Release LevelDB 1.17

- Cleanup: delete unused IntSetToString

  It was added in http://cr/19491949 (and was referenced at the time).
  The last reference was removed in http://cr/19507363.

  This fixes warning/error with pre-release crosstoolv18:
    'std::string leveldb::{anonymous}::IntSetToString(const std::set<long unsigned int>&)' defined but not used [-Werror=unused-function]

- Added arm64 and and armv7s to IOS build as suggested on leveldb mailing list.

- Changed local variable type from int to size_t

  This eliminates compiler warning/error and resolves
  https://code.google.com/p/leveldb/issues/detail?id=140

.gitignore

@@ -1,5 +1,9 @@
 build_config.mk
 *.a
 *.o
+*.dylib*
+*.so
+*.so.*
 *_test
 db_bench
+leveldbutil

AUTHORS

@@ -6,3 +6,7 @@ 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>

Makefile

@@ -2,11 +2,6 @@
 # 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.
 
-# Inherit some settings from environment variables, if available
-CXX ?= g++
-CC  ?= gcc
-INSTALL_PATH ?= $(CURDIR)
-
 #-----------------------------------------------
 # Uncomment exactly one of the lines labelled (A), (B), and (C) below
 # to switch between compilation modes.
@@ -17,14 +12,16 @@ OPT ?= -O2 -DNDEBUG       # (A) Production use (optimized mode)
 #-----------------------------------------------
 
 # detect what platform we're building on
-$(shell sh ./build_detect_platform)
-# this file is generated by build_detect_platform to set build flags and sources
+$(shell CC="$(CC)" CXX="$(CXX)" TARGET_OS="$(TARGET_OS)" \
+    ./build_detect_platform build_config.mk ./)
+# this file is generated by the previous line to set build flags and sources
 include build_config.mk
 
 CFLAGS += -I. -I./include $(PLATFORM_CCFLAGS) $(OPT)
 CXXFLAGS += -I. -I./include $(PLATFORM_CXXFLAGS) $(OPT)
 
 LDFLAGS += $(PLATFORM_LDFLAGS)
+LIBS += $(PLATFORM_LIBS)
 
 LIBOBJECTS = $(SOURCES:.cc=.o)
 MEMENVOBJECTS = $(MEMENV_SOURCES:.cc=.o)
@@ -34,6 +31,8 @@ TESTHARNESS = ./util/testharness.o $(TESTUTIL)
 
 TESTS = \
 	arena_test \
+	autocompact_test \
+	bloom_test \
 	c_test \
 	cache_test \
 	coding_test \
@@ -43,6 +42,9 @@ TESTS = \
 	dbformat_test \
 	env_test \
 	filename_test \
+	filter_block_test \
+	issue178_test \
+	issue200_test \
 	log_test \
 	memenv_test \
 	skiplist_test \
@@ -51,7 +53,7 @@ TESTS = \
 	version_set_test \
 	write_batch_test
 
-PROGRAMS = db_bench $(TESTS)
+PROGRAMS = db_bench leveldbutil $(TESTS)
 BENCHMARKS = db_bench_sqlite3 db_bench_tree_db
 
 LIBRARY = libleveldb.a
@@ -61,21 +63,31 @@ default: all
 
 # Should we build shared libraries?
 ifneq ($(PLATFORM_SHARED_EXT),)
+
+ifneq ($(PLATFORM_SHARED_VERSIONED),true)
+SHARED1 = libleveldb.$(PLATFORM_SHARED_EXT)
+SHARED2 = $(SHARED1)
+SHARED3 = $(SHARED1)
+SHARED = $(SHARED1)
+else
 # Update db.h if you change these.
 SHARED_MAJOR = 1
-SHARED_MINOR = 3
+SHARED_MINOR = 17
 SHARED1 = libleveldb.$(PLATFORM_SHARED_EXT)
 SHARED2 = $(SHARED1).$(SHARED_MAJOR)
 SHARED3 = $(SHARED1).$(SHARED_MAJOR).$(SHARED_MINOR)
 SHARED = $(SHARED1) $(SHARED2) $(SHARED3)
-$(SHARED3):
-	$(CXX) $(LDFLAGS) $(PLATFORM_SHARED_LDFLAGS)$(INSTALL_PATH)/$(SHARED2) $(CXXFLAGS) $(PLATFORM_SHARED_CFLAGS) $(SOURCES) -o $(SHARED3)
-$(SHARED2): $(SHARED3)
-	ln -fs $(SHARED3) $(SHARED2)
 $(SHARED1): $(SHARED3)
 	ln -fs $(SHARED3) $(SHARED1)
+$(SHARED2): $(SHARED3)
+	ln -fs $(SHARED3) $(SHARED2)
 endif
 
+$(SHARED3):
+	$(CXX) $(LDFLAGS) $(PLATFORM_SHARED_LDFLAGS)$(SHARED2) $(CXXFLAGS) $(PLATFORM_SHARED_CFLAGS) $(SOURCES) -o $(SHARED3) $(LIBS)
+
+endif  # PLATFORM_SHARED_EXT
+
 all: $(SHARED) $(LIBRARY)
 
 check: all $(PROGRAMS) $(TESTS)
@@ -90,88 +102,108 @@ $(LIBRARY): $(LIBOBJECTS)
 	$(AR) -rs $@ $(LIBOBJECTS)
 
 db_bench: db/db_bench.o $(LIBOBJECTS) $(TESTUTIL)
-	$(CXX) db/db_bench.o $(LIBOBJECTS) $(TESTUTIL) -o $@  $(LDFLAGS)
+	$(CXX) $(LDFLAGS) db/db_bench.o $(LIBOBJECTS) $(TESTUTIL) -o $@ $(LIBS)
 
 db_bench_sqlite3: doc/bench/db_bench_sqlite3.o $(LIBOBJECTS) $(TESTUTIL)
-	$(CXX) doc/bench/db_bench_sqlite3.o $(LIBOBJECTS) $(TESTUTIL) -o $@ $(LDFLAGS) -lsqlite3
+	$(CXX) $(LDFLAGS) doc/bench/db_bench_sqlite3.o $(LIBOBJECTS) $(TESTUTIL) -o $@ -lsqlite3 $(LIBS)
 
 db_bench_tree_db: doc/bench/db_bench_tree_db.o $(LIBOBJECTS) $(TESTUTIL)
-	$(CXX) doc/bench/db_bench_tree_db.o $(LIBOBJECTS) $(TESTUTIL) -o $@ $(LDFLAGS) -lkyotocabinet
+	$(CXX) $(LDFLAGS) doc/bench/db_bench_tree_db.o $(LIBOBJECTS) $(TESTUTIL) -o $@ -lkyotocabinet $(LIBS)
+
+leveldbutil: db/leveldb_main.o $(LIBOBJECTS)
+	$(CXX) $(LDFLAGS) db/leveldb_main.o $(LIBOBJECTS) -o $@ $(LIBS)
 
 arena_test: util/arena_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) util/arena_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LDFLAGS)
+	$(CXX) $(LDFLAGS) util/arena_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
+
+autocompact_test: db/autocompact_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) $(LDFLAGS) db/autocompact_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
+
+bloom_test: util/bloom_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) $(LDFLAGS) util/bloom_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
 
 c_test: db/c_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/c_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LDFLAGS)
+	$(CXX) $(LDFLAGS) db/c_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
 
 cache_test: util/cache_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) util/cache_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LDFLAGS)
+	$(CXX) $(LDFLAGS) util/cache_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
 
 coding_test: util/coding_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) util/coding_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LDFLAGS)
+	$(CXX) $(LDFLAGS) util/coding_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
 
 corruption_test: db/corruption_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/corruption_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LDFLAGS)
+	$(CXX) $(LDFLAGS) db/corruption_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
 
 crc32c_test: util/crc32c_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) util/crc32c_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LDFLAGS)
+	$(CXX) $(LDFLAGS) util/crc32c_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
 
 db_test: db/db_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/db_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LDFLAGS)
+	$(CXX) $(LDFLAGS) db/db_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
 
 dbformat_test: db/dbformat_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/dbformat_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LDFLAGS)
+	$(CXX) $(LDFLAGS) db/dbformat_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
 
 env_test: util/env_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) util/env_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LDFLAGS)
+	$(CXX) $(LDFLAGS) util/env_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
 
 filename_test: db/filename_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/filename_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LDFLAGS)
+	$(CXX) $(LDFLAGS) db/filename_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
+
+filter_block_test: table/filter_block_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) $(LDFLAGS) table/filter_block_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
+
+issue178_test: issues/issue178_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) $(LDFLAGS) issues/issue178_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
+
+issue200_test: issues/issue200_test.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(CXX) $(LDFLAGS) issues/issue200_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
 
 log_test: db/log_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/log_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LDFLAGS)
+	$(CXX) $(LDFLAGS) db/log_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
 
 table_test: table/table_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) table/table_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LDFLAGS)
+	$(CXX) $(LDFLAGS) table/table_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
 
 skiplist_test: db/skiplist_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/skiplist_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LDFLAGS)
+	$(CXX) $(LDFLAGS) db/skiplist_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
 
 version_edit_test: db/version_edit_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/version_edit_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LDFLAGS)
+	$(CXX) $(LDFLAGS) db/version_edit_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
 
 version_set_test: db/version_set_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/version_set_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LDFLAGS)
+	$(CXX) $(LDFLAGS) db/version_set_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
 
 write_batch_test: db/write_batch_test.o $(LIBOBJECTS) $(TESTHARNESS)
-	$(CXX) db/write_batch_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LDFLAGS)
+	$(CXX) $(LDFLAGS) db/write_batch_test.o $(LIBOBJECTS) $(TESTHARNESS) -o $@ $(LIBS)
 
 $(MEMENVLIBRARY) : $(MEMENVOBJECTS)
 	rm -f $@
 	$(AR) -rs $@ $(MEMENVOBJECTS)
 
 memenv_test : helpers/memenv/memenv_test.o $(MEMENVLIBRARY) $(LIBRARY) $(TESTHARNESS)
-	$(CXX) helpers/memenv/memenv_test.o $(MEMENVLIBRARY) $(LIBRARY) $(TESTHARNESS) -o $@ $(LDFLAGS)
+	$(CXX) $(LDFLAGS) helpers/memenv/memenv_test.o $(MEMENVLIBRARY) $(LIBRARY) $(TESTHARNESS) -o $@ $(LIBS)
 
 ifeq ($(PLATFORM), IOS)
 # For iOS, create universal object files to be used on both the simulator and
 # a device.
-SIMULATORROOT=/Developer/Platforms/iPhoneSimulator.platform/Developer
-DEVICEROOT=/Developer/Platforms/iPhoneOS.platform/Developer
-IOSVERSION=$(shell defaults read /Developer/Platforms/iPhoneOS.platform/version CFBundleShortVersionString)
+PLATFORMSROOT=/Applications/Xcode.app/Contents/Developer/Platforms
+SIMULATORROOT=$(PLATFORMSROOT)/iPhoneSimulator.platform/Developer
+DEVICEROOT=$(PLATFORMSROOT)/iPhoneOS.platform/Developer
+IOSVERSION=$(shell defaults read $(PLATFORMSROOT)/iPhoneOS.platform/version CFBundleShortVersionString)
+IOSARCH=-arch armv6 -arch armv7 -arch armv7s -arch arm64
 
 .cc.o:
 	mkdir -p ios-x86/$(dir $@)
-	$(SIMULATORROOT)/usr/bin/$(CXX) $(CXXFLAGS) -isysroot $(SIMULATORROOT)/SDKs/iPhoneSimulator$(IOSVERSION).sdk -arch i686 -c $< -o ios-x86/$@
+	$(CXX) $(CXXFLAGS) -isysroot $(SIMULATORROOT)/SDKs/iPhoneSimulator$(IOSVERSION).sdk -arch i686 -arch x86_64 -c $< -o ios-x86/$@
 	mkdir -p ios-arm/$(dir $@)
-	$(DEVICEROOT)/usr/bin/$(CXX) $(CXXFLAGS) -isysroot $(DEVICEROOT)/SDKs/iPhoneOS$(IOSVERSION).sdk -arch armv6 -arch armv7 -c $< -o ios-arm/$@
+	xcrun -sdk iphoneos $(CXX) $(CXXFLAGS) -isysroot $(DEVICEROOT)/SDKs/iPhoneOS$(IOSVERSION).sdk $(IOSARCH) -c $< -o ios-arm/$@
 	lipo ios-x86/$@ ios-arm/$@ -create -output $@
 
 .c.o:
 	mkdir -p ios-x86/$(dir $@)
-	$(SIMULATORROOT)/usr/bin/$(CC) $(CFLAGS) -isysroot $(SIMULATORROOT)/SDKs/iPhoneSimulator$(IOSVERSION).sdk -arch i686 -c $< -o ios-x86/$@
+	$(CC) $(CFLAGS) -isysroot $(SIMULATORROOT)/SDKs/iPhoneSimulator$(IOSVERSION).sdk -arch i686 -arch x86_64 -c $< -o ios-x86/$@
 	mkdir -p ios-arm/$(dir $@)
-	$(DEVICEROOT)/usr/bin/$(CC) $(CFLAGS) -isysroot $(DEVICEROOT)/SDKs/iPhoneOS$(IOSVERSION).sdk -arch armv6 -arch armv7 -c $< -o ios-arm/$@
+	xcrun -sdk iphoneos $(CC) $(CFLAGS) -isysroot $(DEVICEROOT)/SDKs/iPhoneOS$(IOSVERSION).sdk $(IOSARCH) -c $< -o ios-arm/$@
 	lipo ios-x86/$@ ios-arm/$@ -create -output $@
 
 else

TODO

@@ -7,6 +7,7 @@ db
   within [start_key..end_key]?  For Chrome, deletion of obsolete
   object stores, etc. can be done in the background anyway, so
   probably not that important.
+- There have been requests for MultiGet.
 
 After a range is completely deleted, what gets rid of the
 corresponding files if we do no future changes to that range.  Make

build_detect_platform

@@ -1,94 +1,148 @@
 #!/bin/sh
 #
-# Detects OS we're compiling on and generates build_config.mk,
-# which in turn gets read while processing Makefile.
+# Detects OS we're compiling on and outputs a file specified by the first
+# argument, which in turn gets read while processing Makefile.
 #
-# build_config.mk will set the following variables:
+# The output will set the following variables:
+#   CC                          C Compiler path
+#   CXX                         C++ Compiler path
 #   PLATFORM_LDFLAGS            Linker flags
+#   PLATFORM_LIBS               Libraries flags
 #   PLATFORM_SHARED_EXT         Extension for shared libraries
 #   PLATFORM_SHARED_LDFLAGS     Flags for building shared library
+#                               This flag is embedded just before the name
+#                               of the shared library without intervening spaces
 #   PLATFORM_SHARED_CFLAGS      Flags for compiling objects for shared library
 #   PLATFORM_CCFLAGS            C compiler flags
 #   PLATFORM_CXXFLAGS           C++ compiler flags.  Will contain:
-#       -DLEVELDB_PLATFORM_POSIX if cstdatomic is present
-#       -DLEVELDB_PLATFORM_NOATOMIC if it is not
+#   PLATFORM_SHARED_VERSIONED   Set to 'true' if platform supports versioned
+#                               shared libraries, empty otherwise.
+#
+# The PLATFORM_CCFLAGS and PLATFORM_CXXFLAGS might include the following:
+#
+#       -DLEVELDB_CSTDATOMIC_PRESENT if <cstdatomic> is present
+#       -DLEVELDB_PLATFORM_POSIX     for Posix-based platforms
+#       -DSNAPPY                     if the Snappy library is present
+#
 
-SCRIPT_DIR=`dirname $0`
+OUTPUT=$1
+PREFIX=$2
+if test -z "$OUTPUT" || test -z "$PREFIX"; then
+  echo "usage: $0 <output-filename> <directory_prefix>" >&2
+  exit 1
+fi
 
-# Delete existing build_config.mk
-rm -f build_config.mk
-touch build_config.mk
+# Delete existing output, if it exists
+rm -f $OUTPUT
+touch $OUTPUT
+
+if test -z "$CC"; then
+    CC=cc
+fi
 
 if test -z "$CXX"; then
     CXX=g++
 fi
 
+if test -z "$TMPDIR"; then
+    TMPDIR=/tmp
+fi
+
 # Detect OS
 if test -z "$TARGET_OS"; then
     TARGET_OS=`uname -s`
 fi
 
 COMMON_FLAGS=
+CROSS_COMPILE=
 PLATFORM_CCFLAGS=
 PLATFORM_CXXFLAGS=
 PLATFORM_LDFLAGS=
+PLATFORM_LIBS=
 PLATFORM_SHARED_EXT="so"
 PLATFORM_SHARED_LDFLAGS="-shared -Wl,-soname -Wl,"
 PLATFORM_SHARED_CFLAGS="-fPIC"
+PLATFORM_SHARED_VERSIONED=true
+
+MEMCMP_FLAG=
+if [ "$CXX" = "g++" ]; then
+    # Use libc's memcmp instead of GCC's memcmp.  This results in ~40%
+    # performance improvement on readrandom under gcc 4.4.3 on Linux/x86.
+    MEMCMP_FLAG="-fno-builtin-memcmp"
+fi
 
-# On GCC, we pick libc's memcmp over GCC's memcmp via -fno-builtin-memcmp
 case "$TARGET_OS" in
     Darwin)
         PLATFORM=OS_MACOSX
-        COMMON_FLAGS="-fno-builtin-memcmp -DOS_MACOSX"
+        COMMON_FLAGS="$MEMCMP_FLAG -DOS_MACOSX"
         PLATFORM_SHARED_EXT=dylib
-        PLATFORM_SHARED_LDFLAGS="-dynamiclib -install_name "
+        [ -z "$INSTALL_PATH" ] && INSTALL_PATH=`pwd`
+        PLATFORM_SHARED_LDFLAGS="-dynamiclib -install_name $INSTALL_PATH/"
         PORT_FILE=port/port_posix.cc
         ;;
     Linux)
         PLATFORM=OS_LINUX
-        COMMON_FLAGS="-fno-builtin-memcmp -pthread -DOS_LINUX"
+        COMMON_FLAGS="$MEMCMP_FLAG -pthread -DOS_LINUX"
         PLATFORM_LDFLAGS="-pthread"
         PORT_FILE=port/port_posix.cc
         ;;
     SunOS)
         PLATFORM=OS_SOLARIS
-        COMMON_FLAGS="-fno-builtin-memcmp -D_REENTRANT -DOS_SOLARIS"
-        PLATFORM_LDFLAGS="-lpthread -lrt"
+        COMMON_FLAGS="$MEMCMP_FLAG -D_REENTRANT -DOS_SOLARIS"
+        PLATFORM_LIBS="-lpthread -lrt"
         PORT_FILE=port/port_posix.cc
         ;;
     FreeBSD)
         PLATFORM=OS_FREEBSD
-        COMMON_FLAGS="-fno-builtin-memcmp -D_REENTRANT -DOS_FREEBSD"
-        PLATFORM_LDFLAGS="-lpthread"
+        COMMON_FLAGS="$MEMCMP_FLAG -D_REENTRANT -DOS_FREEBSD"
+        PLATFORM_LIBS="-lpthread"
         PORT_FILE=port/port_posix.cc
         ;;
     NetBSD)
         PLATFORM=OS_NETBSD
-        COMMON_FLAGS="-fno-builtin-memcmp -D_REENTRANT -DOS_NETBSD"
-        PLATFORM_LDFLAGS="-lpthread -lgcc_s"
+        COMMON_FLAGS="$MEMCMP_FLAG -D_REENTRANT -DOS_NETBSD"
+        PLATFORM_LIBS="-lpthread -lgcc_s"
         PORT_FILE=port/port_posix.cc
         ;;
     OpenBSD)
         PLATFORM=OS_OPENBSD
-        COMMON_FLAGS="-fno-builtin-memcmp -D_REENTRANT -DOS_OPENBSD"
+        COMMON_FLAGS="$MEMCMP_FLAG -D_REENTRANT -DOS_OPENBSD"
         PLATFORM_LDFLAGS="-pthread"
         PORT_FILE=port/port_posix.cc
         ;;
     DragonFly)
         PLATFORM=OS_DRAGONFLYBSD
-        COMMON_FLAGS="-fno-builtin-memcmp -D_REENTRANT -DOS_DRAGONFLYBSD"
-        PLATFORM_LDFLAGS="-lpthread"
+        COMMON_FLAGS="$MEMCMP_FLAG -D_REENTRANT -DOS_DRAGONFLYBSD"
+        PLATFORM_LIBS="-lpthread"
         PORT_FILE=port/port_posix.cc
         ;;
     OS_ANDROID_CROSSCOMPILE)
-        PLATFORM="$TARGET_OS"
-        COMMON_FLAGS=""
-        PLATFORM_LDFLAGS=""
-        PORT_FILE=port/port_android.cc
+        PLATFORM=OS_ANDROID
+        COMMON_FLAGS="$MEMCMP_FLAG -D_REENTRANT -DOS_ANDROID -DLEVELDB_PLATFORM_POSIX"
+        PLATFORM_LDFLAGS=""  # All pthread features are in the Android C library
+        PORT_FILE=port/port_posix.cc
+        CROSS_COMPILE=true
+        ;;
+    HP-UX)
+        PLATFORM=OS_HPUX
+        COMMON_FLAGS="$MEMCMP_FLAG -D_REENTRANT -DOS_HPUX"
+        PLATFORM_LDFLAGS="-pthread"
+        PORT_FILE=port/port_posix.cc
+        # man ld: +h internal_name
+        PLATFORM_SHARED_LDFLAGS="-shared -Wl,+h -Wl,"
+        ;;
+    IOS)
+        PLATFORM=IOS
+        COMMON_FLAGS="$MEMCMP_FLAG -DOS_MACOSX"
+        [ -z "$INSTALL_PATH" ] && INSTALL_PATH=`pwd`
+        PORT_FILE=port/port_posix.cc
+        PLATFORM_SHARED_EXT=
+        PLATFORM_SHARED_LDFLAGS=
+        PLATFORM_SHARED_CFLAGS=
+        PLATFORM_SHARED_VERSIONED=
         ;;
     *)
-        echo "Unknown platform!"
+        echo "Unknown platform!" >&2
         exit 1
 esac
 
@@ -96,24 +150,29 @@ esac
 # 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="$SCRIPT_DIR/util $SCRIPT_DIR/db $SCRIPT_DIR/table"
+DIRS="$PREFIX/db $PREFIX/util $PREFIX/table"
+
 set -f # temporarily disable globbing so that our patterns aren't expanded
 PRUNE_TEST="-name *test*.cc -prune"
 PRUNE_BENCH="-name *_bench.cc -prune"
-PORTABLE_FILES=`find $DIRS $PRUNE_TEST -o $PRUNE_BENCH -o -name '*.cc' -print | sort | tr "\n" " "`
+PRUNE_TOOL="-name leveldb_main.cc -prune"
+PORTABLE_FILES=`find $DIRS $PRUNE_TEST -o $PRUNE_BENCH -o $PRUNE_TOOL -o -name '*.cc' -print | sort | sed "s,^$PREFIX/,," | tr "\n" " "`
+
 set +f # re-enable globbing
 
 # The sources consist of the portable files, plus the platform-specific port
 # file.
-echo "SOURCES=$PORTABLE_FILES $PORT_FILE" >> build_config.mk
-echo "MEMENV_SOURCES=helpers/memenv/memenv.cc" >> build_config.mk
+echo "SOURCES=$PORTABLE_FILES $PORT_FILE" >> $OUTPUT
+echo "MEMENV_SOURCES=helpers/memenv/memenv.cc" >> $OUTPUT
 
-if [ "$PLATFORM" = "OS_ANDROID_CROSSCOMPILE" ]; then
+if [ "$CROSS_COMPILE" = "true" ]; then
     # Cross-compiling; do not try any compilation tests.
     true
 else
+    CXXOUTPUT="${TMPDIR}/leveldb_build_detect_platform-cxx.$$"
+
     # If -std=c++0x works, use <cstdatomic>.  Otherwise use port_posix.h.
-    $CXX $CFLAGS -std=c++0x -x c++ - -o /dev/null 2>/dev/null  <<EOF
+    $CXX $CXXFLAGS -std=c++0x -x c++ - -o $CXXOUTPUT 2>/dev/null  <<EOF
       #include <cstdatomic>
       int main() {}
 EOF
@@ -126,31 +185,37 @@ EOF
 
     # Test whether Snappy library is installed
     # http://code.google.com/p/snappy/
-    $CXX $CFLAGS -x c++ - -o /dev/null 2>/dev/null  <<EOF
+    $CXX $CXXFLAGS -x c++ - -o $CXXOUTPUT 2>/dev/null  <<EOF
       #include <snappy.h>
       int main() {}
 EOF
     if [ "$?" = 0 ]; then
         COMMON_FLAGS="$COMMON_FLAGS -DSNAPPY"
-        PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -lsnappy"
+        PLATFORM_LIBS="$PLATFORM_LIBS -lsnappy"
     fi
 
     # Test whether tcmalloc is available
-    $CXX $CFLAGS -x c++ - -o /dev/null -ltcmalloc 2>/dev/null  <<EOF
+    $CXX $CXXFLAGS -x c++ - -o $CXXOUTPUT -ltcmalloc 2>/dev/null  <<EOF
       int main() {}
 EOF
     if [ "$?" = 0 ]; then
-        PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -ltcmalloc"
+        PLATFORM_LIBS="$PLATFORM_LIBS -ltcmalloc"
     fi
+
+    rm -f $CXXOUTPUT 2>/dev/null
 fi
 
 PLATFORM_CCFLAGS="$PLATFORM_CCFLAGS $COMMON_FLAGS"
 PLATFORM_CXXFLAGS="$PLATFORM_CXXFLAGS $COMMON_FLAGS"
 
-echo "PLATFORM=$PLATFORM" >> build_config.mk
-echo "PLATFORM_LDFLAGS=$PLATFORM_LDFLAGS" >> build_config.mk
-echo "PLATFORM_CCFLAGS=$PLATFORM_CCFLAGS" >> build_config.mk
-echo "PLATFORM_CXXFLAGS=$PLATFORM_CXXFLAGS" >> build_config.mk
-echo "PLATFORM_SHARED_CFLAGS=$PLATFORM_SHARED_CFLAGS" >> build_config.mk
-echo "PLATFORM_SHARED_EXT=$PLATFORM_SHARED_EXT" >> build_config.mk
-echo "PLATFORM_SHARED_LDFLAGS=$PLATFORM_SHARED_LDFLAGS" >> build_config.mk
+echo "CC=$CC" >> $OUTPUT
+echo "CXX=$CXX" >> $OUTPUT
+echo "PLATFORM=$PLATFORM" >> $OUTPUT
+echo "PLATFORM_LDFLAGS=$PLATFORM_LDFLAGS" >> $OUTPUT
+echo "PLATFORM_LIBS=$PLATFORM_LIBS" >> $OUTPUT
+echo "PLATFORM_CCFLAGS=$PLATFORM_CCFLAGS" >> $OUTPUT
+echo "PLATFORM_CXXFLAGS=$PLATFORM_CXXFLAGS" >> $OUTPUT
+echo "PLATFORM_SHARED_CFLAGS=$PLATFORM_SHARED_CFLAGS" >> $OUTPUT
+echo "PLATFORM_SHARED_EXT=$PLATFORM_SHARED_EXT" >> $OUTPUT
+echo "PLATFORM_SHARED_LDFLAGS=$PLATFORM_SHARED_LDFLAGS" >> $OUTPUT
+echo "PLATFORM_SHARED_VERSIONED=$PLATFORM_SHARED_VERSIONED" >> $OUTPUT

db/autocompact_test.cc

@@ -0,0 +1,118 @@
+// Copyright (c) 2013 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 "leveldb/db.h"
+#include "db/db_impl.h"
+#include "leveldb/cache.h"
+#include "util/testharness.h"
+#include "util/testutil.h"
+
+namespace leveldb {
+
+class AutoCompactTest {
+ public:
+  std::string dbname_;
+  Cache* tiny_cache_;
+  Options options_;
+  DB* db_;
+
+  AutoCompactTest() {
+    dbname_ = test::TmpDir() + "/autocompact_test";
+    tiny_cache_ = NewLRUCache(100);
+    options_.block_cache = tiny_cache_;
+    DestroyDB(dbname_, options_);
+    options_.create_if_missing = true;
+    options_.compression = kNoCompression;
+    ASSERT_OK(DB::Open(options_, dbname_, &db_));
+  }
+
+  ~AutoCompactTest() {
+    delete db_;
+    DestroyDB(dbname_, Options());
+    delete tiny_cache_;
+  }
+
+  std::string Key(int i) {
+    char buf[100];
+    snprintf(buf, sizeof(buf), "key%06d", i);
+    return std::string(buf);
+  }
+
+  uint64_t Size(const Slice& start, const Slice& limit) {
+    Range r(start, limit);
+    uint64_t size;
+    db_->GetApproximateSizes(&r, 1, &size);
+    return size;
+  }
+
+  void DoReads(int n);
+};
+
+static const int kValueSize = 200 * 1024;
+static const int kTotalSize = 100 * 1024 * 1024;
+static const int kCount = kTotalSize / kValueSize;
+
+// Read through the first n keys repeatedly and check that they get
+// compacted (verified by checking the size of the key space).
+void AutoCompactTest::DoReads(int n) {
+  std::string value(kValueSize, 'x');
+  DBImpl* dbi = reinterpret_cast<DBImpl*>(db_);
+
+  // Fill database
+  for (int i = 0; i < kCount; i++) {
+    ASSERT_OK(db_->Put(WriteOptions(), Key(i), value));
+  }
+  ASSERT_OK(dbi->TEST_CompactMemTable());
+
+  // Delete everything
+  for (int i = 0; i < kCount; i++) {
+    ASSERT_OK(db_->Delete(WriteOptions(), Key(i)));
+  }
+  ASSERT_OK(dbi->TEST_CompactMemTable());
+
+  // Get initial measurement of the space we will be reading.
+  const int64_t initial_size = Size(Key(0), Key(n));
+  const int64_t initial_other_size = Size(Key(n), Key(kCount));
+
+  // Read until size drops significantly.
+  std::string limit_key = Key(n);
+  for (int read = 0; true; read++) {
+    ASSERT_LT(read, 100) << "Taking too long to compact";
+    Iterator* iter = db_->NewIterator(ReadOptions());
+    for (iter->SeekToFirst();
+         iter->Valid() && iter->key().ToString() < limit_key;
+         iter->Next()) {
+      // Drop data
+    }
+    delete iter;
+    // Wait a little bit to allow any triggered compactions to complete.
+    Env::Default()->SleepForMicroseconds(1000000);
+    uint64_t size = Size(Key(0), Key(n));
+    fprintf(stderr, "iter %3d => %7.3f MB [other %7.3f MB]\n",
+            read+1, size/1048576.0, Size(Key(n), Key(kCount))/1048576.0);
+    if (size <= initial_size/10) {
+      break;
+    }
+  }
+
+  // Verify that the size of the key space not touched by the reads
+  // is pretty much unchanged.
+  const int64_t final_other_size = Size(Key(n), Key(kCount));
+  ASSERT_LE(final_other_size, initial_other_size + 1048576);
+  ASSERT_GE(final_other_size, initial_other_size/5 - 1048576);
+}
+
+TEST(AutoCompactTest, ReadAll) {
+  DoReads(kCount);
+}
+
+TEST(AutoCompactTest, ReadHalf) {
+  DoReads(kCount/2);
+}
+
+}  // namespace leveldb
+
+int main(int argc, char** argv) {
+  return leveldb::test::RunAllTests();
+}

db/c.cc

@@ -10,6 +10,7 @@
 #include "leveldb/comparator.h"
 #include "leveldb/db.h"
 #include "leveldb/env.h"
+#include "leveldb/filter_policy.h"
 #include "leveldb/iterator.h"
 #include "leveldb/options.h"
 #include "leveldb/status.h"
@@ -21,8 +22,12 @@ using leveldb::CompressionType;
 using leveldb::DB;
 using leveldb::Env;
 using leveldb::FileLock;
+using leveldb::FilterPolicy;
 using leveldb::Iterator;
+using leveldb::kMajorVersion;
+using leveldb::kMinorVersion;
 using leveldb::Logger;
+using leveldb::NewBloomFilterPolicy;
 using leveldb::NewLRUCache;
 using leveldb::Options;
 using leveldb::RandomAccessFile;
@@ -78,6 +83,47 @@ struct leveldb_comparator_t : public Comparator {
   virtual void FindShortSuccessor(std::string* key) const { }
 };
 
