Examples with Transformation org.apache.flink.api.dag.Transformation used on opensource projects

Example 6 with Transformation

use of org.apache.flink.api.dag.Transformation in project flink by apache.

the class UpsertKafkaDynamicTableFactoryTest method assertKafkaSource.

private void assertKafkaSource(ScanTableSource.ScanRuntimeProvider provider) {
    assertThat(provider, instanceOf(DataStreamScanProvider.class));
    final DataStreamScanProvider dataStreamScanProvider = (DataStreamScanProvider) provider;
    final Transformation<RowData> transformation = dataStreamScanProvider.produceDataStream(n -> Optional.empty(), StreamExecutionEnvironment.createLocalEnvironment()).getTransformation();
    assertThat(transformation, instanceOf(SourceTransformation.class));
    SourceTransformation<RowData, KafkaPartitionSplit, KafkaSourceEnumState> sourceTransformation = (SourceTransformation<RowData, KafkaPartitionSplit, KafkaSourceEnumState>) transformation;
    assertThat(sourceTransformation.getSource(), instanceOf(KafkaSource.class));
}

Example 7 with Transformation

use of org.apache.flink.api.dag.Transformation in project flink by apache.

the class StreamExecDeduplicate method translateToPlanInternal.

@SuppressWarnings("unchecked")
@Override
protected Transformation<RowData> translateToPlanInternal(PlannerBase planner, ExecNodeConfig config) {
    final ExecEdge inputEdge = getInputEdges().get(0);
    final Transformation<RowData> inputTransform = (Transformation<RowData>) inputEdge.translateToPlan(planner);
    final RowType inputRowType = (RowType) inputEdge.getOutputType();
    final InternalTypeInfo<RowData> rowTypeInfo = (InternalTypeInfo<RowData>) inputTransform.getOutputType();
    final TypeSerializer<RowData> rowSerializer = rowTypeInfo.createSerializer(planner.getExecEnv().getConfig());
    final OneInputStreamOperator<RowData, RowData> operator;
    if (isRowtime) {
        operator = new RowtimeDeduplicateOperatorTranslator(config, rowTypeInfo, rowSerializer, inputRowType, keepLastRow, generateUpdateBefore).createDeduplicateOperator();
    } else {
        operator = new ProcTimeDeduplicateOperatorTranslator(config, rowTypeInfo, rowSerializer, inputRowType, keepLastRow, generateUpdateBefore).createDeduplicateOperator();
    }
    final OneInputTransformation<RowData, RowData> transform = ExecNodeUtil.createOneInputTransformation(inputTransform, createTransformationMeta(DEDUPLICATE_TRANSFORMATION, config), operator, rowTypeInfo, inputTransform.getParallelism());
    final RowDataKeySelector selector = KeySelectorUtil.getRowDataSelector(uniqueKeys, rowTypeInfo);
    transform.setStateKeySelector(selector);
    transform.setStateKeyType(selector.getProducedType());
    return transform;
}

Example 8 with Transformation

use of org.apache.flink.api.dag.Transformation in project flink by apache.

the class StreamExecExchange method translateToPlanInternal.

@SuppressWarnings("unchecked")
@Override
protected Transformation<RowData> translateToPlanInternal(PlannerBase planner, ExecNodeConfig config) {
    final Transformation<RowData> inputTransform = (Transformation<RowData>) getInputEdges().get(0).translateToPlan(planner);
    final StreamPartitioner<RowData> partitioner;
    final int parallelism;
    final InputProperty inputProperty = getInputProperties().get(0);
    final InputProperty.DistributionType distributionType = inputProperty.getRequiredDistribution().getType();
    switch(distributionType) {
        case SINGLETON:
            partitioner = new GlobalPartitioner<>();
            parallelism = 1;
            break;
        case HASH:
            // TODO Eliminate duplicate keys
            int[] keys = ((HashDistribution) inputProperty.getRequiredDistribution()).getKeys();
            InternalTypeInfo<RowData> inputType = (InternalTypeInfo<RowData>) inputTransform.getOutputType();
            RowDataKeySelector keySelector = KeySelectorUtil.getRowDataSelector(keys, inputType);
            partitioner = new KeyGroupStreamPartitioner<>(keySelector, DEFAULT_LOWER_BOUND_MAX_PARALLELISM);
            parallelism = ExecutionConfig.PARALLELISM_DEFAULT;
            break;
        default:
            throw new TableException(String.format("%s is not supported now!", distributionType));
    }
    final Transformation<RowData> transformation = new PartitionTransformation<>(inputTransform, partitioner);
    createTransformationMeta(EXCHANGE_TRANSFORMATION, config).fill(transformation);
    transformation.setParallelism(parallelism);
    transformation.setOutputType(InternalTypeInfo.of(getOutputType()));
    return transformation;
}

