Github user fhueske commented on a diff in the pull request:
https://github.com/apache/flink/pull/3550#discussion_r106546112
--- Diff:
flink-libraries/flink-table/src/main/scala/org/apache/flink/table/plan/nodes/datastream/DataStreamOverAggregate.scala
---
@@ -119,6 +154,64 @@ class DataStreamOverAggregate(
}
+ def createTimeBoundedProcessingTimeOverWindow(
+ inputDS: DataStream[Row]): DataStream[Row] = {
+
+ val overWindow: Group = logicWindow.groups.get(0)
+ val partitionKeys: Array[Int] = overWindow.keys.toArray
+ val namedAggregates: Seq[CalcitePair[AggregateCall, String]] =
generateNamedAggregates
+
+ // final long time_boundary =
+ //
Long.parseLong(windowReference.getConstants().get(1).getValue().toString());
+ val index =
overWindow.lowerBound.getOffset.asInstanceOf[RexInputRef].getIndex
+ val count = input.getRowType().getFieldCount()
+ val lowerboundIndex = index - count
+ val time_boundary = logicWindow.constants.get(lowerboundIndex)
+ .getValue2.asInstanceOf[java.math.BigDecimal].longValue()
+
+
+ val (aggFields, aggregates) =
AggregateUtil.transformToAggregateFunctions(
+ namedAggregates.map(_.getKey),inputType, needRetraction = false)
+
+
+ // As we it is not possible to operate neither on sliding count neither
+ // on sliding time we need to manage the eviction of the events that
+ // expire ourselves based on the proctime (system time). Therefore the
+ // current system time is assign as the timestamp of the event to be
+ // recognize by the evictor
+
+ val inputDataStreamTimed = inputDS
+ .assignTimestampsAndWatermarks(new ProcTimeTimestampExtractor())
+
+ // get the output types
+ val rowTypeInfo =
FlinkTypeFactory.toInternalRowTypeInfo(getRowType).asInstanceOf[RowTypeInfo]
+
+ val result: DataStream[Row] =
+ if (partitionKeys.nonEmpty) {
+ inputDataStreamTimed.keyBy(partitionKeys:_*)
+ .window(GlobalWindows.create())
+ .trigger(CountTrigger.of(1))
+ .evictor(TimeEvictor.of(Time.milliseconds(time_boundary)))
+ .apply(new
DataStreamIncrementalAggregateWindowFunction[GlobalWindow]
--- End diff --
Let's use a process function. We have to change the code anyway once we
want to support `FOLLOWING`. Also a `ProcessFunction` does not need to
aggregate all rows for each row but remember the accumulators of the last row,
add the new row and retract the old ones.
---
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.
---
