cloud-fan commented on a change in pull request #25024:
URL: https://github.com/apache/spark/pull/25024#discussion_r449467645
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/aggregate/udaf.scala
##########
@@ -450,3 +454,63 @@ case class ScalaUDAF(
override def nodeName: String = udaf.getClass.getSimpleName
}
+
+case class ScalaAggregator[IN, BUF, OUT](
+ children: Seq[Expression],
+ agg: Aggregator[IN, BUF, OUT],
+ inputEncoderNR: ExpressionEncoder[IN],
+ nullable: Boolean = true,
+ isDeterministic: Boolean = true,
+ mutableAggBufferOffset: Int = 0,
+ inputAggBufferOffset: Int = 0)
+ extends TypedImperativeAggregate[BUF]
+ with NonSQLExpression
+ with UserDefinedExpression
+ with ImplicitCastInputTypes
+ with Logging {
+
+ private[this] lazy val inputEncoder = inputEncoderNR.resolveAndBind()
Review comment:
I think this is the problem. We shouldn't keep the encoder unresolved in
the query plan, and resolve it in the executor side. We can follow
`ResolveEncodersInUDF`: add a rule to resolve the encoders in `ScalaAggregator`
at driver side.
cc @viirya @dongjoon-hyun
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
