rocksdb/db/multi_cf_iterator_test.cc

1174 lines
45 KiB
C++
Raw Normal View History

Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
// Copyright (c) Meta Platforms, Inc. and affiliates.
// 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).
#include "db/db_test_util.h"
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
#include "rocksdb/attribute_groups.h"
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
namespace ROCKSDB_NAMESPACE {
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
class CoalescingIteratorTest : public DBTestBase {
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
public:
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
CoalescingIteratorTest()
: DBTestBase("coalescing_iterator_test", /*env_do_fsync=*/true) {}
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
// Verify Iteration of CoalescingIterator
// by SeekToFirst() + Next() and SeekToLast() + Prev()
void VerifyCoalescingIterator(const std::vector<ColumnFamilyHandle*>& cfhs,
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
const std::vector<Slice>& expected_keys,
const std::vector<Slice>& expected_values,
const std::optional<std::vector<WideColumns>>&
expected_wide_columns = std::nullopt,
const Slice* lower_bound = nullptr,
const Slice* upper_bound = nullptr,
bool allow_unprepared_value = false) {
const size_t num_keys = expected_keys.size();
ReadOptions read_options;
read_options.iterate_lower_bound = lower_bound;
read_options.iterate_upper_bound = upper_bound;
read_options.allow_unprepared_value = allow_unprepared_value;
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
std::unique_ptr<Iterator> iter =
db_->NewCoalescingIterator(read_options, cfhs);
auto check_iter_entry = [&](size_t idx) {
ASSERT_EQ(iter->key(), expected_keys[idx]);
if (allow_unprepared_value) {
ASSERT_TRUE(iter->value().empty());
ASSERT_TRUE(iter->PrepareValue());
ASSERT_TRUE(iter->Valid());
}
ASSERT_EQ(iter->value(), expected_values[idx]);
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
if (expected_wide_columns.has_value()) {
ASSERT_EQ(iter->columns(), expected_wide_columns.value()[idx]);
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
}
};
{
size_t i = 0;
for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
check_iter_entry(i);
++i;
}
ASSERT_EQ(num_keys, i);
ASSERT_OK(iter->status());
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
}
{
size_t i = 0;
for (iter->SeekToLast(); iter->Valid(); iter->Prev()) {
check_iter_entry(num_keys - 1 - i);
++i;
}
ASSERT_EQ(num_keys, i);
ASSERT_OK(iter->status());
}
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
}
void VerifyExpectedKeys(ColumnFamilyHandle* cfh,
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
const std::vector<Slice>& expected_keys) {
std::unique_ptr<Iterator> iter(db_->NewIterator(ReadOptions(), cfh));
size_t i = 0;
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
ASSERT_EQ(iter->key(), expected_keys[i]);
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
++i;
}
ASSERT_EQ(i, expected_keys.size());
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
ASSERT_OK(iter->status());
}
};
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
TEST_F(CoalescingIteratorTest, InvalidArguments) {
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
Options options = GetDefaultOptions();
{
CreateAndReopenWithCF({"cf_1", "cf_2", "cf_3"}, options);
// Invalid - No CF is provided
std::unique_ptr<Iterator> iter_with_no_cf =
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
db_->NewCoalescingIterator(ReadOptions(), {});
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
ASSERT_NOK(iter_with_no_cf->status());
ASSERT_TRUE(iter_with_no_cf->status().IsInvalidArgument());
}
}
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
TEST_F(CoalescingIteratorTest, SimpleValues) {
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
Options options = GetDefaultOptions();
{
// Case 1: Unique key per CF
CreateAndReopenWithCF({"cf_1", "cf_2", "cf_3"}, options);
ASSERT_OK(Put(0, "key_1", "key_1_cf_0_val"));
ASSERT_OK(Put(1, "key_2", "key_2_cf_1_val"));
ASSERT_OK(Put(2, "key_3", "key_3_cf_2_val"));
ASSERT_OK(Put(3, "key_4", "key_4_cf_3_val"));
std::vector<Slice> expected_keys = {"key_1", "key_2", "key_3", "key_4"};
std::vector<Slice> expected_values = {"key_1_cf_0_val", "key_2_cf_1_val",
"key_3_cf_2_val", "key_4_cf_3_val"};
// Test for iteration over CF default->1->2->3
std::vector<ColumnFamilyHandle*> cfhs_order_0_1_2_3 = {
handles_[0], handles_[1], handles_[2], handles_[3]};
VerifyCoalescingIterator(cfhs_order_0_1_2_3, expected_keys,
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
expected_values);
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
// Test for iteration over CF 3->1->default_cf->2
std::vector<ColumnFamilyHandle*> cfhs_order_3_1_0_2 = {
handles_[3], handles_[1], handles_[0], handles_[2]};
// Iteration order and the return values should be the same since keys are
// unique per CF
VerifyCoalescingIterator(cfhs_order_3_1_0_2, expected_keys,
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
expected_values);
// Verify Seek()
{
std::unique_ptr<Iterator> iter =
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
db_->NewCoalescingIterator(ReadOptions(), cfhs_order_0_1_2_3);
iter->Seek("");
ASSERT_EQ(IterStatus(iter.get()), "key_1->key_1_cf_0_val");
iter->Seek("key_1");
ASSERT_EQ(IterStatus(iter.get()), "key_1->key_1_cf_0_val");
iter->Seek("key_2");
ASSERT_EQ(IterStatus(iter.get()), "key_2->key_2_cf_1_val");
iter->Next();
ASSERT_EQ(IterStatus(iter.get()), "key_3->key_3_cf_2_val");
iter->Seek("key_x");
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
}
// Verify SeekForPrev()
{
std::unique_ptr<Iterator> iter =
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
db_->NewCoalescingIterator(ReadOptions(), cfhs_order_0_1_2_3);
iter->SeekForPrev("");
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
iter->SeekForPrev("key_1");
ASSERT_EQ(IterStatus(iter.get()), "key_1->key_1_cf_0_val");
iter->Next();
ASSERT_EQ(IterStatus(iter.get()), "key_2->key_2_cf_1_val");
iter->SeekForPrev("key_x");
ASSERT_EQ(IterStatus(iter.get()), "key_4->key_4_cf_3_val");
iter->Prev();
ASSERT_EQ(IterStatus(iter.get()), "key_3->key_3_cf_2_val");
iter->Next();
ASSERT_EQ(IterStatus(iter.get()), "key_4->key_4_cf_3_val");
iter->Next();
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
}
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
}
{
// Case 2: Same key in multiple CFs
options = CurrentOptions(options);
DestroyAndReopen(options);
CreateAndReopenWithCF({"cf_1", "cf_2", "cf_3"}, options);
ASSERT_OK(Put(0, "key_1", "key_1_cf_0_val"));
ASSERT_OK(Put(3, "key_1", "key_1_cf_3_val"));
ASSERT_OK(Put(1, "key_2", "key_2_cf_1_val"));
ASSERT_OK(Put(2, "key_2", "key_2_cf_2_val"));
ASSERT_OK(Put(0, "key_3", "key_3_cf_0_val"));
ASSERT_OK(Put(1, "key_3", "key_3_cf_1_val"));
ASSERT_OK(Put(3, "key_3", "key_3_cf_3_val"));
std::vector<Slice> expected_keys = {"key_1", "key_2", "key_3"};
// Test for iteration over CFs default->1->2->3
std::vector<ColumnFamilyHandle*> cfhs_order_0_1_2_3 = {
handles_[0], handles_[1], handles_[2], handles_[3]};
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
std::vector<Slice> expected_values = {"key_1_cf_3_val", "key_2_cf_2_val",
"key_3_cf_3_val"};
VerifyCoalescingIterator(cfhs_order_0_1_2_3, expected_keys,
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
expected_values);
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
// Test for iteration over CFs 3->2->default_cf->1
std::vector<ColumnFamilyHandle*> cfhs_order_3_2_0_1 = {
handles_[3], handles_[2], handles_[0], handles_[1]};
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
expected_values = {"key_1_cf_0_val", "key_2_cf_1_val", "key_3_cf_1_val"};
VerifyCoalescingIterator(cfhs_order_3_2_0_1, expected_keys,