+struct leveldb_filterpolicy_t : public FilterPolicy {
+  void* state_;
+  void (*destructor_)(void*);
+  const char* (*name_)(void*);
+  char* (*create_)(
+      void*,
+      const char* const* key_array, const size_t* key_length_array,
+      int num_keys,
+      size_t* filter_length);
+  unsigned char (*key_match_)(
+      void*,
+      const char* key, size_t length,
+      const char* filter, size_t filter_length);
+
+  virtual ~leveldb_filterpolicy_t() {
+    (*destructor_)(state_);
+  }
+
+  virtual const char* Name() const {
+    return (*name_)(state_);
+  }
+
+  virtual void CreateFilter(const Slice* keys, int n, std::string* dst) const {
+    std::vector<const char*> key_pointers(n);
+    std::vector<size_t> key_sizes(n);
+    for (int i = 0; i < n; i++) {
+      key_pointers[i] = keys[i].data();
+      key_sizes[i] = keys[i].size();
+    }
+    size_t len;
+    char* filter = (*create_)(state_, &key_pointers[0], &key_sizes[0], n, &len);
+    dst->append(filter, len);
+    free(filter);
+  }
+
+  virtual bool KeyMayMatch(const Slice& key, const Slice& filter) const {
+    return (*key_match_)(state_, key.data(), key.size(),
+                         filter.data(), filter.size());
+  }
+};
+
 struct leveldb_env_t {
   Env* rep;
   bool is_default;
@@ -218,6 +264,17 @@ void leveldb_approximate_sizes(
   delete[] ranges;
 }
 
+void leveldb_compact_range(
+    leveldb_t* db,
+    const char* start_key, size_t start_key_len,
+    const char* limit_key, size_t limit_key_len) {
+  Slice a, b;
+  db->rep->CompactRange(
+      // Pass NULL Slice if corresponding "const char*" is NULL
+      (start_key ? (a = Slice(start_key, start_key_len), &a) : NULL),
+      (limit_key ? (b = Slice(limit_key, limit_key_len), &b) : NULL));
+}
+
 void leveldb_destroy_db(
     const leveldb_options_t* options,
     const char* name,
@@ -340,6 +397,12 @@ void leveldb_options_set_comparator(
   opt->rep.comparator = cmp;
 }
 
+void leveldb_options_set_filter_policy(
+    leveldb_options_t* opt,
+    leveldb_filterpolicy_t* policy) {
+  opt->rep.filter_policy = policy;
+}
+
 void leveldb_options_set_create_if_missing(
     leveldb_options_t* opt, unsigned char v) {
   opt->rep.create_if_missing = v;
@@ -407,6 +470,55 @@ void leveldb_comparator_destroy(leveldb_comparator_t* cmp) {
   delete cmp;
 }
 
+leveldb_filterpolicy_t* leveldb_filterpolicy_create(
+    void* state,
+    void (*destructor)(void*),
+    char* (*create_filter)(
+        void*,
+        const char* const* key_array, const size_t* key_length_array,
+        int num_keys,
+        size_t* filter_length),
+    unsigned char (*key_may_match)(
+        void*,
+        const char* key, size_t length,
+        const char* filter, size_t filter_length),
+    const char* (*name)(void*)) {
+  leveldb_filterpolicy_t* result = new leveldb_filterpolicy_t;
+  result->state_ = state;
+  result->destructor_ = destructor;
+  result->create_ = create_filter;
+  result->key_match_ = key_may_match;
+  result->name_ = name;
+  return result;
+}
+
+void leveldb_filterpolicy_destroy(leveldb_filterpolicy_t* filter) {
+  delete filter;
+}
+
+leveldb_filterpolicy_t* leveldb_filterpolicy_create_bloom(int bits_per_key) {
+  // Make a leveldb_filterpolicy_t, but override all of its methods so
+  // they delegate to a NewBloomFilterPolicy() instead of user
+  // supplied C functions.
+  struct Wrapper : public leveldb_filterpolicy_t {
+    const FilterPolicy* rep_;
+    ~Wrapper() { delete rep_; }
+    const char* Name() const { return rep_->Name(); }
+    void CreateFilter(const Slice* keys, int n, std::string* dst) const {
+      return rep_->CreateFilter(keys, n, dst);
+    }
+    bool KeyMayMatch(const Slice& key, const Slice& filter) const {
+      return rep_->KeyMayMatch(key, filter);
+    }
+    static void DoNothing(void*) { }
+  };
+  Wrapper* wrapper = new Wrapper;
+  wrapper->rep_ = NewBloomFilterPolicy(bits_per_key);
+  wrapper->state_ = NULL;
+  wrapper->destructor_ = &Wrapper::DoNothing;
+  return wrapper;
+}
+
 leveldb_readoptions_t* leveldb_readoptions_create() {
   return new leveldb_readoptions_t;
 }
@@ -468,4 +580,16 @@ void leveldb_env_destroy(leveldb_env_t* env) {
   delete env;
 }
 
+void leveldb_free(void* ptr) {
+  free(ptr);
+}
+
+int leveldb_major_version() {
+  return kMajorVersion;
+}
+
+int leveldb_minor_version() {
+  return kMinorVersion;
+}
+
 }  // end extern "C"

db/c_test.c

@@ -19,6 +19,13 @@ static void StartPhase(const char* name) {
   phase = name;
 }
 
+static const char* GetTempDir(void) {
+    const char* ret = getenv("TEST_TMPDIR");
+    if (ret == NULL || ret[0] == '\0')
+        ret = "/tmp";
+    return ret;
+}
+
 #define CheckNoError(err)                                               \
   if ((err) != NULL) {                                                  \
     fprintf(stderr, "%s:%d: %s: %s\n", __FILE__, __LINE__, phase, (err)); \
@@ -122,6 +129,31 @@ static const char* CmpName(void* arg) {
   return "foo";
 }
 
+// Custom filter policy
+static unsigned char fake_filter_result = 1;
+static void FilterDestroy(void* arg) { }
+static const char* FilterName(void* arg) {
+  return "TestFilter";
+}
+static char* FilterCreate(
+    void* arg,
+    const char* const* key_array, const size_t* key_length_array,
+    int num_keys,
+    size_t* filter_length) {
+  *filter_length = 4;
+  char* result = malloc(4);
+  memcpy(result, "fake", 4);
+  return result;
+}
+unsigned char FilterKeyMatch(
+    void* arg,
+    const char* key, size_t length,
+    const char* filter, size_t filter_length) {
+  CheckCondition(filter_length == 4);
+  CheckCondition(memcmp(filter, "fake", 4) == 0);
+  return fake_filter_result;
+}
+
 int main(int argc, char** argv) {
   leveldb_t* db;
   leveldb_comparator_t* cmp;
@@ -131,8 +163,14 @@ int main(int argc, char** argv) {
   leveldb_readoptions_t* roptions;
   leveldb_writeoptions_t* woptions;
   char* err = NULL;
+  int run = -1;
 
-  snprintf(dbname, sizeof(dbname), "/tmp/leveldb_c_test-%d",
+  CheckCondition(leveldb_major_version() >= 1);
+  CheckCondition(leveldb_minor_version() >= 1);
+
+  snprintf(dbname, sizeof(dbname),
+           "%s/leveldb_c_test-%d",
+           GetTempDir(),
            ((int) geteuid()));
 
   StartPhase("create_objects");
@@ -169,6 +207,12 @@ int main(int argc, char** argv) {
   CheckCondition(err != NULL);
   Free(&err);
 
+  StartPhase("leveldb_free");
+  db = leveldb_open(options, dbname, &err);
+  CheckCondition(err != NULL);
+  leveldb_free(err);
+  err = NULL;
+
   StartPhase("open");
   leveldb_options_set_create_if_missing(options, 1);
   db = leveldb_open(options, dbname, &err);
@@ -180,6 +224,14 @@ int main(int argc, char** argv) {
   CheckNoError(err);
   CheckGet(db, roptions, "foo", "hello");
 
+  StartPhase("compactall");
+  leveldb_compact_range(db, NULL, 0, NULL, 0);
+  CheckGet(db, roptions, "foo", "hello");
+
+  StartPhase("compactrange");
+  leveldb_compact_range(db, "a", 1, "z", 1);
+  CheckGet(db, roptions, "foo", "hello");
+
   StartPhase("writebatch");
   {
     leveldb_writebatch_t* wb = leveldb_writebatch_create();
@@ -279,6 +331,49 @@ int main(int argc, char** argv) {
     CheckGet(db, roptions, "foo", NULL);
     CheckGet(db, roptions, "bar", NULL);
     CheckGet(db, roptions, "box", "c");
+    leveldb_options_set_create_if_missing(options, 1);
+    leveldb_options_set_error_if_exists(options, 1);
+  }
+
+  StartPhase("filter");
+  for (run = 0; run < 2; run++) {
+    // First run uses custom filter, second run uses bloom filter
+    CheckNoError(err);
+    leveldb_filterpolicy_t* policy;
+    if (run == 0) {
+      policy = leveldb_filterpolicy_create(
+          NULL, FilterDestroy, FilterCreate, FilterKeyMatch, FilterName);
+    } else {
+      policy = leveldb_filterpolicy_create_bloom(10);
+    }
+
+    // Create new database
+    leveldb_close(db);
+    leveldb_destroy_db(options, dbname, &err);
+    leveldb_options_set_filter_policy(options, policy);
+    db = leveldb_open(options, dbname, &err);
+    CheckNoError(err);
+    leveldb_put(db, woptions, "foo", 3, "foovalue", 8, &err);
+    CheckNoError(err);
+    leveldb_put(db, woptions, "bar", 3, "barvalue", 8, &err);
+    CheckNoError(err);
+    leveldb_compact_range(db, NULL, 0, NULL, 0);
+
+    fake_filter_result = 1;
+    CheckGet(db, roptions, "foo", "foovalue");
+    CheckGet(db, roptions, "bar", "barvalue");
+    if (phase == 0) {
+      // Must not find value when custom filter returns false
+      fake_filter_result = 0;
+      CheckGet(db, roptions, "foo", NULL);
+      CheckGet(db, roptions, "bar", NULL);
+      fake_filter_result = 1;
+
+      CheckGet(db, roptions, "foo", "foovalue");
+      CheckGet(db, roptions, "bar", "barvalue");
+    }
+    leveldb_options_set_filter_policy(options, NULL);
+    leveldb_filterpolicy_destroy(policy);
   }
 
   StartPhase("cleanup");

db/corruption_test.cc

@@ -35,6 +35,7 @@ class CorruptionTest {
   CorruptionTest() {
     tiny_cache_ = NewLRUCache(100);
     options_.env = &env_;
+    options_.block_cache = tiny_cache_;
     dbname_ = test::TmpDir() + "/db_test";
     DestroyDB(dbname_, options_);
 
@@ -50,17 +51,14 @@ class CorruptionTest {
      delete tiny_cache_;
   }
 
-  Status TryReopen(Options* options = NULL) {
+  Status TryReopen() {
     delete db_;
     db_ = NULL;
-    Options opt = (options ? *options : options_);
-    opt.env = &env_;
-    opt.block_cache = tiny_cache_;
-    return DB::Open(opt, dbname_, &db_);
+    return DB::Open(options_, dbname_, &db_);
   }
 
-  void Reopen(Options* options = NULL) {
-    ASSERT_OK(TryReopen(options));
+  void Reopen() {
+    ASSERT_OK(TryReopen());
   }
 
   void RepairDB() {
@@ -77,7 +75,13 @@ class CorruptionTest {
       Slice key = Key(i, &key_space);
       batch.Clear();
       batch.Put(key, Value(i, &value_space));
-      ASSERT_OK(db_->Write(WriteOptions(), &batch));
+      WriteOptions options;
+      // Corrupt() doesn't work without this sync on windows; stat reports 0 for
+      // the file size.
+      if (i == n - 1) {
+        options.sync = true;
+      }
+      ASSERT_OK(db_->Write(options, &batch));
     }
   }
 
@@ -92,6 +96,10 @@ class CorruptionTest {
     for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
       uint64_t key;
       Slice in(iter->key());
+      if (in == "" || in == "~") {
+        // Ignore boundary keys.
+        continue;
+      }
       if (!ConsumeDecimalNumber(&in, &key) ||
           !in.empty() ||
           key < next_expected) {
@@ -123,7 +131,7 @@ class CorruptionTest {
     FileType type;
     std::string fname;
     int picked_number = -1;
-    for (int i = 0; i < filenames.size(); i++) {
+    for (size_t i = 0; i < filenames.size(); i++) {
       if (ParseFileName(filenames[i], &number, &type) &&
           type == filetype &&
           int(number) > picked_number) {  // Pick latest file
@@ -233,7 +241,23 @@ TEST(CorruptionTest, TableFile) {
   dbi->TEST_CompactRange(1, NULL, NULL);
 
   Corrupt(kTableFile, 100, 1);
-  Check(99, 99);
+  Check(90, 99);
+}
+
+TEST(CorruptionTest, TableFileRepair) {
+  options_.block_size = 2 * kValueSize;  // Limit scope of corruption
+  options_.paranoid_checks = true;
+  Reopen();
+  Build(100);
+  DBImpl* dbi = reinterpret_cast<DBImpl*>(db_);
+  dbi->TEST_CompactMemTable();
+  dbi->TEST_CompactRange(0, NULL, NULL);
+  dbi->TEST_CompactRange(1, NULL, NULL);
+
+  Corrupt(kTableFile, 100, 1);
+  RepairDB();
+  Reopen();
+  Check(95, 99);
 }
 
 TEST(CorruptionTest, TableFileIndexData) {
@@ -299,7 +323,7 @@ TEST(CorruptionTest, CompactionInputError) {
   ASSERT_EQ(1, Property("leveldb.num-files-at-level" + NumberToString(last)));
 
   Corrupt(kTableFile, 100, 1);
-  Check(9, 9);
+  Check(5, 9);
 
   // Force compactions by writing lots of values
   Build(10000);
@@ -307,32 +331,23 @@ TEST(CorruptionTest, CompactionInputError) {
 }
 
 TEST(CorruptionTest, CompactionInputErrorParanoid) {
-  Options options;
-  options.paranoid_checks = true;
-  options.write_buffer_size = 1048576;
-  Reopen(&options);
+  options_.paranoid_checks = true;
+  options_.write_buffer_size = 512 << 10;
+  Reopen();
   DBImpl* dbi = reinterpret_cast<DBImpl*>(db_);
 
-  // Fill levels >= 1 so memtable compaction outputs to level 1
-  for (int level = 1; level < config::kNumLevels; level++) {
-    dbi->Put(WriteOptions(), "", "begin");
-    dbi->Put(WriteOptions(), "~", "end");
+  // Make multiple inputs so we need to compact.
+  for (int i = 0; i < 2; i++) {
+    Build(10);
     dbi->TEST_CompactMemTable();
+    Corrupt(kTableFile, 100, 1);
+    env_.SleepForMicroseconds(100000);
   }
+  dbi->CompactRange(NULL, NULL);
 
-  Build(10);
-  dbi->TEST_CompactMemTable();
-  ASSERT_EQ(1, Property("leveldb.num-files-at-level0"));
-
-  Corrupt(kTableFile, 100, 1);
-  Check(9, 9);
-
-  // Write must eventually fail because of corrupted table
-  Status s;
+  // Write must fail because of corrupted table
   std::string tmp1, tmp2;
-  for (int i = 0; i < 10000 && s.ok(); i++) {
-    s = db_->Put(WriteOptions(), Key(i, &tmp1), Value(i, &tmp2));
-  }
+  Status s = db_->Put(WriteOptions(), Key(5, &tmp1), Value(5, &tmp2));
   ASSERT_TRUE(!s.ok()) << "write did not fail in corrupted paranoid db";
 }
 

db/db_bench.cc

@@ -25,15 +25,20 @@
 //      overwrite     -- overwrite N values in random key order in async mode
 //      fillsync      -- write N/100 values in random key order in sync mode
 //      fill100K      -- write N/1000 100K values in random order in async mode
+//      deleteseq     -- delete N keys in sequential order
+//      deleterandom  -- delete N keys in random order
 //      readseq       -- read N times sequentially
 //      readreverse   -- read N times in reverse order
 //      readrandom    -- read N times in random order
+//      readmissing   -- read N missing keys in random order
 //      readhot       -- read N times in random order from 1% section of DB
+//      seekrandom    -- N random seeks
 //      crc32c        -- repeated crc32c of 4K of data
 //      acquireload   -- load N*1000 times
 //   Meta operations:
 //      compact     -- Compact the entire DB
 //      stats       -- Print DB stats
+//      sstables    -- Print sstable info
 //      heapprofile -- Dump a heap profile (if supported by this port)
 static const char* FLAGS_benchmarks =
     "fillseq,"
@@ -85,13 +90,17 @@ static int FLAGS_cache_size = -1;
 // Maximum number of files to keep open at the same time (use default if == 0)
 static int FLAGS_open_files = 0;
 
+// Bloom filter bits per key.
+// Negative means use default settings.
+static int FLAGS_bloom_bits = -1;
+
 // If true, do not destroy the existing database.  If you set this
 // flag and also specify a benchmark that wants a fresh database, that
 // benchmark will fail.
 static bool FLAGS_use_existing_db = false;
 
 // Use the db with the following name.
-static const char* FLAGS_db = "/tmp/dbbench";
+static const char* FLAGS_db = NULL;
 
 namespace leveldb {
 
@@ -119,7 +128,7 @@ class RandomGenerator {
     pos_ = 0;
   }
 
-  Slice Generate(int len) {
+  Slice Generate(size_t len) {
     if (pos_ + len > data_.size()) {
       pos_ = 0;
       assert(len < data_.size());
@@ -130,11 +139,11 @@ class RandomGenerator {
 };
 
 static Slice TrimSpace(Slice s) {
-  int start = 0;
+  size_t start = 0;
   while (start < s.size() && isspace(s[start])) {
     start++;
   }
-  int limit = s.size();
+  size_t limit = s.size();
   while (limit > start && isspace(s[limit-1])) {
     limit--;
   }
@@ -293,6 +302,7 @@ struct ThreadState {
 class Benchmark {
  private:
   Cache* cache_;
+  const FilterPolicy* filter_policy_;
   DB* db_;
   int num_;
   int value_size_;
@@ -378,6 +388,9 @@ class Benchmark {
  public:
   Benchmark()
   : cache_(FLAGS_cache_size >= 0 ? NewLRUCache(FLAGS_cache_size) : NULL),
+    filter_policy_(FLAGS_bloom_bits >= 0
+                   ? NewBloomFilterPolicy(FLAGS_bloom_bits)
+                   : NULL),
     db_(NULL),
     num_(FLAGS_num),
     value_size_(FLAGS_value_size),
@@ -386,7 +399,7 @@ class Benchmark {
     heap_counter_(0) {
     std::vector<std::string> files;
     Env::Default()->GetChildren(FLAGS_db, &files);
-    for (int i = 0; i < files.size(); i++) {
+    for (size_t i = 0; i < files.size(); i++) {
       if (Slice(files[i]).starts_with("heap-")) {
         Env::Default()->DeleteFile(std::string(FLAGS_db) + "/" + files[i]);
       }
@@ -399,6 +412,7 @@ class Benchmark {
   ~Benchmark() {
     delete db_;
     delete cache_;
+    delete filter_policy_;
   }
 
   void Run() {
@@ -457,11 +471,19 @@ class Benchmark {
         method = &Benchmark::ReadReverse;
       } else if (name == Slice("readrandom")) {
         method = &Benchmark::ReadRandom;
+      } else if (name == Slice("readmissing")) {
+        method = &Benchmark::ReadMissing;
+      } else if (name == Slice("seekrandom")) {
+        method = &Benchmark::SeekRandom;
       } else if (name == Slice("readhot")) {
         method = &Benchmark::ReadHot;
       } else if (name == Slice("readrandomsmall")) {
         reads_ /= 1000;
         method = &Benchmark::ReadRandom;
+      } else if (name == Slice("deleteseq")) {
+        method = &Benchmark::DeleteSeq;
+      } else if (name == Slice("deleterandom")) {
+        method = &Benchmark::DeleteRandom;
       } else if (name == Slice("readwhilewriting")) {
         num_threads++;  // Add extra thread for writing
         method = &Benchmark::ReadWhileWriting;
@@ -478,7 +500,9 @@ class Benchmark {
       } else if (name == Slice("heapprofile")) {
         HeapProfile();
       } else if (name == Slice("stats")) {
-        PrintStats();
+        PrintStats("leveldb.stats");
+      } else if (name == Slice("sstables")) {
+        PrintStats("leveldb.sstables");
       } else {
         if (name != Slice()) {  // No error message for empty name
           fprintf(stderr, "unknown benchmark '%s'\n", name.ToString().c_str());
@@ -669,6 +693,8 @@ class Benchmark {
     options.create_if_missing = !FLAGS_use_existing_db;
     options.block_cache = cache_;
     options.write_buffer_size = FLAGS_write_buffer_size;
+    options.max_open_files = FLAGS_open_files;
+    options.filter_policy = filter_policy_;
     Status s = DB::Open(options, FLAGS_db, &db_);
     if (!s.ok()) {
       fprintf(stderr, "open error: %s\n", s.ToString().c_str());
@@ -743,10 +769,28 @@ class Benchmark {
   void ReadRandom(ThreadState* thread) {
     ReadOptions options;
     std::string value;
+    int found = 0;
     for (int i = 0; i < reads_; i++) {
       char key[100];
       const int k = thread->rand.Next() % FLAGS_num;
       snprintf(key, sizeof(key), "%016d", k);
+      if (db_->Get(options, key, &value).ok()) {
+        found++;
+      }
+      thread->stats.FinishedSingleOp();
+    }
+    char msg[100];
+    snprintf(msg, sizeof(msg), "(%d of %d found)", found, num_);
+    thread->stats.AddMessage(msg);
+  }
+
+  void ReadMissing(ThreadState* thread) {
+    ReadOptions options;
+    std::string value;
+    for (int i = 0; i < reads_; i++) {
+      char key[100];
+      const int k = thread->rand.Next() % FLAGS_num;
+      snprintf(key, sizeof(key), "%016d.", k);
       db_->Get(options, key, &value);
       thread->stats.FinishedSingleOp();
     }
@@ -765,6 +809,54 @@ class Benchmark {
     }
   }
 
+  void SeekRandom(ThreadState* thread) {
+    ReadOptions options;
+    std::string value;
+    int found = 0;
+    for (int i = 0; i < reads_; i++) {
+      Iterator* iter = db_->NewIterator(options);
+      char key[100];
+      const int k = thread->rand.Next() % FLAGS_num;
+      snprintf(key, sizeof(key), "%016d", k);
+      iter->Seek(key);
+      if (iter->Valid() && iter->key() == key) found++;
+      delete iter;
+      thread->stats.FinishedSingleOp();
+    }
+    char msg[100];
+    snprintf(msg, sizeof(msg), "(%d of %d found)", found, num_);
+    thread->stats.AddMessage(msg);
+  }
+
+  void DoDelete(ThreadState* thread, bool seq) {
+    RandomGenerator gen;
+    WriteBatch batch;
+    Status s;
+    for (int i = 0; i < num_; i += entries_per_batch_) {
+      batch.Clear();
+      for (int j = 0; j < entries_per_batch_; j++) {
+        const int k = seq ? i+j : (thread->rand.Next() % FLAGS_num);
+        char key[100];
+        snprintf(key, sizeof(key), "%016d", k);
+        batch.Delete(key);
+        thread->stats.FinishedSingleOp();
+      }
+      s = db_->Write(write_options_, &batch);
+      if (!s.ok()) {
+        fprintf(stderr, "del error: %s\n", s.ToString().c_str());
+        exit(1);
+      }
+    }
+  }
+
+  void DeleteSeq(ThreadState* thread) {
+    DoDelete(thread, true);
+  }
+
+  void DeleteRandom(ThreadState* thread) {
+    DoDelete(thread, false);
+  }
+
   void ReadWhileWriting(ThreadState* thread) {
     if (thread->tid > 0) {
       ReadRandom(thread);
@@ -799,9 +891,9 @@ class Benchmark {
     db_->CompactRange(NULL, NULL);
   }
 
-  void PrintStats() {
+  void PrintStats(const char* key) {
     std::string stats;
-    if (!db_->GetProperty("leveldb.stats", &stats)) {
+    if (!db_->GetProperty(key, &stats)) {
       stats = "(failed)";
     }
     fprintf(stdout, "\n%s\n", stats.c_str());
@@ -834,6 +926,7 @@ class Benchmark {
 int main(int argc, char** argv) {
   FLAGS_write_buffer_size = leveldb::Options().write_buffer_size;
   FLAGS_open_files = leveldb::Options().max_open_files;
+  std::string default_db_path;
 
   for (int i = 1; i < argc; i++) {
     double d;
@@ -861,6 +954,8 @@ int main(int argc, char** argv) {
       FLAGS_write_buffer_size = n;
     } else if (sscanf(argv[i], "--cache_size=%d%c", &n, &junk) == 1) {
       FLAGS_cache_size = n;
+    } else if (sscanf(argv[i], "--bloom_bits=%d%c", &n, &junk) == 1) {
+      FLAGS_bloom_bits = n;
     } else if (sscanf(argv[i], "--open_files=%d%c", &n, &junk) == 1) {
       FLAGS_open_files = n;
     } else if (strncmp(argv[i], "--db=", 5) == 0) {
@@ -871,6 +966,13 @@ int main(int argc, char** argv) {
     }
   }
 
+  // Choose a location for the test database if none given with --db=<path>
+  if (FLAGS_db == NULL) {
+      leveldb::Env::Default()->GetTestDirectory(&default_db_path);
+      default_db_path += "/dbbench";
+      FLAGS_db = default_db_path.c_str();
+  }
+
   leveldb::Benchmark benchmark;
   benchmark.Run();
   return 0;

db/db_impl.cc

@@ -35,6 +35,8 @@
 
 namespace leveldb {
 
+const int kNumNonTableCacheFiles = 10;
+
 // Information kept for every waiting writer
 struct DBImpl::Writer {
   Status status;
@@ -87,12 +89,14 @@ static void ClipToRange(T* ptr, V minvalue, V maxvalue) {
 }
 Options SanitizeOptions(const std::string& dbname,
                         const InternalKeyComparator* icmp,
+                        const InternalFilterPolicy* ipolicy,
                         const Options& src) {
   Options result = src;
   result.comparator = icmp;
-  ClipToRange(&result.max_open_files,           20,     50000);
-  ClipToRange(&result.write_buffer_size,        64<<10, 1<<30);
-  ClipToRange(&result.block_size,               1<<10,  4<<20);
+  result.filter_policy = (src.filter_policy != NULL) ? ipolicy : NULL;
+  ClipToRange(&result.max_open_files,    64 + kNumNonTableCacheFiles, 50000);
+  ClipToRange(&result.write_buffer_size, 64<<10,                      1<<30);
+  ClipToRange(&result.block_size,        1<<10,                       4<<20);
   if (result.info_log == NULL) {
     // Open a log file in the same directory as the db
     src.env->CreateDir(dbname);  // In case it does not exist
@@ -109,12 +113,14 @@ Options SanitizeOptions(const std::string& dbname,
   return result;
 }
 
