Disable warning as error

Summary: IterKey set buffer_size_ to a wrong initial value, causing it to always allocate values from heap instead of stack if the key size is smaller. Fix it.

Test Plan: make all check

Reviewers: haobo, ljin

Reviewed By: haobo

CC: igor, dhruba, yhchiang, leveldb

Differential Revision: https://reviews.facebook.net/D18279

Makefile

@@ -44,7 +44,6 @@ else
 	PLATFORM_CCFLAGS += $(JEMALLOC_INCLUDE) -DHAVE_JEMALLOC
 endif
 
-WARNING_FLAGS = -Wall -Werror -Wno-sign-compare
 CFLAGS += $(WARNING_FLAGS) -I. -I./include $(PLATFORM_CCFLAGS) $(OPT)
 CXXFLAGS += $(WARNING_FLAGS) -I. -I./include $(PLATFORM_CXXFLAGS) $(OPT) -Woverloaded-virtual
 
@@ -170,7 +169,7 @@ dbg: $(LIBRARY) $(PROGRAMS)
 # Will also generate shared libraries.
 release:
 	$(MAKE) clean
-	OPT="-DNDEBUG -O2" $(MAKE) all -j32
+	OPT="-DNDEBUG -O2" $(MAKE) $(LIBRARY) db_bench sst_dump ldb -j32
 
 coverage:
 	$(MAKE) clean

build_tools/build_detect_platform

@@ -47,13 +47,7 @@ COMMON_FLAGS="-DROCKSDB_PLATFORM_POSIX"
 if [ -d /mnt/gvfs/third-party -a -z "$CXX" ]; then
     FBCODE_BUILD="true"
     if [ -z "$USE_CLANG" ]; then
-        CENTOS_VERSION=`rpm -q --qf "%{VERSION}" \
+        source $PWD/build_tools/fbcode_config.sh
-          $(rpm -q --whatprovides redhat-release)`
-        if [ "$CENTOS_VERSION" = "6" ]; then
-          source $PWD/build_tools/fbcode.gcc481.sh
-        else
-          source $PWD/build_tools/fbcode.gcc471.sh
-        fi
     else
         source $PWD/build_tools/fbcode.clang31.sh
     fi
@@ -228,7 +222,7 @@ EOF
       int main() {}
 EOF
     if [ "$?" = 0 ]; then
-        COMMON_FLAGS="$COMMON_FLAGS -DGFLAGS"
+        COMMON_FLAGS="$COMMON_FLAGS -DGFLAGS=gflags"
         PLATFORM_LDFLAGS="$PLATFORM_LDFLAGS -lgflags"
     fi
 

build_tools/dependencies.sh

@@ -0,0 +1,19 @@
+# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
+GCC_BASE=/mnt/gvfs/third-party2/gcc/7331085db891a2ef4a88a48a751d834e8d68f4cb/7.x/centos7-native/b2ef2b6
+CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/963d9aeda70cc4779885b1277484fe7544a04e3e/9.0.0/platform007/9e92d53/
+LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/6ace84e956873d53638c738b6f65f3f469cca74c/7.x/platform007/5620abc
+GLIBC_BASE=/mnt/gvfs/third-party2/glibc/192b0f42d63dcf6210d6ceae387b49af049e6e0c/2.26/platform007/f259413
+SNAPPY_BASE=/mnt/gvfs/third-party2/snappy/7f9bdaada18f59bc27ec2b0871eb8a6144343aef/1.1.3/platform007/ca4da3d
+ZLIB_BASE=/mnt/gvfs/third-party2/zlib/2d9f0b9a4274cc21f61272a9e89bdb859bce8f1f/1.2.8/platform007/ca4da3d
+BZIP2_BASE=/mnt/gvfs/third-party2/bzip2/dc49a21c5fceec6456a7a28a94dcd16690af1337/1.0.6/platform007/ca4da3d
+LZ4_BASE=/mnt/gvfs/third-party2/lz4/0f607f8fc442ea7d6b876931b1898bb573d5e5da/1.9.1/platform007/ca4da3d
+ZSTD_BASE=/mnt/gvfs/third-party2/zstd/ca22bc441a4eb709e9e0b1f9fec9750fed7b31c5/1.4.x/platform007/15a3614
+GFLAGS_BASE=/mnt/gvfs/third-party2/gflags/0b9929d2588991c65a57168bf88aff2db87c5d48/2.2.0/platform007/ca4da3d
+JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/c26f08f47ac35fc31da2633b7da92d6b863246eb/master/platform007/c26c002
+NUMA_BASE=/mnt/gvfs/third-party2/numa/3f3fb57a5ccc5fd21c66416c0b83e0aa76a05376/2.0.11/platform007/ca4da3d
+LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/40c73d874898b386a71847f1b99115d93822d11f/1.4/platform007/6f3e0a9
+TBB_BASE=/mnt/gvfs/third-party2/tbb/4ce8e8dba77cdbd81b75d6f0c32fd7a1b76a11ec/2018_U5/platform007/ca4da3d
+KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/fb251ecd2f5ae16f8671f7014c246e52a748fe0b/fb/platform007/da39a3e
+BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/ab9f09bba370e7066cafd4eb59752db93f2e8312/2.29.1/platform007/15a3614
+VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/d42d152a15636529b0861ec493927200ebebca8e/3.15.0/platform007/ca4da3d
+LUA_BASE=/mnt/gvfs/third-party2/lua/f0cd714433206d5139df61659eb7b28b1dea6683/5.3.4/platform007/5007832

build_tools/fbcode.gcc481.sh

@@ -4,74 +4,82 @@
 # fbcode settings.  It uses the latest g++ compiler and also
 # uses jemalloc
 
-TOOLCHAIN_REV=53dc1fe83f84e9145b9ffb81b81aa7f6a49c87cc
+GCC_BASE=/mnt/gvfs/third-party2/gcc/8219ec1bcedf8ad9da05e121e193364de2cc4f61/5.x/centos6-native/c447969
-CENTOS_VERSION=`rpm -q --qf "%{VERSION}" $(rpm -q --whatprovides redhat-release)`
+CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/64d8d58e3d84f8bde7a029763d4f5baf39d0d5b9/stable/centos6-native/6aaf4de
-if [ "$CENTOS_VERSION" = "6" ]; then
+LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/ba9be983c81de7299b59fe71950c664a84dcb5f8/5.x/gcc-5-glibc-2.23/339d858
-  TOOLCHAIN_EXECUTABLES="/mnt/gvfs/third-party/$TOOLCHAIN_REV/centos6-native"
+GLIBC_BASE=/mnt/gvfs/third-party2/glibc/f20197cf3d4bd50339c9777aaa0b2ccadad9e2cb/2.23/gcc-5-glibc-2.23/ca1d1c0
-else
+SNAPPY_BASE=/mnt/gvfs/third-party2/snappy/6427ce8c7496e4ab06c2da81543b94c0de8be3d0/1.1.3/gcc-5-glibc-2.23/9bc6787
-  TOOLCHAIN_EXECUTABLES="/mnt/gvfs/third-party/$TOOLCHAIN_REV/centos5.2-native"
+ZLIB_BASE=/mnt/gvfs/third-party2/zlib/8f1e8b867d26efef93eac2fabbdb2e1d512665d7/1.2.8/gcc-5-glibc-2.23/9bc6787
-fi
+BZIP2_BASE=/mnt/gvfs/third-party2/bzip2/70471c0571559fe0af7db6d7e8860b93a7eadfe1/1.0.6/gcc-5-glibc-2.23/9bc6787
-TOOLCHAIN_LIB_BASE="/mnt/gvfs/third-party/$TOOLCHAIN_REV/gcc-4.8.1-glibc-2.17"
+LZ4_BASE=/mnt/gvfs/third-party2/lz4/453c89d6f0e68cdf1c151c769197fabedad9cac8/r131/gcc-5-glibc-2.23/9bc6787
+ZSTD_BASE=/mnt/gvfs/third-party2/zstd/00a40fa5f8bd2cd0622f2e868552793aef37ccf4/1.3.0/gcc-5-glibc-2.23/03859b5
+GFLAGS_BASE=/mnt/gvfs/third-party2/gflags/47eef08f9acb77de982fbda6047c26d330739538/2.2.0/gcc-5-glibc-2.23/9bc6787
+JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/4414ddc78df8008b35cc4adac23590ad29148584/master/gcc-5-glibc-2.23/d506c82
+NUMA_BASE=/mnt/gvfs/third-party2/numa/9d7ae2693d05d62f9a579cb21e6b717cf257a75d/2.0.11/gcc-5-glibc-2.23/9bc6787
+LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/2b2dd58e3a52ccf2c1d827def59e5f740de0ad15/1.2/gcc-5-glibc-2.23/b443de1
+TBB_BASE=/mnt/gvfs/third-party2/tbb/379addf7ab2468a2b4293b47456cfcd1c9cb318d/4.3/gcc-5-glibc-2.23/9bc6787
+KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/3f68f5fe65a85b7c2d3e66852268fbd1efdb3151/4.0.9-36_fbk5_2933_gd092e3f/gcc-5-glibc-2.23/da39a3e
+BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/b9fab0aec99d9c36408e810b2677e91c12807afd/2.28/centos6-native/da39a3e
+VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/423431d61786b20bcc3bde8972901130cb29e6b3/3.11.0/gcc-5-glibc-2.23/9bc6787
+LUA_BASE=/mnt/gvfs/third-party2/lua/3b0bb3bd9a0f690a069c479fcc0f7424fc7456d2/5.2.3/gcc-5-glibc-2.23/65372bd
 
-# location of libhdfs libraries
-if test "$USE_HDFS"; then
-  JAVA_HOME="/usr/local/jdk-6u22-64"
-  JINCLUDE="-I$JAVA_HOME/include -I$JAVA_HOME/include/linux"
-  GLIBC_RUNTIME_PATH="/usr/local/fbcode/gcc-4.8.1-glibc-2.17"
-  HDFSLIB=" -Wl,--no-whole-archive hdfs/libhdfs.a -L$JAVA_HOME/jre/lib/amd64 "
-  HDFSLIB+=" -L$JAVA_HOME/jre/lib/amd64/server -L$GLIBC_RUNTIME_PATH/lib "
-  HDFSLIB+=" -ldl -lverify -ljava -ljvm "
-fi
 
 # location of libgcc
-LIBGCC_INCLUDE=" -I $TOOLCHAIN_LIB_BASE/libgcc/libgcc-4.8.1/8aac7fc/include"
+LIBGCC_INCLUDE=" -I $LIBGCC_BASE/include"
-LIBGCC_LIBS=" -L $TOOLCHAIN_LIB_BASE/libgcc/libgcc-4.8.1/8aac7fc/libs"
+LIBGCC_LIBS=" -L $LIBGCC_BASE/lib"
 
 # location of glibc
-GLIBC_INCLUDE=" -I $TOOLCHAIN_LIB_BASE/glibc/glibc-2.17/99df8fc/include"
+GLIBC_INCLUDE=" -I $GLIBC_BASE/include"
-GLIBC_LIBS=" -L $TOOLCHAIN_LIB_BASE/glibc/glibc-2.17/99df8fc/lib"
+GLIBC_LIBS=" -L $GLIBC_BASE/lib"
 
 # location of snappy headers and libraries
-SNAPPY_INCLUDE=" -I $TOOLCHAIN_LIB_BASE/snappy/snappy-1.0.3/43d84e2/include"
+SNAPPY_INCLUDE=" -I $SNAPPY_BASE/include/"
-SNAPPY_LIBS=" $TOOLCHAIN_LIB_BASE/snappy/snappy-1.0.3/43d84e2/lib/libsnappy.a"
+SNAPPY_LIBS=" $SNAPPY_BASE/lib/libsnappy.a"
 
 # location of zlib headers and libraries
-ZLIB_INCLUDE=" -I $TOOLCHAIN_LIB_BASE/zlib/zlib-1.2.5/c3f970a/include"
+ZLIB_INCLUDE=" -I $ZLIB_BASE/include/"
-ZLIB_LIBS=" $TOOLCHAIN_LIB_BASE/zlib/zlib-1.2.5/c3f970a/lib/libz.a"
+ZLIB_LIBS=" $ZLIB_BASE/lib/libz.a"
+
 
 # location of bzip headers and libraries
-BZIP_INCLUDE=" -I $TOOLCHAIN_LIB_BASE/bzip2/bzip2-1.0.6/c3f970a/include"
+BZIP_INCLUDE=" -I $BZIP2_BASE/include"
-BZIP_LIBS=" $TOOLCHAIN_LIB_BASE/bzip2/bzip2-1.0.6/c3f970a/lib/libbz2.a"
+BZIP_LIBS=" $BZIP2_BASE/lib/libbz2.a"
+
+LZ4_INCLUDE=" -I $LZ4_BASE/include/"
+LZ4_LIBS=" $LZ4_BASE/lib/liblz4.a"
+
 
 # location of gflags headers and libraries
-GFLAGS_INCLUDE=" -I $TOOLCHAIN_LIB_BASE/gflags/gflags-1.6/c3f970a/include"
+GFLAGS_INCLUDE=" -I $GFLAGS_BASE/include/"
-GFLAGS_LIBS=" $TOOLCHAIN_LIB_BASE/gflags/gflags-1.6/c3f970a/lib/libgflags.a"
+GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags.a"
 
 # location of jemalloc
-JEMALLOC_INCLUDE=" -I $TOOLCHAIN_LIB_BASE/jemalloc/jemalloc-3.4.1/4d53c6f/include/"
+JEMALLOC_INCLUDE=" -I $JEMALLOC_BASE/include/"
-JEMALLOC_LIB=" -Wl,--whole-archive $TOOLCHAIN_LIB_BASE/jemalloc/jemalloc-3.4.1/4d53c6f/lib/libjemalloc.a"
+JEMALLOC_LIB=" -Wl,--whole-archive  $JEMALLOC_BASE/lib/libjemalloc.a"
 
 # use Intel SSE support for checksum calculations
 export USE_SSE=" -msse -msse4.2 "
 
-CC="$TOOLCHAIN_EXECUTABLES/gcc/gcc-4.8.1/cc6c9dc/bin/gcc"
+CC="$GCC_BASE/bin/gcc"
-CXX="$TOOLCHAIN_EXECUTABLES/gcc/gcc-4.8.1/cc6c9dc/bin/g++ $JINCLUDE $SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $GFLAGS_INCLUDE"
+CXX="$GCC_BASE/bin/g++ $JINCLUDE $SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $GFLAGS_INCLUDE"
-AR=$TOOLCHAIN_EXECUTABLES/binutils/binutils-2.21.1/da39a3e/bin/ar
+AR="$BINUTILS_BASE/bin/ar"
-RANLIB=$TOOLCHAIN_EXECUTABLES/binutils/binutils-2.21.1/da39a3e/bin/ranlib
+RANLIB="$BINUTILS_BASE/bin/ranlib"
 