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
expected_values);
// Verify Seek()
{
std::unique_ptr<Iterator> iter =
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
db_->NewCoalescingIterator(ReadOptions(), cfhs_order_3_2_0_1);
iter->Seek("");
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
ASSERT_EQ(IterStatus(iter.get()), "key_1->key_1_cf_0_val");
iter->Seek("key_1");
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
ASSERT_EQ(IterStatus(iter.get()), "key_1->key_1_cf_0_val");
iter->Seek("key_2");
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
ASSERT_EQ(IterStatus(iter.get()), "key_2->key_2_cf_1_val");
iter->Next();
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
ASSERT_EQ(IterStatus(iter.get()), "key_3->key_3_cf_1_val");
iter->Seek("key_x");
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
}
// Verify SeekForPrev()
{
std::unique_ptr<Iterator> iter =
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
db_->NewCoalescingIterator(ReadOptions(), cfhs_order_3_2_0_1);
iter->SeekForPrev("");
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
iter->SeekForPrev("key_1");
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
ASSERT_EQ(IterStatus(iter.get()), "key_1->key_1_cf_0_val");
iter->Next();
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
ASSERT_EQ(IterStatus(iter.get()), "key_2->key_2_cf_1_val");
iter->SeekForPrev("key_x");
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
ASSERT_EQ(IterStatus(iter.get()), "key_3->key_3_cf_1_val");
iter->Next();
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
}
}
}
TEST_F(CoalescingIteratorTest, LowerAndUpperBounds) {
Options options = GetDefaultOptions();
{
// Case 1: Unique key per CF
CreateAndReopenWithCF({"cf_1", "cf_2", "cf_3"}, options);
ASSERT_OK(Put(0, "key_1", "key_1_cf_0_val"));
ASSERT_OK(Put(1, "key_2", "key_2_cf_1_val"));
ASSERT_OK(Put(2, "key_3", "key_3_cf_2_val"));
ASSERT_OK(Put(3, "key_4", "key_4_cf_3_val"));
std::vector<ColumnFamilyHandle*> cfhs_order_0_1_2_3 = {
handles_[0], handles_[1], handles_[2], handles_[3]};
// with lower_bound
{
// lower_bound is inclusive
Slice lb = Slice("key_2");
std::vector<Slice> expected_keys = {"key_2", "key_3", "key_4"};
std::vector<Slice> expected_values = {"key_2_cf_1_val", "key_3_cf_2_val",
"key_4_cf_3_val"};
VerifyCoalescingIterator(cfhs_order_0_1_2_3, expected_keys,
expected_values, std::nullopt, &lb);
}
// with upper_bound
{
// upper_bound is exclusive
Slice ub = Slice("key_3");
std::vector<Slice> expected_keys = {"key_1", "key_2"};
std::vector<Slice> expected_values = {"key_1_cf_0_val", "key_2_cf_1_val"};
VerifyCoalescingIterator(cfhs_order_0_1_2_3, expected_keys,
expected_values, std::nullopt, nullptr, &ub);
}
// with lower and upper bound
{
Slice lb = Slice("key_2");
Slice ub = Slice("key_4");
std::vector<Slice> expected_keys = {"key_2", "key_3"};
std::vector<Slice> expected_values = {"key_2_cf_1_val", "key_3_cf_2_val"};
VerifyCoalescingIterator(cfhs_order_0_1_2_3, expected_keys,
expected_values, std::nullopt, &lb, &ub);
}
{
Slice lb = Slice("key_2");
Slice ub = Slice("key_4");
ReadOptions read_options;
read_options.iterate_lower_bound = &lb;
read_options.iterate_upper_bound = &ub;
// Verify Seek() with bounds
{
std::unique_ptr<Iterator> iter =
db_->NewCoalescingIterator(read_options, cfhs_order_0_1_2_3);
iter->Seek("");
ASSERT_EQ(IterStatus(iter.get()), "key_2->key_2_cf_1_val");
iter->Next();
ASSERT_EQ(IterStatus(iter.get()), "key_3->key_3_cf_2_val");
iter->Seek("key_x");
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
}
// Verify SeekForPrev() with bounds
{
std::unique_ptr<Iterator> iter =
db_->NewCoalescingIterator(read_options, cfhs_order_0_1_2_3);
iter->SeekForPrev("");
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
iter->SeekForPrev("key_1");
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
iter->SeekForPrev("key_2");
ASSERT_EQ(IterStatus(iter.get()), "key_2->key_2_cf_1_val");
iter->Next();
ASSERT_EQ(IterStatus(iter.get()), "key_3->key_3_cf_2_val");
iter->SeekForPrev("key_x");
ASSERT_EQ(IterStatus(iter.get()), "key_3->key_3_cf_2_val");
iter->Prev();
ASSERT_EQ(IterStatus(iter.get()), "key_2->key_2_cf_1_val");
iter->Next();
ASSERT_EQ(IterStatus(iter.get()), "key_3->key_3_cf_2_val");
iter->Next();
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
}
}
}
{
// Case 2: Same key in multiple CFs
options = CurrentOptions(options);
DestroyAndReopen(options);
CreateAndReopenWithCF({"cf_1", "cf_2", "cf_3"}, options);
ASSERT_OK(Put(0, "key_1", "key_1_cf_0_val"));
ASSERT_OK(Put(3, "key_1", "key_1_cf_3_val"));
ASSERT_OK(Put(1, "key_2", "key_2_cf_1_val"));
ASSERT_OK(Put(2, "key_2", "key_2_cf_2_val"));
ASSERT_OK(Put(0, "key_3", "key_3_cf_0_val"));
ASSERT_OK(Put(1, "key_3", "key_3_cf_1_val"));
ASSERT_OK(Put(3, "key_3", "key_3_cf_3_val"));
// Test for iteration over CFs default->1->2->3
std::vector<ColumnFamilyHandle*> cfhs_order_0_1_2_3 = {
handles_[0], handles_[1], handles_[2], handles_[3]};
// with lower_bound
{
// lower_bound is inclusive
Slice lb = Slice("key_2");
std::vector<Slice> expected_keys = {"key_2", "key_3"};
std::vector<Slice> expected_values = {"key_2_cf_2_val", "key_3_cf_3_val"};
VerifyCoalescingIterator(cfhs_order_0_1_2_3, expected_keys,
expected_values, std::nullopt, &lb);
}
// with upper_bound
{
// upper_bound is exclusive
Slice ub = Slice("key_3");
std::vector<Slice> expected_keys = {"key_1", "key_2"};
std::vector<Slice> expected_values = {"key_1_cf_3_val", "key_2_cf_2_val"};
VerifyCoalescingIterator(cfhs_order_0_1_2_3, expected_keys,
expected_values, std::nullopt, nullptr, &ub);
}
// with lower and upper bound
{
Slice lb = Slice("key_2");
Slice ub = Slice("key_3");
std::vector<Slice> expected_keys = {"key_2"};
std::vector<Slice> expected_values = {"key_2_cf_2_val"};
VerifyCoalescingIterator(cfhs_order_0_1_2_3, expected_keys,
expected_values, std::nullopt, &lb, &ub);
}
// Test for iteration over CFs 3->2->default_cf->1
std::vector<ColumnFamilyHandle*> cfhs_order_3_2_0_1 = {
handles_[3], handles_[2], handles_[0], handles_[1]};
{
// lower_bound is inclusive
Slice lb = Slice("key_2");
std::vector<Slice> expected_keys = {"key_2", "key_3"};
std::vector<Slice> expected_values = {"key_2_cf_1_val", "key_3_cf_1_val"};
VerifyCoalescingIterator(cfhs_order_3_2_0_1, expected_keys,
expected_values, std::nullopt, &lb);
}
// with upper_bound
{
// upper_bound is exclusive
Slice ub = Slice("key_3");
std::vector<Slice> expected_keys = {"key_1", "key_2"};
std::vector<Slice> expected_values = {"key_1_cf_0_val", "key_2_cf_1_val"};
VerifyCoalescingIterator(cfhs_order_3_2_0_1, expected_keys,
expected_values, std::nullopt, nullptr, &ub);
}
// with lower and upper bound
{
Slice lb = Slice("key_2");
Slice ub = Slice("key_3");
std::vector<Slice> expected_keys = {"key_2"};
std::vector<Slice> expected_values = {"key_2_cf_1_val"};
VerifyCoalescingIterator(cfhs_order_3_2_0_1, expected_keys,
expected_values, std::nullopt, &lb, &ub);
}
{
Slice lb = Slice("key_2");
Slice ub = Slice("key_3");
ReadOptions read_options;
read_options.iterate_lower_bound = &lb;
read_options.iterate_upper_bound = &ub;
// Verify Seek() with bounds
{
std::unique_ptr<Iterator> iter =
db_->NewCoalescingIterator(read_options, cfhs_order_3_2_0_1);
iter->Seek("");
ASSERT_EQ(IterStatus(iter.get()), "key_2->key_2_cf_1_val");
iter->Next();
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
iter->Seek("key_x");
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
}
// Verify SeekForPrev() with bounds
{
std::unique_ptr<Iterator> iter =
db_->NewCoalescingIterator(read_options, cfhs_order_3_2_0_1);
iter->SeekForPrev("");
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
iter->SeekForPrev("key_1");
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
iter->SeekForPrev("key_2");