-DBImpl::DBImpl(const Options& options, const std::string& dbname)
-    : env_(options.env),
-      internal_comparator_(options.comparator),
-      options_(SanitizeOptions(dbname, &internal_comparator_, options)),
-      owns_info_log_(options_.info_log != options.info_log),
-      owns_cache_(options_.block_cache != options.block_cache),
+DBImpl::DBImpl(const Options& raw_options, const std::string& dbname)
+    : env_(raw_options.env),
+      internal_comparator_(raw_options.comparator),
+      internal_filter_policy_(raw_options.filter_policy),
+      options_(SanitizeOptions(dbname, &internal_comparator_,
+                               &internal_filter_policy_, raw_options)),
+      owns_info_log_(options_.info_log != raw_options.info_log),
+      owns_cache_(options_.block_cache != raw_options.block_cache),
       dbname_(dbname),
       db_lock_(NULL),
       shutting_down_(NULL),
@@ -124,6 +130,7 @@ DBImpl::DBImpl(const Options& options, const std::string& dbname)
       logfile_(NULL),
       logfile_number_(0),
       log_(NULL),
+      seed_(0),
       tmp_batch_(new WriteBatch),
       bg_compaction_scheduled_(false),
       manual_compaction_(NULL) {
@@ -131,7 +138,7 @@ DBImpl::DBImpl(const Options& options, const std::string& dbname)
   has_imm_.Release_Store(NULL);
 
   // Reserve ten files or so for other uses and give the rest to TableCache.
-  const int table_cache_size = options.max_open_files - 10;
+  const int table_cache_size = options_.max_open_files - kNumNonTableCacheFiles;
   table_cache_ = new TableCache(dbname_, &options_, table_cache_size);
 
   versions_ = new VersionSet(dbname_, &options_, table_cache_,
@@ -209,6 +216,12 @@ void DBImpl::MaybeIgnoreError(Status* s) const {
 }
 
 void DBImpl::DeleteObsoleteFiles() {
+  if (!bg_error_.ok()) {
+    // After a background error, we don't know whether a new version may
+    // or may not have been committed, so we cannot safely garbage collect.
+    return;
+  }
+
   // Make a set of all of the live files
   std::set<uint64_t> live = pending_outputs_;
   versions_->AddLiveFiles(&live);
@@ -306,16 +319,24 @@ Status DBImpl::Recover(VersionEdit* edit) {
     if (!s.ok()) {
       return s;
     }
+    std::set<uint64_t> expected;
+    versions_->AddLiveFiles(&expected);
     uint64_t number;
     FileType type;
     std::vector<uint64_t> logs;
     for (size_t i = 0; i < filenames.size(); i++) {
-      if (ParseFileName(filenames[i], &number, &type)
-          && type == kLogFile
-          && ((number >= min_log) || (number == prev_log))) {
-        logs.push_back(number);
+      if (ParseFileName(filenames[i], &number, &type)) {
+        expected.erase(number);
+        if (type == kLogFile && ((number >= min_log) || (number == prev_log)))
+          logs.push_back(number);
       }
     }
+    if (!expected.empty()) {
+      char buf[50];
+      snprintf(buf, sizeof(buf), "%d missing files; e.g.",
+               static_cast<int>(expected.size()));
+      return Status::Corruption(buf, TableFileName(dbname_, *(expected.begin())));
+    }
 
     // Recover in the order in which the logs were generated
     std::sort(logs.begin(), logs.end());
@@ -479,7 +500,7 @@ Status DBImpl::WriteLevel0Table(MemTable* mem, VersionEdit* edit,
   return s;
 }
 
-Status DBImpl::CompactMemTable() {
+void DBImpl::CompactMemTable() {
   mutex_.AssertHeld();
   assert(imm_ != NULL);
 
@@ -507,9 +528,9 @@ Status DBImpl::CompactMemTable() {
     imm_ = NULL;
     has_imm_.Release_Store(NULL);
     DeleteObsoleteFiles();
+  } else {
+    RecordBackgroundError(s);
   }
-
-  return s;
 }
 
 void DBImpl::CompactRange(const Slice* begin, const Slice* end) {
@@ -552,16 +573,18 @@ void DBImpl::TEST_CompactRange(int level, const Slice* begin,const Slice* end) {
   }
 
   MutexLock l(&mutex_);
-  while (!manual.done) {
-    while (manual_compaction_ != NULL) {
-      bg_cv_.Wait();
-    }
-    manual_compaction_ = &manual;
-    MaybeScheduleCompaction();
-    while (manual_compaction_ == &manual) {
+  while (!manual.done && !shutting_down_.Acquire_Load() && bg_error_.ok()) {
+    if (manual_compaction_ == NULL) {  // Idle
+      manual_compaction_ = &manual;
+      MaybeScheduleCompaction();
+    } else {  // Running either my compaction or another compaction.
       bg_cv_.Wait();
     }
   }
+  if (manual_compaction_ == &manual) {
+    // Cancel my manual compaction since we aborted early for some reason.
+    manual_compaction_ = NULL;
+  }
 }
 
 Status DBImpl::TEST_CompactMemTable() {
@@ -580,12 +603,22 @@ Status DBImpl::TEST_CompactMemTable() {
   return s;
 }
 
+void DBImpl::RecordBackgroundError(const Status& s) {
+  mutex_.AssertHeld();
+  if (bg_error_.ok()) {
+    bg_error_ = s;
+    bg_cv_.SignalAll();
+  }
+}
+
 void DBImpl::MaybeScheduleCompaction() {
   mutex_.AssertHeld();
   if (bg_compaction_scheduled_) {
     // Already scheduled
   } else if (shutting_down_.Acquire_Load()) {
     // DB is being deleted; no more background compactions
+  } else if (!bg_error_.ok()) {
+    // Already got an error; no more changes
   } else if (imm_ == NULL &&
              manual_compaction_ == NULL &&
              !versions_->NeedsCompaction()) {
@@ -603,9 +636,14 @@ void DBImpl::BGWork(void* db) {
 void DBImpl::BackgroundCall() {
   MutexLock l(&mutex_);
   assert(bg_compaction_scheduled_);
-  if (!shutting_down_.Acquire_Load()) {
+  if (shutting_down_.Acquire_Load()) {
+    // No more background work when shutting down.
+  } else if (!bg_error_.ok()) {
+    // No more background work after a background error.
+  } else {
     BackgroundCompaction();
   }
+
   bg_compaction_scheduled_ = false;
 
   // Previous compaction may have produced too many files in a level,
@@ -653,6 +691,9 @@ void DBImpl::BackgroundCompaction() {
     c->edit()->AddFile(c->level() + 1, f->number, f->file_size,
                        f->smallest, f->largest);
     status = versions_->LogAndApply(c->edit(), &mutex_);
+    if (!status.ok()) {
+      RecordBackgroundError(status);
+    }
     VersionSet::LevelSummaryStorage tmp;
     Log(options_.info_log, "Moved #%lld to level-%d %lld bytes %s: %s\n",
         static_cast<unsigned long long>(f->number),
@@ -663,6 +704,9 @@ void DBImpl::BackgroundCompaction() {
   } else {
     CompactionState* compact = new CompactionState(c);
     status = DoCompactionWork(compact);
+    if (!status.ok()) {
+      RecordBackgroundError(status);
+    }
     CleanupCompaction(compact);
     c->ReleaseInputs();
     DeleteObsoleteFiles();
@@ -676,9 +720,6 @@ void DBImpl::BackgroundCompaction() {
   } else {
     Log(options_.info_log,
         "Compaction error: %s", status.ToString().c_str());
-    if (options_.paranoid_checks && bg_error_.ok()) {
-      bg_error_ = status;
-    }
   }
 
   if (is_manual) {
@@ -963,6 +1004,9 @@ Status DBImpl::DoCompactionWork(CompactionState* compact) {
   if (status.ok()) {
     status = InstallCompactionResults(compact);
   }
+  if (!status.ok()) {
+    RecordBackgroundError(status);
+  }
   VersionSet::LevelSummaryStorage tmp;
   Log(options_.info_log,
       "compacted to: %s", versions_->LevelSummary(&tmp));
@@ -989,7 +1033,8 @@ static void CleanupIteratorState(void* arg1, void* arg2) {
 }  // namespace
 
 Iterator* DBImpl::NewInternalIterator(const ReadOptions& options,
-                                      SequenceNumber* latest_snapshot) {
+                                      SequenceNumber* latest_snapshot,
+                                      uint32_t* seed) {
   IterState* cleanup = new IterState;
   mutex_.Lock();
   *latest_snapshot = versions_->LastSequence();
@@ -1013,13 +1058,15 @@ Iterator* DBImpl::NewInternalIterator(const ReadOptions& options,
   cleanup->version = versions_->current();
   internal_iter->RegisterCleanup(CleanupIteratorState, cleanup, NULL);
 
+  *seed = ++seed_;
   mutex_.Unlock();
   return internal_iter;
 }
 
 Iterator* DBImpl::TEST_NewInternalIterator() {
   SequenceNumber ignored;
-  return NewInternalIterator(ReadOptions(), &ignored);
+  uint32_t ignored_seed;
+  return NewInternalIterator(ReadOptions(), &ignored, &ignored_seed);
 }
 
 int64_t DBImpl::TEST_MaxNextLevelOverlappingBytes() {
@@ -1076,12 +1123,21 @@ Status DBImpl::Get(const ReadOptions& options,
 
 Iterator* DBImpl::NewIterator(const ReadOptions& options) {
   SequenceNumber latest_snapshot;
-  Iterator* internal_iter = NewInternalIterator(options, &latest_snapshot);
+  uint32_t seed;
+  Iterator* iter = NewInternalIterator(options, &latest_snapshot, &seed);
   return NewDBIterator(
-      &dbname_, env_, user_comparator(), internal_iter,
+      this, user_comparator(), iter,
       (options.snapshot != NULL
        ? reinterpret_cast<const SnapshotImpl*>(options.snapshot)->number_
-       : latest_snapshot));
+       : latest_snapshot),
+      seed);
+}
+
+void DBImpl::RecordReadSample(Slice key) {
+  MutexLock l(&mutex_);
+  if (versions_->current()->RecordReadSample(key)) {
+    MaybeScheduleCompaction();
+  }
 }
 
 const Snapshot* DBImpl::GetSnapshot() {
@@ -1134,13 +1190,23 @@ Status DBImpl::Write(const WriteOptions& options, WriteBatch* my_batch) {
     {
       mutex_.Unlock();
       status = log_->AddRecord(WriteBatchInternal::Contents(updates));
+      bool sync_error = false;
       if (status.ok() && options.sync) {
         status = logfile_->Sync();
+        if (!status.ok()) {
+          sync_error = true;
+        }
       }
       if (status.ok()) {
         status = WriteBatchInternal::InsertInto(updates, mem_);
       }
       mutex_.Lock();
+      if (sync_error) {
+        // The state of the log file is indeterminate: the log record we
+        // just added may or may not show up when the DB is re-opened.
+        // So we force the DB into a mode where all future writes fail.
+        RecordBackgroundError(status);
+      }
     }
     if (updates == tmp_batch_) tmp_batch_->Clear();
 
@@ -1247,10 +1313,11 @@ Status DBImpl::MakeRoomForWrite(bool force) {
     } else if (imm_ != NULL) {
       // We have filled up the current memtable, but the previous
       // one is still being compacted, so we wait.
+      Log(options_.info_log, "Current memtable full; waiting...\n");
       bg_cv_.Wait();
     } else if (versions_->NumLevelFiles(0) >= config::kL0_StopWritesTrigger) {
       // There are too many level-0 files.
-      Log(options_.info_log, "waiting...\n");
+      Log(options_.info_log, "Too many L0 files; waiting...\n");
       bg_cv_.Wait();
     } else {
       // Attempt to switch to a new memtable and trigger compaction of old
@@ -1259,6 +1326,8 @@ Status DBImpl::MakeRoomForWrite(bool force) {
       WritableFile* lfile = NULL;
       s = env_->NewWritableFile(LogFileName(dbname_, new_log_number), &lfile);
       if (!s.ok()) {
+        // Avoid chewing through file number space in a tight loop.
+        versions_->ReuseFileNumber(new_log_number);
         break;
       }
       delete log_;

db/db_impl.h

@@ -13,6 +13,7 @@
 #include "leveldb/db.h"
 #include "leveldb/env.h"
 #include "port/port.h"
+#include "port/thread_annotations.h"
 
 namespace leveldb {
 
@@ -58,20 +59,26 @@ class DBImpl : public DB {
   // file at a level >= 1.
   int64_t TEST_MaxNextLevelOverlappingBytes();
 
+  // Record a sample of bytes read at the specified internal key.
+  // Samples are taken approximately once every config::kReadBytesPeriod
+  // bytes.
+  void RecordReadSample(Slice key);
+
  private:
   friend class DB;
   struct CompactionState;
   struct Writer;
 
   Iterator* NewInternalIterator(const ReadOptions&,
-                                SequenceNumber* latest_snapshot);
+                                SequenceNumber* latest_snapshot,
+                                uint32_t* seed);
 
   Status NewDB();
 
   // Recover the descriptor from persistent storage.  May do a significant
   // amount of work to recover recently logged updates.  Any changes to
   // be made to the descriptor are added to *edit.
-  Status Recover(VersionEdit* edit);
+  Status Recover(VersionEdit* edit) EXCLUSIVE_LOCKS_REQUIRED(mutex_);
 
   void MaybeIgnoreError(Status* s) const;
 
@@ -80,31 +87,41 @@ class DBImpl : public DB {
 
   // Compact the in-memory write buffer to disk.  Switches to a new
   // log-file/memtable and writes a new descriptor iff successful.
-  Status CompactMemTable();
+  // Errors are recorded in bg_error_.
+  void CompactMemTable() EXCLUSIVE_LOCKS_REQUIRED(mutex_);
 
   Status RecoverLogFile(uint64_t log_number,
                         VersionEdit* edit,
-                        SequenceNumber* max_sequence);
+                        SequenceNumber* max_sequence)
+      EXCLUSIVE_LOCKS_REQUIRED(mutex_);
 
-  Status WriteLevel0Table(MemTable* mem, VersionEdit* edit, Version* base);
+  Status WriteLevel0Table(MemTable* mem, VersionEdit* edit, Version* base)
+      EXCLUSIVE_LOCKS_REQUIRED(mutex_);
 
-  Status MakeRoomForWrite(bool force /* compact even if there is room? */);
+  Status MakeRoomForWrite(bool force /* compact even if there is room? */)
+      EXCLUSIVE_LOCKS_REQUIRED(mutex_);
   WriteBatch* BuildBatchGroup(Writer** last_writer);
 
-  void MaybeScheduleCompaction();
+  void RecordBackgroundError(const Status& s);
+
+  void MaybeScheduleCompaction() EXCLUSIVE_LOCKS_REQUIRED(mutex_);
   static void BGWork(void* db);
   void BackgroundCall();
-  void BackgroundCompaction();
-  void CleanupCompaction(CompactionState* compact);
-  Status DoCompactionWork(CompactionState* compact);
+  void  BackgroundCompaction() EXCLUSIVE_LOCKS_REQUIRED(mutex_);
+  void CleanupCompaction(CompactionState* compact)
+      EXCLUSIVE_LOCKS_REQUIRED(mutex_);
+  Status DoCompactionWork(CompactionState* compact)
+      EXCLUSIVE_LOCKS_REQUIRED(mutex_);
 
   Status OpenCompactionOutputFile(CompactionState* compact);
   Status FinishCompactionOutputFile(CompactionState* compact, Iterator* input);
-  Status InstallCompactionResults(CompactionState* compact);
+  Status InstallCompactionResults(CompactionState* compact)
+      EXCLUSIVE_LOCKS_REQUIRED(mutex_);
 
   // Constant after construction
   Env* const env_;
   const InternalKeyComparator internal_comparator_;
+  const InternalFilterPolicy internal_filter_policy_;
   const Options options_;  // options_.comparator == &internal_comparator_
   bool owns_info_log_;
   bool owns_cache_;
@@ -126,6 +143,7 @@ class DBImpl : public DB {
   WritableFile* logfile_;
   uint64_t logfile_number_;
   log::Writer* log_;
+  uint32_t seed_;                // For sampling.
 
   // Queue of writers.
   std::deque<Writer*> writers_;
@@ -185,6 +203,7 @@ class DBImpl : public DB {
 // it is not equal to src.info_log.
 extern Options SanitizeOptions(const std::string& db,
                                const InternalKeyComparator* icmp,
+                               const InternalFilterPolicy* ipolicy,
                                const Options& src);
 
 }  // namespace leveldb

db/db_iter.cc

@@ -5,12 +5,14 @@
 #include "db/db_iter.h"
 
 #include "db/filename.h"
+#include "db/db_impl.h"
 #include "db/dbformat.h"
 #include "leveldb/env.h"
 #include "leveldb/iterator.h"
 #include "port/port.h"
 #include "util/logging.h"
 #include "util/mutexlock.h"
+#include "util/random.h"
 
 namespace leveldb {
 
@@ -46,15 +48,16 @@ class DBIter: public Iterator {
     kReverse
   };
 
-  DBIter(const std::string* dbname, Env* env,
-         const Comparator* cmp, Iterator* iter, SequenceNumber s)
-      : dbname_(dbname),
-        env_(env),
+  DBIter(DBImpl* db, const Comparator* cmp, Iterator* iter, SequenceNumber s,
+         uint32_t seed)
+      : db_(db),
         user_comparator_(cmp),
         iter_(iter),
         sequence_(s),
         direction_(kForward),
-        valid_(false) {
+        valid_(false),
+        rnd_(seed),
+        bytes_counter_(RandomPeriod()) {
   }
   virtual ~DBIter() {
     delete iter_;
@@ -100,8 +103,12 @@ class DBIter: public Iterator {
     }
   }
 
-  const std::string* const dbname_;
-  Env* const env_;
+  // Pick next gap with average value of config::kReadBytesPeriod.
+  ssize_t RandomPeriod() {
+    return rnd_.Uniform(2*config::kReadBytesPeriod);
+  }
+
+  DBImpl* db_;
   const Comparator* const user_comparator_;
   Iterator* const iter_;
   SequenceNumber const sequence_;
@@ -112,13 +119,23 @@ class DBIter: public Iterator {
   Direction direction_;
   bool valid_;
 
+  Random rnd_;
+  ssize_t bytes_counter_;
+
   // No copying allowed
   DBIter(const DBIter&);
   void operator=(const DBIter&);
 };
 
 inline bool DBIter::ParseKey(ParsedInternalKey* ikey) {
-  if (!ParseInternalKey(iter_->key(), ikey)) {
+  Slice k = iter_->key();
+  ssize_t n = k.size() + iter_->value().size();
+  bytes_counter_ -= n;
+  while (bytes_counter_ < 0) {
+    bytes_counter_ += RandomPeriod();
+    db_->RecordReadSample(k);
+  }
+  if (!ParseInternalKey(k, ikey)) {
     status_ = Status::Corruption("corrupted internal key in DBIter");
     return false;
   } else {
@@ -144,12 +161,13 @@ void DBIter::Next() {
       saved_key_.clear();
       return;
     }
+    // saved_key_ already contains the key to skip past.
+  } else {
+    // Store in saved_key_ the current key so we skip it below.
+    SaveKey(ExtractUserKey(iter_->key()), &saved_key_);
   }
 
-  // Temporarily use saved_key_ as storage for key to skip.
-  std::string* skip = &saved_key_;
-  SaveKey(ExtractUserKey(iter_->key()), skip);
-  FindNextUserEntry(true, skip);
+  FindNextUserEntry(true, &saved_key_);
 }
 
 void DBIter::FindNextUserEntry(bool skipping, std::string* skip) {
@@ -288,12 +306,12 @@ void DBIter::SeekToLast() {
 }  // anonymous namespace
 
 Iterator* NewDBIterator(
-    const std::string* dbname,
-    Env* env,
+    DBImpl* db,
     const Comparator* user_key_comparator,
     Iterator* internal_iter,
-    const SequenceNumber& sequence) {
-  return new DBIter(dbname, env, user_key_comparator, internal_iter, sequence);
+    SequenceNumber sequence,
+    uint32_t seed) {
+  return new DBIter(db, user_key_comparator, internal_iter, sequence, seed);
 }
 
 }  // namespace leveldb

db/db_iter.h

@@ -11,15 +11,17 @@
 
 namespace leveldb {
 
+class DBImpl;
+
 // 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(
-    const std::string* dbname,
-    Env* env,
+    DBImpl* db,
     const Comparator* user_key_comparator,
     Iterator* internal_iter,
-    const SequenceNumber& sequence);
+    SequenceNumber sequence,
+    uint32_t seed);
 
 }  // namespace leveldb
 

db/dbformat.cc

@@ -26,7 +26,7 @@ std::string ParsedInternalKey::DebugString() const {
            (unsigned long long) sequence,
            int(type));
   std::string result = "'";
-  result += user_key.ToString();
+  result += EscapeString(user_key.ToString());
   result += buf;
   return result;
 }
@@ -98,6 +98,26 @@ 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

db/dbformat.h

@@ -8,6 +8,7 @@
 #include <stdio.h>
 #include "leveldb/comparator.h"
 #include "leveldb/db.h"
+#include "leveldb/filter_policy.h"
 #include "leveldb/slice.h"
 #include "leveldb/table_builder.h"
 #include "util/coding.h"
@@ -37,6 +38,9 @@ static const int kL0_StopWritesTrigger = 12;
 // space if the same key space is being repeatedly overwritten.
 static const int kMaxMemCompactLevel = 2;
 
+// Approximate gap in bytes between samples of data read during iteration.
+static const int kReadBytesPeriod = 1048576;
+
 }  // namespace config
 
 class InternalKey;
@@ -123,6 +127,17 @@ class InternalKeyComparator : public Comparator {
   int Compare(const InternalKey& a, const InternalKey& 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.

db/filename.cc

@@ -30,6 +30,11 @@ std::string LogFileName(const std::string& name, uint64_t number) {
 }
 
 std::string TableFileName(const std::string& name, uint64_t number) {
+  assert(number > 0);
+  return MakeFileName(name, number, "ldb");
+}
+
+std::string SSTTableFileName(const std::string& name, uint64_t number) {
   assert(number > 0);
   return MakeFileName(name, number, "sst");
 }
@@ -71,7 +76,7 @@ std::string OldInfoLogFileName(const std::string& dbname) {
 //    dbname/LOG
 //    dbname/LOG.old
 //    dbname/MANIFEST-[0-9]+
-//    dbname/[0-9]+.(log|sst)
+//    dbname/[0-9]+.(log|sst|ldb)
 bool ParseFileName(const std::string& fname,
                    uint64_t* number,
                    FileType* type) {
@@ -106,7 +111,7 @@ bool ParseFileName(const std::string& fname,
     Slice suffix = rest;
     if (suffix == Slice(".log")) {
       *type = kLogFile;
-    } else if (suffix == Slice(".sst")) {
+    } else if (suffix == Slice(".sst") || suffix == Slice(".ldb")) {
       *type = kTableFile;
     } else if (suffix == Slice(".dbtmp")) {
       *type = kTempFile;

db/filename.h

@@ -37,6 +37,11 @@ extern std::string LogFileName(const std::string& dbname, uint64_t number);
 // "dbname".
 extern std::string TableFileName(const std::string& dbname, uint64_t number);
 
+// Return the legacy file name for an sstable with the specified number
+// in the db named by "dbname". The result will be prefixed with
+// "dbname".
+extern std::string SSTTableFileName(const std::string& dbname, uint64_t number);
+
 // Return the name of the descriptor file for the db named by
 // "dbname" and the specified incarnation number.  The result will be
 // prefixed with "dbname".

db/filename_test.cc

@@ -27,6 +27,7 @@ TEST(FileNameTest, Parse) {
     { "100.log",            100,   kLogFile },
     { "0.log",              0,     kLogFile },
     { "0.sst",              0,     kTableFile },
+    { "0.ldb",              0,     kTableFile },
     { "CURRENT",            0,     kCurrentFile },
     { "LOCK",               0,     kDBLockFile },
     { "MANIFEST-2",         2,     kDescriptorFile },
@@ -70,7 +71,7 @@ TEST(FileNameTest, Parse) {
   for (int i = 0; i < sizeof(errors) / sizeof(errors[0]); i++) {
     std::string f = errors[i];
     ASSERT_TRUE(!ParseFileName(f, &number, &type)) << f;
-  };
+  }
 }
 
