2016-02-10 00:12:00 +01:00
|
|
|
// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
|
2017-07-16 01:03:42 +02:00
|
|
|
// 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).
|
2014-08-19 00:19:17 +02:00
|
|
|
|
2015-06-19 00:54:05 +02:00
|
|
|
#ifndef ROCKSDB_LITE
|
|
|
|
|
2014-08-19 00:19:17 +02:00
|
|
|
#include "rocksdb/utilities/write_batch_with_index.h"
|
2014-10-11 01:21:34 +02:00
|
|
|
|
2015-07-02 01:13:49 +02:00
|
|
|
#include <limits>
|
2015-09-02 22:58:22 +02:00
|
|
|
#include <memory>
|
2014-10-11 01:21:34 +02:00
|
|
|
|
2014-08-19 00:19:17 +02:00
|
|
|
#include "db/column_family.h"
|
2016-10-18 22:19:26 +02:00
|
|
|
#include "db/db_impl.h"
|
2015-05-11 23:51:51 +02:00
|
|
|
#include "db/merge_context.h"
|
|
|
|
#include "db/merge_helper.h"
|
2017-04-06 22:59:31 +02:00
|
|
|
#include "memtable/skiplist.h"
|
2017-04-06 04:02:00 +02:00
|
|
|
#include "options/db_options.h"
|
2015-09-02 22:58:22 +02:00
|
|
|
#include "rocksdb/comparator.h"
|
|
|
|
#include "rocksdb/iterator.h"
|
2014-08-19 00:19:17 +02:00
|
|
|
#include "util/arena.h"
|
2015-05-11 23:51:51 +02:00
|
|
|
#include "utilities/write_batch_with_index/write_batch_with_index_internal.h"
|
2014-08-19 00:19:17 +02:00
|
|
|
|
|
|
|
namespace rocksdb {
|
2014-10-07 18:47:16 +02:00
|
|
|
|
2014-10-11 01:21:34 +02:00
|
|
|
// when direction == forward
|
|
|
|
// * current_at_base_ <=> base_iterator > delta_iterator
|
|
|
|
// when direction == backwards
|
|
|
|
// * current_at_base_ <=> base_iterator < delta_iterator
|
|
|
|
// always:
|
|
|
|
// * equal_keys_ <=> base_iterator == delta_iterator
|
|
|
|
class BaseDeltaIterator : public Iterator {
|
|
|
|
public:
|
|
|
|
BaseDeltaIterator(Iterator* base_iterator, WBWIIterator* delta_iterator,
|
|
|
|
const Comparator* comparator)
|
|
|
|
: forward_(true),
|
|
|
|
current_at_base_(true),
|
|
|
|
equal_keys_(false),
|
|
|
|
status_(Status::OK()),
|
|
|
|
base_iterator_(base_iterator),
|
|
|
|
delta_iterator_(delta_iterator),
|
|
|
|
comparator_(comparator) {}
|
|
|
|
|
|
|
|
virtual ~BaseDeltaIterator() {}
|
|
|
|
|
|
|
|
bool Valid() const override {
|
|
|
|
return current_at_base_ ? BaseValid() : DeltaValid();
|
|
|
|
}
|
|
|
|
|
|
|
|
void SeekToFirst() override {
|
|
|
|
forward_ = true;
|
|
|
|
base_iterator_->SeekToFirst();
|
|
|
|
delta_iterator_->SeekToFirst();
|
|
|
|
UpdateCurrent();
|
|
|
|
}
|
|
|
|
|
|
|
|
void SeekToLast() override {
|
|
|
|
forward_ = false;
|
|
|
|
base_iterator_->SeekToLast();
|
|
|
|
delta_iterator_->SeekToLast();
|
|
|
|
UpdateCurrent();
|
|
|
|
}
|
|
|
|
|
2014-11-06 20:14:28 +01:00
|
|
|
void Seek(const Slice& k) override {
|
2014-10-11 01:21:34 +02:00
|
|
|
forward_ = true;
|
2014-11-06 20:14:28 +01:00
|
|
|
base_iterator_->Seek(k);
|
|
|
|
delta_iterator_->Seek(k);
|
2014-10-11 01:21:34 +02:00
|
|
|
UpdateCurrent();
|
|
|
|
}
|
|
|
|
|
2016-09-28 03:20:57 +02:00
|
|
|
void SeekForPrev(const Slice& k) override {
|
|
|
|
forward_ = false;
|
|
|
|
base_iterator_->SeekForPrev(k);
|
|
|
|
delta_iterator_->SeekForPrev(k);
|
|
|
|
UpdateCurrent();
|
|
|
|
}
|
|
|
|
|
2014-10-11 01:21:34 +02:00
|
|
|
void Next() override {
|
|
|
|
if (!Valid()) {
|
|
|
|
status_ = Status::NotSupported("Next() on invalid iterator");
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!forward_) {
|
|
|
|
// Need to change direction
|
|
|
|
// if our direction was backward and we're not equal, we have two states:
|
|
|
|
// * both iterators are valid: we're already in a good state (current
|
|
|
|
// shows to smaller)
|
|
|
|
// * only one iterator is valid: we need to advance that iterator
|
|
|
|
forward_ = true;
|
|
|
|
equal_keys_ = false;
|
|
|
|
if (!BaseValid()) {
|
|
|
|
assert(DeltaValid());
|
|
|
|
base_iterator_->SeekToFirst();
|
|
|
|
} else if (!DeltaValid()) {
|
|
|
|
delta_iterator_->SeekToFirst();
|
|
|
|
} else if (current_at_base_) {
|
|
|
|
// Change delta from larger than base to smaller
|
|
|
|
AdvanceDelta();
|
|
|
|
} else {
|
|
|
|
// Change base from larger than delta to smaller
|
|
|
|
AdvanceBase();
|
|
|
|
}
|
|
|
|
if (DeltaValid() && BaseValid()) {
|
2015-09-09 00:30:49 +02:00
|
|
|
if (comparator_->Equal(delta_iterator_->Entry().key,
|
|
|
|
base_iterator_->key())) {
|
2014-10-11 01:21:34 +02:00
|
|
|
equal_keys_ = true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
Advance();
|
|
|
|
}
|
|
|
|
|
|
|
|
void Prev() override {
|
|
|
|
if (!Valid()) {
|
|
|
|
status_ = Status::NotSupported("Prev() on invalid iterator");
|
|
|
|
}
|
|
|
|
|
|
|
|
if (forward_) {
|
|
|
|
// Need to change direction
|
|
|
|
// if our direction was backward and we're not equal, we have two states:
|
|
|
|
// * both iterators are valid: we're already in a good state (current
|
|
|
|
// shows to smaller)
|
|
|
|
// * only one iterator is valid: we need to advance that iterator
|
|
|
|
forward_ = false;
|
|
|
|
equal_keys_ = false;
|
|
|
|
if (!BaseValid()) {
|
|
|
|
assert(DeltaValid());
|
|
|
|
base_iterator_->SeekToLast();