-CFLAGS="-B$TOOLCHAIN_EXECUTABLES/binutils/binutils-2.21.1/da39a3e/bin/gold -m64 -mtune=generic"
+CFLAGS="-B$BINUTILS/bin/gold -m64 -mtune=generic"
 CFLAGS+=" $LIBGCC_INCLUDE $GLIBC_INCLUDE"
 CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_ATOMIC_PRESENT -DROCKSDB_FALLOCATE_PRESENT"
 CFLAGS+=" -DSNAPPY -DGFLAGS -DZLIB -DBZIP2"
 
-EXEC_LDFLAGS="-Wl,--dynamic-linker,/usr/local/fbcode/gcc-4.8.1-glibc-2.17/lib/ld.so"
+EXEC_LDFLAGS+=" -B$BINUTILS_BASE/bin/gold"
-EXEC_LDFLAGS+=" -Wl,--no-whole-archive $TOOLCHAIN_LIB_BASE/libunwind/libunwind-1.0.1/675d945/lib/libunwind.a"
+EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/gcc-5-glibc-2.23/lib/ld.so"
+EXEC_LDFLAGS+=" $LIBUNWIND"
+EXEC_LDFLAGS+=" -Wl,-rpath=/usr/local/fbcode/gcc-5-glibc-2.23/lib"
 EXEC_LDFLAGS+=" $HDFSLIB $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $GFLAGS_LIBS"
 
 PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS "
 
 EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $GFLAGS_LIBS"
 
-VALGRIND_VER="$TOOLCHAIN_LIB_BASE/valgrind/valgrind-3.8.1/c3f970a/bin/"
+VALGRIND_VER="$VALGRIND_BASE/bin/"
 
 export CC CXX AR RANLIB CFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE

build_tools/fbcode_config.sh

@@ -0,0 +1,139 @@
+#!/bin/sh
+#
+# Set environment variables so that we can compile rocksdb using
+# fbcode settings.  It uses the latest g++ and clang compilers and also
+# uses jemalloc
+# Environment variables that change the behavior of this script:
+# PIC_BUILD -- if true, it will only take pic versions of libraries from fbcode. libraries that don't have pic variant will not be included
+
+
+BASEDIR=`dirname $BASH_SOURCE`
+source "$BASEDIR/dependencies.sh"
+
+CFLAGS=""
+
+# libgcc
+LIBGCC_INCLUDE="$LIBGCC_BASE/include/c++/7.3.0"
+LIBGCC_LIBS=" -L $LIBGCC_BASE/lib"
+
+# glibc
+GLIBC_INCLUDE="$GLIBC_BASE/include"
+GLIBC_LIBS=" -L $GLIBC_BASE/lib"
+
+# snappy
+SNAPPY_INCLUDE=" -I $SNAPPY_BASE/include/"
+if test -z $PIC_BUILD; then
+  SNAPPY_LIBS=" $SNAPPY_BASE/lib/libsnappy.a"
+else
+  SNAPPY_LIBS=" $SNAPPY_BASE/lib/libsnappy_pic.a"
+fi
+CFLAGS+=" -DSNAPPY"
+
+if test -z $PIC_BUILD; then
+  # location of zlib headers and libraries
+  ZLIB_INCLUDE=" -I $ZLIB_BASE/include/"
+  ZLIB_LIBS=" $ZLIB_BASE/lib/libz.a"
+  CFLAGS+=" -DZLIB"
+
+  LZ4_INCLUDE=" -I $LZ4_BASE/include/"
+  LZ4_LIBS=" $LZ4_BASE/lib/liblz4.a"
+  CFLAGS+=" -DLZ4"
+fi
+
+# location of gflags headers and libraries
+GFLAGS_INCLUDE=" -I $GFLAGS_BASE/include/"
+if test -z $PIC_BUILD; then
+  GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags.a"
+else
+  GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags_pic.a"
+fi
+CFLAGS+=" -DGFLAGS=gflags"
+
+# location of jemalloc
+JEMALLOC_INCLUDE=" -I $JEMALLOC_BASE/include/"
+JEMALLOC_LIB=" $JEMALLOC_BASE/lib/libjemalloc.a"
+
+if test -z $PIC_BUILD; then
+  # location of numa
+  NUMA_INCLUDE=" -I $NUMA_BASE/include/"
+  NUMA_LIB=" $NUMA_BASE/lib/libnuma.a"
+  CFLAGS+=" -DNUMA"
+
+  # location of libunwind
+  LIBUNWIND="$LIBUNWIND_BASE/lib/libunwind.a"
+fi
+
+# location of TBB
+TBB_INCLUDE=" -isystem $TBB_BASE/include/"
+if test -z $PIC_BUILD; then
+  TBB_LIBS="$TBB_BASE/lib/libtbb.a"
+else
+  TBB_LIBS="$TBB_BASE/lib/libtbb_pic.a"
+fi
+CFLAGS+=" -DTBB"
+
+# use Intel SSE support for checksum calculations
+export USE_SSE=" -msse -msse4.2 "
+
+BINUTILS="$BINUTILS_BASE/bin"
+AR="$BINUTILS/ar"
+
+DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $LZ4_INCLUDE $GFLAGS_INCLUDE"
+
+STDLIBS="-L $GCC_BASE/lib64"
+
+CLANG_BIN="$CLANG_BASE/bin"
+CLANG_LIB="$CLANG_BASE/lib"
+CLANG_SRC="$CLANG_BASE/../../src"
+
+CLANG_ANALYZER="$CLANG_BIN/clang++"
+CLANG_SCAN_BUILD="$CLANG_SRC/llvm/tools/clang/tools/scan-build/bin/scan-build"
+
+if [ -z "$USE_CLANG" ]; then
+  # gcc
+  CC="$GCC_BASE/bin/gcc"
+  CXX="$GCC_BASE/bin/g++"
+
+  CFLAGS+=" -B$BINUTILS/gold"
+  CFLAGS+=" -isystem $LIBGCC_INCLUDE"
+  CFLAGS+=" -isystem $GLIBC_INCLUDE"
+  JEMALLOC=1
+else
+  # clang
+  CLANG_INCLUDE="$CLANG_LIB/clang/stable/include"
+  CC="$CLANG_BIN/clang"
+  CXX="$CLANG_BIN/clang++"
+
+  KERNEL_HEADERS_INCLUDE="$KERNEL_HEADERS_BASE/include"
+
+  CFLAGS+=" -B$BINUTILS/gold -nostdinc -nostdlib"
+  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/7.x "
+  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/7.x/x86_64-facebook-linux "
+  CFLAGS+=" -isystem $GLIBC_INCLUDE"
+  CFLAGS+=" -isystem $LIBGCC_INCLUDE"
+  CFLAGS+=" -isystem $CLANG_INCLUDE"
+  CFLAGS+=" -isystem $KERNEL_HEADERS_INCLUDE/linux "
+  CFLAGS+=" -isystem $KERNEL_HEADERS_INCLUDE "
+  CFLAGS+=" -Wno-expansion-to-defined "
+  CXXFLAGS="-nostdinc++"
+fi
+
+CFLAGS+=" $DEPS_INCLUDE"
+CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE -DROCKSDB_RANGESYNC_PRESENT -DROCKSDB_SCHED_GETCPU_PRESENT -DROCKSDB_SUPPORT_THREAD_LOCAL -DHAVE_SSE42"
+CXXFLAGS+=" $CFLAGS"
+
+EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB $TBB_LIBS"
+EXEC_LDFLAGS+=" -B$BINUTILS/gold"
+EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/platform007/lib/ld.so"
+EXEC_LDFLAGS+=" $LIBUNWIND"
+EXEC_LDFLAGS+=" -Wl,-rpath=/usr/local/fbcode/platform007/lib"
+# required by libtbb
+EXEC_LDFLAGS+=" -ldl"
+
+PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS $STDLIBS -lgcc -lstdc++"
+
+EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $TBB_LIBS"
+
+VALGRIND_VER="$VALGRIND_BASE/bin/"
+
+export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE CLANG_ANALYZER CLANG_SCAN_BUILD 

db/db_bench.cc

@@ -378,7 +378,7 @@ static bool ValidateCompressionLevel(const char* flagname, int32_t value) {
 }
 
 static const bool FLAGS_compression_level_dummy __attribute__((unused)) =
-    google::RegisterFlagValidator(&FLAGS_compression_level,
+    GFLAGS::RegisterFlagValidator(&FLAGS_compression_level,
                                   &ValidateCompressionLevel);
 
 DEFINE_int32(min_level_to_compress, -1, "If non-negative, compression starts"
@@ -535,32 +535,32 @@ DEFINE_string(merge_operator, "", "The merge operator to use with the database."
               " utilities/merge_operators.h");
 
 static const bool FLAGS_soft_rate_limit_dummy __attribute__((unused)) =
-  google::RegisterFlagValidator(&FLAGS_soft_rate_limit,
+  GFLAGS::RegisterFlagValidator(&FLAGS_soft_rate_limit,
                                 &ValidateRateLimit);
 
 static const bool FLAGS_hard_rate_limit_dummy __attribute__((unused)) =
-  google::RegisterFlagValidator(&FLAGS_hard_rate_limit, &ValidateRateLimit);
+  GFLAGS::RegisterFlagValidator(&FLAGS_hard_rate_limit, &ValidateRateLimit);
 
 static const bool FLAGS_prefix_size_dummy __attribute__((unused)) =
-  google::RegisterFlagValidator(&FLAGS_prefix_size, &ValidatePrefixSize);
+  GFLAGS::RegisterFlagValidator(&FLAGS_prefix_size, &ValidatePrefixSize);
 
 static const bool FLAGS_key_size_dummy __attribute__((unused)) =
-  google::RegisterFlagValidator(&FLAGS_key_size, &ValidateKeySize);
+  GFLAGS::RegisterFlagValidator(&FLAGS_key_size, &ValidateKeySize);
 
 static const bool FLAGS_cache_numshardbits_dummy __attribute__((unused)) =
-  google::RegisterFlagValidator(&FLAGS_cache_numshardbits,
+  GFLAGS::RegisterFlagValidator(&FLAGS_cache_numshardbits,
                                 &ValidateCacheNumshardbits);
 
 static const bool FLAGS_readwritepercent_dummy __attribute__((unused)) =
-  google::RegisterFlagValidator(&FLAGS_readwritepercent,
+  GFLAGS::RegisterFlagValidator(&FLAGS_readwritepercent,
                                 &ValidateInt32Percent);
 
 static const bool FLAGS_deletepercent_dummy __attribute__((unused)) =
-  google::RegisterFlagValidator(&FLAGS_deletepercent,
+  GFLAGS::RegisterFlagValidator(&FLAGS_deletepercent,
                                 &ValidateInt32Percent);
 static const bool
   FLAGS_table_cache_numshardbits_dummy __attribute__((unused)) =
-  google::RegisterFlagValidator(&FLAGS_table_cache_numshardbits,
+  GFLAGS::RegisterFlagValidator(&FLAGS_table_cache_numshardbits,
                                 &ValidateTableCacheNumshardbits);
 
 namespace rocksdb {
@@ -2775,9 +2775,9 @@ class Benchmark {
 
 int main(int argc, char** argv) {
   rocksdb::InstallStackTraceHandler();
-  google::SetUsageMessage(std::string("\nUSAGE:\n") + std::string(argv[0]) +
+  GFLAGS::SetUsageMessage(std::string("\nUSAGE:\n") + std::string(argv[0]) +
                           " [OPTIONS]...");
-  google::ParseCommandLineFlags(&argc, &argv, true);
+  GFLAGS::ParseCommandLineFlags(&argc, &argv, true);
 
   FLAGS_compaction_style_e = (rocksdb::CompactionStyle) FLAGS_compaction_style;
   if (FLAGS_statistics) {

db/db_impl.cc

@@ -2540,9 +2540,9 @@ Status DBImpl::ProcessKeyValueCompaction(
   Status status;
   std::string compaction_filter_value;
   ParsedInternalKey ikey;
-  std::string current_user_key;
+  IterKey current_user_key;
   bool has_current_user_key = false;
-  std::vector<char> delete_key;   // for compaction filter
+  IterKey delete_key;
   SequenceNumber last_sequence_for_key __attribute__((unused)) =
     kMaxSequenceNumber;
   SequenceNumber visible_in_snapshot = kMaxSequenceNumber;
@@ -2612,16 +2612,16 @@ Status DBImpl::ProcessKeyValueCompaction(
       // Do not hide error keys
       // TODO: error key stays in db forever? Figure out the intention/rationale
       // v10 error v8 : we cannot hide v8 even though it's pretty obvious.
-      current_user_key.clear();
+      current_user_key.Clear();
       has_current_user_key = false;
       last_sequence_for_key = kMaxSequenceNumber;
       visible_in_snapshot = kMaxSequenceNumber;
     } else {
       if (!has_current_user_key ||
           user_comparator()->Compare(ikey.user_key,
-            Slice(current_user_key)) != 0) {
+            current_user_key.GetKey()) != 0) {
         // First occurrence of this user key
-        current_user_key.assign(ikey.user_key.data(), ikey.user_key.size());
+        current_user_key.SetUserKey(ikey.user_key);
         has_current_user_key = true;
         last_sequence_for_key = kMaxSequenceNumber;
         visible_in_snapshot = kMaxSequenceNumber;
@@ -2642,13 +2642,11 @@ Status DBImpl::ProcessKeyValueCompaction(
                 &compaction_filter_value,
                 &value_changed);
           if (to_delete) {
-            // make a copy of the original key
+            // make a copy of the original key and convert it to a delete
-            delete_key.assign(key.data(), key.data() + key.size());
+            delete_key.SetInternalKey(ExtractUserKey(key), ikey.sequence,
-            // convert it to a delete
+                                      kTypeDeletion);
-            UpdateInternalKey(&delete_key[0], delete_key.size(),
-                ikey.sequence, kTypeDeletion);
             // anchor the key again
-            key = Slice(&delete_key[0], delete_key.size());
+            key = delete_key.GetKey();
             // needed because ikey is backed by key
             ParseInternalKey(key, &ikey);
             // no value associated with delete
@@ -3455,7 +3453,7 @@ Status DBImpl::GetImpl(const ReadOptions& options,
     StartPerfTimer(&from_files_timer);
 
     sv->current->Get(options, lkey, value, &s, &merge_context, &stats,
-                     options_, value_found);
+                     value_found);
     have_stat_update = true;
     BumpPerfTime(&perf_context.get_from_output_files_time, &from_files_timer);
     RecordTick(options_.statistics.get(), MEMTABLE_MISS);
@@ -3559,7 +3557,7 @@ std::vector<Status> DBImpl::MultiGet(const ReadOptions& options,
       // Done
     } else {
       get_version->current->Get(options, lkey, value, &s, &merge_context,
-                                &stats, options_);
+                                &stats);
       have_stat_update = true;
     }
 

