ulysses-you commented on a change in pull request #30368:
URL: https://github.com/apache/spark/pull/30368#discussion_r524888701
##########
File path:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala
##########
@@ -1452,11 +1452,27 @@ object PushPredicateThroughJoin extends
Rule[LogicalPlan] with PredicateHelper {
}
/**
- * Combines two adjacent [[Limit]] operators into one, merging the
- * expressions into one single expression.
+ * 1. Eliminate [[Limit]] operators if it's child max row <= limit.
+ * 2. Combines two adjacent [[Limit]] operators into one, merging the
+ * expressions into one single expression.
*/
-object CombineLimits extends Rule[LogicalPlan] {
- def apply(plan: LogicalPlan): LogicalPlan = plan transform {
+object EliminateLimits extends Rule[LogicalPlan] {
+ private def canEliminate(limitExpr: Expression, child: LogicalPlan): Boolean
= {
+ // We skip such case that Sort is after Limit since
+ // SparkStrategies will convert them to TakeOrderedAndProjectExec
+ val skipEliminate = child match {
+ case Sort(_, true, _) => true
Review comment:
I'm ok to skip eliminate this case if we afraid the potential
regression. But the benchmark show it's no harmful, I also believe that, for
large data size range shuffle sort is faster than singe partition sort.
And to clarify, we can choose the follow option
1. skip eliminate limit if child is sort, to avoid potential regression
about `TakeOrderedAndProjectExec`.
2. eliminate all matched limit, since benchmark support it.
3. add a config to control the eliminate limit ? e.g.
`spark.sql.optimizer.eliminateLimits.enable`.
----------------------------------------------------------------
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:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
