sigmod commented on code in PR #35864:
URL: https://github.com/apache/spark/pull/35864#discussion_r842226772
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/NestedColumnAliasing.scala:
##########
@@ -312,6 +312,32 @@ object NestedColumnAliasing {
}
}
+object GeneratorUnrequiredChildrenPruning {
+ def unapply(plan: LogicalPlan): Option[LogicalPlan] = plan match {
+ case p@Project(_, g: Generate) =>
+ val requiredAttrs = p.references ++ g.generator.references
+ var pruned = false
+ val newChild = if (!g.child.outputSet.subsetOf(requiredAttrs)) {
+ pruned = true
+ Project(g.child.output.filter(requiredAttrs.contains), g.child)
+ } else {
+ g.child
+ }
+ val unrequired = g.generator.references -- p.references
+ val unrequiredIndices = newChild.output.zipWithIndex.filter(t =>
unrequired.contains(t._1))
+ .map(_._2)
+ if (unrequiredIndices.toSet != g.unrequiredChildIndex.toSet) {
+ pruned = true
+ }
+ if (pruned) {
+ Some(p.copy(child = g.copy(child = newChild, unrequiredChildIndex =
unrequiredIndices)))
+ } else {
+ None
+ }
Review Comment:
>> The previous implementation is not incorrect, but it is redundant IMO.
I think the meaning of `prunedChild` is a bit vague, particularly what
`allReferences` mean.
Therefore, "but it is redundant" is based on reading the content of
`prunedChild`, rather than per the contract of `prunedChild`.
A cleaner version of it imo looks like the following:
```
def prunedChild(c: LogicalPlan, requiredAttributes: AttributeSet):
LogicalPlan =
require(requiredAttributes.subsetOf(c.outputSet))
if (c.outputSet != requiredAttributes) {
Project(requiredAttributes, c)
} else {
c
}
```
I'm fine with go with your current code. We can update `prunedChild` in a
followup PR -- maybe it'll fail some tests.
--
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]