ASSERT_EQ(IterStatus(iter.get()), "key_2->key_2_cf_1_val");
iter->SeekForPrev("key_x");
ASSERT_EQ(IterStatus(iter.get()), "key_2->key_2_cf_1_val");
iter->Next();
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
}
}
}
}
MultiCFSnapshot for NewIterators() API (#12573) Summary: As mentioned in https://github.com/facebook/rocksdb/issues/12561 and https://github.com/facebook/rocksdb/issues/12566 , `NewIterators()` API has not been providing consistent view of the db across multiple column families. This PR addresses it by utilizing `MultiCFSnapshot()` function which has been used for `MultiGet()` APIs. To be able to obtain the thread-local super version with ref, `sv_exclusive_access` parameter has been added to `MultiCFSnapshot()` so that we could call `GetReferencedSuperVersion()` or `GetAndRefSuperVersion()` depending on the param and support `Refresh()` API for MultiCfIterators Pull Request resolved: https://github.com/facebook/rocksdb/pull/12573 Test Plan: **Unit Tests Added** ``` ./db_iterator_test --gtest_filter="*IteratorsConsistentView*" ``` ``` ./multi_cf_iterator_test -- --gtest_filter="*ConsistentView*" ``` **Performance Check** Setup ``` make -j64 release TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=10000000 -compression_type=none ``` Run ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="multireadrandom" -cache_size=10485760000 ``` Before the change ``` DB path: [/dev/shm/db_bench/dbbench] multireadrandom : 6.374 micros/op 156892 ops/sec 6.374 seconds 1000000 operations; (0 of 1000000 found) ``` After the change ``` DB path: [/dev/shm/db_bench/dbbench] multireadrandom : 6.265 micros/op 159627 ops/sec 6.265 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: jowlyzhang Differential Revision: D56444066 Pulled By: jaykorean fbshipit-source-id: 327ce73c072da30c221e18d4f3389f49115b8f99
2024-04-24 22:28:55 +00:00
TEST_F(CoalescingIteratorTest, ConsistentViewExplicitSnapshot) {
Options options = GetDefaultOptions();
options.atomic_flush = true;
MultiCFSnapshot for NewIterators() API (#12573) Summary: As mentioned in https://github.com/facebook/rocksdb/issues/12561 and https://github.com/facebook/rocksdb/issues/12566 , `NewIterators()` API has not been providing consistent view of the db across multiple column families. This PR addresses it by utilizing `MultiCFSnapshot()` function which has been used for `MultiGet()` APIs. To be able to obtain the thread-local super version with ref, `sv_exclusive_access` parameter has been added to `MultiCFSnapshot()` so that we could call `GetReferencedSuperVersion()` or `GetAndRefSuperVersion()` depending on the param and support `Refresh()` API for MultiCfIterators Pull Request resolved: https://github.com/facebook/rocksdb/pull/12573 Test Plan: **Unit Tests Added** ``` ./db_iterator_test --gtest_filter="*IteratorsConsistentView*" ``` ``` ./multi_cf_iterator_test -- --gtest_filter="*ConsistentView*" ``` **Performance Check** Setup ``` make -j64 release TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=10000000 -compression_type=none ``` Run ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="multireadrandom" -cache_size=10485760000 ``` Before the change ``` DB path: [/dev/shm/db_bench/dbbench] multireadrandom : 6.374 micros/op 156892 ops/sec 6.374 seconds 1000000 operations; (0 of 1000000 found) ``` After the change ``` DB path: [/dev/shm/db_bench/dbbench] multireadrandom : 6.265 micros/op 159627 ops/sec 6.265 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: jowlyzhang Differential Revision: D56444066 Pulled By: jaykorean fbshipit-source-id: 327ce73c072da30c221e18d4f3389f49115b8f99
2024-04-24 22:28:55 +00:00
CreateAndReopenWithCF({"cf_1", "cf_2", "cf_3"}, options);
for (int i = 0; i < 4; ++i) {
ASSERT_OK(Put(i, "cf" + std::to_string(i) + "_key",
"cf" + std::to_string(i) + "_val"));
}
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
{{"DBImpl::BGWorkFlush:done",
"DBImpl::MultiCFSnapshot::BeforeCheckingSnapshot"}});
MultiCFSnapshot for NewIterators() API (#12573) Summary: As mentioned in https://github.com/facebook/rocksdb/issues/12561 and https://github.com/facebook/rocksdb/issues/12566 , `NewIterators()` API has not been providing consistent view of the db across multiple column families. This PR addresses it by utilizing `MultiCFSnapshot()` function which has been used for `MultiGet()` APIs. To be able to obtain the thread-local super version with ref, `sv_exclusive_access` parameter has been added to `MultiCFSnapshot()` so that we could call `GetReferencedSuperVersion()` or `GetAndRefSuperVersion()` depending on the param and support `Refresh()` API for MultiCfIterators Pull Request resolved: https://github.com/facebook/rocksdb/pull/12573 Test Plan: **Unit Tests Added** ``` ./db_iterator_test --gtest_filter="*IteratorsConsistentView*" ``` ``` ./multi_cf_iterator_test -- --gtest_filter="*ConsistentView*" ``` **Performance Check** Setup ``` make -j64 release TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=10000000 -compression_type=none ``` Run ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="multireadrandom" -cache_size=10485760000 ``` Before the change ``` DB path: [/dev/shm/db_bench/dbbench] multireadrandom : 6.374 micros/op 156892 ops/sec 6.374 seconds 1000000 operations; (0 of 1000000 found) ``` After the change ``` DB path: [/dev/shm/db_bench/dbbench] multireadrandom : 6.265 micros/op 159627 ops/sec 6.265 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: jowlyzhang Differential Revision: D56444066 Pulled By: jaykorean fbshipit-source-id: 327ce73c072da30c221e18d4f3389f49115b8f99
2024-04-24 22:28:55 +00:00
bool flushed = false;
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
"DBImpl::MultiCFSnapshot::AfterRefSV", [&](void* /*arg*/) {
if (!flushed) {
for (int i = 0; i < 4; ++i) {
ASSERT_OK(Put(i, "cf" + std::to_string(i) + "_key",
"cf" + std::to_string(i) + "_val_new"));
}
ASSERT_OK(Flush());
MultiCFSnapshot for NewIterators() API (#12573) Summary: As mentioned in https://github.com/facebook/rocksdb/issues/12561 and https://github.com/facebook/rocksdb/issues/12566 , `NewIterators()` API has not been providing consistent view of the db across multiple column families. This PR addresses it by utilizing `MultiCFSnapshot()` function which has been used for `MultiGet()` APIs. To be able to obtain the thread-local super version with ref, `sv_exclusive_access` parameter has been added to `MultiCFSnapshot()` so that we could call `GetReferencedSuperVersion()` or `GetAndRefSuperVersion()` depending on the param and support `Refresh()` API for MultiCfIterators Pull Request resolved: https://github.com/facebook/rocksdb/pull/12573 Test Plan: **Unit Tests Added** ``` ./db_iterator_test --gtest_filter="*IteratorsConsistentView*" ``` ``` ./multi_cf_iterator_test -- --gtest_filter="*ConsistentView*" ``` **Performance Check** Setup ``` make -j64 release TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=10000000 -compression_type=none ``` Run ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="multireadrandom" -cache_size=10485760000 ``` Before the change ``` DB path: [/dev/shm/db_bench/dbbench] multireadrandom : 6.374 micros/op 156892 ops/sec 6.374 seconds 1000000 operations; (0 of 1000000 found) ``` After the change ``` DB path: [/dev/shm/db_bench/dbbench] multireadrandom : 6.265 micros/op 159627 ops/sec 6.265 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: jowlyzhang Differential Revision: D56444066 Pulled By: jaykorean fbshipit-source-id: 327ce73c072da30c221e18d4f3389f49115b8f99
2024-04-24 22:28:55 +00:00
flushed = true;
}
});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
std::vector<ColumnFamilyHandle*> cfhs_order_0_1_2_3 = {
handles_[0], handles_[1], handles_[2], handles_[3]};
ReadOptions read_options;
const Snapshot* snapshot = db_->GetSnapshot();
