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).
|
2013-10-16 23:59:46 +02:00
|
|
|
//
|
2011-03-18 23:37:00 +01:00
|
|
|
// 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.
|
|
|
|
|
2013-07-23 23:42:27 +02:00
|
|
|
#include <memory>
|
2020-11-17 02:58:59 +01:00
|
|
|
|
2014-03-14 21:40:06 +01:00
|
|
|
#include "db/column_family.h"
|
2020-11-17 02:58:59 +01:00
|
|
|
#include "db/db_test_util.h"
|
2016-06-21 03:01:03 +02:00
|
|
|
#include "db/memtable.h"
|
2011-03-18 23:37:00 +01:00
|
|
|
#include "db/write_batch_internal.h"
|
2020-11-17 02:58:59 +01:00
|
|
|
#include "rocksdb/db.h"
|
2013-08-23 17:38:13 +02:00
|
|
|
#include "rocksdb/env.h"
|
|
|
|
#include "rocksdb/memtablerep.h"
|
2014-08-19 00:19:17 +02:00
|
|
|
#include "rocksdb/utilities/write_batch_with_index.h"
|
2016-06-21 03:01:03 +02:00
|
|
|
#include "rocksdb/write_buffer_manager.h"
|
2015-10-13 00:06:38 +02:00
|
|
|
#include "table/scoped_arena_iterator.h"
|
2019-05-30 20:21:38 +02:00
|
|
|
#include "test_util/testharness.h"
|
2019-05-31 02:39:43 +02:00
|
|
|
#include "util/string_util.h"
|
2011-03-18 23:37:00 +01:00
|
|
|
|
2020-02-20 21:07:53 +01:00
|
|
|
namespace ROCKSDB_NAMESPACE {
|
2011-03-18 23:37:00 +01:00
|
|
|
|
2020-11-17 02:58:59 +01:00
|
|
|
static std::string PrintContents(WriteBatch* b,
|
|
|
|
bool merge_operator_supported = true) {
|
2011-03-18 23:37:00 +01:00
|
|
|
InternalKeyComparator cmp(BytewiseComparator());
|
2013-07-23 23:42:27 +02:00
|
|
|
auto factory = std::make_shared<SkipListFactory>();
|
2014-01-15 00:32:37 +01:00
|
|
|
Options options;
|
|
|
|
options.memtable_factory = factory;
|
2020-11-17 02:58:59 +01:00
|
|
|
if (merge_operator_supported) {
|
|
|
|
options.merge_operator.reset(new TestPutOperator());
|
|
|
|
}
|
2021-05-05 22:59:21 +02:00
|
|
|
ImmutableOptions ioptions(options);
|
2016-06-21 03:01:03 +02:00
|
|
|
WriteBufferManager wb(options.db_write_buffer_size);
|
2016-09-14 06:11:59 +02:00
|
|
|
MemTable* mem = new MemTable(cmp, ioptions, MutableCFOptions(options), &wb,
|
2017-06-02 21:08:01 +02:00
|
|
|
kMaxSequenceNumber, 0 /* column_family_id */);
|
2011-05-21 04:17:43 +02:00
|
|
|
mem->Ref();
|
2011-03-18 23:37:00 +01:00
|
|
|
std::string state;
|
2014-11-18 19:20:10 +01:00
|
|
|
ColumnFamilyMemTablesDefault cf_mems_default(mem);
|
Refactor trimming logic for immutable memtables (#5022)
Summary:
MyRocks currently sets `max_write_buffer_number_to_maintain` in order to maintain enough history for transaction conflict checking. The effectiveness of this approach depends on the size of memtables. When memtables are small, it may not keep enough history; when memtables are large, this may consume too much memory.
We are proposing a new way to configure memtable list history: by limiting the memory usage of immutable memtables. The new option is `max_write_buffer_size_to_maintain` and it will take precedence over the old `max_write_buffer_number_to_maintain` if they are both set to non-zero values. The new option accounts for the total memory usage of flushed immutable memtables and mutable memtable. When the total usage exceeds the limit, RocksDB may start dropping immutable memtables (which is also called trimming history), starting from the oldest one.
The semantics of the old option actually works both as an upper bound and lower bound. History trimming will start if number of immutable memtables exceeds the limit, but it will never go below (limit-1) due to history trimming.
In order the mimic the behavior with the new option, history trimming will stop if dropping the next immutable memtable causes the total memory usage go below the size limit. For example, assuming the size limit is set to 64MB, and there are 3 immutable memtables with sizes of 20, 30, 30. Although the total memory usage is 80MB > 64MB, dropping the oldest memtable will reduce the memory usage to 60MB < 64MB, so in this case no memtable will be dropped.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5022
Differential Revision: D14394062
Pulled By: miasantreble
fbshipit-source-id: 60457a509c6af89d0993f988c9b5c2aa9e45f5c5
2019-08-23 22:54:09 +02:00
|
|
|
Status s =
|
|
|
|
WriteBatchInternal::InsertInto(b, &cf_mems_default, nullptr, nullptr);
|
2019-09-09 20:22:28 +02:00
|
|
|
uint32_t count = 0;
|
2015-11-06 16:03:30 +01:00
|
|
|
int put_count = 0;
|
|
|
|
int delete_count = 0;
|
|
|
|
int single_delete_count = 0;
|
2016-08-16 17:16:04 +02:00
|
|
|
int delete_range_count = 0;
|
2015-11-06 16:03:30 +01:00
|
|
|
int merge_count = 0;
|
2016-09-12 23:14:40 +02:00
|
|
|
for (int i = 0; i < 2; ++i) {
|
|
|
|
Arena arena;
|
2016-11-19 23:14:35 +01:00
|
|
|
ScopedArenaIterator arena_iter_guard;
|
|
|
|
std::unique_ptr<InternalIterator> iter_guard;
|
|
|
|
InternalIterator* iter;
|
|
|
|
if (i == 0) {
|
|
|
|
iter = mem->NewIterator(ReadOptions(), &arena);
|
|
|
|
arena_iter_guard.set(iter);
|
|
|
|
} else {
|
2018-11-29 00:26:56 +01:00
|
|
|
iter = mem->NewRangeTombstoneIterator(ReadOptions(),
|
|
|
|
kMaxSequenceNumber /* read_seq */);
|
2016-11-19 23:14:35 +01:00
|
|
|
iter_guard.reset(iter);
|
|
|
|
}
|
2016-11-21 21:07:09 +01:00
|
|
|
if (iter == nullptr) {
|
|
|
|
continue;
|
|
|
|
}
|
2020-12-09 00:53:59 +01:00
|
|
|
EXPECT_OK(iter->status());
|
2016-09-12 23:14:40 +02:00
|
|
|
for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
|
|
|
|
ParsedInternalKey ikey;
|
2017-07-24 20:28:20 +02:00
|
|
|
ikey.clear();
|
2020-10-28 18:11:13 +01:00
|
|
|
EXPECT_OK(ParseInternalKey(iter->key(), &ikey, true /* log_err_key */));
|
2016-09-12 23:14:40 +02:00
|
|
|
switch (ikey.type) {
|
|
|
|
case kTypeValue:
|
|
|
|
state.append("Put(");
|
|
|
|
state.append(ikey.user_key.ToString());
|
|
|
|
state.append(", ");
|
|
|
|
state.append(iter->value().ToString());
|
|
|
|
state.append(")");
|
|
|
|
count++;
|
|
|
|
put_count++;
|
|
|
|
break;
|
|
|
|
case kTypeDeletion:
|
|
|
|
state.append("Delete(");
|
|
|
|
state.append(ikey.user_key.ToString());
|
|
|
|
state.append(")");
