Killing Transform Rep

Summary:
Let's get rid of TransformRep and it's children. We have confirmed that HashSkipListRep works better with multifeed, so there is no benefit to keeping this around.

This diff is mostly just deleting references to obsoleted functions. I also have a diff for fbcode that we'll need to push when we switch to new release.

I had to expose HashSkipListRepFactory in the client header files because db_impl.cc needs access to GetTransform() function for SanitizeOptions.

Test Plan: make check

Reviewers: dhruba, haobo, kailiu, sdong

Reviewed By: dhruba

CC: leveldb

Differential Revision: https://reviews.facebook.net/D14397
This commit is contained in:
Igor Canadi
2013-12-03 12:42:15 -08:00
parent 043fc14c3e
commit eb12e47e0e
11 changed files with 161 additions and 696 deletions

View File

@@ -431,12 +431,11 @@ static bool ValidatePrefixSize(const char* flagname, int32_t value) {
}
return true;
}
DEFINE_int32(prefix_size, 0, "Control the prefix size for PrefixHashRep");
DEFINE_int32(prefix_size, 0, "Control the prefix size for HashSkipList");
enum RepFactory {
kSkipList,
kPrefixHash,
kUnsorted,
kVectorRep
};
enum RepFactory StringToRepFactory(const char* ctype) {
@@ -446,8 +445,6 @@ enum RepFactory StringToRepFactory(const char* ctype) {
return kSkipList;
else if (!strcasecmp(ctype, "prefix_hash"))
return kPrefixHash;
else if (!strcasecmp(ctype, "unsorted"))
return kUnsorted;
else if (!strcasecmp(ctype, "vector"))
return kVectorRep;
@@ -803,9 +800,6 @@ class Benchmark {
case kSkipList:
fprintf(stdout, "Memtablerep: skip_list\n");
break;
case kUnsorted:
fprintf(stdout, "Memtablerep: unsorted\n");
break;
case kVectorRep:
fprintf(stdout, "Memtablerep: vector\n");
break;
@@ -1328,14 +1322,8 @@ class Benchmark {
}
switch (FLAGS_rep_factory) {
case kPrefixHash:
options.memtable_factory.reset(
new PrefixHashRepFactory(NewFixedPrefixTransform(FLAGS_prefix_size))
);
break;
case kUnsorted:
options.memtable_factory.reset(
new UnsortedRepFactory
);
options.memtable_factory.reset(NewHashSkipListRepFactory(
NewFixedPrefixTransform(FLAGS_prefix_size)));
break;
case kSkipList:
// no need to do anything

View File

@@ -50,6 +50,7 @@
#include "util/auto_roll_logger.h"
#include "util/build_version.h"
#include "util/coding.h"
#include "util/hash_skiplist_rep.h"
#include "util/logging.h"
#include "util/mutexlock.h"
#include "util/perf_context_imp.h"
@@ -162,10 +163,10 @@ Options SanitizeOptions(const std::string& dbname,
Log(result.info_log, "Compaction filter specified, ignore factory");
}
if (result.prefix_extractor) {
// If a prefix extractor has been supplied and a PrefixHashRepFactory is
// If a prefix extractor has been supplied and a HashSkipListRepFactory is
// being used, make sure that the latter uses the former as its transform
// function.
auto factory = dynamic_cast<PrefixHashRepFactory*>(
auto factory = dynamic_cast<HashSkipListRepFactory*>(
result.memtable_factory.get());
if (factory &&
factory->GetTransform() != result.prefix_extractor) {