 TEST(FileNameTest, Construction) {

db/leveldb_main.cc

@@ -0,0 +1,238 @@
+// Copyright (c) 2012 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 <stdio.h>
+#include "db/dbformat.h"
+#include "db/filename.h"
+#include "db/log_reader.h"
+#include "db/version_edit.h"
+#include "db/write_batch_internal.h"
+#include "leveldb/env.h"
+#include "leveldb/iterator.h"
+#include "leveldb/options.h"
+#include "leveldb/status.h"
+#include "leveldb/table.h"
+#include "leveldb/write_batch.h"
+#include "util/logging.h"
+
+namespace leveldb {
+
+namespace {
+
+bool GuessType(const std::string& fname, FileType* type) {
+  size_t pos = fname.rfind('/');
+  std::string basename;
+  if (pos == std::string::npos) {
+    basename = fname;
+  } else {
+    basename = std::string(fname.data() + pos + 1, fname.size() - pos - 1);
+  }
+  uint64_t ignored;
+  return ParseFileName(basename, &ignored, type);
+}
+
+// Notified when log reader encounters corruption.
+class CorruptionReporter : public log::Reader::Reporter {
+ public:
+  virtual void Corruption(size_t bytes, const Status& status) {
+    printf("corruption: %d bytes; %s\n",
+            static_cast<int>(bytes),
+            status.ToString().c_str());
+  }
+};
+
+// Print contents of a log file. (*func)() is called on every record.
+bool PrintLogContents(Env* env, const std::string& fname,
+                      void (*func)(Slice)) {
+  SequentialFile* file;
+  Status s = env->NewSequentialFile(fname, &file);
+  if (!s.ok()) {
+    fprintf(stderr, "%s\n", s.ToString().c_str());
+    return false;
+  }
+  CorruptionReporter reporter;
+  log::Reader reader(file, &reporter, true, 0);
+  Slice record;
+  std::string scratch;
+  while (reader.ReadRecord(&record, &scratch)) {
+    printf("--- offset %llu; ",
+           static_cast<unsigned long long>(reader.LastRecordOffset()));
+    (*func)(record);
+  }
+  delete file;
+  return true;
+}
+
+// Called on every item found in a WriteBatch.
+class WriteBatchItemPrinter : public WriteBatch::Handler {
+ public:
+  uint64_t offset_;
+  uint64_t sequence_;
+
+  virtual void Put(const Slice& key, const Slice& value) {
+    printf("  put '%s' '%s'\n",
+           EscapeString(key).c_str(),
+           EscapeString(value).c_str());
+  }
+  virtual void Delete(const Slice& key) {
+    printf("  del '%s'\n",
+           EscapeString(key).c_str());
+  }
+};
+
+
+// Called on every log record (each one of which is a WriteBatch)
+// found in a kLogFile.
+static void WriteBatchPrinter(Slice record) {
+  if (record.size() < 12) {
+    printf("log record length %d is too small\n",
+           static_cast<int>(record.size()));
+    return;
+  }
+  WriteBatch batch;
+  WriteBatchInternal::SetContents(&batch, record);
+  printf("sequence %llu\n",
+         static_cast<unsigned long long>(WriteBatchInternal::Sequence(&batch)));
+  WriteBatchItemPrinter batch_item_printer;
+  Status s = batch.Iterate(&batch_item_printer);
+  if (!s.ok()) {
+    printf("  error: %s\n", s.ToString().c_str());
+  }
+}
+
+bool DumpLog(Env* env, const std::string& fname) {
+  return PrintLogContents(env, fname, WriteBatchPrinter);
+}
+
+// Called on every log record (each one of which is a WriteBatch)
+// found in a kDescriptorFile.
+static void VersionEditPrinter(Slice record) {
+  VersionEdit edit;
+  Status s = edit.DecodeFrom(record);
+  if (!s.ok()) {
+    printf("%s\n", s.ToString().c_str());
+    return;
+  }
+  printf("%s", edit.DebugString().c_str());
+}
+
+bool DumpDescriptor(Env* env, const std::string& fname) {
+  return PrintLogContents(env, fname, VersionEditPrinter);
+}
+
+bool DumpTable(Env* env, const std::string& fname) {
+  uint64_t file_size;
+  RandomAccessFile* file = NULL;
+  Table* table = NULL;
+  Status s = env->GetFileSize(fname, &file_size);
+  if (s.ok()) {
+    s = env->NewRandomAccessFile(fname, &file);
+  }
+  if (s.ok()) {
+    // We use the default comparator, which may or may not match the
+    // comparator used in this database. However this should not cause
+    // problems since we only use Table operations that do not require
+    // any comparisons.  In particular, we do not call Seek or Prev.
+    s = Table::Open(Options(), file, file_size, &table);
+  }
+  if (!s.ok()) {
+    fprintf(stderr, "%s\n", s.ToString().c_str());
+    delete table;
+    delete file;
+    return false;
+  }
+
+  ReadOptions ro;
+  ro.fill_cache = false;
+  Iterator* iter = table->NewIterator(ro);
+  for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
+    ParsedInternalKey key;
+    if (!ParseInternalKey(iter->key(), &key)) {
+      printf("badkey '%s' => '%s'\n",
+             EscapeString(iter->key()).c_str(),
+             EscapeString(iter->value()).c_str());
+    } else {
+      char kbuf[20];
+      const char* type;
+      if (key.type == kTypeDeletion) {
+        type = "del";
+      } else if (key.type == kTypeValue) {
+        type = "val";
+      } else {
+        snprintf(kbuf, sizeof(kbuf), "%d", static_cast<int>(key.type));
+        type = kbuf;
+      }
+      printf("'%s' @ %8llu : %s => '%s'\n",
+             EscapeString(key.user_key).c_str(),
+             static_cast<unsigned long long>(key.sequence),
+             type,
+             EscapeString(iter->value()).c_str());
+    }
+  }
+  s = iter->status();
+  if (!s.ok()) {
+    printf("iterator error: %s\n", s.ToString().c_str());
+  }
+
+  delete iter;
+  delete table;
+  delete file;
+  return true;
+}
+
+bool DumpFile(Env* env, const std::string& fname) {
+  FileType ftype;
+  if (!GuessType(fname, &ftype)) {
+    fprintf(stderr, "%s: unknown file type\n", fname.c_str());
+    return false;
+  }
+  switch (ftype) {
+    case kLogFile:         return DumpLog(env, fname);
+    case kDescriptorFile:  return DumpDescriptor(env, fname);
+    case kTableFile:       return DumpTable(env, fname);
+
+    default: {
+      fprintf(stderr, "%s: not a dump-able file type\n", fname.c_str());
+      break;
+    }
+  }
+  return false;
+}
+
+bool HandleDumpCommand(Env* env, char** files, int num) {
+  bool ok = true;
+  for (int i = 0; i < num; i++) {
+    ok &= DumpFile(env, files[i]);
+  }
+  return ok;
+}
+
+}
+}  // namespace leveldb
+
+static void Usage() {
+  fprintf(
+      stderr,
+      "Usage: leveldbutil command...\n"
+      "   dump files...         -- dump contents of specified files\n"
+      );
+}
+
+int main(int argc, char** argv) {
+  leveldb::Env* env = leveldb::Env::Default();
+  bool ok = true;
+  if (argc < 2) {
+    Usage();
+    ok = false;
+  } else {
+    std::string command = argv[1];
+    if (command == "dump") {
+      ok = leveldb::HandleDumpCommand(env, argv+2, argc-2);
+    } else {
+      Usage();
+      ok = false;
+    }
+  }
+  return (ok ? 0 : 1);
+}

db/log_reader.cc

@@ -133,7 +133,9 @@ bool Reader::ReadRecord(Slice* record, std::string* scratch) {
 
       case kEof:
         if (in_fragmented_record) {
-          ReportCorruption(scratch->size(), "partial record without end(3)");
+          // This can be caused by the writer dying immediately after
+          // writing a physical record but before completing the next; don't
+          // treat it as a corruption, just ignore the entire logical record.
           scratch->clear();
         }
         return false;
@@ -193,13 +195,12 @@ unsigned int Reader::ReadPhysicalRecord(Slice* result) {
           eof_ = true;
         }
         continue;
-      } else if (buffer_.size() == 0) {
-        // End of file
-        return kEof;
       } else {
-        size_t drop_size = buffer_.size();
+        // Note that if buffer_ is non-empty, we have a truncated header at the
+        // end of the file, which can be caused by the writer crashing in the
+        // middle of writing the header. Instead of considering this an error,
+        // just report EOF.
         buffer_.clear();
-        ReportCorruption(drop_size, "truncated record at end of file");
         return kEof;
       }
     }
@@ -213,8 +214,14 @@ unsigned int Reader::ReadPhysicalRecord(Slice* result) {
     if (kHeaderSize + length > buffer_.size()) {
       size_t drop_size = buffer_.size();
       buffer_.clear();
-      ReportCorruption(drop_size, "bad record length");
-      return kBadRecord;
+      if (!eof_) {
+        ReportCorruption(drop_size, "bad record length");
+        return kBadRecord;
+      }
+      // If the end of the file has been reached without reading |length| bytes
+      // of payload, assume the writer died in the middle of writing the record.
+      // Don't report a corruption.
+      return kEof;
     }
 
     if (type == kZeroType && length == 0) {

db/log_test.cc

@@ -351,20 +351,32 @@ TEST(LogTest, BadRecordType) {
   ASSERT_EQ("OK", MatchError("unknown record type"));
 }
 
-TEST(LogTest, TruncatedTrailingRecord) {
+TEST(LogTest, TruncatedTrailingRecordIsIgnored) {
   Write("foo");
   ShrinkSize(4);   // Drop all payload as well as a header byte
   ASSERT_EQ("EOF", Read());
-  ASSERT_EQ(kHeaderSize - 1, DroppedBytes());
-  ASSERT_EQ("OK", MatchError("truncated record at end of file"));
+  // Truncated last record is ignored, not treated as an error.
+  ASSERT_EQ(0, DroppedBytes());
+  ASSERT_EQ("", ReportMessage());
 }
 
 TEST(LogTest, BadLength) {
+  const int kPayloadSize = kBlockSize - kHeaderSize;
+  Write(BigString("bar", kPayloadSize));
+  Write("foo");
+  // Least significant size byte is stored in header[4].
+  IncrementByte(4, 1);
+  ASSERT_EQ("foo", Read());
+  ASSERT_EQ(kBlockSize, DroppedBytes());
+  ASSERT_EQ("OK", MatchError("bad record length"));
+}
+
+TEST(LogTest, BadLengthAtEndIsIgnored) {
   Write("foo");
   ShrinkSize(1);
   ASSERT_EQ("EOF", Read());
-  ASSERT_EQ(kHeaderSize + 2, DroppedBytes());
-  ASSERT_EQ("OK", MatchError("bad record length"));
+  ASSERT_EQ(0, DroppedBytes());
+  ASSERT_EQ("", ReportMessage());
 }
 
 TEST(LogTest, ChecksumMismatch) {
@@ -415,6 +427,24 @@ TEST(LogTest, UnexpectedFirstType) {
   ASSERT_EQ("OK", MatchError("partial record without end"));
 }
 
+TEST(LogTest, MissingLastIsIgnored) {
+  Write(BigString("bar", kBlockSize));
+  // Remove the LAST block, including header.
+  ShrinkSize(14);
+  ASSERT_EQ("EOF", Read());
+  ASSERT_EQ("", ReportMessage());
+  ASSERT_EQ(0, DroppedBytes());
+}
+
+TEST(LogTest, PartialLastIsIgnored) {
+  Write(BigString("bar", kBlockSize));
+  // Cause a bad record length in the LAST block.
+  ShrinkSize(1);
+  ASSERT_EQ("EOF", Read());
+  ASSERT_EQ("", ReportMessage());
+  ASSERT_EQ(0, DroppedBytes());
+}
+
 TEST(LogTest, ErrorJoinsRecords) {
   // Consider two fragmented records:
   //    first(R1) last(R1) first(R2) last(R2)

db/repair.cc

@@ -48,7 +48,8 @@ class Repairer {
       : dbname_(dbname),
         env_(options.env),
         icmp_(options.comparator),
-        options_(SanitizeOptions(dbname, &icmp_, options)),
+        ipolicy_(options.filter_policy),
+        options_(SanitizeOptions(dbname, &icmp_, &ipolicy_, options)),
         owns_info_log_(options_.info_log != options.info_log),
         owns_cache_(options_.block_cache != options.block_cache),
         next_file_number_(1) {
@@ -99,6 +100,7 @@ class Repairer {
   std::string const dbname_;
   Env* const env_;
   InternalKeyComparator const icmp_;
+  InternalFilterPolicy const ipolicy_;
   Options const options_;
   bool owns_info_log_;
   bool owns_cache_;
@@ -240,62 +242,134 @@ class Repairer {
   }
 
   void ExtractMetaData() {
-    std::vector<TableInfo> kept;
     for (size_t i = 0; i < table_numbers_.size(); i++) {
-      TableInfo t;
-      t.meta.number = table_numbers_[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],
-            status.ToString().c_str());
-        ArchiveFile(fname);
-      } else {
-        tables_.push_back(t);
-      }
+      ScanTable(table_numbers_[i]);
     }
   }
 
-  Status ScanTable(TableInfo* t) {
-    std::string fname = TableFileName(dbname_, t->meta.number);
-    int counter = 0;
-    Status status = env_->GetFileSize(fname, &t->meta.file_size);
-    if (status.ok()) {
-      Iterator* iter = table_cache_->NewIterator(
-          ReadOptions(), t->meta.number, t->meta.file_size);
-      bool empty = true;
-      ParsedInternalKey parsed;
-      t->max_sequence = 0;
-      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());
-          continue;
-        }
+  Iterator* NewTableIterator(const FileMetaData& meta) {
+    // Same as compaction iterators: if paranoid_checks are on, turn
+    // on checksum verification.
+    ReadOptions r;
+    r.verify_checksums = options_.paranoid_checks;
+    return table_cache_->NewIterator(r, meta.number, meta.file_size);
+  }
 
-        counter++;
-        if (empty) {
-          empty = false;
-          t->meta.smallest.DecodeFrom(key);
-        }
-        t->meta.largest.DecodeFrom(key);
-        if (parsed.sequence > t->max_sequence) {
-          t->max_sequence = parsed.sequence;
-        }
+  void ScanTable(uint64_t number) {
+    TableInfo t;
+    t.meta.number = number;
+    std::string fname = TableFileName(dbname_, number);
+    Status status = env_->GetFileSize(fname, &t.meta.file_size);
+    if (!status.ok()) {
+      // Try alternate file name.
+      fname = SSTTableFileName(dbname_, number);
+      Status s2 = env_->GetFileSize(fname, &t.meta.file_size);
+      if (s2.ok()) {
+        status = Status::OK();
       }
-      if (!iter->status().ok()) {
-        status = iter->status();
-      }
-      delete iter;
     }
+    if (!status.ok()) {
+      ArchiveFile(TableFileName(dbname_, number));
+      ArchiveFile(SSTTableFileName(dbname_, number));
+      Log(options_.info_log, "Table #%llu: dropped: %s",
+          (unsigned long long) t.meta.number,
+          status.ToString().c_str());
+      return;
+    }
+
+    // Extract metadata by scanning through table.
+    int counter = 0;
+    Iterator* iter = NewTableIterator(t.meta);
+    bool empty = true;
+    ParsedInternalKey parsed;
+    t.max_sequence = 0;
+    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());
+        continue;
+      }
+
+      counter++;
+      if (empty) {
+        empty = false;
+        t.meta.smallest.DecodeFrom(key);
+      }
+      t.meta.largest.DecodeFrom(key);
+      if (parsed.sequence > t.max_sequence) {
+        t.max_sequence = parsed.sequence;
+      }
+    }
+    if (!iter->status().ok()) {
+      status = iter->status();
+    }
+    delete iter;
     Log(options_.info_log, "Table #%llu: %d entries %s",
-        (unsigned long long) t->meta.number,
+        (unsigned long long) t.meta.number,
         counter,
         status.ToString().c_str());
-    return status;
+
+    if (status.ok()) {
+      tables_.push_back(t);
+    } else {
+      RepairTable(fname, t);  // RepairTable archives input file.
+    }
+  }
+
+  void RepairTable(const std::string& src, TableInfo t) {
+    // We will copy src contents to a new table and then rename the
+    // new table over the source.
+
+    // Create builder.
+    std::string copy = TableFileName(dbname_, next_file_number_++);
+    WritableFile* file;
+    Status s = env_->NewWritableFile(copy, &file);
+    if (!s.ok()) {
+      return;
+    }
+    TableBuilder* builder = new TableBuilder(options_, file);
+
+    // Copy data.
+    Iterator* iter = NewTableIterator(t.meta);
+    int counter = 0;
+    for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
+      builder->Add(iter->key(), iter->value());
+      counter++;
+    }
+    delete iter;
+
+    ArchiveFile(src);
+    if (counter == 0) {
+      builder->Abandon();  // Nothing to save
+    } else {
+      s = builder->Finish();
+      if (s.ok()) {
+        t.meta.file_size = builder->FileSize();
+      }
+    }
+    delete builder;
+    builder = NULL;
+
+    if (s.ok()) {
+      s = file->Close();
+    }
+    delete file;
+    file = NULL;
+
+    if (counter > 0 && s.ok()) {
+      std::string orig = TableFileName(dbname_, t.meta.number);
+      s = env_->RenameFile(copy, orig);
+      if (s.ok()) {
+        Log(options_.info_log, "Table #%llu: %d entries repaired",
+            (unsigned long long) t.meta.number, counter);
+        tables_.push_back(t);
+      }
+    }
+    if (!s.ok()) {
+      env_->DeleteFile(copy);
+    }
   }
 
   Status WriteDescriptor() {

db/table_cache.cc

@@ -42,23 +42,24 @@ TableCache::~TableCache() {
   delete cache_;
 }
 
-Iterator* TableCache::NewIterator(const ReadOptions& options,
-                                  uint64_t file_number,
-                                  uint64_t file_size,
-                                  Table** tableptr) {
-  if (tableptr != NULL) {
-    *tableptr = NULL;
-  }
-
+Status TableCache::FindTable(uint64_t file_number, uint64_t file_size,
+                             Cache::Handle** handle) {
+  Status s;
   char buf[sizeof(file_number)];
   EncodeFixed64(buf, file_number);
   Slice key(buf, sizeof(buf));
-  Cache::Handle* handle = cache_->Lookup(key);
-  if (handle == NULL) {
+  *handle = cache_->Lookup(key);
+  if (*handle == NULL) {
     std::string fname = TableFileName(dbname_, file_number);
     RandomAccessFile* file = NULL;
     Table* table = NULL;
-    Status s = env_->NewRandomAccessFile(fname, &file);
+    s = env_->NewRandomAccessFile(fname, &file);
+    if (!s.ok()) {
+      std::string old_fname = SSTTableFileName(dbname_, file_number);
+      if (env_->NewRandomAccessFile(old_fname, &file).ok()) {
+        s = Status::OK();
+      }
+    }
     if (s.ok()) {
       s = Table::Open(*options_, file, file_size, &table);
     }
@@ -68,13 +69,28 @@ Iterator* TableCache::NewIterator(const ReadOptions& options,
       delete file;
       // We do not cache error results so that if the error is transient,
       // or somebody repairs the file, we recover automatically.
-      return NewErrorIterator(s);
+    } else {
+      TableAndFile* tf = new TableAndFile;
+      tf->file = file;
+      tf->table = table;
+      *handle = cache_->Insert(key, tf, 1, &DeleteEntry);
     }
+  }
+  return s;
+}
 
-    TableAndFile* tf = new TableAndFile;
-    tf->file = file;
-    tf->table = table;
-    handle = cache_->Insert(key, tf, 1, &DeleteEntry);
+Iterator* TableCache::NewIterator(const ReadOptions& options,
+                                  uint64_t file_number,
+                                  uint64_t file_size,
+                                  Table** tableptr) {
+  if (tableptr != NULL) {
+    *tableptr = NULL;
+  }
+
+  Cache::Handle* handle = NULL;
+  Status s = FindTable(file_number, file_size, &handle);
+  if (!s.ok()) {
+    return NewErrorIterator(s);
   }
 
   Table* table = reinterpret_cast<TableAndFile*>(cache_->Value(handle))->table;
@@ -86,6 +102,22 @@ Iterator* TableCache::NewIterator(const ReadOptions& options,
   return result;
 }
 
+Status TableCache::Get(const ReadOptions& options,
+                       uint64_t file_number,
+                       uint64_t file_size,
+                       const Slice& k,
+                       void* arg,
+                       void (*saver)(void*, const Slice&, const Slice&)) {
+  Cache::Handle* handle = NULL;
+  Status s = FindTable(file_number, file_size, &handle);
+  if (s.ok()) {
+    Table* t = reinterpret_cast<TableAndFile*>(cache_->Value(handle))->table;
+    s = t->InternalGet(options, k, arg, saver);
+    cache_->Release(handle);
+  }
+  return s;
+}
+
 void TableCache::Evict(uint64_t file_number) {
   char buf[sizeof(file_number)];
   EncodeFixed64(buf, file_number);

db/table_cache.h

@@ -35,6 +35,15 @@ class TableCache {
                         uint64_t file_size,
                         Table** tableptr = NULL);
 
+  // If a seek to internal key "k" in specified file finds an entry,
+  // call (*handle_result)(arg, found_key, found_value).
+  Status Get(const ReadOptions& options,
+             uint64_t file_number,
+             uint64_t file_size,
+             const Slice& k,
+             void* arg,
+             void (*handle_result)(void*, const Slice&, const Slice&));
+
   // Evict any entry for the specified file number
   void Evict(uint64_t file_number);
 
@@ -43,6 +52,8 @@ class TableCache {
   const std::string dbname_;
   const Options* options_;
   Cache* cache_;
+
+  Status FindTable(uint64_t file_number, uint64_t file_size, Cache::Handle**);
 };
 
 }  // namespace leveldb

db/version_set.cc

@@ -54,20 +54,6 @@ static int64_t TotalFileSize(const std::vector<FileMetaData*>& files) {
   return sum;
 }
 
-namespace {
-std::string IntSetToString(const std::set<uint64_t>& s) {
-  std::string result = "{";
-  for (std::set<uint64_t>::const_iterator it = s.begin();
-       it != s.end();
-       ++it) {
-    result += (result.size() > 1) ? "," : "";
-    result += NumberToString(*it);
-  }
-  result += "}";
-  return result;
-}
-}  // namespace
-
 Version::~Version() {
   assert(refs_ == 0);
 
@@ -132,7 +118,7 @@ bool SomeFileOverlapsRange(
   const Comparator* ucmp = icmp.user_comparator();
   if (!disjoint_sorted_files) {
     // Need to check against all files
-    for (int i = 0; i < files.size(); i++) {
+    for (size_t i = 0; i < files.size(); i++) {
       const FileMetaData* f = files[i];
       if (AfterFile(ucmp, smallest_user_key, f) ||
           BeforeFile(ucmp, largest_user_key, f)) {
@@ -255,41 +241,85 @@ void Version::AddIterators(const ReadOptions& options,
   }
 }
 
-// If "*iter" points at a value or deletion for user_key, store
-// either the value, or a NotFound error and return true.
-// Else return false.
-static bool GetValue(const Comparator* cmp,
-                     Iterator* iter, const Slice& user_key,
-                     std::string* value,
-                     Status* s) {
-  if (!iter->Valid()) {
-    return false;
-  }
+// Callback from TableCache::Get()
+namespace {
+enum SaverState {
+  kNotFound,
+  kFound,
+  kDeleted,
+  kCorrupt,
+};
+struct Saver {
+  SaverState state;
+  const Comparator* ucmp;
+  Slice user_key;
+  std::string* value;
+};
+}
+static void SaveValue(void* arg, const Slice& ikey, const Slice& v) {
+  Saver* s = reinterpret_cast<Saver*>(arg);
   ParsedInternalKey parsed_key;
-  if (!ParseInternalKey(iter->key(), &parsed_key)) {
-    *s = Status::Corruption("corrupted key for ", user_key);
-    return true;
-  }
-  if (cmp->Compare(parsed_key.user_key, user_key) != 0) {
-    return false;
-  }
-  switch (parsed_key.type) {
-    case kTypeDeletion:
-      *s = Status::NotFound(Slice());  // Use an empty error message for speed
-      break;
-    case kTypeValue: {
-      Slice v = iter->value();
-      value->assign(v.data(), v.size());
-      break;
+  if (!ParseInternalKey(ikey, &parsed_key)) {
+    s->state = kCorrupt;
+  } else {
+    if (s->ucmp->Compare(parsed_key.user_key, s->user_key) == 0) {
+      s->state = (parsed_key.type == kTypeValue) ? kFound : kDeleted;
+      if (s->state == kFound) {
+        s->value->assign(v.data(), v.size());
+      }
     }
   }
-  return true;
 }
 
 static bool NewestFirst(FileMetaData* a, FileMetaData* b) {
   return a->number > b->number;
 }
 
+void Version::ForEachOverlapping(Slice user_key, Slice internal_key,
+                                 void* arg,
+                                 bool (*func)(void*, int, FileMetaData*)) {
+  // TODO(sanjay): Change Version::Get() to use this function.
+  const Comparator* ucmp = vset_->icmp_.user_comparator();
+
+  // Search level-0 in order from newest to oldest.
+  std::vector<FileMetaData*> tmp;
+  tmp.reserve(files_[0].size());
+  for (uint32_t i = 0; i < files_[0].size(); i++) {
+    FileMetaData* f = files_[0][i];
+    if (ucmp->Compare(user_key, f->smallest.user_key()) >= 0 &&
+        ucmp->Compare(user_key, f->largest.user_key()) <= 0) {
+      tmp.push_back(f);
+    }
+  }
+  if (!tmp.empty()) {
+    std::sort(tmp.begin(), tmp.end(), NewestFirst);
+    for (uint32_t i = 0; i < tmp.size(); i++) {
+      if (!(*func)(arg, 0, tmp[i])) {
+        return;
+      }
+    }
+  }
+
+  // Search other levels.
+  for (int level = 1; level < config::kNumLevels; level++) {
+    size_t num_files = files_[level].size();
+    if (num_files == 0) continue;
+
+    // Binary search to find earliest index whose largest key >= internal_key.
+    uint32_t index = FindFile(vset_->icmp_, files_[level], internal_key);
+    if (index < num_files) {
+      FileMetaData* f = files_[level][index];
+      if (ucmp->Compare(user_key, f->smallest.user_key()) < 0) {
+        // All of "f" is past any data for user_key
+      } else {
+        if (!(*func)(arg, level, f)) {
+          return;
+        }
+      }
+    }
+  }
+}
+
 Status Version::Get(const ReadOptions& options,
                     const LookupKey& k,
                     std::string* value,
@@ -361,21 +391,27 @@ Status Version::Get(const ReadOptions& options,
       last_file_read = f;
       last_file_read_level = level;
 
-      Iterator* iter = vset_->table_cache_->NewIterator(
-          options,
-          f->number,
-          f->file_size);
-      iter->Seek(ikey);
-      const bool done = GetValue(ucmp, iter, user_key, value, &s);
-      if (!iter->status().ok()) {
-        s = iter->status();
-        delete iter;
+      Saver saver;
+      saver.state = kNotFound;
+      saver.ucmp = ucmp;
+      saver.user_key = user_key;
+      saver.value = value;
+      s = vset_->table_cache_->Get(options, f->number, f->file_size,
+                                   ikey, &saver, SaveValue);
+      if (!s.ok()) {
         return s;
-      } else {
-        delete iter;
-        if (done) {
+      }
+      switch (saver.state) {
+        case kNotFound:
+          break;      // Keep searching in other files
+        case kFound:
+          return s;
+        case kDeleted:
+          s = Status::NotFound(Slice());  // Use empty error message for speed
+          return s;
+        case kCorrupt:
+          s = Status::Corruption("corrupted key for ", user_key);
           return s;
-        }
       }
     }
   }
@@ -396,6 +432,44 @@ bool Version::UpdateStats(const GetStats& stats) {
   return false;
 }
 
+bool Version::RecordReadSample(Slice internal_key) {
+  ParsedInternalKey ikey;
+  if (!ParseInternalKey(internal_key, &ikey)) {
+    return false;
+  }
+
+  struct State {
+    GetStats stats;  // Holds first matching file
+    int matches;
+
+    static bool Match(void* arg, int level, FileMetaData* f) {
+      State* state = reinterpret_cast<State*>(arg);
+      state->matches++;
+      if (state->matches == 1) {
+        // Remember first match.
+        state->stats.seek_file = f;
+        state->stats.seek_file_level = level;
+      }
+      // We can stop iterating once we have a second match.
+      return state->matches < 2;
+    }
+  };
+
+  State state;
+  state.matches = 0;
+  ForEachOverlapping(ikey.user_key, internal_key, &state, &State::Match);
+
+  // Must have at least two matches since we want to merge across
+  // files. But what if we have a single file that contains many
+  // overwrites and deletions?  Should we have another mechanism for
+  // finding such files?
+  if (state.matches >= 2) {
+    // 1MB cost is about 1 seek (see comment in Builder::Apply).
+    return UpdateStats(state.stats);
+  }
+  return false;
+}
+
 void Version::Ref() {
   ++refs_;
 }
@@ -430,10 +504,13 @@ int Version::PickLevelForMemTableOutput(
       if (OverlapInLevel(level + 1, &smallest_user_key, &largest_user_key)) {
         break;
       }
-      GetOverlappingInputs(level + 2, &start, &limit, &overlaps);
-      const int64_t sum = TotalFileSize(overlaps);
-      if (sum > kMaxGrandParentOverlapBytes) {
-        break;
+      if (level + 2 < config::kNumLevels) {
+        // Check that file does not overlap too many grandparent bytes.
+        GetOverlappingInputs(level + 2, &start, &limit, &overlaps);
+        const int64_t sum = TotalFileSize(overlaps);
+        if (sum > kMaxGrandParentOverlapBytes) {
+          break;
+        }
       }
       level++;
     }
@@ -447,6 +524,8 @@ void Version::GetOverlappingInputs(
     const InternalKey* begin,
     const InternalKey* end,
     std::vector<FileMetaData*>* inputs) {
+  assert(level >= 0);
+  assert(level < config::kNumLevels);
   inputs->clear();
   Slice user_begin, user_end;
   if (begin != NULL) {
@@ -781,6 +860,9 @@ Status VersionSet::LogAndApply(VersionEdit* edit, port::Mutex* mu) {
       if (s.ok()) {
         s = descriptor_file_->Sync();
       }
+      if (!s.ok()) {
+        Log(options_->info_log, "MANIFEST write: %s\n", s.ToString().c_str());
+      }
     }
 
     // If we just created a new descriptor file, install it by writing a
@@ -860,7 +942,7 @@ Status VersionSet::Recover() {
         if (edit.has_comparator_ &&
             edit.comparator_ != icmp_.user_comparator()->Name()) {
           s = Status::InvalidArgument(
-              edit.comparator_ + "does not match existing comparator ",
+              edit.comparator_ + " does not match existing comparator ",
               icmp_.user_comparator()->Name());
         }
       }
@@ -1290,14 +1372,19 @@ Compaction* VersionSet::CompactRange(
   }
 
   // Avoid compacting too much in one shot in case the range is large.
-  const uint64_t limit = MaxFileSizeForLevel(level);
-  uint64_t total = 0;
-  for (int i = 0; i < inputs.size(); i++) {
-    uint64_t s = inputs[i]->file_size;
-    total += s;
-    if (total >= limit) {
-      inputs.resize(i + 1);
-      break;
+  // But we cannot do this for level-0 since level-0 files can overlap
+  // and we must not pick one file and drop another older file if the
+  // two files overlap.
+  if (level > 0) {
+    const uint64_t limit = MaxFileSizeForLevel(level);
+    uint64_t total = 0;
+    for (size_t i = 0; i < inputs.size(); i++) {
+      uint64_t s = inputs[i]->file_size;
+      total += s;
+      if (total >= limit) {
+        inputs.resize(i + 1);
+        break;
+      }
     }
   }
 

db/version_set.h

@@ -21,6 +21,7 @@
 #include "db/dbformat.h"
 #include "db/version_edit.h"
 #include "port/port.h"
+#include "port/thread_annotations.h"
 
 namespace leveldb {
 
@@ -77,6 +78,12 @@ class Version {
   // REQUIRES: lock is held
   bool UpdateStats(const GetStats& stats);
 
+  // Record a sample of bytes read at the specified internal key.
+  // Samples are taken approximately once every config::kReadBytesPeriod
+  // bytes.  Returns true if a new compaction may need to be triggered.
+  // REQUIRES: lock is held
+  bool RecordReadSample(Slice key);
+
   // Reference count management (so Versions do not disappear out from
   // under live iterators)
   void Ref();
@@ -113,6 +120,15 @@ class Version {
   class LevelFileNumIterator;
   Iterator* NewConcatenatingIterator(const ReadOptions&, int level) const;
 
+  // Call func(arg, level, f) for every file that overlaps user_key in
+  // order from newest to oldest.  If an invocation of func returns
+  // false, makes no more calls.
+  //
+  // REQUIRES: user portion of internal_key == user_key.
+  void ForEachOverlapping(Slice user_key, Slice internal_key,
+                          void* arg,
+                          bool (*func)(void*, int, FileMetaData*));
+
   VersionSet* vset_;            // VersionSet to which this Version belongs
   Version* next_;               // Next version in linked list
   Version* prev_;               // Previous version in linked list
@@ -159,7 +175,8 @@ class VersionSet {
   // current version.  Will release *mu while actually writing to the file.
   // REQUIRES: *mu is held on entry.
   // REQUIRES: no other thread concurrently calls LogAndApply()
-  Status LogAndApply(VersionEdit* edit, port::Mutex* mu);
+  Status LogAndApply(VersionEdit* edit, port::Mutex* mu)
+      EXCLUSIVE_LOCKS_REQUIRED(mu);
 
