wuchong commented on a change in pull request #8203:
[Flink-12208][table-planner-blink] Support translation from StreamExecSort /
TemporalSort / SortLimit/ Limit to StreamTransformation.
URL: https://github.com/apache/flink/pull/8203#discussion_r277192734
##########
File path:
flink-table/flink-table-planner-blink/src/main/scala/org/apache/flink/table/plan/nodes/physical/stream/StreamExecSortLimit.scala
##########
@@ -97,5 +114,105 @@ class StreamExecSortLimit(
}
}
- // TODO check `fetch` is not null in translateToPlan
+ //~ ExecNode methods
-----------------------------------------------------------
+
+ override def getInputNodes: util.List[ExecNode[StreamTableEnvironment, _]] =
{
+ List(getInput.asInstanceOf[ExecNode[StreamTableEnvironment, _]])
+ }
+
+ override protected def translateToPlanInternal(
+ tableEnv: StreamTableEnvironment): StreamTransformation[BaseRow] = {
+ if (fetch == null) {
+ throw new TableException(
+ "FETCH is missed, which on streaming table is not supported currently")
+ }
+ val inputRowTypeInfo =
FlinkTypeFactory.toInternalRowType(getInput.getRowType).toTypeInfo
+ val fieldCollations = sortCollation.getFieldCollations
+ val (sortFields, sortDirections, nullsIsLast) =
SortUtil.getKeysAndOrders(fieldCollations)
+ val sortKeySelector = KeySelectorUtil.getBaseRowSelector(sortFields,
inputRowTypeInfo)
+ val sortKeyType = sortKeySelector.getProducedType
+ val tableConfig = tableEnv.getConfig
+ val sortKeyComparator = ComparatorCodeGenerator.gen(tableConfig,
"StreamExecSortComparator",
+ sortFields.indices.toArray, sortKeyType.getInternalTypes,
sortDirections, nullsIsLast)
+ val generateRetraction = StreamExecRetractionRules.isAccRetract(this)
+ val cacheSize =
tableConfig.getConf.getLong(TableConfigOptions.SQL_EXEC_TOPN_CACHE_SIZE)
+ val minIdleStateRetentionTime = tableConfig.getMinIdleStateRetentionTime
+ val maxIdleStateRetentionTime = tableConfig.getMaxIdleStateRetentionTime
+
+ // rankStart begin with 1
+ val rankRange = new ConstantRankRange(limitStart + 1, limitEnd)
+ val rankType = RankType.ROW_NUMBER
+ val outputRankNumber = false
+
+ // Use RankFunction underlying StreamExecSortLimit
+ val processFunction = getStrategy(true) match {
+ case AppendFastStrategy =>
+ new AppendTopNFunction(
+ minIdleStateRetentionTime,
+ maxIdleStateRetentionTime,
+ inputRowTypeInfo,
+ sortKeyComparator,
+ sortKeySelector,
+ rankType,
+ rankRange,
+ generateRetraction,
+ outputRankNumber,
+ cacheSize)
+
+ case UpdateFastStrategy(primaryKeys) =>
+ val rowKeySelector = KeySelectorUtil.getBaseRowSelector(primaryKeys,
inputRowTypeInfo)
+ new FastTopNFunction(
+ minIdleStateRetentionTime,
+ maxIdleStateRetentionTime,
+ inputRowTypeInfo,
+ rowKeySelector,
+ sortKeyComparator,
+ sortKeySelector,
+ rankType,
+ rankRange,
+ generateRetraction,
+ outputRankNumber,
+ cacheSize)
+
+ // TODO Use UnaryUpdateTopNFunction after SortedMapState is merged
+ case RetractStrategy | UnaryUpdateStrategy(_) =>
+ val equaliserCodeGen = new
EqualiserCodeGenerator(inputRowTypeInfo.getInternalTypes)
+ val generatedEqualiser =
equaliserCodeGen.generateRecordEqualiser("RankValueEqualiser")
+ new RetractTopNFunction(
+ minIdleStateRetentionTime,
+ maxIdleStateRetentionTime,
+ inputRowTypeInfo,
+ sortKeyComparator,
+ sortKeySelector,
+ rankType,
+ rankRange,
+ generatedEqualiser,
+ generateRetraction,
+ outputRankNumber)
+ }
+ val operator = new KeyedProcessOperator(processFunction)
+ processFunction.setKeyContext(operator)
+
+ val inputTransform = getInputNodes.get(0).translateToPlan(tableEnv)
+ .asInstanceOf[StreamTransformation[BaseRow]]
+ val outputRowTypeInfo =
FlinkTypeFactory.toInternalRowType(getRowType).toTypeInfo
+ val ret = new OneInputTransformation(
+ inputTransform,
+ getOperatorName,
+ operator,
+ outputRowTypeInfo,
+ inputTransform.getParallelism)
Review comment:
Force parallelism to 1? As there is no partition key here.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