read_options.snapshot = snapshot;
// Verify Seek()
{
std::unique_ptr<Iterator> iter =
db_->NewCoalescingIterator(read_options, cfhs_order_0_1_2_3);
iter->Seek("");
ASSERT_EQ(IterStatus(iter.get()), "cf0_key->cf0_val");
iter->Next();
ASSERT_EQ(IterStatus(iter.get()), "cf1_key->cf1_val");
}
// Verify SeekForPrev()
{
std::unique_ptr<Iterator> iter =
db_->NewCoalescingIterator(read_options, cfhs_order_0_1_2_3);
iter->SeekForPrev("");
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
iter->SeekForPrev("cf2_key");
ASSERT_EQ(IterStatus(iter.get()), "cf2_key->cf2_val");
iter->Prev();
ASSERT_EQ(IterStatus(iter.get()), "cf1_key->cf1_val");
}
db_->ReleaseSnapshot(snapshot);
}
TEST_F(CoalescingIteratorTest, ConsistentViewImplicitSnapshot) {
Options options = GetDefaultOptions();
CreateAndReopenWithCF({"cf_1", "cf_2", "cf_3"}, options);
for (int i = 0; i < 4; ++i) {
ASSERT_OK(Put(i, "cf" + std::to_string(i) + "_key",
"cf" + std::to_string(i) + "_val"));
}
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
{{"DBImpl::BGWorkFlush:done",
"DBImpl::MultiCFSnapshot::BeforeCheckingSnapshot"}});
MultiCFSnapshot for NewIterators() API (#12573) Summary: As mentioned in https://github.com/facebook/rocksdb/issues/12561 and https://github.com/facebook/rocksdb/issues/12566 , `NewIterators()` API has not been providing consistent view of the db across multiple column families. This PR addresses it by utilizing `MultiCFSnapshot()` function which has been used for `MultiGet()` APIs. To be able to obtain the thread-local super version with ref, `sv_exclusive_access` parameter has been added to `MultiCFSnapshot()` so that we could call `GetReferencedSuperVersion()` or `GetAndRefSuperVersion()` depending on the param and support `Refresh()` API for MultiCfIterators Pull Request resolved: https://github.com/facebook/rocksdb/pull/12573 Test Plan: **Unit Tests Added** ``` ./db_iterator_test --gtest_filter="*IteratorsConsistentView*" ``` ``` ./multi_cf_iterator_test -- --gtest_filter="*ConsistentView*" ``` **Performance Check** Setup ``` make -j64 release TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=10000000 -compression_type=none ``` Run ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="multireadrandom" -cache_size=10485760000 ``` Before the change ``` DB path: [/dev/shm/db_bench/dbbench] multireadrandom : 6.374 micros/op 156892 ops/sec 6.374 seconds 1000000 operations; (0 of 1000000 found) ``` After the change ``` DB path: [/dev/shm/db_bench/dbbench] multireadrandom : 6.265 micros/op 159627 ops/sec 6.265 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: jowlyzhang Differential Revision: D56444066 Pulled By: jaykorean fbshipit-source-id: 327ce73c072da30c221e18d4f3389f49115b8f99
2024-04-24 22:28:55 +00:00
bool flushed = false;
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
"DBImpl::MultiCFSnapshot::AfterRefSV", [&](void* /*arg*/) {
if (!flushed) {
for (int i = 0; i < 4; ++i) {
ASSERT_OK(Put(i, "cf" + std::to_string(i) + "_key",
"cf" + std::to_string(i) + "_val_new"));
}
ASSERT_OK(Flush(1));
MultiCFSnapshot for NewIterators() API (#12573) Summary: As mentioned in https://github.com/facebook/rocksdb/issues/12561 and https://github.com/facebook/rocksdb/issues/12566 , `NewIterators()` API has not been providing consistent view of the db across multiple column families. This PR addresses it by utilizing `MultiCFSnapshot()` function which has been used for `MultiGet()` APIs. To be able to obtain the thread-local super version with ref, `sv_exclusive_access` parameter has been added to `MultiCFSnapshot()` so that we could call `GetReferencedSuperVersion()` or `GetAndRefSuperVersion()` depending on the param and support `Refresh()` API for MultiCfIterators Pull Request resolved: https://github.com/facebook/rocksdb/pull/12573 Test Plan: **Unit Tests Added** ``` ./db_iterator_test --gtest_filter="*IteratorsConsistentView*" ``` ``` ./multi_cf_iterator_test -- --gtest_filter="*ConsistentView*" ``` **Performance Check** Setup ``` make -j64 release TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=10000000 -compression_type=none ``` Run ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="multireadrandom" -cache_size=10485760000 ``` Before the change ``` DB path: [/dev/shm/db_bench/dbbench] multireadrandom : 6.374 micros/op 156892 ops/sec 6.374 seconds 1000000 operations; (0 of 1000000 found) ``` After the change ``` DB path: [/dev/shm/db_bench/dbbench] multireadrandom : 6.265 micros/op 159627 ops/sec 6.265 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: jowlyzhang Differential Revision: D56444066 Pulled By: jaykorean fbshipit-source-id: 327ce73c072da30c221e18d4f3389f49115b8f99
2024-04-24 22:28:55 +00:00
flushed = true;
}
});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
std::vector<ColumnFamilyHandle*> cfhs_order_0_1_2_3 = {
handles_[0], handles_[1], handles_[2], handles_[3]};
// Verify Seek()
{
std::unique_ptr<Iterator> iter =
db_->NewCoalescingIterator(ReadOptions(), cfhs_order_0_1_2_3);
iter->Seek("cf2_key");
ASSERT_EQ(IterStatus(iter.get()), "cf2_key->cf2_val_new");
iter->Next();
ASSERT_EQ(IterStatus(iter.get()), "cf3_key->cf3_val_new");
}
// Verify SeekForPrev()
{
std::unique_ptr<Iterator> iter =
db_->NewCoalescingIterator(ReadOptions(), cfhs_order_0_1_2_3);
iter->SeekForPrev("");
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
iter->SeekForPrev("cf1_key");
ASSERT_EQ(IterStatus(iter.get()), "cf1_key->cf1_val_new");
iter->Prev();
ASSERT_EQ(IterStatus(iter.get()), "cf0_key->cf0_val_new");
}
}
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
TEST_F(CoalescingIteratorTest, EmptyCfs) {
Options options = GetDefaultOptions();
{
// Case 1: No keys in any of the CFs
CreateAndReopenWithCF({"cf_1", "cf_2", "cf_3"}, options);
std::unique_ptr<Iterator> iter =
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
db_->NewCoalescingIterator(ReadOptions(), handles_);
iter->SeekToFirst();
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
iter->SeekToLast();
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
iter->Seek("foo");
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
iter->SeekForPrev("foo");
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
ASSERT_OK(iter->status());
}
{
// Case 2: A single key exists in only one of the CF. Rest CFs are empty.
ASSERT_OK(Put(1, "key_1", "key_1_cf_1_val"));
std::unique_ptr<Iterator> iter =
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
db_->NewCoalescingIterator(ReadOptions(), handles_);
iter->SeekToFirst();
ASSERT_EQ(IterStatus(iter.get()), "key_1->key_1_cf_1_val");
iter->Next();
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
iter->SeekToLast();
ASSERT_EQ(IterStatus(iter.get()), "key_1->key_1_cf_1_val");
iter->Prev();
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
}
{
// Case 3: same key exists in all of the CFs except one (cf_2)
ASSERT_OK(Put(0, "key_1", "key_1_cf_0_val"));
ASSERT_OK(Put(3, "key_1", "key_1_cf_3_val"));
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
// handles_ are in the order of 0->1->2->3
std::unique_ptr<Iterator> iter =
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
db_->NewCoalescingIterator(ReadOptions(), handles_);
iter->SeekToFirst();
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
ASSERT_EQ(IterStatus(iter.get()), "key_1->key_1_cf_3_val");
iter->Next();
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
iter->SeekToLast();
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
ASSERT_EQ(IterStatus(iter.get()), "key_1->key_1_cf_3_val");
iter->Prev();
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
}
}
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
TEST_F(CoalescingIteratorTest, WideColumns) {
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
// Set up the DB and Column Families