db/db_impl_readonly.cc

@@ -64,7 +64,7 @@ Status DBImplReadOnly::Get(const ReadOptions& options,
   } else {
     Version::GetStats stats;
     super_version->current->Get(options, lkey, value, &s, &merge_context,
-                                &stats, options_);
+                                &stats);
   }
   return s;
 }

db/db_iter.cc

@@ -39,71 +39,6 @@ static void DumpInternalIter(Iterator* iter) {
 
 namespace {
 
-class IterLookupKey {
- public:
-  IterLookupKey() : key_(space_), buf_size_(sizeof(space_)), key_size_(0) {}
-
-  ~IterLookupKey() { Clear(); }
-
-  Slice GetKey() const {
-    if (key_ != nullptr) {
-      return Slice(key_, key_size_);
-    } else {
-      return Slice();
-    }
-  }
-
-  bool Valid() const { return key_ != nullptr; }
-
-  void Clear() {
-    if (key_ != nullptr && key_ != space_) {
-      delete[] key_;
-    }
-    key_ = space_;
-    buf_size_ = sizeof(buf_size_);
-  }
-
-  // Enlarge the buffer size if needed based on key_size.
-  // By default, static allocated buffer is used. Once there is a key
-  // larger than the static allocated buffer, another buffer is dynamically
-  // allocated, until a larger key buffer is requested. In that case, we
-  // reallocate buffer and delete the old one.
-  void EnlargeBufferIfNeeded(size_t key_size) {
-    // If size is smaller than buffer size, continue using current buffer,
-    // or the static allocated one, as default
-    if (key_size > buf_size_) {
-      // Need to enlarge the buffer.
-      Clear();
-      key_ = new char[key_size];
-      buf_size_ = key_size;
-    }
-    key_size_ = key_size;
-  }
-
-  void SetUserKey(const Slice& user_key) {
-    size_t size = user_key.size();
-    EnlargeBufferIfNeeded(size);
-    memcpy(key_, user_key.data(), size);
-  }
-
-  void SetInternalKey(const Slice& user_key, SequenceNumber s) {
-    size_t usize = user_key.size();
-    EnlargeBufferIfNeeded(usize + sizeof(uint64_t));
-    memcpy(key_, user_key.data(), usize);
-    EncodeFixed64(key_ + usize, PackSequenceAndType(s, kValueTypeForSeek));
-  }
-
- private:
-  char* key_;
-  size_t buf_size_;
-  size_t key_size_;
-  char space_[32];  // Avoid allocation for short keys
-
-  // No copying allowed
-  IterLookupKey(const IterLookupKey&) = delete;
-  void operator=(const LookupKey&) = delete;
-};
-
 // Memtables and sstables that make the DB representation contain
 // (userkey,seq,type) => uservalue entries.  DBIter
 // combines multiple entries for the same userkey found in the DB
@@ -191,7 +126,7 @@ class DBIter: public Iterator {
   SequenceNumber const sequence_;
 
   Status status_;
-  IterLookupKey saved_key_;   // == current key when direction_==kReverse
+  IterKey saved_key_;   // == current key when direction_==kReverse
   std::string saved_value_;   // == current raw value when direction_==kReverse
   std::string skip_key_;
   Direction direction_;

db/db_test.cc

@@ -266,6 +266,9 @@ class DBTest {
   // Sequence of option configurations to try
   enum OptionConfig {
     kDefault,
+    kBlockBasedTableWithPrefixHashIndex,
+    // TODO(kailiu) figure this out
+    // kBlockBasedTableWithWholeKeyHashIndex,
     kPlainTableFirstBytePrefix,
     kPlainTableAllBytesPrefix,
     kVectorRep,
@@ -302,7 +305,8 @@ class DBTest {
     kSkipDeletesFilterFirst = 1,
     kSkipUniversalCompaction = 2,
     kSkipMergePut = 4,
-    kSkipPlainTable = 8
+    kSkipPlainTable = 8,
+    kSkipHashIndex = 16
   };
 
   DBTest() : option_config_(kDefault),
@@ -343,6 +347,11 @@ class DBTest {
               || option_config_ == kPlainTableFirstBytePrefix)) {
         continue;
       }
+      if ((skip_mask & kSkipPlainTable) &&
+          option_config_ == kBlockBasedTableWithPrefixHashIndex) {
+        continue;
+      }
+
       break;
     }
 
@@ -441,6 +450,21 @@ class DBTest {
       case kInfiniteMaxOpenFiles:
         options.max_open_files = -1;
         break;
+      case kBlockBasedTableWithPrefixHashIndex: {
+        BlockBasedTableOptions table_options;
+        table_options.index_type = BlockBasedTableOptions::kHashSearch;
+        options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+        options.prefix_extractor.reset(NewFixedPrefixTransform(1));
+        break;
+      }
+      // TODO(kailiu) figure out why it's failing and fix
+      // case kBlockBasedTableWithWholeKeyHashIndex: {
+      //   BlockBasedTableOptions table_options;
+      //   table_options.index_type = BlockBasedTableOptions::kHashSearch;
+      //   options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+      //   options.prefix_extractor.reset(NewNoopTransform());
+      //   break;
+      // }
       default:
         break;
     }
@@ -885,13 +909,29 @@ TEST(DBTest, Empty) {
     options.write_buffer_size = 100000;  // Small write buffer
     Reopen(&options);
 
+    std::string num;
+    ASSERT_TRUE(dbfull()->GetProperty(
+        "rocksdb.num-entries-active-mem-table", &num));
+    ASSERT_EQ("0", num);
+
     ASSERT_OK(Put("foo", "v1"));
     ASSERT_EQ("v1", Get("foo"));
+    ASSERT_TRUE(dbfull()->GetProperty(
+        "rocksdb.num-entries-active-mem-table", &num));
+    ASSERT_EQ("1", num);
 
     env_->delay_sstable_sync_.Release_Store(env_);   // Block sync calls
     Put("k1", std::string(100000, 'x'));             // Fill memtable
+    ASSERT_TRUE(dbfull()->GetProperty(
+        "rocksdb.num-entries-active-mem-table", &num));
+    ASSERT_EQ("2", num);
+
     Put("k2", std::string(100000, 'y'));             // Trigger compaction
     ASSERT_EQ("v1", Get("foo"));
+    ASSERT_TRUE(dbfull()->GetProperty(
+        "rocksdb.num-entries-active-mem-table", &num));
+    ASSERT_EQ("1", num);
+
     env_->delay_sstable_sync_.Release_Store(nullptr);   // Release sync calls
   } while (ChangeOptions());
 }
@@ -1251,7 +1291,7 @@ TEST(DBTest, KeyMayExist) {
 
     // KeyMayExist function only checks data in block caches, which is not used
     // by plain table format.
-  } while (ChangeOptions(kSkipPlainTable));
+  } while (ChangeOptions(kSkipPlainTable | kSkipHashIndex));
 }
 
 TEST(DBTest, NonBlockingIteration) {
@@ -2064,6 +2104,9 @@ TEST(DBTest, NumImmutableMemTable) {
     ASSERT_OK(dbfull()->Put(writeOpt, "k1", big_value));
     ASSERT_TRUE(dbfull()->GetProperty("rocksdb.num-immutable-mem-table", &num));
     ASSERT_EQ(num, "0");
+    ASSERT_TRUE(dbfull()->GetProperty(
+        "rocksdb.num-entries-active-mem-table", &num));
+    ASSERT_EQ(num, "1");
     perf_context.Reset();
     Get("k1");
     ASSERT_EQ(1, (int) perf_context.get_from_memtable_count);
@@ -2071,6 +2114,13 @@ TEST(DBTest, NumImmutableMemTable) {
     ASSERT_OK(dbfull()->Put(writeOpt, "k2", big_value));
     ASSERT_TRUE(dbfull()->GetProperty("rocksdb.num-immutable-mem-table", &num));
     ASSERT_EQ(num, "1");
+    ASSERT_TRUE(dbfull()->GetProperty(
+        "rocksdb.num-entries-active-mem-table", &num));
+    ASSERT_EQ(num, "1");
+    ASSERT_TRUE(dbfull()->GetProperty(
+        "rocksdb.num-entries-imm-mem-tables", &num));
+    ASSERT_EQ(num, "1");
+
     perf_context.Reset();
     Get("k1");
     ASSERT_EQ(2, (int) perf_context.get_from_memtable_count);
@@ -2083,6 +2133,12 @@ TEST(DBTest, NumImmutableMemTable) {
                                       &num));
     ASSERT_TRUE(dbfull()->GetProperty("rocksdb.num-immutable-mem-table", &num));
     ASSERT_EQ(num, "2");
+    ASSERT_TRUE(dbfull()->GetProperty(
+       "rocksdb.num-entries-active-mem-table", &num));
+    ASSERT_EQ(num, "1");
+    ASSERT_TRUE(dbfull()->GetProperty(
+        "rocksdb.num-entries-imm-mem-tables", &num));
+    ASSERT_EQ(num, "2");
     perf_context.Reset();
     Get("k2");
     ASSERT_EQ(2, (int) perf_context.get_from_memtable_count);
@@ -5882,7 +5938,8 @@ TEST(DBTest, Randomized) {
       int minimum = 0;
       if (option_config_ == kHashSkipList ||
           option_config_ == kHashLinkList ||
-          option_config_ == kPlainTableFirstBytePrefix) {
+          option_config_ == kPlainTableFirstBytePrefix ||
+          option_config_ == kBlockBasedTableWithPrefixHashIndex) {
         minimum = 1;
       }
       if (p < 45) {                               // Put
@@ -5924,6 +5981,7 @@ TEST(DBTest, Randomized) {
       if ((step % 100) == 0) {
         ASSERT_TRUE(CompareIterators(step, &model, db_, nullptr, nullptr));
         ASSERT_TRUE(CompareIterators(step, &model, db_, model_snap, db_snap));
+
         // Save a snapshot from each DB this time that we'll use next
         // time we compare things, to make sure the current state is
         // preserved with the snapshot

db/dbformat.h

@@ -13,6 +13,7 @@
 #include "rocksdb/db.h"
 #include "rocksdb/filter_policy.h"
 #include "rocksdb/slice.h"
+#include "rocksdb/slice_transform.h"
 #include "rocksdb/table.h"
 #include "rocksdb/types.h"
 #include "util/coding.h"
@@ -235,4 +236,100 @@ inline LookupKey::~LookupKey() {
   if (start_ != space_) delete[] start_;
 }
 
+class IterKey {
+ public:
+  IterKey() : key_(space_), buf_size_(sizeof(space_)), key_size_(0) {}
+
+  ~IterKey() { ResetBuffer(); }
+
+  Slice GetKey() const { return Slice(key_, key_size_); }
+
+  void Clear() { key_size_ = 0; }
+
+  void SetUserKey(const Slice& user_key) {
+    size_t size = user_key.size();
+    EnlargeBufferIfNeeded(size);
+    memcpy(key_, user_key.data(), size);
+    key_size_ = size;
+  }
+
+  void SetInternalKey(const Slice& user_key, SequenceNumber s,
+                      ValueType value_type = kValueTypeForSeek) {
+    size_t usize = user_key.size();
+    EnlargeBufferIfNeeded(usize + sizeof(uint64_t));
+    memcpy(key_, user_key.data(), usize);
+    EncodeFixed64(key_ + usize, PackSequenceAndType(s, value_type));
+    key_size_ = usize + sizeof(uint64_t);
+  }
+
+  void SetInternalKey(const ParsedInternalKey& parsed_key) {
+    SetInternalKey(parsed_key.user_key, parsed_key.sequence, parsed_key.type);
+  }
+
+ private:
+  char* key_;
+  size_t buf_size_;
+  size_t key_size_;
+  char space_[32];  // Avoid allocation for short keys
+
+  void ResetBuffer() {
+    if (key_ != nullptr && key_ != space_) {
+      delete[] key_;
+    }
+    key_ = space_;
+    buf_size_ = sizeof(space_);
+    key_size_ = 0;
+  }
+
+  // Enlarge the buffer size if needed based on key_size.
+  // By default, static allocated buffer is used. Once there is a key
+  // larger than the static allocated buffer, another buffer is dynamically
+  // allocated, until a larger key buffer is requested. In that case, we
+  // reallocate buffer and delete the old one.
+  void EnlargeBufferIfNeeded(size_t key_size) {
+    // If size is smaller than buffer size, continue using current buffer,
+    // or the static allocated one, as default
+    if (key_size > buf_size_) {
+      // Need to enlarge the buffer.
+      ResetBuffer();
+      key_ = new char[key_size];
+      buf_size_ = key_size;
+    }
+  }
+
+  // No copying allowed
+  IterKey(const IterKey&) = delete;
+  void operator=(const IterKey&) = delete;
+};
+
+class InternalKeySliceTransform : public SliceTransform {
+ public:
+  explicit InternalKeySliceTransform(const SliceTransform* transform)
+      : transform_(transform) {}
+
+  virtual const char* Name() const { return transform_->Name(); }
+
+  virtual Slice Transform(const Slice& src) const {
+    auto user_key = ExtractUserKey(src);
+    return transform_->Transform(user_key);
+  }
+
+  virtual bool InDomain(const Slice& src) const {
+    auto user_key = ExtractUserKey(src);
+    return transform_->InDomain(user_key);
+  }
+
+  virtual bool InRange(const Slice& dst) const {
+    auto user_key = ExtractUserKey(dst);
+    return transform_->InRange(user_key);
+  }
+
+  const SliceTransform* user_prefix_extractor() const { return transform_; }
+
+ private:
+  // Like comparator, InternalKeySliceTransform will not take care of the
+  // deletion of transform_
+  const SliceTransform* const transform_;
+};
+
 }  // namespace rocksdb

db/internal_stats.cc

@@ -38,6 +38,10 @@ DBPropertyType GetPropertyType(const Slice& property) {
     return kBackgroundErrors;
   } else if (in == "cur-size-active-mem-table") {
     return kCurSizeActiveMemTable;
+  } else if (in == "num-entries-active-mem-table") {
+    return kNumEntriesInMutableMemtable;
+  } else if (in == "num-entries-imm-mem-tables") {
+    return kNumEntriesInImmutableMemtable;
   }
   return kUnknown;
 }
@@ -47,7 +51,7 @@ bool InternalStats::GetProperty(DBPropertyType property_type,
                                 DBImpl* db) {
   VersionSet* version_set = db->versions_.get();
   Version* current = version_set->current();
-  const MemTableList& imm = db->imm_;
+  MemTableList& imm = db->imm_;
   Slice in = property;
 
