rocksdb/util/comparator.cc

304 lines
9.6 KiB
C++
Raw Normal View History

// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
// This source code is licensed under both the GPLv2 (found in the
// COPYING file in the root directory) and Apache 2.0 License
// (found in the LICENSE.Apache file in the root directory).
//
// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file. See the AUTHORS file for names of contributors.
#include "rocksdb/comparator.h"
#include <stdint.h>
#include <algorithm>
#include <memory>
#include <mutex>
#include "options/configurable_helper.h"
#include "port/port.h"
#include "rocksdb/slice.h"
#include "rocksdb/utilities/object_registry.h"
namespace ROCKSDB_NAMESPACE {
namespace {
class BytewiseComparatorImpl : public Comparator {
public:
BytewiseComparatorImpl() { }
static const char* kClassName() { return "leveldb.BytewiseComparator"; }
const char* Name() const override { return kClassName(); }
int Compare(const Slice& a, const Slice& b) const override {
return a.compare(b);
}
bool Equal(const Slice& a, const Slice& b) const override { return a == b; }
void FindShortestSeparator(std::string* start,
const Slice& limit) const override {
// Find length of common prefix
size_t min_length = std::min(start->size(), limit.size());
size_t diff_index = 0;
while ((diff_index < min_length) &&
((*start)[diff_index] == limit[diff_index])) {
diff_index++;
}
if (diff_index >= min_length) {
// Do not shorten if one string is a prefix of the other
} else {
uint8_t start_byte = static_cast<uint8_t>((*start)[diff_index]);
uint8_t limit_byte = static_cast<uint8_t>(limit[diff_index]);
if (start_byte >= limit_byte) {
// Cannot shorten since limit is smaller than start or start is
// already the shortest possible.
return;
}
assert(start_byte < limit_byte);
if (diff_index < limit.size() - 1 || start_byte + 1 < limit_byte) {
(*start)[diff_index]++;
start->resize(diff_index + 1);
} else {
// v
// A A 1 A A A
// A A 2
//
// Incrementing the current byte will make start bigger than limit, we
// will skip this byte, and find the first non 0xFF byte in start and
// increment it.
diff_index++;
while (diff_index < start->size()) {
// Keep moving until we find the first non 0xFF byte to
// increment it
if (static_cast<uint8_t>((*start)[diff_index]) <
static_cast<uint8_t>(0xff)) {
(*start)[diff_index]++;
start->resize(diff_index + 1);
break;
}
diff_index++;
}
}
assert(Compare(*start, limit) < 0);
}
}
void FindShortSuccessor(std::string* key) const override {
// Find first character that can be incremented
size_t n = key->size();
for (size_t i = 0; i < n; i++) {
const uint8_t byte = (*key)[i];
if (byte != static_cast<uint8_t>(0xff)) {
(*key)[i] = byte + 1;
key->resize(i+1);
return;
}
}
// *key is a run of 0xffs. Leave it alone.
}
bool IsSameLengthImmediateSuccessor(const Slice& s,
const Slice& t) const override {
if (s.size() != t.size() || s.size() == 0) {
return false;
}
size_t diff_ind = s.difference_offset(t);
// same slice
if (diff_ind >= s.size()) return false;
uint8_t byte_s = static_cast<uint8_t>(s[diff_ind]);
uint8_t byte_t = static_cast<uint8_t>(t[diff_ind]);
// first different byte must be consecutive, and remaining bytes must be
// 0xff for s and 0x00 for t
if (byte_s != uint8_t{0xff} && byte_s + 1 == byte_t) {
for (size_t i = diff_ind + 1; i < s.size(); ++i) {
byte_s = static_cast<uint8_t>(s[i]);
byte_t = static_cast<uint8_t>(t[i]);
if (byte_s != uint8_t{0xff} || byte_t != uint8_t{0x00}) {
return false;
}
}
return true;
} else {
return false;
}
}
bool CanKeysWithDifferentByteContentsBeEqual() const override {
return false;
}
