cloud-fan commented on a change in pull request #31984:
URL: https://github.com/apache/spark/pull/31984#discussion_r607928080
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/dynamicpruning/PartitionPruning.scala
##########
@@ -247,13 +246,15 @@ object PartitionPruning extends Rule[LogicalPlan] with
PredicateHelper with Join
// otherwise the pruning will not trigger
var partScan = getPartitionTableScan(l, left)
if (partScan.isDefined && canPruneLeft(joinType) &&
- hasPartitionPruningFilter(right)) {
+ hasPartitionPruningFilter(right) &&
+ (canBroadcastBySize(right, conf) ||
hintToBroadcastRight(hint))) {
newLeft = insertPredicate(l, newLeft, r, right, rightKeys,
partScan.get,
canBuildBroadcastRight(joinType))
} else {
partScan = getPartitionTableScan(r, right)
if (partScan.isDefined && canPruneRight(joinType) &&
- hasPartitionPruningFilter(left) ) {
+ hasPartitionPruningFilter(left) &&
+ (canBroadcastBySize(left, conf) ||
hintToBroadcastLeft(hint))) {
Review comment:
Let me revert it first. "broadcast hint" is still a valid case but that
might be too minor. We need to think more about how to improve the cost model
of DPP.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
