompactRange(ColumnFamilyHandle* column_family,
                              const Slice* begin, const Slice* end,
-                              bool reduce_level = false,
-                              int target_level = -1) {
+                              bool reduce_level = false, int target_level = -1,
+                              uint32_t target_path_id = 0) override {
    return Status::NotSupported("Not supported operation in read only mode.");
  }
-  virtual Status DisableFileDeletions() {
+
+#ifndef ROCKSDB_LITE
+  virtual Status DisableFileDeletions() override {
    return Status::NotSupported("Not supported operation in read only mode.");
  }
-  virtual Status EnableFileDeletions(bool force) {
+  virtual Status EnableFileDeletions(bool force) override {
    return Status::NotSupported("Not supported operation in read only mode.");
  }
  virtual Status GetLiveFiles(std::vector<std::string>&,
                              uint64_t* manifest_file_size,
-                              bool flush_memtable = true) {
+                              bool flush_memtable = true) override {
    return Status::NotSupported("Not supported operation in read only mode.");
  }
+#endif  // ROCKSDB_LITE
+
  using DBImpl::Flush;
  virtual Status Flush(const FlushOptions& options,
-                       ColumnFamilyHandle* column_family) {
+                       ColumnFamilyHandle* column_family) override {
    return Status::NotSupported("Not supported operation in read only mode.");
  }

@@ -10,6 +10,8 @@
 #include "db/db_iter.h"
 #include <stdexcept>
 #include <deque>
+#include <string>
+#include <limits>

 #include "db/filename.h"
 #include "db/dbformat.h"
@@ -18,6 +20,7 @@
 #include "rocksdb/iterator.h"
 #include "rocksdb/merge_operator.h"
 #include "port/port.h"
+#include "util/arena.h"
 #include "util/logging.h"
 #include "util/mutexlock.h"
 #include "util/perf_context_imp.h"
@@ -37,8 +40,6 @@ static void DumpInternalIter(Iterator* iter) {
 }
 #endif

-namespace {
-
 // Memtables and sstables that make the DB representation contain
 // (userkey,seq,type) => uservalue entries.  DBIter
 // combines multiple entries for the same userkey found in the DB
@@ -57,24 +58,37 @@ class DBIter: public Iterator {
    kReverse
  };

-  DBIter(Env* env, const Options& options,
-         const Comparator* cmp, Iterator* iter, SequenceNumber s)
-      : env_(env),
-        logger_(options.info_log.get()),
+  DBIter(Env* env, const ImmutableCFOptions& ioptions,
+         const Comparator* cmp, Iterator* iter, SequenceNumber s,
+         bool arena_mode, uint64_t max_sequential_skip_in_iterations,
+         const Slice* iterate_upper_bound = nullptr)
+      : arena_mode_(arena_mode),
+        env_(env),
+        logger_(ioptions.info_log),
        user_comparator_(cmp),
-        user_merge_operator_(options.merge_operator.get()),
+        user_merge_operator_(ioptions.merge_operator),
        iter_(iter),
        sequence_(s),
        direction_(kForward),
        valid_(false),
        current_entry_is_merged_(false),
-        statistics_(options.statistics.get()) {
-    RecordTick(statistics_, NO_ITERATORS, 1);
-    max_skip_ = options.max_sequential_skip_in_iterations;
+        statistics_(ioptions.statistics),
+        iterate_upper_bound_(iterate_upper_bound) {
+    RecordTick(statistics_, NO_ITERATORS);
+    prefix_extractor_ = ioptions.prefix_extractor;
+    max_skip_ = max_sequential_skip_in_iterations;
  }
  virtual ~DBIter() {
    RecordTick(statistics_, NO_ITERATORS, -1);
-    delete iter_;
+    if (!arena_mode_) {
+      delete iter_;
+    } else {
+      iter_->~Iterator();
+    }
+  }
+  virtual void SetIter(Iterator* iter) {
+    assert(iter_ == nullptr);
+    iter_ = iter;
  }
  virtual bool Valid() const { return valid_; }
  virtual Slice key() const {
@@ -101,9 +115,14 @@ class DBIter: public Iterator {
  virtual void SeekToLast();

 private:
+  void PrevInternal();
+  void FindParseableKey(ParsedInternalKey* ikey, Direction direction);
+  bool FindValueForCurrentKey();
+  bool FindValueForCurrentKeyUsingSeek();
+  void FindPrevUserKey();
+  void FindNextUserKey();
  inline void FindNextUserEntry(bool skipping);
  void FindNextUserEntryInternal(bool skipping);
-  void FindPrevUserEntry();
  bool ParseKey(ParsedInternalKey* key);
  void MergeValuesNewToOld();

@@ -116,21 +135,24 @@ class DBIter: public Iterator {
    }
  }

+  const SliceTransform* prefix_extractor_;
+  bool arena_mode_;
  Env* const env_;
  Logger* logger_;
  const Comparator* const user_comparator_;
  const MergeOperator* const user_merge_operator_;
-  Iterator* const iter_;
+  Iterator* iter_;
  SequenceNumber const sequence_;

  Status status_;
-  IterKey saved_key_;   // == current key when direction_==kReverse
-  std::string saved_value_;   // == current raw value when direction_==kReverse
+  IterKey saved_key_;
+  std::string saved_value_;
  Direction direction_;
  bool valid_;
  bool current_entry_is_merged_;
  Statistics* statistics_;
  uint64_t max_skip_;
+  const Slice* iterate_upper_bound_;

  // No copying allowed
  DBIter(const DBIter&);
@@ -151,20 +173,11 @@ inline bool DBIter::ParseKey(ParsedInternalKey* ikey) {
 void DBIter::Next() {
  assert(valid_);

-  if (direction_ == kReverse) {  // Switch directions?
+  if (direction_ == kReverse) {
+    FindNextUserKey();
    direction_ = kForward;
-    // iter_ is pointing just before the entries for this->key(),
-    // so advance into the range of entries for this->key() and then
-    // use the normal skipping code below.
    if (!iter_->Valid()) {
      iter_->SeekToFirst();
-    } else {
-      iter_->Next();
-    }
-    if (!iter_->Valid()) {
-      valid_ = false;
-      saved_key_.Clear();
-      return;
    }
  }

@@ -176,7 +189,6 @@ void DBIter::Next() {
  FindNextUserEntry(true /* skipping the current user key */);
 }

-
 // PRE: saved_key_ has the current user key if skipping
 // POST: saved_key_ should have the next user key if valid_,
 //       if the current entry is a result of merge
@@ -186,9 +198,8 @@ void DBIter::Next() {
 // NOTE: In between, saved_key_ can point to a user key that has
 //       a delete marker
 inline void DBIter::FindNextUserEntry(bool skipping) {
-  PERF_TIMER_AUTO(find_next_user_entry_time);
+  PERF_TIMER_GUARD(find_next_user_entry_time);
  FindNextUserEntryInternal(skipping);
-  PERF_TIMER_STOP(find_next_user_entry_time);
 }

 // Actual implementation of DBIter::FindNextUserEntry()
@@ -200,36 +211,44 @@ void DBIter::FindNextUserEntryInternal(bool skipping) {
  uint64_t num_skipped = 0;
  do {
    ParsedInternalKey ikey;
-    if (ParseKey(&ikey) && ikey.sequence <= sequence_) {
-      if (skipping &&
-          user_comparator_->Compare(ikey.user_key, saved_key_.GetKey()) <= 0) {
-        num_skipped++; // skip this entry
-        PERF_COUNTER_ADD(internal_key_skipped_count, 1);
-      } else {
-        skipping = false;
-        switch (ikey.type) {
-          case kTypeDeletion:
-            // Arrange to skip all upcoming entries for this key since
-            // they are hidden by this deletion.
-            saved_key_.SetUserKey(ikey.user_key);
-            skipping = true;
-            num_skipped = 0;
-            PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
-            break;
-          case kTypeValue:
-            valid_ = true;
-            saved_key_.SetUserKey(ikey.user_key);
-            return;
-          case kTypeMerge:
-            // By now, we are sure the current ikey is going to yield a value
-            saved_key_.SetUserKey(ikey.user_key);
-            current_entry_is_merged_ = true;
-            valid_ = true;
-            MergeValuesNewToOld();  // Go to a different state machine
-            return;
-          default:
-            assert(false);
-            break;
+
+    if (ParseKey(&ikey)) {
+      if (iterate_upper_bound_ != nullptr &&
+          ikey.user_key.compare(*iterate_upper_bound_) >= 0) {
+        break;
+      }
+
+      if (ikey.sequence <= sequence_) {
+        if (skipping &&
+           user_comparator_->Compare(ikey.user_key, saved_key_.GetKey()) <= 0) {
+          num_skipped++;  // skip this entry
+          PERF_COUNTER_ADD(internal_key_skipped_count, 1);
+        } else {
+          skipping = false;
+          switch (ikey.type) {
+            case kTypeDeletion:
+              // Arrange to skip all upcoming entries for this key since
+              // they are hidden by this deletion.
+              saved_key_.SetKey(ikey.user_key);
+              skipping = true;
+              num_skipped = 0;
+              PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
+              break;
+            case kTypeValue:
+              valid_ = true;
+              saved_key_.SetKey(ikey.user_key);
+              return;
+            case kTypeMerge:
+              // By now, we are sure the current ikey is going to yield a value
+              saved_key_.SetKey(ikey.user_key);
+              current_entry_is_merged_ = true;
+              valid_ = true;
+              MergeValuesNewToOld();  // Go to a different state machine
+              return;
+            default:
+              assert(false);
+              break;
+          }
        }
      }
    }
@@ -268,7 +287,6 @@ void DBIter::MergeValuesNewToOld() {
  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)) {
@@ -318,110 +336,259 @@ void DBIter::MergeValuesNewToOld() {

 void DBIter::Prev() {
  assert(valid_);
-
-  // 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"
-                           " if merge_operator is provided");
-  }
-
-  if (direction_ == kForward) {  // Switch directions?
-    // iter_ is pointing at the current entry.  Scan backwards until
-    // the key changes so we can use the normal reverse scanning code.
-    assert(iter_->Valid());  // Otherwise valid_ would have been false
-    saved_key_.SetUserKey(ExtractUserKey(iter_->key()));
-    while (true) {
-      iter_->Prev();
-      if (!iter_->Valid()) {
-        valid_ = false;
-        saved_key_.Clear();
-        ClearSavedValue();
-        return;
-      }
-      if (user_comparator_->Compare(ExtractUserKey(iter_->key()),
-                                    saved_key_.GetKey()) < 0) {
-        break;
-      }
-    }
+  if (direction_ == kForward) {
+    FindPrevUserKey();
    direction_ = kReverse;
  }
-
-  FindPrevUserEntry();
+  PrevInternal();
 }

-void DBIter::FindPrevUserEntry() {
-  assert(direction_ == kReverse);
-  uint64_t num_skipped = 0;
-
-  ValueType value_type = kTypeDeletion;
-  bool saved_key_valid = true;
-  if (iter_->Valid()) {
-    do {
-      ParsedInternalKey ikey;
-      if (ParseKey(&ikey) && ikey.sequence <= sequence_) {
-        if ((value_type != kTypeDeletion) &&
-            user_comparator_->Compare(ikey.user_key, saved_key_.GetKey()) < 0) {
-          // We encountered a non-deleted value in entries for previous keys,
-          break;
-        }
-        value_type = ikey.type;
-        if (value_type == kTypeDeletion) {
-          saved_key_.Clear();
-          ClearSavedValue();
-          saved_key_valid = false;
-        } else {
-          Slice raw_value = iter_->value();
-          if (saved_value_.capacity() > raw_value.size() + 1048576) {
-            std::string empty;
-            swap(empty, saved_value_);
-          }
-          saved_key_.SetUserKey(ExtractUserKey(iter_->key()));
-          saved_value_.assign(raw_value.data(), raw_value.size());
-        }
-      } else {
-        // In the case of ikey.sequence > sequence_, we might have already
-        // iterated to a different user key.
-        saved_key_valid = false;
-      }
-      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 (saved_key_valid && num_skipped > max_skip_) {
-        num_skipped = 0;
-        std::string last_key;
-        AppendInternalKey(&last_key, ParsedInternalKey(saved_key_.GetKey(),
-                                                       kMaxSequenceNumber,
-                                                       kValueTypeForSeek));
-        iter_->Seek(last_key);
-        RecordTick(statistics_, NUMBER_OF_RESEEKS_IN_ITERATION);
-      } else {
-        iter_->Prev();
-      }
-    } while (iter_->Valid());
+void DBIter::PrevInternal() {
+  if (!iter_->Valid()) {
+    valid_ = false;
+    return;
  }

-  if (value_type == kTypeDeletion) {
-    // End
+  ParsedInternalKey ikey;
+
+  while (iter_->Valid()) {
+    saved_key_.SetKey(ExtractUserKey(iter_->key()));
+    if (FindValueForCurrentKey()) {
+      valid_ = true;
+      if (!iter_->Valid()) {
+        return;
+      }
+      FindParseableKey(&ikey, kReverse);
+      if (user_comparator_->Compare(ikey.user_key, saved_key_.GetKey()) == 0) {
+        FindPrevUserKey();
+      }
+      return;
+    }
+    if (!iter_->Valid()) {
+      break;
+    }
+    FindParseableKey(&ikey, kReverse);
+    if (user_comparator_->Compare(ikey.user_key, saved_key_.GetKey()) == 0) {
+
+      FindPrevUserKey();
+    }
+  }
+  // We haven't found any key - iterator is not valid
+  assert(!iter_->Valid());
+  valid_ = false;
+}
+
+// This function checks, if the entry with biggest sequence_number <= sequence_
+// is non kTypeDeletion. If it's not, we save value in saved_value_
+bool DBIter::FindValueForCurrentKey() {
+  assert(iter_->Valid());
+  // Contains operands for merge operator.
+  std::deque<std::string> operands;
+  // last entry before merge (could be kTypeDeletion or kTypeValue)
+  ValueType last_not_merge_type = kTypeDeletion;
+  ValueType last_key_entry_type = kTypeDeletion;
+
+  ParsedInternalKey ikey;
+  FindParseableKey(&ikey, kReverse);
+
+  size_t num_skipped = 0;
+  while (iter_->Valid() && ikey.sequence <= sequence_ &&
+         (user_comparator_->Compare(ikey.user_key, saved_key_.GetKey()) == 0)) {
+    // We iterate too much: let's use Seek() to avoid too much key comparisons
+    if (num_skipped >= max_skip_) {
+      return FindValueForCurrentKeyUsingSeek();
+    }
+
+    last_key_entry_type = ikey.type;
+    switch (last_key_entry_type) {
+      case kTypeValue:
+        operands.clear();
+        saved_value_ = iter_->value().ToString();
+        last_not_merge_type = kTypeValue;
+        break;
+      case kTypeDeletion:
+        operands.clear();
+        last_not_merge_type = kTypeDeletion;
+        PERF_COUNTER_ADD(internal_delete_skipped_count, 1);
+        break;
+      case kTypeMerge:
+        assert(user_merge_operator_ != nullptr);
+        operands.push_back(iter_->value().ToString());
+        break;
+      default:
+        assert(false);
+    }
+
+    PERF_COUNTER_ADD(internal_key_skipped_count, 1);
+    assert(user_comparator_->Compare(ikey.user_key, saved_key_.GetKey()) == 0);
+    iter_->Prev();
+    ++num_skipped;
+    FindParseableKey(&ikey, kReverse);
+  }
+
+  switch (last_key_entry_type) {
+    case kTypeDeletion:
+      valid_ = false;
+      return false;
+    case kTypeMerge:
+      if (last_not_merge_type == kTypeDeletion) {
+        user_merge_operator_->FullMerge(saved_key_.GetKey(), nullptr, operands,
+                                        &saved_value_, logger_);
+      } else {
+        assert(last_not_merge_type == kTypeValue);
+        std::string last_put_value = saved_value_;
+        Slice temp_slice(last_put_value);
+        user_merge_operator_->FullMerge(saved_key_.GetKey(), &temp_slice,
+                                        operands, &saved_value_, logger_);
+      }
+      break;
+    case kTypeValue:
+      // do nothing - we've already has value in saved_value_
+      break;
+    default:
+      assert(false);
+      break;
+  }
+  valid_ = true;
+  return true;
+}
+
+// This function is used in FindValueForCurrentKey.
+// We use Seek() function instead of Prev() to find necessary value
+bool DBIter::FindValueForCurrentKeyUsingSeek() {
+  std::string last_key;
+  AppendInternalKey(&last_key, ParsedInternalKey(saved_key_.GetKey(), sequence_,
+                                                 kValueTypeForSeek));
+  iter_->Seek(last_key);
+  RecordTick(statistics_, NUMBER_OF_RESEEKS_IN_ITERATION);
+
+  // assume there is at least one parseable key for this user key
+  ParsedInternalKey ikey;
+  FindParseableKey(&ikey, kForward);
+
+  if (ikey.type == kTypeValue || ikey.type == kTypeDeletion) {
+    if (ikey.type == kTypeValue) {
+      saved_value_ = iter_->value().ToString();
+      valid_ = true;
+      return true;
+    }
    valid_ = false;
-    saved_key_.Clear();
-    ClearSavedValue();
-    direction_ = kForward;
-  } else {
+    return false;
+  }
+
+  // kTypeMerge. We need to collect all kTypeMerge values and save them
+  // in operands
+  std::deque<std::string> operands;
+  while (iter_->Valid() &&
+         (user_comparator_->Compare(ikey.user_key, saved_key_.GetKey()) == 0) &&
+         ikey.type == kTypeMerge) {
+    operands.push_front(iter_->value().ToString());
+    iter_->Next();
+    FindParseableKey(&ikey, kForward);
+  }
+
+  if (!iter_->Valid() ||
+      (user_comparator_->Compare(ikey.user_key, saved_key_.GetKey()) != 0) ||
+      ikey.type == kTypeDeletion) {
+    user_merge_operator_->FullMerge(saved_key_.GetKey(), nullptr, operands,
+                                    &saved_value_, logger_);
+
+    // Make iter_ valid and point to saved_key_
+    if (!iter_->Valid() ||
+        (user_comparator_->Compare(ikey.user_key, saved_key_.GetKey()) != 0)) {
+      iter_->Seek(last_key);
+      RecordTick(statistics_, NUMBER_OF_RESEEKS_IN_ITERATION);
+    }
    valid_ = true;
+    return true;
+  }
+
+  const Slice& value = iter_->value();
+  user_merge_operator_->FullMerge(saved_key_.GetKey(), &value, operands,
+                                  &saved_value_, logger_);
+  valid_ = true;
+  return true;
+}
+
+// Used in Next to change directions
+// Go to next user key
+// Don't use Seek(),
+// because next user key will be very close
+void DBIter::FindNextUserKey() {
+  if (!iter_->Valid()) {
+    return;
+  }
+  ParsedInternalKey ikey;
+  FindParseableKey(&ikey, kForward);
+  while (iter_->Valid() &&
+         user_comparator_->Compare(ikey.user_key, saved_key_.GetKey()) != 0) {
+    iter_->Next();
+    FindParseableKey(&ikey, kForward);
+  }
+}
+
+// Go to previous user_key
+void DBIter::FindPrevUserKey() {
+  if (!iter_->Valid()) {
+    return;
+  }
+  size_t num_skipped = 0;
+  ParsedInternalKey ikey;
+  FindParseableKey(&ikey, kReverse);
+  while (iter_->Valid() &&
+         user_comparator_->Compare(ikey.user_key, saved_key_.GetKey()) == 0) {
+    if (num_skipped >= max_skip_) {
+      num_skipped = 0;
+      IterKey last_key;
+      last_key.SetInternalKey(ParsedInternalKey(
+          saved_key_.GetKey(), kMaxSequenceNumber, kValueTypeForSeek));
+      iter_->Seek(last_key.GetKey());
+      RecordTick(statistics_, NUMBER_OF_RESEEKS_IN_ITERATION);
+    }
+
+    iter_->Prev();
+    ++num_skipped;
+    FindParseableKey(&ikey, kReverse);
+  }
+}
+
+// Skip all unparseable keys
+void DBIter::FindParseableKey(ParsedInternalKey* ikey, Direction direction) {
+  while (iter_->Valid() && !ParseKey(ikey)) {
+    if (direction == kReverse) {
+      iter_->Prev();
+    } else {
+      iter_->Next();
+    }
  }
 }

 void DBIter::Seek(const Slice& target) {
+  StopWatch sw(env_, statistics_, DB_SEEK);
+
+  // total ordering is not guaranteed if prefix_extractor is set
+  // hence prefix based seeks will not give correct results
+  if (iterate_upper_bound_ != nullptr && prefix_extractor_ != nullptr) {
+    if (!prefix_extractor_->InDomain(*iterate_upper_bound_) ||
+        !prefix_extractor_->InDomain(target) ||
+        prefix_extractor_->Transform(*iterate_upper_bound_).compare(
+          prefix_extractor_->Transform(target)) != 0) {
+      status_ = Status::InvalidArgument("read_options.iterate_*_bound "
+                  " and seek target need to have the same prefix.");
+      valid_ = false;
+      return;
+    }
+  }
+
  saved_key_.Clear();
  // now savved_key is used to store internal key.
  saved_key_.SetInternalKey(target, sequence_);
-  PERF_TIMER_AUTO(seek_internal_seek_time);
-  iter_->Seek(saved_key_.GetKey());
-  PERF_TIMER_STOP(seek_internal_seek_time);
+
+  {
+    PERF_TIMER_GUARD(seek_internal_seek_time);
+    iter_->Seek(saved_key_.GetKey());
+  }
+
  if (iter_->Valid()) {
    direction_ = kForward;
    ClearSavedValue();
@@ -432,11 +599,19 @@ void DBIter::Seek(const Slice& target) {
 }

 void DBIter::SeekToFirst() {
+  // Don't use iter_::Seek() if we set a prefix extractor
+  // because prefix seek wiil be used.
+  if (prefix_extractor_ != nullptr) {
+    max_skip_ = std::numeric_limits<uint64_t>::max();
+  }
  direction_ = kForward;
  ClearSavedValue();
-  PERF_TIMER_AUTO(seek_internal_seek_time);
-  iter_->SeekToFirst();
-  PERF_TIMER_STOP(seek_internal_seek_time);
+
+  {
+    PERF_TIMER_GUARD(seek_internal_seek_time);
+    iter_->SeekToFirst();
+  }
+
  if (iter_->Valid()) {
    FindNextUserEntry(false /* not skipping */);
  } else {
@@ -445,32 +620,73 @@ void DBIter::SeekToFirst() {
 }

 void DBIter::SeekToLast() {
-  // Throw an exception for now if merge_operator is provided
-  // TODO: support backward iteration
-  if (user_merge_operator_) {
-    Log(logger_, "SeekToLast not supported yet if merge_operator is provided");
-    throw std::logic_error("DBIter::SeekToLast: backward iteration not"
-                           " supported if merge_operator is provided");
+  // Don't use iter_::Seek() if we set a prefix extractor
+  // because prefix seek wiil be used.
+  if (prefix_extractor_ != nullptr) {
+    max_skip_ = std::numeric_limits<uint64_t>::max();
  }
-
  direction_ = kReverse;
  ClearSavedValue();
-  PERF_TIMER_AUTO(seek_internal_seek_time);
-  iter_->SeekToLast();
-  PERF_TIMER_STOP(seek_internal_seek_time);
-  FindPrevUserEntry();
+
+  {
+    PERF_TIMER_GUARD(seek_internal_seek_time);
+    iter_->SeekToLast();
+  }
+
+  PrevInternal();
 }

-}  // anonymous namespace
+Iterator* NewDBIterator(Env* env, const ImmutableCFOptions& ioptions,
+                        const Comparator* user_key_comparator,
+                        Iterator* internal_iter,
+                        const SequenceNumber& sequence,
+                        uint64_t max_sequential_skip_in_iterations,
+                        const Slice* iterate_upper_bound) {
+  return new DBIter(env, ioptions, user_key_comparator, internal_iter, sequence,
+                    false, max_sequential_skip_in_iterations,
+                    iterate_upper_bound);
+}

-Iterator* NewDBIterator(
-    Env* env,
-    const Options& options,
-    const Comparator *user_key_comparator,
-    Iterator* internal_iter,
-    const SequenceNumber& sequence) {
-  return new DBIter(env, options, user_key_comparator,
-                    internal_iter, sequence);
+ArenaWrappedDBIter::~ArenaWrappedDBIter() { db_iter_->~DBIter(); }
+
+void ArenaWrappedDBIter::SetDBIter(DBIter* iter) { db_iter_ = iter; }
+
+void ArenaWrappedDBIter::SetIterUnderDBIter(Iterator* iter) {
+  static_cast<DBIter*>(db_iter_)->SetIter(iter);
+}
+
+inline bool ArenaWrappedDBIter::Valid() const { return db_iter_->Valid(); }
+inline void ArenaWrappedDBIter::SeekToFirst() { db_iter_->SeekToFirst(); }
+inline void ArenaWrappedDBIter::SeekToLast() { db_iter_->SeekToLast(); }
+inline void ArenaWrappedDBIter::Seek(const Slice& target) {
+  db_iter_->Seek(target);
+}
+inline void ArenaWrappedDBIter::Next() { db_iter_->Next(); }
+inline void ArenaWrappedDBIter::Prev() { db_iter_->Prev(); }
+inline Slice ArenaWrappedDBIter::key() const { return db_iter_->key(); }
+inline Slice ArenaWrappedDBIter::value() const { return db_iter_->value(); }
+inline Status ArenaWrappedDBIter::status() const { return db_iter_->status(); }
+void ArenaWrappedDBIter::RegisterCleanup(CleanupFunction function, void* arg1,
+                                         void* arg2) {
+  db_iter_->RegisterCleanup(function, arg1, arg2);
+}
+
+ArenaWrappedDBIter* NewArenaWrappedDbIterator(
+    Env* env, const ImmutableCFOptions& ioptions,
+    const Comparator* user_key_comparator,
+    const SequenceNumber& sequence,
+    uint64_t max_sequential_skip_in_iterations,
+    const Slice* iterate_upper_bound) {
+  ArenaWrappedDBIter* iter = new ArenaWrappedDBIter();
+  Arena* arena = iter->GetArena();
+  auto mem = arena->AllocateAligned(sizeof(DBIter));
+  DBIter* db_iter = new (mem) DBIter(env, ioptions, user_key_comparator,
+      nullptr, sequence, true, max_sequential_skip_in_iterations,
+      iterate_upper_bound);
+
+  iter->SetDBIter(db_iter);
+
+  return iter;
 }

 }  // namespace rocksdb
@@ -11,17 +11,67 @@
 #include <stdint.h>
 #include "rocksdb/db.h"
 #include "db/dbformat.h"
+#include "util/arena.h"
+#include "util/autovector.h"

 namespace rocksdb {

+class Arena;
+class DBIter;
+
 // Return a new iterator that converts internal keys (yielded by
 // "*internal_iter") that were live at the specified "sequence" number
 // into appropriate user keys.
 extern Iterator* NewDBIterator(
    Env* env,
-    const Options& options,
+    const ImmutableCFOptions& options,
    const Comparator *user_key_comparator,
    Iterator* internal_iter,
-    const SequenceNumber& sequence);
+    const SequenceNumber& sequence,
+    uint64_t max_sequential_skip_in_iterations,
+    const Slice* iterate_upper_bound = nullptr);
+
+// A wrapper iterator which wraps DB Iterator and the arena, with which the DB
+// iterator is supposed be allocated. This class is used as an entry point of
+// a iterator hierarchy whose memory can be allocated inline. In that way,
+// accessing the iterator tree can be more cache friendly. It is also faster
+// to allocate.
+class ArenaWrappedDBIter : public Iterator {
+ public:
+  virtual ~ArenaWrappedDBIter();
+
+  // Get the arena to be used to allocate memory for DBIter to be wrapped,
+  // as well as child iterators in it.
+  virtual Arena* GetArena() { return &arena_; }
+
+  // Set the DB Iterator to be wrapped
+
+  virtual void SetDBIter(DBIter* iter);
+
+  // Set the internal iterator wrapped inside the DB Iterator. Usually it is
+  // a merging iterator.
+  virtual void SetIterUnderDBIter(Iterator* iter);
+  virtual bool Valid() const override;
+  virtual void SeekToFirst() override;
+  virtual void SeekToLast() override;
+  virtual void Seek(const Slice& target) override;
+  virtual void Next() override;
+  virtual void Prev() override;
+  virtual Slice key() const override;
+  virtual Slice value() const override;
+  virtual Status status() const override;
+  void RegisterCleanup(CleanupFunction function, void* arg1, void* arg2);
+
+ private:
+  DBIter* db_iter_;
+  Arena arena_;
+};
+
+// Generate the arena wrapped iterator class.
+extern ArenaWrappedDBIter* NewArenaWrappedDbIterator(
+    Env* env, const ImmutableCFOptions& options,
+    const Comparator* user_key_comparator,
+    const SequenceNumber& sequence, uint64_t max_sequential_skip_in_iterations,
+    const Slice* iterate_upper_bound = nullptr);

 }  // namespace rocksdb
@@ -1,95 +0,0 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
-//  This source code is licensed under the BSD-style license found in the
-//  LICENSE file in the root directory of this source tree. An additional grant
-//  of patent rights can be found in the PATENTS file in the same directory.
-//
-// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file. See the AUTHORS file for names of contributors.
-
-#include "db/db_impl.h"
-#include <string>
-#include <stdint.h>
-#include <stdio.h>
-#include "db/version_set.h"
-#include "rocksdb/db.h"
-#include "rocksdb/env.h"
-#include "port/port.h"
-#include "util/mutexlock.h"
-
-namespace rocksdb {
-
-void DBImpl::MaybeScheduleLogDBDeployStats() {
-// we did say maybe
-#ifndef ROCKSDB_LITE
-  // There is a lock in the actual logger.
-  if (!logger_ || options_.db_stats_log_interval < 0
-      || host_name_.empty()) {
-    return;
-  }
-
-  if(bg_logstats_scheduled_ || shutting_down_.Acquire_Load()) {
-    // Already scheduled
-  } else {
-    int64_t current_ts = 0;
-    Status st = env_->GetCurrentTime(&current_ts);
-    if (!st.ok()) {
-      return;
-    }
-    if ((current_ts - last_log_ts) < options_.db_stats_log_interval) {
-      return;
-    }
-    last_log_ts = current_ts;
-    bg_logstats_scheduled_ = true;
-    env_->Schedule(&DBImpl::BGLogDBDeployStats, this);
-  }
-}
-
-void DBImpl::BGLogDBDeployStats(void* db) {
-  DBImpl* db_inst = reinterpret_cast<DBImpl*>(db);
-  db_inst->LogDBDeployStats();
-}
-
-void DBImpl::LogDBDeployStats() {
-  mutex_.Lock();
-
-  if (shutting_down_.Acquire_Load()) {
-    bg_logstats_scheduled_ = false;
-    bg_cv_.SignalAll();
-    mutex_.Unlock();
-    return;
-  }
-
-  char tmp_ver[100];
-  sprintf(tmp_ver, "%d.%d", kMajorVersion, kMinorVersion);
-  std::string version_info(tmp_ver);
-
-  uint64_t file_total_size = 0;
-  uint32_t file_total_num = 0;
-  Version* current = default_cf_handle_->cfd()->current();
-  for (int i = 0; i < current->NumberLevels(); i++) {
-    file_total_num += current->NumLevelFiles(i);
-    file_total_size += current->NumLevelBytes(i);
-  }
-
-  Version::LevelSummaryStorage scratch;
-  const char* file_num_summary = current->LevelSummary(&scratch);
-  std::string file_num_per_level(file_num_summary);
-  std::string data_size_per_level(file_num_summary);
-
-  mutex_.Unlock();
-
-  int64_t unix_ts;
-  env_->GetCurrentTime(&unix_ts);
-
-  logger_->Log_Deploy_Stats(version_info, host_name_,
-      db_absolute_path_, file_total_size, file_total_num, file_num_per_level,
-      data_size_per_level, unix_ts);
-
-  mutex_.Lock();
-  bg_logstats_scheduled_ = false;
-  bg_cv_.SignalAll();
-  mutex_.Unlock();
-#endif
-}
-}
@@ -127,26 +127,6 @@ void InternalKeyComparator::FindShortSuccessor(std::string* key) const {
  }
 }

-const char* InternalFilterPolicy::Name() const {
-  return user_policy_->Name();
-}
-
-void InternalFilterPolicy::CreateFilter(const Slice* keys, int n,
-                                        std::string* dst) const {
-  // We rely on the fact that the code in table.cc does not mind us
-  // adjusting keys[].
-  Slice* mkey = const_cast<Slice*>(keys);
-  for (int i = 0; i < n; i++) {
-    mkey[i] = ExtractUserKey(keys[i]);
-    // TODO(sanjay): Suppress dups?
-  }
-  user_policy_->CreateFilter(keys, n, dst);
-}
-
-bool InternalFilterPolicy::KeyMayMatch(const Slice& key, const Slice& f) const {
-  return user_policy_->KeyMayMatch(ExtractUserKey(key), f);
-}
-
 LookupKey::LookupKey(const Slice& user_key, SequenceNumber s) {
  size_t usize = user_key.size();
  size_t needed = usize + 13;  // A conservative estimate
@@ -9,6 +9,7 @@

 #pragma once
 #include <stdio.h>
+#include <string>
 #include "rocksdb/comparator.h"
 #include "rocksdb/db.h"
 #include "rocksdb/filter_policy.h"
@@ -123,17 +124,6 @@ class InternalKeyComparator : public Comparator {
  int Compare(const ParsedInternalKey& a, const ParsedInternalKey& b) const;
 };

-// Filter policy wrapper that converts from internal keys to user keys
-class InternalFilterPolicy : public FilterPolicy {
- private:
-  const FilterPolicy* const user_policy_;
- public:
-  explicit InternalFilterPolicy(const FilterPolicy* p) : user_policy_(p) { }
-  virtual const char* Name() const;
-  virtual void CreateFilter(const Slice* keys, int n, std::string* dst) const;
-  virtual bool KeyMayMatch(const Slice& key, const Slice& filter) const;
-};
-
 // Modules in this directory should keep internal keys wrapped inside
 // the following class instead of plain strings so that we do not
 // incorrectly use string comparisons instead of an InternalKeyComparator.
@@ -254,26 +244,84 @@ class IterKey {

  Slice GetKey() const { return Slice(key_, key_size_); }

+  size_t Size() { return key_size_; }
+
  void Clear() { key_size_ = 0; }

-  void SetUserKey(const Slice& user_key) {
-    size_t size = user_key.size();
+  // Append "non_shared_data" to its back, from "shared_len"
+  // This function is used in Block::Iter::ParseNextKey
+  // shared_len: bytes in [0, shard_len-1] would be remained
+  // non_shared_data: data to be append, its length must be >= non_shared_len
+  void TrimAppend(const size_t shared_len, const char* non_shared_data,
+                  const size_t non_shared_len) {
+    assert(shared_len <= key_size_);
+
+    size_t total_size = shared_len + non_shared_len;
+    if (total_size <= buf_size_) {
+      key_size_ = total_size;
+    } else {
+      // Need to allocate space, delete previous space
+      char* p = new char[total_size];
+      memcpy(p, key_, shared_len);
+
+      if (key_ != nullptr && key_ != space_) {
+        delete[] key_;
+      }
+
+      key_ = p;
+      key_size_ = total_size;
+      buf_size_ = total_size;
+    }
+
+    memcpy(key_ + shared_len, non_shared_data, non_shared_len);
+  }
+
+  void SetKey(const Slice& key) {
+    size_t size = key.size();
    EnlargeBufferIfNeeded(size);
-    memcpy(key_, user_key.data(), size);
+    memcpy(key_, key.data(), size);
    key_size_ = size;
  }

+  void SetInternalKey(const Slice& key_prefix, const Slice& user_key,
+                      SequenceNumber s,
+                      ValueType value_type = kValueTypeForSeek) {
+    size_t psize = key_prefix.size();
+    size_t usize = user_key.size();
+    EnlargeBufferIfNeeded(psize + usize + sizeof(uint64_t));
+    if (psize > 0) {
+      memcpy(key_, key_prefix.data(), psize);
+    }
+    memcpy(key_ + psize, user_key.data(), usize);
+    EncodeFixed64(key_ + usize + psize, PackSequenceAndType(s, value_type));
+    key_size_ = psize + usize + sizeof(uint64_t);
+  }
+
  void SetInternalKey(const Slice& user_key, SequenceNumber s,
                      ValueType value_type = kValueTypeForSeek) {
-    size_t usize = user_key.size();
-    EnlargeBufferIfNeeded(usize + sizeof(uint64_t));
-    memcpy(key_, user_key.data(), usize);
-    EncodeFixed64(key_ + usize, PackSequenceAndType(s, value_type));
-    key_size_ = usize + sizeof(uint64_t);
+    SetInternalKey(Slice(), user_key, s, value_type);
+  }
+
+  void Reserve(size_t size) {
+    EnlargeBufferIfNeeded(size);
+    key_size_ = size;
  }

  void SetInternalKey(const ParsedInternalKey& parsed_key) {
-    SetInternalKey(parsed_key.user_key, parsed_key.sequence, parsed_key.type);
+    SetInternalKey(Slice(), parsed_key);
+  }
+
+  void SetInternalKey(const Slice& key_prefix,
+                      const ParsedInternalKey& parsed_key_suffix) {
+    SetInternalKey(key_prefix, parsed_key_suffix.user_key,
+                   parsed_key_suffix.sequence, parsed_key_suffix.type);
+  }
+
+  void EncodeLengthPrefixedKey(const Slice& key) {
+    auto size = key.size();
+    EnlargeBufferIfNeeded(size + VarintLength(size));
+    char* ptr = EncodeVarint32(key_, size);
+    memcpy(ptr, key.data(), size);
  }

 private:
@@ -342,4 +390,12 @@ class InternalKeySliceTransform : public SliceTransform {
  const SliceTransform* const transform_;
 };

+// Read record from a write batch piece from input.
+// tag, column_family, key, value and blob are return values. Callers own the
+// Slice they point to.
+// Tag is defined as ValueType.
+// input will be advanced to after the record.
+extern Status ReadRecordFromWriteBatch(Slice* input, char* tag,
+                                       uint32_t* column_family, Slice* key,
+                                       Slice* value, Slice* blob);
 }  // namespace rocksdb
@@ -110,6 +110,45 @@ TEST(FormatTest, InternalKeyShortestSuccessor) {
            ShortSuccessor(IKey("\xff\xff", 100, kTypeValue)));
 }

+TEST(FormatTest, IterKeyOperation) {
+  IterKey k;
+  const char p[] = "abcdefghijklmnopqrstuvwxyz";
+  const char q[] = "0123456789";
+
+  ASSERT_EQ(std::string(k.GetKey().data(), k.GetKey().size()),
+            std::string(""));
+
+  k.TrimAppend(0, p, 3);
+  ASSERT_EQ(std::string(k.GetKey().data(), k.GetKey().size()),
+            std::string("abc"));
+
+  k.TrimAppend(1, p, 3);
+  ASSERT_EQ(std::string(k.GetKey().data(), k.GetKey().size()),
+            std::string("aabc"));
+
+  k.TrimAppend(0, p, 26);
+  ASSERT_EQ(std::string(k.GetKey().data(), k.GetKey().size()),
+            std::string("abcdefghijklmnopqrstuvwxyz"));
+
+  k.TrimAppend(26, q, 10);
+  ASSERT_EQ(std::string(k.GetKey().data(), k.GetKey().size()),
+            std::string("abcdefghijklmnopqrstuvwxyz0123456789"));
+
+  k.TrimAppend(36, q, 1);
+  ASSERT_EQ(std::string(k.GetKey().data(), k.GetKey().size()),
+            std::string("abcdefghijklmnopqrstuvwxyz01234567890"));
+
+  k.TrimAppend(26, q, 1);
+  ASSERT_EQ(std::string(k.GetKey().data(), k.GetKey().size()),
+            std::string("abcdefghijklmnopqrstuvwxyz0"));
+
+  // Size going up, memory allocation is triggered
+  k.TrimAppend(27, p, 26);
+  ASSERT_EQ(std::string(k.GetKey().data(), k.GetKey().size()),
+            std::string("abcdefghijklmnopqrstuvwxyz0"
+              "abcdefghijklmnopqrstuvwxyz"));
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -34,6 +34,7 @@ class DeleteFileTest {
  DeleteFileTest() {
    db_ = nullptr;
    env_ = Env::Default();
+    options_.max_background_flushes = 0;
    options_.write_buffer_size = 1024*1024*1000;
    options_.target_file_size_base = 1024*1024*1000;
    options_.max_bytes_for_level_base = 1024*1024*1000;
@@ -147,7 +148,6 @@ class DeleteFileTest {
 TEST(DeleteFileTest, AddKeysAndQueryLevels) {
  CreateTwoLevels();
  std::vector<LiveFileMetaData> metadata;
-  std::vector<int> keysinlevel;
  db_->GetLiveFilesMetaData(&metadata);

  std::string level1file = "";
@@ -14,21 +14,15 @@

 namespace rocksdb {

-FileIndexer::FileIndexer(const uint32_t num_levels,
-                         const Comparator* ucmp)
-  : num_levels_(num_levels),
-    ucmp_(ucmp),
-    next_level_index_(num_levels),
-    level_rb_(num_levels, -1) {
-}
-
+FileIndexer::FileIndexer(const Comparator* ucmp)
+    : num_levels_(0), ucmp_(ucmp), level_rb_(nullptr) {}

 uint32_t FileIndexer::NumLevelIndex() {
  return next_level_index_.size();
 }

 uint32_t FileIndexer::LevelIndexSize(uint32_t level) {
-  return next_level_index_[level].size();
+  return next_level_index_[level].num_index;
 }

 void FileIndexer::GetNextLevelIndex(
@@ -46,11 +40,13 @@ void FileIndexer::GetNextLevelIndex(
  assert(level < num_levels_ - 1);
  assert(static_cast<int32_t>(file_index) <= level_rb_[level]);

-  const auto& index = next_level_index_[level][file_index];
+  const IndexUnit* index_units = next_level_index_[level].index_units;
+  const auto& index = index_units[file_index];

  if (cmp_smallest < 0) {
-    *left_bound = (level > 0 && file_index > 0) ?
-      next_level_index_[level][file_index - 1].largest_lb : 0;
+    *left_bound = (level > 0 && file_index > 0)
+                      ? index_units[file_index - 1].largest_lb
+                      : 0;
    *right_bound = index.smallest_rb;
  } else if (cmp_smallest == 0) {
    *left_bound = index.smallest_lb;
@@ -73,16 +69,25 @@ void FileIndexer::GetNextLevelIndex(
  assert(*right_bound <= level_rb_[level + 1]);
 }

-void FileIndexer::ClearIndex() {
-  for (uint32_t level = 1; level < num_levels_; ++level) {
-    next_level_index_[level].clear();
-  }
-}
-
-void FileIndexer::UpdateIndex(std::vector<FileMetaData*>* const files) {
+void FileIndexer::UpdateIndex(Arena* arena, const uint32_t num_levels,
+                              std::vector<FileMetaData*>* const files) {
  if (files == nullptr) {
    return;
  }
+  if (num_levels == 0) {  // uint_32 0-1 would cause bad behavior
+    num_levels_ = num_levels;
+    return;
+  }
+  assert(level_rb_ == nullptr);  // level_rb_ should be init here
+
+  num_levels_ = num_levels;
+  next_level_index_.resize(num_levels);
+
+  char* mem = arena->AllocateAligned(num_levels_ * sizeof(int32_t));
+  level_rb_ = new (mem) int32_t[num_levels_];
+  for (size_t i = 0; i < num_levels_; i++) {
+    level_rb_[i] = -1;
+  }

  // L1 - Ln-1
  for (uint32_t level = 1; level < num_levels_ - 1; ++level) {
@@ -93,55 +98,56 @@ void FileIndexer::UpdateIndex(std::vector<FileMetaData*>* const files) {
    if (upper_size == 0) {
      continue;
    }
-    auto& index = next_level_index_[level];
-    index.resize(upper_size);
+    IndexLevel& index_level = next_level_index_[level];
+    index_level.num_index = upper_size;
+    char* mem = arena->AllocateAligned(upper_size * sizeof(IndexUnit));
+    index_level.index_units = new (mem) IndexUnit[upper_size];

-    CalculateLB(upper_files, lower_files, &index,
-        [this](const FileMetaData* a, const FileMetaData* b) -> int {
+    CalculateLB(
+        upper_files, lower_files, &index_level,
+        [this](const FileMetaData * a, const FileMetaData * b)->int {
          return ucmp_->Compare(a->smallest.user_key(), b->largest.user_key());
        },
-        [](IndexUnit* index, int32_t f_idx) {
-          index->smallest_lb = f_idx;
-        });
-    CalculateLB(upper_files, lower_files, &index,
-        [this](const FileMetaData* a, const FileMetaData* b) -> int {
+        [](IndexUnit* index, int32_t f_idx) { index->smallest_lb = f_idx; });
+    CalculateLB(
+        upper_files, lower_files, &index_level,
+        [this](const FileMetaData * a, const FileMetaData * b)->int {
          return ucmp_->Compare(a->largest.user_key(), b->largest.user_key());
        },
-        [](IndexUnit* index, int32_t f_idx) {
-          index->largest_lb = f_idx;
-        });
-    CalculateRB(upper_files, lower_files, &index,
-        [this](const FileMetaData* a, const FileMetaData* b) -> int {
+        [](IndexUnit* index, int32_t f_idx) { index->largest_lb = f_idx; });
+    CalculateRB(
+        upper_files, lower_files, &index_level,
+        [this](const FileMetaData * a, const FileMetaData * b)->int {
          return ucmp_->Compare(a->smallest.user_key(), b->smallest.user_key());
        },
-        [](IndexUnit* index, int32_t f_idx) {
-          index->smallest_rb = f_idx;
-        });
-    CalculateRB(upper_files, lower_files, &index,
-        [this](const FileMetaData* a, const FileMetaData* b) -> int {
+        [](IndexUnit* index, int32_t f_idx) { index->smallest_rb = f_idx; });
+    CalculateRB(
+        upper_files, lower_files, &index_level,
+        [this](const FileMetaData * a, const FileMetaData * b)->int {
          return ucmp_->Compare(a->largest.user_key(), b->smallest.user_key());
        },
-        [](IndexUnit* index, int32_t f_idx) {
-          index->largest_rb = f_idx;
-        });
+        [](IndexUnit* index, int32_t f_idx) { index->largest_rb = f_idx; });
  }
+
  level_rb_[num_levels_ - 1] = files[num_levels_ - 1].size() - 1;
 }

-void FileIndexer::CalculateLB(const std::vector<FileMetaData*>& upper_files,
-    const std::vector<FileMetaData*>& lower_files,
-    std::vector<IndexUnit>* index,
+void FileIndexer::CalculateLB(
+    const std::vector<FileMetaData*>& upper_files,
+    const std::vector<FileMetaData*>& lower_files, IndexLevel* index_level,
    std::function<int(const FileMetaData*, const FileMetaData*)> cmp_op,
    std::function<void(IndexUnit*, int32_t)> set_index) {
  const int32_t upper_size = upper_files.size();
  const int32_t lower_size = lower_files.size();
  int32_t upper_idx = 0;
  int32_t lower_idx = 0;
+
+  IndexUnit* index = index_level->index_units;
  while (upper_idx < upper_size && lower_idx < lower_size) {
    int cmp = cmp_op(upper_files[upper_idx], lower_files[lower_idx]);

    if (cmp == 0) {
-      set_index(&(*index)[upper_idx], lower_idx);
+      set_index(&index[upper_idx], lower_idx);
      ++upper_idx;
      ++lower_idx;
    } else if (cmp > 0) {
@@ -151,7 +157,7 @@ void FileIndexer::CalculateLB(const std::vector<FileMetaData*>& upper_files,
    } else {
      // Lower level's file becomes larger, update the index, and
      // move to the next upper file
-      set_index(&(*index)[upper_idx], lower_idx);
+      set_index(&index[upper_idx], lower_idx);
      ++upper_idx;
    }
  }
@@ -159,25 +165,27 @@ void FileIndexer::CalculateLB(const std::vector<FileMetaData*>& upper_files,
  while (upper_idx < upper_size) {
    // Lower files are exhausted, that means the remaining upper files are
    // greater than any lower files. Set the index to be the lower level size.
-    set_index(&(*index)[upper_idx], lower_size);
+    set_index(&index[upper_idx], lower_size);
    ++upper_idx;
  }
 }