|
|
|
|
} else if (!DeltaValid()) {
|
|
|
|
delta_iterator_->SeekToLast();
|
|
|
|
} else if (current_at_base_) {
|
|
|
|
// Change delta from less advanced than base to more advanced
|
|
|
|
AdvanceDelta();
|
|
|
|
} else {
|
|
|
|
// Change base from less advanced than delta to more advanced
|
|
|
|
AdvanceBase();
|
|
|
|
}
|
|
|
|
if (DeltaValid() && BaseValid()) {
|
2015-09-09 00:30:49 +02:00
|
|
|
if (comparator_->Equal(delta_iterator_->Entry().key,
|
|
|
|
base_iterator_->key())) {
|
2014-10-11 01:21:34 +02:00
|
|
|
equal_keys_ = true;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
Advance();
|
|
|
|
}
|
|
|
|
|
|
|
|
Slice key() const override {
|
|
|
|
return current_at_base_ ? base_iterator_->key()
|
|
|
|
: delta_iterator_->Entry().key;
|
|
|
|
}
|
|
|
|
|
|
|
|
Slice value() const override {
|
|
|
|
return current_at_base_ ? base_iterator_->value()
|
|
|
|
: delta_iterator_->Entry().value;
|
|
|
|
}
|
|
|
|
|
2015-02-26 20:28:41 +01:00
|
|
|
Status status() const override {
|
2014-10-11 01:21:34 +02:00
|
|
|
if (!status_.ok()) {
|
|
|
|
return status_;
|
|
|
|
}
|
|
|
|
if (!base_iterator_->status().ok()) {
|
|
|
|
return base_iterator_->status();
|
|
|
|
}
|
|
|
|
return delta_iterator_->status();
|
|
|
|
}
|
|
|
|
|
|
|
|
private:
|
|
|
|
void AssertInvariants() {
|
|
|
|
#ifndef NDEBUG
|
|
|
|
if (!Valid()) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
if (!BaseValid()) {
|
|
|
|
assert(!current_at_base_ && delta_iterator_->Valid());
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
if (!DeltaValid()) {
|
|
|
|
assert(current_at_base_ && base_iterator_->Valid());
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
// we don't support those yet
|
|
|
|
assert(delta_iterator_->Entry().type != kMergeRecord &&
|
|
|
|
delta_iterator_->Entry().type != kLogDataRecord);
|
|
|
|
int compare = comparator_->Compare(delta_iterator_->Entry().key,
|
|
|
|
base_iterator_->key());
|
|
|
|
if (forward_) {
|
|
|
|
// current_at_base -> compare < 0
|
|
|
|
assert(!current_at_base_ || compare < 0);
|
|
|
|
// !current_at_base -> compare <= 0
|
|
|
|
assert(current_at_base_ && compare >= 0);
|
|
|
|
} else {
|
|
|
|
// current_at_base -> compare > 0
|
|
|
|
assert(!current_at_base_ || compare > 0);
|
|
|
|
// !current_at_base -> compare <= 0
|
|
|
|
assert(current_at_base_ && compare <= 0);
|
|
|
|
}
|
|
|
|
// equal_keys_ <=> compare == 0
|
|
|
|
assert((equal_keys_ || compare != 0) && (!equal_keys_ || compare == 0));
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
|
|
|
void Advance() {
|
|
|
|
if (equal_keys_) {
|
|
|
|
assert(BaseValid() && DeltaValid());
|
|
|
|
AdvanceBase();
|
|
|
|
AdvanceDelta();
|
|
|
|
} else {
|
|
|
|
if (current_at_base_) {
|
|
|
|
assert(BaseValid());
|
|
|
|
AdvanceBase();
|
|
|
|
} else {
|
|
|
|
assert(DeltaValid());
|
|
|
|
AdvanceDelta();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
UpdateCurrent();
|
|
|
|
}
|
|
|
|
|
|
|
|
void AdvanceDelta() {
|
|
|
|
if (forward_) {
|
|
|
|
delta_iterator_->Next();
|
|
|
|
} else {
|
|
|
|
delta_iterator_->Prev();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
void AdvanceBase() {
|
|
|
|
if (forward_) {
|
|
|
|
base_iterator_->Next();
|
|
|
|
} else {
|
|
|
|
base_iterator_->Prev();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
bool BaseValid() const { return base_iterator_->Valid(); }
|
|
|
|
bool DeltaValid() const { return delta_iterator_->Valid(); }
|
|
|
|
void UpdateCurrent() {
|
2016-07-09 02:50:51 +02:00
|
|
|
// Suppress false positive clang analyzer warnings.
|
|
|
|
#ifndef __clang_analyzer__
|
2014-10-11 01:21:34 +02:00
|
|
|
while (true) {
|
2015-06-10 21:57:38 +02:00
|
|
|
WriteEntry delta_entry;
|
|
|
|
if (DeltaValid()) {
|
|
|
|
delta_entry = delta_iterator_->Entry();
|
|
|
|
}
|
2014-10-11 01:21:34 +02:00
|
|
|
equal_keys_ = false;
|
|
|
|
if (!BaseValid()) {
|
|
|
|
// Base has finished.
|
|
|
|
if (!DeltaValid()) {
|
|
|
|
// Finished
|
|
|
|
return;
|
|
|
|
}
|
2015-09-25 21:23:07 +02:00
|
|
|
if (delta_entry.type == kDeleteRecord ||
|
|
|
|
delta_entry.type == kSingleDeleteRecord) {
|
2014-10-11 01:21:34 +02:00
|
|
|
AdvanceDelta();
|
|
|
|
} else {
|
|
|
|
current_at_base_ = false;
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
} else if (!DeltaValid()) {
|
|
|
|
// Delta has finished.
|
|
|
|
current_at_base_ = true;
|
|
|
|
return;
|
|
|
|
} else {
|
2015-06-10 21:57:38 +02:00
|
|
|
int compare =
|
|
|
|
(forward_ ? 1 : -1) *
|
|
|
|
comparator_->Compare(delta_entry.key, base_iterator_->key());
|
2014-10-11 01:21:34 +02:00
|
|
|
if (compare <= 0) { // delta bigger or equal
|
|
|
|
if (compare == 0) {
|
|
|
|
equal_keys_ = true;
|
|
|
|
}
|
2015-09-25 21:23:07 +02:00
|
|
|
if (delta_entry.type != kDeleteRecord &&
|
|
|
|
delta_entry.type != kSingleDeleteRecord) {
|
2014-10-11 01:21:34 +02:00
|
|
|
current_at_base_ = false;
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
// Delta is less advanced and is delete.