|
|
|
|
count++;
|
|
|
|
delete_count++;
|
|
|
|
break;
|
|
|
|
case kTypeSingleDeletion:
|
|
|
|
state.append("SingleDelete(");
|
|
|
|
state.append(ikey.user_key.ToString());
|
|
|
|
state.append(")");
|
|
|
|
count++;
|
|
|
|
single_delete_count++;
|
|
|
|
break;
|
|
|
|
case kTypeRangeDeletion:
|
|
|
|
state.append("DeleteRange(");
|
|
|
|
state.append(ikey.user_key.ToString());
|
|
|
|
state.append(", ");
|
|
|
|
state.append(iter->value().ToString());
|
|
|
|
state.append(")");
|
|
|
|
count++;
|
|
|
|
delete_range_count++;
|
|
|
|
break;
|
|
|
|
case kTypeMerge:
|
|
|
|
state.append("Merge(");
|
|
|
|
state.append(ikey.user_key.ToString());
|
|
|
|
state.append(", ");
|
|
|
|
state.append(iter->value().ToString());
|
|
|
|
state.append(")");
|
|
|
|
count++;
|
|
|
|
merge_count++;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
assert(false);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
state.append("@");
|
2021-09-01 23:14:50 +02:00
|
|
|
state.append(ToString(ikey.sequence));
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
2020-12-09 00:53:59 +01:00
|
|
|
EXPECT_OK(iter->status());
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
2020-11-17 02:58:59 +01:00
|
|
|
if (s.ok()) {
|
|
|
|
EXPECT_EQ(b->HasPut(), put_count > 0);
|
|
|
|
EXPECT_EQ(b->HasDelete(), delete_count > 0);
|
|
|
|
EXPECT_EQ(b->HasSingleDelete(), single_delete_count > 0);
|
|
|
|
EXPECT_EQ(b->HasDeleteRange(), delete_range_count > 0);
|
|
|
|
EXPECT_EQ(b->HasMerge(), merge_count > 0);
|
|
|
|
if (count != WriteBatchInternal::Count(b)) {
|
|
|
|
state.append("CountMismatch()");
|
|
|
|
}
|
|
|
|
} else {
|
2013-08-15 01:32:46 +02:00
|
|
|
state.append(s.ToString());
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
2013-12-02 06:23:44 +01:00
|
|
|
delete mem->Unref();
|
2011-03-18 23:37:00 +01:00
|
|
|
return state;
|
|
|
|
}
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
class WriteBatchTest : public testing::Test {};
|
2011-03-18 23:37:00 +01:00
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(WriteBatchTest, Empty) {
|
2011-03-18 23:37:00 +01:00
|
|
|
WriteBatch batch;
|
|
|
|
ASSERT_EQ("", PrintContents(&batch));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(0u, WriteBatchInternal::Count(&batch));
|
|
|
|
ASSERT_EQ(0u, batch.Count());
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(WriteBatchTest, Multiple) {
|
2011-03-18 23:37:00 +01:00
|
|
|
WriteBatch batch;
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Put(Slice("foo"), Slice("bar")));
|
|
|
|
ASSERT_OK(batch.Delete(Slice("box")));
|
|
|
|
ASSERT_OK(batch.DeleteRange(Slice("bar"), Slice("foo")));
|
|
|
|
ASSERT_OK(batch.Put(Slice("baz"), Slice("boo")));
|
2011-03-18 23:37:00 +01:00
|
|
|
WriteBatchInternal::SetSequence(&batch, 100);
|
2012-11-06 21:02:18 +01:00
|
|
|
ASSERT_EQ(100U, WriteBatchInternal::Sequence(&batch));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(4u, WriteBatchInternal::Count(&batch));
|
2016-08-16 17:16:04 +02:00
|
|
|
ASSERT_EQ(
|
|
|
|
"Put(baz, boo)@103"
|
|
|
|
"Delete(box)@101"
|
2016-09-12 23:14:40 +02:00
|
|
|
"Put(foo, bar)@100"
|
|
|
|
"DeleteRange(bar, foo)@102",
|
2016-08-16 17:16:04 +02:00
|
|
|
PrintContents(&batch));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(4u, batch.Count());
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(WriteBatchTest, Corruption) {
|
2011-03-18 23:37:00 +01:00
|
|
|
WriteBatch batch;
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Put(Slice("foo"), Slice("bar")));
|
|
|
|
ASSERT_OK(batch.Delete(Slice("box")));
|
2011-03-18 23:37:00 +01:00
|
|
|
WriteBatchInternal::SetSequence(&batch, 200);
|
|
|
|
Slice contents = WriteBatchInternal::Contents(&batch);
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(WriteBatchInternal::SetContents(
|
|
|
|
&batch, Slice(contents.data(), contents.size() - 1)));
|
2011-03-18 23:37:00 +01:00
|
|
|
ASSERT_EQ("Put(foo, bar)@200"
|
2013-08-15 01:32:46 +02:00
|
|
|
"Corruption: bad WriteBatch Delete",
|
2011-03-18 23:37:00 +01:00
|
|
|
PrintContents(&batch));
|
|
|
|
}
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(WriteBatchTest, Append) {
|
2012-03-09 01:23:21 +01:00
|
|
|
WriteBatch b1, b2;
|
|
|
|
WriteBatchInternal::SetSequence(&b1, 200);
|
|
|
|
WriteBatchInternal::SetSequence(&b2, 300);
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(WriteBatchInternal::Append(&b1, &b2));
|
2012-03-09 01:23:21 +01:00
|
|
|
ASSERT_EQ("",
|
|
|
|
PrintContents(&b1));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(0u, b1.Count());
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(b2.Put("a", "va"));
|
|
|
|
ASSERT_OK(WriteBatchInternal::Append(&b1, &b2));
|
2012-03-09 01:23:21 +01:00
|
|
|
ASSERT_EQ("Put(a, va)@200",
|
|
|
|
PrintContents(&b1));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(1u, b1.Count());
|
2012-03-09 01:23:21 +01:00
|
|
|
b2.Clear();
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(b2.Put("b", "vb"));
|
|
|
|
ASSERT_OK(WriteBatchInternal::Append(&b1, &b2));
|
2012-03-09 01:23:21 +01:00
|
|
|
ASSERT_EQ("Put(a, va)@200"
|
|
|
|
"Put(b, vb)@201",
|
|
|
|
PrintContents(&b1));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(2u, b1.Count());
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(b2.Delete("foo"));
|
|
|
|
ASSERT_OK(WriteBatchInternal::Append(&b1, &b2));
|
2012-03-09 01:23:21 +01:00
|
|
|
ASSERT_EQ("Put(a, va)@200"
|
|
|
|
"Put(b, vb)@202"
|
|
|
|
"Put(b, vb)@201"
|
|
|
|
"Delete(foo)@203",
|
|
|
|
PrintContents(&b1));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(4u, b1.Count());
|
Add facility to write only a portion of WriteBatch to WAL
Summary:
When constructing a write batch a client may now call MarkWalTerminationPoint() on that batch. No batch operations after this call will be added written to the WAL but will still be inserted into the Memtable. This facility is used to remove one of the three WriteImpl calls in 2PC transactions. This produces a ~1% perf improvement.
