You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@quickstep.apache.org by zu...@apache.org on 2017/01/25 21:45:17 UTC
incubator-quickstep git commit: Added limited optimizer support for
Partitioned Hash Joins.
Repository: incubator-quickstep
Updated Branches:
refs/heads/reorder-partitioned-hash-join [created] 0b73be174
Added limited optimizer support for Partitioned Hash Joins.
Project: http://git-wip-us.apache.org/repos/asf/incubator-quickstep/repo
Commit: http://git-wip-us.apache.org/repos/asf/incubator-quickstep/commit/0b73be17
Tree: http://git-wip-us.apache.org/repos/asf/incubator-quickstep/tree/0b73be17
Diff: http://git-wip-us.apache.org/repos/asf/incubator-quickstep/diff/0b73be17
Branch: refs/heads/reorder-partitioned-hash-join
Commit: 0b73be174d2c3efead2499a3a6258aac7b73b8b9
Parents: 7bce0b8
Author: Zuyu Zhang <zu...@apache.org>
Authored: Wed Jan 25 01:49:28 2017 -0800
Committer: Zuyu Zhang <zu...@apache.org>
Committed: Wed Jan 25 12:32:59 2017 -0800
----------------------------------------------------------------------
query_optimizer/ExecutionGenerator.cpp | 143 +++++++++++++++----
.../tests/execution_generator/Partition.test | 51 +++++++
relational_operators/HashJoinOperator.cpp | 32 ++---
relational_operators/HashJoinOperator.hpp | 136 ++++++++++--------
relational_operators/WorkOrder.proto | 2 +-
relational_operators/WorkOrderFactory.cpp | 14 +-
types/TypedValue.hpp | 19 +++
7 files changed, 284 insertions(+), 113 deletions(-)
----------------------------------------------------------------------
http://git-wip-us.apache.org/repos/asf/incubator-quickstep/blob/0b73be17/query_optimizer/ExecutionGenerator.cpp
----------------------------------------------------------------------
diff --git a/query_optimizer/ExecutionGenerator.cpp b/query_optimizer/ExecutionGenerator.cpp
index e25b8ad..aced3e1 100644
--- a/query_optimizer/ExecutionGenerator.cpp
+++ b/query_optimizer/ExecutionGenerator.cpp
@@ -132,8 +132,10 @@
#include "gflags/gflags.h"
#include "glog/logging.h"
+using std::find;
using std::move;
using std::static_pointer_cast;
+using std::swap;
using std::unique_ptr;
using std::unordered_map;
using std::vector;
@@ -608,6 +610,44 @@ void ExecutionGenerator::convertSharedSubplanReference(const physical::SharedSub
}
}
+namespace {
+
+bool areSamePartitionSchemeHeaders(const PartitionSchemeHeader &lhs_partition_header,
+ const CatalogRelationSchema &lhs_scheme,
+ const PartitionSchemeHeader &rhs_partition_header,
+ const CatalogRelationSchema &rhs_scheme) {
+ if (lhs_partition_header.getPartitionType() != rhs_partition_header.getPartitionType()) {
+ return false;
+ }
+
+ if (lhs_partition_header.getNumPartitions() != rhs_partition_header.getNumPartitions()) {
+ return false;
+ }
+
+ // Check whether the underlying types in CatalogAttribute are the same.
+ if (!lhs_scheme.getAttributeById(lhs_partition_header.getPartitionAttributeId())->getType().equals(
+ rhs_scheme.getAttributeById(rhs_partition_header.getPartitionAttributeId())->getType())) {
+ return false;
+ }
+
+ switch (lhs_partition_header.getPartitionType()) {
+ case PartitionSchemeHeader::PartitionType::kHash:
+ return true;
+ case PartitionSchemeHeader::PartitionType::kRange: {
+ const vector<TypedValue> &lhs_ranges =
+ static_cast<const RangePartitionSchemeHeader&>(lhs_partition_header).getPartitionRangeBoundaries();
+ const vector<TypedValue> &rhs_ranges =
+ static_cast<const RangePartitionSchemeHeader&>(rhs_partition_header).getPartitionRangeBoundaries();
+
+ return lhs_ranges == rhs_ranges;
+ }
+ }
+
+ return false;
+}
+
+} // namespace
+
void ExecutionGenerator::convertHashJoin(const P::HashJoinPtr &physical_plan) {
// HashJoin is converted to three operators:
// BuildHash, HashJoin, DestroyHash. The second is the primary operator.
@@ -618,13 +658,10 @@ void ExecutionGenerator::convertHashJoin(const P::HashJoinPtr &physical_plan) {
std::vector<attribute_id> probe_attribute_ids;
std::vector<attribute_id> build_attribute_ids;
- std::size_t build_cardinality =
- cost_model_for_hash_join_->estimateCardinality(build_physical);
-
bool any_probe_attributes_nullable = false;
bool any_build_attributes_nullable = false;
- const std::vector<E::AttributeReferencePtr> &left_join_attributes =
+ std::vector<E::AttributeReferencePtr> left_join_attributes =
physical_plan->left_join_attributes();
for (const E::AttributeReferencePtr &left_join_attribute : left_join_attributes) {
const CatalogAttribute *probe_catalog_attribute
@@ -636,7 +673,7 @@ void ExecutionGenerator::convertHashJoin(const P::HashJoinPtr &physical_plan) {
}
}
- const std::vector<E::AttributeReferencePtr> &right_join_attributes =
+ std::vector<E::AttributeReferencePtr> right_join_attributes =
physical_plan->right_join_attributes();
for (const E::AttributeReferencePtr &right_join_attribute : right_join_attributes) {
const CatalogAttribute *build_catalog_attribute
@@ -648,6 +685,76 @@ void ExecutionGenerator::convertHashJoin(const P::HashJoinPtr &physical_plan) {
}
}
+ const CatalogRelationInfo *build_relation_info =
+ findRelationInfoOutputByPhysical(build_physical);
+ const CatalogRelationInfo *probe_operator_info =
+ findRelationInfoOutputByPhysical(probe_physical);
+
+ const CatalogRelation *build_relation = build_relation_info->relation;
+ const CatalogRelation *probe_relation = probe_operator_info->relation;
+
+ const PartitionScheme *build_partition_scheme = build_relation->getPartitionScheme();
+ const PartitionScheme *probe_partition_scheme = probe_relation->getPartitionScheme();
+
+ // Check whether we need repartition for both build and probe.