Add support for timestamp in Get/Put (#5079) Summary: It's useful to be able to (optionally) associate key-value pairs with user-provided timestamps. This PR is an early effort towards this goal and continues the work of facebook#4942. A suite of new unit tests exist in DBBasicTestWithTimestampWithParam. Support for timestamp requires the user to provide timestamp as a slice in `ReadOptions` and `WriteOptions`. All timestamps of the same database must share the same length, format, etc. The format of the timestamp is the same throughout the same database, and the user is responsible for providing a comparator function (Comparator) to order the <key, timestamp> tuples. Once created, the format and length of the timestamp cannot change (at least for now). Test plan (on devserver): ``` $COMPILE_WITH_ASAN=1 make -j32 all $./db_basic_test --gtest_filter=Timestamp/DBBasicTestWithTimestampWithParam.PutAndGet/* $make check ``` All tests must pass. We also run the following db_bench tests to verify whether there is regression on Get/Put while timestamp is not enabled. ``` $TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=fillseq,readrandom -num=1000000 $TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=fillrandom -num=1000000 ``` Repeat for 6 times for both versions. Results are as follows: ``` | | readrandom | fillrandom | | master | 16.77 MB/s | 47.05 MB/s | | PR5079 | 16.44 MB/s | 47.03 MB/s | ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/5079 Differential Revision: D15132946 Pulled By: riversand963 fbshipit-source-id: 833a0d657eac21182f0f206c910a6438154c742c
2019-06-06 06:07:28 +00:00
Iterator with timestamp (#6255) Summary: Preliminary support for iterator with user timestamp. Current implementation does not consider merge operator and reverse iterator. Auto compaction is also disabled in unit tests. Create an iterator with timestamp. ``` ... read_opts.timestamp = &ts; auto* iter = db->NewIterator(read_opts); // target is key without timestamp. for (iter->Seek(target); iter->Valid(); iter->Next()) {} for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {} delete iter; read_opts.timestamp = &ts1; // lower_bound and upper_bound are without timestamp. read_opts.iterate_lower_bound = &lower_bound; read_opts.iterate_upper_bound = &upper_bound; auto* iter1 = db->NewIterator(read_opts); // Do Seek or SeekToFirst() delete iter1; ``` Test plan (dev server) ``` $make check ``` Simple benchmarking (dev server) 1. The overhead introduced by this PR even when timestamp is disabled. key size: 16 bytes value size: 100 bytes Entries: 1000000 Data reside in main memory, and try to stress iterator. Repeated three times on master and this PR. - Seek without next ``` ./db_bench -db=/dev/shm/rocksdbtest-1000 -benchmarks=fillseq,seekrandom -enable_pipelined_write=false -disable_wal=true -format_version=3 ``` master: 159047.0 ops/sec this PR: 158922.3 ops/sec (2% drop in throughput) - Seek and next 10 times ``` ./db_bench -db=/dev/shm/rocksdbtest-1000 -benchmarks=fillseq,seekrandom -enable_pipelined_write=false -disable_wal=true -format_version=3 -seek_nexts=10 ``` master: 109539.3 ops/sec this PR: 107519.7 ops/sec (2% drop in throughput) Pull Request resolved: https://github.com/facebook/rocksdb/pull/6255 Differential Revision: D19438227 Pulled By: riversand963 fbshipit-source-id: b66b4979486f8474619f4aa6bdd88598870b0746
2020-03-07 00:21:03 +00:00
using Comparator::CompareWithoutTimestamp;
int CompareWithoutTimestamp(const Slice& a, bool /*a_has_ts*/, const Slice& b,
bool /*b_has_ts*/) const override {
Add support for timestamp in Get/Put (#5079) Summary: It's useful to be able to (optionally) associate key-value pairs with user-provided timestamps. This PR is an early effort towards this goal and continues the work of facebook#4942. A suite of new unit tests exist in DBBasicTestWithTimestampWithParam. Support for timestamp requires the user to provide timestamp as a slice in `ReadOptions` and `WriteOptions`. All timestamps of the same database must share the same length, format, etc. The format of the timestamp is the same throughout the same database, and the user is responsible for providing a comparator function (Comparator) to order the <key, timestamp> tuples. Once