Options options = GetDefaultOptions();
CreateAndReopenWithCF({"cf_1", "cf_2", "cf_3"}, options);
constexpr char key_1[] = "key_1";
WideColumns key_1_columns_in_cf_2{
{kDefaultWideColumnName, "cf_2_col_val_0_key_1"},
{"cf_2_col_name_1", "cf_2_col_val_1_key_1"},
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
{"cf_2_col_name_2", "cf_2_col_val_2_key_1"},
{"cf_overlap_col_name", "cf_2_overlap_value_key_1"}};
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
WideColumns key_1_columns_in_cf_3{
{"cf_3_col_name_1", "cf_3_col_val_1_key_1"},
{"cf_3_col_name_2", "cf_3_col_val_2_key_1"},
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
{"cf_3_col_name_3", "cf_3_col_val_3_key_1"},
{"cf_overlap_col_name", "cf_3_overlap_value_key_1"}};
WideColumns key_1_expected_columns_cfh_order_2_3{
{kDefaultWideColumnName, "cf_2_col_val_0_key_1"},
{"cf_2_col_name_1", "cf_2_col_val_1_key_1"},
{"cf_2_col_name_2", "cf_2_col_val_2_key_1"},
{"cf_3_col_name_1", "cf_3_col_val_1_key_1"},
{"cf_3_col_name_2", "cf_3_col_val_2_key_1"},
{"cf_3_col_name_3", "cf_3_col_val_3_key_1"},
{"cf_overlap_col_name", "cf_3_overlap_value_key_1"}};
WideColumns key_1_expected_columns_cfh_order_3_2{
{kDefaultWideColumnName, "cf_2_col_val_0_key_1"},
{"cf_2_col_name_1", "cf_2_col_val_1_key_1"},
{"cf_2_col_name_2", "cf_2_col_val_2_key_1"},
{"cf_3_col_name_1", "cf_3_col_val_1_key_1"},
{"cf_3_col_name_2", "cf_3_col_val_2_key_1"},
{"cf_3_col_name_3", "cf_3_col_val_3_key_1"},
{"cf_overlap_col_name", "cf_2_overlap_value_key_1"}};
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
constexpr char key_2[] = "key_2";
WideColumns key_2_columns_in_cf_1{
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
{"cf_overlap_col_name", "cf_1_overlap_value_key_2"}};
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
WideColumns key_2_columns_in_cf_2{
{"cf_2_col_name_1", "cf_2_col_val_1_key_2"},
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
{"cf_2_col_name_2", "cf_2_col_val_2_key_2"},
{"cf_overlap_col_name", "cf_2_overlap_value_key_2"}};
WideColumns key_2_expected_columns_cfh_order_1_2{
{"cf_2_col_name_1", "cf_2_col_val_1_key_2"},
{"cf_2_col_name_2", "cf_2_col_val_2_key_2"},
{"cf_overlap_col_name", "cf_2_overlap_value_key_2"}};
WideColumns key_2_expected_columns_cfh_order_2_1{
{"cf_2_col_name_1", "cf_2_col_val_1_key_2"},
{"cf_2_col_name_2", "cf_2_col_val_2_key_2"},
{"cf_overlap_col_name", "cf_1_overlap_value_key_2"}};
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
constexpr char key_3[] = "key_3";
WideColumns key_3_columns_in_cf_1{
{"cf_1_col_name_1", "cf_1_col_val_1_key_3"}};
WideColumns key_3_columns_in_cf_3{
{"cf_3_col_name_1", "cf_3_col_val_1_key_3"}};
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
WideColumns key_3_expected_columns{
{"cf_1_col_name_1", "cf_1_col_val_1_key_3"},
{"cf_3_col_name_1", "cf_3_col_val_1_key_3"},
};
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
constexpr char key_4[] = "key_4";
WideColumns key_4_columns_in_cf_0{
{"cf_0_col_name_1", "cf_0_col_val_1_key_4"}};
WideColumns key_4_columns_in_cf_2{
{"cf_2_col_name_1", "cf_2_col_val_1_key_4"}};
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
WideColumns key_4_expected_columns{
{"cf_0_col_name_1", "cf_0_col_val_1_key_4"},
{"cf_2_col_name_1", "cf_2_col_val_1_key_4"},
};
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
// Use AttributeGroup PutEntity API to insert them together
AttributeGroups key_1_attribute_groups{
AttributeGroup(handles_[2], key_1_columns_in_cf_2),
AttributeGroup(handles_[3], key_1_columns_in_cf_3)};
AttributeGroups key_2_attribute_groups{
AttributeGroup(handles_[1], key_2_columns_in_cf_1),
AttributeGroup(handles_[2], key_2_columns_in_cf_2)};
AttributeGroups key_3_attribute_groups{
AttributeGroup(handles_[1], key_3_columns_in_cf_1),
AttributeGroup(handles_[3], key_3_columns_in_cf_3)};
AttributeGroups key_4_attribute_groups{
AttributeGroup(handles_[0], key_4_columns_in_cf_0),
AttributeGroup(handles_[2], key_4_columns_in_cf_2)};
ASSERT_OK(db_->PutEntity(WriteOptions(), key_1, key_1_attribute_groups));
ASSERT_OK(db_->PutEntity(WriteOptions(), key_2, key_2_attribute_groups));
ASSERT_OK(db_->PutEntity(WriteOptions(), key_3, key_3_attribute_groups));
ASSERT_OK(db_->PutEntity(WriteOptions(), key_4, key_4_attribute_groups));
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
// Keys should be returned in order regardless of cfh order
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
std::vector<Slice> expected_keys = {key_1, key_2, key_3, key_4};
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
// Since value for kDefaultWideColumnName only exists for key_1, rest will
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
// return empty value after coalesced
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
std::vector<Slice> expected_values = {"cf_2_col_val_0_key_1", "", "", ""};
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
// Test for iteration over CF default->1->2->3
{
std::vector<ColumnFamilyHandle*> cfhs_order_0_1_2_3 = {
handles_[0], handles_[1], handles_[2], handles_[3]};
// Coalesced columns
std::vector<WideColumns> expected_wide_columns_0_1_2_3 = {
key_1_expected_columns_cfh_order_2_3,
key_2_expected_columns_cfh_order_1_2, key_3_expected_columns,
key_4_expected_columns};
VerifyCoalescingIterator(cfhs_order_0_1_2_3, expected_keys, expected_values,
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
expected_wide_columns_0_1_2_3);
}
// Test for iteration over CF 3->2->default->1
{
std::vector<ColumnFamilyHandle*> cfhs_order_3_2_0_1 = {
handles_[3], handles_[2], handles_[0], handles_[1]};
// Coalesced columns
std::vector<WideColumns> expected_wide_columns_3_2_0_1 = {
key_1_expected_columns_cfh_order_3_2,
key_2_expected_columns_cfh_order_2_1, key_3_expected_columns,
key_4_expected_columns};
VerifyCoalescingIterator(cfhs_order_3_2_0_1, expected_keys, expected_values,
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
expected_wide_columns_3_2_0_1);
}
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
}
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
TEST_F(CoalescingIteratorTest, DifferentComparatorsInMultiCFs) {
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
// This test creates two column families with two different comparators.
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
// Attempting to create the CoalescingIterator should fail.
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
Options options = GetDefaultOptions();
options.create_if_missing = true;
DestroyAndReopen(options);
options.comparator = BytewiseComparator();
CreateColumnFamilies({"cf_forward"}, options);
options.comparator = ReverseBytewiseComparator();
CreateColumnFamilies({"cf_reverse"}, options);
ASSERT_OK(Put(0, "key_1", "value_1"));
ASSERT_OK(Put(0, "key_2", "value_2"));
ASSERT_OK(Put(0, "key_3", "value_3"));
ASSERT_OK(Put(1, "key_1", "value_1"));
ASSERT_OK(Put(1, "key_2", "value_2"));
ASSERT_OK(Put(1, "key_3", "value_3"));
VerifyExpectedKeys(handles_[0], {"key_1", "key_2", "key_3"});
VerifyExpectedKeys(handles_[1], {"key_3", "key_2", "key_1"});
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
std::unique_ptr<Iterator> iter =
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
db_->NewCoalescingIterator(ReadOptions(), handles_);
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
ASSERT_NOK(iter->status());
ASSERT_TRUE(iter->status().IsInvalidArgument());
}
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
TEST_F(CoalescingIteratorTest, CustomComparatorsInMultiCFs) {
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
// This test creates two column families with the same custom test
// comparators (but instantiated independently). Attempting to create the
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
// CoalescingIterator should not fail.
