Github user marmbrus commented on a diff in the pull request:
https://github.com/apache/spark/pull/482#discussion_r12036349
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/Optimizer.scala
---
@@ -91,10 +91,53 @@ object ColumnPruning extends Rule[LogicalPlan] {
*/
object ConstantFolding extends Rule[LogicalPlan] {
def apply(plan: LogicalPlan): LogicalPlan = plan transform {
- case q: LogicalPlan => q transformExpressionsDown {
+ case q: LogicalPlan => q transformExpressionsUp {
// Skip redundant folding of literals.
case l: Literal => l
+ // if it's foldable
case e if e.foldable => Literal(e.eval(null), e.dataType)
+ case e @ Count(Literal(null, _)) => Literal(null, e.dataType)
+ case e @ Sum(Literal(null, _)) => Literal(null, e.dataType)
+ case e @ Average(Literal(null, _)) => Literal(null, e.dataType)
+ case e @ IsNull(Literal(null, _)) => Literal(true, BooleanType)
+ case e @ IsNull(Literal(_, _)) => Literal(false, BooleanType)
+ case e @ IsNull(c @ Rand) => Literal(false, BooleanType)
+ case e @ IsNotNull(Literal(null, _)) => Literal(false, BooleanType)
+ case e @ IsNotNull(Literal(_, _)) => Literal(true, BooleanType)
+ case e @ IsNotNull(c @ Rand) => Literal(true, BooleanType)
+ case e @ GetItem(Literal(null, _), _) => Literal(null, e.dataType)
+ case e @ GetItem(_, Literal(null, _)) => Literal(null, e.dataType)
+ case e @ GetField(Literal(null, _), _) => Literal(null, e.dataType)
+ case e @ Coalesce(children) => {
+ val newChildren = children.filter(c => c match {
+ case Literal(null, _) => false
+ case _ => true
+ })
+ if(newChildren.length == null) {
+ Literal(null, e.dataType)
+ } else if(newChildren.length == children.length){
+ e
+ } else {
+ Coalesce(newChildren)
+ }
+ }
+ case e @ If(Literal(v, _), trueValue, falseValue) => if(v == true)
trueValue else falseValue
+ case e @ In(Literal(v, _), list) if(list.exists(c => c match {
+ case Literal(candidate, _) if(candidate == v) => true
+ case _ => false
+ })) => Literal(true, BooleanType)
+
+ case e @ SortOrder(_, _) => e
+ // put exceptional cases(Unary & Binary Expression) before here.
+ case e: UnaryExpression => e.child match {
--- End diff --
I think its reasonable to enforce this nullability semantic on unary and
binary nodes, but we should add something to their scaladoc. Maybe also just
make `SortOrder` not a UnaryNode.
---
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.
---