created, the format and length of the timestamp cannot change (at least for now). Test plan (on devserver): ``` $COMPILE_WITH_ASAN=1 make -j32 all $./db_basic_test --gtest_filter=Timestamp/DBBasicTestWithTimestampWithParam.PutAndGet/* $make check ``` All tests must pass. We also run the following db_bench tests to verify whether there is regression on Get/Put while timestamp is not enabled. ``` $TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=fillseq,readrandom -num=1000000 $TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=fillrandom -num=1000000 ``` Repeat for 6 times for both versions. Results are as follows: ``` | | readrandom | fillrandom | | master | 16.77 MB/s | 47.05 MB/s | | PR5079 | 16.44 MB/s | 47.03 MB/s | ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/5079 Differential Revision: D15132946 Pulled By: riversand963 fbshipit-source-id: 833a0d657eac21182f0f206c910a6438154c742c
2019-06-06 06:07:28 +00:00
return a.compare(b);
}
Enable backward iterator for keys with user-defined timestamp (#8035) Summary: This PR does the following: - Enable backward iteration for keys with user-defined timestamp. Note that merge, single delete, range delete are not supported yet. - Introduces a new helper API `Comparator::EqualWithoutTimestamp()`. - Fix a typo in `SetTimestamp()`. - Add/update unit tests Run db_bench (built with DEBUG_LEVEL=0) to demonstrate that no overhead is introduced for CPU-intensive workloads with a lot of `Prev()`. Also provided results of iterating keys with timestamps. 1. Disable timestamp, run: ``` ./db_bench -db=/dev/shm/rocksdb -disable_wal=1 -benchmarks=fillseq,seekrandom[-W1-X6] -reverse_iterator=1 -seek_nexts=5 ``` Results: > Baseline > - seekrandom [AVG 6 runs] : 96115 ops/sec; 53.2 MB/sec > - seekrandom [MEDIAN 6 runs] : 98075 ops/sec; 54.2 MB/sec > > This PR > - seekrandom [AVG 6 runs] : 95521 ops/sec; 52.8 MB/sec > - seekrandom [MEDIAN 6 runs] : 96338 ops/sec; 53.3 MB/sec 2. Enable timestamp, run: ``` ./db_bench -user_timestamp_size=8 -db=/dev/shm/rocksdb -disable_wal=1 -benchmarks=fillseq,seekrandom[-W1-X6] -reverse_iterator=1 -seek_nexts=5 ``` Result: > Baseline: not supported > > This PR > - seekrandom [AVG 6 runs] : 90514 ops/sec; 50.1 MB/sec > - seekrandom [MEDIAN 6 runs] : 90834 ops/sec; 50.2 MB/sec Pull Request resolved: https://github.com/facebook/rocksdb/pull/8035 Reviewed By: ltamasi Differential Revision: D26926668 Pulled By: riversand963 fbshipit-source-id: 95330cc2242397c03e09d29e5417dfb0adc98ef5
2021-03-10 19:13:55 +00:00
bool EqualWithoutTimestamp(const Slice& a, const Slice& b) const override {
return a == b;
}
};
class ReverseBytewiseComparatorImpl : public BytewiseComparatorImpl {
public:
ReverseBytewiseComparatorImpl() { }
static const char* kClassName() {
return "rocksdb.ReverseBytewiseComparator";
}
const char* Name() const override { return kClassName(); }
int Compare(const Slice& a, const Slice& b) const override {
return -a.compare(b);
}
void FindShortestSeparator(std::string* start,
const Slice& limit) const override {
// Find length of common prefix
size_t min_length = std::min(start->size(), limit.size());
size_t diff_index = 0;
while ((diff_index < min_length) &&
((*start)[diff_index] == limit[diff_index])) {
diff_index++;
}
assert(diff_index <= min_length);
if (diff_index == min_length) {
// Do not shorten if one string is a prefix of the other
//
// We could handle cases like:
// V
// A A 2 X Y
// A A 2
// in a similar way as BytewiseComparator::FindShortestSeparator().
// We keep it simple by not implementing it. We can come back to it
// later when needed.
} else {
uint8_t start_byte = static_cast<uint8_t>((*start)[diff_index]);
uint8_t limit_byte = static_cast<uint8_t>(limit[diff_index]);
if (start_byte > limit_byte && diff_index < start->size() - 1) {
// Case like
// V
// A A 3 A A
// A A 1 B B
//
// or
// v
// A A 2 A A
// A A 1 B B
// In this case "AA2" will be good.
