[ 
https://issues.apache.org/jira/browse/APEXMALHAR-2106?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15341489#comment-15341489
 ] 

ASF GitHub Bot commented on APEXMALHAR-2106:
--------------------------------------------

Github user bhupeshchawda commented on a diff in the pull request:

    https://github.com/apache/apex-malhar/pull/309#discussion_r67841300
  
    --- Diff: 
library/src/main/java/com/datatorrent/lib/stream/MultipleStreamMerger.java ---
    @@ -0,0 +1,215 @@
    +/**
    + * 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 com.datatorrent.lib.stream;
    +
    +import java.util.ArrayList;
    +import java.util.Queue;
    +
    +import org.eclipse.jetty.util.ArrayQueue;
    +
    +import org.apache.hadoop.conf.Configuration;
    +
    +import com.datatorrent.api.DAG;
    +import com.datatorrent.api.DefaultInputPort;
    +import com.datatorrent.api.DefaultOutputPort;
    +
    +/**
    + * A helper class  that adds functionality to bypass the platform 
limitations of combining more than two
    + * streams at a time with Stream Merger.
    + *
    + * Usage:
    + *
    + * dag.addOperator("Stream_1", op1);
    + * dag.addOperator("Stream_2", op2);
    + * dag.addOperator("Stream_3", op3);
    + *
    + * MultipleStreamMerger merger = new MultipleStreamMerger();
    + * DefaultOutputPort streamOut = merger.merge(op1.out)
    + * .merge(op2.out)
    + * .merge(op3.out)
    + * .mergeStreams(dag, conf);
    + *
    + * dag.addStream("merger-counter", streamOut, counterOp.counter);
    + *
    + * @param <K>
    + */
    +public class MultipleStreamMerger<K>
    +{
    +  public class Stream
    +  {
    +    DefaultInputPort<K> destPort;
    +    SourcedOutputPort sourcePort;
    +    String name;
    +
    +    public Stream(String name, SourcedOutputPort sourcePort, 
DefaultInputPort<K> destPort)
    +    {
    +      this.destPort = destPort;
    +      this.sourcePort = sourcePort;
    +      this.name = name;
    +    }
    +  }
    +
    +  public class NamedMerger
    +  {
    +    StreamMerger<K> merger;
    +    String name;
    +
    +    public NamedMerger(String name, StreamMerger<K> merger)
    +    {
    +      this.merger = merger;
    +      this.name = name;
    +    }
    +  }
    +
    +  /**
    +   * A simple class to allow us to track whether the port to be merged is 
internal (allowing it to be thread local)
    +   * or external
    +   */
    +  public class SourcedOutputPort
    +  {
    +    boolean internal;
    +    DefaultOutputPort<K> port;
    +
    +    public SourcedOutputPort(DefaultOutputPort<K> port)
    +    {
    +      this.internal = false;
    +      this.port = port;
    +    }
    +
    +    public SourcedOutputPort(boolean internal, DefaultOutputPort<K> port)
    +    {
    +      this.internal = internal;
    +      this.port = port;
    +    }
    +  }
    +
    +  ArrayList<DefaultOutputPort<K>> streamsToMerge = new ArrayList<>();
    +
    +  private DefaultOutputPort<K> streamOutput = new DefaultOutputPort<>();
    +
    +  /**
    +   * Used to define all the sources to be merged into a single stream.
    +   *
    +   * @param sourcePort - The output port from the upstream operator that 
provides data
    +   * @return The updated MultipleStreamMerger object that tracks which 
streams should be unified.
    +   */
    +  public MultipleStreamMerger<K> merge(DefaultOutputPort<K> sourcePort)
    +  {
    +    streamsToMerge.add(sourcePort);
    +    return this;
    +  }
    +
    +
    +  /**
    +   * Given the streams to merge have been selected with {@link 
#merge(DefaultOutputPort)}, create a subDAG and add it
    +   * to an existing DAG.
    +   *
    +   * To merge more than two streams at a time, we construct a tiered 
hierarchy of thread-local StreamMerger operators
    +   * E.g.
    +   *
    +   * Stream 0 ->
    +   *            StreamMerger_1 ->