-void FileIndexer::CalculateRB(const std::vector<FileMetaData*>& upper_files,
-    const std::vector<FileMetaData*>& lower_files,
-    std::vector<IndexUnit>* index,
+void FileIndexer::CalculateRB(
+    const std::vector<FileMetaData*>& upper_files,
+    const std::vector<FileMetaData*>& lower_files, IndexLevel* index_level,
    std::function<int(const FileMetaData*, const FileMetaData*)> cmp_op,
    std::function<void(IndexUnit*, int32_t)> set_index) {
  const int32_t upper_size = upper_files.size();
  const int32_t lower_size = lower_files.size();
  int32_t upper_idx = upper_size - 1;
  int32_t lower_idx = lower_size - 1;
+
+  IndexUnit* index = index_level->index_units;
  while (upper_idx >= 0 && lower_idx >= 0) {
    int cmp = cmp_op(upper_files[upper_idx], lower_files[lower_idx]);

    if (cmp == 0) {
-      set_index(&(*index)[upper_idx], lower_idx);
+      set_index(&index[upper_idx], lower_idx);
      --upper_idx;
      --lower_idx;
    } else if (cmp < 0) {
@@ -187,14 +195,14 @@ void FileIndexer::CalculateRB(const std::vector<FileMetaData*>& upper_files,
    } else {
      // Lower level's file becomes smaller, update the index, and move to
      // the next the upper file
-      set_index(&(*index)[upper_idx], lower_idx);
+      set_index(&index[upper_idx], lower_idx);
      --upper_idx;
    }
  }
  while (upper_idx >= 0) {
    // Lower files are exhausted, that means the remaining upper files are
    // smaller than any lower files. Set it to -1.
-    set_index(&(*index)[upper_idx], -1);
+    set_index(&index[upper_idx], -1);
    --upper_idx;
  }
 }
@@ -12,11 +12,15 @@
 #include <functional>
 #include <limits>
 #include <vector>
+#include "util/arena.h"
+#include "util/autovector.h"

 namespace rocksdb {

 class Comparator;
 struct FileMetaData;
+struct FdWithKeyRange;
+struct FileLevel;

 // The file tree structure in Version is prebuilt and the range of each file
 // is known. On Version::Get(), it uses binary search to find a potential file
@@ -36,7 +40,7 @@ struct FileMetaData;
 // naive approach.
 class FileIndexer {
 public:
-  FileIndexer(const uint32_t num_levels, const Comparator* ucmp);
+  explicit FileIndexer(const Comparator* ucmp);

  uint32_t NumLevelIndex();

@@ -50,16 +54,15 @@ class FileIndexer {
    const uint32_t level, const uint32_t file_index, const int cmp_smallest,
    const int cmp_largest, int32_t* left_bound, int32_t* right_bound);

-  void ClearIndex();
-
-  void UpdateIndex(std::vector<FileMetaData*>* const files);
+  void UpdateIndex(Arena* arena, const uint32_t num_levels,
+                   std::vector<FileMetaData*>* const files);

  enum {
    kLevelMaxIndex = std::numeric_limits<int32_t>::max()
  };

 private:
-  const uint32_t num_levels_;
+  uint32_t num_levels_;
  const Comparator* ucmp_;

  struct IndexUnit {
@@ -110,20 +113,28 @@ class FileIndexer {
    int32_t largest_rb;
  };

-  void CalculateLB(const std::vector<FileMetaData*>& upper_files,
-    const std::vector<FileMetaData*>& lower_files,
-    std::vector<IndexUnit>* index,
-    std::function<int(const FileMetaData*, const FileMetaData*)> cmp_op,
-    std::function<void(IndexUnit*, int32_t)> set_index);
+  // Data structure to store IndexUnits in a whole level
+  struct IndexLevel {
+    size_t num_index;
+    IndexUnit* index_units;

-  void CalculateRB(const std::vector<FileMetaData*>& upper_files,
-    const std::vector<FileMetaData*>& lower_files,
-    std::vector<IndexUnit>* index,
-    std::function<int(const FileMetaData*, const FileMetaData*)> cmp_op,
-    std::function<void(IndexUnit*, int32_t)> set_index);
+    IndexLevel() : num_index(0), index_units(nullptr) {}
+  };

-  std::vector<std::vector<IndexUnit>> next_level_index_;
-  std::vector<int32_t> level_rb_;
+  void CalculateLB(
+      const std::vector<FileMetaData*>& upper_files,
+      const std::vector<FileMetaData*>& lower_files, IndexLevel* index_level,
+      std::function<int(const FileMetaData*, const FileMetaData*)> cmp_op,
+      std::function<void(IndexUnit*, int32_t)> set_index);
+
+  void CalculateRB(
+      const std::vector<FileMetaData*>& upper_files,
+      const std::vector<FileMetaData*>& lower_files, IndexLevel* index_level,
+      std::function<int(const FileMetaData*, const FileMetaData*)> cmp_op,
+      std::function<void(IndexUnit*, int32_t)> set_index);
+
+  autovector<IndexLevel> next_level_index_;
+  int32_t* level_rb_;
 };

 }  // namespace rocksdb
@@ -35,16 +35,13 @@ class IntComparator : public Comparator {
  void FindShortSuccessor(std::string* key) const {}
 };

-
 struct FileIndexerTest {
 public:
-  FileIndexerTest() :
-    kNumLevels(4), indexer(kNumLevels, &ucmp),
-    files(new std::vector<FileMetaData*>[kNumLevels]) {
-  }
+  FileIndexerTest()
+      : kNumLevels(4), files(new std::vector<FileMetaData*>[kNumLevels]) {}

  ~FileIndexerTest() {
-    Reset();
+    ClearFiles();
    delete[] files;
  }

@@ -59,14 +56,13 @@ struct FileIndexerTest {
    return InternalKey(Slice(reinterpret_cast<char*>(&v), 8), 0, kTypeValue);
  }

-  void Reset() {
+  void ClearFiles() {
    for (uint32_t i = 0; i < kNumLevels; ++i) {
      for (auto* f : files[i]) {
        delete f;
      }
      files[i].clear();
    }
-    indexer.ClearIndex();
  }

  void GetNextLevelIndex(const uint32_t level, const uint32_t file_index,
@@ -74,23 +70,32 @@ struct FileIndexerTest {
      int32_t* right_index) {
    *left_index = 100;
    *right_index = 100;
-    indexer.GetNextLevelIndex(level, file_index, cmp_smallest, cmp_largest,
-        left_index, right_index);
+    indexer->GetNextLevelIndex(level, file_index, cmp_smallest, cmp_largest,
+                               left_index, right_index);
  }

+  int32_t left = 100;
+  int32_t right = 100;
  const uint32_t kNumLevels;
  IntComparator ucmp;
-  FileIndexer indexer;
+  FileIndexer* indexer;

  std::vector<FileMetaData*>* files;
 };

-TEST(FileIndexerTest, next_level_hint) {
-  for (uint32_t i = 0; i < kNumLevels; ++i) {
-    ASSERT_EQ(0U, indexer.LevelIndexSize(i));
-  }
+// Case 0: Empty
+TEST(FileIndexerTest, Empty) {
+  Arena arena;
+  indexer = new FileIndexer(&ucmp);
+  indexer->UpdateIndex(&arena, 0, files);
+  delete indexer;
+}

-  // Case 1: no overlap, files are on the left of next level files
+// Case 1: no overlap, files are on the left of next level files
+TEST(FileIndexerTest, no_overlap_left) {
+  Arena arena;
+  uint32_t kNumLevels = 4;
+  indexer = new FileIndexer(&ucmp);
  // level 1
  AddFile(1, 100, 200);
  AddFile(1, 300, 400);
@@ -103,9 +108,7 @@ TEST(FileIndexerTest, next_level_hint) {
  AddFile(3, 2500, 2600);
  AddFile(3, 2601, 2699);
  AddFile(3, 2700, 2800);
-  indexer.UpdateIndex(files);
-  int32_t left = 100;
-  int32_t right = 100;
+  indexer->UpdateIndex(&arena, kNumLevels, files);
  for (uint32_t level = 1; level < 3; ++level) {
    for (uint32_t f = 0; f < 3; ++f) {
      GetNextLevelIndex(level, f, -1, -1, &left, &right);
@@ -125,12 +128,15 @@ TEST(FileIndexerTest, next_level_hint) {
      ASSERT_EQ(2, right);
    }
  }
+  delete indexer;
+  ClearFiles();
+}

-  // Case 2: no overlap, files are on the right of next level files
-  Reset();
-  for (uint32_t i = 1; i < kNumLevels; ++i) {
-    ASSERT_EQ(0U, indexer.LevelIndexSize(i));
-  }
+// Case 2: no overlap, files are on the right of next level files
+TEST(FileIndexerTest, no_overlap_right) {
+  Arena arena;
+  uint32_t kNumLevels = 4;
+  indexer = new FileIndexer(&ucmp);
  // level 1
  AddFile(1, 2100, 2200);
  AddFile(1, 2300, 2400);
@@ -143,7 +149,7 @@ TEST(FileIndexerTest, next_level_hint) {
  AddFile(3, 500, 600);
  AddFile(3, 501, 699);
  AddFile(3, 700, 800);
-  indexer.UpdateIndex(files);
+  indexer->UpdateIndex(&arena, kNumLevels, files);
  for (uint32_t level = 1; level < 3; ++level) {
    for (uint32_t f = 0; f < 3; ++f) {
      GetNextLevelIndex(level, f, -1, -1, &left, &right);
@@ -166,11 +172,16 @@ TEST(FileIndexerTest, next_level_hint) {
      ASSERT_EQ(2, right);
    }
  }
+  delete indexer;
+}

-  // Case 3: empty L2
-  Reset();
+// Case 3: empty L2
+TEST(FileIndexerTest, empty_L2) {
+  Arena arena;
+  uint32_t kNumLevels = 4;
+  indexer = new FileIndexer(&ucmp);
  for (uint32_t i = 1; i < kNumLevels; ++i) {
-    ASSERT_EQ(0U, indexer.LevelIndexSize(i));
+    ASSERT_EQ(0U, indexer->LevelIndexSize(i));
  }
  // level 1
  AddFile(1, 2100, 2200);
@@ -180,7 +191,7 @@ TEST(FileIndexerTest, next_level_hint) {
  AddFile(3, 500, 600);
  AddFile(3, 501, 699);
  AddFile(3, 700, 800);
-  indexer.UpdateIndex(files);
+  indexer->UpdateIndex(&arena, kNumLevels, files);
  for (uint32_t f = 0; f < 3; ++f) {
    GetNextLevelIndex(1, f, -1, -1, &left, &right);
    ASSERT_EQ(0, left);
@@ -201,13 +212,14 @@ TEST(FileIndexerTest, next_level_hint) {
    ASSERT_EQ(0, left);
    ASSERT_EQ(-1, right);
  }
+  delete indexer;
+  ClearFiles();
+}

-
-  // Case 4: mixed
-  Reset();
-  for (uint32_t i = 1; i < kNumLevels; ++i) {
-    ASSERT_EQ(0U, indexer.LevelIndexSize(i));
-  }
+// Case 4: mixed
+TEST(FileIndexerTest, mixed) {
+  Arena arena;
+  indexer = new FileIndexer(&ucmp);
  // level 1
  AddFile(1, 100, 200);
  AddFile(1, 250, 400);
@@ -222,7 +234,7 @@ TEST(FileIndexerTest, next_level_hint) {
  AddFile(3, 0, 50);
  AddFile(3, 100, 200);
  AddFile(3, 201, 250);
-  indexer.UpdateIndex(files);
+  indexer->UpdateIndex(&arena, kNumLevels, files);
  // level 1, 0
  GetNextLevelIndex(1, 0, -1, -1, &left, &right);
  ASSERT_EQ(0, left);
@@ -321,6 +333,8 @@ TEST(FileIndexerTest, next_level_hint) {
    ASSERT_EQ(3, left);
    ASSERT_EQ(2, right);
  }
+  delete indexer;
+  ClearFiles();
 }

 }  // namespace rocksdb
@@ -6,11 +6,16 @@
 // 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 __STDC_FORMAT_MACROS
+#define __STDC_FORMAT_MACROS
+#endif

 #include "db/filename.h"
+#include <inttypes.h>

 #include <ctype.h>
 #include <stdio.h>
+#include <vector>
 #include "db/dbformat.h"
 #include "rocksdb/env.h"
 #include "util/logging.h"
@@ -18,14 +23,16 @@
 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.
+// {[0-9,a-z,A-Z,-,_,.]} with _. And append '_LOG\0' at the end.
 // Return the number of chars stored in dest not including the trailing '\0'.
-static int FlattenPath(const std::string& path, char* dest, int len) {
-  int write_idx = 0;
-  int i = 0;
-  int src_len = path.size();
+static size_t GetInfoLogPrefix(const std::string& path, char* dest, int len) {
+  const char suffix[] = "_LOG";

-  while (i < src_len && write_idx < len - 1) {
+  size_t write_idx = 0;
+  size_t i = 0;
+  size_t src_len = path.size();
+
+  while (i < src_len && write_idx < len - sizeof(suffix)) {
    if ((path[i] >= 'a' && path[i] <= 'z') ||
        (path[i] >= '0' && path[i] <= '9') ||
        (path[i] >= 'A' && path[i] <= 'Z') ||
@@ -39,8 +46,10 @@ static int FlattenPath(const std::string& path, char* dest, int len) {
    }
    i++;
  }
-
-  dest[write_idx] = '\0';
+  assert(sizeof(suffix) <= len - write_idx);
+  // "\0" is automatically added by snprintf
+  snprintf(dest + write_idx, len - write_idx, suffix);
+  write_idx += sizeof(suffix) - 1;
  return write_idx;
 }

@@ -66,9 +75,34 @@ std::string ArchivedLogFileName(const std::string& name, uint64_t number) {
  return MakeFileName(name + "/" + ARCHIVAL_DIR, number, "log");
 }

-std::string TableFileName(const std::string& name, uint64_t number) {
+std::string MakeTableFileName(const std::string& path, uint64_t number) {
+  return MakeFileName(path, number, "sst");
+}
+
+std::string TableFileName(const std::vector<DbPath>& db_paths, uint64_t number,
+                          uint32_t path_id) {
  assert(number > 0);
-  return MakeFileName(name, number, "sst");
+  std::string path;
+  if (path_id >= db_paths.size()) {
+    path = db_paths.back().path;
+  } else {
+    path = db_paths[path_id].path;
+  }
+  return MakeTableFileName(path, number);
+}
+
+const size_t kFormatFileNumberBufSize = 38;
+
+void FormatFileNumber(uint64_t number, uint32_t path_id, char* out_buf,
+                      size_t out_buf_size) {
+  if (path_id == 0) {
+    snprintf(out_buf, out_buf_size, "%" PRIu64, number);
+  } else {
+    snprintf(out_buf, out_buf_size, "%" PRIu64
+                                    "(path "
+                                    "%" PRIu32 ")",
+             number, path_id);
+  }
 }

 std::string DescriptorFileName(const std::string& dbname, uint64_t number) {
@@ -91,14 +125,26 @@ std::string TempFileName(const std::string& dbname, uint64_t number) {
  return MakeFileName(dbname, number, "dbtmp");
 }

+InfoLogPrefix::InfoLogPrefix(bool has_log_dir,
+                             const std::string& db_absolute_path) {
+  if (!has_log_dir) {
+    const char kInfoLogPrefix[] = "LOG";
+    // "\0" is automatically added to the end
+    snprintf(buf, sizeof(buf), kInfoLogPrefix);
+    prefix = Slice(buf, sizeof(kInfoLogPrefix) - 1);
+  } else {
+    size_t len = GetInfoLogPrefix(db_absolute_path, buf, sizeof(buf));
+    prefix = Slice(buf, len);
+  }
+}
+
 std::string InfoLogFileName(const std::string& dbname,
    const std::string& db_path, const std::string& log_dir) {
  if (log_dir.empty())
    return dbname + "/LOG";

-  char flatten_db_path[256];
-  FlattenPath(db_path, flatten_db_path, 256);
-  return log_dir + "/" + flatten_db_path + "_LOG";
+  InfoLogPrefix info_log_prefix(true, db_path);
+  return log_dir + "/" + info_log_prefix.buf;
 }

 // Return the name of the old info log file for "dbname".
@@ -110,9 +156,8 @@ std::string OldInfoLogFileName(const std::string& dbname, uint64_t ts,
  if (log_dir.empty())
    return dbname + "/LOG.old." + buf;

-  char flatten_db_path[256];
-  FlattenPath(db_path, flatten_db_path, 256);
-  return log_dir + "/" + flatten_db_path + "_LOG.old." + buf;
+  InfoLogPrefix info_log_prefix(true, db_path);
+  return log_dir + "/" + info_log_prefix.buf + ".old." + buf;
 }

 std::string MetaDatabaseName(const std::string& dbname, uint64_t number) {
@@ -130,8 +175,8 @@ std::string IdentityFileName(const std::string& dbname) {
 //    dbname/IDENTITY
 //    dbname/CURRENT
 //    dbname/LOCK
-//    dbname/LOG
-//    dbname/LOG.old.[0-9]+
+//    dbname/<info_log_name_prefix>
+//    dbname/<info_log_name_prefix>.old.[0-9]+
 //    dbname/MANIFEST-[0-9]+
 //    dbname/[0-9]+.(log|sst)
 //    dbname/METADB-[0-9]+
@@ -140,6 +185,12 @@ bool ParseFileName(const std::string& fname,
                   uint64_t* number,
                   FileType* type,
                   WalFileType* log_type) {
+  return ParseFileName(fname, number, "", type, log_type);
+}
+
+bool ParseFileName(const std::string& fname, uint64_t* number,
+                   const Slice& info_log_name_prefix, FileType* type,
+                   WalFileType* log_type) {
  Slice rest(fname);
  if (fname.length() > 1 && fname[0] == '/') {
    rest.remove_prefix(1);
@@ -153,18 +204,22 @@ bool ParseFileName(const std::string& fname,
  } else if (rest == "LOCK") {
    *number = 0;
    *type = kDBLockFile;
-  } else if (rest == "LOG" || rest == "LOG.old") {
-    *number = 0;
-    *type = kInfoLogFile;
-  } else if (rest.starts_with("LOG.old.")) {
-    uint64_t ts_suffix;
-    // sizeof also counts the trailing '\0'.
-    rest.remove_prefix(sizeof("LOG.old.") - 1);
-    if (!ConsumeDecimalNumber(&rest, &ts_suffix)) {
-      return false;
+  } else if (info_log_name_prefix.size() > 0 &&
+             rest.starts_with(info_log_name_prefix)) {
+    rest.remove_prefix(info_log_name_prefix.size());
+    if (rest == "" || rest == ".old") {
+      *number = 0;
+      *type = kInfoLogFile;
+    } else if (rest.starts_with(".old.")) {
+      uint64_t ts_suffix;
+      // sizeof also counts the trailing '\0'.
+      rest.remove_prefix(sizeof(".old.") - 1);
+      if (!ConsumeDecimalNumber(&rest, &ts_suffix)) {
+        return false;
+      }
+      *number = ts_suffix;
+      *type = kInfoLogFile;
    }
-    *number = ts_suffix;
-    *type = kInfoLogFile;
  } else if (rest.starts_with("MANIFEST-")) {
    rest.remove_prefix(strlen("MANIFEST-"));
    uint64_t num;
@@ -11,7 +11,9 @@

 #pragma once
 #include <stdint.h>
+#include <unordered_map>
 #include <string>
+#include <vector>
 #include "rocksdb/slice.h"
 #include "rocksdb/status.h"
 #include "rocksdb/transaction_log.h"
@@ -34,6 +36,9 @@ enum FileType {
  kIdentityFile
 };

+// map from file number to path ID.
+typedef std::unordered_map<uint64_t, uint32_t> FileNumToPathIdMap;
+
 // Return the name of the log file with the specified number
 // in the db named by "dbname".  The result will be prefixed with
 // "dbname".
@@ -48,10 +53,19 @@ extern std::string ArchivalDirectory(const std::string& dbname);
 extern std::string ArchivedLogFileName(const std::string& dbname,
                                       uint64_t num);

+extern std::string MakeTableFileName(const std::string& name, uint64_t number);
+
 // Return the name of the sstable with the specified number
 // in the db named by "dbname".  The result will be prefixed with
 // "dbname".
-extern std::string TableFileName(const std::string& dbname, uint64_t number);
+extern std::string TableFileName(const std::vector<DbPath>& db_paths,
+                                 uint64_t number, uint32_t path_id);
+
+// Sufficient buffer size for FormatFileNumber.
+extern const size_t kFormatFileNumberBufSize;
+
+extern void FormatFileNumber(uint64_t number, uint32_t path_id, char* out_buf,
+                             size_t out_buf_size);

 // Return the name of the descriptor file for the db named by
 // "dbname" and the specified incarnation number.  The result will be
@@ -72,6 +86,16 @@ extern std::string LockFileName(const std::string& dbname);
 // The result will be prefixed with "dbname".
 extern std::string TempFileName(const std::string& dbname, uint64_t number);

+// A helper structure for prefix of info log names.
+struct InfoLogPrefix {
+  char buf[260];
+  Slice prefix;
+  // Prefix with DB absolute path encoded
+  explicit InfoLogPrefix(bool has_log_dir, const std::string& db_absolute_path);
+  // Default Prefix
+  explicit InfoLogPrefix();
+};
+
 // Return the name of the info log file for "dbname".
 extern std::string InfoLogFileName(const std::string& dbname,
    const std::string& db_path="", const std::string& log_dir="");
@@ -93,10 +117,13 @@ extern std::string IdentityFileName(const std::string& dbname);
 // If filename is a rocksdb file, store the type of the file in *type.
 // The number encoded in the filename is stored in *number.  If the
 // filename was successfully parsed, returns true.  Else return false.
-extern bool ParseFileName(const std::string& filename,
-                          uint64_t* number,
-                          FileType* type,
+// info_log_name_prefix is the path of info logs.
+extern bool ParseFileName(const std::string& filename, uint64_t* number,
+                          const Slice& info_log_name_prefix, FileType* type,
                          WalFileType* log_type = nullptr);
+// Same as previous function, but skip info log files.
+extern bool ParseFileName(const std::string& filename, uint64_t* number,
+                          FileType* type, WalFileType* log_type = nullptr);

 // Make the CURRENT file point to the descriptor file with the
 // specified number.
@@ -23,30 +23,50 @@ TEST(FileNameTest, Parse) {
  FileType type;
  uint64_t number;

+  char kDefautInfoLogDir = 1;
+  char kDifferentInfoLogDir = 2;
+  char kNoCheckLogDir = 4;
+  char kAllMode = kDefautInfoLogDir | kDifferentInfoLogDir | kNoCheckLogDir;
+
  // Successful parses
  static struct {
    const char* fname;
    uint64_t number;
    FileType type;
+    char mode;
  } cases[] = {
-    { "100.log",            100,   kLogFile },
-    { "0.log",              0,     kLogFile },
-    { "0.sst",              0,     kTableFile },
-    { "CURRENT",            0,     kCurrentFile },
-    { "LOCK",               0,     kDBLockFile },
-    { "MANIFEST-2",         2,     kDescriptorFile },
-    { "MANIFEST-7",         7,     kDescriptorFile },
-    { "METADB-2",           2,     kMetaDatabase },
-    { "METADB-7",           7,     kMetaDatabase },
-    { "LOG",                0,     kInfoLogFile },
-    { "LOG.old",            0,     kInfoLogFile },
-    { "18446744073709551615.log", 18446744073709551615ull, kLogFile },
-  };
-  for (unsigned int i = 0; i < sizeof(cases) / sizeof(cases[0]); i++) {
-    std::string f = cases[i].fname;
-    ASSERT_TRUE(ParseFileName(f, &number, &type)) << f;
-    ASSERT_EQ(cases[i].type, type) << f;
-    ASSERT_EQ(cases[i].number, number) << f;
+        {"100.log", 100, kLogFile, kAllMode},
+        {"0.log", 0, kLogFile, kAllMode},
+        {"0.sst", 0, kTableFile, kAllMode},
+        {"CURRENT", 0, kCurrentFile, kAllMode},
+        {"LOCK", 0, kDBLockFile, kAllMode},
+        {"MANIFEST-2", 2, kDescriptorFile, kAllMode},
+        {"MANIFEST-7", 7, kDescriptorFile, kAllMode},
+        {"METADB-2", 2, kMetaDatabase, kAllMode},
+        {"METADB-7", 7, kMetaDatabase, kAllMode},
+        {"LOG", 0, kInfoLogFile, kDefautInfoLogDir},
+        {"LOG.old", 0, kInfoLogFile, kDefautInfoLogDir},
+        {"LOG.old.6688", 6688, kInfoLogFile, kDefautInfoLogDir},
+        {"rocksdb_dir_LOG", 0, kInfoLogFile, kDifferentInfoLogDir},
+        {"rocksdb_dir_LOG.old", 0, kInfoLogFile, kDifferentInfoLogDir},
+        {"rocksdb_dir_LOG.old.6688", 6688, kInfoLogFile, kDifferentInfoLogDir},
+        {"18446744073709551615.log", 18446744073709551615ull, kLogFile,
+         kAllMode}, };
+  for (char mode : {kDifferentInfoLogDir, kDefautInfoLogDir, kNoCheckLogDir}) {
+    for (unsigned int i = 0; i < sizeof(cases) / sizeof(cases[0]); i++) {
+      InfoLogPrefix info_log_prefix(mode != kDefautInfoLogDir, "/rocksdb/dir");
+      if (cases[i].mode & mode) {
+        std::string f = cases[i].fname;
+        if (mode == kNoCheckLogDir) {
+          ASSERT_TRUE(ParseFileName(f, &number, &type)) << f;
+        } else {
+          ASSERT_TRUE(ParseFileName(f, &number, info_log_prefix.prefix, &type))
+              << f;
+        }
+        ASSERT_EQ(cases[i].type, type) << f;
+        ASSERT_EQ(cases[i].number, number) << f;
+      }
+    }
  }

  // Errors
@@ -85,6 +105,22 @@ TEST(FileNameTest, Parse) {
  };
 }

+TEST(FileNameTest, InfoLogFileName) {
+  std::string dbname = ("/data/rocksdb");
+  std::string db_absolute_path;
+  Env::Default()->GetAbsolutePath(dbname, &db_absolute_path);
+
+  ASSERT_EQ("/data/rocksdb/LOG", InfoLogFileName(dbname, db_absolute_path, ""));
+  ASSERT_EQ("/data/rocksdb/LOG.old.666",
+            OldInfoLogFileName(dbname, 666u, db_absolute_path, ""));
+
+  ASSERT_EQ("/data/rocksdb_log/data_rocksdb_LOG",
+            InfoLogFileName(dbname, db_absolute_path, "/data/rocksdb_log"));
+  ASSERT_EQ(
+      "/data/rocksdb_log/data_rocksdb_LOG.old.666",
+      OldInfoLogFileName(dbname, 666u, db_absolute_path, "/data/rocksdb_log"));
+}
+
 TEST(FileNameTest, Construction) {
  uint64_t number;
  FileType type;
@@ -108,7 +144,10 @@ TEST(FileNameTest, Construction) {
  ASSERT_EQ(192U, number);
  ASSERT_EQ(kLogFile, type);

-  fname = TableFileName("bar", 200);
+  fname = TableFileName({DbPath("bar", 0)}, 200, 0);
+  std::string fname1 =
+      TableFileName({DbPath("foo", 0), DbPath("bar", 0)}, 200, 1);
+  ASSERT_EQ(fname, fname1);
  ASSERT_EQ("bar/", std::string(fname.data(), 4));
  ASSERT_TRUE(ParseFileName(fname.c_str() + 4, &number, &type));
  ASSERT_EQ(200U, number);
@@ -0,0 +1,63 @@
+//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+
+#include "db/flush_scheduler.h"
+
+#include <cassert>
+
+#include "db/column_family.h"
+
+namespace rocksdb {
+
+void FlushScheduler::ScheduleFlush(ColumnFamilyData* cfd) {
+#ifndef NDEBUG
+  assert(column_families_set_.find(cfd) == column_families_set_.end());
+  column_families_set_.insert(cfd);
+#endif  // NDEBUG
+  cfd->Ref();
+  column_families_.push_back(cfd);
+}
+
+ColumnFamilyData* FlushScheduler::GetNextColumnFamily() {
+  ColumnFamilyData* cfd = nullptr;
+  while (column_families_.size() > 0) {
+    cfd = column_families_.front();
+    column_families_.pop_front();
+    if (cfd->IsDropped()) {
+      if (cfd->Unref()) {
+        delete cfd;
+        cfd = nullptr;
+      }
+    } else {
+      break;
+    }
+  }
+#ifndef NDEBUG
+  if (cfd != nullptr) {
+    auto itr = column_families_set_.find(cfd);
+    assert(itr != column_families_set_.end());
+    column_families_set_.erase(itr);
+  }
+#endif  // NDEBUG
+  return cfd;
+}
+
+bool FlushScheduler::Empty() { return column_families_.empty(); }
+
+void FlushScheduler::Clear() {
+  for (auto cfd : column_families_) {
+#ifndef NDEBUG
+    auto itr = column_families_set_.find(cfd);
+    assert(itr != column_families_set_.end());
+    column_families_set_.erase(itr);
+#endif  // NDEBUG
+    if (cfd->Unref()) {
+      delete cfd;
+    }
+  }
+  column_families_.clear();
+}
+
+}  // namespace rocksdb
@@ -0,0 +1,39 @@
+//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+
+#pragma once
+
+#include <stdint.h>
+#include <deque>
+#include <set>
+#include <vector>
+
+namespace rocksdb {
+
+class ColumnFamilyData;
+
+// This class is thread-compatible. It's should only be accessed from single
+// write thread (between BeginWrite() and EndWrite())
+class FlushScheduler {
+ public:
+  FlushScheduler() = default;
+  ~FlushScheduler() = default;
+
+  void ScheduleFlush(ColumnFamilyData* cfd);
+  // Returns Ref()-ed column family. Client needs to Unref()
+  ColumnFamilyData* GetNextColumnFamily();
+
+  bool Empty();
+
+  void Clear();
+
+ private:
+  std::deque<ColumnFamilyData*> column_families_;
+#ifndef NDEBUG
+  std::set<ColumnFamilyData*> column_families_set_;
+#endif  // NDEBUG
+};
+
+}  // namespace rocksdb
@@ -0,0 +1,547 @@
+//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+
+#ifndef ROCKSDB_LITE
+#include "db/forward_iterator.h"
+
+#include <limits>
+#include <string>
+#include <utility>
+
+#include "db/db_impl.h"
+#include "db/db_iter.h"
+#include "db/column_family.h"
+#include "rocksdb/env.h"
+#include "rocksdb/slice.h"
+#include "rocksdb/slice_transform.h"
+#include "table/merger.h"
+#include "db/dbformat.h"
+
+namespace rocksdb {
+
+// Usage:
+//     LevelIterator iter;
+//     iter.SetFileIndex(file_index);
+//     iter.Seek(target);
+//     iter.Next()
+class LevelIterator : public Iterator {
+ public:
+  LevelIterator(const ColumnFamilyData* const cfd,
+      const ReadOptions& read_options,
+      const std::vector<FileMetaData*>& files)
+    : cfd_(cfd), read_options_(read_options), files_(files), valid_(false),
+      file_index_(std::numeric_limits<uint32_t>::max()) {}
+
+  void SetFileIndex(uint32_t file_index) {
+    assert(file_index < files_.size());
+    if (file_index != file_index_) {
+      file_index_ = file_index;
+      Reset();
+    }
+    valid_ = false;
+  }
+  void Reset() {
+    assert(file_index_ < files_.size());
+    file_iter_.reset(cfd_->table_cache()->NewIterator(
+        read_options_, *(cfd_->soptions()), cfd_->internal_comparator(),
+        files_[file_index_]->fd, nullptr /* table_reader_ptr */, false));
+  }
+  void SeekToLast() override {
+    status_ = Status::NotSupported("LevelIterator::SeekToLast()");
+    valid_ = false;
+  }
+  void Prev() {
+    status_ = Status::NotSupported("LevelIterator::Prev()");
+    valid_ = false;
+  }
+  bool Valid() const override {
+    return valid_;
+  }
+  void SeekToFirst() override {
+    SetFileIndex(0);
+    file_iter_->SeekToFirst();
+    valid_ = file_iter_->Valid();
+  }
+  void Seek(const Slice& internal_key) override {
+    assert(file_iter_ != nullptr);
+    file_iter_->Seek(internal_key);
+    valid_ = file_iter_->Valid();
+  }
+  void Next() override {
+    assert(valid_);
+    file_iter_->Next();
+    for (;;) {
+      if (file_iter_->status().IsIncomplete() || file_iter_->Valid()) {
+        valid_ = !file_iter_->status().IsIncomplete();
+        return;
+      }
+      if (file_index_ + 1 >= files_.size()) {
+        valid_ = false;
+        return;
+      }
+      SetFileIndex(file_index_ + 1);
+      file_iter_->SeekToFirst();
+    }
+  }
+  Slice key() const override {
+    assert(valid_);
+    return file_iter_->key();
+  }
+  Slice value() const override {
+    assert(valid_);
+    return file_iter_->value();
+  }
+  Status status() const override {
+    if (!status_.ok()) {
+      return status_;
+    } else if (file_iter_ && !file_iter_->status().ok()) {
+      return file_iter_->status();
+    }
+    return Status::OK();
+  }
+
+ private:
+  const ColumnFamilyData* const cfd_;
+  const ReadOptions& read_options_;
+  const std::vector<FileMetaData*>& files_;
+
+  bool valid_;
+  uint32_t file_index_;
+  Status status_;
+  std::unique_ptr<Iterator> file_iter_;
+};
+
+ForwardIterator::ForwardIterator(DBImpl* db, const ReadOptions& read_options,
+                                 ColumnFamilyData* cfd)
+    : db_(db),
+      read_options_(read_options),
+      cfd_(cfd),
+      prefix_extractor_(cfd->options()->prefix_extractor.get()),
+      user_comparator_(cfd->user_comparator()),
+      immutable_min_heap_(MinIterComparator(&cfd_->internal_comparator())),
+      sv_(nullptr),
+      mutable_iter_(nullptr),
+      current_(nullptr),
+      valid_(false),
+      is_prev_set_(false),
+      is_prev_inclusive_(false) {}
+
+ForwardIterator::~ForwardIterator() {
+  Cleanup();
+}
+
+void ForwardIterator::Cleanup() {
+  if (mutable_iter_ != nullptr) {
+    mutable_iter_->~Iterator();
+  }
+  for (auto* m : imm_iters_) {
+    m->~Iterator();
+  }
+  imm_iters_.clear();
+  for (auto* f : l0_iters_) {
+    delete f;
+  }
+  l0_iters_.clear();
+  for (auto* l : level_iters_) {
+    delete l;
+  }
+  level_iters_.clear();
+
+  if (sv_ != nullptr && sv_->Unref()) {
+    DBImpl::DeletionState deletion_state;
+    db_->mutex_.Lock();
+    sv_->Cleanup();
+    db_->FindObsoleteFiles(deletion_state, false, true);
+    db_->mutex_.Unlock();
+    delete sv_;
+    if (deletion_state.HaveSomethingToDelete()) {
+      db_->PurgeObsoleteFiles(deletion_state);
+    }
+  }
+}
+
+bool ForwardIterator::Valid() const {
+  return valid_;
+}
+
+void ForwardIterator::SeekToFirst() {
+  if (sv_ == nullptr ||
+      sv_ ->version_number != cfd_->GetSuperVersionNumber()) {
+    RebuildIterators();
+  } else if (status_.IsIncomplete()) {
+    ResetIncompleteIterators();
+  }
+  SeekInternal(Slice(), true);
+}
+
+void ForwardIterator::Seek(const Slice& internal_key) {
+  if (sv_ == nullptr ||
+      sv_ ->version_number != cfd_->GetSuperVersionNumber()) {
+    RebuildIterators();
+  } else if (status_.IsIncomplete()) {
+    ResetIncompleteIterators();
+  }
+  SeekInternal(internal_key, false);
+}
+
+void ForwardIterator::SeekInternal(const Slice& internal_key,
+                                   bool seek_to_first) {
+  // mutable
+  seek_to_first ? mutable_iter_->SeekToFirst() :
+                  mutable_iter_->Seek(internal_key);
+
+  // immutable
+  // TODO(ljin): NeedToSeekImmutable has negative impact on performance
+  // if it turns to need to seek immutable often. We probably want to have
+  // an option to turn it off.
+  if (seek_to_first || NeedToSeekImmutable(internal_key)) {
+    {
+      auto tmp = MinIterHeap(MinIterComparator(&cfd_->internal_comparator()));
+      immutable_min_heap_.swap(tmp);
+    }
+    for (auto* m : imm_iters_) {
+      seek_to_first ? m->SeekToFirst() : m->Seek(internal_key);
+      if (m->Valid()) {
+        immutable_min_heap_.push(m);
+      }
+    }
+
+    Slice user_key;
+    if (!seek_to_first) {
+      user_key = ExtractUserKey(internal_key);
+    }
+    auto* files = sv_->current->files_;
+    for (uint32_t i = 0; i < files[0].size(); ++i) {
+      if (seek_to_first) {
+        l0_iters_[i]->SeekToFirst();
+      } else {
+        // If the target key passes over the larget key, we are sure Next()
+        // won't go over this file.
+        if (user_comparator_->Compare(user_key,
+              files[0][i]->largest.user_key()) > 0) {
+          continue;
+        }
+        l0_iters_[i]->Seek(internal_key);
+      }
+
+      if (l0_iters_[i]->status().IsIncomplete()) {
+        // if any of the immutable iterators is incomplete (no-io option was
+        // used), we are unable to reliably find the smallest key
+        assert(read_options_.read_tier == kBlockCacheTier);
+        status_ = l0_iters_[i]->status();
+        valid_ = false;
+        return;
+      } else if (l0_iters_[i]->Valid()) {
+        immutable_min_heap_.push(l0_iters_[i]);
+      }
+    }
+
+    int32_t search_left_bound = 0;
+    int32_t search_right_bound = FileIndexer::kLevelMaxIndex;
+    for (int32_t level = 1; level < sv_->current->NumberLevels(); ++level) {
+      if (files[level].empty()) {
+        search_left_bound = 0;
+        search_right_bound = FileIndexer::kLevelMaxIndex;
+        continue;
+      }
+      assert(level_iters_[level - 1] != nullptr);
+      uint32_t f_idx = 0;
+      if (!seek_to_first) {
+        // TODO(ljin): remove before committing
+        // f_idx = FindFileInRange(
+        //    files[level], internal_key, 0, files[level].size());
+
+        if (search_left_bound == search_right_bound) {
+          f_idx = search_left_bound;
+        } else if (search_left_bound < search_right_bound) {
+          f_idx = FindFileInRange(
+              files[level], internal_key, search_left_bound,
+              search_right_bound == FileIndexer::kLevelMaxIndex ?
+                files[level].size() : search_right_bound);
+        } else {
+          // search_left_bound > search_right_bound
+          // There are only 2 cases this can happen:
+          // (1) target key is smaller than left most file
+          // (2) target key is larger than right most file
+          assert(search_left_bound == (int32_t)files[level].size() ||
+                 search_right_bound == -1);
+          if (search_right_bound == -1) {
+            assert(search_left_bound == 0);
+            f_idx = 0;
+          } else {
+            sv_->current->file_indexer_.GetNextLevelIndex(
+                level, files[level].size() - 1,
+                1, 1, &search_left_bound, &search_right_bound);
+            continue;
+          }
+        }
+
+        // Prepare hints for the next level
+        if (f_idx < files[level].size()) {
+          int cmp_smallest = user_comparator_->Compare(
+              user_key, files[level][f_idx]->smallest.user_key());
+          int cmp_largest = -1;
+          if (cmp_smallest >= 0) {
+            cmp_smallest = user_comparator_->Compare(
+                user_key, files[level][f_idx]->smallest.user_key());
+          }
+          sv_->current->file_indexer_.GetNextLevelIndex(level, f_idx,
+              cmp_smallest, cmp_largest,
+              &search_left_bound, &search_right_bound);
+        } else {
+          sv_->current->file_indexer_.GetNextLevelIndex(
+              level, files[level].size() - 1,
+              1, 1, &search_left_bound, &search_right_bound);
+        }
+      }
+
+      // Seek
+      if (f_idx < files[level].size()) {
+        level_iters_[level - 1]->SetFileIndex(f_idx);
+        seek_to_first ? level_iters_[level - 1]->SeekToFirst() :
+                        level_iters_[level - 1]->Seek(internal_key);
+
+        if (level_iters_[level - 1]->status().IsIncomplete()) {
+          // see above
+          assert(read_options_.read_tier == kBlockCacheTier);
+          status_ = level_iters_[level - 1]->status();
+          valid_ = false;
+          return;
+        } else if (level_iters_[level - 1]->Valid()) {
+          immutable_min_heap_.push(level_iters_[level - 1]);
+        }
+      }
+    }
+
+    if (seek_to_first) {
+      is_prev_set_ = false;
+    } else {
+      prev_key_.SetKey(internal_key);
+      is_prev_set_ = true;
+      is_prev_inclusive_ = true;
+    }
+  } else if (current_ && current_ != mutable_iter_) {
+    // current_ is one of immutable iterators, push it back to the heap
+    immutable_min_heap_.push(current_);
+  }
+
+  UpdateCurrent();
+}
+
+void ForwardIterator::Next() {
+  assert(valid_);
+
+  if (sv_ == nullptr ||
+      sv_->version_number != cfd_->GetSuperVersionNumber()) {
+    std::string current_key = key().ToString();
+    Slice old_key(current_key.data(), current_key.size());
+
+    RebuildIterators();
+    SeekInternal(old_key, false);
+    if (!valid_ || key().compare(old_key) != 0) {
+      return;
+    }
+  } else if (current_ != mutable_iter_) {
+    // It is going to advance immutable iterator
+
+    bool update_prev_key = true;
+    if (is_prev_set_ && prefix_extractor_) {
+      // advance prev_key_ to current_ only if they share the same prefix
+      update_prev_key =
+        prefix_extractor_->Transform(prev_key_.GetKey()).compare(
+          prefix_extractor_->Transform(current_->key())) == 0;
+    }
+
+    if (update_prev_key) {
+      prev_key_.SetKey(current_->key());
+      is_prev_set_ = true;
+      is_prev_inclusive_ = false;
+    }
+  }
+
+  current_->Next();
+  if (current_ != mutable_iter_) {
+    if (current_->status().IsIncomplete()) {
+      assert(read_options_.read_tier == kBlockCacheTier);
+      status_ = current_->status();
+      valid_ = false;
+      return;
+    } else if (current_->Valid()) {
+      immutable_min_heap_.push(current_);
+    }
+  }
+
+  UpdateCurrent();
+}
+
+Slice ForwardIterator::key() const {
+  assert(valid_);
+  return current_->key();
+}
+
+Slice ForwardIterator::value() const {
+  assert(valid_);
+  return current_->value();
+}
+
+Status ForwardIterator::status() const {
+  if (!status_.ok()) {
+    return status_;
+  } else if (!mutable_iter_->status().ok()) {
+    return mutable_iter_->status();
+  }
+
+  for (auto *it : imm_iters_) {
+    if (it && !it->status().ok()) {
+      return it->status();
+    }
+  }
+  for (auto *it : l0_iters_) {
+    if (it && !it->status().ok()) {
+      return it->status();
+    }
+  }
+  for (auto *it : level_iters_) {
+    if (it && !it->status().ok()) {
+      return it->status();
+    }
+  }
+
+  return Status::OK();
+}
+
+void ForwardIterator::RebuildIterators() {
+  // Clean up
+  Cleanup();
+  // New
+  sv_ = cfd_->GetReferencedSuperVersion(&(db_->mutex_));
+  mutable_iter_ = sv_->mem->NewIterator(read_options_, &arena_);
+  sv_->imm->AddIterators(read_options_, &imm_iters_, &arena_);
+  const auto& l0_files = sv_->current->files_[0];
+  l0_iters_.reserve(l0_files.size());
+  for (const auto* l0 : l0_files) {
+    l0_iters_.push_back(cfd_->table_cache()->NewIterator(
+        read_options_, *cfd_->soptions(), cfd_->internal_comparator(), l0->fd));
+  }
+  level_iters_.reserve(sv_->current->NumberLevels() - 1);
+  for (int32_t level = 1; level < sv_->current->NumberLevels(); ++level) {
+    if (sv_->current->files_[level].empty()) {
+      level_iters_.push_back(nullptr);
+    } else {
+      level_iters_.push_back(new LevelIterator(cfd_, read_options_,
+          sv_->current->files_[level]));
+    }
+  }
+
+  current_ = nullptr;
+  is_prev_set_ = false;
+}
+
+void ForwardIterator::ResetIncompleteIterators() {
+  const auto& l0_files = sv_->current->files_[0];
+  for (uint32_t i = 0; i < l0_iters_.size(); ++i) {
+    assert(i < l0_files.size());
+    if (!l0_iters_[i]->status().IsIncomplete()) {
+      continue;
+    }
+    delete l0_iters_[i];
+    l0_iters_[i] = cfd_->table_cache()->NewIterator(
+        read_options_, *cfd_->soptions(), cfd_->internal_comparator(),
+        l0_files[i]->fd);
+  }
+
+  for (auto* level_iter : level_iters_) {
+    if (level_iter && level_iter->status().IsIncomplete()) {
+      level_iter->Reset();
+    }
+  }
+
+  current_ = nullptr;
+  is_prev_set_ = false;
+}
+
+void ForwardIterator::UpdateCurrent() {
+  if (immutable_min_heap_.empty() && !mutable_iter_->Valid()) {
+    current_ = nullptr;
+  } else if (immutable_min_heap_.empty()) {
+    current_ = mutable_iter_;
+  } else if (!mutable_iter_->Valid()) {
+    current_ = immutable_min_heap_.top();
+    immutable_min_heap_.pop();
+  } else {
+    current_ = immutable_min_heap_.top();
+    assert(current_ != nullptr);
+    assert(current_->Valid());
+    int cmp = cfd_->internal_comparator().InternalKeyComparator::Compare(
+        mutable_iter_->key(), current_->key());
+    assert(cmp != 0);
+    if (cmp > 0) {
+      immutable_min_heap_.pop();
+    } else {
+      current_ = mutable_iter_;
+    }
+  }
+  valid_ = (current_ != nullptr);
+  if (!status_.ok()) {
+    status_ = Status::OK();
+  }
+}
+
+bool ForwardIterator::NeedToSeekImmutable(const Slice& target) {
+  // We maintain the interval (prev_key_, immutable_min_heap_.top()->key())
+  // such that there are no records with keys within that range in
+  // immutable_min_heap_. Since immutable structures (SST files and immutable
+  // memtables) can't change in this version, we don't need to do a seek if
+  // 'target' belongs to that interval (immutable_min_heap_.top() is already
+  // at the correct position).
+
+  if (!valid_ || !current_ || !is_prev_set_) {
+    return true;
+  }
+  Slice prev_key = prev_key_.GetKey();
+  if (prefix_extractor_ && prefix_extractor_->Transform(target).compare(
+    prefix_extractor_->Transform(prev_key)) != 0) {
+    return true;
+  }
+  if (cfd_->internal_comparator().InternalKeyComparator::Compare(
+        prev_key, target) >= (is_prev_inclusive_ ? 1 : 0)) {
+    return true;
+  }
+
+  if (immutable_min_heap_.empty() && current_ == mutable_iter_) {
+    // Nothing to seek on.
+    return false;
+  }
+  if (cfd_->internal_comparator().InternalKeyComparator::Compare(
+        target, current_ == mutable_iter_ ? immutable_min_heap_.top()->key()
+                                          : current_->key()) > 0) {
+    return true;
+  }
+  return false;
+}
+
+uint32_t ForwardIterator::FindFileInRange(
+    const std::vector<FileMetaData*>& files, const Slice& internal_key,
+    uint32_t left, uint32_t right) {
+  while (left < right) {
+    uint32_t mid = (left + right) / 2;
+    const FileMetaData* f = files[mid];
+    if (cfd_->internal_comparator().InternalKeyComparator::Compare(
+          f->largest.Encode(), internal_key) < 0) {
+      // Key at "mid.largest" is < "target".  Therefore all
+      // files at or before "mid" are uninteresting.
+      left = mid + 1;
+    } else {
+      // Key at "mid.largest" is >= "target".  Therefore all files
+      // after "mid" are uninteresting.
+      right = mid;
+    }
+  }
+  return right;
+}
+
+}  // namespace rocksdb
+
+#endif  // ROCKSDB_LITE
@@ -0,0 +1,109 @@
+//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+#pragma once
+
+#ifndef ROCKSDB_LITE
+
+#include <string>
+#include <vector>
+#include <queue>
+
+#include "rocksdb/db.h"
+#include "rocksdb/iterator.h"
+#include "rocksdb/options.h"
+#include "db/dbformat.h"
+#include "util/arena.h"
+
+namespace rocksdb {
+
+class DBImpl;
+class Env;
+struct SuperVersion;
+class ColumnFamilyData;
+class LevelIterator;
+struct FileMetaData;
+
+class MinIterComparator {
+ public:
+  explicit MinIterComparator(const Comparator* comparator) :
+    comparator_(comparator) {}
+
+  bool operator()(Iterator* a, Iterator* b) {
+    return comparator_->Compare(a->key(), b->key()) > 0;
+  }
+ private:
+  const Comparator* comparator_;
+};
+
+typedef std::priority_queue<Iterator*,
+          std::vector<Iterator*>,
+          MinIterComparator> MinIterHeap;
+
+/**
+ * ForwardIterator is a special type of iterator that only supports Seek()
+ * and Next(). It is expected to perform better than TailingIterator by
+ * removing the encapsulation and making all information accessible within
+ * the iterator. At the current implementation, snapshot is taken at the
+ * time Seek() is called. The Next() followed do not see new values after.
+ */
+class ForwardIterator : public Iterator {
+ public:
+  ForwardIterator(DBImpl* db, const ReadOptions& read_options,
+                  ColumnFamilyData* cfd);
+  virtual ~ForwardIterator();
+
+  void SeekToLast() override {
+    status_ = Status::NotSupported("ForwardIterator::SeekToLast()");
+    valid_ = false;
+  }
+  void Prev() {
+    status_ = Status::NotSupported("ForwardIterator::Prev");
+    valid_ = false;
+  }
+
+  virtual bool Valid() const override;
+  void SeekToFirst() override;
+  virtual void Seek(const Slice& target) override;
+  virtual void Next() override;
+  virtual Slice key() const override;
+  virtual Slice value() const override;
+  virtual Status status() const override;
+
+ private:
+  void Cleanup();
+  void RebuildIterators();
+  void ResetIncompleteIterators();
+  void SeekInternal(const Slice& internal_key, bool seek_to_first);
+  void UpdateCurrent();
+  bool NeedToSeekImmutable(const Slice& internal_key);
+  uint32_t FindFileInRange(
+    const std::vector<FileMetaData*>& files, const Slice& internal_key,
+    uint32_t left, uint32_t right);
+
+  DBImpl* const db_;
+  const ReadOptions read_options_;
+  ColumnFamilyData* const cfd_;
+  const SliceTransform* const prefix_extractor_;
+  const Comparator* user_comparator_;
+  MinIterHeap immutable_min_heap_;
+
+  SuperVersion* sv_;
+  Iterator* mutable_iter_;
+  std::vector<Iterator*> imm_iters_;
+  std::vector<Iterator*> l0_iters_;
+  std::vector<LevelIterator*> level_iters_;
+  Iterator* current_;
+  // internal iterator status
+  Status status_;
+  bool valid_;
+
+  IterKey prev_key_;
+  bool is_prev_set_;
+  bool is_prev_inclusive_;
+  Arena arena_;
+};
+
+}  // namespace rocksdb
+#endif  // ROCKSDB_LITE
@@ -7,16 +7,105 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.