   switch (property_type) {
@@ -353,6 +357,14 @@ bool InternalStats::GetProperty(DBPropertyType property_type,
       // Current size of the active memtable
       *value = std::to_string(db->mem_->ApproximateMemoryUsage());
       return true;
+    case kNumEntriesInMutableMemtable:
+      // Current size of the active memtable
+      *value = std::to_string(db->mem_->GetNumEntries());
+      return true;
+    case kNumEntriesInImmutableMemtable:
+      // Current size of the active memtable
+      *value = std::to_string(imm.current()->GetTotalNumEntries());
+      return true;
     default:
       return false;
   }

db/internal_stats.h

@@ -26,13 +26,16 @@ enum DBPropertyType {
   kLevelStats,       // Return number of files and total sizes of each level
   kStats,            // Return general statitistics of DB
   kSsTables,         // Return a human readable string of current SST files
-  kNumImmutableMemTable,  // Return number of immutable mem tables
+  kNumImmutableMemTable,   // Return number of immutable mem tables
-  kMemtableFlushPending,  // Return 1 if mem table flushing is pending,
+  kMemtableFlushPending,   // Return 1 if mem table flushing is pending,
-                          // otherwise
+                           // otherwise 0.
-                          // 0.
+  kCompactionPending,      // Return 1 if a compaction is pending. Otherwise 0.
-  kCompactionPending,     // Return 1 if a compaction is pending. Otherwise 0.
+  kBackgroundErrors,       // Return accumulated background errors encountered.
-  kBackgroundErrors,      // Return accumulated background errors encountered.
   kCurSizeActiveMemTable,  // Return current size of the active memtable
+  kNumEntriesInMutableMemtable,    // Return number of entries in the mutable
+                                   // memtable.
+  kNumEntriesInImmutableMemtable,  // Return sum of number of entries in all
+                                   // the immutable mem tables.
   kUnknown,
 };
 

db/memtable.cc

@@ -37,6 +37,7 @@ MemTable::MemTable(const InternalKeyComparator& cmp, const Options& options)
       arena_(options.arena_block_size),
       table_(options.memtable_factory->CreateMemTableRep(
           comparator_, &arena_, options.prefix_extractor.get())),
+      num_entries_(0),
       flush_in_progress_(false),
       flush_completed_(false),
       file_number_(0),
@@ -260,6 +261,7 @@ void MemTable::Add(SequenceNumber s, ValueType type,
   memcpy(p, value.data(), val_size);
   assert((unsigned)(p + val_size - buf) == (unsigned)encoded_len);
   table_->Insert(handle);
+  num_entries_++;
 
   if (prefix_bloom_) {
     assert(prefix_extractor_);

db/memtable.h

@@ -132,6 +132,9 @@ class MemTable {
   // key in the memtable.
   size_t CountSuccessiveMergeEntries(const LookupKey& key);
 
+  // Get total number of entries in the mem table.
+  uint64_t GetNumEntries() const { return num_entries_; }
+
   // Returns the edits area that is needed for flushing the memtable
   VersionEdit* GetEdits() { return &edit_; }
 
@@ -182,6 +185,8 @@ class MemTable {
   Arena arena_;
   unique_ptr<MemTableRep> table_;
 
+  uint64_t num_entries_;
+
   // These are used to manage memtable flushes to storage
   bool flush_in_progress_; // started the flush
   bool flush_completed_;   // finished the flush

db/memtable_list.cc

@@ -77,6 +77,14 @@ void MemTableListVersion::AddIterators(const ReadOptions& options,
   }
 }
 
+uint64_t MemTableListVersion::GetTotalNumEntries() const {
+  uint64_t total_num = 0;
+  for (auto& m : memlist_) {
+    total_num += m->GetNumEntries();
+  }
+  return total_num;
+}
+
 // caller is responsible for referencing m
 void MemTableListVersion::Add(MemTable* m) {
   assert(refs_ == 1);  // only when refs_ == 1 is MemTableListVersion mutable

db/memtable_list.h

@@ -43,6 +43,8 @@ class MemTableListVersion {
   void AddIterators(const ReadOptions& options,
                     std::vector<Iterator*>* iterator_list);
 
+  uint64_t GetTotalNumEntries() const;
+
  private:
   // REQUIRE: m is mutable memtable
   void Add(MemTable* m);

db/plain_table_db_test.cc

@@ -190,7 +190,7 @@ class TestPlainTableReader : public PlainTableReader {
                          file_size, bloom_bits_per_key, hash_table_ratio,
                          index_sparseness, table_properties),
         expect_bloom_not_match_(expect_bloom_not_match) {
-    Status s = PopulateIndex();
+    Status s = PopulateIndex(const_cast<TableProperties*>(table_properties));
     ASSERT_TRUE(s.ok());
   }
 
@@ -265,6 +265,19 @@ TEST(PlainTableDBTest, Flush) {
       ASSERT_OK(Put("0000000000000bar", "v2"));
       ASSERT_OK(Put("1000000000000foo", "v3"));
       dbfull()->TEST_FlushMemTable();
+
+      TablePropertiesCollection ptc;
+      reinterpret_cast<DB*>(dbfull())->GetPropertiesOfAllTables(&ptc);
+      ASSERT_EQ(1U, ptc.size());
+      auto row = ptc.begin();
+      auto tp = row->second;
+      ASSERT_EQ(
+          total_order ? "4" : "12",
+          (tp->user_collected_properties).at("plain_table_hash_table_size"));
+      ASSERT_EQ(
+          total_order ? "9" : "0",
+          (tp->user_collected_properties).at("plain_table_sub_index_size"));
+
       ASSERT_EQ("v3", Get("1000000000000foo"));
       ASSERT_EQ("v2", Get("0000000000000bar"));
     }
@@ -429,6 +442,48 @@ TEST(PlainTableDBTest, Iterator) {
   }
 }
 
+std::string MakeLongKey(size_t length, char c) {
+  return std::string(length, c);
+}
+
+TEST(PlainTableDBTest, IteratorLargeKeys) {
+  Options options = CurrentOptions();
+  options.table_factory.reset(NewTotalOrderPlainTableFactory(0, 0, 16));
+  options.create_if_missing = true;
+  options.prefix_extractor.reset();
+  DestroyAndReopen(&options);
+
+  std::string key_list[] = {
+      MakeLongKey(30, '0'),
+      MakeLongKey(16, '1'),
+      MakeLongKey(32, '2'),
+      MakeLongKey(60, '3'),
+      MakeLongKey(90, '4'),
+      MakeLongKey(50, '5'),
+      MakeLongKey(26, '6')
+  };
+
+  for (size_t i = 0; i < 7; i++) {
+    ASSERT_OK(Put(key_list[i], std::to_string(i)));
+  }
+
+  dbfull()->TEST_FlushMemTable();
+
+  Iterator* iter = dbfull()->NewIterator(ro_);
+  iter->Seek(key_list[0]);
+
+  for (size_t i = 0; i < 7; i++) {
+    ASSERT_TRUE(iter->Valid());
+    ASSERT_EQ(key_list[i], iter->key().ToString());
+    ASSERT_EQ(std::to_string(i), iter->value().ToString());
+    iter->Next();
+  }
+
+  ASSERT_TRUE(!iter->Valid());
+
+  delete iter;
+}
+
 // A test comparator which compare two strings in this way:
 // (1) first compare prefix of 8 bytes in alphabet order,
 // (2) if two strings share the same prefix, sort the other part of the string

db/prefix_test.cc

@@ -620,7 +620,7 @@ TEST(PrefixTest, PrefixHash) {
 }
 
 int main(int argc, char** argv) {
-  google::ParseCommandLineFlags(&argc, &argv, true);
+  GFLAGS::ParseCommandLineFlags(&argc, &argv, true);
   std::cout << kDbName << "\n";
 
   rocksdb::test::RunAllTests();

db/table_cache.cc

@@ -111,19 +111,20 @@ Iterator* TableCache::NewIterator(const ReadOptions& options,
   if (table_reader_ptr != nullptr) {
     *table_reader_ptr = nullptr;
   }
-  Cache::Handle* handle = file_meta.table_reader_handle;
+  TableReader* table_reader = file_meta.table_reader;
+  Cache::Handle* handle = nullptr;
   Status s;
-  if (!handle) {
+  if (table_reader == nullptr) {
     s = FindTable(toptions, icomparator, file_meta.number, file_meta.file_size,
                   &handle, nullptr, options.read_tier == kBlockCacheTier);
-  }
+    if (!s.ok()) {
-  if (!s.ok()) {
+      return NewErrorIterator(s);
-    return NewErrorIterator(s);
+    }
+    table_reader = GetTableReaderFromHandle(handle);
   }
 
-  TableReader* table_reader = GetTableReaderFromHandle(handle);
   Iterator* result = table_reader->NewIterator(options);
-  if (!file_meta.table_reader_handle) {
+  if (handle != nullptr) {
     result->RegisterCleanup(&UnrefEntry, cache_.get(), handle);
   }
   if (table_reader_ptr != nullptr) {
@@ -143,17 +144,20 @@ Status TableCache::Get(const ReadOptions& options,
                        bool (*saver)(void*, const ParsedInternalKey&,
                                      const Slice&, bool),
                        bool* table_io, void (*mark_key_may_exist)(void*)) {
-  Cache::Handle* handle = file_meta.table_reader_handle;
+  TableReader* t = file_meta.table_reader;
   Status s;
-  if (!handle) {
+  Cache::Handle* handle = nullptr;
+  if (!t) {
     s = FindTable(storage_options_, internal_comparator, file_meta.number,
                   file_meta.file_size, &handle, table_io,
                   options.read_tier == kBlockCacheTier);
+    if (s.ok()) {
+      t = GetTableReaderFromHandle(handle);
+    }
   }
   if (s.ok()) {
-    TableReader* t = GetTableReaderFromHandle(handle);
     s = t->Get(options, k, arg, saver, mark_key_may_exist);
-    if (!file_meta.table_reader_handle) {
+    if (handle != nullptr) {
       ReleaseHandle(handle);
     }
   } else if (options.read_tier && s.IsIncomplete()) {
@@ -169,15 +173,16 @@ Status TableCache::GetTableProperties(
     const FileMetaData& file_meta,
     std::shared_ptr<const TableProperties>* properties, bool no_io) {
   Status s;
-  auto table_handle = file_meta.table_reader_handle;
+  auto table_reader = file_meta.table_reader;
   // table already been pre-loaded?
-  if (table_handle) {
+  if (table_reader) {
-    auto table = GetTableReaderFromHandle(table_handle);
+    *properties = table_reader->GetTableProperties();
-    *properties = table->GetTableProperties();
+
     return s;
   }
 
   bool table_io;
+  Cache::Handle* table_handle = nullptr;
   s = FindTable(toptions, internal_comparator, file_meta.number,
                 file_meta.file_size, &table_handle, &table_io, no_io);
   if (!s.ok()) {
@@ -195,20 +200,21 @@ bool TableCache::PrefixMayMatch(const ReadOptions& options,
                                 const FileMetaData& file_meta,
                                 const Slice& internal_prefix, bool* table_io) {
   bool may_match = true;
-  auto table_handle = file_meta.table_reader_handle;
+  auto table_reader = file_meta.table_reader;
-  if (table_handle == nullptr) {
+  Cache::Handle* table_handle = nullptr;
+  if (table_reader == nullptr) {
     // Need to get table handle from file number
     Status s = FindTable(storage_options_, icomparator, file_meta.number,
                          file_meta.file_size, &table_handle, table_io);
     if (!s.ok()) {
       return may_match;
     }
+    table_reader = GetTableReaderFromHandle(table_handle);
   }
 
-  auto table = GetTableReaderFromHandle(table_handle);
+  may_match = table_reader->PrefixMayMatch(internal_prefix);
-  may_match = table->PrefixMayMatch(internal_prefix);
 
-  if (file_meta.table_reader_handle == nullptr) {
+  if (table_handle != nullptr) {
     // Need to release handle if it is generated from here.
     ReleaseHandle(table_handle);
   }

db/version_edit.h

@@ -31,10 +31,13 @@ struct FileMetaData {
 
   // Needs to be disposed when refs becomes 0.
   Cache::Handle* table_reader_handle;
+  // Table reader in table_reader_handle
+  TableReader* table_reader;
 
   FileMetaData(uint64_t number, uint64_t file_size) :
       refs(0), allowed_seeks(1 << 30), number(number), file_size(file_size),
-      being_compacted(false), table_reader_handle(nullptr) {
+      being_compacted(false), table_reader_handle(nullptr),
+      table_reader(nullptr) {
   }
   FileMetaData() : FileMetaData(0, 0) { }
 };

db/version_set.cc

@@ -148,7 +148,7 @@ namespace {
 struct EncodedFileMetaData {
   uint64_t number;   // file number
   uint64_t file_size;   // file size
-  Cache::Handle* table_reader_handle;   // cached table reader's handler
+  TableReader* table_reader;   // cached table reader
 };
 }  // namespace
 
@@ -196,7 +196,7 @@ class Version::LevelFileNumIterator : public Iterator {
     auto* file_meta = (*flist_)[index_];
     current_value_.number = file_meta->number;
     current_value_.file_size = file_meta->file_size;
-    current_value_.table_reader_handle = file_meta->table_reader_handle;
+    current_value_.table_reader = file_meta->table_reader;
     return Slice(reinterpret_cast<const char*>(&current_value_),
                  sizeof(EncodedFileMetaData));
   }
@@ -228,7 +228,7 @@ static Iterator* GetFileIterator(void* arg, const ReadOptions& options,
     const EncodedFileMetaData* encoded_meta =
         reinterpret_cast<const EncodedFileMetaData*>(file_value.data());
     FileMetaData meta(encoded_meta->number, encoded_meta->file_size);
-    meta.table_reader_handle = encoded_meta->table_reader_handle;
+    meta.table_reader = encoded_meta->table_reader;
     return cache->NewIterator(
         options.prefix ? options_copy : options, soptions, icomparator, meta,
         nullptr /* don't need reference to table*/, for_compaction);
@@ -254,7 +254,7 @@ bool Version::PrefixMayMatch(const ReadOptions& options,
         reinterpret_cast<const EncodedFileMetaData*>(
             level_iter->value().data());
     FileMetaData meta(encoded_meta->number, encoded_meta->file_size);
-    meta.table_reader_handle = encoded_meta->table_reader_handle;
+    meta.table_reader = encoded_meta->table_reader;
     may_match = vset_->table_cache_->PrefixMayMatch(options, vset_->icmp_, meta,
                                                     internal_prefix, nullptr);
   }
@@ -478,6 +478,12 @@ bool BySmallestKey(FileMetaData* a, FileMetaData* b,
 