   // Recover the last saved descriptor from persistent storage.
   Status Recover();
@@ -173,6 +190,15 @@ class VersionSet {
   // Allocate and return a new file number
   uint64_t NewFileNumber() { return next_file_number_++; }
 
+  // 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) {
+    if (next_file_number_ == file_number + 1) {
+      next_file_number_ = file_number;
+    }
+  }
+
   // Return the number of Table files at the specified level.
   int NumLevelFiles(int level) const;
 

doc/bench/db_bench_sqlite3.cc

@@ -75,6 +75,9 @@ static bool FLAGS_transaction = true;
 // If true, we enable Write-Ahead Logging
 static bool FLAGS_WAL_enabled = true;
 
+// Use the db with the following name.
+static const char* FLAGS_db = NULL;
+
 inline
 static void ExecErrorCheck(int status, char *err_msg) {
   if (status != SQLITE_OK) {
@@ -317,11 +320,16 @@ class Benchmark {
     bytes_(0),
     rand_(301) {
     std::vector<std::string> files;
-    Env::Default()->GetChildren("/tmp", &files);
+    std::string test_dir;
+    Env::Default()->GetTestDirectory(&test_dir);
+    Env::Default()->GetChildren(test_dir, &files);
     if (!FLAGS_use_existing_db) {
       for (int i = 0; i < files.size(); i++) {
         if (Slice(files[i]).starts_with("dbbench_sqlite3")) {
-          Env::Default()->DeleteFile("/tmp/" + files[i]);
+          std::string file_name(test_dir);
+          file_name += "/";
+          file_name += files[i];
+          Env::Default()->DeleteFile(file_name.c_str());
         }
       }
     }
@@ -415,7 +423,11 @@ class Benchmark {
     db_num_++;
 
     // Open database
-    snprintf(file_name, sizeof(file_name), "/tmp/dbbench_sqlite3-%d.db",
+    std::string tmp_dir;
+    Env::Default()->GetTestDirectory(&tmp_dir);
+    snprintf(file_name, sizeof(file_name),
+             "%s/dbbench_sqlite3-%d.db",
+             tmp_dir.c_str(),
              db_num_);
     status = sqlite3_open(file_name, &db_);
     if (status) {
@@ -606,7 +618,7 @@ class Benchmark {
         ErrorCheck(status);
 
         // Execute read statement
-        while ((status = sqlite3_step(read_stmt)) == SQLITE_ROW);
+        while ((status = sqlite3_step(read_stmt)) == SQLITE_ROW) {}
         StepErrorCheck(status);
 
         // Reset SQLite statement for another use
@@ -655,6 +667,7 @@ class Benchmark {
 }  // namespace leveldb
 
 int main(int argc, char** argv) {
+  std::string default_db_path;
   for (int i = 1; i < argc; i++) {
     double d;
     int n;
@@ -684,12 +697,21 @@ int main(int argc, char** argv) {
     } else if (sscanf(argv[i], "--WAL_enabled=%d%c", &n, &junk) == 1 &&
                (n == 0 || n == 1)) {
       FLAGS_WAL_enabled = n;
+    } else if (strncmp(argv[i], "--db=", 5) == 0) {
+      FLAGS_db = argv[i] + 5;
     } else {
       fprintf(stderr, "Invalid flag '%s'\n", argv[i]);
       exit(1);
     }
   }
 
+  // Choose a location for the test database if none given with --db=<path>
+  if (FLAGS_db == NULL) {
+      leveldb::Env::Default()->GetTestDirectory(&default_db_path);
+      default_db_path += "/dbbench";
+      FLAGS_db = default_db_path.c_str();
+  }
+
   leveldb::Benchmark benchmark;
   benchmark.Run();
   return 0;

doc/bench/db_bench_tree_db.cc

@@ -68,6 +68,9 @@ static bool FLAGS_use_existing_db = false;
 // is off.
 static bool FLAGS_compression = true;
 
+// Use the db with the following name.
+static const char* FLAGS_db = NULL;
+
 inline
 static void DBSynchronize(kyotocabinet::TreeDB* db_)
 {
@@ -292,11 +295,16 @@ class Benchmark {
     bytes_(0),
     rand_(301) {
     std::vector<std::string> files;
-    Env::Default()->GetChildren("/tmp", &files);
+    std::string test_dir;
+    Env::Default()->GetTestDirectory(&test_dir);
+    Env::Default()->GetChildren(test_dir.c_str(), &files);
     if (!FLAGS_use_existing_db) {
       for (int i = 0; i < files.size(); i++) {
         if (Slice(files[i]).starts_with("dbbench_polyDB")) {
-          Env::Default()->DeleteFile("/tmp/" + files[i]);
+          std::string file_name(test_dir);
+          file_name += "/";
+          file_name += files[i];
+          Env::Default()->DeleteFile(file_name.c_str());
         }
       }
     }
@@ -385,8 +393,12 @@ class Benchmark {
     db_ = new kyotocabinet::TreeDB();
     char file_name[100];
     db_num_++;
-    snprintf(file_name, sizeof(file_name), "/tmp/dbbench_polyDB-%d.kct",
-                               db_num_);
+    std::string test_dir;
+    Env::Default()->GetTestDirectory(&test_dir);
+    snprintf(file_name, sizeof(file_name),
+             "%s/dbbench_polyDB-%d.kct",
+             test_dir.c_str(),
+             db_num_);
 
     // Create tuning options and open the database
     int open_options = kyotocabinet::PolyDB::OWRITER |
@@ -470,6 +482,7 @@ class Benchmark {
 }  // namespace leveldb
 
 int main(int argc, char** argv) {
+  std::string default_db_path;
   for (int i = 1; i < argc; i++) {
     double d;
     int n;
@@ -494,12 +507,21 @@ int main(int argc, char** argv) {
     } else if (sscanf(argv[i], "--compression=%d%c", &n, &junk) == 1 &&
                (n == 0 || n == 1)) {
       FLAGS_compression = (n == 1) ? true : false;
+    } else if (strncmp(argv[i], "--db=", 5) == 0) {
+      FLAGS_db = argv[i] + 5;
     } else {
       fprintf(stderr, "Invalid flag '%s'\n", argv[i]);
       exit(1);
     }
   }
 
+  // Choose a location for the test database if none given with --db=<path>
+  if (FLAGS_db == NULL) {
+      leveldb::Env::Default()->GetTestDirectory(&default_db_path);
+      default_db_path += "/dbbench";
+      FLAGS_db = default_db_path.c_str();
+  }
+
   leveldb::Benchmark benchmark;
   benchmark.Run();
   return 0;

doc/impl.html

@@ -11,7 +11,7 @@
 
 The implementation of leveldb is similar in spirit to the
 representation of a single
-<a href="http://labs.google.com/papers/bigtable.html">
+<a href="http://research.google.com/archive/bigtable.html">
 Bigtable tablet (section 5.3)</a>.
 However the organization of the files that make up the representation
 is somewhat different and is explained below.

doc/index.html

@@ -400,6 +400,69 @@ We might want to prefix <code>filename</code> keys with one letter (say '/') and
 over just the metadata do not force us to fetch and cache bulky file
 contents.
 <p>
+<h2>Filters</h2>
+<p>
+Because of the way <code>leveldb</code> data is organized on disk,
+a single <code>Get()</code> call may involve multiple reads from disk.
+The optional <code>FilterPolicy</code> mechanism can be used to reduce
+the number of disk reads substantially.
+<pre>
+   leveldb::Options options;
+   options.filter_policy = NewBloomFilterPolicy(10);
+   leveldb::DB* db;
+   leveldb::DB::Open(options, "/tmp/testdb", &amp;db);
+   ... use the database ...
+   delete db;
+   delete options.filter_policy;
+</pre>
+The preceding code associates a
+<a href="http://en.wikipedia.org/wiki/Bloom_filter">Bloom filter</a>
+based filtering policy with the database.  Bloom filter based
+filtering relies on keeping some number of bits of data in memory per
+key (in this case 10 bits per key since that is the argument we passed
+to NewBloomFilterPolicy).  This filter will reduce the number of unnecessary
+disk reads needed for <code>Get()</code> calls by a factor of
+approximately a 100.  Increasing the bits per key will lead to a
+larger reduction at the cost of more memory usage.  We recommend that
+applications whose working set does not fit in memory and that do a
+lot of random reads set a filter policy.
+<p>
+If you are using a custom comparator, you should ensure that the filter
+policy you are using is compatible with your comparator.  For example,
+consider a comparator that ignores trailing spaces when comparing keys.
+<code>NewBloomFilterPolicy</code> must not be used with such a comparator.
+Instead, the application should provide a custom filter policy that
+also ignores trailing spaces.  For example:
+<pre>
+  class CustomFilterPolicy : public leveldb::FilterPolicy {
+   private:
+    FilterPolicy* builtin_policy_;
+   public:
+    CustomFilterPolicy() : builtin_policy_(NewBloomFilterPolicy(10)) { }
+    ~CustomFilterPolicy() { delete builtin_policy_; }
+
+    const char* Name() const { return "IgnoreTrailingSpacesFilter"; }
+
+    void CreateFilter(const Slice* keys, int n, std::string* dst) const {
+      // Use builtin bloom filter code after removing trailing spaces
+      std::vector&lt;Slice&gt; trimmed(n);
+      for (int i = 0; i &lt; n; i++) {
+        trimmed[i] = RemoveTrailingSpaces(keys[i]);
+      }
+      return builtin_policy_-&gt;CreateFilter(&amp;trimmed[i], n, dst);
+    }
+
+    bool KeyMayMatch(const Slice& key, const Slice& filter) const {
+      // Use builtin bloom filter code after removing trailing spaces
+      return builtin_policy_-&gt;KeyMayMatch(RemoveTrailingSpaces(key), filter);
+    }
+  };
+</pre>
+<p>
+Advanced applications may provide a filter policy that does not use
+a bloom filter but uses some other mechanism for summarizing a set
+of keys.  See <code>leveldb/filter_policy.h</code> for detail.
+<p>
 <h1>Checksums</h1>
 <p>
 <code>leveldb</code> associates checksums with all data it stores in the file system.

doc/log_format.txt

@@ -4,8 +4,8 @@ exception is that the tail of the file may contain a partial block.
 Each block consists of a sequence of records:
    block := record* trailer?
    record :=
-	checksum: uint32	// crc32c of type and data[]
-	length: uint16
+	checksum: uint32	// crc32c of type and data[] ; little-endian
+	length: uint16		// little-endian
 	type: uint8		// One of FULL, FIRST, MIDDLE, LAST
 	data: uint8[length]
 

doc/table_format.txt

@@ -18,6 +18,8 @@ The file contains internal pointers.  Each such pointer is called
 a BlockHandle and contains the following information:
   offset:	    varint64
   size:		    varint64
+See https://developers.google.com/protocol-buffers/docs/encoding#varints
+for an explanation of varint64 format.
 
 (1) The sequence of key/value pairs in the file are stored in sorted
 order and partitioned into a sequence of data blocks.  These blocks
@@ -41,11 +43,52 @@ BlockHandle for the data block.
 
 (6) At the very end of the file is a fixed length footer that contains
 the BlockHandle of the metaindex and index blocks as well as a magic number.
-       metaindex_handle:       char[p];    // Block handle for metaindex
-       index_handle:	       char[q];    // Block handle for index
-       padding:		       char[40-p-q]; // 0 bytes to make fixed length
-       			 	       // (40==2*BlockHandle::kMaxEncodedLength)
-       magic:		       fixed64;    // == 0xdb4775248b80fb57
+       metaindex_handle: char[p];    // Block handle for metaindex
+       index_handle:     char[q];    // Block handle for index
+       padding:          char[40-p-q]; // zeroed bytes to make fixed length
+                                       // (40==2*BlockHandle::kMaxEncodedLength)
+       magic:            fixed64;    // == 0xdb4775248b80fb57 (little-endian)
+
+"filter" Meta Block
+-------------------
+
+If a "FilterPolicy" was specified when the database was opened, a
+filter block is stored in each table.  The "metaindex" block contains
+an entry that maps from "filter.<N>" to the BlockHandle for the filter
+block where "<N>" is the string returned by the filter policy's
+"Name()" method.
+
+The filter block stores a sequence of filters, where filter i contains
+the output of FilterPolicy::CreateFilter() on all keys that are stored
+in a block whose file offset falls within the range
+
+    [ i*base ... (i+1)*base-1 ]
+
+Currently, "base" is 2KB.  So for example, if blocks X and Y start in
+the range [ 0KB .. 2KB-1 ], all of the keys in X and Y will be
+converted to a filter by calling FilterPolicy::CreateFilter(), and the
+resulting filter will be stored as the first filter in the filter
+block.
+
+The filter block is formatted as follows:
+
+     [filter 0]
+     [filter 1]
+     [filter 2]
+     ...
+     [filter N-1]
+
+     [offset of filter 0]                  : 4 bytes
+     [offset of filter 1]                  : 4 bytes
+     [offset of filter 2]                  : 4 bytes
+     ...
+     [offset of filter N-1]                : 4 bytes
+
+     [offset of beginning of offset array] : 4 bytes
+     lg(base)                              : 1 byte
+
+The offset array at the end of the filter block allows efficient
+mapping from a data block offset to the corresponding filter.
 
 "stats" Meta Block
 ------------------

helpers/memenv/memenv.cc

@@ -221,6 +221,11 @@ class WritableFileImpl : public WritableFile {
   FileState* file_;
 };
 
+class NoOpLogger : public Logger {
+ public:
+  virtual void Logv(const char* format, va_list ap) { }
+};
+
 class InMemoryEnv : public EnvWrapper {
  public:
   explicit InMemoryEnv(Env* base_env) : EnvWrapper(base_env) { }
@@ -358,6 +363,11 @@ class InMemoryEnv : public EnvWrapper {
     return Status::OK();
   }
 
+  virtual Status NewLogger(const std::string& fname, Logger** result) {
+    *result = new NoOpLogger;
+    return Status::OK();
+  }
+
  private:
   // Map from filenames to FileState objects, representing a simple file system.
   typedef std::map<std::string, FileState*> FileSystem;

include/leveldb/c.h

@@ -9,7 +9,6 @@
   Does not support:
   . getters for the option types
   . custom comparators that implement key shortening
-  . capturing post-write-snapshot
   . custom iter, db, env, cache implementations using just the C bindings
 
   Some conventions:
@@ -28,6 +27,7 @@
   be true on entry:
      *errptr == NULL
      *errptr points to a malloc()ed null-terminated error message
+       (On Windows, *errptr must have been malloc()-ed by this library.)
   On success, a leveldb routine leaves *errptr unchanged.
   On failure, leveldb frees the old value of *errptr and
   set *errptr to a malloc()ed error message.
@@ -55,6 +55,7 @@ typedef struct leveldb_cache_t         leveldb_cache_t;
 typedef struct leveldb_comparator_t    leveldb_comparator_t;
 typedef struct leveldb_env_t           leveldb_env_t;
 typedef struct leveldb_filelock_t      leveldb_filelock_t;
+typedef struct leveldb_filterpolicy_t  leveldb_filterpolicy_t;
 typedef struct leveldb_iterator_t      leveldb_iterator_t;
 typedef struct leveldb_logger_t        leveldb_logger_t;
 typedef struct leveldb_options_t       leveldb_options_t;
@@ -127,6 +128,11 @@ extern void leveldb_approximate_sizes(
     const char* const* range_limit_key, const size_t* range_limit_key_len,
     uint64_t* sizes);
 
+extern void leveldb_compact_range(
+    leveldb_t* db,
+    const char* start_key, size_t start_key_len,
+    const char* limit_key, size_t limit_key_len);
+
 /* Management operations */
 
 extern void leveldb_destroy_db(
@@ -177,6 +183,9 @@ extern void leveldb_options_destroy(leveldb_options_t*);
 extern void leveldb_options_set_comparator(
     leveldb_options_t*,
     leveldb_comparator_t*);
+extern void leveldb_options_set_filter_policy(
+    leveldb_options_t*,
+    leveldb_filterpolicy_t*);
 extern void leveldb_options_set_create_if_missing(
     leveldb_options_t*, unsigned char);
 extern void leveldb_options_set_error_if_exists(
@@ -209,6 +218,26 @@ extern leveldb_comparator_t* leveldb_comparator_create(
     const char* (*name)(void*));
 extern void leveldb_comparator_destroy(leveldb_comparator_t*);
 
+/* Filter policy */
+
+extern leveldb_filterpolicy_t* leveldb_filterpolicy_create(
+    void* state,
+    void (*destructor)(void*),
+    char* (*create_filter)(
+        void*,
+        const char* const* key_array, const size_t* key_length_array,
+        int num_keys,
+        size_t* filter_length),
+    unsigned char (*key_may_match)(
+        void*,
+        const char* key, size_t length,
+        const char* filter, size_t filter_length),
+    const char* (*name)(void*));
+extern void leveldb_filterpolicy_destroy(leveldb_filterpolicy_t*);
+
+extern leveldb_filterpolicy_t* leveldb_filterpolicy_create_bloom(
+    int bits_per_key);
+
 /* Read options */
 
 extern leveldb_readoptions_t* leveldb_readoptions_create();
@@ -239,6 +268,21 @@ extern void leveldb_cache_destroy(leveldb_cache_t* cache);
 extern leveldb_env_t* leveldb_create_default_env();
 extern void leveldb_env_destroy(leveldb_env_t*);
 
+/* Utility */
+
+/* Calls free(ptr).
+   REQUIRES: ptr was malloc()-ed and returned by one of the routines
+   in this file.  Note that in certain cases (typically on Windows), you
+   may need to call this routine instead of free(ptr) to dispose of
+   malloc()-ed memory returned by this library. */
+extern void leveldb_free(void* ptr);
+
+/* Return the major version number for this release. */
+extern int leveldb_major_version();
+
+/* Return the minor version number for this release. */
+extern int leveldb_minor_version();
+
 #ifdef __cplusplus
 }  /* end extern "C" */
 #endif

include/leveldb/db.h

@@ -14,7 +14,7 @@ namespace leveldb {
 
 // Update Makefile if you change these
 static const int kMajorVersion = 1;
-static const int kMinorVersion = 3;
+static const int kMinorVersion = 17;
 
 struct Options;
 struct ReadOptions;

include/leveldb/env.h

@@ -13,9 +13,9 @@
 #ifndef STORAGE_LEVELDB_INCLUDE_ENV_H_
 #define STORAGE_LEVELDB_INCLUDE_ENV_H_
 
-#include <cstdarg>
 #include <string>
 #include <vector>
+#include <stdarg.h>
 #include <stdint.h>
 #include "leveldb/status.h"
 
@@ -175,6 +175,11 @@ class SequentialFile {
   //
   // REQUIRES: External synchronization
   virtual Status Skip(uint64_t n) = 0;
+
+ private:
+  // No copying allowed
+  SequentialFile(const SequentialFile&);
+  void operator=(const SequentialFile&);
 };
 
 // A file abstraction for randomly reading the contents of a file.
@@ -194,6 +199,11 @@ class RandomAccessFile {
   // Safe for concurrent use by multiple threads.
   virtual Status Read(uint64_t offset, size_t n, Slice* result,
                       char* scratch) const = 0;
+
+ private:
+  // No copying allowed
+  RandomAccessFile(const RandomAccessFile&);
+  void operator=(const RandomAccessFile&);
 };
 
 // A file abstraction for sequential writing.  The implementation

include/leveldb/filter_policy.h

@@ -0,0 +1,70 @@
+// Copyright (c) 2012 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.
+//
+// A database can be configured with a custom FilterPolicy object.
+// This object is responsible for creating a small filter from a set
+// of keys.  These filters are stored in leveldb and are consulted
+// automatically by leveldb to decide whether or not to read some
+// information from disk. In many cases, a filter can cut down the
+// number of disk seeks form a handful to a single disk seek per
+// DB::Get() call.
+//
+// Most people will want to use the builtin bloom filter support (see
+// NewBloomFilterPolicy() below).
+
+#ifndef STORAGE_LEVELDB_INCLUDE_FILTER_POLICY_H_
+#define STORAGE_LEVELDB_INCLUDE_FILTER_POLICY_H_
+
+#include <string>
+
+namespace leveldb {
+
+class Slice;
+
+class FilterPolicy {
+ public:
+  virtual ~FilterPolicy();
+
+  // Return the name of this policy.  Note that if the filter encoding
+  // changes in an incompatible way, the name returned by this method
+  // must be changed.  Otherwise, old incompatible filters may be
+  // passed to methods of this type.
+  virtual const char* Name() const = 0;
+
+  // keys[0,n-1] contains a list of keys (potentially with duplicates)
+  // that are ordered according to the user supplied comparator.
+  // Append a filter that summarizes keys[0,n-1] to *dst.
+  //
+  // Warning: do not change the initial contents of *dst.  Instead,
+  // append the newly constructed filter to *dst.
+  virtual void CreateFilter(const Slice* keys, int n, std::string* dst)
+      const = 0;
+
+  // "filter" contains the data appended by a preceding call to
+  // CreateFilter() on this class.  This method must return true if
+  // the key was in the list of keys passed to CreateFilter().
+  // 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;
+};
+
+// 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
+// is 10, which yields a filter with ~ 1% false positive rate.
+//
+// Callers must delete the result after any database that is using the
+// result has been closed.
+//
+// Note: if you are using a custom comparator that ignores some parts
+// of the keys being compared, you must not use NewBloomFilterPolicy()
+// and must provide your own FilterPolicy that also ignores the
+// corresponding parts of the keys.  For example, if the comparator
+// 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);
+
+}
+
+#endif  // STORAGE_LEVELDB_INCLUDE_FILTER_POLICY_H_

include/leveldb/options.h

@@ -12,6 +12,7 @@ namespace leveldb {
 class Cache;
 class Comparator;
 class Env;
+class FilterPolicy;
 class Logger;
 class Snapshot;
 
@@ -127,6 +128,13 @@ struct Options {
   // efficiently detect that and will switch to uncompressed mode.
   CompressionType compression;
 
+  // If non-NULL, use the specified filter policy to reduce disk reads.
+  // Many applications will benefit from passing the result of
+  // NewBloomFilterPolicy() here.
+  //
+  // Default: NULL
+  const FilterPolicy* filter_policy;
+
   // Create an Options object with default values for all fields.
   Options();
 };

include/leveldb/slice.h

@@ -94,7 +94,7 @@ inline bool operator!=(const Slice& x, const Slice& y) {
 }
 
 inline int Slice::compare(const Slice& b) const {
-  const int min_len = (size_ < b.size_) ? size_ : b.size_;
+  const size_t min_len = (size_ < b.size_) ? size_ : b.size_;
   int r = memcmp(data_, b.data_, min_len);
   if (r == 0) {
     if (size_ < b.size_) r = -1;

include/leveldb/table.h

@@ -12,9 +12,11 @@ namespace leveldb {
 
 class Block;
 class BlockHandle;
+class Footer;
 struct Options;
 class RandomAccessFile;
 struct ReadOptions;
+class TableCache;
 
 // A Table is a sorted map from strings to strings.  Tables are
 // immutable and persistent.  A Table may be safely accessed from
@@ -60,6 +62,19 @@ class Table {
   explicit Table(Rep* rep) { rep_ = rep; }
   static Iterator* BlockReader(void*, const ReadOptions&, const Slice&);
 
+  // Calls (*handle_result)(arg, ...) with the entry found after a call
+  // to Seek(key).  May not make such a call if filter policy says
+  // that key is not present.
+  friend class TableCache;
+  Status InternalGet(
+      const ReadOptions&, const Slice& key,
+      void* arg,
+      void (*handle_result)(void* arg, const Slice& k, const Slice& v));
+
+
+  void ReadMeta(const Footer& footer);
+  void ReadFilter(const Slice& filter_handle_value);
+
   // No copying allowed
   Table(const Table&);
   void operator=(const Table&);

include/leveldb/table_builder.h

@@ -77,6 +77,7 @@ class TableBuilder {
  private:
   bool ok() const { return status().ok(); }
   void WriteBlock(BlockBuilder* block, BlockHandle* handle);
+  void WriteRawBlock(const Slice& data, CompressionType, BlockHandle* handle);
 
   struct Rep;
   Rep* rep_;

issues/issue178_test.cc

@@ -0,0 +1,92 @@
+// Copyright (c) 2013 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.
+
+// Test for issue 178: a manual compaction causes deleted data to reappear.
+#include <iostream>
+#include <sstream>
+#include <cstdlib>
+
+#include "leveldb/db.h"
+#include "leveldb/write_batch.h"
+#include "util/testharness.h"
+
+namespace {
+
+const int kNumKeys = 1100000;
+
+std::string Key1(int i) {
+  char buf[100];
+  snprintf(buf, sizeof(buf), "my_key_%d", i);
+  return buf;
+}
+
+std::string Key2(int i) {
+  return Key1(i) + "_xxx";
+}
+
+class Issue178 { };
+
+TEST(Issue178, Test) {
+  // Get rid of any state from an old run.
+  std::string dbpath = leveldb::test::TmpDir() + "/leveldb_cbug_test";
+  DestroyDB(dbpath, leveldb::Options());
+
+  // Open database.  Disable compression since it affects the creation
+  // of layers and the code below is trying to test against a very
+  // specific scenario.
+  leveldb::DB* db;
+  leveldb::Options db_options;
+  db_options.create_if_missing = true;
+  db_options.compression = leveldb::kNoCompression;
+  ASSERT_OK(leveldb::DB::Open(db_options, dbpath, &db));
+
+  // create first key range
+  leveldb::WriteBatch batch;
+  for (size_t i = 0; i < kNumKeys; i++) {
+    batch.Put(Key1(i), "value for range 1 key");
+  }
+  ASSERT_OK(db->Write(leveldb::WriteOptions(), &batch));
+
+  // create second key range
+  batch.Clear();
+  for (size_t i = 0; i < kNumKeys; i++) {
+    batch.Put(Key2(i), "value for range 2 key");
+  }
+  ASSERT_OK(db->Write(leveldb::WriteOptions(), &batch));
+
+  // delete second key range
+  batch.Clear();
+  for (size_t i = 0; i < kNumKeys; i++) {
+    batch.Delete(Key2(i));
+  }
+  ASSERT_OK(db->Write(leveldb::WriteOptions(), &batch));
+
+  // compact database
+  std::string start_key = Key1(0);
+  std::string end_key = Key1(kNumKeys - 1);
+  leveldb::Slice least(start_key.data(), start_key.size());
+  leveldb::Slice greatest(end_key.data(), end_key.size());
+
+  // commenting out the line below causes the example to work correctly
+  db->CompactRange(&least, &greatest);
+
+  // count the keys
+  leveldb::Iterator* iter = db->NewIterator(leveldb::ReadOptions());
+  size_t num_keys = 0;
+  for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
+    num_keys++;
+  }
+  delete iter;
+  ASSERT_EQ(kNumKeys, num_keys) << "Bad number of keys";
+
+  // close database
+  delete db;
+  DestroyDB(dbpath, leveldb::Options());
+}
+
+}  // anonymous namespace
+
+int main(int argc, char** argv) {
+  return leveldb::test::RunAllTests();
+}

issues/issue200_test.cc

@@ -0,0 +1,59 @@
+// Copyright (c) 2013 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.
+
+// Test for issue 200: when iterator switches direction from backward
+// to forward, the current key can be yielded unexpectedly if a new
+// mutation has been added just before the current key.
+
+#include "leveldb/db.h"
+#include "util/testharness.h"
+
+namespace leveldb {
+
+class Issue200 { };
+
+TEST(Issue200, Test) {
+  // Get rid of any state from an old run.
+  std::string dbpath = test::TmpDir() + "/leveldb_issue200_test";
+  DestroyDB(dbpath, Options());
+
+  DB *db;
+  Options options;
+  options.create_if_missing = true;
+  ASSERT_OK(DB::Open(options, dbpath, &db));
+
+  WriteOptions write_options;
+  ASSERT_OK(db->Put(write_options, "1", "b"));
+  ASSERT_OK(db->Put(write_options, "2", "c"));
+  ASSERT_OK(db->Put(write_options, "3", "d"));
+  ASSERT_OK(db->Put(write_options, "4", "e"));
+  ASSERT_OK(db->Put(write_options, "5", "f"));
+
+  ReadOptions read_options;
+  Iterator *iter = db->NewIterator(read_options);
+
+  // Add an element that should not be reflected in the iterator.
+  ASSERT_OK(db->Put(write_options, "25", "cd"));
+
+  iter->Seek("5");
+  ASSERT_EQ(iter->key().ToString(), "5");
+  iter->Prev();
+  ASSERT_EQ(iter->key().ToString(), "4");
+  iter->Prev();
+  ASSERT_EQ(iter->key().ToString(), "3");
+  iter->Next();
+  ASSERT_EQ(iter->key().ToString(), "4");
+  iter->Next();
+  ASSERT_EQ(iter->key().ToString(), "5");
+
+  delete iter;
+  delete db;
+  DestroyDB(dbpath, options);
+}
+
+}  // namespace leveldb
+
+int main(int argc, char** argv) {
+  return leveldb::test::RunAllTests();
+}

port/atomic_pointer.h

@@ -36,6 +36,8 @@
 #define ARCH_CPU_X86_FAMILY 1
 #elif defined(__ARMEL__)
 #define ARCH_CPU_ARM_FAMILY 1
+#elif defined(__ppc__) || defined(__powerpc__) || defined(__powerpc64__)
+#define ARCH_CPU_PPC_FAMILY 1
 #endif
 
 namespace leveldb {
@@ -48,6 +50,13 @@ namespace port {
 // http://msdn.microsoft.com/en-us/library/ms684208(v=vs.85).aspx
 #define LEVELDB_HAVE_MEMORY_BARRIER
 