 #include "db/internal_stats.h"
-#include "db/column_family.h"

+#ifndef __STDC_FORMAT_MACROS
+#define __STDC_FORMAT_MACROS
+#endif
+
+#include <inttypes.h>
 #include <vector>
+#include "db/column_family.h"
+#include "db/db_impl.h"

 namespace rocksdb {

-DBPropertyType GetPropertyType(const Slice& property) {
+namespace {
+const double kMB = 1048576.0;
+const double kGB = kMB * 1024;
+
+void PrintLevelStatsHeader(char* buf, size_t len, const std::string& cf_name) {
+  snprintf(
+      buf, len,
+      "\n** Compaction Stats [%s] **\n"
+      "Level   Files   Size(MB) Score Read(GB)  Rn(GB) Rnp1(GB) "
+      "Write(GB) Wnew(GB) RW-Amp W-Amp Rd(MB/s) Wr(MB/s)  Rn(cnt) "
+      "Rnp1(cnt) Wnp1(cnt) Wnew(cnt)  Comp(sec) Comp(cnt) Avg(sec) "
+      "Stall(sec) Stall(cnt) Avg(ms) RecordIn RecordDrop\n"
+      "--------------------------------------------------------------------"
+      "--------------------------------------------------------------------"
+      "--------------------------------------------------------------------\n",
+      cf_name.c_str());
+}
+
+void PrintLevelStats(char* buf, size_t len, const std::string& name,
+    int num_files, int being_compacted, double total_file_size, double score,
+    double rw_amp, double w_amp, double stall_us, uint64_t stalls,
+    const InternalStats::CompactionStats& stats) {
+  uint64_t bytes_read = stats.bytes_readn + stats.bytes_readnp1;
+  uint64_t bytes_new = stats.bytes_written - stats.bytes_readnp1;
+  double elapsed = (stats.micros + 1) / 1000000.0;
+
+  snprintf(buf, len,
+    "%4s %5d/%-3d %8.0f %5.1f " /* Level, Files, Size(MB), Score */
+    "%8.1f " /* Read(GB) */
+    "%7.1f " /* Rn(GB) */
+    "%8.1f " /* Rnp1(GB) */
+    "%9.1f " /* Write(GB) */
+    "%8.1f " /* Wnew(GB) */
+    "%6.1f " /* RW-Amp */
+    "%5.1f " /* W-Amp */
+    "%8.1f " /* Rd(MB/s) */
+    "%8.1f " /* Wr(MB/s) */
+    "%8d " /* Rn(cnt) */
+    "%9d " /* Rnp1(cnt) */
+    "%9d " /* Wnp1(cnt) */
+    "%9d " /* Wnew(cnt) */
+    "%10.0f " /* Comp(sec) */
+    "%9d " /* Comp(cnt) */
+    "%8.3f " /* Avg(sec) */
+    "%10.2f " /* Stall(sec) */
+    "%10" PRIu64 " " /* Stall(cnt) */
+    "%7.2f" /* Avg(ms) */
+    "%8d " /* input entries */
+    "%10d\n" /* number of records reduced */,
+    name.c_str(), num_files, being_compacted, total_file_size / kMB, score,
+    bytes_read / kGB,
+    stats.bytes_readn / kGB,
+    stats.bytes_readnp1 / kGB,
+    stats.bytes_written / kGB,
+    bytes_new / kGB,
+    rw_amp,
+    w_amp,
+    bytes_read / kMB / elapsed,
+    stats.bytes_written / kMB / elapsed,
+    stats.files_in_leveln,
+    stats.files_in_levelnp1,
+    stats.files_out_levelnp1,
+    stats.files_out_levelnp1 - stats.files_in_levelnp1,
+    stats.micros / 1000000.0,
+    stats.count,
+    stats.count == 0 ? 0 : stats.micros / 1000000.0 / stats.count,
+    stall_us / 1000000.0,
+    stalls,
+    stalls == 0 ? 0 : stall_us / 1000.0 / stalls,
+    stats.num_input_records,
+    stats.num_dropped_records);
+}
+
+
+}
+
+DBPropertyType GetPropertyType(const Slice& property, bool* is_int_property,
+                               bool* need_out_of_mutex) {
+  assert(is_int_property != nullptr);
+  assert(need_out_of_mutex != nullptr);
  Slice in = property;
  Slice prefix("rocksdb.");
-  if (!in.starts_with(prefix)) return kUnknown;
+  *need_out_of_mutex = false;
+  *is_int_property = false;
+  if (!in.starts_with(prefix)) {
+    return kUnknown;
+  }
  in.remove_prefix(prefix.size());

  if (in.starts_with("num-files-at-level")) {
@@ -25,9 +114,16 @@ DBPropertyType GetPropertyType(const Slice& property) {
    return kLevelStats;
  } else if (in == "stats") {
    return kStats;
+  } else if (in == "cfstats") {
+    return kCFStats;
+  } else if (in == "dbstats") {
+    return kDBStats;
  } else if (in == "sstables") {
    return kSsTables;
-  } else if (in == "num-immutable-mem-table") {
+  }
+
+  *is_int_property = true;
+  if (in == "num-immutable-mem-table") {
    return kNumImmutableMemTable;
  } else if (in == "mem-table-flush-pending") {
    return kMemtableFlushPending;
@@ -41,14 +137,37 @@ DBPropertyType GetPropertyType(const Slice& property) {
    return kNumEntriesInMutableMemtable;
  } else if (in == "num-entries-imm-mem-tables") {
    return kNumEntriesInImmutableMemtable;
+  } else if (in == "estimate-num-keys") {
+    return kEstimatedNumKeys;
+  } else if (in == "estimate-table-readers-mem") {
+    *need_out_of_mutex = true;
+    return kEstimatedUsageByTableReaders;
+  } else if (in == "is-file-deletions-enabled") {
+    return kIsFileDeletionEnabled;
  }
  return kUnknown;
 }

-bool InternalStats::GetProperty(DBPropertyType property_type,
-                                const Slice& property, std::string* value,
-                                ColumnFamilyData* cfd) {
-  Version* current = cfd->current();
+bool InternalStats::GetIntPropertyOutOfMutex(DBPropertyType property_type,
+                                             Version* version,
+                                             uint64_t* value) const {
+  assert(value != nullptr);
+  if (property_type != kEstimatedUsageByTableReaders) {
+    return false;
+  }
+  if (version == nullptr) {
+    *value = 0;
+  } else {
+    *value = version->GetMemoryUsageByTableReaders();
+  }
+  return true;
+}
+
+bool InternalStats::GetStringProperty(DBPropertyType property_type,
+                                      const Slice& property,
+                                      std::string* value) {
+  assert(value != nullptr);
+  Version* current = cfd_->current();
  Slice in = property;

  switch (property_type) {
@@ -76,294 +195,292 @@ bool InternalStats::GetProperty(DBPropertyType property_type,
      for (int level = 0; level < number_levels_; level++) {
        snprintf(buf, sizeof(buf), "%3d %8d %8.0f\n", level,
                 current->NumLevelFiles(level),
-                 current->NumLevelBytes(level) / 1048576.0);
+                 current->NumLevelBytes(level) / kMB);
        value->append(buf);
      }
      return true;
    }
    case kStats: {
-      char buf[1000];
-
-      uint64_t wal_bytes = 0;
-      uint64_t wal_synced = 0;
-      uint64_t user_bytes_written = 0;
-      uint64_t write_other = 0;
-      uint64_t write_self = 0;
-      uint64_t write_with_wal = 0;
-      uint64_t total_bytes_written = 0;
-      uint64_t total_bytes_read = 0;
-      uint64_t micros_up = env_->NowMicros() - started_at_;
-      // Add "+1" to make sure seconds_up is > 0 and avoid NaN later
-      double seconds_up = (micros_up + 1) / 1000000.0;
-      uint64_t total_slowdown = 0;
-      uint64_t total_slowdown_count = 0;
-      uint64_t interval_bytes_written = 0;
-      uint64_t interval_bytes_read = 0;
-      uint64_t interval_bytes_new = 0;
-      double interval_seconds_up = 0;
-
-      if (statistics_) {
-        wal_bytes = statistics_->getTickerCount(WAL_FILE_BYTES);
-        wal_synced = statistics_->getTickerCount(WAL_FILE_SYNCED);
-        user_bytes_written = statistics_->getTickerCount(BYTES_WRITTEN);
-        write_other = statistics_->getTickerCount(WRITE_DONE_BY_OTHER);
-        write_self = statistics_->getTickerCount(WRITE_DONE_BY_SELF);
-        write_with_wal = statistics_->getTickerCount(WRITE_WITH_WAL);
+      if (!GetStringProperty(kCFStats, "rocksdb.cfstats", value)) {
+        return false;
      }
-
-      snprintf(
-          buf, sizeof(buf),
-          "                               Compactions\n"
-          "Level  Files Size(MB) Score Time(sec)  Read(MB) Write(MB)    Rn(MB) "
-          " "
-          "Rnp1(MB)  Wnew(MB) RW-Amplify Read(MB/s) Write(MB/s)      Rn     "
-          "Rnp1 "
-          "    Wnp1     NewW    Count   msComp   msStall  Ln-stall Stall-cnt\n"
-          "--------------------------------------------------------------------"
-          "--"
-          "--------------------------------------------------------------------"
-          "--"
-          "----------------------------------------------------------------\n");
-      value->append(buf);
-      for (int level = 0; level < number_levels_; level++) {
-        int files = current->NumLevelFiles(level);
-        if (compaction_stats_[level].micros > 0 || files > 0) {
-          int64_t bytes_read = compaction_stats_[level].bytes_readn +
-                               compaction_stats_[level].bytes_readnp1;
-          int64_t bytes_new = compaction_stats_[level].bytes_written -
-                              compaction_stats_[level].bytes_readnp1;
-          double amplify =
-              (compaction_stats_[level].bytes_readn == 0)
-                  ? 0.0
-                  : (compaction_stats_[level].bytes_written +
-                     compaction_stats_[level].bytes_readnp1 +
-                     compaction_stats_[level].bytes_readn) /
-                        (double)compaction_stats_[level].bytes_readn;
-
-          total_bytes_read += bytes_read;
-          total_bytes_written += compaction_stats_[level].bytes_written;
-
-          uint64_t stalls = level == 0 ? (stall_counts_[LEVEL0_SLOWDOWN] +
-                                          stall_counts_[LEVEL0_NUM_FILES] +
-                                          stall_counts_[MEMTABLE_COMPACTION])
-                                       : stall_leveln_slowdown_count_[level];
-
-          double stall_us = level == 0 ? (stall_micros_[LEVEL0_SLOWDOWN] +
-                                          stall_micros_[LEVEL0_NUM_FILES] +
-                                          stall_micros_[MEMTABLE_COMPACTION])
-                                       : stall_leveln_slowdown_[level];
-
-          snprintf(buf, sizeof(buf),
-                   "%3d %8d %8.0f %5.1f %9.0f %9.0f %9.0f %9.0f %9.0f %9.0f "
-                   "%10.1f %9.1f %11.1f %8d %8d %8d %8d %8d %8d %9.1f %9.1f "
-                   "%9lu\n",
-                   level, files, current->NumLevelBytes(level) / 1048576.0,
-                   current->NumLevelBytes(level) /
-                       cfd->compaction_picker()->MaxBytesForLevel(level),
-                   compaction_stats_[level].micros / 1e6,
-                   bytes_read / 1048576.0,
-                   compaction_stats_[level].bytes_written / 1048576.0,
-                   compaction_stats_[level].bytes_readn / 1048576.0,
-                   compaction_stats_[level].bytes_readnp1 / 1048576.0,
-                   bytes_new / 1048576.0, amplify,
-                   // +1 to avoid division by 0
-                   (bytes_read / 1048576.0) /
-                       ((compaction_stats_[level].micros + 1) / 1000000.0),
-                   (compaction_stats_[level].bytes_written / 1048576.0) /
-                       ((compaction_stats_[level].micros + 1) / 1000000.0),
-                   compaction_stats_[level].files_in_leveln,
-                   compaction_stats_[level].files_in_levelnp1,
-                   compaction_stats_[level].files_out_levelnp1,
-                   compaction_stats_[level].files_out_levelnp1 -
-                       compaction_stats_[level].files_in_levelnp1,
-                   compaction_stats_[level].count,
-                   (int)((double)compaction_stats_[level].micros / 1000.0 /
-                         (compaction_stats_[level].count + 1)),
-                   (double)stall_us / 1000.0 / (stalls + 1),
-                   stall_us / 1000000.0, (unsigned long)stalls);
-          total_slowdown += stall_leveln_slowdown_[level];
-          total_slowdown_count += stall_leveln_slowdown_count_[level];
-          value->append(buf);
-        }
+      if (!GetStringProperty(kDBStats, "rocksdb.dbstats", value)) {
+        return false;
      }
-
-      interval_bytes_new = user_bytes_written - last_stats_.ingest_bytes_;
-      interval_bytes_read =
-          total_bytes_read - last_stats_.compaction_bytes_read_;
-      interval_bytes_written =
-          total_bytes_written - last_stats_.compaction_bytes_written_;
-      interval_seconds_up = seconds_up - last_stats_.seconds_up_;
-
-      snprintf(buf, sizeof(buf), "Uptime(secs): %.1f total, %.1f interval\n",
-               seconds_up, interval_seconds_up);
-      value->append(buf);
-
-      snprintf(buf, sizeof(buf),
-               "Writes cumulative: %llu total, %llu batches, "
-               "%.1f per batch, %.2f ingest GB\n",
-               (unsigned long long)(write_other + write_self),
-               (unsigned long long)write_self,
-               (write_other + write_self) / (double)(write_self + 1),
-               user_bytes_written / (1048576.0 * 1024));
-      value->append(buf);
-
-      snprintf(buf, sizeof(buf),
-               "WAL cumulative: %llu WAL writes, %llu WAL syncs, "
-               "%.2f writes per sync, %.2f GB written\n",
-               (unsigned long long)write_with_wal,
-               (unsigned long long)wal_synced,
-               write_with_wal / (double)(wal_synced + 1),
-               wal_bytes / (1048576.0 * 1024));
-      value->append(buf);
-
-      snprintf(buf, sizeof(buf),
-               "Compaction IO cumulative (GB): "
-               "%.2f new, %.2f read, %.2f write, %.2f read+write\n",
-               user_bytes_written / (1048576.0 * 1024),
-               total_bytes_read / (1048576.0 * 1024),
-               total_bytes_written / (1048576.0 * 1024),
-               (total_bytes_read + total_bytes_written) / (1048576.0 * 1024));
-      value->append(buf);
-
-      snprintf(
-          buf, sizeof(buf),
-          "Compaction IO cumulative (MB/sec): "
-          "%.1f new, %.1f read, %.1f write, %.1f read+write\n",
-          user_bytes_written / 1048576.0 / seconds_up,
-          total_bytes_read / 1048576.0 / seconds_up,
-          total_bytes_written / 1048576.0 / seconds_up,
-          (total_bytes_read + total_bytes_written) / 1048576.0 / seconds_up);
-      value->append(buf);
-
-      // +1 to avoid divide by 0 and NaN
-      snprintf(
-          buf, sizeof(buf),
-          "Amplification cumulative: %.1f write, %.1f compaction\n",
-          (double)(total_bytes_written + wal_bytes) / (user_bytes_written + 1),
-          (double)(total_bytes_written + total_bytes_read + wal_bytes) /
-              (user_bytes_written + 1));
-      value->append(buf);
-
-      uint64_t interval_write_other = write_other - last_stats_.write_other_;
-      uint64_t interval_write_self = write_self - last_stats_.write_self_;
-
-      snprintf(buf, sizeof(buf),
-               "Writes interval: %llu total, %llu batches, "
-               "%.1f per batch, %.1f ingest MB\n",
-               (unsigned long long)(interval_write_other + interval_write_self),
-               (unsigned long long)interval_write_self,
-               (double)(interval_write_other + interval_write_self) /
-                   (interval_write_self + 1),
-               (user_bytes_written - last_stats_.ingest_bytes_) / 1048576.0);
-      value->append(buf);
-
-      uint64_t interval_write_with_wal =
-          write_with_wal - last_stats_.write_with_wal_;
-
-      uint64_t interval_wal_synced = wal_synced - last_stats_.wal_synced_;
-      uint64_t interval_wal_bytes = wal_bytes - last_stats_.wal_bytes_;
-
-      snprintf(buf, sizeof(buf),
-               "WAL interval: %llu WAL writes, %llu WAL syncs, "
-               "%.2f writes per sync, %.2f MB written\n",
-               (unsigned long long)interval_write_with_wal,
-               (unsigned long long)interval_wal_synced,
-               interval_write_with_wal / (double)(interval_wal_synced + 1),
-               interval_wal_bytes / (1048576.0 * 1024));
-      value->append(buf);
-
-      snprintf(buf, sizeof(buf),
-               "Compaction IO interval (MB): "
-               "%.2f new, %.2f read, %.2f write, %.2f read+write\n",
-               interval_bytes_new / 1048576.0, interval_bytes_read / 1048576.0,
-               interval_bytes_written / 1048576.0,
-               (interval_bytes_read + interval_bytes_written) / 1048576.0);
-      value->append(buf);
-
-      snprintf(buf, sizeof(buf),
-               "Compaction IO interval (MB/sec): "
-               "%.1f new, %.1f read, %.1f write, %.1f read+write\n",
-               interval_bytes_new / 1048576.0 / interval_seconds_up,
-               interval_bytes_read / 1048576.0 / interval_seconds_up,
-               interval_bytes_written / 1048576.0 / interval_seconds_up,
-               (interval_bytes_read + interval_bytes_written) / 1048576.0 /
-                   interval_seconds_up);
-      value->append(buf);
-
-      // +1 to avoid divide by 0 and NaN
-      snprintf(
-          buf, sizeof(buf),
-          "Amplification interval: %.1f write, %.1f compaction\n",
-          (double)(interval_bytes_written + wal_bytes) /
-              (interval_bytes_new + 1),
-          (double)(interval_bytes_written + interval_bytes_read + wal_bytes) /
-              (interval_bytes_new + 1));
-      value->append(buf);
-
-      snprintf(buf, sizeof(buf),
-               "Stalls(secs): %.3f level0_slowdown, %.3f level0_numfiles, "
-               "%.3f memtable_compaction, %.3f leveln_slowdown\n",
-               stall_micros_[LEVEL0_SLOWDOWN] / 1000000.0,
-               stall_micros_[LEVEL0_NUM_FILES] / 1000000.0,
-               stall_micros_[MEMTABLE_COMPACTION] / 1000000.0,
-               total_slowdown / 1000000.0);
-      value->append(buf);
-
-      snprintf(buf, sizeof(buf),
-               "Stalls(count): %lu level0_slowdown, %lu level0_numfiles, "
-               "%lu memtable_compaction, %lu leveln_slowdown\n",
-               (unsigned long)stall_counts_[LEVEL0_SLOWDOWN],
-               (unsigned long)stall_counts_[LEVEL0_NUM_FILES],
-               (unsigned long)stall_counts_[MEMTABLE_COMPACTION],
-               (unsigned long)total_slowdown_count);
-      value->append(buf);
-
-      last_stats_.compaction_bytes_read_ = total_bytes_read;
-      last_stats_.compaction_bytes_written_ = total_bytes_written;
-      last_stats_.ingest_bytes_ = user_bytes_written;
-      last_stats_.seconds_up_ = seconds_up;
-      last_stats_.wal_bytes_ = wal_bytes;
-      last_stats_.wal_synced_ = wal_synced;
-      last_stats_.write_with_wal_ = write_with_wal;
-      last_stats_.write_other_ = write_other;
-      last_stats_.write_self_ = write_self;
-
+      return true;
+    }
+    case kCFStats: {
+      DumpCFStats(value);
+      return true;
+    }
+    case kDBStats: {
+      DumpDBStats(value);
      return true;
    }
    case kSsTables:
      *value = current->DebugString();
      return true;
-    case kNumImmutableMemTable:
-      *value = std::to_string(cfd->imm()->size());
-      return true;
-    case kMemtableFlushPending:
-      // Return number of mem tables that are ready to flush (made immutable)
-      *value = std::to_string(cfd->imm()->IsFlushPending() ? 1 : 0);
-      return true;
-    case kCompactionPending:
-      // 1 if the system already determines at least one compacdtion is needed.
-      // 0 otherwise,
-      *value = std::to_string(current->NeedsCompaction() ? 1 : 0);
-      return true;
-    case kBackgroundErrors:
-      // Accumulated number of  errors in background flushes or compactions.
-      *value = std::to_string(GetBackgroundErrorCount());
-      return true;
-    case kCurSizeActiveMemTable:
-      // Current size of the active memtable
-      *value = std::to_string(cfd->mem()->ApproximateMemoryUsage());
-      return true;
-    case kNumEntriesInMutableMemtable:
-      // Current size of the active memtable
-      *value = std::to_string(cfd->mem()->GetNumEntries());
-      return true;
-    case kNumEntriesInImmutableMemtable:
-      // Current size of the active memtable
-      *value = std::to_string(cfd->imm()->current()->GetTotalNumEntries());
-      return true;
    default:
      return false;
  }
 }

+bool InternalStats::GetIntProperty(DBPropertyType property_type,
+                                   uint64_t* value, DBImpl* db) const {
+  Version* current = cfd_->current();
+
+  switch (property_type) {
+    case kNumImmutableMemTable:
+      *value = cfd_->imm()->size();
+      return true;
+    case kMemtableFlushPending:
+      // Return number of mem tables that are ready to flush (made immutable)
+      *value = (cfd_->imm()->IsFlushPending() ? 1 : 0);
+      return true;
+    case kCompactionPending:
+      // 1 if the system already determines at least one compacdtion is needed.
+      // 0 otherwise,
+      *value = (current->NeedsCompaction() ? 1 : 0);
+      return true;
+    case kBackgroundErrors:
+      // Accumulated number of  errors in background flushes or compactions.
+      *value = GetBackgroundErrorCount();
+      return true;
+    case kCurSizeActiveMemTable:
+      // Current size of the active memtable
+      *value = cfd_->mem()->ApproximateMemoryUsage();
+      return true;
+    case kNumEntriesInMutableMemtable:
+      // Current size of the active memtable
+      *value = cfd_->mem()->GetNumEntries();
+      return true;
+    case kNumEntriesInImmutableMemtable:
+      // Current size of the active memtable
+      *value = cfd_->imm()->current()->GetTotalNumEntries();
+      return true;
+    case kEstimatedNumKeys:
+      // Estimate number of entries in the column family:
+      // Use estimated entries in tables + total entries in memtables.
+      *value = cfd_->mem()->GetNumEntries() +
+               cfd_->imm()->current()->GetTotalNumEntries() +
+               current->GetEstimatedActiveKeys();
+      return true;
+#ifndef ROCKSDB_LITE
+    case kIsFileDeletionEnabled:
+      *value = db->IsFileDeletionsEnabled();
+      return true;
+#endif
+    default:
+      return false;
+  }
+}
+
+void InternalStats::DumpDBStats(std::string* value) {
+  char buf[1000];
+  // DB-level stats, only available from default column family
+  double seconds_up = (env_->NowMicros() - started_at_ + 1) / 1000000.0;
+  double interval_seconds_up = seconds_up - db_stats_snapshot_.seconds_up;
+  snprintf(buf, sizeof(buf),
+           "\n** DB Stats **\nUptime(secs): %.1f total, %.1f interval\n",
+           seconds_up, interval_seconds_up);
+  value->append(buf);
+  // Cumulative
+  uint64_t user_bytes_written = db_stats_[InternalStats::BYTES_WRITTEN];
+  uint64_t write_other = db_stats_[InternalStats::WRITE_DONE_BY_OTHER];
+  uint64_t write_self = db_stats_[InternalStats::WRITE_DONE_BY_SELF];
+  uint64_t wal_bytes = db_stats_[InternalStats::WAL_FILE_BYTES];
+  uint64_t wal_synced = db_stats_[InternalStats::WAL_FILE_SYNCED];
+  uint64_t write_with_wal = db_stats_[InternalStats::WRITE_WITH_WAL];
+  // Data
+  snprintf(buf, sizeof(buf),
+           "Cumulative writes: %" PRIu64 " writes, %" PRIu64 " batches, "
+           "%.1f writes per batch, %.2f GB user ingest\n",
+           write_other + write_self, write_self,
+           (write_other + write_self) / static_cast<double>(write_self + 1),
+           user_bytes_written / kGB);
+  value->append(buf);
+  // WAL
+  snprintf(buf, sizeof(buf),
+           "Cumulative WAL: %" PRIu64 " writes, %" PRIu64 " syncs, "
+           "%.2f writes per sync, %.2f GB written\n",
+           write_with_wal, wal_synced,
+           write_with_wal / static_cast<double>(wal_synced + 1),
+           wal_bytes / kGB);
+  value->append(buf);
+
+  // Interval
+  uint64_t interval_write_other = write_other - db_stats_snapshot_.write_other;
+  uint64_t interval_write_self = write_self - db_stats_snapshot_.write_self;
+  snprintf(buf, sizeof(buf),
+           "Interval writes: %" PRIu64 " writes, %" PRIu64 " batches, "
+           "%.1f writes per batch, %.1f MB user ingest\n",
+           interval_write_other + interval_write_self,
+           interval_write_self,
+           static_cast<double>(interval_write_other + interval_write_self) /
+               (interval_write_self + 1),
+           (user_bytes_written - db_stats_snapshot_.ingest_bytes) / kMB);
+  value->append(buf);
+
+  uint64_t interval_write_with_wal =
+      write_with_wal - db_stats_snapshot_.write_with_wal;
+  uint64_t interval_wal_synced = wal_synced - db_stats_snapshot_.wal_synced;
+  uint64_t interval_wal_bytes = wal_bytes - db_stats_snapshot_.wal_bytes;
+
+  snprintf(buf, sizeof(buf),
+           "Interval WAL: %" PRIu64 " writes, %" PRIu64 " syncs, "
+           "%.2f writes per sync, %.2f MB written\n",
+           interval_write_with_wal,
+           interval_wal_synced,
+           interval_write_with_wal /
+              static_cast<double>(interval_wal_synced + 1),
+           interval_wal_bytes / kGB);
+  value->append(buf);
+
+  db_stats_snapshot_.seconds_up = seconds_up;
+  db_stats_snapshot_.ingest_bytes = user_bytes_written;
+  db_stats_snapshot_.write_other = write_other;
+  db_stats_snapshot_.write_self = write_self;
+  db_stats_snapshot_.wal_bytes = wal_bytes;
+  db_stats_snapshot_.wal_synced = wal_synced;
+  db_stats_snapshot_.write_with_wal = write_with_wal;
+}
+
+void InternalStats::DumpCFStats(std::string* value) {
+  Version* current = cfd_->current();
+
+  int num_levels_to_check =
+      (cfd_->options()->compaction_style != kCompactionStyleUniversal &&
+       cfd_->options()->compaction_style != kCompactionStyleFIFO)
+          ? current->NumberLevels() - 1
+          : 1;
+  // Compaction scores are sorted base on its value. Restore them to the
+  // level order
+  std::vector<double> compaction_score(number_levels_, 0);
+  for (int i = 0; i < num_levels_to_check; ++i) {
+    compaction_score[current->compaction_level_[i]] =
+      current->compaction_score_[i];
+  }
+  // Count # of files being compacted for each level
+  std::vector<int> files_being_compacted(number_levels_, 0);
+  for (int level = 0; level < num_levels_to_check; ++level) {
+    for (auto* f : current->files_[level]) {
+      if (f->being_compacted) {
+        ++files_being_compacted[level];
+      }
+    }
+  }
+
+  char buf[1000];
+  // Per-ColumnFamily stats
+  PrintLevelStatsHeader(buf, sizeof(buf), cfd_->GetName());
+  value->append(buf);
+
+  CompactionStats stats_sum(0);
+  int total_files = 0;
+  int total_files_being_compacted = 0;
+  double total_file_size = 0;
+  uint64_t total_slowdown_soft = 0;
+  uint64_t total_slowdown_count_soft = 0;
+  uint64_t total_slowdown_hard = 0;
+  uint64_t total_slowdown_count_hard = 0;
+  uint64_t total_stall_count = 0;
+  double total_stall_us = 0;
+  for (int level = 0; level < number_levels_; level++) {
+    int files = current->NumLevelFiles(level);
+    total_files += files;
+    total_files_being_compacted += files_being_compacted[level];
+    if (comp_stats_[level].micros > 0 || files > 0) {
+      uint64_t stalls = level == 0 ?
+        (cf_stats_count_[LEVEL0_SLOWDOWN] +
+         cf_stats_count_[LEVEL0_NUM_FILES] +
+         cf_stats_count_[MEMTABLE_COMPACTION])
+        : (stall_leveln_slowdown_count_soft_[level] +
+           stall_leveln_slowdown_count_hard_[level]);
+
+      double stall_us = level == 0 ?
+         (cf_stats_value_[LEVEL0_SLOWDOWN] +
+          cf_stats_value_[LEVEL0_NUM_FILES] +
+          cf_stats_value_[MEMTABLE_COMPACTION])
+         : (stall_leveln_slowdown_soft_[level] +
+            stall_leveln_slowdown_hard_[level]);
+
+      stats_sum.Add(comp_stats_[level]);
+      total_file_size += current->NumLevelBytes(level);
+      total_stall_us += stall_us;
+      total_stall_count += stalls;
+      total_slowdown_soft += stall_leveln_slowdown_soft_[level];
+      total_slowdown_count_soft += stall_leveln_slowdown_count_soft_[level];
+      total_slowdown_hard += stall_leveln_slowdown_hard_[level];
+      total_slowdown_count_hard += stall_leveln_slowdown_count_hard_[level];
+      int64_t bytes_read = comp_stats_[level].bytes_readn +
+                           comp_stats_[level].bytes_readnp1;
+      double rw_amp = (comp_stats_[level].bytes_readn == 0) ? 0.0
+          : (comp_stats_[level].bytes_written + bytes_read) /
+            static_cast<double>(comp_stats_[level].bytes_readn);
+      double w_amp = (comp_stats_[level].bytes_readn == 0) ? 0.0
+          : comp_stats_[level].bytes_written /
+            static_cast<double>(comp_stats_[level].bytes_readn);
+      PrintLevelStats(buf, sizeof(buf), "L" + std::to_string(level),
+          files, files_being_compacted[level], current->NumLevelBytes(level),
+          compaction_score[level], rw_amp, w_amp, stall_us, stalls,
+          comp_stats_[level]);
+      value->append(buf);
+    }
+  }
+  uint64_t curr_ingest = cf_stats_value_[BYTES_FLUSHED];
+  // Cumulative summary
+  double rw_amp = (stats_sum.bytes_written + stats_sum.bytes_readn +
+      stats_sum.bytes_readnp1) / static_cast<double>(curr_ingest + 1);
+  double w_amp = stats_sum.bytes_written / static_cast<double>(curr_ingest + 1);
+  // Stats summary across levels
+  PrintLevelStats(buf, sizeof(buf), "Sum", total_files,
+      total_files_being_compacted, total_file_size, 0, rw_amp, w_amp,
+      total_stall_us, total_stall_count, stats_sum);
+  value->append(buf);
+  // Interval summary
+  uint64_t interval_ingest =
+      curr_ingest - cf_stats_snapshot_.ingest_bytes + 1;
+  CompactionStats interval_stats(stats_sum);
+  interval_stats.Subtract(cf_stats_snapshot_.comp_stats);
+  rw_amp = (interval_stats.bytes_written +
+      interval_stats.bytes_readn + interval_stats.bytes_readnp1) /
+      static_cast<double>(interval_ingest);
+  w_amp = interval_stats.bytes_written / static_cast<double>(interval_ingest);
+  PrintLevelStats(buf, sizeof(buf), "Int", 0, 0, 0, 0,
+      rw_amp, w_amp, total_stall_us - cf_stats_snapshot_.stall_us,
+      total_stall_count - cf_stats_snapshot_.stall_count, interval_stats);
+  value->append(buf);
+
+  snprintf(buf, sizeof(buf),
+           "Flush(GB): accumulative %.3f, interval %.3f\n",
+           curr_ingest / kGB, interval_ingest / kGB);
+  value->append(buf);
+  snprintf(buf, sizeof(buf),
+           "Stalls(secs): %.3f level0_slowdown, %.3f level0_numfiles, "
+           "%.3f memtable_compaction, %.3f leveln_slowdown_soft, "
+           "%.3f leveln_slowdown_hard\n",
+           cf_stats_value_[LEVEL0_SLOWDOWN] / 1000000.0,
+           cf_stats_value_[LEVEL0_NUM_FILES] / 1000000.0,
+           cf_stats_value_[MEMTABLE_COMPACTION] / 1000000.0,
+           total_slowdown_soft / 1000000.0,
+           total_slowdown_hard / 1000000.0);
+  value->append(buf);
+
+  snprintf(buf, sizeof(buf),
+           "Stalls(count): %" PRIu64 " level0_slowdown, "
+           "%" PRIu64 " level0_numfiles, %" PRIu64 " memtable_compaction, "
+           "%" PRIu64 " leveln_slowdown_soft, "
+           "%" PRIu64 " leveln_slowdown_hard\n",
+           cf_stats_count_[LEVEL0_SLOWDOWN],
+           cf_stats_count_[LEVEL0_NUM_FILES],
+           cf_stats_count_[MEMTABLE_COMPACTION],
+           total_slowdown_count_soft, total_slowdown_count_hard);
+  value->append(buf);
+
+  cf_stats_snapshot_.ingest_bytes = curr_ingest;
+  cf_stats_snapshot_.comp_stats = stats_sum;
+  cf_stats_snapshot_.stall_us = total_stall_us;
+  cf_stats_snapshot_.stall_count = total_stall_count;
+}
+
 }  // namespace rocksdb
@@ -9,8 +9,6 @@
 //

 #pragma once
-#include "rocksdb/statistics.h"
-#include "util/statistics.h"
 #include "db/version_set.h"

 #include <vector>
@@ -23,11 +21,15 @@ namespace rocksdb {
 class MemTableList;
 class DBImpl;

-enum DBPropertyType {
+enum DBPropertyType : uint32_t {
+  kUnknown,
  kNumFilesAtLevel,  // Number of files at a specific level
  kLevelStats,       // Return number of files and total sizes of each level
-  kStats,            // Return general statitistics of DB
+  kCFStats,          // Return general statitistics of CF
+  kDBStats,          // Return general statitistics of DB
+  kStats,            // Return general statitistics of both DB and CF
  kSsTables,         // Return a human readable string of current SST files
+  kStartIntTypes,    // ---- Dummy value to indicate the start of integer values
  kNumImmutableMemTable,   // Return number of immutable mem tables
  kMemtableFlushPending,   // Return 1 if mem table flushing is pending,
                           // otherwise 0.
@@ -38,45 +40,79 @@ enum DBPropertyType {
                                   // memtable.
  kNumEntriesInImmutableMemtable,  // Return sum of number of entries in all
                                   // the immutable mem tables.
-  kUnknown,
+  kEstimatedNumKeys,  // Estimated total number of keys in the database.
+  kEstimatedUsageByTableReaders,  // Estimated memory by table readers.
+  kIsFileDeletionEnabled,         // Equals disable_delete_obsolete_files_,
+                                  // 0 means file deletions enabled
 };

-extern DBPropertyType GetPropertyType(const Slice& property);
+extern DBPropertyType GetPropertyType(const Slice& property,
+                                      bool* is_int_property,
+                                      bool* need_out_of_mutex);

 class InternalStats {
 public:
-  enum WriteStallType {
+  enum InternalCFStatsType {
    LEVEL0_SLOWDOWN,
    MEMTABLE_COMPACTION,
    LEVEL0_NUM_FILES,
    WRITE_STALLS_ENUM_MAX,
+    BYTES_FLUSHED,
+    INTERNAL_CF_STATS_ENUM_MAX,
  };

-  InternalStats(int num_levels, Env* env, Statistics* statistics)
-      : compaction_stats_(num_levels),
-        stall_micros_(WRITE_STALLS_ENUM_MAX, 0),
-        stall_counts_(WRITE_STALLS_ENUM_MAX, 0),
-        stall_leveln_slowdown_(num_levels, 0),
-        stall_leveln_slowdown_count_(num_levels, 0),
+  enum InternalDBStatsType {
+    WAL_FILE_BYTES,
+    WAL_FILE_SYNCED,
+    BYTES_WRITTEN,
+    WRITE_DONE_BY_OTHER,
+    WRITE_DONE_BY_SELF,
+    WRITE_WITH_WAL,
+    INTERNAL_DB_STATS_ENUM_MAX,
+  };
+
+  InternalStats(int num_levels, Env* env, ColumnFamilyData* cfd)
+      : db_stats_(INTERNAL_DB_STATS_ENUM_MAX),
+        cf_stats_value_(INTERNAL_CF_STATS_ENUM_MAX),
+        cf_stats_count_(INTERNAL_CF_STATS_ENUM_MAX),
+        comp_stats_(num_levels),
+        stall_leveln_slowdown_hard_(num_levels),
+        stall_leveln_slowdown_count_hard_(num_levels),
+        stall_leveln_slowdown_soft_(num_levels),
+        stall_leveln_slowdown_count_soft_(num_levels),
        bg_error_count_(0),
        number_levels_(num_levels),
-        statistics_(statistics),
        env_(env),
-        started_at_(env->NowMicros()) {}
+        cfd_(cfd),
+        started_at_(env->NowMicros()) {
+    for (int i = 0; i< INTERNAL_DB_STATS_ENUM_MAX; ++i) {
+      db_stats_[i] = 0;
+    }
+    for (int i = 0; i< INTERNAL_CF_STATS_ENUM_MAX; ++i) {
+      cf_stats_value_[i] = 0;
+      cf_stats_count_[i] = 0;
+    }
+    for (int i = 0; i < num_levels; ++i) {
+      stall_leveln_slowdown_hard_[i] = 0;
+      stall_leveln_slowdown_count_hard_[i] = 0;
+      stall_leveln_slowdown_soft_[i] = 0;
+      stall_leveln_slowdown_count_soft_[i] = 0;
+    }
+  }

-  // Per level compaction stats.  compaction_stats_[level] stores the stats for
+  // Per level compaction stats.  comp_stats_[level] stores the stats for
  // compactions that produced data for the specified "level".
  struct CompactionStats {
    uint64_t micros;

    // Bytes read from level N during compaction between levels N and N+1
-    int64_t bytes_readn;
+    uint64_t bytes_readn;

    // Bytes read from level N+1 during compaction between levels N and N+1
-    int64_t bytes_readnp1;
+    uint64_t bytes_readnp1;

    // Total bytes written during compaction between levels N and N+1
-    int64_t bytes_written;
+    uint64_t bytes_written;

    // Files read from level N during compaction between levels N and N+1
    int files_in_leveln;
@@ -87,10 +123,17 @@ class InternalStats {
    // Files written during compaction between levels N and N+1
    int files_out_levelnp1;

+    // Total incoming entries during compaction between levels N and N+1
+    int num_input_records;
+
+    // Accumulated diff number of entries
+    // (num input entries - num output entires) for compaction  levels N and N+1
+    int num_dropped_records;
+
    // Number of compactions done
    int count;

-    CompactionStats()
+    explicit CompactionStats(int count = 0)
        : micros(0),
          bytes_readn(0),
          bytes_readnp1(0),
@@ -98,7 +141,21 @@ class InternalStats {
          files_in_leveln(0),
          files_in_levelnp1(0),
          files_out_levelnp1(0),
-          count(0) {}
+          num_input_records(0),
+          num_dropped_records(0),
+          count(count) {}
+
+    explicit CompactionStats(const CompactionStats& c)
+        : micros(c.micros),
+          bytes_readn(c.bytes_readn),
+          bytes_readnp1(c.bytes_readnp1),
+          bytes_written(c.bytes_written),
+          files_in_leveln(c.files_in_leveln),
+          files_in_levelnp1(c.files_in_levelnp1),
+          files_out_levelnp1(c.files_out_levelnp1),
+          num_input_records(c.num_input_records),
+          num_dropped_records(c.num_dropped_records),
+          count(c.count) {}

    void Add(const CompactionStats& c) {
      this->micros += c.micros;
@@ -108,68 +165,114 @@ class InternalStats {
      this->files_in_leveln += c.files_in_leveln;
      this->files_in_levelnp1 += c.files_in_levelnp1;
      this->files_out_levelnp1 += c.files_out_levelnp1;
-      this->count += 1;
+      this->num_input_records += c.num_input_records;
+      this->num_dropped_records += c.num_dropped_records;
+      this->count += c.count;
+    }
+
+    void Subtract(const CompactionStats& c) {
+      this->micros -= c.micros;
+      this->bytes_readn -= c.bytes_readn;
+      this->bytes_readnp1 -= c.bytes_readnp1;
+      this->bytes_written -= c.bytes_written;
+      this->files_in_leveln -= c.files_in_leveln;
+      this->files_in_levelnp1 -= c.files_in_levelnp1;
+      this->files_out_levelnp1 -= c.files_out_levelnp1;
+      this->num_input_records -= c.num_input_records;
+      this->num_dropped_records -= c.num_dropped_records;
+      this->count -= c.count;
    }
  };

  void AddCompactionStats(int level, const CompactionStats& stats) {
-    compaction_stats_[level].Add(stats);
+    comp_stats_[level].Add(stats);
  }

-  void RecordWriteStall(WriteStallType write_stall_type, uint64_t micros) {
-    stall_micros_[write_stall_type] += micros;
-    stall_counts_[write_stall_type]++;
+  void RecordLevelNSlowdown(int level, uint64_t micros, bool soft) {
+    if (soft) {
+      stall_leveln_slowdown_soft_[level] += micros;
+      ++stall_leveln_slowdown_count_soft_[level];
+    } else {
+      stall_leveln_slowdown_hard_[level] += micros;
+      ++stall_leveln_slowdown_count_hard_[level];
+    }
  }

-  void RecordLevelNSlowdown(int level, uint64_t micros) {
-    stall_leveln_slowdown_[level] += micros;
-    stall_leveln_slowdown_count_[level] += micros;
+  void AddCFStats(InternalCFStatsType type, uint64_t value) {
+    cf_stats_value_[type] += value;
+    ++cf_stats_count_[type];
+  }
+
+  void AddDBStats(InternalDBStatsType type, uint64_t value) {
+    db_stats_[type] += value;
  }

  uint64_t GetBackgroundErrorCount() const { return bg_error_count_; }

  uint64_t BumpAndGetBackgroundErrorCount() { return ++bg_error_count_; }

-  bool GetProperty(DBPropertyType property_type, const Slice& property,
-                   std::string* value, ColumnFamilyData* cfd);
+  bool GetStringProperty(DBPropertyType property_type, const Slice& property,
+                         std::string* value);
+
+  bool GetIntProperty(DBPropertyType property_type, uint64_t* value,
+                      DBImpl* db) const;
+
+  bool GetIntPropertyOutOfMutex(DBPropertyType property_type, Version* version,
+                                uint64_t* value) const;

 private:
-  std::vector<CompactionStats> compaction_stats_;
+  void DumpDBStats(std::string* value);
+  void DumpCFStats(std::string* value);
+
+  // Per-DB stats
+  std::vector<uint64_t> db_stats_;
+  // Per-ColumnFamily stats
+  std::vector<uint64_t> cf_stats_value_;
+  std::vector<uint64_t> cf_stats_count_;
+  // Per-ColumnFamily/level compaction stats
+  std::vector<CompactionStats> comp_stats_;
+  // These count the number of microseconds for which MakeRoomForWrite stalls.
+  std::vector<uint64_t> stall_leveln_slowdown_hard_;
+  std::vector<uint64_t> stall_leveln_slowdown_count_hard_;
+  std::vector<uint64_t> stall_leveln_slowdown_soft_;
+  std::vector<uint64_t> stall_leveln_slowdown_count_soft_;

