holdenk commented on code in PR #46143:
URL: https://github.com/apache/spark/pull/46143#discussion_r2529175212
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala:
##########
@@ -2053,17 +2053,117 @@ object PushDownPredicates extends Rule[LogicalPlan] {
object PushPredicateThroughNonJoin extends Rule[LogicalPlan] with
PredicateHelper {
def apply(plan: LogicalPlan): LogicalPlan = plan transform applyLocally
+ // Match the column ordering. It's not something we guarantee but seems like
it
+ // could be useful especially for folks writing out to "raw" files.
+ private def matchColumnOrdering(original: Seq[NamedExpression], updated:
Seq[NamedExpression]):
+ Seq[NamedExpression] = {
+ val nameToIndex = original.map(_.name).zipWithIndex.toMap
+ val response = updated.toArray
+ updated.foreach { e =>
+ val idx = nameToIndex(e.name)
+ response(idx) = e
+ }
+ response.toSeq
+ }
+
val applyLocally: PartialFunction[LogicalPlan, LogicalPlan] = {
// SPARK-13473: We can't push the predicate down when the underlying
projection output non-
// deterministic field(s). Non-deterministic expressions are essentially
stateful. This
// implies that, for a given input row, the output are determined by the
expression's initial
// state and all the input rows processed before. In another word, the
order of input rows
// matters for non-deterministic expressions, while pushing down
predicates changes the order.
// This also applies to Aggregate.
- case Filter(condition, project @ Project(fields, grandChild))
+ case f @ Filter(condition, project @ Project(fields, grandChild))
if fields.forall(_.deterministic) && canPushThroughCondition(grandChild,
condition) =>
val aliasMap = getAliasMap(project)
- project.copy(child = Filter(replaceAlias(condition, aliasMap),
grandChild))
Review Comment:
Ah yes I remember now, we still would probably want to push a filter + a
projection through in the case where we have an filter condition referencing an
expensive alias because we could potentially eliminate some additional
projections being computed by filtering on it before computing the rest of
aliases. e.g. `filter(expensive_regex_ref1) -> project(expensive_regex_ref1,
expensive_regex_ref2)` we want to get `project(expensive_regex_ref2) ->
filter(expensive_regex_ref1) -> project(expensive_regex_ref1)`
--
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]
