wayang-platforms/wayang-spark/src/main/java/org/apache/wayang/spark/operators/SparkMaterializedGroupByOperator.java

/*
 * 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.wayang.spark.operators;

import java.util.Arrays;
import java.util.Collection;
import java.util.Collections;
import java.util.List;
import java.util.Optional;
import org.apache.spark.api.java.JavaPairRDD;
import org.apache.spark.api.java.JavaRDD;
import org.apache.spark.api.java.function.Function;
import org.apache.wayang.basic.operators.MaterializedGroupByOperator;
import org.apache.wayang.core.api.Configuration;
import org.apache.wayang.core.function.TransformationDescriptor;
import org.apache.wayang.core.optimizer.OptimizationContext;
import org.apache.wayang.core.optimizer.costs.LoadProfileEstimator;
import org.apache.wayang.core.optimizer.costs.LoadProfileEstimators;
import org.apache.wayang.core.plan.wayangplan.ExecutionOperator;
import org.apache.wayang.core.platform.ChannelDescriptor;
import org.apache.wayang.core.platform.ChannelInstance;
import org.apache.wayang.core.platform.lineage.ExecutionLineageNode;
import org.apache.wayang.core.types.DataSetType;
import org.apache.wayang.core.util.Tuple;
import org.apache.wayang.spark.channels.RddChannel;
import org.apache.wayang.spark.execution.SparkExecutor;
import scala.Tuple2;


/**
 * Spark implementation of the {@link MaterializedGroupByOperator}.
 */
public class SparkMaterializedGroupByOperator<Type, KeyType>
        extends MaterializedGroupByOperator<Type, KeyType>
        implements SparkExecutionOperator {


    public SparkMaterializedGroupByOperator(TransformationDescriptor<Type, KeyType> keyDescriptor,
                                            DataSetType<Type> inputType,
                                            DataSetType<Iterable<Type>> outputType) {
        super(keyDescriptor, inputType, outputType);
    }

    /**
     * Copies an instance (exclusive of broadcasts).
     *
     * @param that that should be copied
     */
    public SparkMaterializedGroupByOperator(MaterializedGroupByOperator<Type, KeyType> that) {
        super(that);
    }

    @Override
    public Tuple<Collection<ExecutionLineageNode>, Collection<ChannelInstance>> evaluate(
            ChannelInstance[] inputs,
            ChannelInstance[] outputs,
            SparkExecutor sparkExecutor,
            OptimizationContext.OperatorContext operatorContext) {
        assert inputs.length == this.getNumInputs();
        assert outputs.length == this.getNumOutputs();

        RddChannel.Instance input = (RddChannel.Instance) inputs[0];
        RddChannel.Instance output = (RddChannel.Instance) outputs[0];

        final JavaRDD<Type> inputRdd = input.provideRdd();
        final Function<Type, KeyType> keyExtractor =
                sparkExecutor.getCompiler().compile(this.keyDescriptor, this, operatorContext, inputs);
        final Function<scala.Tuple2<KeyType, Iterable<Type>>, Iterable<Type>> projector = new GroupProjector<>();
        final JavaPairRDD<KeyType, Iterable<Type>> groupedKeyRdd = inputRdd.groupBy(keyExtractor, sparkExecutor.getNumDefaultPartitions());
        this.name(groupedKeyRdd);
        final JavaRDD<Iterable<Type>> outputRdd = groupedKeyRdd.map(projector);
        this.name(outputRdd);

        output.accept(outputRdd, sparkExecutor);

        return ExecutionOperator.modelLazyExecution(inputs, outputs, operatorContext);
    }

    @Override
    protected ExecutionOperator createCopy() {
        return new SparkMaterializedGroupByOperator<>(this.getKeyDescriptor(), this.getInputType(), this.getOutputType());
    }

    private static class GroupProjector<Key, Type> implements Function<scala.Tuple2<Key, Iterable<Type>>, Iterable<Type>> {

        @Override
        public Iterable<Type> call(Tuple2<Key, Iterable<Type>> groupWithKey) throws Exception {
            return groupWithKey._2;
        }

    }

    @Override
    public String getLoadProfileEstimatorConfigurationKey() {
        return "wayang.spark.groupby.load";
    }

    @Override
    public Optional<LoadProfileEstimator> createLoadProfileEstimator(Configuration configuration) {
        final Optional<LoadProfileEstimator> optEstimator =
                SparkExecutionOperator.super.createLoadProfileEstimator(configuration);
        LoadProfileEstimators.nestUdfEstimator(optEstimator, this.keyDescriptor, configuration);
        return optEstimator;
    }

    @Override
    public List<ChannelDescriptor> getSupportedInputChannels(int index) {
        return Arrays.asList(RddChannel.UNCACHED_DESCRIPTOR, RddChannel.CACHED_DESCRIPTOR);
    }

    @Override
    public List<ChannelDescriptor> getSupportedOutputChannels(int index) {
        return Collections.singletonList(RddChannel.UNCACHED_DESCRIPTOR);
    }

    @Override
    public boolean containsAction() {
        return false;
    }

}