```
RocksDB - unoptimized 2pc, sync_binlog=1, disable_2pc=off
INFO 2016-08-31 14:30:38,814 [main]: REQUEST PHASE COMPLETED. 75000000 requests done in 2619 seconds. Requests/second = 28628
RocksDB - optimized 2pc , sync_binlog=1, disable_2pc=off
INFO 2016-08-31 16:26:59,442 [main]: REQUEST PHASE COMPLETED. 75000000 requests done in 2581 seconds. Requests/second = 29054
```
Test Plan: Two unit tests added.
Reviewers: sdong, yiwu, IslamAbdelRahman
Reviewed By: yiwu
Subscribers: hermanlee4, dhruba, andrewkr
Differential Revision: https://reviews.facebook.net/D64599
2016-10-07 20:31:26 +02:00
|
|
|
b2.Clear();
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(b2.Put("c", "cc"));
|
|
|
|
ASSERT_OK(b2.Put("d", "dd"));
|
Add facility to write only a portion of WriteBatch to WAL
Summary:
When constructing a write batch a client may now call MarkWalTerminationPoint() on that batch. No batch operations after this call will be added written to the WAL but will still be inserted into the Memtable. This facility is used to remove one of the three WriteImpl calls in 2PC transactions. This produces a ~1% perf improvement.
```
RocksDB - unoptimized 2pc, sync_binlog=1, disable_2pc=off
INFO 2016-08-31 14:30:38,814 [main]: REQUEST PHASE COMPLETED. 75000000 requests done in 2619 seconds. Requests/second = 28628
RocksDB - optimized 2pc , sync_binlog=1, disable_2pc=off
INFO 2016-08-31 16:26:59,442 [main]: REQUEST PHASE COMPLETED. 75000000 requests done in 2581 seconds. Requests/second = 29054
```
Test Plan: Two unit tests added.
Reviewers: sdong, yiwu, IslamAbdelRahman
Reviewed By: yiwu
Subscribers: hermanlee4, dhruba, andrewkr
Differential Revision: https://reviews.facebook.net/D64599
2016-10-07 20:31:26 +02:00
|
|
|
b2.MarkWalTerminationPoint();
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(b2.Put("e", "ee"));
|
|
|
|
ASSERT_OK(WriteBatchInternal::Append(&b1, &b2, /*wal only*/ true));
|
Add facility to write only a portion of WriteBatch to WAL
Summary:
When constructing a write batch a client may now call MarkWalTerminationPoint() on that batch. No batch operations after this call will be added written to the WAL but will still be inserted into the Memtable. This facility is used to remove one of the three WriteImpl calls in 2PC transactions. This produces a ~1% perf improvement.
```
RocksDB - unoptimized 2pc, sync_binlog=1, disable_2pc=off
INFO 2016-08-31 14:30:38,814 [main]: REQUEST PHASE COMPLETED. 75000000 requests done in 2619 seconds. Requests/second = 28628
RocksDB - optimized 2pc , sync_binlog=1, disable_2pc=off
INFO 2016-08-31 16:26:59,442 [main]: REQUEST PHASE COMPLETED. 75000000 requests done in 2581 seconds. Requests/second = 29054
```
Test Plan: Two unit tests added.
Reviewers: sdong, yiwu, IslamAbdelRahman
Reviewed By: yiwu
Subscribers: hermanlee4, dhruba, andrewkr
Differential Revision: https://reviews.facebook.net/D64599
2016-10-07 20:31:26 +02:00
|
|
|
ASSERT_EQ(
|
|
|
|
"Put(a, va)@200"
|
|
|
|
"Put(b, vb)@202"
|
|
|
|
"Put(b, vb)@201"
|
|
|
|
"Put(c, cc)@204"
|
|
|
|
"Put(d, dd)@205"
|
|
|
|
"Delete(foo)@203",
|
|
|
|
PrintContents(&b1));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(6u, b1.Count());
|
Add facility to write only a portion of WriteBatch to WAL
Summary:
When constructing a write batch a client may now call MarkWalTerminationPoint() on that batch. No batch operations after this call will be added written to the WAL but will still be inserted into the Memtable. This facility is used to remove one of the three WriteImpl calls in 2PC transactions. This produces a ~1% perf improvement.
```
RocksDB - unoptimized 2pc, sync_binlog=1, disable_2pc=off
INFO 2016-08-31 14:30:38,814 [main]: REQUEST PHASE COMPLETED. 75000000 requests done in 2619 seconds. Requests/second = 28628
RocksDB - optimized 2pc , sync_binlog=1, disable_2pc=off
INFO 2016-08-31 16:26:59,442 [main]: REQUEST PHASE COMPLETED. 75000000 requests done in 2581 seconds. Requests/second = 29054
```
Test Plan: Two unit tests added.
Reviewers: sdong, yiwu, IslamAbdelRahman
Reviewed By: yiwu
Subscribers: hermanlee4, dhruba, andrewkr
Differential Revision: https://reviews.facebook.net/D64599
2016-10-07 20:31:26 +02:00
|
|
|
ASSERT_EQ(
|
|
|
|
"Put(c, cc)@0"
|
|
|
|
"Put(d, dd)@1"
|
|
|
|
"Put(e, ee)@2",
|
|
|
|
PrintContents(&b2));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(3u, b2.Count());
|
2012-03-09 01:23:21 +01:00
|
|
|
}
|
|
|
|
|
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
|
|
|
TEST_F(WriteBatchTest, SingleDeletion) {
|
|
|
|
WriteBatch batch;
|
|
|
|
WriteBatchInternal::SetSequence(&batch, 100);
|
|
|
|
ASSERT_EQ("", PrintContents(&batch));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(0u, batch.Count());
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Put("a", "va"));
|
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
|
|
|
ASSERT_EQ("Put(a, va)@100", PrintContents(&batch));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(1u, batch.Count());
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.SingleDelete("a"));
|
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
|
|
|
ASSERT_EQ(
|
|
|
|
"SingleDelete(a)@101"
|
|
|
|
"Put(a, va)@100",
|
|
|
|
PrintContents(&batch));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(2u, batch.Count());
|
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
|
|
|
}
|
|
|
|
|
2013-08-22 03:27:48 +02:00
|
|
|
namespace {
|
|
|
|
struct TestHandler : public WriteBatch::Handler {
|
|
|
|
std::string seen;
|
2019-02-19 22:36:04 +01:00
|
|
|
Status PutCF(uint32_t column_family_id, const Slice& key,
|
|
|
|
const Slice& value) override {
|
2014-01-07 23:41:42 +01:00
|
|
|
if (column_family_id == 0) {
|
|
|
|
seen += "Put(" + key.ToString() + ", " + value.ToString() + ")";
|
|
|
|
} else {
|
2014-11-25 05:44:49 +01:00
|
|
|
seen += "PutCF(" + ToString(column_family_id) + ", " +
|
2014-01-07 23:41:42 +01:00
|
|
|
key.ToString() + ", " + value.ToString() + ")";
|
|
|
|
}
|
2014-02-26 02:30:54 +01:00
|
|
|
return Status::OK();
|
2013-08-22 03:27:48 +02:00
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
