Search in sources :

Example 1 with MiniBatchGroupAggFunction

use of org.apache.flink.table.runtime.operators.aggregate.MiniBatchGroupAggFunction in project flink by apache.

the class StreamExecGroupAggregate method translateToPlanInternal.

@SuppressWarnings("unchecked")
@Override
protected Transformation<RowData> translateToPlanInternal(PlannerBase planner, ExecNodeConfig config) {
    if (grouping.length > 0 && config.getStateRetentionTime() < 0) {
        LOG.warn("No state retention interval configured for a query which accumulates state. " + "Please provide a query configuration with valid retention interval to prevent excessive " + "state size. You may specify a retention time of 0 to not clean up the state.");
    }
    final ExecEdge inputEdge = getInputEdges().get(0);
    final Transformation<RowData> inputTransform = (Transformation<RowData>) inputEdge.translateToPlan(planner);
    final RowType inputRowType = (RowType) inputEdge.getOutputType();
    final AggsHandlerCodeGenerator generator = new AggsHandlerCodeGenerator(new CodeGeneratorContext(config.getTableConfig()), planner.getRelBuilder(), JavaScalaConversionUtil.toScala(inputRowType.getChildren()), // TODO: but other operators do not copy this input field.....
    true).needAccumulate();
    if (needRetraction) {
        generator.needRetract();
    }
    final AggregateInfoList aggInfoList = AggregateUtil.transformToStreamAggregateInfoList(inputRowType, JavaScalaConversionUtil.toScala(Arrays.asList(aggCalls)), aggCallNeedRetractions, needRetraction, true, true);
    final GeneratedAggsHandleFunction aggsHandler = generator.generateAggsHandler("GroupAggsHandler", aggInfoList);
    final LogicalType[] accTypes = Arrays.stream(aggInfoList.getAccTypes()).map(LogicalTypeDataTypeConverter::fromDataTypeToLogicalType).toArray(LogicalType[]::new);
    final LogicalType[] aggValueTypes = Arrays.stream(aggInfoList.getActualValueTypes()).map(LogicalTypeDataTypeConverter::fromDataTypeToLogicalType).toArray(LogicalType[]::new);
    final GeneratedRecordEqualiser recordEqualiser = new EqualiserCodeGenerator(aggValueTypes).generateRecordEqualiser("GroupAggValueEqualiser");
    final int inputCountIndex = aggInfoList.getIndexOfCountStar();
    final boolean isMiniBatchEnabled = config.get(ExecutionConfigOptions.TABLE_EXEC_MINIBATCH_ENABLED);
    final OneInputStreamOperator<RowData, RowData> operator;
    if (isMiniBatchEnabled) {
        MiniBatchGroupAggFunction aggFunction = new MiniBatchGroupAggFunction(aggsHandler, recordEqualiser, accTypes, inputRowType, inputCountIndex, generateUpdateBefore, config.getStateRetentionTime());
        operator = new KeyedMapBundleOperator<>(aggFunction, AggregateUtil.createMiniBatchTrigger(config));
    } else {
        GroupAggFunction aggFunction = new GroupAggFunction(aggsHandler, recordEqualiser, accTypes, inputCountIndex, generateUpdateBefore, config.getStateRetentionTime());
        operator = new KeyedProcessOperator<>(aggFunction);
    }
    // partitioned aggregation
    final OneInputTransformation<RowData, RowData> transform = ExecNodeUtil.createOneInputTransformation(inputTransform, createTransformationMeta(GROUP_AGGREGATE_TRANSFORMATION, config), operator, InternalTypeInfo.of(getOutputType()), inputTransform.getParallelism());
    // set KeyType and Selector for state
    final RowDataKeySelector selector = KeySelectorUtil.getRowDataSelector(grouping, InternalTypeInfo.of(inputRowType));
    transform.setStateKeySelector(selector);
    transform.setStateKeyType(selector.getProducedType());
    return transform;
}
Also used : OneInputTransformation(org.apache.flink.streaming.api.transformations.OneInputTransformation) Transformation(org.apache.flink.api.dag.Transformation) AggregateInfoList(org.apache.flink.table.planner.plan.utils.AggregateInfoList) ExecEdge(org.apache.flink.table.planner.plan.nodes.exec.ExecEdge) CodeGeneratorContext(org.apache.flink.table.planner.codegen.CodeGeneratorContext) GroupAggFunction(org.apache.flink.table.runtime.operators.aggregate.GroupAggFunction) MiniBatchGroupAggFunction(org.apache.flink.table.runtime.operators.aggregate.MiniBatchGroupAggFunction) MiniBatchGroupAggFunction(org.apache.flink.table.runtime.operators.aggregate.MiniBatchGroupAggFunction) RowType(org.apache.flink.table.types.logical.RowType) AggsHandlerCodeGenerator(org.apache.flink.table.planner.codegen.agg.AggsHandlerCodeGenerator) LogicalType(org.apache.flink.table.types.logical.LogicalType) GeneratedAggsHandleFunction(org.apache.flink.table.runtime.generated.GeneratedAggsHandleFunction) EqualiserCodeGenerator(org.apache.flink.table.planner.codegen.EqualiserCodeGenerator) GeneratedRecordEqualiser(org.apache.flink.table.runtime.generated.GeneratedRecordEqualiser) RowData(org.apache.flink.table.data.RowData) RowDataKeySelector(org.apache.flink.table.runtime.keyselector.RowDataKeySelector)

Aggregations

Transformation (org.apache.flink.api.dag.Transformation)1 OneInputTransformation (org.apache.flink.streaming.api.transformations.OneInputTransformation)1 RowData (org.apache.flink.table.data.RowData)1 CodeGeneratorContext (org.apache.flink.table.planner.codegen.CodeGeneratorContext)1 EqualiserCodeGenerator (org.apache.flink.table.planner.codegen.EqualiserCodeGenerator)1 AggsHandlerCodeGenerator (org.apache.flink.table.planner.codegen.agg.AggsHandlerCodeGenerator)1 ExecEdge (org.apache.flink.table.planner.plan.nodes.exec.ExecEdge)1 AggregateInfoList (org.apache.flink.table.planner.plan.utils.AggregateInfoList)1 GeneratedAggsHandleFunction (org.apache.flink.table.runtime.generated.GeneratedAggsHandleFunction)1 GeneratedRecordEqualiser (org.apache.flink.table.runtime.generated.GeneratedRecordEqualiser)1 RowDataKeySelector (org.apache.flink.table.runtime.keyselector.RowDataKeySelector)1 GroupAggFunction (org.apache.flink.table.runtime.operators.aggregate.GroupAggFunction)1 MiniBatchGroupAggFunction (org.apache.flink.table.runtime.operators.aggregate.MiniBatchGroupAggFunction)1 LogicalType (org.apache.flink.table.types.logical.LogicalType)1 RowType (org.apache.flink.table.types.logical.RowType)1