Github user cloud-fan commented on a diff in the pull request:
https://github.com/apache/spark/pull/10451#discussion_r48432615
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala
---
@@ -153,6 +153,15 @@ object SetOperationPushDown extends Rule[LogicalPlan]
with PredicateHelper {
)
)
+ // Push down limit into union
+ case Limit(exp, Union(left, right)) =>
+ Limit(exp,
+ Union(
+ CombineLimits(Limit(exp, left)),
+ CombineLimits(Limit(exp, right))
+ )
+ )
--- End diff --
After think it more, there may be a problem: If `left` or `right` is an
operator that can push `Limit` down(currently there is no such operator, but we
can't guarantee there won't be). Then every time you push down a `Limit` here,
it will be pushed down further. Thus the `CombineLimits` can NOT detect that
you have already pushed the `Limit` down, and keeps generating new `Limit` and
pushing it down.
I think we should have a better way to detect whether we have pushed
`Limit` down or not, or add some comments to say that this rule assumes the
newly added `Limit` on top of `left` and `right` won't be removed by other
optimization rules.
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