Status DeleteCF(uint32_t column_family_id, const Slice& key) override {
|
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
|
|
|
if (column_family_id == 0) {
|
|
|
|
seen += "Delete(" + key.ToString() + ")";
|
|
|
|
} else {
|
|
|
|
seen += "DeleteCF(" + ToString(column_family_id) + ", " +
|
|
|
|
key.ToString() + ")";
|
|
|
|
}
|
|
|
|
return Status::OK();
|
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
Status SingleDeleteCF(uint32_t column_family_id,
|
|
|
|
const Slice& key) override {
|
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
|
|
|
if (column_family_id == 0) {
|
|
|
|
seen += "SingleDelete(" + key.ToString() + ")";
|
|
|
|
} else {
|
|
|
|
seen += "SingleDeleteCF(" + ToString(column_family_id) + ", " +
|
|
|
|
key.ToString() + ")";
|
|
|
|
}
|
|
|
|
return Status::OK();
|
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
Status DeleteRangeCF(uint32_t column_family_id, const Slice& begin_key,
|
|
|
|
const Slice& end_key) override {
|
2016-08-16 17:16:04 +02:00
|
|
|
if (column_family_id == 0) {
|
|
|
|
seen += "DeleteRange(" + begin_key.ToString() + ", " +
|
|
|
|
end_key.ToString() + ")";
|
|
|
|
} else {
|
|
|
|
seen += "DeleteRangeCF(" + ToString(column_family_id) + ", " +
|
|
|
|
begin_key.ToString() + ", " + end_key.ToString() + ")";
|
|
|
|
}
|
|
|
|
return Status::OK();
|
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
Status MergeCF(uint32_t column_family_id, const Slice& key,
|
|
|
|
const Slice& value) override {
|
2014-01-07 23:41:42 +01:00
|
|
|
if (column_family_id == 0) {
|
|
|
|
seen += "Merge(" + key.ToString() + ", " + value.ToString() + ")";
|
|
|
|
} else {
|
2014-11-25 05:44:49 +01:00
|
|
|
seen += "MergeCF(" + ToString(column_family_id) + ", " +
|
2014-01-07 23:41:42 +01:00
|
|
|
key.ToString() + ", " + value.ToString() + ")";
|
|
|
|
}
|
2014-02-26 02:30:54 +01:00
|
|
|
return Status::OK();
|
2013-08-22 03:27:48 +02:00
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
void LogData(const Slice& blob) override {
|
2013-08-22 03:27:48 +02:00
|
|
|
seen += "LogData(" + blob.ToString() + ")";
|
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
Status MarkBeginPrepare(bool unprepare) override {
|
2018-07-07 02:17:36 +02:00
|
|
|
seen +=
|
|
|
|
"MarkBeginPrepare(" + std::string(unprepare ? "true" : "false") + ")";
|
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
|
|
|
return Status::OK();
|
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
Status MarkEndPrepare(const Slice& xid) override {
|
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
|
|
|
seen += "MarkEndPrepare(" + xid.ToString() + ")";
|
|
|
|
return Status::OK();
|
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
Status MarkNoop(bool empty_batch) override {
|
2017-09-29 01:43:04 +02:00
|
|
|
seen += "MarkNoop(" + std::string(empty_batch ? "true" : "false") + ")";
|
2017-09-18 23:36:53 +02:00
|
|
|
return Status::OK();
|
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
Status MarkCommit(const Slice& xid) override {
|
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
|
|
|
seen += "MarkCommit(" + xid.ToString() + ")";
|
|
|
|
return Status::OK();
|
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
Status MarkRollback(const Slice& xid) override {
|
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
|
|
|
seen += "MarkRollback(" + xid.ToString() + ")";
|
|
|
|
return Status::OK();
|
|
|
|
}
|
2013-08-22 03:27:48 +02:00
|
|
|
};
|
|
|
|
}
|
|
|
|
|
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
|
|
|
TEST_F(WriteBatchTest, PutNotImplemented) {
|
2014-12-04 21:01:55 +01:00
|
|
|
WriteBatch batch;
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Put(Slice("k1"), Slice("v1")));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(1u, batch.Count());
|
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
|
|
|
ASSERT_EQ("Put(k1, v1)@0", PrintContents(&batch));
|
2014-12-04 21:01:55 +01:00
|
|
|
|
|
|
|
WriteBatch::Handler handler;
|
|
|
|
ASSERT_OK(batch.Iterate(&handler));
|
|
|
|
}
|
|
|
|
|
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
|
|
|
TEST_F(WriteBatchTest, DeleteNotImplemented) {
|
2014-12-04 21:01:55 +01:00
|
|
|
WriteBatch batch;
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Delete(Slice("k2")));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(1u, batch.Count());
|
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
|
|
|
ASSERT_EQ("Delete(k2)@0", PrintContents(&batch));
|
2014-12-04 21:01:55 +01:00
|
|
|
|
|
|
|
WriteBatch::Handler handler;
|
|
|
|
ASSERT_OK(batch.Iterate(&handler));
|
|
|
|
}
|
|
|
|
|
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
|
|
|
TEST_F(WriteBatchTest, SingleDeleteNotImplemented) {
|
2014-12-04 21:01:55 +01:00
|
|
|
WriteBatch batch;
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.SingleDelete(Slice("k2")));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(1u, batch.Count());
|
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
|
|
|
ASSERT_EQ("SingleDelete(k2)@0", PrintContents(&batch));
|
|
|
|
|
|
|
|
WriteBatch::Handler handler;
|
|
|
|
ASSERT_OK(batch.Iterate(&handler));
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_F(WriteBatchTest, MergeNotImplemented) {
|
|
|
|
WriteBatch batch;
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Merge(Slice("foo"), Slice("bar")));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(1u, batch.Count());
|
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
|
|
|
ASSERT_EQ("Merge(foo, bar)@0", PrintContents(&batch));
|
2014-12-04 21:01:55 +01:00
|
|
|
|
|
|
|
WriteBatch::Handler handler;
|
|
|
|
ASSERT_OK(batch.Iterate(&handler));
|
|
|
|
}
|
|
|
|
|
2020-11-17 02:58:59 +01:00
|
|
|
TEST_F(WriteBatchTest, MergeWithoutOperatorInsertionFailure) {
|
|
|
|
WriteBatch batch;
|
|
|
|
ASSERT_OK(batch.Merge(Slice("foo"), Slice("bar")));
|
|
|
|
ASSERT_EQ(1u, batch.Count());
|
|
|
|
ASSERT_EQ(
|
|
|
|
"Invalid argument: Merge requires `ColumnFamilyOptions::merge_operator "
|
|
|
|
"!= nullptr`",
|
|
|
|
PrintContents(&batch, false /* merge_operator_supported */));
|
|
|
|
}
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(WriteBatchTest, Blob) {