+ if (build_partition_scheme) {
+ const PartitionSchemeHeader &build_partition_scheme_header = build_partition_scheme->getPartitionSchemeHeader();
+ switch (build_partition_scheme_header.getPartitionType()) {
+ case PartitionSchemeHeader::PartitionType::kRange:
+ THROW_SQL_ERROR() << "Re-partition for range-based is not supported";
+ case PartitionSchemeHeader::PartitionType::kHash: {
+ const attribute_id build_partition_attr = build_partition_scheme_header.getPartitionAttributeId();
+ if (find(build_attribute_ids.begin(), build_attribute_ids.end(), build_partition_attr) ==
+ build_attribute_ids.end()) {
+ THROW_SQL_ERROR() << "Re-partition for hash-based is not supported";
+ }
+ break;
+ }
+ }
+ }
+
+ if (probe_partition_scheme) {
+ const PartitionSchemeHeader &probe_partition_scheme_header = probe_partition_scheme->getPartitionSchemeHeader();
+ switch (probe_partition_scheme_header.getPartitionType()) {
+ case PartitionSchemeHeader::PartitionType::kRange:
+ THROW_SQL_ERROR() << "Re-partition for range-based is not supported";
+ case PartitionSchemeHeader::PartitionType::kHash: {
+ const attribute_id probe_partition_attr = probe_partition_scheme_header.getPartitionAttributeId();
+ if (find(probe_attribute_ids.begin(), probe_attribute_ids.end(), probe_partition_attr) ==
+ probe_attribute_ids.end()) {
+ THROW_SQL_ERROR() << "Re-partition for hash-based is not supported";
+ }
+ break;
+ }
+ }
+ }
+
+ // No partition.
+ std::size_t num_partitions = 1;
+
+ // Set 'num_partitions', and swap the build and probe, if needed.
+ if (build_partition_scheme && probe_partition_scheme) {
+ const PartitionSchemeHeader &build_partition_scheme_header = build_partition_scheme->getPartitionSchemeHeader();
+ const PartitionSchemeHeader &probe_partition_scheme_header = probe_partition_scheme->getPartitionSchemeHeader();
+ if (!areSamePartitionSchemeHeaders(build_partition_scheme_header, *build_relation,
+ probe_partition_scheme_header, *probe_relation)) {
+ THROW_SQL_ERROR() << "Hash Join of two different HashPartitionSchemeHeaders is not supported";
+ }
+
+ num_partitions = build_partition_scheme->getPartitionSchemeHeader().getNumPartitions();
+ } else if (build_partition_scheme) {
+ num_partitions = build_partition_scheme->getPartitionSchemeHeader().getNumPartitions();
+ } else if (probe_partition_scheme) {
+ num_partitions = probe_partition_scheme->getPartitionSchemeHeader().getNumPartitions();
+
+ swap(probe_physical, build_physical);
+ swap(probe_attribute_ids, build_attribute_ids);
+ swap(any_probe_attributes_nullable, any_build_attributes_nullable);
+ swap(left_join_attributes, right_join_attributes);
+ swap(probe_operator_info, build_relation_info);
+ swap(probe_relation, build_relation);
+ }
+
// Remember key types for call to SimplifyHashTableImplTypeProto() below.
std::vector<const Type*> key_types;
for (std::vector<E::AttributeReferencePtr>::size_type attr_idx = 0;
@@ -677,11 +784,6 @@ void ExecutionGenerator::convertHashJoin(const P::HashJoinPtr &physical_plan) {
convertNamedExpressions(physical_plan->project_expressions(),
query_context_proto_->add_scalar_groups());
- const CatalogRelationInfo *build_relation_info =
- findRelationInfoOutputByPhysical(build_physical);
- const CatalogRelationInfo *probe_operator_info =
- findRelationInfoOutputByPhysical(probe_physical);
-
// Create a vector that indicates whether each project expression is using
// attributes from the build relation as input. This information is required
// by the current implementation of hash left outer join
@@ -694,10 +796,8 @@ void ExecutionGenerator::convertHashJoin(const P::HashJoinPtr &physical_plan) {
build_physical->getOutputAttributes())));
}
- const CatalogRelation *build_relation = build_relation_info->relation;
-
// FIXME(quickstep-team): Add support for self-join.
- if (build_relation == probe_operator_info->relation) {
+ if (build_relation == probe_relation) {
THROW_SQL_ERROR() << "Self-join is not supported";
}
@@ -706,18 +806,7 @@ void ExecutionGenerator::convertHashJoin(const P::HashJoinPtr &physical_plan) {
query_context_proto_->join_hash_tables_size();
S::QueryContext::HashTableContext *hash_table_context_proto =
query_context_proto_->add_join_hash_tables();
-
- // No partition.
- std::size_t num_partitions = 1;
- if (build_relation->hasPartitionScheme() &&
- build_attribute_ids.size() == 1) {
- const PartitionSchemeHeader &partition_scheme_header =
- build_relation->getPartitionScheme()->getPartitionSchemeHeader();
- if (build_attribute_ids[0] == partition_scheme_header.getPartitionAttributeId()) {
- // TODO(zuyu): add optimizer support for partitioned hash joins.
- hash_table_context_proto->set_num_partitions(num_partitions);
- }
- }
+ hash_table_context_proto->set_num_partitions(num_partitions);
S::HashTable *hash_table_proto = hash_table_context_proto->mutable_join_hash_table();
@@ -734,6 +823,8 @@ void ExecutionGenerator::convertHashJoin(const P::HashJoinPtr &physical_plan) {
build_relation->getAttributeById(build_attribute)->getType().getProto());
}
+ const std::size_t build_cardinality =
+ cost_model_for_hash_join_->estimateCardinality(build_physical);
hash_table_proto->set_estimated_num_entries(build_cardinality);
// Create three operators.
