Github user mallman commented on a diff in the pull request: https://github.com/apache/spark/pull/16578#discussion_r140366890 --- Diff: sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/AggregateFieldExtractionPushdown.scala --- @@ -0,0 +1,77 @@ +/* + * 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.spark.sql.catalyst.optimizer + +import org.apache.spark.sql.catalyst.expressions.{Attribute, AttributeSet, NamedExpression} +import org.apache.spark.sql.catalyst.plans.logical.{Aggregate, LogicalPlan, Project} + +/** + * Pushes down aliases to [[expressions.GetStructField]] expressions in an aggregate's grouping and + * aggregate expressions into a projection over its children. The original + * [[expressions.GetStructField]] expressions are replaced with references to the pushed down + * aliases. + */ +object AggregateFieldExtractionPushdown extends FieldExtractionPushdown { + override def apply(plan: LogicalPlan): LogicalPlan = + plan transformDown { + case agg @ Aggregate(groupingExpressions, aggregateExpressions, child) => + val expressions = groupingExpressions ++ aggregateExpressions + val attributes = AttributeSet(expressions.collect { case att: Attribute => att }) + val childAttributes = AttributeSet(child.expressions) + val fieldExtractors0 = + expressions + .flatMap(getFieldExtractors) + .distinct + val fieldExtractors1 = + fieldExtractors0 + .filter(_.collectFirst { case att: Attribute => att } + .filter(attributes.contains).isEmpty) --- End diff -- > Why do we need to trim the extractors which contain attributes referred from `groupingExpressions ++ aggregateExpressions`? Consider this query: ```sql select a, a.b, count(1) from r1 group by a, a.b ``` The grouping fields are `a` and `a.b`. `a` is an `Attribute`. `a.b` is a `GetStructField`. Since we need all of `a` to answer this query, it doesn't make sense to attempt to push down `a.b`. At the same time, `fieldExtractors0` includes all `GetStructField` instances. This includes `a.b`. The code you refer to above filters out the `a.b` `GetStructField` because our query requires all of `a`. If we do not filter out `a.b`, then the child (projection) of the new `Aggregate` will not contain `a` in its `output`. The query planner will barf. The logic for the creation of the new child projection is here: https://github.com/apache/spark/blob/00ab80c9b78c45c1a8f8c202c5bab04a62cda2ef/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/AggregateFieldExtractionPushdown.scala#L63-L70 This case is tested by https://github.com/apache/spark/blob/00ab80c9b78c45c1a8f8c202c5bab04a62cda2ef/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/optimizer/AggregateFieldExtractionPushdownSuite.scala#L60-L76
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org