Github user fhueske commented on a diff in the pull request:
https://github.com/apache/flink/pull/2792#discussion_r88657597
--- Diff:
flink-libraries/flink-table/src/main/scala/org/apache/flink/api/table/plan/nodes/datastream/DataStreamAggregate.scala
---
@@ -135,50 +128,123 @@ class DataStreamAggregate(
namedProperties)
val prepareOpName = s"prepare select: ($aggString)"
- val mappedInput = inputDS
- .map(aggregateResult._1)
- .name(prepareOpName)
-
- val groupReduceFunction = aggregateResult._2
- val rowTypeInfo = new RowTypeInfo(fieldTypes)
-
- val result = {
- // grouped / keyed aggregation
- if (groupingKeys.length > 0) {
- val aggOpName = s"groupBy: (${groupingToString(inputType,
grouping)}), " +
- s"window: ($window), " +
- s"select: ($aggString)"
- val aggregateFunction =
- createWindowAggregationFunction(window, namedProperties,
groupReduceFunction)
-
- val keyedStream = mappedInput.keyBy(groupingKeys: _*)
-
- val windowedStream = createKeyedWindowedStream(window, keyedStream)
- .asInstanceOf[WindowedStream[Row, Tuple, DataStreamWindow]]
-
- windowedStream
- .apply(aggregateFunction)
- .returns(rowTypeInfo)
- .name(aggOpName)
- .asInstanceOf[DataStream[Any]]
+ val keyedAggOpName = s"groupBy: (${groupingToString(inputType,
grouping)}), " +
+ s"window: ($window), " +
+ s"select: ($aggString)"
+ val nonKeyedAggOpName = s"window: ($window), select: ($aggString)"
+
+ val (aggFieldIndexes, aggregates) =
+ AggregateUtil.transformToAggregateFunctions(
+ namedAggregates.map(_.getKey), inputType, grouping.length)
+
+ val result: DataStream[Any] = {
+ // check whether all aggregates support partial aggregate
+ if (aggregates.forall(_.supportPartial)){
+ // do Incremental Aggregation
+ // add grouping fields, position keys in the input, and input type
+ val (mapFunction,
+ reduceFunction,
+ groupingOffsetMapping,
+ aggOffsetMapping,
+ intermediateRowArity) =
AggregateUtil.createOperatorFunctionsForIncrementalAggregates(
--- End diff --
Can we add a separate method to create the preparing `MapFunction` to
`AggregateUtil`?
This is code that is shared for all aggregations (batch, streaming),
(incremental, non-incremental), etc.
Would be nice to have that extracted and the mapper applied outside of this
large condition. Would be great if you could refactor the DataSetAggregate code
on the way as well.
---
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.
---
