flink-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF GitHub Bot (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (FLINK-1293) Add support for out-of-place aggregations
Date Mon, 01 Dec 2014 16:40:12 GMT

    [ https://issues.apache.org/jira/browse/FLINK-1293?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14230007#comment-14230007
] 

ASF GitHub Bot commented on FLINK-1293:
---------------------------------------

Github user he-sk commented on a diff in the pull request:

    https://github.com/apache/incubator-flink/pull/243#discussion_r21100381
  
    --- Diff: flink-java/src/main/java/org/apache/flink/api/java/aggregation/AggregationOperatorFactory.java
---
    @@ -0,0 +1,297 @@
    +/*
    + * 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.api.java.aggregation;
    +
    +import static java.lang.String.format;
    +import static java.util.Arrays.asList;
    +import static org.apache.flink.api.java.aggregation.Aggregations.key;
    +
    +import java.util.ArrayList;
    +import java.util.List;
    +import java.util.Vector;
    +
    +import org.apache.commons.lang3.ArrayUtils;
    +import org.apache.commons.lang3.Validate;
    +import org.apache.flink.api.common.typeinfo.BasicTypeInfo;
    +import org.apache.flink.api.common.typeinfo.TypeInformation;
    +import org.apache.flink.api.java.DataSet;
    +import org.apache.flink.api.java.aggregation.AggregationFunction.ResultTypeBehavior;
    +import org.apache.flink.api.java.operators.AggregationOperator;
    +import org.apache.flink.api.java.operators.UnsortedGrouping;
    +import org.apache.flink.api.java.tuple.Tuple;
    +import org.apache.flink.api.java.typeutils.TupleTypeInfo;
    +import org.apache.flink.api.java.typeutils.TupleTypeInfoBase;
    +
    +import com.google.common.primitives.Ints;
    +
    +/**
    + * Factory method container to construct an
    + * {@link AggregationOperator} from a {@link DataSet} or
    + * {@link UnsortedGrouping}.
    + * 
    + * <p>The factory performs the following tasks:
    + * 
    + * <ol>
    + *  <li>Decompose composite aggregation functions into intermediates.
    + *  <li>Insert missing key aggregation function for any group keys. 
    + *  <li>Set intermediate, and output position for each aggregation function.
    + *  <li>Map any group keys to their position in the intermediate tuple.
    + *  <li>Compute the types of intermediate tuple and aggregation result.
    + *  <li>Create the aggregation operator.
    + * </ol>
    + *
    + * <p>Note: Tasks are implemented in a member class in order to
    + * facilitate testing. 
    + */
    +public class AggregationOperatorFactory {
    +	
    +	private static final AggregationOperatorFactory INSTANCE = new AggregationOperatorFactory();
    +	private AggregationFunctionPreprocessor aggregationFunctionPreprocessor = new AggregationFunctionPreprocessor();
    +	private ResultTypeFactory resultTypeFactory = new ResultTypeFactory();
    +	
    +	
    +	/**
    +	 * Construct an {@link AggregationOperator} that implements the
    +	 * aggregation functions listed in {@code functions} on the
    +	 * (ungrouped) DataSet {@code input}. 
    +	 * @param input	An (ungrouped) DataSet.
    +	 * @param functions The aggregation functions that should be computed.
    +	 * @return An AggregationOperator representing the specified aggregations.
    +	 */
    +	public <T, R extends Tuple> AggregationOperator<T, R> aggregate(DataSet<T>
input, AggregationFunction<?, ?>[] functions) {
    +		AggregationOperator<T, R> op = createAggregationOperator(input, new int[0], functions);
    +		return op;
    +	}
    +
    +	/**
    +	 * Construct an {@link AggregationOperator} that implements the
    +	 * aggregation functions listed in {@code functions} on the grouped
    +	 * DataSet {@code input}.
    +	 * 
    +	 * <p>If there are no {@link Aggregations.keys} specified in
    +	 * {@code functions} then a {@code key()} aggregation function is
    +	 * inserted for each group key.
    +	 *  
    +	 * @param input	An grouped DataSet.
    +	 * @param functions The aggregation functions that should be computed.
    +	 * @return An AggregationOperator representing the specified aggregations.
    +	 */
    +	public <T, R extends Tuple> AggregationOperator<T, R> aggregate(UnsortedGrouping<T>
grouping, AggregationFunction<?, ?>[] functions) {
    +		DataSet<T> input = grouping.getDataSet();
    +		int[] groupKeys = grouping.getKeys().computeLogicalKeyPositions();
    +		AggregationOperator<T, R> op = createAggregationOperator(input, groupKeys, functions);
    +		return op;
    +	}
    +	
    +	// TODO if sum and/or count are present, use these to compute average
    +	<T, R extends Tuple> AggregationOperator<T, R> createAggregationOperator(DataSet<T>
input, int[] groupKeys, AggregationFunction<?, ?>[] functions) {
    +		AggregationFunction<?, ?>[] functionsWithExpandedKeys = aggregationFunctionPreprocessor.expandKeys(functions,
groupKeys);
    +		AggregationFunction<?, ?>[] intermediateFunctions = aggregationFunctionPreprocessor.createIntermediateFunctions(functionsWithExpandedKeys,
groupKeys);
    +		int[] intermediateGroupKeys = aggregationFunctionPreprocessor.createIntermediateGroupKeys(intermediateFunctions);
    +		TypeInformation<R> resultType = resultTypeFactory.createAggregationResultType(input.getType(),
functionsWithExpandedKeys);
    +		TypeInformation<Tuple> intermediateType = resultTypeFactory.createAggregationResultType(input.getType(),
intermediateFunctions);
    +		AggregationOperator<T, R> op = new AggregationOperator<T, R>(input, resultType,
intermediateType, intermediateGroupKeys, functionsWithExpandedKeys, intermediateFunctions);
    +		return op;
    +	}
    +
    +	static class AggregationFunctionPreprocessor {
    +
    +		public AggregationFunction<?, ?>[] expandKeys(AggregationFunction<?, ?>[]
functions, int[] groupKeys) {
    +			Vector<AggregationFunction<?, ?>> expanded = new Vector<AggregationFunction<?,?>>();
    +
    +			// test where keys should be included and save keys defined by user
    +			int insertionPosition = -1;
    +			int currentPosition = 0;
    +			List<Integer> definedByUser = new ArrayList<Integer>();
    +			for (AggregationFunction<?, ?> function : functions) {
    +				if (function instanceof KeySelectionAggregationFunction) {
    +					if (function == KeySelectionAggregationFunction.INCLUDE_ALL_KEYS_FUNCTION) {
    +						if (insertionPosition == -1) {
    +							insertionPosition = currentPosition;
    +						}
    +						continue;
    +					} else {
    +						int field = function.getInputPosition();
    +						Validate.isTrue(ArrayUtils.contains(groupKeys, field),
    +								format("The key %d is not in the grouping %s", 
    +										field, asList(groupKeys)));
    +						definedByUser.add(field);
    +					}
    +				}
    +				expanded.add(function);
    +				currentPosition += 1;
    +			}
    +
    +			// insert missing keys if requested
    +			AggregationFunction<?, ?>[] result = null;
    +			if (insertionPosition != -1) {
    +				for (int groupKey : groupKeys) {
    +					if ( ! definedByUser.contains(groupKey) ) {
    +						AggregationFunction<?, ?> key = key(groupKey);
    +						expanded.insertElementAt(key, insertionPosition);
    +						insertionPosition += 1;
    +					}
    +				}
    +				result = new AggregationFunction<?, ?>[expanded.size()];
    +				expanded.toArray(result);
    +			} else {
    +				result = functions;
    +			}
    +			return result;
    +		}
    +
    +		public AggregationFunction<?, ?>[] createIntermediateFunctions(AggregationFunction<?,
?>[] functions, int[] groupKeys) {
    +			List<AggregationFunction<?, ?>> intermediates = new ArrayList<AggregationFunction<?,?>>();
    +			List<CompositeAggregationFunction<?, ?>> composites = new ArrayList<CompositeAggregationFunction<?,
?>>();
    +			int outputPosition = 0;
    +			for (AggregationFunction<?, ?> function : functions) {
    +
    +				// set output position according to the order specified by the user
    +				function.setOutputPosition(outputPosition);
    +				outputPosition += 1;
    +
    +				// check if key() is used without groupBy
    +				if (groupKeys.length == 0
    +						&& function instanceof KeySelectionAggregationFunction) {
    +					throw new IllegalArgumentException("Key selection aggregation function can only
be used on grouped DataSets.");
    +				}
    +
    +				// separate composites
    +				if (function instanceof CompositeAggregationFunction) {
    +					composites.add((CompositeAggregationFunction<?, ?>) function);
    --- End diff --
    