@@ -784,7 +875,7 @@ void ExecutionGenerator::convertHashJoin(const P::HashJoinPtr &physical_plan) {
new HashJoinOperator(
query_handle_->query_id(),
*build_relation,
- *probe_operator_info->relation,
+ *probe_relation,
probe_operator_info->isStoredRelation(),
probe_attribute_ids,
any_probe_attributes_nullable,
http://git-wip-us.apache.org/repos/asf/incubator-quickstep/blob/0b73be17/query_optimizer/tests/execution_generator/Partition.test
----------------------------------------------------------------------
diff --git a/query_optimizer/tests/execution_generator/Partition.test b/query_optimizer/tests/execution_generator/Partition.test
index ab05391..8e94316 100644
--- a/query_optimizer/tests/execution_generator/Partition.test
+++ b/query_optimizer/tests/execution_generator/Partition.test
@@ -18,12 +18,20 @@
CREATE TABLE foo (id INT NULL,
name CHAR(20))
PARTITION BY HASH(id) PARTITIONS 4;
+CREATE TABLE bar (id INT NULL,
+ score DOUBLE NULL)
+PARTITION BY HASH(id) PARTITIONS 4;
INSERT INTO foo
SELECT int_col, char_col
FROM test
WHERE int_col > 0 OR int_col < 0;
+INSERT INTO bar
+SELECT int_col, double_col
+FROM test
+WHERE int_col % 2 = 0;
+
SELECT * FROM foo;
--
+-----------+--------------------+
@@ -52,3 +60,46 @@ SELECT * FROM foo;
| -17| -17 4.123106|
| -21| -21 4.582576|
+-----------+--------------------+
+==
+
+# Partitioned Hash Join.
+SELECT bar.id, foo.name AS "char_col"
+FROM foo JOIN bar ON foo.id = bar.id;
+--
++-----------+--------------------+
+|id |char_col |
++-----------+--------------------+
+| 4| 4 2.000000|
+| 8| 8 2.828427|
+| 12| 12 3.464102|
+| 16| 16 4.000000|
+| 24| 24 4.898979|
+| 2| 2 1.414214|
+| 6| 6 2.449490|
+| 14| 14 3.741657|
+| 18| 18 4.242641|
+| 22| 22 4.690416|
++-----------+--------------------+
+==
+
+# Hash Join with two stored relations, one of which is partitioned.
+SELECT bar.id, test.char_col
+FROM test JOIN bar ON test.int_col = bar.id;
+--
+[same as above]
+==
+
+# Same as above, but has a different join order.
+SELECT bar.id, test.char_col
+FROM bar JOIN test ON bar.id = test.int_col;
+--
+[same as above]
+==
+
+# Hash Join with one stored, partitioned relation,
+# and a non-stored, non-partitioned one.
+SELECT bar.id, test.char_col
+FROM bar JOIN test ON bar.id = test.int_col
+WHERE test.int_col % 2 = 0;
+--
+[same as above]
http://git-wip-us.apache.org/repos/asf/incubator-quickstep/blob/0b73be17/relational_operators/HashJoinOperator.cpp
----------------------------------------------------------------------
diff --git a/relational_operators/HashJoinOperator.cpp b/relational_operators/HashJoinOperator.cpp
index 0062b93..7fd46c9 100644
--- a/relational_operators/HashJoinOperator.cpp
+++ b/relational_operators/HashJoinOperator.cpp
@@ -213,14 +213,14 @@ bool HashJoinOperator::getAllNonOuterJoinWorkOrders(
return true;
}
- for (std::size_t part_id = 0; part_id < num_partitions_; ++part_id) {
+ for (std::size_t part_id = 0; part_id < build_num_partitions_; ++part_id) {
const JoinHashTable &hash_table =
*(query_context->getJoinHashTable(hash_table_index_, part_id));
for (const block_id probe_block_id : probe_relation_block_ids_[part_id]) {
container->addNormalWorkOrder(
new JoinWorkOrderClass(query_id_, build_relation_, probe_relation_, join_key_attributes_,
- any_join_key_attributes_nullable_, num_partitions_, part_id, probe_block_id,
+ any_join_key_attributes_nullable_, build_num_partitions_, part_id, probe_block_id,
residual_predicate, selection, hash_table, output_destination, storage_manager,
CreateLIPFilterAdaptiveProberHelper(lip_deployment_index_, query_context)),
op_index_);
@@ -229,14 +229,14 @@ bool HashJoinOperator::getAllNonOuterJoinWorkOrders(
started_ = true;
return true;
} else {
- for (std::size_t part_id = 0; part_id < num_partitions_; ++part_id) {
+ for (std::size_t part_id = 0; part_id < build_num_partitions_; ++part_id) {
const JoinHashTable &hash_table =
*(query_context->getJoinHashTable(hash_table_index_, part_id));
while (num_workorders_generated_[part_id] < probe_relation_block_ids_[part_id].size()) {
container->addNormalWorkOrder(
new JoinWorkOrderClass(query_id_, build_relation_, probe_relation_, join_key_attributes_,
- any_join_key_attributes_nullable_, num_partitions_, part_id,
+ any_join_key_attributes_nullable_, build_num_partitions_, part_id,
probe_relation_block_ids_[part_id][num_workorders_generated_[part_id]],
residual_predicate, selection, hash_table, output_destination, storage_manager,
CreateLIPFilterAdaptiveProberHelper(lip_deployment_index_, query_context)),
@@ -269,16 +269,16 @@ bool HashJoinOperator::getAllOuterJoinWorkOrders(
return true;
}
- for (std::size_t part_id = 0; part_id < num_partitions_; ++part_id) {
+ for (std::size_t part_id = 0; part_id < build_num_partitions_; ++part_id) {
const JoinHashTable &hash_table =
*(query_context->getJoinHashTable(hash_table_index_, part_id));
for (const block_id probe_block_id : probe_relation_block_ids_[part_id]) {
container->addNormalWorkOrder(
new HashOuterJoinWorkOrder(query_id_, build_relation_, probe_relation_, join_key_attributes_,
- any_join_key_attributes_nullable_, num_partitions_, part_id, probe_block_id,
- selection, is_selection_on_build_, hash_table, output_destination,