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
Options options = GetDefaultOptions();
options.create_if_missing = true;
DestroyAndReopen(options);
static auto comparator_1 =
std::make_unique<test::SimpleSuffixReverseComparator>(
test::SimpleSuffixReverseComparator());
static auto comparator_2 =
std::make_unique<test::SimpleSuffixReverseComparator>(
test::SimpleSuffixReverseComparator());
ASSERT_NE(comparator_1, comparator_2);
options.comparator = comparator_1.get();
CreateColumnFamilies({"cf_1"}, options);
options.comparator = comparator_2.get();
CreateColumnFamilies({"cf_2"}, options);
ASSERT_OK(Put(0, "key_001_001", "value_0_3"));
ASSERT_OK(Put(0, "key_001_002", "value_0_2"));
ASSERT_OK(Put(0, "key_001_003", "value_0_1"));
ASSERT_OK(Put(0, "key_002_001", "value_0_6"));
ASSERT_OK(Put(0, "key_002_002", "value_0_5"));
ASSERT_OK(Put(0, "key_002_003", "value_0_4"));
ASSERT_OK(Put(1, "key_001_001", "value_1_3"));
ASSERT_OK(Put(1, "key_001_002", "value_1_2"));
ASSERT_OK(Put(1, "key_001_003", "value_1_1"));
ASSERT_OK(Put(1, "key_003_004", "value_1_6"));
ASSERT_OK(Put(1, "key_003_005", "value_1_5"));
ASSERT_OK(Put(1, "key_003_006", "value_1_4"));
VerifyExpectedKeys(
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
handles_[0], {"key_001_003", "key_001_002", "key_001_001", "key_002_003",
"key_002_002", "key_002_001"});
VerifyExpectedKeys(
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
handles_[1], {"key_001_003", "key_001_002", "key_001_001", "key_003_006",
"key_003_005", "key_003_004"});
std::vector<Slice> expected_keys = {
"key_001_003", "key_001_002", "key_001_001", "key_002_003", "key_002_002",
"key_002_001", "key_003_006", "key_003_005", "key_003_004"};
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
std::vector<Slice> expected_values = {"value_1_1", "value_1_2", "value_1_3",
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
"value_0_4", "value_0_5", "value_0_6",
"value_1_4", "value_1_5", "value_1_6"};
std::unique_ptr<Iterator> iter =
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
db_->NewCoalescingIterator(ReadOptions(), handles_);
size_t i = 0;
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
ASSERT_EQ(expected_keys[i], iter->key());
ASSERT_EQ(expected_values[i], iter->value());
++i;
}
ASSERT_EQ(expected_keys.size(), i);
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
ASSERT_OK(iter->status());
}
TEST_F(CoalescingIteratorTest, AllowUnpreparedValue) {
Options options = GetDefaultOptions();
options.enable_blob_files = true;
CreateAndReopenWithCF({"cf_1", "cf_2", "cf_3"}, options);
ASSERT_OK(Put(0, "key_1", "key_1_cf_0_val"));
ASSERT_OK(Put(3, "key_1", "key_1_cf_3_val"));
ASSERT_OK(Put(1, "key_2", "key_2_cf_1_val"));
ASSERT_OK(Put(2, "key_2", "key_2_cf_2_val"));
ASSERT_OK(Put(0, "key_3", "key_3_cf_0_val"));
ASSERT_OK(Put(1, "key_3", "key_3_cf_1_val"));
ASSERT_OK(Put(3, "key_3", "key_3_cf_3_val"));
ASSERT_OK(Flush());
std::vector<ColumnFamilyHandle*> cfhs_order_3_2_0_1{handles_[3], handles_[2],
handles_[0], handles_[1]};
std::vector<Slice> expected_keys{"key_1", "key_2", "key_3"};
std::vector<Slice> expected_values{"key_1_cf_0_val", "key_2_cf_1_val",
"key_3_cf_1_val"};
VerifyCoalescingIterator(cfhs_order_3_2_0_1, expected_keys, expected_values,
/* expected_wide_columns */ std::nullopt,
/* lower_bound */ nullptr, /* upper_bound */ nullptr,
/* allow_unprepared_value */ true);
ReadOptions read_options;
read_options.allow_unprepared_value = true;
{
std::unique_ptr<Iterator> iter =
db_->NewCoalescingIterator(read_options, cfhs_order_3_2_0_1);
iter->Seek("");
ASSERT_EQ(IterStatus(iter.get()), "key_1->");
ASSERT_TRUE(iter->PrepareValue());
ASSERT_EQ(IterStatus(iter.get()), "key_1->key_1_cf_0_val");
iter->Seek("key_1");
ASSERT_EQ(IterStatus(iter.get()), "key_1->");
ASSERT_TRUE(iter->PrepareValue());
ASSERT_EQ(IterStatus(iter.get()), "key_1->key_1_cf_0_val");
iter->Seek("key_2");
ASSERT_EQ(IterStatus(iter.get()), "key_2->");
ASSERT_TRUE(iter->PrepareValue());
ASSERT_EQ(IterStatus(iter.get()), "key_2->key_2_cf_1_val");
iter->Next();
ASSERT_EQ(IterStatus(iter.get()), "key_3->");
ASSERT_TRUE(iter->PrepareValue());
ASSERT_EQ(IterStatus(iter.get()), "key_3->key_3_cf_1_val");
iter->Seek("key_x");
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
}
{
std::unique_ptr<Iterator> iter =
db_->NewCoalescingIterator(read_options, cfhs_order_3_2_0_1);
iter->SeekForPrev("");
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
iter->SeekForPrev("key_1");
ASSERT_EQ(IterStatus(iter.get()), "key_1->");
ASSERT_TRUE(iter->PrepareValue());
ASSERT_EQ(IterStatus(iter.get()), "key_1->key_1_cf_0_val");
iter->Next();
ASSERT_EQ(IterStatus(iter.get()), "key_2->");
ASSERT_TRUE(iter->PrepareValue());
ASSERT_EQ(IterStatus(iter.get()), "key_2->key_2_cf_1_val");
iter->SeekForPrev("key_x");
ASSERT_EQ(IterStatus(iter.get()), "key_3->");
ASSERT_TRUE(iter->PrepareValue());
ASSERT_EQ(IterStatus(iter.get()), "key_3->key_3_cf_1_val");
iter->Next();
ASSERT_EQ(IterStatus(iter.get()), "(invalid)");
}
}
TEST_F(CoalescingIteratorTest, AllowUnpreparedValue_Corruption) {
Options options = GetDefaultOptions();
options.enable_blob_files = true;
CreateAndReopenWithCF({"cf_1", "cf_2", "cf_3"}, options);
ASSERT_OK(Put(0, "key_1", "key_1_cf_0_val"));
ASSERT_OK(Put(3, "key_1", "key_1_cf_3_val"));
ASSERT_OK(Put(1, "key_2", "key_2_cf_1_val"));
ASSERT_OK(Put(2, "key_2", "key_2_cf_2_val"));
ASSERT_OK(Put(0, "key_3", "key_3_cf_0_val"));
ASSERT_OK(Put(1, "key_3", "key_3_cf_1_val"));
ASSERT_OK(Put(3, "key_3", "key_3_cf_3_val"));
ASSERT_OK(Flush());
ReadOptions read_options;
read_options.allow_unprepared_value = true;
std::vector<ColumnFamilyHandle*> cfhs_order_3_2_0_1{handles_[3], handles_[2],
handles_[0], handles_[1]};
std::unique_ptr<Iterator> iter =
db_->NewCoalescingIterator(read_options, cfhs_order_3_2_0_1);
iter->SeekToFirst();
ASSERT_TRUE(iter->Valid());
ASSERT_OK(iter->status());
ASSERT_EQ(iter->key(), "key_1");
ASSERT_TRUE(iter->value().empty());
SyncPoint::GetInstance()->SetCallBack(
"BlobFileReader::GetBlob:TamperWithResult", [](void* arg) {
Slice* const blob_index = static_cast<Slice*>(arg);
assert(blob_index);
assert(!blob_index->empty());
blob_index->remove_prefix(1);
});
SyncPoint::GetInstance()->EnableProcessing();
ASSERT_FALSE(iter->PrepareValue());
ASSERT_FALSE(iter->Valid());
ASSERT_TRUE(iter->status().IsCorruption());
SyncPoint::GetInstance()->DisableProcessing();
SyncPoint::GetInstance()->ClearAllCallBacks();
}
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
class AttributeGroupIteratorTest : public DBTestBase {
public:
AttributeGroupIteratorTest()
: DBTestBase("attribute_group_iterator_test", /*env_do_fsync=*/true) {}
void VerifyAttributeGroupIterator(
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
const std::vector<ColumnFamilyHandle*>& cfhs,
const std::vector<Slice>& expected_keys,
const std::vector<IteratorAttributeGroups>& expected_attribute_groups,
const Slice* lower_bound = nullptr, const Slice* upper_bound = nullptr,
bool allow_unprepared_value = false) {
const size_t num_keys = expected_keys.size();
ReadOptions read_options;
read_options.iterate_lower_bound = lower_bound;
read_options.iterate_upper_bound = upper_bound;
read_options.allow_unprepared_value = allow_unprepared_value;
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
std::unique_ptr<AttributeGroupIterator> iter =
db_->NewAttributeGroupIterator(read_options, cfhs);
auto check_iter_entry = [&](size_t idx) {
ASSERT_EQ(iter->key(), expected_keys[idx]);
if (allow_unprepared_value) {
ASSERT_TRUE(iter->attribute_groups().empty());
ASSERT_TRUE(iter->PrepareValue());
ASSERT_TRUE(iter->Valid());
}
ASSERT_EQ(iter->attribute_groups(), expected_attribute_groups[idx]);
};
{
size_t i = 0;
for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
check_iter_entry(i);
++i;
