sigmod commented on code in PR #36177:
URL: https://github.com/apache/spark/pull/36177#discussion_r849692135
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/joins.scala:
##########
@@ -184,6 +202,23 @@ object EliminateOuterJoin extends Rule[LogicalPlan] with
PredicateHelper {
case a @ Aggregate(_, _, p @ Project(_, Join(_, right, RightOuter, _, _)))
if a.groupOnly && p.references.subsetOf(right.outputSet) =>
a.copy(child = p.copy(child = right))
+
+ case a @ Aggregate(_, aggExprs, Join(left, _, LeftOuter, _, _))
+ if a.references.subsetOf(left.outputSet) &&
+ !hasNotDuplicateAgnosticAggregateFunction(aggExprs) =>
Review Comment:
Nit: how about:
`allDuplicateAgnostic(aggExprs)`
Double negatives seem harder to read.
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/joins.scala:
##########
@@ -166,6 +176,14 @@ object EliminateOuterJoin extends Rule[LogicalPlan] with
PredicateHelper {
}
}
+ private def hasNotDuplicateAgnosticAggregateFunction(
+ aggregateExpressions: Seq[NamedExpression]): Boolean = {
+ aggregateExpressions.exists(_.exists {
+ case agg: AggregateFunction =>
!EliminateDistinct.isDuplicateAgnostic(agg)
Review Comment:
Does it make sense to all check `deterministic` in `isDuplicateAgnostic`?
E.g., the rewrite doesn't work for
`SELECT t1.c1, min(t1.c2 * rand()) FROM t1 LEFT JOIN t2 ON t1.c1 = t2.c1
GROUP BY t1.c1`
or
`SELECT t1.c1, min(udf(t1.c2)) FROM t1 LEFT JOIN t2 ON t1.c1 = t2.c1 GROUP
BY t1.c1`
where udf is non deterministic?
--
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]
For additional commands, e-mail: [email protected]