- storage_manager,
+ any_join_key_attributes_nullable_, build_num_partitions_, part_id,
+ probe_block_id, selection, is_selection_on_build_, hash_table,
+ output_destination, storage_manager,
CreateLIPFilterAdaptiveProberHelper(lip_deployment_index_, query_context)),
op_index_);
}
@@ -286,14 +286,14 @@ bool HashJoinOperator::getAllOuterJoinWorkOrders(
started_ = true;
return true;
} else {
- for (std::size_t part_id = 0; part_id < num_partitions_; ++part_id) {
+ for (std::size_t part_id = 0; part_id < build_num_partitions_; ++part_id) {
const JoinHashTable &hash_table =
*(query_context->getJoinHashTable(hash_table_index_, part_id));
while (num_workorders_generated_[part_id] < probe_relation_block_ids_[part_id].size()) {
container->addNormalWorkOrder(
new HashOuterJoinWorkOrder(query_id_, build_relation_, probe_relation_, join_key_attributes_,
- any_join_key_attributes_nullable_, num_partitions_, part_id,
+ any_join_key_attributes_nullable_, build_num_partitions_, part_id,
probe_relation_block_ids_[part_id][num_workorders_generated_[part_id]],
selection, is_selection_on_build_, hash_table, output_destination,
storage_manager,
@@ -336,7 +336,7 @@ bool HashJoinOperator::getAllNonOuterJoinWorkOrderProtos(
return true;
}
- for (std::size_t part_id = 0; part_id < num_partitions_; ++part_id) {
+ for (std::size_t part_id = 0; part_id < build_num_partitions_; ++part_id) {
for (const block_id probe_block_id : probe_relation_block_ids_[part_id]) {
container->addWorkOrderProto(
createNonOuterJoinWorkOrderProto(hash_join_type, probe_block_id, part_id),
@@ -346,7 +346,7 @@ bool HashJoinOperator::getAllNonOuterJoinWorkOrderProtos(
started_ = true;
return true;
} else {
- for (std::size_t part_id = 0; part_id < num_partitions_; ++part_id) {
+ for (std::size_t part_id = 0; part_id < build_num_partitions_; ++part_id) {
while (num_workorders_generated_[part_id] < probe_relation_block_ids_[part_id].size()) {
container->addWorkOrderProto(
createNonOuterJoinWorkOrderProto(hash_join_type,
@@ -376,7 +376,7 @@ serialization::WorkOrder* HashJoinOperator::createNonOuterJoinWorkOrderProto(
}
proto->SetExtension(serialization::HashJoinWorkOrder::any_join_key_attributes_nullable,
any_join_key_attributes_nullable_);
- proto->SetExtension(serialization::HashJoinWorkOrder::num_partitions, num_partitions_);
+ proto->SetExtension(serialization::HashJoinWorkOrder::build_num_partitions, build_num_partitions_);
proto->SetExtension(serialization::HashJoinWorkOrder::insert_destination_index, output_destination_index_);
proto->SetExtension(serialization::HashJoinWorkOrder::join_hash_table_index, hash_table_index_);
proto->SetExtension(serialization::HashJoinWorkOrder::partition_id, part_id);
@@ -399,7 +399,7 @@ bool HashJoinOperator::getAllOuterJoinWorkOrderProtos(WorkOrderProtosContainer *
return true;
}
- for (std::size_t part_id = 0; part_id < num_partitions_; ++part_id) {
+ for (std::size_t part_id = 0; part_id < build_num_partitions_; ++part_id) {
for (const block_id probe_block_id : probe_relation_block_ids_[part_id]) {
container->addWorkOrderProto(createOuterJoinWorkOrderProto(probe_block_id, part_id), op_index_);
}
@@ -407,7 +407,7 @@ bool HashJoinOperator::getAllOuterJoinWorkOrderProtos(WorkOrderProtosContainer *
started_ = true;
return true;
} else {
- for (std::size_t part_id = 0; part_id < num_partitions_; ++part_id) {
+ for (std::size_t part_id = 0; part_id < build_num_partitions_; ++part_id) {
while (num_workorders_generated_[part_id] < probe_relation_block_ids_[part_id].size()) {
container->addWorkOrderProto(
createOuterJoinWorkOrderProto(probe_relation_block_ids_[part_id][num_workorders_generated_[part_id]],
@@ -436,7 +436,7 @@ serialization::WorkOrder* HashJoinOperator::createOuterJoinWorkOrderProto(const
}
proto->SetExtension(serialization::HashJoinWorkOrder::any_join_key_attributes_nullable,
any_join_key_attributes_nullable_);
- proto->SetExtension(serialization::HashJoinWorkOrder::num_partitions, num_partitions_);
+ proto->SetExtension(serialization::HashJoinWorkOrder::build_num_partitions, build_num_partitions_);
proto->SetExtension(serialization::HashJoinWorkOrder::insert_destination_index, output_destination_index_);
proto->SetExtension(serialization::HashJoinWorkOrder::join_hash_table_index, hash_table_index_);
proto->SetExtension(serialization::HashJoinWorkOrder::selection_index, selection_index_);
http://git-wip-us.apache.org/repos/asf/incubator-quickstep/blob/0b73be17/relational_operators/HashJoinOperator.hpp
----------------------------------------------------------------------
diff --git a/relational_operators/HashJoinOperator.hpp b/relational_operators/HashJoinOperator.hpp
index e655f70..5cf7cb2 100644
--- a/relational_operators/HashJoinOperator.hpp
+++ b/relational_operators/HashJoinOperator.hpp
@@ -101,8 +101,8 @@ class HashJoinOperator : public RelationalOperator {
* @param join_key_attributes The IDs of equijoin attributes in
* probe_relation.
* @param any_join_key_attributes_nullable If any attribute is nullable.
- * @param num_partitions The number of partitions in 'input_relation'. If no
- * partitions, it is one.
+ * @param build_num_partitions The number of partitions in 'build_relation'.
