cloud-fan commented on a change in pull request #24068: [SPARK-27105][SQL]
Optimize away exponential complexity in ORC predicate conversion
URL: https://github.com/apache/spark/pull/24068#discussion_r282045908
##########
File path:
sql/core/v2.3.4/src/main/scala/org/apache/spark/sql/execution/datasources/orc/OrcFilters.scala
##########
@@ -153,92 +188,140 @@ private[sql] object OrcFilters extends OrcFiltersBase {
// Pushing one side of AND down is only safe to do at the top level or
in the child
// AND before hitting NOT or OR conditions, and in this case, the
unsupported predicate
// can be safely removed.
- val leftBuilderOption =
- createBuilder(dataTypeMap, left, newBuilder,
canPartialPushDownConjuncts)
- val rightBuilderOption =
- createBuilder(dataTypeMap, right, newBuilder,
canPartialPushDownConjuncts)
- (leftBuilderOption, rightBuilderOption) match {
- case (Some(_), Some(_)) =>
- for {
- lhs <- createBuilder(dataTypeMap, left,
- builder.startAnd(), canPartialPushDownConjuncts)
- rhs <- createBuilder(dataTypeMap, right, lhs,
canPartialPushDownConjuncts)
- } yield rhs.end()
-
- case (Some(_), None) if canPartialPushDownConjuncts =>
- createBuilder(dataTypeMap, left, builder,
canPartialPushDownConjuncts)
-
- case (None, Some(_)) if canPartialPushDownConjuncts =>
- createBuilder(dataTypeMap, right, builder,
canPartialPushDownConjuncts)
-
+ val lhs =
+ trimNonConvertibleSubtreesImpl(dataTypeMap, left,
canPartialPushDownConjuncts = true)
+ val rhs =
+ trimNonConvertibleSubtreesImpl(dataTypeMap, right,
canPartialPushDownConjuncts = true)
+ (lhs, rhs) match {
+ case (Some(l), Some(r)) => Some(And(l, r))
+ case (Some(_), None) if canPartialPushDownConjuncts => lhs
+ case (None, Some(_)) if canPartialPushDownConjuncts => rhs
case _ => None
}
case Or(left, right) =>
for {
- _ <- createBuilder(dataTypeMap, left, newBuilder,
canPartialPushDownConjuncts = false)
- _ <- createBuilder(dataTypeMap, right, newBuilder,
canPartialPushDownConjuncts = false)
- lhs <- createBuilder(dataTypeMap, left,
- builder.startOr(), canPartialPushDownConjuncts = false)
- rhs <- createBuilder(dataTypeMap, right, lhs,
canPartialPushDownConjuncts = false)
- } yield rhs.end()
+ lhs <-
+ trimNonConvertibleSubtreesImpl(dataTypeMap, left,
canPartialPushDownConjuncts = false)
+ rhs <-
+ trimNonConvertibleSubtreesImpl(dataTypeMap, right,
canPartialPushDownConjuncts = false)
+ } yield Or(lhs, rhs)
case Not(child) =>
- for {
- _ <- createBuilder(dataTypeMap, child, newBuilder,
canPartialPushDownConjuncts = false)
- negate <- createBuilder(dataTypeMap,
- child, builder.startNot(), canPartialPushDownConjuncts = false)
- } yield negate.end()
+ trimNonConvertibleSubtreesImpl(dataTypeMap, child,
canPartialPushDownConjuncts = false)
+ .map(Not)
// NOTE: For all case branches dealing with leaf predicates below, the
additional `startAnd()`
Review comment:
need to update comment?
----------------------------------------------------------------
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]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
