Github user fhueske commented on a diff in the pull request:
https://github.com/apache/flink/pull/3386#discussion_r107719825
--- Diff:
flink-libraries/flink-table/src/main/scala/org/apache/flink/table/runtime/aggregate/UnboundedEventTimeOverProcessFunction.scala
---
@@ -0,0 +1,200 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package org.apache.flink.table.runtime.aggregate
+
+import java.util.{ArrayList, LinkedList, List => JList}
+
+import org.apache.flink.api.common.typeinfo.{BasicTypeInfo,
TypeInformation}
+import org.apache.flink.configuration.Configuration
+import org.apache.flink.types.Row
+import org.apache.flink.streaming.api.functions.ProcessFunction
+import org.apache.flink.util.{Collector, Preconditions}
+import org.apache.flink.api.common.state._
+import org.apache.flink.api.java.typeutils.{ListTypeInfo}
+import org.apache.flink.streaming.api.operators.TimestampedCollector
+import org.apache.flink.table.functions.{Accumulator, AggregateFunction}
+
+
+/**
+ * A ProcessFunction to support unbounded event-time over-window
+ *
+ * @param aggregates the aggregate functions
+ * @param aggFields the filed index which the aggregate functions use
+ * @param forwardedFieldCount the input fields count
+ * @param intermediateType the intermediate row tye which the state saved
+ * @param inputType the input row tye which the state saved
+ *
+ */
+class UnboundedEventTimeOverProcessFunction(
+ private val aggregates: Array[AggregateFunction[_]],
+ private val aggFields: Array[Int],
+ private val forwardedFieldCount: Int,
+ private val intermediateType: TypeInformation[Row],
+ private val inputType: TypeInformation[Row])
+ extends ProcessFunction[Row, Row]{
+
+ Preconditions.checkNotNull(aggregates)
+ Preconditions.checkNotNull(aggFields)
+ Preconditions.checkArgument(aggregates.length == aggFields.length)
+
+ private var output: Row = _
+ private var accumulatorState: ValueState[Row] = _
+ private var rowMapState: MapState[Long, JList[Row]] = _
+ private var sortList: LinkedList[Long] = _
+
+
+ override def open(config: Configuration) {
+ output = new Row(forwardedFieldCount + aggregates.length)
+ sortList = new LinkedList[Long]()
+
+ val stateDescriptor: ValueStateDescriptor[Row] =
+ new ValueStateDescriptor[Row]("accumulatorstate", intermediateType)
+ accumulatorState = getRuntimeContext.getState[Row](stateDescriptor)
+
+ val rowListTypeInfo: TypeInformation[JList[Row]] =
+ new
ListTypeInfo[Row](inputType).asInstanceOf[TypeInformation[JList[Row]]]
+ val mapStateDescriptor: MapStateDescriptor[Long, JList[Row]] =
+ new MapStateDescriptor[Long, JList[Row]]("rowmapstate",
+ BasicTypeInfo.LONG_TYPE_INFO.asInstanceOf[TypeInformation[Long]],
rowListTypeInfo)
+ rowMapState = getRuntimeContext.getMapState(mapStateDescriptor)
+ }
+
+ /**
+ * Process one element from the input stream, not emit the output
+ *
+ * @param input The input value.
+ * @param ctx The ctx to register timer or get current time
+ * @param out The collector for returning result values.
+ *
+ */
+ override def processElement(
+ input: Row,
+ ctx: ProcessFunction[Row, Row]#Context,
+ out: Collector[Row]): Unit = {
+
+ // discard later record
+ if (ctx.timestamp() >= ctx.timerService().currentWatermark()) {
+ // ensure every key just register on timer
+
ctx.timerService.registerEventTimeTimer(ctx.timerService.currentWatermark + 1)
+
+ val rowList =
+ if (rowMapState.contains(ctx.timestamp))
rowMapState.get(ctx.timestamp)
+ else new ArrayList[Row]()
+ rowList.add(input)
+ rowMapState.put(ctx.timestamp, rowList)
+ }
+ }
+
+ /**
+ * Called when a timer set fires, sort current records according the
timestamp
+ * and emit the output
+ *
+ * @param timestamp The timestamp of the firing timer.
+ * @param ctx The ctx to register timer or get current time
+ * @param out The collector for returning result values.
+ */
+ override def onTimer(
+ timestamp: Long,
+ ctx: ProcessFunction[Row, Row]#OnTimerContext,
+ out: Collector[Row]): Unit = {
+
+
Preconditions.checkArgument(out.isInstanceOf[TimestampedCollector[Row]])
+ val collector = out.asInstanceOf[TimestampedCollector[Row]]
+
+ val mapIter = rowMapState.iterator
+ if (mapIter.hasNext) {
+ val curWatermark = ctx.timerService.currentWatermark
+ var i = 0
+
+ // sort record according timestamp
+ do {
+ val entry = mapIter.next
+ if (entry.getKey <= curWatermark) {
--- End diff --
we could set an `earlyRows` flag to `true` and avoid to get an iterator on
the Map state to figure out that we have rows left.
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