  // Used to compute per-interval statistics
-  struct StatsSnapshot {
-    uint64_t compaction_bytes_read_;     // Bytes read by compaction
-    uint64_t compaction_bytes_written_;  // Bytes written by compaction
-    uint64_t ingest_bytes_;              // Bytes written by user
-    uint64_t wal_bytes_;                 // Bytes written to WAL
-    uint64_t wal_synced_;                // Number of times WAL is synced
-    uint64_t write_with_wal_;            // Number of writes that request WAL
+  struct CFStatsSnapshot {
+    // ColumnFamily-level stats
+    CompactionStats comp_stats;
+    uint64_t ingest_bytes;            // Bytes written to L0
+    uint64_t stall_us;                // Stall time in micro-seconds
+    uint64_t stall_count;             // Stall count
+
+    CFStatsSnapshot()
+        : comp_stats(0),
+          ingest_bytes(0),
+          stall_us(0),
+          stall_count(0) {}
+  } cf_stats_snapshot_;
+
+  struct DBStatsSnapshot {
+    // DB-level stats
+    uint64_t ingest_bytes;            // Bytes written by user
+    uint64_t wal_bytes;               // Bytes written to WAL
+    uint64_t wal_synced;              // Number of times WAL is synced
+    uint64_t write_with_wal;          // Number of writes that request WAL
    // These count the number of writes processed by the calling thread or
    // another thread.
-    uint64_t write_other_;
-    uint64_t write_self_;
-    double seconds_up_;
+    uint64_t write_other;
+    uint64_t write_self;
+    double seconds_up;

-    StatsSnapshot()
-        : compaction_bytes_read_(0),
-          compaction_bytes_written_(0),
-          ingest_bytes_(0),
-          wal_bytes_(0),
-          wal_synced_(0),
-          write_with_wal_(0),
-          write_other_(0),
-          write_self_(0),
-          seconds_up_(0) {}
-  };
-
-  // Counters from the previous time per-interval stats were computed
-  StatsSnapshot last_stats_;
-
-  // These count the number of microseconds for which MakeRoomForWrite stalls.
-  std::vector<uint64_t> stall_micros_;
-  std::vector<uint64_t> stall_counts_;
-  std::vector<uint64_t> stall_leveln_slowdown_;
-  std::vector<uint64_t> stall_leveln_slowdown_count_;
+    DBStatsSnapshot()
+        : ingest_bytes(0),
+          wal_bytes(0),
+          wal_synced(0),
+          write_with_wal(0),
+          write_other(0),
+          write_self(0),
+          seconds_up(0) {}
+  } db_stats_snapshot_;

  // Total number of background errors encountered. Every time a flush task
  // or compaction task fails, this counter is incremented. The failure can
@@ -178,10 +281,10 @@ class InternalStats {
  // or compaction will cause the counter to increase too.
  uint64_t bg_error_count_;

-  int number_levels_;
-  Statistics* statistics_;
+  const int number_levels_;
  Env* env_;
-  uint64_t started_at_;
+  ColumnFamilyData* cfd_;
+  const uint64_t started_at_;
 };

 }  // namespace rocksdb
@@ -9,6 +9,7 @@
 #include "util/testharness.h"
 #include "util/benchharness.h"
 #include "db/version_set.h"
+#include "db/write_controller.h"
 #include "util/mutexlock.h"

 namespace rocksdb {
@@ -21,6 +22,7 @@ std::string MakeKey(unsigned int num) {

 void BM_LogAndApply(int iters, int num_base_files) {
  VersionSet* vset;
+  WriteController wc;
  ColumnFamilyData* default_cfd;
  uint64_t fnum = 1;
  port::Mutex mu;
@@ -42,7 +44,12 @@ void BM_LogAndApply(int iters, int num_base_files) {

    Options options;
    EnvOptions sopt;
-    vset = new VersionSet(dbname, &options, sopt, nullptr);
+    // Notice we are using the default options not through SanitizeOptions().
+    // We might want to initialize some options manually if needed.
+    options.db_paths.emplace_back(dbname, 0);
+    // The parameter of table cache is passed in as null, so any file I/O
+    // operation is likely to fail.
+    vset = new VersionSet(dbname, &options, sopt, nullptr, &wc);
    std::vector<ColumnFamilyDescriptor> dummy;
    dummy.push_back(ColumnFamilyDescriptor());
    ASSERT_OK(vset->Recover(dummy));
@@ -51,9 +58,10 @@ void BM_LogAndApply(int iters, int num_base_files) {
    for (int i = 0; i < num_base_files; i++) {
      InternalKey start(MakeKey(2 * fnum), 1, kTypeValue);
      InternalKey limit(MakeKey(2 * fnum + 1), 1, kTypeDeletion);
-      vbase.AddFile(2, ++fnum, 1 /* file size */, start, limit, 1, 1);
+      vbase.AddFile(2, ++fnum, 0, 1 /* file size */, start, limit, 1, 1);
    }
-    ASSERT_OK(vset->LogAndApply(default_cfd, &vbase, &mu));
+    ASSERT_OK(vset->LogAndApply(default_cfd,
+          *default_cfd->GetLatestMutableCFOptions(), &vbase, &mu));
  }

  for (int i = 0; i < iters; i++) {
@@ -61,9 +69,11 @@ void BM_LogAndApply(int iters, int num_base_files) {
    vedit.DeleteFile(2, fnum);
    InternalKey start(MakeKey(2 * fnum), 1, kTypeValue);
    InternalKey limit(MakeKey(2 * fnum + 1), 1, kTypeDeletion);
-    vedit.AddFile(2, ++fnum, 1 /* file size */, start, limit, 1, 1);
-    vset->LogAndApply(default_cfd, &vedit, &mu);
+    vedit.AddFile(2, ++fnum, 0, 1 /* file size */, start, limit, 1, 1);
+    vset->LogAndApply(default_cfd, *default_cfd->GetLatestMutableCFOptions(),
+                      &vedit, &mu);
  }
+  delete vset;
 }

 BENCHMARK_NAMED_PARAM(BM_LogAndApply, 1000_iters_1_file, 1000, 1)
@@ -20,6 +20,7 @@
 #include "rocksdb/iterator.h"
 #include "rocksdb/merge_operator.h"
 #include "rocksdb/slice_transform.h"
+#include "table/merger.h"
 #include "util/arena.h"
 #include "util/coding.h"
 #include "util/murmurhash.h"
@@ -30,40 +31,57 @@

 namespace rocksdb {

-MemTable::MemTable(const InternalKeyComparator& cmp, const Options& options)
+MemTableOptions::MemTableOptions(
+    const MutableCFOptions& mutable_cf_options, const Options& options)
+  : write_buffer_size(mutable_cf_options.write_buffer_size),
+    arena_block_size(mutable_cf_options.arena_block_size),
+    memtable_prefix_bloom_bits(mutable_cf_options.memtable_prefix_bloom_bits),
+    memtable_prefix_bloom_probes(
+        mutable_cf_options.memtable_prefix_bloom_probes),
+    memtable_prefix_bloom_huge_page_tlb_size(
+        mutable_cf_options.memtable_prefix_bloom_huge_page_tlb_size),
+    inplace_update_support(options.inplace_update_support),
+    inplace_update_num_locks(options.inplace_update_num_locks),
+    inplace_callback(options.inplace_callback),
+    max_successive_merges(mutable_cf_options.max_successive_merges),
+    filter_deletes(mutable_cf_options.filter_deletes) {}
+
+MemTable::MemTable(const InternalKeyComparator& cmp,
+                   const ImmutableCFOptions& ioptions,
+                   const MemTableOptions& moptions)
    : comparator_(cmp),
+      ioptions_(ioptions),
+      moptions_(moptions),
      refs_(0),
-      kArenaBlockSize(OptimizeBlockSize(options.arena_block_size)),
-      kWriteBufferSize(options.write_buffer_size),
-      arena_(options.arena_block_size),
-      table_(options.memtable_factory->CreateMemTableRep(
-          comparator_, &arena_, options.prefix_extractor.get(),
-          options.info_log.get())),
+      kArenaBlockSize(OptimizeBlockSize(moptions.arena_block_size)),
+      arena_(moptions.arena_block_size),
+      table_(ioptions.memtable_factory->CreateMemTableRep(
+          comparator_, &arena_, ioptions.prefix_extractor, ioptions.info_log)),
      num_entries_(0),
      flush_in_progress_(false),
      flush_completed_(false),
      file_number_(0),
      first_seqno_(0),
      mem_next_logfile_number_(0),
-      locks_(options.inplace_update_support ? options.inplace_update_num_locks
-                                            : 0),
-      prefix_extractor_(options.prefix_extractor.get()),
-      should_flush_(ShouldFlushNow()) {
+      locks_(moptions.inplace_update_support ? moptions.inplace_update_num_locks
+                                             : 0),
+      prefix_extractor_(ioptions.prefix_extractor),
+      should_flush_(ShouldFlushNow()),
+      flush_scheduled_(false) {
  // if should_flush_ == true without an entry inserted, something must have
  // gone wrong already.
  assert(!should_flush_);
-  if (prefix_extractor_ && options.memtable_prefix_bloom_bits > 0) {
+  if (prefix_extractor_ && moptions.memtable_prefix_bloom_bits > 0) {
    prefix_bloom_.reset(new DynamicBloom(
-        options.memtable_prefix_bloom_bits, options.bloom_locality,
-        options.memtable_prefix_bloom_probes, nullptr,
-        options.memtable_prefix_bloom_huge_page_tlb_size,
-        options.info_log.get()));
+        &arena_,
+        moptions.memtable_prefix_bloom_bits, ioptions.bloom_locality,
+        moptions.memtable_prefix_bloom_probes, nullptr,
+        moptions.memtable_prefix_bloom_huge_page_tlb_size,
+        ioptions.info_log));
  }
 }

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

 size_t MemTable::ApproximateMemoryUsage() {
  size_t arena_usage = arena_.ApproximateMemoryUsage();
@@ -95,14 +113,16 @@ bool MemTable::ShouldFlushNow() const {
  // if we can still allocate one more block without exceeding the
  // over-allocation ratio, then we should not flush.
  if (allocated_memory + kArenaBlockSize <
-      kWriteBufferSize + kArenaBlockSize * kAllowOverAllocationRatio) {
+      moptions_.write_buffer_size +
+      kArenaBlockSize * kAllowOverAllocationRatio) {
    return false;
  }

-  // if user keeps adding entries that exceeds kWriteBufferSize, we need to
-  // flush earlier even though we still have much available memory left.
-  if (allocated_memory >
-      kWriteBufferSize + kArenaBlockSize * kAllowOverAllocationRatio) {
+  // if user keeps adding entries that exceeds moptions.write_buffer_size,
+  // we need to flush earlier even though we still have much available
+  // memory left.
+  if (allocated_memory > moptions_.write_buffer_size +
+      kArenaBlockSize * kAllowOverAllocationRatio) {
    return true;
  }

@@ -172,16 +192,25 @@ const char* EncodeKey(std::string* scratch, const Slice& target) {

 class MemTableIterator: public Iterator {
 public:
-  MemTableIterator(const MemTable& mem, const ReadOptions& options,
-                   bool enforce_total_order)
+  MemTableIterator(
+      const MemTable& mem, const ReadOptions& read_options, Arena* arena)
      : bloom_(nullptr),
        prefix_extractor_(mem.prefix_extractor_),
-        valid_(false) {
-    if (prefix_extractor_ != nullptr && !enforce_total_order) {
+        valid_(false),
+        arena_mode_(arena != nullptr) {
+    if (prefix_extractor_ != nullptr && !read_options.total_order_seek) {
      bloom_ = mem.prefix_bloom_.get();
-      iter_.reset(mem.table_->GetDynamicPrefixIterator());
+      iter_ = mem.table_->GetDynamicPrefixIterator(arena);
    } else {
-      iter_.reset(mem.table_->GetIterator());
+      iter_ = mem.table_->GetIterator(arena);
+    }
+  }
+
+  ~MemTableIterator() {
+    if (arena_mode_) {
+      iter_->~Iterator();
+    } else {
+      delete iter_;
    }
  }

@@ -228,17 +257,19 @@ class MemTableIterator: public Iterator {
 private:
  DynamicBloom* bloom_;
  const SliceTransform* const prefix_extractor_;
-  std::unique_ptr<MemTableRep::Iterator> iter_;
+  MemTableRep::Iterator* iter_;
  bool valid_;
+  bool arena_mode_;

  // No copying allowed
  MemTableIterator(const MemTableIterator&);
  void operator=(const MemTableIterator&);
 };

-Iterator* MemTable::NewIterator(const ReadOptions& options,
-    bool enforce_total_order) {
-  return new MemTableIterator(*this, options, enforce_total_order);
+Iterator* MemTable::NewIterator(const ReadOptions& read_options, Arena* arena) {
+  assert(arena != nullptr);
+  auto mem = arena->AllocateAligned(sizeof(MemTableIterator));
+  return new (mem) MemTableIterator(*this, read_options, arena);
 }

 port::RWMutex* MemTable::GetLock(const Slice& key) {
@@ -349,7 +380,7 @@ static bool SaveValue(void* arg, const char* entry) {
          s->value->assign(v.data(), v.size());
        }
        if (s->inplace_update_support) {
-          s->mem->GetLock(s->key->user_key())->Unlock();
+          s->mem->GetLock(s->key->user_key())->ReadUnlock();
        }
        *(s->found_final_value) = true;
        return false;
@@ -372,7 +403,16 @@ static bool SaveValue(void* arg, const char* entry) {
        return false;
      }
      case kTypeMerge: {
-        std::string merge_result;  // temporary area for merge results later
+        if (!merge_operator) {
+          *(s->status) = Status::InvalidArgument(
+              "merge_operator is not properly initialized.");
+          // Normally we continue the loop (return true) when we see a merge
+          // operand.  But in case of an error, we should stop the loop
+          // immediately and pretend we have found the value to stop further
+          // seek.  Otherwise, the later call will override this error status.
+          *(s->found_final_value) = true;
+          return false;
+        }
        Slice v = GetLengthPrefixedSlice(key_ptr + key_length);
        *(s->merge_in_progress) = true;
        merge_context->PushOperand(v);
@@ -389,8 +429,13 @@ static bool SaveValue(void* arg, const char* entry) {
 }

 bool MemTable::Get(const LookupKey& key, std::string* value, Status* s,
-                   MergeContext& merge_context, const Options& options) {
-  PERF_TIMER_AUTO(get_from_memtable_time);
+                   MergeContext* merge_context) {
+  // The sequence number is updated synchronously in version_set.h
+  if (IsEmpty()) {
+    // Avoiding recording stats for speed.
+    return false;
+  }
+  PERF_TIMER_GUARD(get_from_memtable_time);

  Slice user_key = key.user_key();
  bool found_final_value = false;
@@ -408,11 +453,11 @@ bool MemTable::Get(const LookupKey& key, std::string* value, Status* s,
    saver.value = value;
    saver.status = s;
    saver.mem = this;
-    saver.merge_context = &merge_context;
-    saver.merge_operator = options.merge_operator.get();
-    saver.logger = options.info_log.get();
-    saver.inplace_update_support = options.inplace_update_support;
-    saver.statistics = options.statistics.get();
+    saver.merge_context = merge_context;
+    saver.merge_operator = ioptions_.merge_operator;
+    saver.logger = ioptions_.info_log;
+    saver.inplace_update_support = moptions_.inplace_update_support;
+    saver.statistics = ioptions_.statistics;
    table_->Get(key, &saver, SaveValue);
  }

@@ -420,7 +465,6 @@ bool MemTable::Get(const LookupKey& key, std::string* value, Status* s,
  if (!found_final_value && merge_in_progress) {
    *s = Status::MergeInProgress("");
  }
-  PERF_TIMER_STOP(get_from_memtable_time);
  PERF_COUNTER_ADD(get_from_memtable_count, 1);
  return found_final_value;
 }
@@ -432,7 +476,7 @@ void MemTable::Update(SequenceNumber seq,
  Slice mem_key = lkey.memtable_key();

  std::unique_ptr<MemTableRep::Iterator> iter(
-    table_->GetIterator(lkey.user_key()));
+      table_->GetDynamicPrefixIterator());
  iter->Seek(lkey.internal_key(), mem_key.data());

  if (iter->Valid()) {
@@ -485,13 +529,12 @@ void MemTable::Update(SequenceNumber seq,

 bool MemTable::UpdateCallback(SequenceNumber seq,
                              const Slice& key,
-                              const Slice& delta,
-                              const Options& options) {
+                              const Slice& delta) {
  LookupKey lkey(key, seq);
  Slice memkey = lkey.memtable_key();

  std::unique_ptr<MemTableRep::Iterator> iter(
-    table_->GetIterator(lkey.user_key()));
+      table_->GetDynamicPrefixIterator());
  iter->Seek(lkey.internal_key(), memkey.data());

  if (iter->Valid()) {
@@ -521,8 +564,8 @@ bool MemTable::UpdateCallback(SequenceNumber seq,

          std::string str_value;
          WriteLock wl(GetLock(lkey.user_key()));
-          auto status = options.inplace_callback(prev_buffer, &new_prev_size,
-                                                    delta, &str_value);
+          auto status = moptions_.inplace_callback(prev_buffer, &new_prev_size,
+                                                   delta, &str_value);
          if (status == UpdateStatus::UPDATED_INPLACE) {
            // Value already updated by callback.
            assert(new_prev_size <= prev_size);
@@ -535,12 +578,12 @@ bool MemTable::UpdateCallback(SequenceNumber seq,
                memcpy(p, prev_buffer, new_prev_size);
              }
            }
-            RecordTick(options.statistics.get(), NUMBER_KEYS_UPDATED);
+            RecordTick(ioptions_.statistics, NUMBER_KEYS_UPDATED);
            should_flush_ = ShouldFlushNow();
            return true;
          } else if (status == UpdateStatus::UPDATED) {
            Add(seq, kTypeValue, key, Slice(str_value));
-            RecordTick(options.statistics.get(), NUMBER_KEYS_WRITTEN);
+            RecordTick(ioptions_.statistics, NUMBER_KEYS_WRITTEN);
            should_flush_ = ShouldFlushNow();
            return true;
          } else if (status == UpdateStatus::UPDATE_FAILED) {
@@ -566,7 +609,7 @@ size_t MemTable::CountSuccessiveMergeEntries(const LookupKey& key) {
  // reps). By passing in the user key, we allow efficient iterator creation.
  // The iterator only needs to be ordered within the same user key.
  std::unique_ptr<MemTableRep::Iterator> iter(
-      table_->GetIterator(key.user_key()));
+      table_->GetDynamicPrefixIterator());
  iter->Seek(key.internal_key(), memkey.data());

  size_t num_successive_merges = 0;
@@ -593,7 +636,7 @@ size_t MemTable::CountSuccessiveMergeEntries(const LookupKey& key) {

 void MemTableRep::Get(const LookupKey& k, void* callback_args,
                      bool (*callback_func)(void* arg, const char* entry)) {
-  auto iter = GetIterator(k.user_key());
+  auto iter = GetDynamicPrefixIterator();
  for (iter->Seek(k.internal_key(), k.memtable_key().data());
       iter->Valid() && callback_func(callback_args, iter->key());
       iter->Next()) {
@@ -10,21 +10,45 @@
 #pragma once
 #include <string>
 #include <memory>
+#include <functional>
 #include <deque>
+#include <vector>
 #include "db/dbformat.h"
 #include "db/skiplist.h"
 #include "db/version_edit.h"
 #include "rocksdb/db.h"
 #include "rocksdb/memtablerep.h"
+#include "rocksdb/immutable_options.h"
 #include "util/arena.h"
 #include "util/dynamic_bloom.h"
+#include "util/mutable_cf_options.h"

 namespace rocksdb {

+class Arena;
 class Mutex;
 class MemTableIterator;
 class MergeContext;

+struct MemTableOptions {
+  explicit MemTableOptions(
+      const MutableCFOptions& mutable_cf_options,
+      const Options& options);
+  size_t write_buffer_size;
+  size_t arena_block_size;
+  uint32_t memtable_prefix_bloom_bits;
+  uint32_t memtable_prefix_bloom_probes;
+  size_t memtable_prefix_bloom_huge_page_tlb_size;
+  bool inplace_update_support;
+  size_t inplace_update_num_locks;
+  UpdateStatus (*inplace_callback)(char* existing_value,
+                                   uint32_t* existing_value_size,
+                                   Slice delta_value,
+                                   std::string* merged_value);
+  size_t max_successive_merges;
+  bool filter_deletes;
+};
+
 class MemTable {
 public:
  struct KeyComparator : public MemTableRep::KeyComparator {
@@ -39,7 +63,8 @@ class MemTable {
  // MemTables are reference counted.  The initial reference count
  // is zero and the caller must call Ref() at least once.
  explicit MemTable(const InternalKeyComparator& comparator,
-                    const Options& options);
+                    const ImmutableCFOptions& ioptions,
+                    const MemTableOptions& moptions);

  ~MemTable();

@@ -66,7 +91,11 @@ class MemTable {

  // This method heuristically determines if the memtable should continue to
  // host more data.
-  bool ShouldFlush() const { return should_flush_; }
+  bool ShouldScheduleFlush() const {
+    return flush_scheduled_ == false && should_flush_;
+  }
+
+  void MarkFlushScheduled() { flush_scheduled_ = true; }

  // Return an iterator that yields the contents of the memtable.
  //
@@ -77,8 +106,10 @@ class MemTable {
  //
  // By default, it returns an iterator for prefix seek if prefix_extractor
  // is configured in Options.
-  Iterator* NewIterator(const ReadOptions& options,
-                        bool enforce_total_order = false);
+  // arena: If not null, the arena needs to be used to allocate the Iterator.
+  //        Calling ~Iterator of the iterator will destroy all the states but
+  //        those allocated in arena.
+  Iterator* NewIterator(const ReadOptions& read_options, Arena* arena);

  // Add an entry into memtable that maps key to value at the
  // specified sequence number and with the specified type.
@@ -96,7 +127,7 @@ class MemTable {
  //   store MergeInProgress in s, and return false.
  // Else, return false.
  bool Get(const LookupKey& key, std::string* value, Status* s,
-           MergeContext& merge_context, const Options& options);
+           MergeContext* merge_context);

  // Attempts to update the new_value inplace, else does normal Add
  // Pseudocode
@@ -120,8 +151,7 @@ class MemTable {
  //   else return false
  bool UpdateCallback(SequenceNumber seq,
                      const Slice& key,
-                      const Slice& delta,
-                      const Options& options);
+                      const Slice& delta);

  // Returns the number of successive merge entries starting from the newest
  // entry for the key up to the last non-merge entry or last entry for the
@@ -134,6 +164,9 @@ class MemTable {
  // Returns the edits area that is needed for flushing the memtable
  VersionEdit* GetEdits() { return &edit_; }

+  // Returns if there is no entry inserted to the mem table.
+  bool IsEmpty() const { return first_seqno_ == 0; }
+
  // Returns the sequence number of the first element that was inserted
  // into the memtable
  SequenceNumber GetFirstSequenceNumber() { return first_seqno_; }
@@ -166,8 +199,11 @@ class MemTable {

  const Arena& TEST_GetArena() const { return arena_; }

+  const ImmutableCFOptions* GetImmutableOptions() const { return &ioptions_; }
+  const MemTableOptions* GetMemTableOptions() const { return &moptions_; }
+
 private:
-  // Dynamically check if we can add more incoming entries.
+  // Dynamically check if we can add more incoming entries
  bool ShouldFlushNow() const;

  friend class MemTableIterator;
@@ -175,9 +211,10 @@ class MemTable {
  friend class MemTableList;

  KeyComparator comparator_;
+  const ImmutableCFOptions& ioptions_;
+  const MemTableOptions moptions_;
  int refs_;
  const size_t kArenaBlockSize;
-  const size_t kWriteBufferSize;
  Arena arena_;
  unique_ptr<MemTableRep> table_;

@@ -210,6 +247,9 @@ class MemTable {

  // a flag indicating if a memtable has met the criteria to flush
  bool should_flush_;
+
+  // a flag indicating if flush has been scheduled
+  bool flush_scheduled_;
 };

 extern const char* EncodeKey(std::string* scratch, const Slice& target);
@@ -11,6 +11,7 @@
 #include "db/version_set.h"
 #include "rocksdb/env.h"
 #include "rocksdb/iterator.h"
+#include "table/merger.h"
 #include "util/coding.h"
 #include "util/log_buffer.h"

@@ -61,10 +62,9 @@ int MemTableList::size() const {
 // Return the most recent value found, if any.
 // Operands stores the list of merge operations to apply, so far.
 bool MemTableListVersion::Get(const LookupKey& key, std::string* value,
-                              Status* s, MergeContext& merge_context,
-                              const Options& options) {
+                              Status* s, MergeContext* merge_context) {
  for (auto& memtable : memlist_) {
-    if (memtable->Get(key, value, s, merge_context, options)) {
+    if (memtable->Get(key, value, s, merge_context)) {
      return true;
    }
  }
@@ -72,9 +72,18 @@ bool MemTableListVersion::Get(const LookupKey& key, std::string* value,
 }

 void MemTableListVersion::AddIterators(const ReadOptions& options,
-                                       std::vector<Iterator*>* iterator_list) {
+                                       std::vector<Iterator*>* iterator_list,
+                                       Arena* arena) {
  for (auto& m : memlist_) {
-    iterator_list->push_back(m->NewIterator(options));
+    iterator_list->push_back(m->NewIterator(options, arena));
+  }
+}
+
+void MemTableListVersion::AddIterators(
+    const ReadOptions& options, MergeIteratorBuilder* merge_iter_builder) {
+  for (auto& m : memlist_) {
+    merge_iter_builder->AddIterator(
+        m->NewIterator(options, merge_iter_builder->GetArena()));
  }
 }

@@ -131,7 +140,7 @@ void MemTableList::PickMemtablesToFlush(autovector<MemTable*>* ret) {

 void MemTableList::RollbackMemtableFlush(const autovector<MemTable*>& mems,
                                         uint64_t file_number,
-                                         std::set<uint64_t>* pending_outputs) {
+                                         FileNumToPathIdMap* pending_outputs) {
  assert(!mems.empty());

  // If the flush was not successful, then just reset state.
@@ -151,9 +160,10 @@ void MemTableList::RollbackMemtableFlush(const autovector<MemTable*>& mems,

 // Record a successful flush in the manifest file
 Status MemTableList::InstallMemtableFlushResults(
-    ColumnFamilyData* cfd, const autovector<MemTable*>& mems, VersionSet* vset,
+    ColumnFamilyData* cfd, const MutableCFOptions& mutable_cf_options,
+    const autovector<MemTable*>& mems, VersionSet* vset,
    port::Mutex* mu, Logger* info_log, uint64_t file_number,
-    std::set<uint64_t>& pending_outputs, autovector<MemTable*>* to_delete,
+    FileNumToPathIdMap* pending_outputs, autovector<MemTable*>* to_delete,
    Directory* db_directory, LogBuffer* log_buffer) {
  mu->AssertHeld();

@@ -188,7 +198,7 @@ Status MemTableList::InstallMemtableFlushResults(
                cfd->GetName().c_str(), (unsigned long)m->file_number_);

    // this can release and reacquire the mutex.
-    s = vset->LogAndApply(cfd, &m->edit_, mu, db_directory);
+    s = vset->LogAndApply(cfd, mutable_cf_options, &m->edit_, mu, db_directory);

    // we will be changing the version in the next code path,
    // so we better create a new one, since versions are immutable
@@ -210,7 +220,7 @@ Status MemTableList::InstallMemtableFlushResults(
        // has been written to a committed version so that other concurrently
        // executing compaction threads do not mistakenly assume that this
        // file is not live.
-        pending_outputs.erase(m->file_number_);
+        pending_outputs->erase(m->file_number_);
        if (m->Unref() != nullptr) {
          to_delete->push_back(m);
        }
@@ -224,7 +234,7 @@ Status MemTableList::InstallMemtableFlushResults(
        m->flush_in_progress_ = false;
        m->edit_.Clear();
        num_flush_not_started_++;
-        pending_outputs.erase(m->file_number_);
+        pending_outputs->erase(m->file_number_);
        m->file_number_ = 0;
        imm_flush_needed.Release_Store((void *)1);
      }
@@ -15,6 +15,7 @@
 #include "rocksdb/iterator.h"

 #include "db/dbformat.h"
+#include "db/filename.h"
 #include "db/skiplist.h"
 #include "db/memtable.h"
 #include "rocksdb/db.h"
@@ -28,6 +29,7 @@ namespace rocksdb {
 class ColumnFamilyData;
 class InternalKeyComparator;
 class Mutex;
+class MergeIteratorBuilder;

 // keeps a list of immutable memtables in a vector. the list is immutable
 // if refcount is bigger than one. It is used as a state for Get() and
@@ -44,10 +46,13 @@ class MemTableListVersion {
  // 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,
-           MergeContext& merge_context, const Options& options);
+           MergeContext* merge_context);

  void AddIterators(const ReadOptions& options,
-                    std::vector<Iterator*>* iterator_list);
+                    std::vector<Iterator*>* iterator_list, Arena* arena);
+
+  void AddIterators(const ReadOptions& options,
+                    MergeIteratorBuilder* merge_iter_builder);

  uint64_t GetTotalNumEntries() const;

@@ -104,17 +109,15 @@ class MemTableList {
  // they can get picked up again on the next round of flush.
  void RollbackMemtableFlush(const autovector<MemTable*>& mems,
                             uint64_t file_number,
-                             std::set<uint64_t>* pending_outputs);
+                             FileNumToPathIdMap* pending_outputs);

  // Commit a successful flush in the manifest file
-  Status InstallMemtableFlushResults(ColumnFamilyData* cfd,
-                                     const autovector<MemTable*>& m,
-                                     VersionSet* vset, port::Mutex* mu,
-                                     Logger* info_log, uint64_t file_number,
-                                     std::set<uint64_t>& pending_outputs,
-                                     autovector<MemTable*>* to_delete,
-                                     Directory* db_directory,
-                                     LogBuffer* log_buffer);
+  Status InstallMemtableFlushResults(
+      ColumnFamilyData* cfd, const MutableCFOptions& mutable_cf_options,
+      const autovector<MemTable*>& m, VersionSet* vset,
+      port::Mutex* mu, Logger* info_log, uint64_t file_number,
+      FileNumToPathIdMap* pending_outputs, autovector<MemTable*>* to_delete,
+      Directory* db_directory, LogBuffer* log_buffer);

  // New memtables are inserted at the front of the list.
  // Takes ownership of the referenced held on *m by the caller of Add().
@@ -24,10 +24,12 @@ void MergeHelper::MergeUntil(Iterator* iter, SequenceNumber stop_before,
                             bool at_bottom, Statistics* stats, int* steps) {
  // Get a copy of the internal key, before it's invalidated by iter->Next()
  // Also maintain the list of merge operands seen.
+  assert(HasOperator());
  keys_.clear();
  operands_.clear();
  keys_.push_front(iter->key().ToString());
  operands_.push_front(iter->value().ToString());
+  assert(user_merge_operator_);

  success_ = false;   // Will become true if we hit Put/Delete or bottom

@@ -78,13 +78,16 @@ class MergeHelper {
  //   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() {
+  bool IsSuccess() const { return success_; }
+  Slice key() const { assert(success_); return Slice(keys_.back()); }
+  Slice value() const { assert(success_); return Slice(operands_.back()); }
+  const std::deque<std::string>& keys() const {
+    assert(!success_); return keys_;
+  }
+  const std::deque<std::string>& values() const {
    assert(!success_); return operands_;
  }
+  bool HasOperator() const { return user_merge_operator_ != nullptr; }

 private:
  const Comparator* user_comparator_;
@@ -12,12 +12,12 @@
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
 #include "rocksdb/merge_operator.h"
+#include "rocksdb/utilities/db_ttl.h"
 #include "db/dbformat.h"
 #include "db/db_impl.h"
 #include "db/write_batch_internal.h"
 #include "utilities/merge_operators.h"
 #include "util/testharness.h"
-#include "utilities/db_ttl.h"

 using namespace std;
 using namespace rocksdb;
@@ -212,7 +212,7 @@ class Counters {
  void assert_add(const string& key, uint64_t value) {
    int result = add(key, value);
    assert(result);
-    if (result == 0) exit(1); // Disable unused variable warning. 
+    if (result == 0) exit(1); // Disable unused variable warning.
  }
 };

@@ -460,6 +460,41 @@ void runTest(int argc, const string& dbname, const bool use_ttl = false) {
      }
    }
  }
+
+  {
+    cout << "Test merge-operator not set after reopen\n";
+    {
+      auto db = OpenDb(dbname);
+      MergeBasedCounters counters(db, 0);
+      counters.add("test-key", 1);
+      counters.add("test-key", 1);
+      counters.add("test-key", 1);
+      db->CompactRange(nullptr, nullptr);
+    }
+
+    DB* reopen_db;
+    ASSERT_OK(DB::Open(Options(), dbname, &reopen_db));
+    std::string value;
+    ASSERT_TRUE(!(reopen_db->Get(ReadOptions(), "test-key", &value).ok()));
+    delete reopen_db;
+    DestroyDB(dbname, Options());
+  }
+
+  /* Temporary remove this test
+  {
+    cout << "Test merge-operator not set after reopen (recovery case)\n";
+    {
+      auto db = OpenDb(dbname);
+      MergeBasedCounters counters(db, 0);
+      counters.add("test-key", 1);
+      counters.add("test-key", 1);
+      counters.add("test-key", 1);
+    }
+
+    DB* reopen_db;
+    ASSERT_TRUE(DB::Open(Options(), dbname, &reopen_db).IsInvalidArgument());
+  }
+  */
 }
 }  // namespace

@@ -6,7 +6,6 @@
 #include <algorithm>
 #include <iostream>
 #include <vector>
-#include "/usr/include/valgrind/callgrind.h"

 #include "rocksdb/db.h"
 #include "rocksdb/perf_context.h"
@@ -29,7 +28,7 @@ const std::string kDbName = rocksdb::test::TmpDir() + "/perf_context_test";

 namespace rocksdb {

-std::shared_ptr<DB> OpenDb() {
+std::shared_ptr<DB> OpenDb(bool read_only = false) {
    DB* db;
    Options options;
    options.create_if_missing = true;
@@ -39,12 +38,16 @@ std::shared_ptr<DB> OpenDb() {
      FLAGS_min_write_buffer_number_to_merge;

    if (FLAGS_use_set_based_memetable) {
-      auto prefix_extractor = rocksdb::NewFixedPrefixTransform(0);
-      options.memtable_factory.reset(
-          NewHashSkipListRepFactory(prefix_extractor));
+      options.prefix_extractor.reset(rocksdb::NewFixedPrefixTransform(0));
+      options.memtable_factory.reset(NewHashSkipListRepFactory());
    }

-    Status s = DB::Open(options, kDbName,  &db);
+    Status s;
+    if (!read_only) {
+      s = DB::Open(options, kDbName, &db);
+    } else {
+      s = DB::OpenForReadOnly(options, kDbName, &db);
+    }
    ASSERT_OK(s);
    return std::shared_ptr<DB>(db);
 }
@@ -76,7 +79,8 @@ TEST(PerfContextTest, SeekIntoDeletion) {
    std::string value;

    perf_context.Reset();
-    StopWatchNano timer(Env::Default(), true);
+    StopWatchNano timer(Env::Default());
+    timer.Start();
    auto status = db->Get(read_options, key, &value);
    auto elapsed_nanos = timer.ElapsedNanos();
    ASSERT_TRUE(status.IsNotFound());
@@ -149,11 +153,12 @@ TEST(PerfContextTest, StopWatchNanoOverhead) {
 TEST(PerfContextTest, StopWatchOverhead) {
  // profile the timer cost by itself!
  const int kTotalIterations = 1000000;
+  uint64_t elapsed = 0;
  std::vector<uint64_t> timings(kTotalIterations);

-  StopWatch timer(Env::Default());
+  StopWatch timer(Env::Default(), nullptr, 0, &elapsed);
  for (auto& timing : timings) {
-    timing = timer.ElapsedMicros();
+    timing = elapsed;
  }

  HistogramImpl histogram;
@@ -166,7 +171,7 @@ TEST(PerfContextTest, StopWatchOverhead) {
  std::cout << histogram.ToString();
 }

-void ProfileKeyComparison() {
+void ProfileQueries(bool enabled_time = false) {
  DestroyDB(kDbName, Options());    // Start this test with a fresh DB

  auto db = OpenDb();
@@ -175,11 +180,21 @@ void ProfileKeyComparison() {
  ReadOptions read_options;

  HistogramImpl hist_put;
+
  HistogramImpl hist_get;
  HistogramImpl hist_get_snapshot;
  HistogramImpl hist_get_memtable;
+  HistogramImpl hist_get_files;
  HistogramImpl hist_get_post_process;
  HistogramImpl hist_num_memtable_checked;
+
+  HistogramImpl hist_mget;
+  HistogramImpl hist_mget_snapshot;
+  HistogramImpl hist_mget_memtable;
+  HistogramImpl hist_mget_files;
+  HistogramImpl hist_mget_post_process;
+  HistogramImpl hist_mget_num_memtable_checked;
+
  HistogramImpl hist_write_pre_post;
  HistogramImpl hist_write_wal_time;
  HistogramImpl hist_write_memtable_time;
@@ -187,8 +202,13 @@ void ProfileKeyComparison() {
  std::cout << "Inserting " << FLAGS_total_keys << " key/value pairs\n...\n";

  std::vector<int> keys;
+  const int kFlushFlag = -1;
  for (int i = 0; i < FLAGS_total_keys; ++i) {
    keys.push_back(i);
+    if (i == FLAGS_total_keys / 2) {
+      // Issuing a flush in the middle.
+      keys.push_back(kFlushFlag);
+    }
  }

  if (FLAGS_random_key) {
@@ -196,27 +216,54 @@ void ProfileKeyComparison() {
  }

  for (const int i : keys) {
+    if (i == kFlushFlag) {
+      FlushOptions fo;
+      db->Flush(fo);
+      continue;
+    }
    std::string key = "k" + std::to_string(i);
    std::string value = "v" + std::to_string(i);

+    std::vector<Slice> keys = {Slice(key)};
+    std::vector<std::string> values;
+
    perf_context.Reset();
    db->Put(write_options, key, value);
    hist_write_pre_post.Add(perf_context.write_pre_and_post_process_time);
    hist_write_wal_time.Add(perf_context.write_wal_time);
    hist_write_memtable_time.Add(perf_context.write_memtable_time);
    hist_put.Add(perf_context.user_key_comparison_count);
+  }
+
+  for (const int i : keys) {
+    std::string key = "k" + std::to_string(i);
+    std::string value = "v" + std::to_string(i);
+
+    std::vector<Slice> keys = {Slice(key)};
+    std::vector<std::string> values;

    perf_context.Reset();
    db->Get(read_options, key, &value);
    hist_get_snapshot.Add(perf_context.get_snapshot_time);
    hist_get_memtable.Add(perf_context.get_from_memtable_time);
+    hist_get_files.Add(perf_context.get_from_output_files_time);
    hist_num_memtable_checked.Add(perf_context.get_from_memtable_count);
    hist_get_post_process.Add(perf_context.get_post_process_time);
    hist_get.Add(perf_context.user_key_comparison_count);
+
+    perf_context.Reset();
+    db->MultiGet(read_options, keys, &values);
+    hist_mget_snapshot.Add(perf_context.get_snapshot_time);
+    hist_mget_memtable.Add(perf_context.get_from_memtable_time);
+    hist_mget_files.Add(perf_context.get_from_output_files_time);
+    hist_mget_num_memtable_checked.Add(perf_context.get_from_memtable_count);
+    hist_mget_post_process.Add(perf_context.get_post_process_time);
+    hist_mget.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();
+            << "Get uesr key comparison: \n" << hist_get.ToString()
+            << "MultiGet uesr key comparison: \n" << hist_get.ToString();
  std::cout << "Put(): Pre and Post Process Time: \n"
            << hist_write_pre_post.ToString()
            << " Writing WAL time: \n"
@@ -224,25 +271,139 @@ void ProfileKeyComparison() {
            << " Writing Mem Table time: \n"
            << hist_write_memtable_time.ToString() << "\n";

-  std::cout << "Get(): Time to get snapshot: \n"
+  std::cout << "Get(): Time to get snapshot: \n" << hist_get_snapshot.ToString()
+            << " Time to get value from memtables: \n"
+            << hist_get_memtable.ToString() << "\n"
+            << " Time to get value from output files: \n"
+            << hist_get_files.ToString() << "\n"
+            << " Number of memtables checked: \n"
+            << hist_num_memtable_checked.ToString() << "\n"
+            << " Time to post process: \n" << hist_get_post_process.ToString()
+            << "\n";
+
+  std::cout << "MultiGet(): Time to get snapshot: \n"
+            << hist_mget_snapshot.ToString()
+            << " Time to get value from memtables: \n"
+            << hist_mget_memtable.ToString() << "\n"
+            << " Time to get value from output files: \n"
+            << hist_mget_files.ToString() << "\n"
+            << " Number of memtables checked: \n"
+            << hist_mget_num_memtable_checked.ToString() << "\n"
+            << " Time to post process: \n" << hist_mget_post_process.ToString()
+            << "\n";
+
+  if (enabled_time) {
+    ASSERT_GT(hist_get.Average(), 0);
+    ASSERT_GT(hist_get_snapshot.Average(), 0);
+    ASSERT_GT(hist_get_memtable.Average(), 0);
+    ASSERT_GT(hist_get_files.Average(), 0);
+    ASSERT_GT(hist_get_post_process.Average(), 0);
+    ASSERT_GT(hist_num_memtable_checked.Average(), 0);
+
+    ASSERT_GT(hist_mget.Average(), 0);
+    ASSERT_GT(hist_mget_snapshot.Average(), 0);
+    ASSERT_GT(hist_mget_memtable.Average(), 0);
+    ASSERT_GT(hist_mget_files.Average(), 0);
+    ASSERT_GT(hist_mget_post_process.Average(), 0);
+    ASSERT_GT(hist_mget_num_memtable_checked.Average(), 0);
+  }
+
+  db.reset();
+  db = OpenDb(true);
+
+  hist_get.Clear();
+  hist_get_snapshot.Clear();
+  hist_get_memtable.Clear();
+  hist_get_files.Clear();
+  hist_get_post_process.Clear();
+  hist_num_memtable_checked.Clear();
+
+  hist_mget.Clear();
+  hist_mget_snapshot.Clear();
+  hist_mget_memtable.Clear();
+  hist_mget_files.Clear();
+  hist_mget_post_process.Clear();
+  hist_mget_num_memtable_checked.Clear();
+
+  for (const int i : keys) {
+    std::string key = "k" + std::to_string(i);
+    std::string value = "v" + std::to_string(i);
+
+    std::vector<Slice> keys = {Slice(key)};
+    std::vector<std::string> values;
+
+    perf_context.Reset();
+    db->Get(read_options, key, &value);
+    hist_get_snapshot.Add(perf_context.get_snapshot_time);
+    hist_get_memtable.Add(perf_context.get_from_memtable_time);
+    hist_get_files.Add(perf_context.get_from_output_files_time);
+    hist_num_memtable_checked.Add(perf_context.get_from_memtable_count);
+    hist_get_post_process.Add(perf_context.get_post_process_time);
+    hist_get.Add(perf_context.user_key_comparison_count);
+
+    perf_context.Reset();
+    db->MultiGet(read_options, keys, &values);
+    hist_mget_snapshot.Add(perf_context.get_snapshot_time);
+    hist_mget_memtable.Add(perf_context.get_from_memtable_time);
+    hist_mget_files.Add(perf_context.get_from_output_files_time);
+    hist_mget_num_memtable_checked.Add(perf_context.get_from_memtable_count);
+    hist_mget_post_process.Add(perf_context.get_post_process_time);
+    hist_mget.Add(perf_context.user_key_comparison_count);
+  }
+
+  std::cout << "ReadOnly Get uesr key comparison: \n" << hist_get.ToString()
+            << "ReadOnly MultiGet uesr key comparison: \n"
+            << hist_mget.ToString();
+
+  std::cout << "ReadOnly Get(): Time to get snapshot: \n"
            << hist_get_snapshot.ToString()
            << " Time to get value from memtables: \n"
            << hist_get_memtable.ToString() << "\n"
+            << " Time to get value from output files: \n"
+            << hist_get_files.ToString() << "\n"
            << " Number of memtables checked: \n"
            << hist_num_memtable_checked.ToString() << "\n"
-            << " Time to post process: \n"
-            << hist_get_post_process.ToString() << "\n";
+            << " Time to post process: \n" << hist_get_post_process.ToString()
+            << "\n";
+
+  std::cout << "ReadOnly MultiGet(): Time to get snapshot: \n"
+            << hist_mget_snapshot.ToString()
+            << " Time to get value from memtables: \n"
+            << hist_mget_memtable.ToString() << "\n"
+            << " Time to get value from output files: \n"
+            << hist_mget_files.ToString() << "\n"
+            << " Number of memtables checked: \n"
+            << hist_mget_num_memtable_checked.ToString() << "\n"
+            << " Time to post process: \n" << hist_mget_post_process.ToString()
+            << "\n";
+
+  if (enabled_time) {
+    ASSERT_GT(hist_get.Average(), 0);
+    ASSERT_GT(hist_get_memtable.Average(), 0);
+    ASSERT_GT(hist_get_files.Average(), 0);
+    ASSERT_GT(hist_num_memtable_checked.Average(), 0);
+    // In read-only mode Get(), no super version operation is needed
+    ASSERT_EQ(hist_get_post_process.Average(), 0);
+    ASSERT_EQ(hist_get_snapshot.Average(), 0);
+
+    ASSERT_GT(hist_mget.Average(), 0);
+    ASSERT_GT(hist_mget_snapshot.Average(), 0);
+    ASSERT_GT(hist_mget_memtable.Average(), 0);
+    ASSERT_GT(hist_mget_files.Average(), 0);
+    ASSERT_GT(hist_mget_post_process.Average(), 0);
+    ASSERT_GT(hist_mget_num_memtable_checked.Average(), 0);
+  }
 }

 TEST(PerfContextTest, KeyComparisonCount) {
  SetPerfLevel(kEnableCount);
-  ProfileKeyComparison();
+  ProfileQueries();

  SetPerfLevel(kDisable);
-  ProfileKeyComparison();
+  ProfileQueries();

  SetPerfLevel(kEnableTime);
-  ProfileKeyComparison();
+  ProfileQueries(true);
 }

 // make perf_context_test
@@ -23,6 +23,7 @@
 #include "rocksdb/slice_transform.h"
 #include "rocksdb/table.h"
 #include "table/meta_blocks.h"
+#include "table/bloom_block.h"
 #include "table/plain_table_factory.h"
 #include "table/plain_table_reader.h"
 #include "util/hash.h"
@@ -61,7 +62,20 @@ class PlainTableDBTest {
  // Return the current option configuration.
  Options CurrentOptions() {
    Options options;
-    options.table_factory.reset(NewPlainTableFactory(16, 2, 0.8, 3));
+
+    PlainTableOptions plain_table_options;
+    plain_table_options.user_key_len = 0;
+    plain_table_options.bloom_bits_per_key = 2;
+    plain_table_options.hash_table_ratio = 0.8;
+    plain_table_options.index_sparseness = 3;
+    plain_table_options.huge_page_tlb_size = 0;
+    plain_table_options.encoding_type = kPrefix;
+    plain_table_options.full_scan_mode = false;
+    plain_table_options.store_index_in_file = false;
+
+    options.table_factory.reset(NewPlainTableFactory(plain_table_options));
+    options.memtable_factory.reset(NewHashLinkListRepFactory(4, 0, 3, true));
+
    options.prefix_extractor.reset(NewFixedPrefixTransform(8));
    options.allow_mmap_reads = true;
    return options;
@@ -174,21 +188,43 @@ TEST(PlainTableDBTest, Empty) {
  ASSERT_EQ("NOT_FOUND", Get("0000000000000foo"));
 }

