[ https://issues.apache.org/jira/browse/FLINK-5159?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15703158#comment-15703158 ]
ASF GitHub Bot commented on FLINK-5159: --------------------------------------- Github user fhueske commented on a diff in the pull request: https://github.com/apache/flink/pull/2811#discussion_r89884351 --- Diff: flink-libraries/flink-table/src/main/scala/org/apache/flink/api/table/plan/rules/dataSet/DataSetSingleRowJoinRule.scala --- @@ -0,0 +1,88 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.flink.api.table.plan.rules.dataSet + +import org.apache.calcite.plan.volcano.RelSubset +import org.apache.calcite.plan.{Convention, RelOptRule, RelOptRuleCall} +import org.apache.calcite.rel.RelNode +import org.apache.calcite.rel.convert.ConverterRule +import org.apache.calcite.rel.core.JoinRelType +import org.apache.calcite.rel.logical.{LogicalAggregate, LogicalJoin} +import org.apache.flink.api.table.plan.nodes.dataset.{DataSetConvention, DataSetSingleRowJoin} + +import scala.collection.JavaConversions._ + +class DataSetSingleRowJoinRule + extends ConverterRule( + classOf[LogicalJoin], + Convention.NONE, + DataSetConvention.INSTANCE, + "DataSetSingleRowCrossRule") { + + override def matches(call: RelOptRuleCall): Boolean = { + val join = call.rel(0).asInstanceOf[LogicalJoin] + + if (isInnerJoin(join)) { + isGlobalAggregation(join.getRight.asInstanceOf[RelSubset].getOriginal) || + isGlobalAggregation(join.getLeft.asInstanceOf[RelSubset].getOriginal) + } else { + false + } + } + + private def isInnerJoin(join: LogicalJoin) = { + join.getJoinType == JoinRelType.INNER + } + + private def isGlobalAggregation(node: RelNode) = { + node.isInstanceOf[LogicalAggregate] && + isSingleLine(node.asInstanceOf[LogicalAggregate]) + } + + private def isSingleLine(agg: LogicalAggregate) = { + agg.getGroupSets == null || + (agg.getGroupSets.size() == 1 && + agg.getGroupSets.get(0).isEmpty && + agg.getGroupSet.isEmpty) + } + + override def convert(rel: RelNode): RelNode = { + val join = rel.asInstanceOf[LogicalJoin] + val traitSet = rel.getTraitSet.replace(DataSetConvention.INSTANCE) + val dataSetLeftNode = RelOptRule.convert(join.getLeft, DataSetConvention.INSTANCE) + val dataSetRightNode = RelOptRule.convert(join.getRight, DataSetConvention.INSTANCE) + val leftIsSingle = isGlobalAggregation(join.getLeft.asInstanceOf[RelSubset].getOriginal) + + new DataSetSingleRowJoin( + rel.getCluster, + traitSet, + dataSetLeftNode, + dataSetRightNode, + leftIsSingle, + rel.getRowType, + join.getCondition, + join.getRowType, + join.analyzeCondition.pairs.toList, --- End diff -- this parameter can be removed. `joinCondition` includes the complete join predicate that we need to evaluate. > Improve perfomance of inner joins with a single row input > --------------------------------------------------------- > > Key: FLINK-5159 > URL: https://issues.apache.org/jira/browse/FLINK-5159 > Project: Flink > Issue Type: Improvement > Components: Table API & SQL > Reporter: Alexander Shoshin > Assignee: Alexander Shoshin > Priority: Minor > > All inner joins (including a cross join) can be implemented as a > {{MapFunction}} if one of their inputs is a single row. This row can be > passed to a {{MapFunction}} as a {{BroadcastSet}}. > This approach is going to be more lightweight than the other current > strategies. -- This message was sent by Atlassian JIRA (v6.3.4#6332)