|
2013-08-15 01:32:46 +02:00
|
|
|
WriteBatch batch;
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Put(Slice("k1"), Slice("v1")));
|
|
|
|
ASSERT_OK(batch.Put(Slice("k2"), Slice("v2")));
|
|
|
|
ASSERT_OK(batch.Put(Slice("k3"), Slice("v3")));
|
|
|
|
ASSERT_OK(batch.PutLogData(Slice("blob1")));
|
|
|
|
ASSERT_OK(batch.Delete(Slice("k2")));
|
|
|
|
ASSERT_OK(batch.SingleDelete(Slice("k3")));
|
|
|
|
ASSERT_OK(batch.PutLogData(Slice("blob2")));
|
|
|
|
ASSERT_OK(batch.Merge(Slice("foo"), Slice("bar")));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(6u, batch.Count());
|
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
|
|
|
ASSERT_EQ(
|
|
|
|
"Merge(foo, bar)@5"
|
|
|
|
"Put(k1, v1)@0"
|
|
|
|
"Delete(k2)@3"
|
|
|
|
"Put(k2, v2)@1"
|
|
|
|
"SingleDelete(k3)@4"
|
|
|
|
"Put(k3, v3)@2",
|
|
|
|
PrintContents(&batch));
|
2013-08-15 01:32:46 +02:00
|
|
|
|
2013-08-22 03:27:48 +02:00
|
|
|
TestHandler handler;
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Iterate(&handler));
|
2013-08-22 03:27:48 +02:00
|
|
|
ASSERT_EQ(
|
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
|
|
|
"Put(k1, v1)"
|
|
|
|
"Put(k2, v2)"
|
|
|
|
"Put(k3, v3)"
|
|
|
|
"LogData(blob1)"
|
|
|
|
"Delete(k2)"
|
|
|
|
"SingleDelete(k3)"
|
|
|
|
"LogData(blob2)"
|
|
|
|
"Merge(foo, bar)",
|
|
|
|
handler.seen);
|
2013-08-22 03:27:48 +02:00
|
|
|
}
|
|
|
|
|
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
|
|
|
TEST_F(WriteBatchTest, PrepareCommit) {
|
|
|
|
WriteBatch batch;
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(WriteBatchInternal::InsertNoop(&batch));
|
|
|
|
ASSERT_OK(batch.Put(Slice("k1"), Slice("v1")));
|
|
|
|
ASSERT_OK(batch.Put(Slice("k2"), Slice("v2")));
|
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
|
|
|
batch.SetSavePoint();
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(WriteBatchInternal::MarkEndPrepare(&batch, Slice("xid1")));
|
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
|
|
|
Status s = batch.RollbackToSavePoint();
|
|
|
|
ASSERT_EQ(s, Status::NotFound());
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(WriteBatchInternal::MarkCommit(&batch, Slice("xid1")));
|
|
|
|
ASSERT_OK(WriteBatchInternal::MarkRollback(&batch, Slice("xid1")));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(2u, batch.Count());
|
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
|
|
|
|
|
|
|
TestHandler handler;
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Iterate(&handler));
|
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
|
|
|
ASSERT_EQ(
|
2018-07-07 02:17:36 +02:00
|
|
|
"MarkBeginPrepare(false)"
|
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
|
|
|
"Put(k1, v1)"
|
|
|
|
"Put(k2, v2)"
|
|
|
|
"MarkEndPrepare(xid1)"
|
|
|
|
"MarkCommit(xid1)"
|
|
|
|
"MarkRollback(xid1)",
|
|
|
|
handler.seen);
|
|
|
|
}
|
|
|
|
|
2016-02-01 20:03:28 +01:00
|
|
|
// It requires more than 30GB of memory to run the test. With single memory
|
|
|
|
// allocation of more than 30GB.
|
|
|
|
// Not all platform can run it. Also it runs a long time. So disable it.
|
|
|
|
TEST_F(WriteBatchTest, DISABLED_ManyUpdates) {
|
|
|
|
// Insert key and value of 3GB and push total batch size to 12GB.
|
2016-02-02 01:07:53 +01:00
|
|
|
static const size_t kKeyValueSize = 4u;
|
2020-09-24 00:24:57 +02:00
|
|
|
static const uint32_t kNumUpdates = uint32_t{3} << 30;
|
2016-02-01 20:03:28 +01:00
|
|
|
std::string raw(kKeyValueSize, 'A');
|
|
|
|
WriteBatch batch(kNumUpdates * (4 + kKeyValueSize * 2) + 1024u);
|
|
|
|
char c = 'A';
|
|
|
|
for (uint32_t i = 0; i < kNumUpdates; i++) {
|
|
|
|
if (c > 'Z') {
|
|
|
|
c = 'A';
|
|
|
|
}
|
|
|
|
raw[0] = c;
|
|
|
|
raw[raw.length() - 1] = c;
|
|
|
|
c++;
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Put(raw, raw));
|
2016-02-01 20:03:28 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
ASSERT_EQ(kNumUpdates, batch.Count());
|
|
|
|
|
|
|
|
struct NoopHandler : public WriteBatch::Handler {
|
|
|
|
uint32_t num_seen = 0;
|
|
|
|
char expected_char = 'A';
|
2019-02-19 22:36:04 +01:00
|
|
|
Status PutCF(uint32_t /*column_family_id*/, const Slice& key,
|
|
|
|
const Slice& value) override {
|
2016-02-01 20:03:28 +01:00
|
|
|
EXPECT_EQ(kKeyValueSize, key.size());
|
|
|
|
EXPECT_EQ(kKeyValueSize, value.size());
|
|
|
|
EXPECT_EQ(expected_char, key[0]);
|
|
|
|
EXPECT_EQ(expected_char, value[0]);
|
|
|
|
EXPECT_EQ(expected_char, key[kKeyValueSize - 1]);
|
|
|
|
EXPECT_EQ(expected_char, value[kKeyValueSize - 1]);
|
|
|
|
expected_char++;
|
|
|
|
if (expected_char > 'Z') {
|
|
|
|
expected_char = 'A';
|
|
|
|
}
|
|
|
|
++num_seen;
|
|
|
|
return Status::OK();
|
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
Status DeleteCF(uint32_t /*column_family_id*/,
|
|
|
|
const Slice& /*key*/) override {
|
2017-07-17 06:23:33 +02:00
|
|
|
ADD_FAILURE();
|
2016-02-01 20:03:28 +01:00
|
|
|
return Status::OK();
|
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
Status SingleDeleteCF(uint32_t /*column_family_id*/,
|
|
|
|
const Slice& /*key*/) override {
|
2017-07-17 06:23:33 +02:00
|
|
|
ADD_FAILURE();
|
2016-02-01 20:03:28 +01:00
|
|
|
return Status::OK();
|
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
Status MergeCF(uint32_t /*column_family_id*/, const Slice& /*key*/,
|
|
|
|
const Slice& /*value*/) override {
|
2017-07-17 06:23:33 +02:00
|
|
|
ADD_FAILURE();
|
2016-02-01 20:03:28 +01:00
|
|
|
return Status::OK();
|
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
void LogData(const Slice& /*blob*/) override { ADD_FAILURE(); }
|
|
|
|
bool Continue() override { return num_seen < kNumUpdates; }
|
2016-02-01 20:03:28 +01:00
|
|
|
} handler;
|
|
|
|
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Iterate(&handler));
|
2016-02-01 20:03:28 +01:00
|
|
|
ASSERT_EQ(kNumUpdates, handler.num_seen);
|
|
|
|
}
|
|
|
|
|
|
|
|
// The test requires more than 18GB memory to run it, with single memory
|
|
|
|
// allocation of more than 12GB. Not all the platform can run it. So disable it.
|
|
|
|
TEST_F(WriteBatchTest, DISABLED_LargeKeyValue) {
|
|
|
|
// Insert key and value of 3GB and push total batch size to 12GB.