+extern const uint64_t kPlainTableMagicNumber;
+
 class TestPlainTableReader : public PlainTableReader {
 public:
-  TestPlainTableReader(const EnvOptions& storage_options,
+  TestPlainTableReader(const EnvOptions& env_options,
                       const InternalKeyComparator& icomparator,
-                       uint64_t file_size, int bloom_bits_per_key,
-                       double hash_table_ratio, size_t index_sparseness,
+                       EncodingType encoding_type, uint64_t file_size,
+                       int bloom_bits_per_key, double hash_table_ratio,
+                       size_t index_sparseness,
                       const TableProperties* table_properties,
                       unique_ptr<RandomAccessFile>&& file,
-                       const Options& options, bool* expect_bloom_not_match)
-      : PlainTableReader(options, std::move(file), storage_options, icomparator,
-                         file_size, bloom_bits_per_key, hash_table_ratio,
-                         index_sparseness, table_properties, 2 * 1024 * 1024),
+                       const ImmutableCFOptions& ioptions,
+                       bool* expect_bloom_not_match,
+                       bool store_index_in_file)
+      : PlainTableReader(ioptions, std::move(file), env_options, icomparator,
+                         encoding_type, file_size, table_properties),
        expect_bloom_not_match_(expect_bloom_not_match) {
-    Status s = PopulateIndex(const_cast<TableProperties*>(table_properties));
+    Status s = MmapDataFile();
    ASSERT_TRUE(s.ok());
+
+    s = PopulateIndex(const_cast<TableProperties*>(table_properties),
+                      bloom_bits_per_key, hash_table_ratio, index_sparseness,
+                      2 * 1024 * 1024);
+    ASSERT_TRUE(s.ok());
+
+    TableProperties* props = const_cast<TableProperties*>(table_properties);
+    if (store_index_in_file) {
+      auto bloom_version_ptr = props->user_collected_properties.find(
+          PlainTablePropertyNames::kBloomVersion);
+      ASSERT_TRUE(bloom_version_ptr != props->user_collected_properties.end());
+      ASSERT_EQ(bloom_version_ptr->second, std::string("1"));
+      if (ioptions.bloom_locality > 0) {
+        auto num_blocks_ptr = props->user_collected_properties.find(
+            PlainTablePropertyNames::kNumBloomBlocks);
+        ASSERT_TRUE(num_blocks_ptr != props->user_collected_properties.end());
+      }
+    }
  }

  virtual ~TestPlainTableReader() {}
@@ -196,7 +232,11 @@ class TestPlainTableReader : public PlainTableReader {
 private:
  virtual bool MatchBloom(uint32_t hash) const override {
    bool ret = PlainTableReader::MatchBloom(hash);
-    ASSERT_TRUE(!*expect_bloom_not_match_ || !ret);
+    if (*expect_bloom_not_match_) {
+      ASSERT_TRUE(!ret);
+    } else {
+      ASSERT_TRUE(ret);
+    }
    return ret;
  }
  bool* expect_bloom_not_match_;
@@ -206,30 +246,50 @@ extern const uint64_t kPlainTableMagicNumber;
 class TestPlainTableFactory : public PlainTableFactory {
 public:
  explicit TestPlainTableFactory(bool* expect_bloom_not_match,
-                                 uint32_t user_key_len, int bloom_bits_per_key,
-                                 double hash_table_ratio,
-                                 size_t index_sparseness,
-                                 size_t huge_page_tlb_size)
-      : PlainTableFactory(user_key_len, user_key_len, hash_table_ratio,
-                          index_sparseness, huge_page_tlb_size),
-        bloom_bits_per_key_(bloom_bits_per_key),
-        hash_table_ratio_(hash_table_ratio),
-        index_sparseness_(index_sparseness),
+                                 const PlainTableOptions& options)
+      : PlainTableFactory(options),
+        bloom_bits_per_key_(options.bloom_bits_per_key),
+        hash_table_ratio_(options.hash_table_ratio),
+        index_sparseness_(options.index_sparseness),
+        store_index_in_file_(options.store_index_in_file),
        expect_bloom_not_match_(expect_bloom_not_match) {}

-  Status NewTableReader(const Options& options, const EnvOptions& soptions,
+  Status NewTableReader(const ImmutableCFOptions& ioptions,
+                        const EnvOptions& env_options,
                        const InternalKeyComparator& internal_comparator,
                        unique_ptr<RandomAccessFile>&& file, uint64_t file_size,
                        unique_ptr<TableReader>* table) const override {
    TableProperties* props = nullptr;
    auto s = ReadTableProperties(file.get(), file_size, kPlainTableMagicNumber,
-                                 options.env, options.info_log.get(), &props);
+                                 ioptions.env, ioptions.info_log, &props);
    ASSERT_TRUE(s.ok());

+    if (store_index_in_file_) {
+      BlockHandle bloom_block_handle;
+      s = FindMetaBlock(file.get(), file_size, kPlainTableMagicNumber,
+                        ioptions.env, BloomBlockBuilder::kBloomBlock,
+                        &bloom_block_handle);
+      ASSERT_TRUE(s.ok());
+
+      BlockHandle index_block_handle;
+      s = FindMetaBlock(
+          file.get(), file_size, kPlainTableMagicNumber, ioptions.env,
+          PlainTableIndexBuilder::kPlainTableIndexBlock, &index_block_handle);
+      ASSERT_TRUE(s.ok());
+    }
+
+    auto& user_props = props->user_collected_properties;
+    auto encoding_type_prop =
+        user_props.find(PlainTablePropertyNames::kEncodingType);
+    assert(encoding_type_prop != user_props.end());
+    EncodingType encoding_type = static_cast<EncodingType>(
+        DecodeFixed32(encoding_type_prop->second.c_str()));
+
    std::unique_ptr<PlainTableReader> new_reader(new TestPlainTableReader(
-        soptions, internal_comparator, file_size, bloom_bits_per_key_,
-        hash_table_ratio_, index_sparseness_, props, std::move(file), options,
-        expect_bloom_not_match_));
+        env_options, internal_comparator, encoding_type, file_size,
+        bloom_bits_per_key_, hash_table_ratio_, index_sparseness_, props,
+        std::move(file), ioptions, expect_bloom_not_match_,
+        store_index_in_file_));

    *table = std::move(new_reader);
    return s;
@@ -239,44 +299,92 @@ class TestPlainTableFactory : public PlainTableFactory {
  int bloom_bits_per_key_;
  double hash_table_ratio_;
  size_t index_sparseness_;
+  bool store_index_in_file_;
  bool* expect_bloom_not_match_;
 };

 TEST(PlainTableDBTest, Flush) {
  for (size_t huge_page_tlb_size = 0; huge_page_tlb_size <= 2 * 1024 * 1024;
       huge_page_tlb_size += 2 * 1024 * 1024) {
+    for (EncodingType encoding_type : {kPlain, kPrefix}) {
    for (int bloom_bits = 0; bloom_bits <= 117; bloom_bits += 117) {
      for (int total_order = 0; total_order <= 1; total_order++) {
-        Options options = CurrentOptions();
-        options.create_if_missing = true;
-        // Set only one bucket to force bucket conflict.
-        // Test index interval for the same prefix to be 1, 2 and 4
-        if (total_order) {
-          options.table_factory.reset(NewTotalOrderPlainTableFactory(
-              16, bloom_bits, 2, huge_page_tlb_size));
-        } else {
-          options.table_factory.reset(NewPlainTableFactory(
-              16, bloom_bits, 0.75, 16, huge_page_tlb_size));
+        for (int store_index_in_file = 0; store_index_in_file <= 1;
+             ++store_index_in_file) {
+          if (!bloom_bits && store_index_in_file) {
+            continue;
+          }
+
+          Options options = CurrentOptions();
+          options.create_if_missing = true;
+          // Set only one bucket to force bucket conflict.
+          // Test index interval for the same prefix to be 1, 2 and 4
+          if (total_order) {
+            options.prefix_extractor.reset();
+
+            PlainTableOptions plain_table_options;
+            plain_table_options.user_key_len = 0;
+            plain_table_options.bloom_bits_per_key = bloom_bits;
+            plain_table_options.hash_table_ratio = 0;
+            plain_table_options.index_sparseness = 2;
+            plain_table_options.huge_page_tlb_size = huge_page_tlb_size;
+            plain_table_options.encoding_type = encoding_type;
+            plain_table_options.full_scan_mode = false;
+            plain_table_options.store_index_in_file = store_index_in_file;
+
+            options.table_factory.reset(
+                NewPlainTableFactory(plain_table_options));
+          } else {
+            PlainTableOptions plain_table_options;
+            plain_table_options.user_key_len = 0;
+            plain_table_options.bloom_bits_per_key = bloom_bits;
+            plain_table_options.hash_table_ratio = 0.75;
+            plain_table_options.index_sparseness = 16;
+            plain_table_options.huge_page_tlb_size = huge_page_tlb_size;
+            plain_table_options.encoding_type = encoding_type;
+            plain_table_options.full_scan_mode = false;
+            plain_table_options.store_index_in_file = store_index_in_file;
+
+            options.table_factory.reset(
+                NewPlainTableFactory(plain_table_options));
+          }
+          DestroyAndReopen(&options);
+          uint64_t int_num;
+          ASSERT_TRUE(dbfull()->GetIntProperty(
+              "rocksdb.estimate-table-readers-mem", &int_num));
+          ASSERT_EQ(int_num, 0U);
+
+          ASSERT_OK(Put("1000000000000foo", "v1"));
+          ASSERT_OK(Put("0000000000000bar", "v2"));
+          ASSERT_OK(Put("1000000000000foo", "v3"));
+          dbfull()->TEST_FlushMemTable();
+
+          ASSERT_TRUE(dbfull()->GetIntProperty(
+              "rocksdb.estimate-table-readers-mem", &int_num));
+          ASSERT_GT(int_num, 0U);
+
+          TablePropertiesCollection ptc;
+          reinterpret_cast<DB*>(dbfull())->GetPropertiesOfAllTables(&ptc);
+          ASSERT_EQ(1U, ptc.size());
+          auto row = ptc.begin();
+          auto tp = row->second;
+
+          if (!store_index_in_file) {
+            ASSERT_EQ(total_order ? "4" : "12",
+                      (tp->user_collected_properties)
+                          .at("plain_table_hash_table_size"));
+            ASSERT_EQ("0", (tp->user_collected_properties)
+                               .at("plain_table_sub_index_size"));
+          } else {
+            ASSERT_EQ("0", (tp->user_collected_properties)
+                               .at("plain_table_hash_table_size"));
+            ASSERT_EQ("0", (tp->user_collected_properties)
+                               .at("plain_table_sub_index_size"));
+          }
+          ASSERT_EQ("v3", Get("1000000000000foo"));
+          ASSERT_EQ("v2", Get("0000000000000bar"));
+        }
        }
-        DestroyAndReopen(&options);
-
-        ASSERT_OK(Put("1000000000000foo", "v1"));
-        ASSERT_OK(Put("0000000000000bar", "v2"));
-        ASSERT_OK(Put("1000000000000foo", "v3"));
-        dbfull()->TEST_FlushMemTable();
-
-        TablePropertiesCollection ptc;
-        reinterpret_cast<DB*>(dbfull())->GetPropertiesOfAllTables(&ptc);
-        ASSERT_EQ(1U, ptc.size());
-        auto row = ptc.begin();
-        auto tp = row->second;
-        ASSERT_EQ(total_order ? "4" : "12", (tp->user_collected_properties).at(
-                                                "plain_table_hash_table_size"));
-        ASSERT_EQ(total_order ? "9" : "0", (tp->user_collected_properties).at(
-                                               "plain_table_sub_index_size"));
-
-        ASSERT_EQ("v3", Get("1000000000000foo"));
-        ASSERT_EQ("v2", Get("0000000000000bar"));
      }
    }
  }
@@ -285,23 +393,42 @@ TEST(PlainTableDBTest, Flush) {
 TEST(PlainTableDBTest, Flush2) {
  for (size_t huge_page_tlb_size = 0; huge_page_tlb_size <= 2 * 1024 * 1024;
       huge_page_tlb_size += 2 * 1024 * 1024) {
+    for (EncodingType encoding_type : {kPlain, kPrefix}) {
    for (int bloom_bits = 0; bloom_bits <= 117; bloom_bits += 117) {
      for (int total_order = 0; total_order <= 1; total_order++) {
+        for (int store_index_in_file = 0; store_index_in_file <= 1;
+             ++store_index_in_file) {
+          if (encoding_type == kPrefix && total_order) {
+            continue;
+          }
+          if (!bloom_bits && store_index_in_file) {
+            continue;
+          }
+          if (total_order && store_index_in_file) {
+          continue;
+        }
        bool expect_bloom_not_match = false;
        Options options = CurrentOptions();
        options.create_if_missing = true;
        // Set only one bucket to force bucket conflict.
        // Test index interval for the same prefix to be 1, 2 and 4
+        PlainTableOptions plain_table_options;
        if (total_order) {
          options.prefix_extractor = nullptr;
-          options.table_factory.reset(
-              new TestPlainTableFactory(&expect_bloom_not_match, 16, bloom_bits,
-                                        0, 2, huge_page_tlb_size));
+          plain_table_options.hash_table_ratio = 0;
+          plain_table_options.index_sparseness = 2;
        } else {
-          options.table_factory.reset(
-              new TestPlainTableFactory(&expect_bloom_not_match, 16, bloom_bits,
-                                        0.75, 16, huge_page_tlb_size));
+          plain_table_options.hash_table_ratio = 0.75;
+          plain_table_options.index_sparseness = 16;
        }
+        plain_table_options.user_key_len = kPlainTableVariableLength;
+        plain_table_options.bloom_bits_per_key = bloom_bits;
+        plain_table_options.huge_page_tlb_size = huge_page_tlb_size;
+        plain_table_options.encoding_type = encoding_type;
+        plain_table_options.store_index_in_file = store_index_in_file;
+        options.table_factory.reset(new TestPlainTableFactory(
+            &expect_bloom_not_match, plain_table_options));
+
        DestroyAndReopen(&options);
        ASSERT_OK(Put("0000000000000bar", "b"));
        ASSERT_OK(Put("1000000000000foo", "v1"));
@@ -329,7 +456,6 @@ TEST(PlainTableDBTest, Flush2) {
          // Neither key nor value should exist.
          expect_bloom_not_match = true;
          ASSERT_EQ("NOT_FOUND", Get("5_not00000000bar"));
-
          // Key doesn't exist any more but prefix exists.
          if (total_order) {
            ASSERT_EQ("NOT_FOUND", Get("1000000000000not"));
@@ -338,6 +464,8 @@ TEST(PlainTableDBTest, Flush2) {
          expect_bloom_not_match = false;
        }
      }
+      }
+    }
    }
  }
 }
@@ -345,8 +473,12 @@ TEST(PlainTableDBTest, Flush2) {
 TEST(PlainTableDBTest, Iterator) {
  for (size_t huge_page_tlb_size = 0; huge_page_tlb_size <= 2 * 1024 * 1024;
       huge_page_tlb_size += 2 * 1024 * 1024) {
+    for (EncodingType encoding_type : {kPlain, kPrefix}) {
    for (int bloom_bits = 0; bloom_bits <= 117; bloom_bits += 117) {
      for (int total_order = 0; total_order <= 1; total_order++) {
+        if (encoding_type == kPrefix && total_order == 1) {
+          continue;
+        }
        bool expect_bloom_not_match = false;
        Options options = CurrentOptions();
        options.create_if_missing = true;
@@ -354,13 +486,28 @@ TEST(PlainTableDBTest, Iterator) {
        // Test index interval for the same prefix to be 1, 2 and 4
        if (total_order) {
          options.prefix_extractor = nullptr;
-          options.table_factory.reset(
-              new TestPlainTableFactory(&expect_bloom_not_match, 16, bloom_bits,
-                                        0, 2, huge_page_tlb_size));
+
+          PlainTableOptions plain_table_options;
+          plain_table_options.user_key_len = 16;
+          plain_table_options.bloom_bits_per_key = bloom_bits;
+          plain_table_options.hash_table_ratio = 0;
+          plain_table_options.index_sparseness = 2;
+          plain_table_options.huge_page_tlb_size = huge_page_tlb_size;
+          plain_table_options.encoding_type = encoding_type;
+
+          options.table_factory.reset(new TestPlainTableFactory(
+              &expect_bloom_not_match, plain_table_options));
        } else {
-          options.table_factory.reset(
-              new TestPlainTableFactory(&expect_bloom_not_match, 16, bloom_bits,
-                                        0.75, 16, huge_page_tlb_size));
+          PlainTableOptions plain_table_options;
+          plain_table_options.user_key_len = 16;
+          plain_table_options.bloom_bits_per_key = bloom_bits;
+          plain_table_options.hash_table_ratio = 0.75;
+          plain_table_options.index_sparseness = 16;
+          plain_table_options.huge_page_tlb_size = huge_page_tlb_size;
+          plain_table_options.encoding_type = encoding_type;
+
+          options.table_factory.reset(new TestPlainTableFactory(
+              &expect_bloom_not_match, plain_table_options));
        }
        DestroyAndReopen(&options);

@@ -447,6 +594,7 @@ TEST(PlainTableDBTest, Iterator) {
        delete iter;
      }
    }
+    }
  }
 }

@@ -458,7 +606,13 @@ std::string MakeLongKey(size_t length, char c) {

 TEST(PlainTableDBTest, IteratorLargeKeys) {
  Options options = CurrentOptions();
-  options.table_factory.reset(NewTotalOrderPlainTableFactory(0, 0, 16));
+
+  PlainTableOptions plain_table_options;
+  plain_table_options.user_key_len = 0;
+  plain_table_options.bloom_bits_per_key = 0;
+  plain_table_options.hash_table_ratio = 0;
+
+  options.table_factory.reset(NewPlainTableFactory(plain_table_options));
  options.create_if_missing = true;
  options.prefix_extractor.reset();
  DestroyAndReopen(&options);
@@ -494,6 +648,54 @@ TEST(PlainTableDBTest, IteratorLargeKeys) {
  delete iter;
 }

+namespace {
+std::string MakeLongKeyWithPrefix(size_t length, char c) {
+  return "00000000" + std::string(length - 8, c);
+}
+}  // namespace
+
+TEST(PlainTableDBTest, IteratorLargeKeysWithPrefix) {
+  Options options = CurrentOptions();
+
+  PlainTableOptions plain_table_options;
+  plain_table_options.user_key_len = 16;
+  plain_table_options.bloom_bits_per_key = 0;
+  plain_table_options.hash_table_ratio = 0.8;
+  plain_table_options.index_sparseness = 3;
+  plain_table_options.huge_page_tlb_size = 0;
+  plain_table_options.encoding_type = kPrefix;
+
+  options.table_factory.reset(NewPlainTableFactory(plain_table_options));
+  options.create_if_missing = true;
+  DestroyAndReopen(&options);
+
+  std::string key_list[] = {
+      MakeLongKeyWithPrefix(30, '0'), MakeLongKeyWithPrefix(16, '1'),
+      MakeLongKeyWithPrefix(32, '2'), MakeLongKeyWithPrefix(60, '3'),
+      MakeLongKeyWithPrefix(90, '4'), MakeLongKeyWithPrefix(50, '5'),
+      MakeLongKeyWithPrefix(26, '6')};
+
+  for (size_t i = 0; i < 7; i++) {
+    ASSERT_OK(Put(key_list[i], std::to_string(i)));
+  }
+
+  dbfull()->TEST_FlushMemTable();
+
+  Iterator* iter = dbfull()->NewIterator(ReadOptions());
+  iter->Seek(key_list[0]);
+
+  for (size_t i = 0; i < 7; i++) {
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ(key_list[i], iter->key().ToString());
+    ASSERT_EQ(std::to_string(i), iter->value().ToString());
+    iter->Next();
+  }
+
+  ASSERT_TRUE(!iter->Valid());
+
+  delete iter;
+}
+
 // A test comparator which compare two strings in this way:
 // (1) first compare prefix of 8 bytes in alphabet order,
 // (2) if two strings share the same prefix, sort the other part of the string
@@ -599,8 +801,16 @@ TEST(PlainTableDBTest, HashBucketConflict) {
      options.create_if_missing = true;
      // Set only one bucket to force bucket conflict.
      // Test index interval for the same prefix to be 1, 2 and 4
-      options.table_factory.reset(
-          NewTotalOrderPlainTableFactory(16, 0, 2 ^ i, huge_page_tlb_size));
+
+      PlainTableOptions plain_table_options;
+      plain_table_options.user_key_len = 16;
+      plain_table_options.bloom_bits_per_key = 0;
+      plain_table_options.hash_table_ratio = 0;
+      plain_table_options.index_sparseness = 2 ^ i;
+      plain_table_options.huge_page_tlb_size = huge_page_tlb_size;
+
+      options.table_factory.reset(NewPlainTableFactory(plain_table_options));
+
      DestroyAndReopen(&options);
      ASSERT_OK(Put("5000000000000fo0", "v1"));
      ASSERT_OK(Put("5000000000000fo1", "v2"));
@@ -686,8 +896,15 @@ TEST(PlainTableDBTest, HashBucketConflictReverseSuffixComparator) {
      options.comparator = &comp;
      // Set only one bucket to force bucket conflict.
      // Test index interval for the same prefix to be 1, 2 and 4
-      options.table_factory.reset(
-          NewTotalOrderPlainTableFactory(16, 0, 2 ^ i, huge_page_tlb_size));
+
+      PlainTableOptions plain_table_options;
+      plain_table_options.user_key_len = 16;
+      plain_table_options.bloom_bits_per_key = 0;
+      plain_table_options.hash_table_ratio = 0;
+      plain_table_options.index_sparseness = 2 ^ i;
+      plain_table_options.huge_page_tlb_size = huge_page_tlb_size;
+
+      options.table_factory.reset(NewPlainTableFactory(plain_table_options));
      DestroyAndReopen(&options);
      ASSERT_OK(Put("5000000000000fo0", "v1"));
      ASSERT_OK(Put("5000000000000fo1", "v2"));
@@ -767,7 +984,13 @@ TEST(PlainTableDBTest, NonExistingKeyToNonEmptyBucket) {
  options.create_if_missing = true;
  // Set only one bucket to force bucket conflict.
  // Test index interval for the same prefix to be 1, 2 and 4
-  options.table_factory.reset(NewTotalOrderPlainTableFactory(16, 0, 5));
+  PlainTableOptions plain_table_options;
+  plain_table_options.user_key_len = 16;
+  plain_table_options.bloom_bits_per_key = 0;
+  plain_table_options.hash_table_ratio = 0;
+  plain_table_options.index_sparseness = 5;
+
+  options.table_factory.reset(NewPlainTableFactory(plain_table_options));
  DestroyAndReopen(&options);
  ASSERT_OK(Put("5000000000000fo0", "v1"));
  ASSERT_OK(Put("5000000000000fo1", "v2"));
@@ -846,6 +1069,49 @@ TEST(PlainTableDBTest, CompactionTrigger) {
  ASSERT_EQ(NumTableFilesAtLevel(1), 1);
 }

+TEST(PlainTableDBTest, AdaptiveTable) {
+  Options options = CurrentOptions();
+  options.create_if_missing = true;
+
+  options.table_factory.reset(NewPlainTableFactory());
+  DestroyAndReopen(&options);
+
+  ASSERT_OK(Put("1000000000000foo", "v1"));
+  ASSERT_OK(Put("0000000000000bar", "v2"));
+  ASSERT_OK(Put("1000000000000foo", "v3"));
+  dbfull()->TEST_FlushMemTable();
+
+  options.create_if_missing = false;
+  std::shared_ptr<TableFactory> dummy_factory;
+  std::shared_ptr<TableFactory> block_based_factory(
+      NewBlockBasedTableFactory());
+  options.table_factory.reset(NewAdaptiveTableFactory(
+      block_based_factory, dummy_factory, dummy_factory));
+  Reopen(&options);
+  ASSERT_EQ("v3", Get("1000000000000foo"));
+  ASSERT_EQ("v2", Get("0000000000000bar"));
+
+  ASSERT_OK(Put("2000000000000foo", "v4"));
+  ASSERT_OK(Put("3000000000000bar", "v5"));
+  dbfull()->TEST_FlushMemTable();
+  ASSERT_EQ("v4", Get("2000000000000foo"));
+  ASSERT_EQ("v5", Get("3000000000000bar"));
+
+  Reopen(&options);
+  ASSERT_EQ("v3", Get("1000000000000foo"));
+  ASSERT_EQ("v2", Get("0000000000000bar"));
+  ASSERT_EQ("v4", Get("2000000000000foo"));
+  ASSERT_EQ("v5", Get("3000000000000bar"));
+
+  options.table_factory.reset(NewBlockBasedTableFactory());
+  Reopen(&options);
+  ASSERT_NE("v3", Get("1000000000000foo"));
+
+  options.table_factory.reset(NewPlainTableFactory());
+  Reopen(&options);
+  ASSERT_NE("v5", Get("3000000000000bar"));
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -189,6 +189,10 @@ class PrefixTest {
          options.memtable_factory.reset(
              NewHashLinkListRepFactory(bucket_count, 2 * 1024 * 1024));
          return true;
+        case kHashLinkListTriggerSkipList:
+          options.memtable_factory.reset(
+              NewHashLinkListRepFactory(bucket_count, 0, 3));
+          return true;
        default:
          return false;
      }
@@ -208,6 +212,7 @@ class PrefixTest {
    kHashSkipList,
    kHashLinkList,
    kHashLinkListHugePageTlb,
+    kHashLinkListTriggerSkipList,
    kEnd
  };
  int option_config_;
@@ -31,6 +31,11 @@

 #ifndef ROCKSDB_LITE

+#ifndef __STDC_FORMAT_MACROS
+#define __STDC_FORMAT_MACROS
+#endif
+
+#include <inttypes.h>
 #include "db/builder.h"
 #include "db/db_impl.h"
 #include "db/dbformat.h"
@@ -44,6 +49,9 @@
 #include "rocksdb/comparator.h"
 #include "rocksdb/db.h"
 #include "rocksdb/env.h"
+#include "rocksdb/options.h"
+#include "rocksdb/immutable_options.h"
+#include "util/scoped_arena_iterator.h"

 namespace rocksdb {

@@ -55,16 +63,16 @@ class Repairer {
      : dbname_(dbname),
        env_(options.env),
        icmp_(options.comparator),
-        ipolicy_(options.filter_policy),
-        options_(SanitizeOptions(dbname, &icmp_, &ipolicy_, options)),
+        options_(SanitizeOptions(dbname, &icmp_, options)),
+        ioptions_(options_),
        raw_table_cache_(
            // TableCache can be small since we expect each table to be opened
            // once.
            NewLRUCache(10, options_.table_cache_numshardbits,
                        options_.table_cache_remove_scan_count_limit)),
        next_file_number_(1) {
-    table_cache_ = new TableCache(dbname_, &options_, storage_options_,
-                                  raw_table_cache_.get());
+    table_cache_ =
+        new TableCache(ioptions_, env_options_, raw_table_cache_.get());
    edit_ = new VersionEdit();
  }

@@ -82,18 +90,17 @@ class Repairer {
      status = WriteDescriptor();
    }
    if (status.ok()) {
-      unsigned long long bytes = 0;
+      uint64_t bytes = 0;
      for (size_t i = 0; i < tables_.size(); i++) {
-        bytes += tables_[i].meta.file_size;
+        bytes += tables_[i].meta.fd.GetFileSize();
      }
      Log(options_.info_log,
          "**** Repaired rocksdb %s; "
-          "recovered %d files; %llu bytes. "
+          "recovered %zu files; %" PRIu64
+          "bytes. "
          "Some data may have been lost. "
          "****",
-          dbname_.c_str(),
-          static_cast<int>(tables_.size()),
-          bytes);
+          dbname_.c_str(), tables_.size(), bytes);
    }
    return status;
  }
@@ -107,51 +114,60 @@ class Repairer {

  std::string const dbname_;
  Env* const env_;
-  InternalKeyComparator const icmp_;
-  InternalFilterPolicy const ipolicy_;
-  Options const options_;
+  const InternalKeyComparator icmp_;
+  const Options options_;
+  const ImmutableCFOptions ioptions_;
  std::shared_ptr<Cache> raw_table_cache_;
  TableCache* table_cache_;
  VersionEdit* edit_;

  std::vector<std::string> manifests_;
-  std::vector<uint64_t> table_numbers_;
+  std::vector<FileDescriptor> table_fds_;
  std::vector<uint64_t> logs_;
  std::vector<TableInfo> tables_;
  uint64_t next_file_number_;
-  const EnvOptions storage_options_;
+  const EnvOptions env_options_;

  Status FindFiles() {
    std::vector<std::string> filenames;
-    Status status = env_->GetChildren(dbname_, &filenames);
-    if (!status.ok()) {
-      return status;
-    }
-    if (filenames.empty()) {
-      return Status::Corruption(dbname_, "repair found no files");
-    }
+    bool found_file = false;
+    for (uint32_t path_id = 0; path_id < options_.db_paths.size(); path_id++) {
+      Status status =
+          env_->GetChildren(options_.db_paths[path_id].path, &filenames);
+      if (!status.ok()) {
+        return status;
+      }
+      if (!filenames.empty()) {
+        found_file = true;
+      }

-    uint64_t number;
-    FileType type;
-    for (size_t i = 0; i < filenames.size(); i++) {
-      if (ParseFileName(filenames[i], &number, &type)) {
-        if (type == kDescriptorFile) {
-          manifests_.push_back(filenames[i]);
-        } else {
-          if (number + 1 > next_file_number_) {
-            next_file_number_ = number + 1;
-          }
-          if (type == kLogFile) {
-            logs_.push_back(number);
-          } else if (type == kTableFile) {
-            table_numbers_.push_back(number);
+      uint64_t number;
+      FileType type;
+      for (size_t i = 0; i < filenames.size(); i++) {
+        if (ParseFileName(filenames[i], &number, &type)) {
+          if (type == kDescriptorFile) {
+            assert(path_id == 0);
+            manifests_.push_back(filenames[i]);
          } else {
-            // Ignore other files
+            if (number + 1 > next_file_number_) {
+              next_file_number_ = number + 1;
+            }
+            if (type == kLogFile) {
+              assert(path_id == 0);
+              logs_.push_back(number);
+            } else if (type == kTableFile) {
+              table_fds_.emplace_back(number, path_id, 0);
+            } else {
+              // Ignore other files
+            }
          }
        }
      }
    }
-    return status;
+    if (!found_file) {
+      return Status::Corruption(dbname_, "repair found no files");
+    }
+    return Status::OK();
  }

  void ConvertLogFilesToTables() {
@@ -159,8 +175,8 @@ class Repairer {
      std::string logname = LogFileName(dbname_, logs_[i]);
      Status status = ConvertLogToTable(logs_[i]);
      if (!status.ok()) {
-        Log(options_.info_log, "Log #%llu: ignoring conversion error: %s",
-            (unsigned long long) logs_[i],
+        Log(options_.info_log,
+            "Log #%" PRIu64 ": ignoring conversion error: %s", logs_[i],
            status.ToString().c_str());
      }
      ArchiveFile(logname);
@@ -174,17 +190,15 @@ class Repairer {
      uint64_t lognum;
      virtual void Corruption(size_t bytes, const Status& s) {
        // We print error messages for corruption, but continue repairing.
-        Log(info_log, "Log #%llu: dropping %d bytes; %s",
-            (unsigned long long) lognum,
-            static_cast<int>(bytes),
-            s.ToString().c_str());
+        Log(info_log, "Log #%" PRIu64 ": dropping %d bytes; %s", lognum,
+            static_cast<int>(bytes), s.ToString().c_str());
      }
    };

    // Open the log file
    std::string logname = LogFileName(dbname_, log);
    unique_ptr<SequentialFile> lfile;
-    Status status = env_->NewSequentialFile(logname, &lfile, storage_options_);
+    Status status = env_->NewSequentialFile(logname, &lfile, env_options_);
    if (!status.ok()) {
      return status;
    }
@@ -205,7 +219,8 @@ class Repairer {
    std::string scratch;
    Slice record;
    WriteBatch batch;
-    MemTable* mem = new MemTable(icmp_, options_);
+    MemTable* mem = new MemTable(icmp_, ioptions_,
+        MemTableOptions(MutableCFOptions(options_, ioptions_), options_));
    auto cf_mems_default = new ColumnFamilyMemTablesDefault(mem, &options_);
    mem->Ref();
    int counter = 0;
@@ -220,8 +235,7 @@ class Repairer {
      if (status.ok()) {
        counter += WriteBatchInternal::Count(&batch);
      } else {
-        Log(options_.info_log, "Log #%llu: ignoring %s",
-            (unsigned long long) log,
+        Log(options_.info_log, "Log #%" PRIu64 ": ignoring %s", log,
            status.ToString().c_str());
        status = Status::OK();  // Keep going with rest of file
      }
@@ -230,37 +244,42 @@ class Repairer {
    // Do not record a version edit for this conversion to a Table
    // since ExtractMetaData() will also generate edits.
    FileMetaData meta;
-    meta.number = next_file_number_++;
-    ReadOptions ro;
-    Iterator* iter = mem->NewIterator(ro, true /* enforce_total_order */);
-    status = BuildTable(dbname_, env_, options_, storage_options_, table_cache_,
-                        iter, &meta, icmp_, 0, 0, kNoCompression);
-    delete iter;
+    meta.fd = FileDescriptor(next_file_number_++, 0, 0);
+    {
+      ReadOptions ro;
+      ro.total_order_seek = true;
+      Arena arena;
+      ScopedArenaIterator iter(mem->NewIterator(ro, &arena));
+      status = BuildTable(dbname_, env_, ioptions_, env_options_, table_cache_,
+                          iter.get(), &meta, icmp_, 0, 0, kNoCompression,
+                          CompressionOptions());
+    }
    delete mem->Unref();
    delete cf_mems_default;
    mem = nullptr;
    if (status.ok()) {
-      if (meta.file_size > 0) {
-        table_numbers_.push_back(meta.number);
+      if (meta.fd.GetFileSize() > 0) {
+        table_fds_.push_back(meta.fd);
      }
    }
-    Log(options_.info_log, "Log #%llu: %d ops saved to Table #%llu %s",
-        (unsigned long long) log,
-        counter,
-        (unsigned long long) meta.number,
-        status.ToString().c_str());
+    Log(options_.info_log,
+        "Log #%" PRIu64 ": %d ops saved to Table #%" PRIu64 " %s", log, counter,
+        meta.fd.GetNumber(), status.ToString().c_str());
    return status;
  }

  void ExtractMetaData() {
-    for (size_t i = 0; i < table_numbers_.size(); i++) {
+    for (size_t i = 0; i < table_fds_.size(); i++) {
      TableInfo t;
-      t.meta.number = table_numbers_[i];
+      t.meta.fd = table_fds_[i];
      Status status = ScanTable(&t);
      if (!status.ok()) {
-        std::string fname = TableFileName(dbname_, table_numbers_[i]);
-        Log(options_.info_log, "Table #%llu: ignoring %s",
-            (unsigned long long) table_numbers_[i],
+        std::string fname = TableFileName(
+            options_.db_paths, t.meta.fd.GetNumber(), t.meta.fd.GetPathId());
+        char file_num_buf[kFormatFileNumberBufSize];
+        FormatFileNumber(t.meta.fd.GetNumber(), t.meta.fd.GetPathId(),
+                         file_num_buf, sizeof(file_num_buf));
+        Log(options_.info_log, "Table #%s: ignoring %s", file_num_buf,
            status.ToString().c_str());
        ArchiveFile(fname);
      } else {
@@ -270,13 +289,16 @@ class Repairer {
  }

  Status ScanTable(TableInfo* t) {
-    std::string fname = TableFileName(dbname_, t->meta.number);
+    std::string fname = TableFileName(options_.db_paths, t->meta.fd.GetNumber(),
+                                      t->meta.fd.GetPathId());
    int counter = 0;
-    Status status = env_->GetFileSize(fname, &t->meta.file_size);
+    uint64_t file_size;
+    Status status = env_->GetFileSize(fname, &file_size);
+    t->meta.fd = FileDescriptor(t->meta.fd.GetNumber(), t->meta.fd.GetPathId(),
+                                file_size);
    if (status.ok()) {
-      FileMetaData dummy_meta(t->meta.number, t->meta.file_size);
      Iterator* iter = table_cache_->NewIterator(
-          ReadOptions(), storage_options_, icmp_, dummy_meta);
+          ReadOptions(), env_options_, icmp_, t->meta.fd);
      bool empty = true;
      ParsedInternalKey parsed;
      t->min_sequence = 0;
@@ -284,9 +306,8 @@ class Repairer {
      for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
        Slice key = iter->key();
        if (!ParseInternalKey(key, &parsed)) {
-          Log(options_.info_log, "Table #%llu: unparsable key %s",
-              (unsigned long long) t->meta.number,
-              EscapeString(key).c_str());
+          Log(options_.info_log, "Table #%" PRIu64 ": unparsable key %s",
+              t->meta.fd.GetNumber(), EscapeString(key).c_str());
          continue;
        }

@@ -308,10 +329,8 @@ class Repairer {
      }
      delete iter;
    }
-    Log(options_.info_log, "Table #%llu: %d entries %s",
-        (unsigned long long) t->meta.number,
-        counter,
-        status.ToString().c_str());
+    Log(options_.info_log, "Table #%" PRIu64 ": %d entries %s",
+        t->meta.fd.GetNumber(), counter, status.ToString().c_str());
    return status;
  }

@@ -319,7 +338,7 @@ class Repairer {
    std::string tmp = TempFileName(dbname_, 1);
    unique_ptr<WritableFile> file;
    Status status = env_->NewWritableFile(
-        tmp, &file, env_->OptimizeForManifestWrite(storage_options_));
+        tmp, &file, env_->OptimizeForManifestWrite(env_options_));
    if (!status.ok()) {
      return status;
    }
@@ -339,9 +358,9 @@ class Repairer {
    for (size_t i = 0; i < tables_.size(); i++) {
      // 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.min_sequence, t.max_sequence);
+      edit_->AddFile(0, t.meta.fd.GetNumber(), t.meta.fd.GetPathId(),
+                     t.meta.fd.GetFileSize(), t.meta.smallest, t.meta.largest,
+                     t.min_sequence, t.max_sequence);
    }