 Version::Version(VersionSet* vset, uint64_t version_number)
     : vset_(vset),
+      internal_comparator_(&(vset->icmp_)),
+      user_comparator_(internal_comparator_->user_comparator()),
+      table_cache_(vset->table_cache_),
+      merge_operator_(vset->options_->merge_operator.get()),
+      info_log_(vset->options_->info_log.get()),
+      db_statistics_(vset->options_->statistics.get()),
       next_(this),
       prev_(this),
       refs_(0),
@@ -497,27 +503,22 @@ void Version::Get(const ReadOptions& options,
                   Status* status,
                   MergeContext* merge_context,
                   GetStats* stats,
-                  const Options& db_options,
                   bool* value_found) {
   Slice ikey = k.internal_key();
   Slice user_key = k.user_key();
-  const Comparator* ucmp = vset_->icmp_.user_comparator();
-
-  auto merge_operator = db_options.merge_operator.get();
-  auto logger = db_options.info_log.get();
 
   assert(status->ok() || status->IsMergeInProgress());
   Saver saver;
   saver.state = status->ok()? kNotFound : kMerge;
-  saver.ucmp = ucmp;
+  saver.ucmp = user_comparator_;
   saver.user_key = user_key;
   saver.value_found = value_found;
   saver.value = value;
-  saver.merge_operator = merge_operator;
+  saver.merge_operator = merge_operator_;
   saver.merge_context = merge_context;
-  saver.logger = logger;
+  saver.logger = info_log_;
   saver.didIO = false;
-  saver.statistics = db_options.statistics.get();
+  saver.statistics = db_statistics_;
 
   stats->seek_file = nullptr;
   stats->seek_file_level = -1;
@@ -548,7 +549,7 @@ void Version::Get(const ReadOptions& options,
       // On Level-n (n>=1), files are sorted.
       // Binary search to find earliest index whose largest key >= ikey.
       // We will also stop when the file no longer overlaps ikey
-      start_index = FindFile(vset_->icmp_, files_[level], ikey);
+      start_index = FindFile(*internal_comparator_, files_[level], ikey);
     }
 
     // Traverse each relevant file to find the desired key
@@ -557,8 +558,10 @@ void Version::Get(const ReadOptions& options,
 #endif
     for (uint32_t i = start_index; i < num_files; ++i) {
       FileMetaData* f = files[i];
-      if (ucmp->Compare(user_key, f->smallest.user_key()) < 0 ||
+      // Skip key range filtering for levle 0 if there are few level 0 files.
-          ucmp->Compare(user_key, f->largest.user_key()) > 0) {
+      if ((level > 0 || num_files > 2) &&
+          (user_comparator_->Compare(user_key, f->smallest.user_key()) < 0 ||
+           user_comparator_->Compare(user_key, f->largest.user_key()) > 0)) {
         // Only process overlapping files.
         if (level > 0) {
           // If on Level-n (n>=1) then the files are sorted.
@@ -574,7 +577,8 @@ void Version::Get(const ReadOptions& options,
       // Sanity check to make sure that the files are correctly sorted
       if (prev_file) {
         if (level != 0) {
-          int comp_sign = vset_->icmp_.Compare(prev_file->largest, f->smallest);
+          int comp_sign =
+              internal_comparator_->Compare(prev_file->largest, f->smallest);
           assert(comp_sign < 0);
         } else {
           // level == 0, the current file cannot be newer than the previous one.
@@ -588,9 +592,8 @@ void Version::Get(const ReadOptions& options,
       prev_file = f;
 #endif
       bool tableIO = false;
-      *status =
+      *status = table_cache_->Get(options, *internal_comparator_, *f, ikey,
-          vset_->table_cache_->Get(options, vset_->icmp_, *f, ikey, &saver,
+                                  &saver, SaveValue, &tableIO, MarkKeyMayExist);
-                                   SaveValue, &tableIO, MarkKeyMayExist);
       // TODO: examine the behavior for corrupted key
       if (!status->ok()) {
         return;
@@ -635,12 +638,12 @@ void Version::Get(const ReadOptions& options,
   if (kMerge == saver.state) {
     // merge_operands are in saver and we hit the beginning of the key history
     // do a final merge of nullptr and operands;
-    if (merge_operator->FullMerge(user_key, nullptr,
+    if (merge_operator_->FullMerge(user_key, nullptr,
-                                  saver.merge_context->GetOperands(),
+                                   saver.merge_context->GetOperands(), value,
-                                  value, logger)) {
+                                   info_log_)) {
       *status = Status::OK();
     } else {
-      RecordTick(db_options.statistics.get(), NUMBER_MERGE_FAILURES);
+      RecordTick(db_statistics_, NUMBER_MERGE_FAILURES);
       *status = Status::Corruption("could not perform end-of-key merge for ",
                                    user_key);
     }
@@ -1447,6 +1450,12 @@ class VersionSet::Builder {
                                        file_meta->number, file_meta->file_size,
                                        &file_meta->table_reader_handle,
                                        &table_io, false);
+        if (file_meta->table_reader_handle != nullptr) {
+          // Load table_reader
+          file_meta->table_reader =
+              vset_->table_cache_->GetTableReaderFromHandle(
+                  file_meta->table_reader_handle);
+        }
       }
     }
   }

db/version_set.h

@@ -83,8 +83,7 @@ class Version {
     int seek_file_level;
   };
   void Get(const ReadOptions&, const LookupKey& key, std::string* val,
-           Status* status, MergeContext* merge_context,
+           Status* status, MergeContext* merge_context, GetStats* stats,
-           GetStats* stats, const Options& db_option,
            bool* value_found = nullptr);
 
   // Adds "stats" into the current state.  Returns true if a new
@@ -224,6 +223,12 @@ class Version {
   void UpdateFilesBySize();
 
   VersionSet* vset_;            // VersionSet to which this Version belongs
+  const InternalKeyComparator* internal_comparator_;
+  const Comparator* user_comparator_;
+  TableCache* table_cache_;
+  const MergeOperator* merge_operator_;
+  Logger* info_log_;
+  Statistics* db_statistics_;
   Version* next_;               // Next version in linked list
   Version* prev_;               // Previous version in linked list
   int refs_;                    // Number of live refs to this version

include/rocksdb/table.h

@@ -60,6 +60,12 @@ struct BlockBasedTableOptions {
     // A space efficient index block that is optimized for
     // binary-search-based index.
     kBinarySearch,
+
+    // The hash index, if enabled, will do the hash lookup when
+    // `ReadOption.prefix_seek == true`. User should also specify
+    // `Options.prefix_extractor` to allow the index block to correctly
+    // extract the prefix of the given key and perform hash table lookup.
+    kHashSearch,
   };
 
   IndexType index_type = kBinarySearch;

include/rocksdb/table_properties.h

@@ -23,7 +23,7 @@ namespace rocksdb {
 //      ++pos) {
 //   ...
 // }
-typedef std::map<std::string, std::string> UserCollectedProperties;
+typedef std::map<const std::string, std::string> UserCollectedProperties;
 
 // TableProperties contains a bunch of read-only properties of its associated
 // table.

port/stack_trace.cc

@@ -10,7 +10,9 @@
 #include <execinfo.h>
 #include <signal.h>
 #include <stdio.h>
-#include <stdlib.h>
+// It's odd that including this breaks in GCC 7 but the build doesn't break
+// if I remove it even under GCC 4.8.
+// #include <stdlib.h>
 #include <string.h>
 #include <unistd.h>
 

table/block.cc

@@ -11,16 +11,20 @@
 
 #include "table/block.h"
 
-#include <vector>
 #include <algorithm>
+#include <string>
+#include <unordered_map>
+#include <vector>
+
 #include "rocksdb/comparator.h"
+#include "table/block_hash_index.h"
 #include "table/format.h"
 #include "util/coding.h"
 #include "util/logging.h"
 
 namespace rocksdb {
 
-inline uint32_t Block::NumRestarts() const {
+uint32_t Block::NumRestarts() const {
   assert(size_ >= 2*sizeof(uint32_t));
   return DecodeFixed32(data_ + size_ - sizeof(uint32_t));
 }
@@ -92,6 +96,7 @@ class Block::Iter : public Iterator {
   std::string key_;
   Slice value_;
   Status status_;
+  BlockHashIndex* hash_index_;
 
   inline int Compare(const Slice& a, const Slice& b) const {
     return comparator_->Compare(a, b);
@@ -118,16 +123,15 @@ class Block::Iter : public Iterator {
   }
 
  public:
-  Iter(const Comparator* comparator,
+  Iter(const Comparator* comparator, const char* data, uint32_t restarts,
-       const char* data,
+       uint32_t num_restarts, BlockHashIndex* hash_index)
-       uint32_t restarts,
-       uint32_t num_restarts)
       : comparator_(comparator),
         data_(data),
         restarts_(restarts),
         num_restarts_(num_restarts),
         current_(restarts_),
-        restart_index_(num_restarts_) {
+        restart_index_(num_restarts_),
+        hash_index_(hash_index) {
     assert(num_restarts_ > 0);
   }
 
@@ -169,45 +173,22 @@ class Block::Iter : public Iterator {
   }
 
   virtual void Seek(const Slice& target) {
-    // Binary search in restart array to find the first restart point
+    uint32_t index = 0;
-    // with a key >= target
+    bool ok = hash_index_ ? HashSeek(target, &index)
-    uint32_t left = 0;
+                          : BinarySeek(target, 0, num_restarts_ - 1, &index);
-    uint32_t right = num_restarts_ - 1;
-    while (left < right) {
-      uint32_t mid = (left + right + 1) / 2;
-      uint32_t region_offset = GetRestartPoint(mid);
-      uint32_t shared, non_shared, value_length;
-      const char* key_ptr = DecodeEntry(data_ + region_offset,
-                                        data_ + restarts_,
-                                        &shared, &non_shared, &value_length);
-      if (key_ptr == nullptr || (shared != 0)) {
-        CorruptionError();
-        return;
-      }
-      Slice mid_key(key_ptr, non_shared);
-      if (Compare(mid_key, target) < 0) {
-        // Key at "mid" is smaller than "target".  Therefore all
-        // blocks before "mid" are uninteresting.
-        left = mid;
-      } else {
-        // Key at "mid" is >= "target".  Therefore all blocks at or
-        // after "mid" are uninteresting.
-        right = mid - 1;
-      }
-    }
 
+    if (!ok) {
+      return;
+    }
+    SeekToRestartPoint(index);
     // Linear search (within restart block) for first key >= target
-    SeekToRestartPoint(left);
+
     while (true) {
-      if (!ParseNextKey()) {
+      if (!ParseNextKey() || Compare(key_, target) >= 0) {
-        return;
-      }
-      if (Compare(key_, target) >= 0) {
         return;
       }
     }
   }
-
   virtual void SeekToFirst() {
     SeekToRestartPoint(0);
     ParseNextKey();
@@ -257,6 +238,53 @@ class Block::Iter : public Iterator {
       return true;
     }
   }
+  // Binary search in restart array to find the first restart point
+  // with a key >= target
+  bool BinarySeek(const Slice& target, uint32_t left, uint32_t right,
+                  uint32_t* index) {
+    assert(left <= right);
+
+    while (left < right) {
+      uint32_t mid = (left + right + 1) / 2;
+      uint32_t region_offset = GetRestartPoint(mid);
+      uint32_t shared, non_shared, value_length;
+      const char* key_ptr =
+          DecodeEntry(data_ + region_offset, data_ + restarts_, &shared,
+                      &non_shared, &value_length);
+      if (key_ptr == nullptr || (shared != 0)) {
+        CorruptionError();
+        return false;
+      }
+      Slice mid_key(key_ptr, non_shared);
+      if (Compare(mid_key, target) < 0) {
+        // Key at "mid" is smaller than "target". Therefore all
+        // blocks before "mid" are uninteresting.
+        left = mid;
+      } else {
+        // Key at "mid" is >= "target". Therefore all blocks at or
+        // after "mid" are uninteresting.
+        right = mid - 1;
+      }
+    }
+
+    *index = left;
+    return true;
+  }
+
+  bool HashSeek(const Slice& target, uint32_t* index) {
+    assert(hash_index_);
+    auto restart_index = hash_index_->GetRestartIndex(target);
+    if (restart_index == nullptr) {
+      current_ = restarts_;
+      return 0;
+    }
+
+    // the elements in restart_array[index : index + num_blocks]
+    // are all with same prefix. We'll do binary search in that small range.
+    auto left = restart_index->first_index;
+    auto right = restart_index->first_index + restart_index->num_blocks - 1;
+    return BinarySeek(target, left, right, index);
+  }
 };
 
 Iterator* Block::NewIterator(const Comparator* cmp) {
@@ -267,8 +295,13 @@ Iterator* Block::NewIterator(const Comparator* cmp) {
   if (num_restarts == 0) {
     return NewEmptyIterator();
   } else {
-    return new Iter(cmp, data_, restart_offset_, num_restarts);
+    return new Iter(cmp, data_, restart_offset_, num_restarts,
+                    hash_index_.get());
   }
 }
 
+void Block::SetBlockHashIndex(BlockHashIndex* hash_index) {
+  hash_index_.reset(hash_index);
+}
+
 }  // namespace rocksdb

table/block.h

@@ -10,6 +10,7 @@
 #pragma once
 #include <stddef.h>
 #include <stdint.h>
+
 #include "rocksdb/iterator.h"
 #include "rocksdb/options.h"
 
@@ -17,6 +18,7 @@ namespace rocksdb {
 
 struct BlockContents;
 class Comparator;
+class BlockHashIndex;
 
 class Block {
  public:
@@ -26,20 +28,28 @@ class Block {
   ~Block();
 
   size_t size() const { return size_; }
-  bool   cachable() const { return cachable_; }
+  const char* data() const { return data_; }
+  bool cachable() const { return cachable_; }
+  uint32_t NumRestarts() const;
   CompressionType compression_type() const { return compression_type_; }
+
+  // If hash index lookup is enabled and `use_hash_index` is true. This block
+  // will do hash lookup for the key prefix.
+  //
+  // NOTE: for the hash based lookup, if a key prefix doesn't match any key,
+  // the iterator will simply be set as "invalid", rather than returning
+  // the key that is just pass the target key.
   Iterator* NewIterator(const Comparator* comparator);
-  const char* data() { return data_; }
+  void SetBlockHashIndex(BlockHashIndex* hash_index);
 
  private:
-  uint32_t NumRestarts() const;
-
   const char* data_;
   size_t size_;
   uint32_t restart_offset_;     // Offset in data_ of restart array
   bool owned_;                  // Block owns data_[]
   bool cachable_;
   CompressionType compression_type_;
+  std::unique_ptr<BlockHashIndex> hash_index_;
 
   // No copying allowed
   Block(const Block&);

table/block_based_table_builder.cc

@@ -88,8 +88,7 @@ class IndexBuilder {
   const Comparator* comparator_;
 };
 