|
2016-02-02 01:07:53 +01:00
|
|
|
static const size_t kKeyValueSize = 3221225472u;
|
2016-02-01 20:03:28 +01:00
|
|
|
std::string raw(kKeyValueSize, 'A');
|
2017-03-31 01:47:19 +02:00
|
|
|
WriteBatch batch(size_t(12884901888ull + 1024u));
|
2016-02-01 20:03:28 +01:00
|
|
|
for (char i = 0; i < 2; i++) {
|
|
|
|
raw[0] = 'A' + i;
|
|
|
|
raw[raw.length() - 1] = 'A' - i;
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Put(raw, raw));
|
2016-02-01 20:03:28 +01:00
|
|
|
}
|
|
|
|
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(2u, batch.Count());
|
2016-02-01 20:03:28 +01:00
|
|
|
|
|
|
|
struct NoopHandler : public WriteBatch::Handler {
|
|
|
|
int num_seen = 0;
|
2019-02-19 22:36:04 +01:00
|
|
|
Status PutCF(uint32_t /*column_family_id*/, const Slice& key,
|
|
|
|
const Slice& value) override {
|
2016-02-01 20:03:28 +01:00
|
|
|
EXPECT_EQ(kKeyValueSize, key.size());
|
|
|
|
EXPECT_EQ(kKeyValueSize, value.size());
|
|
|
|
EXPECT_EQ('A' + num_seen, key[0]);
|
|
|
|
EXPECT_EQ('A' + num_seen, value[0]);
|
|
|
|
EXPECT_EQ('A' - num_seen, key[kKeyValueSize - 1]);
|
|
|
|
EXPECT_EQ('A' - num_seen, value[kKeyValueSize - 1]);
|
|
|
|
++num_seen;
|
|
|
|
return Status::OK();
|
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
Status DeleteCF(uint32_t /*column_family_id*/,
|
|
|
|
const Slice& /*key*/) override {
|
2017-07-17 06:23:33 +02:00
|
|
|
ADD_FAILURE();
|
2016-02-01 20:03:28 +01:00
|
|
|
return Status::OK();
|
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
Status SingleDeleteCF(uint32_t /*column_family_id*/,
|
|
|
|
const Slice& /*key*/) override {
|
2017-07-17 06:23:33 +02:00
|
|
|
ADD_FAILURE();
|
2016-02-01 20:03:28 +01:00
|
|
|
return Status::OK();
|
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
Status MergeCF(uint32_t /*column_family_id*/, const Slice& /*key*/,
|
|
|
|
const Slice& /*value*/) override {
|
2017-07-17 06:23:33 +02:00
|
|
|
ADD_FAILURE();
|
2016-02-01 20:03:28 +01:00
|
|
|
return Status::OK();
|
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
void LogData(const Slice& /*blob*/) override { ADD_FAILURE(); }
|
|
|
|
bool Continue() override { return num_seen < 2; }
|
2016-02-01 20:03:28 +01:00
|
|
|
} handler;
|
|
|
|
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Iterate(&handler));
|
2016-02-01 20:03:28 +01:00
|
|
|
ASSERT_EQ(2, handler.num_seen);
|
|
|
|
}
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(WriteBatchTest, Continue) {
|
2013-08-22 03:27:48 +02:00
|
|
|
WriteBatch batch;
|
|
|
|
|
|
|
|
struct Handler : public TestHandler {
|
|
|
|
int num_seen = 0;
|
2019-02-19 22:36:04 +01:00
|
|
|
Status PutCF(uint32_t column_family_id, const Slice& key,
|
|
|
|
const Slice& value) override {
|
2013-08-22 03:27:48 +02:00
|
|
|
++num_seen;
|
2014-02-26 02:30:54 +01:00
|
|
|
return TestHandler::PutCF(column_family_id, key, value);
|
2013-08-15 01:32:46 +02:00
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
Status DeleteCF(uint32_t column_family_id, const Slice& key) override {
|
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
|
|
|
++num_seen;
|
|
|
|
return TestHandler::DeleteCF(column_family_id, key);
|
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
Status SingleDeleteCF(uint32_t column_family_id,
|
|
|
|
const Slice& key) override {
|
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
|
|
|
++num_seen;
|
|
|
|
return TestHandler::SingleDeleteCF(column_family_id, key);
|
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
Status MergeCF(uint32_t column_family_id, const Slice& key,
|
|
|
|
const Slice& value) override {
|
2013-08-22 03:27:48 +02:00
|
|
|
++num_seen;
|
2014-02-26 02:30:54 +01:00
|
|
|
return TestHandler::MergeCF(column_family_id, key, value);
|
2013-08-15 01:32:46 +02:00
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
void LogData(const Slice& blob) override {
|
2013-08-22 03:27:48 +02:00
|
|
|
++num_seen;
|
|
|
|
TestHandler::LogData(blob);
|
2013-08-15 01:32:46 +02:00
|
|
|
}
|
2019-02-19 22:36:04 +01:00
|
|
|
bool Continue() override { return num_seen < 5; }
|
2013-08-15 01:32:46 +02:00
|
|
|
} handler;
|
2013-08-22 03:27:48 +02:00
|
|
|
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Put(Slice("k1"), Slice("v1")));
|
|
|
|
ASSERT_OK(batch.Put(Slice("k2"), Slice("v2")));
|
|
|
|
ASSERT_OK(batch.PutLogData(Slice("blob1")));
|
|
|
|
ASSERT_OK(batch.Delete(Slice("k1")));
|
|
|
|
ASSERT_OK(batch.SingleDelete(Slice("k2")));
|
|
|
|
ASSERT_OK(batch.PutLogData(Slice("blob2")));
|
|
|
|
ASSERT_OK(batch.Merge(Slice("foo"), Slice("bar")));
|
|
|
|
ASSERT_OK(batch.Iterate(&handler));
|
2013-08-15 01:32:46 +02:00
|
|
|
ASSERT_EQ(
|
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
|
|
|
"Put(k1, v1)"
|
|
|
|
"Put(k2, v2)"
|
|
|
|
"LogData(blob1)"
|
|
|
|
"Delete(k1)"
|
|
|
|
"SingleDelete(k2)",
|
|
|
|
handler.seen);
|
2013-08-15 01:32:46 +02:00
|
|
|
}
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(WriteBatchTest, PutGatherSlices) {
|
2013-11-07 21:37:58 +01:00
|
|
|
WriteBatch batch;
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Put(Slice("foo"), Slice("bar")));
|
2013-11-07 21:37:58 +01:00
|
|
|
|
|
|
|
{
|
|
|
|
// Try a write where the key is one slice but the value is two
|
|
|
|
Slice key_slice("baz");
|
|
|
|
Slice value_slices[2] = { Slice("header"), Slice("payload") };
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(
|
|
|
|
batch.Put(SliceParts(&key_slice, 1), SliceParts(value_slices, 2)));
|
2013-11-07 21:37:58 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
// One where the key is composite but the value is a single slice
|
|
|
|
Slice key_slices[3] = { Slice("key"), Slice("part2"), Slice("part3") };
|
|
|
|
Slice value_slice("value");
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(