View File

@@ -244,7 +244,6 @@ class DBTest {
enum OptionConfig {
kDefault,
kVectorRep,
kUnsortedRep,
kMergePut,
kFilter,
kUncompressed,
@@ -255,7 +254,7 @@ class DBTest {
kCompactOnFlush,
kPerfOptions,
kDeletesFilterFirst,
kPrefixHashRep,
kHashSkipList,
kUniversalCompaction,
kCompressedBlockCache,
kEnd
@@ -339,9 +338,9 @@ class DBTest {
Options CurrentOptions() {
Options options;
switch (option_config_) {
case kPrefixHashRep:
options.memtable_factory.reset(new
PrefixHashRepFactory(NewFixedPrefixTransform(1)));
case kHashSkipList:
options.memtable_factory.reset(
NewHashSkipListRepFactory(NewFixedPrefixTransform(1)));
break;
case kMergePut:
options.merge_operator = MergeOperators::CreatePutOperator();
@@ -375,9 +374,6 @@ class DBTest {
case kDeletesFilterFirst:
options.filter_deletes = true;
break;
case kUnsortedRep:
options.memtable_factory.reset(new UnsortedRepFactory);
break;
case kVectorRep:
options.memtable_factory.reset(new VectorRepFactory(100));
break;
@@ -4600,7 +4596,7 @@ TEST(DBTest, Randomized) {
// TODO(sanjay): Test Get() works
int p = rnd.Uniform(100);
int minimum = 0;
if (option_config_ == kPrefixHashRep) {
if (option_config_ == kHashSkipList) {
minimum = 1;
}
if (p < 45) { // Put
@@ -4770,90 +4766,82 @@ void PrefixScanInit(DBTest *dbtest) {
}
TEST(DBTest, PrefixScan) {
for (int it = 0; it < 2; ++it) {
ReadOptions ro = ReadOptions();
int count;
Slice prefix;
Slice key;
char buf[100];
Iterator* iter;
snprintf(buf, sizeof(buf), "03______:");
prefix = Slice(buf, 8);
key = Slice(buf, 9);
auto prefix_extractor = NewFixedPrefixTransform(8);
// db configs
env_->count_random_reads_ = true;
Options options = CurrentOptions();
options.env = env_;
options.no_block_cache = true;
options.filter_policy = NewBloomFilterPolicy(10);
options.prefix_extractor = prefix_extractor;
options.whole_key_filtering = false;
options.disable_auto_compactions = true;
options.max_background_compactions = 2;
options.create_if_missing = true;
options.disable_seek_compaction = true;
if (it == 0) {
options.memtable_factory.reset(NewHashSkipListRepFactory(
prefix_extractor));
} else {
options.memtable_factory = std::make_shared<PrefixHashRepFactory>(
prefix_extractor);
}
ReadOptions ro = ReadOptions();
int count;
Slice prefix;
Slice key;
char buf[100];
Iterator* iter;
snprintf(buf, sizeof(buf), "03______:");
prefix = Slice(buf, 8);
key = Slice(buf, 9);
auto prefix_extractor = NewFixedPrefixTransform(8);
// db configs
env_->count_random_reads_ = true;
Options options = CurrentOptions();
options.env = env_;
options.no_block_cache = true;
options.filter_policy = NewBloomFilterPolicy(10);
options.prefix_extractor = prefix_extractor;
options.whole_key_filtering = false;
options.disable_auto_compactions = true;
options.max_background_compactions = 2;
options.create_if_missing = true;
options.disable_seek_compaction = true;
options.memtable_factory.reset(NewHashSkipListRepFactory(prefix_extractor));
// prefix specified, with blooms: 2 RAND I/Os
// SeekToFirst
DestroyAndReopen(&options);
PrefixScanInit(this);
count = 0;
env_->random_read_counter_.Reset();
ro.prefix = &prefix;
iter = db_->NewIterator(ro);
for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
assert(iter->key().starts_with(prefix));
count++;
}
ASSERT_OK(iter->status());
delete iter;
ASSERT_EQ(count, 2);
ASSERT_EQ(env_->random_read_counter_.Read(), 2);
// prefix specified, with blooms: 2 RAND I/Os
// Seek
DestroyAndReopen(&options);
PrefixScanInit(this);
count = 0;
env_->random_read_counter_.Reset();
ro.prefix = &prefix;
iter = db_->NewIterator(ro);
for (iter->Seek(key); iter->Valid(); iter->Next()) {
assert(iter->key().starts_with(prefix));
count++;
}
ASSERT_OK(iter->status());
delete iter;
ASSERT_EQ(count, 2);
ASSERT_EQ(env_->random_read_counter_.Read(), 2);
// no prefix specified: 11 RAND I/Os
DestroyAndReopen(&options);
PrefixScanInit(this);
count = 0;
env_->random_read_counter_.Reset();
iter = db_->NewIterator(ReadOptions());
for (iter->Seek(prefix); iter->Valid(); iter->Next()) {
if (! iter->key().starts_with(prefix)) {
break;
}
count++;
}
ASSERT_OK(iter->status());
delete iter;
ASSERT_EQ(count, 2);
ASSERT_EQ(env_->random_read_counter_.Read(), 11);
Close();
delete options.filter_policy;
// prefix specified, with blooms: 2 RAND I/Os
// SeekToFirst
DestroyAndReopen(&options);
PrefixScanInit(this);
count = 0;
env_->random_read_counter_.Reset();
ro.prefix = &prefix;
iter = db_->NewIterator(ro);
for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
assert(iter->key().starts_with(prefix));
count++;
}
ASSERT_OK(iter->status());
delete iter;
ASSERT_EQ(count, 2);
ASSERT_EQ(env_->random_read_counter_.Read(), 2);
// prefix specified, with blooms: 2 RAND I/Os
// Seek
DestroyAndReopen(&options);
PrefixScanInit(this);
count = 0;
env_->random_read_counter_.Reset();
ro.prefix = &prefix;
iter = db_->NewIterator(ro);
for (iter->Seek(key); iter->Valid(); iter->Next()) {
assert(iter->key().starts_with(prefix));
count++;
}
ASSERT_OK(iter->status());
delete iter;
ASSERT_EQ(count, 2);
ASSERT_EQ(env_->random_read_counter_.Read(), 2);
// no prefix specified: 11 RAND I/Os
DestroyAndReopen(&options);
PrefixScanInit(this);
count = 0;
env_->random_read_counter_.Reset();
iter = db_->NewIterator(ReadOptions());
for (iter->Seek(prefix); iter->Valid(); iter->Next()) {
if (! iter->key().starts_with(prefix)) {
break;
}
count++;
}
ASSERT_OK(iter->status());
delete iter;
ASSERT_EQ(count, 2);
ASSERT_EQ(env_->random_read_counter_.Read(), 11);
Close();
delete options.filter_policy;
}
std::string MakeKey(unsigned int num) {