    //fprintf(stderr, "NewDescriptor:\n%s\n", edit_.DebugString().c_str());
@@ -1,794 +0,0 @@
-// 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.
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
-//  This source code is licensed under the BSD-style license found in the
-//  LICENSE file in the root directory of this source tree. An additional grant
-//  of patent rights can be found in the PATENTS file in the same directory.
-//
-// 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 <algorithm>
-#include <set>
-
-#include "rocksdb/db.h"
-#include "rocksdb/filter_policy.h"
-#include "db/db_impl.h"
-#include "db/filename.h"
-#include "db/version_set.h"
-#include "db/write_batch_internal.h"
-#include "rocksdb/statistics.h"
-#include "rocksdb/cache.h"
-#include "rocksdb/compaction_filter.h"
-#include "rocksdb/env.h"
-#include "rocksdb/table.h"
-#include "rocksdb/table_properties.h"
-#include "table/table_builder.h"
-#include "util/hash.h"
-#include "util/logging.h"
-#include "util/mutexlock.h"
-#include "util/testharness.h"
-#include "util/testutil.h"
-#include "utilities/merge_operators.h"
-
-using std::unique_ptr;
-
-// IS THIS FILE STILL NEEDED?
-namespace rocksdb {
-
-// SimpleTable is a simple table format for UNIT TEST ONLY. It is not built
-// as production quality.
-// SimpleTable requires the input key size to be fixed 16 bytes, value cannot
-// be longer than 150000 bytes and stored data on disk in this format:
-// +--------------------------------------------+  <= key1 offset
-// | key1            | value_size (4 bytes) |   |
-// +----------------------------------------+   |
-// | value1                                     |
-// |                                            |
-// +----------------------------------------+---+  <= key2 offset
-// | key2            | value_size (4 bytes) |   |
-// +----------------------------------------+   |
-// | value2                                     |
-// |                                            |
-// |        ......                              |
-// +-----------------+--------------------------+   <= index_block_offset
-// | key1            | key1 offset (8 bytes)    |
-// +-----------------+--------------------------+
-// | key2            | key2 offset (8 bytes)    |
-// +-----------------+--------------------------+
-// | key3            | key3 offset (8 bytes)    |
-// +-----------------+--------------------------+
-// |        ......                              |
-// +-----------------+------------+-------------+
-// | index_block_offset (8 bytes) |
-// +------------------------------+
-
-// SimpleTable is a simple table format for UNIT TEST ONLY. It is not built
-// as production quality.
-class SimpleTableReader: public TableReader {
-public:
-  // Attempt to open the table that is stored in bytes [0..file_size)
-  // of "file", and read the metadata entries necessary to allow
-  // retrieving data from the table.
-  //
-  // If successful, returns ok and sets "*table" to the newly opened
-  // table.  The client should delete "*table" when no longer needed.
-  // If there was an error while initializing the table, sets "*table"
-  // to nullptr and returns a non-ok status.  Does not take ownership of
-  // "*source", but the client must ensure that "source" remains live
-  // for the duration of the returned table's lifetime.
-  //
-  // *file must remain live while this Table is in use.
-  static Status Open(const Options& options, const EnvOptions& soptions,
-                     unique_ptr<RandomAccessFile> && file, uint64_t file_size,
-                     unique_ptr<TableReader>* table_reader);
-
-  Iterator* NewIterator(const ReadOptions&) override;
-
-  Status Get(const ReadOptions&, const Slice& key, void* arg,
-             bool (*handle_result)(void* arg, const ParsedInternalKey& k,
-                                   const Slice& v, bool),
-             void (*mark_key_may_exist)(void*) = nullptr) override;
-
-  uint64_t ApproximateOffsetOf(const Slice& key) override;
-
-  void SetupForCompaction() override;
-
-  std::shared_ptr<const TableProperties> GetTableProperties() const override;
-
-  ~SimpleTableReader();
-
-private:
-  struct Rep;
-  Rep* rep_;
-
-  explicit SimpleTableReader(Rep* rep) {
-    rep_ = rep;
-  }
-  friend class TableCache;
-  friend class SimpleTableIterator;
-
-  Status GetOffset(const Slice& target, uint64_t* offset);
-
-  // No copying allowed
-  explicit SimpleTableReader(const TableReader&) = delete;
-  void operator=(const TableReader&) = delete;
-};
-
-// Iterator to iterate SimpleTable
-class SimpleTableIterator: public Iterator {
-public:
-  explicit SimpleTableIterator(SimpleTableReader* table);
-  ~SimpleTableIterator();
-
-  bool Valid() const;
-
-  void SeekToFirst();
-
-  void SeekToLast();
-
-  void Seek(const Slice& target);
-
-  void Next();
-
-  void Prev();
-
-  Slice key() const;
-
-  Slice value() const;
-
-  Status status() const;
-
-private:
-  SimpleTableReader* table_;
-  uint64_t offset_;
-  uint64_t next_offset_;
-  Slice key_;
-  Slice value_;
-  char tmp_str_[4];
-  char* key_str_;
-  char* value_str_;
-  int value_str_len_;
-  Status status_;
-  // No copying allowed
-  SimpleTableIterator(const SimpleTableIterator&) = delete;
-  void operator=(const Iterator&) = delete;
-};
-
-struct SimpleTableReader::Rep {
-  ~Rep() {
-  }
-  Rep(const EnvOptions& storage_options, uint64_t index_start_offset,
-      int num_entries) :
-      soptions(storage_options), index_start_offset(index_start_offset),
-      num_entries(num_entries) {
-  }
-
-  Options options;
-  const EnvOptions& soptions;
-  Status status;
-  unique_ptr<RandomAccessFile> file;
-  uint64_t index_start_offset;
-  int num_entries;
-  std::shared_ptr<TableProperties> table_properties;
-
-  const static int user_key_size = 16;
-  const static int offset_length = 8;
-  const static int key_footer_len = 8;
-
-  static int GetInternalKeyLength() {
-    return user_key_size + key_footer_len;
-  }
-};
-
-SimpleTableReader::~SimpleTableReader() {
-  delete rep_;
-}
-
-Status SimpleTableReader::Open(const Options& options,
-                               const EnvOptions& soptions,
-                               unique_ptr<RandomAccessFile> && file,
-                               uint64_t size,
-                               unique_ptr<TableReader>* table_reader) {
-  char footer_space[Rep::offset_length];
-  Slice footer_input;
-  Status s = file->Read(size - Rep::offset_length, Rep::offset_length,
-                        &footer_input, footer_space);
-  if (s.ok()) {
-    uint64_t index_start_offset = DecodeFixed64(footer_space);
-
-    int num_entries = (size - Rep::offset_length - index_start_offset)
-        / (Rep::GetInternalKeyLength() + Rep::offset_length);
-    SimpleTableReader::Rep* rep = new SimpleTableReader::Rep(soptions,
-                                                             index_start_offset,
-                                                             num_entries);
-
-    rep->file = std::move(file);
-    rep->options = options;
-    table_reader->reset(new SimpleTableReader(rep));
-  }
-  return s;
-}
-
-void SimpleTableReader::SetupForCompaction() {
-}
-
-std::shared_ptr<const TableProperties> SimpleTableReader::GetTableProperties()
-    const {
-  return rep_->table_properties;
-}
-
-Iterator* SimpleTableReader::NewIterator(const ReadOptions& options) {
-  return new SimpleTableIterator(this);
-}
-
-Status SimpleTableReader::GetOffset(const Slice& target, uint64_t* offset) {
-  uint32_t left = 0;
-  uint32_t right = rep_->num_entries - 1;
-  char key_chars[Rep::GetInternalKeyLength()];
-  Slice tmp_slice;
-
-  uint32_t target_offset = 0;
-  while (left <= right) {
-    uint32_t mid = (left + right + 1) / 2;
-
-    uint64_t offset_to_read = rep_->index_start_offset
-        + (Rep::GetInternalKeyLength() + Rep::offset_length) * mid;
-    Status s = rep_->file->Read(offset_to_read, Rep::GetInternalKeyLength(),
-                                &tmp_slice, key_chars);
-    if (!s.ok()) {
-      return s;
-    }
-
-    InternalKeyComparator ikc(rep_->options.comparator);
-    int compare_result = ikc.Compare(tmp_slice, target);
-
-    if (compare_result < 0) {
-      if (left == right) {
-        target_offset = right + 1;
-        break;
-      }
-      left = mid;
-    } else {
-      if (left == right) {
-        target_offset = left;
-        break;
-      }
-      right = mid - 1;
-    }
-  }
-
-  if (target_offset >= (uint32_t) rep_->num_entries) {
-    *offset = rep_->index_start_offset;
-    return Status::OK();
-  }
-
-  char value_offset_chars[Rep::offset_length];
-
-  int64_t offset_for_value_offset = rep_->index_start_offset
-      + (Rep::GetInternalKeyLength() + Rep::offset_length) * target_offset
-      + Rep::GetInternalKeyLength();
-  Status s = rep_->file->Read(offset_for_value_offset, Rep::offset_length,
-                              &tmp_slice, value_offset_chars);
-  if (s.ok()) {
-    *offset = DecodeFixed64(value_offset_chars);
-  }
-  return s;
-}
-
-Status SimpleTableReader::Get(const ReadOptions& options, const Slice& k,
-                              void* arg,
-                              bool (*saver)(void*, const ParsedInternalKey&,
-                                            const Slice&, bool),
-                              void (*mark_key_may_exist)(void*)) {
-  Status s;
-  SimpleTableIterator* iter = new SimpleTableIterator(this);
-  for (iter->Seek(k); iter->Valid(); iter->Next()) {
-    ParsedInternalKey parsed_key;
-    if (!ParseInternalKey(iter->key(), &parsed_key)) {
-      return Status::Corruption(Slice());
-    }
-
-    if (!(*saver)(arg, parsed_key, iter->value(), true)) {
-      break;
-    }
-  }
-  s = iter->status();
-  delete iter;
-  return s;
-}
-
-uint64_t SimpleTableReader::ApproximateOffsetOf(const Slice& key) {
-  return 0;
-}
-
-SimpleTableIterator::SimpleTableIterator(SimpleTableReader* table) :
-    table_(table) {
-  key_str_ = new char[SimpleTableReader::Rep::GetInternalKeyLength()];
-  value_str_len_ = -1;
-  SeekToFirst();
-}
-
-SimpleTableIterator::~SimpleTableIterator() {
- delete[] key_str_;
- if (value_str_len_ >= 0) {
-   delete[] value_str_;
- }
-}
-
-bool SimpleTableIterator::Valid() const {
-  return offset_ < table_->rep_->index_start_offset;
-}
-
-void SimpleTableIterator::SeekToFirst() {
-  next_offset_ = 0;
-  Next();
-}
-
-void SimpleTableIterator::SeekToLast() {
-  assert(false);
-}
-
-void SimpleTableIterator::Seek(const Slice& target) {
-  Status s = table_->GetOffset(target, &next_offset_);
-  if (!s.ok()) {
-    status_ = s;
-  }
-  Next();
-}
-
-void SimpleTableIterator::Next() {
-  offset_ = next_offset_;
-  if (offset_ >= table_->rep_->index_start_offset) {
-    return;
-  }
-  Slice result;
-  int internal_key_size = SimpleTableReader::Rep::GetInternalKeyLength();
-
-  Status s = table_->rep_->file->Read(next_offset_, internal_key_size, &result,
-                                      key_str_);
-  next_offset_ += internal_key_size;
-  key_ = result;
-
-  Slice value_size_slice;
-  s = table_->rep_->file->Read(next_offset_, 4, &value_size_slice, tmp_str_);
-  next_offset_ += 4;
-  uint32_t value_size = DecodeFixed32(tmp_str_);
-
-  Slice value_slice;
-  if ((int) value_size > value_str_len_) {
-    if (value_str_len_ >= 0) {
-      delete[] value_str_;
-    }
-    value_str_ = new char[value_size];
-    value_str_len_ = value_size;
-  }
-  s = table_->rep_->file->Read(next_offset_, value_size, &value_slice,
-                               value_str_);
-  next_offset_ += value_size;
-  value_ = value_slice;
-}
-
-void SimpleTableIterator::Prev() {
-  assert(false);
-}
-
-Slice SimpleTableIterator::key() const {
-  Log(table_->rep_->options.info_log, "key!!!!");
-  return key_;
-}
-
-Slice SimpleTableIterator::value() const {
-  return value_;
-}
-
-Status SimpleTableIterator::status() const {
-  return status_;
-}
-
-class SimpleTableBuilder: public TableBuilder {
-public:
-  // Create a builder that will store the contents of the table it is
-  // building in *file.  Does not close the file.  It is up to the
-  // caller to close the file after calling Finish(). The output file
-  // will be part of level specified by 'level'.  A value of -1 means
-  // that the caller does not know which level the output file will reside.
-  SimpleTableBuilder(const Options& options, WritableFile* file,
-                     CompressionType compression_type);
-
-  // REQUIRES: Either Finish() or Abandon() has been called.
-  ~SimpleTableBuilder();
-
-  // Add key,value to the table being constructed.
-  // REQUIRES: key is after any previously added key according to comparator.
-  // REQUIRES: Finish(), Abandon() have not been called
-  void Add(const Slice& key, const Slice& value) override;
-
-  // Return non-ok iff some error has been detected.
-  Status status() const override;
-
-  // Finish building the table.  Stops using the file passed to the
-  // constructor after this function returns.
-  // REQUIRES: Finish(), Abandon() have not been called
-  Status Finish() override;
-
-  // Indicate that the contents of this builder should be abandoned.  Stops
-  // using the file passed to the constructor after this function returns.
-  // If the caller is not going to call Finish(), it must call Abandon()
-  // before destroying this builder.
-  // REQUIRES: Finish(), Abandon() have not been called
-  void Abandon() override;
-
-  // Number of calls to Add() so far.
-  uint64_t NumEntries() const override;
-
-  // Size of the file generated so far.  If invoked after a successful
-  // Finish() call, returns the size of the final generated file.
-  uint64_t FileSize() const override;
-
-private:
-  struct Rep;
-  Rep* rep_;
-
-  // No copying allowed
-  SimpleTableBuilder(const SimpleTableBuilder&) = delete;
-  void operator=(const SimpleTableBuilder&) = delete;
-};
-
-struct SimpleTableBuilder::Rep {
-  Options options;
-  WritableFile* file;
-  uint64_t offset = 0;
-  Status status;
-
-  uint64_t num_entries = 0;
-
-  bool closed = false;  // Either Finish() or Abandon() has been called.
-
-  const static int user_key_size = 16;
-  const static int offset_length = 8;
-  const static int key_footer_len = 8;
-
-  static int GetInternalKeyLength() {
-    return user_key_size + key_footer_len;
-  }
-
-  std::string index;
-
-  Rep(const Options& opt, WritableFile* f) :
-      options(opt), file(f) {
-  }
-  ~Rep() {
-  }
-};
-
-SimpleTableBuilder::SimpleTableBuilder(const Options& options,
-                                       WritableFile* file,
-                                       CompressionType compression_type) :
-    rep_(new SimpleTableBuilder::Rep(options, file)) {
-}
-
-SimpleTableBuilder::~SimpleTableBuilder() {
-  delete (rep_);
-}
-
-void SimpleTableBuilder::Add(const Slice& key, const Slice& value) {
-  assert((int ) key.size() == Rep::GetInternalKeyLength());
-
-  // Update index
-  rep_->index.append(key.data(), key.size());
-  PutFixed64(&(rep_->index), rep_->offset);
-
-  // Write key-value pair
-  rep_->file->Append(key);
-  rep_->offset += Rep::GetInternalKeyLength();
-
-  std::string size;
-  int value_size = value.size();
-  PutFixed32(&size, value_size);
-  Slice sizeSlice(size);
-  rep_->file->Append(sizeSlice);
-  rep_->file->Append(value);
-  rep_->offset += value_size + 4;
-
-  rep_->num_entries++;
-}
-
-Status SimpleTableBuilder::status() const {
-  return Status::OK();
-}
-
-Status SimpleTableBuilder::Finish() {
-  Rep* r = rep_;
-  assert(!r->closed);
-  r->closed = true;
-
-  uint64_t index_offset = rep_->offset;
-  Slice index_slice(rep_->index);
-  rep_->file->Append(index_slice);
-  rep_->offset += index_slice.size();
-
-  std::string index_offset_str;
-  PutFixed64(&index_offset_str, index_offset);
-  Slice foot_slice(index_offset_str);
-  rep_->file->Append(foot_slice);
-  rep_->offset += foot_slice.size();
-
-  return Status::OK();
-}
-
-void SimpleTableBuilder::Abandon() {
-  rep_->closed = true;
-}
-
-uint64_t SimpleTableBuilder::NumEntries() const {
-  return rep_->num_entries;
-}
-
-uint64_t SimpleTableBuilder::FileSize() const {
-  return rep_->offset;
-}
-
-class SimpleTableFactory: public TableFactory {
-public:
-  ~SimpleTableFactory() {
-  }
-  SimpleTableFactory() {
-  }
-  const char* Name() const override {
-    return "SimpleTable";
-  }
-  Status NewTableReader(const Options& options, const EnvOptions& soptions,
-                        const InternalKeyComparator& internal_key,
-                        unique_ptr<RandomAccessFile>&& file, uint64_t file_size,
-                        unique_ptr<TableReader>* table_reader) const;
-
-  TableBuilder* NewTableBuilder(const Options& options,
-                                const InternalKeyComparator& internal_key,
-                                WritableFile* file,
-                                CompressionType compression_type) const;
-};
-
-Status SimpleTableFactory::NewTableReader(
-    const Options& options, const EnvOptions& soptions,
-    const InternalKeyComparator& internal_key,
-    unique_ptr<RandomAccessFile>&& file, uint64_t file_size,
-    unique_ptr<TableReader>* table_reader) const {
-
-  return SimpleTableReader::Open(options, soptions, std::move(file), file_size,
-                                 table_reader);
-}
-
-TableBuilder* SimpleTableFactory::NewTableBuilder(
-    const Options& options, const InternalKeyComparator& internal_key,
-    WritableFile* file, CompressionType compression_type) const {
-  return new SimpleTableBuilder(options, file, compression_type);
-}
-
-class SimpleTableDBTest {
-protected:
-public:
-  std::string dbname_;
-  Env* env_;
-  DB* db_;
-
-  Options last_options_;
-
-  SimpleTableDBTest() :
-      env_(Env::Default()) {
-    dbname_ = test::TmpDir() + "/simple_table_db_test";
-    ASSERT_OK(DestroyDB(dbname_, Options()));
-    db_ = nullptr;
-    Reopen();
-  }
-
-  ~SimpleTableDBTest() {
-    delete db_;
-    ASSERT_OK(DestroyDB(dbname_, Options()));
-  }
-
-  // Return the current option configuration.
-  Options CurrentOptions() {
-    Options options;
-    options.table_factory.reset(new SimpleTableFactory());
-    return options;
-  }
-
-  DBImpl* dbfull() {
-    return reinterpret_cast<DBImpl*>(db_);
-  }
-
-  void Reopen(Options* options = nullptr) {
-    ASSERT_OK(TryReopen(options));
-  }
-
-  void Close() {
-    delete db_;
-    db_ = nullptr;
-  }
-
-  void DestroyAndReopen(Options* options = nullptr) {
-    //Destroy using last options
-    Destroy(&last_options_);
-    ASSERT_OK(TryReopen(options));
-  }
-
-  void Destroy(Options* options) {
-    delete db_;
-    db_ = nullptr;
-    ASSERT_OK(DestroyDB(dbname_, *options));
-  }
-
-  Status PureReopen(Options* options, DB** db) {
-    return DB::Open(*options, dbname_, db);
-  }
-
-  Status TryReopen(Options* options = nullptr) {
-    delete db_;
-    db_ = nullptr;
-    Options opts;
-    if (options != nullptr) {
-      opts = *options;
-    } else {
-      opts = CurrentOptions();
-      opts.create_if_missing = true;
-    }
-    last_options_ = opts;
-
-    return DB::Open(opts, dbname_, &db_);
-  }
-
-  Status Put(const Slice& k, const Slice& v) {
-    return db_->Put(WriteOptions(), k, v);
-  }
-
-  Status Delete(const std::string& k) {
-    return db_->Delete(WriteOptions(), k);
-  }
-
-  std::string Get(const std::string& k, const Snapshot* snapshot = nullptr) {
-    ReadOptions options;
-    options.snapshot = snapshot;
-    std::string result;
-    Status s = db_->Get(options, k, &result);
-    if (s.IsNotFound()) {
-      result = "NOT_FOUND";
-    } else if (!s.ok()) {
-      result = s.ToString();
-    }
-    return result;
-  }
-
-
-  int NumTableFilesAtLevel(int level) {
-    std::string property;
-    ASSERT_TRUE(
-        db_->GetProperty("rocksdb.num-files-at-level" + NumberToString(level),
-                         &property));
-    return atoi(property.c_str());
-  }
-
-  // Return spread of files per level
-  std::string FilesPerLevel() {
-    std::string result;
-    int last_non_zero_offset = 0;
-    for (int level = 0; level < db_->NumberLevels(); level++) {
-      int f = NumTableFilesAtLevel(level);
-      char buf[100];
-      snprintf(buf, sizeof(buf), "%s%d", (level ? "," : ""), f);
-      result += buf;
-      if (f > 0) {
-        last_non_zero_offset = result.size();
-      }
-    }
-    result.resize(last_non_zero_offset);
-    return result;
-  }
-
-  std::string IterStatus(Iterator* iter) {
-    std::string result;
-    if (iter->Valid()) {
-      result = iter->key().ToString() + "->" + iter->value().ToString();
-    } else {
-      result = "(invalid)";
-    }
-    return result;
-  }
-};
-
-TEST(SimpleTableDBTest, Empty) {
-  ASSERT_TRUE(db_ != nullptr);
-  ASSERT_EQ("NOT_FOUND", Get("0000000000000foo"));
-}
-
-TEST(SimpleTableDBTest, ReadWrite) {
-  ASSERT_OK(Put("0000000000000foo", "v1"));
-  ASSERT_EQ("v1", Get("0000000000000foo"));
-  ASSERT_OK(Put("0000000000000bar", "v2"));
-  ASSERT_OK(Put("0000000000000foo", "v3"));
-  ASSERT_EQ("v3", Get("0000000000000foo"));
-  ASSERT_EQ("v2", Get("0000000000000bar"));
-}
-
-TEST(SimpleTableDBTest, Flush) {
-  ASSERT_OK(Put("0000000000000foo", "v1"));
-  ASSERT_OK(Put("0000000000000bar", "v2"));
-  ASSERT_OK(Put("0000000000000foo", "v3"));
-  dbfull()->TEST_FlushMemTable();
-  ASSERT_EQ("v3", Get("0000000000000foo"));
-  ASSERT_EQ("v2", Get("0000000000000bar"));
-}
-
-TEST(SimpleTableDBTest, Flush2) {
-  ASSERT_OK(Put("0000000000000bar", "b"));
-  ASSERT_OK(Put("0000000000000foo", "v1"));
-  dbfull()->TEST_FlushMemTable();
-
-  ASSERT_OK(Put("0000000000000foo", "v2"));
-  dbfull()->TEST_FlushMemTable();
-  ASSERT_EQ("v2", Get("0000000000000foo"));
-
-  ASSERT_OK(Put("0000000000000eee", "v3"));
-  dbfull()->TEST_FlushMemTable();
-  ASSERT_EQ("v3", Get("0000000000000eee"));
-
-  ASSERT_OK(Delete("0000000000000bar"));
-  dbfull()->TEST_FlushMemTable();
-  ASSERT_EQ("NOT_FOUND", Get("0000000000000bar"));
-
-  ASSERT_OK(Put("0000000000000eee", "v5"));
-  dbfull()->TEST_FlushMemTable();
-  ASSERT_EQ("v5", Get("0000000000000eee"));
-}
-
-static std::string Key(int i) {
-  char buf[100];
-  snprintf(buf, sizeof(buf), "key_______%06d", i);
-  return std::string(buf);
-}
-
-static std::string RandomString(Random* rnd, int len) {
-  std::string r;
-  test::RandomString(rnd, len, &r);
-  return r;
-}
-
-TEST(SimpleTableDBTest, CompactionTrigger) {
-  Options options = CurrentOptions();
-  options.write_buffer_size = 100 << 10; //100KB
-  options.num_levels = 3;
-  options.max_mem_compaction_level = 0;
-  options.level0_file_num_compaction_trigger = 3;
-  Reopen(&options);
-
-  Random rnd(301);
-
-  for (int num = 0; num < options.level0_file_num_compaction_trigger - 1;
-      num++) {
-    std::vector<std::string> values;
-    // Write 120KB (12 values, each 10K)
-    for (int i = 0; i < 12; i++) {
-      values.push_back(RandomString(&rnd, 10000));
-      ASSERT_OK(Put(Key(i), values[i]));
-    }
-    dbfull()->TEST_WaitForFlushMemTable();
-    ASSERT_EQ(NumTableFilesAtLevel(0), num + 1);
-  }
-
-  //generate one more file in level-0, and should trigger level-0 compaction
-  std::vector<std::string> values;
-  for (int i = 0; i < 12; i++) {
-    values.push_back(RandomString(&rnd, 10000));
-    ASSERT_OK(Put(Key(i), values[i]));
-  }
-  dbfull()->TEST_WaitForCompact();
-
-  ASSERT_EQ(NumTableFilesAtLevel(0), 0);
-  ASSERT_EQ(NumTableFilesAtLevel(1), 1);
-}
-
-}  // namespace rocksdb
-
-int main(int argc, char** argv) {
-  return rocksdb::test::RunAllTests();
-}
@@ -71,7 +71,7 @@ class SnapshotList {
  }

  // get the sequence number of the most recent snapshot
-  const SequenceNumber GetNewest() {
+  SequenceNumber GetNewest() {
    if (empty()) {
      return 0;
    }
@@ -13,7 +13,9 @@
 #include "db/version_edit.h"

 #include "rocksdb/statistics.h"
+#include "table/iterator_wrapper.h"
 #include "table/table_reader.h"
+#include "table/get_context.h"
 #include "util/coding.h"
 #include "util/stop_watch.h"

@@ -30,17 +32,15 @@ static void UnrefEntry(void* arg1, void* arg2) {
  cache->Release(h);
 }

-static Slice GetSliceForFileNumber(uint64_t* file_number) {
+static Slice GetSliceForFileNumber(const uint64_t* file_number) {
  return Slice(reinterpret_cast<const char*>(file_number),
               sizeof(*file_number));
 }

-TableCache::TableCache(const std::string& dbname, const Options* options,
-                       const EnvOptions& storage_options, Cache* const cache)
-    : env_(options->env),
-      dbname_(dbname),
-      options_(options),
-      storage_options_(storage_options),
+TableCache::TableCache(const ImmutableCFOptions& ioptions,
+                       const EnvOptions& env_options, Cache* const cache)
+    : ioptions_(ioptions),
+      env_options_(env_options),
      cache_(cache) {}

 TableCache::~TableCache() {
@@ -54,39 +54,37 @@ void TableCache::ReleaseHandle(Cache::Handle* handle) {
  cache_->Release(handle);
 }

-Status TableCache::FindTable(const EnvOptions& toptions,
+Status TableCache::FindTable(const EnvOptions& env_options,
                             const InternalKeyComparator& internal_comparator,
-                             uint64_t file_number, uint64_t file_size,
-                             Cache::Handle** handle, bool* table_io,
+                             const FileDescriptor& fd, Cache::Handle** handle,
                             const bool no_io) {
  Status s;
-  Slice key = GetSliceForFileNumber(&file_number);
+  uint64_t number = fd.GetNumber();
+  Slice key = GetSliceForFileNumber(&number);
  *handle = cache_->Lookup(key);
  if (*handle == nullptr) {
    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);
+    std::string fname =
+        TableFileName(ioptions_.db_paths, fd.GetNumber(), fd.GetPathId());
    unique_ptr<RandomAccessFile> file;
    unique_ptr<TableReader> table_reader;
-    s = env_->NewRandomAccessFile(fname, &file, toptions);
-    RecordTick(options_->statistics.get(), NO_FILE_OPENS);
+    s = ioptions_.env->NewRandomAccessFile(fname, &file, env_options);
+    RecordTick(ioptions_.statistics, NO_FILE_OPENS);
    if (s.ok()) {
-      if (options_->advise_random_on_open) {
+      if (ioptions_.advise_random_on_open) {
        file->Hint(RandomAccessFile::RANDOM);
      }
-      StopWatch sw(env_, options_->statistics.get(), TABLE_OPEN_IO_MICROS);
-      s = options_->table_factory->NewTableReader(
-          *options_, toptions, internal_comparator, std::move(file), file_size,
-          &table_reader);
+      StopWatch sw(ioptions_.env, ioptions_.statistics, TABLE_OPEN_IO_MICROS);
+      s = ioptions_.table_factory->NewTableReader(
+          ioptions_, env_options, internal_comparator, std::move(file),
+          fd.GetFileSize(), &table_reader);
    }

    if (!s.ok()) {
      assert(table_reader == nullptr);
-      RecordTick(options_->statistics.get(), NO_FILE_ERRORS);
+      RecordTick(ioptions_.statistics, NO_FILE_ERRORS);
      // We do not cache error results so that if the error is transient,
      // or somebody repairs the file, we recover automatically.
    } else {
@@ -98,27 +96,27 @@ Status TableCache::FindTable(const EnvOptions& toptions,
 }

 Iterator* TableCache::NewIterator(const ReadOptions& options,
-                                  const EnvOptions& toptions,
+                                  const EnvOptions& env_options,
                                  const InternalKeyComparator& icomparator,
-                                  const FileMetaData& file_meta,
+                                  const FileDescriptor& fd,
                                  TableReader** table_reader_ptr,
-                                  bool for_compaction) {
+                                  bool for_compaction, Arena* arena) {
  if (table_reader_ptr != nullptr) {
    *table_reader_ptr = nullptr;
  }
-  TableReader* table_reader = file_meta.table_reader;
+  TableReader* table_reader = fd.table_reader;
  Cache::Handle* handle = nullptr;
  Status s;
  if (table_reader == nullptr) {
-    s = FindTable(toptions, icomparator, file_meta.number, file_meta.file_size,
-                  &handle, nullptr, options.read_tier == kBlockCacheTier);
+    s = FindTable(env_options, icomparator, fd, &handle,
+                  options.read_tier == kBlockCacheTier);
    if (!s.ok()) {
-      return NewErrorIterator(s);
+      return NewErrorIterator(s, arena);
    }
    table_reader = GetTableReaderFromHandle(handle);
  }

-  Iterator* result = table_reader->NewIterator(options);
+  Iterator* result = table_reader->NewIterator(options, arena);
  if (handle != nullptr) {
    result->RegisterCleanup(&UnrefEntry, cache_, handle);
  }
@@ -135,40 +133,37 @@ Iterator* TableCache::NewIterator(const ReadOptions& options,

 Status TableCache::Get(const ReadOptions& options,
                       const InternalKeyComparator& internal_comparator,
-                       const FileMetaData& file_meta, const Slice& k, void* arg,
-                       bool (*saver)(void*, const ParsedInternalKey&,
-                                     const Slice&, bool),
-                       bool* table_io, void (*mark_key_may_exist)(void*)) {
-  TableReader* t = file_meta.table_reader;
+                       const FileDescriptor& fd, const Slice& k,
+                       GetContext* get_context) {
+  TableReader* t = fd.table_reader;
  Status s;
  Cache::Handle* handle = nullptr;
  if (!t) {
-    s = FindTable(storage_options_, internal_comparator, file_meta.number,
-                  file_meta.file_size, &handle, table_io,
+    s = FindTable(env_options_, internal_comparator, fd, &handle,
                  options.read_tier == kBlockCacheTier);
    if (s.ok()) {
      t = GetTableReaderFromHandle(handle);
    }
  }
  if (s.ok()) {
-    s = t->Get(options, k, arg, saver, mark_key_may_exist);
+    s = t->Get(options, k, get_context);
    if (handle != nullptr) {
      ReleaseHandle(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);
+    get_context->MarkKeyMayExist();
    return Status::OK();
  }
  return s;
 }
+
 Status TableCache::GetTableProperties(
-    const EnvOptions& toptions,
-    const InternalKeyComparator& internal_comparator,
-    const FileMetaData& file_meta,
+    const EnvOptions& env_options,
+    const InternalKeyComparator& internal_comparator, const FileDescriptor& fd,
    std::shared_ptr<const TableProperties>* properties, bool no_io) {
  Status s;
-  auto table_reader = file_meta.table_reader;
+  auto table_reader = fd.table_reader;
  // table already been pre-loaded?
  if (table_reader) {
    *properties = table_reader->GetTableProperties();
@@ -176,10 +171,8 @@ Status TableCache::GetTableProperties(
    return s;
  }

-  bool table_io;
  Cache::Handle* table_handle = nullptr;
-  s = FindTable(toptions, internal_comparator, file_meta.number,
-                file_meta.file_size, &table_handle, &table_io, no_io);
+  s = FindTable(env_options, internal_comparator, fd, &table_handle, no_io);
  if (!s.ok()) {
    return s;
  }
@@ -190,6 +183,29 @@ Status TableCache::GetTableProperties(
  return s;
 }

+size_t TableCache::GetMemoryUsageByTableReader(
+    const EnvOptions& env_options,
+    const InternalKeyComparator& internal_comparator,
+    const FileDescriptor& fd) {
+  Status s;
+  auto table_reader = fd.table_reader;
+  // table already been pre-loaded?
+  if (table_reader) {
+    return table_reader->ApproximateMemoryUsage();
+  }
+
+  Cache::Handle* table_handle = nullptr;
+  s = FindTable(env_options, internal_comparator, fd, &table_handle, true);
+  if (!s.ok()) {
+    return 0;
+  }
+  assert(table_handle);
+  auto table = GetTableReaderFromHandle(table_handle);
+  auto ret = table->ApproximateMemoryUsage();
+  ReleaseHandle(table_handle);
+  return ret;
+}
+
 void TableCache::Evict(Cache* cache, uint64_t file_number) {
  cache->Erase(GetSliceForFileNumber(&file_number));
 }
@@ -11,6 +11,7 @@

 #pragma once
 #include <string>
+#include <vector>
 #include <stdint.h>

 #include "db/dbformat.h"
@@ -18,18 +19,19 @@
 #include "rocksdb/cache.h"
 #include "rocksdb/env.h"
 #include "rocksdb/table.h"
+#include "rocksdb/options.h"
 #include "table/table_reader.h"

 namespace rocksdb {

 class Env;
-struct FileMetaData;
+class Arena;
+struct FileDescriptor;
+class GetContext;

-// TODO(sdong): try to come up with a better API to pass the file information
-//              other than simply passing FileMetaData.
 class TableCache {
 public:
-  TableCache(const std::string& dbname, const Options* options,
+  TableCache(const ImmutableCFOptions& ioptions,
             const EnvOptions& storage_options, Cache* cache);
  ~TableCache();

@@ -42,19 +44,17 @@ class TableCache {
  // returned iterator is live.
  Iterator* NewIterator(const ReadOptions& options, const EnvOptions& toptions,
                        const InternalKeyComparator& internal_comparator,
-                        const FileMetaData& file_meta,
+                        const FileDescriptor& file_fd,
                        TableReader** table_reader_ptr = nullptr,
-                        bool for_compaction = false);
+                        bool for_compaction = false, Arena* arena = nullptr);

  // If a seek to internal key "k" in specified file finds an entry,
  // call (*handle_result)(arg, found_key, found_value) repeatedly until
  // it returns false.
  Status Get(const ReadOptions& options,
             const InternalKeyComparator& internal_comparator,
-             const FileMetaData& file_meta, const Slice& k, void* arg,
-             bool (*handle_result)(void*, const ParsedInternalKey&,
-                                   const Slice&, bool),
-             bool* table_io, void (*mark_key_may_exist)(void*) = nullptr);
+             const FileDescriptor& file_fd, const Slice& k,
+             GetContext* get_context);

  // Evict any entry for the specified file number
  static void Evict(Cache* cache, uint64_t file_number);
@@ -62,8 +62,8 @@ class TableCache {
  // Find table reader
  Status FindTable(const EnvOptions& toptions,
                   const InternalKeyComparator& internal_comparator,
-                   uint64_t file_number, uint64_t file_size, Cache::Handle**,
-                   bool* table_io = nullptr, const bool no_io = false);
+                   const FileDescriptor& file_fd, Cache::Handle**,
+                   const bool no_io = false);

  // Get TableReader from a cache handle.
  TableReader* GetTableReaderFromHandle(Cache::Handle* handle);
@@ -76,18 +76,23 @@ class TableCache {
  //            we set `no_io` to be true.
  Status GetTableProperties(const EnvOptions& toptions,
                            const InternalKeyComparator& internal_comparator,
-                            const FileMetaData& file_meta,
+                            const FileDescriptor& file_meta,
                            std::shared_ptr<const TableProperties>* properties,
                            bool no_io = false);

+  // Return total memory usage of the table reader of the file.
+  // 0 of table reader of the file is not loaded.
+  size_t GetMemoryUsageByTableReader(
+      const EnvOptions& toptions,
+      const InternalKeyComparator& internal_comparator,
+      const FileDescriptor& fd);
+
  // Release the handle from a cache
  void ReleaseHandle(Cache::Handle* handle);

 private:
-  Env* const env_;
-  const std::string dbname_;
-  const Options* options_;
-  const EnvOptions& storage_options_;
+  const ImmutableCFOptions& ioptions_;
+  const EnvOptions& env_options_;
  Cache* const cache_;
 };

@@ -11,6 +11,7 @@
 #include "db/dbformat.h"
 #include "db/table_properties_collector.h"
 #include "rocksdb/table.h"
+#include "rocksdb/immutable_options.h"
 #include "table/block_based_table_factory.h"
 #include "table/meta_blocks.h"
 #include "table/plain_table_factory.h"
@@ -85,12 +86,14 @@ class DumbLogger : public Logger {
 // Utilities test functions
 namespace {
 void MakeBuilder(const Options& options,
+                 const ImmutableCFOptions& ioptions,
                 const InternalKeyComparator& internal_comparator,
                 std::unique_ptr<FakeWritableFile>* writable,
                 std::unique_ptr<TableBuilder>* builder) {
  writable->reset(new FakeWritableFile);
-  builder->reset(options.table_factory->NewTableBuilder(
-      options, internal_comparator, writable->get(), options.compression));
+  builder->reset(ioptions.table_factory->NewTableBuilder(
+      ioptions, internal_comparator, writable->get(),
+      options.compression, options.compression_opts));
 }
 }  // namespace

@@ -153,7 +156,8 @@ void TestCustomizedTablePropertiesCollector(
  // -- Step 1: build table
  std::unique_ptr<TableBuilder> builder;
  std::unique_ptr<FakeWritableFile> writable;
-  MakeBuilder(options, internal_comparator, &writable, &builder);
+  const ImmutableCFOptions ioptions(options);
+  MakeBuilder(options, ioptions, internal_comparator, &writable, &builder);

  for (const auto& kv : kvs) {
    if (encode_as_internal) {
@@ -213,7 +217,14 @@ TEST(TablePropertiesTest, CustomizedTablePropertiesCollector) {
  Options options;
  options.table_properties_collector_factories.emplace_back(
      new RegularKeysStartWithAFactory());
-  options.table_factory = std::make_shared<PlainTableFactory>(8, 8, 0);
+
+  PlainTableOptions plain_table_options;
+  plain_table_options.user_key_len = 8;
+  plain_table_options.bloom_bits_per_key = 8;
+  plain_table_options.hash_table_ratio = 0;
+
+  options.table_factory =
+      std::make_shared<PlainTableFactory>(plain_table_options);
  test::PlainInternalKeyComparator ikc(options.comparator);
  TestCustomizedTablePropertiesCollector(kPlainTableMagicNumber, true, options,
                                         ikc);
@@ -250,16 +261,17 @@ void TestInternalKeyPropertiesCollector(
    // SanitizeOptions().
    options.info_log = std::make_shared<DumbLogger>();
    options = SanitizeOptions("db",            // just a place holder
-                              &pikc, nullptr,  // don't care filter policy
+                              &pikc,
                              options);
    options.comparator = comparator;
  } else {
    options.table_properties_collector_factories = {
        std::make_shared<InternalKeyPropertiesCollectorFactory>()};
  }
+  const ImmutableCFOptions ioptions(options);

  for (int iter = 0; iter < 2; ++iter) {
-    MakeBuilder(options, pikc, &writable, &builder);
+    MakeBuilder(options, ioptions, pikc, &writable, &builder);
    for (const auto& k : keys) {
      builder->Add(k.Encode(), "val");
    }
@@ -299,11 +311,15 @@ TEST(TablePropertiesTest, InternalKeyPropertiesCollector) {
      true /* not sanitize */,
      std::make_shared<BlockBasedTableFactory>()
  );
+
+  PlainTableOptions plain_table_options;
+  plain_table_options.user_key_len = 8;
+  plain_table_options.bloom_bits_per_key = 8;
+  plain_table_options.hash_table_ratio = 0;
+
  TestInternalKeyPropertiesCollector(
-      kPlainTableMagicNumber,
-      false /* not sanitize */,
-      std::make_shared<PlainTableFactory>(8, 8, 0)
-  );
+      kPlainTableMagicNumber, false /* not sanitize */,
+      std::make_shared<PlainTableFactory>(plain_table_options));
 }

 }  // namespace rocksdb
@@ -1,221 +0,0 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
-//  This source code is licensed under the BSD-style license found in the
-//  LICENSE file in the root directory of this source tree. An additional grant
-//  of patent rights can be found in the PATENTS file in the same directory.
-
-#ifndef ROCKSDB_LITE
-#include "db/tailing_iter.h"
-
-#include <string>
-#include <utility>
-#include <vector>
-#include "db/db_impl.h"
-#include "db/db_iter.h"
-#include "db/column_family.h"
-#include "rocksdb/env.h"
-#include "rocksdb/slice.h"
-#include "rocksdb/slice_transform.h"
-#include "table/merger.h"
-
-namespace rocksdb {
-
-TailingIterator::TailingIterator(Env* const env, DBImpl* db,
-    const ReadOptions& read_options, ColumnFamilyData* cfd)
-    : env_(env),
-      db_(db),
-      read_options_(read_options),
-      cfd_(cfd),
-      super_version_(nullptr),
-      current_(nullptr),
-      status_(Status::InvalidArgument("Seek() not called on this iterator")) {}
-
-TailingIterator::~TailingIterator() {
-  Cleanup();
-}
-
-bool TailingIterator::Valid() const {
-  return current_ != nullptr;
-}
-
-void TailingIterator::SeekToFirst() {
-  if (!IsCurrentVersion()) {
-    CreateIterators();
-  }
-
-  mutable_->SeekToFirst();
-  immutable_->SeekToFirst();
-  UpdateCurrent();
-}
-
-void TailingIterator::Seek(const Slice& target) {
-  if (!IsCurrentVersion()) {
-    CreateIterators();
-  }
-
-  mutable_->Seek(target);
-
-  // We maintain the interval (prev_key_, immutable_->key()] such that there
-  // are no records with keys within that range in immutable_ other than
-  // immutable_->key(). Since immutable_ can't change in this version, we don't
-  // need to do a seek if 'target' belongs to that interval (i.e. immutable_ is
-  // already at the correct position)!
-  //
-  // If prefix seek is used and immutable_ is not valid, seek if target has a
-  // different prefix than prev_key.
-  //
-  // prev_key_ is updated by Next(). SeekImmutable() sets prev_key_ to
-  // 'target' -- in this case, prev_key_ is included in the interval, so
-  // prev_inclusive_ has to be set.
-
-  const Comparator* cmp = cfd_->user_comparator();
-  if (!is_prev_set_ || cmp->Compare(prev_key_, target) >= !is_prev_inclusive_ ||
-      (immutable_->Valid() && cmp->Compare(target, immutable_->key()) > 0) ||
-      (cfd_->options()->prefix_extractor != nullptr && !IsSamePrefix(target))) {
-    SeekImmutable(target);
-  }
-
-  UpdateCurrent();
-}
-
-void TailingIterator::Next() {
-  assert(Valid());
-
-  if (!IsCurrentVersion()) {
-    // save the current key, create new iterators and then seek
-    std::string current_key = key().ToString();
-    Slice key_slice(current_key.data(), current_key.size());
-
-    CreateIterators();
-    Seek(key_slice);
-
-    if (!Valid() || key().compare(key_slice) != 0) {
-      // record with current_key no longer exists
-      return;
-    }
-
-  } else if (current_ == immutable_.get()) {
-    // immutable iterator is advanced -- update prev_key_
-    prev_key_ = key().ToString();
-    is_prev_inclusive_ = false;
-    is_prev_set_ = true;
-  }
-
-  current_->Next();
-  UpdateCurrent();
-}
-
-Slice TailingIterator::key() const {
-  assert(Valid());
-  return current_->key();
-}
-
-Slice TailingIterator::value() const {
-  assert(Valid());
-  return current_->value();
-}
-
-Status TailingIterator::status() const {
-  if (!status_.ok()) {
-    return status_;
-  } else if (!mutable_->status().ok()) {
-    return mutable_->status();
-  } else {
-    return immutable_->status();
-  }
-}
-
-void TailingIterator::Prev() {
-  status_ = Status::NotSupported("This iterator doesn't support Prev()");
-}
-
-void TailingIterator::SeekToLast() {
-  status_ = Status::NotSupported("This iterator doesn't support SeekToLast()");
-}
-
-void TailingIterator::Cleanup() {
-  // Release old super version if necessary
-  mutable_.reset();
-  immutable_.reset();
-  if (super_version_ != nullptr && super_version_->Unref()) {
-    DBImpl::DeletionState deletion_state;
-    db_->mutex_.Lock();
-    super_version_->Cleanup();
-    db_->FindObsoleteFiles(deletion_state, false, true);
-    db_->mutex_.Unlock();
-    delete super_version_;
-    if (deletion_state.HaveSomethingToDelete()) {
-      db_->PurgeObsoleteFiles(deletion_state);
-    }
-  }
-}
-
-void TailingIterator::CreateIterators() {
-  Cleanup();
-  super_version_= cfd_->GetReferencedSuperVersion(&(db_->mutex_));
-
-  Iterator* mutable_iter = super_version_->mem->NewIterator(read_options_);
-  // create a DBIter that only uses memtable content; see NewIterator()
-  mutable_.reset(
-      NewDBIterator(env_, *cfd_->options(), cfd_->user_comparator(),
-                    mutable_iter, kMaxSequenceNumber));
-
-  std::vector<Iterator*> list;
-  super_version_->imm->AddIterators(read_options_, &list);
-  super_version_->current->AddIterators(
-      read_options_, *cfd_->soptions(), &list);
-  Iterator* immutable_iter =
-      NewMergingIterator(&cfd_->internal_comparator(), &list[0], list.size());
-
-  // create a DBIter that only uses memtable content; see NewIterator()
-  immutable_.reset(
-      NewDBIterator(env_, *cfd_->options(), cfd_->user_comparator(),
-                    immutable_iter, kMaxSequenceNumber));
-
-  current_ = nullptr;
-  is_prev_set_ = false;
-}
-
-void TailingIterator::UpdateCurrent() {
-  current_ = nullptr;
-
-  if (mutable_->Valid()) {
-    current_ = mutable_.get();
-  }
-  const Comparator* cmp = cfd_->user_comparator();
-  if (immutable_->Valid() &&
-      (current_ == nullptr ||
-       cmp->Compare(immutable_->key(), current_->key()) < 0)) {
-    current_ = immutable_.get();
-  }
-
-  if (!status_.ok()) {
-    // reset status that was set by Prev() or SeekToLast()
-    status_ = Status::OK();
-  }
-}
-
-bool TailingIterator::IsCurrentVersion() const {
-  return super_version_ != nullptr &&
-         super_version_->version_number == cfd_->GetSuperVersionNumber();
-}
-
-bool TailingIterator::IsSamePrefix(const Slice& target) const {
-  const SliceTransform* extractor = cfd_->options()->prefix_extractor.get();
-
-  assert(extractor);
-  assert(is_prev_set_);
-
-  return extractor->Transform(target)
-    .compare(extractor->Transform(prev_key_)) == 0;
-}
-
-void TailingIterator::SeekImmutable(const Slice& target) {
-  prev_key_ = target.ToString();
-  is_prev_inclusive_ = true;
-  is_prev_set_ = true;
-
-  immutable_->Seek(target);
-}
-
-}  // namespace rocksdb
-#endif  // ROCKSDB_LITE
@@ -1,97 +0,0 @@
-//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
-//  This source code is licensed under the BSD-style license found in the
-//  LICENSE file in the root directory of this source tree. An additional grant
-//  of patent rights can be found in the PATENTS file in the same directory.
-#pragma once
-
-#ifndef ROCKSDB_LITE
-
-#include <string>
-
-#include "rocksdb/db.h"
-#include "rocksdb/iterator.h"
-#include "rocksdb/options.h"
-
-namespace rocksdb {
-
-class DBImpl;
-class Env;
-struct SuperVersion;
-class ColumnFamilyData;
-
-/**
- * TailingIterator is a special type of iterator that doesn't use an (implicit)
- * snapshot. In other words, it can be used to read data that was added to the
- * db after the iterator had been created.
- *
- * TailingIterator is optimized for sequential reading. It doesn't support
- * Prev() and SeekToLast() operations.
- */
-class TailingIterator : public Iterator {
- public:
-  TailingIterator(Env* const env, DBImpl* db, const ReadOptions& read_options,
-                  ColumnFamilyData* cfd);
-  virtual ~TailingIterator();
-
-  virtual bool Valid() const override;
-  virtual void SeekToFirst() override;
-  virtual void SeekToLast() override;
-  virtual void Seek(const Slice& target) override;
-  virtual void Next() override;
-  virtual void Prev() override;
-  virtual Slice key() const override;
-  virtual Slice value() const override;
-  virtual Status status() const override;
-
- private:
-  void Cleanup();
-
-  Env* const env_;
-  DBImpl* const db_;
-  const ReadOptions read_options_;
-  ColumnFamilyData* const cfd_;
-  SuperVersion* super_version_;
-
-  // TailingIterator merges the contents of the two iterators below (one using
-  // mutable memtable contents only, other over SSTs and immutable memtables).
-  // See DBIter::GetTailingIteratorPair().
-  std::unique_ptr<Iterator> mutable_;
-  std::unique_ptr<Iterator> immutable_;
-
-  // points to either mutable_ or immutable_
-  Iterator* current_;
-
-  // key that precedes immutable iterator's current key
-  std::string prev_key_;
-
-  // unless prev_set is true, prev_key/prev_head is not valid and shouldn't be
-  // used; reset by createIterators()
-  bool is_prev_set_;
-
-  // prev_key_ was set by SeekImmutable(), which means that the interval of
-  // keys covered by immutable_ is [prev_key_, current], i.e. it includes the
-  // left endpoint
-  bool is_prev_inclusive_;
-
-  // internal iterator status
-  Status status_;
-
-  // check if this iterator's version matches DB's version
-  bool IsCurrentVersion() const;
-
-  // check if SeekImmutable() is needed due to target having a different prefix
-  // than prev_key_ (used when in prefix seek mode)
-  bool IsSamePrefix(const Slice& target) const;
-
-  // creates mutable_ and immutable_ iterators and updates version_number_
-  void CreateIterators();
-
-  // set current_ to be one of the iterators with the smallest key
-  void UpdateCurrent();
-
-  // seek on immutable_ and update prev_key
-  void SeekImmutable(const Slice& target);
-};
-
-}  // namespace rocksdb
-#endif  // ROCKSDB_LITE
@@ -18,25 +18,30 @@ namespace rocksdb {
 // Tag numbers for serialized VersionEdit.  These numbers are written to
 // disk and should not be changed.
 enum Tag {
-  kComparator           = 1,
-  kLogNumber            = 2,
-  kNextFileNumber       = 3,
-  kLastSequence         = 4,
-  kCompactPointer       = 5,
-  kDeletedFile          = 6,
-  kNewFile              = 7,
+  kComparator = 1,
+  kLogNumber = 2,
+  kNextFileNumber = 3,
+  kLastSequence = 4,
+  kCompactPointer = 5,
+  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
-
-  kColumnFamily         = 200,  // specify column family for version edit
-  kColumnFamilyAdd      = 201,
-  kColumnFamilyDrop     = 202,
-  kMaxColumnFamily      = 203,
+  kNewFile2 = 100,
+  kNewFile3 = 102,
+  kColumnFamily = 200,  // specify column family for version edit
+  kColumnFamilyAdd = 201,
+  kColumnFamilyDrop = 202,
+  kMaxColumnFamily = 203,
 };

+uint64_t PackFileNumberAndPathId(uint64_t number, uint64_t path_id) {
+  assert(number <= kFileNumberMask);
+  return number | (path_id * (kFileNumberMask + 1));
+}
+
 void VersionEdit::Clear() {
  comparator_.clear();
  max_level_ = 0;
@@ -59,7 +64,7 @@ void VersionEdit::Clear() {
  column_family_name_.clear();
 }

-void VersionEdit::EncodeTo(std::string* dst) const {
+bool VersionEdit::EncodeTo(std::string* dst) const {
  if (has_comparator_) {
    PutVarint32(dst, kComparator);
    PutLengthPrefixedSlice(dst, comparator_);
@@ -93,10 +98,22 @@ 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, kNewFile2);
+    if (!f.smallest.Valid() || !f.largest.Valid()) {
+      return false;
+    }
+    if (f.fd.GetPathId() == 0) {
+      // Use older format to make sure user can roll back the build if they
+      // don't config multiple DB paths.
+      PutVarint32(dst, kNewFile2);
+    } else {
+      PutVarint32(dst, kNewFile3);
+    }
    PutVarint32(dst, new_files_[i].first);  // level
-    PutVarint64(dst, f.number);
-    PutVarint64(dst, f.file_size);
+    PutVarint64(dst, f.fd.GetNumber());
+    if (f.fd.GetPathId() != 0) {
+      PutVarint32(dst, f.fd.GetPathId());
+    }
+    PutVarint64(dst, f.fd.GetFileSize());
    PutLengthPrefixedSlice(dst, f.smallest.Encode());
    PutLengthPrefixedSlice(dst, f.largest.Encode());
    PutVarint64(dst, f.smallest_seqno);
@@ -117,6 +134,7 @@ void VersionEdit::EncodeTo(std::string* dst) const {
  if (is_column_family_drop_) {
    PutVarint32(dst, kColumnFamilyDrop);
  }
+  return true;
 }

 static bool GetInternalKey(Slice* input, InternalKey* dst) {
@@ -230,12 +248,14 @@ Status VersionEdit::DecodeFrom(const Slice& src) {
        }
        break;

-      case kNewFile:
-        if (GetLevel(&input, &level, &msg) &&
-            GetVarint64(&input, &f.number) &&
-            GetVarint64(&input, &f.file_size) &&
+      case kNewFile: {
+        uint64_t number;
+        uint64_t file_size;
+        if (GetLevel(&input, &level, &msg) && GetVarint64(&input, &number) &&
+            GetVarint64(&input, &file_size) &&
            GetInternalKey(&input, &f.smallest) &&
            GetInternalKey(&input, &f.largest)) {
+          f.fd = FileDescriptor(number, 0, file_size);
          new_files_.push_back(std::make_pair(level, f));
        } else {
          if (!msg) {
@@ -243,15 +263,17 @@ Status VersionEdit::DecodeFrom(const Slice& src) {
          }
        }
        break;
-
-      case kNewFile2:
-        if (GetLevel(&input, &level, &msg) &&
-            GetVarint64(&input, &f.number) &&
-            GetVarint64(&input, &f.file_size) &&
+      }
+      case kNewFile2: {
+        uint64_t number;
+        uint64_t file_size;
+        if (GetLevel(&input, &level, &msg) && GetVarint64(&input, &number) &&
+            GetVarint64(&input, &file_size) &&
            GetInternalKey(&input, &f.smallest) &&
            GetInternalKey(&input, &f.largest) &&
            GetVarint64(&input, &f.smallest_seqno) &&
-            GetVarint64(&input, &f.largest_seqno) ) {
+            GetVarint64(&input, &f.largest_seqno)) {
+          f.fd = FileDescriptor(number, 0, file_size);
          new_files_.push_back(std::make_pair(level, f));
        } else {
          if (!msg) {
@@ -259,6 +281,27 @@ Status VersionEdit::DecodeFrom(const Slice& src) {
          }
        }
        break;
+      }
+
+      case kNewFile3: {
+        uint64_t number;
+        uint32_t path_id;
+        uint64_t file_size;
+        if (GetLevel(&input, &level, &msg) && GetVarint64(&input, &number) &&
+            GetVarint32(&input, &path_id) && GetVarint64(&input, &file_size) &&
+            GetInternalKey(&input, &f.smallest) &&
+            GetInternalKey(&input, &f.largest) &&
+            GetVarint64(&input, &f.smallest_seqno) &&
+            GetVarint64(&input, &f.largest_seqno)) {
+          f.fd = FileDescriptor(number, path_id, file_size);
+          new_files_.push_back(std::make_pair(level, f));
+        } else {
+          if (!msg) {
+            msg = "new-file2 entry";
+          }
+        }
+        break;
+      }

      case kColumnFamily:
        if (!GetVarint32(&input, &column_family_)) {
@@ -336,9 +379,9 @@ std::string VersionEdit::DebugString(bool hex_key) const {
    r.append("\n  AddFile: ");
    AppendNumberTo(&r, new_files_[i].first);
    r.append(" ");
-    AppendNumberTo(&r, f.number);
+    AppendNumberTo(&r, f.fd.GetNumber());
    r.append(" ");
-    AppendNumberTo(&r, f.file_size);
+    AppendNumberTo(&r, f.fd.GetFileSize());
    r.append(" ");
    r.append(f.smallest.DebugString(hex_key));
    r.append(" .. ");
@@ -14,36 +14,118 @@
 #include <string>
 #include "rocksdb/cache.h"
 #include "db/dbformat.h"
+#include "util/arena.h"
+#include "util/autovector.h"

 namespace rocksdb {

 class VersionSet;

