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