|
|
|
|
AdvanceDelta();
|
|
|
|
if (equal_keys_) {
|
|
|
|
AdvanceBase();
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
current_at_base_ = true;
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
AssertInvariants();
|
2016-07-09 02:50:51 +02:00
|
|
|
#endif // __clang_analyzer__
|
2014-10-11 01:21:34 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
bool forward_;
|
|
|
|
bool current_at_base_;
|
|
|
|
bool equal_keys_;
|
|
|
|
Status status_;
|
|
|
|
std::unique_ptr<Iterator> base_iterator_;
|
|
|
|
std::unique_ptr<WBWIIterator> delta_iterator_;
|
|
|
|
const Comparator* comparator_; // not owned
|
|
|
|
};
|
|
|
|
|
2014-08-19 00:19:17 +02:00
|
|
|
typedef SkipList<WriteBatchIndexEntry*, const WriteBatchEntryComparator&>
|
|
|
|
WriteBatchEntrySkipList;
|
|
|
|
|
|
|
|
class WBWIIteratorImpl : public WBWIIterator {
|
|
|
|
public:
|
|
|
|
WBWIIteratorImpl(uint32_t column_family_id,
|
|
|
|
WriteBatchEntrySkipList* skip_list,
|
|
|
|
const ReadableWriteBatch* write_batch)
|
|
|
|
: column_family_id_(column_family_id),
|
|
|
|
skip_list_iter_(skip_list),
|
2015-06-10 21:57:38 +02:00
|
|
|
write_batch_(write_batch) {}
|
2014-08-19 00:19:17 +02:00
|
|
|
|
|
|
|
virtual ~WBWIIteratorImpl() {}
|
|
|
|
|
2015-06-10 21:57:38 +02:00
|
|
|
virtual bool Valid() const override {
|
|
|
|
if (!skip_list_iter_.Valid()) {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
const WriteBatchIndexEntry* iter_entry = skip_list_iter_.key();
|
|
|
|
return (iter_entry != nullptr &&
|
|
|
|
iter_entry->column_family == column_family_id_);
|
|
|
|
}
|
2014-08-19 00:19:17 +02:00
|
|
|
|
2015-02-26 20:28:41 +01:00
|
|
|
virtual void SeekToFirst() override {
|
2014-10-15 04:46:19 +02:00
|
|
|
WriteBatchIndexEntry search_entry(WriteBatchIndexEntry::kFlagMin,
|
2016-04-02 00:23:46 +02:00
|
|
|
column_family_id_, 0, 0);
|
2014-10-11 01:11:40 +02:00
|
|
|
skip_list_iter_.Seek(&search_entry);
|
|
|
|
}
|
|
|
|
|
2015-02-26 20:28:41 +01:00
|
|
|
virtual void SeekToLast() override {
|
2014-10-15 04:46:19 +02:00
|
|
|
WriteBatchIndexEntry search_entry(WriteBatchIndexEntry::kFlagMin,
|
2016-04-02 00:23:46 +02:00
|
|
|
column_family_id_ + 1, 0, 0);
|
2014-10-11 01:11:40 +02:00
|
|
|
skip_list_iter_.Seek(&search_entry);
|
|
|
|
if (!skip_list_iter_.Valid()) {
|
|
|
|
skip_list_iter_.SeekToLast();
|
|
|
|
} else {
|
|
|
|
skip_list_iter_.Prev();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-08-19 00:19:17 +02:00
|
|
|
virtual void Seek(const Slice& key) override {
|
|
|
|
WriteBatchIndexEntry search_entry(&key, column_family_id_);
|
|
|
|
skip_list_iter_.Seek(&search_entry);
|
|
|
|
}
|
|
|
|
|
2016-09-28 03:20:57 +02:00
|
|
|
virtual void SeekForPrev(const Slice& key) override {
|
|
|
|
WriteBatchIndexEntry search_entry(&key, column_family_id_);
|
|
|
|
skip_list_iter_.SeekForPrev(&search_entry);
|
|
|
|
}
|
|
|
|
|
2015-06-10 21:57:38 +02:00
|
|
|
virtual void Next() override { skip_list_iter_.Next(); }
|
2014-08-19 00:19:17 +02:00
|
|
|
|
2015-06-10 21:57:38 +02:00
|
|
|
virtual void Prev() override { skip_list_iter_.Prev(); }
|
2014-08-19 00:19:17 +02:00
|
|
|
|
2015-06-10 21:57:38 +02:00
|
|
|
virtual WriteEntry Entry() const override {
|
|
|
|
WriteEntry ret;
|
Modification of WriteBatch to support two phase commit
Summary: Adds three new WriteBatch data types: Prepare(xid), Commit(xid), Rollback(xid). Prepare(xid) should precede the (single) operation to which is applies. There can obviously be multiple Prepare(xid) markers. There should only be one Rollback(xid) or Commit(xid) marker yet not both. None of this logic is currently enforced and will most likely be implemented further up such as in the memtableinserter. All three markers are similar to PutLogData in that they are writebatch meta-data, ie stored but not counted. All three markers differ from PutLogData in that they will actually be written to disk. As for WriteBatchWithIndex, Prepare, Commit, Rollback are all implemented just as PutLogData and none are tested just as PutLogData.
Test Plan: single unit test in write_batch_test.
Reviewers: hermanlee4, sdong, anthony
Subscribers: leveldb, dhruba, vasilep, andrewkr
Differential Revision: https://reviews.facebook.net/D57867
2016-04-08 08:35:51 +02:00
|
|
|
Slice blob, xid;
|
2015-06-10 21:57:38 +02:00
|
|
|
const WriteBatchIndexEntry* iter_entry = skip_list_iter_.key();
|
|
|
|
// this is guaranteed with Valid()
|
|
|
|
assert(iter_entry != nullptr &&
|
|
|
|
iter_entry->column_family == column_family_id_);
|
Modification of WriteBatch to support two phase commit
Summary: Adds three new WriteBatch data types: Prepare(xid), Commit(xid), Rollback(xid). Prepare(xid) should precede the (single) operation to which is applies. There can obviously be multiple Prepare(xid) markers. There should only be one Rollback(xid) or Commit(xid) marker yet not both. None of this logic is currently enforced and will most likely be implemented further up such as in the memtableinserter. All three markers are similar to PutLogData in that they are writebatch meta-data, ie stored but not counted. All three markers differ from PutLogData in that they will actually be written to disk. As for WriteBatchWithIndex, Prepare, Commit, Rollback are all implemented just as PutLogData and none are tested just as PutLogData.
Test Plan: single unit test in write_batch_test.
Reviewers: hermanlee4, sdong, anthony
Subscribers: leveldb, dhruba, vasilep, andrewkr
Differential Revision: https://reviews.facebook.net/D57867
2016-04-08 08:35:51 +02:00
|
|
|
auto s = write_batch_->GetEntryFromDataOffset(
|
|
|
|
iter_entry->offset, &ret.type, &ret.key, &ret.value, &blob, &xid);
|
2015-06-10 21:57:38 +02:00
|
|
|
assert(s.ok());
|
|
|
|
assert(ret.type == kPutRecord || ret.type == kDeleteRecord ||
|
2016-08-16 17:16:04 +02:00
|
|
|
ret.type == kSingleDeleteRecord || ret.type == kDeleteRangeRecord ||
|
|
|
|
ret.type == kMergeRecord);
|
2015-06-10 21:57:38 +02:00
|
|
|
return ret;
|
|
|
|
}
|
2015-06-10 20:05:09 +02:00
|
|
|
|
2015-06-10 21:57:38 +02:00
|
|
|
virtual Status status() const override {
|
|
|
|
// this is in-memory data structure, so the only way status can be non-ok is
|
|
|
|
// through memory corruption
|
|
|
|
return Status::OK();
|
|
|
|
}
|
2015-06-10 20:05:09 +02:00
|
|
|
|
2014-10-10 22:31:28 +02:00
|
|
|
const WriteBatchIndexEntry* GetRawEntry() const {
|
|
|
|
return skip_list_iter_.key();
|
|
|
|
}
|
|
|
|
|
2014-08-19 00:19:17 +02:00
|
|
|
private:
|
|
|
|
uint32_t column_family_id_;
|
|
|
|
WriteBatchEntrySkipList::Iterator skip_list_iter_;
|
|
|
|
const ReadableWriteBatch* write_batch_;
|
|
|
|
};
|
2014-09-22 20:37:35 +02:00
|
|
|
|
|
|
|
struct WriteBatchWithIndex::Rep {
|
2017-08-23 19:01:17 +02:00
|
|
|
explicit Rep(const Comparator* index_comparator, size_t reserved_bytes = 0,
|
|
|
|
size_t max_bytes = 0, bool _overwrite_key = false)
|
2017-04-11 00:38:34 +02:00
|
|
|
: write_batch(reserved_bytes, max_bytes),
|
2014-09-22 20:37:35 +02:00
|
|
|
comparator(index_comparator, &write_batch),
|
2014-10-10 22:31:28 +02:00
|
|
|
skip_list(comparator, &arena),
|
2014-10-31 19:59:54 +01:00
|
|
|
overwrite_key(_overwrite_key),
|
2014-10-10 22:31:28 +02:00
|
|
|
last_entry_offset(0) {}
|
2014-09-22 20:37:35 +02:00
|
|
|
ReadableWriteBatch write_batch;
|
|
|
|
WriteBatchEntryComparator comparator;
|
|
|
|
Arena arena;
|
|
|
|
WriteBatchEntrySkipList skip_list;
|
2014-10-10 22:31:28 +02:00
|
|
|
bool overwrite_key;
|
|
|
|
size_t last_entry_offset;
|
|
|
|
|
|
|
|
// Remember current offset of internal write batch, which is used as
|
|
|
|
// the starting offset of the next record.
|
|
|
|
void SetLastEntryOffset() { last_entry_offset = write_batch.GetDataSize(); }
|
|
|
|
|
|
|
|
// In overwrite mode, find the existing entry for the same key and update it
|
|
|
|
// to point to the current entry.