+// Mac OS
+#elif defined(OS_MACOSX)
+inline void MemoryBarrier() {
+  OSMemoryBarrier();
+}
+#define LEVELDB_HAVE_MEMORY_BARRIER
+
 // Gcc on x86
 #elif defined(ARCH_CPU_X86_FAMILY) && defined(__GNUC__)
 inline void MemoryBarrier() {
@@ -66,20 +75,30 @@ inline void MemoryBarrier() {
 }
 #define LEVELDB_HAVE_MEMORY_BARRIER
 
-// Mac OS
-#elif defined(OS_MACOSX)
+// ARM Linux
+#elif defined(ARCH_CPU_ARM_FAMILY) && defined(__linux__)
+typedef void (*LinuxKernelMemoryBarrierFunc)(void);
+// The Linux ARM kernel provides a highly optimized device-specific memory
+// barrier function at a fixed memory address that is mapped in every
+// user-level process.
+//
+// This beats using CPU-specific instructions which are, on single-core
+// devices, un-necessary and very costly (e.g. ARMv7-A "dmb" takes more
+// than 180ns on a Cortex-A8 like the one on a Nexus One). Benchmarking
+// shows that the extra function call cost is completely negligible on
+// multi-core devices.
+//
 inline void MemoryBarrier() {
-  OSMemoryBarrier();
+  (*(LinuxKernelMemoryBarrierFunc)0xffff0fa0)();
 }
 #define LEVELDB_HAVE_MEMORY_BARRIER
 
-// ARM
-#elif defined(ARCH_CPU_ARM_FAMILY)
-typedef void (*LinuxKernelMemoryBarrierFunc)(void);
-LinuxKernelMemoryBarrierFunc pLinuxKernelMemoryBarrier __attribute__((weak)) =
-    (LinuxKernelMemoryBarrierFunc) 0xffff0fa0;
+// PPC
+#elif defined(ARCH_CPU_PPC_FAMILY) && defined(__GNUC__)
 inline void MemoryBarrier() {
-  pLinuxKernelMemoryBarrier();
+  // TODO for some powerpc expert: is there a cheaper suitable variant?
+  // Perhaps by having separate barriers for acquire and release ops.
+  asm volatile("sync" : : : "memory");
 }
 #define LEVELDB_HAVE_MEMORY_BARRIER
 
@@ -128,6 +147,66 @@ class AtomicPointer {
   }
 };
 
+// Atomic pointer based on sparc memory barriers
+#elif defined(__sparcv9) && defined(__GNUC__)
+class AtomicPointer {
+ private:
+  void* rep_;
+ public:
+  AtomicPointer() { }
+  explicit AtomicPointer(void* v) : rep_(v) { }
+  inline void* Acquire_Load() const {
+    void* val;
+    __asm__ __volatile__ (
+        "ldx [%[rep_]], %[val] \n\t"
+         "membar #LoadLoad|#LoadStore \n\t"
+        : [val] "=r" (val)
+        : [rep_] "r" (&rep_)
+        : "memory");
+    return val;
+  }
+  inline void Release_Store(void* v) {
+    __asm__ __volatile__ (
+        "membar #LoadStore|#StoreStore \n\t"
+        "stx %[v], [%[rep_]] \n\t"
+        :
+        : [rep_] "r" (&rep_), [v] "r" (v)
+        : "memory");
+  }
+  inline void* NoBarrier_Load() const { return rep_; }
+  inline void NoBarrier_Store(void* v) { rep_ = v; }
+};
+
+// Atomic pointer based on ia64 acq/rel
+#elif defined(__ia64) && defined(__GNUC__)
+class AtomicPointer {
+ private:
+  void* rep_;
+ public:
+  AtomicPointer() { }
+  explicit AtomicPointer(void* v) : rep_(v) { }
+  inline void* Acquire_Load() const {
+    void* val    ;
+    __asm__ __volatile__ (
+        "ld8.acq %[val] = [%[rep_]] \n\t"
+        : [val] "=r" (val)
+        : [rep_] "r" (&rep_)
+        : "memory"
+        );
+    return val;
+  }
+  inline void Release_Store(void* v) {
+    __asm__ __volatile__ (
+        "st8.rel [%[rep_]] = %[v]  \n\t"
+        :
+        : [rep_] "r" (&rep_), [v] "r" (v)
+        : "memory"
+        );
+  }
+  inline void* NoBarrier_Load() const { return rep_; }
+  inline void NoBarrier_Store(void* v) { rep_ = v; }
+};
+
 // We have neither MemoryBarrier(), nor <cstdatomic>
 #else
 #error Please implement AtomicPointer for this platform.
@@ -137,6 +216,7 @@ class AtomicPointer {
 #undef LEVELDB_HAVE_MEMORY_BARRIER
 #undef ARCH_CPU_X86_FAMILY
 #undef ARCH_CPU_ARM_FAMILY
+#undef ARCH_CPU_PPC_FAMILY
 
 }  // namespace port
 }  // namespace leveldb

port/port.h

@@ -14,8 +14,6 @@
 #  include "port/port_posix.h"
 #elif defined(LEVELDB_PLATFORM_CHROMIUM)
 #  include "port/port_chromium.h"
-#elif defined(LEVELDB_PLATFORM_ANDROID)
-#  include "port/port_android.h"
 #endif
 
 #endif  // STORAGE_LEVELDB_PORT_PORT_H_

port/port_android.cc

@@ -1,64 +0,0 @@
-// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file. See the AUTHORS file for names of contributors.
-
-#include "port/port_android.h"
-
-#include <cstdlib>
-
-extern "C" {
-size_t fread_unlocked(void *a, size_t b, size_t c, FILE *d) {
-  return fread(a, b, c, d);
-}
-
-size_t fwrite_unlocked(const void *a, size_t b, size_t c, FILE *d) {
-  return fwrite(a, b, c, d);
-}
-
-int fflush_unlocked(FILE *f) {
-  return fflush(f);
-}
-
-int fdatasync(int fd) {
-  return fsync(fd);
-}
-}
-
-namespace leveldb {
-namespace port {
-
-static void PthreadCall(const char* label, int result) {
-  if (result != 0) {
-    fprintf(stderr, "pthread %s: %s\n", label, strerror(result));
-    abort();
-  }
-}
-
-Mutex::Mutex() { PthreadCall("init mutex", pthread_mutex_init(&mu_, NULL)); }
-Mutex::~Mutex() { PthreadCall("destroy mutex", pthread_mutex_destroy(&mu_)); }
-void Mutex::Lock() { PthreadCall("lock", pthread_mutex_lock(&mu_)); }
-void Mutex::Unlock() { PthreadCall("unlock", pthread_mutex_unlock(&mu_)); }
-
-CondVar::CondVar(Mutex* mu)
-    : mu_(mu) {
-  PthreadCall("init cv", pthread_cond_init(&cv_, NULL));
-}
-
-CondVar::~CondVar() { 
-  PthreadCall("destroy cv", pthread_cond_destroy(&cv_));
-}
-
-void CondVar::Wait() {
-  PthreadCall("wait", pthread_cond_wait(&cv_, &mu_->mu_));
-}
-
-void CondVar::Signal(){
-  PthreadCall("signal", pthread_cond_signal(&cv_));
-}
-
-void CondVar::SignalAll() {
-  PthreadCall("broadcast", pthread_cond_broadcast(&cv_));
-}
-
-}  // namespace port
-}  // namespace leveldb

port/port_android.h

@@ -1,156 +0,0 @@
-// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file. See the AUTHORS file for names of contributors.
-//
-// See port_example.h for documentation for the following types/functions.
-
-#ifndef STORAGE_LEVELDB_PORT_PORT_ANDROID_H_
-#define STORAGE_LEVELDB_PORT_PORT_ANDROID_H_
-
-#include <endian.h>
-#include <pthread.h>
-#include <stdint.h>
-#include <unistd.h>
-#include <string>
-#include <cctype>
-
-// Collapse the plethora of ARM flavors available to an easier to manage set
-// Defs reference is at https://wiki.edubuntu.org/ARM/Thumb2PortingHowto
-#if defined(__ARM_ARCH_6__) || \
-    defined(__ARM_ARCH_6J__) || \
-    defined(__ARM_ARCH_6K__) || \
-    defined(__ARM_ARCH_6Z__) || \
-    defined(__ARM_ARCH_6T2__) || \
-    defined(__ARM_ARCH_6ZK__) || \
-    defined(__ARM_ARCH_7__) || \
-    defined(__ARM_ARCH_7R__) || \
-    defined(__ARM_ARCH_7A__)
-#define ARMV6_OR_7 1
-#endif
-
-extern "C" {
-  size_t fread_unlocked(void *a, size_t b, size_t c, FILE *d);
-  size_t fwrite_unlocked(const void *a, size_t b, size_t c, FILE *d);
-  int fflush_unlocked(FILE *f);
-  int fdatasync (int fd);
-}
-
-namespace leveldb {
-namespace port {
-
-static const bool kLittleEndian = __BYTE_ORDER == __LITTLE_ENDIAN;
-
-class CondVar;
-
-class Mutex {
- public:
-  Mutex();
-  ~Mutex();
-
-  void Lock();
-  void Unlock();
-  void AssertHeld() {
-    //TODO(gabor): How can I implement this?
-  }
-
- private:
-  friend class CondVar;
-  pthread_mutex_t mu_;
-
-  // No copying
-  Mutex(const Mutex&);
-  void operator=(const Mutex&);
-};
-
-class CondVar {
- public:
-  explicit CondVar(Mutex* mu);
-  ~CondVar();
-  void Wait();
-  void Signal();
-  void SignalAll();
- private:
-  Mutex* mu_;
-  pthread_cond_t cv_;
-};
-
-#ifndef ARMV6_OR_7
-// On ARM chipsets <V6, 0xffff0fa0 is the hard coded address of a 
-// memory barrier function provided by the kernel.
-typedef void (*LinuxKernelMemoryBarrierFunc)(void);
-LinuxKernelMemoryBarrierFunc pLinuxKernelMemoryBarrier ATTRIBUTE_WEAK =
-    (LinuxKernelMemoryBarrierFunc) 0xffff0fa0;
-#endif
-
-// Storage for a lock-free pointer
-class AtomicPointer {
- private:
-  void* rep_;
-
-  inline void MemoryBarrier() const {
-    // TODO(gabor): This only works on Android instruction sets >= V6
-#ifdef ARMV6_OR_7
-    __asm__ __volatile__("dmb" : : : "memory");
-#else
-    pLinuxKernelMemoryBarrier();
-#endif
-  }
-
- public:
-  AtomicPointer() { }
-  explicit AtomicPointer(void* v) : rep_(v) { }
-  inline void* Acquire_Load() const {
-    void* r = rep_;
-    MemoryBarrier();
-    return r;
-  }
-  inline void Release_Store(void* v) {
-    MemoryBarrier();
-    rep_ = v;
-  }
-  inline void* NoBarrier_Load() const {
-    void* r = rep_;
-    return r;
-  }
-  inline void NoBarrier_Store(void* v) {
-    rep_ = v;
-  }
-};
-
-// TODO(gabor): Implement compress
-inline bool Snappy_Compress(
-    const char* input,
-    size_t input_length,
-    std::string* output) {
-  return false;
-}
-
-// TODO(gabor): Implement uncompress
-inline bool Snappy_GetUncompressedLength(const char* input, size_t length,
-                                         size_t* result) {
-  return false;
-}
-
-// TODO(gabor): Implement uncompress
-inline bool Snappy_Uncompress(
-    const char* input_data,
-    size_t input_length,
-    char* output) {
-  return false;
-}
-
-inline uint64_t ThreadIdentifier() {
-  pthread_t tid = pthread_self();
-  uint64_t r = 0;
-  memcpy(&r, &tid, sizeof(r) < sizeof(tid) ? sizeof(r) : sizeof(tid));
-  return r;
-}
-
-inline bool GetHeapProfile(void (*func)(void*, const char*, int), void* arg) {
-  return false;
-}
-
-}  // namespace port
-}  // namespace leveldb
-
-#endif  // STORAGE_LEVELDB_PORT_PORT_ANDROID_H_

port/port_example.h

@@ -60,6 +60,16 @@ class CondVar {
   void SignallAll();
 };
 
+// Thread-safe initialization.
+// Used as follows:
+//      static port::OnceType init_control = LEVELDB_ONCE_INIT;
+//      static void Initializer() { ... do something ...; }
+//      ...
+//      port::InitOnce(&init_control, &Initializer);
+typedef intptr_t OnceType;
+#define LEVELDB_ONCE_INIT 0
+extern void InitOnce(port::OnceType*, void (*initializer)());
+
 // A type that holds a pointer that can be read or written atomically
 // (i.e., without word-tearing.)
 class AtomicPointer {

port/port_posix.cc

@@ -46,5 +46,9 @@ void CondVar::SignalAll() {
   PthreadCall("broadcast", pthread_cond_broadcast(&cv_));
 }
 
+void InitOnce(OnceType* once, void (*initializer)()) {
+  PthreadCall("once", pthread_once(once, initializer));
+}
+
 }  // namespace port
 }  // namespace leveldb

port/port_posix.h

@@ -7,22 +7,40 @@
 #ifndef STORAGE_LEVELDB_PORT_PORT_POSIX_H_
 #define STORAGE_LEVELDB_PORT_PORT_POSIX_H_
 
+#undef PLATFORM_IS_LITTLE_ENDIAN
 #if defined(OS_MACOSX)
   #include <machine/endian.h>
+  #if defined(__DARWIN_LITTLE_ENDIAN) && defined(__DARWIN_BYTE_ORDER)
+    #define PLATFORM_IS_LITTLE_ENDIAN \
+        (__DARWIN_BYTE_ORDER == __DARWIN_LITTLE_ENDIAN)
+  #endif
 #elif defined(OS_SOLARIS)
   #include <sys/isa_defs.h>
   #ifdef _LITTLE_ENDIAN
-    #define LITTLE_ENDIAN
+    #define PLATFORM_IS_LITTLE_ENDIAN true
   #else
-    #define BIG_ENDIAN
+    #define PLATFORM_IS_LITTLE_ENDIAN false
   #endif
-#elif defined(OS_FREEBSD) || defined(OS_OPENBSD) || defined(OS_NETBSD) ||\
+#elif defined(OS_FREEBSD)
+  #include <sys/types.h>
+  #include <sys/endian.h>
+  #define PLATFORM_IS_LITTLE_ENDIAN (_BYTE_ORDER == _LITTLE_ENDIAN)
+#elif defined(OS_OPENBSD) || defined(OS_NETBSD) ||\
       defined(OS_DRAGONFLYBSD)
   #include <sys/types.h>
   #include <sys/endian.h>
+#elif defined(OS_HPUX)
+  #define PLATFORM_IS_LITTLE_ENDIAN false
+#elif defined(OS_ANDROID)
+  // Due to a bug in the NDK x86 <sys/endian.h> definition,
+  // _BYTE_ORDER must be used instead of __BYTE_ORDER on Android.
+  // See http://code.google.com/p/android/issues/detail?id=39824
+  #include <endian.h>
+  #define PLATFORM_IS_LITTLE_ENDIAN  (_BYTE_ORDER == _LITTLE_ENDIAN)
 #else
   #include <endian.h>
 #endif
+
 #include <pthread.h>
 #ifdef SNAPPY
 #include <snappy.h>
@@ -31,14 +49,13 @@
 #include <string>
 #include "port/atomic_pointer.h"
 
-#ifdef LITTLE_ENDIAN
-#define IS_LITTLE_ENDIAN true
-#else
-#define IS_LITTLE_ENDIAN (__BYTE_ORDER == __LITTLE_ENDIAN)
+#ifndef PLATFORM_IS_LITTLE_ENDIAN
+#define PLATFORM_IS_LITTLE_ENDIAN (__BYTE_ORDER == __LITTLE_ENDIAN)
 #endif
 
 #if defined(OS_MACOSX) || defined(OS_SOLARIS) || defined(OS_FREEBSD) ||\
-    defined(OS_NETBSD) || defined(OS_OPENBSD) || defined(OS_DRAGONFLYBSD)
+    defined(OS_NETBSD) || defined(OS_OPENBSD) || defined(OS_DRAGONFLYBSD) ||\
+    defined(OS_ANDROID) || defined(OS_HPUX)
 // Use fread/fwrite/fflush on platforms without _unlocked variants
 #define fread_unlocked fread
 #define fwrite_unlocked fwrite
@@ -51,10 +68,17 @@
 #define fdatasync fsync
 #endif
 
+#if defined(OS_ANDROID) && __ANDROID_API__ < 9
+// fdatasync() was only introduced in API level 9 on Android. Use fsync()
+// when targetting older platforms.
+#define fdatasync fsync
+#endif
+
 namespace leveldb {
 namespace port {
 
-static const bool kLittleEndian = IS_LITTLE_ENDIAN;
+static const bool kLittleEndian = PLATFORM_IS_LITTLE_ENDIAN;
+#undef PLATFORM_IS_LITTLE_ENDIAN
 
 class CondVar;
 
@@ -88,6 +112,10 @@ class CondVar {
   Mutex* mu_;
 };
 
+typedef pthread_once_t OnceType;
+#define LEVELDB_ONCE_INIT PTHREAD_ONCE_INIT
+extern void InitOnce(OnceType* once, void (*initializer)());
+
 inline bool Snappy_Compress(const char* input, size_t length,
                             ::std::string* output) {
 #ifdef SNAPPY

port/thread_annotations.h

@@ -0,0 +1,59 @@
+// Copyright (c) 2012 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#ifndef STORAGE_LEVELDB_PORT_THREAD_ANNOTATIONS_H
+
+// Some environments provide custom macros to aid in static thread-safety
+// analysis.  Provide empty definitions of such macros unless they are already
+// defined.
+
+#ifndef EXCLUSIVE_LOCKS_REQUIRED
+#define EXCLUSIVE_LOCKS_REQUIRED(...)
+#endif
+
+#ifndef SHARED_LOCKS_REQUIRED
+#define SHARED_LOCKS_REQUIRED(...)
+#endif
+
+#ifndef LOCKS_EXCLUDED
+#define LOCKS_EXCLUDED(...)
+#endif
+
+#ifndef LOCK_RETURNED
+#define LOCK_RETURNED(x)
+#endif
+
+#ifndef LOCKABLE
+#define LOCKABLE
+#endif
+
+#ifndef SCOPED_LOCKABLE
+#define SCOPED_LOCKABLE
+#endif
+
+#ifndef EXCLUSIVE_LOCK_FUNCTION
+#define EXCLUSIVE_LOCK_FUNCTION(...)
+#endif
+
+#ifndef SHARED_LOCK_FUNCTION
+#define SHARED_LOCK_FUNCTION(...)
+#endif
+
+#ifndef EXCLUSIVE_TRYLOCK_FUNCTION
+#define EXCLUSIVE_TRYLOCK_FUNCTION(...)
+#endif
+
+#ifndef SHARED_TRYLOCK_FUNCTION
+#define SHARED_TRYLOCK_FUNCTION(...)
+#endif
+
+#ifndef UNLOCK_FUNCTION
+#define UNLOCK_FUNCTION(...)
+#endif
+
+#ifndef NO_THREAD_SAFETY_ANALYSIS
+#define NO_THREAD_SAFETY_ANALYSIS
+#endif
+
+#endif  // STORAGE_LEVELDB_PORT_THREAD_ANNOTATIONS_H

table/block.cc

@@ -9,28 +9,30 @@
 #include <vector>
 #include <algorithm>
 #include "leveldb/comparator.h"
+#include "table/format.h"
 #include "util/coding.h"
 #include "util/logging.h"
 
 namespace leveldb {
 
 inline uint32_t Block::NumRestarts() const {
-  assert(size_ >= 2*sizeof(uint32_t));
+  assert(size_ >= sizeof(uint32_t));
   return DecodeFixed32(data_ + size_ - sizeof(uint32_t));
 }
 
-Block::Block(const char* data, size_t size, bool take_ownership)
-    : data_(data),
-      size_(size),
-      owned_(take_ownership) {
+Block::Block(const BlockContents& contents)
+    : data_(contents.data.data()),
+      size_(contents.data.size()),
+      owned_(contents.heap_allocated) {
   if (size_ < sizeof(uint32_t)) {
     size_ = 0;  // Error marker
   } else {
-    restart_offset_ = size_ - (1 + NumRestarts()) * sizeof(uint32_t);
-    if (restart_offset_ > size_ - sizeof(uint32_t)) {
-      // The size is too small for NumRestarts() and therefore
-      // restart_offset_ wrapped around.
+    size_t max_restarts_allowed = (size_-sizeof(uint32_t)) / sizeof(uint32_t);
+    if (NumRestarts() > max_restarts_allowed) {
+      // The size is too small for NumRestarts()
       size_ = 0;
+    } else {
+      restart_offset_ = size_ - (1 + NumRestarts()) * sizeof(uint32_t);
     }
   }
 }
@@ -161,8 +163,8 @@ class Block::Iter : public Iterator {
   }
 
   virtual void Seek(const Slice& target) {
-    // Binary search in restart array to find the first restart point
-    // with a key >= target
+    // Binary search in restart array to find the last restart point
+    // with a key < target
     uint32_t left = 0;
     uint32_t right = num_restarts_ - 1;
     while (left < right) {
@@ -252,7 +254,7 @@ class Block::Iter : public Iterator {
 };
 
 Iterator* Block::NewIterator(const Comparator* cmp) {
-  if (size_ < 2*sizeof(uint32_t)) {
+  if (size_ < sizeof(uint32_t)) {
     return NewErrorIterator(Status::Corruption("bad block contents"));
   }
   const uint32_t num_restarts = NumRestarts();

table/block.h

@@ -11,14 +11,13 @@
 
 namespace leveldb {
 
+struct BlockContents;
 class Comparator;
 
 class Block {
  public:
   // Initialize the block with the specified contents.
-  // Takes ownership of data[] and will delete[] it when done iff
-  // "take_ownership is true.
-  Block(const char* data, size_t size, bool take_ownership);
+  explicit Block(const BlockContents& contents);
 
   ~Block();
 

table/filter_block.cc

@@ -0,0 +1,111 @@
+// Copyright (c) 2012 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 "table/filter_block.h"
+
+#include "leveldb/filter_policy.h"
+#include "util/coding.h"
+
+namespace leveldb {
+
+// See doc/table_format.txt for an explanation of the filter block format.
+
+// Generate new filter every 2KB of data
+static const size_t kFilterBaseLg = 11;
+static const size_t kFilterBase = 1 << kFilterBaseLg;
+
+FilterBlockBuilder::FilterBlockBuilder(const FilterPolicy* policy)
+    : policy_(policy) {
+}
+
+void FilterBlockBuilder::StartBlock(uint64_t block_offset) {
+  uint64_t filter_index = (block_offset / kFilterBase);
+  assert(filter_index >= filter_offsets_.size());
+  while (filter_index > filter_offsets_.size()) {
+    GenerateFilter();
+  }
+}
+
+void FilterBlockBuilder::AddKey(const Slice& key) {
+  Slice k = key;
+  start_.push_back(keys_.size());
+  keys_.append(k.data(), k.size());
+}
+
+Slice FilterBlockBuilder::Finish() {
+  if (!start_.empty()) {
+    GenerateFilter();
+  }
+
+  // Append array of per-filter offsets
+  const uint32_t array_offset = result_.size();
+  for (size_t i = 0; i < filter_offsets_.size(); i++) {
+    PutFixed32(&result_, filter_offsets_[i]);
+  }
+
+  PutFixed32(&result_, array_offset);
+  result_.push_back(kFilterBaseLg);  // Save encoding parameter in result
+  return Slice(result_);
+}
+
+void FilterBlockBuilder::GenerateFilter() {
+  const size_t num_keys = start_.size();
+  if (num_keys == 0) {
+    // Fast path if there are no keys for this filter
+    filter_offsets_.push_back(result_.size());
+    return;
+  }
+
+  // Make list of keys from flattened key structure
+  start_.push_back(keys_.size());  // Simplify length computation
+  tmp_keys_.resize(num_keys);
+  for (size_t i = 0; i < num_keys; i++) {
+    const char* base = keys_.data() + start_[i];
+    size_t length = start_[i+1] - start_[i];
+    tmp_keys_[i] = Slice(base, length);
+  }
+
+  // Generate filter for current set of keys and append to result_.
+  filter_offsets_.push_back(result_.size());
+  policy_->CreateFilter(&tmp_keys_[0], num_keys, &result_);
+
+  tmp_keys_.clear();
+  keys_.clear();
+  start_.clear();
+}
+
+FilterBlockReader::FilterBlockReader(const FilterPolicy* policy,
+                                     const Slice& contents)
+    : policy_(policy),
+      data_(NULL),
+      offset_(NULL),
+      num_(0),
+      base_lg_(0) {
+  size_t n = contents.size();
+  if (n < 5) return;  // 1 byte for base_lg_ and 4 for start of offset array
+  base_lg_ = contents[n-1];
+  uint32_t last_word = DecodeFixed32(contents.data() + n - 5);
+  if (last_word > n - 5) return;
+  data_ = contents.data();
+  offset_ = data_ + last_word;
+  num_ = (n - 5 - last_word) / 4;
+}
+
+bool FilterBlockReader::KeyMayMatch(uint64_t block_offset, const Slice& key) {
+  uint64_t index = block_offset >> base_lg_;
+  if (index < num_) {
+    uint32_t start = DecodeFixed32(offset_ + index*4);
+    uint32_t limit = DecodeFixed32(offset_ + index*4 + 4);
+    if (start <= limit && limit <= (offset_ - data_)) {
+      Slice filter = Slice(data_ + start, limit - start);
+      return policy_->KeyMayMatch(key, filter);
+    } else if (start == limit) {
+      // Empty filters do not match any keys
+      return false;
+    }
+  }
+  return true;  // Errors are treated as potential matches
+}
+
+}

table/filter_block.h

@@ -0,0 +1,68 @@
+// Copyright (c) 2012 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.
+//
+// A filter block is stored near the end of a Table file.  It contains
+// filters (e.g., bloom filters) for all data blocks in the table combined
+// into a single filter block.
+
+#ifndef STORAGE_LEVELDB_TABLE_FILTER_BLOCK_H_
+#define STORAGE_LEVELDB_TABLE_FILTER_BLOCK_H_
+
+#include <stddef.h>
+#include <stdint.h>
+#include <string>
+#include <vector>
+#include "leveldb/slice.h"
+#include "util/hash.h"
+
+namespace leveldb {
+
+class FilterPolicy;
+
+// A FilterBlockBuilder is used to construct all of the filters for a
+// particular Table.  It generates a single string which is stored as
+// a special block in the Table.
+//
+// The sequence of calls to FilterBlockBuilder must match the regexp:
+//      (StartBlock AddKey*)* Finish
+class FilterBlockBuilder {
+ public:
+  explicit FilterBlockBuilder(const FilterPolicy*);
+
+  void StartBlock(uint64_t block_offset);
+  void AddKey(const Slice& key);
+  Slice Finish();
+
+ private:
+  void GenerateFilter();
+
+  const FilterPolicy* policy_;
+  std::string keys_;              // Flattened key contents
+  std::vector<size_t> start_;     // Starting index in keys_ of each key
+  std::string result_;            // Filter data computed so far
+  std::vector<Slice> tmp_keys_;   // policy_->CreateFilter() argument
+  std::vector<uint32_t> filter_offsets_;
+
+  // No copying allowed
+  FilterBlockBuilder(const FilterBlockBuilder&);
+  void operator=(const FilterBlockBuilder&);
+};
+
+class FilterBlockReader {
+ public:
+ // REQUIRES: "contents" and *policy must stay live while *this is live.
+  FilterBlockReader(const FilterPolicy* policy, const Slice& contents);
+  bool KeyMayMatch(uint64_t block_offset, const Slice& key);
+
+ private:
+  const FilterPolicy* policy_;
+  const char* data_;    // Pointer to filter data (at block-start)
+  const char* offset_;  // Pointer to beginning of offset array (at block-end)
+  size_t num_;          // Number of entries in offset array
+  size_t base_lg_;      // Encoding parameter (see kFilterBaseLg in .cc file)
+};
+
+}
+
+#endif  // STORAGE_LEVELDB_TABLE_FILTER_BLOCK_H_