View File

@@ -38,8 +38,8 @@ std::shared_ptr<DB> OpenDb() {
if (FLAGS_use_set_based_memetable) {
auto prefix_extractor = rocksdb::NewFixedPrefixTransform(0);
options.memtable_factory =
std::make_shared<rocksdb::PrefixHashRepFactory>(prefix_extractor);
options.memtable_factory.reset(
NewHashSkipListRepFactory(prefix_extractor));
}
Status s = DB::Open(options, kDbName, &db);

View File

@@ -11,7 +11,6 @@
#include "util/testharness.h"
DEFINE_bool(use_prefix_hash_memtable, true, "");
DEFINE_bool(use_nolock_version, true, "");
DEFINE_bool(trigger_deadlock, false,
"issue delete in range scan to trigger PrefixHashMap deadlock");
DEFINE_uint64(bucket_count, 100000, "number of buckets");
@@ -109,14 +108,8 @@ class PrefixTest {
if (FLAGS_use_prefix_hash_memtable) {
auto prefix_extractor = NewFixedPrefixTransform(8);
options.prefix_extractor = prefix_extractor;
if (FLAGS_use_nolock_version) {
options.memtable_factory.reset(NewHashSkipListRepFactory(
prefix_extractor, FLAGS_bucket_count));
} else {
options.memtable_factory =
std::make_shared<rocksdb::PrefixHashRepFactory>(
prefix_extractor, FLAGS_bucket_count, FLAGS_num_locks);
}
options.memtable_factory.reset(NewHashSkipListRepFactory(
prefix_extractor, FLAGS_bucket_count));
}
Status s = DB::Open(options, kDbName, &db);