|
|
|
|
// Return true if the key is found and updated.
|
|
|
|
bool UpdateExistingEntry(ColumnFamilyHandle* column_family, const Slice& key);
|
|
|
|
bool UpdateExistingEntryWithCfId(uint32_t column_family_id, const Slice& key);
|
|
|
|
|
|
|
|
// Add the recent entry to the update.
|
|
|
|
// In overwrite mode, if key already exists in the index, update it.
|
|
|
|
void AddOrUpdateIndex(ColumnFamilyHandle* column_family, const Slice& key);
|
|
|
|
void AddOrUpdateIndex(const Slice& key);
|
|
|
|
|
|
|
|
// Allocate an index entry pointing to the last entry in the write batch and
|
|
|
|
// put it to skip list.
|
|
|
|
void AddNewEntry(uint32_t column_family_id);
|
2015-02-24 02:49:23 +01:00
|
|
|
|
|
|
|
// Clear all updates buffered in this batch.
|
|
|
|
void Clear();
|
2015-07-11 05:15:45 +02:00
|
|
|
void ClearIndex();
|
|
|
|
|
|
|
|
// Rebuild index by reading all records from the batch.
|
|
|
|
// Returns non-ok status on corruption.
|
|
|
|
Status ReBuildIndex();
|
2014-10-10 22:31:28 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
bool WriteBatchWithIndex::Rep::UpdateExistingEntry(
|
|
|
|
ColumnFamilyHandle* column_family, const Slice& key) {
|
|
|
|
uint32_t cf_id = GetColumnFamilyID(column_family);
|
|
|
|
return UpdateExistingEntryWithCfId(cf_id, key);
|
|
|
|
}
|
|
|
|
|
|
|
|
bool WriteBatchWithIndex::Rep::UpdateExistingEntryWithCfId(
|
|
|
|
uint32_t column_family_id, const Slice& key) {
|
|
|
|
if (!overwrite_key) {
|
|
|
|
return false;
|
|
|
|
}
|
2014-09-22 20:37:35 +02:00
|
|
|
|
2014-10-10 22:31:28 +02:00
|
|
|
WBWIIteratorImpl iter(column_family_id, &skip_list, &write_batch);
|
|
|
|
iter.Seek(key);
|
|
|
|
if (!iter.Valid()) {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
if (comparator.CompareKey(column_family_id, key, iter.Entry().key) != 0) {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
WriteBatchIndexEntry* non_const_entry =
|
|
|
|
const_cast<WriteBatchIndexEntry*>(iter.GetRawEntry());
|
|
|
|
non_const_entry->offset = last_entry_offset;
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
void WriteBatchWithIndex::Rep::AddOrUpdateIndex(
|
|
|
|
ColumnFamilyHandle* column_family, const Slice& key) {
|
|
|
|
if (!UpdateExistingEntry(column_family, key)) {
|
2014-09-22 20:37:35 +02:00
|
|
|
uint32_t cf_id = GetColumnFamilyID(column_family);
|
|
|
|
const auto* cf_cmp = GetColumnFamilyUserComparator(column_family);
|
|
|
|
if (cf_cmp != nullptr) {
|
|
|
|
comparator.SetComparatorForCF(cf_id, cf_cmp);
|
|
|
|
}
|
2014-10-10 22:31:28 +02:00
|
|
|
AddNewEntry(cf_id);
|
|
|
|
}
|
|
|
|
}
|
2014-09-22 20:37:35 +02:00
|
|
|
|
2014-10-10 22:31:28 +02:00
|
|
|
void WriteBatchWithIndex::Rep::AddOrUpdateIndex(const Slice& key) {
|
|
|
|
if (!UpdateExistingEntryWithCfId(0, key)) {
|
|
|
|
AddNewEntry(0);
|
2014-09-22 20:37:35 +02:00
|
|
|
}
|
2014-10-10 22:31:28 +02:00
|
|
|
}
|
2014-09-22 20:37:35 +02:00
|
|
|
|
2014-10-10 22:31:28 +02:00
|
|
|
void WriteBatchWithIndex::Rep::AddNewEntry(uint32_t column_family_id) {
|
2016-04-02 00:23:46 +02:00
|
|
|
const std::string& wb_data = write_batch.Data();
|
|
|
|
Slice entry_ptr = Slice(wb_data.data() + last_entry_offset,
|
|
|
|
wb_data.size() - last_entry_offset);
|
|
|
|
// Extract key
|
|
|
|
Slice key;
|
2016-04-04 20:10:46 +02:00
|
|
|
bool success __attribute__((__unused__)) =
|
|
|
|
ReadKeyFromWriteBatchEntry(&entry_ptr, &key, column_family_id != 0);
|
2016-04-02 00:23:46 +02:00
|
|
|
assert(success);
|
|
|
|
|
2014-09-22 20:37:35 +02:00
|
|
|
auto* mem = arena.Allocate(sizeof(WriteBatchIndexEntry));
|
2014-10-10 22:31:28 +02:00
|
|
|
auto* index_entry =
|
2016-04-02 00:23:46 +02:00
|
|
|
new (mem) WriteBatchIndexEntry(last_entry_offset, column_family_id,
|
|
|
|
key.data() - wb_data.data(), key.size());
|
2014-10-10 22:31:28 +02:00
|
|
|
skip_list.Insert(index_entry);
|
2014-09-22 20:37:35 +02:00
|
|
|
}
|
2014-08-19 00:19:17 +02:00
|
|
|
|
2015-02-24 02:49:23 +01:00
|
|
|
void WriteBatchWithIndex::Rep::Clear() {
|
|
|
|
write_batch.Clear();
|
2015-07-11 05:15:45 +02:00
|
|
|
ClearIndex();
|
|
|
|
}
|
|
|
|
|
|
|
|
void WriteBatchWithIndex::Rep::ClearIndex() {
|
|
|
|
skip_list.~WriteBatchEntrySkipList();
|
2015-02-24 02:49:23 +01:00
|
|
|
arena.~Arena();
|
|
|
|
new (&arena) Arena();
|
|
|
|
new (&skip_list) WriteBatchEntrySkipList(comparator, &arena);
|
|
|
|
last_entry_offset = 0;
|
|
|
|
}
|
|
|
|
|
2015-07-11 05:15:45 +02:00
|
|
|
Status WriteBatchWithIndex::Rep::ReBuildIndex() {
|
|
|
|
Status s;
|
|
|
|
|
|
|
|
ClearIndex();
|
|
|
|
|
|
|
|
if (write_batch.Count() == 0) {
|
|
|
|
// Nothing to re-index
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
size_t offset = WriteBatchInternal::GetFirstOffset(&write_batch);
|
|
|
|
|
|
|
|
Slice input(write_batch.Data());
|
|
|
|
input.remove_prefix(offset);
|
|
|
|
|
|
|
|
// Loop through all entries in Rep and add each one to the index
|
|
|
|
int found = 0;
|
|
|
|
while (s.ok() && !input.empty()) {
|
Modification of WriteBatch to support two phase commit
Summary: Adds three new WriteBatch data types: Prepare(xid), Commit(xid), Rollback(xid). Prepare(xid) should precede the (single) operation to which is applies. There can obviously be multiple Prepare(xid) markers. There should only be one Rollback(xid) or Commit(xid) marker yet not both. None of this logic is currently enforced and will most likely be implemented further up such as in the memtableinserter. All three markers are similar to PutLogData in that they are writebatch meta-data, ie stored but not counted. All three markers differ from PutLogData in that they will actually be written to disk. As for WriteBatchWithIndex, Prepare, Commit, Rollback are all implemented just as PutLogData and none are tested just as PutLogData.
Test Plan: single unit test in write_batch_test.