+const uint64_t kFileNumberMask = 0x3FFFFFFFFFFFFFFF;
+
+extern uint64_t PackFileNumberAndPathId(uint64_t number, uint64_t path_id);
+
+// A copyable structure contains information needed to read data from an SST
+// file. It can contains a pointer to a table reader opened for the file, or
+// file number and size, which can be used to create a new table reader for it.
+// The behavior is undefined when a copied of the structure is used when the
+// file is not in any live version any more.
+struct FileDescriptor {
+  // Table reader in table_reader_handle
+  TableReader* table_reader;
+  uint64_t packed_number_and_path_id;
+  uint64_t file_size;  // File size in bytes
+
+  FileDescriptor() : FileDescriptor(0, 0, 0) {}
+
+  FileDescriptor(uint64_t number, uint32_t path_id, uint64_t file_size)
+      : table_reader(nullptr),
+        packed_number_and_path_id(PackFileNumberAndPathId(number, path_id)),
+        file_size(file_size) {}
+
+  FileDescriptor& operator=(const FileDescriptor& fd) {
+    table_reader = fd.table_reader;
+    packed_number_and_path_id = fd.packed_number_and_path_id;
+    file_size = fd.file_size;
+    return *this;
+  }
+
+  uint64_t GetNumber() const {
+    return packed_number_and_path_id & kFileNumberMask;
+  }
+  uint32_t GetPathId() const {
+    return packed_number_and_path_id / (kFileNumberMask + 1);
+  }
+  uint64_t GetFileSize() const { return file_size; }
+};
+
 struct FileMetaData {
  int refs;
-  int allowed_seeks;          // Seeks allowed until compaction
-  uint64_t number;
-  uint64_t file_size;         // File size in bytes
-  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
+  FileDescriptor fd;
+  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

  // Needs to be disposed when refs becomes 0.
  Cache::Handle* table_reader_handle;
-  // Table reader in table_reader_handle
-  TableReader* table_reader;

-  FileMetaData(uint64_t number, uint64_t file_size)
+  // Stats for compensating deletion entries during compaction
+
+  // File size compensated by deletion entry.
+  // This is updated in Version::UpdateTemporaryStats() first time when the
+  // file is created or loaded.  After it is updated, it is immutable.
+  uint64_t compensated_file_size;
+  uint64_t num_entries;            // the number of entries.
+  uint64_t num_deletions;          // the number of deletion entries.
+  uint64_t raw_key_size;           // total uncompressed key size.
+  uint64_t raw_value_size;         // total uncompressed value size.
+  bool init_stats_from_file;   // true if the data-entry stats of this file
+                               // has initialized from file.
+
+  FileMetaData()
      : refs(0),
-        allowed_seeks(1 << 30),
-        number(number),
-        file_size(file_size),
        being_compacted(false),
        table_reader_handle(nullptr),
-        table_reader(nullptr) {}
-  FileMetaData() : FileMetaData(0, 0) {}
+        compensated_file_size(0),
+        num_entries(0),
+        num_deletions(0),
+        raw_key_size(0),
+        raw_value_size(0),
+        init_stats_from_file(false) {}
+};
+
+// A compressed copy of file meta data that just contain
+// smallest and largest key's slice
+struct FdWithKeyRange {
+  FileDescriptor fd;
+  Slice smallest_key;    // slice that contain smallest key
+  Slice largest_key;     // slice that contain largest key
+
+  FdWithKeyRange()
+      : fd(),
+        smallest_key(),
+        largest_key() {
+  }
+
+  FdWithKeyRange(FileDescriptor fd,
+      Slice smallest_key, Slice largest_key)
+      : fd(fd),
+        smallest_key(smallest_key),
+        largest_key(largest_key) {
+  }
+};
+
+// Data structure to store an array of FdWithKeyRange in one level
+// Actual data is guaranteed to be stored closely
+struct FileLevel {
+  size_t num_files;
+  FdWithKeyRange* files;
+  FileLevel() {
+    num_files = 0;
+    files = nullptr;
+  }
 };

 class VersionEdit {
@@ -81,16 +163,13 @@ class VersionEdit {
  // Add the specified file at the specified number.
  // REQUIRES: This version has not been saved (see VersionSet::SaveTo)
  // REQUIRES: "smallest" and "largest" are smallest and largest keys in file
-  void AddFile(int level, uint64_t file,
-               uint64_t file_size,
-               const InternalKey& smallest,
-               const InternalKey& largest,
-               const SequenceNumber& smallest_seqno,
+  void AddFile(int level, uint64_t file, uint64_t file_path_id,
+               uint64_t file_size, const InternalKey& smallest,
+               const InternalKey& largest, const SequenceNumber& smallest_seqno,
               const SequenceNumber& largest_seqno) {
    assert(smallest_seqno <= largest_seqno);
    FileMetaData f;
-    f.number = file;
-    f.file_size = file_size;
+    f.fd = FileDescriptor(file, file_path_id, file_size);
    f.smallest = smallest;
    f.largest = largest;
    f.smallest_seqno = smallest_seqno;
@@ -133,13 +212,15 @@ class VersionEdit {
    is_column_family_drop_ = true;
  }

-  void EncodeTo(std::string* dst) const;
+  // return true on success.
+  bool EncodeTo(std::string* dst) const;
  Status DecodeFrom(const Slice& src);

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

 private:
  friend class VersionSet;
+  friend class Version;

  typedef std::set< std::pair<int, uint64_t>> DeletedFileSet;

@@ -30,11 +30,10 @@ TEST(VersionEditTest, EncodeDecode) {
  VersionEdit edit;
  for (int i = 0; i < 4; i++) {
    TestEncodeDecode(edit);
-    edit.AddFile(3, kBig + 300 + i, kBig + 400 + i,
+    edit.AddFile(3, kBig + 300 + i, kBig + 400 + i, 0,
                 InternalKey("foo", kBig + 500 + i, kTypeValue),
                 InternalKey("zoo", kBig + 600 + i, kTypeDeletion),
-                 kBig + 500 + i,
-                 kBig + 600 + i);
+                 kBig + 500 + i, kBig + 600 + i);
    edit.DeleteFile(4, kBig + 700 + i);
  }

@@ -45,6 +44,16 @@ TEST(VersionEditTest, EncodeDecode) {
  TestEncodeDecode(edit);
 }

+TEST(VersionEditTest, EncodeEmptyFile) {
+  VersionEdit edit;
+  edit.AddFile(0, 0, 0, 0,
+               InternalKey(),
+               InternalKey(),
+               0, 0);
+  std::string buffer;
+  ASSERT_TRUE(!edit.EncodeTo(&buffer));
+}
+
 TEST(VersionEditTest, ColumnFamilyTest) {
  VersionEdit edit;
  edit.SetColumnFamily(2);
@@ -34,6 +34,7 @@
 #include "db/column_family.h"
 #include "db/log_reader.h"
 #include "db/file_indexer.h"
+#include "db/write_controller.h"

 namespace rocksdb {

@@ -51,57 +52,69 @@ class MergeContext;
 class ColumnFamilyData;
 class ColumnFamilySet;
 class TableCache;
+class MergeIteratorBuilder;

-// Return the smallest index i such that files[i]->largest >= key.
-// Return files.size() if there is no such file.
-// REQUIRES: "files" contains a sorted list of non-overlapping files.
+// Return the smallest index i such that file_level.files[i]->largest >= key.
+// Return file_level.num_files if there is no such file.
+// REQUIRES: "file_level.files" contains a sorted list of
+// non-overlapping files.
 extern int FindFile(const InternalKeyComparator& icmp,
-                    const std::vector<FileMetaData*>& files,
+                    const FileLevel& file_level,
                    const Slice& key);

 // Returns true iff some file in "files" overlaps the user key range
 // [*smallest,*largest].
 // smallest==nullptr represents a key smaller than all keys in the DB.
 // largest==nullptr represents a key largest than all keys in the DB.
-// REQUIRES: If disjoint_sorted_files, files[] contains disjoint ranges
-//           in sorted order.
+// REQUIRES: If disjoint_sorted_files, file_level.files[]
+// contains disjoint ranges in sorted order.
 extern bool SomeFileOverlapsRange(
    const InternalKeyComparator& icmp,
    bool disjoint_sorted_files,
-    const std::vector<FileMetaData*>& files,
+    const FileLevel& file_level,
    const Slice* smallest_user_key,
    const Slice* largest_user_key);

+// Generate FileLevel from vector<FdWithKeyRange*>
+// Would copy smallest_key and largest_key data to sequential memory
+// arena: Arena used to allocate the memory
+extern void DoGenerateFileLevel(FileLevel* file_level,
+        const std::vector<FileMetaData*>& files,
+        Arena* arena);
+
 class Version {
 public:
  // Append to *iters a sequence of iterators that will
  // yield the contents of this Version when merged together.
  // REQUIRES: This version has been saved (see VersionSet::SaveTo)
+
  void AddIterators(const ReadOptions&, const EnvOptions& soptions,
-                    std::vector<Iterator*>* iters);
+                    MergeIteratorBuilder* merger_iter_builder);

  // Lookup the value for key.  If found, store it in *val and
-  // return OK.  Else return a non-OK status.  Fills *stats.
+  // return OK.  Else return a non-OK status.
  // 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, MergeContext* merge_context, GetStats* stats,
+           Status* status, MergeContext* merge_context,
           bool* value_found = nullptr);

-  // Adds "stats" into the current state.  Returns true if a new
-  // compaction may need to be triggered, false otherwise.
-  // REQUIRES: lock is held
-  bool UpdateStats(const GetStats& stats);
-
  // Updates internal structures that keep track of compaction scores
  // We use compaction scores to figure out which compaction to do next
  // REQUIRES: If Version is not yet saved to current_, it can be called without
  // a lock. Once a version is saved to current_, call only with mutex held
-  void ComputeCompactionScore(std::vector<uint64_t>& size_being_compacted);
+  void ComputeCompactionScore(
+      const MutableCFOptions& mutable_cf_options,
+      std::vector<uint64_t>& size_being_compacted);
+
+  // Generate file_levels_ from files_
+  void GenerateFileLevels();
+
+  // Update scores, pre-calculated variables. It needs to be called before
+  // applying the version to the version set.
+  void PrepareApply(
+      const MutableCFOptions& mutable_cf_options,
+      std::vector<uint64_t>& size_being_compacted);

  // Reference count management (so Versions do not disappear out from
  // under live iterators)
@@ -160,7 +173,8 @@ class Version {

  // 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,
+  int PickLevelForMemTableOutput(const MutableCFOptions& mutable_cf_options,
+                                 const Slice& smallest_user_key,
                                 const Slice& largest_user_key);

  int NumberLevels() const { return num_levels_; }
@@ -169,15 +183,15 @@ class Version {
  int NumLevelFiles(int level) const { return files_[level].size(); }

  // Return the combined file size of all files at the specified level.
-  int64_t NumLevelBytes(int level) const;
+  uint64_t NumLevelBytes(int level) const;

  // Return a human-readable short (single-line) summary of the number
  // of files per level.  Uses *scratch as backing store.
  struct LevelSummaryStorage {
-    char buffer[100];
+    char buffer[1000];
  };
  struct FileSummaryStorage {
-    char buffer[1000];
+    char buffer[3000];
  };
  const char* LevelSummary(LevelSummaryStorage* scratch) const;
  // Return a human-readable short (single-line) summary of files
@@ -189,7 +203,7 @@ class Version {
  int64_t MaxNextLevelOverlappingBytes();

  // Add all files listed in the current version to *live.
-  void AddLiveFiles(std::set<uint64_t>* live);
+  void AddLiveFiles(std::vector<FileDescriptor>* live);

  // Return a human readable string that describes this version's contents.
  std::string DebugString(bool hex = false) const;
@@ -197,12 +211,35 @@ class Version {
  // Returns the version nuber of this version
  uint64_t GetVersionNumber() const { return version_number_; }

+  uint64_t GetAverageValueSize() const {
+    if (num_non_deletions_ == 0) {
+      return 0;
+    }
+    assert(total_raw_key_size_ + total_raw_value_size_ > 0);
+    assert(total_file_size_ > 0);
+    return total_raw_value_size_ / num_non_deletions_ * total_file_size_ /
+           (total_raw_key_size_ + total_raw_value_size_);
+  }
+
+  // REQUIRES: lock is held
+  // On success, "tp" will contains the table properties of the file
+  // specified in "file_meta".  If the file name of "file_meta" is
+  // known ahread, passing it by a non-null "fname" can save a
+  // file-name conversion.
+  Status GetTableProperties(std::shared_ptr<const TableProperties>* tp,
+                            const FileMetaData* file_meta,
+                            const std::string* fname = nullptr);
+
  // REQUIRES: lock is held
  // On success, *props will be populated with all SSTables' table properties.
  // The keys of `props` are the sst file name, the values of `props` are the
  // tables' propertis, represented as shared_ptr.
  Status GetPropertiesOfAllTables(TablePropertiesCollection* props);

+  uint64_t GetEstimatedActiveKeys();
+
+  size_t GetMemoryUsageByTableReaders();
+
  // used to sort files by size
  struct Fsize {
    int index;
@@ -213,18 +250,33 @@ class Version {
  friend class Compaction;
  friend class VersionSet;
  friend class DBImpl;
+  friend class CompactedDBImpl;
  friend class ColumnFamilyData;
  friend class CompactionPicker;
  friend class LevelCompactionPicker;
  friend class UniversalCompactionPicker;
  friend class FIFOCompactionPicker;
+  friend class ForwardIterator;
+  friend class InternalStats;

  class LevelFileNumIterator;
  class LevelFileIteratorState;

-  bool PrefixMayMatch(const ReadOptions& options, Iterator* level_iter,
+  bool PrefixMayMatch(const ReadOptions& read_options, Iterator* level_iter,
                      const Slice& internal_prefix) const;

+  // Update num_non_empty_levels_.
+  void UpdateNumNonEmptyLevels();
+
+  // The helper function of UpdateTemporaryStats, which may fill the missing
+  // fields of file_mata from its associated TableProperties.
+  // Returns true if it does initialize FileMetaData.
+  bool MaybeInitializeFileMetaData(FileMetaData* file_meta);
+
+  // Update the temporary stats associated with the current version.
+  // This temporary stats will be used in compaction.
+  void UpdateTemporaryStats();
+
  // Sort all files for this version based on their file size and
  // record results in files_by_size_. The largest files are listed first.
  void UpdateFilesBySize();
@@ -234,13 +286,19 @@ class Version {
  const Comparator* user_comparator_;
  TableCache* table_cache_;
  const MergeOperator* merge_operator_;
+
+  autovector<FileLevel> file_levels_;   // A copy of list of files per level
  Logger* info_log_;
  Statistics* db_statistics_;
+  int num_levels_;              // Number of levels
+  int num_non_empty_levels_;    // Number of levels. Any level larger than it
+                                // is guaranteed to be empty.
+  FileIndexer file_indexer_;
  VersionSet* vset_;            // VersionSet to which this Version belongs
+  Arena arena_;                 // Used to allocate space for file_levels_
  Version* next_;               // Next version in linked list
  Version* prev_;               // Previous version in linked list
  int refs_;                    // Number of live refs to this version
-  int num_levels_;              // Number of levels

  // List of files per level, files in each level are arranged
  // in increasing order of keys
@@ -261,11 +319,7 @@ class Version {
  // that on a running system, we need to look at only the first
  // few largest files because a new version is created every few
  // seconds/minutes (because of concurrent compactions).
-  static const int number_of_files_to_sort_ = 50;
-
-  // Next file to compact based on seek stats.
-  FileMetaData* file_to_compact_;
-  int file_to_compact_level_;
+  static const size_t number_of_files_to_sort_ = 50;

  // Level that should be compacted next and its compaction score.
  // Score < 1 means compaction is not strictly needed.  These fields
@@ -274,15 +328,25 @@ class Version {
  // These are used to pick the best compaction level
  std::vector<double> compaction_score_;
  std::vector<int> compaction_level_;
-  double max_compaction_score_; // max score in l1 to ln-1
-  int max_compaction_score_level_; // level on which max score occurs
+  double max_compaction_score_ = 0.0;   // max score in l1 to ln-1
+  int max_compaction_score_level_ = 0;  // level on which max score occurs

  // A version number that uniquely represents this version. This is
  // used for debugging and logging purposes only.
  uint64_t version_number_;

  Version(ColumnFamilyData* cfd, VersionSet* vset, uint64_t version_number = 0);
-  FileIndexer file_indexer_;
+
+  // total file size
+  uint64_t total_file_size_;
+  // the total size of all raw keys.
+  uint64_t total_raw_key_size_;
+  // the total size of all raw values.
+  uint64_t total_raw_value_size_;
+  // total number of non-deletion entries
+  uint64_t num_non_deletions_;
+  // total number of deletion entries
+  uint64_t num_deletions_;

  ~Version();

@@ -299,8 +363,9 @@ class Version {

 class VersionSet {
 public:
-  VersionSet(const std::string& dbname, const DBOptions* options,
-             const EnvOptions& storage_options, Cache* table_cache);
+  VersionSet(const std::string& dbname, const DBOptions* db_options,
+             const EnvOptions& env_options, Cache* table_cache,
+             WriteController* write_controller);
  ~VersionSet();

  // Apply *edit to the current version to form a new descriptor that
@@ -309,7 +374,9 @@ class VersionSet {
  // column_family_options has to be set if edit is column family add
  // REQUIRES: *mu is held on entry.
  // REQUIRES: no other thread concurrently calls LogAndApply()
-  Status LogAndApply(ColumnFamilyData* column_family_data, VersionEdit* edit,
+  Status LogAndApply(ColumnFamilyData* column_family_data,
+                     const MutableCFOptions& mutable_cf_options,
+                     VersionEdit* edit,
                     port::Mutex* mu, Directory* db_directory = nullptr,
                     bool new_descriptor_log = false,
                     const ColumnFamilyOptions* column_family_options =
@@ -338,7 +405,7 @@ class VersionSet {
  // among [4-6] contains files.
  static Status ReduceNumberOfLevels(const std::string& dbname,
                                     const Options* options,
-                                     const EnvOptions& storage_options,
+                                     const EnvOptions& env_options,
                                     int new_levels);

  // printf contents (for debugging)
@@ -360,7 +427,7 @@ class VersionSet {
  // Arrange to reuse "file_number" unless a newer file number has
  // already been allocated.
  // REQUIRES: "file_number" was returned by a call to NewFileNumber().
-  void ReuseFileNumber(uint64_t file_number) {
+  void ReuseLogFileNumber(uint64_t file_number) {
    if (next_file_number_ == file_number + 1) {
      next_file_number_ = file_number;
    }
@@ -401,7 +468,7 @@ class VersionSet {
  Iterator* MakeInputIterator(Compaction* c);

  // Add all files listed in any live version to *live.
-  void AddLiveFiles(std::vector<uint64_t>* live_list);
+  void AddLiveFiles(std::vector<FileDescriptor>* live_list);

  // Return the approximate offset in the database of the data for
  // "key" as of version "v".
@@ -447,14 +514,14 @@ class VersionSet {
  bool ManifestContains(uint64_t manifest_file_number,
                        const std::string& record) const;

-  ColumnFamilyData* CreateColumnFamily(const ColumnFamilyOptions& options,
+  ColumnFamilyData* CreateColumnFamily(const ColumnFamilyOptions& cf_options,
                                       VersionEdit* edit);

  std::unique_ptr<ColumnFamilySet> column_family_set_;

  Env* const env_;
  const std::string dbname_;
-  const DBOptions* const options_;
+  const DBOptions* const db_options_;
  uint64_t next_file_number_;
  uint64_t manifest_file_number_;
  uint64_t pending_manifest_file_number_;
@@ -475,12 +542,12 @@ class VersionSet {

  std::vector<FileMetaData*> obsolete_files_;

-  // storage options for all reads and writes except compactions
-  const EnvOptions& storage_options_;
+  // env options for all reads and writes except compactions
+  const EnvOptions& env_options_;

-  // storage options used for compactions. This is a copy of
-  // storage_options_ but with readaheads set to readahead_compactions_.
-  const EnvOptions storage_options_compactions_;
+  // env options used for compactions. This is a copy of
+  // env_options_ but with readaheads set to readahead_compactions_.
+  const EnvOptions env_options_compactions_;

  // No copying allowed
  VersionSet(const VersionSet&);
@@ -14,14 +14,15 @@

 namespace rocksdb {

-class FindFileTest {
+class GenerateFileLevelTest {
 public:
  std::vector<FileMetaData*> files_;
-  bool disjoint_sorted_files_;
+  FileLevel file_level_;
+  Arena arena_;

-  FindFileTest() : disjoint_sorted_files_(true) { }
+  GenerateFileLevelTest() { }

-  ~FindFileTest() {
+  ~GenerateFileLevelTest() {
    for (unsigned int i = 0; i < files_.size(); i++) {
      delete files_[i];
    }
@@ -31,29 +32,108 @@ class FindFileTest {
           SequenceNumber smallest_seq = 100,
           SequenceNumber largest_seq = 100) {
    FileMetaData* f = new FileMetaData;
-    f->number = files_.size() + 1;
+    f->fd = FileDescriptor(files_.size() + 1, 0, 0);
    f->smallest = InternalKey(smallest, smallest_seq, kTypeValue);
    f->largest = InternalKey(largest, largest_seq, kTypeValue);
    files_.push_back(f);
  }

+  int Compare() {
+    int diff = 0;
+    for (size_t i = 0; i < files_.size(); i++) {
+      if (file_level_.files[i].fd.GetNumber() != files_[i]->fd.GetNumber()) {
+        diff++;
+      }
+    }
+    return diff;
+  }
+};
+
+TEST(GenerateFileLevelTest, Empty) {
+  DoGenerateFileLevel(&file_level_, files_, &arena_);
+  ASSERT_EQ(0u, file_level_.num_files);
+  ASSERT_EQ(0, Compare());
+}
+
+TEST(GenerateFileLevelTest, Single) {
+  Add("p", "q");
+  DoGenerateFileLevel(&file_level_, files_, &arena_);
+  ASSERT_EQ(1u, file_level_.num_files);
+  ASSERT_EQ(0, Compare());
+}
+
+
+TEST(GenerateFileLevelTest, Multiple) {
+  Add("150", "200");
+  Add("200", "250");
+  Add("300", "350");
+  Add("400", "450");
+  DoGenerateFileLevel(&file_level_, files_, &arena_);
+  ASSERT_EQ(4u, file_level_.num_files);
+  ASSERT_EQ(0, Compare());
+}
+
+class FindLevelFileTest {
+ public:
+  FileLevel file_level_;
+  bool disjoint_sorted_files_;
+  Arena arena_;
+
+  FindLevelFileTest() : disjoint_sorted_files_(true) { }
+
+  ~FindLevelFileTest() {
+  }
+
+  void LevelFileInit(size_t num = 0) {
+    char* mem = arena_.AllocateAligned(num * sizeof(FdWithKeyRange));
+    file_level_.files = new (mem)FdWithKeyRange[num];
+    file_level_.num_files = 0;
+  }
+
+  void Add(const char* smallest, const char* largest,
+           SequenceNumber smallest_seq = 100,
+           SequenceNumber largest_seq = 100) {
+    InternalKey smallest_key = InternalKey(smallest, smallest_seq, kTypeValue);
+    InternalKey largest_key = InternalKey(largest, largest_seq, kTypeValue);
+
+    Slice smallest_slice = smallest_key.Encode();
+    Slice largest_slice = largest_key.Encode();
+
+    char* mem = arena_.AllocateAligned(
+        smallest_slice.size() + largest_slice.size());
+    memcpy(mem, smallest_slice.data(), smallest_slice.size());
+    memcpy(mem + smallest_slice.size(), largest_slice.data(),
+        largest_slice.size());
+
+    // add to file_level_
+    size_t num = file_level_.num_files;
+    auto& file = file_level_.files[num];
+    file.fd = FileDescriptor(num + 1, 0, 0);
+    file.smallest_key = Slice(mem, smallest_slice.size());
+    file.largest_key = Slice(mem + smallest_slice.size(),
+        largest_slice.size());
+    file_level_.num_files++;
+  }
+
  int Find(const char* key) {
    InternalKey target(key, 100, kTypeValue);
    InternalKeyComparator cmp(BytewiseComparator());
-    return FindFile(cmp, files_, target.Encode());
+    return FindFile(cmp, file_level_, target.Encode());
  }

  bool Overlaps(const char* smallest, const char* largest) {
    InternalKeyComparator cmp(BytewiseComparator());
    Slice s(smallest != nullptr ? smallest : "");
    Slice l(largest != nullptr ? largest : "");
-    return SomeFileOverlapsRange(cmp, disjoint_sorted_files_, files_,
+    return SomeFileOverlapsRange(cmp, disjoint_sorted_files_, file_level_,
                                 (smallest != nullptr ? &s : nullptr),
                                 (largest != nullptr ? &l : nullptr));
  }
 };

-TEST(FindFileTest, Empty) {
+TEST(FindLevelFileTest, LevelEmpty) {
+  LevelFileInit(0);
+
  ASSERT_EQ(0, Find("foo"));
  ASSERT_TRUE(! Overlaps("a", "z"));
  ASSERT_TRUE(! Overlaps(nullptr, "z"));
@@ -61,7 +141,9 @@ TEST(FindFileTest, Empty) {
  ASSERT_TRUE(! Overlaps(nullptr, nullptr));
 }

-TEST(FindFileTest, Single) {
+TEST(FindLevelFileTest, LevelSingle) {
+  LevelFileInit(1);
+
  Add("p", "q");
  ASSERT_EQ(0, Find("a"));
  ASSERT_EQ(0, Find("p"));
@@ -91,8 +173,9 @@ TEST(FindFileTest, Single) {
  ASSERT_TRUE(Overlaps(nullptr, nullptr));
 }

+TEST(FindLevelFileTest, LevelMultiple) {
+  LevelFileInit(4);

-TEST(FindFileTest, Multiple) {
  Add("150", "200");
  Add("200", "250");
  Add("300", "350");
@@ -130,7 +213,9 @@ TEST(FindFileTest, Multiple) {
  ASSERT_TRUE(Overlaps("450", "500"));
 }

-TEST(FindFileTest, MultipleNullBoundaries) {
+TEST(FindLevelFileTest, LevelMultipleNullBoundaries) {
+  LevelFileInit(4);
+
  Add("150", "200");
  Add("200", "250");
  Add("300", "350");
@@ -150,7 +235,9 @@ TEST(FindFileTest, MultipleNullBoundaries) {
  ASSERT_TRUE(Overlaps("450", nullptr));
 }

-TEST(FindFileTest, OverlapSequenceChecks) {
+TEST(FindLevelFileTest, LevelOverlapSequenceChecks) {
+  LevelFileInit(1);
+
  Add("200", "200", 5000, 3000);
  ASSERT_TRUE(! Overlaps("199", "199"));
  ASSERT_TRUE(! Overlaps("201", "300"));
@@ -159,7 +246,9 @@ TEST(FindFileTest, OverlapSequenceChecks) {
  ASSERT_TRUE(Overlaps("200", "210"));
 }

-TEST(FindFileTest, OverlappingFiles) {
+TEST(FindLevelFileTest, LevelOverlappingFiles) {
+  LevelFileInit(2);
+
  Add("150", "600");
  Add("400", "500");
  disjoint_sorted_files_ = false;
@@ -23,10 +23,10 @@
 //    data: uint8[len]

 #include "rocksdb/write_batch.h"
-#include "rocksdb/options.h"
 #include "rocksdb/merge_operator.h"
 #include "db/dbformat.h"
 #include "db/db_impl.h"
+#include "db/column_family.h"
 #include "db/memtable.h"
 #include "db/snapshot.h"
 #include "db/write_batch_internal.h"
@@ -80,6 +80,58 @@ int WriteBatch::Count() const {
  return WriteBatchInternal::Count(this);
 }

+Status ReadRecordFromWriteBatch(Slice* input, char* tag,
+                                uint32_t* column_family, Slice* key,
+                                Slice* value, Slice* blob) {
+  assert(key != nullptr && value != nullptr);
+  *tag = (*input)[0];
+  input->remove_prefix(1);
+  *column_family = 0;  // default
+  switch (*tag) {
+    case kTypeColumnFamilyValue:
+      if (!GetVarint32(input, column_family)) {
+        return Status::Corruption("bad WriteBatch Put");
+      }
+    // intentional fallthrough
+    case kTypeValue:
+      if (!GetLengthPrefixedSlice(input, key) ||
+          !GetLengthPrefixedSlice(input, value)) {
+        return Status::Corruption("bad WriteBatch Put");
+      }
+      break;
+    case kTypeColumnFamilyDeletion:
+      if (!GetVarint32(input, column_family)) {
+        return Status::Corruption("bad WriteBatch Delete");
+      }
+    // intentional fallthrough
+    case kTypeDeletion:
+      if (!GetLengthPrefixedSlice(input, key)) {
+        return Status::Corruption("bad WriteBatch Delete");
+      }
+      break;
+    case kTypeColumnFamilyMerge:
+      if (!GetVarint32(input, column_family)) {
+        return Status::Corruption("bad WriteBatch Merge");
+      }
+    // intentional fallthrough
+    case kTypeMerge:
+      if (!GetLengthPrefixedSlice(input, key) ||
+          !GetLengthPrefixedSlice(input, value)) {
+        return Status::Corruption("bad WriteBatch Merge");
+      }
+      break;
+    case kTypeLogData:
+      assert(blob != nullptr);
+      if (!GetLengthPrefixedSlice(input, blob)) {
+        return Status::Corruption("bad WriteBatch Blob");
+      }
+      break;
+    default:
+      return Status::Corruption("unknown WriteBatch tag");
+  }
+  return Status::OK();
+}
+
 Status WriteBatch::Iterate(Handler* handler) const {
  Slice input(rep_);
  if (input.size() < kHeader) {
@@ -91,57 +143,33 @@ Status WriteBatch::Iterate(Handler* handler) const {
  int found = 0;
  Status s;
  while (s.ok() && !input.empty() && handler->Continue()) {
-    char tag = input[0];
-    input.remove_prefix(1);
+    char tag = 0;
    uint32_t column_family = 0;  // default
+
+    s = ReadRecordFromWriteBatch(&input, &tag, &column_family, &key, &value,
+                                 &blob);
+    if (!s.ok()) {
+      return s;
+    }
+
    switch (tag) {
      case kTypeColumnFamilyValue:
-        if (!GetVarint32(&input, &column_family)) {
-          return Status::Corruption("bad WriteBatch Put");
-        }
-      // intentional fallthrough
      case kTypeValue:
-        if (GetLengthPrefixedSlice(&input, &key) &&
-            GetLengthPrefixedSlice(&input, &value)) {
-          s = handler->PutCF(column_family, key, value);
-          found++;
-        } else {
-          return Status::Corruption("bad WriteBatch Put");
-        }
+        s = handler->PutCF(column_family, key, value);
+        found++;
        break;
      case kTypeColumnFamilyDeletion:
-        if (!GetVarint32(&input, &column_family)) {
-          return Status::Corruption("bad WriteBatch Delete");
-        }
-      // intentional fallthrough
      case kTypeDeletion:
-        if (GetLengthPrefixedSlice(&input, &key)) {
-          s = handler->DeleteCF(column_family, key);
-          found++;
-        } else {
-          return Status::Corruption("bad WriteBatch Delete");
-        }
+        s = handler->DeleteCF(column_family, key);
+        found++;
        break;
      case kTypeColumnFamilyMerge:
-        if (!GetVarint32(&input, &column_family)) {
-          return Status::Corruption("bad WriteBatch Merge");
-        }
-      // intentional fallthrough
      case kTypeMerge:
-        if (GetLengthPrefixedSlice(&input, &key) &&
-            GetLengthPrefixedSlice(&input, &value)) {
-          s = handler->MergeCF(column_family, key, value);
-          found++;
-        } else {
-          return Status::Corruption("bad WriteBatch Merge");
-        }
+        s = handler->MergeCF(column_family, key, value);
+        found++;
        break;
      case kTypeLogData:
-        if (GetLengthPrefixedSlice(&input, &blob)) {
-          handler->LogData(blob);
-        } else {
-          return Status::Corruption("bad WriteBatch Blob");
-        }
+        handler->LogData(blob);
        break;
      default:
        return Status::Corruption("unknown WriteBatch tag");
@@ -186,17 +214,6 @@ void WriteBatchInternal::Put(WriteBatch* b, uint32_t column_family_id,
  PutLengthPrefixedSlice(&b->rep_, value);
 }

-namespace {
-inline uint32_t GetColumnFamilyID(ColumnFamilyHandle* column_family) {
-  uint32_t column_family_id = 0;
-  if (column_family != nullptr) {
-    auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
-    column_family_id = cfh->GetID();
-  }
-  return column_family_id;
-}
-}  // namespace
-
 void WriteBatch::Put(ColumnFamilyHandle* column_family, const Slice& key,
                     const Slice& value) {
  WriteBatchInternal::Put(this, GetColumnFamilyID(column_family), key, value);
@@ -236,6 +253,23 @@ void WriteBatch::Delete(ColumnFamilyHandle* column_family, const Slice& key) {
  WriteBatchInternal::Delete(this, GetColumnFamilyID(column_family), key);
 }

+void WriteBatchInternal::Delete(WriteBatch* b, uint32_t column_family_id,
+                                const SliceParts& key) {
+  WriteBatchInternal::SetCount(b, WriteBatchInternal::Count(b) + 1);
+  if (column_family_id == 0) {
+    b->rep_.push_back(static_cast<char>(kTypeDeletion));
+  } else {
+    b->rep_.push_back(static_cast<char>(kTypeColumnFamilyDeletion));
+    PutVarint32(&b->rep_, column_family_id);
+  }
+  PutLengthPrefixedSliceParts(&b->rep_, key);
+}
+
+void WriteBatch::Delete(ColumnFamilyHandle* column_family,
+                        const SliceParts& key) {
+  WriteBatchInternal::Delete(this, GetColumnFamilyID(column_family), key);
+}
+
 void WriteBatchInternal::Merge(WriteBatch* b, uint32_t column_family_id,
                               const Slice& key, const Slice& value) {
  WriteBatchInternal::SetCount(b, WriteBatchInternal::Count(b) + 1);
@@ -264,17 +298,17 @@ class MemTableInserter : public WriteBatch::Handler {
 public:
  SequenceNumber sequence_;
  ColumnFamilyMemTables* cf_mems_;
-  bool recovery_;
+  bool ignore_missing_column_families_;
  uint64_t log_number_;
  DBImpl* db_;
  const bool dont_filter_deletes_;

  MemTableInserter(SequenceNumber sequence, ColumnFamilyMemTables* cf_mems,
-                   bool recovery, uint64_t log_number, DB* db,
-                   const bool dont_filter_deletes)
+                   bool ignore_missing_column_families, uint64_t log_number,
+                   DB* db, const bool dont_filter_deletes)
      : sequence_(sequence),
        cf_mems_(cf_mems),
-        recovery_(recovery),
+        ignore_missing_column_families_(ignore_missing_column_families),
        log_number_(log_number),
        db_(reinterpret_cast<DBImpl*>(db)),
        dont_filter_deletes_(dont_filter_deletes) {
@@ -286,12 +320,18 @@ class MemTableInserter : public WriteBatch::Handler {

  bool SeekToColumnFamily(uint32_t column_family_id, Status* s) {
    bool found = cf_mems_->Seek(column_family_id);
-    if (recovery_ && (!found || log_number_ < cf_mems_->GetLogNumber())) {
-      // if in recovery envoronment:
-      // * If column family was not found, it might mean that the WAL write
-      // batch references to the column family that was dropped after the
-      // insert. We don't want to fail the whole write batch in that case -- we
-      // just ignore the update.
+    if (!found) {
+      if (ignore_missing_column_families_) {
+        *s = Status::OK();
+      } else {
+        *s = Status::InvalidArgument(
+            "Invalid column family specified in write batch");
+      }
+      return false;
+    }
+    if (log_number_ != 0 && log_number_ < cf_mems_->GetLogNumber()) {
+      // This is true only in recovery environment (log_number_ is always 0 in
+      // non-recovery, regular write code-path)
      // * If log_number_ < cf_mems_->GetLogNumber(), this means that column
      // family already contains updates from this log. We can't apply updates
      // twice because of update-in-place or merge workloads -- ignore the
@@ -299,18 +339,8 @@ class MemTableInserter : public WriteBatch::Handler {
      *s = Status::OK();
      return false;
    }
-    if (!found) {
-      assert(!recovery_);
-      // If the column family was not found in non-recovery enviornment
-      // (client's write code-path), we have to fail the write and return
-      // the failure status to the client.
-      *s = Status::InvalidArgument(
-          "Invalid column family specified in write batch");
-      return false;
-    }
    return true;
  }
-
  virtual Status PutCF(uint32_t column_family_id, const Slice& key,
                       const Slice& value) {
    Status seek_status;
@@ -319,14 +349,15 @@ class MemTableInserter : public WriteBatch::Handler {
      return seek_status;
    }
    MemTable* mem = cf_mems_->GetMemTable();
-    const Options* options = cf_mems_->GetOptions();
-    if (!options->inplace_update_support) {
+    auto* ioptions = mem->GetImmutableOptions();
+    auto* moptions = mem->GetMemTableOptions();
+    if (!moptions->inplace_update_support) {
      mem->Add(sequence_, kTypeValue, key, value);
-    } else if (options->inplace_callback == nullptr) {
+    } else if (moptions->inplace_callback == nullptr) {
      mem->Update(sequence_, key, value);
-      RecordTick(options->statistics.get(), NUMBER_KEYS_UPDATED);
+      RecordTick(ioptions->statistics, NUMBER_KEYS_UPDATED);
    } else {
-      if (mem->UpdateCallback(sequence_, key, value, *options)) {
+      if (mem->UpdateCallback(sequence_, key, value)) {
      } else {
        // key not found in memtable. Do sst get, update, add
        SnapshotImpl read_from_snapshot;
@@ -345,17 +376,17 @@ class MemTableInserter : public WriteBatch::Handler {

        char* prev_buffer = const_cast<char*>(prev_value.c_str());
        uint32_t prev_size = prev_value.size();
-        auto status = options->inplace_callback(s.ok() ? prev_buffer : nullptr,
-                                                s.ok() ? &prev_size : nullptr,
-                                                value, &merged_value);
+        auto status = moptions->inplace_callback(s.ok() ? prev_buffer : nullptr,
+                                                 s.ok() ? &prev_size : nullptr,
+                                                 value, &merged_value);
        if (status == UpdateStatus::UPDATED_INPLACE) {
          // prev_value is updated in-place with final value.
          mem->Add(sequence_, kTypeValue, key, Slice(prev_buffer, prev_size));
-          RecordTick(options->statistics.get(), NUMBER_KEYS_WRITTEN);
+          RecordTick(ioptions->statistics, NUMBER_KEYS_WRITTEN);
        } else if (status == UpdateStatus::UPDATED) {
          // merged_value contains the final value.
          mem->Add(sequence_, kTypeValue, key, Slice(merged_value));
-          RecordTick(options->statistics.get(), NUMBER_KEYS_WRITTEN);
+          RecordTick(ioptions->statistics, NUMBER_KEYS_WRITTEN);
        }
      }
    }
@@ -363,6 +394,7 @@ class MemTableInserter : public WriteBatch::Handler {
    // sequence number. Even if the update eventually fails and does not result
    // in memtable add/update.
    sequence_++;
+    cf_mems_->CheckMemtableFull();
    return Status::OK();
  }

@@ -374,17 +406,18 @@ class MemTableInserter : public WriteBatch::Handler {
      return seek_status;
    }
    MemTable* mem = cf_mems_->GetMemTable();
-    const Options* options = cf_mems_->GetOptions();
+    auto* ioptions = mem->GetImmutableOptions();
+    auto* moptions = mem->GetMemTableOptions();
    bool perform_merge = false;

-    if (options->max_successive_merges > 0 && db_ != nullptr) {
+    if (moptions->max_successive_merges > 0 && db_ != nullptr) {
      LookupKey lkey(key, sequence_);

      // Count the number of successive merges at the head
      // of the key in the memtable
      size_t num_merges = mem->CountSuccessiveMergeEntries(lkey);

-      if (num_merges >= options->max_successive_merges) {
+      if (num_merges >= moptions->max_successive_merges) {
        perform_merge = true;
      }
    }
@@ -408,19 +441,19 @@ class MemTableInserter : public WriteBatch::Handler {
      Slice get_value_slice = Slice(get_value);

      // 2) Apply this merge
-      auto merge_operator = options->merge_operator.get();
+      auto merge_operator = ioptions->merge_operator;
      assert(merge_operator);

      std::deque<std::string> operands;
      operands.push_front(value.ToString());
      std::string new_value;
      if (!merge_operator->FullMerge(key, &get_value_slice, operands,
-                                     &new_value, options->info_log.get())) {
+                                     &new_value, ioptions->info_log)) {
          // Failed to merge!
-        RecordTick(options->statistics.get(), NUMBER_MERGE_FAILURES);
+        RecordTick(ioptions->statistics, NUMBER_MERGE_FAILURES);

-          // Store the delta in memtable
-          perform_merge = false;
+        // Store the delta in memtable
+        perform_merge = false;
      } else {
        // 3) Add value to memtable
        mem->Add(sequence_, kTypeValue, key, new_value);
@@ -433,6 +466,7 @@ class MemTableInserter : public WriteBatch::Handler {
    }

    sequence_++;
+    cf_mems_->CheckMemtableFull();
    return Status::OK();
  }

@@ -443,8 +477,9 @@ class MemTableInserter : public WriteBatch::Handler {
      return seek_status;
    }
    MemTable* mem = cf_mems_->GetMemTable();
-    const Options* options = cf_mems_->GetOptions();
-    if (!dont_filter_deletes_ && options->filter_deletes) {
+    auto* ioptions = mem->GetImmutableOptions();
+    auto* moptions = mem->GetMemTableOptions();
+    if (!dont_filter_deletes_ && moptions->filter_deletes) {
      SnapshotImpl read_from_snapshot;
      read_from_snapshot.number_ = sequence_;
      ReadOptions ropts;
@@ -455,12 +490,13 @@ class MemTableInserter : public WriteBatch::Handler {
        cf_handle = db_->DefaultColumnFamily();
      }
      if (!db_->KeyMayExist(ropts, cf_handle, key, &value)) {
-        RecordTick(options->statistics.get(), NUMBER_FILTERED_DELETES);
+        RecordTick(ioptions->statistics, NUMBER_FILTERED_DELETES);
        return Status::OK();
      }
    }
    mem->Add(sequence_, kTypeDeletion, key, Slice());
    sequence_++;
+    cf_mems_->CheckMemtableFull();
    return Status::OK();
  }
 };
@@ -468,10 +504,12 @@ class MemTableInserter : public WriteBatch::Handler {

 Status WriteBatchInternal::InsertInto(const WriteBatch* b,
                                      ColumnFamilyMemTables* memtables,
-                                      bool recovery, uint64_t log_number,
-                                      DB* db, const bool dont_filter_deletes) {
+                                      bool ignore_missing_column_families,
+                                      uint64_t log_number, DB* db,
+                                      const bool dont_filter_deletes) {
  MemTableInserter inserter(WriteBatchInternal::Sequence(b), memtables,
-                            recovery, log_number, db, dont_filter_deletes);
+                            ignore_missing_column_families, log_number, db,
+                            dont_filter_deletes);
  return b->Iterate(&inserter);
 }

@@ -28,6 +28,7 @@ class ColumnFamilyMemTables {
  virtual MemTable* GetMemTable() const = 0;
  virtual const Options* GetOptions() const = 0;
  virtual ColumnFamilyHandle* GetColumnFamilyHandle() = 0;
+  virtual void CheckMemtableFull() = 0;
 };

 class ColumnFamilyMemTablesDefault : public ColumnFamilyMemTables {
@@ -54,6 +55,8 @@ class ColumnFamilyMemTablesDefault : public ColumnFamilyMemTables {

  ColumnFamilyHandle* GetColumnFamilyHandle() override { return nullptr; }

+  void CheckMemtableFull() override {}
+
 private:
  bool ok_;
  MemTable* mem_;
@@ -71,6 +74,9 @@ class WriteBatchInternal {
  static void Put(WriteBatch* batch, uint32_t column_family_id,
                  const SliceParts& key, const SliceParts& value);

+  static void Delete(WriteBatch* batch, uint32_t column_family_id,
+                     const SliceParts& key);
+
  static void Delete(WriteBatch* batch, uint32_t column_family_id,
                     const Slice& key);

@@ -103,18 +109,18 @@ class WriteBatchInternal {
  // Inserts batch entries into memtable
  // If dont_filter_deletes is false AND options.filter_deletes is true,
  // then --> Drops deletes in batch if db->KeyMayExist returns false
-  // If recovery == true, this means InsertInto is executed on a recovery
-  // code-path. WriteBatch referencing a dropped column family can be
-  // found on a recovery code-path and should be ignored (recovery should not
-  // fail). Additionally, the memtable will be updated only if
+  // If ignore_missing_column_families == true. WriteBatch referencing
+  // non-existing column family should be ignored.
+  // However, if ignore_missing_column_families == false, any WriteBatch
+  // referencing non-existing column family will return a InvalidArgument()
+  // failure.
+  //
+  // If log_number is non-zero, the memtable will be updated only if
  // memtables->GetLogNumber() >= log_number
-  // However, if recovery == false, any WriteBatch referencing
-  // non-existing column family will return a failure. Also, log_number is
-  // ignored in that case
  static Status InsertInto(const WriteBatch* batch,
                           ColumnFamilyMemTables* memtables,
-                           bool recovery = false, uint64_t log_number = 0,
-                           DB* db = nullptr,
+                           bool ignore_missing_column_families = false,
+                           uint64_t log_number = 0, DB* db = nullptr,
                           const bool dont_filter_deletes = true);

  static void Append(WriteBatch* dst, const WriteBatch* src);
@@ -15,8 +15,10 @@
 #include "db/write_batch_internal.h"
 #include "rocksdb/env.h"
 #include "rocksdb/memtablerep.h"
+#include "rocksdb/utilities/write_batch_with_index.h"
 #include "util/logging.h"
 #include "util/testharness.h"
+#include "util/scoped_arena_iterator.h"

 namespace rocksdb {

@@ -25,13 +27,16 @@ static std::string PrintContents(WriteBatch* b) {
  auto factory = std::make_shared<SkipListFactory>();
  Options options;
  options.memtable_factory = factory;
-  MemTable* mem = new MemTable(cmp, options);
+  ImmutableCFOptions ioptions(options);
+  MemTable* mem = new MemTable(cmp, ioptions,
+      MemTableOptions(MutableCFOptions(options, ioptions), options));
  mem->Ref();
  std::string state;
  ColumnFamilyMemTablesDefault cf_mems_default(mem, &options);
  Status s = WriteBatchInternal::InsertInto(b, &cf_mems_default);
  int count = 0;
-  Iterator* iter = mem->NewIterator(ReadOptions());
+  Arena arena;
+  ScopedArenaIterator iter(mem->NewIterator(ReadOptions(), &arena));
  for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
    ParsedInternalKey ikey;
    memset((void *)&ikey, 0, sizeof(ikey));
@@ -66,7 +71,6 @@ static std::string PrintContents(WriteBatch* b) {
    state.append("@");
    state.append(NumberToString(ikey.sequence));
  }
-  delete iter;
  if (!s.ok()) {
    state.append(s.ToString());
  } else if (count != WriteBatchInternal::Count(b)) {
@@ -286,6 +290,9 @@ class ColumnFamilyHandleImplDummy : public ColumnFamilyHandleImpl {
  explicit ColumnFamilyHandleImplDummy(int id)
      : ColumnFamilyHandleImpl(nullptr, nullptr, nullptr), id_(id) {}
  uint32_t GetID() const override { return id_; }
+  const Comparator* user_comparator() const override {
+    return BytewiseComparator();
+  }

 private:
  uint32_t id_;
@@ -316,6 +323,88 @@ TEST(WriteBatchTest, ColumnFamiliesBatchTest) {
      handler.seen);
 }