+ * If no partitions, it is one.
* @param output_relation The output relation.
* @param output_destination_index The index of the InsertDestination in the
* QueryContext to insert the join results.
@@ -128,7 +128,7 @@ class HashJoinOperator : public RelationalOperator {
const bool probe_relation_is_stored,
const std::vector<attribute_id> &join_key_attributes,
const bool any_join_key_attributes_nullable,
- const std::size_t num_partitions,
+ const std::size_t build_num_partitions,
const CatalogRelation &output_relation,
const QueryContext::insert_destination_id output_destination_index,
const QueryContext::join_hash_table_id hash_table_index,
@@ -142,7 +142,7 @@ class HashJoinOperator : public RelationalOperator {
probe_relation_is_stored_(probe_relation_is_stored),
join_key_attributes_(join_key_attributes),
any_join_key_attributes_nullable_(any_join_key_attributes_nullable),
- num_partitions_(num_partitions),
+ build_num_partitions_(build_num_partitions),
output_relation_(output_relation),
output_destination_index_(output_destination_index),
hash_table_index_(hash_table_index),
@@ -152,8 +152,8 @@ class HashJoinOperator : public RelationalOperator {
? std::vector<bool>()
: *is_selection_on_build),
join_type_(join_type),
- probe_relation_block_ids_(num_partitions),
- num_workorders_generated_(num_partitions),
+ probe_relation_block_ids_(build_num_partitions),
+ num_workorders_generated_(build_num_partitions),
started_(false) {
DCHECK(join_type != JoinType::kLeftOuterJoin ||
(is_selection_on_build != nullptr &&
@@ -162,12 +162,15 @@ class HashJoinOperator : public RelationalOperator {
if (probe_relation_is_stored) {
if (probe_relation.hasPartitionScheme()) {
const PartitionScheme &part_scheme = *probe_relation.getPartitionScheme();
- for (std::size_t part_id = 0; part_id < num_partitions_; ++part_id) {
+ DCHECK_EQ(build_num_partitions_, part_scheme.getPartitionSchemeHeader().getNumPartitions());
+ for (std::size_t part_id = 0; part_id < build_num_partitions_; ++part_id) {
probe_relation_block_ids_[part_id] = part_scheme.getBlocksInPartition(part_id);
}
} else {
- // No partition.
- probe_relation_block_ids_[0] = probe_relation.getBlocksSnapshot();
+ // Broadcast join if probe has no partitions.
+ for (std::size_t part_id = 0; part_id < build_num_partitions_; ++part_id) {
+ probe_relation_block_ids_[part_id] = probe_relation.getBlocksSnapshot();
+ }
}
}
}
@@ -209,7 +212,14 @@ class HashJoinOperator : public RelationalOperator {
const partition_id part_id) override {
DCHECK_EQ(probe_relation_.getID(), input_relation_id);
- probe_relation_block_ids_[part_id].push_back(input_block_id);
+ if (probe_relation_.hasPartitionScheme()) {
+ probe_relation_block_ids_[part_id].push_back(input_block_id);
+ } else {
+ // Broadcast join if probe has no partitions.
+ for (std::size_t build_part_id = 0; build_part_id < build_num_partitions_; ++build_part_id) {
+ probe_relation_block_ids_[build_part_id].push_back(input_block_id);
+ }
+ }
}
QueryContext::insert_destination_id getInsertDestinationID() const override {
@@ -261,7 +271,7 @@ class HashJoinOperator : public RelationalOperator {
const bool probe_relation_is_stored_;
const std::vector<attribute_id> join_key_attributes_;
const bool any_join_key_attributes_nullable_;
- const std::size_t num_partitions_;
+ const std::size_t build_num_partitions_;
const CatalogRelation &output_relation_;
const QueryContext::insert_destination_id output_destination_index_;
const QueryContext::join_hash_table_id hash_table_index_;
@@ -295,8 +305,8 @@ class HashInnerJoinWorkOrder : public WorkOrder {
* @param join_key_attributes The IDs of equijoin attributes in \c
* probe_relation.
* @param any_join_key_attributes_nullable If any attribute is nullable.
- * @param num_partitions The number of partitions in 'probe_relation'. If no
- * partitions, it is one.
+ * @param build_num_partitions The number of partitions in 'build_relation'.
+ * If no partitions, it is one.
* @param part_id The partition id of 'probe_relation'.
* @param lookup_block_id The block id of the probe_relation.
* @param residual_predicate If non-null, apply as an additional filter to
@@ -317,7 +327,7 @@ class HashInnerJoinWorkOrder : public WorkOrder {
const CatalogRelationSchema &probe_relation,
const std::vector<attribute_id> &join_key_attributes,
const bool any_join_key_attributes_nullable,
- const std::size_t num_partitions,
+ const std::size_t build_num_partitions,
const partition_id part_id,
const block_id lookup_block_id,
const Predicate *residual_predicate,
@@ -331,7 +341,7 @@ class HashInnerJoinWorkOrder : public WorkOrder {
probe_relation_(probe_relation),
join_key_attributes_(join_key_attributes),
any_join_key_attributes_nullable_(any_join_key_attributes_nullable),
- num_partitions_(num_partitions),
+ build_num_partitions_(build_num_partitions),
part_id_(part_id),
block_id_(lookup_block_id),
residual_predicate_(residual_predicate),
@@ -352,8 +362,8 @@ class HashInnerJoinWorkOrder : public WorkOrder {
* @param join_key_attributes The IDs of equijoin attributes in \c
* probe_relation.
* @param any_join_key_attributes_nullable If any attribute is nullable.
- * @param num_partitions The number of partitions in 'probe_relation'. If no
- * partitions, it is one.
+ * @param build_num_partitions The number of partitions in 'build_relation'.
+ * If no partitions, it is one.
* @param part_id The partition id of 'probe_relation'.
* @param lookup_block_id The block id of the probe_relation.
