Github user JoshRosen commented on a diff in the pull request:
https://github.com/apache/spark/pull/7560#discussion_r35073547
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/GeneratedAggregate.scala
---
@@ -266,7 +266,18 @@ case class GeneratedAggregate(
val joinedRow = new JoinedRow3
- if (groupingExpressions.isEmpty) {
+ if (!iter.hasNext) {
+ // This is an empty input, so return early so that we do not
allocate data structures
+ // that won't be cleaned up (see SPARK-8357).
+ if (groupingExpressions.isEmpty) {
--- End diff --
Here, I made a slight simplification compared to @navis's original patch:
if `groupingExpressions` is empty and the input is empty, then always return an
empty aggregation buffer. @navis's patch contained an additional branch here
which would skip this output if `partial = true`, but I think that is an
unnecessary performance optimization given that the non-generated-Aggregate
operator still outputs an empty row even on empty inputs. Removing this branch
means fewer cases to have to test.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
