Github user JoshRosen commented on a diff in the pull request:

    https://github.com/apache/spark/pull/7179#discussion_r33749911
  
    --- Diff: 
sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/expressions/CodeGenerationSuite.scala
 ---
    @@ -42,4 +47,47 @@ class CodeGenerationSuite extends SparkFunSuite {
     
         futures.foreach(Await.result(_, 10.seconds))
       }
    +
    +  // Test GenerateOrdering for all common types. For each type, we 
construct random input rows that
    +  // contain two columns of that type, then for pairs of 
randomly-generated rows we check that
    +  // GenerateOrdering agrees with RowOrdering.
    +  (DataTypeTestUtils.atomicTypes ++ Set(NullType)).foreach { dataType =>
    +    test(s"GenerateOrdering with $dataType") {
    +      val rowOrdering = RowOrdering.forSchema(Seq(dataType, dataType))
    +      val genOrdering = GenerateOrdering.generate(
    +        BoundReference(0, dataType, nullable = true).asc ::
    +        BoundReference(1, dataType, nullable = true).asc :: Nil)
    +      val rowType = StructType(
    +        StructField("a", dataType, nullable = true) ::
    +        StructField("b", dataType, nullable = true) :: Nil)
    +      val toCatalyst = 
CatalystTypeConverters.createToCatalystConverter(rowType)
    +      // Sort ordering is not defined for NaN, so skip any random inputs 
that contain it:
    +      def isIncomparable(v: Any): Boolean = v match {
    --- End diff --
    
    While working on this, I discovered that `RowOrdering` and 
`GenerateOrdering` disagree for inputs containing NaN. This isn't a bug per-se, 
since many systems have undefined behavior when sorting on NaN.  For this 
reason, I think that some databases prohibit NaN and Infinity from being used.


---
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.
---

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to