-// This index builder builds space-efficient index block for binary-search-based
+// This index builder builds space-efficient index block.
-// index.
 //
 // Optimizations:
 //  1. Made block's `block_restart_interval` to be 1, which will avoid linear
@@ -97,9 +96,9 @@ class IndexBuilder {
 //  2. Shorten the key length for index block. Other than honestly using the
 //     last key in the data block as the index key, we instead find a shortest
 //     substitute key that serves the same function.
-class BinarySearchIndexBuilder : public IndexBuilder {
+class ShortenedIndexBuilder : public IndexBuilder {
  public:
-  explicit BinarySearchIndexBuilder(const Comparator* comparator)
+  explicit ShortenedIndexBuilder(const Comparator* comparator)
       : IndexBuilder(comparator),
         index_block_builder_(1 /* block_restart_interval == 1 */, comparator) {}
 
@@ -128,11 +127,37 @@ class BinarySearchIndexBuilder : public IndexBuilder {
   BlockBuilder index_block_builder_;
 };
 
+// FullKeyIndexBuilder is also based on BlockBuilder. It works pretty much like
+// ShortenedIndexBuilder, but preserves the full key instead the substitude key.
+class FullKeyIndexBuilder : public IndexBuilder {
+ public:
+  explicit FullKeyIndexBuilder(const Comparator* comparator)
+      : IndexBuilder(comparator),
+        index_block_builder_(1 /* block_restart_interval == 1 */, comparator) {}
+
+  virtual void AddEntry(std::string* last_key_in_current_block,
+                        const Slice* first_key_in_next_block,
+                        const BlockHandle& block_handle) override {
+    std::string handle_encoding;
+    block_handle.EncodeTo(&handle_encoding);
+    index_block_builder_.Add(*last_key_in_current_block, handle_encoding);
+  }
+
+  virtual Slice Finish() override { return index_block_builder_.Finish(); }
+
+  virtual size_t EstimatedSize() const {
+    return index_block_builder_.CurrentSizeEstimate();
+  }
+
+ private:
+  BlockBuilder index_block_builder_;
+};
+
 // Create a index builder based on its type.
 IndexBuilder* CreateIndexBuilder(IndexType type, const Comparator* comparator) {
   switch (type) {
     case BlockBasedTableOptions::kBinarySearch: {
-      return new BinarySearchIndexBuilder(comparator);
+      return new ShortenedIndexBuilder(comparator);
     }
     default: {
       assert(!"Do not recognize the index type ");
@@ -296,13 +321,16 @@ struct BlockBasedTableBuilder::Rep {
   }
 };
 
+// TODO(sdong): Currently only write out binary search index. In
+// BlockBasedTableReader, Hash index will be built using binary search index.
 BlockBasedTableBuilder::BlockBasedTableBuilder(
     const Options& options, const BlockBasedTableOptions& table_options,
     const InternalKeyComparator& internal_comparator, WritableFile* file,
     CompressionType compression_type)
     : rep_(new Rep(options, internal_comparator, file,
                    table_options.flush_block_policy_factory.get(),
-                   compression_type, table_options.index_type)) {
+                   compression_type,
+                   BlockBasedTableOptions::IndexType::kBinarySearch)) {
   if (rep_->filter_block != nullptr) {
     rep_->filter_block->StartBlock(0);
   }

table/block_based_table_reader.cc

@@ -9,6 +9,7 @@
 
 #include "table/block_based_table_reader.h"
 
+#include <functional>
 #include <string>
 #include <utility>
 
@@ -25,6 +26,7 @@
 
 #include "table/block.h"
 #include "table/filter_block.h"
+#include "table/block_hash_index.h"
 #include "table/format.h"
 #include "table/meta_blocks.h"
 #include "table/two_level_iterator.h"
@@ -180,19 +182,51 @@ class BinarySearchIndexReader : public IndexReader {
   std::unique_ptr<Block> index_block_;
 };
 
-// TODO(kailiu) This class is only a stub for now. And the comment below is also
-// not completed.
 // Index that leverages an internal hash table to quicken the lookup for a given
 // key.
+// @param data_iter_gen, equavalent to BlockBasedTable::NewIterator(). But that
+// functions requires index to be initalized. To avoid this problem external
+// caller will pass a function that can create the iterator over the entries
+// without the table to be fully initialized.
 class HashIndexReader : public IndexReader {
  public:
   static Status Create(RandomAccessFile* file, const BlockHandle& index_handle,
                        Env* env, const Comparator* comparator,
-                       BlockBasedTable* table,
+                       std::function<Iterator*(Iterator*)> data_iter_gen,
                        const SliceTransform* prefix_extractor,
                        IndexReader** index_reader) {
-    return Status::NotSupported("not implemented yet!");
+    assert(prefix_extractor);
+    Block* index_block = nullptr;
+    auto s =
+        ReadBlockFromFile(file, ReadOptions(), index_handle, &index_block, env);
+
+    if (!s.ok()) {
+      return s;
+    }
+
+    *index_reader = new HashIndexReader(comparator, index_block);
+    std::unique_ptr<Iterator> index_iter(index_block->NewIterator(nullptr));
+    std::unique_ptr<Iterator> data_iter(
+        data_iter_gen(index_block->NewIterator(nullptr)));
+    auto hash_index = CreateBlockHashIndex(index_iter.get(), data_iter.get(),
+                                           index_block->NumRestarts(),
+                                           comparator, prefix_extractor);
+    index_block->SetBlockHashIndex(hash_index);
+    return s;
   }
+
+  virtual Iterator* NewIterator() override {
+    return index_block_->NewIterator(comparator_);
+  }
+
+  virtual size_t size() const override { return index_block_->size(); }
+
+ private:
+  HashIndexReader(const Comparator* comparator, Block* index_block)
+      : IndexReader(comparator), index_block_(index_block) {
+    assert(index_block_ != nullptr);
+  }
+  std::unique_ptr<Block> index_block_;
 };
 
 
@@ -223,6 +257,11 @@ struct BlockBasedTable::Rep {
 
   std::shared_ptr<const TableProperties> table_properties;
   BlockBasedTableOptions::IndexType index_type;
+  // TODO(kailiu) It is very ugly to use internal key in table, since table
+  // module should not be relying on db module. However to make things easier
+  // and compatible with existing code, we introduce a wrapper that allows
+  // block to extract prefix without knowing if a key is internal or not.
+  unique_ptr<SliceTransform> internal_prefix_transform;
 };
 
 BlockBasedTable::~BlockBasedTable() {
@@ -327,8 +366,20 @@ Status BlockBasedTable::Open(const Options& options, const EnvOptions& soptions,
   s = ReadMetaBlock(rep, &meta, &meta_iter);
 
   // Read the properties
+  bool found_properties_block = true;
   meta_iter->Seek(kPropertiesBlock);
-  if (meta_iter->Valid() && meta_iter->key() == kPropertiesBlock) {
+  if (meta_iter->status().ok() &&
+      (!meta_iter->Valid() || meta_iter->key() != kPropertiesBlock)) {
+    meta_iter->Seek(kPropertiesBlockOldName);
+    if (meta_iter->status().ok() &&
+        (!meta_iter->Valid() || meta_iter->key() != kPropertiesBlockOldName)) {
+      found_properties_block = false;
+      Log(WARN, rep->options.info_log,
+          "Cannot find Properties block from file.");
+    }
+  }
+
+  if (found_properties_block) {
     s = meta_iter->status();
     TableProperties* table_properties = nullptr;
     if (s.ok()) {
@@ -747,8 +798,7 @@ BlockBasedTable::CachableEntry<FilterBlockReader> BlockBasedTable::GetFilter(
   return { filter, cache_handle };
 }
 
-Iterator* BlockBasedTable::NewIndexIterator(const ReadOptions& read_options)
+Iterator* BlockBasedTable::NewIndexIterator(const ReadOptions& read_options) {
-    const {
   // index reader has already been pre-populated.
   if (rep_->index_reader) {
     return rep_->index_reader->NewIterator();
@@ -954,11 +1004,14 @@ Status BlockBasedTable::Get(
   return s;
 }
 
+namespace {
 bool SaveDidIO(void* arg, const ParsedInternalKey& key, const Slice& value,
                bool didIO) {
   *reinterpret_cast<bool*>(arg) = didIO;
   return false;
 }
+}  // namespace
+
 bool BlockBasedTable::TEST_KeyInCache(const ReadOptions& options,
                                       const Slice& key) {
   // We use Get() as it has logic that checks whether we read the
@@ -975,22 +1028,51 @@ bool BlockBasedTable::TEST_KeyInCache(const ReadOptions& options,
 //  3. options
 //  4. internal_comparator
 //  5. index_type
-Status BlockBasedTable::CreateIndexReader(IndexReader** index_reader) const {
+Status BlockBasedTable::CreateIndexReader(IndexReader** index_reader) {
   // Some old version of block-based tables don't have index type present in
   // table properties. If that's the case we can safely use the kBinarySearch.
-  auto index_type = BlockBasedTableOptions::kBinarySearch;
+
-  auto& props = rep_->table_properties->user_collected_properties;
+  auto index_type_on_file = BlockBasedTableOptions::kBinarySearch;
-  auto pos = props.find(BlockBasedTablePropertyNames::kIndexType);
+  if (rep_->table_properties) {
-  if (pos != props.end()) {
+    auto& props = rep_->table_properties->user_collected_properties;
-    index_type = static_cast<BlockBasedTableOptions::IndexType>(
+    auto pos = props.find(BlockBasedTablePropertyNames::kIndexType);
-        DecodeFixed32(pos->second.c_str()));
+    if (pos != props.end()) {
+      index_type_on_file = static_cast<BlockBasedTableOptions::IndexType>(
+          DecodeFixed32(pos->second.c_str()));
+    }
   }
 
-  switch (index_type) {
+  // TODO(sdong): Currently binary index is the only index type we support in
+  // files. Hash index is built on top of binary index too.
+  if (index_type_on_file != BlockBasedTableOptions::kBinarySearch) {
+    return Status::NotSupported("File Contains not supported index type: ",
+                                std::to_string(index_type_on_file));
+  }
+
+  auto file = rep_->file.get();
+  const auto& index_handle = rep_->index_handle;
+  auto env = rep_->options.env;
+  auto comparator = &rep_->internal_comparator;
+
+  switch (rep_->index_type) {
     case BlockBasedTableOptions::kBinarySearch: {
-      return BinarySearchIndexReader::Create(
+      return BinarySearchIndexReader::Create(file, index_handle, env,
-          rep_->file.get(), rep_->index_handle, rep_->options.env,
+                                             comparator, index_reader);
-          &rep_->internal_comparator, index_reader);
+    }
+    case BlockBasedTableOptions::kHashSearch: {
+      // We need to wrap data with internal_prefix_transform to make sure it can
+      // handle prefix correctly.
+      rep_->internal_prefix_transform.reset(
+          new InternalKeySliceTransform(rep_->options.prefix_extractor.get()));
+      return HashIndexReader::Create(
+          file, index_handle, env, comparator,
+          [&](Iterator* index_iter) {
+            return NewTwoLevelIterator(
+                index_iter, &BlockBasedTable::DataBlockReader,
+                const_cast<BlockBasedTable*>(this), ReadOptions(),
+                rep_->soptions, rep_->internal_comparator);
+          },
+          rep_->internal_prefix_transform.get(), index_reader);
     }
     default: {
       std::string error_message =
@@ -1023,7 +1105,10 @@ uint64_t BlockBasedTable::ApproximateOffsetOf(const Slice& key) {
     // key is past the last key in the file. If table_properties is not
     // available, approximate the offset by returning the offset of the
     // metaindex block (which is right near the end of the file).
-    result = rep_->table_properties->data_size;
+    result = 0;
+    if (rep_->table_properties) {
+      result = rep_->table_properties->data_size;
+    }
     // table_properties is not present in the table.
     if (result == 0) {
       result = rep_->metaindex_handle.offset();

table/block_based_table_reader.h

@@ -12,6 +12,7 @@
 #include <stdint.h>
 #include <memory>
 #include <utility>
+#include <string>
 
 #include "rocksdb/statistics.h"
 #include "rocksdb/status.h"
@@ -131,7 +132,7 @@ class BlockBasedTable : public TableReader {
   //  2. index is not present in block cache.
   //  3. We disallowed any io to be performed, that is, read_options ==
   //     kBlockCacheTier
-  Iterator* NewIndexIterator(const ReadOptions& read_options) const;
+  Iterator* NewIndexIterator(const ReadOptions& read_options);
 
   // Read block cache from block caches (if set): block_cache and
   // block_cache_compressed.
@@ -164,7 +165,7 @@ class BlockBasedTable : public TableReader {
 
   void ReadMeta(const Footer& footer);
   void ReadFilter(const Slice& filter_handle_value);
-  Status CreateIndexReader(IndexReader** index_reader) const;
+  Status CreateIndexReader(IndexReader** index_reader);
 
   // Read the meta block from sst.
   static Status ReadMetaBlock(
@@ -198,4 +199,8 @@ class BlockBasedTable : public TableReader {
   void operator=(const TableReader&) = delete;
 };
 
+// Backward compatible properties block name. Limited in block based
+// table.
+extern const std::string kPropertiesBlockOldName;
+
 }  // namespace rocksdb

table/block_test.cc

@@ -3,7 +3,10 @@
 //  LICENSE file in the root directory of this source tree. An additional grant
 //  of patent rights can be found in the PATENTS file in the same directory.
 //
+#include <stdio.h>
 #include <string>
+#include <vector>
+
 #include "db/dbformat.h"
 #include "db/memtable.h"
 #include "db/write_batch_internal.h"
@@ -11,9 +14,11 @@
 #include "rocksdb/env.h"
 #include "rocksdb/iterator.h"
 #include "rocksdb/table.h"
+#include "rocksdb/slice_transform.h"
 #include "table/block.h"
 #include "table/block_builder.h"
 #include "table/format.h"
+#include "table/block_hash_index.h"
 #include "util/random.h"
 #include "util/testharness.h"
 #include "util/testutil.h"
@@ -25,6 +30,40 @@ static std::string RandomString(Random* rnd, int len) {
   test::RandomString(rnd, len, &r);
   return r;
 }
+std::string GenerateKey(int primary_key, int secondary_key, int padding_size,
+                        Random *rnd) {
+  char buf[50];
+  char *p = &buf[0];
+  snprintf(buf, sizeof(buf), "%6d%4d", primary_key, secondary_key);
+  std::string k(p);
+  if (padding_size) {
+    k += RandomString(rnd, padding_size);
+  }
+
+  return k;
+}
+
+// Generate random key value pairs.
+// The generated key will be sorted. You can tune the parameters to generated
+// different kinds of test key/value pairs for different scenario.
+void GenerateRandomKVs(std::vector<std::string> *keys,
+                       std::vector<std::string> *values, const int from,
+                       const int len, const int step = 1,
+                       const int padding_size = 0,
+                       const int keys_share_prefix = 1) {
+  Random rnd(302);
+
+  // generate different prefix
+  for (int i = from; i < from + len; i += step) {
+    // generating keys that shares the prefix
+    for (int j = 0; j < keys_share_prefix; ++j) {
+      keys->emplace_back(GenerateKey(i, j, padding_size, &rnd));
+
+      // 100 bytes values
+      values->emplace_back(RandomString(&rnd, 100));
+    }
+  }
+}
 
 class BlockTest {};
 
