Github user fhueske commented on a diff in the pull request:
https://github.com/apache/flink/pull/5555#discussion_r182479163
--- Diff:
flink-libraries/flink-table/src/main/scala/org/apache/flink/table/codegen/AggregationCodeGenerator.scala
---
@@ -93,6 +97,8 @@ class AggregationCodeGenerator(
aggregates: Array[AggregateFunction[_ <: Any, _ <: Any]],
aggFields: Array[Array[Int]],
aggMapping: Array[Int],
+ distinctAggs: Array[Seq[DataViewSpec[_]]],
+ isStateBackedDataViews: Boolean,
--- End diff --
We should add a constructor check for `if (partialResults &&
isStateBackedDataViews)` and throw an exception if `true`. `partialResults`
means that the `Row` with the accumulators has to be emitted which won't work
well for state-backed distinct maps that are probably too big.
---
