Github user fhueske commented on a diff in the pull request:
https://github.com/apache/flink/pull/4355#discussion_r130126823
--- Diff:
flink-libraries/flink-table/src/test/java/org/apache/flink/table/runtime/utils/JavaUserDefinedAggFunctions.java
---
@@ -135,4 +138,172 @@ public void retract(WeightedAvgAccum accumulator, int
iValue, int iWeight) {
accumulator.count -= iWeight;
}
}
+
+ /**
+ * CountDistinct accumulator.
+ */
+ public static class CountDistinctAccum {
+ public MapView<String, Integer> map;
+ public long count;
+ }
+
+ /**
+ * CountDistinct aggregate.
+ */
+ public static class CountDistinct extends AggregateFunction<Long,
CountDistinctAccum> {
--- End diff --
I don't think we should implement `COUNT DISTINCT` as a special
`AggregateFunction`. At least not in the long term.
I think it would be better to handle this inside of the
`GeneratedAggregations` and only accumulate and retract distinct values from
user-defined aggregate functions. With this approach, any aggregation function
can be used with `DISTINCT` and the state for distinction can also be shared
across multiple aggregation functions. This is also the approach that has been
started in PR #3783.
For now this is fine, but in the long run we should go for something like
PR #3783 (which also requires the `GeneratedAggregations.initialize()` method.)
---
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.
---