Example 9 with Transformation

use of org.apache.flink.api.dag.Transformation in project flink by apache.

the class StreamExecGlobalGroupAggregate 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 AggregateInfoList localAggInfoList = AggregateUtil.transformToStreamAggregateInfoList(localAggInputRowType, JavaScalaConversionUtil.toScala(Arrays.asList(aggCalls)), aggCallNeedRetractions, needRetraction, JavaScalaConversionUtil.toScala(Optional.ofNullable(indexOfCountStar)), // isStateBackendDataViews
    false, // needDistinctInfo
    true);
    final AggregateInfoList globalAggInfoList = AggregateUtil.transformToStreamAggregateInfoList(localAggInputRowType, JavaScalaConversionUtil.toScala(Arrays.asList(aggCalls)), aggCallNeedRetractions, needRetraction, JavaScalaConversionUtil.toScala(Optional.ofNullable(indexOfCountStar)), // isStateBackendDataViews
    true, // needDistinctInfo
    true);
    final GeneratedAggsHandleFunction localAggsHandler = generateAggsHandler("LocalGroupAggsHandler", localAggInfoList, grouping.length, localAggInfoList.getAccTypes(), config, planner.getRelBuilder());
    final GeneratedAggsHandleFunction globalAggsHandler = generateAggsHandler("GlobalGroupAggsHandler", globalAggInfoList, // mergedAccOffset
    0, localAggInfoList.getAccTypes(), config, planner.getRelBuilder());
    final int indexOfCountStar = globalAggInfoList.getIndexOfCountStar();
    final LogicalType[] globalAccTypes = Arrays.stream(globalAggInfoList.getAccTypes()).map(LogicalTypeDataTypeConverter::fromDataTypeToLogicalType).toArray(LogicalType[]::new);
    final LogicalType[] globalAggValueTypes = Arrays.stream(globalAggInfoList.getActualValueTypes()).map(LogicalTypeDataTypeConverter::fromDataTypeToLogicalType).toArray(LogicalType[]::new);
    final GeneratedRecordEqualiser recordEqualiser = new EqualiserCodeGenerator(globalAggValueTypes).generateRecordEqualiser("GroupAggValueEqualiser");
    final OneInputStreamOperator<RowData, RowData> operator;
    final boolean isMiniBatchEnabled = config.get(ExecutionConfigOptions.TABLE_EXEC_MINIBATCH_ENABLED);
    if (isMiniBatchEnabled) {
        MiniBatchGlobalGroupAggFunction aggFunction = new MiniBatchGlobalGroupAggFunction(localAggsHandler, globalAggsHandler, recordEqualiser, globalAccTypes, indexOfCountStar, generateUpdateBefore, config.getStateRetentionTime());
        operator = new KeyedMapBundleOperator<>(aggFunction, AggregateUtil.createMiniBatchTrigger(config));
    } else {
        throw new TableException("Local-Global optimization is only worked in miniBatch mode");
    }
    // partitioned aggregation
    final OneInputTransformation<RowData, RowData> transform = ExecNodeUtil.createOneInputTransformation(inputTransform, createTransformationMeta(GLOBAL_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;
}

Example 10 with Transformation

use of org.apache.flink.api.dag.Transformation 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;
}