Reviewers: hermanlee4, sdong, anthony
Subscribers: leveldb, dhruba, vasilep, andrewkr
Differential Revision: https://reviews.facebook.net/D57867
2016-04-08 08:35:51 +02:00
|
|
|
Slice key, value, blob, xid;
|
2015-07-11 05:15:45 +02:00
|
|
|
uint32_t column_family_id = 0; // default
|
|
|
|
char tag = 0;
|
|
|
|
|
|
|
|
// set offset of current entry for call to AddNewEntry()
|
|
|
|
last_entry_offset = input.data() - write_batch.Data().data();
|
|
|
|
|
|
|
|
s = ReadRecordFromWriteBatch(&input, &tag, &column_family_id, &key,
|
Modification of WriteBatch to support two phase commit
Summary: Adds three new WriteBatch data types: Prepare(xid), Commit(xid), Rollback(xid). Prepare(xid) should precede the (single) operation to which is applies. There can obviously be multiple Prepare(xid) markers. There should only be one Rollback(xid) or Commit(xid) marker yet not both. None of this logic is currently enforced and will most likely be implemented further up such as in the memtableinserter. All three markers are similar to PutLogData in that they are writebatch meta-data, ie stored but not counted. All three markers differ from PutLogData in that they will actually be written to disk. As for WriteBatchWithIndex, Prepare, Commit, Rollback are all implemented just as PutLogData and none are tested just as PutLogData.
Test Plan: single unit test in write_batch_test.
Reviewers: hermanlee4, sdong, anthony
Subscribers: leveldb, dhruba, vasilep, andrewkr
Differential Revision: https://reviews.facebook.net/D57867
2016-04-08 08:35:51 +02:00
|
|
|
&value, &blob, &xid);
|
2015-07-11 05:15:45 +02:00
|
|
|
if (!s.ok()) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
switch (tag) {
|
|
|
|
case kTypeColumnFamilyValue:
|
|
|
|
case kTypeValue:
|
|
|
|
case kTypeColumnFamilyDeletion:
|
|
|
|
case kTypeDeletion:
|
2015-09-25 21:23:07 +02:00
|
|
|
case kTypeColumnFamilySingleDeletion:
|
|
|
|
case kTypeSingleDeletion:
|
2015-07-11 05:15:45 +02:00
|
|
|
case kTypeColumnFamilyMerge:
|
|
|
|
case kTypeMerge:
|
|
|
|
found++;
|
|
|
|
if (!UpdateExistingEntryWithCfId(column_family_id, key)) {
|
|
|
|
AddNewEntry(column_family_id);
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
case kTypeLogData:
|
Modification of WriteBatch to support two phase commit
Summary: Adds three new WriteBatch data types: Prepare(xid), Commit(xid), Rollback(xid). Prepare(xid) should precede the (single) operation to which is applies. There can obviously be multiple Prepare(xid) markers. There should only be one Rollback(xid) or Commit(xid) marker yet not both. None of this logic is currently enforced and will most likely be implemented further up such as in the memtableinserter. All three markers are similar to PutLogData in that they are writebatch meta-data, ie stored but not counted. All three markers differ from PutLogData in that they will actually be written to disk. As for WriteBatchWithIndex, Prepare, Commit, Rollback are all implemented just as PutLogData and none are tested just as PutLogData.
Test Plan: single unit test in write_batch_test.
Reviewers: hermanlee4, sdong, anthony
Subscribers: leveldb, dhruba, vasilep, andrewkr
Differential Revision: https://reviews.facebook.net/D57867
2016-04-08 08:35:51 +02:00
|
|
|
case kTypeBeginPrepareXID:
|
|
|
|
case kTypeEndPrepareXID:
|
|
|
|
case kTypeCommitXID:
|
|
|
|
case kTypeRollbackXID:
|
|
|
|
case kTypeNoop:
|
2015-07-11 05:15:45 +02:00
|
|
|
break;
|
|
|
|
default:
|
|
|
|
return Status::Corruption("unknown WriteBatch tag");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (s.ok() && found != write_batch.Count()) {
|
|
|
|
s = Status::Corruption("WriteBatch has wrong count");
|
|
|
|
}
|
|
|
|
|
|
|
|
return s;
|
|
|
|
}
|
2014-08-19 00:19:17 +02:00
|
|
|
|
2017-04-11 00:38:34 +02:00
|
|
|
WriteBatchWithIndex::WriteBatchWithIndex(
|
|
|
|
const Comparator* default_index_comparator, size_t reserved_bytes,
|
|
|
|
bool overwrite_key, size_t max_bytes)
|
|
|
|
: rep(new Rep(default_index_comparator, reserved_bytes, max_bytes,
|
|
|
|
overwrite_key)) {}
|
2014-08-19 00:19:17 +02:00
|
|
|
|
2017-04-11 00:38:34 +02:00
|
|
|
WriteBatchWithIndex::~WriteBatchWithIndex() {}
|
2014-08-19 00:19:17 +02:00
|
|
|
|
2017-04-11 00:38:34 +02:00
|
|
|
WriteBatch* WriteBatchWithIndex::GetWriteBatch() { return &rep->write_batch; }
|
2014-08-19 00:19:17 +02:00
|
|
|
|
2017-04-11 00:38:34 +02:00
|
|
|
WBWIIterator* WriteBatchWithIndex::NewIterator() {
|
|
|
|
return new WBWIIteratorImpl(0, &(rep->skip_list), &rep->write_batch);
|
2014-08-19 00:19:17 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
WBWIIterator* WriteBatchWithIndex::NewIterator(
|
|
|
|
ColumnFamilyHandle* column_family) {
|
|
|
|
return new WBWIIteratorImpl(GetColumnFamilyID(column_family),
|
|
|
|
&(rep->skip_list), &rep->write_batch);
|
|
|
|
}
|
|
|
|
|
2014-10-11 01:21:34 +02:00
|
|
|
Iterator* WriteBatchWithIndex::NewIteratorWithBase(
|
|
|
|
ColumnFamilyHandle* column_family, Iterator* base_iterator) {
|
|
|
|
if (rep->overwrite_key == false) {
|
|
|
|
assert(false);
|
|
|
|
return nullptr;
|
|
|
|
}
|
|
|
|
return new BaseDeltaIterator(base_iterator, NewIterator(column_family),
|
|
|
|
GetColumnFamilyUserComparator(column_family));
|
|
|
|
}
|
|
|
|
|
2015-02-03 07:29:43 +01:00
|
|
|
Iterator* WriteBatchWithIndex::NewIteratorWithBase(Iterator* base_iterator) {
|
|
|
|
if (rep->overwrite_key == false) {
|
|
|
|
assert(false);
|
|
|
|
return nullptr;
|
|
|
|
}
|
|
|
|
// default column family's comparator
|
|
|
|
return new BaseDeltaIterator(base_iterator, NewIterator(),
|
|
|
|
rep->comparator.default_comparator());
|
|
|
|
}
|
|
|
|
|
2017-04-11 00:38:34 +02:00
|
|
|
