huaxingao commented on code in PR #36043:
URL: https://github.com/apache/spark/pull/36043#discussion_r841285841
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/v2/V2ScanRelationPushDown.scala:
##########
@@ -380,27 +380,32 @@ object V2ScanRelationPushDown extends Rule[LogicalPlan]
with PredicateHelper wit
sHolder.pushedLimit = Some(limit)
sHolder.sortOrders = orders
if (isPartiallyPushed) {
- s
+ (s, isPartiallyPushed)
} else {
- operation
+ (operation, isPartiallyPushed)
}
} else {
- s
+ (s, true)
}
} else {
- s
+ (s, true)
}
case p: Project =>
- val newChild = pushDownLimit(p.child, limit)
- p.withNewChildren(Seq(newChild))
- case other => other
+ val (newChild, isPartiallyPushed) = pushDownLimit(p.child, limit)
+ (p.withNewChildren(Seq(newChild)), isPartiallyPushed)
+ case other => (other, true)
}
def pushDownLimits(plan: LogicalPlan): LogicalPlan = plan.transform {
case globalLimit @ Limit(IntegerLiteral(limitValue), child) =>
- val newChild = pushDownLimit(child, limitValue)
- val newLocalLimit =
globalLimit.child.asInstanceOf[LocalLimit].withNewChildren(Seq(newChild))
- globalLimit.withNewChildren(Seq(newLocalLimit))
+ val (newChild, isPartiallyPushed) = pushDownLimit(child, limitValue)
+ if (isPartiallyPushed) {
+ val newLocalLimit =
+
globalLimit.child.asInstanceOf[LocalLimit].withNewChildren(Seq(newChild))
+ globalLimit.withNewChildren(Seq(newLocalLimit))
+ } else {
+ newChild
Review Comment:
I think there is a problem here. If `isPartiallyPushed` is false, it is
assumed that `Limit` is completely pushed down so Spark doesn't do `Limit` any
more. However, the `isPartiallyPushed` false could come from the default case
in `PushDownUtils`.`pushLimit`
```
def pushLimit(scanBuilder: ScanBuilder, limit: Int): (Boolean, Boolean) = {
scanBuilder match {
case s: SupportsPushDownLimit if s.pushLimit(limit) =>
(true, s.isPartiallyPushed)
case _ => (false, false)
}
}
```
In this case, the `Limit` at Spark is removed wrongly.
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe e-mail: [email protected]