    +   * Stream 1 ->
    +   *                               StreamMerger_Final -> Out
    +   * Stream 2 ->
    +   *            StreamMerger_2 ->
    +   * Stream 3 ->
    +   * Note that we don't use the populateDAG function because that is only 
used to flatten the
    +   * module when what we really need to do is to define the connections to 
the parent DAG.
    +   * The populateDAG does not contain operators outside of the module and 
thus cannot connect
    +   * to the external operators.
    +   *
    +   * @param dag - The DAG to update
    +   * @param conf - The configuration
    +   *
    +   */
    +  public DefaultOutputPort<K> mergeStreams(DAG dag, Configuration conf)
    +  {
    +    if (streamsToMerge.size() < 2) {
    +      throw new IllegalArgumentException("Not enough streams to merge, at 
least two streams must be selected for " +
    +          "merging with `.merge()`.");
    +    }
    +
    +    ArrayList<Stream> streamsToAddToDag = new ArrayList<>();
    +    ArrayList<NamedMerger> operatorsToAdd = new ArrayList<>();
    +
    +    // Determine operators and streams to add to the DAG
    +    constructMergeTree(streamsToAddToDag, operatorsToAdd);
    +
    +    for (NamedMerger m : operatorsToAdd) {
    +      dag.addOperator(m.name, m.merger);
    +    }
    +
    +    for (Stream s : streamsToAddToDag) {
    +      DAG.StreamMeta stream = dag.addStream(s.name, s.sourcePort.port, 
s.destPort);
    +      if (s.sourcePort.internal) {
    +        stream.setLocality(DAG.Locality.CONTAINER_LOCAL);
    +      }
    +    }
    +
    +    return streamOutput;
    +  }
    +
    +  /**
    +   * Given a set of streams to be merged (defined via {@link 
#merge(DefaultOutputPort)}), compute the optimal
    +   * structure of cascading mergers that need to be instantiated, added to 
the dag, and linked together.
    +   * @param streamsToAddToDag - (output)  A list that is populated with 
streams that should be added to the  DAG
    +   * @param operatorsToAdd - (output) A list that is populated with 
operators to be added to the DAG
    +   */
    +  public void constructMergeTree(
    +      ArrayList<Stream> streamsToAddToDag,
    +      ArrayList<NamedMerger> operatorsToAdd)
    +  {
    +    if (streamsToMerge.size() < 2) {
    +      throw new IllegalArgumentException("Not enough streams to merge. 
Ensure `.merge` was called for each stream " +
    +          "to be added.");
    +    }
    +
    +    /**
    +     * We can unify all unmerged streams by using a Queue. Unconnected 
ports are pushed to a queue (FIFO) and then for
    +     * every pair of unconnected ports, we create a Merger. If only one 
port remains, we simply use the last added
    +     * merger.
    +     *
    +     * We use SourcedOutputPorts to track whether a given port comes from 
inside or outside the module, allowing us to
    +     * set operator locality for those ports that are strictly from inside 
the module.
    +     */
    +    Queue<SourcedOutputPort> unconnectedPorts = new ArrayQueue<>();
    --- End diff --
    
    Why using this implementation of queue?


> Support merging multiple streams with StreamMerger 
> ---------------------------------------------------
>
>                 Key: APEXMALHAR-2106
>                 URL: https://issues.apache.org/jira/browse/APEXMALHAR-2106
>             Project: Apache Apex Malhar
>          Issue Type: New Feature
>            Reporter: Ilya Ganelin
>            Assignee: Ilya Ganelin
>
> To properly implement the Flatten transformation (and other Stream 
> combination operations), Apex must support merging data from multiple 
> sources. The StreamMerger operator can be improved to merge multiple streams, 
> rather than just the two streams it can handle in the present implementation. 



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

Reply via email to