#ifndef NDEBUG
std::string old_start = *start;
#endif
start->resize(diff_index + 1);
#ifndef NDEBUG
assert(old_start >= *start);
#endif
assert(Slice(*start).compare(limit) > 0);
}
}
}
void FindShortSuccessor(std::string* /*key*/) const override {
// Don't do anything for simplicity.
}
bool CanKeysWithDifferentByteContentsBeEqual() const override {
return false;
}
Add support for timestamp in Get/Put (#5079) Summary: It's useful to be able to (optionally) associate key-value pairs with user-provided timestamps. This PR is an early effort towards this goal and continues the work of facebook#4942. A suite of new unit tests exist in DBBasicTestWithTimestampWithParam. Support for timestamp requires the user to provide timestamp as a slice in `ReadOptions` and `WriteOptions`. All timestamps of the same database must share the same length, format, etc. The format of the timestamp is the same throughout the same database, and the user is responsible for providing a comparator function (Comparator) to order the <key, timestamp> tuples. Once created, the format and length of the timestamp cannot change (at least for now). Test plan (on devserver): ``` $COMPILE_WITH_ASAN=1 make -j32 all $./db_basic_test --gtest_filter=Timestamp/DBBasicTestWithTimestampWithParam.PutAndGet/* $make check ``` All tests must pass. We also run the following db_bench tests to verify whether there is regression on Get/Put while timestamp is not enabled. ``` $TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=fillseq,readrandom -num=1000000 $TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=fillrandom -num=1000000 ``` Repeat for 6 times for both versions. Results are as follows: ``` | | readrandom | fillrandom | | master | 16.77 MB/s | 47.05 MB/s | | PR5079 | 16.44 MB/s | 47.03 MB/s | ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/5079 Differential Revision: D15132946 Pulled By: riversand963 fbshipit-source-id: 833a0d657eac21182f0f206c910a6438154c742c
2019-06-06 06:07:28 +00:00
Iterator with timestamp (#6255) Summary: Preliminary support for iterator with user timestamp. Current implementation does not consider merge operator and reverse iterator. Auto compaction is also disabled in unit tests. Create an iterator with timestamp. ``` ... read_opts.timestamp = &ts; auto* iter = db->NewIterator(read_opts); // target is key without timestamp. for (iter->Seek(target); iter->Valid(); iter->Next()) {} for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {} delete iter; read_opts.timestamp = &ts1; // lower_bound and upper_bound are without timestamp. read_opts.iterate_lower_bound = &lower_bound; read_opts.iterate_upper_bound = &upper_bound; auto* iter1 = db->NewIterator(read_opts); // Do Seek or SeekToFirst() delete iter1; ``` Test plan (dev server) ``` $make check ``` Simple benchmarking (dev server) 1. The overhead introduced by this PR even when timestamp is disabled. key size: 16 bytes value size: 100 bytes Entries: 1000000 Data reside in main memory, and try to stress iterator. Repeated three times on master and this PR. - Seek without next ``` ./db_bench -db=/dev/shm/rocksdbtest-1000 -benchmarks=fillseq,seekrandom -enable_pipelined_write=false -disable_wal=true -format_version=3 ``` master: 159047.0 ops/sec this PR: 158922.3 ops/sec (2% drop in throughput) - Seek and next 10 times ``` ./db_bench -db=/dev/shm/rocksdbtest-1000 -benchmarks=fillseq,seekrandom -enable_pipelined_write=false -disable_wal=true -format_version=3 -seek_nexts=10 ``` master: 109539.3 ops/sec this PR: 107519.7 ops/sec (2% drop in throughput) Pull Request resolved: https://github.com/facebook/rocksdb/pull/6255 Differential Revision: D19438227 Pulled By: riversand963 fbshipit-source-id: b66b4979486f8474619f4aa6bdd88598870b0746
2020-03-07 00:21:03 +00:00
using Comparator::CompareWithoutTimestamp;
int CompareWithoutTimestamp(const Slice& a, bool /*a_has_ts*/, const Slice& b,
bool /*b_has_ts*/) const override {