@@ -39,24 +78,11 @@ TEST(BlockTest, SimpleTest) {
   std::vector<std::string> values;
   BlockBuilder builder(options, ic.get());
   int num_records = 100000;
-  char buf[10];
-  char* p = &buf[0];
 
+  GenerateRandomKVs(&keys, &values, 0, num_records);
   // add a bunch of records to a block
   for (int i = 0; i < num_records; i++) {
-    // generate random kvs
+    builder.Add(keys[i], values[i]);
-    sprintf(p, "%6d", i);
-    std::string k(p);
-    std::string v = RandomString(&rnd, 100); // 100 byte values
-
-    // write kvs to the block
-    Slice key(k);
-    Slice value(v);
-    builder.Add(key, value);
-
-    // remember kvs in a lookaside array
-    keys.push_back(k);
-    values.push_back(v);
   }
 
   // read serialized contents of the block
@@ -101,6 +127,114 @@ TEST(BlockTest, SimpleTest) {
   delete iter;
 }
 
+// return the block contents
+BlockContents GetBlockContents(std::unique_ptr<BlockBuilder> *builder,
+                               const std::vector<std::string> &keys,
+                               const std::vector<std::string> &values,
+                               const int prefix_group_size = 1) {
+  builder->reset(
+      new BlockBuilder(1 /* restart interval */, BytewiseComparator()));
+
+  // Add only half of the keys
+  for (size_t i = 0; i < keys.size(); ++i) {
+    (*builder)->Add(keys[i], values[i]);
+  }
+  Slice rawblock = (*builder)->Finish();
+
+  BlockContents contents;
+  contents.data = rawblock;
+  contents.cachable = false;
+  contents.heap_allocated = false;
+
+  return contents;
+}
+
+void CheckBlockContents(BlockContents contents, const int max_key,
+                        const std::vector<std::string> &keys,
+                        const std::vector<std::string> &values) {
+  const size_t prefix_size = 6;
+  // create block reader
+  Block reader1(contents);
+  Block reader2(contents);
+
+  std::unique_ptr<const SliceTransform> prefix_extractor(
+      NewFixedPrefixTransform(prefix_size));
+
+  {
+    auto iter1 = reader1.NewIterator(nullptr);
+    auto iter2 = reader1.NewIterator(nullptr);
+    reader1.SetBlockHashIndex(CreateBlockHashIndex(iter1, iter2, keys.size(),
+                                                   BytewiseComparator(),
+                                                   prefix_extractor.get()));
+
+    delete iter1;
+    delete iter2;
+  }
+
+  std::unique_ptr<Iterator> hash_iter(
+      reader1.NewIterator(BytewiseComparator()));
+
+  std::unique_ptr<Iterator> regular_iter(
+      reader2.NewIterator(BytewiseComparator()));
+
+  // Seek existent keys
+  for (size_t i = 0; i < keys.size(); i++) {
+    hash_iter->Seek(keys[i]);
+    ASSERT_OK(hash_iter->status());
+    ASSERT_TRUE(hash_iter->Valid());
+
+    Slice v = hash_iter->value();
+    ASSERT_EQ(v.ToString().compare(values[i]), 0);
+  }
+
+  // Seek non-existent keys.
+  // For hash index, if no key with a given prefix is not found, iterator will
+  // simply be set as invalid; whereas the binary search based iterator will
+  // return the one that is closest.
+  for (int i = 1; i < max_key - 1; i += 2) {
+    auto key = GenerateKey(i, 0, 0, nullptr);
+    hash_iter->Seek(key);
+    ASSERT_TRUE(!hash_iter->Valid());
+
+    regular_iter->Seek(key);
+    ASSERT_TRUE(regular_iter->Valid());
+  }
+}
+
+// In this test case, no two key share same prefix.
+TEST(BlockTest, SimpleIndexHash) {
+  const int kMaxKey = 100000;
+  std::vector<std::string> keys;
+  std::vector<std::string> values;
+  GenerateRandomKVs(&keys, &values, 0 /* first key id */,
+                    kMaxKey /* last key id */, 2 /* step */,
+                    8 /* padding size (8 bytes randomly generated suffix) */);
+
+  std::unique_ptr<BlockBuilder> builder;
+  auto contents = GetBlockContents(&builder, keys, values);
+
+  CheckBlockContents(contents, kMaxKey, keys, values);
+}
+
+TEST(BlockTest, IndexHashWithSharedPrefix) {
+  const int kMaxKey = 100000;
+  // for each prefix, there will be 5 keys starts with it.
+  const int kPrefixGroup = 5;
+  std::vector<std::string> keys;
+  std::vector<std::string> values;
+  // Generate keys with same prefix.
+  GenerateRandomKVs(&keys, &values, 0,  // first key id
+                    kMaxKey,            // last key id
+                    2,                  // step
+                    10,                 // padding size,
+                    kPrefixGroup);
+
+  std::unique_ptr<BlockBuilder> builder;
+  auto contents = GetBlockContents(&builder, keys, values, kPrefixGroup);
+
+  CheckBlockContents(contents, kMaxKey, keys, values);
+}
+
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {

table/meta_blocks.cc

@@ -244,6 +244,8 @@ Status ReadTableProperties(RandomAccessFile* file, uint64_t file_size,
       metaindex_block.NewIterator(BytewiseComparator()));
 
   // -- Read property block
+  // This function is not used by BlockBasedTable, so we don't have to
+  // worry about old properties block name.
   meta_iter->Seek(kPropertiesBlock);
   TableProperties table_properties;
   if (meta_iter->Valid() &&

table/plain_table_builder.cc

@@ -74,10 +74,12 @@ void PlainTableBuilder::Add(const Slice& key, const Slice& value) {
 
   if (!IsFixedLength()) {
     // Write key length
-    key_size_str_.clear();
+    char key_size_buf[5];  // tmp buffer for key size as varint32
-    PutVarint32(&key_size_str_, user_key_size);
+    char* ptr = EncodeVarint32(key_size_buf, user_key_size);
-    file_->Append(key_size_str_);
+    assert(ptr <= key_size_buf + sizeof(key_size_buf));
-    offset_ += key_size_str_.length();
+    auto len = ptr - key_size_buf;
+    file_->Append(Slice(key_size_buf, len));
+    offset_ += len;
   }
 
   // Write key
@@ -86,25 +88,32 @@ void PlainTableBuilder::Add(const Slice& key, const Slice& value) {
     status_ = Status::Corruption(Slice());
     return;
   }
+  // For value size as varint32 (up to 5 bytes).
+  // If the row is of value type with seqId 0, flush the special flag together
+  // in this buffer to safe one file append call, which takes 1 byte.
+  char value_size_buf[6];
+  size_t value_size_buf_size = 0;
   if (parsed_key.sequence == 0 && parsed_key.type == kTypeValue) {
     file_->Append(Slice(key.data(), user_key_size));
-    char tmp_char = PlainTableFactory::kValueTypeSeqId0;
+    offset_ += user_key_size;
-    file_->Append(Slice(&tmp_char, 1));
+    value_size_buf[0] = PlainTableFactory::kValueTypeSeqId0;
-    offset_ += key.size() - 7;
+    value_size_buf_size = 1;
   } else {
     file_->Append(key);
     offset_ += key.size();
   }
 
   // Write value length
-  value_size_str_.clear();
   int value_size = value.size();
-  PutVarint32(&value_size_str_, value_size);
+  char* end_ptr =
-  file_->Append(value_size_str_);
+      EncodeVarint32(value_size_buf + value_size_buf_size, value_size);
+  assert(end_ptr <= value_size_buf + sizeof(value_size_buf));
+  value_size_buf_size = end_ptr - value_size_buf;
+  file_->Append(Slice(value_size_buf, value_size_buf_size));
 
   // Write value
   file_->Append(value);
-  offset_ += value_size + value_size_str_.length();
+  offset_ += value_size + value_size_buf_size;
 
   properties_.num_entries++;
   properties_.raw_key_size += key.size();

table/plain_table_builder.h

@@ -69,9 +69,6 @@ private:
   const size_t user_key_len_;
   bool closed_ = false;  // Either Finish() or Abandon() has been called.
 
-  std::string key_size_str_;
-  std::string value_size_str_;
-
   bool IsFixedLength() const {
     return user_key_len_ > 0;
   }

table/plain_table_reader.cc

@@ -81,10 +81,9 @@ class PlainTableIterator : public Iterator {
   bool use_prefix_seek_;
   uint32_t offset_;
   uint32_t next_offset_;
-  Slice key_;
+  IterKey key_;
   Slice value_;
   Status status_;
-  std::string tmp_str_;
   // No copying allowed
   PlainTableIterator(const PlainTableIterator&) = delete;
   void operator=(const Iterator&) = delete;
@@ -104,8 +103,8 @@ PlainTableReader::PlainTableReader(
       kHashTableRatio(hash_table_ratio),
       kBloomBitsPerKey(bloom_bits_per_key),
       kIndexIntervalForSamePrefixKeys(index_sparseness),
-      table_properties_(table_properties),
+      table_properties_(nullptr),
-      data_end_offset_(table_properties_->data_size),
+      data_end_offset_(table_properties->data_size),
       user_key_len_(table_properties->fixed_key_len) {
   assert(kHashTableRatio >= 0.0);
 }
@@ -137,7 +136,7 @@ Status PlainTableReader::Open(
       bloom_bits_per_key, hash_table_ratio, index_sparseness, props));
 
   // -- Populate Index
-  s = new_reader->PopulateIndex();
+  s = new_reader->PopulateIndex(props);
   if (!s.ok()) {
     return s;
   }
@@ -364,7 +363,10 @@ void PlainTableReader::FillIndexes(
       index_size_, kSubIndexSize);
 }
 
-Status PlainTableReader::PopulateIndex() {
+Status PlainTableReader::PopulateIndex(TableProperties* props) {
+  assert(props != nullptr);
+  table_properties_.reset(props);
+
   // options.prefix_extractor is requried for a hash-based look-up.
   if (options_.prefix_extractor.get() == nullptr && kHashTableRatio != 0) {
     return Status::NotSupported(
@@ -409,6 +411,14 @@ Status PlainTableReader::PopulateIndex() {
   // From the temp data structure, populate indexes.
   FillIndexes(sub_index_size_needed, hash_to_offsets, entries_per_bucket);
 
+  // Fill two table properties.
+  // TODO(sdong): after we have the feature of storing index in file, this
+  // properties need to be populated to index_size instead.
+  props->user_collected_properties["plain_table_hash_table_size"] =
+      std::to_string(index_size_ * 4U);
+  props->user_collected_properties["plain_table_sub_index_size"] =
+      std::to_string(sub_index_size_needed);
+
   return Status::OK();
 }
 
@@ -720,9 +730,7 @@ void PlainTableIterator::Next() {
     status_ = table_->Next(&next_offset_, &parsed_key, &value_);
     if (status_.ok()) {
       // Make a copy in this case. TODO optimize.
-      tmp_str_.clear();
+      key_.SetInternalKey(parsed_key);
-      AppendInternalKey(&tmp_str_, parsed_key);
-      key_ = Slice(tmp_str_);
     } else {
       offset_ = next_offset_ = table_->data_end_offset_;
     }
@@ -735,7 +743,7 @@ void PlainTableIterator::Prev() {
 
 Slice PlainTableIterator::key() const {
   assert(Valid());
-  return key_;
+  return key_.GetKey();
 }
 
 Slice PlainTableIterator::value() const {

table/plain_table_reader.h

@@ -86,6 +86,9 @@ class PlainTableReader: public TableReader {
   // PopulateIndex() builds index of keys. It must be called before any query
   // to the table.
   //
+  // props: the table properties object that need to be stored. Ownership of
+  //        the object will be passed.
+  //
   // index_ contains buckets size of index_size_, each is a
   // 32-bit integer. The lower 31 bits contain an offset value (explained below)
   // and the first bit of the integer indicates type of the offset.
@@ -121,7 +124,7 @@ class PlainTableReader: public TableReader {
   //    ....
   //   record N file offset:  fixedint32
   // <end>
-  Status PopulateIndex();
+  Status PopulateIndex(TableProperties* props);
 
  private:
   struct IndexRecord;

table/table_properties.cc

@@ -91,5 +91,7 @@ const std::string TablePropertiesNames::kFixedKeyLen =
     "rocksdb.fixed.key.length";
 
 extern const std::string kPropertiesBlock = "rocksdb.properties";
+// Old property block name for backward compatibility
+extern const std::string kPropertiesBlockOldName = "rocksdb.stats";
 
 }  // namespace rocksdb

table/table_reader_bench.cc

@@ -233,9 +233,9 @@ DEFINE_string(time_unit, "microsecond",
               "`microsecond` (default) or `nanosecond`");
 
 int main(int argc, char** argv) {
-  google::SetUsageMessage(std::string("\nUSAGE:\n") + std::string(argv[0]) +
+  GFLAGS::SetUsageMessage(std::string("\nUSAGE:\n") + std::string(argv[0]) +
                           " [OPTIONS]...");
-  google::ParseCommandLineFlags(&argc, &argv, true);
+  GFLAGS::ParseCommandLineFlags(&argc, &argv, true);
 
   rocksdb::TableFactory* tf = new rocksdb::BlockBasedTableFactory();
   rocksdb::Options options;

table/table_test.cc

@@ -1055,6 +1055,116 @@ static std::string RandomString(Random* rnd, int len) {
   return r;
 }
 