* @param residual_predicate If non-null, apply as an additional filter to
@@ -374,7 +384,7 @@ class HashInnerJoinWorkOrder : public WorkOrder {
const CatalogRelationSchema &probe_relation,
std::vector<attribute_id> &&join_key_attributes,
const bool any_join_key_attributes_nullable,
- const std::size_t num_partitions,
+ const std::size_t build_num_partitions,
const partition_id part_id,
const block_id lookup_block_id,
const Predicate *residual_predicate,
@@ -388,7 +398,7 @@ class HashInnerJoinWorkOrder : public WorkOrder {
probe_relation_(probe_relation),
join_key_attributes_(std::move(join_key_attributes)),
any_join_key_attributes_nullable_(any_join_key_attributes_nullable),
- num_partitions_(num_partitions),
+ build_num_partitions_(build_num_partitions),
part_id_(part_id),
block_id_(lookup_block_id),
residual_predicate_(residual_predicate),
@@ -411,12 +421,12 @@ class HashInnerJoinWorkOrder : public WorkOrder {
void execute() override;
/**
- * @brief Get the number of partitions.
+ * @brief Get the number of partitions in 'build_relation'.
*
- * @return The number of partitions.
+ * @return The number of partitions in 'build_relation'.
*/
- std::size_t num_partitions() const {
- return num_partitions_;
+ std::size_t build_num_partitions() const {
+ return build_num_partitions_;
}
/**
@@ -433,7 +443,7 @@ class HashInnerJoinWorkOrder : public WorkOrder {
const CatalogRelationSchema &probe_relation_;
const std::vector<attribute_id> join_key_attributes_;
const bool any_join_key_attributes_nullable_;
- const std::size_t num_partitions_;
+ const std::size_t build_num_partitions_;
const partition_id part_id_;
const block_id block_id_;
const Predicate *residual_predicate_;
@@ -465,8 +475,8 @@ class HashSemiJoinWorkOrder : public WorkOrder {
* @param join_key_attributes The IDs of equijoin attributes in \c
* probe_relation.
* @param any_join_key_attributes_nullable If any attribute is nullable.
- * @param num_partitions The number of partitions in 'probe_relation'. If no
- * partitions, it is one.
+ * @param build_num_partitions The number of partitions in 'build_relation'.
+ * If no partitions, it is one.
* @param part_id The partition id of 'probe_relation'.
* @param lookup_block_id The block id of the probe_relation.
* @param residual_predicate If non-null, apply as an additional filter to
@@ -487,7 +497,7 @@ class HashSemiJoinWorkOrder : public WorkOrder {
const CatalogRelationSchema &probe_relation,
const std::vector<attribute_id> &join_key_attributes,
const bool any_join_key_attributes_nullable,
- const std::size_t num_partitions,
+ const std::size_t build_num_partitions,
const partition_id part_id,
const block_id lookup_block_id,
const Predicate *residual_predicate,
@@ -501,7 +511,7 @@ class HashSemiJoinWorkOrder : public WorkOrder {
probe_relation_(probe_relation),
join_key_attributes_(join_key_attributes),
any_join_key_attributes_nullable_(any_join_key_attributes_nullable),
- num_partitions_(num_partitions),
+ build_num_partitions_(build_num_partitions),
part_id_(part_id),
block_id_(lookup_block_id),
residual_predicate_(residual_predicate),
@@ -522,8 +532,8 @@ class HashSemiJoinWorkOrder : public WorkOrder {
* @param join_key_attributes The IDs of equijoin attributes in \c
* probe_relation.
* @param any_join_key_attributes_nullable If any attribute is nullable.
- * @param num_partitions The number of partitions in 'probe_relation'. If no
- * partitions, it is one.
+ * @param build_num_partitions The number of partitions in 'build_relation'.
+ * If no partitions, it is one.
* @param part_id The partition id of 'probe_relation'.
* @param lookup_block_id The block id of the probe_relation.
* @param residual_predicate If non-null, apply as an additional filter to
@@ -544,7 +554,7 @@ class HashSemiJoinWorkOrder : public WorkOrder {
const CatalogRelationSchema &probe_relation,
std::vector<attribute_id> &&join_key_attributes,
const bool any_join_key_attributes_nullable,
- const std::size_t num_partitions,
+ const std::size_t build_num_partitions,
const partition_id part_id,
const block_id lookup_block_id,
const Predicate *residual_predicate,
@@ -558,7 +568,7 @@ class HashSemiJoinWorkOrder : public WorkOrder {
probe_relation_(probe_relation),
join_key_attributes_(std::move(join_key_attributes)),
any_join_key_attributes_nullable_(any_join_key_attributes_nullable),
- num_partitions_(num_partitions),
+ build_num_partitions_(build_num_partitions),
part_id_(part_id),
block_id_(lookup_block_id),
residual_predicate_(residual_predicate),
@@ -573,12 +583,12 @@ class HashSemiJoinWorkOrder : public WorkOrder {
void execute() override;
/**
- * @brief Get the number of partitions.
+ * @brief Get the number of partitions in 'build_relation'.
*
- * @return The number of partitions.
+ * @return The number of partitions in 'build_relation'.
*/
- std::size_t num_partitions() const {
- return num_partitions_;
+ std::size_t build_num_partitions() const {
+ return build_num_partitions_;
}
/**
@@ -599,7 +609,7 @@ class HashSemiJoinWorkOrder : public WorkOrder {
const CatalogRelationSchema &probe_relation_;
const std::vector<attribute_id> join_key_attributes_;
const bool any_join_key_attributes_nullable_;
- const std::size_t num_partitions_;
+ const std::size_t build_num_partitions_;
const partition_id part_id_;
const block_id block_id_;
const Predicate *residual_predicate_;
@@ -631,8 +641,8 @@ class HashAntiJoinWorkOrder : public WorkOrder {
* @param join_key_attributes The IDs of equijoin attributes in \c
* probe_relation.
* @param any_join_key_attributes_nullable If any attribute is nullable.
- * @param num_partitions The number of partitions in 'probe_relation'. If no
- * partitions, it is one.
+ * @param build_num_partitions The number of partitions in 'build_relation'.
+ * If no partitions, it is one.
* @param part_id The partition id of 'probe_relation'.
* @param lookup_block_id The block id of the probe_relation.