Add support for timestamp in Get/Put (#5079) Summary: It's useful to be able to (optionally) associate key-value pairs with user-provided timestamps. This PR is an early effort towards this goal and continues the work of facebook#4942. A suite of new unit tests exist in DBBasicTestWithTimestampWithParam. Support for timestamp requires the user to provide timestamp as a slice in `ReadOptions` and `WriteOptions`. All timestamps of the same database must share the same length, format, etc. The format of the timestamp is the same throughout the same database, and the user is responsible for providing a comparator function (Comparator) to order the <key, timestamp> tuples. Once created, the format and length of the timestamp cannot change (at least for now). Test plan (on devserver): ``` $COMPILE_WITH_ASAN=1 make -j32 all $./db_basic_test --gtest_filter=Timestamp/DBBasicTestWithTimestampWithParam.PutAndGet/* $make check ``` All tests must pass. We also run the following db_bench tests to verify whether there is regression on Get/Put while timestamp is not enabled. ``` $TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=fillseq,readrandom -num=1000000 $TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=fillrandom -num=1000000 ``` Repeat for 6 times for both versions. Results are as follows: ``` | | readrandom | fillrandom | | master | 16.77 MB/s | 47.05 MB/s | | PR5079 | 16.44 MB/s | 47.03 MB/s | ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/5079 Differential Revision: D15132946 Pulled By: riversand963 fbshipit-source-id: 833a0d657eac21182f0f206c910a6438154c742c
2019-06-06 06:07:28 +00:00
return -a.compare(b);
}
};
}// namespace
const Comparator* BytewiseComparator() {
static BytewiseComparatorImpl bytewise;
return &bytewise;
}
const Comparator* ReverseBytewiseComparator() {
static ReverseBytewiseComparatorImpl rbytewise;
return &rbytewise;
}
#ifndef ROCKSDB_LITE
static int RegisterBuiltinComparators(ObjectLibrary& library,
const std::string& /*arg*/) {
library.Register<const Comparator>(
BytewiseComparatorImpl::kClassName(),
[](const std::string& /*uri*/,
std::unique_ptr<const Comparator>* /*guard */,
std::string* /* errmsg */) { return BytewiseComparator(); });
library.Register<const Comparator>(
ReverseBytewiseComparatorImpl::kClassName(),
[](const std::string& /*uri*/,
std::unique_ptr<const Comparator>* /*guard */,
std::string* /* errmsg */) { return ReverseBytewiseComparator(); });
return 2;
}
#endif // ROCKSDB_LITE
Status Comparator::CreateFromString(const ConfigOptions& config_options,
const std::string& value,
const Comparator** result) {
#ifndef ROCKSDB_LITE
static std::once_flag once;
std::call_once(once, [&]() {
RegisterBuiltinComparators(*(ObjectLibrary::Default().get()), "");
});
#endif // ROCKSDB_LITE
std::string id;
std::unordered_map<std::string, std::string> opt_map;
Status status =
ConfigurableHelper::GetOptionsMap(value, *result, &id, &opt_map);
if (!status.ok()) { // GetOptionsMap failed
return status;
}
std::string curr_opts;
#ifndef ROCKSDB_LITE
if (*result != nullptr && (*result)->GetId() == id) {
// Try to get the existing options, ignoring any errors
ConfigOptions embedded = config_options;
embedded.delimiter = ";";
(*result)->GetOptionString(embedded, &curr_opts).PermitUncheckedError();
}
#endif
if (id == BytewiseComparatorImpl::kClassName()) {
*result = BytewiseComparator();
} else if (id == ReverseBytewiseComparatorImpl::kClassName()) {
*result = ReverseBytewiseComparator();
} else if (value.empty()) {
// No Id and no options. Clear the object
*result = nullptr;
return Status::OK();
} else if (id.empty()) { // We have no Id but have options. Not good
return Status::NotSupported("Cannot reset object ", id);
} else {
#ifndef ROCKSDB_LITE
status = config_options.registry->NewStaticObject(id, result);
#else
status = Status::NotSupported("Cannot load object in LITE mode ", id);
#endif // ROCKSDB_LITE
if (!status.ok()) {
if (config_options.ignore_unsupported_options &&
status.IsNotSupported()) {
return Status::OK();
} else {
return status;
}
} else if (!curr_opts.empty() || !opt_map.empty()) {
Comparator* comparator = const_cast<Comparator*>(*result);
status = ConfigurableHelper::ConfigureNewObject(
config_options, comparator, id, curr_opts, opt_map);
}
}
return status;
}
} // namespace ROCKSDB_NAMESPACE