table/filter_block_test.cc

@@ -0,0 +1,128 @@
+// Copyright (c) 2012 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 "table/filter_block.h"
+
+#include "leveldb/filter_policy.h"
+#include "util/coding.h"
+#include "util/hash.h"
+#include "util/logging.h"
+#include "util/testharness.h"
+#include "util/testutil.h"
+
+namespace leveldb {
+
+// For testing: emit an array with one hash value per key
+class TestHashFilter : public FilterPolicy {
+ public:
+  virtual const char* Name() const {
+    return "TestHashFilter";
+  }
+
+  virtual void CreateFilter(const Slice* keys, int n, std::string* dst) const {
+    for (int i = 0; i < n; i++) {
+      uint32_t h = Hash(keys[i].data(), keys[i].size(), 1);
+      PutFixed32(dst, h);
+    }
+  }
+
+  virtual bool KeyMayMatch(const Slice& key, const Slice& filter) const {
+    uint32_t h = Hash(key.data(), key.size(), 1);
+    for (size_t i = 0; i + 4 <= filter.size(); i += 4) {
+      if (h == DecodeFixed32(filter.data() + i)) {
+        return true;
+      }
+    }
+    return false;
+  }
+};
+
+class FilterBlockTest {
+ public:
+  TestHashFilter policy_;
+};
+
+TEST(FilterBlockTest, EmptyBuilder) {
+  FilterBlockBuilder builder(&policy_);
+  Slice block = builder.Finish();
+  ASSERT_EQ("\\x00\\x00\\x00\\x00\\x0b", EscapeString(block));
+  FilterBlockReader reader(&policy_, block);
+  ASSERT_TRUE(reader.KeyMayMatch(0, "foo"));
+  ASSERT_TRUE(reader.KeyMayMatch(100000, "foo"));
+}
+
+TEST(FilterBlockTest, SingleChunk) {
+  FilterBlockBuilder builder(&policy_);
+  builder.StartBlock(100);
+  builder.AddKey("foo");
+  builder.AddKey("bar");
+  builder.AddKey("box");
+  builder.StartBlock(200);
+  builder.AddKey("box");
+  builder.StartBlock(300);
+  builder.AddKey("hello");
+  Slice block = builder.Finish();
+  FilterBlockReader reader(&policy_, block);
+  ASSERT_TRUE(reader.KeyMayMatch(100, "foo"));
+  ASSERT_TRUE(reader.KeyMayMatch(100, "bar"));
+  ASSERT_TRUE(reader.KeyMayMatch(100, "box"));
+  ASSERT_TRUE(reader.KeyMayMatch(100, "hello"));
+  ASSERT_TRUE(reader.KeyMayMatch(100, "foo"));
+  ASSERT_TRUE(! reader.KeyMayMatch(100, "missing"));
+  ASSERT_TRUE(! reader.KeyMayMatch(100, "other"));
+}
+
+TEST(FilterBlockTest, MultiChunk) {
+  FilterBlockBuilder builder(&policy_);
+
+  // First filter
+  builder.StartBlock(0);
+  builder.AddKey("foo");
+  builder.StartBlock(2000);
+  builder.AddKey("bar");
+
+  // Second filter
+  builder.StartBlock(3100);
+  builder.AddKey("box");
+
+  // Third filter is empty
+
+  // Last filter
+  builder.StartBlock(9000);
+  builder.AddKey("box");
+  builder.AddKey("hello");
+
+  Slice block = builder.Finish();
+  FilterBlockReader reader(&policy_, block);
+
+  // Check first filter
+  ASSERT_TRUE(reader.KeyMayMatch(0, "foo"));
+  ASSERT_TRUE(reader.KeyMayMatch(2000, "bar"));
+  ASSERT_TRUE(! reader.KeyMayMatch(0, "box"));
+  ASSERT_TRUE(! reader.KeyMayMatch(0, "hello"));
+
+  // Check second filter
+  ASSERT_TRUE(reader.KeyMayMatch(3100, "box"));
+  ASSERT_TRUE(! reader.KeyMayMatch(3100, "foo"));
+  ASSERT_TRUE(! reader.KeyMayMatch(3100, "bar"));
+  ASSERT_TRUE(! reader.KeyMayMatch(3100, "hello"));
+
+  // Check third filter (empty)
+  ASSERT_TRUE(! reader.KeyMayMatch(4100, "foo"));
+  ASSERT_TRUE(! reader.KeyMayMatch(4100, "bar"));
+  ASSERT_TRUE(! reader.KeyMayMatch(4100, "box"));
+  ASSERT_TRUE(! reader.KeyMayMatch(4100, "hello"));
+
+  // Check last filter
+  ASSERT_TRUE(reader.KeyMayMatch(9000, "box"));
+  ASSERT_TRUE(reader.KeyMayMatch(9000, "hello"));
+  ASSERT_TRUE(! reader.KeyMayMatch(9000, "foo"));
+  ASSERT_TRUE(! reader.KeyMayMatch(9000, "bar"));
+}
+
+}  // namespace leveldb
+
+int main(int argc, char** argv) {
+  return leveldb::test::RunAllTests();
+}

table/format.cc

@@ -66,10 +66,10 @@ Status Footer::DecodeFrom(Slice* input) {
 Status ReadBlock(RandomAccessFile* file,
                  const ReadOptions& options,
                  const BlockHandle& handle,
-                 Block** block,
-                 bool* may_cache) {
-  *block = NULL;
-  *may_cache = false;
+                 BlockContents* result) {
+  result->data = Slice();
+  result->cachable = false;
+  result->heap_allocated = false;
 
   // Read the block contents as well as the type/crc footer.
   // See table_builder.cc for the code that built this structure.
@@ -105,11 +105,13 @@ Status ReadBlock(RandomAccessFile* file,
         // Use it directly under the assumption that it will be live
         // while the file is open.
         delete[] buf;
-        *block = new Block(data, n, false /* do not take ownership */);
-        *may_cache = false;  // Do not double-cache
+        result->data = Slice(data, n);
+        result->heap_allocated = false;
+        result->cachable = false;  // Do not double-cache
       } else {
-        *block = new Block(buf, n, true /* take ownership */);
-        *may_cache = true;
+        result->data = Slice(buf, n);
+        result->heap_allocated = true;
+        result->cachable = true;
       }
 
       // Ok
@@ -127,8 +129,9 @@ Status ReadBlock(RandomAccessFile* file,
         return Status::Corruption("corrupted compressed block contents");
       }
       delete[] buf;
-      *block = new Block(ubuf, ulength, true /* take ownership */);
-      *may_cache = true;
+      result->data = Slice(ubuf, ulength);
+      result->heap_allocated = true;
+      result->cachable = true;
       break;
     }
     default:

table/format.h

@@ -83,16 +83,18 @@ static const uint64_t kTableMagicNumber = 0xdb4775248b80fb57ull;
 // 1-byte type + 32-bit crc
 static const size_t kBlockTrailerSize = 5;
 
-// Read the block identified by "handle" from "file".  On success,
-// store a pointer to the heap-allocated result in *block and return
-// OK.  On failure store NULL in *block and return non-OK.
-// On success, stores true in *may_cache if the result may be
-// cached, false if it must not be cached.
+struct BlockContents {
+  Slice data;           // Actual contents of data
+  bool cachable;        // True iff data can be cached
+  bool heap_allocated;  // True iff caller should delete[] data.data()
+};
+
+// Read the block identified by "handle" from "file".  On failure
+// return non-OK.  On success fill *result and return OK.
 extern Status ReadBlock(RandomAccessFile* file,
                         const ReadOptions& options,
                         const BlockHandle& handle,
-                        Block** block,
-                        bool* may_cache);
+                        BlockContents* result);
 
 // Implementation details follow.  Clients should ignore,
 

table/table.cc

@@ -5,8 +5,12 @@
 #include "leveldb/table.h"
 
 #include "leveldb/cache.h"
+#include "leveldb/comparator.h"
 #include "leveldb/env.h"
+#include "leveldb/filter_policy.h"
+#include "leveldb/options.h"
 #include "table/block.h"
+#include "table/filter_block.h"
 #include "table/format.h"
 #include "table/two_level_iterator.h"
 #include "util/coding.h"
@@ -15,6 +19,8 @@ namespace leveldb {
 
 struct Table::Rep {
   ~Rep() {
+    delete filter;
+    delete [] filter_data;
     delete index_block;
   }
 
@@ -22,6 +28,8 @@ struct Table::Rep {
   Status status;
   RandomAccessFile* file;
   uint64_t cache_id;
+  FilterBlockReader* filter;
+  const char* filter_data;
 
   BlockHandle metaindex_handle;  // Handle to metaindex_block: saved from footer
   Block* index_block;
@@ -47,11 +55,13 @@ Status Table::Open(const Options& options,
   if (!s.ok()) return s;
 
   // Read the index block
+  BlockContents contents;
   Block* index_block = NULL;
   if (s.ok()) {
-    bool may_cache;  // Ignored result
-    s = ReadBlock(file, ReadOptions(), footer.index_handle(), &index_block,
-                  &may_cache);
+    s = ReadBlock(file, ReadOptions(), footer.index_handle(), &contents);
+    if (s.ok()) {
+      index_block = new Block(contents);
+    }
   }
 
   if (s.ok()) {
@@ -63,7 +73,10 @@ Status Table::Open(const Options& options,
     rep->metaindex_handle = footer.metaindex_handle();
     rep->index_block = index_block;
     rep->cache_id = (options.block_cache ? options.block_cache->NewId() : 0);
+    rep->filter_data = NULL;
+    rep->filter = NULL;
     *table = new Table(rep);
+    (*table)->ReadMeta(footer);
   } else {
     if (index_block) delete index_block;
   }
@@ -71,6 +84,52 @@ Status Table::Open(const Options& options,
   return s;
 }
 
+void Table::ReadMeta(const Footer& footer) {
+  if (rep_->options.filter_policy == NULL) {
+    return;  // Do not need any metadata
+  }
+
+  // TODO(sanjay): Skip this if footer.metaindex_handle() size indicates
+  // it is an empty block.
+  ReadOptions opt;
+  BlockContents contents;
+  if (!ReadBlock(rep_->file, opt, footer.metaindex_handle(), &contents).ok()) {
+    // Do not propagate errors since meta info is not needed for operation
+    return;
+  }
+  Block* meta = new Block(contents);
+
+  Iterator* iter = meta->NewIterator(BytewiseComparator());
+  std::string key = "filter.";
+  key.append(rep_->options.filter_policy->Name());
+  iter->Seek(key);
+  if (iter->Valid() && iter->key() == Slice(key)) {
+    ReadFilter(iter->value());
+  }
+  delete iter;
+  delete meta;
+}
+
+void Table::ReadFilter(const Slice& filter_handle_value) {
+  Slice v = filter_handle_value;
+  BlockHandle filter_handle;
+  if (!filter_handle.DecodeFrom(&v).ok()) {
+    return;
+  }
+
+  // We might want to unify with ReadBlock() if we start
+  // requiring checksum verification in Table::Open.
+  ReadOptions opt;
+  BlockContents block;
+  if (!ReadBlock(rep_->file, opt, filter_handle, &block).ok()) {
+    return;
+  }
+  if (block.heap_allocated) {
+    rep_->filter_data = block.data.data();     // Will need to delete later
+  }
+  rep_->filter = new FilterBlockReader(rep_->options.filter_policy, block.data);
+}
+
 Table::~Table() {
   delete rep_;
 }
@@ -107,7 +166,7 @@ Iterator* Table::BlockReader(void* arg,
   // can add more features in the future.
 
   if (s.ok()) {
-    bool may_cache;
+    BlockContents contents;
     if (block_cache != NULL) {
       char cache_key_buffer[16];
       EncodeFixed64(cache_key_buffer, table->rep_->cache_id);
@@ -117,14 +176,20 @@ Iterator* Table::BlockReader(void* arg,
       if (cache_handle != NULL) {
         block = reinterpret_cast<Block*>(block_cache->Value(cache_handle));
       } else {
-        s = ReadBlock(table->rep_->file, options, handle, &block, &may_cache);
-        if (s.ok() && may_cache && options.fill_cache) {
-          cache_handle = block_cache->Insert(
-              key, block, block->size(), &DeleteCachedBlock);
+        s = ReadBlock(table->rep_->file, options, handle, &contents);
+        if (s.ok()) {
+          block = new Block(contents);
+          if (contents.cachable && options.fill_cache) {
+            cache_handle = block_cache->Insert(
+                key, block, block->size(), &DeleteCachedBlock);
+          }
         }
       }
     } else {
-      s = ReadBlock(table->rep_->file, options, handle, &block, &may_cache);
+      s = ReadBlock(table->rep_->file, options, handle, &contents);
+      if (s.ok()) {
+        block = new Block(contents);
+      }
     }
   }
 
@@ -148,6 +213,38 @@ Iterator* Table::NewIterator(const ReadOptions& options) const {
       &Table::BlockReader, const_cast<Table*>(this), options);
 }
 
+Status Table::InternalGet(const ReadOptions& options, const Slice& k,
+                          void* arg,
+                          void (*saver)(void*, const Slice&, const Slice&)) {
+  Status s;
+  Iterator* iiter = rep_->index_block->NewIterator(rep_->options.comparator);
+  iiter->Seek(k);
+  if (iiter->Valid()) {
+    Slice handle_value = iiter->value();
+    FilterBlockReader* filter = rep_->filter;
+    BlockHandle handle;
+    if (filter != NULL &&
+        handle.DecodeFrom(&handle_value).ok() &&
+        !filter->KeyMayMatch(handle.offset(), k)) {
+      // Not found
+    } else {
+      Iterator* block_iter = BlockReader(this, options, iiter->value());
+      block_iter->Seek(k);
+      if (block_iter->Valid()) {
+        (*saver)(arg, block_iter->key(), block_iter->value());
+      }
+      s = block_iter->status();
+      delete block_iter;
+    }
+  }
+  if (s.ok()) {
+    s = iiter->status();
+  }
+  delete iiter;
+  return s;
+}
+
+
 uint64_t Table::ApproximateOffsetOf(const Slice& key) const {
   Iterator* index_iter =
       rep_->index_block->NewIterator(rep_->options.comparator);

table/table_builder.cc

@@ -5,14 +5,15 @@
 #include "leveldb/table_builder.h"
 
 #include <assert.h>
-#include <stdio.h>
 #include "leveldb/comparator.h"
 #include "leveldb/env.h"
+#include "leveldb/filter_policy.h"
+#include "leveldb/options.h"
 #include "table/block_builder.h"
+#include "table/filter_block.h"
 #include "table/format.h"
 #include "util/coding.h"
 #include "util/crc32c.h"
-#include "util/logging.h"
 
 namespace leveldb {
 
@@ -27,6 +28,7 @@ struct TableBuilder::Rep {
   std::string last_key;
   int64_t num_entries;
   bool closed;          // Either Finish() or Abandon() has been called.
+  FilterBlockBuilder* filter_block;
 
   // We do not emit the index entry for a block until we have seen the
   // first key for the next data block.  This allows us to use shorter
@@ -51,6 +53,8 @@ struct TableBuilder::Rep {
         index_block(&index_block_options),
         num_entries(0),
         closed(false),
+        filter_block(opt.filter_policy == NULL ? NULL
+                     : new FilterBlockBuilder(opt.filter_policy)),
         pending_index_entry(false) {
     index_block_options.block_restart_interval = 1;
   }
@@ -58,10 +62,14 @@ struct TableBuilder::Rep {
 
 TableBuilder::TableBuilder(const Options& options, WritableFile* file)
     : rep_(new Rep(options, file)) {
+  if (rep_->filter_block != NULL) {
+    rep_->filter_block->StartBlock(0);
+  }
 }
 
 TableBuilder::~TableBuilder() {
   assert(rep_->closed);  // Catch errors where caller forgot to call Finish()
+  delete rep_->filter_block;
   delete rep_;
 }
 
@@ -98,6 +106,10 @@ void TableBuilder::Add(const Slice& key, const Slice& value) {
     r->pending_index_entry = false;
   }
 
+  if (r->filter_block != NULL) {
+    r->filter_block->AddKey(key);
+  }
+
   r->last_key.assign(key.data(), key.size());
   r->num_entries++;
   r->data_block.Add(key, value);
@@ -119,6 +131,9 @@ void TableBuilder::Flush() {
     r->pending_index_entry = true;
     r->status = r->file->Flush();
   }
+  if (r->filter_block != NULL) {
+    r->filter_block->StartBlock(r->offset);
+  }
 }
 
 void TableBuilder::WriteBlock(BlockBuilder* block, BlockHandle* handle) {
@@ -152,6 +167,15 @@ void TableBuilder::WriteBlock(BlockBuilder* block, BlockHandle* handle) {
       break;
     }
   }
+  WriteRawBlock(block_contents, type, handle);
+  r->compressed_output.clear();
+  block->Reset();
+}
+
+void TableBuilder::WriteRawBlock(const Slice& block_contents,
+                                 CompressionType type,
+                                 BlockHandle* handle) {
+  Rep* r = rep_;
   handle->set_offset(r->offset);
   handle->set_size(block_contents.size());
   r->status = r->file->Append(block_contents);
@@ -166,8 +190,6 @@ void TableBuilder::WriteBlock(BlockBuilder* block, BlockHandle* handle) {
       r->offset += block_contents.size() + kBlockTrailerSize;
     }
   }
-  r->compressed_output.clear();
-  block->Reset();
 }
 
 Status TableBuilder::status() const {
@@ -179,13 +201,32 @@ Status TableBuilder::Finish() {
   Flush();
   assert(!r->closed);
   r->closed = true;
-  BlockHandle metaindex_block_handle;
-  BlockHandle index_block_handle;
+
+  BlockHandle filter_block_handle, metaindex_block_handle, index_block_handle;
+
+  // Write filter block
+  if (ok() && r->filter_block != NULL) {
+    WriteRawBlock(r->filter_block->Finish(), kNoCompression,
+                  &filter_block_handle);
+  }
+
+  // Write metaindex block
   if (ok()) {
     BlockBuilder meta_index_block(&r->options);
+    if (r->filter_block != NULL) {
+      // Add mapping from "filter.Name" to location of filter data
+      std::string key = "filter.";
+      key.append(r->options.filter_policy->Name());
+      std::string handle_encoding;
+      filter_block_handle.EncodeTo(&handle_encoding);
+      meta_index_block.Add(key, handle_encoding);
+    }
+
     // TODO(postrelease): Add stats and other meta blocks
     WriteBlock(&meta_index_block, &metaindex_block_handle);
   }
+
+  // Write index block
   if (ok()) {
     if (r->pending_index_entry) {
       r->options.comparator->FindShortSuccessor(&r->last_key);
@@ -196,6 +237,8 @@ Status TableBuilder::Finish() {
     }
     WriteBlock(&r->index_block, &index_block_handle);
   }
+
+  // Write footer
   if (ok()) {
     Footer footer;
     footer.set_metaindex_handle(metaindex_block_handle);

table/table_test.cc

@@ -168,8 +168,6 @@ class Constructor {
   // Construct the data structure from the data in "data"
   virtual Status FinishImpl(const Options& options, const KVMap& data) = 0;
 
-  virtual size_t NumBytes() const = 0;
-
   virtual Iterator* NewIterator() const = 0;
 
   virtual const KVMap& data() { return data_; }
@@ -185,7 +183,6 @@ class BlockConstructor: public Constructor {
   explicit BlockConstructor(const Comparator* cmp)
       : Constructor(cmp),
         comparator_(cmp),
-        block_size_(-1),
         block_(NULL) { }
   ~BlockConstructor() {
     delete block_;
@@ -201,22 +198,21 @@ class BlockConstructor: public Constructor {
       builder.Add(it->first, it->second);
     }
     // Open the block
-    Slice block_data = builder.Finish();
-    block_size_ = block_data.size();
-    char* block_data_copy = new char[block_size_];
-    memcpy(block_data_copy, block_data.data(), block_size_);
-    block_ = new Block(block_data_copy, block_size_, true /* take ownership */);
+    data_ = builder.Finish().ToString();
+    BlockContents contents;
+    contents.data = data_;
+    contents.cachable = false;
+    contents.heap_allocated = false;
+    block_ = new Block(contents);
     return Status::OK();
   }
-  virtual size_t NumBytes() const { return block_size_; }
-
   virtual Iterator* NewIterator() const {
     return block_->NewIterator(comparator_);
   }
 
  private:
   const Comparator* comparator_;
-  int block_size_;
+  std::string data_;
   Block* block_;
 
   BlockConstructor();
@@ -253,7 +249,6 @@ class TableConstructor: public Constructor {
     table_options.comparator = options.comparator;
     return Table::Open(table_options, source_, sink.contents().size(), &table_);
   }
-  virtual size_t NumBytes() const { return source_->Size(); }
 
   virtual Iterator* NewIterator() const {
     return table_->NewIterator(ReadOptions());
@@ -342,10 +337,6 @@ class MemTableConstructor: public Constructor {
     }
     return Status::OK();
   }
-  virtual size_t NumBytes() const {
-    return memtable_->ApproximateMemoryUsage();
-  }
-
   virtual Iterator* NewIterator() const {
     return new KeyConvertingIterator(memtable_->NewIterator());
   }
@@ -379,13 +370,6 @@ class DBConstructor: public Constructor {
     }
     return Status::OK();
   }
-  virtual size_t NumBytes() const {
-    Range r("", "\xff\xff");
-    uint64_t size;
-    db_->GetApproximateSizes(&r, 1, &size);
-    return size;
-  }
-
   virtual Iterator* NewIterator() const {
     return db_->NewIterator(ReadOptions());
   }
@@ -660,6 +644,36 @@ class Harness {
   Constructor* constructor_;
 };
 
+// Test empty table/block.
+TEST(Harness, Empty) {
+  for (int i = 0; i < kNumTestArgs; i++) {
+    Init(kTestArgList[i]);
+    Random rnd(test::RandomSeed() + 1);
+    Test(&rnd);
+  }
+}
+
+// Special test for a block with no restart entries.  The C++ leveldb
+// code never generates such blocks, but the Java version of leveldb
+// seems to.
+TEST(Harness, ZeroRestartPointsInBlock) {
+  char data[sizeof(uint32_t)];
+  memset(data, 0, sizeof(data));
+  BlockContents contents;
+  contents.data = Slice(data, sizeof(data));
+  contents.cachable = false;
+  contents.heap_allocated = false;
+  Block block(contents);
+  Iterator* iter = block.NewIterator(BytewiseComparator());
+  iter->SeekToFirst();
+  ASSERT_TRUE(!iter->Valid());
+  iter->SeekToLast();
+  ASSERT_TRUE(!iter->Valid());
+  iter->Seek("foo");
+  ASSERT_TRUE(!iter->Valid());
+  delete iter;
+}
+
 // Test the empty key
 TEST(Harness, SimpleEmptyKey) {
   for (int i = 0; i < kNumTestArgs; i++) {
@@ -809,7 +823,7 @@ TEST(TableTest, ApproximateOffsetOfPlain) {
   ASSERT_TRUE(Between(c.ApproximateOffsetOf("k05"),  210000, 211000));
   ASSERT_TRUE(Between(c.ApproximateOffsetOf("k06"),  510000, 511000));
   ASSERT_TRUE(Between(c.ApproximateOffsetOf("k07"),  510000, 511000));
-  ASSERT_TRUE(Between(c.ApproximateOffsetOf("xyz"),  610000, 611000));
+  ASSERT_TRUE(Between(c.ApproximateOffsetOf("xyz"),  610000, 612000));
 
 }
 

util/arena.cc

@@ -40,7 +40,7 @@ char* Arena::AllocateFallback(size_t bytes) {
 }
 
 char* Arena::AllocateAligned(size_t bytes) {
-  const int align = sizeof(void*);    // We'll align to pointer size
+  const int align = (sizeof(void*) > 8) ? sizeof(void*) : 8;
   assert((align & (align-1)) == 0);   // Pointer size should be a power of 2
   size_t current_mod = reinterpret_cast<uintptr_t>(alloc_ptr_) & (align-1);
   size_t slop = (current_mod == 0 ? 0 : align - current_mod);

util/arena.h

@@ -5,9 +5,9 @@
 #ifndef STORAGE_LEVELDB_UTIL_ARENA_H_
 #define STORAGE_LEVELDB_UTIL_ARENA_H_
 
-#include <cstddef>
 #include <vector>
 #include <assert.h>
+#include <stddef.h>
 #include <stdint.h>
 
 namespace leveldb {

util/arena_test.cc

@@ -40,7 +40,7 @@ TEST(ArenaTest, Simple) {
       r = arena.Allocate(s);
     }
 
-    for (int b = 0; b < s; b++) {
+    for (size_t b = 0; b < s; b++) {
       // Fill the "i"th allocation with a known bit pattern
       r[b] = i % 256;
     }
@@ -51,10 +51,10 @@ TEST(ArenaTest, Simple) {
       ASSERT_LE(arena.MemoryUsage(), bytes * 1.10);
     }
   }
-  for (int i = 0; i < allocated.size(); i++) {
+  for (size_t i = 0; i < allocated.size(); i++) {
     size_t num_bytes = allocated[i].first;
     const char* p = allocated[i].second;
-    for (int b = 0; b < num_bytes; b++) {
+    for (size_t b = 0; b < num_bytes; b++) {
       // Check the "i"th allocation for the known bit pattern
       ASSERT_EQ(int(p[b]) & 0xff, i % 256);
     }

util/bloom.cc

@@ -0,0 +1,95 @@
+// Copyright (c) 2012 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 "leveldb/filter_policy.h"
+
+#include "leveldb/slice.h"
+#include "util/hash.h"
+
+namespace leveldb {
+
+namespace {
+static uint32_t BloomHash(const Slice& key) {
+  return Hash(key.data(), key.size(), 0xbc9f1d34);
+}
+
+class BloomFilterPolicy : public FilterPolicy {
+ private:
+  size_t bits_per_key_;
+  size_t k_;
+
+ public:
+  explicit BloomFilterPolicy(int bits_per_key)
+      : bits_per_key_(bits_per_key) {
+    // We intentionally round down to reduce probing cost a little bit
+    k_ = static_cast<size_t>(bits_per_key * 0.69);  // 0.69 =~ ln(2)
+    if (k_ < 1) k_ = 1;
+    if (k_ > 30) k_ = 30;
+  }
+
+  virtual const char* Name() const {
+    return "leveldb.BuiltinBloomFilter";
+  }
+
+  virtual void CreateFilter(const Slice* keys, int n, std::string* dst) const {
+    // Compute bloom filter size (in both bits and bytes)
+    size_t bits = n * bits_per_key_;
+
+    // For small n, we can see a very high false positive rate.  Fix it
+    // by enforcing a minimum bloom filter length.
+    if (bits < 64) bits = 64;
+
+    size_t bytes = (bits + 7) / 8;
+    bits = bytes * 8;
+
+    const size_t init_size = dst->size();
+    dst->resize(init_size + bytes, 0);
+    dst->push_back(static_cast<char>(k_));  // Remember # of probes in filter
+    char* array = &(*dst)[init_size];
+    for (size_t i = 0; i < n; i++) {
+      // Use double-hashing to generate a sequence of hash values.
+      // See analysis in [Kirsch,Mitzenmacher 2006].
+      uint32_t h = BloomHash(keys[i]);
+      const uint32_t delta = (h >> 17) | (h << 15);  // Rotate right 17 bits
+      for (size_t j = 0; j < k_; j++) {
+        const uint32_t bitpos = h % bits;
+        array[bitpos/8] |= (1 << (bitpos % 8));
+        h += delta;
+      }
+    }
+  }
+
+  virtual bool KeyMayMatch(const Slice& key, const Slice& bloom_filter) const {
+    const size_t len = bloom_filter.size();
+    if (len < 2) return false;
+
+    const char* array = bloom_filter.data();
+    const size_t bits = (len - 1) * 8;
+
+    // Use the encoded k so that we can read filters generated by
+    // bloom filters created using different parameters.
+    const size_t k = array[len-1];
+    if (k > 30) {
+      // Reserved for potentially new encodings for short bloom filters.
+      // Consider it a match.
+      return true;
+    }
+
+    uint32_t h = BloomHash(key);
+    const uint32_t delta = (h >> 17) | (h << 15);  // Rotate right 17 bits
+    for (size_t j = 0; j < k; j++) {
+      const uint32_t bitpos = h % bits;
+      if ((array[bitpos/8] & (1 << (bitpos % 8))) == 0) return false;
+      h += delta;
+    }
+    return true;
+  }
+};
+}
+
+const FilterPolicy* NewBloomFilterPolicy(int bits_per_key) {
+  return new BloomFilterPolicy(bits_per_key);
+}
+
+}  // namespace leveldb

util/bloom_test.cc

@@ -0,0 +1,161 @@
+// Copyright (c) 2012 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 "leveldb/filter_policy.h"
+
+#include "util/coding.h"
+#include "util/logging.h"
+#include "util/testharness.h"
+#include "util/testutil.h"
+
+namespace leveldb {
+
+static const int kVerbose = 1;
+
+static Slice Key(int i, char* buffer) {
+  EncodeFixed32(buffer, i);
+  return Slice(buffer, sizeof(uint32_t));
+}
+
+class BloomTest {
+ private:
+  const FilterPolicy* policy_;
+  std::string filter_;
+  std::vector<std::string> keys_;
+
+ public:
+  BloomTest() : policy_(NewBloomFilterPolicy(10)) { }
+
+  ~BloomTest() {
+    delete policy_;
+  }
+
+  void Reset() {
+    keys_.clear();
+    filter_.clear();
+  }
+
+  void Add(const Slice& s) {
+    keys_.push_back(s.ToString());
+  }
+
+  void Build() {
+    std::vector<Slice> key_slices;
+    for (size_t i = 0; i < keys_.size(); i++) {
+      key_slices.push_back(Slice(keys_[i]));
+    }
+    filter_.clear();
+    policy_->CreateFilter(&key_slices[0], key_slices.size(), &filter_);
+    keys_.clear();
+    if (kVerbose >= 2) DumpFilter();
+  }
+
+  size_t FilterSize() const {
+    return filter_.size();
+  }
+
+  void DumpFilter() {
+    fprintf(stderr, "F(");
+    for (size_t i = 0; i+1 < filter_.size(); i++) {
+      const unsigned int c = static_cast<unsigned int>(filter_[i]);
+      for (int j = 0; j < 8; j++) {
+        fprintf(stderr, "%c", (c & (1 <<j)) ? '1' : '.');
+      }
+    }
+    fprintf(stderr, ")\n");
+  }
+
+  bool Matches(const Slice& s) {
+    if (!keys_.empty()) {
+      Build();
+    }
+    return policy_->KeyMayMatch(s, filter_);
+  }
+
+  double FalsePositiveRate() {
+    char buffer[sizeof(int)];
+    int result = 0;
+    for (int i = 0; i < 10000; i++) {
+      if (Matches(Key(i + 1000000000, buffer))) {
+        result++;
+      }
+    }
+    return result / 10000.0;
+  }
+};
+
+TEST(BloomTest, EmptyFilter) {
+  ASSERT_TRUE(! Matches("hello"));
+  ASSERT_TRUE(! Matches("world"));
+}
+
+TEST(BloomTest, Small) {
+  Add("hello");
+  Add("world");
+  ASSERT_TRUE(Matches("hello"));
+  ASSERT_TRUE(Matches("world"));
+  ASSERT_TRUE(! Matches("x"));
+  ASSERT_TRUE(! Matches("foo"));
+}
+
+static int NextLength(int length) {
+  if (length < 10) {
+    length += 1;
+  } else if (length < 100) {
+    length += 10;
+  } else if (length < 1000) {
+    length += 100;
+  } else {
+    length += 1000;
+  }
+  return length;
+}
+
+TEST(BloomTest, VaryingLengths) {
+  char buffer[sizeof(int)];
+
+  // Count number of filters that significantly exceed the false positive rate
+  int mediocre_filters = 0;
+  int good_filters = 0;
+
+  for (int length = 1; length <= 10000; length = NextLength(length)) {
+    Reset();
+    for (int i = 0; i < length; i++) {
+      Add(Key(i, buffer));
+    }
+    Build();
+
+    ASSERT_LE(FilterSize(), static_cast<size_t>((length * 10 / 8) + 40))
+        << length;
+
+    // All added keys must match
+    for (int i = 0; i < length; i++) {
+      ASSERT_TRUE(Matches(Key(i, buffer)))
+          << "Length " << length << "; key " << i;
+    }
+
+    // Check false positive rate
+    double rate = FalsePositiveRate();
+    if (kVerbose >= 1) {
+      fprintf(stderr, "False positives: %5.2f%% @ length = %6d ; bytes = %6d\n",
+              rate*100.0, length, static_cast<int>(FilterSize()));
+    }
+    ASSERT_LE(rate, 0.02);   // Must not be over 2%
+    if (rate > 0.0125) mediocre_filters++;  // Allowed, but not too often
+    else good_filters++;
+  }
+  if (kVerbose >= 1) {
+    fprintf(stderr, "Filters: %d good, %d mediocre\n",
+            good_filters, mediocre_filters);
+  }
+  ASSERT_LE(mediocre_filters, good_filters/5);
+}
+
+// Different bits-per-byte
+
+}  // namespace leveldb
+
+int main(int argc, char** argv) {
+  return leveldb::test::RunAllTests();
+}

util/cache.cc

@@ -116,7 +116,6 @@ class HandleTable {
       LRUHandle* h = list_[i];
       while (h != NULL) {
         LRUHandle* next = h->next_hash;
-        Slice key = h->key();
         uint32_t hash = h->hash;
         LRUHandle** ptr = &new_list[hash & (new_length - 1)];
         h->next_hash = *ptr;
@@ -160,7 +159,6 @@ class LRUCache {
   // mutex_ protects the following state.
   port::Mutex mutex_;
   size_t usage_;
-  uint64_t last_id_;
 