* @param residual_predicate If non-null, apply as an additional filter to
@@ -653,7 +663,7 @@ class HashAntiJoinWorkOrder : public WorkOrder {
const CatalogRelationSchema &probe_relation,
const std::vector<attribute_id> &join_key_attributes,
const bool any_join_key_attributes_nullable,
- const std::size_t num_partitions,
+ const std::size_t build_num_partitions,
const partition_id part_id,
const block_id lookup_block_id,
const Predicate *residual_predicate,
@@ -667,7 +677,7 @@ class HashAntiJoinWorkOrder : public WorkOrder {
probe_relation_(probe_relation),
join_key_attributes_(join_key_attributes),
any_join_key_attributes_nullable_(any_join_key_attributes_nullable),
- num_partitions_(num_partitions),
+ build_num_partitions_(build_num_partitions),
part_id_(part_id),
block_id_(lookup_block_id),
residual_predicate_(residual_predicate),
@@ -688,8 +698,8 @@ class HashAntiJoinWorkOrder : public WorkOrder {
* @param join_key_attributes The IDs of equijoin attributes in \c
* probe_relation.
* @param any_join_key_attributes_nullable If any attribute is nullable.
- * @param num_partitions The number of partitions in 'probe_relation'. If no
- * partitions, it is one.
+ * @param build_num_partitions The number of partitions in 'build_relation'.
+ * If no partitions, it is one.
* @param part_id The partition id of 'probe_relation'.
* @param lookup_block_id The block id of the probe_relation.
* @param residual_predicate If non-null, apply as an additional filter to
@@ -710,7 +720,7 @@ class HashAntiJoinWorkOrder : public WorkOrder {
const CatalogRelationSchema &probe_relation,
std::vector<attribute_id> &&join_key_attributes,
const bool any_join_key_attributes_nullable,
- const std::size_t num_partitions,
+ const std::size_t build_num_partitions,
const partition_id part_id,
const block_id lookup_block_id,
const Predicate *residual_predicate,
@@ -724,7 +734,7 @@ class HashAntiJoinWorkOrder : public WorkOrder {
probe_relation_(probe_relation),
join_key_attributes_(std::move(join_key_attributes)),
any_join_key_attributes_nullable_(any_join_key_attributes_nullable),
- num_partitions_(num_partitions),
+ build_num_partitions_(build_num_partitions),
part_id_(part_id),
block_id_(lookup_block_id),
residual_predicate_(residual_predicate),
@@ -745,12 +755,12 @@ class HashAntiJoinWorkOrder : public WorkOrder {
}
/**
- * @brief Get the number of partitions.
+ * @brief Get the number of partitions in 'build_relation'.
*
- * @return The number of partitions.
+ * @return The number of partitions in 'build_relation'.
*/
- std::size_t num_partitions() const {
- return num_partitions_;
+ std::size_t build_num_partitions() const {
+ return build_num_partitions_;
}
/**
@@ -771,7 +781,7 @@ class HashAntiJoinWorkOrder : public WorkOrder {
const CatalogRelationSchema &probe_relation_;
const std::vector<attribute_id> join_key_attributes_;
const bool any_join_key_attributes_nullable_;
- const std::size_t num_partitions_;
+ const std::size_t build_num_partitions_;
const partition_id part_id_;
const block_id block_id_;
const Predicate *residual_predicate_;
@@ -802,8 +812,8 @@ class HashOuterJoinWorkOrder : public WorkOrder {
* @param join_key_attributes The IDs of equijoin attributes in \c
* probe_relation.
* @param any_join_key_attributes_nullable If any attribute is nullable.
- * @param num_partitions The number of partitions in 'probe_relation'. If no
- * partitions, it is one.
+ * @param build_num_partitions The number of partitions in 'build_relation'.
+ * If no partitions, it is one.
* @param part_id The partition id of 'probe_relation'.
* @param lookup_block_id The block id of the probe_relation.
* @param selection A list of Scalars corresponding to the relation attributes
@@ -823,7 +833,7 @@ class HashOuterJoinWorkOrder : public WorkOrder {
const CatalogRelationSchema &probe_relation,
const std::vector<attribute_id> &join_key_attributes,
const bool any_join_key_attributes_nullable,
- const std::size_t num_partitions,
+ const std::size_t build_num_partitions,
const partition_id part_id,
const block_id lookup_block_id,
const std::vector<std::unique_ptr<const Scalar>> &selection,
@@ -837,7 +847,7 @@ class HashOuterJoinWorkOrder : public WorkOrder {
probe_relation_(probe_relation),
join_key_attributes_(join_key_attributes),
any_join_key_attributes_nullable_(any_join_key_attributes_nullable),
- num_partitions_(num_partitions),
+ build_num_partitions_(build_num_partitions),
part_id_(part_id),
block_id_(lookup_block_id),
selection_(selection),
@@ -858,8 +868,8 @@ class HashOuterJoinWorkOrder : public WorkOrder {
* @param join_key_attributes The IDs of equijoin attributes in \c
* probe_relation.
* @param any_join_key_attributes_nullable If any attribute is nullable.
- * @param num_partitions The number of partitions in 'probe_relation'. If no
- * partitions, it is one.
+ * @param build_num_partitions The number of partitions in 'build_relation'.
+ * If no partitions, it is one.
* @param part_id The partition id of 'probe_relation'.
* @param lookup_block_id The block id of the probe_relation.
* @param selection A list of Scalars corresponding to the relation attributes
@@ -878,7 +888,7 @@ class HashOuterJoinWorkOrder : public WorkOrder {
const CatalogRelationSchema &probe_relation,
std::vector<attribute_id> &&join_key_attributes,
const bool any_join_key_attributes_nullable,
- const std::size_t num_partitions,
+ const std::size_t build_num_partitions,
const partition_id part_id,
const block_id lookup_block_id,
const std::vector<std::unique_ptr<const Scalar>> &selection,
@@ -892,7 +902,7 @@ class HashOuterJoinWorkOrder : public WorkOrder {
probe_relation_(probe_relation),
join_key_attributes_(std::move(join_key_attributes)),
any_join_key_attributes_nullable_(any_join_key_attributes_nullable),
- num_partitions_(num_partitions),
+ build_num_partitions_(build_num_partitions),
part_id_(part_id),
block_id_(lookup_block_id),
selection_(selection),
@@ -907,12 +917,12 @@ class HashOuterJoinWorkOrder : public WorkOrder {
void execute() override;
/**
- * @brief Get the number of partitions.
