maropu commented on a change in pull request #26656: [SPARK-27986][SQL] Support
ANSI SQL filter clause for aggregate expression
URL: https://github.com/apache/spark/pull/26656#discussion_r355819860
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/aggregate/AggUtils.scala
##########
@@ -135,19 +135,25 @@ object AggUtils {
}
val distinctAttributes = namedDistinctExpressions.map(_.toAttribute)
val groupingAttributes = groupingExpressions.map(_.toAttribute)
+ val filterWithDistinctAttributes =
functionsWithDistinct.flatMap(_.filterAttributes.toSeq)
// 1. Create an Aggregate Operator for partial aggregations.
val partialAggregate: SparkPlan = {
val aggregateExpressions = functionsWithoutDistinct.map(_.copy(mode =
Partial))
val aggregateAttributes = aggregateExpressions.map(_.resultAttribute)
// We will group by the original grouping expression, plus an additional
expression for the
- // DISTINCT column. For example, for AVG(DISTINCT value) GROUP BY key,
the grouping
- // expressions will be [key, value].
+ // DISTINCT column and the referred attributes in the FILTER clause
associated with each
+ // aggregate function. For example:
+ // 1.for the AVG (DISTINCT value) GROUP BY key, the grouping expression
will be [key, value];
+ // 2.for AVG (DISTINCT value) Filter (WHERE value2> 20) GROUP BY key,
the grouping expression
+ // will be [key, value, value2].
Review comment:
Ah, I see... It seems we cannot simply filter out aggregate input rows in
the first (aggregate) phase because each aggregate function can have different
filter conditions and their actual aggregate processing can happen in different
phases. I think we need to study more about this impl. and add exhaustive tests
for distinct cases.
So, how about dropping support `distinct + filter aggregate` in this pr and
moving this discussion into a next PR to support `distinct + filter aggregate`?
WDYT, @cloud-fan ? We should include the support in this pr?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