Status WriteBatchWithIndex::Put(ColumnFamilyHandle* column_family,
|
|
|
|
const Slice& key, const Slice& value) {
|
2014-10-10 22:31:28 +02:00
|
|
|
rep->SetLastEntryOffset();
|
2017-04-11 00:38:34 +02:00
|
|
|
auto s = rep->write_batch.Put(column_family, key, value);
|
|
|
|
if (s.ok()) {
|
|
|
|
rep->AddOrUpdateIndex(column_family, key);
|
|
|
|
}
|
|
|
|
return s;
|
2014-08-19 00:19:17 +02:00
|
|
|
}
|
|
|
|
|
2017-04-11 00:38:34 +02:00
|
|
|
Status WriteBatchWithIndex::Put(const Slice& key, const Slice& value) {
|
2014-10-10 22:31:28 +02:00
|
|
|
rep->SetLastEntryOffset();
|
2017-04-11 00:38:34 +02:00
|
|
|
auto s = rep->write_batch.Put(key, value);
|
|
|
|
if (s.ok()) {
|
|
|
|
rep->AddOrUpdateIndex(key);
|
|
|
|
}
|
|
|
|
return s;
|
2014-08-19 00:19:17 +02:00
|
|
|
}
|
|
|
|
|
2017-04-11 00:38:34 +02:00
|
|
|
Status WriteBatchWithIndex::Delete(ColumnFamilyHandle* column_family,
|
|
|
|
const Slice& key) {
|
2014-10-10 22:31:28 +02:00
|
|
|
rep->SetLastEntryOffset();
|
2017-04-11 00:38:34 +02:00
|
|
|
auto s = rep->write_batch.Delete(column_family, key);
|
|
|
|
if (s.ok()) {
|
|
|
|
rep->AddOrUpdateIndex(column_family, key);
|
|
|
|
}
|
|
|
|
return s;
|
2014-08-19 00:19:17 +02:00
|
|
|
}
|
|
|
|
|
2017-04-11 00:38:34 +02:00
|
|
|
Status WriteBatchWithIndex::Delete(const Slice& key) {
|
2014-10-10 22:31:28 +02:00
|
|
|
rep->SetLastEntryOffset();
|
2017-04-11 00:38:34 +02:00
|
|
|
auto s = rep->write_batch.Delete(key);
|
|
|
|
if (s.ok()) {
|
|
|
|
rep->AddOrUpdateIndex(key);
|
|
|
|
}
|
|
|
|
return s;
|
2014-08-19 00:19:17 +02:00
|
|
|
}
|
|
|
|
|
2017-04-11 00:38:34 +02:00
|
|
|
Status WriteBatchWithIndex::SingleDelete(ColumnFamilyHandle* column_family,
|
|
|
|
const Slice& key) {
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
2015-09-17 20:42:56 +02:00
|
|
|
rep->SetLastEntryOffset();
|
2017-04-11 00:38:34 +02:00
|
|
|
auto s = rep->write_batch.SingleDelete(column_family, key);
|
|
|
|
if (s.ok()) {
|
|
|
|
rep->AddOrUpdateIndex(column_family, key);
|
|
|
|
}
|
|
|
|
return s;
|
2014-08-19 00:19:17 +02:00
|
|
|
}
|
|
|
|
|
2017-04-11 00:38:34 +02:00
|
|
|
Status WriteBatchWithIndex::SingleDelete(const Slice& key) {
|
2014-10-10 22:31:28 +02:00
|
|
|
rep->SetLastEntryOffset();
|
2017-04-11 00:38:34 +02:00
|
|
|
auto s = rep->write_batch.SingleDelete(key);
|
|
|
|
if (s.ok()) {
|
|
|
|
rep->AddOrUpdateIndex(key);
|
|
|
|
}
|
|
|
|
return s;
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
2015-09-17 20:42:56 +02:00
|
|
|
}
|
|
|
|
|
2017-04-11 00:38:34 +02:00
|
|
|
Status WriteBatchWithIndex::DeleteRange(ColumnFamilyHandle* column_family,
|
|
|
|
const Slice& begin_key,
|
|
|
|
const Slice& end_key) {
|
2016-08-16 17:16:04 +02:00
|
|
|
rep->SetLastEntryOffset();
|
2017-04-11 00:38:34 +02:00
|
|
|
auto s = rep->write_batch.DeleteRange(column_family, begin_key, end_key);
|
|
|
|
if (s.ok()) {
|
|
|
|
rep->AddOrUpdateIndex(column_family, begin_key);
|
|
|
|
}
|
|
|
|
return s;
|
2016-08-16 17:16:04 +02:00
|
|
|
}
|
|
|
|
|
2017-04-11 00:38:34 +02:00
|
|
|
Status WriteBatchWithIndex::DeleteRange(const Slice& begin_key,
|
|
|
|
const Slice& end_key) {
|
2016-08-16 17:16:04 +02:00
|
|
|
rep->SetLastEntryOffset();
|
2017-04-11 00:38:34 +02:00
|
|
|
auto s = rep->write_batch.DeleteRange(begin_key, end_key);
|
|
|
|
if (s.ok()) {
|
|
|
|
rep->AddOrUpdateIndex(begin_key);
|
|
|
|
}
|
|
|
|
return s;
|
2016-08-16 17:16:04 +02:00
|
|
|
}
|
|
|
|
|
2017-04-11 00:38:34 +02:00
|
|
|
Status WriteBatchWithIndex::Merge(ColumnFamilyHandle* column_family,
|
|
|
|
const Slice& key, const Slice& value) {
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
2015-09-17 20:42:56 +02:00
|
|
|
rep->SetLastEntryOffset();
|
2017-04-11 00:38:34 +02:00
|
|
|
auto s = rep->write_batch.Merge(column_family, key, value);
|
|
|
|
if (s.ok()) {
|
|
|
|
rep->AddOrUpdateIndex(column_family, key);
|
|
|
|
}
|
|
|
|
return s;
|
2014-08-19 00:19:17 +02:00
|
|
|
}
|
|
|
|
|
2017-04-11 00:38:34 +02:00
|
|
|
Status WriteBatchWithIndex::Merge(const Slice& key, const Slice& value) {
|
2014-10-10 22:31:28 +02:00
|
|
|
rep->SetLastEntryOffset();
|
2017-04-11 00:38:34 +02:00
|
|
|
auto s = rep->write_batch.Merge(key, value);
|
|
|
|
if (s.ok()) {
|
|
|
|
rep->AddOrUpdateIndex(key);
|
|
|
|
}
|
|
|
|
return s;
|
2014-08-19 00:19:17 +02:00
|
|
|
}
|
|
|
|
|
2017-04-11 00:38:34 +02:00
|
|
|
Status WriteBatchWithIndex::PutLogData(const Slice& blob) {
|
|
|
|
return rep->write_batch.PutLogData(blob);
|
Support for SingleDelete()
Summary:
This patch fixes #7460559. It introduces SingleDelete as a new database
operation. This operation can be used to delete keys that were never
overwritten (no put following another put of the same key). If an overwritten
key is single deleted the behavior is undefined. Single deletion of a
non-existent key has no effect but multiple consecutive single deletions are
not allowed (see limitations).
In contrast to the conventional Delete() operation, the deletion entry is
removed along with the value when the two are lined up in a compaction. Note:
The semantics are similar to @igor's prototype that allowed to have this
behavior on the granularity of a column family (
https://reviews.facebook.net/D42093 ). This new patch, however, is more
aggressive when it comes to removing tombstones: It removes the SingleDelete
together with the value whenever there is no snapshot between them while the
older patch only did this when the sequence number of the deletion was older
than the earliest snapshot.
Most of the complex additions are in the Compaction Iterator, all other changes
should be relatively straightforward. The patch also includes basic support for
single deletions in db_stress and db_bench.