|
|
|
|
batch.Put(SliceParts(key_slices, 3), SliceParts(&value_slice, 1)));
|
2013-11-07 21:37:58 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
WriteBatchInternal::SetSequence(&batch, 100);
|
|
|
|
ASSERT_EQ("Put(baz, headerpayload)@101"
|
|
|
|
"Put(foo, bar)@100"
|
|
|
|
"Put(keypart2part3, value)@102",
|
|
|
|
PrintContents(&batch));
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(3u, batch.Count());
|
2013-11-07 21:37:58 +01:00
|
|
|
}
|
|
|
|
|
2014-03-14 21:40:06 +01:00
|
|
|
namespace {
|
|
|
|
class ColumnFamilyHandleImplDummy : public ColumnFamilyHandleImpl {
|
|
|
|
public:
|
2014-04-25 21:21:34 +02:00
|
|
|
explicit ColumnFamilyHandleImplDummy(int id)
|
2014-03-14 21:40:06 +01:00
|
|
|
: ColumnFamilyHandleImpl(nullptr, nullptr, nullptr), id_(id) {}
|
|
|
|
uint32_t GetID() const override { return id_; }
|
2016-07-26 00:06:11 +02:00
|
|
|
const Comparator* GetComparator() const override {
|
2014-09-22 20:37:35 +02:00
|
|
|
return BytewiseComparator();
|
|
|
|
}
|
2014-03-14 21:40:06 +01:00
|
|
|
|
|
|
|
private:
|
|
|
|
uint32_t id_;
|
|
|
|
};
|
|
|
|
} // namespace anonymous
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(WriteBatchTest, ColumnFamiliesBatchTest) {
|
2014-01-07 23:41:42 +01:00
|
|
|
WriteBatch batch;
|
2014-03-14 21:40:06 +01:00
|
|
|
ColumnFamilyHandleImplDummy zero(0), two(2), three(3), eight(8);
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Put(&zero, Slice("foo"), Slice("bar")));
|
|
|
|
ASSERT_OK(batch.Put(&two, Slice("twofoo"), Slice("bar2")));
|
|
|
|
ASSERT_OK(batch.Put(&eight, Slice("eightfoo"), Slice("bar8")));
|
|
|
|
ASSERT_OK(batch.Delete(&eight, Slice("eightfoo")));
|
|
|
|
ASSERT_OK(batch.SingleDelete(&two, Slice("twofoo")));
|
|
|
|
ASSERT_OK(batch.DeleteRange(&two, Slice("3foo"), Slice("4foo")));
|
|
|
|
ASSERT_OK(batch.Merge(&three, Slice("threethree"), Slice("3three")));
|
|
|
|
ASSERT_OK(batch.Put(&zero, Slice("foo"), Slice("bar")));
|
|
|
|
ASSERT_OK(batch.Merge(Slice("omom"), Slice("nom")));
|
2014-01-07 23:41:42 +01:00
|
|
|
|
|
|
|
TestHandler handler;
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Iterate(&handler));
|
2014-01-07 23:41:42 +01:00
|
|
|
ASSERT_EQ(
|
|
|
|
"Put(foo, bar)"
|
|
|
|
"PutCF(2, twofoo, bar2)"
|
|
|
|
"PutCF(8, eightfoo, bar8)"
|
|
|
|
"DeleteCF(8, eightfoo)"
|
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
|
|
|
"SingleDeleteCF(2, twofoo)"
|
2016-08-16 17:16:04 +02:00
|
|
|
"DeleteRangeCF(2, 3foo, 4foo)"
|
2014-01-07 23:41:42 +01:00
|
|
|
"MergeCF(3, threethree, 3three)"
|
|
|
|
"Put(foo, bar)"
|
|
|
|
"Merge(omom, nom)",
|
|
|
|
handler.seen);
|
|
|
|
}
|
|
|
|
|
2015-07-14 19:35:35 +02:00
|
|
|
#ifndef ROCKSDB_LITE
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(WriteBatchTest, ColumnFamiliesBatchWithIndexTest) {
|
2014-09-22 20:37:35 +02:00
|
|
|
WriteBatchWithIndex batch;
|
2014-08-19 00:19:17 +02:00
|
|
|
ColumnFamilyHandleImplDummy zero(0), two(2), three(3), eight(8);
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Put(&zero, Slice("foo"), Slice("bar")));
|
|
|
|
ASSERT_OK(batch.Put(&two, Slice("twofoo"), Slice("bar2")));
|
|
|
|
ASSERT_OK(batch.Put(&eight, Slice("eightfoo"), Slice("bar8")));
|
|
|
|
ASSERT_OK(batch.Delete(&eight, Slice("eightfoo")));
|
|
|
|
ASSERT_OK(batch.SingleDelete(&two, Slice("twofoo")));
|
|
|
|
ASSERT_OK(batch.Merge(&three, Slice("threethree"), Slice("3three")));
|
|
|
|
ASSERT_OK(batch.Put(&zero, Slice("foo"), Slice("bar")));
|
|
|
|
ASSERT_OK(batch.Merge(Slice("omom"), Slice("nom")));
|
2014-08-19 00:19:17 +02:00
|
|
|
|
|
|
|
std::unique_ptr<WBWIIterator> iter;
|
|
|
|
|
|
|
|
iter.reset(batch.NewIterator(&eight));
|
|
|
|
iter->Seek("eightfoo");
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
ASSERT_EQ(WriteType::kPutRecord, iter->Entry().type);
|
|
|
|
ASSERT_EQ("eightfoo", iter->Entry().key.ToString());
|
|
|
|
ASSERT_EQ("bar8", iter->Entry().value.ToString());
|
|
|
|
|
|
|
|
iter->Next();
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
ASSERT_EQ(WriteType::kDeleteRecord, iter->Entry().type);
|
|
|
|
ASSERT_EQ("eightfoo", iter->Entry().key.ToString());
|
|
|
|
|
|
|
|
iter->Next();
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(!iter->Valid());
|
|
|
|
|
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
|
|
|
iter.reset(batch.NewIterator(&two));
|
|
|
|
iter->Seek("twofoo");
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
ASSERT_EQ(WriteType::kPutRecord, iter->Entry().type);
|
|
|
|
ASSERT_EQ("twofoo", iter->Entry().key.ToString());
|
|
|
|
ASSERT_EQ("bar2", iter->Entry().value.ToString());
|
|
|
|
|
|
|
|
iter->Next();
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
ASSERT_EQ(WriteType::kSingleDeleteRecord, iter->Entry().type);
|
|
|
|
ASSERT_EQ("twofoo", iter->Entry().key.ToString());
|
|
|
|
|
|
|
|
iter->Next();
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(!iter->Valid());
|
|
|
|
|
2014-08-19 00:19:17 +02:00
|
|
|
iter.reset(batch.NewIterator());
|
|
|
|
iter->Seek("gggg");
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
ASSERT_EQ(WriteType::kMergeRecord, iter->Entry().type);
|
|
|
|
ASSERT_EQ("omom", iter->Entry().key.ToString());
|
|
|
|
ASSERT_EQ("nom", iter->Entry().value.ToString());
|
|
|
|
|
|
|
|
iter->Next();
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(!iter->Valid());
|
|
|
|
|
|
|
|
iter.reset(batch.NewIterator(&zero));
|
|
|
|
iter->Seek("foo");
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
ASSERT_EQ(WriteType::kPutRecord, iter->Entry().type);
|
|
|
|
ASSERT_EQ("foo", iter->Entry().key.ToString());
|
|
|
|
ASSERT_EQ("bar", iter->Entry().value.ToString());
|
|
|
|
|
|
|
|
iter->Next();
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
ASSERT_EQ(WriteType::kPutRecord, iter->Entry().type);
