Github user mgaido91 commented on a diff in the pull request:
https://github.com/apache/spark/pull/22029#discussion_r208835030
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/predicates.scala
---
@@ -240,20 +242,25 @@ case class In(value: Expression, list:
Seq[Expression]) extends Predicate {
lazy val inSetConvertible = list.forall(_.isInstanceOf[Literal])
private lazy val ordering =
TypeUtils.getInterpretedOrdering(value.dataType)
- override def nullable: Boolean = children.exists(_.nullable)
+ override def nullable: Boolean = value.dataType match {
+ case _: StructType if !SQLConf.get.inFalseForNullField =>
+ children.exists(_.nullable) ||
+
children.exists(_.dataType.asInstanceOf[StructType].exists(_.nullable))
--- End diff --
why do you think it is unsafe? We are doing it only if the `dataType` of
`value` match `StructType` and according also to what is enforced in the
`checkInputDataTypes`, therefore all the children must be `StructType`s.
I will add anyway a test case for the scenario you described, thanks.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