Limitations:
- Not compatible with cuckoo hash tables
- Single deletions cannot be used in combination with merges and normal
deletions on the same key (other keys are not affected by this)
- Consecutive single deletions are currently not allowed (and older version of
this patch supported this so it could be resurrected if needed)
Test Plan: make all check
Reviewers: yhchiang, sdong, rven, anthony, yoshinorim, igor
Reviewed By: igor
Subscribers: maykov, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43179
2015-09-17 20:42:56 +02:00
|
|
|
}
|
|
|
|
|
2015-02-24 02:49:23 +01:00
|
|
|
void WriteBatchWithIndex::Clear() { rep->Clear(); }
|
|
|
|
|
2015-05-11 23:51:51 +02:00
|
|
|
Status WriteBatchWithIndex::GetFromBatch(ColumnFamilyHandle* column_family,
|
|
|
|
const DBOptions& options,
|
|
|
|
const Slice& key, std::string* value) {
|
|
|
|
Status s;
|
|
|
|
MergeContext merge_context;
|
2016-10-18 22:19:26 +02:00
|
|
|
const ImmutableDBOptions immuable_db_options(options);
|
2015-05-11 23:51:51 +02:00
|
|
|
|
|
|
|
WriteBatchWithIndexInternal::Result result =
|
2015-09-29 04:24:44 +02:00
|
|
|
WriteBatchWithIndexInternal::GetFromBatch(
|
2016-10-18 22:19:26 +02:00
|
|
|
immuable_db_options, this, column_family, key, &merge_context,
|
|
|
|
&rep->comparator, value, rep->overwrite_key, &s);
|
2015-05-11 23:51:51 +02:00
|
|
|
|
|
|
|
switch (result) {
|
|
|
|
case WriteBatchWithIndexInternal::Result::kFound:
|
|
|
|
case WriteBatchWithIndexInternal::Result::kError:
|
Pessimistic Transactions
Summary:
Initial implementation of Pessimistic Transactions. This diff contains the api changes discussed in D38913. This diff is pretty large, so let me know if people would prefer to meet up to discuss it.
MyRocks folks: please take a look at the API in include/rocksdb/utilities/transaction[_db].h and let me know if you have any issues.
Also, you'll notice a couple of TODOs in the implementation of RollbackToSavePoint(). After chatting with Siying, I'm going to send out a separate diff for an alternate implementation of this feature that implements the rollback inside of WriteBatch/WriteBatchWithIndex. We can then decide which route is preferable.
Next, I'm planning on doing some perf testing and then integrating this diff into MongoRocks for further testing.
Test Plan: Unit tests, db_bench parallel testing.
Reviewers: igor, rven, sdong, yhchiang, yoshinorim
Reviewed By: sdong
Subscribers: hermanlee4, maykov, spetrunia, leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D40869
2015-05-26 02:37:33 +02:00
|
|
|
// use returned status
|
|
|
|
break;
|
2015-05-11 23:51:51 +02:00
|
|
|
case WriteBatchWithIndexInternal::Result::kDeleted:
|
|
|
|
case WriteBatchWithIndexInternal::Result::kNotFound:
|
Pessimistic Transactions
Summary:
Initial implementation of Pessimistic Transactions. This diff contains the api changes discussed in D38913. This diff is pretty large, so let me know if people would prefer to meet up to discuss it.
MyRocks folks: please take a look at the API in include/rocksdb/utilities/transaction[_db].h and let me know if you have any issues.
Also, you'll notice a couple of TODOs in the implementation of RollbackToSavePoint(). After chatting with Siying, I'm going to send out a separate diff for an alternate implementation of this feature that implements the rollback inside of WriteBatch/WriteBatchWithIndex. We can then decide which route is preferable.
Next, I'm planning on doing some perf testing and then integrating this diff into MongoRocks for further testing.
Test Plan: Unit tests, db_bench parallel testing.
Reviewers: igor, rven, sdong, yhchiang, yoshinorim
Reviewed By: sdong
Subscribers: hermanlee4, maykov, spetrunia, leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D40869
2015-05-26 02:37:33 +02:00
|
|
|
s = Status::NotFound();
|
|
|
|
break;
|
2015-05-11 23:51:51 +02:00
|
|
|
case WriteBatchWithIndexInternal::Result::kMergeInProgress:
|
Simplify querying of merge results
Summary:
While working on supporting mixing merge operators with
single deletes ( https://reviews.facebook.net/D43179 ),
I realized that returning and dealing with merge results
can be made simpler. Submitting this as a separate diff
because it is not directly related to single deletes.
Before, callers of merge helper had to retrieve the merge
result in one of two ways depending on whether the merge
was successful or not (success = result of merge was single
kTypeValue). For successful merges, the caller could query
the resulting key/value pair and for unsuccessful merges,
the result could be retrieved in the form of two deques of
keys and values. However, with single deletes, a successful merge
does not return a single key/value pair (if merge
operands are merged with a single delete, we have to generate
a value and keep the original single delete around to make
sure that we are not accidentially producing a key overwrite).
In addition, the two existing call sites of the merge
helper were taking the same actions independently from whether
the merge was successful or not, so this patch simplifies that.
Test Plan: make clean all check
Reviewers: rven, sdong, yhchiang, anthony, igor
Reviewed By: igor
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D43353
2015-08-18 02:34:38 +02:00
|
|
|
s = Status::MergeInProgress();
|
Pessimistic Transactions
Summary:
Initial implementation of Pessimistic Transactions. This diff contains the api changes discussed in D38913. This diff is pretty large, so let me know if people would prefer to meet up to discuss it.
MyRocks folks: please take a look at the API in include/rocksdb/utilities/transaction[_db].h and let me know if you have any issues.
Also, you'll notice a couple of TODOs in the implementation of RollbackToSavePoint(). After chatting with Siying, I'm going to send out a separate diff for an alternate implementation of this feature that implements the rollback inside of WriteBatch/WriteBatchWithIndex. We can then decide which route is preferable.
Next, I'm planning on doing some perf testing and then integrating this diff into MongoRocks for further testing.
Test Plan: Unit tests, db_bench parallel testing.
Reviewers: igor, rven, sdong, yhchiang, yoshinorim
Reviewed By: sdong
Subscribers: hermanlee4, maykov, spetrunia, leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D40869
2015-05-26 02:37:33 +02:00
|
|
|
break;
|
2015-05-11 23:51:51 +02:00
|
|
|
default:
|
|
|
|
assert(false);
|
2014-08-19 00:19:17 +02:00
|
|
|
}
|
|
|
|
|
2015-05-11 23:51:51 +02:00
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
Status WriteBatchWithIndex::GetFromBatchAndDB(DB* db,
|
|
|
|
const ReadOptions& read_options,
|
|
|
|
const Slice& key,
|
|
|
|
std::string* value) {
|
2017-08-23 19:01:17 +02:00
|
|
|
assert(value != nullptr);
|
|
|
|
PinnableSlice pinnable_val(value);
|
|
|
|
assert(!pinnable_val.IsPinned());
|
|
|
|
auto s = GetFromBatchAndDB(db, read_options, db->DefaultColumnFamily(), key,
|
|
|
|
&pinnable_val);
|
|
|
|
if (s.ok() && pinnable_val.IsPinned()) {
|
|
|
|
value->assign(pinnable_val.data(), pinnable_val.size());
|
|
|
|
} // else value is already assigned
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
Status WriteBatchWithIndex::GetFromBatchAndDB(DB* db,
|
|
|
|
const ReadOptions& read_options,
|
|
|
|
const Slice& key,
|
|
|
|
PinnableSlice* pinnable_val) {
|
2015-05-11 23:51:51 +02:00
|
|
|
return GetFromBatchAndDB(db, read_options, db->DefaultColumnFamily(), key,
|
2017-08-23 19:01:17 +02:00
|
|
|
pinnable_val);
|
2015-05-11 23:51:51 +02:00
|
|
|
}
|
2014-10-15 04:46:19 +02:00
|
|
|
|
2015-05-11 23:51:51 +02:00
|
|
|
Status WriteBatchWithIndex::GetFromBatchAndDB(DB* db,
|
|
|
|
const ReadOptions& read_options,
|
|
|
|
ColumnFamilyHandle* column_family,
|
|
|
|
const Slice& key,
|
|
|
|
std::string* value) {
|
2017-08-23 19:01:17 +02:00
|
|
|
assert(value != nullptr);
|
|
|
|
PinnableSlice pinnable_val(value);
|
|
|
|
assert(!pinnable_val.IsPinned());
|
|
|
|
auto s =
|
|
|
|
GetFromBatchAndDB(db, read_options, column_family, key, &pinnable_val);
|
|
|
|
if (s.ok() && pinnable_val.IsPinned()) {
|
|
|
|
value->assign(pinnable_val.data(), pinnable_val.size());
|
|
|
|
} // else value is already assigned
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
Status WriteBatchWithIndex::GetFromBatchAndDB(DB* db,
|
|
|
|
const ReadOptions& read_options,
|
|
|
|
ColumnFamilyHandle* column_family,
|
|
|
|
const Slice& key,
|
|
|
|
PinnableSlice* pinnable_val) {
|
2017-09-11 17:58:52 +02:00
|
|
|
return GetFromBatchAndDB(db, read_options, column_family, key, pinnable_val,
|
|
|
|
nullptr);
|
|
|
|
}
|
|
|
|
|
|
|
|
Status WriteBatchWithIndex::GetFromBatchAndDB(
|
|
|
|
DB* db, const ReadOptions& read_options, ColumnFamilyHandle* column_family,
|
|
|
|
const Slice& key, PinnableSlice* pinnable_val, ReadCallback* callback) {
|
|
|
|
if (UNLIKELY(db->GetRootDB() != db)) {
|
|
|
|
return Status::NotSupported("The DB must be of DBImpl type");
|
|
|
|
// Otherwise the cast below would fail
|
|
|
|
}
|
2014-08-19 00:19:17 +02:00
|
|
|
Status s;
|
2015-05-11 23:51:51 +02:00
|
|
|
MergeContext merge_context;
|
2016-10-18 22:19:26 +02:00
|
|
|
const ImmutableDBOptions& immuable_db_options =
|
|
|
|
reinterpret_cast<DBImpl*>(db)->immutable_db_options();
|
2015-05-11 23:51:51 +02:00
|
|
|
|
2017-08-23 19:01:17 +02:00
|
|
|
// Since the lifetime of the WriteBatch is the same as that of the transaction
|
|
|
|
// we cannot pin it as otherwise the returned value will not be available
|
|
|
|
// after the transaction finishes.