    I need to have a pass over the function list to construct the list of intermediate (simple)
functions. Otherwise, intermediates that occur after a composite might not be reused by the
composite. For example, in aggregate(average(0), count(), sum(0)), there would be two sum
and count functions if the average were decomposed immediately instead of saving it in a list
and decomposing it later.


> Add support for out-of-place aggregations
> -----------------------------------------
>
>                 Key: FLINK-1293
>                 URL: https://issues.apache.org/jira/browse/FLINK-1293
>             Project: Flink
>          Issue Type: Improvement
>          Components: Java API, Scala API
>    Affects Versions: 0.7.0-incubating
>            Reporter: Viktor Rosenfeld
>            Assignee: Viktor Rosenfeld
>            Priority: Minor
>
> Currently, the output of an aggregation is of the same type as the input. This restriction
has to major drawbacks:
> 1. Every tuple field can only be used in one aggregation because the aggregations result
is stored in the field.
> 2. Aggregations having a return type that is different from the input type, e.g., count
or average, cannot be implemented.
> It would be nice to have the aggregation return any kind of tuple as a result, so the
restrictions above no longer apply.
> See also:
> - https://github.com/stratosphere/stratosphere/wiki/Design-of-Aggregate-Operator
> - http://apache-flink-incubator-mailing-list-archive.1008284.n3.nabble.com/Hi-Aggregation-support-td2311.html



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message