Github user chenghao-intel commented on a diff in the pull request:
https://github.com/apache/spark/pull/16245#discussion_r97211716
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/expressions.scala
---
@@ -514,6 +514,34 @@ case class OptimizeCodegen(conf: CatalystConf) extends
Rule[LogicalPlan] {
/**
+ * Reorders the predicates in `Filter` so more expensive expressions like
UDF can evaluate later.
+ */
+object ReorderPredicatesInFilter extends Rule[LogicalPlan] with
PredicateHelper {
+ def apply(plan: LogicalPlan): LogicalPlan = plan transform {
+ case f @ Filter(pred, child) =>
+ // Extracts deterministic suffix expressions from Filter predicate.
+ val expressions = splitConjunctivePredicates(pred)
+ // The beginning index of the deterministic suffix expressions.
+ var splitIndex = -1
+ (expressions.length - 1 to 0 by -1).foreach { idx =>
+ if (splitIndex == -1 && !expressions(idx).deterministic) {
+ splitIndex = idx + 1
+ }
+ }
+ if (splitIndex == expressions.length) {
+ // All expressions are non-deterministic, no reordering.
+ f
+ } else {
+ val (nonDeterminstics, deterministicExprs) =
expressions.splitAt(splitIndex)
--- End diff --
Hmm, actually that's what I mean, probably some confusing with
`non-deterministic` with `non-foldable`? I think we can skip them both in a
short cut evaluation. as those expressions are not `stateful`(unfortunately,
Spark SQL expression doesn't have the concept of `stateful`), so skip the
evaluation of them are harmless, and this is exactly the short cut logic of
expression `AND`.
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