MultiCfIterator - AttributeGroupIter Impl & CoalescingIter Optimization (#12534) Summary: Continuing from the previous MultiCfIterator Implementations - (https://github.com/facebook/rocksdb/issues/12422, https://github.com/facebook/rocksdb/issues/12480 #12465), this PR completes the `AttributeGroupIterator` by implementing `AttributeGroupIteratorImpl::AddToAttributeGroups()`. While implementing the `AttributeGroupIterator`, we had to make some changes in `MultiCfIteratorImpl` and found an opportunity to improve `Coalesce()` in `CoalescingIterator`. Lifting `UNDER CONSTRUCTION - DO NOT USE` comment by replacing it with `EXPERIMENTAL` Here are some implementation details: - `IteratorAttributeGroups` is introduced to avoid having to copy all `WideColumn` objects during iteration. - `PopulateIterator()` no longer advances non-top iterators that have the same key as the top iterator in the heap. - `AdvanceIterator()` needs to advance the non-top iterators when they have the same key as the top iterator in the heap. - Instead of populating one by one, `PopulateIterator()` now collects all items with the same key and calls `populate_func(items)` at once. - This allowed optimization in `Coalesce()` such that we no longer do K-1 rounds of 2-way merge, but do one K-way merge instead. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12534 Test Plan: Uncommented the assertions in `verifyAttributeGroupIterator()` ``` ./multi_cf_iterator_test ``` Reviewed By: ltamasi Differential Revision: D56089019 Pulled By: jaykorean fbshipit-source-id: 6b0b4247e221f69b40b147d41492008cc9b15054
2024-04-16 15:45:38 +00:00
}
ASSERT_EQ(i, num_keys);
ASSERT_OK(iter->status());
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
}
{
size_t i = 0;
for (iter->SeekToLast(); iter->Valid(); iter->Prev()) {
check_iter_entry(num_keys - 1 - i);
++i;
MultiCfIterator - AttributeGroupIter Impl & CoalescingIter Optimization (#12534) Summary: Continuing from the previous MultiCfIterator Implementations - (https://github.com/facebook/rocksdb/issues/12422, https://github.com/facebook/rocksdb/issues/12480 #12465), this PR completes the `AttributeGroupIterator` by implementing `AttributeGroupIteratorImpl::AddToAttributeGroups()`. While implementing the `AttributeGroupIterator`, we had to make some changes in `MultiCfIteratorImpl` and found an opportunity to improve `Coalesce()` in `CoalescingIterator`. Lifting `UNDER CONSTRUCTION - DO NOT USE` comment by replacing it with `EXPERIMENTAL` Here are some implementation details: - `IteratorAttributeGroups` is introduced to avoid having to copy all `WideColumn` objects during iteration. - `PopulateIterator()` no longer advances non-top iterators that have the same key as the top iterator in the heap. - `AdvanceIterator()` needs to advance the non-top iterators when they have the same key as the top iterator in the heap. - Instead of populating one by one, `PopulateIterator()` now collects all items with the same key and calls `populate_func(items)` at once. - This allowed optimization in `Coalesce()` such that we no longer do K-1 rounds of 2-way merge, but do one K-way merge instead. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12534 Test Plan: Uncommented the assertions in `verifyAttributeGroupIterator()` ``` ./multi_cf_iterator_test ``` Reviewed By: ltamasi Differential Revision: D56089019 Pulled By: jaykorean fbshipit-source-id: 6b0b4247e221f69b40b147d41492008cc9b15054
2024-04-16 15:45:38 +00:00
}
ASSERT_EQ(i, num_keys);
ASSERT_OK(iter->status());
MultiCFIterator Refactor - CoalescingIterator & AttributeGroupIterator (#12480) Summary: There are a couple of reasons to modify the current implementation of the MultiCfIterator, which implements the generic `Iterator` interface. - The default behavior of `value()`/`columns()` returning data from different Column Families for different keys can be prone to errors, even though there might be valid use cases where users do not care about the origin of the value/columns. - The `attribute_groups()` API, which is not yet implemented, will not be useful for a single-CF iterator. In this PR, we are implementing the following changes: - `IteratorBase` introduced, which includes all basic iterator functions except `value()` and `columns()`. - `Iterator`, which now inherits from `IteratorBase`, includes `value()` and `columns()`. - New public interface `AttributeGroupIterator` inherits from `IteratorBase` and additionally includes `attribute_groups()` (to be implemented). - Renamed former `MultiCfIterator` to `CoalescingIterator` which inherits from `Iterator` - Existing MultiCfIteratorTest has been split into two - `CoalescingIteratorTest` and `AttributeGroupIteratorTest`. - Moved AttributeGroup related code from `wide_columns.h` to a new file, `attribute_groups.h`. Some Implementation Details - `MultiCfIteratorImpl` takes two functions - `populate_func` and `reset_func` and use them to populate `value_` and `columns_` in CoalescingIterator and `attribute_groups_` in AttributeGroupIterator. In CoalescingIterator, populate_func is `Coalesce()`, in AttributeGroupIterator populate_func is `AddToAttributeGroups()`. `reset_func` clears populated value_, columns_ and attribute_groups_ accordingly. - `Coalesce()` merge sorts columns from multiple CFs when a key exists in more than on CFs. column that appears in later CF overwrites the prior ones. For example, if CF1 has `"key_1" ==> {"col_1": "foo", "col_2", "baz"}` and CF2 has `"key_1" ==> {"col_2": "quux", "col_3", "bla"}`, and when the iterator is at `key_1`, `columns()` will return `{"col_1": "foo", "col_2", "quux", "col_3", "bla"}` In this example, `value()` will be empty, because none of them have values for `kDefaultColumnName` Pull Request resolved: https://github.com/facebook/rocksdb/pull/12480 Test Plan: ## Unit Test ``` ./multi_cf_iterator_test ``` ## Performance Test To make sure this change does not impact existing `Iterator` performance **Build** ``` $> make -j64 release ``` **Setup** ``` $> TEST_TMPDIR=/dev/shm/db_bench ./db_bench -benchmarks="filluniquerandom" -key_size=32 -value_size=512 -num=1000000 -compression_type=none ``` **Run** ``` TEST_TMPDIR=/dev/shm/db_bench ./db_bench -use_existing_db=1 -benchmarks="newiterator,seekrandom" -cache_size=10485760000 ``` **Before the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.519 micros/op 1927904 ops/sec 0.519 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.302 micros/op 188589 ops/sec 5.303 seconds 1000000 operations; (0 of 1000000 found) ``` **After the change** ``` DB path: [/dev/shm/db_bench/dbbench] newiterator : 0.497 micros/op 2011012 ops/sec 0.497 seconds 1000000 operations; DB path: [/dev/shm/db_bench/dbbench] seekrandom : 5.252 micros/op 190405 ops/sec 5.252 seconds 1000000 operations; (0 of 1000000 found) ``` Reviewed By: ltamasi Differential Revision: D55353909 Pulled By: jaykorean fbshipit-source-id: 8d7786ffee09e022261ce34aa60e8633685e1946
2024-04-11 18:34:04 +00:00
}
}
};
TEST_F(AttributeGroupIteratorTest, IterateAttributeGroups) {
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
// Set up the DB and Column Families
Options options = GetDefaultOptions();
CreateAndReopenWithCF({"cf_1", "cf_2", "cf_3"}, options);
constexpr char key_1[] = "key_1";
WideColumns key_1_columns_in_cf_2{
{kDefaultWideColumnName, "cf_2_col_val_0_key_1"},
{"cf_2_col_name_1", "cf_2_col_val_1_key_1"},
{"cf_2_col_name_2", "cf_2_col_val_2_key_1"}};
WideColumns key_1_columns_in_cf_3{
{"cf_3_col_name_1", "cf_3_col_val_1_key_1"},
{"cf_3_col_name_2", "cf_3_col_val_2_key_1"},
{"cf_3_col_name_3", "cf_3_col_val_3_key_1"}};
constexpr char key_2[] = "key_2";
WideColumns key_2_columns_in_cf_1{
{"cf_1_col_name_1", "cf_1_col_val_1_key_2"}};
WideColumns key_2_columns_in_cf_2{
{"cf_2_col_name_1", "cf_2_col_val_1_key_2"},
{"cf_2_col_name_2", "cf_2_col_val_2_key_2"}};
constexpr char key_3[] = "key_3";
WideColumns key_3_columns_in_cf_1{
{"cf_1_col_name_1", "cf_1_col_val_1_key_3"}};
WideColumns key_3_columns_in_cf_3{
{"cf_3_col_name_1", "cf_3_col_val_1_key_3"}};
constexpr char key_4[] = "key_4";
WideColumns key_4_columns_in_cf_0{
{"cf_0_col_name_1", "cf_0_col_val_1_key_4"}};
WideColumns key_4_columns_in_cf_2{
{"cf_2_col_name_1", "cf_2_col_val_1_key_4"}};
AttributeGroups key_1_attribute_groups{
AttributeGroup(handles_[2], key_1_columns_in_cf_2),
AttributeGroup(handles_[3], key_1_columns_in_cf_3)};
AttributeGroups key_2_attribute_groups{
AttributeGroup(handles_[1], key_2_columns_in_cf_1),
AttributeGroup(handles_[2], key_2_columns_in_cf_2)};
AttributeGroups key_3_attribute_groups{
AttributeGroup(handles_[1], key_3_columns_in_cf_1),
AttributeGroup(handles_[3], key_3_columns_in_cf_3)};
AttributeGroups key_4_attribute_groups{
