Github user rdblue commented on a diff in the pull request:
https://github.com/apache/spark/pull/22009#discussion_r208637663
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/v2/DataSourceRDD.scala
---
@@ -51,18 +58,19 @@ class DataSourceRDD[T: ClassTag](
valuePrepared
}
- override def next(): T = {
+ override def next(): Any = {
if (!hasNext) {
throw new java.util.NoSuchElementException("End of stream")
}
valuePrepared = false
reader.get()
}
}
- new InterruptibleIterator(context, iter)
+ // TODO: get rid of this type hack.
+ new InterruptibleIterator(context,
iter.asInstanceOf[Iterator[InternalRow]])
}
override def getPreferredLocations(split: Partition): Seq[String] = {
-
split.asInstanceOf[DataSourceRDDPartition[T]].inputPartition.preferredLocations()
+
split.asInstanceOf[DataSourceRDDPartition].inputPartition.preferredLocations()
--- End diff --
Makes sense. If it's a bug, then the error message should indicate that
it's a bug to users instead of throwing a `ClassCastException`. Even if someone
goes to the code here, it there's no comment to indicate that it's a Spark bug.
I'd prefer something like this:
```scala
override def getPreferredLocations(split: Partition): Seq[String] = {
split match {
case dsp: DataSourceRDDPartition =>
dsp.inputPartition.preferredLocations
case _ => throw new SparkException(s"[BUG] Not a
DataSourceRDDPartition: $split")
}
}
```
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
