Github user cloud-fan commented on a diff in the pull request:
https://github.com/apache/spark/pull/16404#discussion_r93921170
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/Analyzer.scala
---
@@ -1918,28 +1918,37 @@ class Analyzer(
case p: Project => p
case f: Filter => f
+ case a: Aggregate if a.groupingExpressions.exists(!_.deterministic)
=>
+ val nondeterToAttr = getNondeterToAttr(a.groupingExpressions)
+ val newChild = Project(a.child.output ++ nondeterToAttr.values,
a.child)
+ a.transformExpressions { case e =>
+ nondeterToAttr.get(e).map(_.toAttribute).getOrElse(e)
+ }.copy(child = newChild)
+
// todo: It's hard to write a general rule to pull out
nondeterministic expressions
// from LogicalPlan, currently we only do it for UnaryNode which has
same output
// schema with its child.
case p: UnaryNode if p.output == p.child.output &&
p.expressions.exists(!_.deterministic) =>
- val nondeterministicExprs =
p.expressions.filterNot(_.deterministic).flatMap { expr =>
- val leafNondeterministic = expr.collect {
- case n: Nondeterministic => n
- }
- leafNondeterministic.map { e =>
- val ne = e match {
- case n: NamedExpression => n
- case _ => Alias(e, "_nondeterministic")(isGenerated = true)
- }
- new TreeNodeRef(e) -> ne
- }
- }.toMap
+ val nondeterToAttr = getNondeterToAttr(p.expressions)
val newPlan = p.transformExpressions { case e =>
- nondeterministicExprs.get(new
TreeNodeRef(e)).map(_.toAttribute).getOrElse(e)
+ nondeterToAttr.get(e).map(_.toAttribute).getOrElse(e)
}
- val newChild = Project(p.child.output ++
nondeterministicExprs.values, p.child)
+ val newChild = Project(p.child.output ++ nondeterToAttr.values,
p.child)
Project(p.output, newPlan.withNewChildren(newChild :: Nil))
}
+
+ private def getNondeterToAttr(exprs: Seq[Expression]): Map[Expression,
NamedExpression] = {
+ exprs.filterNot(_.deterministic).flatMap { expr =>
+ val leafNondeterministic = expr.collect { case n: Nondeterministic
=> n }
--- End diff --
1. this is same with the previous behavior
2. according to the variable name: `leafNondeterministic`, it seems
reasonable to collect `Nondeterministic` here.
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