+ * @brief Get the number of partitions in 'build_relation'.
*
- * @return The number of partitions.
+ * @return The number of partitions in 'build_relation'.
*/
- std::size_t num_partitions() const {
- return num_partitions_;
+ std::size_t build_num_partitions() const {
+ return build_num_partitions_;
}
/**
@@ -929,7 +939,7 @@ class HashOuterJoinWorkOrder : public WorkOrder {
const CatalogRelationSchema &probe_relation_;
const std::vector<attribute_id> join_key_attributes_;
const bool any_join_key_attributes_nullable_;
- const std::size_t num_partitions_;
+ const std::size_t build_num_partitions_;
const partition_id part_id_;
const block_id block_id_;
const std::vector<std::unique_ptr<const Scalar>> &selection_;
http://git-wip-us.apache.org/repos/asf/incubator-quickstep/blob/0b73be17/relational_operators/WorkOrder.proto
----------------------------------------------------------------------
diff --git a/relational_operators/WorkOrder.proto b/relational_operators/WorkOrder.proto
index b914fce..b42d377 100644
--- a/relational_operators/WorkOrder.proto
+++ b/relational_operators/WorkOrder.proto
@@ -129,7 +129,7 @@ message HashJoinWorkOrder {
optional int32 probe_relation_id = 162;
repeated int32 join_key_attributes = 163;
optional bool any_join_key_attributes_nullable = 164;
- optional uint64 num_partitions = 172;
+ optional uint64 build_num_partitions = 172;
optional int32 insert_destination_index = 165;
optional uint32 join_hash_table_index = 166;
optional uint64 partition_id = 173;
http://git-wip-us.apache.org/repos/asf/incubator-quickstep/blob/0b73be17/relational_operators/WorkOrderFactory.cpp
----------------------------------------------------------------------
diff --git a/relational_operators/WorkOrderFactory.cpp b/relational_operators/WorkOrderFactory.cpp
index c09bcbe..d6d0818 100644
--- a/relational_operators/WorkOrderFactory.cpp
+++ b/relational_operators/WorkOrderFactory.cpp
@@ -199,8 +199,8 @@ WorkOrder* WorkOrderFactory::ReconstructFromProto(const serialization::WorkOrder
const block_id lookup_block_id =
proto.GetExtension(serialization::HashJoinWorkOrder::block_id);
- const std::size_t num_partitions =
- proto.GetExtension(serialization::HashJoinWorkOrder::num_partitions);
+ const std::size_t build_num_partitions =
+ proto.GetExtension(serialization::HashJoinWorkOrder::build_num_partitions);
const Predicate *residual_predicate = nullptr;
if (hash_join_work_order_type != serialization::HashJoinWorkOrder::HASH_OUTER_JOIN) {
@@ -233,7 +233,7 @@ WorkOrder* WorkOrderFactory::ReconstructFromProto(const serialization::WorkOrder
probe_relation,
move(join_key_attributes),
any_join_key_attributes_nullable,
- num_partitions,
+ build_num_partitions,
part_id,
lookup_block_id,
residual_predicate,
@@ -251,7 +251,7 @@ WorkOrder* WorkOrderFactory::ReconstructFromProto(const serialization::WorkOrder
probe_relation,
move(join_key_attributes),
any_join_key_attributes_nullable,
- num_partitions,
+ build_num_partitions,
part_id,
lookup_block_id,
residual_predicate,
@@ -277,7 +277,7 @@ WorkOrder* WorkOrderFactory::ReconstructFromProto(const serialization::WorkOrder
probe_relation,
move(join_key_attributes),
any_join_key_attributes_nullable,
- num_partitions,
+ build_num_partitions,
part_id,
lookup_block_id,
selection,
@@ -295,7 +295,7 @@ WorkOrder* WorkOrderFactory::ReconstructFromProto(const serialization::WorkOrder
probe_relation,
move(join_key_attributes),
any_join_key_attributes_nullable,
- num_partitions,
+ build_num_partitions,
part_id,
lookup_block_id,
residual_predicate,
@@ -648,7 +648,7 @@ bool WorkOrderFactory::ProtoIsValid(const serialization::WorkOrder &proto,
query_context.isValidInsertDestinationId(
proto.GetExtension(serialization::HashJoinWorkOrder::insert_destination_index)) &&
proto.HasExtension(serialization::HashJoinWorkOrder::join_hash_table_index) &&
- proto.HasExtension(serialization::HashJoinWorkOrder::num_partitions) &&
+ proto.HasExtension(serialization::HashJoinWorkOrder::build_num_partitions) &&
proto.HasExtension(serialization::HashJoinWorkOrder::partition_id) &&
query_context.isValidJoinHashTableId(
proto.GetExtension(serialization::HashJoinWorkOrder::join_hash_table_index),
http://git-wip-us.apache.org/repos/asf/incubator-quickstep/blob/0b73be17/types/TypedValue.hpp
----------------------------------------------------------------------
diff --git a/types/TypedValue.hpp b/types/TypedValue.hpp
index d75720a..832690d 100644
--- a/types/TypedValue.hpp
+++ b/types/TypedValue.hpp
@@ -253,6 +253,25 @@ class TypedValue {
}
/**
+ * @brief Equal operator.
+ **/
+ bool operator==(const TypedValue &rhs) const {
+ if (getTypeID() != rhs.getTypeID()) {
+ return false;
+ }
+
+ if (isNull() != rhs.isNull()) {
+ return false;
+ }
+
+ if (isNull()) {
+ return true;
+ }
+
+ return fastEqualCheck(rhs);
+ }
+
+ /**
* @brief Create a new literal TypedValue with pre-allocated out-of-line
* data.
* @warning The memory at value_ptr must be allocated with malloc() or