[SPARK-21204] RuntimeException with Set and Case Class in Spark 2.1.1

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.1.1
• Fix Version/s: 2.3.0
• Component/s: Optimizer, SQL
• Assignee: L. C. Hsieh
• Reporter: Leo Romanovsky
• Created: 24/Jun/17 23:36
• Updated: 20/Aug/17 08:50
• Resolved: 06/Jul/17 17:08

Description

When attempting to produce a Dataset containing a Set, such as with:

dbData
  .groupBy("userId")
  .agg(functions.collect_set("friendId") as "friendIds")
  .as[(Int, Set[Int])]

An exception occurs. This can be avoided by casting to a Seq, but sometimes it makes more logical sense have a Set, especially when using the collect_set aggregation operation. Additionally, I am unable to write this Dataset to a Cassandra table containing a Set column without first converting to an RDD.

[error] Exception in thread "main" java.lang.UnsupportedOperationException: No Encoder found for Set[Int]
[error] - field (class: "scala.collection.immutable.Set", name: "_2")
[error] - root class: "scala.Tuple2"
[error] 	at org.apache.spark.sql.catalyst.ScalaReflection$.org$apache$spark$sql$catalyst$ScalaReflection$$serializerFor(ScalaReflection.scala:602)
[error] 	at org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$9.apply(ScalaReflection.scala:596)
[error] 	at org.apache.spark.sql.catalyst.ScalaReflection$$anonfun$9.apply(ScalaReflection.scala:587)
[error] 	at scala.collection.TraversableLike$$anonfun$flatMap$1.apply(TraversableLike.scala:252)
[error] 	at scala.collection.TraversableLike$$anonfun$flatMap$1.apply(TraversableLike.scala:252)
[error] 	at scala.collection.immutable.List.foreach(List.scala:381)
[error] 	at scala.collection.TraversableLike$class.flatMap(TraversableLike.scala:252)
[error] 	at scala.collection.immutable.List.flatMap(List.scala:344)
[error] 	at org.apache.spark.sql.catalyst.ScalaReflection$.org$apache$spark$sql$catalyst$ScalaReflection$$serializerFor(ScalaReflection.scala:587)
[error] 	at org.apache.spark.sql.catalyst.ScalaReflection$.serializerFor(ScalaReflection.scala:425)
[error] 	at org.apache.spark.sql.catalyst.encoders.ExpressionEncoder$.apply(ExpressionEncoder.scala:71)
[error] 	at org.apache.spark.sql.Encoders$.product(Encoders.scala:275)
[error] 	at org.apache.spark.sql.SQLImplicits.newProductEncoder(SQLImplicits.scala:49)

I think the resolution to this might be similar to adding the Map type - https://github.com/apache/spark/pull/16986

Issue Links

• is duplicated by: SPARK-17414 Set type is not supported for creating data frames
• links to: [Github] Pull Request #18416 (viirya)