|
|
|
|
ASSERT_EQ("foo", iter->Entry().key.ToString());
|
|
|
|
ASSERT_EQ("bar", iter->Entry().value.ToString());
|
|
|
|
|
|
|
|
iter->Next();
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
ASSERT_EQ(WriteType::kMergeRecord, iter->Entry().type);
|
|
|
|
ASSERT_EQ("omom", iter->Entry().key.ToString());
|
|
|
|
ASSERT_EQ("nom", iter->Entry().value.ToString());
|
|
|
|
|
|
|
|
iter->Next();
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(!iter->Valid());
|
|
|
|
|
|
|
|
TestHandler handler;
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.GetWriteBatch()->Iterate(&handler));
|
2014-08-19 00:19:17 +02:00
|
|
|
ASSERT_EQ(
|
|
|
|
"Put(foo, bar)"
|
|
|
|
"PutCF(2, twofoo, bar2)"
|
|
|
|
"PutCF(8, eightfoo, bar8)"
|
|
|
|
"DeleteCF(8, eightfoo)"
|
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
|
|
|
"SingleDeleteCF(2, twofoo)"
|
2014-08-19 00:19:17 +02:00
|
|
|
"MergeCF(3, threethree, 3three)"
|
|
|
|
"Put(foo, bar)"
|
|
|
|
"Merge(omom, nom)",
|
|
|
|
handler.seen);
|
|
|
|
}
|
2015-07-14 19:35:35 +02:00
|
|
|
#endif // !ROCKSDB_LITE
|
2014-08-19 00:19:17 +02:00
|
|
|
|
2015-07-11 05:15:45 +02:00
|
|
|
TEST_F(WriteBatchTest, SavePointTest) {
|
|
|
|
Status s;
|
|
|
|
WriteBatch batch;
|
|
|
|
batch.SetSavePoint();
|
|
|
|
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Put("A", "a"));
|
|
|
|
ASSERT_OK(batch.Put("B", "b"));
|
2015-07-11 05:15:45 +02:00
|
|
|
batch.SetSavePoint();
|
|
|
|
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Put("C", "c"));
|
|
|
|
ASSERT_OK(batch.Delete("A"));
|
2015-07-11 05:15:45 +02:00
|
|
|
batch.SetSavePoint();
|
|
|
|
batch.SetSavePoint();
|
|
|
|
|
|
|
|
ASSERT_OK(batch.RollbackToSavePoint());
|
|
|
|
ASSERT_EQ(
|
|
|
|
"Delete(A)@3"
|
|
|
|
"Put(A, a)@0"
|
|
|
|
"Put(B, b)@1"
|
|
|
|
"Put(C, c)@2",
|
|
|
|
PrintContents(&batch));
|
|
|
|
|
|
|
|
ASSERT_OK(batch.RollbackToSavePoint());
|
|
|
|
ASSERT_OK(batch.RollbackToSavePoint());
|
|
|
|
ASSERT_EQ(
|
|
|
|
"Put(A, a)@0"
|
|
|
|
"Put(B, b)@1",
|
|
|
|
PrintContents(&batch));
|
|
|
|
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Delete("A"));
|
|
|
|
ASSERT_OK(batch.Put("B", "bb"));
|
2015-07-11 05:15:45 +02:00
|
|
|
|
|
|
|
ASSERT_OK(batch.RollbackToSavePoint());
|
|
|
|
ASSERT_EQ("", PrintContents(&batch));
|
|
|
|
|
|
|
|
s = batch.RollbackToSavePoint();
|
|
|
|
ASSERT_TRUE(s.IsNotFound());
|
|
|
|
ASSERT_EQ("", PrintContents(&batch));
|
|
|
|
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Put("D", "d"));
|
|
|
|
ASSERT_OK(batch.Delete("A"));
|
2015-07-11 05:15:45 +02:00
|
|
|
|
|
|
|
batch.SetSavePoint();
|
|
|
|
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Put("A", "aaa"));
|
2015-07-11 05:15:45 +02:00
|
|
|
|
|
|
|
ASSERT_OK(batch.RollbackToSavePoint());
|
|
|
|
ASSERT_EQ(
|
|
|
|
"Delete(A)@1"
|
|
|
|
"Put(D, d)@0",
|
|
|
|
PrintContents(&batch));
|
|
|
|
|
|
|
|
batch.SetSavePoint();
|
|
|
|
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch.Put("D", "d"));
|
|
|
|
ASSERT_OK(batch.Delete("A"));
|
2015-07-11 05:15:45 +02:00
|
|
|
|
|
|
|
ASSERT_OK(batch.RollbackToSavePoint());
|
|
|
|
ASSERT_EQ(
|
|
|
|
"Delete(A)@1"
|
|
|
|
"Put(D, d)@0",
|
|
|
|
PrintContents(&batch));
|
|
|
|
|
|
|
|
s = batch.RollbackToSavePoint();
|
|
|
|
ASSERT_TRUE(s.IsNotFound());
|
|
|
|
ASSERT_EQ(
|
|
|
|
"Delete(A)@1"
|
|
|
|
"Put(D, d)@0",
|
|
|
|
PrintContents(&batch));
|
|
|
|
|
|
|
|
WriteBatch batch2;
|
|
|
|
|
|
|
|
s = batch2.RollbackToSavePoint();
|
|
|
|
ASSERT_TRUE(s.IsNotFound());
|
|
|
|
ASSERT_EQ("", PrintContents(&batch2));
|
|
|
|
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch2.Delete("A"));
|
2015-07-11 05:15:45 +02:00
|
|
|
batch2.SetSavePoint();
|
|
|
|
|
|
|
|
s = batch2.RollbackToSavePoint();
|
|
|
|
ASSERT_OK(s);
|
|
|
|
ASSERT_EQ("Delete(A)@0", PrintContents(&batch2));
|
|
|
|
|
|
|
|
batch2.Clear();
|
|
|
|
ASSERT_EQ("", PrintContents(&batch2));
|
|
|
|
|
|
|
|
batch2.SetSavePoint();
|
|
|
|
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch2.Delete("B"));
|
2015-07-11 05:15:45 +02:00
|
|
|
ASSERT_EQ("Delete(B)@0", PrintContents(&batch2));
|
|
|
|
|
|
|
|
batch2.SetSavePoint();
|
|
|
|
s = batch2.RollbackToSavePoint();
|
|
|
|
ASSERT_OK(s);
|
|
|
|
ASSERT_EQ("Delete(B)@0", PrintContents(&batch2));
|
|
|
|
|
|
|
|
s = batch2.RollbackToSavePoint();
|
|
|
|
ASSERT_OK(s);
|
|
|
|
ASSERT_EQ("", PrintContents(&batch2));
|
|
|
|
|
|
|
|
s = batch2.RollbackToSavePoint();
|
|
|
|
ASSERT_TRUE(s.IsNotFound());
|
|
|
|
ASSERT_EQ("", PrintContents(&batch2));
|
2017-05-03 19:54:07 +02:00
|
|
|
|
|
|
|
WriteBatch batch3;
|
|
|
|
|
|
|
|
s = batch3.PopSavePoint();
|
|
|
|
ASSERT_TRUE(s.IsNotFound());
|
|
|
|
ASSERT_EQ("", PrintContents(&batch3));
|
|
|
|
|
|
|
|
batch3.SetSavePoint();
|
2020-10-20 22:17:17 +02:00
|
|
|
ASSERT_OK(batch3.Delete("A"));
|
2017-05-03 19:54:07 +02:00
|
|
|
|
|
|
|
s = batch3.PopSavePoint();
|
|
|
|
ASSERT_OK(s);
|
|
|
|
ASSERT_EQ("Delete(A)@0", PrintContents(&batch3));
|
2015-07-11 05:15:45 +02:00
|
|
|
}
|
|
|
|
|
2017-04-11 00:38:34 +02:00
|
|
|
TEST_F(WriteBatchTest, MemoryLimitTest) {
|
|
|
|
Status s;
|
|
|
|
// The header size is 12 bytes. The two Puts take 8 bytes which gives total
|
|
|
|
// of 12 + 8 * 2 = 28 bytes.
|
|
|
|
WriteBatch batch(0, 28);
|
|
|
|
|
|
|
|
ASSERT_OK(batch.Put("a", "...."));
|
|
|
|
ASSERT_OK(batch.Put("b", "...."));
|
|
|
|
s = batch.Put("c", "....");
|
|
|
|
ASSERT_TRUE(s.IsMemoryLimit());
|
|
|
|
}
|
|
|
|
|
2020-02-20 21:07:53 +01:00
|
|
|
} // namespace ROCKSDB_NAMESPACE
|
2011-03-18 23:37:00 +01:00
|
|
|
|
|
|
|
int main(int argc, char** argv) {
|
2015-03-17 22:08:00 +01:00
|
|
|
::testing::InitGoogleTest(&argc, argv);
|
|
|
|
return RUN_ALL_TESTS();
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|