   // Dummy head of LRU list.
   // lru.prev is newest entry, lru.next is oldest entry.
@@ -170,8 +168,7 @@ class LRUCache {
 };
 
 LRUCache::LRUCache()
-    : usage_(0),
-      last_id_(0) {
+    : usage_(0) {
   // Make empty circular linked list
   lru_.next = &lru_;
   lru_.prev = &lru_;

util/coding.cc

@@ -7,29 +7,29 @@
 namespace leveldb {
 
 void EncodeFixed32(char* buf, uint32_t value) {
-#if __BYTE_ORDER == __LITTLE_ENDIAN
-  memcpy(buf, &value, sizeof(value));
-#else
-  buf[0] = value & 0xff;
-  buf[1] = (value >> 8) & 0xff;
-  buf[2] = (value >> 16) & 0xff;
-  buf[3] = (value >> 24) & 0xff;
-#endif
+  if (port::kLittleEndian) {
+    memcpy(buf, &value, sizeof(value));
+  } else {
+    buf[0] = value & 0xff;
+    buf[1] = (value >> 8) & 0xff;
+    buf[2] = (value >> 16) & 0xff;
+    buf[3] = (value >> 24) & 0xff;
+  }
 }
 
 void EncodeFixed64(char* buf, uint64_t value) {
-#if __BYTE_ORDER == __LITTLE_ENDIAN
-  memcpy(buf, &value, sizeof(value));
-#else
-  buf[0] = value & 0xff;
-  buf[1] = (value >> 8) & 0xff;
-  buf[2] = (value >> 16) & 0xff;
-  buf[3] = (value >> 24) & 0xff;
-  buf[4] = (value >> 32) & 0xff;
-  buf[5] = (value >> 40) & 0xff;
-  buf[6] = (value >> 48) & 0xff;
-  buf[7] = (value >> 56) & 0xff;
-#endif
+  if (port::kLittleEndian) {
+    memcpy(buf, &value, sizeof(value));
+  } else {
+    buf[0] = value & 0xff;
+    buf[1] = (value >> 8) & 0xff;
+    buf[2] = (value >> 16) & 0xff;
+    buf[3] = (value >> 24) & 0xff;
+    buf[4] = (value >> 32) & 0xff;
+    buf[5] = (value >> 40) & 0xff;
+    buf[6] = (value >> 48) & 0xff;
+    buf[7] = (value >> 56) & 0xff;
+  }
 }
 
 void PutFixed32(std::string* dst, uint32_t value) {

util/coding_test.cc

@@ -51,6 +51,29 @@ TEST(Coding, Fixed64) {
   }
 }
 
+// Test that encoding routines generate little-endian encodings
+TEST(Coding, EncodingOutput) {
+  std::string dst;
+  PutFixed32(&dst, 0x04030201);
+  ASSERT_EQ(4, dst.size());
+  ASSERT_EQ(0x01, static_cast<int>(dst[0]));
+  ASSERT_EQ(0x02, static_cast<int>(dst[1]));
+  ASSERT_EQ(0x03, static_cast<int>(dst[2]));
+  ASSERT_EQ(0x04, static_cast<int>(dst[3]));
+
+  dst.clear();
+  PutFixed64(&dst, 0x0807060504030201ull);
+  ASSERT_EQ(8, dst.size());
+  ASSERT_EQ(0x01, static_cast<int>(dst[0]));
+  ASSERT_EQ(0x02, static_cast<int>(dst[1]));
+  ASSERT_EQ(0x03, static_cast<int>(dst[2]));
+  ASSERT_EQ(0x04, static_cast<int>(dst[3]));
+  ASSERT_EQ(0x05, static_cast<int>(dst[4]));
+  ASSERT_EQ(0x06, static_cast<int>(dst[5]));
+  ASSERT_EQ(0x07, static_cast<int>(dst[6]));
+  ASSERT_EQ(0x08, static_cast<int>(dst[7]));
+}
+
 TEST(Coding, Varint32) {
   std::string s;
   for (uint32_t i = 0; i < (32 * 32); i++) {
@@ -86,16 +109,16 @@ TEST(Coding, Varint64) {
     values.push_back(power);
     values.push_back(power-1);
     values.push_back(power+1);
-  };
+  }
 
   std::string s;
-  for (int i = 0; i < values.size(); i++) {
+  for (size_t i = 0; i < values.size(); i++) {
     PutVarint64(&s, values[i]);
   }
 
   const char* p = s.data();
   const char* limit = p + s.size();
-  for (int i = 0; i < values.size(); i++) {
+  for (size_t i = 0; i < values.size(); i++) {
     ASSERT_TRUE(p < limit);
     uint64_t actual;
     const char* start = p;
@@ -120,7 +143,7 @@ TEST(Coding, Varint32Truncation) {
   std::string s;
   PutVarint32(&s, large_value);
   uint32_t result;
-  for (int len = 0; len < s.size() - 1; len++) {
+  for (size_t len = 0; len < s.size() - 1; len++) {
     ASSERT_TRUE(GetVarint32Ptr(s.data(), s.data() + len, &result) == NULL);
   }
   ASSERT_TRUE(GetVarint32Ptr(s.data(), s.data() + s.size(), &result) != NULL);
@@ -139,7 +162,7 @@ TEST(Coding, Varint64Truncation) {
   std::string s;
   PutVarint64(&s, large_value);
   uint64_t result;
-  for (int len = 0; len < s.size() - 1; len++) {
+  for (size_t len = 0; len < s.size() - 1; len++) {
     ASSERT_TRUE(GetVarint64Ptr(s.data(), s.data() + len, &result) == NULL);
   }
   ASSERT_TRUE(GetVarint64Ptr(s.data(), s.data() + s.size(), &result) != NULL);

util/comparator.cc

@@ -6,6 +6,7 @@
 #include <stdint.h>
 #include "leveldb/comparator.h"
 #include "leveldb/slice.h"
+#include "port/port.h"
 #include "util/logging.h"
 
 namespace leveldb {
@@ -65,11 +66,15 @@ class BytewiseComparatorImpl : public Comparator {
 };
 }  // namespace
 
-// Intentionally not destroyed to prevent destructor racing
-// with background threads.
-static const Comparator* bytewise = new BytewiseComparatorImpl;
+static port::OnceType once = LEVELDB_ONCE_INIT;
+static const Comparator* bytewise;
+
+static void InitModule() {
+  bytewise = new BytewiseComparatorImpl;
+}
 
 const Comparator* BytewiseComparator() {
+  port::InitOnce(&once, InitModule);
   return bytewise;
 }
 

util/env_posix.cc

@@ -3,6 +3,7 @@
 // found in the LICENSE file. See the AUTHORS file for names of contributors.
 
 #include <deque>
+#include <set>
 #include <dirent.h>
 #include <errno.h>
 #include <fcntl.h>
@@ -23,6 +24,7 @@
 #include "leveldb/slice.h"
 #include "port/port.h"
 #include "util/logging.h"
+#include "util/mutexlock.h"
 #include "util/posix_logger.h"
 
 namespace leveldb {
@@ -90,18 +92,75 @@ class PosixRandomAccessFile: public RandomAccessFile {
   }
 };
 
+// Helper class to limit mmap file usage so that we do not end up
+// running out virtual memory or running into kernel performance
+// problems for very large databases.
+class MmapLimiter {
+ public:
+  // Up to 1000 mmaps for 64-bit binaries; none for smaller pointer sizes.
+  MmapLimiter() {
+    SetAllowed(sizeof(void*) >= 8 ? 1000 : 0);
+  }
+
+  // If another mmap slot is available, acquire it and return true.
+  // Else return false.
+  bool Acquire() {
+    if (GetAllowed() <= 0) {
+      return false;
+    }
+    MutexLock l(&mu_);
+    intptr_t x = GetAllowed();
+    if (x <= 0) {
+      return false;
+    } else {
+      SetAllowed(x - 1);
+      return true;
+    }
+  }
+
+  // Release a slot acquired by a previous call to Acquire() that returned true.
+  void Release() {
+    MutexLock l(&mu_);
+    SetAllowed(GetAllowed() + 1);
+  }
+
+ private:
+  port::Mutex mu_;
+  port::AtomicPointer allowed_;
+
+  intptr_t GetAllowed() const {
+    return reinterpret_cast<intptr_t>(allowed_.Acquire_Load());
+  }
+
+  // REQUIRES: mu_ must be held
+  void SetAllowed(intptr_t v) {
+    allowed_.Release_Store(reinterpret_cast<void*>(v));
+  }
+
+  MmapLimiter(const MmapLimiter&);
+  void operator=(const MmapLimiter&);
+};
+
 // mmap() based random-access
 class PosixMmapReadableFile: public RandomAccessFile {
  private:
   std::string filename_;
   void* mmapped_region_;
   size_t length_;
+  MmapLimiter* limiter_;
 
  public:
   // base[0,length-1] contains the mmapped contents of the file.
-  PosixMmapReadableFile(const std::string& fname, void* base, size_t length)
-      : filename_(fname), mmapped_region_(base), length_(length) { }
-  virtual ~PosixMmapReadableFile() { munmap(mmapped_region_, length_); }
+  PosixMmapReadableFile(const std::string& fname, void* base, size_t length,
+                        MmapLimiter* limiter)
+      : filename_(fname), mmapped_region_(base), length_(length),
+        limiter_(limiter) {
+  }
+
+  virtual ~PosixMmapReadableFile() {
+    munmap(mmapped_region_, length_);
+    limiter_->Release();
+  }
 
   virtual Status Read(uint64_t offset, size_t n, Slice* result,
                       char* scratch) const {
@@ -116,172 +175,83 @@ class PosixMmapReadableFile: public RandomAccessFile {
   }
 };
 
-// We preallocate up to an extra megabyte and use memcpy to append new
-// data to the file.  This is safe since we either properly close the
-// file before reading from it, or for log files, the reading code
-// knows enough to skip zero suffixes.
-class PosixMmapFile : public WritableFile {
+class PosixWritableFile : public WritableFile {
  private:
   std::string filename_;
-  int fd_;
-  size_t page_size_;
-  size_t map_size_;       // How much extra memory to map at a time
-  char* base_;            // The mapped region
-  char* limit_;           // Limit of the mapped region
-  char* dst_;             // Where to write next  (in range [base_,limit_])
-  char* last_sync_;       // Where have we synced up to
-  uint64_t file_offset_;  // Offset of base_ in file
-
-  // Have we done an munmap of unsynced data?
-  bool pending_sync_;
-
-  // Roundup x to a multiple of y
-  static size_t Roundup(size_t x, size_t y) {
-    return ((x + y - 1) / y) * y;
-  }
-
-  size_t TruncateToPageBoundary(size_t s) {
-    s -= (s & (page_size_ - 1));
-    assert((s % page_size_) == 0);
-    return s;
-  }
-
-  bool UnmapCurrentRegion() {
-    bool result = true;
-    if (base_ != NULL) {
-      if (last_sync_ < limit_) {
-        // Defer syncing this data until next Sync() call, if any
-        pending_sync_ = true;
-      }
-      if (munmap(base_, limit_ - base_) != 0) {
-        result = false;
-      }
-      file_offset_ += limit_ - base_;
-      base_ = NULL;
-      limit_ = NULL;
-      last_sync_ = NULL;
-      dst_ = NULL;
-
-      // Increase the amount we map the next time, but capped at 1MB
-      if (map_size_ < (1<<20)) {
-        map_size_ *= 2;
-      }
-    }
-    return result;
-  }
-
-  bool MapNewRegion() {
-    assert(base_ == NULL);
-    if (ftruncate(fd_, file_offset_ + map_size_) < 0) {
-      return false;
-    }
-    void* ptr = mmap(NULL, map_size_, PROT_READ | PROT_WRITE, MAP_SHARED,
-                     fd_, file_offset_);
-    if (ptr == MAP_FAILED) {
-      return false;
-    }
-    base_ = reinterpret_cast<char*>(ptr);
-    limit_ = base_ + map_size_;
-    dst_ = base_;
-    last_sync_ = base_;
-    return true;
-  }
+  FILE* file_;
 
  public:
-  PosixMmapFile(const std::string& fname, int fd, size_t page_size)
-      : filename_(fname),
-        fd_(fd),
-        page_size_(page_size),
-        map_size_(Roundup(65536, page_size)),
-        base_(NULL),
-        limit_(NULL),
-        dst_(NULL),
-        last_sync_(NULL),
-        file_offset_(0),
-        pending_sync_(false) {
-    assert((page_size & (page_size - 1)) == 0);
-  }
+  PosixWritableFile(const std::string& fname, FILE* f)
+      : filename_(fname), file_(f) { }
 
-
-  ~PosixMmapFile() {
-    if (fd_ >= 0) {
-      PosixMmapFile::Close();
+  ~PosixWritableFile() {
+    if (file_ != NULL) {
+      // Ignoring any potential errors
+      fclose(file_);
     }
   }
 
   virtual Status Append(const Slice& data) {
-    const char* src = data.data();
-    size_t left = data.size();
-    while (left > 0) {
-      assert(base_ <= dst_);
-      assert(dst_ <= limit_);
-      size_t avail = limit_ - dst_;
-      if (avail == 0) {
-        if (!UnmapCurrentRegion() ||
-            !MapNewRegion()) {
-          return IOError(filename_, errno);
-        }
-      }
-
-      size_t n = (left <= avail) ? left : avail;
-      memcpy(dst_, src, n);
-      dst_ += n;
-      src += n;
-      left -= n;
+    size_t r = fwrite_unlocked(data.data(), 1, data.size(), file_);
+    if (r != data.size()) {
+      return IOError(filename_, errno);
     }
     return Status::OK();
   }
 
   virtual Status Close() {
-    Status s;
-    size_t unused = limit_ - dst_;
-    if (!UnmapCurrentRegion()) {
-      s = IOError(filename_, errno);
-    } else if (unused > 0) {
-      // Trim the extra space at the end of the file
-      if (ftruncate(fd_, file_offset_ - unused) < 0) {
-        s = IOError(filename_, errno);
-      }
+    Status result;
+    if (fclose(file_) != 0) {
+      result = IOError(filename_, errno);
     }
-
-    if (close(fd_) < 0) {
-      if (s.ok()) {
-        s = IOError(filename_, errno);
-      }
-    }
-
-    fd_ = -1;
-    base_ = NULL;
-    limit_ = NULL;
-    return s;
+    file_ = NULL;
+    return result;
   }
 
   virtual Status Flush() {
+    if (fflush_unlocked(file_) != 0) {
+      return IOError(filename_, errno);
+    }
     return Status::OK();
   }
 
-  virtual Status Sync() {
+  Status SyncDirIfManifest() {
+    const char* f = filename_.c_str();
+    const char* sep = strrchr(f, '/');
+    Slice basename;
+    std::string dir;
+    if (sep == NULL) {
+      dir = ".";
+      basename = f;
+    } else {
+      dir = std::string(f, sep - f);
+      basename = sep + 1;
+    }
     Status s;
-
-    if (pending_sync_) {
-      // Some unmapped data was not synced
-      pending_sync_ = false;
-      if (fdatasync(fd_) < 0) {
-        s = IOError(filename_, errno);
+    if (basename.starts_with("MANIFEST")) {
+      int fd = open(dir.c_str(), O_RDONLY);
+      if (fd < 0) {
+        s = IOError(dir, errno);
+      } else {
+        if (fsync(fd) < 0) {
+          s = IOError(dir, errno);
+        }
+        close(fd);
       }
     }
+    return s;
+  }
 
-    if (dst_ > last_sync_) {
-      // Find the beginnings of the pages that contain the first and last
-      // bytes to be synced.
-      size_t p1 = TruncateToPageBoundary(last_sync_ - base_);
-      size_t p2 = TruncateToPageBoundary(dst_ - base_ - 1);
-      last_sync_ = dst_;
-      if (msync(base_ + p1, p2 - p1 + page_size_, MS_SYNC) < 0) {
-        s = IOError(filename_, errno);
-      }
+  virtual Status Sync() {
+    // Ensure new files referred to by the manifest are in the filesystem.
+    Status s = SyncDirIfManifest();
+    if (!s.ok()) {
+      return s;
+    }
+    if (fflush_unlocked(file_) != 0 ||
+        fdatasync(fileno(file_)) != 0) {
+      s = Status::IOError(filename_, strerror(errno));
     }
-
     return s;
   }
 };
@@ -300,6 +270,25 @@ static int LockOrUnlock(int fd, bool lock) {
 class PosixFileLock : public FileLock {
  public:
   int fd_;
+  std::string name_;
+};
+
+// Set of locked files.  We keep a separate set instead of just
+// relying on fcntrl(F_SETLK) since fcntl(F_SETLK) does not provide
+// any protection against multiple uses from the same process.
+class PosixLockTable {
+ private:
+  port::Mutex mu_;
+  std::set<std::string> locked_files_;
+ public:
+  bool Insert(const std::string& fname) {
+    MutexLock l(&mu_);
+    return locked_files_.insert(fname).second;
+  }
+  void Remove(const std::string& fname) {
+    MutexLock l(&mu_);
+    locked_files_.erase(fname);
+  }
 };
 
 class PosixEnv : public Env {
@@ -307,7 +296,7 @@ class PosixEnv : public Env {
   PosixEnv();
   virtual ~PosixEnv() {
     fprintf(stderr, "Destroying Env::Default()\n");
-    exit(1);
+    abort();
   }
 
   virtual Status NewSequentialFile(const std::string& fname,
@@ -329,19 +318,21 @@ class PosixEnv : public Env {
     int fd = open(fname.c_str(), O_RDONLY);
     if (fd < 0) {
       s = IOError(fname, errno);
-    } else if (sizeof(void*) >= 8) {
-      // Use mmap when virtual address-space is plentiful.
+    } else if (mmap_limit_.Acquire()) {
       uint64_t size;
       s = GetFileSize(fname, &size);
       if (s.ok()) {
         void* base = mmap(NULL, size, PROT_READ, MAP_SHARED, fd, 0);
         if (base != MAP_FAILED) {
-          *result = new PosixMmapReadableFile(fname, base, size);
+          *result = new PosixMmapReadableFile(fname, base, size, &mmap_limit_);
         } else {
           s = IOError(fname, errno);
         }
       }
       close(fd);
+      if (!s.ok()) {
+        mmap_limit_.Release();
+      }
     } else {
       *result = new PosixRandomAccessFile(fname, fd);
     }
@@ -351,12 +342,12 @@ class PosixEnv : public Env {
   virtual Status NewWritableFile(const std::string& fname,
                                  WritableFile** result) {
     Status s;
-    const int fd = open(fname.c_str(), O_CREAT | O_RDWR | O_TRUNC, 0644);
-    if (fd < 0) {
+    FILE* f = fopen(fname.c_str(), "w");
+    if (f == NULL) {
       *result = NULL;
       s = IOError(fname, errno);
     } else {
-      *result = new PosixMmapFile(fname, fd, page_size_);
+      *result = new PosixWritableFile(fname, f);
     }
     return s;
   }
@@ -386,7 +377,7 @@ class PosixEnv : public Env {
       result = IOError(fname, errno);
     }
     return result;
-  };
+  }
 
   virtual Status CreateDir(const std::string& name) {
     Status result;
@@ -394,7 +385,7 @@ class PosixEnv : public Env {
       result = IOError(name, errno);
     }
     return result;
-  };
+  }
 
   virtual Status DeleteDir(const std::string& name) {
     Status result;
@@ -402,7 +393,7 @@ class PosixEnv : public Env {
       result = IOError(name, errno);
     }
     return result;
-  };
+  }
 
   virtual Status GetFileSize(const std::string& fname, uint64_t* size) {
     Status s;
@@ -430,12 +421,17 @@ class PosixEnv : public Env {
     int fd = open(fname.c_str(), O_RDWR | O_CREAT, 0644);
     if (fd < 0) {
       result = IOError(fname, errno);
+    } else if (!locks_.Insert(fname)) {
+      close(fd);
+      result = Status::IOError("lock " + fname, "already held by process");
     } else if (LockOrUnlock(fd, true) == -1) {
       result = IOError("lock " + fname, errno);
       close(fd);
+      locks_.Remove(fname);
     } else {
       PosixFileLock* my_lock = new PosixFileLock;
       my_lock->fd_ = fd;
+      my_lock->name_ = fname;
       *lock = my_lock;
     }
     return result;
@@ -447,6 +443,7 @@ class PosixEnv : public Env {
     if (LockOrUnlock(my_lock->fd_, false) == -1) {
       result = IOError("unlock", errno);
     }
+    locks_.Remove(my_lock->name_);
     close(my_lock->fd_);
     delete my_lock;
     return result;
@@ -502,7 +499,7 @@ class PosixEnv : public Env {
   void PthreadCall(const char* label, int result) {
     if (result != 0) {
       fprintf(stderr, "pthread %s: %s\n", label, strerror(result));
-      exit(1);
+      abort();
     }
   }
 
@@ -513,7 +510,6 @@ class PosixEnv : public Env {
     return NULL;
   }
 
-  size_t page_size_;
   pthread_mutex_t mu_;
   pthread_cond_t bgsignal_;
   pthread_t bgthread_;
@@ -523,10 +519,12 @@ class PosixEnv : public Env {
   struct BGItem { void* arg; void (*function)(void*); };
   typedef std::deque<BGItem> BGQueue;
   BGQueue queue_;
+
+  PosixLockTable locks_;
+  MmapLimiter mmap_limit_;
 };
 
-PosixEnv::PosixEnv() : page_size_(getpagesize()),
-                       started_bgthread_(false) {
+PosixEnv::PosixEnv() : started_bgthread_(false) {
   PthreadCall("mutex_init", pthread_mutex_init(&mu_, NULL));
   PthreadCall("cvar_init", pthread_cond_init(&bgsignal_, NULL));
 }

util/filter_policy.cc

@@ -0,0 +1,11 @@
+// Copyright (c) 2012 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 "leveldb/filter_policy.h"
+
+namespace leveldb {
+
+FilterPolicy::~FilterPolicy() { }
+
+}  // namespace leveldb

util/hash.cc

@@ -6,6 +6,13 @@
 #include "util/coding.h"
 #include "util/hash.h"
 
+// The FALLTHROUGH_INTENDED macro can be used to annotate implicit fall-through
+// between switch labels. The real definition should be provided externally.
+// This one is a fallback version for unsupported compilers.
+#ifndef FALLTHROUGH_INTENDED
+#define FALLTHROUGH_INTENDED do { } while (0)
+#endif
+
 namespace leveldb {
 
 uint32_t Hash(const char* data, size_t n, uint32_t seed) {
@@ -28,10 +35,10 @@ uint32_t Hash(const char* data, size_t n, uint32_t seed) {
   switch (limit - data) {
     case 3:
       h += data[2] << 16;
-      // fall through
+      FALLTHROUGH_INTENDED;
     case 2:
       h += data[1] << 8;
-      // fall through
+      FALLTHROUGH_INTENDED;
     case 1:
       h += data[0];
       h *= m;

util/mutexlock.h

@@ -6,6 +6,7 @@
 #define STORAGE_LEVELDB_UTIL_MUTEXLOCK_H_
 
 #include "port/port.h"
+#include "port/thread_annotations.h"
 
 namespace leveldb {
 
@@ -19,12 +20,13 @@ namespace leveldb {
 //     ... some complex code, possibly with multiple return paths ...
 //   }
 
-class MutexLock {
+class SCOPED_LOCKABLE MutexLock {
  public:
-  explicit MutexLock(port::Mutex *mu) : mu_(mu) {
+  explicit MutexLock(port::Mutex *mu) EXCLUSIVE_LOCK_FUNCTION(mu)
+      : mu_(mu)  {
     this->mu_->Lock();
   }
-  ~MutexLock() { this->mu_->Unlock(); }
+  ~MutexLock() UNLOCK_FUNCTION() { this->mu_->Unlock(); }
 
  private:
   port::Mutex *const mu_;

util/options.cc

@@ -21,7 +21,8 @@ Options::Options()
       block_cache(NULL),
       block_size(4096),
       block_restart_interval(16),
-      compression(kSnappyCompression) {
+      compression(kSnappyCompression),
+      filter_policy(NULL) {
 }
 
 

util/random.h

@@ -16,7 +16,12 @@ class Random {
  private:
   uint32_t seed_;
  public:
-  explicit Random(uint32_t s) : seed_(s & 0x7fffffffu) { }
+  explicit Random(uint32_t s) : seed_(s & 0x7fffffffu) {
+    // Avoid bad seeds.
+    if (seed_ == 0 || seed_ == 2147483647L) {
+      seed_ = 1;
+    }
+  }
   uint32_t Next() {
     static const uint32_t M = 2147483647L;   // 2^31-1
     static const uint64_t A = 16807;  // bits 14, 8, 7, 5, 2, 1, 0

util/testharness.cc

@@ -38,7 +38,7 @@ int RunAllTests() {
 
   int num = 0;
   if (tests != NULL) {
-    for (int i = 0; i < tests->size(); i++) {
+    for (size_t i = 0; i < tests->size(); i++) {
       const Test& t = (*tests)[i];
       if (matcher != NULL) {
         std::string name = t.base;

util/testutil.cc

@@ -32,7 +32,7 @@ std::string RandomKey(Random* rnd, int len) {
 
 
 extern Slice CompressibleString(Random* rnd, double compressed_fraction,
-                                int len, std::string* dst) {
+                                size_t len, std::string* dst) {
   int raw = static_cast<int>(len * compressed_fraction);
   if (raw < 1) raw = 1;
   std::string raw_data;

util/testutil.h

@@ -24,7 +24,7 @@ extern std::string RandomKey(Random* rnd, int len);
 // "N*compressed_fraction" bytes and return a Slice that references
 // the generated data.
 extern Slice CompressibleString(Random* rnd, double compressed_fraction,
-                                int len, std::string* dst);
+                                size_t len, std::string* dst);
 
 // A wrapper that allows injection of errors.
 class ErrorEnv : public EnvWrapper {