+TEST(WriteBatchTest, ColumnFamiliesBatchWithIndexTest) {
+  WriteBatchWithIndex batch;
+  ColumnFamilyHandleImplDummy zero(0), two(2), three(3), eight(8);
+  batch.Put(&zero, Slice("foo"), Slice("bar"));
+  batch.Put(&two, Slice("twofoo"), Slice("bar2"));
+  batch.Put(&eight, Slice("eightfoo"), Slice("bar8"));
+  batch.Delete(&eight, Slice("eightfoo"));
+  batch.Merge(&three, Slice("threethree"), Slice("3three"));
+  batch.Put(&zero, Slice("foo"), Slice("bar"));
+  batch.Merge(Slice("omom"), Slice("nom"));
+
+  std::unique_ptr<WBWIIterator> iter;
+
+  iter.reset(batch.NewIterator(&eight));
+  iter->Seek("eightfoo");
+  ASSERT_OK(iter->status());
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ(WriteType::kPutRecord, iter->Entry().type);
+  ASSERT_EQ("eightfoo", iter->Entry().key.ToString());
+  ASSERT_EQ("bar8", iter->Entry().value.ToString());
+
+  iter->Next();
+  ASSERT_OK(iter->status());
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ(WriteType::kDeleteRecord, iter->Entry().type);
+  ASSERT_EQ("eightfoo", iter->Entry().key.ToString());
+
+  iter->Next();
+  ASSERT_OK(iter->status());
+  ASSERT_TRUE(!iter->Valid());
+
+  iter.reset(batch.NewIterator());
+  iter->Seek("gggg");
+  ASSERT_OK(iter->status());
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ(WriteType::kMergeRecord, iter->Entry().type);
+  ASSERT_EQ("omom", iter->Entry().key.ToString());
+  ASSERT_EQ("nom", iter->Entry().value.ToString());
+
+  iter->Next();
+  ASSERT_OK(iter->status());
+  ASSERT_TRUE(!iter->Valid());
+
+  iter.reset(batch.NewIterator(&zero));
+  iter->Seek("foo");
+  ASSERT_OK(iter->status());
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ(WriteType::kPutRecord, iter->Entry().type);
+  ASSERT_EQ("foo", iter->Entry().key.ToString());
+  ASSERT_EQ("bar", iter->Entry().value.ToString());
+
+  iter->Next();
+  ASSERT_OK(iter->status());
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ(WriteType::kPutRecord, iter->Entry().type);
+  ASSERT_EQ("foo", iter->Entry().key.ToString());
+  ASSERT_EQ("bar", iter->Entry().value.ToString());
+
+  iter->Next();
+  ASSERT_OK(iter->status());
+  ASSERT_TRUE(iter->Valid());
+  ASSERT_EQ(WriteType::kMergeRecord, iter->Entry().type);
+  ASSERT_EQ("omom", iter->Entry().key.ToString());
+  ASSERT_EQ("nom", iter->Entry().value.ToString());
+
+  iter->Next();
+  ASSERT_OK(iter->status());
+  ASSERT_TRUE(!iter->Valid());
+
+  TestHandler handler;
+  batch.GetWriteBatch()->Iterate(&handler);
+  ASSERT_EQ(
+      "Put(foo, bar)"
+      "PutCF(2, twofoo, bar2)"
+      "PutCF(8, eightfoo, bar8)"
+      "DeleteCF(8, eightfoo)"
+      "MergeCF(3, threethree, 3three)"
+      "Put(foo, bar)"
+      "Merge(omom, nom)",
+      handler.seen);
+}
+
 }  // namespace rocksdb

 int main(int argc, char** argv) {
@@ -0,0 +1,37 @@
+//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+
+#include "db/write_controller.h"
+
+#include <cassert>
+
+namespace rocksdb {
+
+std::unique_ptr<WriteControllerToken> WriteController::GetStopToken() {
+  ++total_stopped_;
+  return std::unique_ptr<WriteControllerToken>(new StopWriteToken(this));
+}
+
+std::unique_ptr<WriteControllerToken> WriteController::GetDelayToken(
+    uint64_t delay_us) {
+  total_delay_us_ += delay_us;
+  return std::unique_ptr<WriteControllerToken>(
+      new DelayWriteToken(this, delay_us));
+}
+
+bool WriteController::IsStopped() const { return total_stopped_ > 0; }
+uint64_t WriteController::GetDelay() const { return total_delay_us_; }
+
+StopWriteToken::~StopWriteToken() {
+  assert(controller_->total_stopped_ >= 1);
+  --controller_->total_stopped_;
+}
+
+DelayWriteToken::~DelayWriteToken() {
+  assert(controller_->total_delay_us_ >= delay_us_);
+  controller_->total_delay_us_ -= delay_us_;
+}
+
+}  // namespace rocksdb
@@ -0,0 +1,78 @@
+//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+
+#pragma once
+
+#include <stdint.h>
+
+#include <memory>
+
+namespace rocksdb {
+
+class WriteControllerToken;
+
+// WriteController is controlling write stalls in our write code-path. Write
+// stalls happen when compaction can't keep up with write rate.
+// All of the methods here (including WriteControllerToken's destructors) need
+// to be called while holding DB mutex
+class WriteController {
+ public:
+  WriteController() : total_stopped_(0), total_delay_us_(0) {}
+  ~WriteController() = default;
+
+  // When an actor (column family) requests a stop token, all writes will be
+  // stopped until the stop token is released (deleted)
+  std::unique_ptr<WriteControllerToken> GetStopToken();
+  // When an actor (column family) requests a delay token, total delay for all
+  // writes will be increased by delay_us. The delay will last until delay token
+  // is released
+  std::unique_ptr<WriteControllerToken> GetDelayToken(uint64_t delay_us);
+
+  // these two metods are querying the state of the WriteController
+  bool IsStopped() const;
+  uint64_t GetDelay() const;
+
+ private:
+  friend class WriteControllerToken;
+  friend class StopWriteToken;
+  friend class DelayWriteToken;
+
+  int total_stopped_;
+  uint64_t total_delay_us_;
+};
+
+class WriteControllerToken {
+ public:
+  explicit WriteControllerToken(WriteController* controller)
+      : controller_(controller) {}
+  virtual ~WriteControllerToken() {}
+
+ protected:
+  WriteController* controller_;
+
+ private:
+  // no copying allowed
+  WriteControllerToken(const WriteControllerToken&) = delete;
+  void operator=(const WriteControllerToken&) = delete;
+};
+
+class StopWriteToken : public WriteControllerToken {
+ public:
+  explicit StopWriteToken(WriteController* controller)
+      : WriteControllerToken(controller) {}
+  virtual ~StopWriteToken();
+};
+
+class DelayWriteToken : public WriteControllerToken {
+ public:
+  DelayWriteToken(WriteController* controller, uint64_t delay_us)
+      : WriteControllerToken(controller), delay_us_(delay_us) {}
+  virtual ~DelayWriteToken();
+
+ private:
+  uint64_t delay_us_;
+};
+
+}  // namespace rocksdb
@@ -0,0 +1,40 @@
+//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+//
+#include "db/write_controller.h"
+
+#include "util/testharness.h"
+
+namespace rocksdb {
+
+class WriteControllerTest {};
+
+TEST(WriteControllerTest, SanityTest) {
+  WriteController controller;
+  auto stop_token_1 = controller.GetStopToken();
+  auto stop_token_2 = controller.GetStopToken();
+
+  ASSERT_EQ(true, controller.IsStopped());
+  stop_token_1.reset();
+  ASSERT_EQ(true, controller.IsStopped());
+  stop_token_2.reset();
+  ASSERT_EQ(false, controller.IsStopped());
+
+  auto delay_token_1 = controller.GetDelayToken(5);
+  ASSERT_EQ(static_cast<uint64_t>(5), controller.GetDelay());
+  auto delay_token_2 = controller.GetDelayToken(8);
+  ASSERT_EQ(static_cast<uint64_t>(13), controller.GetDelay());
+
+  delay_token_2.reset();
+  ASSERT_EQ(static_cast<uint64_t>(5), controller.GetDelay());
+  delay_token_1.reset();
+  ASSERT_EQ(static_cast<uint64_t>(0), controller.GetDelay());
+  delay_token_1.reset();
+  ASSERT_EQ(false, controller.IsStopped());
+}
+
+}  // namespace rocksdb
+
+int main(int argc, char** argv) { return rocksdb::test::RunAllTests(); }
@@ -0,0 +1,147 @@
+//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+
+#include "db/write_thread.h"
+
+namespace rocksdb {
+
+Status WriteThread::EnterWriteThread(WriteThread::Writer* w,
+                                     uint64_t expiration_time) {
+  // the following code block pushes the current writer "w" into the writer
+  // queue "writers_" and wait until one of the following conditions met:
+  // 1. the job of "w" has been done by some other writers.
+  // 2. "w" becomes the first writer in "writers_"
+  // 3. "w" timed-out.
+  writers_.push_back(w);
+
+  bool timed_out = false;
+  while (!w->done && w != writers_.front()) {
+    if (expiration_time == 0) {
+      w->cv.Wait();
+    } else if (w->cv.TimedWait(expiration_time)) {
+      if (w->in_batch_group) {
+        // then it means the front writer is currently doing the
+        // write on behalf of this "timed-out" writer.  Then it
+        // should wait until the write completes.
+        expiration_time = 0;
+      } else {
+        timed_out = true;
+        break;
+      }
+    }
+  }
+
+  if (timed_out) {
+#ifndef NDEBUG
+    bool found = false;
+#endif
+    for (auto iter = writers_.begin(); iter != writers_.end(); iter++) {
+      if (*iter == w) {
+        writers_.erase(iter);
+#ifndef NDEBUG
+        found = true;
+#endif
+        break;
+      }
+    }
+#ifndef NDEBUG
+    assert(found);
+#endif
+    // writers_.front() might still be in cond_wait without a time-out.
+    // As a result, we need to signal it to wake it up.  Otherwise no
+    // one else will wake him up, and RocksDB will hang.
+    if (!writers_.empty()) {
+      writers_.front()->cv.Signal();
+    }
+    return Status::TimedOut();
+  }
+  return Status::OK();
+}
+
+void WriteThread::ExitWriteThread(WriteThread::Writer* w,
+                                  WriteThread::Writer* last_writer,
+                                  Status status) {
+  // Pop out the current writer and all writers being pushed before the
+  // current writer from the writer queue.
+  while (!writers_.empty()) {
+    Writer* ready = writers_.front();
+    writers_.pop_front();
+    if (ready != w) {
+      ready->status = status;
+      ready->done = true;
+      ready->cv.Signal();
+    }
+    if (ready == last_writer) break;
+  }
+
+  // Notify new head of write queue
+  if (!writers_.empty()) {
+    writers_.front()->cv.Signal();
+  }
+}
+
+// This function will be called only when the first writer succeeds.
+// All writers in the to-be-built batch group will be processed.
+//
+// REQUIRES: Writer list must be non-empty
+// REQUIRES: First writer must have a non-nullptr batch
+void WriteThread::BuildBatchGroup(WriteThread::Writer** last_writer,
+                                  autovector<WriteBatch*>* write_batch_group) {
+  assert(!writers_.empty());
+  Writer* first = writers_.front();
+  assert(first->batch != nullptr);
+
+  size_t size = WriteBatchInternal::ByteSize(first->batch);
+  write_batch_group->push_back(first->batch);
+
+  // Allow the group to grow up to a maximum size, but if the
+  // original write is small, limit the growth so we do not slow
+  // down the small write too much.
+  size_t max_size = 1 << 20;
+  if (size <= (128<<10)) {
+    max_size = size + (128<<10);
+  }
+
+  *last_writer = first;
+  std::deque<Writer*>::iterator iter = writers_.begin();
+  ++iter;  // Advance past "first"
+  for (; iter != writers_.end(); ++iter) {
+    Writer* w = *iter;
+    if (w->sync && !first->sync) {
+      // Do not include a sync write into a batch handled by a non-sync write.
+      break;
+    }
+
+    if (!w->disableWAL && first->disableWAL) {
+      // Do not include a write that needs WAL into a batch that has
+      // WAL disabled.
+      break;
+    }
+
+    if (w->timeout_hint_us < first->timeout_hint_us) {
+      // Do not include those writes with shorter timeout.  Otherwise, we might
+      // execute a write that should instead be aborted because of timeout.
+      break;
+    }
+
+    if (w->batch == nullptr) {
+      // Do not include those writes with nullptr batch. Those are not writes,
+      // those are something else. They want to be alone
+      break;
+    }
+
+    size += WriteBatchInternal::ByteSize(w->batch);
+    if (size > max_size) {
+      // Do not make batch too big
+      break;
+    }
+
+    write_batch_group->push_back(w->batch);
+    w->in_batch_group = true;
+    *last_writer = w;
+  }
+}
+
+}  // namespace rocksdb
@@ -0,0 +1,80 @@
+//  Copyright (c) 2013, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+
+#pragma once
+
+#include <stdint.h>
+#include <deque>
+#include <limits>
+#include "rocksdb/status.h"
+#include "db/write_batch_internal.h"
+#include "util/autovector.h"
+#include "port/port.h"
+
+namespace rocksdb {
+
+class WriteThread {
+ public:
+  static const uint64_t kNoTimeOut = std::numeric_limits<uint64_t>::max();
+  // Information kept for every waiting writer
+  struct Writer {
+    Status status;
+    WriteBatch* batch;
+    bool sync;
+    bool disableWAL;
+    bool in_batch_group;
+    bool done;
+    uint64_t timeout_hint_us;
+    port::CondVar cv;
+
+    explicit Writer(port::Mutex* mu)
+        : batch(nullptr),
+          sync(false),
+          disableWAL(false),
+          in_batch_group(false),
+          done(false),
+          timeout_hint_us(kNoTimeOut),
+          cv(mu) {}
+  };
+
+  WriteThread() = default;
+  ~WriteThread() = default;
+
+  // Before applying write operation (such as DBImpl::Write, DBImpl::Flush)
+  // thread should grab the mutex_ and be the first on writers queue.
+  // EnterWriteThread is used for it.
+  // Be aware! Writer's job can be done by other thread (see DBImpl::Write
+  // for examples), so check it via w.done before applying changes.
+  //
+  // Writer* w:                writer to be placed in the queue
+  // uint64_t expiration_time: maximum time to be in the queue
+  // See also: ExitWriteThread
+  // REQUIRES: db mutex held
+  Status EnterWriteThread(Writer* w, uint64_t expiration_time);
+
+  // After doing write job, we need to remove already used writers from
+  // writers_ queue and notify head of the queue about it.
+  // ExitWriteThread is used for this.
+  //
+  // Writer* w:           Writer, that was added by EnterWriteThread function
+  // Writer* last_writer: Since we can join a few Writers (as DBImpl::Write
+  //                      does)
+  //                      we should pass last_writer as a parameter to
+  //                      ExitWriteThread
+  //                      (if you don't touch other writers, just pass w)
+  // Status status:       Status of write operation
+  // See also: EnterWriteThread
+  // REQUIRES: db mutex held
+  void ExitWriteThread(Writer* w, Writer* last_writer, Status status);
+
+  void BuildBatchGroup(Writer** last_writer,
+                       autovector<WriteBatch*>* write_batch_group);
+
+ private:
+  // Queue of writers.
+  std::deque<Writer*> writers_;
+};
+
+}  // namespace rocksdb
@@ -642,10 +642,6 @@ Default:1, i.e. pick maxfilesize amount of data as the source of a compaction.
 <li> <code>Options::max_grandparent_overlap_factor</code> -   Control maximum bytes of overlaps in grandparent (i.e., level+2) before we
 stop building a single file in a level->level+1 compaction.
 <p>
-<li> <code>Options::disable_seek_compaction</code> -  Disable compaction triggered by seek.
-With bloomfilter and fast storage, a miss on one level is very cheap if the file handle is cached in table cache
-(which is true if max_open_files is large).
-<p>
 <li> <code>Options::max_background_compactions</code> - Maximum number of concurrent background jobs, submitted to
 the default LOW priority thread pool
 </ul>
@@ -6,8 +6,8 @@
  Use of this source code is governed by a BSD-style license that can be
  found in the LICENSE file. See the AUTHORS file for names of contributors.

-  C bindings for leveldb.  May be useful as a stable ABI that can be
-  used by programs that keep leveldb in a shared library, or for
+  C bindings for rocksdb.  May be useful as a stable ABI that can be
+  used by programs that keep rocksdb in a shared library, or for
  a JNI api.

  Does not support:
@@ -56,8 +56,18 @@ extern "C" {

 typedef struct rocksdb_t                 rocksdb_t;
 typedef struct rocksdb_cache_t           rocksdb_cache_t;
+typedef struct rocksdb_compactionfilter_t rocksdb_compactionfilter_t;
+typedef struct rocksdb_compactionfiltercontext_t
+    rocksdb_compactionfiltercontext_t;
+typedef struct rocksdb_compactionfilterfactory_t
+    rocksdb_compactionfilterfactory_t;
+typedef struct rocksdb_compactionfilterv2_t
+    rocksdb_compactionfilterv2_t;
+typedef struct rocksdb_compactionfilterfactoryv2_t
+    rocksdb_compactionfilterfactoryv2_t;
 typedef struct rocksdb_comparator_t      rocksdb_comparator_t;
 typedef struct rocksdb_env_t             rocksdb_env_t;
+typedef struct rocksdb_fifo_compaction_options_t rocksdb_fifo_compaction_options_t;
 typedef struct rocksdb_filelock_t        rocksdb_filelock_t;
 typedef struct rocksdb_filterpolicy_t    rocksdb_filterpolicy_t;
 typedef struct rocksdb_flushoptions_t    rocksdb_flushoptions_t;
@@ -65,6 +75,8 @@ typedef struct rocksdb_iterator_t        rocksdb_iterator_t;
 typedef struct rocksdb_logger_t          rocksdb_logger_t;
 typedef struct rocksdb_mergeoperator_t   rocksdb_mergeoperator_t;
 typedef struct rocksdb_options_t         rocksdb_options_t;
+typedef struct rocksdb_block_based_table_options_t
+    rocksdb_block_based_table_options_t;
 typedef struct rocksdb_randomfile_t      rocksdb_randomfile_t;
 typedef struct rocksdb_readoptions_t     rocksdb_readoptions_t;
 typedef struct rocksdb_seqfile_t         rocksdb_seqfile_t;
@@ -75,6 +87,7 @@ typedef struct rocksdb_writebatch_t      rocksdb_writebatch_t;
 typedef struct rocksdb_writeoptions_t    rocksdb_writeoptions_t;
 typedef struct rocksdb_universal_compaction_options_t rocksdb_universal_compaction_options_t;
 typedef struct rocksdb_livefiles_t     rocksdb_livefiles_t;
+typedef struct rocksdb_column_family_handle_t rocksdb_column_family_handle_t;

 /* DB operations */

@@ -89,6 +102,45 @@ extern rocksdb_t* rocksdb_open_for_read_only(
    unsigned char error_if_log_file_exist,
    char** errptr);

+extern rocksdb_t* rocksdb_open_column_families(
+    const rocksdb_options_t* options,
+    const char* name,
+    int num_column_families,
+    const char** column_family_names,
+    const rocksdb_options_t** column_family_options,
+    rocksdb_column_family_handle_t** column_family_handles,
+    char** errptr);
+
+extern rocksdb_t* rocksdb_open_for_read_only_column_families(
+    const rocksdb_options_t* options,
+    const char* name,
+    int num_column_families,
+    const char** column_family_names,
+    const rocksdb_options_t** column_family_options,
+    rocksdb_column_family_handle_t** column_family_handles,
+    unsigned char error_if_log_file_exist,
+    char** errptr);
+
+char** rocksdb_list_column_families(
+    const rocksdb_options_t* options,
+    const char* name,
+    size_t* lencf,
+    char** errptr);
+void rocksdb_list_column_families_destroy(char** list, size_t len);
+
+extern rocksdb_column_family_handle_t* rocksdb_create_column_family(
+    rocksdb_t* db,
+    const rocksdb_options_t* column_family_options,
+    const char* column_family_name,
+    char** errptr);
+
+extern void rocksdb_drop_column_family(
+    rocksdb_t* db,
+    rocksdb_column_family_handle_t* handle,
+    char** errptr);
+
+extern void rocksdb_column_family_handle_destroy(rocksdb_column_family_handle_t*);
+
 extern void rocksdb_close(rocksdb_t* db);

 extern void rocksdb_put(
@@ -98,12 +150,27 @@ extern void rocksdb_put(
    const char* val, size_t vallen,
    char** errptr);

+extern void rocksdb_put_cf(
+    rocksdb_t* db,
+    const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t keylen,
+    const char* val, size_t vallen,
+    char** errptr);
+
 extern void rocksdb_delete(
    rocksdb_t* db,
    const rocksdb_writeoptions_t* options,
    const char* key, size_t keylen,
    char** errptr);

+void rocksdb_delete_cf(
+    rocksdb_t* db,
+    const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t keylen,
+    char** errptr);
+
 extern void rocksdb_merge(
    rocksdb_t* db,
    const rocksdb_writeoptions_t* options,
@@ -111,6 +178,14 @@ extern void rocksdb_merge(
    const char* val, size_t vallen,
    char** errptr);

+extern void rocksdb_merge_cf(
+    rocksdb_t* db,
+    const rocksdb_writeoptions_t* options,
+    rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t keylen,
+    const char* val, size_t vallen,
+    char** errptr);
+
 extern void rocksdb_write(
    rocksdb_t* db,
    const rocksdb_writeoptions_t* options,
@@ -126,10 +201,23 @@ extern char* rocksdb_get(
    size_t* vallen,
    char** errptr);

+extern char* rocksdb_get_cf(
+    rocksdb_t* db,
+    const rocksdb_readoptions_t* options,
+    rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t keylen,
+    size_t* vallen,
+    char** errptr);
+
 extern rocksdb_iterator_t* rocksdb_create_iterator(
    rocksdb_t* db,
    const rocksdb_readoptions_t* options);

+extern rocksdb_iterator_t* rocksdb_create_iterator_cf(
+    rocksdb_t* db,
+    const rocksdb_readoptions_t* options,
+    rocksdb_column_family_handle_t* column_family);
+
 extern const rocksdb_snapshot_t* rocksdb_create_snapshot(
    rocksdb_t* db);

@@ -143,6 +231,11 @@ extern char* rocksdb_property_value(
    rocksdb_t* db,
    const char* propname);

+extern char* rocksdb_property_value_cf(
+    rocksdb_t* db,
+    rocksdb_column_family_handle_t* column_family,
+    const char* propname);
+
 extern void rocksdb_approximate_sizes(
    rocksdb_t* db,
    int num_ranges,
@@ -150,11 +243,25 @@ extern void rocksdb_approximate_sizes(
    const char* const* range_limit_key, const size_t* range_limit_key_len,
    uint64_t* sizes);

+extern void rocksdb_approximate_sizes_cf(
+    rocksdb_t* db,
+    rocksdb_column_family_handle_t* column_family,
+    int num_ranges,
+    const char* const* range_start_key, const size_t* range_start_key_len,
+    const char* const* range_limit_key, const size_t* range_limit_key_len,
+    uint64_t* sizes);
+
 extern void rocksdb_compact_range(
    rocksdb_t* db,
    const char* start_key, size_t start_key_len,
    const char* limit_key, size_t limit_key_len);

+extern void rocksdb_compact_range_cf(
+    rocksdb_t* db,
+    rocksdb_column_family_handle_t* column_family,
+    const char* start_key, size_t start_key_len,
+    const char* limit_key, size_t limit_key_len);
+
 extern void rocksdb_delete_file(
    rocksdb_t* db,
    const char* name);
@@ -204,6 +311,8 @@ extern void rocksdb_iter_get_error(const rocksdb_iterator_t*, char** errptr);
 /* Write batch */

 extern rocksdb_writebatch_t* rocksdb_writebatch_create();
+extern rocksdb_writebatch_t* rocksdb_writebatch_create_from(const char* rep,
+                                                            size_t size);
 extern void rocksdb_writebatch_destroy(rocksdb_writebatch_t*);
 extern void rocksdb_writebatch_clear(rocksdb_writebatch_t*);
 extern int rocksdb_writebatch_count(rocksdb_writebatch_t*);
@@ -211,13 +320,27 @@ extern void rocksdb_writebatch_put(
    rocksdb_writebatch_t*,
    const char* key, size_t klen,
    const char* val, size_t vlen);
+extern void rocksdb_writebatch_put_cf(
+    rocksdb_writebatch_t*,
+    rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t klen,
+    const char* val, size_t vlen);
 extern void rocksdb_writebatch_merge(
    rocksdb_writebatch_t*,
    const char* key, size_t klen,
    const char* val, size_t vlen);
+extern void rocksdb_writebatch_merge_cf(
+    rocksdb_writebatch_t*,
+    rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t klen,
+    const char* val, size_t vlen);
 extern void rocksdb_writebatch_delete(
    rocksdb_writebatch_t*,
    const char* key, size_t klen);
+extern void rocksdb_writebatch_delete_cf(
+    rocksdb_writebatch_t*,
+    rocksdb_column_family_handle_t* column_family,
+    const char* key, size_t klen);
 extern void rocksdb_writebatch_iterate(
    rocksdb_writebatch_t*,
    void* state,
@@ -225,24 +348,68 @@ extern void rocksdb_writebatch_iterate(
    void (*deleted)(void*, const char* k, size_t klen));
 extern const char* rocksdb_writebatch_data(rocksdb_writebatch_t*, size_t *size);

+/* Block based table options */
+
+extern rocksdb_block_based_table_options_t*
+    rocksdb_block_based_options_create();
+extern void rocksdb_block_based_options_destroy(
+    rocksdb_block_based_table_options_t* options);
+extern void rocksdb_block_based_options_set_block_size(
+    rocksdb_block_based_table_options_t* options, size_t block_size);
+extern void rocksdb_block_based_options_set_block_size_deviation(
+    rocksdb_block_based_table_options_t* options, int block_size_deviation);
+extern void rocksdb_block_based_options_set_block_restart_interval(
+    rocksdb_block_based_table_options_t* options, int block_restart_interval);
+extern void rocksdb_block_based_options_set_filter_policy(
+    rocksdb_block_based_table_options_t* options,
+    rocksdb_filterpolicy_t* filter_policy);
+extern void rocksdb_block_based_options_set_no_block_cache(
+    rocksdb_block_based_table_options_t* options,
+    unsigned char no_block_cache);
+extern void rocksdb_block_based_options_set_block_cache(
+    rocksdb_block_based_table_options_t* options, rocksdb_cache_t* block_cache);
+extern void rocksdb_block_based_options_set_block_cache_compressed(
+    rocksdb_block_based_table_options_t* options,
+    rocksdb_cache_t* block_cache_compressed);
+extern void rocksdb_block_based_options_set_whole_key_filtering(
+    rocksdb_block_based_table_options_t*, unsigned char);
+extern void rocksdb_options_set_block_based_table_factory(
+    rocksdb_options_t *opt, rocksdb_block_based_table_options_t* table_options);
+
 /* Options */

 extern rocksdb_options_t* rocksdb_options_create();
 extern void rocksdb_options_destroy(rocksdb_options_t*);
+extern void rocksdb_options_increase_parallelism(
+    rocksdb_options_t* opt, int total_threads);
+extern void rocksdb_options_optimize_for_point_lookup(
+    rocksdb_options_t* opt, uint64_t block_cache_size_mb);
+extern void rocksdb_options_optimize_level_style_compaction(
+    rocksdb_options_t* opt, uint64_t memtable_memory_budget);
+extern void rocksdb_options_optimize_universal_style_compaction(
+    rocksdb_options_t* opt, uint64_t memtable_memory_budget);
+extern void rocksdb_options_set_compaction_filter(
+    rocksdb_options_t*,
+    rocksdb_compactionfilter_t*);
+extern void rocksdb_options_set_compaction_filter_factory(
+    rocksdb_options_t*, rocksdb_compactionfilterfactory_t*);
+extern void rocksdb_options_set_compaction_filter_factory_v2(
+    rocksdb_options_t*,
+    rocksdb_compactionfilterfactoryv2_t*);
 extern void rocksdb_options_set_comparator(
    rocksdb_options_t*,
    rocksdb_comparator_t*);
-extern void rocksdb_options_set_merge_operator(rocksdb_options_t*,
-                                               rocksdb_mergeoperator_t*);
+extern void rocksdb_options_set_merge_operator(
+    rocksdb_options_t*,
+    rocksdb_mergeoperator_t*);
 extern void rocksdb_options_set_compression_per_level(
  rocksdb_options_t* opt,
  int* level_values,
  size_t num_levels);
-extern void rocksdb_options_set_filter_policy(
-    rocksdb_options_t*,
-    rocksdb_filterpolicy_t*);
 extern void rocksdb_options_set_create_if_missing(
    rocksdb_options_t*, unsigned char);
+extern void rocksdb_options_set_create_missing_column_families(
+    rocksdb_options_t*, unsigned char);
 extern void rocksdb_options_set_error_if_exists(
    rocksdb_options_t*, unsigned char);
 extern void rocksdb_options_set_paranoid_checks(
@@ -252,13 +419,8 @@ extern void rocksdb_options_set_info_log(rocksdb_options_t*, rocksdb_logger_t*);
 extern void rocksdb_options_set_info_log_level(rocksdb_options_t*, int);
 extern void rocksdb_options_set_write_buffer_size(rocksdb_options_t*, size_t);
 extern void rocksdb_options_set_max_open_files(rocksdb_options_t*, int);
-extern void rocksdb_options_set_cache(rocksdb_options_t*, rocksdb_cache_t*);
-extern void rocksdb_options_set_cache_compressed(rocksdb_options_t*, rocksdb_cache_t*);
-extern void rocksdb_options_set_block_size(rocksdb_options_t*, size_t);
-extern void rocksdb_options_set_block_restart_interval(rocksdb_options_t*, int);
 extern void rocksdb_options_set_compression_options(
    rocksdb_options_t*, int, int, int);
-extern void rocksdb_options_set_whole_key_filtering(rocksdb_options_t*, unsigned char);
 extern void rocksdb_options_set_prefix_extractor(
    rocksdb_options_t*, rocksdb_slicetransform_t*);
 extern void rocksdb_options_set_num_levels(rocksdb_options_t*, int);
@@ -309,8 +471,6 @@ extern void rocksdb_options_set_arena_block_size(
    rocksdb_options_t*, size_t);
 extern void rocksdb_options_set_use_fsync(
    rocksdb_options_t*, int);
-extern void rocksdb_options_set_db_stats_log_interval(
-    rocksdb_options_t*, int);
 extern void rocksdb_options_set_db_log_dir(
    rocksdb_options_t*, const char*);
 extern void rocksdb_options_set_wal_dir(
@@ -353,7 +513,6 @@ extern void rocksdb_options_set_max_sequential_skip_in_iterations(
    rocksdb_options_t*, uint64_t);
 extern void rocksdb_options_set_disable_data_sync(rocksdb_options_t*, int);
 extern void rocksdb_options_set_disable_auto_compactions(rocksdb_options_t*, int);
-extern void rocksdb_options_set_disable_seek_compaction(rocksdb_options_t*, int);
 extern void rocksdb_options_set_delete_obsolete_files_period_micros(
    rocksdb_options_t*, uint64_t);
 extern void rocksdb_options_set_source_compaction_factor(rocksdb_options_t*, int);
@@ -378,8 +537,6 @@ extern void rocksdb_options_set_min_partial_merge_operands(
    rocksdb_options_t*, uint32_t);
 extern void rocksdb_options_set_bloom_locality(
    rocksdb_options_t*, uint32_t);
-extern void rocksdb_options_set_allow_thread_local(
-    rocksdb_options_t*, unsigned char);
 extern void rocksdb_options_set_inplace_update_support(
    rocksdb_options_t*, unsigned char);
 extern void rocksdb_options_set_inplace_update_num_locks(
@@ -397,10 +554,77 @@ extern void rocksdb_options_set_compression(rocksdb_options_t*, int);

 enum {
  rocksdb_level_compaction = 0,
-  rocksdb_universal_compaction = 1
+  rocksdb_universal_compaction = 1,
+  rocksdb_fifo_compaction = 2
 };
 extern void rocksdb_options_set_compaction_style(rocksdb_options_t*, int);
 extern void rocksdb_options_set_universal_compaction_options(rocksdb_options_t*, rocksdb_universal_compaction_options_t*);
+extern void rocksdb_options_set_fifo_compaction_options(rocksdb_options_t* opt,
+    rocksdb_fifo_compaction_options_t* fifo);
+
+/* Compaction Filter */
+
+extern rocksdb_compactionfilter_t* rocksdb_compactionfilter_create(
+    void* state,
+    void (*destructor)(void*),
+    unsigned char (*filter)(
+        void*,
+        int level,
+        const char* key, size_t key_length,
+        const char* existing_value, size_t value_length,
+        char** new_value, size_t *new_value_length,
+        unsigned char* value_changed),
+    const char* (*name)(void*));
+extern void rocksdb_compactionfilter_destroy(rocksdb_compactionfilter_t*);
+
+/* Compaction Filter Context */
+
+extern unsigned char rocksdb_compactionfiltercontext_is_full_compaction(
+    rocksdb_compactionfiltercontext_t* context);
+
+extern unsigned char rocksdb_compactionfiltercontext_is_manual_compaction(
+    rocksdb_compactionfiltercontext_t* context);
+
+/* Compaction Filter Factory */
+
+extern rocksdb_compactionfilterfactory_t*
+    rocksdb_compactionfilterfactory_create(
+        void* state, void (*destructor)(void*),
+        rocksdb_compactionfilter_t* (*create_compaction_filter)(
+            void*, rocksdb_compactionfiltercontext_t* context),
+        const char* (*name)(void*));
+extern void rocksdb_compactionfilterfactory_destroy(
+    rocksdb_compactionfilterfactory_t*);
+
+/* Compaction Filter V2 */
+
+extern rocksdb_compactionfilterv2_t* rocksdb_compactionfilterv2_create(
+    void* state,
+    void (*destructor)(void*),
+    // num_keys specifies the number of array entries in every *list parameter.
+    // New values added to the new_values_list should be malloc'd and will be
+    // freed by the caller. Specify true in the to_delete_list to remove an
+    // entry during compaction; false to keep it.
+    void (*filter)(
+        void*, int level, size_t num_keys,
+        const char* const* keys_list, const size_t* keys_list_sizes,
+        const char* const* existing_values_list, const size_t* existing_values_list_sizes,
+        char** new_values_list, size_t* new_values_list_sizes,
+        unsigned char* to_delete_list),
+    const char* (*name)(void*));
+extern void rocksdb_compactionfilterv2_destroy(rocksdb_compactionfilterv2_t*);
+
+/* Compaction Filter Factory V2 */
+
+extern rocksdb_compactionfilterfactoryv2_t* rocksdb_compactionfilterfactoryv2_create(
+    void* state,
+    rocksdb_slicetransform_t* prefix_extractor,
+    void (*destructor)(void*),
+    rocksdb_compactionfilterv2_t* (*create_compaction_filter_v2)(
+        void*, const rocksdb_compactionfiltercontext_t* context),
+    const char* (*name)(void*));
+extern void rocksdb_compactionfilterfactoryv2_destroy(rocksdb_compactionfilterfactoryv2_t*);
+
 /* Comparator */

 extern rocksdb_comparator_t* rocksdb_comparator_create(
@@ -472,6 +696,10 @@ extern void rocksdb_readoptions_set_fill_cache(
 extern void rocksdb_readoptions_set_snapshot(
    rocksdb_readoptions_t*,
    const rocksdb_snapshot_t*);
+extern void rocksdb_readoptions_set_iterate_upper_bound(
+    rocksdb_readoptions_t*,
+    const char* key,
+    size_t keylen);
 extern void rocksdb_readoptions_set_read_tier(
    rocksdb_readoptions_t*, int);
 extern void rocksdb_readoptions_set_tailing(
@@ -546,6 +774,12 @@ extern void rocksdb_universal_compaction_options_set_stop_style(
 extern void rocksdb_universal_compaction_options_destroy(
  rocksdb_universal_compaction_options_t*);

+extern rocksdb_fifo_compaction_options_t* rocksdb_fifo_compaction_options_create();
+extern void rocksdb_fifo_compaction_options_set_max_table_files_size(
+    rocksdb_fifo_compaction_options_t* fifo_opts, uint64_t size);
+extern void rocksdb_fifo_compaction_options_destroy(
+    rocksdb_fifo_compaction_options_t* fifo_opts);
+
 extern int rocksdb_livefiles_count(
  const rocksdb_livefiles_t*);
 extern const char* rocksdb_livefiles_name(
@@ -127,9 +127,6 @@ class Cache {
  void LRU_Append(Handle* e);
  void Unref(Handle* e);

-  struct Rep;
-  Rep* rep_;
-
  // No copying allowed
  Cache(const Cache&);
  void operator=(const Cache&);
@@ -9,6 +9,7 @@
 #ifndef STORAGE_ROCKSDB_INCLUDE_COMPACTION_FILTER_H_
 #define STORAGE_ROCKSDB_INCLUDE_COMPACTION_FILTER_H_

+#include <memory>
 #include <string>
 #include <vector>

@@ -86,7 +87,7 @@ class CompactionFilterV2 {
  //
  // Each entry in the return vector indicates if the corresponding kv should
  // be preserved in the output of this compaction run. The application can
-  // inspect the exisitng values of the keys and make decision based on it.
+  // inspect the existing values of the keys and make decision based on it.
  //
  // When a value is to be preserved, the application has the option
  // to modify the entry in existing_values and pass it back through an entry
@@ -108,7 +109,7 @@ class CompactionFilterV2 {
 };

 // Each compaction will create a new CompactionFilter allowing the
-// application to know about different campactions
+// application to know about different compactions
 class CompactionFilterFactory {
 public:
  virtual ~CompactionFilterFactory() { }
@@ -120,7 +121,7 @@ class CompactionFilterFactory {
  virtual const char* Name() const = 0;
 };

-// Default implementaion of CompactionFilterFactory which does not
+// Default implementation of CompactionFilterFactory which does not
 // return any filter
 class DefaultCompactionFilterFactory : public CompactionFilterFactory {
 public:
@@ -175,7 +176,7 @@ class CompactionFilterFactoryV2 {
  const SliceTransform* prefix_extractor_;
 };

-// Default implementaion of CompactionFilterFactoryV2 which does not
+// Default implementation of CompactionFilterFactoryV2 which does not
 // return any filter
 class DefaultCompactionFilterFactoryV2 : public CompactionFilterFactoryV2 {
 public:
@@ -62,6 +62,10 @@ class Comparator {
 // must not be deleted.
 extern const Comparator* BytewiseComparator();

+// Return a builtin comparator that uses reverse lexicographic byte-wise
+// ordering. 
+extern const Comparator* ReverseBytewiseComparator();
+
 }  // namespace rocksdb

 #endif  // STORAGE_ROCKSDB_INCLUDE_COMPARATOR_H_
@@ -55,6 +55,7 @@ class Env;
 // Metadata associated with each SST file.
 struct LiveFileMetaData {
  std::string column_family_name;  // Name of the column family
+  std::string db_path;
  std::string name;                // Name of the file
  int level;               // Level at which this file resides.
  size_t size;             // File size in bytes.
@@ -122,7 +123,7 @@ class DB {

  // Open DB with column families.
  // db_options specify database specific options
-  // column_families is the vector of all column families in the databse,
+  // column_families is the vector of all column families in the database,
  // containing column family name and options. You need to open ALL column
  // families in the database. To get the list of column families, you can use
  // ListColumnFamilies(). Also, you can open only a subset of column families
@@ -159,6 +160,7 @@ class DB {
  virtual Status DropColumnFamily(ColumnFamilyHandle* column_family);

  // Set the database entry for "key" to "value".
+  // If "key" already exists, it will be overwritten.
  // Returns OK on success, and a non-OK status on error.
  // Note: consider setting options.sync = true.
  virtual Status Put(const WriteOptions& options,
@@ -305,6 +307,14 @@ class DB {
    return GetProperty(DefaultColumnFamily(), property, value);
  }

+  // Similar to GetProperty(), but only works for a subset of properties whose
+  // return value is an integer. Return the value by integer.
+  virtual bool GetIntProperty(ColumnFamilyHandle* column_family,
+                              const Slice& property, uint64_t* value) = 0;
+  virtual bool GetIntProperty(const Slice& property, uint64_t* value) {
+    return GetIntProperty(DefaultColumnFamily(), property, value);
+  }
+
  // For each i in [0,n-1], store in "sizes[i]", the approximate
  // file system space used by keys in "[range[i].start .. range[i].limit)".
  //
@@ -337,15 +347,25 @@ class DB {
  // 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).
+  // Compaction outputs should be placed in options.db_paths[target_path_id].
+  // Behavior is undefined if target_path_id is out of range.
  virtual Status CompactRange(ColumnFamilyHandle* column_family,
                              const Slice* begin, const Slice* end,
-                              bool reduce_level = false,
-                              int target_level = -1) = 0;
+                              bool reduce_level = false, int target_level = -1,
+                              uint32_t target_path_id = 0) = 0;
  virtual Status CompactRange(const Slice* begin, const Slice* end,
-                              bool reduce_level = false,
-                              int target_level = -1) {
+                              bool reduce_level = false, int target_level = -1,
+                              uint32_t target_path_id = 0) {
    return CompactRange(DefaultColumnFamily(), begin, end, reduce_level,
-                        target_level);
+                        target_level, target_path_id);
+  }
+  virtual bool SetOptions(ColumnFamilyHandle* column_family,
+      const std::unordered_map<std::string, std::string>& new_options) {
+    return true;
+  }
+  virtual bool SetOptions(
+      const std::unordered_map<std::string, std::string>& new_options) {
+    return SetOptions(DefaultColumnFamily(), new_options);
  }

  // Number of levels used for this DB.
@@ -35,6 +35,7 @@ class WritableFile;
 class RandomRWFile;
 class Directory;
 struct DBOptions;
+class RateLimiter;

 using std::unique_ptr;
 using std::shared_ptr;
@@ -74,6 +75,9 @@ struct EnvOptions {
  // write. By default, we set it to true for MANIFEST writes and false for
  // WAL writes
  bool fallocate_with_keep_size = true;
+
+  // If not nullptr, write rate limiting is enabled for flush and compaction
+  RateLimiter* rate_limiter = nullptr;
 };

 class Env {
@@ -194,8 +198,16 @@ class Env {
  // REQUIRES: lock has not already been unlocked.
  virtual Status UnlockFile(FileLock* lock) = 0;

+  // Priority for scheduling job in thread pool
  enum Priority { LOW, HIGH, TOTAL };

+  // Priority for requesting bytes in rate limiter scheduler
+  enum IOPriority {
+    IO_LOW = 0,
+    IO_HIGH = 1,
+    IO_TOTAL = 2
+  };
+
  // 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.
@@ -260,6 +272,9 @@ class Env {
  // default number: 1
  virtual void SetBackgroundThreads(int number, Priority pri = LOW) = 0;

+  // Lower IO priority for threads from the specified pool.
+  virtual void LowerThreadPoolIOPriority(Priority pool = LOW) {}
+
  // Converts seconds-since-Jan-01-1970 to a printable string
  virtual std::string TimeToString(uint64_t time) = 0;

@@ -371,7 +386,10 @@ class RandomAccessFile {
 // at a time to the file.
 class WritableFile {
 public:
-  WritableFile() : last_preallocated_block_(0), preallocation_block_size_ (0) {
+  WritableFile()
+    : last_preallocated_block_(0),
+      preallocation_block_size_(0),
+      io_priority_(Env::IO_TOTAL) {
  }
  virtual ~WritableFile();

@@ -390,6 +408,14 @@ class WritableFile {
    return Sync();
  }

+  /*
+   * Change the priority in rate limiter if rate limiting is enabled.
+   * If rate limiting is not enabled, this call has no effect.
+   */
+  virtual void SetIOPriority(Env::IOPriority pri) {
+    io_priority_ = pri;
+  }
+
  /*
   * Get the size of valid data in the file.
   */
@@ -474,6 +500,9 @@ class WritableFile {
  // No copying allowed
  WritableFile(const WritableFile&);
  void operator=(const WritableFile&);
+
+ protected:
+  Env::IOPriority io_priority_;
 };

 // A file abstraction for random reading and writing.
@@ -753,6 +782,9 @@ class EnvWrapper : public Env {
  void SetBackgroundThreads(int num, Priority pri) {
    return target_->SetBackgroundThreads(num, pri);
  }
+  void LowerThreadPoolIOPriority(Priority pool = LOW) override {
+    target_->LowerThreadPoolIOPriority(pool);
+  }
  std::string TimeToString(uint64_t time) {
    return target_->TimeToString(time);
  }
@@ -21,11 +21,52 @@
 #define STORAGE_ROCKSDB_INCLUDE_FILTER_POLICY_H_

 #include <string>
+#include <memory>

 namespace rocksdb {

 class Slice;

+// A class that takes a bunch of keys, then generates filter
+class FilterBitsBuilder {
+ public:
+  virtual ~FilterBitsBuilder() {}
+
+  // Add Key to filter, you could use any way to store the key.
+  // Such as: storing hashes or original keys
+  // Keys are in sorted order and duplicated keys are possible.
+  virtual void AddKey(const Slice& key) = 0;
+
+  // Generate the filter using the keys that are added
+  // The return value of this function would be the filter bits,
+  // The ownership of actual data is set to buf
+  virtual Slice Finish(std::unique_ptr<const char[]>* buf) = 0;
+};
+
+// A class that checks if a key can be in filter
+// It should be initialized by Slice generated by BitsBuilder
+class FilterBitsReader {
+ public:
+  virtual ~FilterBitsReader() {}
+
+  // Check if the entry match the bits in filter
+  virtual bool MayMatch(const Slice& entry) = 0;
+};
+
+// We add a new format of filter block called full filter block
+// This new interface gives you more space of customization
+//
+// For the full filter block, you can plug in your version by implement
+// the FilterBitsBuilder and FilterBitsReader
+//
+// There are two sets of interface in FilterPolicy
+// Set 1: CreateFilter, KeyMayMatch: used for blockbased filter
+// Set 2: GetFilterBitsBuilder, GetFilterBitsReader, they are used for
+// full filter.
+// Set 1 MUST be implemented correctly, Set 2 is optional
+// RocksDB would first try using functions in Set 2. if they return nullptr,
+// it would use Set 1 instead.
+// You can choose filter type in NewBloomFilterPolicy
 class FilterPolicy {
 public:
  virtual ~FilterPolicy();
@@ -51,11 +92,28 @@ class FilterPolicy {
  // This method may return true or false if the key was not on the
  // list, but it should aim to return false with a high probability.
  virtual bool KeyMayMatch(const Slice& key, const Slice& filter) const = 0;
+
+  // Get the FilterBitsBuilder, which is ONLY used for full filter block
+  // It contains interface to take individual key, then generate filter
+  virtual FilterBitsBuilder* GetFilterBitsBuilder() const {
+    return nullptr;
+  }
+
+  // Get the FilterBitsReader, which is ONLY used for full filter block
+  // It contains interface to tell if key can be in filter
+  // The input slice should NOT be deleted by FilterPolicy
+  virtual FilterBitsReader* GetFilterBitsReader(const Slice& contents) const {
+    return nullptr;
+  }
 };

 // Return a new filter policy that uses a bloom filter with approximately
-// the specified number of bits per key.  A good value for bits_per_key
+// the specified number of bits per key.
+//
+// bits_per_key: bits per key in bloom filter. A good value for bits_per_key
 // is 10, which yields a filter with ~ 1% false positive rate.
+// use_block_based_builder: use block based filter rather than full fiter.
+// If you want to builder full filter, it needs to be set to false.
 //
 // Callers must delete the result after any database that is using the
 // result has been closed.
@@ -67,8 +125,8 @@ class FilterPolicy {
 // ignores trailing spaces, it would be incorrect to use a
 // FilterPolicy (like NewBloomFilterPolicy) that does not ignore
 // trailing spaces in keys.
-extern const FilterPolicy* NewBloomFilterPolicy(int bits_per_key);
-
+extern const FilterPolicy* NewBloomFilterPolicy(int bits_per_key,
+    bool use_block_based_builder = true);
 }

 #endif  // STORAGE_ROCKSDB_INCLUDE_FILTER_POLICY_H_