|
|
|
|
std::string& batch_value = *pinnable_val->GetSelf();
|
2015-05-11 23:51:51 +02:00
|
|
|
WriteBatchWithIndexInternal::Result result =
|
|
|
|
WriteBatchWithIndexInternal::GetFromBatch(
|
2016-10-18 22:19:26 +02:00
|
|
|
immuable_db_options, this, column_family, key, &merge_context,
|
|
|
|
&rep->comparator, &batch_value, rep->overwrite_key, &s);
|
2015-05-11 23:51:51 +02:00
|
|
|
|
|
|
|
if (result == WriteBatchWithIndexInternal::Result::kFound) {
|
2017-08-23 19:01:17 +02:00
|
|
|
pinnable_val->PinSelf();
|
2015-05-11 23:51:51 +02:00
|
|
|
return s;
|
|
|
|
}
|
|
|
|
if (result == WriteBatchWithIndexInternal::Result::kDeleted) {
|
|
|
|
return Status::NotFound();
|
|
|
|
}
|
|
|
|
if (result == WriteBatchWithIndexInternal::Result::kError) {
|
|
|
|
return s;
|
|
|
|
}
|
2015-09-29 04:24:44 +02:00
|
|
|
if (result == WriteBatchWithIndexInternal::Result::kMergeInProgress &&
|
|
|
|
rep->overwrite_key == true) {
|
|
|
|
// Since we've overwritten keys, we do not know what other operations are
|
|
|
|
// in this batch for this key, so we cannot do a Merge to compute the
|
|
|
|
// result. Instead, we will simply return MergeInProgress.
|
|
|
|
return Status::MergeInProgress();
|
|
|
|
}
|
|
|
|
|
2015-05-11 23:51:51 +02:00
|
|
|
assert(result == WriteBatchWithIndexInternal::Result::kMergeInProgress ||
|
|
|
|
result == WriteBatchWithIndexInternal::Result::kNotFound);
|
|
|
|
|
|
|
|
// Did not find key in batch OR could not resolve Merges. Try DB.
|
2017-09-11 17:58:52 +02:00
|
|
|
if (!callback) {
|
|
|
|
s = db->Get(read_options, column_family, key, pinnable_val);
|
|
|
|
} else {
|
|
|
|
s = reinterpret_cast<DBImpl*>(db)->GetImpl(read_options, column_family, key,
|
|
|
|
pinnable_val, nullptr, callback);
|
|
|
|
}
|
2015-05-11 23:51:51 +02:00
|
|
|
|
2015-12-10 17:54:48 +01:00
|
|
|
if (s.ok() || s.IsNotFound()) { // DB Get Succeeded
|
2015-05-11 23:51:51 +02:00
|
|
|
if (result == WriteBatchWithIndexInternal::Result::kMergeInProgress) {
|
|
|
|
// Merge result from DB with merges in Batch
|
|
|
|
auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(column_family);
|
|
|
|
const MergeOperator* merge_operator =
|
|
|
|
cfh->cfd()->ioptions()->merge_operator;
|
2016-10-18 22:19:26 +02:00
|
|
|
Statistics* statistics = immuable_db_options.statistics.get();
|
|
|
|
Env* env = immuable_db_options.env;
|
|
|
|
Logger* logger = immuable_db_options.info_log.get();
|
2015-05-11 23:51:51 +02:00
|
|
|
|
|
|
|
Slice* merge_data;
|
|
|
|
if (s.ok()) {
|
2017-08-23 19:01:17 +02:00
|
|
|
merge_data = pinnable_val;
|
2015-05-11 23:51:51 +02:00
|
|
|
} else { // Key not present in db (s.IsNotFound())
|
|
|
|
merge_data = nullptr;
|
|
|
|
}
|
2014-08-19 00:19:17 +02:00
|
|
|
|
2016-06-14 01:17:26 +02:00
|
|
|
if (merge_operator) {
|
2017-08-23 19:01:17 +02:00
|
|
|
s = MergeHelper::TimedFullMerge(
|
|
|
|
merge_operator, key, merge_data, merge_context.GetOperands(),
|
|
|
|
pinnable_val->GetSelf(), logger, statistics, env);
|
|
|
|
pinnable_val->PinSelf();
|
2016-06-14 01:17:26 +02:00
|
|
|
} else {
|
|
|
|
s = Status::InvalidArgument("Options::merge_operator must be set");
|
|
|
|
}
|
2015-05-11 23:51:51 +02:00
|
|
|
}
|
2014-08-19 00:19:17 +02:00
|
|
|
}
|
|
|
|
|
2015-05-11 23:51:51 +02:00
|
|
|
return s;
|
2014-10-10 22:31:28 +02:00
|
|
|
}
|
|
|
|
|
2015-07-11 05:15:45 +02:00
|
|
|
void WriteBatchWithIndex::SetSavePoint() { rep->write_batch.SetSavePoint(); }
|
|
|
|
|
|
|
|
Status WriteBatchWithIndex::RollbackToSavePoint() {
|
|
|
|
Status s = rep->write_batch.RollbackToSavePoint();
|
|
|
|
|
|
|
|
if (s.ok()) {
|
|
|
|
s = rep->ReBuildIndex();
|
|
|
|
}
|
|
|
|
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
2017-05-03 19:54:07 +02:00
|
|
|
Status WriteBatchWithIndex::PopSavePoint() {
|
|
|
|
return rep->write_batch.PopSavePoint();
|
|
|
|
}
|
|
|
|
|
2017-04-11 00:38:34 +02:00
|
|
|
void WriteBatchWithIndex::SetMaxBytes(size_t max_bytes) {
|
|
|
|
rep->write_batch.SetMaxBytes(max_bytes);
|
|
|
|
}
|
|
|
|
|
2014-08-19 00:19:17 +02:00
|
|
|
} // namespace rocksdb
|
2015-06-19 00:54:05 +02:00
|
|
|
#endif // !ROCKSDB_LITE
|