AttributeGroup(handles_[0], key_4_columns_in_cf_0),
AttributeGroup(handles_[2], key_4_columns_in_cf_2)};
ASSERT_OK(db_->PutEntity(WriteOptions(), key_1, key_1_attribute_groups));
ASSERT_OK(db_->PutEntity(WriteOptions(), key_2, key_2_attribute_groups));
ASSERT_OK(db_->PutEntity(WriteOptions(), key_3, key_3_attribute_groups));
ASSERT_OK(db_->PutEntity(WriteOptions(), key_4, key_4_attribute_groups));
IteratorAttributeGroups key_1_expected_attribute_groups{
IteratorAttributeGroup(key_1_attribute_groups[0]),
IteratorAttributeGroup(key_1_attribute_groups[1])};
IteratorAttributeGroups key_2_expected_attribute_groups{
IteratorAttributeGroup(key_2_attribute_groups[0]),
IteratorAttributeGroup(key_2_attribute_groups[1])};
IteratorAttributeGroups key_3_expected_attribute_groups{
IteratorAttributeGroup(key_3_attribute_groups[0]),
IteratorAttributeGroup(key_3_attribute_groups[1])};
IteratorAttributeGroups key_4_expected_attribute_groups{
IteratorAttributeGroup(key_4_attribute_groups[0]),
IteratorAttributeGroup(key_4_attribute_groups[1])};
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
// Test for iteration over CF default->1->2->3
std::vector<ColumnFamilyHandle*> cfhs_order_0_1_2_3 = {
handles_[0], handles_[1], handles_[2], handles_[3]};
{
std::vector<Slice> expected_keys = {key_1, key_2, key_3, key_4};
std::vector<IteratorAttributeGroups> expected_attribute_groups{
key_1_expected_attribute_groups, key_2_expected_attribute_groups,
key_3_expected_attribute_groups, key_4_expected_attribute_groups};
VerifyAttributeGroupIterator(cfhs_order_0_1_2_3, expected_keys,
expected_attribute_groups);
}
Slice lb = Slice("key_2");
Slice ub = Slice("key_4");
// Test for lower bound only
{
std::vector<Slice> expected_keys = {key_2, key_3, key_4};
std::vector<IteratorAttributeGroups> expected_attribute_groups{
key_2_expected_attribute_groups, key_3_expected_attribute_groups,
key_4_expected_attribute_groups};
VerifyAttributeGroupIterator(cfhs_order_0_1_2_3, expected_keys,
expected_attribute_groups, &lb);
}
// Test for upper bound only
{
std::vector<Slice> expected_keys = {key_1, key_2, key_3};
std::vector<IteratorAttributeGroups> expected_attribute_groups{
key_1_expected_attribute_groups, key_2_expected_attribute_groups,
key_3_expected_attribute_groups};
VerifyAttributeGroupIterator(cfhs_order_0_1_2_3, expected_keys,
expected_attribute_groups, nullptr, &ub);
}
// Test for lower and upper bound
{
std::vector<Slice> expected_keys = {key_2, key_3};
std::vector<IteratorAttributeGroups> expected_attribute_groups{
key_2_expected_attribute_groups, key_3_expected_attribute_groups};
VerifyAttributeGroupIterator(cfhs_order_0_1_2_3, expected_keys,
expected_attribute_groups, &lb, &ub);
}
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
}
TEST_F(AttributeGroupIteratorTest, AllowUnpreparedValue) {
Options options = GetDefaultOptions();
CreateAndReopenWithCF({"cf_1", "cf_2", "cf_3"}, options);
constexpr char key_1[] = "key_1";
WideColumns key_1_columns_in_cf_2{
{kDefaultWideColumnName, "cf_2_col_val_0_key_1"},
{"cf_2_col_name_1", "cf_2_col_val_1_key_1"},
{"cf_2_col_name_2", "cf_2_col_val_2_key_1"}};
WideColumns key_1_columns_in_cf_3{
{"cf_3_col_name_1", "cf_3_col_val_1_key_1"},
{"cf_3_col_name_2", "cf_3_col_val_2_key_1"},
{"cf_3_col_name_3", "cf_3_col_val_3_key_1"}};
constexpr char key_2[] = "key_2";
WideColumns key_2_columns_in_cf_1{
{"cf_1_col_name_1", "cf_1_col_val_1_key_2"}};
WideColumns key_2_columns_in_cf_2{
{"cf_2_col_name_1", "cf_2_col_val_1_key_2"},
{"cf_2_col_name_2", "cf_2_col_val_2_key_2"}};
constexpr char key_3[] = "key_3";
WideColumns key_3_columns_in_cf_1{
{"cf_1_col_name_1", "cf_1_col_val_1_key_3"}};
WideColumns key_3_columns_in_cf_3{
{"cf_3_col_name_1", "cf_3_col_val_1_key_3"}};
constexpr char key_4[] = "key_4";
WideColumns key_4_columns_in_cf_0{
{"cf_0_col_name_1", "cf_0_col_val_1_key_4"}};
WideColumns key_4_columns_in_cf_2{
{"cf_2_col_name_1", "cf_2_col_val_1_key_4"}};
AttributeGroups key_1_attribute_groups{
AttributeGroup(handles_[2], key_1_columns_in_cf_2),
AttributeGroup(handles_[3], key_1_columns_in_cf_3)};
AttributeGroups key_2_attribute_groups{
AttributeGroup(handles_[1], key_2_columns_in_cf_1),
AttributeGroup(handles_[2], key_2_columns_in_cf_2)};
AttributeGroups key_3_attribute_groups{
AttributeGroup(handles_[1], key_3_columns_in_cf_1),
AttributeGroup(handles_[3], key_3_columns_in_cf_3)};
AttributeGroups key_4_attribute_groups{
AttributeGroup(handles_[0], key_4_columns_in_cf_0),
AttributeGroup(handles_[2], key_4_columns_in_cf_2)};
ASSERT_OK(db_->PutEntity(WriteOptions(), key_1, key_1_attribute_groups));
ASSERT_OK(db_->PutEntity(WriteOptions(), key_2, key_2_attribute_groups));
ASSERT_OK(db_->PutEntity(WriteOptions(), key_3, key_3_attribute_groups));
ASSERT_OK(db_->PutEntity(WriteOptions(), key_4, key_4_attribute_groups));
IteratorAttributeGroups key_1_expected_attribute_groups{
IteratorAttributeGroup(key_1_attribute_groups[0]),
IteratorAttributeGroup(key_1_attribute_groups[1])};
IteratorAttributeGroups key_2_expected_attribute_groups{
IteratorAttributeGroup(key_2_attribute_groups[0]),
IteratorAttributeGroup(key_2_attribute_groups[1])};
IteratorAttributeGroups key_3_expected_attribute_groups{
IteratorAttributeGroup(key_3_attribute_groups[0]),
IteratorAttributeGroup(key_3_attribute_groups[1])};
IteratorAttributeGroups key_4_expected_attribute_groups{
IteratorAttributeGroup(key_4_attribute_groups[0]),
IteratorAttributeGroup(key_4_attribute_groups[1])};
std::vector<ColumnFamilyHandle*> cfhs_order_0_1_2_3{handles_[0], handles_[1],
handles_[2], handles_[3]};
std::vector<Slice> expected_keys{key_1, key_2, key_3, key_4};
std::vector<IteratorAttributeGroups> expected_attribute_groups{
key_1_expected_attribute_groups, key_2_expected_attribute_groups,
key_3_expected_attribute_groups, key_4_expected_attribute_groups};
VerifyAttributeGroupIterator(
cfhs_order_0_1_2_3, expected_keys, expected_attribute_groups,
/* lower_bound */ nullptr, /* upper_bound */ nullptr,
/* allow_unprepared_value */ true);
}
Introduce MultiCfIterator (#12153) Summary: This PR introduces a new implementation of `Iterator` via a new public API called `NewMultiCfIterator()`. The new API takes a vector of column family handles to build a cross-column-family iterator, which internally maintains multiple `DBIter`s as child iterators from a consistent database state. When a key exists in multiple column families, the iterator selects the value (and wide columns) from the first column family containing the key, following the order provided in the `column_families` parameter. Similar to the merging iterator, a min heap is used to iterate across the child iterators. Backward iteration and direction change functionalities will be implemented in future PRs. The comparator used to compare keys across different column families will be derived from the iterator of the first column family specified in `column_families`. This comparator will be checked against the comparators from all other column families that the iterator will traverse. If there's a mismatch with any of the comparators, the initialization of the iterator will fail. Please note that this PR is not enough for users to start using `MultiCfIterator`. The `MultiCfIterator` and related APIs are still marked as "**DO NOT USE - UNDER CONSTRUCTION**". This PR is just the first of many PRs that will follow soon. This PR includes the following: - Introduction and partial implementation of the `MultiCfIterator`, which implements the generic `Iterator` interface. The implementation includes the construction of the iterator, `SeekToFirst()`, `Next()`, `Valid()`, `key()`, `value()`, and `columns()`. - Unit tests to verify iteration across multiple column families in two distinct scenarios: (1) keys are unique across all column families, and (2) the same keys exist in multiple column families. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12153 Reviewed By: pdillinger Differential Revision: D52308697 Pulled By: jaykorean fbshipit-source-id: b03e69f13b40af5a8f0598d0f43a0bec01ef8294
2024-03-05 18:22:43 +00:00
} // namespace ROCKSDB_NAMESPACE
int main(int argc, char** argv) {
ROCKSDB_NAMESPACE::port::InstallStackTraceHandler();
::testing::InitGoogleTest(&argc, argv);
return RUN_ALL_TESTS();
}