+void AddInternalKey(TableConstructor* c, const std::string prefix,
+                    int suffix_len = 800) {
+  static Random rnd(1023);
+  InternalKey k(prefix + RandomString(&rnd, 800), 0, kTypeValue);
+  c->Add(k.Encode().ToString(), "v");
+}
+
+TEST(TableTest, HashIndexTest) {
+  TableConstructor c(BytewiseComparator());
+
+  // keys with prefix length 3, make sure the key/value is big enough to fill
+  // one block
+  AddInternalKey(&c, "0015");
+  AddInternalKey(&c, "0035");
+
+  AddInternalKey(&c, "0054");
+  AddInternalKey(&c, "0055");
+
+  AddInternalKey(&c, "0056");
+  AddInternalKey(&c, "0057");
+
+  AddInternalKey(&c, "0058");
+  AddInternalKey(&c, "0075");
+
+  AddInternalKey(&c, "0076");
+  AddInternalKey(&c, "0095");
+
+  std::vector<std::string> keys;
+  KVMap kvmap;
+  Options options;
+  BlockBasedTableOptions table_options;
+  table_options.index_type = BlockBasedTableOptions::kHashSearch;
+  options.table_factory.reset(new BlockBasedTableFactory(table_options));
+
+  options.prefix_extractor.reset(NewFixedPrefixTransform(3));
+  options.block_cache = NewLRUCache(1024);
+  options.block_size = 1700;
+
+  std::unique_ptr<InternalKeyComparator> comparator(
+      new InternalKeyComparator(BytewiseComparator()));
+  c.Finish(options, *comparator, &keys, &kvmap);
+  auto reader = c.table_reader();
+
+  auto props = c.table_reader()->GetTableProperties();
+  ASSERT_EQ(5u, props->num_data_blocks);
+
+  std::unique_ptr<Iterator> hash_iter(reader->NewIterator(ReadOptions()));
+
+  // -- Find keys do not exist, but have common prefix.
+  std::vector<std::string> prefixes = {"001", "003", "005", "007", "009"};
+  std::vector<std::string> lower_bound = {keys[0], keys[1], keys[2],
+                                          keys[7], keys[9], };
+
+  // find the lower bound of the prefix
+  for (size_t i = 0; i < prefixes.size(); ++i) {
+    hash_iter->Seek(InternalKey(prefixes[i], 0, kTypeValue).Encode());
+    ASSERT_OK(hash_iter->status());
+    ASSERT_TRUE(hash_iter->Valid());
+
+    // seek the first element in the block
+    ASSERT_EQ(lower_bound[i], hash_iter->key().ToString());
+    ASSERT_EQ("v", hash_iter->value().ToString());
+  }
+
+  // find the upper bound of prefixes
+  std::vector<std::string> upper_bound = {keys[1], keys[2], keys[7], keys[9], };
+
+  // find existing keys
+  for (const auto& item : kvmap) {
+    auto ukey = ExtractUserKey(item.first).ToString();
+    hash_iter->Seek(ukey);
+
+    // ASSERT_OK(regular_iter->status());
+    ASSERT_OK(hash_iter->status());
+
+    // ASSERT_TRUE(regular_iter->Valid());
+    ASSERT_TRUE(hash_iter->Valid());
+
+    ASSERT_EQ(item.first, hash_iter->key().ToString());
+    ASSERT_EQ(item.second, hash_iter->value().ToString());
+  }
+
+  for (size_t i = 0; i < prefixes.size(); ++i) {
+    // the key is greater than any existing keys.
+    auto key = prefixes[i] + "9";
+    hash_iter->Seek(InternalKey(key, 0, kTypeValue).Encode());
+
+    ASSERT_OK(hash_iter->status());
+    if (i == prefixes.size() - 1) {
+      // last key
+      ASSERT_TRUE(!hash_iter->Valid());
+    } else {
+      ASSERT_TRUE(hash_iter->Valid());
+      // seek the first element in the block
+      ASSERT_EQ(upper_bound[i], hash_iter->key().ToString());
+      ASSERT_EQ("v", hash_iter->value().ToString());
+    }
+  }
+
+  // find keys with prefix that don't match any of the existing prefixes.
+  std::vector<std::string> non_exist_prefixes = {"002", "004", "006", "008"};
+  for (const auto& prefix : non_exist_prefixes) {
+    hash_iter->Seek(InternalKey(prefix, 0, kTypeValue).Encode());
+    // regular_iter->Seek(prefix);
+
+    ASSERT_OK(hash_iter->status());
+    ASSERT_TRUE(!hash_iter->Valid());
+  }
+}
+
 // It's very hard to figure out the index block size of a block accurately.
 // To make sure we get the index size, we just make sure as key number
 // grows, the filter block size also grows.

tools/db_repl_stress.cc

@@ -87,10 +87,10 @@ DEFINE_uint64(wal_size_limit_MB, 10, "the wal size limit for the run"
               "(in MB)");
 
 int main(int argc, const char** argv) {
-  google::SetUsageMessage(std::string("\nUSAGE:\n") + std::string(argv[0]) +
+  GFLAGS::SetUsageMessage(std::string("\nUSAGE:\n") + std::string(argv[0]) +
     " --num_inserts=<num_inserts> --wal_ttl_seconds=<WAL_ttl_seconds>" +
     " --wal_size_limit_MB=<WAL_size_limit_MB>");
-  google::ParseCommandLineFlags(&argc, const_cast<char***>(&argv), true);
+  GFLAGS::ParseCommandLineFlags(&argc, const_cast<char***>(&argv), true);
 
   Env* env = Env::Default();
   std::string default_db_path;

tools/db_stress.cc

@@ -61,7 +61,7 @@ static bool ValidateUint32Range(const char* flagname, uint64_t value) {
 }
 DEFINE_uint64(seed, 2341234, "Seed for PRNG");
 static const bool FLAGS_seed_dummy =
-  google::RegisterFlagValidator(&FLAGS_seed, &ValidateUint32Range);
+  GFLAGS::RegisterFlagValidator(&FLAGS_seed, &ValidateUint32Range);
 
 DEFINE_int64(max_key, 1 * KB * KB * KB,
              "Max number of key/values to place in database");
@@ -171,7 +171,7 @@ static bool ValidateInt32Positive(const char* flagname, int32_t value) {
 }
 DEFINE_int32(reopen, 10, "Number of times database reopens");
 static const bool FLAGS_reopen_dummy =
-  google::RegisterFlagValidator(&FLAGS_reopen, &ValidateInt32Positive);
+  GFLAGS::RegisterFlagValidator(&FLAGS_reopen, &ValidateInt32Positive);
 
 DEFINE_int32(bloom_bits, 10, "Bloom filter bits per key. "
              "Negative means use default settings.");
@@ -199,7 +199,7 @@ DEFINE_int32(kill_random_test, 0,
              "If non-zero, kill at various points in source code with "
              "probability 1/this");
 static const bool FLAGS_kill_random_test_dummy =
-  google::RegisterFlagValidator(&FLAGS_kill_random_test,
+  GFLAGS::RegisterFlagValidator(&FLAGS_kill_random_test,
                                 &ValidateInt32Positive);
 extern int rocksdb_kill_odds;
 
@@ -227,32 +227,32 @@ static bool ValidateInt32Percent(const char* flagname, int32_t value) {
 DEFINE_int32(readpercent, 10,
              "Ratio of reads to total workload (expressed as a percentage)");
 static const bool FLAGS_readpercent_dummy =
-  google::RegisterFlagValidator(&FLAGS_readpercent, &ValidateInt32Percent);
+  GFLAGS::RegisterFlagValidator(&FLAGS_readpercent, &ValidateInt32Percent);
 
 DEFINE_int32(prefixpercent, 20,
              "Ratio of prefix iterators to total workload (expressed as a"
              " percentage)");
 static const bool FLAGS_prefixpercent_dummy =
-  google::RegisterFlagValidator(&FLAGS_prefixpercent, &ValidateInt32Percent);
+  GFLAGS::RegisterFlagValidator(&FLAGS_prefixpercent, &ValidateInt32Percent);
 
 DEFINE_int32(writepercent, 45,
              " Ratio of deletes to total workload (expressed as a percentage)");
 static const bool FLAGS_writepercent_dummy =
-  google::RegisterFlagValidator(&FLAGS_writepercent, &ValidateInt32Percent);
+  GFLAGS::RegisterFlagValidator(&FLAGS_writepercent, &ValidateInt32Percent);
 
 DEFINE_int32(delpercent, 15,
              "Ratio of deletes to total workload (expressed as a percentage)");
 static const bool FLAGS_delpercent_dummy =
-  google::RegisterFlagValidator(&FLAGS_delpercent, &ValidateInt32Percent);
+  GFLAGS::RegisterFlagValidator(&FLAGS_delpercent, &ValidateInt32Percent);
 
 DEFINE_int32(iterpercent, 10, "Ratio of iterations to total workload"
              " (expressed as a percentage)");
 static const bool FLAGS_iterpercent_dummy =
-  google::RegisterFlagValidator(&FLAGS_iterpercent, &ValidateInt32Percent);
+  GFLAGS::RegisterFlagValidator(&FLAGS_iterpercent, &ValidateInt32Percent);
 
 DEFINE_uint64(num_iterations, 10, "Number of iterations per MultiIterate run");
 static const bool FLAGS_num_iterations_dummy =
-  google::RegisterFlagValidator(&FLAGS_num_iterations, &ValidateUint32Range);
+  GFLAGS::RegisterFlagValidator(&FLAGS_num_iterations, &ValidateUint32Range);
 
 DEFINE_bool(disable_seek_compaction, false,
             "Option to disable compation triggered by read.");
@@ -292,18 +292,18 @@ static rocksdb::Env* FLAGS_env = rocksdb::Env::Default();
 
 DEFINE_uint64(ops_per_thread, 600000, "Number of operations per thread.");
 static const bool FLAGS_ops_per_thread_dummy =
-  google::RegisterFlagValidator(&FLAGS_ops_per_thread, &ValidateUint32Range);
+  GFLAGS::RegisterFlagValidator(&FLAGS_ops_per_thread, &ValidateUint32Range);
 
 DEFINE_uint64(log2_keys_per_lock, 2, "Log2 of number of keys per lock");
 static const bool FLAGS_log2_keys_per_lock_dummy =
-  google::RegisterFlagValidator(&FLAGS_log2_keys_per_lock,
+  GFLAGS::RegisterFlagValidator(&FLAGS_log2_keys_per_lock,
                                 &ValidateUint32Range);
 
 DEFINE_int32(purge_redundant_percent, 50,
              "Percentage of times we want to purge redundant keys in memory "
              "before flushing");
 static const bool FLAGS_purge_redundant_percent_dummy =
-  google::RegisterFlagValidator(&FLAGS_purge_redundant_percent,
+  GFLAGS::RegisterFlagValidator(&FLAGS_purge_redundant_percent,
                                 &ValidateInt32Percent);
 
 DEFINE_bool(filter_deletes, false, "On true, deletes use KeyMayExist to drop"
@@ -340,7 +340,7 @@ static bool ValidatePrefixSize(const char* flagname, int32_t value) {
 }
 DEFINE_int32(prefix_size, 7, "Control the prefix size for HashSkipListRep");
 static const bool FLAGS_prefix_size_dummy =
-  google::RegisterFlagValidator(&FLAGS_prefix_size, &ValidatePrefixSize);
+  GFLAGS::RegisterFlagValidator(&FLAGS_prefix_size, &ValidatePrefixSize);
 
 DEFINE_bool(use_merge, false, "On true, replaces all writes with a Merge "
             "that behaves like a Put");
@@ -1502,9 +1502,9 @@ class StressTest {
 
 
 int main(int argc, char** argv) {
-  google::SetUsageMessage(std::string("\nUSAGE:\n") + std::string(argv[0]) +
+  GFLAGS::SetUsageMessage(std::string("\nUSAGE:\n") + std::string(argv[0]) +
                           " [OPTIONS]...");
-  google::ParseCommandLineFlags(&argc, &argv, true);
+  GFLAGS::ParseCommandLineFlags(&argc, &argv, true);
 
   if (FLAGS_statistics) {
     dbstats = rocksdb::CreateDBStatistics();

util/bloom_test.cc

@@ -164,7 +164,7 @@ TEST(BloomTest, VaryingLengths) {
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {
-  google::ParseCommandLineFlags(&argc, &argv, true);
+  GFLAGS::ParseCommandLineFlags(&argc, &argv, true);
 
   return rocksdb::test::RunAllTests();
 }

util/dynamic_bloom_test.cc

@@ -196,7 +196,7 @@ TEST(DynamicBloomTest, perf) {
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {
-  google::ParseCommandLineFlags(&argc, &argv, true);
+  GFLAGS::ParseCommandLineFlags(&argc, &argv, true);
 
   return rocksdb::test::RunAllTests();
 }

util/env_posix.cc

@@ -761,6 +761,10 @@ class PosixWritableFile : public WritableFile {
   }
 
   virtual Status Sync() {
+    Status s = Flush();
+    if (!s.ok()) {
+      return s;
+    }
     TEST_KILL_RANDOM(rocksdb_kill_odds);
     if (pending_sync_ && fdatasync(fd_) < 0) {
       return IOError(filename_, errno);
@@ -771,6 +775,10 @@ class PosixWritableFile : public WritableFile {
   }
 
   virtual Status Fsync() {
+    Status s = Flush();
+    if (!s.ok()) {
+      return s;
+    }
     TEST_KILL_RANDOM(rocksdb_kill_odds);
     if (pending_fsync_ && fsync(fd_) < 0) {
       return IOError(filename_, errno);

util/env_test.cc

@@ -290,7 +290,6 @@ TEST(EnvPosixTest, AllocateTest) {
   // allocate 100 MB
   size_t kPreallocateSize = 100 * 1024 * 1024;
   size_t kBlockSize = 512;
-  size_t kPageSize = 4096;
   std::string data = "test";
   wfile->SetPreallocationBlockSize(kPreallocateSize);
   ASSERT_OK(wfile->Append(Slice(data)));
@@ -299,8 +298,9 @@ TEST(EnvPosixTest, AllocateTest) {
   struct stat f_stat;
   stat(fname.c_str(), &f_stat);
   ASSERT_EQ((unsigned int)data.size(), f_stat.st_size);
+  auto st_blocks = f_stat.st_blocks;
   // verify that blocks are preallocated
-  ASSERT_EQ((unsigned int)(kPreallocateSize / kBlockSize), f_stat.st_blocks);
+  ASSERT_LE((unsigned int)(kPreallocateSize / kBlockSize), st_blocks);
 
   // close the file, should deallocate the blocks
   wfile.reset();
@@ -308,8 +308,7 @@ TEST(EnvPosixTest, AllocateTest) {
   stat(fname.c_str(), &f_stat);
   ASSERT_EQ((unsigned int)data.size(), f_stat.st_size);
   // verify that preallocated blocks were deallocated on file close
-  size_t data_blocks_pages = ((data.size() + kPageSize - 1) / kPageSize);
+  ASSERT_LT(f_stat.st_blocks, st_blocks);
-  ASSERT_EQ((unsigned int)(data_blocks_pages * kPageSize / kBlockSize), f_stat.st_blocks);
 }
 #endif
 

util/thread_local.h

@@ -10,6 +10,7 @@
 #pragma once
 
 #include <atomic>
+#include <functional>
 #include <memory>
 #include <unordered_map>
 #include <vector>